IIT Mandi Mechanism and Robot Book
IIT Mandi Mechanism and Robot Book
IIT Mandi Mechanism and Robot Book
Rajeev Kumar
Vishal S. Chauhan
Mohammad Talha
Himanshu Pathak Editors
Machines,
Mechanism
and Robotics
Proceedings of iNaCoMM 2019
Lecture Notes in Mechanical Engineering
Series Editors
Francisco Cavas-Martínez, Departamento de Estructuras, Universidad Politécnica
de Cartagena, Cartagena, Murcia, Spain
Fakher Chaari, National School of Engineers, University of Sfax, Sfax, Tunisia
Francesco Gherardini, Dipartimento di Ingegneria, Università di Modena e Reggio
Emilia, Modena, Italy
Mohamed Haddar, National School of Engineers of Sfax (ENIS), Sfax, Tunisia
Vitalii Ivanov, Department of Manufacturing Engineering Machine and Tools,
Sumy State University, Sumy, Ukraine
Young W. Kwon, Department of Manufacturing Engineering and Aerospace
Engineering, Graduate School of Engineering and Applied Science, Monterey,
CA, USA
Justyna Trojanowska, Poznan University of Technology, Poznan, Poland
Francesca di Mare, Institute of Energy Technology, Ruhr-Universität Bochum,
Bochum, Nordrhein-Westfalen, Germany
Lecture Notes in Mechanical Engineering (LNME) publishes the latest develop-
ments in Mechanical Engineering—quickly, informally and with high quality.
Original research reported in proceedings and post-proceedings represents the core of
LNME. Volumes published in LNME embrace all aspects, subfields and new
challenges of mechanical engineering. Topics in the series include:
• Engineering Design
• Machinery and Machine Elements
• Mechanical Structures and Stress Analysis
• Automotive Engineering
• Engine Technology
• Aerospace Technology and Astronautics
• Nanotechnology and Microengineering
• Control, Robotics, Mechatronics
• MEMS
• Theoretical and Applied Mechanics
• Dynamical Systems, Control
• Fluid Mechanics
• Engineering Thermodynamics, Heat and Mass Transfer
• Manufacturing
• Precision Engineering, Instrumentation, Measurement
• Materials Engineering
• Tribology and Surface Technology
Machines, Mechanism
and Robotics
Proceedings of iNaCoMM 2019
Editors
Rajeev Kumar Vishal S. Chauhan
School of Engineering School of Engineering
Indian Institute of Technology Mandi Indian Institute of Technology Mandi
Mandi, Himachal Pradesh, India Mandi, Himachal Pradesh, India
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Organization
Conference Proceedings
Chief Guests
Guests of Honour
Shri. Jayant Patil, Whole Time Director (Defence and L&T-NxT), Member of the
Board, Larsen & Toubro Limited
Dr. S. Guruprasad, Distinguished Scientist and Director General—Production
Coordination and Services Interaction (PC and SI)
Commodore Mukesh Bhargava (Retd.), Vice President, Member of the Board, L&T
Defence
Shri. K. B. Batra, General Manager (Engineering and PCRI), BHEL, Haridwar
v
vi Organization
Patron
Advisory Committee
Organizing Committee
Organizing Institute
Preface
It is a great privilege for us to present the proceedings of the 4th International and
19th National in the series of biennial conferences on Machines and Mechanisms
(iNaCoMM 2019) organized under the aegis of Association for Machines and Mech-
anisms (AMM) and International Federation for the Promotion of Mechanism and
Machine Science (IFToMM) to be held at Indian Institute of technology Mandi,
India, during 5–7 December 2019.
iNaCoMM 2019 aims to provide a platform for discussing the issues, challenges,
opportunities and findings of various aspects of design and analysis of machines
and mechanisms by bringing together academicians, researchers, industry experts
and students who are working in these domains. This conference has been a good
opportunity for participants coming from different parts of the country to present
and discuss the practical challenges encountered and the solutions adopted in their
respective research areas.
The responses to the call-for-papers had been overwhelming. Unfortunately, many
papers from reputed institutions could not be accepted due to the reviewing outcomes.
We would like to express our acknowledgment and appreciation to the reviewers
who helped us in maintaining the high quality of manuscripts by making valuable
suggestions for the authors. We would also like to extend our thanks to the members
of the organizing team for their hard work.
The deliberations in the conference consist of plenary sessions and keynote
lectures by the leading experts and oral and poster presentations by the delegates
on a wide range of topics in the field of machines and mechanisms. The exuberant
programme will result in extensive exchange of information and networking among
ix
x Preface
xi
xii Contents
xxvii
xxviii Editors and Contributors
Contributors
1 Introduction
Supported by SERB (Science and Engineering Research Board) Through the Project
YSS/2015/001245
density on the earth [7]. Different studies have been performed to maximize the
levitating force of the standing wave acoustic levitation [8, 9]. Polystyrene particles
have been translated in the water medium along the axial direction of the standing
wave acoustic levitation system [10]. Different studies have been performed in which
the particles have been manipulated by controlling the phase of the ultrasonic sound
sources [11, 12].
In this study, the levitation of polystyrene balls near the pressure nodes of the
standing wave is demonstrated. Further, a one-dimensional array of five transducers
connected through the revolute joints is used to create the focal point. Focal point
of the array can be varied by providing the rotation to the transducers. By gradually
varying the focal point of the transducer array, the mixing of the particles has been
studied numerically.
2 Numerical Modeling
For the numerical study of the standing wave acoustic levitation, 2D model of COM-
SOL Multiphysics software is used. Standing wave acoustic levitation system consists
of a driver and a reflector. Geometry of the simple acoustic levitation system with
one ultrasonic tweezer/driver is shown in Fig. 1a. The diameter of the tweezer casing
(outer sides highlighted in blue) is 16 mm and height is 12 mm. Inside the tweezer
casing, there is a vibrating part, of diameter 10 mm and height 7 mm, which produces
the sound. The material of the driver is taken as aluminum. The medium around the
tweezer is air (highlighted in gray) of width 20 mm. Further, the standing wave
acoustic levitation system with one-dimensional array of five acoustic transducers is
studied numerically. Figure 1b shows the geometry of the levitation system with five
acoustic transducers arranged on a flat surface or having focal point at infinity.
Fig. 1 Geometry of standing wave acoustic levitation system with a One acoustic tweezer and b
One-dimensional array of five acoustic transducers
A Mechanical Contrivance for Acoustic Levitation and Mixing of Particles? 3
2.2 Procedure
In standing wave acoustic levitation system, the distance between the driver and the
reflector should satisfy the condition of resonance. In Fig. 1a, the upper boundary of
the air domain is provided with the sound hard boundary condition and hence works
as a reflector. Plane wave radiation condition is provided at the side boundaries of
the air domain. Sound hard boundary condition and plane wave radiation condition
can be represented by the following expressions respectively
1
− n − (∇ pt − q) = 0. (1)
ρc
1 k i
− n − (∇ pt − q) + i p + T p = Q i . (2)
ρc ρc 2kρc
where k is the wave number, ρc is the density of the medium, q is the dipole source,
Q i is the monopole source of incoming pressure field and T at a given point on the
boundary denotes the Laplace operator in the tangent plane at that particular point.
For the excitation frequency of 40 kHz, distance between the driver and the reflec-
tor for third resonance mode is calculated numerically. Numerical result is validated
with the experimental result. Further, another numerical model with one-dimensional
array of five acoustic transducers is developed. The intersection point of axes of all
the transducers is considered as the focal point. The focal point of the transducer
array is varied by providing the rotation to the transducers. The effect of varying the
focal point of the transducers array is studied.
In this study, finite element approach is used to simulate the problem. Equation (3)
represents the general wave equation
∂2 p
− c2 ∇ 2 p = 0. (3)
∂t 2
To obtain solution using finite element method, Eq. (3) is converted into the weak
formulation
∂2 p
Wi 2 d A − c2 Wi ∇ 2 pd A = 0. (4)
∂t
D D
4 S. Yadav and A. Gupta
where Wi is the trial function. Using the polynomial for approximate solution and
applying the boundary conditions, p can be written as-
N
p= Ci (t)Ni (7)
i=1
where Ci (t) are the time dependent coefficients and Ni are basic functions.
From Eqs. (6) and (7), the weak formulation can be obtained and can be written as-
N
N
∂ 2 Ci (t)
Ni Wi d A + c 2
Ci ∇ Ni .∇Wi d A = 0. (8)
i=1
∂t 2 i=1
D D
From Eq. (8), Ci can be obtained. Putting the values of Ci in the Eq. (7) gives the
final expression of the pressure.
Finite element-based software (COMSOL Multiphysics) is used for the numerical
study. 2D, six nodded free triangular elements are selected to mesh the geometry. To
find out the optimum size of the element, convergence study has been performed for
single acoustic tweezer model as well as the one-dimensional array of five transducers
model. Plot of absolute pressure at a point with respect to the mesh size (number of
elements per wavelength) for both the models is shown in Fig. 2a, b respectively.
As shown in Fig. 2, the solution is almost close to converged solution with element
size λ/15, where λ is the wavelength. Further reducing the element size will lead to
the higher computational cost. So, the element size λ/15 is selected for the simulation
to get satisfactory results for both models without increasing the unnecessary com-
putational cost. Figure 3 shows the meshing of the model of standing wave acoustic
levitation system with one acoustic tweezer.
A Mechanical Contrivance for Acoustic Levitation and Mixing of Particles? 5
Fig. 2 Convergence plot for the numerical model with a Single acoustic tweezer and b One-
dimensional array of five transducers
In standing wave acoustic levitation, particles can be levitated near the pressure nodes
of the standing wave. When the distance between the driver and the reflector is right,
the resonance phenomenon occurs and the standing wave is formed. For the levitation
system with single acoustic tweezer, the distance between the driver surface and the
reflector for third resonance mode is calculated numerically. The distance for third
resonance mode is obtained as 13.56 mm as shown in Fig. 4a. Distance for the third
6 S. Yadav and A. Gupta
Fig. 4 Resonance condition corresponding to the third resonance mode calculated a Numerically
as 13.56 mm and b Experimentally as 13.50 mm
resonance mode is also calculated experimentally as 13.50 mm. For third resonance
mode, two pressure nodes are obtained outside the casing of the tweezer. At each
pressure node, a small polystyrene ball of 3 mm diameter is levitating as shown in
Fig. 4b.
Figure 4 shows that the particles can be trapped near the pressure nodes of the
standing wave. Further, using the same approach, a one-dimensional array of the
five transducers (shown in Fig. 1b) is studied numerically. The excitation frequency
of the transducers is 40 kHz. For the resonance condition, the distance between the
driver surface and the reflector is taken as 130 mm.
Initially, the focal point of the transducer array is taken at a distance of 60 mm
as shown in Fig. 5a. Two standing waves with separate pressure nodes/anti-nodes
are constructed. As the focal point moves away from the transducers, the pressure
nodes/anti-nodes of both the standing waves come closer as shown in Fig. 5b (focal
point at 75 mm) and Fig. 5c (focal point at 90 mm). As the distance of the focal
point increases further, both the standing waves combine into a single standing wave
as shown in Fig. 5d (focal point at 105 mm). In standing wave acoustic levitation,
particle gets trapped near the pressure node. As the pressure node moves, the particle
also moves with it. Hence, two particles can be mixed together by gradually varying
the focal point of the array of the transducers.
4 Summary
In this study, standing wave acoustic levitation using ultrasonic acoustic tweezer is
studied numerically and validated with the experiment. It is found that the particles
can be levitated near the pressure nodes of the standing wave. Further, the similar
approach is used to propose a mechanical design for the acoustic levitation using
an one-dimensional array of five transducers. The effect of varying the focal point
A Mechanical Contrivance for Acoustic Levitation and Mixing of Particles? 7
Fig. 5 Surface plot of total acoustic pressure field with focal point of the transducers array at a 60
mm, b 75 mm, c 90 mm and d 105 mm
of the transducer array is studied numerically. It is observed that the mixing of two
particles can be achieved by gradually varying the focal point of the array of the
transducers.
8 S. Yadav and A. Gupta
References
1. Apfel RE (1976) Technique for measuring the adiabatic compressibility, density, and sound
speed of submicroliter liquid samples. J Acoust Soc Am 59(2):339–343
2. Trinh EH, Apfel RE (1980) Sound velocity of supercooled water down to -33 C using acoustic
levitation. J Chem Phys 72(12):6731–6735
3. Trinh EH, Hsu CJ (1986) Acoustic levitation methods for density measurements. J Acoust Soc
Am 80(6):1757–1761
4. Trinh EH, Ohsaka K (1995) Measurement of density, sound velocity, surface tension, and
viscosity of freely suspended supercooled liquids. Int J Thermophys 16(2):545–555
5. Tian Y, Holt RG, Apfel RE (1995) A new method for measuring liquid surface tension with
acoustic levitation. Rev Sci Instrum 66(5):3349–3354
6. Ohsaka K, Trinh EH (1989) Melting and solidification of acoustically levitated drops. J Cryst
Growth 96:973–978
7. Xie W-J, Cao CD, Lu YJ, Wei B-B (2002) Levitation of iridium and liquid mercury by ultra-
sound. Phys Rev Lett 89(10):104304
8. Xie WJ, Wei B (2001) Parametric study of single-axis acoustic levitation. Appl Phys Lett
79(6):881–883
9. Yadav S, Gupta A (2019) Maximization of acoustic levitating force for a single axis acoustic
levitation system using the finite element method. Chin Phys Lett 36(3):034302
10. Whitworth G, Grundy MA, Coakley WT (1991) Transport and harvesting particles using mod-
ulated of suspended ultrasound. Ultrasonics 29:439–444
11. Matsui T, Ohdaira E, Masuzawa N, Ide M (1995) Translation of an Object Using Phase-
Controlled Sound Sources in Acoustic Levitation. Japanese Journal of Applied Physics
34(5B):2771–2773
12. Marzo A, Seah SA, Drinkwater BW, Sahoo DR, Long B, Subramanian S (2015) Holographic
acoustic elements for manipulation of levitated objects. Nat Commun 6:1–7
Design and Validation of Flexure-Based
Hinges for Space Deployable Antenna
Reflector
Abstract Space exploration arises the demand for launching large-diameter antenna
reflectors to satisfy the need of high-bandwidth telecommunication, earth observation
and deep space interplanetary missions. Launching of monolithic LDR antennas of
sizes 3 m or more is not feasible due to limited launch fairing space of state-of-the-art
launch vehicles. Therefore, the development of innovative deployment mechanisms
is the present need of the hour. Many researchers have proposed various mechanism
options to fold large size antenna reflectors in compact size and deploy in space
to full configuration. Deployment process of antenna is process of transition from
mechanism to structure which is one of the unreliable stages due to availability
of many conventional rotary joints which causes loss of energy due to backlash,
friction and misalignment. This paper proposes a solution of replacement of conven-
tional hinges with flexure hinges in state-of-the-art space deployable configuration
of large size reflectors, reforming it as compliant configuration which eliminates the
factors causing loss of energy. Tape flexures are explored as a suitable candidate
for compliant deployable configuration. The proposed configuration consists of two
tape flexures mounted in such a way that concave curve of each tape will face to
each other. Options with single tape flexure and double tape flexures are explored
and compared. Experimental evaluation of a proposed joint configuration is carried
out for the development of a deployable antenna reflector.
1 Introduction
Deployable structures are used for launching large size space antenna reflectors of
the order of more than 3 m. The main requirement is to fold the antenna reflectors
to compact size so that it can be easily accommodated in launch vehicle fairing
space during launching and later deployed to full configuration in space after launch.
Deployable structures are useful to overcome this type of challenge. Deployable mesh
antenna is the subject of research last two decades, due to its simple configuration,
high thermal stability, repeatability, easy in folding to compact size. Deployment
of this deployable mechanism is most vulnerable stage for the antenna which is
intended to happen in space, if any failure in full deployment, the complete mission
will may be be ruined. This type of failure will be considered as single-point failure.
Deployment of an unfurlable antenna reflector is categorized in three types, viz. self-
actuated initial deployment, motorized deployment and final deployment to deployed
lock position. During all these stages of deployment, actuation power needs to over-
come the energy losses due to friction, backlash and misalignment. To eliminate this
problem, a solution of flexure-based joints is proposed and presented in this paper.
Flexure-based joints consist of flexible hinges which are folded elastically in stowed
configuration. The flexible hinges can self-deploy to certain angle by releasing stored
strain energy which helps in initial deployment when released from stowed configu-
ration. There are many advantages of flexible hinges like easy to manufacture, easy to
assemble, low mass to deployed stiffness ratio, low cost and self-latching capability
at fully deployed configuration. There is a lot of demand of these flexible hinges
which is being widely used in space deployable structures such as synthetic aperture
radars (SAR), solar arrays and deployable antenna booms. Design of flexure hinges
using tape springs is presented in this paper as an objective to form a compliant-based
deployable reflector configuration.
2.1 Description
Simplest type of ring structure used for mesh antenna is known as Astromesh [1] ring
structure, in which four-bar mechanism is used as a basic entity with one diagonal
telescopic member and a cable passing thru the diagonal telescopic member which
helps in deployment of Astromesh configuration. In order to provide a synchronous
deployment, a gear pair is incorporated as represented by schematic view of single
bay of Astromesh deployable mechanism (see Fig. 1).
Design and Validation of Flexure-Based Hinges for Space … 11
Fig. 1 Exploded view of deployable mesh reflector with description of single bay
The bay with diagonal telescopic member and joints at each corner is repeated number
of times to get closed-loop ring structure. Deployment of this configuration is done
by contracting the telescopic member with help of cable winding mechanism. Each
entity is tied with top mesh layer and rear net which are further connected with
each other at multiple points thru unidirectional cable ties. When deployed fully, the
connection with entities pull the mesh and cable ties prestress the mesh layer to take
parabolic shape. Kinematic formulation of deployment path is plotted the different
intermediate position of ring structure (see Fig. 2) which are considered to be vulner-
able in full deployment of antenna mesh reflectors. Red color members are upper
circumferential members, and blue color members represent lower circumferential
members. Vertical members are shown in black color, and telescopic members are
described in green color.
Each bay of deployable configuration consists of two types of joints, viz. 3J and
5J joint configuration. 3J joint configuration comprises a gear pairs for simulta-
neous motion of circumferential links, and 5J joint consists of conventional hinges
connecting circumferential and diagonal links. An attempt has been made to convert
circumferential conventional hinges of 5J joint configuration into flexure hinges using
tape flexures. These flexible hinges will be in fully bend position of tape flexures
maximum up to 90° in stowed configuration during launch and will be allowed to
relax to reach at straight position for full deployment in orbit (see Fig. 3).
12 H. Arora et al.
Conventional 5J joint
Compliant 5J joint
Fig. 3 Stowed and deployed configuration of deployable ring structure with CAD illustrations of
conventional and compliant 5J joints
Tape flexures [2] have been explored for making 5J joint as frictionless compliant
joint because of its geometry simplicity. It consists of simple metallic strips which
have a slight curvature along transverse direction (see Fig. 4). Best example of these
Design and Validation of Flexure-Based Hinges for Space … 13
tape flexures is steel tapes used for measurement. Material of strips could be of low
modulus of elasticity for providing flexibility during folding sequence and hardened
to gain flexure strength to produce bending greater than 90° without having perma-
nent deformation. The curve portion provides the stiffness in straight configuration
which will be useful in final deployed configuration. Tension in mesh to form a
parabolic profile will certainly generate forces in transverse direction on tape flex-
ures. Therefore, stiffness in transverse direction is achieved by taking two tapes with
opposite curved faces, and stiffness can further be enhanced with multiple numbers
of layers of tape flexures.
Buckling behavior of tape flexure [3] exists differently when bending moment applied
on concave and convex curvature (see Fig. 5). When a tape flexure subjects to bending
moment applied on convex curvature side, it shows the higher rigidity initially, then
its stiffness suddenly breaks down with a snap and form an elastic bend up to critical
moment M+ max , and further, bending moment drops down and stabilizes at M+ * .
Table 1 Von Mises stresses (MPa) for various tape flexure configurations
Material Hardened steel (carpenter steel CFRP laminate Be–Cu alloy
tape)
Single-layer tape flexures 403.5 82.6 196.5
Double-layer tape flexures 387.6 75.6 165.1
Similarly, when tape flexure subjects to bending moment applied on concave curva-
ture side, it shows lower rigidity and perform elastic bend with lower small bending
moment M-*. Therefore, combination of tape spring with opposite curvature will
result into smooth bending and provide adequate stiffness.
A set of single-layer tape flexures facing with opposite curvature and double-layered
tape flexures facing with opposite curvature are structurally analyzed [4] with rota-
tional moments applied at ends, and tangential contacts boundary conditions [5] have
been applied (see Fig. 6). This analysis is a kind of nonlinear large deflection analysis
which is solved using Altairs Radioss explicit dynamic simulation tool. Deflection
and stresses are compared with various materials and flexure configuration (see Table
1). Each case is analyzed with 0.12 mm thickness of tape, and rotational moment of
85° is applied at each end.
flexures are deflecting with the tension loads produced by RF mesh, whereas double-
layer flexures are well capable to accommodate mesh stretching tension loads and
behaving as a stable joint in fully deployed configuration which further eliminates
the additional locking requirement. The configuration in folding state is achieved
fully by bending of flexures near to 90°, and flexure rotational stiffness helped in
initial deployment of 10°–15°.
Conventional joints have stiffness in both radial and tangential axis of deployed
antenna reflector, whereas in case of compliant joints, equal stiffness can be varied
by combination of tape spring configuration and adequate layers. Compliant joints
are frictionless joints as compared to friction between mating parts of conventional
joints. There are no chances of loss of energy in compliant configuration due to
rotation. Therefore, chances of obstruction of deployment are eliminated. Compliant
joints are easy to manufacture and assemble as compared to multi-parts configuration
of conventional joints. Auto locking/latching is achieved in compliant joints by its
geometry as an added advantage, whereas additional locking mechanism is required
in case of configuration with conventional joints. Torsion springs are required in
conventional hinges to generate initial torsional moment, whereas initial bending
configuration of tape flexures will itself behave like a spring to produce required
torsional moment which eliminates the requirement of additional torsional springs.
5 Conclusion
This paper has addressed the criticality of deployment path for a mesh reflector to
deploy fully in space. To solve the problem, a solution is proposed to adapt compliant
configuration of deployable reflector by replacing conventional hinges into flexure
hinges. Tape flexures are studied as a prominent option for flexible hinges. Buckling
behavior of curved tapes is studied with a conclusion of using a combination of
two tapes positioned with opposite curve faces. The combinations with single and
16 H. Arora et al.
double layer of tapes are structurally analyzed with various material options and
found that flexible tapes made of CFRP material are more suitable with benign
stresses under maximum rotational moment. Experimental validation of the concept
is demonstrated with modified design of joints incorporating tape flexures. Tape
flexures are ascertained to be a suitable option for deployable configuration of large
size space antenna reflectors.
References
1. Thomson MW (1999) The astromesh deployable reflector. IEEE Antenna Propogation Soc
1516–1519
2. Seffen KA (2001) On the behavior of folded tape-springs. J Appl Mech ASME 68:369
3. Zuo Y, Jin G, Xie P (2018) Calculative and experimental study of the CFRP tape spring. J Mech
Sci Technol 32(8):3603–3609 (Springer)
4. Yang H, Liu R, Wang Y, Deng Z (2015) Experiment and multiobjective optimization design of
tape-spring hinges. Struct Multidisc Optim 51:1373–1384 (Springer)
5. Guan F, Wu Y, Wang X, Y (2010) The mechanical behavior of the double piece of tape spring. In:
international conference on intelligent computing (ICIC) 2010. Communications in computer
and information science book series (CCIS), vol 93, Springer, Berlin Heidelberg, pp 102–110
Effect of Implant Materials on Bone
Remodelling Around Cemented
Acetabular Cup
Abstract Bone resorption around cemented acetabular cup caused failure in the long
tenure. This study aims to investigate the effect of different implant materials on bone
remodelling around the cemented acetabular cup. Finite element (FE) models of intact
and implanted pelvic models were established using CT-scan data sets. Three combi-
nations (acetabular cup-femoral head) of implant materials, UHMWPE-CoCrMo,
CoCrMo-CoCrMo, and alumina-alumina ceramic, were considered to predict the
effect of implant materials on bone remodelling around cemented acetabular cup.
Less bone resorption was observed for UHMWPE acetabular cup. Bone remodelling
was nearly similar for the ceramic and metallic cup. Considering the bone remod-
elling scenario, the results of the current study stated that ceramic could be a good
alternative to metallic implant in the cemented acetabular cup since ceramic has
better wear resistance property than metal.
1 Introduction
Total hip replacement (THR) is an invasive restoration process that has gained much
popularity in the past few years. In THA, the diseased part of the acetabulum is
replaced by an artificial cup commonly known as an acetabular cup. The reports,
as suggested by NJR for England and Wales indicated, around 11% of hip replace-
ment surgery had to go through revision during the period from 1 April 2003 to
31 December 2015 [1]. NJR provided that 38,310 cases out of 78,130 first hip re-
revisions were associated with aseptic loosening of the acetabular cup [1]. The domi-
nant cause for the failure of the cemented acetabular cup is aseptic loosening [1].
Excessive bone resorption is one of the causes of late loosening of the cemented
acetabular cup and that may promote earlier with the improper selection of implant.
Decrease in cup bearing diameter would reduce volumetric wear and subsequently
increase bone resorption. Selection of implant material also shows a decisive role
in implant-induced adaptive bone remodelling. A stiff implant can promote earlier
loosening of the cup due to bone resorption. Elevated cement mantle stress and high
contact stress also increase the risk of failure. Earlier studies suggested that cemented
acetabular cup performs better as compared to uncemented one for elderly patients
[2].
Clinical interest is focused on the more durable acetabular cup that may reduce
loosening. Polyethylene acetabular cup has problem-related to excessive wear gener-
ation and that promote late loosening due to osteolysis. The ceramic implant would
reduce the wear-related problem; however, bone remodelling is the leading causes
of concerns. Owing to less computational cost and economic, finite element (FE)
analysis is generally considered for pre-clinical studies plus the assessment of the
orthopaedic implants. Past research on cemented acetabular cups was focused on the
immediate post-operative state [3–13]. The research was constrained to the influence
of stress/strain shielding [4–6] or stress in cement mantle owing to changes in the
thickness of the cement mantle [3, 10–14] or enhancing the capability of cement–bone
interface or fatigue (repeated variations of stress) failure analysis of cement mantle
[9–11, 15–17]. Long-term analysis, including bone remodelling, around of cemented
acetabular cups are few [18, 19]. A clinical study by Shetty et al. [18] used only one
type of acetabular material (polyethylene) to investigate the bone remodelling around
the cemented acetabular cup. A numerical analysis of bone remodelling around the
cemented acetabular cup specifies concluding bone density decrement around the
acetabulum [19]. However, to date, the investigation of different implant material on
bone remodelling around the cemented acetabular cup is still not well reported.
Enhancement in cement mantle thickness would mitigate tensile stress in the
cement mantle [8]. However, in that study, the only immediate post-operative state
was considered, and the effect of implant material properties was not considered.
Regardless of failure due to osteolysis, it is foreseen that implant material might
affect bone remodelling around the acetabulum. Elevated tensile stress in the cement
mantle can initiate a crack, which can further propagate due to loading conditions and
eventual loosening of the implant [20]. The immediate post-operative state might not
be able to predict the loosening of the implant accurately. Considering this hypothesis
and background, the objectives of this present study is to determine the effect of
different implant materials on bone remodelling around the cemented acetabular
cup.
Considering the success of the cemented acetabular cup for elderly patients, a right
hemipelvis of 62 years, and the female patient was considered for the current study.
Three-dimensional (3D) FE model of this right hemipelvis was made from CT-
scan data like previous evaluations [21–23]. The developing technique of the 3D FE
model of the pelvic bone was corroborated and tested [24, 25]. Ten-noded tetrahedral
Effect of Implant Materials on Bone Remodelling Around Cemented … 19
elements were considered for the meshing the geometry of implanted hemipelvis.
Mesh convergence study was made (established on stress value generated in the
implanted pelvic bone) in such a way that the outcome does not govern by the
element size. Coarse, medium and fine element size was taken for meshing. The
coarse, medium and fine mesh were consisting of 180,217, 312,352 and 376,930
number of elements. So, the second FE model containing 312,352 number of elements
(element edge length varied from 0.5–3 mm) was considered for the present analysis
to reduce the computational cost. The intact and implanted pelvic bone model was
produced to comprehend bone remodelling and fatigue in the cement mantle. The
solid model and FE sub-model of the implanted pelvic bone, along with all cups, are
represented in Fig. 1. Three millimetres (3 mm) uniform or constant cement mantle
thickness polymethylmethacrylate (PMMA) was considered for cemented fixation
of the acetabular cup, which already has been reported as the safe value of cement
thickness for cemented acetabular cups [3, 17]. The acetabular cup was situated and
tilting at an angle of 45-degree inclination angle and 15-degree anteversion angle
[26]. Ceramic (alumina), cobalt-chromium molybdenum (Co–Cr–Mo) and ultra-
high molecular weight polyethylene (UHMWPE) were taken for the acetabular cup.
The dimensions of each acetabular cup kept the same, where the bearing diameter is
28 mm, and the outer diameter is 48 mm. A spherical femoral head was considered
for the presentation of the hip joint reaction force (Fig. 1).
Fixed constraints
The construction of the pelvic bone is close to the sandwich structure. Modelling
technique of the FE model of the pelvic bone was similar to an earlier paper [21]. An
elastic, isotropic and heterogeneous material property was taken for the cancellous
(spongy) part of the bone. Materials property of spongy bone was taken as heteroge-
neous and distributed using CT-scan data as same as an earlier published study [21].
A linear relationship was formed among density and HU and expressed in the below
equation to get the heterogeneous bone density of cancellous part of the bone.
In this equation, 0 HU relates to the bone density of 0.022 g cm−3 , and 1646
corresponds to the highest cortical bone density of 1.73 g cm−3 [21]. After obtaining
the heterogeneous cancellous bone density, the following empirical relation was taken
to determine Young’s modulus (E in MPa) of cancellous part of the bone. Poisson’s
ratio was taken as 0.2.
Direct one-to-one validation was not likely since CT data were based on in vivo.
The current pelvis FE model was corroborated in earlier experimental analysis
accomplished on a composite hemipelvis [24]. The cortical bone was considered to
have, linear, elastic, isotropic and homogeneous material properties. Cortical bone
was considered to have an elastic modulus of 17 GPa, Poisson’s ratio of 0.3 and
density 1.73 gm cm−3 . The elastic modulus of the UHMWPE (polymer), metallic
(Co–Cr–Mo) and ceramic (alumina) acetabular cup is taken as 1.174 GPa, 210 GPa
and 350 GPa. The Poisson’s ratio for subsequent implant material is taken as 0.4, 0.3
and 0.26. The elastic modulus and Poisson’s ratio for cement mantle were taken as
2 GPa and 0.33, respectively [11, 23].
In this investigation, the loading cycle was considered as a normal walking cycle.
Eight load cases (static) were considered in this whole period of the normal walking
cycle [27]. 21 muscle forces, along with the hip joint reaction force, were taken
into account in applied loading boundary conditions for each static load case [21,
27]. The fixed constraint was set at the pubis and sacroiliac joint [14, 21]. In this
analysis, a sub-modelling technique was used to analyse the effect of different implant
materials on bone remodelling [26]. Bone remodelling process used in this analysis
Effect of Implant Materials on Bone Remodelling Around Cemented … 21
was similar to earlier published studies [26, 28]. Though, eight static load cases of a
normal walking cycle were taken in place of two load cases. Months or years were
represented as a result of bone remodelling, and this time was calculated by using
the bone adaptation rate, as explained in an earlier study [29]. Center position of the
spherical femoral head was set as the application point of hip joint reaction force.
The acetabular cup is fixed to the bone and bone cement (PMMA) for cemented
fixation. Fully bonded implant-cement and cement-bone interfacial condition were
taken in this analysis [17]. Therefore, frictionless contact was analysed at the interface
of the acetabular cup and femoral head [17] to characterise the good lubricating
articulating surface. A 0.5 mm redial allowance was given between the femoral
head and acetabular cup. Contact analysis was done using the asymmetric surface to
surface contact elements. Contact stiffness 100 N/mm2 and penetration tolerance 0.1
(factor) were suited in this analysis for convergence [ANSYS Manual]. A sensitivity
analysis was completed with the disparity of contact factors to realise the consequence
of these factors on contact results. These contact parameters were selected in a
system with a particular procedure that these factors do not have any impact on the
simulations results as well as computational time. In this study, all the simulation
and solutions were done using ANSYS v18 (ANSYS Inc. PA, USA).
3 Results
The results of bone remodelling around cemented acetabular cup were presented
in terms of months, and the adaptation rate of bone was used to calculate the time
(months) as shown in Fig. 2. The implant-induced variations in bone density due
to different implant materials in different locations are presented in Fig. 2. It has
been seen that implant material affects the bone density distribution because of
bone remodelling around the acetabular cup (Fig. 2). Bone resorption, fall in bone
density, was seen in cancellous bone from immediate post-operative condition to
after 6 months of bone remodelling in the cemented acetabular cup (Fig. 2). Bone
density decrement is not the same in all the locations (dependent on the location or
sections) as shown in Fig. 2.
There is less bone density variation or reduction among immediate post-operative
and after 6 months of bone remodelling considering polyethylene acetabular cup as
compared to metallic and ceramic acetabular cup. However, there are very fewer
changes (or almost similar) in bone density distribution among metallic and ceramic
acetabular cup (Fig. 2). Comparison of average bone density distribution for imme-
diate post-operative and after 6 months of bone remodelling with different implant
materials is represented in Fig. 3. The average bone density reduced from 0.446 to
22 A. Kumar et al.
Fig. 2 Bone density distribution after 6 months of bone remodelling: equilibrium in bone
remodelling for different implant materials (UHMWPE, Co–Cr–Mo and ceramic)
0.4 after the 6 months of bone remodelling for the polyethylene (UHMWPE) cup.
However, the reduction in bone density was more for metallic and ceramic acetab-
ular cup as compared to polyethylene acetabular cup. The main concern associated
with the polyethylene was the polyethylene wear particles, which was responsible
for osteolysis and loosening in cemented acetabular cup. The average bone density
reduced from 0.446 to 0.36 after the 6 months of bone remodelling is for the metallic
Effect of Implant Materials on Bone Remodelling Around Cemented … 23
(Co–Cr–Mo) acetabular cup. The average bone density reduced from 0.446 to 0.356
after the 6 months of bone remodelling is for the ceramic (alumina) acetabular cup
(Fig. 3).
4 Discussion
Subject-dedicated FE model of implanted pelvic bone was created based on the CT-
scan data. Three different implant materials (UHMWPE, metallic and ceramic) were
considered for this study. The dimensions of each acetabular cup are kept the same.
This investigation aims to predict the outcome of different implant material on bone
remodelling around the cemented acetabular cup. The result of the present study
showed that bone resorption and apposition were found around the acetabular cup
due to bone remodelling. Results of our study also indicated that implant materials
affect the bone density distribution around the cemented acetabular cup. Our study
highlighted the decrement in the bone density distribution in cemented acetabular
cup owing to different implant materials.
Less bone density reduction was observed for UHMWPE acetabular cup as
compared to metallic and ceramic acetabular cup. However, considering wear factor
ceramic acetabular cup would perform better as compared to the metallic acetab-
ular cup. The bone density distribution around the acetabular cup was found to be
almost similar in case of the ceramic and metallic cup. However, after achieving
the equilibrium, a little higher bone densities of 2–7% were detected in the central
part of the acetabulum in the superior region for the ceramic cup as likened to the
metallic cup as shown in Fig. 2. Average bone density reduction is less in polyethy-
lene (UHMWPE) acetabular cup as compared to the metallic and ceramic acetabular
cup (Fig. 3). Result of our study showed that there is approximately 10% reduction
in the bone density among the immediate post-operative and polyethylene acetabular
cup. However, both metallic and ceramic acetabular cup shows the approximately
19.6% reduction in the bone density (Fig. 3).
There is some limitation associated with this study. An adaptation rate of 129.6
gm mm−2 months was considered for the bone remodelling analysis, which was
similar to the femur; however, the actual adaptation rate for the pelvis is unknown.
The adaptation rate appears to be dependent on a particular location as well as might
also change from patient to patient. Bone remodelling is well known to be reliant on
loading boundary conditions. However, another limitation of our study is that we have
considered the eight static load cases of a normal walking cycle. Cancellous bone
was taken to be isotropic; however, pelvic trabecular bone was found to be not very
much anisotropic. In this investigation, we have taken only three implant materials
(UHMWPE, metallic and ceramic) to analyse the bone remodelling around cemented
acetabular cup. However, so many studies in the past literature have been concen-
trated on the advancement of new biomaterial in the arena of synthetic hip joints
considering mechanical strength, biocompatibility, wear resistance and mechanical
dependability as well as to reduce wear and osteolysis [30]. Therefore, a separate
24 A. Kumar et al.
study would be beneficial by considering the stainless steel, zirconia, titanium alloys,
silicon nitride, high crosslinked UHMWPE, zirconia-toughened alumina, metals
with specified coatings and alumina–zirconia composites as the new biomaterials
for different combination of implant materials to better understand the bone remod-
elling analysis of cemented acetabular cup owing to different implant materials. The
ceramic as an implant material has several advantages over metallic and polyethylene
implants; these are as follows:
• The heads of implant prepared from ceramics have high hardness as well as more
scratch counteraction than metals.
• The hard and ultra-smooth surface of ceramic implants also helps to diminish the
wear rate as compared to metallic and polyethylene implants.
• As per the earlier investigations, the wear rate of ceramic implants could be
reduced by 95% less than the metallic and other implants.
• Along with that the liners used in ceramic implants could also further reduce the
wear rate by a very significant quantity.
5 Conclusion
3D FE models of implanted pelvic bone have been really helpful to understand the
bone remodelling around cemented acetabular cup due to different implant materials
(UHMWPE, Co-Cr–Mo and ceramic). Results of this study indicated that ceramic
could be an excellent alternative to metallic implant in the cemented acetabular cup
since ceramic has better wear resistance property than metal. Ceramic also has less
surface fatigue, less wear rate, less osteolysis and harmless wear particle due to direct
solid contact and has long-term results as well as higher survival rate. Ceramic also
provide better clinical outcomes. Likewise, the implants made of ceramic have a
smaller coefficient of friction as compared to metallic implant materials. Ceramic
also has necessary qualities as suitable biocompatibility and simplicity of fabrication.
References
1. National Joint Registry (2016) National Joint Registry for England and Wales, 13th Annual
Report, NJR
2. Fang M, Noiseux N, Linson E, Cram P (2015) The Effect of advancing age on total joint
replacement outcomes. Geriatr Orthop Surg Rehabil 6(3):173–179
3. Lamvohee JMS, Ingle P, Cheah K, Dowell J, Mootanah R (2014) Total hip replacement: tensile
stress in bone cement is influenced by cement mantle thickness, acetabular size, bone quality,
and body mass index. J Comput Sci Syst Biol 7:72–78
Effect of Implant Materials on Bone Remodelling Around Cemented … 25
4. Carter DR, Vasu R, Harris WH (1982) Stress distributions in the acetabular region II: effects
of cement thickness and metal backing of the total hip acetabular component. J Biomech
17(3):165–170
5. Vasu R, Carter DR, Harris WH (1982) Stress distributions in the acetabular region-I. Before
and after total hip replacement. J Biomech 15(3):155–164
6. Crowninshield RD, Brand RA, Pedersen DRA (1983) Stress analysis of acetabular reconstruc-
tion in protrusio acetabuli. J Bone Joint Surg Am 65(4):495–499
7. Oh I, Sander TW, Treharne RW (1985) Total hip acetabular cup flange design and its effect on
cement fixation. Clin Orthop Relat Res 195:304–309
8. Revie I, Wallace M, Orr J (1994) The effect of PMMA thickness on thermal bone necrosis
around acetabular sockets. Proc Inst Mech Eng Part H J Eng Med 208:45–51
9. Mootanah R, Ingle P, Dowell J, Cheah K, Shelton JC (2000) Fixation of the acetabular cup
in cemented total hip replacement: improving the anchorage hole profile using finite element
method. Technol Health Care 8:343–355
10. Mootanah R, Ingle P, Cheah K, Dowell JK, Shelton JC (2004) Total hip replacement: results
of a postal survey of current practice on the cement fixation of the acetabular cup in the UK.
Hip Int 14:155–162
11. Mootanah R, Dowell JK, Cheah K, Ingle P, Shelton JC (2007) Configuration of anchorage
holes affects fixation of the acetabular component in cemented total hip replacement—a finite
element study. Comput Methods Biomech Biomed Eng 10:439–445
12. Kumar, Y.S., Pant, B., Darunkumar, Singh K.: Thickness effects on maximum von-Mises Stress
of a cement mantle in total hip replacement—a finite element study. J Appl Biomater Biomech
7:111–115
13. Coultrup OJ, Hunt C, Wroblewski BM, Taylor M (2010) Computational assessment of the
effect of polyethylene wear rate, mantle thickness, and porosity on the mechanical failure of
the acetabular cement mantle. J Orthop Res 28:565–570
14. Kumar A, Sanjay D, Mondal S, Ghosh R, Kumar R (2020) Influence of interface crack and
non-uniform cement thickness on mixed-mode stress intensity factor and prediction of interface
failure of cemented acetabular cup. Theor Appl Fract Mech 107:102524. https://doi.org/10.
1016/j.tafmec.2020.102524
15. Tong J, Zant NP, Wang JY, Heaton P, Hussell JG (2008) Fatigue in cemented acetabular
replacements. Int J Fatigue 8:1366–1375
16. Heaton-Adegbile P, Zant NP, Tong J (2006) In vitro fatigue behaviour of a cemented acetabular
reconstruction. J Biomech 39:2882–2886
17. Zant NP, Wong CK, Tong J (2007) fatigue failure in the cement mantle of a simplified acetabular
replacement model. Int J Fatigue 29:1245–1252
18. Shetty NR, Hamer AJ, Kerry RM, Stockley I, Eastell R, Wilkinson JM (2006) Bone remodelling
around a cemented polyethylene cup: a longitudinal densitometry study. J Bone Joint Surg Br
88(4):455–459
19. Ghosh R (2016) Assessment of failure of cemented polyethylene acetabular component due to
bone remodeling: a finite element study. J Orthop 13:140–147
20. Mann KA, Gupta S, Race A, Miller MA, Cleary RJ, Ayers DC (2004) Cement microcracks in
thin mantle regions after in vitro fatigue loading. J Arthrop 19:605–612
21. Ghosh R, Pal B, Ghosh D, Gupta S (2015) Finite element analysis of a hemi-pelvis: the effect
of inclusion of cartilage layer on acetabular stresses and strain. Comput Methods Biomech
Biomed Eng 18(7):697–710
22. Ghosh R, Mukherjee K, Gupta S (2013) Bone remodelling around uncemented acetabular
prostheses. Bone Joint J Br 95:184
23. Kumar A, Ghosh R, Kumar R (2020) Effects of interfacial crack and implant material on
mixed-mode stress intensity factor and prediction of interface failure of cemented acetabular
cup. J Biomed Mater Res Part B Appl Biomater 108:1844–1856. https://doi.org/10.1002/jbm.
b.34526
24. Ghosh R, Gupta S, Dickinson A, Browne M (2012) Experimental validation of finite element
models of intact and implanted composite hemi-pelvises using digital image correlation. Trans
ASME J Biomech Eng 134(8):081003 (1–9)
26 A. Kumar et al.
Abstract Lower back pain (LBP) is the common problem prevailing in most of the
individuals. The individuals suffering from LBP would be of any age group. It is
found common in subjects with higher body mass index (BMI), older age, inferior
posture and occupation, etc. This study analyses the influence of ageing and high BMI
on LBP by applying image processing techniques on X-ray images of individuals
of various age groups. In this work, the vital features corresponding to degenerative
discs are measured and analysed. The results indicated that the effects of ageing and
high BMI are influencing LBP.
Keywords Low back pain · Degenerative disc · Medical image processing · Body
mass index
1 Introduction
Lower back pain (LBP) is the common problem all along with the world. The lumbar
spine, or low back, consists of a complex network of interconnecting bones, joints,
nerves, tendons, ligaments, and muscles. Injury to any of these body parts can cause
LBP. It is affecting the people of all ages from children to elders. The pain in such
cases can be acute, sub-acute or chronic. There are few risk factors associated to
LBP, namely age, obesity, body posture, occupation, and BMI, etc. [1]. The various
imaging modalities are being used for identification of musculoskeletal pain. The
imaging modalities include conventional X-ray, digital radiology, ultrasonography
P. Praveen (B)
Dept. of E&IE, JSS Academy of Technical Education, Bengaluru, India
P. Praveen · M. S. Mallikarjunaswamy
Sri Jayachamarajendra College of Engineering, JSS Science & Technology University, Mysuru,
India
e-mail: [email protected]
S. Chandrashekhara
ChanRe Rheumatology & Immunology Center and Research, Bengaluru, India
2 Methodology
The segmental inter-vertebral disc position is essential to understand the LBP. The
images of individuals suffering with LBP were obtained from hospital and processed.
The dataset includes candidates of age 25–85 years which includes both the genders.
The study was reviewed by institutional ethical committee of ChanRe Rheumatology
and Immunology Centre and Research, Bengaluru. The study was informed, and
consent was obtained from participants. The image processing steps which involve
segmentation, feature extraction and analysis are depicted in Fig. 1.
There is a necessity of eliminating noise before the processing of such images
without blurring the edges of inter-vertebral discs. Therefore, the preprocessing oper-
ations were performed to eliminate the noise induced in such X-ray images. The
images were operated using median filter to eliminate noise and also preserve edges
of the inter-vertebral discs. The window size of median filter was 9 × 9, and in the
next step of processing, images were segmented. The images were further processed
to segment inter-vertebral discs. The acquired images of dataset were preprocessed
which involves region localization and enhancement techniques. Then, the images
were segmented for feature extraction of diagnostically pertinent region of inter-
ests, i.e. the herniated disc or degenerative disc which is vital in respective cases.
The images were further processed for the extraction of feature, i.e. inter-vertebral
distance. These extracted features were analysed in subjects of different age groups
and subjects with various BMI. Further, the statistical analysis was done. Finally, the
influence of ageing and BMI on LBP was analysed.
Influence of Ageing and High BMI on Lower Back Pain 29
Preprocessing
Segmentation
Feature Extraction
Analysis
3 Results
X-ray of LBP suffered individuals exhibited the degeneration of the disc. The degen-
erative disc was evident as the inter-vertebral disc distance in LBP suffered indi-
viduals seemed to be less compared to the inter-vertebral distance in the healthy
individuals. Figure 2a shows the X-ray of the lower back of LBP suffered individual.
The preprocessed images and the segmented images of lower back are shown in
Fig. 2b, c, respectively.
The feature extraction in lower back with disc degeneration is shown in Fig. 3.
The inter-vertebral disc (foramen) gets deteriorated by making the inter-vertebral
distance smaller due to the ageing, drying over time, prolong wear and tear. Hence,
this shows the symptoms of degenerative disc. Based on the statistical analysis,
obesity and higher BMI are found to be the massive causes for the herniated disc
as the weight of the upper portion of the body directly impacts on the lower back.
The cases of herniated disc were mostly found in obese individuals compared to the
normal or lower BMI individuals. The measurements are made in region of lateral
posterior, lateral middle and lateral anterior for each of the inter-vertebral disc. The
mean of measurement indicating the features of LBP in typical five cases is tabulated
in Table 1.
The results are grouped and correlated based on the individual age group and
BMI. The two groups of individuals were made on the basis of age, namely less than
50 years and equal to and greater than 50 years of age. Similarly, the subjects were
30 P. Praveen et al.
Fig. 2 a X-ray of the lower back, b preprocessed X-ray image of lower back and c segmented
image of X-ray of lower back
Degenerative disc
grouped on the basis of BMI, namely overweight whose BMI is less than 30 and
obese who’s BMI is equal or greater than 30. The results showed that the subjects
with the age group more than 50 years and the subjects of obese class were the major
victims of the LBP due to degenerative disc. This study infers that the influence of
the ageing and the BMI plays a vital role in causing the lower back pain.
References
1. Taimela S, Kujala UM, Salminen JJ, Viljanen T (1997) The prevalence of low back pain among
children and adolescents: a nationwide, cohort-based questionnaire survey in Finland. Spine
22:1132–1136
2. Grassi W, Filippucci E, Carotti M, Salaffi F (2003) Imaging modalities for identifying the origin
of regional musculoskeletal pain. Best Pract Res Clin Rheumatol 17(1):17–32
3. Darlow S, Forster BB, O’Sullivan K, O’Sullivan P (2017) It is time to stop causing harm within
appropriate imaging for low back pain. Br J Sports Med 5:14–415
4. Borenstein DG, Md (1996) Chronic low back pain. Rheum Dis Clin N Am 22:439–456
5. Hoy D, Bain C (2012) A systematic review of the global prevalence of low back pain. Am Coll
Rheumatol 64(6):2028–2037
6. Luoma K, MD, Riihimäki H (2000) Low back pain in relation to lumbar disc degeneration.
Spine 25(4):487–492
7. Podichetty VK (2007) The aging spine: the role of inflammatory mediators in intervertebral disc
degeneration. Cell Mol Biol (Noisy-le-grand) 53(5): 4–18
8. Dey N, Ashour AS, Althoupety AS (2017) Thermal imaging in medical science. Recent advances
in applied thermal imaging for industrial applications, IGI Global, 87–117
Design and Analysis of a Robotic Lizard
Using Five-Bar Mechanisms
Abstract Legged robots are being used to explore rough terrains as they are capable
of traversing gaps and obstacles. In this paper, a new mechanism is designed to
replicate a robotic lizard using integrated five-bar mechanisms. There are two five-
bar mechanisms from which two more are formed by connecting the links in a
particular order. The legs are attached to the links of the five-bar mechanism such
that, when the mechanism is actuated, they move the robot forward. Position analysis
using vector loop approach has been done for the mechanism. A prototype has been
built and controlled using servomotors to verify the robotic lizard mechanism.
1 Introduction
Mimicking animals for the purpose of solving human centered problems plays an
important role in the field of robotics. Among these, four legged creatures are gaining
attention in the past few decades [1]. Among them, MEMS-based robots are gaining
importance [2]. Miniaturized robots can be used for surveillance and locomotion in
cramped spaces.
A lizard is a commonly seen reptile whose mechanics have been studied in detail
in the following literature [3]. It was found that the walking and trotting gaits [4]
exhibited by the lizards were similar to that of the mammals. The biomechanics and
kinematics of the sprawling pattern exhibited by the lizards were studied extensively
V. S. Rajashekhar (B)
Tentacles Robotic Foundation, Kanchipuram, Tamil Nadu 603202, India
C. K. Dinakar Raj · S. Vishwesh · E. Selva Perumal · M. Nirmal Kumar
Department of Mechanical Engineering, Adhiparasakthi Engineering College, Melmaruvathur,
Kanchipuram, Tamil Nadu 603319, India
in [5]. The sprawling gait has be experimentally shown on a dynamic model in [6].
It is observed that the compliant and flexible trunk of the lizards help in reducing the
peak power [7].
The lizard-based bio-inspired robots have been made in the past, such as a water
running robot [8], a leg mechanism [9], and a quadrupeds [10, 11]. Motion analyis
[12], motion planning [13], and gait planning [14] based on kinematics of the Gecko
have been done on robotic lizards. A seven degrees of freedom robotic Gecko has
been developed for the experimental verification of the kinematic analysis [15].
Although there are works that exist in the literature that try to replicate a real
lizard, they have not been able to mimic them closely. The novelty of this paper lies
in the core mechanism design and its abilty to replicate the gaits of a real lizard.
The real lizard that served as a source of inspiration is shown in Fig. 1a, and the
fabricated robotic lizard is shown in Fig. 1b. Topological design has been done using
the method mentioned in [16]. The position analysis has been done using the vector
loop approach. Finally, the prototype of the robotic lizard has been shown exhibiting
the walking gait.
Fig. 1 a The real lizard that served as an inspiration. b The bio-inspired robotic lizard introduced
in this work
Design and Analysis of a Robotic Lizard … 35
The design of robotic lizards using mechanisms like Watt-I planar linkage mechanism
[17] and CPG-driven modular robots [18] have been done in the past. In this work,
the development of the robotic lizard mechanism using two active and two passive
five-bar mechanisms is done. The degree of freedom analysis is done to find out the
number of degrees of freedom of the mechanism and also to find out whether the
right driving links have been chosen for the mechanism. This is a eight step process,
and it is done using the method given in [16].
m
v
F= fi − ξLj (1)
i=1 j=1
v
j
ξLj = dim.((∩i=1 Mbi ) ∪ Mbj+1 ) (2)
j=1
where,
F—Degree of freedom of parallel manipulator (PM).
fi —Degree of freedom of the ith joint.
m—Total number of joints in the parallel manipulator.
v—Total number of independent loops in the mechanism, where v = m − n + 1.
n—Total number of links in the mechanism.
ξLj —Total number of independent equations of the jth loop.
j
∩i=1 Mbi —Position and orientation characteristic (POC) set generated by the sub-PM
formed by the former j branches.
Mbj+1 —POC set generated by the end link of j + 1 sub-chains.
Calculating the number of independent loops putting n = 13 and m = 16, we get
v = 4. The eight steps involved in the calculation of the degrees of freedom of the
mechanism are as follows.
1. The topological structure is mentioned symbolically here.
Branches:
SOC1 (−R1 ||R2 ||R3 −); SOC2 (−R4 ||R5 −); SOC3 (−R1 ||R7 ||R8 −);
SOC4 (−R11 ||R10 −); SOC5 (−R5 ||R7 ||R9 −); SOC6 (−R12 ||R13 −);
SOC7 (−R11 ||R14 ||R16 −); SOC8 (−R12 ||R15 −)
Platforms:
Fixed platform: R1 , R5 , R11 , R12
Moving platform: The rest are all moving platforms
2. An arbitrary point o is chosen on the moving platform.
t2 t2
ξL1 = dim(Mb1 ∪ Mb2 ) = dim ∪
r 1 (||R1 ) r 1 (||R4 )
t2
= dim =3
r (|| (R1 , R4 ))
1
Similarly, it can be done for the other three loops, and the value of the number
of independent displacement equations is found to be the same in each case.
5. Calculating the DOF of the mechanism
m v
F= fi − ξLj = 16 − (3 + 3 + 3 + 3) = 4
i=1 j=1
6. Finding the inactive pairs.
Based on the calculations done in [16], similar steps were followed. It was found
that there is no inactive pair in the mechanism except for the tail which is regarded
as passive linkages.
7. Determining the position and orientation characteristic set of the robotic lizard
mechanism.
Based on the formula given in [16],
MPa = Mb1 ∩ Mb2
2
t2 t2 t2 t
MPa = ∩ ∩ ... ∩ = 1
r 1 (||R1 ) r 1 (||R4 ) r 1 (||R12 ) r
The degree of freedom of the mechanism is four, and the dimension of the above
POC set is three. Hence, the module has two translational and one rotational
degree of freedom. The discrepancy in the values is due to the fact that there is
one more rotational degree of freedom on the same plane which the position and
orientation characteristic analysis has ignored. This is because both the rotational
degrees of freedom are the same.
8. Select driving pairs
The joints R1 , R5 , R11 and R12 are chosen to be the driving pairs, and hence, they
are fixed. Calculating the degrees of freedom of the new mechanism,
m v
F∗ = fi − ξLj = 12 − (3 + 3 + 3 + 3) = 0
i=1 j=1
∗
Since F = 0, the joints R1 , R5 , R11 and R12 can be used as driving pairs simul-
taneously.
Thus, the topological design of the mechanism has been done.
Design and Analysis of a Robotic Lizard … 37
The robotic lizard mechanism is as shown in Fig. 2c. The vector loop method adopted
from [19] is used for the position analysis of the robotic lizard. There are five vector
loop equations framed, one for the head, two for the body, one for the trunk, and
one for the tail. These equations are solved by keeping the known angles on one side
and the unknown angles on the other side. The position of each part of the body is
known by substituting the known four servomotor angles. These equations will be
presented in detail in this section.
The vector loop equation of the head which forms a five-bar mechanism in the robotic
lizard is written in Eq. 3. It is shown in Fig. 2b.
R2 + R3 − R4 − R5 − R1 = 0 (3)
Collecting the sin and cos terms together, Eqs. 4 and 5 are obtained.
Solving for the θ3 and θ4 by rearranging Eqs. 4 and 5, the following are obtained.
√
B2 − 4AC
−B −
θ4 = 2 arctan (6)
2A
√
−E + E 2 + 4DF
θ3 = 2 arctan (7)
2D
Fig. 2 a The topological design of the robotic lizard mechanism. b The four five-bar mechanisms
that form the robotic lizard. It is used in position analysis. c The prototype of the robotic lizard. d
The graphical user interface developed for the robotic lizard
38 V. S. Rajashekhar et al.
where,
K1 = L5 sin(θ5 ) − L2 sin(θ2 );
K2 = L5 cos(θ5 ) − L2 cos(θ2 ) + L1
K3 = −L5 sin(θ5 ) + L2 sin(θ2 );
K4 = −L5 cos(θ5 ) + L2 cos(θ2 ) + L1
A = L4 −2 K2 L4−L 3 +K2 +K1
2 2 2 2
2
;
B = 2 K1 L4 ;
C = L4 +2 K2 L4 −L 3 +K2 +K1
2 2 2 2
2
D = L4 −L3 +2 K24 L3 −K4 −K3 ;
2 2 2 2
E = 2 K3 L3 ;
F = L4 −L3 −2 K24 L3 −K4 −K3
2 2 2 2
Similarly, for the rest of the three five-bar mechanisms, the vector loop equations are
as follows. For the body side 1, body side 2, and tail,
R6 + R7 − R8 − R9 − R10 = 0 (8)
From the above equations, the values of θ can be found that are similar to Eqs. 6
and 7. Thus, the position analysis has been done for the mechanism.
The four joints are actuated in a particular order so that the robotic lizard mechanism
exhibits the forward walking gait. The equations governing the coordinates of the
linkages of the head are as follows.
x1 = 0
y1 = 0
x2 = L2 cos(θ2 )
y2 = L2 sin(θ2 )
x3 = x2 + L3 cos(θ2 + θ3 )
y3 = y2 + L3 sin(θ2 + θ3 )
x5 = L5 cos(θ5 )
y5 = L5 sin(θ5 )
x4 = x5 + L4 cos(θ5 + θ4 )
y4 = y5 + L4 sin(θ5 + θ4 ) (11)
In the similar way, it can be derived for the left and right side of the body and the tail
of the mechanism.
Design and Analysis of a Robotic Lizard … 39
The workspace of the robot is plotted by writing a Python Code. The algorithm used
for plotting the workspace of the robot is as follows.
Fig. 3 The workspace of the a head, b tail, c body left, d body right
40 V. S. Rajashekhar et al.
Figure 3a–d shows the reachable workspace of the head, tail, body left, and body
right of the robotic lizard mechanism.
The links of the mechanism were fabricated using the balsa wood in a CNC router.
The mechanism was actuated using the four servomotors in a particular order. The
control used here was Arduino UNO, and processing IDE was used as a graphical user
interface (GUI) to control the robot. The GUI is as shown in Fig. 2d. The robotic lizard
was able to move on a flat surface and exhibit the walking gait as shown in Fig. 4.
5 Conclusions
In this work, a new robotic lizard mechanism was developed. The topological design
of the mechanism was done. Then, the position analysis of the mechanism was done
to find the various angles between the links. Finally, a prototype was made and
controlled using servomotors. A graphical user interface was used to operate the
robotic lizard. The future work would be to implement deep reinforcement learning
to obtain a stable gait in unknown environments.
Design and Analysis of a Robotic Lizard … 41
References
1 Introduction
In a nuclear fuel fabrication facility, various process operations are carried out in
specially designed glove boxes as shown in Fig. 1. These glove boxes are dynamic
containment system which acts as a barrier between operator and the radiological
material inside it. Glove boxes are always maintained under dynamic negative pres-
sure to prevent outward leakage of radioactivity. They are a leak-tight system with
a permissible leak rate of 0.05% GB volume per hour as given by relevant stan-
dards [1–3]and are designed as class I component as per the ASME section III NB
guidelines [4]. Process equipment is kept inside the glove boxes to carry out various
fabrication steps. Normally, material transfer between one glove box to another in the
fabrication line is carried out manually by the operators. In case of reprocessed feed
material coming out of fast reactors, radiological doses to the operator are expected
to be higher. Hence, complete automation of material transfer is the need of the hour.
Initially, a prototype system has been planned to connect few sample glove boxes
using developed material handling system. The next section shall highlight design
philosophy involved in the development of automatic material transfer system.
To develop any system for functioning inside a glove box is a challenging task. There
are various technical considerations required before starting the design. Following
are the important considerations for the design of automatic material transfer system
for glove box applications:
1. Modular Design—In case of breakdown of any component, user of the system
shall be able to take out the component, after decontamination, from the bag
out port of the glove box and replace it with a new component. Size of each
Development of an Automated Material Handling System Inside … 45
individual component shall be such that it can be bagged out (removed) from
the glove box using standard Ø 0.3 m bag-out port. Hence, all the component
modules shall be designed based on this size constraint.
2. Standardization—All the important components which may require frequent
maintenance or replacement shall be standardized. For example, rollers in a
roller conveyor, drive motors, fasteners, and supporting channels, etc. This helps
in reducing maintenance downtime of the system.
3. Remote In Situ Maintenance—A system inside a glove box has a very limited
access from the outside. Hence, provision of remote maintenance is required.
Remote maintenance tools shall be developed to fulfil the requirement. In situ
maintenance of conveyor system is possible by using fast changing mounting
fixtures for the rollers. Similarly, quick changing couplers for motors and
drive mechanisms shall be preferred. Quick changing tools for the conveyor
mechanism shall be available during maintenance.
4. Radiation Resistance Components—Environment inside a glove box will have
high radiation field. This requires two prong design philosophy. First design
objective shall be to keep electronic components outside the shielded glove box
as far as possible. If it is not possible, then use radiation-hardened cables and
electronic components inside the glove box.
5. Fail-safe Design—The system design shall be fail safe to avoid any radiological
consequences. The system shall be simple and rugged in construction.
Considering the above-mentioned design objectives, a prototype material handling
system is being designed. The system is being developed considering connection of
two glove boxes with a material transfer tunnel. Once material transfer system is
successfully demonstrated in the prototype; the same module would be scaled up to
cover the entire fabrication line. Next section describes the details of the proposed
system.
3 System Details
In the proposed system, roller type of conveyors is planned to be used for the material
transfer in a tray, as shown in the conceptual drawings from Figs. 2, 3 and 4. As can be
seen in the drawings two bigger glove boxes (Type-VI) are connected with a smaller
glove box (Type-I) with the help of material transfer tunnels. The conveyor passing
through transfer tunnel has been designed to be modular in construction, so that in
case of a radiological emergency, this can be retracted back and tunnel can be closed
using a bunk. Hence, for a module of three interconnected glove boxes, five parts of
conveyor will be connected together to make a complete conveyor system. Smooth
transition of material placed in a carrier tray from one conveyor segment to the other
has to be ensured. In case of power failure, it is expected that operator shall be able
to move the material to a predetermined location. A provision to switch to manual
handling mode using a clutch mechanism has been provided.
46 A. Saraswat and P. S. Somayajulu
Fig. 2 Conceptual scheme of the conveyor system for material transfer inside glove boxes
Roller
Conveyor
Carrier
Tray
Type VI Material
Type I
Glove boxes transfer
Glove box
tunnels
Omni directional
transfer units
motion to the pallet/carrier tray in varied directions making angle of 0°–180° from
the line of motion.
• Carrier Trays: A number of carrier trays have been designed to carry material
containers of different shapes. Total payload to be carried is 20 kg. These trays
shall be provided with pre-defined grooves to position the material containers
accurately inside them as shown in Fig. 4. These trays have been standardized.
• Drive system: All the independent roller conveyor modules have been powered
with the motors. The system requires precise movements and positioning of
the material located in a transfer tray, and hence, motors with VFD have been
provided. Independent drives have been provided to the omnidirectional transfer
units and each individual conveyor module. It shall be noted that in case of any
power failure or manual mode of operation, provision for automatic disengage-
ment of the drive unit form the conveyor has been designed. It can be possible by
using motors with the clutch arrangement. Bidirectional motion of the conveyor
shall be possible with the drive unit.
• Controller: All the drives and actuators required in the system shall be controlled
through a PLC. It shall be possible to generate random and simultaneous sequence
of material/tray from one glove box to the other. Provision for auto and manual
mode of operation has been provided. An HMI has been provided to control the
conveyor system remotely.
4 Conclusions
A material handling system for the transfer of feed materials inside the glove
box system has been developed. It has been designed considering all the possible
constraints available for the operation and maintenance inside a containment system.
The developed system shall be initially tested in a normal environment. Once the
design features of automation and remote maintenance are demonstrated, then the
system shall be installed in a radioactive environment.
References
Rakesha Chandra Dash, Dipak Kumar Maiti, and Bhrigu Nath Singh
1 Introduction
Flow-induced vibrations can be found in structures like airplanes and power trans-
mission lines. A special kind of flow-induced vibration is transverse galloping. It
is induced in prismatic cylinders. A certain threshold value of wind speed causes
galloping to start, and it is called the onset speed of galloping. During the past
few years, many studies have been done to convert surrounding vibrations to elec-
trical energy by using piezoelectric material. This phenomenon is first studied and
explained by Hartog [1]. Aerodynamic forces are described using quasi-steady
hypothesis in his study. This approximation is valid when characteristic time scale
of the airflow is much lower than the characteristic timescale of the structure oscil-
lation. The effect of changing the geometry of the cross section of bluff body on
the amplitude of the structure is studied by Novak [2]. A. Barrero-Gil et al. used
theoretical calculation to extract energy from transverse galloping [3]. They used a
lumped-parameter model to describe the mechanical behavior of the system. A poly-
nomial representation is used based on quasi-steady approximation. Abdelkefi et al.
showed that for relatively high Reynolds number, galloping force can be approxi-
mately represented by a cubic polynomial function of the angle of attack [4]. Effect
of D-shaped cross section geometry on energy harvesting is theoretically and experi-
mentally investigated by Sirohi et al. [5]. A linear and nonlinear analysis is performed
by Yan et al. to determine the effect of the electrical load resistance on the onset speed
of galloping and level of the harvested power [6]. The nonlinear characterization of
energy harvester system applied with combination of vibratory base excitations and
aerodynamic loading is studied by Yan et al. [7]. Other than energy harvesting, Mout-
lana MK et al. (2015) used piezoelectric patches to control cantilever beam with end
mass [8].
In this work, the effect of linear and nonlinear representation of Euler–Bernoulli’s
beam theory on amplitude of vibration and voltage developed is discussed. A
nonlinear distributed parameter model is derived for bimorph configuration of piezo-
electric patches following Bibo et al. [9]. Quasi-steady theory is considered to model
aerodynamic forces. Effects of coefficients in representing the aerodynamic force on
energy harvesting are discussed.
2 Theoretical Background
See Fig. 1.
The dynamics of a beam can be completely described by using the longitu-
dinal displacement, u(s, t), and transversal displacement, w(s, t), following Euler’s
theory as depicted in Fig. 2. Nonlinearity in strain is considered for axial direction
w2
∈x = ∈0 + ∈r = u + − zw (1)
2
Two Cartesian coordinate systems are used in this expression to describe a beam
element at different instant of deformation. Global coordinate system is (x, y, z) ,
and local system is (x, y, z).
Following the linear constitutive relationships, axial σx stress and axial strain ∈x
can be given by
σxp = Y p ∈xp −d31 E 3 (3)
where b and p superscripts stand for structural and piezoelectric layers, respectively.
Ohms law can be used to relate voltage and current via the relation V (t) =
R Q̇ R (t), where Q̇ R (t) is current passing through the resistive load R.
3 Mathematical Formulation
t2
δL + δwext dt = 0 (4)
t1
The total potential energy of the system, U, can be represented by the expression
1 b b 1
U= σx ∈x +σxp ∈xp dv − E 3 D3 dv (7)
2 v 2 v
where v is total domain, and D3 is the electric displacement and can be expressed by
the following linear piezoelectric relation
where e33 is the permittivity of the piezoelectric element at constant strain. Putting
all relations in Eq. (7) and integrating over the cross section area of each layer, we
obtain
L b
1 2 2 2
2 1 2 2 1 2
U= YI w w +w + θ w + w w R Q̇ R ds + C p R Q̇ R
2 2 2
0
(9)
where
Nonlinear Modeling and Stability Analysis of Piezoelectric … 53
3 3
2 p hs hs 2b b h s 3
YI = bY hp + − + Y
3 2 2 3 2
e33 W p L p d31 Y p b 2
Cp = θ= h p + h p hs
tp 2h p
Non-conservative Forces
There are three types of work done due to non-conservative forces considered in this
system, namely lateral aerodynamic force, Fy , mechanical viscous damping, Ca , and
due to resistive load, R. Non-conservative virtual work can be expressed as
L b
δW ext
= Fy δw| L b + Fy L c δw Lb
− (Ca ẇδw)ds − R Q̇ R (t)δ Q R (10)
0
1 ẇ| L b
Fy = ρa C Fy (α)Dt L t U∞
2
C Fy = a1 α + a3 α 3 + a5 α 5 α = w L b +
2 U∞
where ρa , U∞ , C Fy , and L t are the air density, wind speed, coefficient of lift, and
length of bluff body, respectively. Substituting Eqs. (5), (9), and (10) in (4), then
taking virtual variations with respect to w and Q R , and setting the coefficients of δw
and δ Q R to zero, the following equations and boundary conditions can be obtained:
M(s)ẅ + ca ẇ + Y I (s)w + w Y I (s)w w
⎡ ⎡ s ⎤ ⎤
! s
1 2 1
V (t) + ⎣w M(s)⎣ ẅ ds ⎦ds ⎦
2
+ θ (s) 1 + w
2 2
Lb 0
= Fy δ(s − L b ) − L c δ (s − L b ) (11)
⎡L ⎤
b
1 ∂⎣ 1
C p V̇ (t) + V (t) = θ (s)w 1 + w2 ds ⎦ (12)
R ∂t 2
0
where δ(s) is Dirac delta function. The boundary conditions associated with the
above expression are given by
w|s=0 = 0 w s=0 = 0
54 R. C. Dash et al.
!
Dt Dt2
At s = L b Y I (s)w = −Mt ẅ − It + Mt ẅ .
2 4
Reduced-Order Model
We express w(s, t), representing the lateral vibrations of the beam, in the form of a
convergent series given by
∞
"
w(s, t) = φi (s)qi (t) (13)
i=1
where qi (t) is the generalized time-dependent coordinates, and φi (s) are the set of
mass-normalized orthonormal eigenfunctions. These eigenfunction also represents
the mode shapes of defined configuration of beam with bluff body attached at the
end.
Substitution of Eq. (13) into (11) is done, and then, integration is done over
the length of the beam by multiplying φn (s) on both the sides. The orthonormality
properties of the eigenfunctions are used, and the following linearly decoupled set
of nonlinear ordinary differential equations is formed for value of n = 1, 2, 3, ….
∞
"
q̈n + 2ζn ωn q̇n + ωn2 qn + θn∗ V (t) + Ani j qi q j V (t)
i, j
∞
" ∞
"
+ Bni jk qi q̈ j qk + 2q̇i q̇k + q j q̈k + Cni jk qi q j qk = Fn (q, q̇) (14)
i, j,k i, j,k
where
L b L b
ωn2 = Y I φn ds θn∗ = φθ ds
0 0
L b
1
Ani j = φn θ φi φ j ds
2
0
⎡ ⎛ s ⎞ ⎤
L b s L b
φn ⎣
Bni jk = φi M ⎝ ⎠ ⎦
φ j φk ds ds ds Cni jk = φn φi Y I φ j φk ds
2
0 L 0 0
1 Dt
Fn = φn (L b ) + φn (L b ) ρa Dt L t U∞
2
a1 α ∗ + a3 α ∗3 + a5 α ∗5
2 2
Nonlinear Modeling and Stability Analysis of Piezoelectric … 55
∞
∞
" 1 "
∗
α = φi (L b )qi + φi (L b )q̇i
i
U∞ i
Equations 14 and 15 are the coupled governing equations of the system. These
two equations are solved by using fourth-order Runge–kutta method.
Geometric and physical parameters used to calculate the results are given in
Table 1. Value of R is taken to be 172 × 103 [10].
Phase portrait for galloping-based energy harvester before and after onset speed is
plotted in Fig. 5 and 6, respectively. Only linear term in the polynomial expansion
of lateral force is used. Before onset speed given any initial condition, the velocity
and amplitude of vibration go to zero, but after onset speed amplitude and velocity
drastically increases. Initial conditions are taken in order of 10−7 . The amplitude of
vibration remains constant for a particular speed in practice. So, linear representation
of aerodynamic force gives erroneous result. Variation of tip deflection with wind
speed for a linear system is shown in Fig. 4. Taking cubic term also in consideration
variation of tip deflection with wind speed is represented by Fig. 3. System bifurcates
at wind speed 4 m/s and goes into limit cycle oscillation. This type of bifurcation
is called Hopf bifurcation. Figure 8 depicts the variation of amplitude of vibration
of the system with wind speed for a fifth-order representation of aerodynamic force.
The effect of linear and nonlinear representation of Euler–Bernoulli beam theory is
also shown. There is a sudden jump in amplitude due to the present of subcritical
Hopf bifurcation. There is a hysteresis present due to this instability. Variation of
voltage with wind speed is given in Fig. 7.
Electrical power can be calculated using the formula:
V2
P=
R
Therefore, electrical power corresponding to wind speed 7 m/s can be calculated
to be 0.36 mW.
56
Amplitudde (m)
0.03
0.02
0.01
0.00
0 1 2 3 4 5 6
U (m/s)
aerodynamic damping
system
41
6.0x10
Amplitude (m)
41
4.0x10
41
2.0x10
0.0
0 1 2 3 4 5
U (m/s)
5 Conclusions
Voltage (V)
4
0
0 1 2 3 4 5 6 7 8
U (m/s)
0.04
0.03
0.02
0.01
0.00
0 1 2 3 4 5 6 7
U (m/s)
References
6. Yan Z, Abdelkefi A, Hajj MR (2014) Piezoelectric energy harvesting from hybrid vibrations.
Smart Mater Struct 23(2)
7. Yan Z, Abdelkefi A (2014) Nonlinear characterization of concurrent energy harvesting from
galloping and base excitations. Nonlinear Dyn 1171–1189
8. Adali S (2015) Vibration of a cantilever beam with extended tip mass and axial load subject to
piezoelectric control. Res Dev J 31(1):60–65
9. Bibo A, Daqaq MF (2018) Modeling and characterization of a piezoelectric energy harvester
under combined aerodynamic and base excitations. J Vib Acoust 137:1–12
10. Javed U, Abdelkefi A (2017) Impacts of the aerodynamic force representation on the stability
and performance of a galloping-based energy harvester. J Sound Vib 400:213–226
Optimization of Surface Roughness
of Laser Trepanned Hole in ZTA Plate
1 Introduction
holes above 1.2 mm diameter in difficult-to-cut materials [2]. Study of different pulse
duration-based laser drilling process for different structural ceramics has reviewed
by Wang et al. [8]. Bharatish et al. [9] studied the effect of hole diameter, laser
power, pulse frequency and scanning speed on hole taper (HT), hole circularity and
heat-affected zone (HAZ). They found laser power and hole diameter as significant
parameters on hole circularity, while laser frequency and laser power on HAZ and
HT. Adelmann et al. [10] performed laser drilling on aluminum nitride and alumina
ceramics. They reported that lower laser focus position yields lesser HT. Saini et al.
[11] done laser drilling on YSZ ceramic. They determined optimum hole quality
characteristics using artificial neural network. Saini et al. [12] studied recast layer,
microhardness and microcrack width in laser trepanned hole in ZTA. They reported
that trepanning speed and air gas pressure were significant factors for considered
hole characteristics. Zhang et al. [13] have done laser trepan drilling on silicon
nitride ceramic. They obtained high RLT at high assist gas pressure. Murray and
Tyrer [14] have done laser drilling on partially stabilized tetragonal zirconia (PSTZ)
ceramic. They investigated the effect of heating temperature on recast layer microc-
racking. They reported that heat treatment before and after drilling of PSTZ reduced
the formation of microcracks. Preliminary literature review revealed that most of
the researchers have studied hole diameter, circularity, recast layer, HAZ and HT of
laser drilled hole. But study of laser trepanned hole surface quality seldom is found.
The present research paper investigates the optimum set of control factors for surface
roughness in laser trepan drilled hole of 6.0 mm thick ZTA. Regression model has
been developed from experimental observations. Further, developed model has been
used as objective function to elect the optimum set of control factors using artifi-
cial intelligence tool (i.e., genetic algorithm) within range of control factors. The
confirmed optimum results have been compared with best experimental observation.
2 Experimentation
Laser trepan drilling experiments have been conducted on ZTA (6.0 mm thick) using
pulsed Nd:YAG laser. Table 1 shows the range and levels of selected control factors.
Figure 1 shows the measurement procedure of drilled hole surface roughness. The
observed experimental values are drawn by Fig. 2 [15].
Fig. 1 Representation of
surface roughness
measurement of laser
trepanned hole
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Number of experiment
To realize the influence of LTD process on drilled hole surface quality characteristic,
a second-order regression model is developed using response surface method, and it
is represented by Eq. (1) [16].
u
u
u
u
Y K = β0 + βi X i + βii X i2 + βi j X i X j (1)
i=1 i=1 i=1 j=u+1
where Y K is the hole characteristic, βs and β0 are regression coefficient and constant,
and u is number of control factors, while X i = linear, X i2 = square, andX i X j =
interaction of control factors.
In the developed regression model, back elimination approach using p-value has
used for eliminating the insignificant terms. The p-value of significant control factors
64 S. K. Saini et al.
is considered less than 0.05 [16]. Equation (2) represents the developed regression
models for surface roughness.
Fig. 4 SEM images of surface roughness at a non-optimal parameter level, b optimum parameter
level
66 S. K. Saini et al.
4 Conclusions
Trepan drilling experiments have been conducted on 6.0 mm thick ZTA using pulsed
Nd:YAG laser, successfully. Developed second-order regression models found reli-
able at 95% confidence level. Drilled hole surface roughness has been reduced by
17.5% at pulse frequency = 12.5 Hz, pulse width = 4 ms, trepanning speed =
34 mm/min and assist gas pressure = 10 kg/cm2 . The observed SEM images of
drilled hole surface at non-optimum and optimum control factor levels also confirm
the above improvement. The confirmation test result shows errors 4.7%.
Acknowledgements Financial support for present work was granted by Aeronautics R & D Board,
New Delhi, Government of India through the sanction order no. AR&DB/01/2031751/M/I. The
authors also thankfully acknowledge the financial assistance provided by TEQIP-III.
References
1. Wang J, Stevens R (1989) Review zirconia toughened alumina (ZTA) ceramics. J Mater Sci
24:3421–3440
2. Saini SK, Dubey AK, Upadhyay BN, Choubey A (2018) Study of hole characteristics in laser
Trepan drilling of ZTA. Opt Laser Technol 103:330–339
3. Jain VK (2007) Advanced machining processes. Allied Publishers Private Limited Delhi
4. Tuersley IP, Jawaid A, Pashby IR (1994) Review: various methods of machining advanced
ceramic materials. J Mater Process Technol 42:377–390
5. El-Hofy H (2005) Advanced machining processes. McGraw Hill Mechanical Engineering
Series, New York
6. Samant AN, Dahotre NB (2009) Laser machining of structural ceramics—a review. J Eur
Ceram Soc 29:969–993
7. Dubey AK, Yadava V (2008) Laser beam machining—a review. Int J Mach Tools Manuf
48:609–628
8. Wang H, Lin HT, Wang C, Zheng L, Hu X (2017) Laser drilling of structural ceramics-a review.
J Eur Ceram Soc 37:1157–1173
9. Bharatish A, Murthy HNN, Anand B, Madhusoodana CD, Praveena GS, Krishna M (2013)
Characterization of hole circularity and heat affected zone in pulsed CO2 laser drilling of
alumina ceramics. Opt Laser Technol 53:22–32
10. Adelmann B, Hellmann R (2015) Rapid micro hole laser drilling in ceramic substrates using
single mode fiber laser. J Mater Process Technol 221:80–86
11. Saini SK, Dubey AK, Pant P, Upadhyay BN, Choubey A (2017) Study of laser drilled hole
quality of Yttria Stabilized Zirconia. Lasers Manuf Mater Process 4:121–135
12. Saini SK, Dubey AK (2019) Study of material characteristics in laser trepan drilling of ZTA.
J Manuf Process 44:349–358
13. Zhang J, Long Y, Liao S, Lin HT, Wang C (2017) Effect of laser scanning speed on geometrical
features of Nd:YAG laser machined hole in thin silicon nitride sub-strate. Ceram Int 43:2938–
2942
14. Murray AJ, Tyrer JR (1999) Nd:YAG laser cutting and drilling of PSTZ-influence of substrate
heating temperature on recast layer microcracking. J Laser Appl 1:128–135
15. Saini SK, Dubey AK, Upadhyay BN (2019) Study and optimization of recast layer thickness
and surface quality in laser trepan drilled ZTA. Int J Adv Manuf Technol https://doi.org/10.
1007/s00170-019-03704-3
Optimization of Surface Roughness of Laser Trepanned Hole … 67
1 Introduction
The study on modelling and balancing of two-wheeled robots (segway) has gained
momentum over the last decades worldwide in robotics domains. This is due to
the inherent unstable dynamics and nonlinearity of such systems. The principle of
inverted pendulum system has been employed into segway, which is a self-balancing
human transporter and looks like a high technological scooter (see Fig. 1a). The
configuration of segway consists of two wheels and an inverted pendulum as a handle
attached to the vehicle body [1–3]. To move forward or backward, the rider just has
to lean the pendulum forward or backward. The segway consists of a control system,
series of sensors and driving system. Segway is equipped with two sensors, i.e. a
(a) (b)
Fig. 1 a Concept of balancing a broomstick on index finger applied to segway and b schema of a
segway
gyroscopic sensor and an accelerometer for measuring the pitch angle and pitch
angular velocity. The system is stabilized with the help of a feedback gain and
delay [1]. The system is controlled by feedback linearization technique using LQR
controller [2]. Two types of control techniques are used to control the two-wheeled
robotic machine [2]. Gyroscopic sensor with Kalman filter is used to measure the
pitch angle of the vehicle. The dynamics of the model are derived with the help of
Lagrangian mechanics [3]. Control algorithms and signal processing are distributed
among the three microprocessors, one for each drive and one is responsible for the
stabilizing control [4]. One is the proportional derivative control and other one is
the fuzzy logic control to balance the vehicle [5]. By using the liberalized models,
various control laws have been applied to the wheeled inverted pendulum [6, 7].
The two-wheeled inverted pendulum system is implemented and designed with the
compensation of the friction.
The stabilization concept of inverted pendulum system has been employed in this
paper to form a segway. The simulation results for straight line motion as well as
turning motion are also presented. The segway moves forward if the handle bar is
tilted in the front and moves backward if the handle bar is tilted in the rear. For the
left turn, the bar is turned towards left and for the right turn, and the bar is tilted
towards right. The motivation of this paper is the validation through modelling and
simulation of a non-model-based control system design, i.e. energy-based control
technique is used.
Modelling and simulation play a vital role in the area of engineering design
process. The accurate mathematical description of any system provides an easy way
to understand the system in very quick and accurate manner. The data and infor-
mation about any system can be obtained from modelling and simulation before
implementing it in real-life situation. Bond graph (BG) model is an efficient tool
for modelling engineering systems. Bond graph modelling technique is a way to
describe the dynamic behaviour of a physical system graphically. The technique is
very similar to the block diagram and signal flow graph, but the major difference is
that the bonds in the bond graph represent bi-directional energy exchange [8, 9]. The
PI Control-Based Modelling of Segway Using Bond Graph 71
The modelling of segway is done by the combination of five sub-models, i.e. vehicle
body, wheel, handle, electric motor and wheel bushing system. Figure 2 represents
the whole system, segway in word bond graph form where the bonds with two
parallel lines denote multi-bonds between two modules. The bushing system is used
to connect the left and right wheels to the main body of the vehicle. The electric
motors are connected to the main body of the vehicle and the two wheels with the
help of scalar bonds. Likewise, the inverted pendulum (handle) is also connected to
the main vehicle body via scalar bonds.
(a) (b)
Fig. 3 a Sketch for illustration of velocities in the moving frame and b bond graph modelling of
vehicle body and attached components
Vehicle Body. The model of vehicle body is shown in Fig. 3a. The vehicle is
assumed to be symmetrical about the longitudinal axis. The bushing is taken as
a simple linear spring damper system. The main vehicle body’s motion can be
expressed as the three linear displacements about the three body fixed coordinate
axis, and rotational motion of the body is given as yaw, pitch and roll motion. The
Newton–Euler (NE) equations are used to model the main body of the vehicle [12].
These equations contain attached body fixed axis which are aligned to the inertia
principal axis. The Newton–Euler equations are given in [10–12].
Now, further step in the modelling is to compute the velocities at different points
on the rigid body and transforms them to the inertial coordinate system and physical
constraint implementations at chosen points. For three linear velocities of left bushing
reference point (Fig. 3a) in the moving system of axes, the equations can be written
as
Same type of expressions can be written for the right-side bushing. The transfor-
mation of velocity from the moving frame to the inertial frame is given in detail in
the appendix and Euler angle rates from the body fixed angular velocities are given
in [10]. Figure 3b represents the models of main vehicle body and transformation
of velocities (linear to angular) to the bushing reference points. Three moments and
forces (three sets each) act upon the body. The bond graph coordinate transforma-
tions block (CTF) for coordinate transformation from inertial to moving frame and
from moving frame to inertial frame are given in [10] and also explained in the
PI Control-Based Modelling of Segway Using Bond Graph 73
appendix. As the weight of the vehicle body and the aerodynamic forces act in the
inertial frame, these forces are to be converted to body fixed forces through coordi-
nate transformation before being applied to the vehicle body. Here, in this simulation,
the aerodynamic forces are neglected as the speed of the segway is not too high.
Modelling of Handle (Inverted pendulum) for Segway. The bond graph
modelling for inverted pendulum system, which is attached to the vehicle main body
is shown in Fig. 4. The inverted pendulum system is connected to the main body of
the vehicle through the two coordinate transformation blocks. The angle θ shown in
the bond graph of inverted pendulum in Fig. 4 represents the pitch angle, which is to
be controlled by the PI controller.
Model of Wheel. The schematic diagram of wheel is shown in Fig. 5a. The bond
graph model for the wheel is shown in Fig. 5b. It is modelled as a rigid body having
five degrees of freedoms and the rolling about x-axis is neglected here. There are six
input ports in the bond graph modelling of wheel. Three effort inputs are coming for
the bushing and torque is supplied from the differential. The contact point between
the wheel and road is not a fixed point. It changes as per the wheel rotation about
the axle. The gravity force and the normal contact force between the wheel and road
always act in the inertial direction. The weight of the wheel is acting in inertial z-
direction, which can be represented by source of effort (Se: wm g). The mass of the
wheel is represented as (I : wm ) and rw is the radius of the wheel. The tractive force,
which is used to generate the motion between the wheel and road surface, is given
by Se : Ft . The mass moments of inertias are represented by Iwy and Iwz in y- and
z-directions, respectively.
Model of Wheel Bushing. The schematic diagram of wheel bushing is shown in
Fig. 6a. The bond graph model of wheel bushing that is located between the wheel
axle and mounting point is shown in Fig. 6b. The bushing is modelled as a parallel
spring damper system. In the bond graph model, the stiffness (C: K S ), damping
74 A. Kumar et al.
(a) (b)
(a) (b)
(a) (b)
Model of Steering Controller. The steering is provided for the turning motion
of the vehicle. The bond graph modelling for steering controller is given in Fig. 7b.
Steering is basically a moment which has applied to the axles of the two wheels about
z-axis causing the change in the yaw angle of the wheel (also the relative torque on
the body of the vehicle). The steering is provided by changing the rotational speed
of the motors. The speeds of the motors are modulated by and where and are the
angular position of the wheel about z-axis. These angular positions are determined
by steering controller. The modulus of two transformers and as shown in Fig. 7b are
determined by the Ackermann’s formulae [10] in the controller domain and these
are given as
a cos2 θ1 + c tan θ1 cos2 θ1
δ̇i = δ̇ (4)
a cos2 θst − c tan θst cos2 θst
a cos2 θr − c tan θr cos2 θr
δ̇o = δ̇ (5)
a cos2 θst + c tan θst cos2 θst
Initially, single wheel model of segway without controller (Fig. 8) is developed. M2,
M6, M35, M16, M20, M23 and M30 are inductance of motor, flywheel along with
wheel mass moment of inertia, mass of wheel, mass of vehicle (segway), pendulum
mass, mass moment of inertia of pendulum rod and pendulum mass, respectively.
The signal flow graph of the single wheel model is shown in Fig. 9.
76 A. Kumar et al.
Fig. 9 Signal flow graph of single wheel model of segway without controller
The transfer function between the angular velocity of segway arm and the voltage
applied to the motor, when linearized about the vertical position is
(6)
From the open-loop transfer function of the system as given in Eq. (6), it is
found that there is a double pole at the origin, which makes the system unstable.
Rest all the poles are either at imaginary axis or in the left-half plane. Further, the
controllability of the open-loop system is calculated and it is found that all the 11
states are controllable. The bond graph model of single wheel model of segway with
PI controller is shown in Fig. 10. The transfer function between the angular velocity
of segway arm and the voltage applied to the motor with PI controller is
(7)
After implementing the PI controller, all the closed-loop poles are lying in the
left-half plane, which ensure the stability of the system. When PI controller is
added in the forward loop, there is a new pole that lies at the negative real axis
PI Control-Based Modelling of Segway Using Bond Graph 77
(−0.51 + 0. (i)) due to the use of integral term and a new zero is also located at
(−0.64 + 0. (i)) on the same axis.
PI controller will try to take the handle to the reference position (inverted position).
But due to inertia, the pitch angle is displaced in backward direction due to forward
motion of segway. And then due to this change in the pitch angle, segway moves in
the backward direction to minimize the change in pitch angle, and finally, vehicle
comes to the rest after 4 s and handle will come to its reference position after a
little bit oscillations about the reference position in 7 s. The plot for pitch angle with
time (Fig. 11a), plot for linear speed of the vehicle with time (Fig. 11b) and plot for
angular speed of the wheel with time (Fig. 11c) are shown.
PI Control-Based Modelling of Segway Using Bond Graph 79
Fig. 11 Plot for a pitch angle, b linear speed of the vehicle, c angular speed of the wheel for positive
force applied to the handle and d pitch angle for backward force
in Fig. 12d. The angular speeds of left and right wheels are shown in Fig. 12e, f,
respectively. The speed of the left wheel is lesser than the speed of the right wheel
during turning of the segway.
Fig. 12 Plot for a linear speed of vehicle, b yaw angle at the centre, c yaw angle of the vehicle
body, d pitch angle of vehicle, e angular speed of left wheel and f angular speed of right wheel
during steering
PI Control-Based Modelling of Segway Using Bond Graph 81
5 Conclusions
Bond graph modelling technique is very efficient tool for modelling any physical
and mechanical system, and simulating various kinds of conditions for the same. As
a segway is an under-actuated, nonlinear and unstable system, designing controller
for such a system is a very complicated task. In this paper, an effort was made
to develop the stabilization and swing up controller for segway using bond graph
approach. Also, the bond graph modelling of the segway and its various parts such
as vehicle body, handle, wheel, wheel bushing and electric motor with PI control
were developed. The proportional integral (PI) control was well demonstrated to be
capable of measuring the change in the pitch angle from the reference position and
generates a control input signal for forward and backward motion of the segway to
minimize that change.
References
1. Martynenko G, Formal AM (2005) The theory of the control of a monocycle. J Appl Math
Mech 69:516–528
2. Nawawi A, Ahmad MN, Osman JH (2008) Real-time control of a two-wheeled inverted
pendulum mobile robot. Int J Comput Inf Eng 79:316–356
3. Dai F, Gao X, Jiang S, Guo W, Liu Y (2015) A two-wheeled inverted pendulum robot with
friction compensation. Mechatronics 30:116–125
4. Pathak K, Franch J, Agrawal S (2005) Velocity and position control of a wheeled inverted
pendulum by partial feedback linearization. IEEE Trans Rob 21:16–28
5. Goher KM, Tokhi MO, Siddique NH (2011) Dynamic modelling and control of a two wheeled
robotic vehicle with a virtual payload. ARPN J Eng Appl Sci 9:131–150
6. Grepl R (2009) Balancing wheeled robot: effective modelling, sensory processing and
simplified control. Eng Mech 16:141–154
7. Delgado S, Kotyczka P (2015) Energy shaping for the robust stabilization of the wheeled
inverted pendulum, vol 48. IFAC, Elsevier Ltd., pp 093–098
8. Mukherjee A, Karmakar R, Samantaray AK (2006) Bond graph in modelling, simulation and
fault identification. CRC Press, Florida
9. Karnopp DC, Margolis DL, Rosenberg RC (2000) System dynamics, modelling and simulation
of mechatronic system. Wiley, New York
10. Bera TK, Bhattacharya K, Samantaray AK (2011) Evaluation of antilock braking system with
an integrated model of full vehicle dynamics. Simul Model Pract Theory 19:2131–2150
11. Bera TK, Merzouki R, Ould Bouamama B, Samantaray AK (2012) Design and validation of
the reconfiguration strategy for a redundantly actuated intelligent autonomous vehicle. Proc
Inst Mech Eng Part I J Syst Control Eng 226(8):1060–1076
12. Bera TK, Bhattacharya K, Samantaray AK (2011) Bond graph model based evaluation of a
sliding mode controller for combined regenerative and antilock braking system. Proc Inst Mech
Eng Part I J Syst Control Eng 225(7):918–934
Strategic Coordination and Navigation
of Multiple Wheeled Robots
Abstract Multi-robot navigation and coordination are addressed in this paper. All
the robots are subjected to their own kinematic and dynamic constraints. Genetic
algorithm tuned fuzzy logic-based motion planner is compared with the potential
field-based motion planner. To avoid the conflicts during the navigation, two different
coordination schemes, namely strategic and heuristic, are implemented. Results are
compared through computer simulation. Simulation experiments were started with
eight number of robots initially, and the number of robots has been increased up to 17
later on due to the need of coordination scheme for the maximum number of robots.
Strategic coordination scheme along with the genetic fuzzy-based motion planner
is found to perform better than the other combinations concerning the quality of
solutions and time taken to reach the goal positions. Computational complexity of
different methods has also been compared and presented.
1 Introduction
Lots of researchers have put their effort to solve the mobile robot navigation prob-
lem. Meng et al. [14] studied a leader—follower coordinated tracking problem for
multiple heterogeneous Lagrange systems to control a group of followers to track
a time-varying global objective function (often denoted as a leader) by using only
local information interactions. Peng et al. [6] solved the navigation problem con-
sidering it as a mixed-integer nonlinear programming problem. Here, the path of
each agent is decomposed into collision-prone and collision-free segments. But, this
approach is not at all suitable for real-time systems, as the environment is highly
dynamic [6]. Morinaga et al. [15] discussed a motion planning problem for a spher-
ical rolling robot actuated by two internal rotors that are placed on orthogonal axes.
A motion planning strategy composed of two trivial and one nontrivial manoeuvre
is devised. Neural network [8]-based motion planning as well as traditional methods
like potential field method [17, 19] also have been implemented to solve this nav-
igation problem. The artificial neural network (ANN) can learn the scenario of the
environment, and a few researchers have successfully applied the neural network [8]
to develop the motion planning algorithm of mobile robots. Motion planning as well
as coordination techniques have also been developed simultaneously. Recently, Kia
et al. [7] proposed a novel algorithm for networked robots which localize itself in a
global coordinate frame by local dead reckoning and opportunistically corrects its
pose estimate whenever it receives a relative measurement update message from a
server. Load distribution strategy has been developed for cooperative manipulators,
and kinematic constraints imposed for manipulator ensemble have also been done
by Erhart and Hirche [3]. GA fuzzy [11] and neuro fuzzy [5] have been applied to
solve such problems with dynamic obstacles. Obstacle avoidance problem [13] of
mobile robots using FLC has also been studied by many researchers for years. In
such a case, primary fact is to develop a high-efficiency rule set. The performance
of FLC is influenced by its knowledge base (rule set) and the membership functions.
Therefore, it is essential to adjust the rules of fuzzy controller and the membership
functions of input and output variables to get better performance.
Genetic algorithm has been applied to tune the input and output membership
functions of the fuzzy logic controller [20]. Three different algorithms have been
developed with different effects. Very few researchers as like [20] have also used
the GA optimized FLC in this research problem. Most of them experimented on
the regular (circle shape) obstacles for simulation. But these approaches might not
work efficiently in the presence of irregular obstacles. Here, we have done simulation
experiment based on the navigation problem of a wheel mobile robot in the unknown
environment. FLC has been used to sense the data from knowledge base though it
is manually constructed by authors. Then, a GA has been applied to optimize the
membership functions as well as the rules set of the FLC. We have studied some exten-
sive literature survey based on different research direction. Percentage calculation
of different research direction has been presented via Fig. 1. Here, it is evident that
researchers are focusing on coordination characteristics but very few researchers have
successfully implemented strategic coordination on navigation problem [1, 10, 23].
In the present research, different approaches have been tried to solve the motion
planning problem as well as coordination of multi-agent systems in the dynamic
Strategic Coordination and Navigation of Multiple Wheeled Robots 85
environment. For the motion planning purposes, three different approaches have been
developed, and some strategies are proposed to evolve the coordination’s among the
agents. The remaining part of the paper is structured as follows: in Sect. 2, developed
cooperative multi-agent systems have been studied. Results and discussion are pre-
sented and discussed in Sect. 3. Comparison with previous research works has been
depicted in Sect. 4. Finally, some concluding remarks and the scopes for future work
are discussed in Sect. 5.
the obstructing robot which is closest to the planning robot is treated as the critical
robot, and motion of the planning robot is made based on the critical robot. Here,
we consider two inputs (distance, di j and angle, θi for the ith planning robot treating
jth robot as critical one) and two outputs (deviation, φi and acceleration αi for the
ith planning robot) for the motion planner. All the inputs and outputs are restricted
to some of their lower and upper limits. In this manner, the motion of all the robots
is made. After that, positions and orientations of the robots are updated [5], and time
step is updated to next one. This process is repeated for all the robots until or unless
all of them reach the goal. Finally, total travelling time of each robot is calculated
by multiplying with the number of time steps (say, n i for the ith robot) a robot has
taken to reach the goal with the T (time step) value. However, last time step may
not be the full-time step, and it will be a fraction of T say (ki ∈ 0–1) and depend on
the distance between the planning robots to its goal (called as goal distance, dgoal ).
Moreover, if goal distance (dgoal ) is less than the maximum distance (dmax ), motion-
planning scheme is not activated, and no collision with the other robots is presumed.
Total
Rtravelling time taken by the robots to achieve the task for a particular scenario
is i=1 (n i + ki )T , where R denotes the number of robots. The primary require-
ment of this study is to minimize the total travelling time. Overall, motion planning
scheme is explained via activity diagram in Fig. 2.
In a conflicting situation, one robot might be treated as most critical to other
robots. In such case, according to the process explained above, a motion planner will
be providing reduced velocities and deviations for both the robots. It is a conflicting
situation, and reduction of motion of both the robots is unnecessary. One might
move along the goal direction with a maximum velocity of it, and the other may
compromise its motion. But, the question is who will compromise and whom to
be given the upper hand in a competitive environment, where the performance of
each one of them is measured. Also, a robot will get this motivation only when they
get some rewards for this. It demands a cooperation scheme. Proper strategies to be
developed to decide who will cooperate and how much? We have developed two
different motion planning approaches along with two cooperative schemes to solve
this kind of problems, and they are explained below in brief.
This traditional method is very popular and effective. Due to the generation of attrac-
tive and repulsive forces, this method has shown the directions to calculate the devi-
ation angle and other necessary things. The concept of this method has been taken
from the literature [5, 19]. For more details, interested readers may go through [9].
Strategic Coordination and Navigation of Multiple Wheeled Robots 87
The fuzzy theory can deal with uncertain and approximate information. In the field of
mobile robot navigation, this technique is generated additional significance amongst
researchers. In this approach, both the database and rule base of the fuzzy logic-based
motion planner are extracted from the published literature [5]. Four each linguistic
terms for deviation and acceleration and five each linguistic terms for angle and
88 B. Pradhan et al.
deviation is considered. Therefore, a rule base will have twenty antecedents (input
combinations), and for each antecedent, there could be twenty different consequents.
So, four hundred possible rules might exist in the rule base and out of which twenty
rules must be there. Detail of this is available in the book [22].
Due to the non-optimal nature of FLC, it is required to tune the FLC rule base as
well as database. In this paper, knowledge base of FL-based motion planner has been
designed manually. Thus, a binary-coded genetic algorithm (GA) has been used to
tune the two input and two output-based knowledge base of FLC. Working principles
of the genetic fuzzy system may be extracted from the book of Pratihar [22]. The
information regarding GA strings used here has been extracted from [18, 21]. The
essential purpose of GA is to optimize the travelling time taken by the robot and is
mentioned in Eq. 1. Minimize total travelling time
100
N
(T ) = (n i j + ki j )T (1)
j=1 i=1
where n i j and ki j can be understood from Sect. 2. A robot will take the least time to
reach the goal if it does not deviate from its goal line and move with the maximum
possible velocity. Therefore, above function can be minimized indirectly by mini-
mizing the error due to deviation and acceleration of the robots. Therefore, fitness
function of GA is considered as the half mean squared error (HMSE) of all the 100
training scenarios and expressed as
100 N M
1 2
HMSE = amax − ai jm + φi jm 2 (2)
200N × M j=1 i=1 m=1
where ai jm and φi jm are the acceleration and deviation of the ith robot in jth situation
following mth time step. The above function is to be minimized satisfying all the
constraints of the robots.
Let us discuss the answers to these queries. To survive daily life, human beings need
to show coordination with either static things or dynamic elements. The attributes
or elements might be living or non-living. As robots are designed/developed like
human beings, so these characteristics are also to be incorporated. Otherwise, they
may suffer at the time of navigation and cannot handle the conflicting situations.
Different parameters are involved in developing coordination between robots [2, 12,
16]. Suppose, in a multi-agent system, there are four robots where robots 1 and 2 are
treating each other to be critical. In such a situation, motion planner will ask both
the robots to compromise their motion to avoid the collision. However, this is not
required for both the robots, one may negotiate and other may not. The robot which
compromises is called cooperative robot, and other is a selfish robot. Question is how
to segregate them and is there any reward for becoming cooperative or penalty for
showing non-cooperative. There could be at least three possible strategies for a robot.
• No Coordination (NC) If the robot does not want to sacrifice its motion generated
by the motion planner.
• Full Coordination (FC) The robot is ready to sacrifice both the outputs which
might take different deviation and acceleration other than the generated by the
motion planner.
• Partial Coordination (PC) The robot is ready to sacrifice only one output which
might take different deviation but considers the same acceleration as generated by
the motion planner.
The robot is ready to sacrifice only one output which might take different devia-
tion but consider the same acceleration as generated by the motion planner. Partial
coordination is better than the full coordination; however, wheeled robots are limited
by its turning speed and cannot take any amount of turn. Thus, we will have to iden-
tify the best option for a situation. The robot which will cooperate will be awarded
a reward, and the non-cooperative robot will be given a penalty. Reward and penalty
values are presented in Table 1.
Now, the question is who will follow which strategy in a two robot facing the con-
flicting situation. Heuristic coordination can solve the said problem [19]. However,
it lacks adaptability and not optimal in any sense. Thus, we are proposing a strategic
coordination scheme based on game theory in this paper.
Here, both the robots have three alternatives and are competing players. Gains
or losses called pay-off and profit of one robot will be the loss of the other robot.
Pay-off matrix for different strategies adopted by the robots is placed in Table 2. It
suggests that multiple strategies can be developed. But in between nine developed
strategies, the best plan will be selected. The concept of two-person zero-sum pure
strategy games [4] is used to identify the best strategy, and accordingly, the revised
motion of the robots is calculated.
Finally, three motion planning approaches can be clubbed with the two coordi-
nation schemes (Heuristic or Strategic) to solve the navigation problems of multiple
mobile robots and tested through computer simulations.
The efficacy of the designed approach has been experienced through computer sim-
ulation to solve different cases of car-like robots. Three different cases have been
considered with eight, twelve and seventeen robots navigating in a grid of 40 × 40 m2 .
Some of the important parameters of the robots are mentioned below.
• Dimension: Two-wheeled mobile robot (70 mm × 70 mm × 80 mm; wheel diam-
eter: 40 mm)
• Maximum Velocity: 4 m/s
• Acceleration: 0.005–0.05 m/s2
• Time interval(dt): 2 s.
Strategic Coordination and Navigation of Multiple Wheeled Robots 91
The optimal GA parameters are achieved from the GA parametric study and in Table 3
for three different cases. Optimal rule base for three different cases is also obtained
and presented in Table 4. It can be observed that the robots are slowing down the
acceleration and taking more deviation with the increase in the number of robots
present in the environment.
Once the tuning of Approach 3 is over, all the three approaches along with the two
coordination schemes are compared to solve the navigation problem of multiple
mobile robots. Let us first try to identify the need of different coordination schemes.
A robot “ j” can take n i j number of time steps (out of which last time step may be
fractional)
to reach the goal in a particular scenario “i”. Therefore, a robot “ j” can run
100
for a total i=1 n number of time steps, and all the robots as a whole can traverse
8/12/17 100 i j
j=1 i=1 n i j many time steps. It is sure that the motion planning schemes are
not activated for all the time steps.
Table 3 Optimal parameters of GA for the three different cases for Approach 3
Cases Crossover Mutation Population size Maximum
probability probability number of
generations
Case 1 (involves 0.5 0.009 90 90
8 robots)
Case 2 (involves 0.011 70 100
12 robots)
Case 3 (involves 0.011 70 130
17 robots)
92 B. Pradhan et al.
Table 7 presents the data relevant to the number of time steps that the cooperation
schemes have been fired. It is also for sure that not all the robots will be cooperative
for an equal number of times. Therefore, we have tried to identify the robot that has
cooperated maximum times. Although it depends on the position and orientations
of the robots, the maximum velocity for all the robots is same. Therefore, the robot
which has shown more cooperativeness is a more autonomous robot compared to the
other. In that perspective, it can be observed that the Approach 3 in a combination
of strategic coordination scheme (SC2) has provided the best result and generated a
number of cooperative scenarios. The eighth robot has become more cooperative for
most of the cases and therefore said to be most adaptive robot amongst all.
Table 5 Average travel time in seconds of the robots for different approaches
Robots 8-Robots case 12-Robots case 17-Robots case
Approach 1 Approach 2 Approach 3 Approach 1 Approach 2 Approach 3 Approach 1 Approach 2 Approach 3
SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2 SC1 SC2
R1 60.55 60.64 61.28 61.64 61.28 61.25 62.55 62.89 61.53 61.93 61.31 60.63 63.73 63.46 61.79 62.22 61.77 61.16
R2 69.24 69.34 69.89 70.1 69.8 70.35 72.54 72.67 70.31 70.6 69.79 70.85 73.03 73.33 70.88 71.28 70.41 72.15
R3 63.26 63.4 63.82 63.99 63.84 64.17 64.23 64.5 64.06 64.19 64.01 64.54 65.14 64.97 64.51 64.91 64.39 65.21
R4 55.5 55.47 55.88 56.13 55.7 56.34 59.14 59.03 56.16 56.44 55.94 56.11 59.97 59.59 56.69 57.05 56.5 56.92
R5 103.85 104.12 104.99 105.29 104.87 105.37 104.7 105.18 105.32 105.56 104.97 105.74 105.19 105.31 105.64 105.9 105.27 106.1
R6 83.33 83.27 84.07 84.37 83.95 83.93 85.61 85.43 84.26 84.68 84.09 84.29 85.72 85.68 84.63 85.17 84.44 85.08
R7 115.25 115.36 116.23 116.48 116.07 116 115.18 115.36 116.49 116.76 116.33 115 115.24 115.46 116.86 117.37 116.65 115.29
R8 82.96 83.21 83.68 84.15 83.55 83.71 85.33 85.38 83.95 84.39 83.54 83.1 85.22 85.52 84.24 84.8 83.94 82.72
R9 59.81 59.71 54.59 54.89 54.14 54.18 59.8 60.14 55.37 55.94 54.94 56.29
R10 48.49 49.31 43.2 43.48 42.91 43.32 48.62 49.66 43.39 43.68 43.26 43.47
R11 61.38 61.64 61.62 62.1 61.2 61.15 62.37 62.35 62.08 62.95 61.88 62.06
R12 57.48 57.4 53.04 53.25 52.79 52.81 58.57 58.07 53.77 54.12 53.44 53.24
R13 62.36 62.57 59.35 59.85 59.25 59.05
Strategic Coordination and Navigation of Multiple Wheeled Robots
No. of robots Approaches Three best scenarios out Best scenario Travelling time in seconds Average travel time in sec-
of hundred scenarios of different robots for the onds
best scenario
SC1 SC2 SC1 SC2 SC1 SC2
8 1 5, 20, 71 3, 5, 32 5 [115.57, 79.73, [117.28, 82.00, 105.3 106.4
129.03, 87.12, 129.03, 87.12,
116.96, 96.74, 118.01, 97.08,
125.88, 91.09] 125.88, 95.07]
2 31, 46, 94 1, 20, 31 31 [18.94, 33.45, [18.94, 34.55, 73.16 68.94
25.99, 87.00, 25.99, 52.98,
121.54, 87.56, 124.81, 87.89,
128.11, 82.65] 124.24, 82.09]
3 5, 8, 66 1, 7, 49 7 [69.71, 70.58, [69.71, 74.32, 91.1 90.94
95.52, 73.53, 84.15, 73.53,
113.45, 96.73, 113.45, 96.73,
126.96, 82.29] 126.96, 88.68]
12 1 1, 8, 27 1, 3, 32 3 [119.37, 118.89, [121.94, 89.11, 111 84.03
119.03, 119.55, 100.38, 18.27,
121.92, 118.50, 124.80, 82.50,
119.14, 118.99, 119.14, 93.01,
118.49, 20.23, 73.51, 21.52,
118.87, 118.73] 98.87, 65.27]
2 1, 19, 31 2, 20, 31 31 [18.94, 33.45, [18.94, 34.55, 64.94 62.21
25.99, 87.00, 25.99, 48.98,
123.56, 87.31, 124.81, 97.58,
129.58, 80.57, 124.24, 80.57,
33.09, 44.66, 34.91, 44.17,
47.44, 67.68] 47.44, 64.32]
3 19, 39, 71 1, 49, 52 52 [65.16, 36.93, [65.16, 36.93, 70.25 71.64
87.86, 24.12, 86.17, 24.12,
120.85,83.96, 132.65, 83.96,
120.06, 94.06, 119.56, 95.75,
83.93, 32.27, 81.39, 32.27,
60.23, 33.52] 63.77, 37.96]
B. Pradhan et al.
(continued)
Table 6 (continued)
No. of robots Approaches Three best scenarios out Best scenario Travelling time in seconds Average travel time in sec-
of hundred scenarios of different robots for the onds
best scenario
SC1 SC2 SC1 SC2 SC1 SC2
17 1 1, 27, 35 3, 5, 71 71 [81.87, 110.44, [71.43, 119.61, 71.03 71.01
95.38, 48.44, 114.38, 51.56,
87.52, 73.51, 87.52, 73.51,
79.95, 95.43, 84.91, 80.14,
61.75, 55.81, 63.62, 55.81,
46.41, 43.71, 45.59, 30.00,
47.89, 126.03, 44.11, 126.03,
73.37, 4.56, 73.37, 4.53,
75.51] 81.01]
2 5, 16, 19 27, 46, 72 46 [46.10, 103.26, [45.90, 100.23, 65.23 66.98
68.66, 55.23, 68.66, 52.77,
109.4, 93.47, 114.45, 93.47,
100.42, 69.72, 104.12, 69.72,
41.6, 33.74, 41.6, 33.74,
15.19, 46.66, 23.19, 46.66,
35.3, 58.13, 35.30, 61.87,
87.28, 46.73, 95.48, 46.73,
97.96] 104.8]
Strategic Coordination and Navigation of Multiple Wheeled Robots
Average travelling time for all the one hundred scenarios is presented in Table 5. It
has been observed that heuristic coordination scheme (SC1) has resulted in less aver-
age travelling time for the robots in compared to the strategic coordination scheme
(SC2). Increase in the number of cooperative scenarios may be the reason for the
increase in travelling time. However, the difference is not very high. For eight robots
case, performance regarding the travelling time of Approach 1 is the best. But, with
the increase in robots, Approach 3 has superseded the others. It indicates that the
Approach 3 is more adaptive and robust. We have also identified the three best scenar-
ios regarding a higher number of cooperative situations and noted the best scenarios
in which highest number of cooperative schemes is fired (refer to Table 6). It has
been observed that the strategic cooperation scheme has generated the best result
in all such cases. Movement of eight, twelve and seventeen robots for the outstand-
ing scenario of different methods has been envisioned in Figs. 3, 4, 5, 6, 7 and 8,
respectively.
Strategic Coordination and Navigation of Multiple Wheeled Robots 97
Fig. 3 Navigation of eight, twelve, sixteen and twenty robots using PFM (heuristic coordination),
respectively
6 Concluding Remarks
Traditional as well as soft computing-based approaches are found to solve the coor-
dination problems of multiple mobile robots working in a dynamic environment.
Heuristic-based coordination schemes have been considered in our earlier study,
but it lacks in adaptability and robustness. Therefore, a game theory-based strategic
coordination scheme is presented in this paper. Two different types of motion plan-
ning schemes are coined with two coordination strategies to evolve the best-suited
navigation algorithm. Following observations are noted.
98 B. Pradhan et al.
Fig. 4 Navigation of eight, twelve, sixteen and twenty robots using fuzzy (heuristic coordination),
respectively
Fig. 5 Navigation of eight, twelve, sixteen and twenty robots using GA fuzzy (heuristic coordina-
tion), respectively
We could not test the performance of the developed approaches with real robots. We
are presently doing this. Also, we are thinking of collecting data from the real market
scenario. A big data-based approach can be integrated to enhance the complexity of
the existing problem.
100 B. Pradhan et al.
Fig. 6 Navigation of eight, twelve, sixteen and twenty robots using PFM (strategic coordination),
respectively
Strategic Coordination and Navigation of Multiple Wheeled Robots 101
Fig. 7 Navigation of eight, twelve, sixteen and twenty robots using fuzzy (strategic coordination),
respectively
102 B. Pradhan et al.
Fig. 8 Navigation of eight, twelve, sixteen and twenty robots using GA fuzzy (strategic coordina-
tion), respectively
References
1. Cui R, Gao B, Guo J (2012) Pareto-optimal coordination of multiple robots with safety guar-
antees. Auton Robots 32(3):189–205. https://doi.org/10.1007/s10514-011-9265-9
2. Cui R, Guo J, Gao B (2013) Game theory-based negotiation for multiple robots task allocation.
Robotica 31(6):923–934. https://doi.org/10.1017/S0263574713000192
3. Erhart S, Hirche S (2015) Internal force analysis and load distribution for cooperative multi-
robot manipulation. IEEE Trans Robot 31(5):1238–1243. https://doi.org/10.1109/TRO.2015.
2459412
4. Hillier F, Lieberman G (2010) Introduction to operations research. McGraw-Hill, New York,
NY. https://books.google.co.in/books?id=NvE5PgAACAAJ
5. Hui NB, Mahendar V, Pratihar DK (2006) Time-optimal, collision-free navigation of a car-like
mobile robot using neuro-fuzzy approaches. Fuzzy Sets Syst 157(16):2171–2204. https://doi.
org/10.1016/j.fss.2006.04.004
Strategic Coordination and Navigation of Multiple Wheeled Robots 103
6. Peng J, Akella S (2003) Coordinating the motions of multiple robots with kinodynamic
constraints. In: 2003 IEEE international conference on robotics and automation (Cat. No.
03CH37422), vol 3, pp 4066–4073. https://doi.org/10.1109/ROBOT.2003.1242222
7. Kia SS, Hechtbauer J, Gogokhiya D, Martínez S (2018) Server-assisted distributed cooperative
localization over unreliable communication links. IEEE Trans Robot 34(5):1392–1399. https://
doi.org/10.1109/TRO.2018.2830411
8. Low KH, Leow WK, Ang MH (2002) Integrated planning and control of mobile robot with self-
organizing neural network. In: Proceedings 2002 IEEE international conference on robotics
and automation (Cat. No. 02CH37292), vol 4, pp 3870–3875. https://doi.org/10.1109/ROBOT.
2002.1014324
9. Latombe JC (1991) Robot motion planning. Kluwer Academic Publishers, Norwell, MA, USA
10. Li H, Savkin AV (2018) An algorithm for safe navigation of mobile robots by a sensor network
in dynamic cluttered industrial environments. Robot Comput-Integr Manuf 54:65–82. https://
doi.org/10.1016/j.rcim.2018.05.008
11. Liu Q, Lu Y, Xie C (2006) Optimal genetic fuzzy obstacle avoidance controller of autonomous
mobile robot based on ultrasonic sensors. In: 2006 IEEE international conference on robotics
and biomimetics, pp 125–129. https://doi.org/10.1109/ROBIO.2006.340327
12. Liu S, Sun D, Zhu C (2011) Coordinated motion planning for multiple mobile robots along
designed paths with formation requirement. IEEE/ASME Trans Mechatron 16(6):1021–1031.
https://doi.org/10.1109/TMECH.2010.2070843
13. Marín L, Vallés M, Soriano Á, Valera Á, Albertos P (2014) Event-based localization in Ack-
ermann steering limited resource mobile robots. IEEE/ASME Trans Mechatron 19(4):1171–
1182. https://doi.org/10.1109/TMECH.2013.2277271
14. Meng Z, Dimarogonas DV, Johansson KH (2014) Leader-follower coordinated tracking of mul-
tiple heterogeneous Lagrange systems using continuous control. IEEE Trans Robot 30(3):739–
745. https://doi.org/10.1109/TRO.2013.2294060
15. Morinaga A, Svinin M, Yamamoto M (2014) A motion planning strategy for a spherical rolling
robot driven by two internal rotors. IEEE Trans Robot 30(4):993–1002. https://doi.org/10.1109/
TRO.2014.2307112
16. Philip G, Givigi S, Schwartz MH (2014) Cooperative navigation of unknown environments
using potential games. Int J Mechatron Autom 4:173. https://doi.org/10.1504/IJMA.2014.
064098
17. Pradhan B, Nandi A, Hui NB, Roy DS, Rodrigues JJPC (2020) A novel hybrid neural
network-based multirobot path planning with motion coordination. IEEE Trans Vehic Technol
69(2):1319–1327
18. Pradhan B, Hui NB, Roy DS (2020) Heuristic coordination for multi-agent motion planning.
In: Khanna A, Gupta D, Bhattacharyya S, Snasel V, Platos J, Hassanien AE (eds) International
conference on innovative computing and communications. Springer, Singapore, pp 569–578
19. Pradhan B, Sinha Roy D, Baran Hui N (2018) Motion planning and coordination of multi-agent
systems. Int J Comput Vis Robot 8:492. https://doi.org/10.1504/IJCVR.2018.095002
20. Pradhan B, Sinha Roy D, Baran Hui N (2019) Multi-agent navigation and coordination using
GA-Fuzzy approach. In: SocProS 2017, vol 2, pp 793–805
21. Pradhan B, Varadarajan V, Hui N, Sinha Roy D (2019) Intelligent navigation of multiple
coordinated robots. J Intell Fuzzy Syst 36(5):4413–4423. https://doi.org/10.3233/JIFS-169996
22. Pratihar D (2015) Soft computing: fundamentals and applications. Alpha Science International
Limited, Oxford, UK. https://books.google.co.in/books?id=gv4lswEACAAJ
23. Su J, Xie W (2011) Motion planning and coordination for robot systems based on representation
space. IEEE Trans Syst Man Cybern Part B (Cybern) 41(1):248–259. https://doi.org/10.1109/
TSMCB.2010.2051025
Spur Gear Mechanism for Accurate
Angular Indexing and Locking
of Angular Position by Using Additive
Manufacturing
1 Introduction
replace the conventional spring elements that meet the design specifications for self-
locking purposes. Ratchet is gear tooth mechanism with a pawl that engages the
teeth [2]. The main components of a ratchet and pawl mechanism are gear (uniform
asymmetric teeth), pawl and base [3]. When ratchet rotates in forward, pawl slides
over the teeth, and if ratchet rotates in the opposite direction, pawl gets inserted
into the teeth and gets locked restricting backward motions. Ratchet mechanism
provides uni-directional transmission both in linear and rotational direction against
heavy loading conditions [4]. The purpose the paper is to develop a compact novel
mechanism, which is capable of locking in bi-direction with more number of locking
positions and thereby decreasing the interval angles without fasteners.
2 Literature Review
The indexing mechanism converts rotatory motion into a series of step movements of
the output link. Common types of indexing mechanisms are ratchet and pawl, rack
and pinion, Geneva mechanism and cam drive [5]. Ratchet and pawl mechanism
used for large wheels that rotate in one direction along the specified axis [6]. The
most common examples of the ratchet and pawl mechanism are clocks, jacks and
hoists [5]. One of the experimental studies conducted by Jalili (2003), piezoelectric
ceramic-based ratchet actuation device is utilized to control the engine valves. In
the study, it revealed that the opening and closing of engine valves are computer
controlled by cam timings [7]. The usage of electromechanical actuator can be used
to control the ratchet and pawl mechanism [8].
Need for bearings and springs is eliminated by a new design of the ratchet
mechanism [2]. Under that study, the micro-wire EDM process is identified as the
potential manufacturing process for making the prototype model of the mechanism
and inferred that it was a potentially working prototype model. Ratchet mechanism
utilized for preventing the vehicle from descending down [9]. A microgripper with a
ratchet self-locking mechanism manufactured using silicon-on-insulator wafer with
30 µm layer and with a locking interval as 10 µm. With the application of ratchet
locking, damage of the microgripper has been reduced [10]. During locking, pawl
moves in the reverse direction; as a result, it may lead to damage of the pawl due to
the force acting on the ratchet [11]. Thereby, an alternative mechanism is needed to
reduce the damage of the pawl and permits the bi-direction movement.
Gears are basically a toothed profile that transmits power and motion by the gear
tooth interlocked in mesh [12]. Parameters that are considered for the gear tooth
profile are pitch circle diameter, tooth thickness, pressure angle, number of teeth, the
rotational angle of the gear and rotational speed of the gear. Manufacture of gears
requires high accuracy and specialized manufacturing process for production of gear
tooth profile [13]. Spur gear is relatively simple in design and easy to produce for
operating at low-speed applications. Comparing the bending and contact stress for
both normal contact ratio (NCR) and for high contact ratio (HCR), spur gear using
finite element analysis inferred that NRC gear set has less contact ratio than HCR
Spur Gear Mechanism for Accurate Angular Indexing and Locking … 107
gear set [14]. Most of the spur gear applications are with respect to 20° involute teeth
that are manufactured by gear hobbing machine. However, if the involute profile
is less than 17 teeth, it was identified that low involute profile thickness leads to
crack initiated at the root fillet area of the profile geometry [15]. Contact stress and
bending stress on the spur gear are affected by the shape of the gear tooth profile,
and it is identified by nonlinear finite element approach [16]. By using the boundary
element method, effect of circular root fillet over the standard trochoidal root fillet
was studied. The study revealed that teeth having circular fillet gave high bending
strength [17]. Fillet radius of nylon 6/6 gear produced through an injection moulding
process affects the performance of the transmission [18].
The fatigue life of the gear is low when the number of teeth is less and has to
optimize at the earlier stage of gear design [19]. Design iteration work was carried out
to find the optimum tooth profile that reduces vibration and noises [20]. A numerical
fault detection model of a dynamic transmission error (DTE) is used to examine
the effects of tooth crack length for symmetric and asymmetric tooth profiles [21].
The parameters of the spur gear are towards minimizing the transmission errors
that are governed by nominal torque. The study also inferred that low contact ratio
gears will result in more noise and vibrations when used for transmitting power. The
mathematical model is governed by line equation which utilizes simple curve features
for improving the tooth profile and its taking pressure angle as a parametric variable
[22]. Mathematical model of spur gear is based on the curvature that engages with
the tooth profiles. Study revealed that mathematical model developed to enhance the
contact strength decreases the wear of the tooth surfaces [23]. Non-circular gear trains
are used to create an indexing mechanism where the indexing period is different [24].
Conventional gear hobbing process and injection moulding process are involved in
manufacturing of gear tooth profile [25].
Additive manufacturing (AM) deals with developing a wide range of products
with complicated profiles from the CAD model. AM technology having the possi-
bility to produce low volume and small quantities of customizable products that are
in complicated shapes with low cost [26]. AM technology is suited for producing
functional end products that can be used directly [27]. It finds potential applications
in the area of healthcare such as implantable prosthesis, orthotics, bio-engineering
and in surgical planning [28]. AM technology benefitted industries such as auto-
motive, aerospace, consumer products and architectural modelling catering to new
product development [29].
The application of AM technology allows the development of non-assembly
mechanism and for robotic systems. The study revealed that evaluation of the mech-
anism was easy as it was able to visualize the joints clearly, able to visualize the
internal mechanisms and achieve multiple degrees of freedom as in the case of
robotic systems. Mavroidis et al. studied the various kinematic models using rapid
prototyping technologies and found out that rapid prototype products developed are
the exact replica of the physical models of the machines [30]. In a study conducted
by Plooij et al., different locking was identified; in the study, Geneva mechanism
was able to lock at 5 positions with 72° interval. Ratchet mechanism was capable
of locking at 15 positions with 24° interval, while cam follower-based locking is
108 A. Pugalendhi et al.
3 Methodology
Softwares used in this study are SolidWorks 2016 and CatalystEX. Modelling
and simulation of the components are performed by SolidWorks. After the accepted
simulation results, final parts to be printed are obtained from SolidWorks in the STL
format. CatalystEX is used to interact with the machine such as tool path generation,
part orientation, model and support material creation and its layer gap as well as
printing time.
P=Z (1)
Equation 1 infers that the total number of positions is directly proportional to total
number of teeth on gear.
A = 360◦ × Z −1 (2)
P = Z × L (3)
A = 360◦ × Z −1 L −1 (4)
110 A. Pugalendhi et al.
i x 360◦ × L −1 + A , n > 1, n ∈ Z +
LPx = (5)
0 if n = 1
where
P = Total number of positions after increasing the position of the lock
A = Minimized interval angle after the introduction of position of the lock
LPx = Starting position/origin of the provision of the lock
By using Eqs. 3 and 4, if the lock is able to change their positions in four loca-
tions means, it results 72 positions and interval angles of 5°, respectively. Figure 1
illustrates the performance of the proposed spur gear mechanism with four locking
locations. Initially, desired position is achieved by insertion of lock into the gear as
shown in Fig. 1a. Here, interval angle of 20° is possible in both directions. Red colour
solid line represents the origin, and blue colour centre line represents the interval
angle accomplished by modified ratchet mechanism. At this juncture, the apparatus
facilitates the interval angles of 20°, 40° and 60° up to 360°. Adjustment of the 10°
Spur Gear Mechanism for Accurate Angular Indexing and Locking … 111
(for third quadrant = 5° + 5°) angular position of lock from starting position leads
to double the total number of locking positions (36 No’s) and thereby reducing half
of the interval angle (10°). Now, the apparatus facilitates the interval angles of 10°,
30° and 50° up to 350° as shown in Fig. 1c. Similarly, again 5° adjustment of the
position of lock in second (5°) and fourth (5° + 5° + 5°) quadrants and its interval
angles are shown in Fig. 1b and Fig. 1d, respectively. Thereby, the apparatus facili-
tates 5° interval angle and 72 angular positions. Due to space constraint, adjustment
of lock position is performed on upcoming quadrants. Starting position or origin of
the provision of lock is determined by Eq. 5, and the results of first, second, third
and fourth lock position are 0°, 95°, 190° and 285°, respectively.
4 Analysis
In this study, the stress analysis of spur gear mechanism was carried out using Solid-
Works simulation. Both identical part materials are ABS Plus-P430 and its detailed
112 A. Pugalendhi et al.
Fig. 2 Von Mises stress plot (a) and displacement plot (b) of proposed spur gear mechanism
Following the design and simulation of the model, STL file format is converted by
SolidWorks. This final STL file is ready for printing by uPrint machine through
Spur Gear Mechanism for Accurate Angular Indexing and Locking … 113
pre-processing software named as CatalystEX. Layer gap (gap between the adja-
cent layers), part orientation and placement of the printed parts are controlled by
CatalystEx. 3D printed parts strength, surface finish, total printing time and mate-
rials consumptions of both model and support material are mainly affected by part
orientation and its layer gap. Envelope temperature of the uPrint machine and model
material temperature is 70 °C and 210 °C, respectively.
The fabrication of the spur gear mechanism is divided into four parts such as
gear, lock, shaft and housing. Before printing the whole model, the working and
accuracy of the mechanism with the help of sample miniature prototype are checked.
Here, shaft and gear coupled by single part. All parts are manufactured by same
procedure and same fabrication conditions in single print. Solid-type layer gap for
model material and smart-type layer gap for support material are chosen. Due to
minimum Z axis height and less support material consumption, auto orientation is
selected for all the three parts.
All the three parts are printed in single run with desired locations on build tray.
Finally, packed chromeleon backup archive (CMB) file is send to printer for fabrica-
tion. Once the printing was finished, printed parts are removed from build tray and
removing the all support materials using wave wash technology. This fabrication
process consumes 0.48 in3 of model material and 0.24 in3 of support material, and it
took 0.38 h. 3D printed miniature model of spur gear mechanism is shown in Fig. 3.
Individual parts of miniature model are shown in Fig. 3a, and assembly of all parts
are shown in Fig. 3b.
Gear with shaft is placed on bore of the housing, and it is rotated freely in both
directions. Placing the lock into teeth of the spur gear results, arrest the rotation of
gear. Movement of the lock is arrested by slots provided in the outer periphery of the
housing. For easy understanding, pointer position and lock provisions are marked
in both housing and gear. Despite, exact locking is ensured by angle readings on
the gear. For example, 75° angle is achieved by lock engaged in fourth provision of
housing as shown in Fig. 3b.
Fig. 3 Miniature model of spur gear mechanism a individual parts and b assembly
114 A. Pugalendhi et al.
Fig. 4 Final spur gear mechanism fabricated by additive manufacturing a individual parts and
b assembly
After ensuring the performance and functionality of the proposed spur gear mech-
anism, actual prototype is manufactured similar to miniature prototype. 3D printed
individual parts of final spur gear mechanism and assembly are shown in Fig. 4a
and b. This fabrication process consumes 5.27 in3 of model material and 0.47 in3 of
support material, and it took 7.22 h.
Table 3 Difference between traditional ratchet mechanism and proposed spur gear mechanism
S. No. Description Ratchet mechanism Spur gear mechanism
1 Rotary motion Ratchet Spur gear
2 Locking by Pawl Spur teeth
3 Contact between ratchet and Continuous Discontinuous
pawl
4 Direction of rotation Uni-directional Bi-directional
5 Self-locking Yes No
6 Fasteners requirement Yes No
7 Total number of positions 18 72
8 Interval angle 20° 5°
9 % of increases in position NA 75%
10 % of increases in accuracy NA 75%
11 Mode of locking Automatic Manual/automatic
12 Measurement of angle Partially Exactly
13 Play between ratchet and Reverse direction of ratchet Purely eliminated
pawl
14 Size Larger Compact
7 Conclusion
Acknowledgements We gratefully acknowledge the financial support for establishing the Centre
of Excellence in Manufacturing Sciences at Coimbatore Institute of Technology, India from Ministry
of Human Resource Development (MHRD), Government of India where the R&D work is carried
out.
116 A. Pugalendhi et al.
References
21. Dogan Oguz, Karpat Fatih (2019) Crack detection for spur gears with asymmetric teeth based
on the dynamic transmission error. Mech Mach Theory 133:417–431. https://doi.org/10.1016/
j.mechmachtheory.2018.11.026
22. Zhang-Hua F, Ta-Wei C, Chieh-Wen T (2002) Mathematical model for parametric tooth
profile of spur gear using line of action. Math Comput Model 36(4–5):603–614. pii:SO895-
7177102100185-l
23. Liu H, Mao K, Zhu C, Xu X, Liu M (2012) Parametric studies of spur gear lubrication perfor-
mance considering dynamic loads. Proc Inst Mech Eng Part J: J Eng Tribol 226(9):731–737.
https://doi.org/10.1177/1350650112445672
24. Fangyan Z, Lin H, Xinghui H, Li Bo, Dingfang C (2016) Synthesis of indexing mechanisms
with non-circular gears. Mech Mach Theory 105:108–128. https://doi.org/10.1016/j.mechma
chtheory.2016.06.019
25. Tang Y, Chang S, Wang Z, Zhang K (2011) Study on helical tooth profile modification of
planetary gear transmission on the basis of gear transmission error. Appl Mech Mater 86:47–50.
https://doi.org/10.4028/www.scientific.net/AMM.86.47
26. Ngo T, Kashani A, Imbalzano G, Nguyen K, Hui D (2018) Additive manufacturing (3D
printing): a review of materials, methods, applications and challenges. Compos B Eng
143:172–196. https://doi.org/10.1016/j.compositesb.2018.02.012
27. Campbell I, Bourell D, Gibson I (2012) Additive manufacturing: rapid prototyping comes of
age. Rapid Prototyp J 18(4):255–258. https://doi.org/10.1007/978-1-4419-1120-9
28. Webb P (2002) A review of rapid prototyping (RP) techniques in the medical and biomedical
sector. J Med Eng Technol 24(4):149–153. https://doi.org/10.1080/03091900050163427
29. Edgar J, Tint S (2019) Additive manufacturing technologies: 3D printing, rapid prototyping,
and direct digital manufacturing, 2nd edn. http://dx.doi.org/10.1595/205651315X688406
30. Mavroidis C, DeLaurentis K, Won J, Alam M (2001) Fabrication of non-assembly mechanisms
and robotic systems using rapid prototyping. J Mech Des 123(4):516. https://doi.org/10.1115/
1.1415034
31. Plooij M, Mathijssen G, Cherelle P (2015) Review of locking devices used in robotics. IEEE.
https://doi.org/10.1109/MRA.2014.2381368
32. Giorgio F, Pierluigi R, Jorge A (2006) The pure-rolling cam-equivalent of the Geneva mech-
anism. Mech Mach Theory 41:1320–1335. https://doi.org/10.1016/j.mechmachtheory.2006.
01.002
33. Yuhu Y, Jianyong W, Shicai Z, Tian H (2019) Design of a novel coaxial eccentric indexing cam
mechanism. Mech Mach Theory 132:1–12. https://doi.org/10.1016/j.mechmachtheory.2018.
10.012
34. Andrianesis K, Tzes A (2014) Development and control of a multifunctional prosthetic hand
with shape memory alloy actuators. J Intell Rob Syst 78(2):257–289. https://doi.org/10.1007/
s10846-014-0061-6
35. Pons J, Rocon E, Ceres R, Reynaerts D, Saro B, Levin S (2004) The MANUS-HAND dextrous
robotics upper limb prosthesis: mechanical and manipulation aspects. Auton Robots 16(2):143–
163. https://doi.org/10.1023/B:AURO.0000016862.38337.f1
36. Unal R, Behrens S, Carloni R, Hekman E, Stramigioli S, Koopman B (2018) Conceptual design
of a fully passive transfemoral prosthesis to facilitate energy-efficient gait. IEEE Trans Neural
Syst Rehabil Eng 26(12):2360–2366. https://doi.org/10.1109/IEMBS.2011.6090111
37. Saharan L, Tadesse Y (2016) Robotic hand with locking mechanism using TCP muscles for
applications in prosthetic hand and humanoids. Bioinspiration Biomimetics Bioreplication
9797. https://doi.org/10.1117/12.2219535
38. Yang J, Xue Y, Shang J, Luo Z (2014) Research on a new bilateral self-locking mechanism for
an inchworm micro in-pipe robot with large traction. Int J Adv Rob Syst 11(10):174. https://
doi.org/10.5772/59309
Fabrication of Solid Lubricant Coating
and Its Optimization Using Response
Surface Methodology
1 Introduction
between particles was observed to be helpful in achieving higher deposition rate and
thick layered coating.
Beri et al. well defined that the tool electrode (green compact) fabricated by powder
compaction process is called green compact [2]. The loose bonding between particles
facilitates the material deposition over the substrate. EDC is capable to deposit the
thick coating over the workpiece by alternating the various process parameters as
implemented by the Patowari et al. [3]. Krishna and Patowari analyzed melting and
vaporization takes place in EDM process due to the occurrence of multiple electrical
discharges between tool electrode and the work surface in presence of hydrocarbon
oil [4]. Beri et al. also discussed that the electrical discharge process can be applied
in aerospace and automotive industry. At high temperature, the chemical reaction
starts between the eroded tool and carbon present in dielectric oil and results in the
formation of carbide [2]. Goto verified the phenomenon of thick layer formation when
the carbide piled up over the work surface [5]. Solid lubricant was revealed to be the
finest choice for components that require high temperature/vacuum for working [6].
EDC has been widely studied by the researchers but limited research work carried
out on the formation of solid lubricant coating using EDC process through MoS2 +
Cu powder green compact tool.
In this paper, solid lubricant coating of powder mixture (MoS2 + Cu) in green
compact form is done over a mild steel substrate. The green compact is prepared
by compressing the powder at a specific pressure and temperature to provide loose
bonding between powder particles. The effect of various experimental input process
parameters (duty factor, peak current, and mixing ratio) on the output parameters
(tool wear rate (TWR), mass deposition rate (MDR), and coating thickness) is also
studied. For this purpose, response surface methodology (RSM) is applied to apply
the mathematical relation among independent and dependent variables. The surface
plots of all the output responses with change in input have been drawn. Many exper-
iments carried out in this work to find out that if experimental values are in good
agreement with those predictive results.
2 Experimentation
Various trials were conducted to achieve a desirable mixture ratio of MoS2 and Cu
powder for the green compact. Then for the final electrode, to be used in the main
experiment, MoS2 powder was mixed with Cu powder in a ratio of 0.6:1, 1:1 and
1.5:1 by percentage weight. The powder compact tool was made with mixture of
molybdenum disulfide (MoS2 ) and copper (Cu) powders in a hot mounting press.
The compaction of prepared powder mixture was done in hot mounting press at
a pressure 200 kg/cm2 , the temperature 130 °C, heating time 10 min, and cooling
5 min. A powder compact having diameter 15 mm and height 3 mm was formed and
that size depends on the powder mixture amount used. A green compact electrode
shown in Fig. 1a was then pasted with copper rod end using the conductive paste. At
the end, the complete tool electrode was used to perform the coating process. The
Fabrication of Solid Lubricant Coating and Its Optimization … 121
(a) (b)
Tool electrode
Workpiece
Clamping vice
EDC process was carried out using die sink EDM by connecting tool and workpiece
to reverse polarity in presence of hydrocarbon oil as shown in Fig. 1b. The process
started with a series of electrical discharges. A chemical reaction takes place between
dielectric particles and the tool material, as a result of which a layer forms over the
work piece. The tool melts faster than the workpiece and material deposition starts
from the tool to the work surface due to loose bonding between particles. Few EDM
parameters those kept constant were gap voltage: 40 V, time of coating: 4 min, pulse
on time: 29 µs, for all the experiment and output parameters [layer thickness (LT),
tool wear rate (TWR), mass deposition rate (MDR)] were evaluated as given in
Table 1.
y = f (x1 , x2 , x3 , . . . , xn ) ± ε (1)
122 R. Tyagi et al.
n
n
n
f = a0 + ai xi + aii xi2 + ai j xi x j + ε (2)
i=1 i=1 i< j
The ANOVA of TWR, MDR, and LT is given in Tables 2, 3 and 4. The p-value
indicates that probability of F-value > calculated owning to noise. F value is used to
find the significance of values. When p-value goes beyond 0.05, the term becomes
insignificant which lacks fit. It shows that non-significant term is discarded from
the model to allow it to fits well. As the table shows that P values are significant as
all the p values <0.05. The coefficient of determination (R2 ) for TWR is 90.96%.
Similarly, the coefficient of determination (R2 ) for deposition rate is 90.07% and
for layer thickness is 99.58%. Further, the normality test is carried out by plotting
output results. Figure 2 shows that residuals are distributed uniformly because all the
residues lie nearly along the normal probability plot. The results obtained from the
experiments are in such a way that all the results of ANOVA are in good agreement
with the reprised models.
The response of TWR, MDR, and LT with respect to variable parameters is observed
as shown in Fig. 3. The peak current has a major role in EDC process that enhances
the mass deposition rate, tool wear rate, and coating thickness. In EDC process,
the energy of spark depends upon the peak current value. In case of higher peak
124 R. Tyagi et al.
current, more spark energy causes the more melting that increases the mass deposi-
tion over workpiece due to the melting of tool material. Hence, coating layer thick-
ness increases directly with the change in peak current. The variation of TWR and
MDR with respect to the duty factor was observed that gives better performance for
achieving higher coating thickness up to 50% DF So, good quality of coating can be
obtained at 50% DF. The arcing was observed above this value (50%) that lead to
the unstable coating.
As shown in Fig. 3a, at low value of duty factor (30%), only weak sparks were
established that does not allow enough material to be transported over the workpiece.
Hence, high peak current value is needed for required spark energy at low duty factor
values to transfer the particles. The TWR raises as the peak current increases and
further decreases beyond a certain limit at higher duty factor due which may be due
to increase in debris particles that restricts the stable ionization process. Also, at
10 A current increase in duty factor causes the decrease in TWR as only arcing takes
place for these values. In Fig. 3b, it can be observed that at higher concentration of
MoS2 and low current, TWR increases due to lose bonding in particles due to low
concentration of Cu in the powder mixture which led to decrease in green compact
electrode strength. Although, TWR decreases with rise in peak current as the strength
of the electrode is low to bear the spark energy.
In Fig. 3c and d, mass deposition rate shows approximately similar trend as
observed for TWR. The machining was more than coating at low current value (4 A)
126 R. Tyagi et al.
95
90
80
70
Percent
60
50
40
30
20
10
1
-0.050 -0.025 0.000 0.025 0.050 0.075
Residual
95
90
80
70
Percent
60
50
40
30
20
10
5
1
-0.02 -0.01 0.00 0.01 0.02
Residual
95
90
80
70
Percent
60
50
40
30
20
10
1
-0.02 -0.01 0.00 0.01 0.02
Residual
Fig. 4 FESEM images of a top and b cross section of coating with EDS
which leads to decrease in deposition rate even at high duty factor values (Fig. 3c).
Although, increase in current increases the deposition rate. However, the presence
of debris inside the working gap restricts the proper transfer of tool material to be
transported over the workpiece. Also, higher mixing ratio leads to decrease in depo-
sition rate as TWR decreases (Fig. 3d). Similar trend is observed for layer thickness
which is directly proportional to peak current (Fig. 3e and f). Hence, the peak current
plays major role in this technique to achieve the desired coating thickness.
FESEM of coated samples was performed to detect the presence of pores in the
coating. In EDM, the peak current helps in initiating the spark discharge to achieve
a sufficient spark heat to melt and transport the green compact material over the
substrate. Hence, higher value of peak current must be needed to deposit the large
amount of material onto the substrate to obtain the denser coating. Pores at the top
surface were found in less amount at high current; 10 A (Fig. 4a). The maximum
thickness of 0.647 was obtained and EDS analysis in Fig. 4b confirms the presence
of Mo, S, Cu, Fe, and C, over the coating surface.
4 Conclusions
A coating of single powder layer of MoS2 over the work surface was not possible by
the EDC process as it is very dry in nature. So, a second material has to be blended with
it for making the tool. With an increase in peak current due to high spark energy, the
coating thickness, mass deposition rate, and tool wear rate increase as well. Therefore,
the peak current value significantly affects all the coating parameters. An enhanced
surface integrity of coating can be achieved with mixing ratio of MoS2 :Cu/1:1, peak
current at 10 A and DF at 50%. The experiments carried out in this work shows that
experimental values are in good agreement with those predictive results.
Fabrication of Solid Lubricant Coating and Its Optimization … 129
References
1. Murray JW, Algodi SJ, Fay MW, Brown PD, Clare AT (2017) Formation mechanism of electrical
discharge TiC-Fe composite coatings. J Mater Process Technol 243:143–151
2. Beri N, Maheshwari S, Sharma C, Kumar A (2010) Technological advancement in electrical
discharge machining with powder metallurgy processed electrodes: a review. Mater Manuf
Processes 25(10):1186–1197
3. Patowari PK, Saha P, Mishra PK (2010) Artificial neural network model in surface modification
by EDM using tungsten–copper powder metallurgy sintered electrodes. Int J Adv Manuf Technol
51(5–8):627–638
4. Krishna ME, Patowari PK (2014) Parametric study of electric discharge coating using powder
metallurgical green compact electrodes. Mater Manuf Processes 29:1131–1138
5. Goto A (2001) Development of electrical discharge coating method. In: 13th proceedings of the
international symposium for electromaching
6. Zhen J, Cheng J, Zhu S, Hao J, Qiao Z, Yang J, Liu W (2017) High-temperature tribological
behavior of a nickel alloy matrix solid-lubricating composite under vacuum. Tribol Int 110:52–56
Computing and Verification of IPMC
Parameters Through Equivalent Beam
Theory
Abstract In this paper, an effort is made to measure and verify analytical and
experimental aspects of a cantilevered smart material of various dimensions which
is electrically actuated for various applications. The smart material is taken as
ionic polymer-metal composite (IPMC). The supported cantilevered beam model
is analyzed, the relations for block force are provided, and the same is used to derive
displacement at the free end. In order to make a proper agreement between the
analytical results and measured results from an experimental setup, a comparison is
made using a bending beam equation for the cantilevered condition with determined
physical properties. The effect of water content in the parameters like deflection
under the voltage actuation of IPMC is observed during the experimental process for
underwater conditions.
1 Introduction
The recent trend of smart actuation with properties like lightweight, low power
consumption, stealth property, and large bending makes IPMC a good contender for
use in applications like AUVs, underwater surveillance, and many more. The basic
properties and characteristics of IPMC are provided in [1] in their review work which
covers mechanical and electrical properties. The aspect of manufacturing the IPMC
covers many details and [2] explained the various manufacturing techniques related
to IPMC and also reviewed various problems like effects of various cations, force
optimization, manufacturing techniques with better cost minimization, and many
more to enhance the successful manufacturing commercialization of the IPMC. As
per the current scenario where the demand for energy grows, there is a need for smart
materials which can efficiently work. IPMC structure comprises of a soft polymer
base material; here it is Nafion, and it is coated from both the sides with an electrode;
here it is platinum, which not only actuates under the supply of electric potential,
but also is able to store potential when it is twisted or bend in any direction. IPMC
as a smart material proved to be the best alternative for various industrial devices
as explained in the [3] compared with conventional devices for applications like
sensors and actuators. As IPMC actuate on the application of electric field, there
are various ways to model this aspect and [4] described the modeling of IPMC as
actuator considering three basic categories: physics-based model, black box model,
and gray box model. Each category emphasis on certain assumptions and the literature
provides the differences as well. Now, [5] explained a micromechanical physics-
based theory for bending of IPMC governing applied electric field, transport of ions,
and the elastic deformation in order to make assumptions for IPMC responses in air.
Work was done in [6] using a black box model where a physical consideration is less
and more focusses on the curvature of actuation applicable only to some conditions.
The model of the actuation of IPMC based on gray box modeling where physical laws
are combined with empirical models with some physical interpretations using both
analytical and experimental methods was studied in [7]. A micromechanical theory
developed in [8] to understand the response in the air of a water-saturated IPMC by
coupling the electric field, ion transport, and elastic deformation. Their explanation
provided the optimal macroscopic responses while guiding the microstructure of
the material. The bending response of the IPMC which is the main characteristic to
study as for application aspect of this smart material which mainly depends upon
the process of electroplating the polymer, type of element used for electrode, the
main composition of polymer used, and the concentration of the aqueous solution
mainly in pH and salt content. The effects of the above said points are properly
studied in [9]. As the property of IPMC suggested that applying electric stimuli, a
bending response is generated. Here in [10], an electromechanical model in which
mechanical deformation is induced by electric fields was forced to the kinetics of
the characteristic bending response. Comparison is made between the theoretical
model and experimental results by performing experiments simultaneously with the
membrane of varying thickness in different kinds of salt solution. The work is related
to blocked force measurement done in [11] with the use of a microbalance machine.
The force generated at the tip of IPMC acting as an actuator when the tip is fixed is the
blocked force. Here electro-active paper (EAPap) is actuated by DC electric fields,
and the theoretical cantilever beam model is compared with the measured blocked
force showing good agreement between the two. The first work is done using IPMC
as a robotic swimmer utilizing its bending characteristics in [12] provided a proper
design concept for kinematics and dynamics of the robotic actuator. The IPMC
parameters depend on its water content as [13] performed experimental verification
of the fact that the tip deflection of IPMC was higher at the water content of about
70% and low at lower water content as in case of deflection in air condition due to
evaporation of water by successive voltage supply.
Computing and Verification of IPMC Parameters Through … 133
In this paper, an effort is made to measure and verify analytical and experimental
aspects of a cantilevered smart material of various dimensions which is electrically
actuated for various applications. The smart material is taken as ionic polymer-
metal composite (IPMC). The supported cantilevered beam model is analyzed, the
relations for block force are provided, and the same is used to derive displacement at
the free end. In order to make a proper agreement between the analytical results and
measured results from an experimental setup, a comparison is made using a bending
beam and bimorph beam equation for the cantilevered condition with determined
physical properties. The effect of water content in the parameters like deflection
under the voltage actuation of IPMC is observed during the experimental process for
underwater conditions.
2 Mathematical Model
The modeling for IPMC actuated by an input voltage can be done by considering the
condition of a supported cantilever beam.
Figure 1 illustrates that applying beam theory using free body diagram will yield
reaction force as [14],
3pL
RB = (1)
8E I
RB is the reactive force at the supported end, p is the distributed load, L is the
length of cantilevered IPMC, E is Young’s modulus, and I is the second moment of
inertia.
In order to get the distributed loads, since we have used the micro-weighing for
block force calculation, so from (1), we have
8E I F
p= (2)
3L
pL 4
ymax = (3)
8E I
3 Results
The experimental work mainly focuses on Fig. 2 for blocked force measurement
which illustrates that the blocked force measurement of IPMC is done using micro-
weighing machine with the use of a balanced weight.
Following the illustrative figure shown in Fig. 2, the IPMC is connected with the
dynamic signal generator and actuated with the DC power supply. The experimenta-
tion is done on a Nafion-based IPMC which is procured from environmental robots
(www.environmental-robots.com) of dimensions 33 mm × 10 mm × 0.2 mm and
41.44 mm × 10 mm × 0.2 mm.
In order to get the blocked force generated by IPMC, the free end is blocked with
the balanced weight, and the IPMC is actuated under various voltage conditions as
shown in Fig. 3.
Table 1 shows the blocked force measured from the micro-weighing machine and
also showing the trend that the increasing voltage supply increases the value of the
blocked force generated by IPMC, but the limit of actuation voltage is 4–5 V.
Figure 5 shows the displacement results for 33 mm length IPMC attaining the
peak value of around 3.44 mm in the air when actuated at 4 V input voltage at the
particular time showing resonance condition.
Figure 6 shows the displacement results for 41.44 mm length IPMC attaining the
peak value of around 8.51 mm in the air when actuated at 4 V input voltage at the
particular time showing resonance condition.
The curve fitted relationship can be written as:
f (d) = p1 ∗ d + p2 (4)
values of coefficients (with 95% confidence bounds) for both IPMC samples are:
For 33 mm IPMC: p1 = −0.2452, p2 = 0.7191 and
For 41.44 mm IPMC: p1 = −0.3133, p2 = 1.122.
Final relation can be written from Eq. 4 as
−0.2452 ∗ d + 0.7191 for 33 mm IPMC Sample
f (d) = (5)
−0.3133 ∗ d + 1.122 for 41.44 mm IPMC Sample
where f is the measured reaction force and d is the tip displacement for IPMC. Thus
from Eq. (5), the measured displacement d when f = 0 providing tip displacement
for IPMC without any constraint and the measured force f for d = 0 provides the
reaction force.
Figures 7 and 8 show the maximum displacement and force values for respective
IPMC samples actuated at 1.5 V.
The comparison between analytical and measured displacement from IPMC actu-
ator can be seen in Table 2. The experimentation process for displacement measure-
ment is done underwater to examine the change in the water content of the IPMC
and also the different environmental conditions from air to underwater. The same
IPMC sample is used for this experimentation and also the measurement techniques
remain the same.
A beaker filled with water is used here to make underwater conditions for IPMC.
The arrangement is so made that the IPMC is dipped underwater, and when actu-
ated by voltage, its tip starts deflecting, and the deflection is measured by a laser
vibrometer as shown in Fig. 9.
Figures 10 and 11 show the displacement vs time plot for the sample of IPMC
whose length is 33 mm and for different voltage actuation of 3 and 4 V, respectively.
138 A. Gupta and S. Mukherjee
It has been observed that because of the high water content the IPMC deflects more
in underwater conditions than in air conditions as the water content of IPMC starts
decreasing due to evaporation by continuous voltage supply.
4 Conclusion
The work is done in an effort to measure and verify analytical and experimental
aspects of a cantilevered smart material ionic polymer-metal composite (IPMC).
The supported cantilevered beam model is analyzed, the relations for block force are
140 A. Gupta and S. Mukherjee
provided, and the same is used to derive displacement at the free end. In order to
make a proper agreement between the analytical results and measured results from
an experimental setup, a comparison is made using a bending beam equation for the
cantilevered condition with determined physical properties. The curve fitted results
for various length IPMC shows maximum displacement and force values actuated at
particular input voltage supply. Water content in IPMC affects the deflection made
under the actuation of voltage provided is observed in the experimental results for
different voltage inputs which confirms the literature related to this, and in order to
prepare a static beam model for underwater deflection of IPMC, more functions are
needed in beam theory to define damping effects due to water, effect of added mass,
and many more. So, this will provide us a future task to do which will be helpful for
many underwater applications aspects of IPMC.
References
Abstract The metal matrix composites (MMCs) are increasing day by day in the
entire manufacturing sectors due to their unique properties such as high strength
to weight ratio, good mechanical properties, and better durability. The application
of waste material in metal matrix composites has been getting more attention as
they can reinforce particles in a metal matrix which enhance the strength properties
of the composites. In addition, by applying these wastes materials in a useful way
not only save the manufacturing cost of products but also reduce the pollution on
the environment. Aluminium MMCs have found applications in the manufacturing
of various components such as a piston, cylinder block, and brake drum, in which
wear and friction are an important phenomenon. A comprehensive literature review is
carried out on MMCs based on individual reinforcement and multiple reinforcements
including various product applications. It is summarized in form of fields involved in
tribology, fabrication processes/parameters, reinforcement(s), matrix contribution,
tribological testing parameter, and product application areas of MMCs. This paper
attempts to review the different combination of reinforcing material(s), both synthetic
and agro-derived used in the processing of hybrid aluminium matrix composites,
its effect on different mechanical and wear properties of the composite. Different
methods for production of these materials have been studied and discussed in detail.
Compatibility of agro-derived reinforcements with AMCs and research areas for
further improvement of MMCs has been suggested.
1 Introduction
Over the past decade, the world is redesigning itself with a new point of view. The end
product in any field is now not only judged by its performance alone but also with the
resources used making a product, their optimization. The better of engineering today
aims at producing same or better performance parameters with minimal inclusions
of new processes or systems and reviewing and redesigning the existing product
parameters. The automotive industry is one of the most global industries. As we
move into the twenty-first century, it is clear that global environmental issues will
have an increasingly significant impact on the automotive industry’s future growth
and sustainability. The automotive sector worldwide has been practicing use of new
age light materials for the past two decades and has reached a new high in the recent
years. This is because of raising emissions control standards. The performance of
automobiles would definitely increase with some alterations in the aerodynamics or
engine design, but reducing the weight of the automobile would not only contribute
in enhancing the performance but would also help in cutting down emissions [1]. In
addition to this, automakers have been subjected to very stringent fuel requirements,
to which the response of getting the automobiles lighter in weight is proving one
of the best options. The shift of automakers to aluminium frames, engine blocks,
suspension blocks, etc., has proved instrumental in reducing a considerable amount
of body weight of an automobile. Use of aluminium in automobiles has allowed the
automakers to recycle 80–85 per cent of the automobile with as much as just five
per cent of the energy used for initial production of the car [2]. With new trends
setting in a combination of aluminium with composites and nanocomposites, these
have made it possible to further cut down the weight and also have contributed to
increasing the reliability [3]. As the era of composites started to develop, they are
being tailored according to the required need. Amongst composites MMCs have
found direct applications in automobiles, this is because they offer better mechan-
ical properties than conventional material and as light as compared to the later.
Aluminium matrix composites (AMCs) are the ones used widely by automakers
currently. They are lighter than aluminium thus further reducing the weight of the
automobile. AMCs have been researched widely with regards to their compatibility
with reinforcements of different materials [3]. MMCs are either reinforced with
synthetic reinforcement(s) or agro-industrial derived reinforcements. Also, two or
more synthetic or agro-industrial-derived reinforcements are reinforced in a metal
matrix to form a hybrid MMC. The inclusion of agro-industrial-based reinforce-
ments in MMCs has not been so profound over the past few years because of their
high processing temperatures, lack of information regarding the use of the agricul-
tural waste, and questioned the reliability of such manufactured MMCs [4]. The
use of agro-industrial-based reinforcements made from agro-industrial waste also
tackles another important issue. The agricultural waste materials generated are hard
to disposal, and thus a major concern to environmental pollution. Utilization of
such waste materials could reduce contamination and spaces for disposal. Therefore,
processing of waste material by tailoring it into the required form of material for
application in automobile and construction industries is a prime concern amongst
the current researchers [5].
Characterization of Mechanical Properties of Different … 143
MMCs are rapidly finding their way into many engineering applications such as
aerospace, automobile, marine, etc., thereby replacing the conventionally used metals
and their alloys. MMCs are nothing but metals or metallic alloys that are reinforced
with synthetic reinforcements. The metals or metallic alloys that are used commonly
are light metals such as aluminium (Al), titanium (Ti), zinc (Zn), copper (Cu), and
stainless steel [6]. Aluminium and aluminium alloy are gaining huge industrial signif-
icance because of their outstanding combination of mechanical, physical, and tribo-
logical properties over the base alloys [7]. Aluminium and its alloys have been most
commonly used as a matrix, as compared to other metals, and the reason for the same
has been reported [6, 8]. The inclusions of alloying elements such as magnesium (Mg)
and silicon (Si) benefit over all other alloying elements with respect to properties
such as fluidity, heat treatability, strain hardening, weldability, and corrosion resis-
tance of the alloy. The AA (6XXX) is the series of aluminium alloy with inclusions
of Si and Mg. The aluminium alloy of grade 6061(AA6061) has been used widely
in automotive sectors and it has been an alloy of keen interest to researchers when
it comes to MMCs [7]. The MMCs of AA6061 exhibit better mechanical properties
over conventional metal/alloys [9].
The reinforcement(s) used in MMCs are divided into two groups, namely synthetic
reinforcements in the form of ceramic particulates and agro-industrial waste-derived
reinforcements. The properties of hybrid composite depend on the individual rein-
forcement properties and the type of matrix metal/alloy selected [8]. The end proper-
ties also depend upon the processing route taken up for production of the composite.
The processing route on the other hand again depends on the nature of reinforcements
and matrix metal/alloy [6].
Faleh et al. [15] studied the properties and application of Al-Gr composites. It was
reported that as the graphite content in the composite was increased, the wear rate
decreased considerably. However, after increasing the content of graphite beyond 5
to 6% weight, the wear rate increased. The coefficient of friction also decreased, as
the volume % of graphite in the composite was increased. After an increase of about
30%, the coefficient of friction remains the same. Similar observation was made in
case coefficient of thermal expansion (CTE). CTE decreased with increase in the
reinforcement of graphite. They also reported a wide range of application of Al-
Gr composite. The applications stated are engine cylinders, cylinder blocks, liners,
pistons, and connecting rods.
Roseline et al. [16] carried out a numerical evaluation of zirconium reinforced
aluminium matrix composite. Since the end properties of any composite are based
on various parameters, testing for each parameter while other being constant would
take a lot of time and material. The testing specimens were examined, tested exper-
imentally as well as with numerical sequential methods against varying % of zirco-
nium dioxide (5–10–15%) in the composite. It was found out that the 15% of fused
ammonia gives better properties than other alternatives.
Baradeswaran et al. [17] investigated their work on the effect of varying graphite
content in AA6061. The experiments were carried out to study the wear properties
with carrying graphite content, effects of sliding speed and sliding distance. The
specimens were prepared using a conventional casting method. It was found that
with an increase in Gr weight % in the composite, the dry sliding wear reduces up
to 5% increase in Gr weight and then increases thereafter. The coefficient of friction
goes on decreasing up to 15% of Gr content by weight and then steadily increases
with a further addition of graphite. Overall tribological behaviour of AA6061 was
found to be improved under the addition of graphite as reinforcements.
Nuruzzaman et al. [18] have carried out their research in processing and mechan-
ical properties of silicon carbide metal matrix composite in which specimens of
10%, 20%, and 30% silicon carbide reinforcement by volume were prepared under
different compaction loads (10 and 15 ton). The effect of the stated parameters was
investigated separately. The results showed that there was a mild increase in the
density of the composite as the volume % of reinforcement was increased. Also, the
density of composites of respective composition was higher in case of compaction
loads. The density of composites compacted by a load of 15 ton was greater than that
of compacted by a load of 10. Same was the case with the hardness of the composite.
As the volume % or compaction load of SiC was increased in the composite, its
Vickers hardness value is also increased.
Jeevan et al. [19] have studied the properties of Al-SiC composites related to
compaction, sintering, and mechanical aspects. The specimens were prepared from
pure aluminium and SiC following the powder metallurgy route. The particulate rein-
forced metal matrix was analysed for powder morphology and structure of composite.
Micro-hardness and compressive tests were also carried out. The morphology of
the material powder involved was made with scanning electron microscope (SEM)
as shown in Figs. 1, 2, 3, 4, and 5. The microstructure examination showed grain
size, morphology, and distribution of reinforcement within the matrix. The optical
Characterization of Mechanical Properties of Different … 145
The agro-waste derivatives offer some advantages when used in the synthesis of
AMCs. These advantages include low cost, accessibility, low density, and reduced
environmental pollution. A very few waste products and their processing into ashes
have been done to study their compatibility as a reinforcing material [5, 21]. The
coconut shell ash (CSA) has excellent natural structure and has low ash content.
CSA has proved to be a beneficial reinforcement in the production of lightweight
MMCs. They also offer high thermal resistance [22]. Fly ash (FA), rice husk ash
(RHA), palm oil fuel ash (POFA), palm oil clinkers (POC), and sugarcane bagasse
have shown possible compatibility with MMCs in making them lightweight and
improving their mechanical, wear, and thermal properties. The main reason behind
their compatibility is silica content that is common in all the ashes mentioned above,
which is further processed to form SiC which acts like a reinforcement [23–27].
The use of fly ash with AA6061, AA356, and AA (LM25) has proved to enhance
the most of the properties of the base material. Also, the added advantage is the low
cost of reinforcements [28–30]. The use of fly ash with AA6061, AA356, AA(LM25)
has proved to enhance most of the properties of the base material. The use of rice
husk ash RHA with AA356, AA6061, and in the fields of construction cement has
enhanced most of the needed properties. The use of RHA with other materials to
Characterization of Mechanical Properties of Different … 147
prepare MMCs has proved to be beneficial and a promising substitutes to the existing
synthetic reinforcements [31–33]. The coconut shell ash and RHA contain a consid-
erable amount of silica in them which can be further processed into silicon carbide
or can be used as it is. The use of CSA and RHA has been incorporated into MMCs
which has turned into potential automotive applications [34, 35].
Mishra et al. [36] studied the effect of rice husk ash (RHA) derived reinforcements
on AA(LM6) MMC. The composite was prepared by adopting stir casting route.
About 6% by weight RHA was added in the process. The mechanical and wear
properties were investigated. The composite prepared was much lighter than the
original alloy. The hardness of the composite was decreased on addition of RHA as
compared to AA(LM6) alone.
Kenneth et al. [37] investigated the compatibility of RHA with alumina rein-
forced Al–Mg–Si alloy matrix. Their research aimed at preparing high-performance
composite in low cost. The composite was processed using stir casting route. The
alumina particles were added to varying weight % of RHA to prepare the reinforce-
ments. The samples with 10% reinforcements were processed. After testing, it was
found that the hardness of the composite decreased as the % reinforcement of RHA
was increased. The density of the composite decreased considerably as compared to
the original alloy. The specific strength of the composite is increased by 2% than the
original alloy.
148 A. J. Deshmukh et al.
Santosh et al. [38] investigated the properties of Al-SiC-FA hybrid MMC. The
composite was processed by the stir casting method. The reinforcements were
processed by mixing 2–5% of FA in 3–9% of SiC. The samples were prepared
and tested for compression test, tensile test, and Breinell hardness test. The addition
of SiC and FA to the alloy increased its tensile strength by 45%. The compression
strength decreases with the increase in the size of particles of FA. The BHN increases
from 40BHN for base material to 90BHN.
Apasi et al. [39] studied the wear behaviour of Al-Si-Fe and coconut shell ash
(CSA) MMC. The MMC was prepared by following stir casting route by varying %
weight of CSA from 3 to 15. Wear and hardness of the specimen were studied. The
hardness of the composite increased with increased % of CSA and was maximum for
13% CSA composition. The wear rate decreased notably as the % CSA reinforcement
was increased. The wear rate was minimum for composition 13% CSA, thus making
wear resistant.
Nithyanandhan et al. [40] investigate the mechanical properties of Al-boron
carbide CSA hybrid composite. The research aimed at utilizing agricultural waste
inclusions in MMC. The composite was prepared by following the stir casting
method. Boron carbide with an addition of CSA was used as a reinforcement material.
Specimens were processed by varying the weight of reinforcements in the composite.
Tests were done for tensile and hardness measurement. The tensile strength of the
Characterization of Mechanical Properties of Different … 149
Fig. 11 Comparison of
hardness for different
compositions of coconut
shell ash and silicon carbide
[41]
Fig. 12 Comparison of
hardness for different
compositions of coconut
shell ash and silicon carbide
[41]
• Non-food crops and other bio-renewable resources offer an almost limitless supply
of renewable and potentially sustainable raw materials for the production of bio-
composites.
• The cost of production of these agro-derived reinforcements is much less than the
cost of production of synthetic reinforcements.
Characterization of Mechanical Properties of Different … 153
Fig. 13 Comparison of
density for different
compositions of coconut
shell ash and silicon carbide
[41]
• Many of the agro-derived reinforcements have very less density (e.g.: rice husk
ash, fly ash, etc.). This helps in reducing the weight of the composite they are
used in making them more weight efficient.
• Another advantage is that the properties like tensile strength, compact strength,
etc., of composites having agro-derived reinforcements are quite comparable with
composites having synthetic reinforcements.
• In some cases, i.e. in the case of fibre reinforced composites, some natural fibre
reinforcements are less fire resistant than synthetic fibre reinforcements.
• Agro-derived reinforcements offer a good weight to strength ratio, are economical
to process and help environment in recycling the agricultural waste produced, to
be used as reinforcements in composites.
• Synthetic reinforcements though costlier to produce and being high in density are
always superior to the agro-derived reinforcements when it comes to enhancing the
mechanical properties of the composites. Though both the types of reinforcements
give comparable mechanical properties, synthetic reinforcements though not by
much, are better than agro-derived reinforcements.
Most of the metal matrix composites and hybrid MMCs are produced via two routes
solid route and liquid route [42].
154 A. J. Deshmukh et al.
In solid route, the resulting MMCs are formed as a result of bonding metal matrix’s
and the dispersed reinforcements’ mutual diffusion. This diffusion takes place at
elevated temperature and high pressure. These elevated temperatures are though
less than the temperatures encountered in liquid routes. This helps in depressing
undesirable reactions on the boundaries of matrix and reinforcing media. MMCs
can also be deformed after the sintering operation is complete. This can be achieved
by drawing, extrusion, forging, or rolling. These operations can be cold, i.e. below
recrystallization temperature or hot, i.e. above recrystallization temperature. There
are two main groups in solid-state processing of MMCs, namely diffusion bonding
and sintering/powder metallurgy.
Diffusion bonding: It is a solid route process in which the matrix is in the form of
foils and the reinforcements are in the form of long fibres. These fibres are stacked in
a particular order and then they are pressed at high elevated temperatures. The final
product or material is a composite material having multiple structure. The variants
in diffusion bonding are shown in Fig. 14.
Sintering/Powder metallurgy: Sintering is the method involving consolidation
of powder grains by heating the “green” compact part to a high temperature below the
melting point, and the material of the separate particles diffuses to the neighbouring
powder particles that is shown in Fig. 15.
Powder metallurgy processes have become a very aspect of MMCs, this is because
in powder metallurgy more volume fraction of reinforcements can be added than in
liquid routes. The powder metallurgy route also makes it possible to process MMCs
with nano-reinforcements [43, 44]. Various densities of the end product can be gained
by adjusting the compaction pressure. Hence, making powder metallurgy is very
favourable process [45]. End products which do not require further fabrication can
be easily manufactured using powder metallurgy (PM). Various problems related
to wettability, clustering of particulates, agglomeration, etc., are taken care of in
PM [46]. The use of the PM route in production of AMCs has resulted in desirable
uniform distribution of reinforcements. The properties such as density, hardness,
and microstructure can be altered by changing compaction pressure alone. In some
As compared to solid-state processing, liquid state methods are relatively easy, since
most of the AMCs prepared have a low melting point. This is a process wherein
the reinforcements are incorporated into molten metal matrix. Various methods in
liquid route involve at least partial melting of metal matrix before it is brought into
contact with the reinforcement. This facilitates an intimate bond and thus a stronger
bond, but it sometimes also leads to the formation of brittle interfacial layer. Once
156 A. J. Deshmukh et al.
the reinforcements are incorporated into the metal matrix, the process of mixing and
casting into components of different shapes for further fabrication follows. Great
care must be taken while selecting reinforcement material taking into account it’s
compatibility with matrix is considered. In some cases of reinforcements as in case
of some ceramic reinforcements, they are not wetted by the molten metal matrix.
In such cases, appropriate wetting agents should be added to the molten metal or
the ceramic reinforcements should be appropriately coated. In the liquid route, the
particulates are added to the molten form of the base material. The liquid route is
further divided into two methods, namely stir casting and squeeze casting.
If this method is implemented correctly, this route gives a better distribution of
reinforcements in the matrix. The stir casting method is widely used due to ease of
process. The flexibility of complex shapes can be easily processed in stir casting. [50,
51]. Processes such as friction stir casting are now being used to produce high-quality
surface finish metals. [52]. Particle size, wettability, stirring speed, and environment
are some of the parameters one needs to consider while starting of the stir casting
route.
Stir casting: In this process, the solid reinforcement is just mixed with the molten
metal and allowed to solidify in a suitable mould. The mixture can be prepared by
progressively adding required amount of reinforcement into the molten metal, while
the molten metal is being continuously agitated and stirred. The matrix material is
raised to a temperature above its melting point before the reinforcements are added.
It is one of the simplest and most economical process to produce AMCs. The process
set up is illustrated below in Fig. 16.
Squeeze casting: In this process, a pre-specified amount of molten metal is poured
into a preheated die or die cavity which is placed in a hydraulic press. Then, the molten
metal is pressurized by closing the die cavity by activating the press. The die is kept
under pressure until the molten metal completely solidifies. Since pressure applied
directly affects the nucleation of gas porosity, porosity due to dissolved gasses is
restricted. The last step involves removal of die and removal of final product. The
process of squeeze casing is illustrated in Fig. 17.
Name of author Name of paper Matrix and synthetic reinforcements Research work and major findings
used and route of preparation
Mishra et al. [36] Effect of rice husk ash reinforcements on The composite was prepared by adopting The mechanical and wear properties were
mechanical properties of aluminium alloy stir casting route. They studied the effect investigated. The composite prepared was
(LM6) matrix composites of rice husk ash (RHA)-derived much lighter than the original alloy. The
reinforcements on AA(LM6) MMC. hardness of the composite was decreased
About 6% by weight, RHA was added in on addition of RHA as compared to
the process AA(LM6) alone
Kenneth et al. [37] Fabrication characteristics and They investigated the compatibility of Their research aimed at preparing
mechanical behaviour of rice husk RHA with alumina reinforced Al–Mg–Si high-performance composite in low cost.
ash—alumina reinforced Al–Mg–Si alloy alloy matrix. The composite was The alumina particles were added to
matrix hybrid composites processed using stir casting route varying weight % of RHA to prepare the
reinforcements. The samples with 10%
reinforcements were processed. After
testing, it was found that the hardness of
the composite decreased as the %
reinforcement of RHA was increased. The
density of the composite decreased
considerably as compared to the original
alloy. The specific strength of the
composite increased by 2% than the
original alloy
(continued)
A. J. Deshmukh et al.
Table 1 (continued)
Name of author Name of paper Matrix and synthetic reinforcements Research work and major findings
used and route of preparation
Santosh et al. [38] Novel aluminium-SiCp-fly ash hybrid They investigated the properties of The reinforcements were processed by
metal matrix composites: synthesis and Al-SiC-FA hybrid MMC. The composite mixing 2–5% of FA in 3–9% of SiC. The
properties was processed by the stir casting method samples were prepared and tested for
compression test, tensile test, and Breinell
hardness test. The addition of SiC and FA
to the alloy increased its tensile strength
by 45%. The compression strength
decreases with the increase in the size of
particles of FA. The BHN increases from
40BHN for base material to 90BHN
Apasi et al. [39] Wear behaviour of Al-Si-Fe alloy/coconut They studied the wear behaviour of Wear and hardness of the specimen were
shell ash particulate composites Al–Si–Fe and coconut shell ash (CSA) studied. The hardness of the composite
MMC. The MMC was prepared by increased with increased % of CSA and
following stir casting route by varying % was maximum for 13% CSA composition.
weight of CSA from 3 to 15 The wear rate decreased notably as the %
Characterization of Mechanical Properties of Different …
addition of reinforcements
(continued)
Table 1 (continued)
160
Name of author Name of paper Matrix and synthetic reinforcements Research work and major findings
used and route of preparation
Poornesh et al. [41] Comparison of mechanical properties of They studied the properties of CSA-SiC The specimens were prepared by varying
coconut shell ash and SiC reinforced reinforced hybrid aluminium composite. the percentages of SiC and CSA. As the
hybrid aluminium metal matrix The composite was prepared using stir %CSA increased, the density of the
composites casting process composite also decreased is shown in
Fig. 13. The hardness of the MMC
increased with increase in percentage of
reinforcement is shown in Fig. 12. The
tensile strength of the composite increases
with increase in %SiC in the composite is
shown in Fig. 11
A. J. Deshmukh et al.
Table 2 Literature review on agro-derived reinforcements
Name of author Name of paper Matrix and agro-derived reinforcements Research work and major findings
used and route of preparation
Mishra et al. [36] Effect of rice husk ash reinforcements on The composite was prepared by adopting The mechanical and wear properties were
mechanical properties of aluminium alloy stir casting route. They studied the effect investigated. The composite prepared was
(LM6) matrix composites of rice husk ash (RHA)-derived much lighter than the original alloy. The
reinforcements on AA(LM6) MMC. hardness of the composite was decreased
About 6% by weight RHA was added in on addition of RHA as compared to
the process AA(LM6) alone
Kenneth et al. [37] Fabrication characteristics and They investigated the compatibility of Their research aimed at preparing
mechanical behaviour of rice husk RHA with alumina reinforced Al–Mg–Si high-performance composite in low cost.
ash—alumina reinforced Al–Mg–Si alloy alloy matrix. The composite was The alumina particles were added to
matrix hybrid composites processed using stir casting route varying weight % of RHA to prepare the
reinforcements. The samples with 10%
reinforcements were processed. After
testing, it was found that the hardness of
the composite decreased as the %
reinforcement of RHA was increased. The
Characterization of Mechanical Properties of Different …
Name of author Name of paper Matrix and agro-derived reinforcements Research work and major findings
used and route of preparation
Santosh et al. [38] Novel aluminium-SiCp-fly ash hybrid They investigated the properties of The reinforcements were processed by
metal matrix composites: synthesis and Al-SiC-FA hybrid MMC. The composite mixing 2–5% of FA in 3–9% of SiC. The
properties was processed by the stir casting method samples were prepared and tested for
compression test, tensile test, and Breinell
hardness test. The addition of SiC and FA
to the alloy increased its tensile strength
by 45%. The compression strength
decreases with the increase in the size of
particles of FA. The BHN increases from
40BHN for base material to 90BHN
Apasi et al. [39] Wear behaviour of Al-Si–Fe They studied the wear behaviour of Wear and hardness of the specimen were
alloy/coconut shell ash particulate Al–Si–Fe and coconut shell ash (CSA) studied. The hardness of the composite
composites MMC. The MMC was prepared by increased with increased % of CSA and
following stir casting route by varying % was maximum for 13% CSA composition.
weight of CSA from 3 to 15 The wear rate decreased notably as the %
CSA reinforcement was increased. The
wear rate was minimum for composition
13% CSA, thus making wear resistant
Nithyanandhan et al. [40] Investigation of mechanical properties on They investigate the mechanical Boron carbide with an addition of CSA
aluminium based hybrid composites properties of Al-boron carbide CSA was used as a reinforcement material.
hybrid composite. The research aimed at Specimens were processed by varying the
utilizing agricultural waste inclusions in weight of reinforcements in the
MMC. The composite was prepared by composite. Tests were done for tensile
following the stir casting method and hardness measurement. The tensile
strength of the composite reduces with
inclusions of reinforcements as compared
with AA6061. The hardness of the
composite decreased slightly after the
A. J. Deshmukh et al.
addition of reinforcements
(continued)
Table 2 (continued)
Name of author Name of paper Matrix and agro-derived reinforcements Research work and major findings
used and route of preparation
Poornesh et al. [41] Comparison of mechanical properties of They studied the properties of CSA-SiC The specimens were prepared by varying
coconut shell ash and SiC reinforced reinforced hybrid aluminium composite. the percentages of SiC and CSA. As the
hybrid aluminium metal matrix The composite was prepared using stir %CSA increased, the density of the
composites casting process composite also decreased is shown in
Fig. 13. The hardness of the MMC
increased with increase in percentage of
reinforcement is shown in Fig. 12. The
tensile strength of the composite increases
with increase in %SiC in the composite is
shown in Fig. 11
Characterization of Mechanical Properties of Different …
163
164 A. J. Deshmukh et al.
world, the consumption of raw materials has increased, thereby increasing industrial
waste. Along with the industrial waste, wastes from agriculture sector are a growing
concern. Keeping the goals of protecting the environment, increased importance
has been placed on developing recycling techniques for agro-industrial waste prod-
ucts. The use of agro-industrial waste-based reinforcements in MMCs as a substi-
tute for synthetic reinforcements would prove to be cost-effective as well as the
waste would be used appropriately. The future scope involves advanced techniques
of extracting reinforcements from agricultural waste and mathematical modelling
for evaluating properties of MMCs with varying % of reinforcements. This paper
attempts to review the AMCs reinforced with different synthetic and agro-industrial
waste-based reinforcements. With a thorough review, the following conclusions were
drawn out:
The use of different synthetic reinforcements in a metal matrix makes it possible
to tailor the required the need of the application. Agro-industrial waste-based rein-
forcements reinforced in an MMC give enhanced properties than the base metal in the
matrix. Agro-industrial waste-based reinforcements prove to be the cheaper substi-
tute to synthetic reinforcements. The use of agro-derived reinforcements in compos-
ites can help composite manufactures not only in making economic composites but
also weight efficient. Use of agro-derived reinforcements in composites also helps
environment by recycling the agro-derived waste into useful materials. The mechan-
ical properties of composites with reinforced with agro-derived reinforcements and
synthetic reinforcements are quite comparable. Though the synthetic reinforcements
are costlier to produce and have a bit higher strength to weight ratio than agro-derived
reinforcements, the mechanical properties of composites with synthetic reinforce-
ments are always a little better than composites with agro-derived reinforcements.
Taking an overall view, agro-derived reinforcements have great potential to lower the
factors causing environmental pollution and help in manufacturing the composites
with a very little or no reduction in mechanical properties of composites. A hybrid
MMC of synthetic and agro-industrial reinforcements is cost-effective, as with more
than 60% synthetic reinforcements can be replaced by agro-industrial reinforcements
without degrading the performance of MMC. The most common technique used in
the preparation of MMCs is stir casting and powder metallurgy. However, PM has
an added advantage of allowing incorporation of more than 30% of reinforcements
in MMCs. Future scope involves studying different agricultural wastes which can
act as a replacement to synthetic reinforcements with minimal or no processing and
finding out the right combination of synthetic and agro-derived reinforcements in
hybrid composites to give state of art mechanical properties.
References
1. McAuley JW (2003) Global sustainability and key needs in future automotive design. Environ
Sci Technol 37(23):5414–5416
Characterization of Mechanical Properties of Different … 165
24. Omaltola KM, Onojah AD (2009) Elemental analysis of rice husk ash using X-ray fluorescence
technique. Int J Phys Sci 4:189–193
25. Abdul Khalil HPS, Fizree HM, Jawaid M, Alattas OS (2011) Preparation and characterization
of nanostructured materials from oil palm ash: a bio-agricultural waste from oil palm mill.
BioResources 6(4):4537–4546
26. Robani R, Chan CM (2009) Reusing soft soils with cement-palm oil clinker (POC) stabilisation.
In: Proceedings of the international conference on civil engineering and education in the 21st
century (ICEE2009). (Sarawak, Malaysia, 23–25 March 2009) (Institution of Engineers)
27. Sn W (2008) Sugarcane Bagasse: how easy is it to measure its constituents? pp 266–273
28. Bharathi V, Ramachandra M, Satya S (2017) Influence of fly ash content in aluminium matrix
composite produced by stir-squeeze casting on the scratching abrasion resistance, hardness and
density levels. Mater Today: Proc 4:7397–7405. https://doi.org/10.1016/j.matpr.2017.07.070
29. George B, Sankar A (2018) Fabrication and characterization of aluminum hybrid composite.
Int J Eng Sci Res Technol 7(4)
30. Prasad DS (2011) Production and mechanical properties of A356.2 /RHA composites. Int J
Adv Sci Technol 33
31. Hossain MR, Ali M, Amin MA, Kibria G, Ferdous MS (2017) Fabrication and performance test
of aluminium alloy-rice husk ash hybrid metal matrix composite as industrial and construction
material. Int J Eng Mater Manuf 2:94–102. https://doi.org/10.26776/ijemm.02.04.2017.03
32. Chakravarthi P, Ramnath V, Vetrivel M, Ahmed AR, Buhari ASASM, Muthukumaaran C,
Hilary IA (2018) Design and fabrication of impact die for powder metallurgy. Mater Today
Proc 5:329–334. https://doi.org/10.1016/j.matpr.2017.11.089
33. Raji AA, Waziri H, Hassan NM (2014) Aluminium alloy—rice husk ash composites production
and analysis. Leonardo Electr J Practices Technol 13:84–98
34. Madakson PB, Yawas DS, Apasi A (2012) Characterization of coconut shell ash for potential
utilization in metal matrix composites for automotive applications. Int J Eng Sci Technol
(IJEST) 4
35. Gorthy P, Pudukottah M (2004) Production of silicon carbide from rice husks. J Am Ceram
Soc 82:1393–1400. https://doi.org/10.1111/j.1151-2916.1999.tb01929.x
36. Mishra P, Mishra P, Rana RS (2018) Effect of rice husk ash reinforcements on mechanical
properties of aluminium alloy (lM6) matrix composites. Mater Today Proc 5(2)
37. KanayoAlaneme K, Akintunde IB, ApataOlubambi P, Adewalec TM (2013) Fabrication char-
acteristics and mechanical behaviour of rice husk ash—alumina reinforced Al-Mg-Si alloy
matrix hybrid composites. J Market Res 2(1):60–67
38. Santhosh N, Kempaiah UN, Sunil GS, Reddy HN (2017) Novel aluminium-SiCp-fly ash hybrid
metal matrix composites: synthesis and properties. J Aerosp Eng Technol 7:26–33
39. Apasi A, Madakson PB, Yawas DS, Aigbodion VS (2012) Wear behaviour of Al-Si-Fe
alloy/coconut shell ash particulate composites. Tribol Ind 34:36–43
40. Nithyanandhan T, Rohith K, Sidharath CG, Sachin C, Jagadesh S (2017) Investigation of
mechanical properties on aluminium based hybrid composites. Int J Innov Res Sci Eng Technol
6(7)
41. Poornesh M, Saldanha JX, Pinto GM (2017) Comparison of mechanical properties of coconut
shell ash and SiC reinforced hybrid aluminium metal matrix composites. Am J Mater Sci
7(4):116–119
42. Surappa MK (2003) Aluminium matrix composites: challenges and opportunities. Sadhana
28:319–334. https://doi.org/10.1007/BF02717141
43. Miracle DB (2005) Metal matrix composites—from science to technological significance.
Composites Sci Technol 65(15)
44. Rajmohan T, Arumugama S, Palanikumar K (2014) Synthesis and characterization of
sintered hybrid aluminium matrix composites reinforced with nanocopper oxide particles and
microsilicon carbide particles. Compos Part B Engineering 59
45. Heckel RW (1961) Density-pressure relationship in powder compaction. Trans Metall Soc
AIME 221:671–675
Characterization of Mechanical Properties of Different … 167
46. Akhlaghi F, Zarebidaki A (2009) Influence of graphite content on the dry sliding and oil
impregnated sliding wear behavior of Al 2024–graphite composites produced by in situ powder
metallurgy method. Wear 266:37–45. https://doi.org/10.1016/j.wear.2008.05.013
47. Arockiasamy A, German R, Wang P, Horstemeyer M, Morgan W, Park SJ, Otsuka I (2011)
Sintering behaviour of Al-6061 powder produced by rapid solidification process. Powder Metall
54:354–359. https://doi.org/10.1179/003258909X12573447241626
48. Solay A, Mohan B (2012) Effect of particle size, compaction pressure on density and mechanical
properties of elemental 6061Al alloy through powder metallurgical process. Int J Mater Eng
Innov 3:259–268. https://doi.org/10.1504/IJMATEI.2012.049265
49. Dixit M, Srivastava R (2018) Effect of compaction pressure on microstructure, density and
hardness of copper prepared by powder metallurgy route. In: IOP conference series: materials
science and engineering, vol 377, p 012209. https://doi.org/10.1088/1757-899X/377/1/012209
50. Hashim J, Looney L, Hashmi J (1999) Metal matrix composites: production by the stir casting
method. J Mater Process Technol 92–93: 1–7. https://doi.org/10.1016/S0924-0136(99)00118-1
51. Rosso M (2006) Ceramic and metal matrix composites: routes and properties. J Mater Process
Technol 175(1–3)
52. Izadi H, Nolting A, Munro C, Bishop D, Plucknett K, Gerlich A (2013) Friction stir processing
of Al/SiC composites fabricated by powder metallurgy. J Mater Process Technol 213:1900–
1907. https://doi.org/10.1016/j.jmatprotec.2013.05.012
Shoe-Based Energy Harvesting Using
Ionic Polymer Metal Composites
Abstract Mechanical energy harvesting from human motion is new technology for
generating clean and reasonable electrical energy for powering wearable devices
used in health monitoring, gait analysis and so on. In this document, design and
analysis of ionic polymer metal composite-based shoe energy harvester is proposed.
Ionic polymer metal composite (IPMC) is a smart material of electroactive polymer
categories and functionally similar to piezoelectric material and can be used as actu-
ator, sensors or energy harvester. The harvester is based on specially designed two
initially curved plates of spring steel attached together back to back and is placed
under the heel of the shoe. It is found that IPMC in series produced the voltage of
6.4 V and power of 0.699 W across the load resistance of 2 at walking speed of
2 steps per seconds. From these results, it is concluded that IPMC can be used for
energy harvesting application. This work demonstrates the feasibility of IPMC for
shoe-based energy harvesting to power wearable sensors.
1 Introduction
The proposed model is used to predict the optimal working region under geometry
and material strain restriction of IPMC [8].
Tiwari et al. proposed the IPMC-based energy harvester from ambient vibra-
tion. The energy harvester is analyzed for two electrodes (gold and platinum) types
and effect of geometrical parameters (length, width and thickness) to find the best
combinations for energy harvester [9]. Tiwari and Kim applied a new geometry disc-
type IPMC for energy harvesting from multidirectional vibration source [10]. Cha
et al. studied underwater energy harvester from bending of annular IPMC. Energy
harvester is modeled as thin plate, and IPMC sensing is modeled as lumped circuit
model made up resistor, capacitor & a Warburg impedance also assessed geometry
effect on power harvesting, by varying inner radius of the annulus systematically
[11]. Vinh N D presented an ocean-based wave energy harvester and found IPMC as
suitable candidate for energy harvesting from ocean waves as its response is rapid to
the wave parameters such as amplitude, frequency and wavelength [12].
Kymissis et al. examine three different devices (uniform strip of PZT, stave of
multilayer PVDF and rotary magnetic generator) that can build into a shoe used for
harvesting energy while walking [13]. Saha designed a shoe-based energy harvester
to convert the heel strike of human motion mechanical energy into electrical energy
by using PZT bimorph and PVDF stave [14]. Jung designed a curved piezoelectric
generator for enhancing the output from harvester by distributing the applied load
across the piezoelectric layer, and also, present design allows operations at the low
frequency [15]. Qian designed and optimized a force amplification frame to transmit
and amplify the vertical heel-strike force to the inner piezoelectric stack deployed in
the horizontal direction. Two heel-shaped aluminum plates are employed to gather
and transfer the dynamic force over the heel to the sandwiched force amplification
frames [16]. Zhao designed a sandwich structure for shoe-based energy harvester
from mechanical energy originated from human motion. The designed structure is
compatible with shoe and also increases comfort during walking [17]. Singh used
two structures to harness the human motion energy, one is sandwich structure, placed
under the ball of foot, and other is curved or oval-shaped structure, placed under the
heel of foot [18].
This paper is a proof of concept to develop a shoe-based energy harvester based
on ionic polymer metal composite, which readily converts foot pressure exerted by
human motion to electrical energy. The harvester is based on two initially curved-
shaped structure, which can be placed under heel of foot. A series of experiment
is done using DSO to characterize the feasibility of open circuit voltage output and
power output from the energy harvester for powering low-powered electronic devices.
172 S. N. Patel and S. Mukherjee
2 Experimental Setup
Fig. 2 Manufactured
IPMC-based shoe energy
harvester
Table 1 Geometric
S. No Geometric entity Symbol Value (mm)
properties of IPMC and steel
plate used for energy 1 Length of the curved plate L 80
harvester 2 Width of the curved plate b 40
3 Thickness of the curved plate t 01
4 Length of the IPMC L1 40
5 Width of the IPMC b1 10
6 Thickness of the IPMC t1 0.2
7 Mechanical displacement w 10
Shoe-Based Energy Harvesting Using Ionic Polymer Metal Composites 173
The human motion is characterized as mechanical motion with low frequency and
large displacement. To use it effectively, the energy harvester should be equipped
with polymer-based, flexible material such as IPMCs and PVDF because brittle
material can break into small pieces due to large deformation. The structure and
working of the two initially curved IPMC energy harvesters are shown in Fig. 3.
It consists of two separately prepared initially curved IPMC strips connected back
to back, where each generator comprises two curved steel plates and two IPMC
strips pasted over each steel plates. When the IPMC-based double curves energy
harvester is pressed by external human force in the middle, the IPMC and substrate
are subjected to compressive and tensile stresses, respectively, and hence, electrical
voltage is generated.
The shoe is equipped with fabricated energy harvester at the shoe heel, and elec-
trical connection is taken out from the shoe. When anybody wears the shoe and
applies pressure into energy harvester, it goes from different stages of deformation
as shown in Fig. 3. Figure 3a shows the initial position of the harvester, where no
pressure is applied. While in Fig. 3b, it is the in-between position of the harvester
when half load of human weight is applied over the harvester. When pressure applied
at the harvester is full of its designed load, it gets deformed completely means both the
steel plates are flat as shown in Fig. 3c in this case. When pressure applied is released,
plates are reverting to its original position due to its high stiffness to deformation as
shown in Fig. 3d, e. When the load is released completely from the harvester, it is in
its original initial position Fig. 3f. A rectifier is attached to the harvester to convert
the AC output from the harvester into the DC signal, so that it can be stored or used
for powering the devices.
174 S. N. Patel and S. Mukherjee
Fig. 4 Rectified voltage generated during walking condition of 1 step per second by the shoe energy
harvester
4 Conclusion
A shoe-based IPMC energy harvester based on double initially curved beam is devel-
oped in this paper, and it can be integrated in a shoe readily for energy harvesting from
Shoe-Based Energy Harvesting Using Ionic Polymer Metal Composites 175
Fig. 5 Real-time snapshot of shoe energy harvester for voltage generation during walking condition
Fig. 6 Frequency response result of IPMC shoe energy harvester power output across 2 resistor
at walking condition 1 step per second
human motion. The prototype is fabricated and tested for energy harvesting applica-
tion. Results show that IPMC can be the best alternative for energy harvesting form
human motion due to its matching character with human motion. It can be concluded
from the results that human walking with 2 steps per second generated more power
for same harvester compared to when human walking at 1 step per second. Normal
walking condition also generated enough voltage. The work presented is an attempt
to use IPMC as harnessing the energy expended during human walking for wearable
sensors.
176 S. N. Patel and S. Mukherjee
Fig. 7 Frequency response of voltage output from shoe-based energy harvester at 2 steps per second
walking condition
Fig. 8 Frequency response of shoe energy harvester for power output across the 2 resistor
References
1 Introduction
IoT with the advent of emerging technologies keeps the things connected and allow
them to interact and exchange data. The concept is applied here in IHMS to exchange
patient’s information with the doctor and provide immediate medication. This system
is much useful in remote areas. Especially in hilly areas, it is very difficult to provide
immediate medication to the patient in the absence of a doctor. This in turn is
increasing the death rate.
Considering these scenarios IHMS is developed. In the project, the patient’s health
is monitored by the sensors connected to his or her body using Raspberry Pi as the
platform. Raspberry Pi gathers information and sends the data to the cloud. An
android app is developed and is used to bridge the gap between doctor and patient’s
caretaker. So that in case of emergencies doctors can guide them and provide better
medication. Communication can be performed by sending SMS in the absence of
Internet facility using GSM module. The data is logged into Firebase cloud. IHMS
app is developed using Kodular which is used to build apps without coding. This
system is devised to reduce frequent visits to the clinic as well as provide immediate
or emergency medication.
2 Related Work
Based on the reports by World Health Organization (WHO) [1] many people lose their
lives due to lack of proper medication. Immediate medication includes delivering
quality medical care at remote areas like islands, hilly areas where landslides are
frequent and in situations like ship voyages, military camps is quite a challenging task.
Various smart health care systems have been developed to provide basic medication
and various technologies like Bluetooth, Internet, etc. is used. Renouf and Pollard
[2] declared some challenges regarding emergency medicine in remote areas.
According to Mohammadzadeh and Safdari [3] presented many e-Health systems
with various technologies that are being used. Ajay and Prabhakaran [4] mentioned
the role of cell phones in developing health care settings. MobiHealth [5] a project
funded by European Union provides health monitoring but it makes use of a public
operator to inform the details of a doctor or caretaker. Hameed et al. [6] proposed
a health monitoring system where sensor data is collected and sent to the cloud
for further investigation. Shiva Rama Krishnan et al. [7] proposed patient health
monitoring system using Arduino. Ng et al. [8] highlighted the importance of various
wireless technologies for telemedicine. Boulos et al. [9] mentioned how smartphones
are playing a major role in healthcare.
Similarly, many researchers are still working in this field to provide cheaper and
better solutions.
3 Working of IHMS
As smartphone and Internet became daily needs and the advantage of Internet of
Things is considered to develop the project. The project is divided into two stages.
Stage 1 includes a hardware platform here Raspberry Pi is considered and wearable
sensors like temperature, blood pressure, heart beat, ultrasonic sensor, buzzer and
GSM modules are used. Raspberry Pi is chosen as it is a powerful single board
computer and is faster than Arduino and its flexibility to support various operating
systems. The sensors are connected to the patient using Raspberry Pi which collects
sensor data. Figure 1 depicts the hardware connections. Values aggregated from the
Raspberry Pi are uploaded into firebase cloud server via Python shell.
Stage 2 is Android application called IoT Based Health Monitoring System
(IHMS) is developed and is used as a bridge to communicate between doctor and
patient care taker. A chat window is designed in such a way that there is two-way
IoT-Based Health Monitoring System (IHMS) 181
communication. Whenever the sensor value crosses threshold, alert messages are
sent to the doctor as well the care taker. Doctor can check the messages and send the
prescribed medication. This will be helpful if doctor is away and when immediate
medication is required.
IHMS is developed using Kodular [10] which is used to develop simple android
application without the need for any scripting languages. In Kodular, the app can be
built as combination of components which are the basic building elements. Compo-
nents can be used to design user interface, communication with Firebase database.
Once the design is completed APK is built and can be downloaded or exported and
installed on the mobile. Figure 2 is the flowchart that describes the working of IHMS.
In IHMS the absence or unavailability of Internet is also considered. When the
sensor values cross the threshold alert is not only sent using IHMS but also an SMS
is sent to the doctor’s mobile. In hilly areas as 4G services may not be available this
scenario is also considered. For this purpose, a GSM module is also used. For user
flexibility, many network operators are providing SMS services in regional languages
also. A voice message can also be considered.
The patients, doctors and caretakers details are maintained in the Firebase cloud. The
database has the details of multiple patients, doctors and their respective care takers.
In the project, we considered four different patients with their details uploaded in
the cloud and updated frequently. Figure 3 depicts the temperature details of all the
patients. Similarly, various sensors data is collected and updated.
Figure 4 shows the design of Android app without using any coding techniques.
Kodular becomes very handy to design these apps by simply using the blocks and
components. As per the design, an app is constructed and it provides communication
182 P. Ramya and L. Padmalatha
in both ways that is both doctor and caretaker and vice versa. Using this the develop-
ment time is also reduced without much effort for learning any scripting languages
for building an android app. The app contains a login page followed by the details
after successful login and a chat window for communication. Figure 5 shows the
details and chat option.
To ensure security the app is also tested for invalid credentials as shown in Fig. 6.
Obviously, the login fails if the user enters wrong credentials.
IoT-Based Health Monitoring System (IHMS) 183
5 Conclusion
To ensure reliability the system is tested in real-time and is found to be very useful.
This work can be extensively used in remote areas to improve patient’s health and it
remarks the digital assistance of healthcare. Number of sensors and services can be
extended to provide quality healthcare.
184 P. Ramya and L. Padmalatha
Fig. 6 Authentication
session invalidation
Ethical approval “All procedures performed in studies involving human participants were in
accordance with the ethical standards of the institutional and/or national research committee and
with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.”
Informed consent “Informed consent was obtained from all individual participants included in
the study.”
References
Abstract Natural disasters like earthquakes/landslides are sudden events that cause
widespread destruction, major collateral damage including loss of life. The loss
cannot be completely prevented but it can be reduced. People stuck in rubble/debris
from the collapsed buildings can be rescued, but that operation must be performed
in a very short span of time and requires the intervention of skilled personnel to
avoid further risking, and the lives of the victims trapped underneath. In this paper,
we present the design and control of search and reconnaissance robot (SRR)—a
robot that aims at traversing on all terrains and locating the position of survivors
inside the debris for quick rescue operations and cutting down on the necessity of
trained manpower. The standout feature of SRR from other existing ATVs is the
active articulating chassis to allow climbing and overcoming obstacles of size much
greater than its wheel diameter. Module separation allows the robot to go into tight
spaces where the whole body does not fit. The design of the active articulating
chassis, modularity with a locking mechanism, the entire control of the vehicle and
the various modes of operation are presented.
1 Introduction
Natural disasters are catastrophes arising due to natural causes which result in damage
and destruction of government property, personal property and loss of life. The
numbers show that there has been a heavy toll paid in terms of the number of deaths
by these natural disasters. As per the data provided by the emergency events database
(EM-DAT), 95.6 million people were affected, and 9697 were additionally killed from
the total occurrences of 335 natural disasters in the year 2017 [1]. These numbers
include deaths that could not have been prevented and deaths which could have
been avoided if timely help would have reached or effective rescue operations were
carried out. Incidents like a woman, pulled alive from the ruins of a factory building
in Bangladesh in May 2013, 17 days after it collapsed, and a man survived for
12 days underneath the ruins of a shop after an earthquake, suggest there are chances
of survivors underneath the rubble [2]. Most people, however, would not be lucky
enough to survive that long pinned under debris without food and water, moreover,
they might be hemorrhaging and will require help as soon as possible. Carrying out
these rescue operations are not that easy, the reasons being:
• Heavy machines required to clear out the piles of debris formed,
• Trained and skilled manpower to operate those heavy machines,
• Involves very large areas to be searched for survivors and rescued,
• Disaster struck areas are often difficult to access because of road blockages and
other communication issues,
• The entire process must be carried out within a short span of time, as people might
run out of oxygen, has no access to food and water or might die due to loss of
blood.
In this paper, we present the search and reconnaissance robot (SRR)—a modular
semi-autonomous bot with an active articulating chassis that can be controlled wire-
lessly through Wi-Fi and can move in the air pockets under the debris, while sending
live video feed to the pilot/controller through Wi-Fi which then can be used for
finding the people that are stuck. With the help of an active articulating chassis, the
robot can overcome obstacles of higher sizes than its wheel radius. The robot is
semi-autonomous and is wirelessly controlled with live audio and video feedback.
The robot achieves modularity through a locking mechanism, with which the bot can
split up into smaller units and get into smaller air pockets because of less turning
radius. The combined solution of the articulating chassis and modularity has led to
an increase in the surveyed area which is advantageous. These robots also find their
uses in industrial explorations and cave inspections.
2 Existing ATVs
The major challenge in an ATV design has always been strength and agility. The
design of ATV has always been about striking a balance to your advantage. Several
other all-terrain robots are mentioned below.
Legged robots like the PhantomX (a hexapod) can achieve great agility due to their
complex design but lack the strength to bear physical fall [3]. Legged robots are
also unsuitable for locomotion on softer terrain such as sand. Their legs dig into
the sand and a lot of power is wasted in pulling them out. Whereas the SRR uses
Search and Reconnaissance Robot for Disaster Management 189
ten wheels which are strategically placed to distribute weight uniformly and also to
allow the bot to perform when turned upside down. Eight of these wheels have tracks
to increase the surface area reducing the pressure. With minor changes in design,
the bot can also be adapted to other terrains. Moreover, legged robots with their
rigorous and complicated control are limited by obstacles which can be easily scaled
by the SRR unless unreasonably large. For hexapods, accurate control of all the six
legs is required simultaneously and thus places a requirement of skilled labor for
development and maintenance. The control of the SRR is designed to be intuitive,
requiring little or no training. The control is described in further sections.
SuperDroid robots are innovations with old design. Still stuck with a RC car design.
These robots achieve good strength and can withstand falls up to several stories
high. However, they suffer in terms of agility in very rough terrains due to their
overly simple chassis [4]. Certain designs also suffer from the issue of toppling
upside down rendering the robot useless without human help. Achieving the right
balance between agility and strength is the key. SRR does that to a great extent.
Along with a sturdy design and a high endurance material, features like modularity
and inverted operation promises a long way. Active articulation of the chassis allows
for greater agility. Active articulation allows the robot to climb obstacles that are
much higher than its wheels’ radius within reason while the sturdy design allows for
greater strength.
It is an ATV cum quad bike designed for purposes varying from agriculture to search
and rescue. While in the artificial intelligence, its features are impressive and the
ATV hardly shows practical application because of its size and cost. Considering the
terrain and scenarios, we are targeting robot’s size plays a major role in determining
its performance. The Honda’s 3E-D18 [5] offers great speed on ground speed which
often serves as a disadvantage on unstable surfaces. The weight of these robots can
also affect the debris on which they tread on. The SRR was designed keeping all
of this in mind. Its small size allows great reachability inside the debris. It makes
air pockets accessible. The lightweight ensures light footmark on unstable surfaces
which ensures safety in such unpredictable areas.
190 S. Narayan et al.
3 Design
The design of SRR can be broken down into sections, namely material of the body,
articulating chassis and modularity. The body and design of the robot aim at achieving
the following operational and performance objectives:
• The robot should easily overcome obstacle much greater than its wheel radius
thus ensuring maximum reachability in the debris.
• It should endure falls ranging from minor to heights around 1 m. Impact of small
debris should not cause damage to chassis or any component for that matter.
• All the components including peripherals, onboard computers, batteries and
circuitry should be accommodated in the least possible area to ensure minimum
robot size.
• To fit into tight places, which has always been a challenge for ATV designers.
• The overall cost must be kept low so that the loss of these robots inside the debris
can be afforded.
3.1 Material
3.2 Specifications
The dimensions of the individual segments in SRR are specified in Table 1. Table
2 shows the comparison of SRR with other ATVs, in the fields of weight and
dimensions.
The dimensions of the locking mechanism are 38 mm × 143 mm × 47 mm.
Thickness of the aluminum alloy used: 0.2 mm.
Density of aluminum alloy: 2720 kg/m3 .
Search and Reconnaissance Robot for Disaster Management 191
Table 1 Dimensions of
Segment no. Length (mm) Breadth (mm) Height (mm)
individual segments
Segment 1 220 170 80
Segment 2 190 170 80
Segment 3 150 170 80
The robot stands on a segmented chassis. The chassis is segmented into three
segments which have been precisely calculated dimensions to minimize its span
and accommodate all the required components inside the robot. The three segments
are joined through two different active joints. The numbering of the segments is
shown in Fig. 2. This numbering will be referred to throughout the paper. The third
192 S. Narayan et al.
Fig. 2 a Numbering of the segments where segment 1 is the front segment, b climbing using active
articulation, c the detachable joint between second and first segment with links, d location of high
torque servos for active articulation shown in red
and second segment is linked permanently through an active joint which is actuated
by two metal gear servos. The servos are calibrated carefully to move together at
equal angular velocities.
The first segment has a locking mechanism fastened to its back face. This mech-
anism enables the feature of modularity. The functionality of this component is
explained later in the paper. The second segment also carries two high torque metal
gear servos that have arms protruding out from their axes (see Fig. 2). The locking
mechanism latches onto these arms and the chassis articulates at this joint. The active
articulation of chassis in ATVs has not been given its fair amount of thought until now.
The major advantage of this design is that it allows the bot to climb obstacles much
greater than its wheel’s radius. In practice the SRR is recorded to easily climbing
obstacles of above 250 mm, with potential to overcome even bigger obstacles. This
feature of active articulation immensely increases the reachability of robot in the
debris. It alone widens the scope of practical application. SRR can form various
configurations depending on the dimensions and shape of the obstacle encountered
(see Fig. 2). The first segment and last segment move on four wheels of radius 90 mm.
These wheels are bounded by caterpillar tracks for maximum traction. The second
segment moves on just two wheels of 110 mm.
Search and Reconnaissance Robot for Disaster Management 193
Two three-cell lithium polymer batteries are used must power the whole bot. Each
battery pack when fully charged has an EMF of 12.6 V. The first battery pack is
present in the front segment and has a capacity of 2200 mAh. The second battery
is present in the middle segment and has a capacity of 4500 mAh. The reason for
placing the batteries in the first two segments is to shift the center of mass more
toward the front segment. The reason being, when the bot climbs over a tall object, it
first places the first segment over the obstacle, then the second and finally it lifts the
third segment. Having the center of mass shifted toward the front segment, ensures
that the bot will not topple backward when the first two segments are over the obstacle
and the third segment is still on the ground.
The main challenge was to provide three different voltages while ensuring enough
current supply to all the components. Two buck converters have been used to achieve
this (see Table 3). The motors are powered directly from the battery pack. A 9 A
buck converter tuned to output 8.4 V is used to power four high torque servos.
Another 9 A buck converter tuned to output 5.5 V is used to power the onboard
single board computer (Raspberry Pi) and other electronics including Arduino Mega
2560, onboard cameras and three micro servos for pan-tilt and camera retraction
mechanism.
The battery in the first segment powers, a Raspberry Pi and Arduino Mega placed
in the first segment, three LEDs in the headlight, 4 motors in the first segment and
the servo for the locking mechanism. The battery in the second segment powers the
rest of the electronics.
The operating time of the robot greatly depends on the terrain. On moderately rough
terrain, we were able to get an average operating time of 1 h. On highly rough terrain,
the bot must undergo several articulations, thus the average operating time reduces
to 40 min.
This section proposes the feature of modularity and its unique approach of imple-
mentation through a locking mechanism. This feature alone increases the application
set of SRR.
5.1 Modularity
Modularity opens a different horizon for new application sets. Different segments
can be designed and programmed for entirely different functionalities while keeping
the articulation system’s advantages. For instance, interfacing necessary sensors to
a segment could make it specific to collection of data in gas leaks, landslides or
any disaster for that matter. This can be attributed as an advantage of using the
ROS framework which gives us high modularity in designing the robot as ROS’s
nodal structure is inherently modular. Modularity also allows covering of larger areas
through small segments thus speeding up the process of rescue or data collection all
while keeping the basic feature of all-terrain locomotion intact. In the current stage,
the modularity has been implemented through a locking mechanism proposed next
in the paper.
The mechanism is fastened to the hind side of the first segment of the bot which
must detach. The second segment of the bot has two arms (link) fastened to servo
heads extending forward from the second segment (see Fig. 2c). The two links are
extended out into the arms through the guides in order to lock the two modules
together. When we require the bot to split up, the servo turns anticlockwise pulling
the earlier defined latches back inside and freeing the first segment from the rest of
the bot.
However, the individual modules cannot cross larger obstacles and would be
required to combine back together if needed to cross over one. The pilot/operator
controlling the bot will have to manually align the two halves using the feed from
the onboard cameras and then perform docking. This is assisted by the spring like
tension put by the servo over the latch which pulls the modules into alignment even
if the modules were slightly misaligned. This can be used to dock the modules with
ease.
This mechanism also leaves the servo motor free from the torque when the chassis
articulates. This happens because the torque affects the guides instead of the motor.
Thus, providing a cost-effective solution for attachment and the detachment of the
modules.
This section explains the complete control of the bot and how that makes it unique
from other existing all-terrain robots.
Search and Reconnaissance Robot for Disaster Management 197
An ultrasonic sensor has been used for assistive autonomous climbing. Owing to the
highly uncertain terrain, the operator can switch between completely manual control
and assistive autonomous control. If assistive autonomy is enabled, then the ultrasonic
sensor attached to the first segment is used to detect an obstacle within 150 mm of
its range. If an obstacle is detected, it performs the four different climbing modes in
a sequence. With assistive autonomy, forward command is enough to overcome any
obstacle or traverse any terrain. If no object is detected, it continues to respond to
commands as sent by the operator. The bot remains in assistive autonomy until the
198 S. Narayan et al.
The communication between the bot and the control station takes place using the ROS
framework running on a Raspberry Pi 3 model B present on the bot. The Raspberry
Pi communicates with the onboard Arduino via Pi serial. The control commands
along with low latency video transmission occur wirelessly between a PC and the
onboard Raspberry Pi over Wi-Fi. The control commands are given using a controller.
The range of bot operation is limited only by the range of the Wi-Fi router. Once the
commands are received by the Raspberry Pi present in the bot, it sends the commands
to Arduino via serial communication. Arduino Mega 2560 and Arduino Nano 328p
Search and Reconnaissance Robot for Disaster Management 199
are the microcontrollers which are responsible for providing control signals to the
DC motors, servo motors and other peripherals. Basic ROS framework is shown in
Fig. 7.
The bot has two cameras to provide live video feed. Both the cameras are interfaced
to Raspberry Pi. The camera with mic is used for the live audio feed and is fastened to
the front face of the first segment, while the other camera is in the last segment. The
front camera is immobile, but the back camera is placed on a pan and tilt mechanism to
provide 270 viewing angles for better surveillance and navigation. The back camera
has been placed inside the last segment this prevents it from acting as a hindrance
while climbing. The back camera is attached to a coupler which is actuated a servo
placed in the last segment. This allows the operator the freedom of popping the back
camera in and out of the last segment. This allows the operator to control the height
of the back camera thus controlling the angle of view to the operator’s need. Figure 8
narrates the implementation of live video feed through ROS [9].
Fig. 8 ROS nodes involved for video transfer by Raspberry Pi to the control station
mapping (SLAM) [11] can be combined with trial-iteration using sensor fusion [12]
to generate a map of what’s inside the rubble. This could help the rescue team to find
their way through the rubble.
References
11. Accurate monocular visual-inertial SLAM using a Map-assisted EKF approach. Submitted on
12 Jun 2017 (v1), last revised 31 Mar 2018 (this version, v3)
12. Sensor data fusion using Kalman filter. In: Proceedings of the third international conference
on information fusion
Experimental Validation of Various
Existing Impedance Models for Acoustic
Liners
Abstract This work is to find out the suitability of existing impedance models and
their drawbacks. Predicted absorption coefficients using various impedance models
are compared with the experimental results for different configuration of liners. In
experiments, the realistic flow conditions are being imitated; for the same predic-
tion and from various impedance models and experimental results, they have been
compared. It concludes that the Bauer impedance model is more realistic and has
good agreement with the experimental results. Elderge and Dowling method is better
than the transfer matrix method (TMM).
1 Introduction
find out the most consistent and suitable impedance model for acoustic liner and
finding of this work will help to design an acoustic liner. In Sect. 2, various existing
models have been discussed. The experimental measurement technique is described
in Sect. 3. Finally, Sect. 4 presents the results of analytical models and experimental
measurements.
2 Analytical Models
where the normalized real part γ and normalized imaginary part δ [2] are given by
I12 (st) 1 + st1 + 4 2st
π2
e cosh(st)K 12 (st) cosh(st) − sinh(st)
st
γ = (2)
I12 (st) + π42 e2st cosh(st)K 12 (st)
2
I (st)K 1 (st)e2ka
πst 1
δ= (3)
I12 (st) + 4 2st
π2
e cosh(st)K 12 (st)
where t and r are thickness and radius of orifice, St is the Strouhal number, and I1
and K 1 are the modified Bessel functions of first and second kind.
The perforate impedance [3], ξBetts and Rayleigh conductivity K a are given by
√
4μl 2ωμl 1 − σ 2 eff
ξBetts = + + MB
cσ Cdr 2 cσ Cdr σ Cd
√
kl 2ωμl 16
+i + + (4)
σ Cd cσ Cdr 3π ψ
and
Experimental Validation of Various Existing Impedance … 205
n
K a = ik (5)
ζBetts
where
4σ
n= (6)
π d2
where, vB is bias flow velocity in the cavity, a is area of the hole, c is speed of sound
in the normal incidence tube, σ is open area percentage, Cd is discharge coefficient,
and ρ is density of fluid.
3 Experimental Setup
connected to the downstream tube. In the test section, the cavity of 150 mm diameter is
concentric with the liners of 80 mm diameter. As blowers are of root types, therefore,
to ease the pulsation of flow, plenum chambers are used for both grazing and bias
flow separately. To straighten the flow, two settling chambers have been used for both
the grazing flow separately. Differential manometers are used to measure flow rates.
4 Results
The present study compares the different impedance model for perforated liners with
experimental results to identify the best suitable impedance model to design a liner.
To check the suitability of the models for different orifice diameter and porosity with
different flow conditions, a set of experiments have been performed. This study has
been carried out for two different configurations of liner with the mean bias and
grazing flow. Experiments are performed using ASTM two load method [6]. All the
liners used for this work are of 80 mm diameter and 200 mm of Length. Figure 2a, b
shows the result of liner having orifice diameter 3.2 mm and 2.4% porosity at constant
bias flow at 0.1 Mach with two different grazing flow at 0 and 0.05 Mach, respectively.
These results confirm that the Elderge and Dowling method using Bauer impedance
model has better agreement with the experimental results for different combination of
mean flow. Figure 3a, b shows the result of liner having 3% porosity and 5 mm orifice
diameter at two different grazing flow 0 Mach and 0.05 Mach with a constant bias
flow, i.e., 0.1 Mach. These results confirm that Elderge and Dowling method using
Bauer impedance model has better agreement with experiment. Modified Howe and
Betts models have included the effect of bias and thickness of orifice but they do not
consider the effect of grazing flow, whereas Bauer model has included the effect of
bias, grazing flow and thickness of orifice. Therefore, Bauer model is more realistic.
Bauer model has empirical terms that make it more accurate.
Experimental Validation of Various Existing Impedance … 207
1
a
0.8
0.6
α
0.4
ζ = Bauer
ζ = Betts
0.2
Bauer D&E
ζ = Howe
Measured
0
600 800 1000 1200 1400 1600 1800 2000 2200 2400
Frequency(Hz)
1
b
0.8
0.6
α
0.4
ζ = Bauer
ζ = Betts
0.2 Bauer D&E
ζ = Howe
Measured
0
600 800 1000 1200 1400 1600 1800 2000 2200 2400
Frequency(Hz)
Fig. 2 Comparison of absorption coefficient (α) using different impedance models measured at
bias flow M b = 0.1 and grazing flow a M g = 0, b M g = 0.05 for the liner of orifice dia. 3.2 mm
and porosity 2.4%
5 Conclusion
This study shows that Bauer impedance model is better than the other existing
impedance models. Bauer impedance model is more consistent for different flow
conditions. Bauer impedance model is more realistic because it considers the effect
of bias flow as well as grazing flow. And it is more accurate because it has empirical
terms also. Stagnation enthalpy approach using Bauer impedance model gives better
agreement than the transfer matrix method using the same impedance model. To
design an acoustic liner, prediction using stagnation enthalpy approach, i.e., Elderge
and Dowling method, along with Bauer impedance model is more suitable.
208 A. Tripathi et al.
1
a
0.8
0.6
α
0.4
ζ = Bauer
ζ = Betts
0.2 Bauer D&E
ζ = Howe
Measured
0
600 800 1000 1200 1400 1600 1800 2000 2200 2400
Frequency(Hz)
1
b
0.8
0.6
α
0.4
ζ = Bauer
ζ = Betts
0.2
Bauer D&E
ζ = Howe
Measured
0
600 800 1000 1200 1400 1600 1800 2000 2200 2400
Frequency(Hz)
Fig. 3 Comparison of absorption coefficient (α) using different impedance models measured at
bias flow M b = 0.1 and grazing flow a M g = 0, b M g = 0.05 for the liner of orifice dia. 5 mm and
porosity 3%
References
1. Howe MS (1979) On the theory of unsteady high reynolds number flow through a circular
aperture. Proc R Soc Lond A Math Phys Sci 366:205–223
2. Bauer AB (1977) Impedance theory and measurements on porous acoustic liners. J Aircr
14(8):720–728
3. Betts J, Follet J, Kelly J, Thomas R (2000) Evaluation of an impedance model for perforates
including the effect of bias flow
4. Jing X, Sun X (2000) Effect of plate thickness on impedance of perforated plates with bias flow.
AIAA J. 38(9):1573–1578
5. Eldredge JD, Dowling AP (2003) The absorption of axial acoustic waves by a perforated liner
with bias flow. J Fluid Mech 485:307–335
6. ASTM E2611-09, standard test method for measurement of normal incidence sound transmission
of acoustical materials based on the transfer matrix method (2009)
Design and Modeling of Pipeline
Inspection Robot (PIR) for Underground
Pipelines
Abstract The rapid growth of in-pipe inspection robots has been demanding in the
society which is widely used in oil and gas industry, power plant industry, and sewage
system, etc. This kind of robot can be applicable for inspecting the internal surfaces of
underground pipelines like defects, cracks, and internal erosion. To cater this need,
the main focus of this paper is to design and develop a pipeline inspection robot
(PIR). For this purpose, a novel design of PIR is proposed using scissor mechanism
which consists of two cross-links, and the motion of scissor is adjusted by control-
ling the lead screw in the mechanism. This is controlled through DC motors. The
mathematical model for understanding the kinematics and dynamics of the scissor
mechanism has been developed. Further, kinematic and dynamic simulation has been
carried out. By results, it is proven that the scissor mechanism helps it to automatic
adjustment of the height of the robot in the internal pipelines of robot for varying
cross-sectional diameters. The robot can cover a large diameter range (500–1000 mm)
of pipelines. This robot may solve the problems of inspection of the underground
pipelines. This robot can be minimized the problems of the human factor in labor-
intensive or hazardous work and can also act in unreachable environments during
repair and maintenance inside the underground pipelines in various industries and
day-to-day life.
1 Introduction
Nowadays, the robotics has been demanding fast and rapid development in the
various fields like material handling, coil mines, automobile, inspection, power plant,
biomedical, and defense [1]. In the inspection of pipelines in various industries like
power plant, oil, and gas, water supply system, the intensive and dangerous work
which is inaccessible to human can be easily solved by a robot. Pipeline inspection
robot (PIR) is used to detect the pipe condition, and according to damage, repairing
can also be done by this type of robot [2]. From drinking water to crude oil, sewage
to radioactive material, and also for gases, the inside the underground pipelines have
liquid, gases, and water, etc., which can be transferred from one place to other places
every day. It is problematic to inspect and repair those pipes due to geometrical
and geographical difficulties such as degradation, creep, overheating, corrosion, and
others. To cater this need, in-pipe robot inspection is an admirable solution for solving
this type of problems. From the various literature, it is found that in-pipe robot can be
categorized like wall-press robot, wheel-type robot, and caterpillar robot [3]. It is also
noticed that earlier most of the proposed robot have some limitations like complex
geometry and mechanism, the high tendency of slipping and difficult to move in
various bends of pipelines, etc. [2, 4]. To minimize these issues, a novel design of
wall-press adaptable PIR is proposed in this paper which is intended to be used in
dry well as wet/submerged conditions inside the pipe. It is able to move through
vertical and horizontal pipes, and it can easily pass through the elbow of a pipeline.
In order to find the erosion and corrosion of pipelines, this will be measured by the
variation of the thickness of pipes. The main objective of this paper is on the kine-
matic and dynamic analysis of PIR where scissor mechanism plays a vital role during
closing and opening operations inside the pipelines. In order to achieve the adaptable
wall-press operations, the foot press mechanism is designed. This will provide the
self-adjustment capability at the inner surface of the pipelines during traveling the
robot from one place to another place. This will also prevent the stacking of robot
during the motion of the robot inside the pipelines. The movement of the robot will
be controlled by DC motors which one DC motor is connected to lead screw of
scissor mechanism and foot press mechanism. The sewage camera, ultrasonic probe
(thickness measuring instrument), gas sensor, and control units are attached to the
robot. The gas sensor unit is integrated for identification of the harmful gases like
CH4 , CO, CO2 , and H2S. This robot may solve the problems of inspection of the
underground pipelines such as hazardous gases, damage detection, and regular moni-
toring of pipelines. This robot can be used for a large diameter range (500–1000 mm)
of pipelines.
In previous years, some of the researchers have developed and demonstrated a
few in-pipe robots/systems at the international level. These robots are designed as
per the specific requirement. Ryew et al. [5] have presented a robot system using an
articulated structure like a snake. This is controlled through a tether cable for in-pipe
inspection of underground urban gas pipelines. In this robot, at front and rear end
of the system, two active driving are provided along with passive modules such as a
control module. Choi et al. [6] have focused on underground robotic system for gas
pipe inspection. The robot is also configured as an articulated structure. Tatar et al.
[7] have attempted on the two wheeled-type in-pipe mini robots. The main goal of
this system is the easy adaptability to the inner diameters of the pipes (diameter range
140 and 200 mm). The mini robots are characterized by an adaptable structure, based
on linkage mechanisms. Nassiraei et al. [8] have proposed a sewer inspection method
Design and Modeling of Pipeline Inspection Robot (PIR) … 211
A novel design of PIR using the scissor mechanism for inspection of underground
large diameter pipelines is shown in Fig. 1. The robot has a wall-press-type mech-
anism. The major advantages of this kind of mechanism are that it can adopt the
changes inside the pipe diameter for long-range inspection. It can also carry heavy
loads inside the pipelines. Furthermore, it can avoid getting stuck in pipelines at
various cross sections. The major parts of this robot are body structure, scissor mech-
anism, foot mechanism, thruster, cameras, gas sensor, and thickness measurement
device, etc.
In the body structure, the robot is designed using three scissor mechanisms and
foot mechanism along with thrusters at radial position (at 120°) on the circular
pipe. The scissor mechanism consists of two cross-links which help it to self-adjust
the height of the robot when it passes through pipes with varying cross-sectional
diameters. The scissor mechanism is driven by a DC motor for pressing on to the
wall while robot will travel either horizontally or vertically inside the pipe. Two
DC motors are mounted on the robot; one motor will allow for each opening and
closing the system, and another one motor will help to move forward and backward
the system. When it will reach turning points, then one mechanism will support
the turning point and others will remain the proper position. Then, when it comes
to the horizontal position, it will continue the forward movement through the pipe.
Moreover, this robot is able to pass through bends and T-junctions of pipe. The foot
Thruster
Foot Mechanism
mechanism is designed in such a way that it helps to overcome the minor irregularities
in the pipelines during the motion. The motion of the robot will be provided by DC
motors in which one DC motor is connected to lead screw of scissor mechanism and
foot press mechanism. The sewage camera, ultrasonic probe (thickness measuring
instrument), gas sensor, and control units are attached to the robot. The gas sensor
unit is integrated for identification of the harmful gases like CH4 , CO, CO2 , and H2S.
This robot is designed to adjust itself automatically between pipe diameters ranging
from 500 to 1000 mm.
In order to develop the kinematic modeling of the link mechanism, the robot moves
in the pipelines. The schematic diagram is shown in Fig. 2. A scissor mechanism
connects the body structure and foot mechanism which transmit the motion to the
robot during traveling and providing the support to the inner wall of the robot. In
order to analyze the scissor mechanism, it is divided into two parts, i.e., lower half
and upper half. The lower half of the mechanism provides the motion of the body,
and upper half provides the support to foot mechanism and controlling the reaction
forces during motion as shown in Fig. 3. For the motion of left-hand sides to right-
hand sides, a lead screw is provided whereby two controlling nuts (Nut1 and Nut 2)
positions, the mechanism performs the operation of the scissor mechanism. When
the robot will move inside the pipelines, the normal reaction force (N) will act on the
surface of the body since all three scissor mechanisms are mounted on 120 degrees
radial directions as shown in Fig. 4.
Let the mass of the robot (m) = 10 kg (i.e., weight W = mg).
From the diagram, we can write
3μN ≥ W (1)
Foot mechanism
Leas Screw
Scissor mechanism
Nut2
Nut1
Body Structure
N N
where the coefficient of μ = 0.64 (between the steel and rubber) is taken.
N = W/3μ = (10 × 9.81) (3 × 0.64) = 51.09 (2)
N is the normal force to be applied on the pipe wall by each of the scissors.
In order to analyze the forces in the lower half of the scissor mechanism, the
following reaction forces (free body diagram FBD) are shown in Fig. 5.
where F 1x and F 1y are the force acting at B at x- and y-direction, respectively, and
similarly, F 2x and F 2y are the force acting at A at x- and y-direction, where B is the
junction between two links of scissor, and A is the joint between Nut 2 and thread.
From the diagram, we can write that the sum of the moment (M B ) is zero as given
below;
The acting reaction force (N) for two triangular parts in lower scissor mechanism
is considered.
N = 2F1y = 51.09
From Fig. 5, it is shown that the force acting at the point B (F1x ) is the same as a
force acting at point A (F2x ) in the x-direction. Therefore,
and
F3t
F2x
Portion of thread profile
F3n
α
and
F3t = (30 tan θ ) × cos α. (11)
The total force (Ft ) to be applied on the nut to move along the thread profile is
given by
The force (Ft ) is required to rotate the screw so that the nut moves along the thread
profile.
After putting the value of F3n and F3t from Eq. (10) and (11), we find
Ft = (30 tan θ ) × (cos α + 0.3 sin α) (13)
Tcg,Dcg
Ts = 92.34 tan θ
This lead screw is connected to the gear mechanism as shown in Fig. 7. Three
gears are connected to each other and run through a central gear. The central gear
has a number of teeth (T cg ) and diameter (Dcg ). The outer diameter of three gears
has the same number of teeth and diameter.
So, the required tangential force is
Ftog = Ts (Dog /2)
After putting the values of required torque (Ts ) from Eq. (16), we get
Ftog = 2 tan θ (17)
The same tangential force must be generated by the central gear also Ftog = Ftcg .
Hence, the torque (Tm ) to be generated by the motor is,
Tm = Ftcg × (Dcg 2)
For finding the power requirement for the scissor mechanism, the schematic
diagram for displacement is shown in Fig. 8. The lead screw displacements are
X and Y in x and y directions.
From the diagram, we can write
Y X = tan θ (19)
218 R. K. Jain et al.
or
X tan θ = Y.
dX d dY
tan θ + X tan θ =
dt dt dt
= vx and dY
dX
dt dt
= v y are velocities in x- and y-directions.
The angular velocity dθdt
= ω is considered.
After putting these values, we get
tan θ × vx + X (sec θ )2 ω = v y
For finding the velocities (vx and vy ), let s is to be the rotational speed of the
screw. Hence, the speed of the gear attached to the screw will also be equal to s .
The gear attached to the screw has 100 teeth, so we write
s = 100
The gear train has three gears with numbers of teeth 100, and one central gear has
60 teeth. The module of each gear has one.
So from elementary gear equations, we can write
100 60 = T60 T100
Hence,
vx = (T60 T100 ) × ( p 2π ) × 60 = 0.360 (23)
So,
For finding the angular velocity (ω), the displacement X is related to the length
of the link in the following way:
X = l cos θ
dX dθ
= −l sin θ ×
dt dt
The velocity in the X-direction is given as
ω = −0.360 l sin θ (26)
0.3X 60
v y = 0.360 tan θ − (sec θ )2
l sin θ
Solving for 60 , we get
2
60 = v y .3 tan θ − rad/s (27)
sin 2θ
For v y = 1 mm/s
2
60 = 1 .3 tan θ − rad/s
sin 2θ
Pm = Tm m
After putting the values from Eq. (18) and (28), we get
2
Pm = (60 tan θ ) × 1 .3 tan θ − watt (29)
sin 2θ
After formulating the kinematics of the PIR motion, the numerical simulation for
finding the forces and torque has been carried out using ADMAS software which is
discussed in the next section.
In order to find the design parameters of the PIR, a model is developed in ADAMS
software as shown in Fig. 9. During the development of the model, the scissor mech-
anism is made using the aluminum material. The detail of each component where
different types of joints are given in Table 1. The scissor mechanism is connected
with the lead screw mechanism. This lead screw mechanism is controlled by a DC
motor. For this purpose, the rotating motion is given to lead screw and analyzed at
different conditions. The simulation results are shown in Table 2. In order to find the
maximum travel distance, the simulation results of joint forces and torque at lead
screw point are shown in Figs. 10, 11 and 12.
Lower half of
scissor
Lead Screw
Table 1 Detail of
S. No. Components Type of joint Material
components for the
development of model 1 Scissor (lower and upper) Revolute Aluminum
2 Nut rod and scissor Revolute Mild steel
3 Nut and lead screw Screw Mild steel
4 Nut rod and lead screw Sliding Mild steel
5 Lead screw and end plate Revolute Mild steel
It is found that the maximum travel distance of the lead crew is 12.7 mm at
1.04 rad/s. The maximum joint force between the nut and the lead screw is found
to be 25 N in each direction. Similarly, the maximum torque required for driving
the DC motor is found to be 195 N-mm. For finding the time taken for closing
and opening the scissor mechanism, the travel distance for the various diameter of
pipelines has also analyzed as shown in Table 3. The maximum travel distance of the
scissor mechanism is achieved within 12 s for pipe diameter of 1000 mm as shown
in Fig. 13.
From these analyses, it is found that by controlling the motor speeds, the scissors
mechanism will be helpful in pressing the inner wall in pipelines during traveling the
Table 3 Results for time taken to reach the inner wall of pipelines
S. No. Pipeline diameter (mm) The travel distance of lead Time taken to reach the inner
screw (mm) wall of pipe (s)
1 800 7.5 2
2 900 25.7 7
3 1000 43.4 12
PIR inside the pipelines. This mechanism can be helpful in achieving both tractions
and size adaptability during operation. In order to look inside the pipelines, camera
along with control system will be integrated with the robot. The image can be viewed
in the PC at the control station. Lights with variable illumination capacity will be used
to properly look inside the pipe interior and collect the images which may provide
the information regarding damages in the underground pipelines. This will be helpful
in the estimation of the lifetime of pipelines. This robotic system can also be helpful
to overcome the issues of the human labor for dangerous work and inaccessible
environment. Therefore, by regular maintenance of inside, the underground pipelines
can be minimized the rate of accidents in various industry and day-to-day life. This
is a technological contribution toward Make-In-India and Clean-India Mission.
5 Conclusion
In this paper, a novel design of PIR for underground pipelines is discussed. This
robot is designed using a wall-press mechanism where the modeling of the scissor
mechanism has been carried out for finding the travel distance of the robot inside
the pipelines. By simulation results, it is proven that this robot can be used for
inspection of large diameter underground pipelines (diameter 500–1000 mm). The
scissor mechanism shows the adaptability by varying the size of the workspace,
during operation. The major advantages of this mechanism are that it can adapt to
the changes inside the pipe diameter for long-range inspection. It can also carry
heavy loads inside the pipelines if required. It can avoid getting stuck in pipelines
at various cross sections. The analysis and simulation results will be helpful in the
development of the prototype. This robotic system will be developed and remotely
operated through the dashboard. Furthermore, this robotic system will be deployed
in underground pipeline inspection application like coal mines, gas and oil, water
pipelines and power plant. This is the future scope of this paper.
Acknowledgements The authors are thankful to the Director, CSIR-CMERI, India for granting
permission to publish this paper. This work is carried out under entitled project “Development of
mobile robot technology for inspection of underground large diameter pipelines” of “Robotic Inter-
vention for Industrial and Strategic Applications” (Project No. MLP223512/WP-I) and financially
supported by CSIR, India.
References
1. Reddy AH, Kalyan B, Murthy CS (2015) Mine rescue robot system–a review. Procedia Earth
Planet Sci 11:457–462
2. Berman S, Edan Y (2002) Decentralized autonomous AGV system for material handling. Int
J Prod Res 40:3995–4006
224 R. K. Jain et al.
3. Ismail IN, Anuar A, Sahari KSM, Baharuddin MZ, Fairuz M, Jalal A, Saad JM (2012) Devel-
opment of in-pipe inspection robot: a review. In: IEEE conference on sustainable utilization
and development in engineering and technology, Malaysia, pp 310–315, 23–26 Sept 2012
4. Nayak A, Pradhan SK (2014) Design of a new in-pipe inspection robot. Procedia Eng 97:2081–
2091
5. Park JJ et al (2013) Development of the untethered in-pipe inspection robot for natural gas
pipelines. In: 10th International conference on ubiquitous robots and ambient intelligence
(URAI), Jeju, South Korea, pp 55–58, 30 Oct–2 Nov 2013
6. Ryew S, Baik SH, Ryu SW, Jung KM, Roh SG, Choi HR (2000) In-pipe inspection robot
system with active steering mechanism. In: IEEE RSJ international conference on intelligent
robots and systems, vol 3, pp 1652–1657
7. Choi HR, Ryew SM (2002) Robotic system with active steering capability for internal
inspection of urban gas pipelines. Mechatronics 12(5):713–736
8. Tatar O, Mandru D, Ardelean I (2007) Development of mobile mini robots for in pipe inspection
tasks. Mechanika, 60–64
9. Nassiraei AA, Kawamura Y, Ahrary A, Mikuriya Y, Ishii K (2007) Concept and design of a
fully autonomous sewer pipe inspection mobile robot. In: IEEE international conference on
robotics and automation, Roma, Italy, pp 136–143,10–14 Apr 2007
10. Zhang Y, Guozheng Y (2007) In-pipe inspection robot with active pipe-diameter adaptability
and automatic tractive force adjusting. Mech Mach Theory 42:1618–1631
11. Lim H, Taku O (2009) Development of the pipe inspection robot. In: IEEE ICROS-SICE
international joint conference, Fukuoka, Japan, pp. 5717–5721,18–21 Aug 2009
12. Kim DW, Park CH, Kim HK, Kim SB (2009) Force adjustment of an active pipe inspection
robot. In: IEEE ICROS-SICE international joint conference, Fukuoka, Japan, pp 3792–3797,
18–21 Aug 2009
13. Roslin NS, Anuar A, Jalal MFA, Sahari KSM (2012) A review: hybrid locomotion of in-pipe
inspection robot. Procedia Eng 41:1456–1462
14. Kwon Y, Byung JY (2012) Design and motion planning of a two-module collaborative indoor
pipeline inspection robot. IEEE Trans Rob 28(3):681–696
15. Harish J, Sivanathan K (2013) Design and fabrication of pipe robot that can inspect in various
pipe types. Int J Eng Res Technol (IJERT) 2(4):224–229
16. Nee LV, Elamvazuthi I, Ganesan T, Khan MA, Parasuraman S (2015) Development of a
laboratory-scale pipeline inspection robot. In: IEEE international symposium on robotics and
intelligent sensors, Procedia Comput Sci 76:9–14
17. Tourajizadeh H, Rezaei M (2016) Design and control of a steerable screw in-pipe inspection
robot. In: IEEE 4th international conference on robotics and mechatronics (ICROM), Tehran,
Iran, pp 98–104, 26–28 Oct 2016
18. Kaushik MB, Karthikeyan P, Jothilingam A (2016) Design of pipe-inspection robot for all
pipeline configurations. Int J Eng Res Technol (IJERT) 4(26):1–3
19. Satish PV, Harsha DS, Hepsiba S, Samanthula N (2017) Design of an in-pipe inspection robot.
Int J Eng Res Technol (IJERT) 6(3):63–68
20. Sonawane A, Shahajahan S, Rehaman A (2017) Design and fabrication of an inline pipe
inspection robot. Int J Eng Adv Technol (IJEAT) 6(4):58–61
21. Diaz JAI, Ligeralde MI, Antonio MAB, Mascardo PAR, Maningo JMZ, Fernando AH, Bandala
AA (2018) Development of an adaptive in-pipe inspection robot with rust detection and local-
ization. In: IEEE region 10 conference TENCON-2018, Jeju, Korea, pp 28–31, 28–31 Oct
2018
22. Abdellatif M, Mohamed H, Hesham M, Abdelmoneim A, Kamal A, Khaled A (2018) Mecha-
tronics design of an autonomous pipe-inspection robot. In: MATEC web of conferences, EDP
sciences, vol 153, p 02002
23. Waleed D, Mustafa SH, Mukhopadhyay S, Abdel-Hafez MF, Jaradat MAK, Dias KR, Ahmed
JI (2019) An in-pipe leak detection robot with a neural-network-based leak verification system.
IEEE Sens J 19:1153–1165
Mechanism of Material Removal
in Magneto Abrasive Flow Machining
Abstract The material removal in finishing of precision parts and hard to reach
regions comprises a standout amongst the most demanding and costly stages in
a manufacturing process. The magnetic abrasive flow machining (MAFM) is one
of the non-traditional machining techniques which can successfully manage these
problems. The MAFM process provides excellent surface finish at comparative low
cost. The objective of the present investigation has been to understand the mechanism
of the material removal (MR) and roughness improvement rate (RIR) in MAFM on a
non-magnetic, soft and ductile workpiece. Finishing fluid in this study contains iron
particles as magnetic bonders and diamond particles as abrasive part in different mesh
size. Experimentation was conducted on aluminum tubes to investigate the effect of
the key parameters on the MR and RIR and was designed in central composite design
(CCD) of response surface methodology (RSM). It has been found that magnetic flux
density (MFD) and number of cycles have impact on material removal and surface
roughness. Based on experimental results, regression analysis of MR and RIR has
been performed. This analysis has been used to predict MR and RIR compared with
the experimentally obtained results. Both were found in close agreement. Finished
surfaces were also examined using the scanning electron microscope (SEM) to get
the knowledge for the fine surface texture produced by MAFM.
P. Singh (B)
IKGPTU, Jalandhar, Punjab, India
P. Singh · L. Singh
BSB Engineering College, Fatehgarh Sahib, Punjab 140407, India
S. Singh
GND Engineering College, Ludhiana, Punjab 141006, India
1 Introduction
the MRR of brass is higher as compared to aluminum in the AFM process. Singh
et al. [13] finished brass workpiece using magnetic abrasive media containing iron
powder and Al2 O3 abrasive particles. They concluded that MFD significantly affects
change in SR value (Ra), thus improving SF and Ra increases up to 0.6 T of
MFD, after that it declines sharply. Sharma et al. [14] finished SiC micro-channels
with the AFM process. It was reported that surface finish of the micro-channels
improved by more than 55%. The machining performance was evaluated in terms of
the material removal and surface finish. Optimum values of the process parameters
were obtained for better surface quality. Singh et al. [15] experimentally examined
the comparison between AFM and MAFM for processing of aluminum tubes. The
authors used diamond based mechanically alloyed MAPs in MAFM. They concluded
that MAFM showed the considerable extent of improvement over AFM.
2 Experimental Work
The surface finish of the Aluminum tubes was recorded prior to the post processing
through MAFM using a surface roughness tester (LC 0.001 µm) and the weight
loss was recorded using a precision electronic weighing machine (LC 0.1 mg). The
average surface roughness (Ra) was recorded to be 1.0–1.4 µm. Such tubes, in
general, have a rough profile depending on the processing conditions. A schematic
of a work specimen aluminum tube is illustrated in Fig. 1.
The developed MAFM setup consists of the electromagnet (maximum flux density of
0.8 Tesla), two hydraulic chambers, two medium chambers, a fan as a cooling arrange-
ment for maintaining the temperature within specified limits and workpiece tooling
(fixture) at the middle as shown in Fig. 2. The setup is integrated with the power
supply, hydraulic unit to perform reciprocating action and DC motor to command
the rate of media flow. In the present work, polymeric medium [8] was prepared
for the MAFM processing by mixing diamond based mechanically alloyed magnetic
abrasive particles (average: 275 mesh) with naphthenic based processing oil which is
basically a hydrocarbon oil boiling between 30 and 200 °C (color: dark gray, manu-
factured by: standard chemicals). In the MAFM setup, aluminum tubes were placed
in such a way that the medium flows axially. A special fixture was fabricated to fix
the workpiece for the purpose which facilitates the medium flow direction from one
chamber to the other chamber by constricting the flow across the aluminum tube. The
mechanism ensures physical contact of the magnetic abrasive laden medium with
the workpiece surface to be finished which consequently improves surface quality.
Thus, in the MAFM process, a fixture plays an important role in enhancing process
productivity [16].
In the present work, response surface methodology was used to explore the responses
over the complete experimental set. Consequently, a set of thirteen experiments were
performed according to central composite design (CCD) of response surface methods
[7]. The process parameters and their levels selected for the experiments are depicted
in Table 1 and the experimental outcomes are illustrated in Table 2. There are also
some constant set of parameters for experimentation. Two parameters MFD and no.
of cycles were varied with five levels for both the parameters according to the setup
limitations and processing quality conditions.
Analysis of variance (ANOVA) was carried out to find the significant effect of
key parameters on material removal (MR) and roughness improvement rate (RIR)
which are response characteristics. MR is the algebraic difference between final
weight after MAFM and initial weight of workpiece, and RIR is the ratio of the
algebraic difference between final roughness (Ra) and initial roughness (Ra) of the
work specimen and the initial Ra and then multiplied by 100. The experimental
outcomes for responses under different conditions of parameters are given in Table
2. The response characteristics (MR and RIR) for the experimental condition were
also projected with Design Expert software. The corresponding predicted values of
the responses and the errors have been illustrated in Table 1. The estimated errors
for MR and RIR were found to be ±0.48 % and ±0.38 %, respectively.
3 Regression Analysis
The ANOVA was performed at 95% confidence level for material removal (MR) and
roughness improvement rate (RIR). The quadratic regression models for MR and
RIR in terms of actual parameters are illustrated in Table 3.
The adjusted R2 and predicted R2 values are within the close agreement. The
complete ANOVA outcomes for MR and RIR are shown in Tables 4 and 5, respec-
tively. It was observed that the results between the predicted and experimental were
in close agreement for both MR and RIR. This indicates that the model is suitable
and can be used directly on the shop floor.
The developed model for MR is significant having F-value of 8931.62. The signif-
icant model terms have P-values lower than 0.05. Here, MFD (A), no. of cycles (B),
interaction between MFD (A) and no. of cycles (B) and quadrics MFD and no. of
cycles are significant terms. The non-significant terms have values more than 0.1.
The LOF is not significant which is good to fit the model. Adequacy precision gives
the signal to noise (S/N) ratio. The S/N ratio of 330.101 shows an adequacy of the
model.
The developed model for RIR is significant having F-value of 319354.99. Here,
MFD (A), no. of cycles (B), interaction between MFD (A) and no. of cycles (B) and
quadrics MFD and no. of cycles are significant terms. The LOF is not significant
Mechanism of Material Removal in Magneto … 231
Table 3 Regression relations for material removed and roughness improvement rate
Responses Predicted R2 (%) Adjusted R2 (%) Regression model
Material removal (MR) 99.96 99.97 MR = −54.65172 +
172.72213 * MFD +
3.89916 * No. of Cycles −
2.48333 * MFD * No. of
Cycles −- 152.20977 *
MFD2 − 0.016139 * No. of
Cycles2 ± Error
Roughness improvement 99.97 99.99 RIR = + 20.87 + 17.23 *
rate (RIR) MFD + 0.31 * No. of
Cycles + 0.057 * MFD *
No. of Cycles − 0.203 *
MFD2 − 0.00068* No. of
Cycles2 ± Error
which is good to fit the model. Adequacy precision gives the signal to noise (S/N)
ratio. The S/N ratio of 1856.44 shows an adequacy of the model.
232 P. Singh et al.
The statistical data was acquired for finishing of micro-holes of aluminum using
MAFM. The CCD of RSM was utilized for designing the experiments. The effect
of individual parameter and interaction effects of parameters on responses (MR
and RIR) are discussed. Further optimization of parameters on MR and RIR is
represented.
50
40
30
Mean
20
10
0
0.15 0.30 0.45 0.60 0.75 5 10 15 20 25
in alteration of the scattering pattern of MAPs at the inward surface of the hollow
workpiece with the outcome that progressively MAPs partake in the abrasion area
and axial MAPs strike at the small angle on the work specimen surface. The increased
MR from the work specimen could be because of the cumulative influence of above
reasons. The interaction between MFD and no. of cycles is also significant and is
also noticeable from Table 4.
The standard response surfaces describing the combined influence of process
parameters on MR are shown in Fig. 4. It can be seen that MR enhances with increased
no. of cycles at all values of MFD. On the other hand, the material removal increases
at all values of no. of cycles up to 0.45T of MFD beyond that MR decreased. The
MR was maximum at 25 no. of cycles with 0.45 Tesla of MFD.
Surface Roughness in MAFM. It is clear from Fig. 5 that when MFD is increased,
the RIR value is consistently improved for aluminum. As more no. of MAPs might
come in contact with the work specimen surface because of magnetic field, the value
of RIR increases. With increasing cycles, the RIR value of aluminum workpiece
increased. It can also be seen that during the initial cycles, surface finish increases
sharply but because the abrasive edges get blunt with increasing cycles that resulted
in small improvements in surface finish. Statistical analysis (Table 5) reveals that
although MFD significantly affects RIR, the percentage contribution of MFD to
the RIR value is less than no. of cycles for aluminum. A possible explanation for
234 P. Singh et al.
Fig. 4 Simultaneous effect of MFD and no. of cycles on material removal (MR)
this result can be as follows. Some of the MAPs flowing with the media, under
the influence of MFD, change their flow path and impinge the hollow surface at a
small angle. As aluminum is more ductile and soft, MAPs that hit the surface at some
incidence angle would plow the material and cause pits resulting in more MR. Due to
the high ductility of the aluminum, these pits would get filled up with the metal in the
vicinity by the subsequently following MAPs in the flowing media. In other words,
it can be hypothesized that on a continuing basis, the metal is being displaced by
abrasive particles, and craters are formed and filled up by metal flow. Consequently,
material removal is more from the surface with the application of the magnetic field
with the overall Ra value not getting much affected.
Figure 6 shows the combined effect of the parameters on RIR. The 3D graph
shows that the roughness improvement rate increased with increasing MFD at all
values of no. of cycles. Similarly, RIR increased at all values of no. of cycles with
increased MFD. The roughness improvement rate was maximum at higher values of
MFD and no. of cycles.
ANOVA results show that MFD and no. of cycles significantly affect MR and
RIR. The results also depict that there is a substantial increase in MR and RIR in
aluminum tubes. Statistical analysis also reveals that the interaction effect of MFD
and no. of cycles on both MR and RIR is significant.
Mechanism of Material Removal in Magneto … 235
38
36
Mean
34
32
30
28
4.2 Optimization
The objective of optimization was to maximize surface finish and to maximize mate-
rial removal from the machined surface subject to the constraints of the working
limits of two variables (Table 2). The response optimization was performed for opti-
mizing process parameters. No. of cycles has the highest percentage contribution on
MR and RIR. However, in both the response factors, the considerable contribution
was observed by MFD. The optimum conditions/process variables and model vali-
dation of MR and RIR obtained from numerical optimization are depicted in Table
6. It was noticed that the relative error %age of material removal and roughness
improvement rate are 0.03 and −0.36%. The acquired outcomes appear well within
the practically acceptable range. So, the acquired result can be used for predicting
the optimal values of the appropriate outputs in the same experimental state.
the topography becomes significantly finer. Thus, such aluminum tubes should be
more effective in applications like micro-fluidics and bioapplications. The SEM
observations further reveal that the deep tool marks produced during fabricating the
cavity to be finished are reduced resulting in a fine surface texture after MAFM. So
the irregularities have been reduced to a large extent, and only few pits with some
abrasive grain marks and porous zones are visible as depicted in Fig. 8.
5 Conclusions
The present investigation was carried out to assess the performance of the MAFM
process in terms of the material removal and roughness improvement rate in surface
quality of aluminum tubes. The experimentation was performed by employing RSM
for finishing the aluminum tubes having MFD and no. of cycles at different levels
of parameters. The results have been acquired using commercial analysis software.
The following are major conclusions obtained from the present work:
1. A significant improvement in material removal (up to 51.75 mg) and in surface
finish could be achieved utilizing the MAFM process.
2. The MFD and no. of cycles have significant influence on the process.
3. MFD has the highest contribution in finishing and material removal of aluminum
tubes.
4. The process is capable of removing deep boring tool marks and other surface
irregularities with the increasing MFD and no. of cycles as revealed by SEM
observations.
Acknowledgements The authors are thankful to the Department of RIC, IKGPTU Jalandhar and
Department of Mechanical Engineering, Baba Banda Singh Bahadur Engineering College, for
providing experimental setup and other measuring instruments used in the present study.
238 P. Singh et al.
References
1. Butola R, Jain R, Bhangadia P, Bandhu A, Walia RS, Murtaza Q (2018) 2018, Optimization to
the parameters of abrasive flow machining by Taguchi method. Mater Today Proc 5:4720–4729
2. Das M, Jain VK, Ghoshdastidar PS (2008) Fluid flow analysis of magnetorheological abrasive
flow finishing (MRAFF) process. Int J Mach Tools Manuf 48:415–426
3. Das M, Jain VK, Ghoshdastidar PS (2011) The out-of-roundness of the internal surfaces of
stainless steel tubes finished by the Rotational-Magnetorheological abrasive flow finishing
process. J Mater Manuf Process 26(8):1073–1084
4. Gorana VK, Jain VK, Lal GK (2004) Experimental investigation into cutting forces and active
grain density during abrasive flow machining. Int J Mach Tools Manuf 44:201–211
5. Kathiresan S, Mohan B (2017) Experimental analysis of magneto rheological abrasive flow
finishing process on AISI stainless steel 316L. J Mater Manuf Process 33(4):422–432
6. Kamble PD, Untawale SP, Sahare SB (2012) Use of magneto abrasive flow machining to
increase material removal rate and surface finish. VSRD Int J Mech Automob Prod Eng 2:249–
262
7. Montgomery DC (2004) Design and analysis of experiment. Wiley, New York
8. Rajesha S, Venkatesh G, Sharma AK, Kumar P (2010) Performance study of a natural polymer
based media for abrasive flow machining. Indian J Eng Mater Sci 17:407–413
9. Rajesha S (2011) Some studies to enhance the capabilities of abrasive flow machining process.
Ph. D. Thesis, Indian Institute of Technology, Roorkee
10. Singh S, Shan HS (2002) Development of magneto-abrasive flow machining process. Int J
Mach Tools Manuf 42:953–959
11. Singh S, Shan HS, Kumar P (2002) Wear behavior of materials in magnetically assisted abrasive
flow machining. J Mater Process Technol 128:55–161
12. Singh S, Shan HS, Kumar P (2008) Experimental studies on mechanism of material removal
in abrasive flow machining process. J Mater Manuf Process 23(7):714–718
13. Singh R, Walia RS, Suri NM (2013) Study of parameters for magnetic force assisted abrasive
flow machining process. J Eng Technol Educ 7(2):44–48
14. Sharma AK, Venkatesh G, Kumar P (2015) Fine finishing of SiC microchannels using abrasive
flow machining. Indian J Eng Mater Sci 22:297–306
15. Singh P, Singh L, Singh S (2018) Experimental comparison of abrasive flow machining and
magnetic abrasive flow machining for aluminium tubes. Int J Res Eng Appl Manage 4(3):714–
720
16. Walia RS, Shan HS, Kumar P (2009) Enhancing AFM process productivity through improved
fixturing. Int J Adv Manuf Technol 44:700–709
Optimization of Cutting Parameters
of EN9 Steel with Plain Carbide Tool
Using Response Surface Methodology
Abstract The present work deals with the optimization study on cutting perfor-
mance of plain carbide inserts in turning of EN9 steel. The design of experiments
(DOE) was done using response surface methodology (RSM) technique. The cutting
parameters were spindle speed, feed, and depth of cut (DOC). The combined inter-
actions of cutting parameters on crater wear were analyzed. The crater wear was
observed using machine vision system. The work concentrated on the building up
a mathematical relation between cutting parameters and output responses. Analysis
of variance (ANOVA) was used for checking the adequacy of the model. The rela-
tion between input factors and the response parameters is determined using a cubic
regression model. The optimization was done using desirability ramp plots.
1 Introduction
EN9 is an exceptionally famous grade of medium carbon steel. EN9 is reasonable for
the production of parts like shafts, gears, bolts, sprockets, axles, cams, and studs. In
current scenario of industries, the main objective is to make low-cost superb products
with most extreme efficiency in a brief time. The goal of every metal cutting industry
is to achieve better surface roughness at low tooling cost in computer numeric control
(CNC) machines. Turning operation is most widely used technique to transform raw
material in useful products like shaft, wheel rod, and screw jack rod. The single-
point cutting tool is used in turning to remove undesirable material from the outer
diameter of a rotating cylindrical shape work piece [1]. Better turning surfaces can
prompt change in quality properties like thermal resistance, fatigue strength, and
corrosion resistance [2]. In order to generate better turning surface, there is consider-
able role of input parameters. Most important quality characteristic of any machined
part is surface roughness, which is characterized as a no. of peaks and valleys on
the surface [3]. Better surfaces is not only important for better properties; apart of
it, better surfaces provides esthetic look to product, which is expectation of every
customer [4]. The selection of tool is also necessary for better surfaces. Tool wear is
critical problem, which leads to bring down the machining productivity and decrease
tool life. To maintain the economical production and high quality of materials during
machining, flank and crater wear are need to optimize [5]. Numerous examinations
have been led to research the performance of various cutting tool in the turning of
different materials [6]. Researchers researched the machinability of AISI4340 steel
with plain and multilayer coated carbide inserts. It was concluded that the perfor-
mance of coated carbide inserts found better than plain carbide inserts. The tool life
for coated carbide inserts was found greater than plain carbide inserts [7]. Researchers
deal with the hard turning of AISI52100 steel with multilayered coated carbide inserts
to examine the influence of cutting temperature. Central composite design (CCD)
model of RSM technique was used to perform machining experiments. The ANOVA
results revealed that feed has most dominating factor for cutting temperature [8].
Researchers conducted the experiments on the samples of EN24 alloy steel by using
physical vapor deposition (PVD)-coated TiAlN insert and plain carbide insert using
Taguchi method. It was concluded that spindle speed and DOC were influenced
mostly for turning process. For coated insert, the DOC is effective variable for surface
roughness and MRR followed by spindle speed and feed. For plain carbide insert, the
DOC is the most influencing factor in determining the multi-objective characteristics
[9]. Some researchers’ work deals with machinability studies on tool wear and surface
roughness, in hard turning of AISI D2 steel using polycrystalline cubic boron nitride
(PCBN), mixed ceramic, and coated carbide inserts by using RSM. Three cutting
parameters were cutting speed, cutting time, and tool hardness, and two performance
outputs were surface roughness and tool wear. From the results, it was revealed that
the cutting time altered the tool wear. On the other hand, cutting time was the prin-
cipal factor influencing the surface roughness of the work piece succeeded by cutting
speed [10]. Researchers conducted the experiments on EN8 alloy steel material with
the PVD coated TiAlN insert and plain carbide insert by Taguchi method. From
outcomes, it is concluded that spindle speed and DOC are prominent factors, which
influence the turning of EN8 alloy steel. For coated insert, the spindle speed has
the highest influence in determining the multi-objective characteristics succeeded by
feed and DOC. For plain carbide insert, the DOC is the highest influencing parameter
succeeded by spindle speed and feed rate [11]. Researchers examined the orthog-
onal cutting of AISI1030 steel with plain, coated, and cemented carbide inserts at
different feeds and different cutting speeds by keeping DOC constant. The exper-
iments were analyzed by artificial neural network (ANN). The coated carbide tool
was found better than plain and cemented carbide [12]. Researchers optimized the
cutting parameters using Taguchi method and regression analysis during turning
of EN45 spring steel by plain carbide cutting tool. Experiments were designed by
L16 orthogonal array. The surface roughness, MRR, and tool wear were optimized
by considering cutting speed, feed rate, and DOC as machining parameters. From
results, it is analyzed that the feed rate was the most dominating factor for surface
roughness and the cutting speed is the most dominating factor for MRR and tool
Optimization of Cutting Parameters of EN9 Steel with Plain … 241
wear [13]. Researchers concluded that feed rate has dominating effect on surface
roughness, while the interaction among side cutting edge angle (SCEA) and the feed
factors gives the secondary contribution toward output parameters [14]. Researchers
concluded that feed rate has dominating effect for surface roughness and cutting
forces [15]. Researchers concluded that cutting speed is the influencing parameter
which has highest effect on flank wear [16, 17]. Researchers studied tool wear by
using Taguchi method and concluded that DOC and cutting speed are influencing
parameters that affect the tool wear [18]. Researchers analyzed and revealed that
the combined interaction of low DOC and feed with high cutting speed is better for
gaining low machining force [19]. Researchers concluded that spindle speed is most
influencing parameter for surface roughness during machining [20]. Researchers
examined the performance of plain with coated carbide tool. It was found that the
performance of coated inserts was better than plain carbide inserts. The scrutiny of
literature founds that very less work has been done on EN9 steel, as this grade of
steel has huge number of applications in automobile industry. The aim of this study
is to recommend better tool and optimized parameters for turning operation of this
steel.
2 Experimental Methods
EN9 steel was chosen as workpiece material. The energy dispersive spectroscopy
(EDS) test was performed to confirm the composition of material given by supplier.
The EDS image is shown in Fig. 1. The EDS test confirms the alloying elements Fe,
C, Mn, Si, S, and P. It proves that the material is EN9 steel. The hardness of workpiece
and tool was checked on Mitutoyo HM-100 microhardness tester. It confirms that
the hardness of tooling insert is approx. three times than workpiece hardness, which
is recommended for better machining. The results of hardness along with shape/size
of inserts are presented in Table 1. The machining was carried out in dry cutting
conditions using CNC turning lathe. The cutting parameters range is given in Table 2.
The objective function of the optimization is to minimize the crater wear. As in
order to increase the tool life, the crater wear should be minimized as it is an important
parameter. The constraints for the optimization are provided in Table 2 in the form
of goal and range.
Figure 1 represents the EDS image of EN9 steel which describes the composition
of the various elements present in it. The hardness of workpiece and cutting tools
hardness was tested by the Mitutoyo microhardness tester. The crater wear was
evaluated by machine vision system model “Sipcon SVI-IMG 3D.”
3 Experimental Design
RSM is a gathering of factual and scientific methods that are helpful for modeling,
scrutiny of difficulties in which a response of interest is affected by a few factors
and goal is to streamline this variable. In the present examination, the correlation
Optimization of Cutting Parameters of EN9 Steel with Plain … 243
among cutting conditions, i.e., spindle speed, feed rate, and DOC called inputs and
the output Y, defined as the desired machinability (crater wear) is given by:
Y = φ(s, f, d) (1)
Here, Y is the output response factor and s, f , d are the input parameters. If the
response function is unknown, the second-order model is used, as shown in Eq. (2):
k
k
k
Y = bo + bi X i + bi j X i X j + bii X i2 (2)
i=1 ij i=1
where b0 is the free term, the coefficients b1 , b2 , b3 , …, bk are the linear terms,
and b11 , b22 , bkk are quadratic terms, while b12 , b13 , and bk −1 are interacting terms.
X i represents input parameters, i.e., spindle speed, feed rate, and DOC. The output
factor crater wear is also called response factor.
Box–Behnken design (BBD) test matrix was utilized for planning of crater wear
experiments. In present work, 15 experiments with plain carbide tool were performed
on the work material to develop output response models. The experimental design is
developed to evaluate the impact of spindle speed, feed rate, and DOC on the output
parameters.
There are two techniques in RSM for optimization BBD and CCD. As BBD was
selected, it requires less experiments (15 experiments for three factors). BBD have
fewer design points; thus, they are less expensive to run with the factors as it has
three levels per factors.
Tool wear was checked by machine vision system model SIPCON SVI-IMG 3D.
Figure 2a shows the setup of the machine vision system, Fig. 2b shows image of the
plain carbide tool before wear, and Fig. 2c shows image of tool after wear.
A statistical analysis of output responses was done with the purpose of examining
the influence of feed rate, spindle speed, and DOC on the results. ANOVA was
performed by using Design Expert V.10 software. The motivation behind ANOVA is
to consider which outline parameters altogether influence the quality characteristics.
This is gained by isolating the total uncertainty of the S/N proportions and is measured
by the sum of the squared deviations from the total mean S/N proportions, into
244 S. Chauhan and R. Kumar
Fig. 2 Machine vision system images a setup of machine vision system, b image of insert before
wear, and c image of crater wear on plain carbide insert
improvement by each of the design parameters and the error. The model with a p-
value < 0.1 are treated to have a better improvement to the performance part. Values
of “Prob > F” less than 0.0500 show the significance of model terms. The analysis
was out for 5% acceptance level, i.e., for 95% assurance level. If value of F > 4,
then it implies that the difference in the design parameter has expressive impact on
the quality characteristics of component. Adeq precision measures the S/N ratio, and
desirability of this is greater than 4.
Power transform with cubic model is used for the analysis. Table 3 presents the
ANOVA results for crater wear.
The model F-value of 21.35 suggests that the model is significant. There is just
a 4.56% possibility that a F-value this large could happen because of noise. In this
case, Speed × feed, Speed2 , Speed2 × DOC are significant model terms. The ratio of
18.972 indicates an adequate signal. The regression equation for crater wear model
is given in Eq. (3).
The predicted values of output factor are obtained from regression Eq. 3, and the
experimental values with respect to different combinations of machining parameters
are given in Table 3. The comparison between predicted values to the corresponding
experimental values of plain carbide tools is shown in Fig. 3.
The predicted versus actual values for response variables, i.e., crater wear, with
plain carbide tool are shown in Fig. 3. The residuals plot for all output responses
with plain carbide tool defines that all the values lie on the same line which shows
that the model is appropriate and defines the significance of model. The plot shows
that there is a reasonable consistency between predicted and actual values.
246 S. Chauhan and R. Kumar
Figure 4a shows the effect of spindle speed and feed on the crater war. It was observed
that at low feed and low spindle speed, minimal value of crater wear is obtained. As
the spindle speed increases, the value of crater wear increases. In case of higher
feed, the maximum wear occurs, but as the spindle speed increases, the crater wear
value decreases. The minimal crater wear is obtained at combined interaction of low
speed and low feed. Figure 4b represents the effect of spindle speed and DOC on
the crater wear. It was observed that at low DOC and low spindle speed, maximum
value of crater wear is obtained. As the spindle speed increases, the value of crater
wear decreases up to 792 rpm and then starts increasing. In case of higher DOC, as
the speed increases, the value of crater wear decreases. To get the minimum wear
value, the combined interaction of high speed and high DOC is suitable. Figure 4c
represents the effect of feed and DOC on the crater wear. It was observed that at low
DOC and low feed, minimum value of crater wear is obtained. As the feed increases,
the value of crater wear increases. In case of higher DOC, as the feed increases,
276 1000
X1 = A: Spindle speed X1 = B: Feed
X2 = C: Depth of cut X2 = C: Depth of cut 800
X1 = A: Spindle speed 800
X2 = B: Feed 800 Actual Factor
Actual Factor
Crater wear (um)
600
C: Depth of cut = 0.75
600 400
400
400 200
200 0
200
1 970 1 0.33
0.33 970
0.9 881 0.9 0.275
0.275 881
792 0.8 792 0.8
0.22 0.22
703 0.7 703 0.7
0.165
B: Feed (mm/rev) 0.165 614 A: Spindle speed (rpm) C: Depth of cut (mm) 0.6 614 A: Spindle speed (rpm) C: Depth of cut (mm) 0.6 B: Feed (mm/rev)
0.11 525 0.5 525 0.5 0.11
Fig. 4 Effect of input factors on crater wear with plain carbide tool a Crater wear versus feed and
spindle speed, b crater wear versus DOC and spindle speed, and c crater wear versus DOC and feed
Optimization of Cutting Parameters of EN9 Steel with Plain … 247
the value of crater wear decreases up to 0.22 rev/mm and then increases. To get
the minimal results, combined interaction of low feed and low DOC is suitable. This
could be due to the reason that the contact time and cutting length increase which arise
high stresses and temperature that cause the tool failure. The measured responses are
similar with other documented studies [15] in machining of AISI D2 steel.
Table 4 shows the goals and parameter ranges for optimization of cutting conditions
with plain tool. The goal for input parameters is in range and for output responses is
minimum. The importance for input parameters is given a value of 3 and for output
responses a value of 5 is given.
5 Conclusions
References
1. Abhang LB, Hameedullah M (2010) Power prediction model for turning EN-31 steel using
response surface methodology. Eng Sci Technol Rev 3:116–22
2. Nesßeli S, Yaldız S, Türkeş E (2011) Optimization of tool geometry parameters for turning
operations based on the response surface methodology. Measurement 44:580–7
3. Hamdan A, Sarhan AAD, Hamdi M (2012) An optimization method of the machining parame-
ters in high-speed machining of stainless steel using coated carbide tool for best surface finish.
Int J Adv Manuf Technol 81–91
4. Abhang LB, Hameedullah M (2012) Optimization of machining parameters in steel turning
operation by Taguchi method. Procedia Eng 38:40–8. International conference on modeling,
optimization and computing
5. Verma N, Zafar S, Pathak H (2020) Investigations on thermal damage and surface roughness
of laser beam machined nano-hydroxyapatite UHMWPE composites. Manuf Lett 25:81–87
6. Kumar A, Sahoo B (2012) Experimental investigations on machinability aspects in finish hard
turning of AISI 4340 steel using uncoated and multilayer coated carbide inserts. Measurement
45:2153–65
7. Shihab SK, Khan ZA, Mohammad A, Siddiqueed AN (2014) RSM based study of cutting
temperature during hard turning with multilayer coated carbide insert. Procedia Mater Sci
6:1233–1242
8. Jadhav SR, Shaikh AM (2016) Optimization of process parameters for CNC turning using
Taguchi methods for EN24 alloy steel with coated/uncoated tool inserts. Int J Adv Eng Manag
Sci 2:1923–1931
9. Mir MJ, Wani MF (2018) Modelling and analysis of tool wear and surface roughness in hard
turning of AISI D2 steel using response surface methodology
10. Deshpande RR, Pant R (2017) Optimization of process parameters for CNC turning using
Taguchi methods for EN8 alloy steel with coated/uncoated tool inserts. Int Res J Eng Technol
180–8
11. Nalbant M, Gokkaya H, Toktas I, Sur G (2009) The experimental investigation of the effects of
uncoated, PVD and CVD-coated cemented carbide inserts and cutting parameters on surface
roughness in CNC turning and its prediction using artificial neural networks. Rob Comput
Integr Manuf 25:211–23
12. Kumar S, Pabla BS, Madan J (2017) Optimization of machining parameters in turning EN-45
steel using plain carbide tools. Int J Sci Res Sci Eng Technol 3:37–43
13. Noordin MY, Venkatesh VC, Sharif S, Elting S, Abdullah A (2004) Application of response
surface methodology in describing the performance of coated carbide tools when turning AISI
1045 steel. J Mater Process Technol 145:46–58
14. Rao CJ, Rao DN, Srihari P (2013) Influence of cutting parameters on cutting force and surface
finish in turning operation. Procedia Engineering 64:1405–1415 International conference on
design and manufacturing
15. Kaladhar M, Subbaiah KV, Rao CHS (2013) Optimization of surface roughness and tool flank
wear in turning of AISI 304 austenitic stainless steel with CVD coated tool. J Eng Sci Technol
8:165–76
16. Sahu M, Sahu K (2014) Optimization of cutting parameters on tool wear, workpiece surface
temperature and material removal rate in turning of AISI D2 steel. Int J Adv Mech Eng 4:291–8
17. Verma N, Vettivel SC, Rao PS, Zafar S (2019) Processing, tool wear measurement using
machine vision system and optimization of machining parameters of boron carbide and rice
husk ash reinforced AA 7075 hybrid composite. Mater Res Exp 6:0865f3
18. Suresh R, Basavarajappa S, Samuel GL (2012) Some studies on hard turning of AISI 4340
steel using multilayer coated carbide tool. Measurement 45:1872–84
19. Chauhan S, Verma N, Kumar R (2018) Study on optimization of turning parameters on various
steel grades: a review. Int J Emerg Technol Eng Res 6(5):111–116
20. Chaudhari M, Karunamurthy K (2017) Determining the effect of coated and uncoated tool
insert during turning of EN31 steel. Int J Mech Eng Technol 8:1009–19
Design and Development
of a Sit-to-Stand Assistive Device
Shoudho Das, Satyajit Halder, Sourabh Kumar Sahu, Sujatha Srinivasan,
and Sourav Rakshit
Abstract The aim of this work is to conceptualize, design and develop a device to
assist the sit-to-stand (STS) motion for disabled and elderly. The device assists the
human to stand up from sitting position along the natural trajectory of STS motion.
Both human and device are assumed to be functioning fully in the sagittal plane.
The natural trajectory of armpit is found for the STS motion using a motion capture
system. The coordinates are projected onto a sagittal plane passing through the ankle.
A four-bar mechanism is synthesized with four-position motion generation using
Burmester curve theory. Assuming quasi-static motion, forces on joints at different
time steps are found out. Finite element analysis (FEA) is carried out to determine the
required cross section of the links of the structural components of device. Material
and dimensions of STS device components are selected, and the device is fabricated.
Final experiments of STS device with human show that while using the device,
the ground reaction force (GRF) of the human during STS motion decreases by
approximately 25%, thereby indicating a reduction in effort of the human for STS
by the same order.
1 Introduction
In daily lives, sit-to-stand (STS) is a very common activity. It involves large muscle
forces in order to coordinate most of the limbs together to carry out STS motion [1].
With age people lose muscle mass, thereby reducing their capability to perform STS
motion on their own. Inability to generate enough muscle forces and coordinate all the
limbs properly leads to fatal accidents. It also affects older people psychologically as
they become dependent on others for aid in STS motion. Injured and disabled people
who find difficulty in performing STS motion also become dependent on others.
Several devices currently exist for aiding STS. STS assistive devices can be
broadly categorized into active and passive devices. Passive devices provide static
structural support to the user while she/he performs STS, while active devices follow
the motion of the user during STS and actively push/lift them. Passive STS devices
have been designed in different forms, for example, walkers [2, 3], sticks [4] and
attachments to beds/chairs/sofa [5–7] are some of these. Active devices are actuated
by electric or hydraulic actuators. Here again, two features are common, either the
user is strapped to the device and pulled up [8 9] or the user sits on a seat built into the
device which is pushed up [10]. Some devices additionally use knee supports [11].
None of these devices follow the natural trajectory of STS and impose their own path
of motion on user. A few devices exist that attempt to follow natural trajectory of
STS, but they are bulky and costly [12, 13]. Thus, there is need for a cost effective,
compact STS assistive device.
The main motivation of our work is to design and manufacture a low-cost and
compact STS assistive device following natural STS trajectory. The work is carried
out in the following steps:
1. Experimental determination of natural STS trajectory.
2. Synthesis of a four-bar mechanism to follow the natural STS trajectory.
3. Determination of loads on STS device based on quasi-static analysis; selection
of material and dimensions of device based on finite element analysis (FEA).
4. Fabrication of the STS assistive device.
Each of these steps is described as follows.
STS motion involves motion mostly in the sagittal plane as shown in Fig. 1a. The
human is modelled as a three-link planar mechanism as shown in Fig. 1b. It is assumed
that the arms are placed on the chest in crossed form. Thus, the three links represent
the two legs combined, two thighs combined and trunk with arms combined. The
three links are assumed to be connected by ideal pin joints at the ankle, knee and
hip. Subject used for experiment is a 27-year-old healthy male with weight of 90 kg
and 178 cm height.
The device supports the user at the armpit and pushes him/her up. Thus, armpit
trajectory is required to synthesize the required device. To get armpit trajectory, three
LEDs are attached to the arm at the armpit, elbow and wrist. The captured armpit
trajectory is shown in Fig. 1c.
Design and Development of a Sit-to-Stand Assistive Device 251
Fig. 1 a Sagittal plane view of human doing STS. The three links are identified by attaching two
LED markers (yellow circles) on each body part. b Three link model. c Armpit trajectory of human
Before describing our method of mechanism synthesis, we give a brief review of four-
position motion generation for a four-bar linkage. Details can be found in classic texts
[14].
Refer Fig. 2a ABCDE is the four-bar mechanism (AE = link 1, AB = link 2,
BD = link 3, DE = link 4, and C is the coupler point) in the complex plane. C1 ,
C2 , C3 and C4 are the four specified positions of the coupler point with C1 as the
reference position. ABCDE is split up into two dyads WZ formed by AB and BC,
and US formed by ED and DC. Let us consider the dyad WZ. Positions C2 , C3 and
C4 are positioned with respect to reference position C1 by vectors δ2 , δ3 and δ4 ,
respectively. W and Z make angles of β2 , β3 , β4 , γ2 , γ3 and γ4 , respectively, with
reference configuration at the three subsequent positions. The loop closure equation
for the jth ( j = 1, 2, 3) configuration and reference configuration is:
Fig. 2 a Four-position motion generation problem in complex plane. K-points and M-point are
shown. b “Compatibility” linkage corresponding to Eq. (3)
252 S. Das et al.
W eiβ j − 1 + Z eiγ j − 1 = δ j (1)
For motion generation problem, the four positions of coupler BC are known
(vectors δ2 , δ3 and δ4 and corresponding angles γ2 , γ3 and γ4 ). There are three
vector equations implying six scalar equations. The unknowns are two vectors W,
Z and three scalar angles β2 , β3 and β4 . Hence, there are seven unknowns and six
equations. Any of the three angles β2 , β3 and β4 can be taken as a free variable. The
above equations can be written in matrix form as
⎡ ⎤ ⎡ ⎤
eiβ2 − 1 eiγ2 − 1 δ1
⎣ eiβ3 − 1 eiγ3 − 1 ⎦ W = ⎣ δ2 ⎦ (2)
Z
eiβ4 − 1 eiγ4 − 1 δ3
The compatibility condition requires that the determinant of the augmented matrix
must be zero. This leads to the compatibility equation [14]:
Equation (3) looks like loop closure equation of a four-bar “compatibility” linkage
with “fixed” link 1 and “moving” links j , j = 2,3,4, and “link rotations” β j
measured from reference configuration of the compatibility linkage as shown in
Fig. 2b. Suppose β2 is the free variable. Using these two equations, vector loops are
drawn to develop relations to compute β3 and β4 . For each free variable β2 , two pairs
of solutions for β3 and β4 are obtained as (β3 , β4 ) and (β̃3 , β̃34 ) (see Fig. 2b), known as
Burmester point pairs (BPPs). These two sets of BPPs are obtained from vector loop
Eq. (3): β3 = arg +θ3 −arg 3 , β̃3 = arg + θ̃3 −arg 3 , β4 = arg −θ4 −arg 4
and β̃4 = arg + θ̃4 − arg 4 + π . In the above set of four equations, except for
θ3 , θ̃3 , θ4 and θ̃4 all are known because , 3 , 4 and 2 are functions of known
parameters γ2 , γ3 , γ4 , β3 , β4 , δ2 , δ3 and δ4 , θ3 , θ4 , θ̃3 , θ̃4 are found using equations:
24 − 23 − 2
cos θ3 = , θ̃3 = 2π − θ3 ,
23
2 − 24 − 2
cos θ4 = 3 , θ̃4 = 2π − θ4 .
24
Thus, (β3 , β4 ) and (β̃3 , β̃4 ) can be calculated. Finally, by plugging in all the angles
into Eq. (2), the unknown vectors W and Z are found. The ground pivot points are
denoted by M-points, and the corresponding link end is denoted by K-points (Fig. 2a).
The sets of M-points and K-points can be found using equations
K 1 = R1 − Z (4)
M = K1 − W (5)
Design and Development of a Sit-to-Stand Assistive Device 253
Fig. 3 a Chosen positions T1 , T2 , T3 and T4 of the four-bar coupler along the natural trajectory of
armpit ABC. b Synthesized mechanism ABCDE. Coupler BCE is shown yellow hatched. E is the
coupler point tracing the four armpit points. c Schematic diagram of STS assistive device. ABCD
is the four-bar mechanism, and E is the coupler point passing through points T1 , T2 , T3 and T4 in
(a). The linear actuator (shown by red and green) actuating the device is between hinges G and J
where R1 is the known vector from origin to the reference position of coupler point
C1 as shown in Fig. 2a. Vectors W and Z can be found by choosing one convenient
ground pivot M-point from the solution set of M-points and corresponding K-points.
Second dyad can be found using the same set of BPPs. Since the user is supported and
pushed up by the device at the armpit, the device must follow the natural trajectory of
the armpit during STS (Fig. 3a). From the trajectory, we chose four key positions—
T1 , T2 , T3 and T4 as shown in (Fig. 3a) for the four-bar motion generation problem.
The coupler angle γ j (j = 2, 3, 4) is iteratively varied, and K-points (connection point
between coupler and the other links) and M-points (potential ground pivot points)
are found out. Those K and M-points are chosen that are closest to the seat and
the mechanism link lengths AB, BC and CD (Fig. 3b) are obtained along with the
coupler dimensions BE and CE. Point E on coupler curve traces the natural trajectory
of armpit.
The mechanism link lengths and ground pivot coordinates are already determined as
discussed in last section. The device is actuated by a linear actuator of the following
specifications: Stroke length: 450 mm, Load capacity: 1500 N, maximum speed:
16 mm/s, installation length: 650 mm, full extension length: 1100 mm and weight:
3.5 kg. Older people take around 3 s to complete STS motion. This STS device takes
approximately 15 s to complete similar motion.
254 S. Das et al.
Figure 3c shows the schematic diagram of our designed STS assistive device.
Links AB (crank), BC (coupler), CD (follower) and AD (fixed link) form the four-
bar linkage with CED as the coupler and E the coupler point. Crank AB is actuated
by the linear actuator.
From a cost and fabrication point of view, material for all the links and frames is
selected as mild steel (MS) with yield strength of 250 MPa. The links, coupler and
structural frame of the STS device are made from mild steel tubes of 25.4 mm outer
diameter and 2 mm thickness. Bending radius of tubes is set to at least 52.7 mm
to avoid buckling of tubes. Diameter of MS tubes and bending radius is guided by
fabrication aspects. Revolute joints are composed of 10 mm diameter pins of high-
strength steel with yield strength of 790 MPa. Stress analysis of the device is done
at four positions corresponding to the STS motion generation problem. FE analysis
is done in ABAQUS software. Figure 4 shows the fabricated STS assistive device.
Since the STS device is made by joining (welding) tubes, mild steel was chosen as
the material compared to lighter materials like aluminium to avoid problems with
welding. First, the tubes were cut into required lengths using a band saw machine.
Tubes were bent to required angles using a bending machine. Tubes were welded
using arc welding. Then, holes were drilled in tubes, and bushes were placed inside
holes using tungsten inert gas welding (TIG). Hand grinding was done to deburr the
saw cut tubes and finish weld pool undulations. Four 4-inch diameter castor wheels
were fixed to the base to facilitate easy movement of the device. The clearance
between frame and ground is 0.5 mm.
Design and Development of a Sit-to-Stand Assistive Device 255
Fig. 5 Ground reaction forces from STS experiments with a healthy subject. a Without STS assistive
device. b With STS assistive device
We conducted STS experiments on a force plate with a healthy subject (height 180 cm,
weight 70 kg) with our STS assistive device. The results of the STS experiment are
shown in Fig. 5. When the person stands up without the assistive device as shown in
Fig. 5a, the overall STS motion takes place in approximately 3 s. The initial part (2–
3 s) is the reaction force of feet. At approximately 3.5 s at lift-off from seat, there is a
dip, and then, a sudden rise in force as the human rapidly stands up, which ultimately
stabilizes to the weight of the human (700 N). The total STS motion with the assistive
device time is much longer (approximately 27 s) compared to STS motion without
device as shown in Fig. 5b. Here again, initially, we get the reaction force of feet.
Then, approximately at 6 s, the STS device is actuated, and the human is slowly
moved from sit to stand position. The lift-off from seat occurs at approximately
17 s, so the load on the feet increases. This load slowly increases and stabilizes to
approximately 500 N. Hence, approximately, 200 N load is taken by the assistive
device during STS motion. Finally, at around 30 s, the person stands by himself, and
hence, the load coming on the feet increases to full body weight. Hence, the load
borne by the assistive device during STS motion is approximately 28%.
In this paper, we presented the design and development of a prototype STS assistive
device. Our STS device follows the natural STS trajectory to simultaneously reduce
the effort and increase comfortability of the user. It pushes the user at the armpit
256 S. Das et al.
instead of pushing at elbow or pulling up using a belt tied to waist which the current
STS devices employ. The current weight of the assistive device is approximately
10 kg, which can be much reduced by using lighter materials. Also, due to permanent
joining method like welding, the geometry of the device is fixed and cannot be altered
according to height of the user. Future work will also focus on manufacture of a STS
device with user-compliant geometry.
7 Declaration
The ethical approval for this study was granted by Institutional Ethical Committee,
IIT Madras. The participant in the study was provided with prior informed consent.
The authors certify that there is no conflict of interest in the publication of this work.
References
1. Riley OP, Schenkman ML, Mann RW, Hodge WA (1991) Mechanics of a constrained chair-rise.
J Biomech 24(1):77–85
2. T.W. Cunningham, Rising brace for invalid walker. US Patent 5,005,599 A (1991)
3. M.A. Wolferts, Holder for invalid walker. US Patent 3,085,258 A (1963)
4. D. Randall, R.D. Gill, Dual handed walking and uprisal assist device. US Patent 5,785,070 A
(1998)
5. M.E. Boyce, H.E. Boyce, Convalescent aid. US Patent 4,941,495 A (1990)
6. F.H. Bratton, Patient riser. US Patent 3,668,723 A (1972)
7. A.B. Wilson, Aid for use in sitting down or standing up. US Patent 3,739,793 A (1973)
8. M.D. Costello, Assisted lifting, stand and walking device. US Patent 5,502,851 A (1996)
9. R. Hakamiun, D.P. Genereux, M.D. Falin, M.J. Mutka, J.C. Brooke, Stand assist lift. US Patent
6,175,973 B1 (2001)
10. E. Razon, Sit down and stand up walker with seat assembly. US Patent 8,151,812 B2 (2012)
11. Y. Marcoux, A. Veilleux, Lift chair. US Patent 6,213,554 B1 (2001)
12. H.-G. Jun, Y.-Y. Chang, B.-J. Dan, H. Yang, W.-K. Song, J. Kim, Walking and sit-to stand
support system for elderly and, in IEEE International Conference on Robotics and Automation
(2011)
13. A. Purwar, G. Thomas, J. Miles, J. Renert, Portable six-bar apparatus for lifting, lowering and
self-propelled transit. US Patent 8,468,622 B2 (2013)
14. G.N. Sandor, A.G. Erdman, Advanced Mechanism Design Analysis and Synthesis, vol. 2.
(Prentice Hall Inc, 1984)
Effect of Structural Characteristics
on Kinematics of Planar Kinematic
Chains
Abstract Among the large family of planar kinematic chains, all possible distinct
linkages/mechanisms are required for synthesis of kinematic mechanisms. Distinct
structural inversions of an individual kinematic chain can provide distinct motion
characteristics for the same number and type of links due to their different struc-
tural connectivity. To identify the distinct chains and their structural inversions,
characteristics of structural elements of individual kinematic chains are considered
to develop chain motion characteristic matrix. The proposed matrix is applied for
identifying distinct kinematic chains and their inversions. Counter examples of ten-
link and twelve-link co-spectral chains are also tested successfully. Variation of
motion characteristics of individual planar mechanisms of six link kinematic chains
is studied using motion analysis software, CATIA V5, for all inversions of Watt’s
and Stephenson’s kinematic chains. Proposed method thus successfully identifies the
isomorphism among kinematic chains and their inversions and also establishes their
different motion capabilities.
1 Introduction
maximum code (Max code), and other yields to minimum code (Min code). Minimum
code is used to identify isomorphism among kinematic chains. Rao and Varada Raju
[5] proposed hamming distance concept for detecting isomorphism among kine-
matic chains. Sanyal, Rao, and Choubey [6] investigated a novel approach to detect
isomorphism and their inversions among kinematic chains by introducing probability
concept. This research work also involves identification of characteristics of kine-
matic chains such as rating of kinematic chain, best input and output link, and best
inversions. Bedi and Sanyal [7] proposed joint-based method for identification of
isomorphism and inversions among kinematic chains. Link-joint connectivity table
was presented for testing isomorphism among the kinematic chains. Schemes were
prepared from Link-joint connectivity table to detect distinct inversions of kine-
matic chain. Rao and Prasad Raju [8] presented loop-based technique for detec-
tion of isomorphism among kinematic chains, inversions, and type of freedom in
multi degree of freedom chains. Shukla and Sanyal [9] proposed gradient concept
for identification of isomorphism and inversions of kinematic chains. Gradient
matrices are proposed to distinctly denote the structure of each kinematic chain. First
gradient chain string and second gradient chain string are identified for detection of
isomorphism among the planar kinematic chains.
Nielson and Roth [10] solved double butterfly linkage using Dixon determinant
method. For formulation, three vector loops and six equations are formed and solved
using Dixon determinant. Vishwakarma and Ratan [11] proposed an idea of calcu-
lating displacement, velocity, and acceleration by forming vector loop for Stepheson-
I. Vector equations are formed by using these vector loops. These equations are decou-
pled into four more equations for calculating velocity and acceleration of different
links with respect to crank. Workspace concept is important for applications such as
in parallel or platform type robots. Rao and Anne Jagdeesh [12] proposed a mathe-
matical tool which compares the workspace of kinematic chain without performing
dimensional synthesis of kinematic chain.
Unidentified isomorphism among kinematic chains will result in duplication and
loss of resources like time and money due to repetition. To overcome the problem,
a scope for easy, unique, and reliable method still exists. In the present work, joint
value approach is used for proposing the chain motion characteristic matrix for
identification of isomorphism and their inversions. These distinct kinematic chains
have distinct motion characteristics for every possible inversion of the individual
chain because of distinct connectivity between its various types of links and joints.
To establish the effect of structural characteristics, motion of planar kinematic chains
is also analyzed for all the inversions of six-link kinematic chains.
2 Methodology
Joint value approach is proposed for detecting isomorphism and inversions among the
planar kinematic chains. Type of joint and the joint value decide the effective motion
transfer characteristics of that particular joint, which depends upon the number and
Effect of Structural Characteristics on Kinematics … 259
type of links connected to that particular joint. For detection of isomorphism, joint
values between two links along the minimum distance path are considered and chain
motion characteristic matrix (CMC) is obtained for each kinematic chain. Unique
CMC schemes are represented for individual kinematic chains for comparison.
Joint value is proposed as the reciprocal of the number of other joints connected to it.
When two binary links are connected through the joint, corresponding joint is called
as binary—binary (BB) joint. Through BB joint, motion can be transferred to two
nodes (joints), so the joint value corresponding to binary—binary joint is proposed
as ½. Similarly, joint values are calculated for all combinations of distinct type of
links and are noted in Table 1.
Design parameters in terms of length and/or angle are to be specified to use a specific
type of link for designing mechanism. Summation of design parameter of all links in
a specific class of kinematic chain will be constant. Design parameters are fixed for
particular type of link. Design parameter for binary, ternary, and quaternary links is
one, three, and five, respectively.
Mii = 0 (2.1)
and twelve-link kinematic chains. Schemes thus generated are used further for detec-
tion of isomorphism among kinematic chains and distinct inversions of kinematic
chains.
3 Detection of Isomorphism
If the chain characteristic schemes of any two kinematic chains for their chain motion
characteristic matrices are similar, then they are isomorphic otherwise not.
The two schemes generated for the two chains S(1) and S(2) are different and
additionally based on their LCV; it can be concluded that the Watt’s chain has two
inversions and Stephenson’s Chain has three inversions.
The proposed method is applied to eight-link single degree of freedom planar
kinematic chains, Appendix and a total of 16 distinct kinematic chains are identified.
The number of inversions for eight-link single degree of freedom kinematic chains
is 71, and CMC schemes are given in Table 3.
The proposed method is also extended to nine-link double degree of freedom
kinematic chains. It is discovered that total 40 distinct kinematic chains and 254
inversions are possible for nine-link two degree of freedom kinematic chains.
Table 3 CMC schemes for eight-link single degree of freedom kinematic chains
S. No. CMC schemes Inversions
1 3.250 (1, 4, 5, 8), 1.056 (2, 3, 6, 7) 2
2 3.688 (1, 4, 5, 8), 1.222 (2, 3, 6, 7) 2
3 3.688 (1), 1.583 (2), 1.259 (3), 4.521 (4), 1.333 (5), 4.104 (6), 1.056 (7), 8
3.438 (8)
4 1.000 (1, 8), 3.771 (2, 5), 1.243 (3, 4), 3.250 (6, 7) 4
5 1.000 (1, 8), 3.583 (2, 7), 3.438 (3, 6), 1.076 (4, 5) 4
6 1.049 (1, 6), 3.688 (2, 5), 1.271 (3, 4), 3.250/3.639 (7), 3.250 (8) 5
7 1.037 (1), 3.583 (2, 6), 1.104 (3, 5), 4.111 (4), 4.667 (7), 1.083 (8) 6
8 1.000 (1, 4, 7, 8), 3.667 (2, 3, 5, 6) 2
9 3.917 (1, 2, 4, 7), 1.083 (3, 5, 6, 8) 2
10 5.972 (1), 0.903 (2, 6, 7, 8), 3.708 (3, 5), 1.139 (4) 4
11 4.000 (1), 1.153 (2), 1.125 (3), 6.458 (4), 0.931 (5, 7), 3.396 (6), 0.965 (8) 7
12 1.083 (1), 3.757 (2), 0.886 (3, 8), 3.900 (4), 0.963 (5), 1.215 (6), 0.886 (7) 7
13 3.050 (1), 1.059 (2), 0.969 (3), 5.958 (4), 1.123 (5), 1.212 (6), 3.688 (7), 8
0.886 (8)
14 3.060 (1, 7), 1.046 (2, 6), 0.946 (3, 5), 5.500 (4), 0.989 (8) 5
15 0.938 (1, 3, 4, 6, 7, 8), 5.208 (2, 5) 2
16 6.563 (1, 4), 1.094 (2, 3, 5, 6), 0.813 (7, 8) 3
Total = 71
262 A. Biradar and S. Sanyal
5 Path Analysis
Further, variation in shape of coupler curves is also studied for six-link kinematic
mechanisms using motion simulation software. Dimensions of each parts are kept
same in both Watt’s and Stephenson’s kinematic chains and given in Table 6. In
present work, link dimensions are as given in Table 6.
CATIA V5 software is used to generate coupler curve, and coupler area is calcu-
lated. Stephenson I, Stephenson II, Stephenson III, and Watt II mechanisms are
simulated by using Digital Mockup (DMU). Area traced by coupler curves, i.e.,
workspace, is analyzed for six-bar Watt’s and Stephenson’s mechanism, as given
in Table 7. It is concluded that the area covered by Watt’s mechanism is greater
in comparison with Stephenson’s mechanism for same link dimensions as is also
proposed by Rao [12] (Fig. 4).
Effect of Structural Characteristics on Kinematics … 265
Stephenson I Stephenson II
6 Summary
References
1. Crossley ERF (1964) A contribution to Gruebler’s theory in the number synthesis of plane
mechanism. J Eng Ind 1–8
2. Dobrjanskyj F (1967) Some applications of graph theory to the structural analysis of
mechanisms. ASME J Eng Ind B(1)89:153–158
3. Uicker JJ, Raicu A (1975) A method for the identification and recognition of equivalence of
kinematic chains. Mech Mach Theor 10:375–383
4. Ambekar AG, Agrawal VP (1987) Canonical numbering of kinematic chains and isomorphism
problem: min code. Mech Mach Theor 22:453–461
5. Rao AC, Varada Raju D (1991) Application of the hamming number technique to detect
isomorphism among kinematic chains and inversions. Mech Mach Theor 26(1):55–75
6. Sanyal S, Choubey M, Rao AC (1997) Pseudo probabilistic approach to detect isomorphism
among kinematic chains. Trans Canadian Soc Mech Eng 21(2):65–84
Effect of Structural Characteristics on Kinematics … 267
7. Bedi GS, Sanyal S (2010) Joint connectivity: a new approach to detection of isomorphism and
inversions of planar kinematic chains. J Inst Eng (India) 90:23–26
8. Rao AC, Pathapati VVNRPR (2000) Loop based detection of isomorphism among kinematic
chains, inversions and type of freedom in multi degree of freedom chain. ASME J Mech Des
122:31–42
9. Shukla A, Sanyal S (2017) Gradient method for identification of isomorphism of planar
kinematic chains. Aus J Mech Eng
10. Nielson J, Roth B (1999) Solving the input/output problem for planar mechanisms. ASME J
Eng Ind 121B(1)89:206–211
11. Vishwakarma A, Ratan SS (2016) Kinematic analysis of Stephenson-I six-link mechanism.
Thesis, NIT Kurukshetra
12. Rao AC, Anne J (1998) Topology based characteristics of kinematic chains: work space, rigidity,
input-joint and isomorphism. Mech Mach Theor 33(5):625–638
A Combined Experimental/Finite
Element Model Analysis on Compressive
Behavior of Tamarind Pod Shell Filler
Reinforced Composites
Abstract The purpose of this work is to study the compressive behavior of Tamarind
Pod Shell (TPS) reinforced epoxy composite using an experimental and finite element
(FE) modeling approach. The compressive behavior of the composites was investi-
gated based on the different particle sizes (75, 105, 150, and 250 µm) and NaOH
treatment (5 wt%). The developed FE model validated the compressive strength using
the representative volume element (RVE). The surface treatment was implemented
in the FE model to compare with experimentally NaOH-treated TPS composites. The
results revealed that good agreement between the FE and experimental results. This
shows the accuracy and adaptability of the TPS filler reinforced composite model.
1 Introduction
S. Goudar · R. K. Jain
Academy of Scientific and Innovative Research (AcSIR), Ghaziabad 201002, UP, India
S. Goudar · R. K. Jain (B) · D. Das
CSIR-Central Mechanical Engineering Research Institute (CMERI), Durgapur 713209, West
Bengal (WB), India
e-mail: [email protected]
2 Experimental
2.1 Materials
The epoxy resin (LY 556) along with hardener HY 951 (Make: M/s Hindustan Ciba
Geigy; India) was used as matrix. The density of pure epoxy matrix used in this
study was 1.1880 gm/cm3 . Tamarind fruits were collected from North Karnataka
region of India. At first, the non-food waste TPS was separated from the fruits.
The TPS’s were thoroughly cleaned with distilled water in order to remove dirt and
other contaminants. Later, they were dried under sunlight to reduce the moisture
content. The density of TPS used in this study was 1.3236 g/cm3 . The dried TPS’s
were micronized using a grinder and screened using a selected sieve size fraction
of BS 60 (250 µm), BS 100 (150 µm), BS 150 (105 µm), and BS 200 (75 µm).
Before composite manufacturing, the TPS particles were treated with 5 wt% NaOH
to remove the hemicellulose, lignin and improve bonding between TPS particles and
epoxy. The TPS particles were extensively washed with distilled water to remove the
excess of alkali. The fillers were dried in the hot air oven at 50 °C for 24 h and then
stored in glass sealed container to prevent air moisture contamination.
Firstly, a measured quantity of epoxy resin was mixed with a hardener in the weight
ratio of 10:1. Then, TPS filler content was mixed with prepared resin mixture at a
weight ratio of 10:90. The composite mixture was poured into the glass mold which
has cavities to accommodate the composite samples and left at room temperature
until the mixture was cured. For easy removal of composite samples from mold
cavities, the mold cavities were coated with silicon spray (releasing agent).
The following equation can be used to calculate the actual density of matrix, fillers,
and composites using the pycnometry technique [21]:
w1
ρa = ρw (1)
w2 − w3 + w1
where ρa and ρw represent the actual density of the sample and density of distilled
water at 25 ± 3 °C respectively; w1 , w2 and w3 represents the sample weight,
pycnometer filled with distilled water weight, and pycnometer weight containing
the sample and distilled water, respectively.
272 S. Goudar et al.
The theoretical composite density was obtained using the Agarwal and Broutman
equation [22] in terms of weight fractions:
1
ρt = (2)
Wm
ρm
+ Wr
ρr
ρt − ρa
vvoid = (3)
ρt
Compressive tests were carried out using Universal Testing Machine (Tinius Olsen,
Model H50KS, USA). The compression test was performed in accordance with
ASTM D695 standards at a standard crosshead speed of 1 mm/min [23]. Presented
values of compressive strength were considered from an average of at least five
samples of each composite.
Particle phase
Matrix phase
In Digimat-FE module, inter-phase means the matrix zone in the vicinity of a particle
which is influenced by the presence of the particle. In order to define the inter-phase
zone, it is necessary to specify the mass fraction or thickness of cohesive material.
In the present study, the properties of NaOH such as density (1.05 g/cm3 ) and a
mass fraction (5 wt%) were assigned to the inter-phase cohesive material. Shah
et al. [26] considered the inter-phase model as the treated surface and the properties
of polyvinyl alcohol were assigned to the inter-phase zone. The particles with the
cohesive inter-phase zone are shown in Fig. 1.
The Von Mises failure criterion model was considered to describe the damage
initiation of the composite model. It can be expressed as follows:
f i = Fi (σ ), with:
Table 1 represents actual density, theoretical density, and void content of composites.
Generally, the density of composite depends on the relative proportions of fillers and
matrix material. The calculated theoretical composite density values were found to
differ from actual density values. This is due to the voids present in the composites.
A similar trend was observed by Barczewski et al. [28] in the case of sunflower
husk, hazelnut shell, and walnut shell epoxy composites. After NaOH treatment,
it was found that the density of treated TPS is 1.4223 gm/cm3 . This is because of
alkali treatment increases the filler density by removing the lower denser regions
such as hemicelluloses and lignin. This observation is in good agreement with the
results reported in earlier literature [18, 29]. It was observed from the results that
the density of composites increased with the decrease in the particle size due to less
volume being occupied by fillers at lower particle size. A similar trend was observed
by Shinoj et al. [30] in case of oil palm-filled polyethylene composites. It was also
noted that the content of the voids decreased with the decrease in particle size. It was
obvious that the lowest void content was shown by composite with 75 µm particle
size. All the void content values of composites were used to develop the composite
model in Digimat-FE analysis.
particles provided the better transfer of compressive load from the matrix to the rein-
forcement phase by minimizing the space with the matrix phase and also possessed
a larger surface area, thereby enhancing the compressive strength. An increasing
mechanical property with a reduction in particle size was also observed by Gungor
et al. [7], Raju et al. [31], Nabinejad et al. [32], Kim et al. [33].
FE analysis The stress field heterogeneity was investigated at the microstructure
level of the matrix, particle and void phase using the FE model. The results of the FE
analyses are shown in the form of stress distributions that correspond to each material
phase; the results are shown in colored contours in order to represent the stress
distribution at the microstructure levels. Stress distribution contour plots, damage
initiation and deformation of RVE containing a matrix (Epoxy) and inclusions (TPS)
under uniaxial compression are shown in Fig. 2. Figure 2 clearly illustrated how the
stress is distributed in the location of the particles and voids for different particle
size. The compressive behavior of the composites in the FE simulation was similar
to that of the experimental; with a decreased particle size, compression strength
was increased. Also, it was seen that there is no perfect stress transfer between the
particle and matrix in the 250 µm particle size model. However, the stress transfer
from the matrix to the particle phase can be clearly observed (Fig. 2) after particle
size reduction. It was observed that the stress values are approximately zero in the
regions of voids. This is because of ultra-low properties defined in void elements.
Damage value related to 0 is for undamaged material and 1 is for critical damaged
material at which failure is supposed to occur [20]. In fact, failure is deemed to occur
if the damage value reaches or exceeds the critical value of 1.
For all the composites, it was seen that damage was initiated from the void
and propagated across the matrix phase which eventually spreads throughout the
composite model resulting in the failure of the composite.
4 Conclusion
In this work, the compressive behavior of TPS filler reinforced epoxy composites
is studied through experimental and FE analysis. Experimental investigations were
carried out to obtain physical and compressive properties of TPS composites using
four different particle size (75, 105, 150, and 250 µm) and surface treatment. An FEA
276
Fig. 2 Stress distribution contour plots, damage initiation and deformation of RVE containing a matrix (Epoxy) and inclusions (TPS) under uniaxial compression
S. Goudar et al.
for a 250 µm, b 150 µm, c 105 µm and d 75 µm particle size composite model
A Combined Experimental/Finite Element Model Analysis … 277
Acknowledgements The first author is grateful to All India Council of Technical Education
(AICTE) and Indian National Academy of Engineering (INAE) for the financial support under
Teachers Research Fellowship (TRF) Scheme (Ref. No: INAE/201/TRF dated on July 26, 2018).
We express our sincere thanks to the Director CSIR-CMERI for publishing the chapter.
References
1. Naveen J, Jawaid M, Zainudin ES, Sultan MTH, Yahaya R (2019) Mechanical and moisture
diffusion behaviour of hybrid Kevlar/Cocos Nucifera sheath reinforced epoxy composites. J
Mater Res Technol 8:1308–1318. https://doi.org/10.1016/j.jmrt.2018.07.023
2. Babu NBK, Muthukumaran S, Arokiasamy S, Ramesh T (2018) Thermal and mechanical
behavior of the coir powder filled polyester micro-composites. J Nat Fibers 1–11. https://doi.
org/10.1080/15440478.2018.1555503
3. Singh AA, Afrin S, Karim Z (2017) Green composites: versatile material for future
4. Sajith S, Arumugam V, Dhakal HN (2017) Comparison on mechanical properties of lignocel-
lulosic flour epoxy composites prepared by using coconut shell, rice husk and teakwood as
fillers. Polym Test 58:60–69. https://doi.org/10.1016/j.polymertesting.2016.12.015
5. Sathishkumar TP, Kumar SA, Navaneethakrishnan P, Siva I, Rajini N (2018) Synergy of cashew
nut shell filler on tribological behaviors of natural-fiber-reinforced epoxy composite. Sci Eng
Compos Mater 25:761–772. https://doi.org/10.1515/secm-2016-0243
6. Ikladious N, Shukry N, El-Kalyoubi S, Asaad J, Mansour S, Tawfik S, Abou-Zeid R (2017)
Eco-friendly composites based on peanut shell powder/unsaturated polyester resin. Proc Inst
Mech Eng Part L J Mater Des Appl 146442071772237. https://doi.org/10.1177/146442071772
2377
7. Güngör A, Akbay IK, Özdemir T (2019) Waste walnut shell as an alternative bio-based filler for
the EPDM: mechanical, thermal, and kinetic studies. J Mater Cycles Waste Manag 21:145–155.
https://doi.org/10.1007/s10163-018-0778-6
8. Stalin B, Nagaprasad N, Vignesh V, Ravichandran M (2019) Evaluation of mechanical and
thermal properties of tamarind seed filler reinforced vinyl ester composites. J Vinyl Add Technol
25:E114–E128. https://doi.org/10.1002/vnl.21701
9. Tan W, Hao X, Fan Q, Sun L, Xu J, Wang Q, Ou R (2019) Bamboo particle reinforced polypropy-
lene composites made from different fractions of bamboo culm: fiber characterization and
analysis of composite properties. Polym Compos 25329. https://doi.org/10.1002/pc.25329
10. Boran Torun S, Pesman E, Donmez Cavdar A (2019) Effect of alkali treatment on composites
made from recycled polyethylene and chestnut cupula. Polym Compos 25305. https://doi.org/
10.1002/pc.25305
11. Sharma H, Singh I, Misra JP (2019) Mechanical and thermal behaviour of food waste (Citrus
limetta peel) fillers–based novel epoxy composites. Polym Polym Compos 096739111985101.
https://doi.org/10.1177/0967391119851012
278 S. Goudar et al.
12. Lala SD, Deoghare AB, Chatterjee S (2019) Effect of dual pre-treatment on mechan-
ical, morphological, electrical and thermal properties of rubber seed shell-reinforced epoxy
composites. Arab J Sci Eng 44:845–856. https://doi.org/10.1007/s13369-018-3302-3
13. Agunsoye JO, Odumosu AK, Dada O (2019) Novel epoxy-carbonized coconut shell nanopar-
ticles composites for car bumper application. Int J Adv Manuf Technol 102:893–899. https://
doi.org/10.1007/s00170-018-3206-0
14. Salasinska K, Barczewski M, Górny R, Kloziński A (2018) Evaluation of highly filled epoxy
composites modified with walnut shell waste filler. Polym Bull 75:2511–2528. https://doi.org/
10.1007/s00289-017-2163-3
15. Negawo TA, Polat Y, Buyuknalcaci FN, Kilic A, Saba N, Jawaid M (2019) Mechanical, morpho-
logical, structural and dynamic mechanical properties of alkali treated Ensete stem fibers rein-
forced unsaturated polyester composites. Compos Struct 207:589–597. https://doi.org/10.1016/
j.compstruct.2018.09.043
16. Cai M, Takagi H, Nakagaito AN, Li Y, Waterhouse GIN (2016) Effect of alkali treatment
on interfacial bonding in abaca fiber-reinforced composites. Compos Part A Appl Sci Manuf
90:589–597. https://doi.org/10.1016/j.compositesa.2016.08.025
17. Oushabi A, Sair S, Oudrhiri Hassani F, Abboud Y, Tanane O, El Bouari A (2017) The effect
of alkali treatment on mechanical, morphological and thermal properties of date palm fibers
(DPFs): study of the interface of DPF—polyurethane composite. South African J Chem Eng
23:116–123. https://doi.org/10.1016/j.sajce.2017.04.005
18. Devnani GL, Sinha S (2018) African teff straw as a potential reinforcement in polymer compos-
ites for light-weight applications: mechanical, thermal, physical, and chemical characterization
before and after alkali treatment. J Nat Fibers 1–15. https://doi.org/10.1080/15440478.2018.
1546640
19. Shanmugasundaram N, Rajendran I (2016) Characterization of raw and alkali-treated mulberry
fibers as potential reinforcement in polymer composites. J Reinforced Plastics Compos 35:601–
614. https://doi.org/10.1177/0731684415625822
20. Zhang C, Curiel-Sosa JL, Bui TQ (2019) Meso-scale finite element analysis of mechanical
behavior of 3D braided composites subjected to biaxial tension loadings. Appl Compos Mater
26:139–157. https://doi.org/10.1007/s10443-018-9686-0
21. John MJ, Tlili R, Anandjiwala RD, Boudenne A, Ibos L (2012) Effect of amphiphilic coupling
agent on heat flow and dielectric properties of flax–polypropylene composites. Compos Part
Eng 43:526–532. https://doi.org/10.1016/j.compositesb.2011.11.016
22. Shakuntala O, Raghavendra G, Samir Kumar A (2014) Effect of filler loading on mechanical
and tribological properties of wood apple shell reinforced epoxy composite. Adv Mater Sci
Eng 2014:1–9. https://doi.org/10.1155/2014/538651
23. Melkamu A, Kahsay MB, Tesfay AG (2018) Mechanical and water-absorption properties of
sisal fiber (Agave sisalana)—reinforced polyester composite. J Nat Fibers 1–9. https://doi.org/
10.1080/15440478.2018.1441088
24. Naveen J, Jawaid M, Vasanthanathan A, Chandrasekar M ((2019)) Finite element anal-
ysis of natural fiber-reinforced polymer composites. In: Modelling of damage processes in
biocomposites, fibre-reinforced composites and hybrid composites, pp 153–170. Elsevier
25. Ogierman W, Grzegorzek W (2019) The influence of the WC-Co composite microstructure
model on stress field heterogeneity at the microstructure level: FEM based study. Sci Eng
Compos Mater 26:134–146. https://doi.org/10.1515/secm-2017-0421
26. Shah AR, Prabhakar MN, Wang H, Song J-I (2018) The influence of particle size and surface
treatment of filler on the properties of oyster shell powder filled polypropylene composites.
Polym Compos 39:2420–2430. https://doi.org/10.1002/pc.24225
27. Balać I, Čolić K, Milovančević M, Uskoković P, Zrilić M (2012) Modeling of the matrix
porosity influence on the elastic properties of particulate biocomposites. FME Trans 40:81–86
28. Barczewski M, Sałasińska K, Szulc J (2019) Application of sunflower husk, hazelnut shell and
walnut shell as waste agricultural fillers for epoxy-based composites: a study into mechanical
behavior related to structural and rheological properties. Polym Test 75:1–11. https://doi.org/
10.1016/j.polymertesting.2019.01.017
A Combined Experimental/Finite Element Model Analysis … 279
29. Vishnu Vardhini KJ, Murugan R, Tamil Selvi CSR Optimisation of alkali treatment of banana
fibres on lignin removal. Indian J Fibre Text Res 41:156–160
30. Shinoj S, Visvanathan R, Panigrahi S (2010) Towards industrial utilization of oil palm fibre:
physical and dielectric characterization of linear low density polyethylene composites and
comparison with other fibre sources. Biosyst Eng 106:378–388. https://doi.org/10.1016/j.bio
systemseng.2010.04.008
31. Raju GU, Kumarappa S (2011) Experimental study on mechanical properties of groundnut shell
particle-reinforced epoxy composites. J Reinforced Plastics Compos 30:1029–1037. https://
doi.org/10.1177/0731684411410761
32. Nabinejad O, Sujan D, Rahman ME, Davies IJ (2015) Effect of oil palm shell powder on
the mechanical performance and thermal stability of polyester composites. Mater Des 1980–
2015(65):823–830. https://doi.org/10.1016/j.matdes.2014.09.080
33. Kim H-S, Yang H-S, Kim H-J (2005) Biodegradability and mechanical properties of agro-
flour-filled polybutylene succinate biocomposites. J Appl Polym Sci 97:1513–1521. https://
doi.org/10.1002/app.21905
Workspace Evaluation of Robotino-XT
Under Reconfiguration
Abstract The major part of ongoing research in the field of robotics is to mimic the
behaviour of species present in nature. This paper presents the workspace of the multi-
section bionic manipulator Robotino-XT under the failure of one of its bellow. These
types of failure may be due to repeated use, pressure leakage, hysteresis or improper
design. The CBHA part of Robotino-XT has two bending sections, where each section
has three bellows (actuator). Various strategies are proposed to compensate for the
loss in workspace due to failure of one bellow. Also, the workspace obtained after
the failure of the bellow is compared with the full-fledged workspace.
1 Introduction
Bionic engineering is drawing the attention of researchers from the last decade, by
replicating the motion or behaviour of biological species. Although it is the most
challenging task to replicate the exact motion or behaviour that is possessed by the
nature, researchers are trying to develop various mechanism and methods to replicate
their action up to an extent. The outcomes of this research in robotics are known
as soft robots, or soft structures, and the science behind these robots is popularly
known as soft robotics. Although soft structures are safe, flexible, lightweight, have
high dexterity and able to perform under unpredictable or clustered environment but
they are fragile, subject to deflection by its self-weight when made heavy and their
performance can be affected by repeated use.
The era of continuum robots started when Robinson and Davies [1] define the
fundamental difference between discrete, serpentine and continuum robots with out-
lining their areas of applicability and control issues. Jones and Walker [2] proposed
a novel geometrical modelling approach which is applicable to variety of continuum
robots. Trivedi et al. [3] categorized soft robots as pneumatically actuated artifi-
cial muscle robots and hard robots which have soft capabilities such as snake robot
[4] made of several multi-sections modules. Escande et al. [5] develop a geometri-
cal modelling approach for CBHA based on Jones and Walker [2]. The same author
Escande et al. [6] had also experimentally validated their forward modelling approach
on Robotino-XT using trilateration methodology. Their work also includes develop-
ment and validation of the inverse kinematic model for single section continuous
manipulator. Neppalli et al. [7] proposed an analytical inverse kinematics solution
for multi-section continuous robot by compensating the orientation change of end-
points of each section. The methodology gives final-section endpoints only when
per-section endpoints are previously known. Webster and Jones [8] reviewed various
mathematical modelling approaches applicable to soft robots in common frames and
notations, which exhibits indistinguishable outcomes for piecewise constant curva-
ture method.
There are lot of contributions from various researchers to obtain the workspace
of soft structures based on their kinematic model. Li and Du [9] had expanded the
workspace of under-actuated flexible manipulator under bilateral and unilateral con-
straints. Singh et al. [10] had obtained the workspace of two and three CBHA parallel
configuration. Burgner et al. [11] applied Monte Carlo methodology to compute and
characterize the workspace of concentric tube robots. But till now up to best of
authors’ knowledge, no work has been done to estimate the workspace of soft con-
tinuum manipulator when one or multiple actuation module fails. Continuum robots
had gained the attention of researchers due to their light weight structure, excellent
ability of bending and dexterity which ultimately leads their applicability in explor-
ing narrow-unstructured and hazardous environment for minimally invasion surgery
[12], search or inspection operation in collapsed buildings [13] and many more yet
to be explored.
This paper presents the workspace of the Robotino-XT made by Festo Didactic
due to the failure of one of its bellow due to pressure leakage. First section provides
the literature related to soft robot while description of robot is in the second section.
Experimental setup and results are discussed in third and fourth sections, respectively.
Fifth section presents the conclusion and future scope.
2 CBHA Description
three flexible tubes per sections, one rotational part in the form of arc and a com-
pliant gripper inspired from fish tail movement. The innovation in unconventional
manufacturing process such as rapid prototyping leads to successful development of
such types of structures using polyamide. Various positions and orientations of end
efforter are obtained by varying pneumatic pressure inside the bellows controlled by
PID controller. Figure 1 shows the schematic of the CBHA frames, bending sections
and backbone.
Based on piecewise constant curvature, the position of end effector can be obtained
as shown in Eq. (1) [2, 5] where j represents section ( j = 1, 2) and i represents tube
number (i = 1, 2, . . . , 6):
⎡ ⎤ ⎡ ⎤
xj r j cos φ j 1 − cos θ j
⎣ y j ⎦ = ⎣ r j sin φ j 1 − cos θ j ⎦ (1)
zj r j sin θ j
where ⎡ ⎤
⎡ ⎤ (li +li+1 +li+2 )d j
rj ⎢ −1 √D3(lj i+2 −li+1 ) ⎥
⎣ φj ⎦ = ⎢ tan ⎥
⎣ 2li −li+1 −li+2 ⎦
θj Dj
3d j
2
Dj = 2 li + li+1
2
+ li+2
2
− li li+1 − li li+2 − li+1li+2 )
284 V. Pachouri et al.
For tracking position and orientation of the CBHA under different sets of pressure
inputs, four-camera vision system with motive 2.0 is used. Figure 2a shows the pos-
ture of the static CBHA, where centre nodes represent backbone position of various
sections. While Fig. 2b shows position of base, mid and end of the start of first section,
end of first section and end of second section of the CBHA, respectively, by gener-
ating a rigid body using three markers. Figure 3 shows the perspective and camera
view of motive after calibration and various parts of CBHA. The residual mean error
in calibration process is 0.1 mm.
Fig. 2 a Base, mid and end section of CBHA, b Robotino-XT with marker
Fig. 3 a Perspective view of motive, b camera preview of motive, c various parts of CBHA
manipulator
Workspace Evaluation of Robotino-XT Under Reconfiguration 285
4 Experimental Results
The whole experimentation is divided in to two sections. The first section presents the
simulation (filled multi-coloured solid dots as shown in Fig. 4) and experimental (red
coloured plus dots) results of CBHA workspace when all actuating units are working
properly, i.e. no failure condition shown in Fig. 4. Although there is good similarity
between simulated and experimental workspace, positions are not exactly same,
because the varying parameter for simulation is tube lengths, where as the varying
parameter for experimentation is tube pressures. Also, the other reason for differences
is that the tube pressure is restricted to 2 bar. If more pressure would be available, then
simulated and experimental workspace may be same. While second section provides
the reconfigured workspace of CBHA under different failure scenarios of actuation
module as described in Table 1. The pneumatic pressure inside the faulty bellow tube
is assumed to be 0. Figure 5 shows the workspace of CBHA under various scenarios
of tube failure. The design parameters are mentioned in Table 2.
Fig. 4 Simulated and experimental workspace of CBHA when all actuating units working properly
Fig. 5 Workspace of CBHA when a tube number one fails, b tube number two or three fails, c
tube number four fails and d tube number five or six fails
In this paper, the reconfigured workspace of CBHA is compared with the full-fledged
workspace. Based on experimental results shown in Sect. 4, different strategies are
proposed that can be adopted to compensate the reduction of workspace due to failure
of CBHA bellow. The reconfiguration can be done on various parameters of any soft
robotic system, like in terms of stiffness of backbone, contraction perpendicular to
tube backbone and failure of one of its intrinsic part and many more. From the exper-
imental results, it can be concluded that for this type of continuum manipulator, the
base motion of trunk is the best way to compensate the region which end effector has
missed due to failure of one of its bellow. Also, it has been noticed that the reduction
in workspace due to tube failure of second bending section is more prominent than
tube failure of first bending section. The future work is to develop some criteria for
base motion to compensate the reduction in workspace due to tube failure.
Acknowledgements This work is supported by the joint collaborative research project between
DST in India and CNRS in France, Project number DST-CNRS 2015-02/2024.
References
1. Robinson G, Davies JBC (1999) Continuum robots-a state of the art. In: Proceedings 1999
IEEE international conference on robotics and automation (Cat. No. 99CH36288C), vol 4,
USA, pp 2849–2854. https://doi.org/10.1109/ROBOT.1999.774029
2. Jones BA, Walker ID (2006) Kinematics for multisection continuum robots. IEEE Trans Robot
22(1):43–55. https://doi.org/10.1109/TRO.2005.861458
3. Trivedi D, Rahn CD, Kier WM, Walker ID (2008) Soft robotics: biological inspiration, state
of the art, and future research. Appl Bionics Biomech 5(3):99–117. https://doi.org/10.1080/
11762320802557865
4. Simaan N, Taylor R, Flint P (2004) A dexterous system for laryngeal surgery. In: IEEE inter-
national conference on robotics and automation. Proceedings ICRA’04, Louisiana, vol 1, pp
351–357. https://doi.org/10.1109/ROBOT.2004.1307175
5. Escande C, Pathak PM, Merzouki R, Coelen V (2011) Modelling of multisection bionic manip-
ulator: application to RobotinoXT. In: 2011 IEEE international conference on robotics and
biomimetics, USA, pp 92–97. https://doi.org/10.1109/ROBIO.2011.6181268
6. Escande C, Merzouki R, Pathak PM, Coelen V (2012) Geometric modelling of multisection
bionic manipulator: experimental validation on RobotinoXT. In: 2012 IEEE international con-
ference on robotics and biomimetics (ROBIO), China, pp 2006–2011
7. Neppalli S, Csencsits MA, Jones BA, Walker ID (2009) Closed-form inverse kinematics for
continuum manipulators. Adv Robot 23(15):2077–2091
8. Webster RJ, Jones BA (2010) Design and kinematic modeling of constant curvature continuum
robots: a review. Int J Robot Res 29(13):1661–1683
9. Li Z, Du R: Expanding workspace of underactuated flexible manipulators by actively deploy-
ing constraints. In: 2014 IEEE international conference on robotics and automation (ICRA),
Indonesia, pp 2901–2906. https://doi.org/10.1109/ICRA.2014.6907276
10. Singh I, Singh M, Pathak PM, Merzouki R (2017) Optimal work space of parallel continuum
manipulator consisting of compact bionic handling arms. In: 2017 IEEE international con-
ference on robotics and biomimetics (ROBIO), China, pp 258–263. https://doi.org/10.1109/
ROBIO.2017.8324427
288 V. Pachouri et al.
11. Burgner-Kahrs J, Gilbert HB, Granna J, Swaney PJ, Webster RJ (2014) Workspace character-
ization for concentric tube continuum robots. In: 2014 IEEE/RSJ international conference on
intelligent robots and systems, Chicago, pp 1269–1275. https://doi.org/10.1109/IROS.2014.
6942720
12. Cianchetti M, Ranzani T, Gerboni G, Nanayakkara T, Althoefer K, Dasgupta P, Menciassi
A (2014) Soft robotics technologies to address shortcomings in today’s minimally invasive
surgery: the STIFF-FLOP approach. Soft Robot 1(2):122–131. https://doi.org/10.1089/soro.
2014.0001
13. Walker ID (2017) Use of continuum robots for remote inspection operations. In: 2017 comput-
ing conference, London, United Kingdom, pp 1382–1385. https://doi.org/10.1109/SAI.2017.
8252274
Assessment of Surface Water Quality
Using Principal Component Analysis
in the Yamuna River: A Case Study
Abstract Present work studies of surface water quality using principal component
analysis (PCA). Yamuna river is situated in North India. Its water has important role
in irrigation and domestic uses. Yamuna river water is highly polluted from Gautam
Budha Nagar to Agra. The reason is being rapid economic development, urban-
ization, industrialization and agriculture waste and fertilizers. So, water samples
were monitored from Gautam Budha Nagar to Agra for three years (2014, 2015 and
2016). Water sample collection has been done on the basis of past research analysis
and self-survey. These water quality parameters include pH, temperature, electrical
conductivity, dissolved oxygen, turbidity, total dissolved solid, salinity, chloride,
acidity, total alkalinity, total hardness, nitrate ions, biochemical oxygen demand,
chemical oxygen demand and total coliform. PCA technique has been applied in
analysis dataset. Four factors were identified for the data structure explaining of the
total variance (83.73%) of the dataset, in which nutrient factor (41.79%), domestic
sewage contamination (22.41%), physicochemical variability (11.92%) and wastew-
ater from industrial (7.61%) in total variance. Result clearly shows that PCA tech-
nique is useful tool for water pollution resources identification and management.
Anthropogenic activities are responsible for deterioration the river water quality. So,
its needs to spread the awareness for the pollution status of river water publicly.
1 Introduction
Water is natural source on earth. Surface water is important role in ecological balance.
It is found different forms as a like in river, lake, ocean, steams, wells and rainwater
[1]. Yamuna river origin from lower Himalayas (Banderpooch) in North India. Its
water flows from Yamuna Nagar, Panipat, Sonipat, Bagpat Delhi, Gautam Budha
Nagar, Faridabad, Palwal, Aligarh, Mathura, Agra, Etawah and Prayagraj (Alla-
habad) [2]. Yamuna river water is used for irrigation purposes, industrial purposes
and domestic purposes [3]. Largely cities and industries are situated in the Yamuna
river bank; so, its water has been polluted. The reason is being unplanned develop-
ment, urbanization, industrialization and use of fertilizers in agriculture [4]. Yamuna
river water is affected by these anthropogenic activities. In recent past years, many
researchers used PCA technique for identification of water pollution resources on the
different surface and groundwater [5–7]. The object of the present study is assess-
ment of surface water quality using principal component analysis in the Yamuna
river: A case study.
2 Literature Review
Mishra studies of surface water quality using principal component analysis on the
Ganga river water. Result concluded that this technique is useful tool for water
pollution resources identification [8]. Xiaoyun et al. described of water quality in
Pearl river using multivariate statistical techniques at three sites [9]. Mohd Saiful
Samsudin et al. assessment of surface water quality using PCA techniques: A case
study at Perlis river basin Malaysia. The PCA showed nine main pollution sources on
Perlis river which are a mineral component of the river water [10]. Admu, Mustapha
et al. assessment of Getsi river water quality using principal component analysis
and factor analysis [11]. Murugesan, Athimoolam et al. reported of groundwater
quality using principal component and factor analysis. The groundwater samples
were collected from various locations of west, south and central Chennia in Tamil-
nadu during 2012–2013 in pre-monsoon, monsoon and post-monsoon season [12].
Finally, concluded that PCA technique is reliable technique for identification of water
pollution resources and management.
Large urban and industrial centers are situated in the Yamuna river bank. So, these
areas have been polluting. In the present survey, analysis has been carried out for six
sites which are selected as sample stations. These sample stations are Kalindi Kunj
near flyover, Gautam Budha Nagar (S1 ), Chhainsa village near bridge, Faridabad
(S2 ), Yamuna river bank near Gurwari village, Palwal (S3 ), Yamuna river bank near
Pipli village, Aligarh (S4 ), Gokul Barrage near bridge, Mathura (S5 ) and Hathi Ghat
near railway bridge, and Agra (S6 ) water samples were analyzed by standard method
[13, 14]. PCA technique has been used for data analysis.
Assessment of Surface Water Quality Using … 291
All parameters minimum and maximum values have been identified. The pH concen-
tration value varies from 7.08 to 7.98. The temperature ranges from 27 to 40 °C
in the observation period. The EC values varied from 1022 to 1776 µs/cm. The
dissolved oxygen value varies from 1.8 to 4.8 mg/L. The turbidity varies from 12
to 26 NTU, indicating soil runoff in river water. The TDS values varied between
918 and 1560 mg/L. The salinity values varied from 0.307 to 0.56 ppt. The chloride
value varies from 170 to 310 mg/L. The acidity values of water varied from 7.76 to
42.85 mg/L. The total alkalinity values varied from 218 to 540 mg/L indicated that
dissolved carbon dioxide present in river water. The total hardness values ranged
between 226 and 372 mg/L. The nitrate values varied from 18 to 40 mg/L. The BOD
values varied between 20 and 56 mg/L and are not in permissible limit indicating
organic matter contamination in water. The COD values varied from 67 to 130 mg/L.
Its values are above the BIS permissible limit and indicating oxidation of waterborne
organic and inorganic matter present in river water. The TC values varied from 15,000
to 7,800,000 MPN. Mostly, water quality parameters have been found above the BIS
limit (Tables 1 and 2).
Four reasonable factors were identified for the data structure explaining of the total
variance (83.73%) of the dataset, in which nutrient factor (41.79%), domestic sewage
contamination (22.41%), physicochemical variability (11.92%) and wastewater from
industrial (7.61%) in total variance of river water (Fig. 1).
Eigen Value
7
5
Eigen Value
3
Eigen Value
2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Components
5 Conclusion
Yamuna river water is polluted from Gautam Budha Nagar to Agra. PCA technique
has been applied in analysis dataset, in which nutrient factors (41.79%), domestic
sewage contamination (22.41%), physicochemical variability (11.92%) and wastew-
ater from industrial (7.61%) that represents variance in Yamuna river water. Finally,
concluded that river water has polluted. PCA technique is better technique for large
data analysis of surface water and groundwater.
References
1. Singh KP, Malik A, Sinha S (2005) Water quality assessment and apportionment of pollution
sources of Gomati river, India using multivariate statistical techniques: a case study. Anal Chem
Acta 538:355–374
2. Rout Chadetrik (2017) Assessment of water quality: a case study of river Yamuna. Int J Earth
Sci Eng 10(2):398–403
3. CPCB (2006) Assessment and development of river basin series ADSORBS/41/2006-07.
Cental Control Pollution Board, New Delhi
4. Sharma Deepshikha, Kansal Arun (2011) Water quality analysis of river Yamuna using water
quality index in the national capital territory, India (2000–2009). Appl Water Sci 1:147–157
5. Kumar P, Kaushal RK, Nigam AK (2015) Assessment and management of Ganga river water
quality using multivariate statistical techniques in India. Asian J Water Environ Pollut 12(4):61–
69
6. Raj Nikhal, Azeez PA (2009) Spatial and temporal variation in surface water chemistry of a
tropical river the river Bharathapuzha, India. Curr Sci 96(2):245–251
7. Verma RK, Verma KS, Bhardwaj SK (2015) Pollution potential assessment of Markandar river
around Kala-Amb industrial town of Himachal Pradesh, India. Int J Sci Nat 6(4):606–612
8. Mishra A (2010) Assessment of water quality using principle component analysis: a case study
of the river Ganga. J Water Chem Technol 32(4):227–234
9. Fan Xiaoyum, Cui Baoshan, Zhao Hui, Zhang Zhiming, Zhang Honggang (2010) Assessment
of river water quality in Pearl river delta using multivariate statistical techniques. Procedia
Environ Sci 2:1220–1234
10. Samsudin MS, Juahir Hafizan, Zain SM, Adnan NH (2011) Surface river water quality inter-
pretation using environment techniques: case study at Perlis river basin, Malaysia. Int J Environ
Prot 1(5):1–8
11. Mustapha Adamu, Nabegu AB (2011) Surface water pollution sources identification using
principle component analysis and factor analysis in Getsi river, Kango Nigeria. Aust J Basic
Appl Sci 5(12):1507–1512
12. Athimoolam M, Velayutham RB (2014) Hydrochemistry of groundwater in Chennai city using
principal component and factor analysis: a case study. Res J Chem Environ Sci 2(2):78–88
13. Kaul SN, Gautam (2002) A water and waste water analysis. Daya Publishing House, Delhi
14. Maiti SK (2011) Hand book method in environmental studies. Oxford Book Company, New
Delhi
Performance Analysis of Gripper
Assembly of an In-Vessel Fuel Handling
Machine
Abstract This paper deals with the analysis of gripper assembly of a long slender
in-vessel fuel handling machine (IVFHM) of a fast breeder reactor towards assessing
the behaviour of various kinematic links and joints. Finite element analysis is carried
out for two different operating positions viz. gripper assembly freely hanging and
gripper resting on sub-assembly (SA) head. Results of analysis are interpreted to
understand the kinematic behaviour of various kinematic links and joints. Reac-
tive forces and deformation behaviour of gripper assembly corresponding to freely
hanging case and gripper resting on SA case are found out. Role of design features
such as special joint between actuator rod and connecting piece to facilitate rotational
degree of freedom, hard faced central guide with double curvature is explained. The
present study provides valuable information for the design of gripper assembly of
IVFHM with increased offset. Role of design features such as spherical guide arrange-
ment for connecting rod, joint between connecting rod and inner tube permitting
relative rotation and limited lateral displacement for gripper assembly of IVFHM
with increased offset is also discussed briefly.
1 Introduction
Fast breeder reactors (FBRs) are a special class of nuclear reactors having the unique
capability to provide long term energy security by better utilization of uranium
resources [1]. Fuel handling refers to the refueling of reactor core with fresh fuel
sub-assemblies (SA) and is carried out with reactor in shutdown condition [2]. It
involves receipt, inspection, storage and loading of fresh SA into the main vessel of
reactor, in-vessel transfer, discharge, washing and storage of spent SA. Design of fuel
handling machines is very challenging considering the fact that in-vessel handling
has to be carried out under sodium coolant of reactor which is opaque and also has
to be carried out remotely due to the radioactive environment of reactor.
In-vessel handling of fuel SA is carried out through the combined action of two
rotatable plugs and an offset arm type in-vessel fuel handling machine (IVFHM) [3].
IVFHM consists of various mechanisms viz. gripper assembly to grip the required
fuel SA, hoisting mechanism to lift the gripper along with SA, rotation mechanism
to align the gripper to required core position, guide tube arrangement, etc. This paper
deals with the analysis of gripper assembly of IVFHM at two operating positions
and interpretation of results obtained to ensure the smooth performance of various
kinematic elements and joints.
For both the cases, analysis has been carried out by considering fixed type of joint
and joint allowing rotational dof between actuator rod and connecting piece.
Deflected shape of gripper assembly (lateral direction) for case A and case B is shown
in Fig. 2a, b, respectively. Deformation results indicate that inner tube assembly is
much more flexible in bending than outer tube assembly and is restricted to deform
further by the guides in outer tube assembly which leads to development of reactive
forces in guides. For case A, it is observed that for F up to −3 kg, guiding of inner tube
happens between central rail-guide arrangement and the guide at EL 14,953. Further,
for F greater than −3 kg, guiding occurs between central rail-guide arrangement and
the guide at EL 640.
Analysis results indicate that if actuator rod which is part of inner tube assembly
deflects more in lateral direction, more reactive forces can develop in the actuator rod
guiding locations during motion and can affect the smooth movement. The central
rail-guide arrangement (Fig. 3) is provided to reduce the lateral deflection and rotation
of overhanging connecting rod thereby reducing the lateral deflection of actuator rod
attached to it through connecting piece and thus ensuring the smooth movement of
actuator rod even under increased value of F. The maximum value of lateral reactive
force expected in central rail-guide arrangement for case A is 10 kg for an actuator
rod resistive force of F = 10 kg.
Deflection pattern of gripper assembly when gripper rests on fuel SA head is
shown in Fig. 2b. Since the entire weight of 1000 kg is transferred at an offset, outer
tube assembly takes the shape as shown. This, in turn, forces the inner tube assembly
also to take similar shape leading to the development of ~35 kg lateral reactive
force at the central rail-guide arrangement and guide at EL 640. Reactive force
increases to ~48 kg when F value is increased to 10 kg. This aspect is considered
in the geometry selection of contacting surfaces. Since the central rail and guide
are mounted on flexible components, double curvature (R100 and R30) as shown in
Fig. 3 is provided to avoid edge loading and thereby limiting contact stresses. Double
curvature ensures that even if guide undergoes relative rotation within rail, contact
always occur between curved portion of guide and flat surface of rail. In addition,
contacting surfaces are hard faced with Colmonoy-5 up to a depth of 1.5–2 mm
from surface. This ensures that the point experiencing maximum sub-surface shear
stresses are well within the Colmonoy-5 hard faced region having hardness in the
range of 45–50 HRC [4]. In addition, clearances and geometry are chosen so as to
ensure that the rotations are not restricted in a small distance leading to development
of large reactive forces.
300 A. Krishnan et al.
Analysis results (for both case A and case B) considering fixed joint and joint
allowing rotational dof between actuator rod and connecting piece indicate that rota-
tional dof is required to ensure that no additional resistance is developing in actuator
rod. Design feature as shown in Fig. 4 is provided to ensure the same.
The present study provides valuable information for the design of griper assembly of
IVFHM with larger offset. A special joint is provided between actuator rod and
connecting piece that permits relative rotation as well as limited relative lateral
displacement. Thus, the displacement and rotation of connecting piece is not trans-
ferred to actuator rod, thereby ensuring that the chance of development of additional
reactive force in actuator rod is eliminated even for larger values of offset. Instead of
central rail-guide arrangement, connecting rod is guided in outer tube through two
spherical guides. In addition, a special joint permitting relative rotation and limited
relative lateral displacement is introduced between connecting rod and rest of the
inner tube. This ensures that the offset portion of inner tube assembly is isolated from
rest of inner tube thereby reducing reactive forces. Analysis results indicate that up
on introduction of this special joint, reactive forces are independent of whether the
gripper assembly is freely hanging or gripper is resting on SA head.
Performance Analysis of Gripper Assembly … 301
6 Summary
Finite element model has been developed for two operating positions of gripper
assembly of IVFHM. Analysis results indicate that inner tube being more flexible
tries to deflect more and is restrained by outer tube leading to the development of
reactive forces in guides. In resting on SA case, outer tube forces inner tube to take
its deflection pattern which leads to development of nearly 35 kg reactive force in
central rail-guide arrangement. Role of design features such as double curvature
of central guide, special joint between actuator rod and connecting piece allowing
rotational dof enhances the robustness of the machine.
References
Abstract Circuit breaker is a device used to disrupt the flow of current when an
abnormal condition occurs in switchgear. These withdrawable breakers are regularly
disconnected for maintenance, testing and replacement. A novel design of a portable,
electrically driven racking mechanism which is easy to mount on circuit breaker
without any modifications in its enclosure, is presented in this work. The performance
of the designed racking mechanism is verified by conducting finite element analysis
(FEA) simulation and experimental studies. A prototype is constructed based on the
analysis result, and the performance is validated in an actual circuit breaker.
1 Introduction
zone. To minimize hazards to various people engaged with the process and other near
the equipment, installing remote circuit breaker racking systems that allow personnel
to safely rack in and out circuit breakers from a remote safe location is advisable.
Other existing systems use wheeled platform which is positioned in front of the
breaker, and others need an alteration in the door of the switchgear for mounting the
device, etc. The solution proposed in this work is safe and reliable, offers an alternate
method for racking breakers and reduces the need for a service operator to wear a
bulky full body suit.
Zlystra et al. [1] developed a remotely controlled racking mechanism using a
drive with the help of an actuator pin and button pin housed inside the device. Bower
et al. [2] developed a racking device that can be operated at a safe distance, actuating
mechanism to interface the interlocks and a mounting assembly. Hawkins et al. [3]
developed an apparatus along with a frame for attachment. Vaill et al. [4] developed a
device for displacing the circuit breaker with the help of a trolley system to facilitate
movement.
In view of the precautions and difficulties incurred during racking of circuit breaker
devices inside and outside of switchgear cubicle, a portable motorized remotely
racking device is developed. The risk of the operator during racking is virtually
eliminated in as much as the operator is stationed at a distance from the circuit breaker
when racking operation is done. For this purpose, the amount of torque required to
rotate the lead screw in the withdrawal part is to be calculated. The torque requirement
during racking operation, taking into account the friction phenomenon, is calculated
as shown in Table 1.
To serve the problem basically, the device consists of a mounting unit, racking
device and handheld controller. The primary purpose of this unit is to tightly hold
the racking mechanism during racking operation. The mounting unit is applied to
the circuit breaker compartment door to accept the link with the portable driver. The
portable mounting unit is designed to be grasped by an operator to position it near the
access hole on the front panel of the switchgear. Engagement members such as latch
bracket, provided at two sides, may engage with the engagement features (such as
toggle latch) of the remote racking apparatus. A mounting unit is designed keeping in
mind the space constraints, ease of mounting and portability. The mounting unit can
be rotated in an angular motion for proper alignment of the racking device (Fig. 1).
The racking mechanism is comprised of a motor, gearbox, motor brackets,
coupling, shaft and many electronic types of equipment. The motor, integrated with
the gearbox, is fixed using two motor brackets. The motor has a square socket attach-
ment, service as a coupler, which attaches to the motor shaft. A toggle latch is mounted
on suitable pins on the side faces of the racking device. The latch is adjustable, so
that it aligns with the latch bracket provided on the portable mounting unit safely
securing the device onto the front panel to carry out racking operations.
A power supply, such as 48 VDC 4.4A supply, provides supply to various compo-
nents of the remote racking device. The racking mechanism consists of an electric
motor which serves as an actuator for the racking mechanism. A driver is used for
inter-facing motor with the control circuits. In addition, an Arduino UNO is used for
controlling the entire circuit. It also consists of an encoder which senses the number
of rotation, and the motor uses the feedback which it receives from the encoder to
track the rotations. As an added safety measure, there is a provision for sensing the
over-torque condition arising due to various reasons.
The handheld remote racking controller consists of a number of controls and
indicators for the operation. The indicators are used to indicate the state of the device
and the operations. The control members like pushbuttons start or stop the selected
configuration.
To mount the remote device, the operator aligns the mounting unit with the corre-
sponding crank access hole in the front panel of the cabinet. The portable mounting
unit is mounted on the panel by using magnets placed at four limbs of the mounting
unit. The remote racking device is clamped against the portable mounting unit by
using toggle latches. The operator connects the electrical cable between the remote
racking device and the remote control device. The racking mechanism consists of
an electric motor which serves as an actuator for the racking mechanism. It also
consists of an encoder which senses the number of rotation, and the motor uses the
306 A. S. Syriac and S. S. Chiddarwar
feedback which it receives from the encoder to track the rotations. As an added
safety measure, there is a provision for sensing the over-torque condition arising due
to various reasons. The handheld remote racking controller consists of a number of
controls and indicators for the operation.
Closing the switch, the operator actuates the electric motor direction switch to
indicate whether to rotate the motor in forward (e.g., clockwise) direction (such that
breaker is racked into the cabinet into a connected position) or in reverse direction
(such that the circuit breaker is racked outside the cabinet into a disconnection posi-
tion). A signal is sent via cable to the racking device, which causes to turn the motor
in the direction indicated by the operator. When the desired number of rotation is
achieved, power is removed from the motor causing it to stop. Any rise in current or
over-torque condition causes the entire system to shut down automatically. A proto-
type of the remote racking device as shown in Fig. 2 is fabricated for displacing the
breaker during racking operation.
Static analysis is carried out on the remote racking device to check the performance
when attached to the circuit breaker for racking operation. This performance study
Design and Development of a Remote Racking Mechanism … 307
Fig. 3 FEA simulation result of the device with self-weight acting. a Directional deformation when
limbs 40° apart, b total deformation when limbs 40° apart, c directional deformation when limbs
90° apart, d total deformation when limbs 90° apart, e directional deformation when limbs 120°
apart and f total deformation when limbs 120° apart
helps to determine the best configuration that the mounting unit can take to minimize
the vertical deformation due to self-weight. To find the best configuration of the
mounting unit, three positions of the limbs are selected for analysis. The selected
three angular positions between the two limbs of the unit are 40°, 90° and 140°,
respectively, as shown in Fig. 3. The configurations are selected such that symmetry
is maintained. It is observed that as the angle between the limbs decreases, the total
as well as directional deformation decreases. Therefore, the study helps to find out
the configuration which can hold the device with a better gripping force.
Dynamic analysis is carried out on the remote racking device to check the perfor-
mance when attached to the circuit breaker for racking operation. The analysis is
carried out when the limbs are apart by 45°. It is observed that all types of vibration
308 A. S. Syriac and S. S. Chiddarwar
occur in the device such as longitudinal, bending, torsional and combined as shown
in Fig. 4, and Table 3.
Based on the results obtained in simulation, the device is validated by testing it
on a circuit breaker. The preliminary results indicate that real-time monitoring of the
position of breaker is needed for the breaker to return to the disconnected position
when any power failure occurs in the system. Further, the weight of the system can
be reduced to improve gripping.
4 Conclusion
This paper proposed a remote racking device for displacing circuit breaker to the
desired position that can be included in the structure of the breaker to carry out racking
operation. The mechanisms described in this paper have the potential to reduce the
risk faced by an operator in an arc flash explosion by providing a minimum safe
distance between the live switchgear and the operator, thereby providing inherently
safe architectures for applications in physical human–switchgear interaction. The
proposed mechanism includes an electric motor as well as a transmission connectable
Design and Development of a Remote Racking Mechanism … 309
to the motor and to a screw for moving the breaker. The system further includes a
mounting unit for securing the device and handheld controller to operate the device
from a remote location. The developed remote racking device was analyzed and
tested on a circuit breaker in a real environment.
References
1. Zylstra et al (2010) Portable remote racking device for a circuit breaker. US patent 8,395,065
B2, 2 September 2010
2. Bower et al (2012) Remotely-operated switchgear racking device and a mounting method for
the same. US patent 2012/0199450 A1, 8 February 2012
3. Hawkins et al (2011) Circuit breaker racking apparatus, systems, and methods of using same.
US patent 8, 553, 394, B2, 17 March 2011
4. Vaill et al (2005) Racking device and power module therefor. US patent 7,019,230 B1, 3 August
2005
Iwan Model for Bolted Joint with
Residual Macroslip Stiffness and Pinning
1 Introduction
Bolted joints are the building block of many assembled structure. Introduction of
bolted joints brings non-linearity into system due to the presence of friction at the
interface. When a bolted joint is subjected to dynamic loading, its slowly loses its
preload force. This results in internal phenomena such as microslip, macroslip and
pinning into the joint [1]. These phoenomena further adds up to the challenges in
developing a predictive model of bolted joint. Microslip is the partial microscopic
slip taking place at joint interface due to loading, whereas macroslip is the full slip
of bolted joint at interface. Pinning is the interaction of bolt shank with the internal
surface of bolt hole.
To obtain the predictive model, there exit non-parametric and parametric joint
models. The non-parametric model is widely used as no assumption about property
of joint is required. Parametric model needs to have good understanding of property
of joint. Iwan model for bolted joint is an example of parametric model. It was
first developed to model hysteretic behaviour of a material [2]. Iwan considered
a hysteretic material as a combination of Jenkins element to model hysteretic. It
consists of a spring and a slider connected in series. These elements act as an ideal
elasto-plastic member. Further, this model was first applied by Seagalman [3] to
develop a predictive model for the bolted joint. He developed a four-parameter Iwan
model by replacing joint by a combination of Jenkin’s element connected in parallel
as shown in Fig. 1a. Further improvement over Segalman model was achieved by
other researchers [4–6]. Gaul and Lenz [7] experimentally showed that even after
macroslip, there is some residual stifness left in bolted joint. These phenomena were
not addressed by Segalman in his model. Hence, to include these phenomena Li and
Hao [5] developed six parameters Iwan model. This model addresses the microslip
and macroslip with residual stifness after macroslip. However, they did not consider
pinning of bolted joint which is an inherent property of the joint. Brake [4] in his
reduced order Iwan model included the pinning phenomena by using the concept
of Hertzian contact between two cylinders. However, he did not considered residual
macroslip stifness. From the above discussion, it is worthwhile to note that none of
the present models discuss all the inherent property of bolted joint in a single model.
Hence, this work is an attempt towards developing a comprehensive predictive model
for bolted joint. An eight parameter Iwan model has been developed which address
all the inherent properties (microslip, macroslip with residual stifness, pinning) of
bolted joint when it is subjected to sinusoidally varying load.
Fig. 1 a Iwan model, b schematic of bolted joint and c schematic diagram of backbone curve
Iwan Model for Bolted Joint with Residual Macroslip Stiffness and Pinning 313
Figure 1c shows schematic diagram of the backbone curve of bolted joint subjected
to sinusoidally varying load. When joint is under tensile loading, point 1 depicts
initiation of microslip (partial slip), and φmic is the displacement at this point; point
2 shows the initiation of macroslip (full slip) in joint, and φmac is the displacement at
this point. Region 2–3 is the region of macroslip with residual stifness (k∞ ). At point
3, contact of bolt shank with bolt hole initiates, and φ3 is the displacement at this
point. The phenomena of engagement of bolt shank with through hole are known as
pinning. In this work, all the above-stated phenomena are described with analytical
model.
Iwan model was first developed to understand hysteretic behaviour of a material [2]. It
consists of number of Jenkins element, arranged in parallel. Jenkins element consists
of spring and slider connected in series. All Jenkins elements have same stifness (k)
while different yielding displacement (φi ). Iwan model is shown in Fig. 1a. Once
force is applied on bolted joint, some of Jenkins element will yield while other
remains unyielded. So, total force exerted by elements is sum of force exerted by
those elements which have yielded and force exerted by unyielded element. Total
force can be written as [3],
u(t) ∞
F(t) = φρ(φ)dφ + u(t) ρ(φ)dφ, (1)
0 u(t)
where u(t) is relative displacement of joint, and ρ(φ) is the distribution function of
element having yield displacement φ. Differentiating Eq.1 twice with respect to u,
we get,
d2 F(u)
= −ρ(φ). (2)
du 2
The above equation is valid for u = φ. Iwan paper [8] mentioned that distribution
function ρ(φ) can be selected as any random function, or it can be selected based on
experimental data using Eq. (2). The equation depicts the relation between curvature
of stress–strain diagram and distribution function. Li and Hao [5] added additional
Dirac delta function to include the effect of residual macroslip stifness in bolted
joint. Distribution function of Segalman [3] and that of Li and Hao [5] can be shown
in Fig. 2. To develop the distribution function, Segalman [3] used the experimental
power-law relationship. Theoretically, under small amplitude sinusoidally varying
load, energy dissipation varies as F03 (F0 is the amplitude of applied force) [6].
314 P. Ranjan and A. K. Pandey
Fig. 2 a Distribution function for six parameter Iwan model [5], b distribution function for four-
parameter Iwan model [3]
Pinning force initiates in bolted joint once shank of bolt comes in contact with
the inner surface of through holes of bolted joint. This is modelled by Brake [4]
using Hertzian contact theory for cylindrical contact. If it is assumed that plastic
deformation of bolt or shank did not occur, pinning force (F p ) can be modelled using
elastic contact of two cylinders [10].
π
Fp = E ef L A (4)
4
where L is the height of hole, and A is the interference of two surface, A = ϕ − δ p .
E ef is the effective modulus of bolt and shank hole.
−1
1 − μ21 1 − μ22
E ef = + (5)
E1 E2
Iwan Model for Bolted Joint with Residual Macroslip Stiffness and Pinning 315
where μ j and E j are poission ratio and elastic modulus of bolt and shank. In Eqn.(4),
E ef and L are material properties. Thus,
F p = k p A, (6)
where k p = π4 E ef L.
The phenomena taking place in bolted joint can be divided into relative dis-
placement, u,-based phenomenon such as microslip and macroslip and global
displacement-based phenomena such as pinning [4]. Microslip and macroslip always
depend on the relative position of joint with respect to its previous static position (δ0 ).
At the start of loading, δ0 = 0, and once loading is reversed, δ0 = ϕmax . Similarly, δ0
will keep on changing once loading is reversed. While pinning of bolt shank takes
place when global displacement of joint exceeds δ p , i.e. ϕ ≥ δ p for positive loading
(tension) and ϕ ≤ −δ p for negative loading (compression). Relative displacement u
is considered to be positive for forward direction while −u for backward direction.
Thus, global displacement of joint can be related to relative displacement as
ϕ = δ0 ± u. (7)
Finally, backbone curve can be obtained by substituting Eq. 3 into Eq. 1. Its expression
can be wriiten over different domain as [5], for 0 ≤ u < φmic ,
α+1 α+1
Ru(φmac − φmic )
F(u) = + k2 u + k∞ u, (9)
α+1
for u ≥ φmac ,
α+2 α+2
R(φmac − φmic )
F(u) = + k2 φmac + k∞ u. (11)
α+2
Based on discussion in preceeding section, it is clear that pinning force engages when
u ≥ δp.
316 P. Ranjan and A. K. Pandey
To obtain the response to cyclic loading of a material with given backbone curve
can be obtained using Masing rule [12] which can be explained as follow:
Once initial loading curve (backbone curve) is known, reloading curve can be
obtained by scaling the backbone curve twice in both force and displacement direc-
tion and translating it to terminate it at the tip of backbone curve. The unloading
curve can be obtained by reflecting reloading curve along each of the force and
displacement direction.
Masing rule is only valid for cyclic loading with same amplitude. However, this
rule was further modified by Jayakumar [11] to include monotonic loading. Based
on the above rule, equation for reloading and unloading curve can be obtained as,
±ϕ ∓ δ0
F± (u) = ∓Fs ± 2F , (12)
2
where F+ (u) indicates force exerted by Iwan model while reloading, and F− (u) indi-
cates force exerted by Iwan model while unloading. Fs = F(u max ) is the maximum
force exerted by Jenkins element. Finally, for ϕ ≥ δ p , pinning force can be included
in the above equation as,
Fsliding = F p + F± (u). (13)
Fmac (α + 1) α + 1 Fmac (1 + β)
β= α+2
− , kt = α+1
. (15)
Rφmac α+2 φmac β + α+2
3 Conclusion
Acknowledgements First author would like to acknowledge the fellowship provided by Depart-
ment of Higher Education, Ministry of Human Resource Development, Government of India. The
work is also partially supported by AR&DB, India, with project number 1824.
References
1. Groper M (1985) Microslip and macroslip in bolted joints. Exp Mech 25(2):171–174
2. Iwan WD (1966) A distributed-element model for hysteresis and its steady-state dynamic
response. J Appl Mech 33(4):893–900
3. Segalman DJ (2005) A four-parameter Iwan model for lap-type joints. J Appl Mech 72(5):752–
760
318 P. Ranjan and A. K. Pandey
4. Brake M (2017) A reduced Iwan model that includes pinning for bolted joint mechanics.
Nonlinear Dyn 87(2):1335–1349
5. Li Y, Hao Z (2016) A six-parameter Iwan model and its application. Mech Syst Sig Process
68(2):354–365
6. Song Y, Hartwigsen C, McFarland D, Vakakis AF, Bergman L (2004) Simulation of dynamics
of beam structures with bolted joints using adjusted Iwan beam elements. J Sound Vibr 273(1–
2):249–276
7. Gaul L, Lenz J (1997) Nonlinear dynamics of structures assembled by bolted joints. Acta Mech
125(1–4):169–181
8. Iwan WD (1967) On a class of models for the yielding behavior of continuous and composite
systems. ASME
9. Ungar EE (1964) Energy dissipation at structural joints; mechanisms and magnitudes. Bolt
Beranek and Newman Inc, Cambridge MA
10. Johnson KL (1987) Contact mechanics. Cambridge University Press, Cambridge
11. Jayakumar P (1987) Modeling and identification in structural dynamics. California Institute of
Technology
12. Masing G (1926) Self-stretching and hardening for brass. In: Proceeding of 2nd international
congress on applied mechanics, pp 332–335
Kinematics Model of Bionic Manipulator
by Using Elliptic Integral Approach
Abstract This paper suggests a variable curvature kinematics model for the bionic
continuum robots (BCRs) using elliptic integral approach. For the BCR, the manip-
ulator is assumed to be a flexible beam, in which a large deflection takes place due
to external loadings. A six-DoF elephant trunk-like BCR known as Robotino-XT is
considered here for study. The external load is assumed to be an equivalent moment
calculated from the pneumatic forces applied on the bellow tube-like actuators. Opti-
track motion-sensing vision system with Motive 2.0 interface is used to acquire the
deflection and orientation of the tip of the trunk. The analytical results are validated
with experimental data.
1 Introduction
Soft continuum robotics deals with flexible robots with high degrees of freedoms
(DoFs). This particular class of robots is inspired by biological creatures such as
snakes [1], elephant trunks [2] and so forth. These robots accomplish their move-
ment by the principle of high deformation (also called compliant mechanism) of their
members. No assembly, less wear and more workspace are some of the significant
advantages of the compliant mechanisms over rigid systems. Under the same force,
compliant mechanisms provide more significant deflection than their rigid coun-
terparts, thus making the bionic manipulators more dexterous. However, the high
dexterity of these manipulators is achieved at the cost of more material and system
nonlinearities.
To date, forward kinematics developed for BCRs are mostly based on piecewise
constant curvature assumption [3]. This procedure describes the state of the distorted
backbone of the BCRs with a constant curvature curve. However, models developed
with this approach do not correlate well with the biological counterparts because of
the assumed constant circular section of the backbone curve. On the contrary, a vari-
able curvature approach for multi-sectional BCRs has been proposed to achieve the
exact shape of the backbone [4]. The curvature models work with two sub-mappings,
i.e. explicit mapping and typical mapping. There are several other benchmark models
developed to determine the forward kinematics (FK) of a BCR. Cosserat continuum
theory model [5] and modal method [6] are the other methods used to establish
forward kinematics of hyper-redundant manipulators. Although these models are
accurate, they are not suitable for real-time control applications because of their high
computational complexity.
BCRs accomplish their movement by the principle of high deformation of their
members. So, these robots can be assumed as cantilever beams with large deflection
under loading. Large deflection of cantilever beams is solved by minimising the
integral of the residual error of the governing differential equation of the beam [7].
The very same problem can also be solved by using the elliptic integral approach
[8, 9].
In this paper, a novel method is proposed for the FK of the BCRs in Sect. 2.
Section 3 describes the experimental set-up for Robotino-XT. The implementation
and validation of the analytical results with experimental observations are given in
Sect. 4. Finally, the concluding remarks are presented in Sect. 5.
2 System Configuration
The trunk of Robotino-XT consists of two segments, where each segment includes
three individual pressure-actuated bellow tubes, as shown in Fig. 1a. The bellow tubes
are equidistant from the centre of the manipulator with actuator spacing of 120°. The
4.12° tapered conical-shaped bellow tubes are made of polyamide material with a
laser sintering process. The resultant differential pressure force in bellows creates
a bending moment, which bends the segment about x-y axes, but there is no twist
about the z-axis. Also, an inextensible wire passing through the centre of the assembly
constrains the mean length, i.e. the neutral axis of the equivalent beam has a constant
length.
The resultant bending moment at the head of the section is found by taking a
moment of all the forces in Fig. 1b about the centre and obtained as
Kinematics Model of Bionic Manipulator … 321
F11
F12 F13
Head o d1 y
d1 d1
60o
Flexible trunk
(a) (b)
Fig. 1 a Robotino-XT projected view, b forces at the head of the first segment
√
3d1 Fi2 + Fi3
Me = (Fi3 − Fi2 ) î + Fi1 − d1 ĵ (1)
2 2
where the first subscript to forces (F) or pressure (P) denotes the segment number
and the second denotes the tube number.
The current study is focused only on the first section of the trunk. Moreover, it
is assumed that the trunk of the Robotino-XT is rotated only about the y-axis. As
such, the study is strictly directed towards planar motions of the first trunk segment
within the z-x plane. It is also to be noted that such planar motion is conceivable only
if the same actuation pressures are applied in tube-2 and tube-3 simultaneously, as
shown in Fig. 1b. Furthermore, for the sake of convenience, the trunk weight is also
neglected in mathematical modelling.
The trunk of the Robotino-XT is considered to be a flexible cantilever beam
(neutral axis is assumed to inextensible) having a variable cross section, length L
and modulus of elasticity E. The beam is subjected to an equivalent bending moment
M e and a point load W inclined an angle β from the z-axis as shown in Fig. 2. A
new variable s is introduced for the curvilinear coordinate along the deflected axis
of the beam. The exclusion of self-weight leads the external end-effector load W to
be independent of arc length s. By applying the usual Euler–Bernoulli beam theory
assumptions, the equilibrium equation for the moment can be written as
dθ (s)
EI(s) = M(z,x) = Wx (z 0 − z) + Wz (x0 − x) + Me (2)
ds
322 M. K. Mishra et al.
z0
z
Wx W
FBD
x0
β
s
L
Wz
Me
dz
Wz
θ Wx dx
M(s) ds
Wx Wz
x dM ( s )
M (s) +
ds
where (z0 , x 0 ) is the location of the tip of the deflected beam, Wz , Wx are the respective
z and x components of the load W, EI(s) is the variable flexural rigidity and θ(s) is
the slope of the beam.
Differentiating (2), the force balance equation can be written as
d2 θ (s)
E I (s) = −Wz sin(θ (s)) − Wx cos(θ (s)) (3)
ds 2
where dx
ds
= sin(θ (s)), dz
ds
= cos(θ (s)) (from FBD of Fig. 1a).
This cantilever beam is associated with two boundary conditions: θ(s) at fixed
end is 0 and slope dθ(s)
ds
Me
at free end is EI(L) .
2
The equivalent moment is (Me ) = π r1 − L 1 tan φ2 P11 d1 and the variable
moment of inertia is I (s) = I0 α11 (1 − k11 s)4 − α12 (1 − k12 s)4 where I 0 is the
R14 r14
moment of inertia at the fixed end and α11 = R14 −r14
, α12 = R14 −r14
, k11 = tanφ
R1
, and
k12 = tanφ
r1
. So, (3) can be written in elliptic integral form as
ds E dx1
√ =ς . (4)
I (s) W 1 − x12 1 − ς 2 x12
Integrating (4) and applying the boundary conditions, one obtains [9]
Ws
x1 = sn + F(sin(β/2); ς ); ς (5)
EIς
Kinematics Model of Bionic Manipulator … 323
where sn[u; m] is a Jacobian elliptic function (refer [8] for more details). The slope
of the beam in terms of non-dimensional parameters can be written as
ωs
−1
θ (s) = 2sin sn + F(sin(β/2); ς ); ς −β (6)
ς
2
where s = Ls , ω2 = WEIL are non-dimensional parameters. Inserting (6) into dx
ds
=
sin(θ (s)), dz
ds
= cos(θ (s)), and integrating, the tip position can be found.
3 Experimental Set-Up
Figure 3a depicts the experimental set-up which comprises a flexible continuum trunk
(manipulator) and an external observer (Optitrack motion sensor). In this experiment,
the motion tracking system utilises Motive 2.0 software, four Prime 13 cameras and
nine retro-reflective markers to capture the motion of an object. Prime 13 cameras
are equipped with an optical lens of 8 mm F 1.8 and a stock lens of 5.5 mm F 1.8,
which has high-resolution image sensing (1280 × 1024) capacity. These cameras
track any surfaces covered with retro-reflective material, as shown in Fig. 3b. This
material is intended to reflect the approaching light to its source origin. Infrared light
radiated from the camera is reflected by the markers and identified by the camera’s
sensor. The captured reflections are utilised to ascertain the 2D marker positions,
which are utilised by Motive 2.0 to estimate the corresponding 3D positions through
reconstruction. In the Motive 2.0 interface, three markers at each section are selected
together as a rigid body (as shown by the red dashed triangle in Fig. 3b).
Section-1 Section-2
Flexible Trunk Mobile Base Bellow Tubes
(a) (b)
Fig. 3 a Robotino-XT test bench with vision system, b Robotino-XT with retro-reflective markers
324 M. K. Mishra et al.
80 40
numerical 1.7 bar numerical
x-coordinate (in mm)
60 experimental 30 experimental
θy (in deg)
40 20
P11
20 10
0 bar
0 0
0 20 40 60 80 100 120 140 0 0.5 1 1.5 2
z-coordinate (in mm) Pressure(P11) (in bar)
(a) (b)
5 Conclusions
A general closed-form solution for orientation and deflection of the cantilever beam
is obtained by using the elliptic integral approach, which satisfies the boundary condi-
tions. Experimental results are obtained from Robotino-XT by using the Optitrack
motion sensor. Finally, the developed analytical method is validated with the exper-
imental results obtained from the flexible continuum trunk Robotino-XT. A good
match is found between the numerical and experimental results for both tip deflec-
tion and orientation of the trunk. Further study including the weight of the trunk as
well as the spatial movement of the trunk, i.e. inertial forces, will be taken up as
future research work.
Acknowledgements This work was supported by the joint collaborative research project between
DST (India), Indo French Centre for the Promotion of Advanced Research (IFCPAR) and CNRS
in France (PRC CNRS-DST 2016-2018).
References
1. Khan R, Watanabe M, Shafie AA (2010) Kinematics model of snake robot considering snake
scale. Am J Appl Sci 7(5):669–674
2. Robotino XT. https://www.festo.com/group/en/cms/10239.htm
3. Mishra MK, Samantaray AK, Chakraborty G, Jain A, Pathak PM, Merzouki R (2019) Dynamic
modelling of an elephant trunk like flexible bionic manipulator. In: Proceedings of the 2019
IMECE, 4. ASME, Salt Lake City, pp 1–9
4. Mahl T, Hildebrandt A, Sawodny O (2014) A variable curvature continuum kinematics for
kinematic control of the bionic handling assistant. IEEE Trans Rob 30(4):935–949
5. Chirikjian GS (2015) Conformational modeling of continuum structures in robotics and
structural biology: a review. Adv Robot 29(13):817–829
6. Chirikjian GS, Burdick JW (1994) A modal approach to hyper-redundant manipulator kine-
matics. IEEE Trans Robot Autom 10(3):343–354
7. Dado M, Al-Sadder S (2005) A new technique for large deflection analysis of non-prismatic
cantilever beams. Mech Res Commun 32(2005):692–703
8. Abu-Alshaikh I, Alkhaldi HS, Beithou N (2018) Large deflection of prismatic cantilever beam
exposed to combination of end inclined force and tip moment. Mod Appl Sci 12(1):98–110
9. Kimball C, Tsai LW (2002) Modelling of flexural beams subjected to arbitrary end loads. J Mech
Des 124:223–235
A Review on the Effect of Biomechanical
Aspects and the Type of Stability Fixation
on the Bone Fracture Healing Process
Abstract Bone fracture healing is a complex process that involves both biological
and mechanical aspects. There are various factors that influence the time period of
fracture healing. Over the past decade, many researchers reported the different factors
on which the bone fracture healing process depends. These researches also help the
orthopedic surgeon to bring the upgradations in their surgical techniques. The type
of stability fixation given to the fracture during surgery, the flow of body fluid at
the fracture site and direction of interfragmentary motion, i.e., micromotion during
the load-bearing period, are some important factors that are discussed by the various
researchers in the past. The aim of this review is to highlight the existing knowledge
about the biology of bone fracture repair, the recent research on bone fracture and
the effect of biological and mechanical aspects on the bone fracture healing process.
1 Introduction
The bone fracture healing process depends on which type of stability is given to the
fracture, i.e., absolute and relative stability. From the past studies, it is found that
too much motion or too little motion between the fracture gap leads to instability
and an inadequate amount of callus formation which delays the healing process. A
new concept of stability is also developed which is known as absolutive stability.
As the name suggests, absolutive stability is the combination of absolute as well as
relative stability. The absolute stability is below the fixation plate which is fixed, i.e.,
no callus formation and relative stability on the other side of the plate, i.e., callus
formation. From the past studies, it is found that different types of implants can affect
the range of micromotion, strain and stress at the fracture region. So, fracture healing
depends on the type of stability as well as the choice of implant. Also, the supply
of blood and contact of soft tissues at the fracture region are very critical biological
conditions for the fracture healing process.
It is quite understood that strain between the fracture fragments is the most appro-
priate way of predicting the fracture healing [1]. After predicting and understanding
the fracture healing, it becomes easy for the surgeon to improve the treatment. The
condition of absolute stability is achieved when the fracture strain is less than 2%
which involves primary bone healing in which no fracture callus is formed and for
relative stability, the fracture strain is between 2 and 10% which involves secondary
bone healing in which the bony callus formation occurs [2]. The callus is a carti-
laginous and bony material formed at the fracture site which works as a connecting
bridge across the fracture during the healing period. Under the physiological load, in
the absence of micromotion due to interfragmentary compression, absolute stability
involves surgical exposure of the fracture fragments by dissecting the tissues whereas
relative stability leads to the callus formation. It involves the correction of length,
rotation and alignment, i.e., functional reduction [2]. It is quite difficult to predict
the exact amount of interfragmentary movement between the fracture gaps which
is ideal for the adequate callus formation, proper bone union and early or normal
healing of bone. In one of the studies, in a fracture gap of 2 mm interfragmentary
movement of 0.2–1 mm is observed to be ideal for optimal healing. The direction of
the interfragmentary movement also affects the healing time of bone fracture. One
of the studies highlighted that during the weight-bearing period of fracture healing,
the cyclical micro-movement delays fracture healing.
In addition to the stability provided to the fracture, the type of implants also plays
a vital role in the fracture healing process. It is also found that the implants like less
invasive stabilization system (LISS), dynamic condylar screw (DCS) and condylar
buttress plate (CBP) have different values of total motion, strain and stress for the
same loading [3]. Also, implant choice depends on the type of fracture. For the
comminuted fracture in which the bone is broken down in more than two fragments,
generally intramedullary nails are preferred over the fixation plate with screws and
this type of fracture also heals with callus formation because the condition of relative
stability is achieved. Again some variables like the supply of blood and soft tissue
contacts at the fracture region also affect the fracture healing atmosphere. Proper
supply of blood is necessary which improves the healing in a great manner [4]. The
aim of this review is to highlight the importance of various biological and mechanical
factors affecting the fracture healing progress.
The supply of blood and contacts of tissue are necessary conditions for the healing
of fractured bone. But in the early stage, the flow of fluid gives the minimum contri-
butions in the healing process. It is observed that the strain at the fracture site is the
influential stimulus in the early stage. The bone formation may be intramembranous
(direct conversion of tissue into bone) or endochondral (formation of bone through
A Review on the Effect of Biomechanical … 329
the cartilage formation in between) [5]. Again the type of bone formation depends
on the magnitude of strain and stress at the fracture site.
At the early stage of fracture, the hematoma is formed at the fracture junction
which provides a supporting role in the differentiation of mesenchymal stem cells
(MSCs) into the bone, fibrous tissues and cartilage. During the inflammatory reaction,
several biological as well as growth factors are released. These factors in addition to
the mechanical stimulus such as strain and hydrostatic pressure control the activities
of mesenchymal stem cells (MSCs). MSCs are considered to be the most important
members of the bone formation process [6]. The bone formation process follows
certain stages. The four stages in the healing of bone fracture are shown in Fig. 1.
As the healing starts, the formation of soft callus (cartilaginous callus) takes place
which fills the gap between the bony fragments. This soft callus provides temporary
but stable support to the fractured bone. In the later stage, this cartilaginous soft
callus undergoes calcification and revascularization which transformed it into the
hard bony callus. After remodeling, a compact normal bone is reconstructed. The
revascularization is observed to be a very crucial process for successful bone repair
[7].
The finite element analysis on the bone-implant model is done by various authors by
assigning the different material properties to a different portion of the bone [8]. The
femur bone has various portions like head, shaft, trochanter, bone marrow and neck
as shown in Fig. 2. These different portions of bone are having different material
properties. The material properties for the different portions of the bone used by the
researchers are shown in Table 1.
330 S. Rathor and R. Uddanwadiker
Apart from Young’s modulus and Poisson’s ratio, the shear modulus, compressive
strength, tensile strength, torsional strength and elastic behavior of bone are the
important properties [9].
The healing of bone after some weeks is observed by the change in Young’s
modulus and shear modulus of bone material [10]. The compact bone callus at the
end of the healing process has a greater magnitude of Young’s modulus and shear
modulus than the initial bony callus. This is considered to be an indication of the
recovery of the patient.
Pre-assessment of fracture is very important for the surgeon to take the decisions
regarding surgery process. A bone fracture may be simple, multi-fragmentary and
complex. If the surrounding tissues and cartilage at the fracture region are not
damaged, it is called a simple fracture. In this case, there is a proper supply of
blood and this type of fracture is generally treated with anatomical reduction incor-
porating compression plate and screw which provides rigid fixation and fracture heals
with primary or direct healing, i.e., without callus formation. In the case of complex
fracture, the damage of tissues and cartilage also occurs. This type of fracture needs
anatomical reduction, and generally intramedullary nail and bridge plating ar used
to hold the broken pieces of bone together. It provides flexible fixation and fracture
heals with secondary or indirect healing, i.e., with callus formation. Also, in the case
of multi-fragmentary fracture, flexible fixation inhibits the callus formation.
The term reduction refers to bringing back the disturbed morphology of the bone to its
normal position. This is done by using various available implants like plates/screws,
intramedullary nails, external fixation, etc. The reduction may be open or closed
depending on the type of fracture. Open reduction involves surgical exposure of
fracture fragments by cutting the skin open and dissecting the various layers of tissues,
whereas closed reduction does not require surgical exposure. Closed reduction is
applied when the fragments of bone did not penetrate into the skin. The articulate
and complex fracture needs an anatomical reduction. It is quite understandable that
the minimum surgical exposure is beneficial for the early healing as the supply of
blood to the bone is preserved and it can be re-established quickly [11].
332 S. Rathor and R. Uddanwadiker
The method of fixation plays a vital role in bone fracture healing. Fixation provides
stability to the fractured bone. From the past studies, it is observed that a small amount
of motion, i.e., strain between the fracture fragments, speeds up the bone healing
because this strain stimulates the formation of bony callus at the fracture site [4].
This strain or deformation is also called micromotion between the fracture fragments.
An optimal range of micromotion to be required for the early healing of fracture is
given by various researchers, but it is also observed that too little motion and too
much motion can cause a delay in healing and can also result in a nonunion. Thus,
the formation of callus and hence the healing period depend on the type of stability
provided during surgery, i.e., absolute and relative stability. Absolute stability is
achieved by using a compression plate, lag screw, etc. Intramedullary nailing, bridge
plating, external and internal fixators, etc., are generally used to achieve relative
stability. The absence of micromotion in absolute stability results in less formation
of callus, whereas in relative stability, fracture heals with callus formation because
of micromotion between the fracture fragments.
To achieve absolute stability, there should not be any gap between the fracture
and this is achieved anatomically by applying compression using plate and screws.
This is the basic requirement to achieve primary bone healing, but in many cases
there is an unintentional gap that remains and this affects the stability of the bone–
plate construct in a great manner. In one of the case studies, the effect of different
amounts of fracture gap on the stability of bone–plate construct is analyzed by using
the finite element method. The fracture gap of 1, 2, 3, 4 mm and bone defects of 25,
50, 75, 100% were created with CAD software. The sketch of the models is shown in
Fig. 3. All the bone–plate model was simulated for bending, compression and torsion
loading. It was concluded that there is an exponential decrease in the stiffness for
a minimum gap of 1 mm that remains after plating with no contact between the
fracture fragments. It was also observed that the fracture surface with >50% defects
must be in contact to avoid too much stress concentration on the plate [12] and it was
recommended to take additional precautions while treating this type of case for the
protection of the implant plate.
The interfragmentary motion, i.e., deformation, strain and stress at the fracture site,
is the most important parameter which shows how good an implant is for a particular
type of fracture. The various finite element analyses are performed by the keen
researchers on different types of implants like dynamic condylar screw (DCS),
dynamic hip screw (DHS), proximal femur nail (PFN), less invasive stabilization
system (LISS) and condylar buttress plate (CBP) with fractured bone, and it is found
that each implant is having a different value of deformation, stress and strain [3,
A Review on the Effect of Biomechanical … 333
Fig. 3 a Fully compressed model (0 mm/0%) and b–i other models with gap
8]. The clinical observations also show that each type of implant has a different
impact on the fracture healing process. The compression plate and screw provide the
condition of absolute stability if the plate is rigidly fixed with the fractured bone. If
the same implant is used with flexible fixation, the condition of relative stability is
achieved.
For the comminuted fracture, generally, the intramedullary nail is used which is
passed through the longitudinal axis of the bone without any injury to the blood
supply, and hence, the initial hematoma formation at the fracture region is left undis-
turbed. For the same case, if the compression plate and screw are used, it will cause
more injury in the blood supply region and also to the surrounding tissues which are
undesirable. One of the case studies reported that, for the diaphyseal tibial fracture,
the plate fixation does not remain the choice of the orthopedic surgeon because of the
aforementioned complications and intramedullary nailing is preferred for that [13].
5 Discussion
This review offers an explanation of the types of stability and understanding of their
functional outcomes. It is also reported in most of the studies that there should be
minimum and maximum limits for the strain and stress at the fracture gap. The various
factors which affect the mechanical and biological environment at the fracture site are
also discussed. The unwanted effects of the unintentional gap that remain between
the fracture surfaces while achieving primary bone healing are highlighted. The rela-
tive stability can be achieved by both compression plate/screw and intramedullary
nailing. One demands more surgical exposure than the second one. Both the fixation
heals with indirect healing, i.e., with callus formation. It is quite easy to under-
stand the mechanism of fracture healing by considering only one or two affecting
factors at a time, but if all the factors are taken into the consideration, it is very
difficult to understand the process which makes bone fracture healing process a
very complex biomechanical phenomenon. Also, starting from the differentiation of
mesenchymal stem cells into cartilage, fibrous tissues and bony tissues followed by
different stages of callus formation, the bone fracture healing process is considered
as a tissue regeneration process.
6 Conclusion
References
1. Perren SM (2015) Understanding fracture healing biomechanics based on the “strain” concept
and its clinical applications. Acta Chir Orthop Traum Čech 82(2015):253–260
A Review on the Effect of Biomechanical … 335
2. Kojima KE, Pires RES (2017) Absolute and relative stability for fracture fixation. Injury Int J
Care Injured 48S4(S1)
3. Duffy P, Barron L, Dunbar MJ (2006) Assessment of fracture micromotion in distal femur
fracture fixation with RSA. Clin Orthopedics Related Res 448:105–113
4. Norris BL, Russell TA, Rothberg DL (2018) Absolute versus relative fracture fixation: Impact
on fracture healing. J Orthop Trauma 32(3):S12–S16
5. Claes LE, Heigele CA (1999) Magnitudes of local stress and strain along bony surfaces predict
the course and type of fracture healing. J Biomech 32:255–266
6. Ghiasi MS, Chen J, Vaziri A, Rodriguez EK, Nazarian A (2017) Bone fracture healing in
mechanobiology modeling—a review of principles and methods. Bone Rep 6:87–100
7. Marsell R, Einhorn TA (2011) The biology of fracture healing. Injury Int J Care Injured 42:551–
555
8. Sowmianarayanan S, Chandrasekaran A, Krishna Kumar R (2008) Finite element analysis
of a subtrochanteric fractured femur with dynamic hip screw, dynamic condylar screw, and
proximal femur nail implants–a comparative study. J Eng Med Proc IMechE 222(Part H)
9. Wirtz DC, Schiffers N (2000) Critical evaluation of known bone material properties to realize
anisotropic FE-simulation of the proximal femur. J Biomech 33:1325–1330
10. Li J, Zhao X, Xiaojie Hu, Tao C (2018) A theoretical analysis and finite element simulation of
fixator–bone system stiffness on healing progression. J Appl Biomater Funct Mater 16(3):115–
125
11. Perren SM (2002) Evolution of the internal fixation of long bone fractures. J Bone Joint Surgery
84-B(8)
12. Jong-Keon Oh et al (2010) Effect of fracture gap on stability of compression plate fixation: a
finite element study. J Orthopedic Res 28:462–467
13. Sandoval JMC, Rueda JLO, Daza AS (2017) Management of diaphyseal tibial fractures by
plate fixation with absolute or relative stability: a retrospective study of 45 patients. Trauma
Surg Acute Open Care 2:1–5
14. Epari D, Taylor W (2006) Mechanical condition in the initial phase of bone healing. Clin
Biomech 21(2006):646–655
15. Noordeen MHH, Lavy CBD, Shergill NS (1995) Cyclical micromovement and fracture healing.
J Bone Joint Surg 77-B(4)
Mechanical Behaviour of Special Type
Seals Used in the FBR Application
Abstract Leak tightness of the radioactive pressure boundary is one of the important
safety requirements of the nuclear reactors. Dynamic seals present in the various
mechanisms of fast breeder reactors (FBR) play an important role in achieving the
required leak tightness. Dynamic seals allow high relative velocity and offer low
frictional resistance at a moderate temperature between the mating surfaces of these
mechanisms. An in-house designed and developed labyrinth type V-ring seals are
used as dynamic seals in various mechanisms of Indian FBRs. Design of V-ring seal
is frozen based on engineering judgement. Detailed stress analysis of these seals is
the main focus of the present paper. The stress analysis revealed that the design intent
of the seals is only partially met, with the scope for design improvements. Studies
are extended to optimise the seal geometry for enhanced performance.
1 Introduction
One of the key components in various mechanisms of reactor assembly of fast breeder
reactors are dynamic seals, which allow high relative velocity between the mating
surfaces, low leak rate and offer low frictional resistance at a moderate temperature.
Various seals such as O-ring seals, oil seals, inflatable seals and V-ring seals are
tested in dedicated test set-ups to qualify them for reactor operation. Based on the
results, a custom-designed labyrinth type V-ring seal is found meeting the perfor-
mance requirements [1]. This special type seal is used between the stationary tube
and translation tube of control and safety rod drive mechanism and diverse safety
rod drive mechanism to prevent the leakage of reactor cover gas (argon) into the
reactor containment building (RCB) through the annular space. This paper discusses
the stress analysis carried out towards confirming the design of the labyrinth type
V-ring seal for the FBR application. The selected V-ring seal configuration has been
further optimised using the finite element analysis.
2 Seal Description
The photograph of labyrinth type V-ring seal is shown in Fig. 1a and a and schematic
diagram is shown in Fig. 1b. Construction features of V-ring seal include the body,
the flap and the hinge. The labyrinths, which are on the flap, are pressed against the
chrome-plated central tube, producing the necessary sealing force. Labyrinths on the
flap decrease the leakage further by offering higher resistance to leakage flow. The
present study is carried out with silicone rubber as seal material. The angle subtended
by the flap with vertical is known as flap angle (F). The difference between the outer
diameter of the central tube and inner diameter of the seal is defined as compression
depth (CD). The inner diameter of the seal is measured at the top labyrinth position. In
the present geometry, the outer diameter of the central tube is 110 mm, and the inner
diameter is 106 mm, which implies a CD of 4 mm. The essential design requirements
of the seal are that (i) it has to develop sufficient contact pressure with the central tube
for minimum leakage, (ii) the induced stresses in the seal shall be minimum and (iii)
the reaction forces on the central tube shall be minimum. The present configuration
and essential geometrical details of the labyrinth type V-ring seal are given in Fig. 2.
Fig. 1 a Photograph of labyrinth type V-ring seal, b schematic diagram of V-ring seal
Mechanical Behaviour of Special Type Seals … 339
3 Numerical Analysis
Numerical analyses are carried out in two stages. The first stage is established to
develop an appropriate computational model for the V-ring seals. In this analysis, an
appropriate material model for predicting the mechanical behaviour of the silicone
rubber has been identified [2–7]. Accordingly, the hyperelastic material is adopted for
the V-ring seals. An experiment also carried out to validate the key results obtained
during the first stage of analysis. The contact pressure on the labyrinths, peak von
Mises stresses developed in the seal and the reaction force on the central tube is
obtained from this analysis.
The second stage is focused on optimising the seal geometry to enhance its perfor-
mance and design margins. The seal design is optimised by considering different
parameters such as flap angle, CD, number of labyrinths and flap thickness to
minimise the peak stress induced in the seal and reaction force on the central tube
with optimum contact pressure on the joint.
The present geometry of V-ring seal along with central tube is modelled in commer-
cially available FEA software [8]. Based on the literature survey, uniaxial tensile
test data of silicone rubber is taken [9] and stability of data with available hyper-
elastic material models is evaluated using evaluation tool inbuilt in FEA software.
The central tube is modelled as the rigid axisymmetric element while the V-ring
seal is modelled as the deformable axisymmetric element. Surface to surface contact
logic is employed to establish the interaction between them. The rigid element is
assumed as master surface and the surfaces associated with the labyrinths are taken
as slave surfaces. In interaction properties, tangential behaviour is defined using
penalty based technique with 0.5 value of the coefficient of friction, and normal
340 N. Kaushal et al.
Fig. 3 Distribution of von Mises stress a test configuration, b base configuration and c photograph
of seal after experiment
behaviour is defined using the exponential pressure overclosure scheme. Global seed
size is taken as 0.0016 m while local seed size in the vicinity of labyrinths is taken
as 0.0001 m to get converged results. Validation of the numerical model performed
with the benchmark data is obtained from the literature [10].
To gain more insights into the characteristic behaviour of V-ring seals, another
configuration (called test configuration) of V-ring seal is also studied in addition
to base configuration of the existing design. In the test configuration analysis, it is
performed without considering the labyrinths on the flap.
Results and Discussion
Figure 3a, b shows the von Mises stress distribution for test configuration and base
configuration, respectively. Close examination of the results for test configuration
reveals that the hinge portion of the seal got compressed and is bent like cantilever
beam in the downward direction. Peak stress is near the joint between the hinge and
the body. The top portion of the flap is only in contact with the central tube.
Similarly, the close examination of the results for base configuration reveals that
only top labyrinth among the five labyrinths is in contact with mobile assembly.
The peak stress is located at the labyrinth tip. The peak values of von Mises stress,
reaction force and contact pressure for both test and base configurations are tabulated
in Table 1. It can be seen that with the introduction of labyrinths on the flap, the contact
pressure reduced to half, whereas the reaction force is reduced by one by the tenth.
These findings highlight the superior performance of labyrinth type V-ring seals over
the simple V-ring seals.
Mechanical Behaviour of Special Type Seals … 341
Another important outcome of the stage-1 study is that the top labyrinth only
contacts the central tube, which does not meet the key requirement of providing
multiple labyrinth configuration for enhanced leak tightness.
An experiment has been performed to check the contact between the central tube
and seal in the base configuration. In this experiment, the central tube is coated with
a thin layer of grease and the V-ring seal of the base configuration has pulled from
one side to another. The grease impression has found only at the top portion of the
labyrinth. These results validate the analysis results. The photograph of the seal after
the experiment is shown in Fig. 3c.
As part of optimisation, various studies are carried out to gain more insights into the
effect of various design features and their interplay on the seal performance. These
studies also paved the way for recommending optimised configuration.
The design features considered in the optimisation study are the flap angle varia-
tion, CD, flap thickness and the number of labyrinths. Flap angle for various config-
urations is increased till all labyrinths come in contact with the central tube. CD
values of 2, 4 and 6 mm are taken in these studies. Flap thickness is varied from 4 to
6 mm, and the number of labyrinths is changed as 5, 4 and 3 nos.
Results and Discussion
Variation of peak von Mises stress and reaction force with the flap angle is shown in
Fig. 4a, b, respectively.
The following inferences are drawn from all the studies:
• Increasing the flap angle increases the number of labyrinths contacting the central
tube. The variation of performance characteristics such as peak von Mises stresses
and reaction force with flap angle is not uniform due to variation of seal stiffness
upon the establishment of contact of new labyrinths with the central tube as flap
angle increases.
• Similarly, by varying the compression depth and flap thickness changes, the stiff-
ness of the seal in addition to vary the % of seal compression. These changes
affect seal performance characteristics.
• The variation of seal parameters with flap thickness is found small.
• Reducing the number of labyrinths reduces the contact pressure and reaction force.
342 N. Kaushal et al.
Fig. 4 a Variation of peak von Mises stress with flap angle and b variation of reaction force with
flap angle
Based on these studies, the optimised seal configuration has been finalised. The
design features finalised for the optimised configuration are given in Table 2. The
distribution of von Mises stress and contact pressure for optimised configuration
is shown in Fig. 5. The performance characteristics of optimised configuration are
given in Table 3.
Fig. 5 Distribution of a von Mises stress and b contact pressure for optimised geometry
Mechanical Behaviour of Special Type Seals … 343
Table 3 Performance
Configuration Peak von Reaction force Max. contact
characteristics of optimised
Mises stress (N) pressure
configuration
(MPa) (MPa)
Optimised 0.903 153 1.39
It can be seen that there is a reduction of 65%, 67% and 53% in reaction force,
peak von Mises stress and maximum contact pressure, respectively, for optimised
configuration concerning the base configuration.
4 Conclusion
Detailed numerical study of in-house designed labyrinth type V-ring seal has been
carried out. The numerical analysis provided greater insights to optimise the config-
uration of the seals. Optimisation studies revealed that contact with more number of
labyrinths could be established with higher flap angles. Optimised seal configuration
has been recommended for the future application.
References
1 Introduction
Under sodium ultrasonic scanner and sweep arm scanner employing ultrasonic trans-
ducer are used in fast breeder reactor (FBR) for detecting protrusion or growth of core
subassembly (SA) in sodium at 200 °C. These mechanisms are to be tested with SA
arranged identically as in typical FBR and simulating geometric features of ‘above
core plenum’ (which is the plenum between the control plug and SA) in sodium
[1]. One of the important tests is to demonstrate the capability of the transducer to
detect SA growth, if any, in the ‘above core plenum’ in sodium at 200 °C. In order to
simulate SA growth in sodium vessel, different methods/mechanisms are proposed
and studied thoroughly to arrive at the simplest way to perform the said task. Also,
the mechanism, used for simulating SA growth, before being implemented, has to
be qualified and its reliability has to be ensured by testing it in sodium. This paper
gives details of different methods/mechanisms proposed to simulate the SA growth,
their comparison and sodium test results of finalized mechanism.
Fig. 1 Different concepts proposed to simulate SA growth in sodium vessel (a SA head of varying
height, b hanging from the top and c upside-down SA with hanging concept)
Realization of a Simple Mechanism to Simulate … 347
Small
Nut Oper- Flang Top
ating Tool Flange
Tie Rod
Sodium
Level
SA Head
Screw
Guide
Plate
Nut
Base
Plate
Fig. 2 Schematic of LTM in disengaged, engaged, SA raised condition and a photograph of actual
LTM
introduced from the top of the sodium vessel is used to rotate the nut to adjust the
height of the SA head (see Fig. 2). Challenges associated with this method are (i)
selection of material combination to be used for screw and nut to work well in sodium
and (ii) lack of experience of screw-nut operation in sodium. It is proposed to test
screw-nut assembly with a nut operating tool in sodium before being implemented
in actual operation in the sodium vessel.
The test set-up, i.e. lifting tool mechanism (LTM), consists of SA heads, trapezoidal
screw-nut assemblies, nut operating tools and a base plate hanging from the top of test
pot of sodium loop through tie rods. SA head sits on the nut of screw-nut assembly
while the screw is fixed to the base plate. SA heads are guided to prevent the rotation
of it. Nut operating tools are introduced from the top and positioned above each SA
heads. Nut operating tool has two DOFs, viz., raising/lowering and rotation. The nut
operating tool is lowered manually to get engaged with the nut and is then rotated,
resulting in the nut also to rotate. Since the screw is fixed to the base plate, the
nut starts moving up or down based on the direction of rotation. After attaining the
required height, the nut operating tool may be retracted and SA heads remain at that
348 S. Patri et al.
The LTM testing involves the performance testing in air, argon and sodium then
endurance testing in sodium at 200 °C followed by testing in argon again. Finally,
the operation of LTM is checked in sodium once again. During the testing, torque is
measured while raising/lowering and the same is recorded.
The LTM is assembled in the air with the top flange of the test pot. After assembly,
the mechanism is operated to check for any manufacturing defects, its kinematics
and smooth operation. Torque is measured and recorded during the operation of
LTM in the air. Afterwards, the LTM is installed in the test pot of sodium loop and
its performance is checked in argon. Argon testing is an intermediate step before
proceeding for sodium testing. The LTM is operated for 10 cycles in argon. The
performance of the LTM is checked in sodium before endurance testing. Test pot is
filled with sodium at 200 °C, the mechanism is operated for 10 cycles and torque is
recorded. Then, the LTM is left in sodium at 200 °C for an idling period of 7 days.
After idling period, the LTM is taken up for endurance testing. During endurance
testing, the mechanism is operated in sodium at 200 °C for a total of 250 cycles and
one cycle includes raising and lowering of the mechanism by a distance of 130 mm.
After endurance testing, sodium is dumped and the mechanism is left in argon to
cool down to room temperature. Afterwards, the operation of the LTM is checked in
argon at 200 °C for 10 cycles.
Once again the test pot is filled with sodium and the operation of the lifting tool
mechanism is checked. The LTM is operated for 50 cycles and torque is measured
Realization of a Simple Mechanism to Simulate … 349
while operating the mechanism. Smooth operation is ensured again. After performing
all the tests, the sodium is dumped and the mechanism is taken out of the test pot of
the sodium loop.
Not much variation in the measured torque for raising/lowering is observed in air,
argon and sodium. The variation of torque vs. number of cycles, while raising the
LTM during endurance testing is shown in Fig. 3. Various inferences, regarding the
operation of trapezoidal screw-nut assembly in sodium, are drawn from these results.
Endurance test results in sodium highlights that initially the torque is high and
oscillating. It starts reducing after about ~30 cycles of operation and keeps reducing to
the minimum value. The reasons for torque reduction might be due to the smoothening
of the surfaces of screw and nut with the operation and lubrication effect provided
by the sodium. To support the above inference, the torques in air and sodium are
compared. It is found that the torque in the air before endurance testing and the
initial torque during endurance testing in the sodium are comparable, i.e. in the order
of 4–5 Nm. While the torque in the air after endurance testing is slightly higher than
the end torque during endurance testing in sodium but lower than the torque in the
air before endurance testing, i.e. in the order of 3–4 Nm.
After sodium testing, the sodium is dumped and the system is brought to room
temperature in argon. Further operation of LTM is not possible in argon at room
4
Torque (N-m)
0
0 50 100 150 200 250
No. of Cycles
Fig. 4 LTM during removal from test pot showing solidified sodium on its surface
temperature while the nut operating tool, when decoupled from the nut of screw-nut
assembly, is having free movement. This infers that the screw-nut had gone jammed
due to the solidification of sodium. Further, upon increasing the temperature of argon,
the operation of screw-nut assembly is restored. Initial operation of LTM in argon at
200 °C is smooth while torque keeps on increasing with the number of cycles. The
reason is expected to be the lubrication effect of sodium, i.e. initially, the screw-nut
assembly had sodium on its surfaces and as the operation progresses, the sodium
starts draining resulting in torque increase.
After testing in argon, again sodium is filled in the test pot of sodium loop, and
the LTM is operated to check multiple use of LTM in sodium. Behaviour of the
screw-nut operation is again similar to the previous testing and torque value reduces
as the number of operation increases.
After successful operation of the LTM, the sodium is dumped and the mechanism
is taken out of the test pot. Solidified sodium is found in and around LTM, as shown
in Fig. 4. But no bulk collection of solidified sodium inside of SA head or outside of
it is found. Visual inspection of the LTM is performed and no damage to the various
items of the LTM is observed.
6 Conclusion
References
Abstract Low weight and ability to withstand large deformations make tensegrity-
based structures attractive for a wide range of applications. It has been reported that
given a number of bars (struts), strings (cables) and interconnections between bars
and strings, more than one configuration can be arrived at. However, the possibility
of taking a tensegrity from one of stable configurations to another has not been
reported so far. This paper first explores a Monte Carlo-based method of arriving
at different configurations of tensegrity structures having the same number of bars,
strings and their interconnections. The results were validated against analytical solu-
tions available in the literature. It then uses the dynamic relaxation method to show
the intermediate steps for transforming one configuration into another by the applica-
tion of a set of external forces. This can lead to multi-usability of tensegrity structures
by switching from one stable configuration to other configuration.
1 Introduction
Tensegrity structures are structures with tensile and compressive elements, whose
self-stress keeps the structure in equilibrium. Self-stress is defined as the internal
force of the elements distributed over lengths due to which the nodes are in equilib-
rium and enable free-standing of the structure. The first tensegrity structure was built
by Snelson [1], and the term Tensegrity was coined by Fuller [2]. Many researchers
have defined tensegrity in various ways. The most widely accepted definition is the
one given by Pugh [3]: “A tensegrity system is established when a set of discontin-
uous compression components interacts with a set of continuous tensile components
to define a stable volume in space”. Form finding of a tensegrity structure refers to the
determination of its stable configuration. Many researchers have provided various
methods for form finding of tensegrity structures. Tibert and Pellegrino [4] have
provided a review of various form finding methods. Among all those methods, force
density method is one of the most popular method and was first proposed by Schek
[5]. The advantage of this method is that it allows one to linearize the nonlinear
equilibrium equations which makes the task of finding nodal positions for a given
set force densities easy. Later, Zhang et al. [6] proposed a numerical adaptive force
density method to find force densities and nodal position for stable configuration in
an iterative way. Bernes [7] developed dynamic relaxation method for form finding
which was later on used by Motro [8] to solve for cable net structures. Estrada et al.
[9] presented a form finding procedure with only information of connectivity and
type member as input and provides force densities and nodal coordinates for an equi-
librium configuration as output. More recently, many optimization techniques for
form finding by minimizing the potential energy have been proposed, for example,
Monte Carlo simulation by Li et al. [10] and genetic algorithm by Luo et al. [11].
A tensegrity structure can be in equilibrium and capable of having more than one
stable states. This property of multistability—i.e. existence of various configurations
with the same number of bars (struts) and strings (cables)—allows many different
applications to be explored. Luo et al. [11] and Lee et al. [12] identified multi-stable
configurations by using genetic algorithm and force density, respectively.
The work reported in this paper can be considered an extension of the work
done by Li et al. [10]. In addition to the work by Li et al. [10], two algorithms have
been described, one for identifying multiple stable tensegrity structures by exploiting
Monte Carlo method and another for transforming one stable state to another by using
the dynamic relaxation method.
We choose random initial configurations with bars of equal length and connect
them with cables. The free lengths of cables are allowed to change to find a stable
configuration. The forces in each member (sk ) can be calculated from Eq. (1) as
follows.
0 if(lk − l0k ) < 0
sk = , Here ‘lk ’ is the stressed length (1)
K k (lk − l0k ) otherwise
The dimension of each component forces is ‘ne × 1’. Once we get the components
of member forces in each direction, we can then find the net force at a node, ‘j’. Let
‘Oj ’ denote the set of all elements that are connected at node ‘j’ and ‘P j ’ are the other
356 P. K. Malik et al.
nodes connected to node ‘j’. Then the net force ((f j )) at node ‘j’ can be calculated
from Eq. (3) as shown follows:
Oj
Oj
Oj
f jx = −Ck j Skx , f j y = −Ck j Sky , f j z = −Ck j Skz (3)
k=1 k=1 k=1
Here, C kj is the connectivity of the jth node of size of ‘nm × n’. Similarly, the
elastic potential energy (E k ) of the ‘kth’ is given by Eq. (4).
ne
0 if(lk − l0k ) ≤ 0
Ek = , Total energy = E = Ek (4)
2 k (l k
1
K − l0k ) otherwise
2
k=1
Table 1 Comparison of twist angle obtained from Monte Carlo method and analytical method
provided in literature
Number of bar (nb ) θ obtained analytically (°) θ obtained from MC (°)
6 60 60.0501
10 72 71.8836
25 82.8 82.7873
all strings for both expanded octahedron and truncated icosahedron was taken as
‘0.536’ units and ‘0.3’ units, respectively, and the length of bars for both structures
is considered as ‘1’ unit. The stable configurations obtained are similar to those
reported by Luo et al. [11] which they had achieved by much more computationally
intensive genetic algorithm. The strain energies of expanded octahedron for all three
configurations are (a) 0.690, (b) 0.1709 and (c) 0.2373 units, respectively. Whereas
for truncated icosahedron the energy are (a) 0.1449 and (b) 0.2690 units, respectively.
Thus, Monte Carlo is capable to identify multiple stable configurations. Exploring
whether one shape can be transformed into another is the next challenge which is
addressed in the next section.
In this section, the deployment from one stable configuration to another for the
expanded octahedral is reported through DR method. The DR method aims to
progressively reduce the unbalanced forces at the nodes and achieve equilibrium.
The structure in one energy state overcomes an energy barrier to go to another energy
state by application of an external forces. The steps of DR method are: (1) Give inputs
similar to Monte Carlo method. Define threshold for acceptable unbalanced forces at
nodes and define external forces. (2) Calculate element forces and unbalanced nodal
forces (3) Calculate lumped mass matrix for the nodes as mentioned by Bernes [7].
(4) Find out acceleration, velocity and displacement of each node as described by
Bernes [7] and displace each node to find new coordinates. (5) Calculate the kinetic
energy of the system at time, ‘t + δt’. (6) If KE (t + δt ) > KE (t), and conver-
gence criterion is not met, return to step 2 for next iteration. (7). If KE (t + δt ) <
KE (t), set all velocities to zero, return to step 2 for next iteration. A set of random
forces at each node are applied in one of the stable configurations by keeping one
node fixed. This was done to see whether these set of forces take the structure to
a different configuration. The exercise was successful and we were able to achieve
transformation two stable configurations as shown in Fig. 3.
The external force was applied for the first 2000 iterations. The configurations
during the deployment process have been shown in Fig. 3 from (a) to (f). Whereas
Fig. 3a, f is two stable configurations as achieved by Monte Carlo-based method. This
shows that the DR method is indeed capable of transforming a tensegrity structure
from one stable state to another. Recording of the intermediate states (Fig. 3b–e)
allows the process to be reversed and repeated.
3 Conclusion
This paper explores Monte Carlo-based methods for identifying multiple stable
configurations of tensegrity structures with the same number of bars, strings and
Investigation of Multiple Stable States of Tensegrity Structure 359
References
Abstract Short-wave infrared (IR) heaters are used in the aerodynamic heating test
facilities to simulate the fast-transient thermal loads on high speed flying objects at
ground level. Design and development of a flat IR heater of 215 kW capacity is pre-
sented in this manuscript. Major challenges faced during the design and development
of such high capacity electric heater in a 300 mm by 260 mm space with gold plated
reflector are discussed in detail. Special purpose short-wave IR lamp of 350 W/cm
linear power density is designed for this heater. Experiments are also conducted to
identify the heat flux generating capacity of the heater with the use of thin circular
foil type heat flux sensors. A maximum of 2 MW/m2 heat flux is generated with the
developed heater.
1 Introduction
Infra-Red (IR) heaters based on tungsten filaments, clear quartz envelope and highly
polished reflectors have been in use in various industries from past 40 years [1].
Due to their heat generating capacity, inherent properties and improvements in past
years have kept them on top in various fields of applications. Short-wave (SW) and
medium wave electric IR heaters are the most dominant in aerospace, automobile,
drying and process industries. The SWIR heaters have attained more attention in
aerospace industry due to their fast-transient responses and high temperature capa-
bility. However, the theory, design, development and experimental aspects of these
IR heaters are undocumented. No design and analysis of these heating systems with
ultra-high electric capacity such as 215 kW in a 0.30 m by 0.25 m area is known to
the present authors.
Due to the fast transient process cycle times, the processing of thermoplastic
composites has increased the use of these IR heaters [2]. The accurate modeling of
the IR heaters will help in clear understanding and prediction of the heat-up times
required for a particular process [3, 4]. A very few works [2–6] are available in
literature about the modeling of the IR heater and its applications in various fields.
Monte Carlo simulation in the form of ray tracing scheme is utilized by Travis and
Robert [7] to predict the radiant heat flux distribution produced by tungsten filament,
tubular fused-quartz envelope heating systems with reflectors.
In the present work, we have designed and developed an ultra-fast responsive
SWIR flat heater of 215 kW capacity to cater the needs of the fast-transient thermal
testing of aerospace vehicle structures. A number of such heaters are used to cover
the contour of airframe structure to be tested and all the heaters can be controlled
individually or in combination.
2.1 Background
The heat input to the test article simulating the kinetic heating is given by electric
infrared heaters (EIH) to achieve the specified temperatures. The dominant mode of
heat transfer is through radiation in which rapid heating of the test article can be
achieved. As we know that the component responsible for the heating effect in the
visible light is the radiation in the IR region. The wavelength of the IR radiation
is roughly ranging from 0.1 to 100 µm. The heating effect produced by the IR
radiation has been exploited to heat sections and components to desired levels. The
fundamental requirements for the design of EIH are given in Fig. 1.
The source of IR radiation is SWIR lamp. The lamp is a cylindrical tube of quartz
glass, enveloping a filament made of tungsten. Tungsten is preferred as a filament
material because of its high melting point. The quartz glass envelope is filled with a
mixture of an inert gas like neon and halogen. As current passes through the tungsten
filament, it gets heated up, oxidizes—forms tungsten oxide and vaporizes. To prevent
this, a halogen gas is kept inside the enclosure so that as soon as tungsten oxidizes
it forms tungsten—halide which is inherently unstable and hence decomposes back
to tungsten and halogen. Thus, the tungsten is deposited back on the filament. The
SWIR lamp designed for the present heater has the following properties: linear power
density of (350 ± 5%) W/cm, rated power of (9000 ± 5%) W, rated operating voltage
of 415 V, 50 Hz, AC, shape of C-bend, lamp diameter of 10 ± 0.1 mm (standard quartz
Design and Development of a Short-Wave Electric … 363
Fig. 2 a A schematic of EIH assembled with IR lamps and b a cut-view showing the radiative
interactions among reflector, filament, quartz and test object
T3 lamp), heated length of 257 mm (Tolerance +0, −5 mm), lead wire of 200 mm
on either side with high temperature insulating sleeve up to 600 ◦ C without lugs,
end ceramic caps of maximum diameter 12 ± 0.5 mm and minimum diameter of
8 ± 0.5 mm, bend angle of 90◦ and out of plane bending tolerance is ±0.5 mm. The
lamp also has peak wavelength of IR emission in between 0.5 and 2.0 µm range at
1.2 µm. The life of lamp under continuous operation at rated voltage and open rack
condition is 1 h. A schematic of the IR lamp in EIH is shown in Fig. 2a, the cut-view
of lamp is shown in Fig. 2b. Each lamp consumes an electrical power of 9.0 kW at a
rated voltage of 415 V AC, and generates an equivalent radiative power. A maximum
of 2541 ◦ C temperature is obtained on the tungsten filament by measured voltage,
and current to the lamp. These measured voltage and current are used to calculate
the dynamic resistance of the filament and then with the use of resistance versus
temperature relationship of tungsten material, the filament temperature is calculated.
364 S. Narendar and T. Srinivasa Kumar
The electric IR heater is a stack of symmetrically placed IR lamps over the reflector
of heater body. A typical IR lamp used in the present 215 kW electric IR heater
has been shown in Fig. 2a. The heater body is having a closed cooling duct with a
reflector at lamps end and a back cover to hold requisite electrical bus bars for power
distribution to IR lamps in delta format. The function of a reflector is to redirect/reflect
back the stray IR energy on to the test article. The function of cooling is to ensure
the temperature of the heating panels is maintained for efficient working. The power
isolators on the backside of heaters are to mount the bus bars and it isolates the
electrical connections and heater body. Three surfaces, i.e., reflector, quartz glass
tube, filament and test object, have been identified as the major sources of radiation
exchange. These surfaces have been assigned an index (1–5). Radiosity (Ji ), and
irradiation (G i ) for each surface have been indicated with corresponding indices.
The radiation exchange is related to the wavelength (λ) dependent emissivity (ε(λ)),
transmissivity (τ (λ)), and reflectivity (r (λ)). The radiation balance equations are
written as Jλ,1 = rλ,1 G λ,1 + ελ,1 E λ,1 , Jλ,2 = rλ,2 G λ,2 + τλ,3 G λ,3 + ελ,2 E λ,2 , Jλ,3 =
rλ,3 G λ,3 + τλ,2 G λ,2 + ελ,3 E λ,3 , Jλ,4 = rλ,4 G λ,4 + ελ,4
E λ,4 and Jλ,5 = ελ,5 E λ,5 . The
irradiation of different surfaces is given by G λ,i = A1i j (F j−i A j Jλ,i ), where F j−i
is view or shape factor and A j is the area of radiated surface. Using these expressions,
the G λ,i and Jλ,i can be eliminated to get expressions for E λ,i . These results in five
linear equations with five unknowns. The system of linear equations can be solved
using Gaussian eliminations and with these solutions the expressions for G λ,i and
Jλ,i can be obtained. Then the net heat transfer by radiation to different surfaces can
be determined and the summation will give the net radiation flux output. Note that
all these relations are wavelength dependent. The solid model of the heater is shown
in Fig. 3 along with all its components. Along with water cooling to the back of the
gold plated reflector, an air cooling is also incorporated on the front surface of the
reflector. Three-phase delta connection is incorporated for the IR lamps assembles
on the heaters.
3 Experiments on EIH
Fig. 3 Components of a SWIR heater, realized EIH assembled with SWIR lamps and snapshots
during heater operation at 10 and 50% of rated voltage input
close linear fit is obtained as: heat flux, q (in MW/m2 ) = 0.245 × V − 0.299. By using
spline interpolation method, the convolution of the heater characteristics (see Fig. 4a)
and the desired heat flux-time profile (see Fig. 4b), the input gate voltage versus time
relation is calculated by using “spline” and “ppval” commands of MATLAB and
Simulink and output profile is shown in Fig. 4c. Note that the thyristor gate can
operate between 0 and 10 V. Now, the generated digital data of voltage-time profile
is converted into along data and is then fed to thyristor gate. With respect to this
time-voltage profile, the radiation output of the heater is measured at a distance of
50 mm from its surface. The measured heat flux may or may not match exactly with
the desired one. So, a fine-tuning in the obtained voltage-time profile is needed by
doing the experiment repeatedly. A comparison of the desired and achieved heat flux
profiles is shown in Fig. 4d. A very good match is obtained between both the profiles
with the designed heater. It can be observed that the heater can generate a heat flux
of 1.85 MW/m2 in 8 s of time. This is the ultra-fast response of the SWIR heater. The
heater also reaches its maximum capacity at less than 3 s also. At the end of 22–30 s,
the match between the curves is not good due to the cooling part of the sensor. Other
than this everywhere it is perfectly matched.
366 S. Narendar and T. Srinivasa Kumar
2 2
1.8 1.8
1.6 1.6
Heat Flux (MW/m2 )
Fig. 4 Radiation characteristics of EIH: a heat flux variation with gate voltage, b desired heat flux
versus time, c computed gate voltage versus time and d a comparison of desired and achieved heat
flux profiles with time
4 Conclusions
The developed SWIR heater has been tested successfully for its effectiveness in terms
of heat flux generating capability. A maximum of approximately 2 MW/m2 heat flux
has been measured at a distance of 50 mm from the heater surface within a very
short duration. Multiple numbers of such heater can be used to test the large-scale
aerospace airframe structures and systems.
References
1. Hanson HA, Casey JJ (1987) High temperature test technology AFWAL-TR-86-3105. U.S. Air
Force
2. Brogan MT, Monaghan PF (1996) Thermal simulation of quartz tube infra-red heaters used in
the processing of thermoplastic composites. Composites: Part A 27A:301–306
Design and Development of a Short-Wave Electric … 367
3. Cunningham JE, Monaghan PF, Brogan MT, Cassidy SF (1997) Modelling of pre-heating of flat
panels prior to press forming. Composites: Part A 28A:17–24
4. Cunningham JE, Monaghan PF, Brogan MT (1998) Predictions of the temperature profile within
composite sheets during pre-heating. Composites: Part A 29A:51–61
5. Pettersson M, Stenstroem S (2000) Modelling of an electric IR heater at transient and steady
state conditions: Part I: model and validation. J Heat Mass Transf 43:1209–1222
6. Ojala KT, Lampinen MJ (1995) Modeling, measurements and efficiencies of infrared dryers for
paper drying. In: Mujumdar AS (ed) Hand-book of industrial drying, 2nd edn. Marcel Dekker,
New York, pp 931–976
7. Travis LT, Robert LA (1994) Numerical and experimental analyses of the radiant heat flux
produced by quartz heating systems. NASA Technical Paper 3387
Kinematic and Dynamic Modeling
of a Quadruped Robot
Abstract The quadruped robots have numerous advantages over the wheel due
to their agility to exposure to cluster environment. This paper presents the kine-
matic analysis and dynamic modeling of a four-legged robot. The kinematic mecha-
nism is the fundamental approached before proceed to the gait design. The forward
and inverse kinematics equations are derived for legs with D-H transformation
matrices. Also, dynamic modeling for trot gait pattern is established with the Newton–
Euler theory. The computer simulation results are presented for different movement
sequences of trot gait pattern to verify the validity of proposed design.
1 Introduction
In the last three decades, the mobile robot has made a lot of attention because of
exploring in the complex environment, space, rescue operation, and accomplish a
task without human effort, etc. The mobile robot can be broadly classified into three
categories; wheeled robot, tracked robot and legged robot [1]. The Development
of terrestrial locomotion of legged robot has been grown constantly over the few
decades because of more advantages than other robot vehicles. The advantages of
legged locomotion depend on the postures, the number of legs, and the functionality
of the leg [2]. Though wheeled and tracked robots can work in plane terrain, but most
of them couldn’t fit in cluttered terrain, complex and hazardous environments, etc.
The legged robot has more potential to roam almost all the earth surfaces in different
terrains, just like a human and an animal. The quadruped robots are the best choice
among all legged robots related to mobility and stability of locomotion [3]. The four
legs of the robot are easily controlled, designed and maintained as compared to two or
six legs. The conceptual design of the skeleton quadrupedal robot is shown in Fig. 1.
The structure and parameters of each leg are identical with consist of 3 degrees of
freedom (DOF).
In order to find out the position and velocity of end-effector or feet of the quadruped
robot, the kinematic analysis is very much important. Configuration model of
quadrupedal robot is designed by 3-D modelling software, solid works. In this paper
mechanical model of quadrupedal robot is divided into two main part: body or frame
and legs. Each leg is identical in nature and symmetrically joined in the main body.
Each leg is composed of a series of links connected by different joints. In model
each leg has three joints and actuated by electric motor. Starting with the assumption
that each connection between two links has a single degree of freedom. A robot leg
with ‘n’ number of joints means it will have n + 1 links and associate with a joint
variable. Each leg of the robot is behaved like a serial manipulator which is rigidly
attached to the co-ordinate frame to each link. In order to get the co-ordinate of the
origin of the feet/end effector relative to the base frame, by choosing the origin of the
base frame at any arbitrary point. However, it is assumed that no slippage between
feet and ground.
In 1955, Denavit and Hartenberg proposed a D-H transformation matrix for attaching
an arbitrary frame to each link of a spatial linkage [4]. In this D-H convention,
transformation related with the joint z i and xi . This is the most prominent method
for the solving kinematics analysis of robotic system. The four parameters link twist
(αi ),link length (ri ), joint angle (θi ) and link offset (di ) are associated with D-H
Kinematic and Dynamic Modeling of a Quadruped Robot 371
convention method. Where, angle αi is the angle from z i−1 to z i measure about xi ,ri
is the distance from z i−1 to z i measure along xi , di is the offset distance from xi−1
to xi measure along z i−1 and angle θi is from xi−1 to xi measure about z i−1 . The
co-ordinate transformation of each link from previous co-ordinate system can be
represent in notation
i−1
Ti = Rxi (αi−1 ).Txi (ri−1 ).Tzi (di ).Rzi (θi )
⎛ ⎞
Cθi −Sθi 0 ri−1
⎜ Sθi Cαi−1 Cθi Cαi−1 −Sαi−1 −Sαi−1 di ⎟
=⎜⎝ Sθi Sαi−1 Cθi Sαi−1 Cαi−1 −Cαi−1 di ⎠
⎟ (1)
0 0 0 1
where Rot and Trans denote the rotation and translation respectively, and Sθi Cθi
are the shortest form of sin θi and cos θi respectively. The desired model of Denavit-
Hartenberg leg parameters for forward kinematics is given in Table 1 [5, 6]. For a
single leg, each transformation matrices of link from previous co-ordinate system
can be represented in notations like b T0 ,0 T1 , 1 T2 2 T3 and 3 T4 . The deduction of a
homogeneous
transformation matrix near by co-ordinate systems of one leg foot
legl l=1 is as follows (2–6):
Each leg of the robot has identical D-H co-ordinate frames and parameters so that
direct kinematic equation can apply from joint-3 to joint-1. The transformation from
joint (Oio ) to co-ordinate of the base frame (Ob ) for each leg can be expressed by the
constant translational transformation matrix with considering different sign value of
λ and δ in Eq. (6).
0
T1 = Rot(z, θ1 ) (2)
1
T2 = Rot(x, 90◦ )Tran(r1 , 0, 0)Rot(z, θ2 ) (3)
2
T3 = Tran(r2 , 0, 0)Rot(z, θ3 ) (4)
3
T4 = Tran(r3 , 0, 0) (5)
372 P. Biswal and P. K. Mohanty
b
T0 = Trans(λP, −δ Q, −H )Rot(y, 90◦ ) (6)
R3×3 p3×1
0
T4 = 0 T1 .1 T2 .2 T3 .3 T4 = (7)
0 1
where
⎛ ⎞
Cθ1 Cθ23 −Cθ1 Sθ23 Sθ1
R3×3 = ⎝ Sθ1 Cθ23 −Sθ1 Sθ23 −Cθ1 ⎠ P3×1 =
Sθ23 Cθ23 0
⎛ ⎞
r1 Cθ1 + r2 Cθ1 Cθ2 + r3 Cθ1 Cθ23
⎝ r1 Sθ1 + r2 Sθ1 Cθ2 + r3 Sθ1 Cθ23 ⎠.
r2 Sθ2 + r3 Sθ23
Similarly, the co-ordinate of four feet in the base frame can be obtained by multi-
b 0
T0 with T4 transformation matrix. The position co-ordinates of one foot of
plying
l
leg l=1 with respect to the base frame are given in detail from Eqs. (8–10). The x-z
co-ordinates are generated for all combining possible of joint variables using forward
kinematics in MATLAB as shown in Fig. 2.
70
60
50
Z-axis
40
30
20
10
0
-40 -20 0 20 40 60 80 100
X-axis
where, cθi j = cos θi cos θ j − sin θ j sin θi ,sθi j = sin θi cos θ j + cos θi sin θ j and px ,
p y and pz are denoted as the elements of position vector.
In this paper,
we calculate in detail the joint variables by an inverse kinematic equation
of legl l=1 by the use of end-effectors co-ordinates ( px , p y , pz ) in base global point
(Ob ). The three joint variables are given directly as follows in Eqs. (11–13).
Q + pz
θ1 = tan−1 − (11)
H + pz
M 2 + N 2 − r22 − r32
θ2 = cos−1 (12)
2r2 r3
θ3 = a tan 2(N , M) ± a tan 2 (N 2 + M 2 − K 2 , K , Where K = r2 + r3 cθ3
(13)
Langrange(L) = T − U (Energies)
where T and U are the total kinetic and potential energy respectively of the mechan-
ical system. The Lagrange’s equation for each generalized co-ordinate of the dynamic
equation of motion can be written in the form (15)
d ∂L ∂L
− = τi (15)
∂t ∂ θ̇i ∂θi
4 Simulation Results
z-direction are shown in Figs. 7, 8 and 9. The relation between rotational angle of
each joint and time is θ1 = −2t, θ2 = −7t and θ3 = −9t, where 0 ≤ t ≤ 5 the
location of origin point of coordinate system (O0 ) is given as (180, 100, −40 mm)
in co-ordinate system base (Ob ).
376 P. Biswal and P. K. Mohanty
5 Conclusions
• In this paper, a conceptual model of the four-footed robot has been developed.
From this study, a 3 DOF leg with kinematic modeling which used in quadruped
robot and validated through simulation with the actual model.
• The joint torque and end-effector position is calculated by considering the initial
simulation parameter. The role of forward and inverse kinematic mechanism has
been investigated for development and implementation on the quadruped robot.
Kinematic and Dynamic Modeling of a Quadruped Robot 377
• The real-time analysis of joint variables and motion control of quadruped robot
is perceived based on simulation results. In future work, a controller will be
implemented to coordinate the position of all rotational joints consist of 12 DOF
for the quadruped robot.
Acknowledgements The authors would like to acknowledge the support from the Science and
Engineering Research Board (SERB), a statutory body of the Department of Science & Technology
(DST) Government of India.
378 P. Biswal and P. K. Mohanty
References
1. Meng X, Wang S, Cao Z, Zhang L (2016) A review of quadruped robots and environment
perception. In: 2016 35th Chinese control conference (CCC). IEEE, pp 6350–6356
2. Raibert MH (1986) Legged robots that balance. MIT Press
3. Zhuang H, Gao H, Deng Z, Ding L, Liu Z (2014) A review of heavy-duty legged robots. Sci
China Technol Sci 57(2):298–314
4. Siciliano B, Khatib O (eds) (2016) Springer handbook of robotics, Springer
5. Rong XW, Li YB, Ruan JH, Song HJ (2010) Kinematics analysis and simulation of a quadruped
robot. Appl Mech Mater 26:517–522
6. Gao DH, Hu XB (2010) The analysis of kinematic simulation of 6-DOF articulated carrier robot
based on ADAMS. Manuf Inf Eng China 1
7. Bennani M, Giri F (1996) Dynamic modelling of a four-legged robot. J Intell Rob Syst 17(4):419–
428
Nonlinear Joint Stiffness Parameter
Identification
1 Introduction
Engineering structures generally exhibit some degree of nonlinearity, joints are the
major source of introducing nonlinearity in structures, and sometimes the structural
response can get significantly modified due to such nonlinearities.
Linear models of physical systems have limitations of their own and are applicable
only in a very restrictive domain like when the vibration amplitude is very small.
Thus, to accurately identify and understand the dynamic behavior of a structural
system under general loading conditions, it is essential that nonlinearities present in
the system also are modeled and studied.
Various techniques, methods, models and algorithms with many practical exam-
ples of nonlinear dynamic behavior and the effect of joint nonlinearity have been
reported in the engineering literature, e.g., method based on modal identification and
describing functions, for parametrically identifying nonlinearities from measured
frequency response functions in cantilever beam [1]; Fourier–Galerkin algorithm
to study the effect of joint nonlinearity [2]; parametric identification of a nonlinear
element between any two coordinates of the system [3, 4]; a reduced order model
for accurately incorporating the effects of the joint into a structural model [5]; a
parametric model approach using the method of multiple scales [6]; a method for
studying the nonlinear vibration of satellites with local nonlinear elements [7].
A method demonstrated in this paper is applicable to structures having local
nonlinearity that can be modeled as a single nonlinear element. This method basically
exploits the idea that linear modal identification can be applied to pseudo-linear data
obtained by keeping the response amplitude constant and modal parameters can be
obtained for each response level [8].
where (EI) is bending stiffness of the beam, L is length of the beam and λ =
2 4 1/4
ω ρ AL
EI
in which (ρA) is the linear mass density function, and ω is the natural
frequency of flexural vibration of the cantilever beam. The functions F1 , F3 , F5 , F6
are defined in Ingole [9],
K 1∗ = K 1 + K 3L (1)
3
K 3L ∼
= K 3 X x2 (2)
4
A cross coupling stiffness between translational and rotational spring is not consid-
ered in the present method, and hence, β 12 = β 21 = 0. Equation (4) can be further
simplified as
2
L2 1 F4
− (K 1 + K 3L )K 2
EI λ4 F3
382 S. B. Ingole and S. W. Rajurkar
L3 1 F5
− (K 1 + K 3L )
EI λ3 F3
L 1 F6
+ K2 + 1 = 0 (6)
EI λ F3
EI F5 EI F5
K 1 .K 2 + K 2 .K 3L + .λ K1 + .λ K 3L
L F4 NL L F4 NL
2
EI F6 EI F3
− .λ 3
K2 = .λ4 (7)
L3 F4 NL L 2 F4
NL
K 1 .K 2 + K 2 .K 3L + [A x ]NL K 1 + [A x ]NL
K 3L − [Bx ]NL K 2 = [C x ]NL (10)
K 1 .K 2 + A y LL K 1 − B y LL K 2 = C y LL (11)
The linear parameters of the system K 1 and K 2 can be estimated from measured
natural frequencies at very low forcing levels where the nonlinear internal forces will
be negligible. Now only unknown in Eq. (12) is K 3L which can be estimated from
experimentally measured nonlinear response amplitudes of the system.
3 Numerical Simulation
The method presented here is applicable to structures where there is local nonlinearity
that can be modeled as a single nonlinear element. A simple parametric identification
technique is proposed for nonlinear boundary condition of cantilever beam system
as shown in Fig. 1.
For numerical simulation, Eq. (12) is considered here. The equation can be used
to estimate the natural frequencies of the system when the linear and nonlinear
boundary stiffness parameters and material and geometrical parameters of beam are
known. However, we have exploited the equation for inverse analysis, wherein one
can estimate the unknown stiffness parameters from measured natural frequencies. In
general applications, the linear model of the system can be obtained by using FEM,
and only for the identification of nonlinearity experiments can be made. FRF of the
linear system can be obtained from FRF measurements in the system at very low
forcing levels, where the nonlinear internal forces will be negligible [1]. In the method
proposed, the natural frequencies are estimated from FE model of cantilever beam for
assumed values of linear stiffness parameters, and for the sake of brevity, FE model
has not been presented here. These estimated frequencies are treated as measured
natural frequencies and incorporate in sub-structure synthesis model Eq. (11) to
estimate linear joint stiffness parameters. These parameters are then compared with
assumed values; good agreement between estimated and assumed values is observed.
In forward analysis, Eq. (12) is considered, and the value of K 3 is assumed in the
range so that the non-dimensional stiffness
K 3 .F02 K 3 .X x2
Kn = 3
= = 0.01 to 0.05 (13)
K1 K1
Table 2 Natural frequencies estimated with low response amplitude (linear zone)
Assumed values Response Non-dimensional natural frequency
of stiffness amplitude (m) λ1 λ2 λ3 λ4 λ5
parameters
K 1 = 5 × 105 0.0001 1.8499 4.5370 7.2494 9.5660 12.1313
N/m 0.0003 1.8499 4.5371 7.2495 9.5661 12.1314
K 2 = 1.5 × 104
Nm/rad 0.0005 1.8499 4.5371 7.2496 9.5663 12.1315
K 3 = 1 × 109 0.0008 1.8499 4.5371 7.2499 9.5668 12.1318
N/m3
Table 3 Natural frequencies estimated with higher response amplitude (nonlinear zone)
Assumed values Response Non-dimensional natural frequency
of stiffness amplitude (m) λ1 λ2 λ3 λ4 λ5
parameters
K 1 = 5 × 105 0.001 1.8499 4.5372 7.2502 9.5672 12.1320
N/m 0.002 1.8499 4.5377 7.2525 9.5709 12.1341
K 2 = 1.5 × 104
Nm/rad 0.003 1.8500 4.5385 7.2563 9.5769 12.1375
K 3 = 1 × 109 0.004 1.8500 4.5395 7.2616 9.5853 12.1423
N/m3 0.005 1.8501 4.5409 7.2683 9.5960 12.1485
0.006 1.8501 4.5425 7.2762 9.6090 12.1561
0.007 1.8502 4.5443 7.2852 9.6241 12.1651
0.008 1.8502 4.5463 7.2953 9.6412 12.1754
Nonlinear Joint Stiffness Parameter Identification 385
nonlinear behavior. The identified natural frequencies are given in Table 3. The
higher frequencies are plotted in Fig. 3a–c showing linear zone.
Fig. 4 a Variation of
equivalent linearized
stiffness K 3L (N/m) with
response amplitude for mode
3. b Variation of equivalent
linearized stiffness K 3L
(N/m) with response
amplitude for mode 4. c
Variation of equivalent
linearized stiffness K 3L
(N/m) with response
amplitude for mode 5
388 S. B. Ingole and S. W. Rajurkar
Table 5 Estimates and estimation error in K 3L with ± 0.1% perturbation in mode 4 natural
frequency
Assumed Response Estimation of K 3L
values of amplitude (m) Without +0.1% freq. −0.1% freq. Average %
stiffness perturbation perturbation perturbation error
parameters
K 1 = 5 × 105 0.001 739.58 746.050 733.250 0.86
N/m 0.002 3025.0 3051.53 2999.04 0.85
K 2 = 1.5 ×
104 N-m/rad 0.003 6747.3 6806.88 6689.15 0.86
K 3 = 1 × 109 0.004 11,993.0 12,099.7 11,888.8 0.87
N/m3 0.005 18,734.0 18,902.6 18,570.1 0.88
0.006 27,017.0 27,262.3 26,777.1 0.90
0.007 36,768.0 37,107.0 36,437.6 0.91
0.008 47,988.0 48,437.2 47,549.9 0.92
Fig. 5 a Variation of equivalent linearized stiffness K 3L using mode 4 frequency with 0.1%
frequency perturbation. b Variation of equivalent linearized stiffness K 3L using mode 4 frequency
with −0.1% frequency perturbation
Table 7 Estimates of equivalent linearized stiffness K 3L with ±0.1% frequency perturbation for
X x = 0.006 m, for K x = 100 and K y = 20
Assumed values of Mode Estimation of K 3L
stiffness parameters Without +0.1% freq. −0.1% freq.
perturbation perturbation perturbation
K 1 = 4.99 × 104 N/m 2 269.75 272.15 267.38
K 2 = 3593.75 Nm/rad 3 277.25 277.88 276.65
K 3 = 1 × 107 N/m3
4 253.39 251.47 255.41
5 209.30 205.71 213.05
Table 8 Estimates of equivalent linearized stiffness K 3L with ±0.1% frequency perturbation for
X x = 0.006 m, for K x = 1000 and K y = 100
Assumed values of Mode Estimation of K 3L
stiffness parameters Without +0.1% freq. −0.1% freq.
perturbation perturbation perturbation
K 1 = 4.99 × 105 N/m 2 2806.15 3025.73 2609.83
K 2 = 17,968.75 Nm/rad 3 2757.14 2824.4 2692.44
K 3 = 1 × 108 N/m3
4 2702.96 2726.35 2680.13
5 2654.41 2649.03 2660.67
Table 9 Estimates of equivalent linearized stiffness K 3L with ±0.1% frequency perturbation for
X = 0.006 m for K x = 1000 and K y = 500
Assumed values of Mode Estimation of K 3L
stiffness parameters Without +0.1% freq. −0.1% freq.
perturbation perturbation perturbation
K 1 = 4.99 × 105 N/m 2 2705.4 2938.56 2498.93
K 2 = 89,843.75 Nm/rad 3 2700. 4 2766.06 2637.2
K 3 = 1 × 108 N/m3
4 2664.52 2686.41 2643.14
5 2699.53 2692.34 2707.63
4 Conclusion
References
9. Ingole Sanjay B, Chatterjee A (2013) Joint parameter identification of cantilever beam using
sub-structure synthesis and multi-linear regression. Struct Eng Mech 45(4):423–437. https://
doi.org/10.12989/sem.2013.45.4.423
10. Ingole Sanjay B, Chatterjee A (2017) Joint stiffness identification: a three-parameter joint
model of cantilever beam. Int J Acoust Vibr 22(1):3–13. https://doi.org/10.20855/ijav.2017.22.
1445
Design and Development of Reaper
for Harvesting Maize
Abstract Maize crop is one of the major crops in India, and mechanization of
different operations in Maize cultivation is essential in order to increase the produc-
tivity. Harvesting of Maize is one of the operations which is labour intensive and
involves drudgery. Though reapers are available for other crops, and reapers specifi-
cally developed for harvesting Maize are not found in the Indian market. To address
the identified need, a detailed review of the literature related to Maize crop cultiva-
tion practices in general and harvesting of Maize, in particular, was carried out. Field
study visiting different farms in the region of north Karnataka where Maize is culti-
vated largely was carried out. Based on the literature review and data collected from
field study, the complete specification of the reaper for Maize crop suitable for small
farmers was developed. The specification developed included important parameters
like required ground clearance, wheel diameter, engine power, cutter blade diam-
eter, distance between the wheels, weight of the reaper and cutting speed. Different
product concepts were developed to meet the specifications, and the most suitable
one was selected. All the sub-systems of the selected concept, namely reaper mech-
anism, traction system and the frame, were synthesized and designed to meet the
requirements. The fabrication of all the sub-systems and the assembly was carried
out in-house. The developed reaper has been tested for its working and found to be
satisfactory. The estimated reduction in cost of harvesting is found to be in the range
of 80% apart from reducing drudgery.
1 Introduction
The present scenario in India is that due to gradual increase in population, the demand
for food products is increasing but on the other side, availability of agricultural labour
is continuously decreasing. The mechanization in agriculture has been implemented
largely in most of the countries, to achieve higher productivity in order to meet the
ever-increasing demand for food products. But in India, the extent of adaption of
mechanization in agriculture is very low reflecting in lower productivity. The main
reason for the low level of mechanization in Indian agriculture may be attributed to
the non-availability of low cost and compact machines in market suitable for farmers
with small-scale landholding. There is huge scope for design intervention to address
this issue through development of mechanization solutions to meet the requirement
of small farmers, and the present work is attempt in this direction.
In India, maize is the third most important food crops after rice and wheat, culti-
vated in 8.7 Million hectors. Maize production in India contributes nearly 9% in the
national food basket and greatly contributes to the agricultural GDP [1]. One of the
most important operations in Maize cultivation is harvesting and is done manually
using sickles. Scarcity of labour and higher wages during harvesting season is a
serious problem. Typically, 200 man-hours are required to harvest one hectare of
Maize crop. During harvesting season, labour shortage is a common problem and
can only be addressed through mechanization solution.
2 Problem Statement
In order to develop the specifications for the proposed reaper, field visit and interac-
tion with farmers who are involved in Maize cultivation were carried out. From the
field study and interaction with farmers, the important data related to maize crop such
as row spacing, plant to plant spacing, height of the plant, stem diameter, deviation
of the plant from the centre axis and initial soil bed height were gathered. In order to
make the reaper compact and cost effective, the reaper is designed as walk-behind
type with traction to be suitable for cutting single row. Based on the data gathered,
the design specifications of the proposed walk-behind-type reaper were developed
and illustrated in Table 1.
Considering the specifications, three different concepts were developed, namely
rotating cutting disc reaper, reciprocating cutting blade reaper and continuous rotating
Design and Development of Reaper for Harvesting Maize 395
Fig. 1 Concept sketch and CAD model of the rotating disc reaper
blade reaper. The three concepts considered employed different cutting mechanism
like rotating disc, reciprocating blades and chain saw type cutter. The best concept
among the three was selected based on the suitability of cutting mechanism for cutting
thick stems of Maize, energy requirement, mechanical design of the cutter assembly
and the cutting efficiency [2]. The selected concept is the rotating disc-type reaper
for which the concept sketch and the CAD are shown in Fig. 1. The whole reaper
system consists of sub-systems like the power system, the traction system, the reaper
mechanism and the support structure or frame.
4 Detailed Design
The adapted design of the reaper is walk-behind-type arrangement with two wheels.
The source of power is an IC engine, from which power is drawn for traction and
cutting. To match the walking speed of the operator, a gearbox has been employed to
obtain the necessary reduction. The power required for operating the rotary blade
is transmitted from the engine to the cutter through a belt drive. The engagement
between the traction system and the engine is controlled through a centrifugal
clutch. The components/sub-systems required for developing the intended reaper
396 M. M. Ullegaddi et al.
The traction system in the reaper employs an IC engine to drive the wheels through a
gearbox. The engine capacity is decided based on the traction required and the energy
required for cutting. The traction required to run the reaper at walking velocity is
determined using standard traction model [3] given below.
T
= C g = 0.88(1 − e−0.1Bn ) × 1 − e−K 1 s + K 2
r L ∗R
The traction system shown in Fig. 2 includes engine, gearbox and a centrifugal
clutch. Here, the motion from engine is transferred through the centrifugal clutch to
the gearbox, and the wheel axle is integral with the output shaft of the gearbox.
The designed reaper mechanism utilizes standard available rotary cutting blade [6]
of diameter of 0.3 m. The critical speed at which the cutter is to be run is arrived at
by considering the plant material property and the crushing force to be exerted by
the cutter [4]. The power required to achieve cutting action is determined as follows
[5]:
2 × 3.142 × N × T
P=
60
Torque = F × r
r = Cutting disc radius, m.
F = Crushing force to be exerted, N.
i = Rotational speed of disc, rpm.
The rotary cutter is rotated at the required speed through a belt drive drawing
power from the engine. The reaper mechanism assembly is shown in Fig. 3.
The frame is the supporting structure on which all other sub-assemblies are mounted.
The dimensions of the frame are arrived at based on the mounting requirement of
traction system, engine gearbox and the reaper mechanism. The frame is designed
employing hallow rectangular tube of dimension 40 mm × 20 mm and 2 mm thick-
ness. The final design has been arrived at through number of design iterations to keep
398 M. M. Ullegaddi et al.
the design simple and lighter. The complete frame is fabricated through welding. The
fabricated frame is shown in Fig. 4.
The prototype of the Maize harvesting reaper developed is shown in Fig. 5. The
prototype has been developed in-house and tested for its working and found to be
satisfactory. The total cost of the reaper is INR 14,000/- which is very less compared
to existing harvesting reaper for other crops.
Detailed specification of reaper for Maize crop has been successfully developed based
on the field study and literature review. Three alternative concepts were developed to
meet the specifications, and the best concept was selected. The detailed design was
Design and Development of Reaper for Harvesting Maize 399
carried out, and the prototype was developed in-house. The developed reaper has
been tested for its working and found to be satisfactory. By comparing the cost of
manual harvesting and using the developed reaper, it is found that using the developed
machine results in 77% reduction in cost of harvesting process. The developed system
is such that it can be used for other inter-cultivation operations also by removing the
cutter assembly and using different attachments. It is also identified that since Maize
stems are used as fodder, the developed reaper can also be converted to work also as
a chopper with minor modifications.
References
5. Ghahrae O, Ahmad DB (2008) Design and development of special cutting system for sweet
sorghum harvester. Central European Agriculture 9:469–474
6. Ullegaddi MM, Nagarjun MA, Mahendra Babu NC (2020) ’Advances in Mechanical Engi-
neering’ Design and development of a mini sugarcane harvester. Springer, pp. 427–439
Transpose Jacobian Control of Flexible
Joint Upper Limb Exoskeleton System
Abstract In the present work, a two-link upper limb exoskeleton system is consid-
ered with harmonic drive actuators to track a semi-circular and L-shape path. The
usage of harmonic drive actuators raises the problem of joint flexibility and produces
substantial vibration at the tip of the exoskeleton system. The problem of joint flexi-
bility in the system dynamics is considered using nonlinear spring model. Thereafter,
the dynamics for two reduced cases, i.e., linear spring model and the rigid joint model
are presented. At the last, transpose Jacobian control strategy is proposed to track
the paths with different dynamic models. The controlled joint actuator torques for
different cases are found and compared to analyze the system behavior.
1 Introduction
Powered exoskeleton is a wearable device to improve the limb movement and increase
its ability to withstand hardship by using electric, hydraulic and pneumatic actuators
[1, 2]. Moreover, it can be utilized by nurses to lift and carry patients as well as for
the rehabilitation of stroke or spinal cord injury patients [3]. The most challenging
issue for these devices is to select a lightweight actuation system having a better
response [4]. According to the literature, harmonic drive actuators are used in the
exoskeleton system as they favor low weight, low backlash, enhanced torque, good
repeatability and compactness [5, 6]. The main problem with these drives is their
joint flexibility, caused by shaft windup and bearing deformation. The presence of
flexibility sometimes leads to instability of the system [7, 8]. The dynamics of the
two-link rigid and linear flexible joint are well known [7, 9]. However, the analysis
of nonlinear flexibility in elbow and shoulder joint for the two-link exoskeleton
system needs to be explored. In this work, transpose Jacobian (TJ) control strategy,
introduced by Craig [10], is used for controlling a two-link upper limb exoskeleton
system with rigid and flexible joint dynamics. In this way, the effect of flexibility
from the harmonic drives in the exercise-based rehabilitation devices can be reduced.
2 System Dynamics
Fig. 1 Schematic diagram of a exoskeleton system with hand, b flexible joint model
Transpose Jacobian Control of Flexible Joint Upper Limb … 403
The closed form solution of the dynamic equation of motion is derived using Euler-
Lagrange formulation. This formulation is based on the notation of generalized coor-
dinates, kinetic energy, potential energy and the generalized forces. The generalized
coordinate q represents the joints variable and joint angles in the case of revolute
joints. K is the kinetic energy which depends on the position and the velocity of the
robotic arm; V is the potential energy depending on the position of the arm [11].
To design a reasonable control scheme, the joints are considered and modeled as a
spring having nonlinear cubic stiffness. The effect of soft windup is also considered
in the dynamic model of the harmonic drive because it reduces the stiffness of the
system in the region from 0 to 1 Nm of the torque-torsion characteristic as explained
by Nenad in [12]. Moreover, the model of frictional torques [13] is implemented to
form complete dynamic model of the proposed system. The flexible joint dynamics
of the exoskeleton system consists of two nonlinear differential equations one repre-
sents the link dynamics and the other is the actuators dynamic and defined by the
Lagrangian using Eqs. (1–8):
L q j , q̇ j = K q j , q̇ j − V q j , j = 1, 2 (1)
K q j , q̇ j = K l + K m j = 1, 2 (2)
1
n n
Kl = M jk q̇ j q̇k , j = 1, 2; k = 1, 2; n = 2 (3)
2 j=1 k=1
1
n n
Km = Jm jk q̇m j q̇mk , j = 1, 2; k = 1, 2; n = 2 (4)
2 j=1 k=1
V q j = Vl + Vm j = 1, 2 (5)
n
Vl = m j gh j (q), j = 1, 2; n = 2 (6)
j=1
1
n n
Vm = k jk (q j − qm j )(qk − qmk ), j = 1, 2; k = 1, 2; n (7)
2 j=1 k=1
d ∂L ∂L
τj = − , j = 1, . . . , n (8)
dt ∂ q̇ j ∂q j
Since the generalized coordinate of the revolute joints is the joint’s angle;
therefore, the dynamic Eqs. (9–22) of the proposed system can be written as follows:
404 J. Narayan et al.
M(θ )θ̈ + Sθ̈ m + C θ , θ̇ θ̇ + G(θ ) + f θ̇ − k(θ , θ m )(θ m − θ) = 0 (9)
e−asw (θm1 −θ1 )
2
(θm1 − θ1 )2 0
k(θ , θ m ) = a1 2 + a 2 − k sw −asw (θm2 −θ2 )2
(11)
(θm2 − θ2 ) 0 e
M11 M12
M(θ ) = (12)
M21 M22
M11 = m 1lc1
2
+ m 2 (l12 + lc2
2
+ 2l1lc2 cos θ2 ) + I1 + I2 (13)
m j l 2j
M22 = m 2 lc2
2
+ I2 , I j = (15)
12
C11 C12
C θ, θ̇ = (16)
C21 C22
C11 = −m 2 l1l2 sin θ2 θ̇2 , C12 = −m 2 l1l2 sin θ2 θ̇1 + θ̇2 (17)
G1
G(θ) = (19)
G2
f θ̇ = β1 tan h β2 θ̇ − tan h β3 θ̇ + β4 tan h β5 θ̇ + β6 θ̇ (22)
where L q j , q̇ j is the Lagrangian, K l , K m and Vl , Vm are the kinetic and potential
energy of the links and the motors, respectively, τ represents control torque vector,
M(θ ) is n × n positive definite and symmetric rigid inertia matrix, θ is the vector of
joint angles, S indicates upper triangular matrix and represents the inertial couplings
for the motor and the link acceleration,
θ m be the vector indicating the angular
displacements of the motor shaft, C θ, θ̇ is n × n matrix comprises the Coriolis
and centrifugal effects, G(θ) is n × 1 matrix denotes the gravitational effect, f θ̇
Transpose Jacobian Control of Flexible Joint Upper Limb … 405
is friction torque matrix, k(θ , θ m ) matrix indicates the stiffness of the flexible joints,
θ − θ m denotes the vector of elasticity of joint vibrations, J m indicates the motor
inertia matrix, a1 , a2 diagonal matrix of stiffness coefficients, k sw represents the soft
windup factor and βi for i = 1, . . . ., 6 represents the different friction components.
While m 1 and m 2 are the masses of the first and second link, respectively, g is the
acceleration due to gravity, lc1 and lc2 denote the distance from the previous joint
to the center of gravity of the first and second link, respectively.I j represents the
moment of inertia of the links about an axis perpendicular to the x-y plane passing
through the center of gravity of the link.
Case1: Linear joint model
The dynamic analysis of the exoskeleton system with linear joint flexibility can be
obtained by modeling the joints as a linear spring with constant stiffness and ignoring
the soft windup as well as the frictional effects [14], as shown in Eq. (23):
k(θ , θ m ) = a2 = k, S = 0, f θ̇ = 0, a1 = 0, ksw = 0 (23)
J m θ̈ m + k(θ m − θ ) = τ (25)
3 Control Scheme
In this paper, the transpose Jacobian methodology is applied to the rigid joint model
of exoskeleton system as the first step. Then, the same control strategy is modified
and adopted for flexible joint exoskeleton dynamics to track the desired paths.
406 J. Narayan et al.
Fig. 2 Block diagram of rigid joint model with transpose Jacobian control strategy
This scheme is introduced for the first time by Craig [10] due to its simplicity, the ease
of implementation in the hardware model and the ability to reduce the positional error
in the motion tracking task. Equations 28, 29 provide the control law of transpose
Jacobian controller in terms of actuating torque to be applied at the joints. Here, τ is
the output of the controller, called the control law, and provides the vector of torques to
be applied at the joints of the system, J(θ ) is the Jacobian matrix of the exoskeleton
system. K p and K d are the proportional and derivative matrixes of control gain,
respectively. xe and ye are the errors in the tracking position between the actual and
desired tip position of the exoskeleton system, while ẋe and ẏe are the errors in the
tracking velocity between the actual and desired velocity of the same. This control
law is applied to the dynamic equations of the exoskeleton system resulting in the
angular acceleration vector which is integrated twice to get the actual position of the
tip of the exoskeleton system as illustrated in Fig. 2.
xe ẋ
τ = J(θ ) K p
T
+ Kd e (28)
ye ẏe
J J
J(θ ) = 1 2 (29)
J3 J4
The control law given by the Eq. (28) with the assumption of joint rigidity can
be combined with another control term which is devoted to eliminate the elastic
oscillation that occur at the joints due to joint flexibility [15]. This term is called
the fast term τ f and the previous one is the slow term τ s which controls the rigid
dynamic model of the system. The modified control law is given by Eqs. (30–33),
whereas kv is the control gain of the fast section which is combined with the slow
Transpose Jacobian Control of Flexible Joint Upper Limb … 407
Fig. 3 Block diagram of flexible joint model with transpose Jacobian control strategy
section resulting in the control law τ . This controller output is applied to the dynamic
model of the exoskeleton system to reduce the positional error at the tip of the system
as presented in Fig. 3.
τ = τs + τ f (30)
xe ẋ
τ s = J(θ ) T
Kp + Kd e (31)
ye ẏe
τ f = kv θ̇ − θ̇ m (32)
xe ẋ
τ = J(θ ) T
Kp + Kd e + kv θ̇ − θ̇ m (33)
ye ẏe
The obtained results are summarized in this section by applying the rigid transpose
Jacobian control law to the rigid joint model of the exoskeleton system and by
applying the modified transpose Jacobian control law to the flexible joint model. The
tip of the exoskeleton system is proposed to follow two different paths (semi-circular
and L-shape). The parameters of the exoskeleton system are summarized previously
in the Sect. 2.1. While the parameters of nonlinear flexible joint model were taken
by assuming a highly flexible joints as follows:
a1 = a2 = diag[500], k sw = diag[10], asw = 3000, β1 = 0.5, β2 = 150, β3 =
0 Jm2
50, β4 = 2, β5 = 100, β6 = 0.5, S = , and the controller gains are
0 0
calculated based on the trial and error approach and having the following values for L-
shape path:k p = diag[100],k d = diag[22.5],kv = diag[30]. For semi-circular path,
k p = diag[30], k d = diag[24], kv = diag[12.1]. Figure 4 illustrates the position of
408 J. Narayan et al.
the tip of exoskeleton system tracking the desired trajectory with different dynamic
consideration.
According to Fig. 4a, the tip follows a semi-circular path of 0.3 m radius, starting
from the right corner at the coordinates of (0.3, 0) within 20 s. It is obvious that
when the system is modeled as a rigid joint dynamic, the deviation at the tip position
according to the desired path is small with 0.04 m overshoot at the beginning of
the path. This deviation increases as the flexibility of the joint comes to the picture
with 0.07 m and 0.1 m overshoot for the linear and nonlinear flexible joint models,
respectively. Figure 4b shows the tip position of the system while tracking L-shape
path, starting from the position (0.15, −0.15) at the right corner within 20 s. The
positional error between the actual and desired tip position is very small according
to the rigid joint model. This error increases as the flexible joint model of the system
is considered with 0.005 m and 0.015 m overshoot corresponding to linear flexible
model and nonlinear flexible model, respectively. More deviation from the desired
path at the upper right-side corner of the trajectory is observed because of the sudden
change of the applied torque. Figure 5 illustrates the control torques applied at each
joint of the exoskeleton system in the case of semi-circular path. This control effort is
devoted to reduce the position error at the tip of the system. Moreover, these torques
are small in the case of rigid joint model with maximum 0.03 Nm and 0.06 Nm
for the first and second joint, respectively. It increases as the flexibility of the joint
is considered with maximum 0.06 Nm and 0.15 Nm for the first and second joint,
respectively, of the linear flexible joint model and 0.067 Nm and 0.164 Nm for the
nonlinear flexible joint model.
Figure 6 represents the torques applied at each joint to reduce the tip positional
error of the proposed system as well as to keep tracking the desired L-shape path.
One can observe that the torques devoted to track the L-shape path is higher than
the required to track the semi-circular one, which is satisfied because of the smooth
change in the semi-circular path and there is no sudden change in the positional
Transpose Jacobian Control of Flexible Joint Upper Limb … 409
Fig. 5 Control torque applied at the joints for the tracking of semi-circular path
Fig. 6 Control torque applied at the joints for the tracking of L-shape path
coordinates. The maximum torque value for the rigid joint model is equal to 0.28 Nm
for the first joint, and 0.37 Nm for the second joint at the beginning of the path. On
the other hand, the maximum value reaches to 0.38 Nm and 0.61 for the first and
second joint, respectively, for linear flexible joint model and 0.64 Nm and 0.85 Nm
for the same in the case of the nonlinear flexible joint model. When the tip of the
system reaches the first corner of the trajectory after 10 s, the torque profile shows
a sudden change in the torque with high overshoot in the case of nonlinear flexible
joint model of the exoskeleton system.
410 J. Narayan et al.
5 Conclusion
References
13. Makkar C, Dixon WE, Sawyer WG, Hu G (2005) A new continuously differentiable fric-
tion model for control systems design. In: International conference on advanced intelligent
mechatronics, pp 600–605. IEEE, Monterey, CA, USA
14. Spong MW (1987) Modeling and control of elastic joint robots. J Dyn Syst Meas Contr
109(4):310–319. https://doi.org/10.1115/1.3143860
15. Spong MW (1989) Adaptive control of flexible joint manipulators. Syst Control Lett 13(1):15–
21
Defect-Free Synthesis, Analysis
and Optimization of Planar Lower Limb
Assistive Device for Gait Rehabilitation
Abstract Aging is the natural phenomenon that can be seen around the world. It is
responsible for changing the muscle strength and body composition and limits the
capacity to perform routine activities. Hence, elderly are more prone diseases, central
nervous system disorder, strokes, etc. Also, the rehabilitation and hospitalization
costs are most expensive for strokes among other diseases. Contemporary devices
have many limitations such as they are bulky, heavy, complex, and require special
training to operate; consequently, they are not preferred by patients, therapists, and
small clinics. To address these issues, a new assistive device called “caster walker
gait trainer (CGT)” is proposed for gait rehabilitation.
1 Introduction
Stroke is the major cause of physical disability with hemiparesis that affect survivors.
It results in causing low velocity, asymmetric, and disordered gait [1]. For regaining
the gait mobility, stroke patients are required to undergo gait rehabilitation. The
manual therapy is time consuming and hectic for both the patients and the therapists
[2].
Robotic rehabilitation devices may be used in the early phases of recovery for
overcoming the limitations of conventional manual therapy [3]. It may be combined
with the physiotherapy to escalate the possibility of regaining the natural walk [4].
Various gait rehabilitation robotic devices have been developed. These devices may
This paper endeavors to design a mechanism that can be combined with a caster
walker to follow the Bobath (approach of rehabilitation) by making the patients to
follow the normal gait patterns, starting from their own to the more normal, and even-
tually effective gait pattern. The device caster walker gait trainer (CGT) is focused
on providing assistance to the patients with hemiparesis, especially at the beginning
stage of rehabilitation. The device is designed to be useful for small clinics and
individuals. Some notable specifications of CGT are as follows:
Defect-Free Synthesis, Analysis and Optimization … 415
900
Y-Coordiante
800
700
600
500
600 800 1000 1200
X-Coordinate
416 R. Singh et al.
The loop-closure Eqs. (1)–(2) can be solved [23] for the given orientations of the
link vector r21 to find the orientations other links r31 , r41 , r52 , and r62 . Then
of the
the coordinates of the footplate P Px , Py can be expressed in the global (fixed)
reference frame as in Eqs. (3)–(4):
Px_generated = Aox + r21 cos θ21 + r32 cos θ32 + lcp cos(θ52 + α3 ) (3)
Py_generated = Aoy + r21 sin θ21 + r32 sin θ32 + lcp sin(θ52 + α3 ) (4)
The goal is to design a Stephenson III six-bar mechanism that generates a path similar
to the desired path. In this work, tracking a desired ankle trajectory is considered
over tracking finite precision points accurately. Therefore, tracking error objective
function is posed as:
Defect-Free Synthesis, Analysis and Optimization … 417
n
2 2
Min f (x) = i
px_desier d − px_generated
i
+ piy_deisr ed − piy_generated
i=1
Subject to : g1 (x) = r21 + r11 − r31 − r41 < 0
j j+1
g2 (x) = θ21 − θ21 < 0, j = 1, . . . , N − 1
g3 (x) = r22 + r12 + r32 − r52 − r62 < 0
L k < xk < Uk , k = 1, . . . , n
where n is the total number of design variables, N is the number of precision points,
and x is the design vector.
Aox , Aoy , r11 , r21 , r31 , r41 , α2 , r52 , r62 , α3 , r32 , . . .
x=
lcp , Ao Do , θ12 , θ11 , θ21
1
, . . . , θ21
n
The objective function, f (x), is subjected to Grashof’s, g1 (x) and g3 (x) and order,
g2 (x), constraints to avoid any defects while synthesizing a feasible mechanism.
A nature-inspired technique known as teaching-learning-based optimization
(TLBO) algorithm or any metaphor-less technique [24] may be used to solve the
optimization problem. The simulation is performed 25 times, and the results of the
tracking error and best convergence are presented in Figs. 3 and 4, whereas, Table 1
defines the proposed design values, respectively. The proposed caster walker design
is presented in Fig. 5.
900
Y-Coordiante
800
700
600
500
600 800 1000 1200
X-Coordinate
418 R. Singh et al.
1500
1000
500
0
1 10 25 40 55 70
Iterations
5 Conclusions
The paper proposed a mechanism for a cater walker which can be used for gait
rehabilitation and training. The Stephenson six-bar linkage has been synthesized
optimally using a nature-inspired algorithm, i.e., TLBO. The proposed mechanism
is able to track the ankle joint trajectory with less error. Thus, the mechanism can be
embedded with a caster walker or it may be used in other lower limb rehabilitation
devices.
Defect-Free Synthesis, Analysis and Optimization … 419
1400
D
1200
C
Y-Coordinate
1000
A
B D
o o
800
A
o
B P
600
400
200 400 600 800 1000 1200
X-Coordinate
References
1. Flansbjer UB, Holmbäck AM, Downham D, Patten C, Lexell J (2005) Reliability of gait
performance tests in men and women with hemiparesis after stroke. J Rehabil Med 37(2):75–82
2. Chen G, Qi P, Guo Z, Yu H (2016) Mechanical design and evaluation of a compact portable
knee–ankle–foot robot for gait rehabilitation. Mech Mach Theory 1(103):51–64
3. Pennycott A, Wyss D, Vallery H, Klamroth-Marganska V, Riener R (2012) Towards more
effective robotic gait training for stroke rehabilitation: a review. J Neuroengineering Rehabil
9(1):65
4. Mehrholz J, Thomas S, Werner C, Kugler J, Pohl M, Elsner B (2017) Electromechanical-
assisted training for walking after stroke. Cochrane Database Syst Rev 5
5. Veneman JF, Kruidhof R, Hekman EE, Ekkelenkamp R, Van Asseldonk EH, Van Der Kooij H
(2007) Design and evaluation of the LOPES exoskeleton robot for interactive gait rehabilitation.
IEEE Trans Neural Syst Rehabil Eng 15(3):379–386
6. Cherry MS, Choi DJ, Deng KJ, Kota S, Ferris DP (2006) Design and fabrication of an elastic
knee orthosis: preliminary results. In: ASME 2006 international design engineering technical
conferences and computers and information in engineering conference 2006 Jan 1, pp 565–573.
American Society of Mechanical Engineers
7. Hussain S, Xie SQ, Jamwal PK, Parsons J (2012) An intrinsically compliant robotic orthosis
for treadmill training. Med Eng Phys 34(10):1448–1453
8. Hesse S, Waldner A, Tomelleri C (2010) Innovative gait robot for the repetitive practice of
floor walking and stair climbing up and down in stroke patients. J Neuroengineering Rehabil
7(1):30
9. Freivogel S, Mehrholz J, Husak-Sotomayor T, Schmalohr D (2008) Gait training with the newly
developed ‘LokoHelp’-system is feasible for non-ambulatory patients after stroke, spinal cord
and brain injury. Feasibility Study Brain Injury 22(7–8):625–632
420 R. Singh et al.
10. Shyu JH, Chen CK, Yu CC, Luo YJ (2011) Research and development of an adjustable elliptical
exerciser. Adv Mater Res 308:2078–2083. Trans Tech Publications
11. Kora K, Stinear J, McDaid A (2017) Design, analysis, and optimization of an acute stroke gait
rehabilitation device. J Med Devices. 1;11(1):014503
12. Lee SJ, Hidler J (2008) Biomechanics of overground versus treadmill walking in healthy
individuals. J Appl Physiol. 104(3):747–55
13. Bruni MF, Melegari C, De Cola MC, Bramanti A, Bramanti P, Calabrò RS (2018) What does
best evidence tell us about robotic gait rehabilitation in stroke patients: a systematic review
and meta-analysis. J Clin Neurosci 1(48):11–17
14. Mehrholz J, Thomas S, Werner C, Kugler J, Pohl M, Elsner B (2017) Electromechanical-
assisted training for walking after stroke. Cochrane Database Syst Rev (5)
15. Lo K, Stephenson M, Lockwood C (2017) Effectiveness of robotic assisted rehabilitation for
mobility and functional ability in adult stroke patients: a systematic review. JBI Database Syst
Rev Implementation Rep 15(12):3049–3091
16. Singh R, Chaudhary H, Singh AK (2017) Defect-free optimal synthesis of crank-rocker linkage
using nature-inspired optimization algorithms. Mech Mach Theory 1(116):105–122
17. Singh R, Chaudhary H, Singh AK (2018) A novel gait-based synthesis procedure for the design
of 4-bar exoskeleton with natural trajectories. J Orthop Transl 1(12):6–15
18. Singh R, Chaudhary H, Singh AK (2019) A novel gait-inspired four-bar lower limb exoskeleton
to guide the walking movement. J Mech Med Biol 1950020
19. Jardim B, Siqueira AA (2009) Development of series elastic actuators for impedance control
of an active ankle foot orthosis. In: 20th International congress of mechanical engineering, 15
Nov 2009, pp 15–20
20. McDaid AJ (2017) Design, analysis, and multicriteria optimization of an overground pediatric
robotic gait trainer. IEEE/ASME Trans Mechatron 22(4):1674–1684
21. Majernik J (2013) Normative human gait databases. Stat Res Lett 2(3):69–74
22. Perry J, Davids JR (1992) Gait analysis: normal and pathological function. J Pediatr Orthop
12(6):815
23. Waldron KJ, Kinzel GL, Agrawal SK (2016) Kinematics, dynamics, and design of machinery.
Wiley
24. Singh P, Chaudhary H (2019) Optimal shape synthesis of a metallic flywheel using non-
dominated sorting Jaya algorithm. Soft Computing 1–2
Theoretical and Experimental
Investigation of Friction in Hydraulic
Actuators
Abstract This paper deals with the experimental verification of friction models
used in the estimation of frictional forces in a hydraulic actuator. The friction force
is calculated with the measured readings of pressure in the cylinder chambers and
the acceleration of the piston. In order to predict friction forces in the hydraulic
actuator, mathematical models, namely LuGre model and modified LuGre model,
are extensively used. These models are simulated to predict the friction forces in
the test actuator. The physical test setup and the simulation models are commanded
with identical velocity inputs, and the results in each test case are compared. The
experiment results show that the modified LuGre model with dimensionless fluid
film thickness parameter predicts the friction forces in hydraulic actuator with a
good accuracy when compared with the other models.
1 Introduction
2 Friction Models
In the literature, the LuGre model predicts the most known friction effects like pre-
sliding displacement, response lag and stick-slip motion has been used for fric-
tion compensation in hydraulic servo systems. The LuGre model is represented by
following equations:
dz σ0 z
=v− v (1)
dt g(v)
dz
Fr = σ0 z + σ1 + σ2 v (3)
dt
where z denotes the additionally introduced state to model the bristle average defor-
mation. The bending of the bristles generates the friction force σ0 z + σ1 ż where σ0
and σ1 represent the bristle contact stiffness and the bristle damping co-efficient. The
term σ2 v represents the viscous friction and σ2 is the viscous friction co-efficient.
For the constant velocity condition, the steady-state friction force can be given by,
Though the LuGre model allows good description of the constant velocity behavior
and offers smooth transition at velocity reversals, it is seen that it is too dissipative
of energy in the pre-sliding regime (1, 3). In order to overcome this, Yanada and
Sekikawa (1) introduced a dimensionless lubricant film thickness parameter, ‘h’,
into the Stribeck function and represented the averaged bristle deflection as:
dz σ0 z
=v− v (5)
dt g(v, h)
dz
Fr = σ0 z + σ1 + σ2 v (7)
dt
The fluid film thickness can be given by
dh
= τh (h ss − h) (8)
dt
where
⎧
⎨ τhp (v = 0, h ≤ h ss )
τh = τhn (v = 0, h > h ss ) (9)
⎩
τh0 (v = 0.)
K f |v|2/3 (|v| ≤ |vb |)(|v| ≤ |vb |)
h ss = (10)
K f |vb |2/3 (|v| > |vb |)
FC
Kf = 1− |vb |−2/3 (11)
FS
The term ‘hss ’ is the dimensionless steady-state lubricant film thickness parameter,
‘K f ’ is the proportional constant for lubricant film thickness, ‘vb ’ is the velocity at
which the friction force becomes minimum in the steady-state curve, and ‘τ hp ’,
‘τ hn ’, ‘τ h0 ’ are the time constants for acceleration, deceleration and dwell periods,
respectively. The condition h ≤ h ss in Eq. (9) represents the acceleration condition
where the lubricant film thickness is less than its steady-state value, and the condition
h > h ss represents deceleration condition when the fluid film thickness is greater
than steady-state value. Further, when the lubricant film thickness is included, the
steady-state friction characteristics are given by:
Figure 1 shows the schematic of the experimental setup. The powered hydraulic fluid
is supplied by an axial piston pump at 25 LPM and 70 bar supply pressure. A double
acting hydraulic linear actuator of which, internal diameter 25 mm, outer diameter
40 mm and the stroke equal to ±35 mm are used. The actuator is fixed horizontally
to a test bed. The current input to the servo amplifier is given in sinusoidal and step
input formats through a 12-bit RMC 70 controller. The position of the hydraulic
piston which varies according to the servo valve current input is recorded through
an internal LVDT. The pressures P1 and P2 in the cylinder chambers are measured
through piezoelectric type pressure transducers. All the signals from the sensors are
recorded by a NI DAQ card through 16-bit analogue-to-digital converter.
For identification of the steady-state characteristics, step signals of magnitude
ranging from 0.1 to 1.5 mA are commanded to the servo valve, and the plot for friction
force vs. sliding velocity as shown in Fig. 2 is plotted. For studying the unsteady-state
characteristics and comparison of the friction models with the experimental data, a
sinusoidal velocity command of ±10 mm/s is commanded to the actuator. In this
case, the pressure readings P1 and P2 from the pressure transducers are recorded,
and the friction force is calculated using the force-balance equation:
Fr = (P1 − P2 )A − ma (13)
where ‘m’ is the mass of the piston equal to 2.23 kg and ‘a’ is the piston acceleration.
piston position and the velocity are computed by calculating the time rate of change
in piston position. The positive velocity region corresponds to the extension stroke
and the negative velocity region corresponds to the retraction stroke.
In the negative sliding regime, the minimum friction force is observed at sliding
velocity of 16 mm/s after which it increases linearly. Thus, from identification of
the Stribeck curve presented in Fig. 2, the Stribeck velocity in the positive sliding
regime is identified to be 17 mm/s (0.019 m/s), and in the negative sliding regime,
the Stribeck velocity is identified to be 16 mm/s (0.016 m/s), respectively. All other
identified steady-state parameters are presented in Table 1.
The comparison between the experimental data and simulation results of modified
LuGre model is shown in Fig. 3, which clearly indicates that the modified LuGre
model can predict various dynamic friction characteristics present in the hydraulic
actuator with a relatively good accuracy. The pre-sliding behavior, response to
velocity reversal and the decrease of maximum friction force after the first cycle
are well captured by this model.
The incorporation of varying film thickness parameter accounts for the prediction
of reduction in the maximum friction force after the first cycle. In the negative resis-
tance regime, the lubrication film thickness allows the prediction of Stribeck effect
426 J. Vyas et al.
Table 1 Steady-state
Parameters Extension stroke Retraction stroke
parameters
Fs (N) 221.57 −111.76
Fc (N) 144.49 −52.36
Fv (Ns/m) 30.81 16.67
Vb (m/s) 0.02849 −0.02309
Vs (m/s) 0.01759 −0.016
N 0.2 0.4
σ0 (N/m) 1.00E + 07
σ1 (Ns/m) 0.1
τhp (s) 0.4
τhn (s) 2.6
τh0 (s) 10
Fig. 3 Comparison of LuGre and modified LuGre model with experimental data
which predicts decreasing friction force with increasing velocity. The comparison of
the recent models with the experimental curve is presented in Fig. 3.
5 Conclusion
In this paper, the friction characteristics of the hydraulic actuator are simulated using
various friction models. The accuracy of the friction models under study, namely
LuGre model and modified LuGre model, in predicting the friction forces in a
Theoretical and Experimental Investigation of Friction in … 427
hydraulic actuator is carried out by comparing the experimental results with the
simulation results. The steady-state parameters required for simulation are identified
experimentally, and the methodology for the same has been elaborately discussed.
The LuGre model though capable of predicting the pre-sliding behavior fails to
capture the lag due to the presence of lubricant film. The LuGre model also predicts
higher friction forces when compared to experimentally identified friction forces.
The modified LuGre model which incorporates the dynamics of fluid film thickness
predicts the friction forces with good accuracy when compared with other models.
References
1 Introduction
Squirrel cage induction motors (SCIM) drive the modern industries owing to its
ruggedness, low maintenance and availability of efficient drives. Early detection of
electrical and mechanical faults is essential for minimal downtime. Broken rotor bar
(BRB) is common faults in a heavy starting torque requirement applications like in
mine winders, mine ventilations, centrifugal pumps and several other applications.
Broken rotor bar is among main faults in SCIM [3, 7, 20, 24]. Stator current mon-
itoring is considered as a non-invasive technique for fault monitoring in SCIM [4].
Most researches are performed by analysing the stator current using methods like
fast Fourier transform (FFT), Hilbert transformation, wavelets analysis, model-based
techniques, and other statistical analysis [9, 17, 19, 24]. The classical BRB classi-
fiers are sensitive to load variations [3]. In [25], authors have developed an intelligent
system for fault detection with the help of multi-class SVM and support vector data
description (SVDD). SVM-based fault classifier for induction has been presented in
[16]. In [12], authors have used SVM-based fault detection system with one-against-
one method. In [6], authors have presented a detailed review of the state-of-the-art
methods available for condition monitoring along with the latest developments.
The cost of monitoring of SCIM is also important parameter for its usage. Many
commercially available methods are expensive when deployed for low rating motors.
The availability of highly efficient computational machines drawn the attention of
researchers to look beyond conventional methods. The machine learning techniques
like artificial neural networks, regression techniques, decision trees and many others
were used by researchers for developing fault detection systems [2, 23]. The main
idea is to develop an efficient system for broken rotor bar fault detection under all
possible loading condition of SCIM. The aim of this paper is to develop a SVM
algorithm-based classifier which can detect broken bars. The novelty of this paper
lies in the fact that it can detect broken rotor bar fault under all possible loading
conditions.
This paper consists of five sections for clear illustration of BRB faults, recent
developments and proposed work. Section 1 is introduction, Sect. 2 gives brief intro-
duction of BRB faults, Sect. 3 presents the fast Fourier transform technique, Sect. 4
gives details of support vector machines. It also includes the development of proposed
SVM classifier and implementation of LabVIEW-based data acquisition system for
fault classification. Finally, in Sect. 5, conclusions are made based on the work and
findings.
The three-phase symmetrical winding supplied from a balanced supply will produce
the resultant forward rotating magnetic field. The rotor current generates an effective
three-phase magnetic field with identical number of poles as on the stator side but
rotating at a slip frequency s f s . In healthy SCIM, i.e. in symmetrical cage wind-
ing, only forward rotating magnetic field exists. BRB causes asymmetry in SCIM.
Asymmetry caused by broken bar will lead to a resultant backward rotating field.
The backward rotating magnetic field will induce an emf, and current in the stator
side of the frequency which is called sidebands, formed in the either side of prin-
ciple slot harmonics. Number of broken bar has no effect on the position of theses
characteristic harmonics. However, amplitude at these frequencies depends on the
number of broken bars.
Fast Fourier transformation (FFT) is a computational tool which simplifies the signal
processing and analysis. Motor current signature analysis is a non-invasive fault
monitoring technique for fault detection, and it searches for periodicity of the signal
Support Vector Classifier-Based Broken Rotor Bar Detection … 431
[2, 10, 15]. The sideband frequencies caused by broken rotor bar fault in the stator
current are given by [10]:
f lsb = f (1 − 2ks)Hz (1)
where k = 1 for principle slot harmonic, f is the supply frequency, f lsb is the left
sideband component and f rsb is the right sideband component. The analysis of stator
current can help in evaluating the health of motor. The variation of amplitude is
roughly proportional to the number of broken bars. It is suitable for contiguous broken
bar detection. The magnitude of sideband frequency component is proportional to
the number of contiguous BRBs [4, 5]. Based on these equation, frequencies of
the sidebands are calculated considering k = 1. The amplitude at these frequency
components shows the existence of the broken bars. However, if there are non-
contiguous broken bars, FFT -based current analysis may fail to detect the BRB fault
[5]. The factors like stator supply, loading, external noise and fault condition affect
the current spectrum. This affects the quality of the fault detection system [5]. Based
on the past studies, an empirical relation can be drawn on the left sideband values
(LSB) of the amplitude-frequency spectrum of the stator current. From this empirical
relation, following conclusion is drawn: (i) magnitude of less than −50 dB typically
denotes a healthy motor [4], (ii) magnitude higher than −45 dB corresponds to a
faulty motor [20], (iii) the 54-45 rule has been proposed which states the amplitude
of sidebands to less than −54 dB for a healthy motors, and higher than −45 dB for a
motor with BRB, (iv) the sideband amplitude between −45 and −54 dB is marginal
for broken rotor bars [18]. An empirical approach-based dependence and calculation
of sidebands amplitude under various loading condition is shown in [18].
margin between boundary and nearest data point of both the classes. The margin
is kept as trade-off between the margin level and generalization error. There can
be number of hyperplane drawn for separating two classes (as shown in Fig. 1);
however, the hyperplane having the maximum distance from the nearest data point
on the either side of classes will be chosen for the best fit and accuracy. Figure 2 shows
the hyperplane which segregate the two classes optimally, i.e. positive (+) sign and
negative sign (−). Finally, SVM classification problem is solved by considering an
optimization technique for maximizing the margin and consequently minimizing the
generalization error. Lastly, SVM model is developed which can predict the class of
new data. The traditional approach in machine learning is to minimize the error on
the training data set and it is called empirical risk minimization. However, support
vector machines (SVM) is based on structural risk minimization [8, 21]. SVM can
handle large data sets efficiently. The training of SVM is done so that the dimension
of classified vector does not influence the performance of SVM and it differentiates it
from conventional classifiers. Generally, this method provides better generalization
capabilities with respect to methods based on empirical risk minimization. This
is advantageous in fault classification problem as number of features need not be
limited.
Generally, these classifiers have good classification properties as compared to
conventional classifiers. The support vector contains the information needed for
defining the classifier, and when support vectors are properly selected, rest feature
is not of much interest. It can be used for solving linear and nonlinear classification
problems. For nonlinear classification, kernel functions like Gaussian, quadratic and
cubic are required.
Support Vector Classifier-Based Broken Rotor Bar Detection … 433
The SVM-based fault classifier is proposed for broken rotor bar fault detection.
Several authors [13, 14] have proposed artificial neural network-based classifier for
BRB detection based on the features extracted from current spectrum calculated
by FFT and [11] has used SVM-based classifier. These classifiers were based on
two-dimensional feature vectors based on the LSB and RSB. However, these fea-
tures provide good functionality under nominal loading conditions only. The features
which are robust under varying load conditions can be selected for an efficient algo-
rithm. In this paper, those features are selected which are robust to load variations
and applicable to real industrial environment. The training set has 2 × 202 samples,
the two features and 202 observations. Based on the training of SVM, it gives indica-
tion regarding the health of SCIM. The developed SVM-based fault indicator is for
broken rotor bar fault detection only. The accuracy of SVM classifier depends upon
the selected kernel functions [1]. The schematic diagram for BRB fault detection has
been shown in Fig. 3.
The experimental setup has LabVIEW-based data acquisition system, 4 pole 5.5 kW
squirrel cage induction motor and personal computer. The experimental setup is
shown in Fig. 4. SCIM has 36 stator slots and 28 rotor slots. The DC generator coupled
434 P. Kumar and A. S. Hati
Fig. 3 Schematic diagram for fault detection in squirrel cage induction motor
with SCIM allows load variation from 0 to 100%. The stator current is acquired by hall
effect current sensors and NI DAQ 6230. The rotational frequency varies from 24.83
to 23.66 Hz. The data is collected and processed with the LabVIEW software. The
data collected is for the healthy motor and faulty motor (BRB fault) under different
loading conditions. The collected data is processed by FFT and input vector for
SVM classifier is prepared. This data is then used for developing SVM classifier.
The training set has 2 × 202 samples, the two features and 202 observations. The
features are extracted from the FFT analysis of stator current. The implementation
of SVM classifier is done in MATLAB 2019a with linear, quadratic and Gaussian
function.
4.3 Results
The machine learning-based approach is based on two things: training and testing.
The purpose of training is to develop a function which separates the faulty class and
healthy class. In this paper, SVM is trained based on the acquired data from the setup.
Table 1 shows the performance accuracy of SVM with different kernel functions. The
selection of the kernel functions is important for minimizing the error rate. When
Gaussian kernel function is used, the accuracy is good and this classifier can be used
for fault classification. Area under the receiver operating characteristics (AUROC)
tells about the classification property of SVM classifier. Higher AUROC means the
fault classifier has good classification properties. AUROC of the developed classifier
is 0.95. Furthermore, fivefold cross-validation is done to validate the accuracy of
SVM. Figure 5 shows the model output and testing of broken bar classifier with five-
fold cross validation, and Fig. 6 shows the receiver operating characteristics (ROC)
curve of the SVM classifier. AUC shows the ability of the classifier to segregate the
healthy class and faulty class. For BRB fault detection, input can be fed to the classi-
fier, and it can detect the fault based on the training. This classifier can be efficiently
used for fault detection in SCIM.
Model predictions
BRB - Incorrect
0 BRB - Correct
Heathy - Incorrect
Heathy - Correct
-20
Amplitude
-40
-60
-80
-100
0 100 200 300 400 500 600 700 800 900 1000
Frequency
Fig. 6 ROC curve (positive class: healthy and negative class: BRB)
Support Vector Classifier-Based Broken Rotor Bar Detection … 437
5 Conclusion
This paper aims to design an efficient SVM-based classifier for BRB detection only.
Proposed SVM classifiers can efficiently detect BRB in SCIM. The two features
which are used can detect faults reliably. Based on the proposed work, it can be
concluded that classifier can efficiently detect broken bars. The merit of this method is
its applicability in various loading conditions. Also, novelty of the proposed method
lies in its accuracy, simplicity and applicability under all loading conditions. The
proposed method is simple, inexpensive and reliable. It is also shown that Gaussian
kernel function can detect faults with good accuracy as compared to other kernel
functions.
References
1. Abe S (2005) Support vector machines for pattern classification, vol 2. Springer, New York
2. Ayhan B, Chow MY, Song MH (2006) Multiple discriminant analysis and neural-network-
based monolith and partition fault-detection schemes for broken rotor bar in induction motors.
IEEE Trans Ind Electron 53(4):1298–1308
3. Bellini A, Filippetti F, Franceschini G, Tassoni C, Kliman GB (2001) Quantitative evaluation
of induction motor broken bars by means of electrical signature analysis. IEEE Trans Ind Appl
37(5):1248–1255
4. Benbouzid MEH (2000) A review of induction motors signature analysis as a medium for faults
detection. IEEE Trans Ind Electron 47(5):984–993
5. Benbouzid MEH, Kliman GB (2003) What stator current processing-based technique to use
for induction motor rotor faults diagnosis? IEEE Trans Energy Convers 18(2):238–244
6. Choudhary A, Goyal D, Shimi SL, Akula A (2018) Condition monitoring and fault diagnosis
of induction motors: a review. Arch Comput Methods Eng 1–18
7. Cunha CM, Lyra RO et al (2005) Simulation and analysis of induction machines with rotor
asymmetries. IEEE Trans Ind Appl 41(1):18–24
8. Kecman V (2001) Learning and soft computing: support vector machines, neural networks,
and fuzzy logic models. MIT Press, Cambridge, MA
9. Khan MAS, Rahman MA (2008) Development and implementation of a novel fault diagnostic
and protection technique for IPM motor drives. IEEE Trans Ind Electron 56(1):85–92
10. Kliman G, Stein J (1992) Methods of motor current signature analysis. Electr Mach Power
Syst 20(5):463–474
11. Kurek J, Osowski S (2010) Support vector machine for fault diagnosis of the broken rotor bars
of squirrel-cage induction motor. Neural Comput Appl 19(4):557–564
12. Martínez-Morales JD, Palacios-Hernández ER, Campos-Delgado D (2018) Multiple-fault diag-
nosis in induction motors through support vector machine classification at variable operating
conditions. Electr Eng 100(1):59–73
13. Matić D, Kulić F, Bugarski V (2010) Survey of the methods for online broken bar induction
motor fault detection. J Process Energy Agric 14(2):90–92
14. Matic D, Kulic F, Climente-Alarcon V, Puche-Panadero R (2010) Artificial neural networks
broken rotor bars induction motor fault detection. In: 10th symposium on neural network
applications in electrical engineering. IEEE, pp 49–53
15. Messaoudi M, Sbita L (2010) Multiple faults diagnosis in induction motor using the MCSA
method. Int J Signal Image Process 1(3)
438 P. Kumar and A. S. Hati
16. Mohsun NA (2017) Broken rotor bar fault classification for induction motor based on support
vector machine-SVM. In: 2017 international conference on engineering & MIS (ICEMIS).
IEEE, pp 1–6
17. Schoen RR, Habetler TG (1996) Evaluation and implementation of a system to eliminate arbi-
trary load effects in current-based monitoring of induction machines. In: IAS’96. Conference
record of the 1996 IEEE industry applications conference thirty-first IAS annual meeting, vol 1.
IEEE, pp 671–678
18. Siau J, Graff A, Soong W, Ertugrul N (2004) Broken bar detection in induction motors using
current and flux spectral analysis. Aust J Electr Electron Eng 1(3):171–178
19. Su H, Chong KT (2007) Induction machine condition monitoring using neural network mod-
eling. IEEE Trans Ind Electron 54(1):241–249
20. Thomson WT, Fenger M (2001) Current signature analysis to detect induction motor faults.
IEEE Ind Appl Mag 7(4):26–34
21. Vapnik V (2013) The nature of statistical learning theory. Springer, Berlin
22. Widodo A, Yang BS (2007) Support vector machine in machine condition monitoring and fault
diagnosis. Mech Syst Signal Process 21(6):2560–2574
23. Yang BS, Oh MS, Tan ACC et al (2009) Fault diagnosis of induction motor based on decision
trees and adaptive neuro-fuzzy inference. Expert Syst Appl 36(2):1840–1849
24. Yazici B, Kliman GB (1999) An adaptive statistical time-frequency method for detection of
broken bars and bearing faults in motors using stator current. IEEE Trans Ind Appl 35(2):442–
452
25. Zgarni S, Keskes H, Braham A (2018) Nested SVDD in DAG SVM for induction motor
condition monitoring. Eng Appl Artif Intell 71:210–215
Kinematic and Dynamic Analysis
of Primary FCS Circuits of Typical 25
Seater Transport Aircraft
Abstract This research work presents kinematic and dynamic analysis of primary
flight control surface actuation mechanisms including aileron, elevator and rudder
of typical mechanically controlled transport aircraft. The 3D CAD models of the
actuation mechanisms are constructed using SolidWorks. The circuit assemblies are
exported to MSC ADAMS for detailed kinematic and force analysis. An elaborate
simulation model is built in MSC ADAMS to understand the kinematic relations
and to estimate the forces acting on all links, brackets and joints under the action of
aerodynamic loads on the control surfaces. Analytical formulations are developed to
verify the results of the force analysis. Further, the effect of inertia of control surfaces
on the pilot forces is studied in detail and discussed.
1 Introduction
In a typical mechanically actuated transport aircraft, the cockpit controls are mechan-
ically linked to the primary flight control surfaces, namely aileron, elevator and the
rudder. Controlled motions applied in the cockpit controls result in the movement of
flight controls and vice versa. These mechanical actuation systems typically consist
of push pull rods, steel wire ropes and pulleys [1]. In this work, the 3D CAD models
of components of mechanically actuated primary flight control surface actuation
mechanisms of a typical transport aircraft are constructed using SolidWorks.
The kinematic linkages of flight control surface (FCS) circuit assemblies are config-
ured, and CAD modeling is carried out based on detailed literature survey [2, 3] of
typical 25 seater passenger aircraft. The aerodynamic loads on the control surfaces are
calculated as per appropriate standards, and the hinge moments are applied accord-
ingly. Detailed force analysis of the actuation mechanisms is carried out to extract
forces acting on all joints.
The control circuits of the primary control surfaces such as aileron, elevator and
rudder consist of many numbers of mechanical components. These components
of each circuit mainly include push pull rods, lever and bracket assemblies. The
geometric modeling and assembly of each actuation mechanism are carried out by
applying the appropriate standard and advanced mates using SolidWorks.
The construction of these kinematic assemblies to obtain the specified deflections
of control surfaces is a challenging task. This demands proper positioning of the links
and brackets during assembly. But the mechanisms are successfully assembled, and
kinematic relations are confirmed in SolidWorks before carrying out the further
analysis. As an illustration, the kinematic assembly of aileron actuation mechanism
constructed using SolidWorks is shown in Fig. 1.
To facilitate the operation of control surfaces by either of the pilots (pilot and
co-pilot), the cables and mechanical linkages (push pull rods) are interconnected
through the torque tubes (quadrant region) located forward of the front spar near the
wing center as shown in Fig. 1. The control surfaces are represented as square bars
of appropriate length to facilitate aerodynamic load simulation. Separate hydraulic
actuators are attached with the square bars using lugs and brackets to resist the motion
of the control surface.
Similarly, the kinematic assemblies of elevator and rudder including representative
control surfaces to facilitate aerodynamic load simulation are also constructed in
SolidWorks. These kinematic models are found to be functioning satisfactorily.
Appropriate input motions are applied to the joints at the cockpit controls as a function
of time on the above-mentioned kinematic models of aileron, elevator and rudder
actuation mechanisms. This applied motion is transmitted to the control surface
through mechanical linkages used in the respective control circuits.
442 C. Manjunath et al.
During flight, due to forward velocity of the aircraft, deflection of the control surfaces
with respect to the neutral position increases the aerodynamic load. The drag and
lift forces are the two major forces that act on the control surfaces at its center of
pressure. The resultant of these forces develops a net moment about the hinge line
which is referred as the hinge moment of the control surface.
The maximum hinge moment acting on each primary control surface is calculated
by using following relation,
H = q S f C f C h Nm (1)
where,
q Freestream dynamic pressure
ρ Freestream density (kg/m3 )
V Freestream velocity (m/s)
Sf Control surface reference area (m2 )
Cf Control surface reference chord (m)
Ch Hinge moment coefficient
Based on the literature survey [4], the inputs required are appropriately assumed,
and the hinge moments for aileron, elevator and rudder surfaces are estimated.
Kinematic and Dynamic Analysis of Primary FCS Circuits … 443
The resistive actuators are used to simulate the aerodynamic loads on the control
surfaces to oppose its movements. The actuator loads are calculated based on the
hinge moments acting on the control surfaces along with the provided moment arms
in respective surfaces. The numerical values of loads are obtained both for 100 and
150% of hinge moments. The actuators stroke required to provide the necessary
resistive loads from neutral to maximum deflections of the control surfaces are also
obtained using 1D sketch drawn in SolidWorks.
Figure 2 shows the kinematic and force model of elevator actuation mechanism. The
aerodynamic loads are applied on the control surfaces by simulating loads through
resistive actuators. The applied forces are linearly varying corresponding to control
surface deflections. Force analysis is carried out by applying 100% of calculated
loads and also for 150% loads in all the actuation mechanisms. With these loads, the
model is run, and the joint forces are extracted. The pilot controls such as control
wheel, control column and pedals used in aileron, elevator and rudder, respectively,
are also represented in Fig. 2.
Fig. 2 MSC ADAMS kinematic and force model of elevator actuation mechanism
444 C. Manjunath et al.
Figure 3 shows the brackets/typical joints of the elevator and push pull rods of the
rudder actuation mechanisms with nomenclature. Figure 4 shows the bracket and
push pull rod with corresponding reaction loads acting due to 150% actuator load as
an example.
Similar to the above bracket E B6 , all six components of reaction loads are
extracted for all 46 brackets and also reaction forces acting on all 30 push pull
rods of aileron, elevator and rudder mechanisms. These loads can be used for SOM
design or finite element analysis of the airframe/iron bird structure [5]. Also, the
pilot loads due to the simulated hinge moments are found to be in order as per FAR
23.
Fig. 3 Brackets/typical joints of the elevator and push pull rods of the rudder mechanism
Kinematic and Dynamic Analysis of Primary FCS Circuits … 445
Fig. 4 Bracket and push pull rod with corresponding reaction loads
The lump masses are added to square bar to simulate mass moment of inertia of
the control surface, and force simulation is carried out in all the cases with control
surfaces actuating at a rate of 90°/s. It is found that the reaction forces acting on
rudder circuit joints are increased by 3% approximately but the influence of lump
masses on the aileron, and elevator circuit joints are found to be almost negligible.
This is due to lower moment of inertia simulated in these two cases.
7 Conclusions
Kinematic models of the primary flight control surface actuation mechanisms are
successfully configured using SolidWorks and MSC ADAMS. Joint forces extracted
from the force analysis are validated with the analytical formulations. The forces
extracted on the joints can be directly used for SOM design and/or FE (finite element)
analysis of the airframe or the iron bird structure. Further, the effect of inertia of
control surfaces on the pilot forces is studied, and results are discussed.
References
1. Roskam J, Airplane flight dynamics and automatic flight controls, design, analysis and research
corporation. Part I, 238 (DAR Corporation)
2. Cook MV (2007) flight dynamics principles. Elsevier aerospace engineering series
3. Nam Y, Hong SK (2002) Force control system design for aerodynamic load simulator. Control
Eng Pract 10(2002):549–558
4. https://ruag.picturepark.com/Go/ZeJQA0rC/V/9757/1?purgechache
5. Gerkins T, Fritsch G, Seyffert C, Buttner J, Mugtussidis I, Morrise I (2004) The Iron Bird for
the Fairchild-Dornier 728. Aerosp Sci Technol 8
Kinematic Synthesis and Optimization
of a Double-Slotted Fowler Flap
Mechanism
1 Introduction
Flaps are one of the high lift devices incorporated on the trailing edge of the aircraft
wings. Fowler flaps have series of slots between flaps to add energy to the airflow.
Fowler flap operates with the basic principle of variable area, variable camber and a
gap between the wing and fully extended flap.
In a double-slotted Fowler flap, the flap is split into flap and aft-flap. This allows
the air flow to happen in two slots [1]. One is the gap between main wing and the
fore-flap, and other one is the gap between fore-flap and the aft-flap.
2 Problem Definition
The position of the flap with respect to the wing is specified in terms of gap, overlap
and flap angle [2]. Hence, there are three independent kinematic parameters required
to define the position of flap with respect to the wing. In a double-slotted Fowler flap,
the aft-flap also takes three independent positions with respect to the flap during the
three stages of flight. This adds additional complexity in synthesizing the mechanism,
as it is expected to satisfy totally 18 independent parameters. Combining these two
motions, positions of flap and aft-flap with respect to wing can be specified as shown
in Fig. 1.
The challenge in this kinematic synthesis is, all these positions of the flap and
aft-flap or all the 18 independent kinematic parameters of the flap and aft-flap should
be achieved with one input parameter or actuation.
3 Methodology
After an elaborate literature survey, it is found that six-bar mechanism is best suited
for synthesizing the flap positions, and a double-slotted mechanism is best suited
for synthesizing the aft-flap positions. A four-bar mechanism is required to bridge
and synchronize the motions of these two independent mechanisms. The conceptual
model of the mechanism is shown in Fig. 2.
The complete flap actuation mechanism consists of a six-bar, four-bar and a
double-slotted mechanisms suitably integrated and synchronized. The linkages are
configured accordingly, and as per Grubler–Kutzbach criteria, the degree of freedom
of the mechanism is found to be 1 [3]. Figure 3 shows the line sketch of the flap
mechanism.
To carry out the dimensional synthesis, line sketch model of the six-bar mechanism
is created with all dimensions or parameters as shown in Fig. 4. Totally nine linear
dimensions and three angle dimensions can be identified from the figure. It is required
to synthesis all the 12 dimensions as shown in the figure which will satisfy the posi-
tions of the flap (which are represented by nine independent kinematic parameters)
in all three stages of flight.
To facilitate the synthesis procedure, positions of all the joints of the mechanism
are obtained as a function of given input angle using vector mathematics and coded
as MATLAB program. The mathematical procedure and the MATLAB code are
checked and verified using line sketches in CAD software.
Fig. 5 Plots of percentage change in link dimensions versus output positions (for input f1)
MATLAB algorithm may give many possible combinations of dimensions which will
satisfy the flap positions within the error limits. From these combinations, the opti-
mization of the mechanism is performed to achieve a compact mechanism by mini-
mizing the projection from bottom surface of the wing. The algorithm for synthesis
and optimization of the mechanism is presented in Fig. 6.
The challenge in aft-flap mechanism synthesis is that the aft-flap has to achieve
different positions with respect to the flap without any separate actuator; hence,
relative motion in one of the linkages of the six-bar mechanism with respect to the
flap is selected and used for actuation of the aft-flap.
The aft-flap mechanism synthesis is carried out in two phases. In the first phase,
a double-slotted mechanism is configured to achieve three given positions of the
aft-flap with reference to the flap. In the second phase, a four-bar mechanism is
452 K. Vinod Kumar et al.
synthesized to synchronize the flap movement with respect to the wing to the aft-flap
position. Both these procedures are integrated to have constrained motion of the aft-
flap with respect to the crank motion of the six-bar mechanism. Figure 7 shows the
dimensions of the aft-flap mechanism which includes seven linear dimensions, one
angle dimension and the centers of the curved slots. A novel synthesis procedure is
developed to synthesis all these dimensions of the aft-flap mechanism.
To verify the synthesis procedure, an aircraft wing model is considered with flaps and
aft-flaps based on a typical 90 seater transport aircraft. Dimensions of the wing, flap,
aft-flaps and flap positions during three stages of flight are considered appropriately.
The synthesis procedure is carried out using the mathematical procedures, and all
the unknown dimensions of the mechanism are extracted.
Using the output of the MATLAB code, a 2D planar model and a 3D kinematic
model are developed and simulated using SolidWorks and MSC ADAMS. The assem-
bled mechanism is checked against order defect and circuit defect. During simulation,
it is observed that the mechanism is moving to all three stages within a maximum
positional error value of ± 1% of the chord length of flap positions and ±1° for flap
angle.
The assembled model is functioning satisfactorily. The flap and aft-flap positions
are achieved as synthesized and as configured in the CAD model. Thus, the novel
kinematics synthesis procedure developed for the mechanism is validated.
6 Conclusion
Acknowledgements The authors would like to thank Director, CSIR-NAL, and Structural Tech-
nologies Division (STTD), NAL. Also the authors would like to acknowledge Mr. Mahesh Kadam,
Senior Scientist, NAL for his support in realizing the 3D printed prototype models.
References
1. Rudolph PKC (1996) High-lift systems on commercial subsonic airliners. Contract A46374D
(LAS) September 1996 NASA Contractor Report 4746
2. Puulson JW Jr (1976) Wind tunnel investigation of a Fowler flap and spoiler for an advanced
general aviation wing. NASA Technical Note, NASA TN D-8236, June 1976
3. Shigley JE, Uicker JJ Jr (1981) Theory of machines and mechanisms. McGraw-Hill Book
Company
4. Erdman AG (1981) Three and four precision point kinematic synthesis of planar linkages. Mech
Mach Theory 16:227–245
5. Nafees KK, Mohammad A (2016) Dimensional synthesis of six-bar Stephenson II linkage for
fifteen precision points path generation. Perspect Sci 8:485–487
Intelligent Modeling of Dilution Percent
in Laser Surface Alloying
of Alx Cu0.5 FeNiTi High Entropy Alloy
Abstract Laser surface alloying is one of the techniques to create high entropy
alloy (HEA). HEAs are multi-element solid solution alloys stabilized due to high
mixing entropy. These alloys are superior compared to the existing materials used
in combustion engines, gas turbine components, and medical implants. The dilution
percent is one of the important parameters in laser surface alloying. It has effects on
various properties of alloyed components. This paper presents artificial intelligence
(AI)-based intelligent model for dilution percent. AI-based model is also compared
with a response surface model (RSM). It is observed that RSM is adequate in the
modeling of laser surface alloying.
1 Introduction
are essential parameters in LSA [2–5]. The effect of input process parameters can
be realized on the quality parameters. Dilution percent is one of the most important
quality parameters in LSA. Dilution percent is a measure of the extent of mixing
between alloy and substrate materials. In laser LSA of a multi-component system,
proper mixing of laser alloyed layer and substrate is essential for forming HEA,
good bonding, and modification of surface characteristics. Harooni et al. [6], during
the cladding of Zr on zirconium alloy, observed that with an increase in energy
per unit area, there was an increase in dilution percent. Morville et al. [7] used
COMSOL Multiphysics software to construct a 2D model of laser metal deposition.
They demonstrated that dilution percent increased with an increase in laser power and
a decrease in scan speed and powder feed rate. Recently, Zhai et al. [8] observed that
the dilution percent decreases with the application of an electromagnetic compound
field during laser cladding.
From the above discussion, it is clear that dilution percent is a critical prop-
erty controlled during laser material processing. A comparative study of developed
models in the case of LSA has not been done. This paper discusses the processing of
HEA (Alx Cu0.5 FeNiTi) by LSA. Analytical models using RSM and ANN have been
developed. A comparison between these developed models and experimental values
is also made.
2 Experimental Method
Fiber Optic
Laser Source
CNC Controller
Fig. 1 Laser surface alloying setup (YLS-2000; IPG Photonics Company) available at LMP
Division, RRCAT, Indore
An artificial neural network (ANN) can help calculate the output of a multifaceted
system by network training via trial data. ANN can be a suitable option for modeling
manufacturing systems where no analytical model having high suitability can be
Intelligent Modeling of Dilution Percent in Laser Surface … 459
developed [10]. The essential advantage of using neural networks is that it is not
constrained by the preset mathematical relation among variables of interest. Thus,
ANN is capable of modeling any random complex nonlinear link [11]. In this
network, three inputs and one output are modeled with a single hidden layer having
four neurons. We used an experimentally observed dataset of 15 values. These trial
data were distributed for training, test, and validation in the ratio of 70:15:15. The
mean square error method is used to train the network, and the feed-forward neural
network with the back-propagation (Leven–Marquardt) algorithm is used to forecast
the output. For the developed model, R-value is close to the target. Hence, model fits
the data. Table 2 presents the modeled values of dilution percent (Fig. 2).
The developed RSM and ANN models are compared with obtained experimental
values, and it is observed that the RSM model is more precise in the prediction of
dilution percent. Table 3 shows the obtained average values and error percent in both
models.
Figure 3 shows the comparison of experimental, RSM, and ANN predicted results
for dilution percent. The predicted average responses for dilution percent, calculated
using RSM and ANN, are presented in Table 3. Based on the results of Table 3,
the comparison shows that the average percentage of prediction errors for dilution
percent is 0.013 and 1.09 for RSM and ANN models, respectively. Thus, in our study,
RSM model outperforms the ANN model.
42
Experimental
40 RSM
Neural Network
38
Dilution percent
36
34
32
30
28
26
0 2 4 6 8 10 12 14 16
Experiment Number
Fig. 3 Variation of dilution percent for experimental, RSM, and neural network
4 Conclusions
In this work, we have reported the laser surface alloying of high entropy alloy
(Alx Cu0.5 FeNiTi). The obtained experimental observation for dilution percent is used
to fit the RSM and ANN model. Comparing trial outcomes displays that the average
percentage of prediction errors for dilution percent is 0.013 and 1.09 for RSM and
ANN models, respectively. The results indicate higher accuracy in the RSM model
than the ANN model for this experiment.
References
1. Qiu XW, Liu CG (2013) Microstructure and properties of Al2 CrFeCoCuTiNix high-entropy
alloys prepared by laser cladding. J Alloy Compd 553:216–220
2. Gopinath M, Thota P, Nath AK (2019) Role of molten pool thermo cycle in laser surface
alloying of AISI 1020 steel with in-situ synthesized TiN. Surf Coat Technol 362:150–166
3. Bax B, Rajput R, Kellet R, Reisacher M (2018) Systematic evaluation of process parameter
maps for laser cladding and directed energy deposition. Addit Manuf 21:487–494
4. Nabhani M, Razavi RS, Barekat M (2018) An empirical-statistical model for laser cladding of
Ti-6Al-4V powder on Ti-6Al-4V substrate. Opt Laser Technol 100:265–271
5. Erfanmanesh M, Pour HA, Semnani HM (2017) An empirical-statistical model for laser
cladding of WC-12Co powder on AISI 321 stainless steel. Opt Laser Technol 97:180–186
Intelligent Modeling of Dilution Percent in Laser Surface … 461
6. Harooni A, Nasiri AM, Gerlich AP, Khajepour A, Khalifa A, King JM (2016) Processing
window development for laser cladding of zirconium on zirconium alloy. J Mater Process
Technol 230:263–271
7. Morville S, Carin M, Carron D, Le Masson P, Gharbi M, Peyre P, Fabbro R (2012) Numerical
modeling of powder flow during coaxial laser direct metal deposition—comparison between
Ti-6Al-4V alloy and stainless steel. In: COMSOL conference
8. Zhai L, Ban C, Zhang J, Yao X (2019) Characteristics of dilution and microstructure in laser
cladding Ni-Cr B-Si coating assisted by electromagnetic compound field. Mater. Lett. 243:195–
198
9. Siddiqui AA, Dubey AK, Paul CP (2019) Study of geometrical characteristics in laser surface
alloying of high entropy alloy. Lasers Eng. 43:237–259
10. Fausett L (1994) Fundamentals of neural networks: architectures, algorithms, and applications
11. White H (1989) Learning in artificial neural networks: a statistical perspective. Neural Comput
1:425–464
Spinal Needles Insertion and Traversal
Based on Fiber Bragg Gratings—From
Conceptual Approach to Prototype
Development
1 Introduction
S. Ambastha (B)
CSIR-Central Mechanical Engineering Research Institute (CMERI), Durgapur, India
S. Umesh
ISRO-Laboratory for Electro-Optics Systems (LEOS), Bengaluru, India
S. Ambastha · S. Umesh · S. Asokan
Indian Institute of Science, Bengaluru, India
for handling the needle for its traversal, as compared to the conventional method of
direct hold of the needle by hand, and also provides real-time positioning informa-
tion by measurement of force on the needle based on variation in compressive force
due to stiffness variation of the tissues. The main difficulty faced in compressive
force measurement on the needle is the superimposition of additional axial force due
to bending of needle. The present design helps in its negation and reflects only the
longitudinal force.
In consultation with medical practitioners working in orthopedic surgical field, the
preliminary conceptual design has been further enhanced for prototype development.
The paper presents the limitations of preliminary design and differences in design
aspects between preliminary and prototype developed with feedback from medical
expertise.
An optical fiber-based grating sensor has been used in the present work, where the
working is in the form of Bragg wavelength shift due to application of strain on
the fiber Bragg gratings (FBG) sensor experienced as a compressive force reflecting
the needle traversal. The FBG acts as a reflective mirror to a broadband source of
light, launched into an optical fiber, that reflects back a narrow band of wavelength
satisfying Bragg’s condition, and transmits the rest. Reflected Bragg wavelength (λB )
is given as
λ B = 2 n eff (1)
where neff is the effective refractive index of the fundamental mode of the core of the
fiber and is the periodicity of the grating. The FBG used in the work have been
fabricated in-house using phase mask technique [4, 5].
The conceptual mechanism proposes real-time monitoring of force experienced
by a spinal needle during its penetration through different tissue layers of varying
stiffness using fiber Bragg grating (FBG) sensor. A preliminary design for a handheld
device is developed which has capability of mounting spinal needle and can be used
for direct skin puncturing in a normal clinical way.
The preliminary conceptual design for the device in its simplest form consisted
of a sensing wire on the surface of which a FBG of gauge length 3 mm is bonded.
The sensing wire has been positioned between two T-shaped aluminum parts with
head and tail ends, consisting of slot to hold spinal needle on its top and a hole
to route the communicating optical fiber consisting the FBG for data acquisition
as indicated in the inner design schematic in Fig. 1a. The entire sensing setup has
been enclosed inside a Teflon hollow tube of inner diameter maintained same as
that of the maximum diameter of the alumunium parts and is closed on both the
ends with provisions for the needle protrusion and optical fiber routing as shown
schematically in Fig. 1b. This entire mechanism has been established in order to
Spinal Needles Insertion and Traversal Based on Fiber Bragg … 465
Fig. 1 a Inner schematic of conceptual sensing design and b complete device schematic
ensure one dimensional movement of the sensing wire which has been essentially
required to negate the bending effect on the needle. The partwise pictorial view of
the preliminary device developed based on the conceptual design, making the setup
a handheld device, is shown in Fig. 2, where the components of the device such as
the hollow Teflon tube with two end caps the sensing wire bonded with FBG sensor
and placed between two alumunium parts along with the spinal needle are shown.
The preliminary device developed has been subjected for a force calibration test on
a Micro Universal Testing Machine (UTM) as shown in Fig. 3, with application of
continuous load in the range of 0–10 N. The device has been placed vertically with a
support, and the continuous load is applied through programmed vertical movement
of Micro UTM at 1 mm/min of constant displacement rate. The simultaneous UTM
response and the response of the FBG sensor bonded over the device are acquired
and recorded with a micron optics interrogator (MOI) (SM 130–700) having 1 pm
resolution for measuring Bragg wavelength shift.
466 S. Ambastha et al.
Fig. 3 Calibration of
preliminary device
The response of the preliminary device bonded with FBG corresponding to the
Micro UTM response is found to be in good agreement as shown in Fig. 4a, and the
comparison data obtained from both the devices indicates a linear response between
the FBG wavelength shift and force on Micro UTM, with a correlation coefficient
of 0.99 as in Fig. 4b. The sensitivity of the developed device with FBG sensor is
obtained from the slope of the calibration curve (−28.5 pm wavelength shift due to
compressive force) as 28.5 pm per 1 N of applied force and the resolution is computed
to be 0.035 N.
As a preliminary feasibility test of the conceptual design, the experimental trials have
been performed on commercially available 200 g fresh meat piece contained firmly
inside a transparent cubic plastic box of 70 mm and kept open on one side for needle
insertion. The device is held in the hands by the performer as in Fig. 5 and the spinal
needle mounted on the device is penetrated to establish force monitoring with tissue
puncture. As the needle is punctured inside the specimen due to the tissue stiffness,
it experiences force during its initial puncture, and further continuous penetration of
the needle inside the tissue is performed. The continuous force experienced by the
needle gets translated into strain variation on the FBG, monitored and recorded on
the MOI.
The complete experimental procedure for needle puncture and penetration on the
specimen from initial layer till its end is conducted for nearly 23 s and the continuous
real-time force monitored by the FBG sensor is plotted as shown in Fig. 6, where the
entire procedure has been segmented into three phases.
Phase 1 is the initial phase where the needle is not in contact with the tissue and is
kept in air, thus no force is recorded. The initialization of the experiment contributes
Spinal Needles Insertion and Traversal Based on Fiber Bragg … 467
Fig. 5 Experimental
procedure on fresh meat
tissue
468 S. Ambastha et al.
Phase 2 of the activity where the initial force is recorded as the highest of 2.15 N due
to the sudden interaction of the needle from air to a stiffer material (tissue). As soon
as the first interaction with the tissue is over, the force on the needle reduces and
further penetration of the needle leads to a continuous increase in the force on it due
to the uniform consistency in stiffness of the tissue. At the point the needle hits the
boundary walls of the transparent box, a peak in the force is recorded due to change
in stiffness from tissue to the box material and further the force is reduced due to
loss of needle–tissue interaction and end of further needle penetration contributing
to the third phase of the experiment.
Limitations and Prototype Development. The preliminary design as shown in
Figs. 1 and 2 has been further modified for the application, based on the consultation
and feedback received from orthopedic surgeon. As per the suggestions received, the
working principle of the preliminary device was apt for the application, rather the
material selected for device development and the dimensions were unsuitable for a
handheld operation due to its larger weight and size, which would cause discomfort
for a practitioner, who have to modify the procedure from their conventional way
of direct needle holding to holding a device mounted with the needle. Other critical
feedback was the limitation in the preliminary design for removal or interchange-
ability of the spinal needle during the ongoing procedure without removal of the
entire device.
This leads to the development of a prototype for the application with stainless
steel material and compact dimension providing reduced weight for ease in handling.
Change in design was also considered for providing flexibility of online interchange-
ability or removal of the needle with use of a separate spinal needle fixture mounting
spinal needle and attached externally with the sensing device [6, 7], as shown in Fig. 7.
Spinal Needles Insertion and Traversal Based on Fiber Bragg … 469
The device has been further used for online needle–tissue interaction force study on
human cadaver specimen [6] and further for demonstrating its interchangeability
capability and facility to perform comparative study on force monitoring [7].
4 Conclusion
Acknowledgements The authors are thankful to Dr. Sundaresh Dabir, Orthopedic Surgeon,
MSRUAS for the valuable suggestions and feedback received and Mr. Sajith Somanath, L&T Tech.
Services for help received in performance of the work. First author acknowledges CSIR funded
project MLP 223512 WP II for research submission.
470 S. Ambastha et al.
References
1. Ng KW, Goh JQ, Foo SL, Ting PH, Lee TK (2013) Needle insertion forces studies for optimal
surgical modeling. Int J Biosci Biochem Bioinform 3(3):187–191
2. Okamura AM, Simone C, Mark D (2004) Force modeling for needle insertion into soft tissue.
IEEE Trans Biomed Eng 10(51):1707–1716
3. Gorman P, Krummel T, Webster R, Smith M, Hutchens D (2000) A prototype haptic lumbar
puncture simulator. In: Medicine meets virtual reality. IOS Press, Amsterdam, pp 106–109
4. Othonos A (1997) Fiber Bragg gratings. Rev Sci Instrum 68:4309–4341
5. Kashyap R (1999) Fiber Bragg gratings. Academic, San Diego
6. Ambastha S, Umesh S, Dabir S, Asokan S (2016) Spinal needle force monitoring during lumbar
puncture using Fiber Bragg grating force device. SPIE J Biomed Opt 21(11):117002
7. Ambastha S, Umesh S, Dabir S, Asokan S (2018) Comparison of force required for lumbar
puncture with different gauges of spinal needle using Fiber Bragg grating force device. IEEE
Sens J 18(19):8028–8033
Kinematic and Dynamic Analysis
of Sliding Door Operating Mechanism
for Internal Weapon Bay
Abstract The paper presents a kinematic and dynamic analysis of a sliding door
operating mechanism used in the internal weapon bay of fighter aircraft it is an
enclosure used to store, hold and drop weapons. While ejecting the weapons, doors
of the internal weapon bay are required to be opened and closed after dropping
of weapons in a short period of time. Traditionally, the hinged door configuration
with actuators is used which protrudes outside, thus experiencing the drag force.
In the present study, an innovative sliding door operating mechanism is devised
such that doors slide within the internal weapon bay and are minimally exposed
outside. The kinematic analysis of the mechanism has been carried out analytically
and calculated results for displacement, velocity, and acceleration are compared with
SOLIDWORKS motion software results.
1 Introduction
The internal weapon bay is a rectangular enclosure in the fighter aircraft belly having
doors at the bottom. Sliding door operating mechanism is mounted at front and rear
end of the internal weapon bay. Doors in partial open conditions are shown in Fig. 1.
The mechanism comprises guide rails, actuators, connecting links, and door panels
[1–3].
Guide rails are mounted on the four corners of the internal weapon bay compart-
ment. One end of the door panel is connected with the connecting link and other with
guide rail block. Mechanism is operated via two actuators. In the present configura-
tion for open the door both actuators are retracted. Doors in fully open and closed
conditions are shown in Fig. 2a, b respectively.
The nomenclature of different links of the mechanism is shown in Fig. 3a. Consider
link—a is connecting link, link-b is door link, link-c is actuator link, B is slider of
guide rail. O is the pivot point of connecting link, ‘P’ is the pivot point of actuator
link, h is the fixed distance between actuator pivot and slider path line. ‘e’ is the fixed
distance between connecting link pivot and slider path line. ‘θ ’ is the angle made by
connecting link with horizontal line passing through pivot ‘O’ and ‘β’ is the angle
made by door link with the horizontal line passing through joint ‘A’, d is the distance
traveled by the slider ‘B’ in Y direction. Line diagram of the mechanism is shown in
Fig. 3b [4 and 5].
Kinematic analysis is carried out using the Solid Works software. It is assumed that
the Link ‘a’ (connecting link) is an input link and it follows the motion profile as
shown in Fig. 4. As the doors are to be opened within time period of 1 s. The Link ‘a’
of the mechanism is rotated from initial angle of 11.50 deg to final angle of 139.50
deg, hence the total angular displacement of link ‘a’ in 1 s is 128 deg.
−bαb sin β cos β − bωb2 cos2 β + aαa sin θ cos β + aωa2 cos θ cos β = 0
bαb cos β sin β − bωb2 sin2 β − aαa cos θ sin β + aωa2 sin θ sin β = d̈ sin β
Now adding,
Therefore,
From (7),
Linear displacement, velocity, and acceleration of the point-B (slider) with respect
to pivot point ‘O’ is shown in Fig. 5. Total linear displacement of slider is 292 mm
and it follows straight line path in Y direction.
476 C. K. Waghmare et al.
a 300.0
AnalyƟcal
250.0
SoŌware
Displacement (mm)
200.0
150.0
100.0
50.0
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
-50.0 Time (sec)
b
1000.0
AnalyƟcal
900.0
SoŌware
800.0
700.0
Velocity (mm/sec)
600.0
500.0
400.0
300.0
200.0
100.0
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Time (sec)
c 15000.0
AnalyƟcal
12500.0
SoŌware
10000.0
7500.0
Acceleration (mm/sec )
2
5000.0
2500.0
0.0
-2500.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
-5000.0
-7500.0 Time (sec)
-10000.0
600 340
Linear Velocity
330
500 Linear Displacement 320
Velocity (mm/sec)
Displacement (mm)
400 310
300
300
290
200 280
270
100
260
0 250
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (sec)
To determine the actuator parameters, kinematic data of point ‘E’ i.e. joint between
actuator and door panel is prime important. Displacement and velocity plot of the
point ‘E’ with respect to point ‘P’ is shown in Fig. 6.
4 Conclusion
Above study can be further extended to perform the detailed time analysis and design
of joints and linkages of the mechanism.
478 C. K. Waghmare et al.
16000.0 340
Actuator force Point-E
14000.0 (newton)
Actuator Force (newton)
330
12000.0 Displacement Point-E (mm)
Displacement (mm)
320
10000.0
8000.0 310
6000.0 300
4000.0
290
2000.0
280
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
-2000.0 270
Time (sec)
References
Abstract For synthesizing kinematic chains of a given number of links and degree
of freedom, detection of isomorphism is essential. Two kinematic chains are called
isomorphic if they have equivalent topological configuration. The exertion for detec-
tion of isomorphism among planar kinematic chains is being researched for a long
time to avoid redundancy. In the past, researchers have used several methods like
characteristic polynomial-based approaches, Hamming number method, path and
distance-based techniques, and many more complex algorithms to detect isomor-
phism. In this work, an index for detection of isomorphism among planar kine-
matic chains is proposed by considering the distance matrix and degree of links
matrix. Kinematic chains are isomorphic, if they have same index else they are non-
isomorphic. Hence, to detect isomorphism among different kinematic chains, the
proposed technique is implemented successfully for 8-link single degree of freedom
and 9-link double degree of freedom kinematic chains.
1 Introduction
Structural synthesis and analysis of mechanisms are essential for the invention and
innovation of mechanisms. One key aspect of the structural synthesis is to develop all
possible arrays of kinematic chain and consequent mechanisms for a given number of
links and degrees of freedom. It provides freedom to the designer while selecting the
best or optimum mechanisms according to suitable application and need. In the course
of development of kinematic chain and mechanisms, duplication is prominent which
causes redundant effort and duplicate solutions. Kinematic chains are isomorphic
if they have the same topological organization. The topology of a kinematic chain
or mechanism portrays the information about which type of link is connected to
which other type links by what type of joint. In other words, two kinematic chains
or mechanisms are said to be isomorphic if there exists a one-to-one correspondence
among the links of one kinematic chain to that of the second kinematic chain. A
pair of 8-link kinematic chains shown in Fig. 1 are isomorphic. There is one-to-one
correspondence among the links, and this is given by 1 ≈ 4, 2 ≈ 5, 3 ≈ 6, 4 ≈ 7,
5 ≈ 8, 6 ≈ 3, 7 ≈ 2, 8 ≈ 1.
Study of kinematic chains and mechanism has fascinated the researchers for
decades. Over the past decades, much work on isomorphism in mechanism synthesis
has been reported in literature. The motivation behind these studies is the identifica-
tion of possible distinct mechanisms for a given degree of freedom and number of
links. One critical problem encountered during the synthesis of kinematic chains
is identification of possible structural equivalence or isomorphism among kine-
matic chains. There are several methods available in the literature for isomorphism
identification such as graphical methods based on visual inspection, methods based
on eigenvector of adjacency matrix, code-based approaches, character polynomial-
based approaches, Hamming number-based approaches, distance or path-based
approaches.
Crossley et al. [1], Freudenstein et al. [2], and Woo et al. [3] introduced graph
theory to study kinematic chains and followed inspection-based approach for isomor-
phism detection. Based on graph theory, Uicker et al. [4] first time proposed char-
acteristic polynomial of link-link adjacency matrix (CPAM) of a graph of the kine-
matic chain as an index of isomorphism detection. Later, Mruthyunjaya et al. [5] have
proven that CPAM-based approach failed to distinguish non-isomorphic graphs. They
proposed through matrix algebra that invariance of the characteristic polynomial is
necessary not sufficient condition for two graphs to be isomorphic. Characteristic
polynomial-based approaches were computationally efficient but failed in several
A Method to Detect Isomorphism in Planar Kinematic Chains 481
isomorphic tests. Again, in this approach, Dube et al. [6] proposed another method,
which is based on characteristic polynomial of distance matrix of a kinematic chain
for isomorphism detection. Apart from this, Ambekar et al. [7] developed another
isomorphism detection based on code-based approach. In this method, the adjacency
matrix of a pair of kinematic chain is transformed to canonical form and then their
code is compared. Later in another approach, Rao et al. [8] introduced the concept
of Hamming distance for isomorphism detection among kinematic chains. In the
link-link adjacency matrix, each row is called the hamming code of the link. The
number of digits by which two rows differ is called Hamming distance between links.
They defined a square symmetric Hamming distance matrix and proposed Hamming
value of a chain as an index for isomorphism. In link- and path-based approaches,
Yadav et al. [9] proposed link-link distance matrix for binary kinematic chains. They
assumed, the array of sum of rows in descending order as an index of isomorphism
for kinematic chains. Yadav et al. [10] have also proposed link–path code approach
based on number of paths of different length from a link to all other links. These
all discussed methods have their own shortcomings, like detection based on visual
inspection is tedious and can lead to error. In addition, most of the methods are
based on some complex algorithm. Therefore, the quest for reliable and computa-
tionally efficient methods is still on. In this proposed research work, link-link distance
matrix and degree of links are used as an index of isomorphism. The section below
describes the basic terminologies used in the proposed method, and the subsequent
section explains the method in detail.
Every part of machines, which is having some relative motion with respect to some
other part, is known as kinematic link or element. It is necessary for the link to be a
resistant body for transmission of power and motion. Link is also a rigid connection
between two or more elements of different kinematic pairs. Based on this, there are
different types of like binary (connected to two other links), ternary (connected to
three other links), and quaternary links (connected to four other links). The connection
between two links is always a joint or pair (turning pair, sliding pair, rolling pair, screw
pair, spherical pair, etc.) but this pair will be said kinematic pair if the relative motion
between the links is completely constrained. A kinematic chain is a combination of
kinematic pairs in which each element or link forms part of two pairs and in which the
relative motion is completely constrained. Mechanisms are an assemblage of resistant
bodies, coupled by movable joints, to form a closed kinematic chain with one link
fixed (frame or base link) and having the purpose of transmitting or transforming
motion. The mobility or degree of freedom of a mechanism is minimum number of
independent variables required to define the position or motion of the system. Based
on this, binary, ternary, quaternary, and quenary links have degree of freedom two,
three, four, and five, respectively. Connectivity pattern of a kinematic chain can be
characterized using adjacency matrix and distance matrices. Adjacency matrix (A)
482 A. Dwivedi et al.
In this matrix (A), always aii = 0, since a link-i cannot be connected to itself;
ai j = 1 implies link-i is directly connected to link-j and ai j = 0 implies link-i is not
directly connected to link-j.
Also, in a closed kinematic chain, every link has a distinct association with another
link in form of a distance (shortest) between them, which is constant. It can be
represented in a matrix form called distance matrix (D).
Figure 2 shows a four-bar kinematic chain. Therefore by following the above
definitions, adjacency and distance matrices can be written as:
⎡ ⎤
0101
⎢1 0 1 0⎥
Adjacency Matrix (A) = ⎢ ⎥
⎣ 0 1 0 1 ⎦.
1010
⎡ ⎤
0 121
⎢1 0 1 2⎥
Distance Matrix (D) = ⎢
⎣2
⎥.
1 0 1⎦
1 210
This method is based on distance matrix and the degree of links matrix for a kinematic
chain. These two matrices determine an index of isomorphism for kinematic chains
known as kinematic chain label (KCL). The method can be summarized in following
steps:
i. Representation of Kinematic Chains (KC): The given kinematic chain is
represented with proper numbering of links.
A Method to Detect Isomorphism in Planar Kinematic Chains 483
ii. Distance Matrix Formation (D): Distance matrix D is written using the
definition for a given kinematic chain.
iii. Degree of links Matrix Formation (L): Degree of links matrix L is written in
form a column matrix using degree of links in a given kinematic chain.
iv. Labeling of Links and Chain: A matrix V is defined such that:
V = D×L (2)
where elements of matrix ‘V’ (which will always be a column matrix) represent
label of links in kinematic chain and column matrix V represents kinematic
chain label (KCL).
v. Detection of Isomorphism using KCL: The elements of matrix ‘V’ are
arranged in ascending or descending order. This sequence is called kinematic
chain label (KCL) and is used as an index of isomorphism. If KCL of two
kinematic chains is same, then kinematic chains are isomorphic otherwise
non-isomorphic.
Table 1 illustrates the indexing methodology to determine KCL using the proposed
methodology for four-bar kinematic chain.
Step 2: Distance and degree of links matrix Step 4: Detection of Isomorphism using KCL
formation So here for four-bar kinematic chain
⎡ ⎤ ⎡ ⎤ KCL = 8 8 8 8
0121 2
⎢ ⎥ ⎢ ⎥ Elements of matrix represent label of links of
⎢1 0 1 2⎥ ⎢2⎥
D=⎢ ⎢
⎥ and
⎥ L=⎢ ⎢ ⎥
⎥ four-bar kinematic chain and matrix itself is
⎣2 1 0 1⎦ ⎣2⎦ KCL of four-bar kinematic chain
1210 2
Distaance Matrix Degree of links matrix
484 A. Dwivedi et al.
KC representation KC representation
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
012 321 2 012 211 3
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ 101 232⎥ ⎢ 2⎥ ⎢ 101 222⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ 210 121⎥ ⎢ 3⎥ ⎢ 210 121⎥ ⎢ 3⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
D=⎢ ⎥ and L=⎢ ⎥ D=⎢ ⎥ and L=⎢ ⎥
⎢ 321 012⎥ ⎢ 2⎥ ⎢ 221 012⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ 232 101⎦ ⎣ 2⎦ ⎣ 122 102⎦ ⎣ 2⎦
121 210 3 121 220 2
Distance matrix Degree of links matrix Distance matrix Degree of links matrix
KCL = 15 15 21 21 21 21 KCL = 16 16 18 18 19 19
Table 2 illustrates the indexing methodology to determine KCL using the proposed
methodology for two given kinematic chains. The distance matrix (D) is multiplied
with the degree of links matrix (L) to obtain the column matrix V for both Watt and
Stephenson chain. The elements of matrix V are subsequently sorted in ascending
order to obtain the KCL corresponding to both the kinematic chains, which shows
that Watt and Stephenson chains are non-isomorphic as the KCL of both the chains
are not equivalent.
KC representation KC representation
⎡ ⎤ ⎡ ⎤
01 22 21 21 01 21 22 12
⎢ ⎥ ⎢ ⎥
⎢ 10 12 32 32 ⎥ ⎢ 10 12 33 22 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 21 01 23 32 ⎥ ⎢ 21 01 23 21 ⎥
⎢ ⎥ ⎢ ⎥
⎢ 12 21 ⎥ ⎢ 12 22 ⎥
⎢ 22 10 ⎥ ⎢ 12 10 ⎥
D=⎢ ⎥ and D=⎢ ⎥ and
⎢ 23 21 01 22 ⎥ ⎢ 23 21 01 23 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 12 32 10 12 ⎥ ⎢ 23 32 10 12 ⎥
⎢ ⎥ ⎢ ⎥
⎢ 21 01 ⎥ ⎢ 21 01 ⎥
⎣ 23 32 ⎦ ⎣ 12 22 ⎦
12 21 22 10 22 12 32 10
Distance Matrix Distance Matrix
⎡ ⎤ ⎡ ⎤
3 3
⎢ ⎥ ⎢ ⎥
⎢ 2 ⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 2 ⎥ ⎢ 3⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 3⎥
⎢ 3 ⎥ ⎢ ⎥
L=⎢ ⎥ L=⎢ ⎥
⎢ 2 ⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 3 ⎥ ⎢ 2⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 3⎥
⎣ 2 ⎦ ⎣ ⎦
3 2
Degree of links matrix Degree of links matrix
KCL = 26 26 27 29 32 34 35 36 KCL = 26 26 27 29 32 34 35 36
Fig. 3 Kinematic chain (KC) representation of possible configurations of 8-link single degree of
freedom chains
Similarly, another example which considers 9-link double degree of freedom kine-
matic chains is illustrated in Fig. 4, and the corresponding KCL is listed in Table
5.
From above results, it is obvious that all the 9-link double degree of freedom
possible kinematic chains are non-isomorphic as KCL of all chains not equivalent.
5 Conclusion
In this research work, distance matrix and degree of links matrix, which are the
fundamental techniques to characterize the topology of a kinematic chain, are taken
as an index of isomorphism for planar kinematic chains. Starting from the elementary
four-bar kinematic chain, various mechanisms like Watt, Stephenson, and isomorphic
chains (a pair of 8-link single degree of freedom) are analyzed using the proposed
method. This method has also been implemented on 8-link single degree of freedom
and 9-link double degree of freedom kinematic chains. The method successfully
distinguishes all non-isomorphic chains. If we are taking adjacency matrix and degree
of links matrix as an index of isomorphism, then the method fails to distinguish
possible non-isomorphic kinematic chains. Using the proposed method, topological
configuration of kinematic chains can be easily verified to avoid redundancy, while
synthesizing kinematic chain for a given number of links and degree of freedom.
Additionally, the method can also be made computationally efficient by sponta-
neously generating distance matrices using techniques like graph theory in python.
Hence, through the proposed index of isomorphism (KCL), which has very simple
algorithm; structural depiction of a kinematic chain can be repossessed.
488 A. Dwivedi et al.
Fig. 4 Kinematic chain (KC) representation of possible configurations of 9-link double degree of
freedom chains
A Method to Detect Isomorphism in Planar Kinematic Chains 489
References
1. Crossley FRE (1965) The permutations of kinematic chains of eight member or less from the
graph-theoretic viewpoint. Develop Theor Appl Mech 2:467–486
2. Freudenstein F, Dobrjanskyj L (1966) On a theory for the type synthesis of mechanisms. In:
Applied mechanics. Springer, Berlin, pp 420–428
3. Woo LS (1967) Type synthesis of plane linkages. J Eng Ind 89:159–170
4. Uicker JJ Jr, Raicu A (1975) A method for the identification and recognition of equivalence of
kinematic chains. Mech Mach Theory 10:375–383
5. Mruthyunjaya TS, Raghavan MR (1979) Structural analysis of kinematic chains and mecha-
nisms based on matrix representation. J Mech Des 101:488–494
6. Dube RK, Rao AC (2013) New characteristic polynomial-a reliable index to detect isomorphism
between kinematic chains. J Indian Inst Sci 66:227
7. Ambekar AG, Agrawal VP (1987) Canonical numbering of kinematic chains and isomorphism
problem: min code. Mech Mach Theory 22:453–461
8. Rao AC (1988) Comparison of plane and spatial kinematic chains for structural-error
performance using pseudo hamming distances. Indian J Technol 26:155–160
9. Yadav JN, Pratap CR, Agrawal VP (1996) Computer aided detection of isomorphism among
binary chains using the link-link multiplicity distance concept. Mech Mach Theory 31:873–877
10. Yadav JN, Pratap CR, Agrawal VP (1996) Mechanisms of a kinematic chain and the degree of
structural similarity based on the concept of link-path code. Mech Mach Theory 31:865–871
11. Mruthyunjaya TS (2003) Kinematic structure of mechanisms revisited. Mech Mach Theory
38:279–320
12. Mruthyunjaya TS, Balasubramanian HR (1987) In quest of a reliable and efficient computa-
tional test for detection of isomorphism in kinematic chains. Mech Mach Theory 22:131–139
Generation of Coupler Curves for Planar
Kinematic Chains Using Link Joint
Equations
Abstract Analysis of four bar kinematic chain for generation of its coupler curves
and functions generation is well researched. However, limited amount of work is
reported for higher link mechanisms. In the present work, an attempt has been made
to propose a simple method to generate coupler curves of kinematic chains. Present
method is based on identifying the joint locations of all fixed and moving joint. Simul-
taneous equations of quadratic nature are proposed for each and every moving links
to provide dimensional constraints with respect to the fixed link. MATLAB—math-
ematical software—has been used to solve the non-linear simultaneous equations.
Analysis of the cases referred are carried out and coupler curves are generated for four
bar mechanism, all five inversions of six link mechanism and eight link mechanism.
Keywords Kinematic chains · Fixed and moving joints · Link joint quadratic
equations · Relative angles · Coupler curves
Nomenclature
1 Introduction
Kinematic chains made up of links and joints are used to generate constrained relative
motion. To fulfill the above objective, one of the links of the kinematic chain is fixed.
Generation of coupler curve is required to know motion of links. By fixing various
links of a kinematic chain various coupler paths for different floating links can be
generated.
Researchers have proposed methods graphical and analytical to generate coupler
curves for planar kinematic chain. First notable contribution is of Hrones and Nelson
[1], who have given an atlas of four bar coupler curves containing approximately
7300 curves for engineers to use. Much work has been reported on four bar coupler
curves. Robert- Chebyshev’s theorem [2] for four bar linkages gave generation to
three different four bar cognate linkages. Dijksman [3, 4] has generated symmet-
rical coupler curves for Watt-I kinematic chain. Chatterjee and Mallik [5] have given
formulation for position error of a four bar coupler curve as function of crank angles,
link length and errors in the link length. Various geometric properties of four bar
coupler curves are analyzed by Wang and Xiao [6]. Condition for coupler curves with
different characteristics, such as cusp, inflection point, and crunodes. are obtained by
them. Existence of complementary curve governing the existence of multiple points
on the coupler curves of four (hinged) linkages is given by Hernandez et al. [7].
Dhingra et al. [8] have given closed-form displacement analysis of planar multi-loop
mechanisms along with equation for coupler curves using Grobner bases. Numerical
examples for 8, 10 and 12 link mechanism are given by them. Lan et al. [9] divided
the coupler plane into crank-zone, rocker-zone, coupler-zone and cross-zone. Shapes
and features of coupler curves in different zone are shown by them. Chung [9] charac-
terized the coupler curves further in terms of its order and circularity. Nagarajan and
Bandyopadhyay [10] have given a meta-programming language MML for modeling
and analysis of mechanism. Method proposed is based on generating and solving
loop closure equations of the mechanism. A recent work by Bai and Angeles [11]
has given a method for path synthesis of four bar linkages with smaller system of
equations of lower complexity.
Methods proposed by earlier researchers either do not include direct determination
of coupler curves of different floating links of kinematic chain having more number of
links or are computationally difficult. Additionally, any method including all joints
coordinate values of a kinematic chain will be better for identification of coupler
curves. In view of the above, a computationally simple and reliable method based
on link joint equations is proposed to generate coupler curves of a kinematic chain.
Method incorporates structural details of the kinematic chain up to the last level of
connectivity in form of coordinate values of joints. The method is applied successfully
to four, six and eight bar planar kinematic chains.
Generation of Coupler Curves for Planar Kinematic Chains … 493
2 Method
Present method is based on identifying the joint locations, i.e., of fixed joints and
moving joints—moving known joint positions and moving unknown joint position.
To determine the unknown joint positions quadratic equations for each and every
moving links are generated so as to provide dimensional constraints with respect to
the fixed link for the particular mechanism.
The proposed method is used for generating coupler curves of planar kinematic
chains. The method is introduced through its application to Stephenson’s-II six link
planar kinematic chain, Fig. 1. Stephenson’s chain has seven joints (j1 , j2 , …, j7 ),
four binary link l17 , l 25 , l 34 and l 67 and two ternary links l 123 and l 456.
A. Fixed Joints: Joints in the fixed links are identified as fixed Joints. In the present
example of Stephenson’s-II six link kinematic chain Joints j1 and j7 are fixed
joint. Since L 17 is chosen as fixed links ternary l 123 or binary link l 6 7 can be as
input or output link.
B. Moving Joint: Joints other than fixed joints are moving joints for example in
Fig. 1 Stephenson’s-II six link kinematic chain joints j2 , j3 , j4 , j5 and j6 are
moving joint.
Identification of input link and output link decides the position of known and
unknown moving joints.
B1. Known Moving Joints: Moving joints at one links distance connected
with input link is identified as known moving joints. In the present example
j2 , j3 are moving known joints since they are connected with ternary link.
B2. Unknown Moving Joints: All other moving joints, which are not a
constituent member of input link, are unknown moving joints. In the present
example j4 , j5 , j6 are unknown moving joints.
C. Link Joint Equations: To determine, unknown moving joint position
constraints in the form of fixed link length are introduced as quadratic equa-
tions. Depending upon the number of unknown moving joints these equations
will be formulated. In the present example six such equations, Eqs. (1)–(6) are
generated.
2
( j2x − j5x )2 + j2y − j5y = (l25 )2 (1)
2
( j3x − j4x )2 + j3y − j4y = (l34 )2 (2)
2
( j6x − j7x )2 + j6y − j7y = (l67 )2 (3)
2
( j4x − j6x )2 + j4y − j6y = (l46 )2 (4)
2
( j4x − j5x )2 + j4y − j5y = (l45 )2 (5)
2
( j5x − j6x )2 + j5y − j6y = (l56 )2 (6)
Fig. 2 Coupler curves of all floating links of Stephenson’s-II six link kinematic chain
Above method has been explained with Stephenson’s-II six link mechanism and is
applied on all five inversions of six link mechanisms, i.e., Watt-I & Watt-II, and
Stephenson’s-I & Stephenson’s-III. To verify the results, method is also applied to
generate Watt-I symmetrical coupler curves, generated by Dijksman [1, 2]. Method
is also extended to eight link chains. Results so obtained are shown through images
as earlier start from four links, six links and eight links kinematic chains. All cases of
six link chain though solved are not shown in the results are available with authors.
The method is applied to four link planar kinematic chain, open and cross
configuration both. The results obtained are shown through images, Fig. 3.
Fig. 3 Coupler curves of floating link of four bar open and cross kinematic chain
Generation of Coupler Curves for Planar Kinematic Chains … 497
Image three in Fig. 3 shows coupler curves of l 23 , while image six in Fig. 3 shows
the coupler curves of flipped four bar chain and can be concluded that “though there
is flip condition in four bar chain, there will be no change is shape and size of coupler
curves, with only change in orientation of coupler curves.”
The method is applied to six link planar kinematic chains Watt-I & Watt-II.
B.1 W-I Kinematic Chain
Figure 4 includes seven images. The first image shows the Watt-I six link mech-
anism in which l12 is fixed link, l23 as input link, l147 as output link and remaining
links are floating links. Remaining images are showing the path of different joints
and intermediate points between the joints on link.
Fig. 4 Coupler curves of all floating links of Watt-I six link kinematic chain
498 H. S. Yadav and S. Sanyal
Fig. 5 Coupler curves of all floating links of Watt-II six link kinematic chain
The method is also applied to generate symmetrical curves of Watt-I kinematic chain,
and obtained result is verified with the results of Dijksman [3, 4] (Fig. 6).
The method is also applied to eight link kinematic chains and the results in the form
of coupler curves have been obtained (Fig. 7).
Generation of Coupler Curves for Planar Kinematic Chains … 499
4 Conclusions
From the above results, it can be concluded that the method proposed is simple and
exhaustive in nature. Generated link joint equations can be formulated for higher
number of links mechanism. Potential of the method has been shown on few known
cases including six link inversions and eight links kinematic chains. The method has
the potential of generating coupler curves for the kinematic chains having more and
more number of links. The method is applied successfully to four, six and eight link
kinematic chains. The results are obtained for four; six and eight link mechanism
reflects the concepts given by earlier researchers.
500 H. S. Yadav and S. Sanyal
References
1. Hrones JA, Nelson GL (1951) Analysis of the four bar linkages. MIT Press and Wiley, New
York
2. Hartenberg RS, Denavit J (1964) Kinematics synthesis of linkages. TMH
3. Dijksman EA (1980) Highest-order Coupler points of Watt-I linkages, tracing symmetrical
6-bar curves. Mech Mach Theory 15:421–434
4. Dijksman EA (1984) An unsymmetrical Watt-I linkages generating a family of symmetrical
curves. Mech Mach Theory 19(3):297–306
5. Chatterjee GB, Mallik AK (1987) Mechanical error of a four-bar linkage Coupler curve. Mech
Mach Theory 22(1):85–88
6. Wang DL, Xiao DZ (1993) Distribution of Coupler curves for Crank-Rocker linkages. Mech
Mach Theory 28(5):671–684
7. Hernandez A, Amezua E, Ajuria MB, Llorente JI (1994) Multiple points on the Coupler curves
of transitional four-hinge planar linkages. Mech Mach Theory 29(7):1015–1032
Generation of Coupler Curves for Planar Kinematic Chains … 501
8. Dhingra AK, Almadi AN, Kohli D (2001) Closed-form displacement and Coupler curve
analysis of planar multi-loop mechanism using Grobner base. Mech Mach Theory 36:273–298
9. Lan Z, Zou H, Lu L (2002) Kinematic decomposition of Coupler plane and the study on the
formation and distribution of Coupler curves. Mech Mach Theory 37:115–126
10. Chung W-Y (2005) The characteristics of a Coupler curve. Mech Mach Theory 40:1099-1106
11. Nagarajan A, Bandyopadhyay S (2011) A framework for analysis and dynamic visualisation
of mechanisms. In: 15th National conference on machines and mechanisms (NaCoMM), p 190
12. Bai S, Angeles J (2015) Coupler-curve synthesis of four-bar linkages via a novel formulation.
Mech Mach Theory 94:177–187
Reference to a Book
1 Introduction
PEEK, PTFE and MoS2 are selected in powder form in this study because of their
invariant tribological properties. As known from the literature, PEEK has excellent
anti-wear properties, and PTFE has outstanding anti-friction properties [8]. MoS2
materials are known for their anti-wear properties, and they also enhance the mechan-
ical properties of the materials [10]. Natural PEEK 1000 rod (provided by Quadrant
Tribological Behaviour of Polymer-Based Composite … 505
EPP Surlon Pvt. Ltd., Ghaziabad, UP, India) with a density of about 1320 kg/m3
and PTFE materials having a density of 2200 kg/m3 has been used in this work.
The Particle size of these materials comes in micron. MoS2 powder (Sigma Aldrich,
<2 µm, 99%) with density is of about 5060 kg/m3 with melting point 648 K (2375 °C,
4307 °F) has been used. Individual powder mixed with acetone goes through stirring
for 15 min. After that, three solutions mixed simultaneously and leave for acetone
vapourization. These polymer-based composites are prepared by using compression
moulding (Flowmech Engineers Private Limited, New Delhi). These materials are
moulded in a cylindrical pin shape with 10 mm diameter and 20 mm length with
the help of a metal dye. Before the moulding process, this has to be cleaned perfectly
with acetone and then filled with the mixed composite powder. The whole setup was
kept under the machine to maintain a temperature up to 380 °C, without any pressure.
Once it touches the optimum temperature, a 100 Psi pressure was applied for 3–5 h.
After completion of these process, the shape and surface of this pin were maintained
with the help of grinding and emery paper.
The wear and friction behaviour of the composites are evaluated using pin-on-disc
configuration on a universal Tribometer (CSM Instruments, Switzerland). In this
process, the composite pins with 10 mm diameter are rubbed against the flat coun-
terface 316 SS disc (polished surface, 58 mm diameter). All the tests are conducted
in an unlubricated sliding contact at room temperature and relative humidity (RH)
45 ± 5%. In this case, different input parameters such as load, speed and time were
selected as per the requirement, which is described in the result section. Each test
was repeated three to four times to ensure a relevant statistical evaluation. Friction
and displacement of the pin material were recorded and represented in terms of mean
or average values. Transfer layer deposition on the disc surface and markings of wear
mechanisms on pin surface is also examined using an optical microscope of MCT
setup.
Fig. 1 Worn-out pin examined under an optical microscope, and the schematic of an elliptical
segment of worn-out surface boundary
face is BW , which can be easily measured with the help of an optical microscope,
represented in Fig. 1. A CAD model was developed with a similar pin, and the
elliptical wear-out section was cut from this CAD model pin. But the necessary
parameters required for wear-out plane angle was derived mathematically by using
L W and BW , which is described in the result section. Then the difference in mass and
volume of this wear-out CAD model has been calculated, and this value is related to
the wear of the materials for a different condition.
In this experimental study, PEEK-based composite filled with PTFE and inorganic
MoS2 has been taken under consideration. According to Hufenbach et al., PEEK
filled with PTFE showed better characteristics in both frictions and wear resistance
with a composition of 70 wt% of PEEK and 30 wt% of PTFE [6]. In this study,
a solid inorganic material named Molybdenum disulphide (MoS2 ) was mixed in
different wt% with the exact composition of PEEK and PTFE. This inorganic solid
lubricant was used to improve the tribological characteristics of the polymer. From
Zalaznik et al. observations on PEEK filled MoS2 with 2–10 wt%, the composition
with intermediate wt% of MoS2 showed the best behaviour of friction coefficient
and wear volume [10].
Therefore, considering previous observations, three different wt% of MoS2 with 3,
6, and 10 wt% were taken with PEEK-PTFE composition to develop a new polymer
composite. Although pin form of PEEK and PTFE is available commercially, to
have a better comparison, pure PEEK and PTFE pins were also manufactured by
the same compression moulding process. Table 1 presents the designation of newly
Tribological Behaviour of Polymer-Based Composite … 507
Table 1 Designation of
Materials composition Samples name
newly developed polymeric
materials PEEK_Commercial PEC
PTFE_Commercial PTC
PEEK_Manufactured PEM
PTFE_Manufactured PTM
PEEK (70%)–PTFE (30%) PPM703000
PEEK (68%)–PTFE (29%)–MoS2 (3%) PPM682903
PEEK (66%)–PTFE (28%)–MoS2 (6%) PPM662806
PEEK (63%)–PTFE (27%)–MoS2 (10%) PPM632710
developed materials. Each experiment is repeated three times with the same set of
input parameters to minimize error in results.
The friction and wear characteristics of the composites with a dissimilar proportion
of MoS2 was studied in pin-on-disc setup on tribometer at a load of 2, 6 and 10 N with
speed 50 cm/s for 30 min. In Fig. 2, the approximation of friction of these composites
is shown as a function of load. From this observation, the coefficient of friction of all
the composites reduces slowly with an increase in load. This is due to the formation
of tribo-film on the counter-surface with an increase in applied pressure.
The friction characteristics are observed to be comparable for the composite with
an increase in the percentage of MoS2. The coefficient of friction of PEEK is higher as
compared to the PTFE. The friction of PEEK and PTFE-based composite is retained
in between the COF of pure PEEK and PTFE. Based on the wt% of MoS2 , a signif-
icant difference in wear was studied at the concentration of 3 wt%, i.e. PPM68903.
In each loading case, the COF of this composite was reducing or almost equal to the
PTFE value. A further increase in the concentration of MoS2, the material resulted
in an undesirable friction value. An essential difference in wear properties of the
composites was observed in this research, shown in Fig. 3. The wear volume of every
composite and pure materials increases with an increase in applied load, which other
researchers also observe [4]. This variation in wear volume of the materials is because
of the formation of a thin layer of wear-out material, i.e. transfer film formation, which
is shown in Fig. 4. In this case, PEEK shows good wear behaviour as compared to
PTFE. PEEK-PTFE-based composite filled without MoS2 increase the wear char-
acteristics of the PTFE is almost 90%, i.e. the volumetric wear changes drastically
from 1 to 0.07 mm3 for 2 N load. Similar enhancement was observed in different
loading conditions. In this configuration, the wear characteristics at higher load such
as 10 N has lightly increased as compared to the pure PEEK material.
508 S. R. Guru et al.
0.6
Samples; PEM PTM PPM703000
PPM682903 PPM662806 PPM632710
0.5
Coefficient of friction
0.4
0.3
0.2
0.1
0.0
0 2 4 6 8 10 12
Load (N)
Fig. 2 The coefficient friction of all the moulded composited as a function of load
0.030 7
Samples; PEM PPM703000 PPM682903
PPM662806 PPM632710 PTM 6
0.025
5
Volumetric Wear (mm )
0.020
4
0.015 3
2
0.010
1
3
0.005
0
0.000 -1
0 2 4 6 8 10 12
Load (N)
Sliding Direc on
Tribological Behaviour of Polymer-Based Composite …
4 Conclusion
In this paper, the tribological properties of the PEEK-PTFE blend have been improved
by the inclusion of MoS2 . The effective wt% of MoS2 for PEEK-PTFE composite
is analysed with different loading condition. From this observation, three wt% of
MoS2 gives a satisfactory performance in the tribological field. The enhancement in
tribological properties of these materials occurs due to this reinforcement which is
nearly 90–100% as compared with pure PTFE and 30–40% as compared with pure
PEEK material. Thus, the lower wt% of MoS2 (i.e. 3%) is effective for on an excellent
designing material, further increase the wt% of MoS2 , it becomes unsatisfactory in
both friction and wears due to third body action. The improvement in wear and friction
of PEEK and PTFE occurs due to the transfer film formation of mixed particles.
References
1. Sawyer LC, Grubb DT, Meyers GF (2008) Polymer Microscopy. Springer Nature, Berlin.
https://doi.org/10.1007/978-0-387-72628-1
2. Samad A, Fareed MI (2016) Improving the friction and wear of poly-ether-etherketone (PEEK)
by using thin nano-composite coatings. Wear 365:154–162
3. Lu ZP, Friedrich K (1995) On sliding friction and wear of PEEK and its composites. Wear
183:624–631
4. Lal B, Alam S, Mathur GN (2007) Tribo-investigation on PTFE lubricated PEEK in harsh
operating conditions. Tribol Lett 25:71–77. https://doi.org/10.1007/s11249-006-9158-9
5. Stolarski TA (1986) The friction and wear of poly(tetrafluoroethylene)-poly(etheretherketon)
composites: an initial appraisal of the optimum composition*. Wear 108:357–374
6. Hufenbach W, Institutfur KK (2003) Sliding wear behaviour of PEEK-PTFE blends. J Synth
Lubr 20:227–240
7. Xie GY, Zhuang GS, Sui GX, Yang R (2010) Tribological behavior of PEEK/PTFE compos-
ites reinforced with potassium titanate whiskers. Wear 268:424–430. https://doi.org/10.1016/
j.wear.2009.08.032
8. Burris DL, Sawyer WG (2006) A low friction and ultra low wear rate PEEK/PTFE composite.
Wear 261:410–418. https://doi.org/10.1016/j.wear.2005.12.016
Tribological Behaviour of Polymer-Based Composite … 511
9. Mazza L, Trivella A, Grassi R, Malucelli G (2015) A comparison of the relative friction and
wear responses of PTFE and a PTFE-based composite when tested using three different types of
sliding wear machines. Tribiology Int 90:15–21. https://doi.org/10.1016/j.triboint.2015.04.001
10. Zalaznik M, Novak S, Huski M, Kalin M (2016) Tribological behaviour of a PEEK polymer
containing solid MoS2 lubricants. Lubr Sci 28:27–42. https://doi.org/10.1002/ls.1299
Prediction of the Blood Flow Through
Stenosis in AVF for Hemodialysis
1 Introduction
Fig. 1 a Actual MRI copy of stenosis in AVF, b stenosis showed in blood vessels of AVF by image
processing [1], c Schematic diagram of stenosis
Figure 1a shows the Magneto Resonance Image of venous stenosis in the AVF,
model of the actual AVF and corresponding vessels with the stenosis developed by
image processing (Fig. 1b) and then steady-unsteady comparative study performed
for the end-to-end type of AVF over a end-to-side [1]. Figure 1c shows simple
schematic diagram of AVF. Therefore in this study, we created the similar type of
stenosis models with varying diameter and varying lengths. The validation of CFD-
based analysis to evaluate the hemodynamic significance of access stenosis was done
with the fixed outlet pressure and varying inlet pressure from 50 to 160 mmHg [2].
The present study aims at the effect of stenosis, its diameter, and its length on the
outlet flow rate by applying the inlet flow velocity as the velocity of each linear time
step of the pulse rate like 1.3, 1.5, 2.66 m/s, etc. and outlet fixed pressure set as
10 mmHg [3].
2 Geometrical Model
The geometric model for the numerical analysis is prepared in ICEM-CFD. The
diameter of the stenosis is varied from 1 to 3 mm with the difference of 1 mm and
the stenosis length varied from 30 to 60 mm with the difference of 15 mm for each
stenosis diameter. So a total of 9 models is created. Hex element is used for structured
meshing to each model. A total number of nodes and elements vary from 6000 to
14,400 and 5811–14,001, respectively. Near the wall spacing, 0.001 with the ratio
of 1.12 used to well capture the near-wall behavior as shown in Fig. 2.
3 Numerical Modeling
The test case was set in the OpenFOAM environment, version 3. The test case is set
for each case to run the simulation. CFD analysis of the stenosis models is performed
in OpenFOAM environment using the geometric models created in ICEM-CFD. In
the calculation, blood is treated as a non-Newtonian fluid described by a power-law
model with consistency coefficient of 0.012171 Pa, and maximum dynamics viscosity
Prediction of the Blood Flow Through Stenosis … 515
Fig. 2 The structured mesh of stenosis of diameter 3 mm and length 60 mm with refined near-wall
mesh
∂U
+ ∇ · (U) = ∇ · (v∇U) + ∇U · ∇v
∂t
4 Result
For the validation, the outlet pressure of the stenosis of AVF is set as 10 mmHg as
a typical central venous pressure. The mean inlet pressure is varied by 50, 70, 100,
120, 140 and 160 mmHg to model the flow, which is usually seen in the dialysis
treatment. This was done stenosis diameter 3 mm and length 60 mm. The numerical
data depicted in Table 1.
The average difference is 14% in the literature and OpenFOAM data (Fig. 3).
Then the study is done for the inlet flow velocity as the velocity of each linear time
Table 1 Outflow(ml/min)
Pressure (mmHg) Sten 30 mm Sten 45 mm Sten 60 mm
through the stenosis of 3 mm
diameter at varying inlet 50 796 708 654
pressure and varying stenosis 70 1161 1050 978
length
100 1654 1541 1452
120 1949 1850 1758
140 2218 2034 2058
160 2423 2112 2346
516 S. Shembekar et al.
Fig. 3 Comparison of computational fluid dynamics results from OpenFOAM (orange line) and
literature [2] (blue line) for stenosis of diameter 3 mm and length 30 mm
step of the pulse rate like 1.3, 1.5, 2.66 m/s, etc. and outlet fixed pressure set as
10 mmHg and found the numeric results as shown in Table 2.
In post-processing, the velocity contour as shown in Fig. 5a showed that the
maximum velocity through the stenosis is gone up to 8.5 m/s. the maximum velocity
appears at the middle of the geometry and gradually decreases to the range of 2.5–
3.5 m/s (Fig. 4).
Table 2 Outflow rate through the stenosis of diameter 3 mm and length 60 mm at varying time
step
Linear time step 0 0.05 0.1 0.26 0.37 0.54 0.7 0.8
Velocity (m/s) 1.25 1.5 2.66 1.9 1.75 1.45 1.4 1.25
Outflow rate(ml/min) 947 1137 1974 1440 1327 1099 1099 947
2000
inlet velocity at each time
step 1500
1000
500
0
0 0.2 0.4 0.6 0.8 1
linear time step
Prediction of the Blood Flow Through Stenosis … 517
Fig. 5 For the stenosis of diameter 3 mm and length 60 mm a velocity contour, b streamline
behavior, c pressure contour
5 Conclusion
The flow rate through the various diameter and various length was observed and it is
found that the flow rate increases with the diameter and decreases with the length of
stenosis. The flow rate for the 1 mm diameter stenosis is very less compare to another
diameter of the stenosis. The stenosis of diameter 3 mm is capable of handling the
flow rate from 700 to 3000 ml/min. CFD with OpenFOAM is a very powerful and
handy tool to analyze the blood flow through the blood vessels. Our hope is to create
a tool which will help the nephrologist to predict the hemodynamic significance of
the stenosis based on the location and configuration.
518 S. Shembekar et al.
References
1. Jodko D, Obidowski D, Reorowicz P, Jozwik K (2014) Simulations of the blood flow in the
arterio-venous fistula for hemodialysis. Acta Bioeng Biomech 16(1):69–74
2. Hoganson DM, Hinkle CJ, Agarwal RK (2015) validation of computational fluid dynamics-based
analysis to evaluate hemodynamics significance of access stenosis. Jvasc Access 15(5):409–414.
https://doi.org/10.5301/jca.5000226
3. Jodko D, Obidowski D (2014) Simulation of the blood flow in the arterio-venous fistula for
haemodialysis. Acta Bioeng Biomech 16(1):69–74. https://doi.org/10.5277/abb140109
Tool Quality Monitoring in Friction Stir
Welding Process
Abstract The present research aims at monitoring the quality of the tool in an
advanced joining technique named friction stir welding by analyzing the force
signals. Five tools with different qualities, namely a cracked tool, a tool with a
half-pin, material sticking to the pin, tool without a pin, and a healthy tool, have been
selected. Discrete wavelet transform has been applied on the acquired signal, and
statistical feature of the wavelet coefficients has been extracted which aid in differen-
tiating the selected tools’ condition. The depicted idea will improve the weld quality,
reduce rejection of product, and subsequently will help avoid machine downtime
which may occur because of a faulty tool.
Keywords Tool quality monitoring · Friction stir welding · Axial force · Discrete
wavelet transform
1 Introduction
Condition monitoring has gained the urge from the industries for high-quality and
low-cost production [1]. The quality of a product is of prime importance, and as such,
the manufacturing industries invest a lot on determining the quality of their products.
There is still a probability that the quality of the product might not be up to the mark
even after determining its optimum process parameters required for manufacturing
and lead to wastage of resources. The use of sensors to monitor the process in real
time will generate signals, which can be processed using various signal processing
algorithms to infer and extract information about the process. The present work deals
with the sensor-based monitoring of an advanced welding technique named friction
stir welding (FSW). The following paragraph discusses in detail about FSW.
In FSW, the materials to be welded experience heat and undergo plastic defor-
mation due to the action of a rotating tool [2, 3]. The tool does not get consumed in
the process, and the material which is selected for fabricating this tool is of higher
strength than that of the base materials. Pin and shoulder are the two features of
this tool. The base materials are fixed in a suitable fixture without maintaining any
gap in between them. A schematic diagram representing FSW process in a butt joint
configuration has been shown in Fig. 1. The pin of the tool descends into the base
materials at the joint line [4]. Since the tool is rotating and is in contact with the
workpieces, it generates frictional heat and causes the materials to get deformed.
The pin disrupts the faying surface, and the shoulder generates the frictional heat at
the top surface of the materials to be welded [5]. A schematic view of the tool used
in FSW has been depicted in Fig. 2. In the diagram shown here in Fig. 2, resembles
a cylindrical shape pin. Researchers have used tools of varying pin and shoulder
profile in order to improve the material flow and achieve high-quality welds [6, 7].
A linear motion is provided to the machine bed, which results in transportation of
the deformed material from one side of the weld to the other, creating the joint. The
weld quality is majorly affected by the amount of tool rotational speed (ω) and linear
movement of the bed (v) [5]. This is followed by the geometry of the tool which is
being used for the welding [5].
Earlier, various research works have been published which discuss the use of
sensors in monitoring and controlling FSW. The following paragraph discusses those
sensor-based works which have been performed to monitor and control FSW. The
acoustic emission (AE) signals have been used to monitor gap defects by analyzing
the signal through discrete wavelet transform (DWT), and the feature extracted was
the energy of the signal [8]. AE signal has also been analyzed by fast Fourier transform
(FFT) to determine the effect of varying process parameters, i.e., ω, v, and plunge
depth (p), by determining the power spectral density (PSD) [9]. PSD was found to be
increasing with increase in ω, without any major change in the frequency of the signal.
This was because of the increasing temperature of the weld zone with increasing ω
which resulted in higher thermal stress and was reflected in the AE signals. In another
work, an algorithm was developed to measure the noise in different parts of the weld
and monitor the process [10]. The gap defects have also been monitored by using
axial force (F z ) signals [11]. Samples with notches in the abutting edge were prepared
to be welded. When the tool approached the notch, with the absence of material, a
drop in the F z signal was observed. Wormhole is another type of weld defect in FSW,
which occurs because of the loss of volume of material from the weld zone [3]. It is
also referred as volumetric defects. It occurs because of improper combination of ω
and v leading to availability of excessive heat in the weld zone or an insufficient heat
condition [3]. Force signals acquired in longitudinal (F x ) and transverse directions
(F y ) have been utilized to monitor this defect [12]. The tracking of tool position
over the joint line has also been monitored by using F z and F x , where an increase
in F z value was observed as the tool shifted from the joint line toward the either
side of the base materials [13]. Other studies concerned about the identification of
defects in FSW welds [14, 15]. In these works, the force and torque signal have
been analyzed through DWT. Feature extracted is the mean of square of errors of
the coefficients of three levels. A significant difference in the values of the extracted
feature was observed with respect to the location with and without weld defects. A
comprehensive review on the various works related to the sensor-based monitoring
and control of FSW has been reported in [16].
From the reported literature, it can be observed that researchers have attempted
to detect several types of welding defects, predict the position of tool, and study
the effect of varying process parameters through indirect ways of monitoring. The
automation aims to run the machine unattended. In order to ensure this for FSW, the
quality of the tool also needs to be ensured in addition to detection of weld defects,
and the art of this is lacking in the literature. This analysis is important because tool is
the one which is plunging inside the materials, generating frictional heat and stirring
the deformed material. Sudden breakage of the pin or even a slight damage to it will
deteriorate the weld quality. Thus, in the present work, the tool quality monitoring
has been attempted to be investigated for the automation of FSW process.
2 Experimental Procedure
The materials selected in this study were 3 mm thick sheets of aluminum alloy, 6061,
joined in a butt configuration. The machine utilized is a computer numeric-controlled
FSW machine (ETA Technology, WS004), equipped with strain gauge-type load cells
of 10 kN capacity. The sampling rate for data acquisition is factory set, and the value
of the same is 10 Hz, and the data acquisition is achieved through LabVIEW. The
parameters chosen in this study for welding were ω of 1200 rpm, v of 50 mm/min,
tilt angle (α) of 2°, and a p of 0.1 mm. These parameters have been chosen based
522 D. Mishra et al.
Fig. 4 Selected types of tool: a cracked tool, b tool with half-pin, c material sticking on pin, d tool
without a pin, e healthy tool
DWT is a signal processing technique which makes use of mother wavelets which
are irregular in shape, exist for a short period of time, and have a zero mean value.
DWT provides a measure of both the time and frequency associated with a signal
through detail and approximation coefficients which are calculated by shifting and
scaling the mother wavelet by powers of 2, as depicted in Eqs. (1) and (2) [18].
1 t − k · 2j
ϕ(t) = √ ϕ (1)
2j 2j
∞
1 t − k · 2j
γ = x(t) √ ϕ dt (2)
2j 2j
−∞
where ϕ(t) is the mother wavelet function, k is the shift parameter, j is the scale
parameter, and x(t) is the original signal.
The high frequencies in a signal can be signified as the ones that rapidly change
with time. Thus, they could precisely capture the information about weld defects
which occur because of sudden change in the process. The low frequencies on the
other hand are the ones which slowly change over the time. Since the aim of the
present work is to monitor quality of the tool over the time, the approximation
coefficients are analyzed.
The mother wavelet selected belongs to Daubechies family (db4), and the decom-
position has been carried out up to three levels. The selection of mother wavelet and
level of decomposition has been performed with the help of the literature, where
researchers have utilized force signal to identify defects in FSW weld through DWT
[14]. The approximation coefficients have been analyzed through statistical features,
which help visualizing the differentiation of the signals.
The time domain force signal and the corresponding decomposed coefficients
(approximation-a1, first level) of various tools have been shown in Fig. 6a–e. The
physical welded samples obtained with these tools have been shown in Fig. 7. As
expected, while the F z of the healthy tool seems to have very less variation, the
case is not same with others. The sample welded by using healthy tool has a smooth
surface appearance with no surface defects. Large variations can be observed with
the tool with half-pin because of the severe cutting action being provided and the
improper material flow, which is reflected over F z . The inadequate material mixing
can be seen in the welded sample. The worst case is seen with the tool without a
pin. In this case, as the pin was absent, the material, in thickness direction, did not
get plasticized, which occurred due to the sole movement of the shoulder. This tool
failed to create the joint. The tool with material sticking on the pin might not always
be detrimental, because, if, the material that sticks is of the same grade as of base
materials, then it also may get deformed during plunging. However, if the case is not
so, then these sticking material will destroy the weld quality, and similar observations
Tool Quality Monitoring in Friction Stir … 525
Fig. 6 Time domain signal and the corresponding approximation coefficients (a1) obtained with
selected tools: a cracked tool, b tool with half-pin, c material sticking on pin, d tool without a pin,
e healthy tool
can be seen here in F z . Though the cracked tool employed did not break during or
after the welding, however, it can affect the process if it fails. Nevertheless, some
variations can be seen in F z which may be probably because of the crack portion
which affected the flow of the material.
In order to compare the behavior of the five tools through F z , a statistical feature
(mean) has been extracted out of the approximation coefficients (a1 corresponding
526 D. Mishra et al.
Fig. 6 (continued)
Fig. 6 (continued)
Fig. 7 Physical samples welded by using the five different selected tools: a cracked tool, b tool
with half-pin, c material sticking on pin, d tool without a pin, e healthy tool
4 Conclusion
The present research focuses on monitoring the tool quality in FSW by processing
the F z signals acquired during the welding process through DWT and analyzing the
obtained approximation coefficients. The research marks the importance of force
signal in monitoring of FSW. A threshold value has been identified which clearly
differentiates the healthy and faulty tools. This idea can be utilized in online models
being developed for controlling FSW to better monitor the process and health of
the tool as well. The future direction for this will be to identify the force signal
characteristics with different parametric combinations to have more insights into the
differentiation.
References
1. Wang L, Gao RX (2006) Condition monitoring and control for intelligent manufacturing.
https://doi.org/10.1007/1-84628-269-1
2. Thomas W, Nicholas E (1997) Friction stir welding for the transportation industries. Mater
Des 18:269–273. https://doi.org/10.1016/S0261-3069(97)00062-9
3. Mishra D, Sahu SK, Mahto RP, Pal SK, Pal K (2019) Friction stir welding for joining of
polymers. Springer, Singapore, pp 123–162. https://doi.org/10.1007/978-981-13-0378-4_6
4. Mishra RS, Ma ZY (2005) Friction stir welding and processing, vol 50. https://doi.org/10.1016/
j.mser.2005.07.001
5. Jain R, Kumari K, Kesharwani RK, Kumar S, Pal SK, Singh SB et al (2015) Friction stir
welding: scope and recent development, pp 179–229. https://doi.org/10.1007/978-3-319-201
52-8_6
6. Sahu SK, Mishra D, Mahto RP, Sharma VM, Pal SK, Pal K et al (2018) Friction stir welding of
polypropylene sheet. Eng Sci Technol Int J 21:245–254. https://doi.org/10.1016/j.jestch.2018.
03.002
7. Elangovan K, Balasubramanian V (2008) Influences of tool pin profile and welding speed on
the formation of friction stir processing zone in AA2219 aluminium alloy. J Mater Process
Technol 200:163–175. https://doi.org/10.1016/j.jmatprotec.2007.09.019
8. Chen C, Kovacevic R, Jandgric D (2003) Wavelet transform analysis of acoustic emission in
monitoring friction stir welding of 6061 aluminum. Int J Mach Tools Manuf 43:1383–1390.
https://doi.org/10.1016/S0890-6955(03)00130-5
9. Soundararajan V, Atharifar H, Kovacevic R (2006) Monitoring and processing the acoustic
emission signals from the friction-stir-welding process. Proc Inst Mech Eng Part B J Eng
Manuf 220:1673–1685. https://doi.org/10.1243/09544054JEM586
10. Kleiner D, Bird CR (2004) Signal processing for quality assurance in friction stir welds. Insight
46:85–87
11. Fleming P, Lammlein D, Wilkes D, Bloodworth T, Cook G, Strauss A et al (2008) In-process
gap detection in friction stir welding. Sens Rev 28(1):62–67. https://doi.org/10.1108/026022
80810850044
12. Jene T, Dobmann G, Wagner G, Eifler D (2008) Monitoring of the friction stir welding process
to describe parameter effects on joint quality. Mater Sci 5454:1–11. https://doi.org/10.1007/
BF03266668
13. Fleming PA, Lammlein DH, Wilkes DM, Cook GE, Strauss AM, Delapp DR et al (2009)
Misalignment detection and enabling of seam tracking for friction stir welding. Sci Technol
Weld Join 14:93–96. https://doi.org/10.1179/136217108X372568
530 D. Mishra et al.
14. Kumar U, Yadav I, Kumari S, Kumari K, Ranjan N (2015) Defect identification in friction stir
welding using discrete wavelet analysis. Adv Eng Softw 85:43–50. https://doi.org/10.1016/j.
advengsoft.2015.02.001
15. Kumari S, Jain R, Kumar U, Yadav I, Ranjan N, Kumari K et al (2016) Defect identification in
friction stir welding using continuous wavelet transform. J Intell Manuf 30:1–12. https://doi.
org/10.1007/s10845-016-1259-1
16. Mishra D, Roy RB, Dutta S, Pal SK, Chakravarty D (2018) A review on sensor based monitoring
and control of friction stir welding process and a roadmap to Industry 4.0. J Manuf Process
36:373–397. https://doi.org/10.1016/j.jmapro.2018.10.016
17. Chauhan P, Jain R, Pal SK, Singh SB (2018) Modeling of defects in friction stir welding using
coupled Eulerian and Lagrangian method. J Manuf Process 34:158–166. https://doi.org/10.
1016/j.jmapro.2018.05.022
18. Gao RX, Yan R. Wavelets: theory and applications for manufacturing, pp 1–224. https://doi.
org/10.1007/978-1-4419-1545-0
Condition Monitoring and Fault
Diagnosis of Induction Motor in Electric
Vehicle
Abstract The twenty-first century is witnessing the growth of electric vehicles due
to the declining level of petroleum products and legal concern for clean technology to
take care of environmental pollution. Battery and electric motor are the two impor-
tant components in the electric vehicle. An electric motor is a prime component
responsible for the propulsion of a vehicle. Because of the continuous operation
and load variation, the motor is subjected to different types of faults. Thus, condi-
tion monitoring and on-board diagnosis of an electric motor in the electric vehicle
is essential to avoid catastrophic failure. In India, it is observed that the induction
motor is commonly used in electric vehicles for propulsion. This article proposes the
methodology for condition monitoring and fault identification of components in the
induction motor using an on-board diagnostics in an electric vehicle.
1 Introduction
of an electric motor based on the single physical parameter. Akin suggested the elec-
tric implementation of fault diagnosis in hybrid electric vehicle based on reference
frame theory at a standstill or idle condition of the electric vehicle [2]. Watkins
proposed the new sensor embedded in the stator winding of electric vehicle motor
to provide a useful remaining life of stator winding based on actual vehicle opera-
tion and environmental conditions [3]. Kunthong proposed IOT-based traction motor
drive condition monitoring in an electric vehicle, where temperature, vibration, and
current signals data are acquired and forwarded to the server using Wi-Fi module
[1]. For an electric vehicle traveling in remote areas, it is difficult to get Internet
connectivity. This drawback can be avoided by providing the on-board diagnosis in
the vehicle itself.
In this article, a methodology for OBD of induction motor with the help of
Arduino-based microcontroller is proposed to indicate failure. Using a single phys-
ical parameter, the fault diagnosis cannot be guaranteed; thus, the use of vibration
sensors and the current sensors is proposed for sensing vibration and current, respec-
tively. The dashboard display will indicate the presence of a fault in the induction
motor based on the acquired signal.
For an all-electric vehicle type, the vehicle is dependent on the electric motor for
smooth working, unlike the hybrid electric vehicle. In an electric vehicle due to speed
and load variations, an electric motor is subjected to different types of stresses. These
stresses result in different types of faults in an induction motor. Faults in the induction
motor can be either a mechanical fault or an electrical fault. Mechanical fault includes
bearing fault, rotor bar damage, eccentricity, unbalance, etc. The electrical fault
includes stator damage, shortening of winding, etc.
Bearing faults are responsible for 40–45% failure of induction motor [4]. In an
electric vehicle, depending upon the road condition and torque requirements, the
vehicle gets subjected to shocks and impulses. Thus, the bearing may get affected
due to these impacts and the fault may generate in the bearing. Based on the location
of the defect, bearing defect can be an outer race defect, inner race defect, ball defect,
and cage defect, etc. Bearing defect may affect the performance of the entire electric
vehicle as the propulsion gets affected.
Stator faults are responsible for 30–40% failure of induction motor [4]. The stator
fault can be a frame fault or winding fault. The coating is provided on the windings
of electric motors. The purpose of the winding coating is to withstand the electrical
stress. But, sometimes due to continuous run for a longer time, an electric vehicle may
get subjected to mechanical and environmental stresses. These stresses can affect the
winding coating.
The broken rotor bar fault is responsible for 5–10% failure of the induction motor
[4]. Most of the rotors used in the induction motors are cast rotors. These cast rotors
are hardly repaired, once the faults like cracked or broken rotor bar developed in them.
Condition Monitoring and Fault Diagnosis of Induction … 533
The rotor bar may fail due to thermal, mechanical, magnetic, or residual stresses.
Once the rotor bar gets broken or damaged, excessive current will flow through the
bar adjacent to it, and thus, electrical stresses get generated in the adjacent area. As
these stresses exceed the allowable limit, the life of the motor gets shortened.
Eccentricity faults are due to a nonuniform air gap between the stator and the
rotor. Eccentricity can be static or dynamic depending on the fixed or variable air
gap between the stator and rotor. Loose bolts, improper mounting, misalignment,
and rotor unbalance are responsible for the eccentric fault in an induction motor.
The presence of any of the faults will affect the performance of an electric vehicle.
The catastrophic failure in the electric motor of the electric vehicle may result in a
dangerous situation during driving, especially on the highway.
Based on the parameters used, condition monitoring techniques are vibration moni-
toring, motor current signature analysis, thermography, acoustic monitoring, stray
flux analysis, etc. Mechanical faults can be better diagnosed with vibration moni-
toring while electrical faults can be better diagnosed with motor current signature
analysis. Table 1 indicates the suitable monitoring technique for different fault in the
induction motor which can be used for on-board diagnosis.
In the electric vehicle, the traditional method of continuous vibration monitoring
for an induction motor is not much useful due to the vehicle’s vibration during
propulsion. In running conditions, due to the vehicle’s vibration, the air gap between
the stator and the rotor gets disturbed. Thus, electromagnetic flux gets changed and
the current spectrum becomes noisy. To get a better indication of a fault, it is necessary
to do the condition monitoring at startup or idle condition of an electric vehicle [2].
For a fault diagnosis of induction motor, researchers have suggested different
techniques like a fast Fourier transform (FFT), short-time Fourier Transform (STFT),
multiple signal classification (MUSIC), and digital frequency locked loop (DFLL)
[16]. But, these techniques need an initial knowledge base of a user to diagnose
the fault. It is difficult for a driver to understand the presence of fault based on the
spectrum analysis. Time-domain statistical features may help to diagnose the fault
Vibration Sensor
On
Induction Microcontroller Board
Display
Motor
Current Sensor
with better indication on the dashboard display screen in the form of a numerical
value, but its interpretation and standardization would be a difficult task.
4 Proposed Methodology
It is difficult to diagnose the fault based on a single physical parameter for an induction
motor in an electric vehicle. Even the mechanical vibration of an electric vehicle can
disturb the fault diagnosis of induction motor. Thus, the use of two types of the
sensors is proposed, vibration sensor and current sensor. The vibration sensor will
be mounted over the induction motor casing above the drive end bearing. Hall effect
current sensor will acquire the current signal taken by the induction motor from
the battery through inverter arrangement. The vibration sensor and current sensor
will acquire the signal in the time domain. The statistical features (mean, root mean
square, kurtosis, variance, skewness) will be calculated from the acquired dataset
of vibration and current sensors. These features will be compared with the healthy
condition feature vectors. Based on the difference between the values of statistical
feature vectors, the faults like stator fault, bearing fault, rotor fault, eccentricity fault
Condition Monitoring and Fault Diagnosis of Induction … 535
can be identified. The artificial intelligence module can also be applied for decision
making.
6 On-Board Diagnosis
It is essential and necessary for an electric vehicle to monitor the critical equipment
continuously. The existing dashboard of the vehicle indicates battery level, speed,
km driven, gear indication, and battery maintenance alarm. The proposed on-board
diagnosis helps to give a warning and monitor the faults of the motor in the vehicle.
The proposed dashboard as shown in Fig. 2 will indicate the types of faults in
the motor at the incipient level. Also, the motor of the electric vehicle has a definite
range for the operating temperature. So, by providing the temperature indicator of
the electric motor in a vehicle, heating of the motor can be avoided which will help
to sustain the windings coating for a longer time. In the proposed dashboard, the
indication will be given about the presence of fault so that driver will be alert and
the failure can be avoided at the initial stage.
7 Conclusion
Considering the safety and cost-effective maintenance, the need for condition moni-
toring and fault diagnosis of the electric motor in an electric vehicle is discussed. In
this article, possible mechanical faults and electric faults in an induction motor are
elaborated. The various techniques that can be applied for these fault identifications
are discussed along with its suitability for an electric vehicle. Thus, an on-board
diagnostic technique is proposed as a proactive approach to avoid any catastrophic
failure of components in the induction motor, and the condition of the motor can be
monitored.
References
10. Bazan GH, Scalassara PR, Endo W, Goedtel A, Godoy WF, Palácios RHC (2017) Stator
fault analysis of three-phase induction motors using information measures and artificial neural
networks. Electric Power Syst Res 143:347–356
11. Gu F, Wang T, Alwodai A, Tian X, Shao Y, Ball A (2015) A new method of accurate broken
rotor bar diagnosis based on modulation signal bispectrum analysis of motor current signals.
Mech Syst Signal Process 50:400–413
12. Ishkova I, Vítek O (2015) Diagnosis of eccentricity and broken rotor bar related faults of
induction motor by means of motor current signature analysis. In: 2015 16th international
scientific conference on electric power engineering (EPE). IEEE, pp 682–686
13. Bellini A, Filippetti F, Franceschini G, Tassoni C, Kliman GB (2001) Quantitative evaluation
of induction motor broken bars by means of electrical signature analysis. IEEE Trans Ind Appl
37(5):1248–1255
14. Gupta RB, Singh SK (2019) Detection of crack and unbalancing in a rotor system using artificial
neural network. In: Advances in engineering design. Springer, Berlin, pp 607–618
15. Pinheiro AA, Brandao IM, Da Costa C (2019) Vibration analysis in turbomachines using
machine learning techniques. Europ J Eng Res Sci 4(2):12–16
16. Nandi S, Toliyat HA (2002) Novel frequency-domain-based technique to detect stator interturn
faults in induction machines using stator-induced voltages after switch-off. IEEE Trans Ind
Appl 38(1):101–109
Mechanical Design Calculations
of Flywheel Generator
1 Introduction
Pulsed current has applications like electromagnetic propulsion, where the release
of energy takes place in milliseconds. This pulsed current can be achieved through
capacitor banks or flywheel generators. Flywheel generator has a higher energy
density compared to conventional capacitor banks. Flywheel energy storage system
(FESS), with a capacity of 10 MJ at 17,000 rpm with a 10% discharge rate per cycle, is
to be constructed at IIT Delhi. The planned setup will have an energy storage density
of 77.5 J/g and a power density of 1.94 kW/g. This energy density is much higher
than that of capacitor bank 1 J/cc [1]. In this paper, we discussed the mechanical
design calculations for FESS. Issues like stresses and air drag loss (windage loss)
have been analysed.
In Sect. 2, we discussed the design schematic of the setup. Next, design calcula-
tion is presented along with the modelling of various components. In this section,
we started with a consideration of NdFeB magnets followed by the calculation for
interference amongst the rotor layers. Then, we proceed with stress analysis of rotor
and calculation of static and dynamic deflections of the assembly. We discussed the
effect of anisotropic degree on hoop stress and deflection. Further, we presented a
modified version of Rayleigh’s method to calculate the critical speed of complete
assembly. At last, losses due to air drag and the effect of evacuation are discussed.
2 Design Schematic
The flywheel generator uses a spindle motor of 175 kW to drive the rotor. The
motor torque is transmitted to the rotor through a hub mounted over a steel shaft.
The rotor consists of aluminium layer for shielding, Neodymium magnets (NdFeB)
permanent magnets for the magnetic field, titanium sleeve for pre-compression of
magnets, and filament wound carbon fibre-epoxy composite rotor for providing the
inertia. High-speed ceramic angular contact bearings support this rotor assembly
via flexible flanges. These angular contact bearing are pre-compressed to increase
the stiffness of bearing. This whole assembly is mounted over a stator consisting of
winding for generating the electric pulse. The design schematic is shown in Fig. 1.
The entire setup is then packed into an evacuated chamber.
For applying a magnetic field, Neodymium Iron Boron (NdFeB) magnets are used due
to high magnetic strength. NdFeB magnets have a low tensile strength of ~80 MPa,
limiting the operational speed; therefore, it is compressed using titanium sleeve to
~600 MPa (compressive strength ~850 MPa) such that upon rotation, compressive
Mechanical Design Calculations of Flywheel Generator 541
stress is relieved allowing to achieve speed higher than 17,000 rpm. Magnets, along
with non-magnetic filler material (aluminium), form a cylindrical layer whose equiv-
alent modulus is given by Eq. (1). This equivalent modulus is used to calculate the
required interference fit between magnets and titanium sleeve.
1 χpm 1 − χpm
= + (1)
E eq E pm E Al
where E eq is the modulus of the equivalent layer, χpm is length fraction of NdFeB in
the layer, E pm is the modulus of PM, E Al is the modulus of aluminium.
Interference Calculation. To compress the magnets, interference is calculated using
thin cylinder approximation. Interference amongst the segments not only keeps the
magnet layer in compression but also reduces the radial stress among other segments
(Table 1).
At high speed, stresses become critical in rotors. There are four cylindrical layers in
the flywheel rotor (as discussed before) with different interference fits. The strength
and radial region of each segment are given in Table 2. The hoops stress, radial
stress, and radial deflection are calculated using equations for rotating anisotropic
disc along with boundary conditions [2–4]. The results are shown in Fig. 2.
The PM layer is pre-compressed to ~600 MPa, which is relieved to ~200 MPa
(compressive) at 17,000 rpm. The maximum circumferential stress in the system is
~500 MPa (tensile) in the titanium sleeve and composite rotor. The stresses in all
layers are found to be within their respective allowable limits.
The rotor assembly consists of five horizontal sections with different section prop-
erties. These sections are modelled as connected beams, and loading conditions are
applied with boundary conditions to compute the deflection (say y). The assembly
bends due to gravitational force (say G) and centrifugal force (say C(y)). Gravita-
tional force is independent of deflection, while the centrifugal force is a function of
deflection (y) and changes direction during rotation. These two forces are balanced
by the restoring force (say K) of the rotor, which also depends on the deflected value.
Maximum deflection occurs when centrifugal force aligns with the gravitational
force. The deflection is calculated using an iterative method, and the solution
converges to give the result. However, near critical speed, the solution does not
converge. This is consistent with the theory that near critical speed the deflection
value tends to infinity.
In the calculation, it is assumed that the rotor has an eccentricity of 1 µm (as per
the facility available for balancing the rotor). Figure 3 shows the static deflection (due
to gravitational force) and maximum possible deflection (due to centrifugal force and
gravitational force combined).
Fig. 4 Trend of deflection and hoop stress with changing anisotropic degree (λ)
544 Md Z. Anwar et al.
Fig. 5 Effect of anisotropic degree on max deflection and max hoop stress in a rotating disc
The anisotropic degree can be varied to achieve the optimised value. At 17,000 rpm
and E θ = 120 Gpa, for minimum hoop stress optimised value, is λ = 6, while for
minimum deflection is λ = 2.5 as shown in Fig. 5. Depending upon the design
requirement, the anisotropic properties are designed to achieve the optimised result.
The carbon-epoxy filament wound rotor in the flywheel has a value of λ = 3.65.
The value of anisotropic degree λ can be optimised by selecting modulus of the
fibre and epoxy and by varying volume fraction of the fibre and matrix [3].
E θ = E f V f + E m Vm (3)
E f Em
Er = (4)
Vm E f + V f E m
where E f , E m are elastic moduli of fibre and matrix, respectively; V f , Vm are volume
fraction of fibre and matrix.
The critical speed of the assembly, consisting of five sections having different prop-
erties, is calculated using a similar approach as of Rayleigh’s method, but instead of
assuming a shape function, here actual deflection curve is used. Rayleigh’s method
considers deflection as per the shape function without considering the differences in
stiffness [5]. This results in inaccurate potential and kinetic energy of each section.
l1 2 l2 2
d2 W (x) d2 W (x)
0 E 1 I1 dx 2
dx + l1 E 2 I2 dx 2
dx + · · ·
ω 2 = l1 l2 (5)
0 ρ1 A1 (W (x))2 dx + l1 ρ2 A2 (W (x))2 dx + · · ·
The Rayleigh method with assumed shape function gives a critical speed of
122.7 krpm, while the above approach gives 54 krpm. In Rayleigh’s method with
assumed shape function the flywheel section stores 96% of total PE, but in the real
case, this section is rigid and has negligible deformation as compared to other parts,
and hence, PE which is due to strain is just 6% of the total. This is the reason that
Rayleigh’s method gives the inaccurate results, where the section properties vary
drastically.
Air drag on rotor surfaces causes significant power loss. At atmospheric pressure, the
power loss is ~200 kW. Power loss at surfaces depends on fluid properties and rotor
dimensions (Eq. 6) [6, 7]. The density of air depends on pressure; hence, evacuating
the chamber reduces the power loss. Power loss reduces to ~50 kW at 10 kPa (Fig. 6).
Pw = πCd ρr 4 ω3 L (6)
Cd 2.04 + 1.768 ln Re Cd − 1 = 0 (7)
where Pw is power loss or windage loss (W), ρ is the density of the fluid (kg/m3 ),
r is the radius of cylinder (m), ω is angular velocity(rad/sec), Cd is skin friction
coefficient, L is the length of the rotor, Re is Reynold’s number.
.
4 Conclusion
In this paper, design calculation of the main parameters of the flywheel assembly has
been discussed. The design was done based on energy storage capacity of 10 MJ at
17,000 rpm. The hoops and radial stresses in the flywheel rotor are investigated along
with the modelling of the NdFeB magnet layer and filler material as an equivalent
layer. Benefits of anisotropic material over isotropic were analysed. The deflection in
the rotor assembly has been calculated using an iterative method, which was further
used in finding critical speed. The critical speed of the assembly was determined with
a modified version of Raleigh’s method and found out that assuming shape function
can result in a significant deviation from actual critical speed. Air drag, which is the
primary cause of the power loss in a high-speed flywheel, has been discussed and
shown that evacuating the chamber results in a significant reduction of power loss.
Overall through this paper, general design parameter has been discussed. However,
detailed simulation results need to be considered in the next step to build the proto-
type. Further work is to be done to improve the energy storage capacity and achieve
higher operating speed.
References
1. Curtiss DH, Mongeau PP, Puterbaugh RL (1995) Advanced Composite flywheel structural design
for a pulsed disk alternator. IEEE Trans Magn 31(1):26–31
2. Vrancik JE (1968) Prediction of windage power loss in alternators. National Aeronautics and
Space Administration, Washington, D. C.
3. Zhang Q, Wu S, Yu C, Cui S, Song L (2011) Design of a Model-scale air-core compulsator.
IEEE Trans Plasma Sci 39(1):346–353
4. Zhang F, Du G, Wang T, Liu G, Cao W (2015) Rotor retaining sleeve design for a 1.12-MW
high-speed PM machine. IEEE Trans Ind Appl 51(5):3675–3685
5. Maslen EH, Schweitzer G (2009) Aerodynamic losses, windage losses. In: Magnetic bearings.
Springer, Berlin, pp 141–145
6. Rao SS (2004) Rayleigh s method. In: Mechanical vibrations. Prentice Hall, Upper Saddle River,
pp 742–745
7. Pradeep Kumar V, Swarup S, Rajput S, Kumar G, Nomula AP, Jadhav KB, Taral SY, Daniel KJ,
Datar S (2019) Design and development of 4-MJ capacitor bank-based pulsed power system for
electromagnetic launcher. IEEE Trans Plasma Sci 1–9
Numerical Analysis and Reduction
of Blade-Vortex Interaction (BVI) Noise
in Helicopter Using Numerical
Simulation
1 Introduction
The continuous increase in operations for both military and commercial rotorcrafts,
especially in developing and developed countries demand the need for noise reduc-
tion technologies. Helicopters are a unique machine with some peculiar capabilities
like vertical take-off, hover, etc. that makes it distinctive from other rotorcrafts and
fixed-wing aircraft. Helicopters are used for various tasks like hire hazards, evac-
uation operations on rough terrains, rescue mission, etc. which are not possible by
fixed-wing aircraft. In addition to the generation of lift and thrust, the helicopter
rotor produces undesirable noise. Environmental regulations limit the permissible
sound level for helicopter noise. Low frequency sound travels large distances without
any attenuation, which is of major concern for military rotorcraft. This undesirable
consequence results in the forewarning and early detection of rotorcraft. Hence it is
necessary to predict and reduce helicopter noise.
The flow around the rotor is highly 3D dimensional, unsteady, complex with
the formation of tip vortices [1]. The rotor sits in the region of its wake, aged tip
vortices resulting in unsteady airloads on the blade surfaces. The main rotor system
Fig. 1 Various noise sources in a helicopter. Source J. Gordon Leishman, Principles of Helicopter
Aerodynamics; Second Edition, Cambridge Aerospace Series
is the primary noise source generator of the overall helicopter noise. Helicopter rotor
noise is mainly dominant by High-speed Impulsive (HSI) noise and Blade-Vortex
interaction (BVI) noise, in addition to other noises like tail rotor noise, trailing edge
noise, broadband noise, turbulent noise, etc. as shown in Fig. 1. BVI is one of the
primary sources of noise to the overall helicopter noise. The BVI noise occurs due
to the interaction of strongly concentrated tip vortices with the rotor blade. The
pressure difference below and above the rotor blade surface near the blade tip results
in the curing of air into concentrated vortices called tip vortices. These vortices
flow at proximity to the succeeding blade, which causes a change of angle of attack
along the blade spanwise direction, which in turns affects the thrust of the rotor.
The interaction of the vortices with the leading edge of the blade results in sudden
pressure fluctuation on the surface which leads to strong impulsive BVI noise.
Distinctive features of BVI noise are the pressure fluctuations occur mainly near
the leading edge of the blade section and noise directivity is forward and below
the rotor plane [2]. From previous studies, BVI noise depends on various parame-
ters like Tip Mach no., advance ratio, intersecting angle, tip plane angle, etc. [3].
These parameters alter the distance of tip vortices with the blade surface. The sound
spectrum of helicopter noise shows several peaks at Blade passage frequency (BPF)
and its harmonics. Low frequency components propagate to larger distances without
any attenuation resulting in forewarning and early detection of the rotorcraft. This
problem is of major concern for military activities. For the last two decades, studies
on BVI phenomena and its noise have been studied extensively both theoretically as
well as experimentally. Various experiments have been conducted on the Eurocopter
BO-105 helicopter at German-Dutch Wind tunnel and Ames Rotor Test Apparatus
to capture the far-field noise at different advance ratios. Widnall [4], Lowson and
Ollerhead [5] were the first people to study BVI noise generation by theoretical
studies.
Numerical Analysis and Reduction of Blade-Vortex Interaction … 549
In the past decades, various techniques like passive control, active control tech-
niques to reduce BVI noise has been proposed and studied. Passive control techniques
are based on modifying the blade specification like sweep, anhedral, etc. Landgrebe
et al. [6] studied the effect of thinning and tapering of the rotor blade tip using ogee
tips experimentally. Altering the net force balance on the helicopter changes the
miss-distance and therefore reduces the BVI noise [7]. Active control techniques
like Higher Harmonic Control (HHC), Individual blade control (IBC), trailing edge
tap, etc. are based on the external supply of energy and change the pitch angle of
rotor blade at different azimuthal angles [8–10]. These techniques showed signifi-
cant reduction of helicopter noise. Modulated blade spacing was another approach
studied to understand the effect on BVI noise.
Computational aero-acoustics (CAA) is the branch of aero-acoustics used for
modelling noise generation and far-field propagation by various acoustical analogy
techniques. The advancement of high-performance computing facilities makes
computational aero-acoustical simulation feasible. Coupling between CAA and CFD
helps to get a greater picture of complex turbulence phenomenon with noise gener-
ation by various noise sources and noise propagation to far-field. Ffowcs Williams
and Hawkings formulated the FW-H equation [11], which is used for far-field sound
due to motion of the noise source. The far-field pressure fluctuation is calculated by
Farassat 1A formulation of the FW-H equation [12] in this study.
2 Computational Methodology
To study the BVI and its noise, two NACA 23,012 airfoil at zero angle of attack of
chord length c = 0.0605 m is kept at a distance of pi/2 m. Domain dimension of 1.2
× 3.5 m is taken as the computational domain for this study. For the flow control
method, the slit of width 2 mm is placed near the trailing edge on the 1st airfoil.
The top and bottom domain surface is 10c from the airfoil centre. The suction at
slit airfoil alters and diffuses the downstream vortices. Figure 2 shows the geometry
specifications with and without the slit. An unstructured trimmer mesh with surface
Fig. 2 Computational
domain specification
550 J. S. Syriac and N. Vinod
Table 1 Boundary
Boundaries Condition
conditions for the study
Boundary Value
condition
Inlet (U) Velocity inlet 150 m/s
Top Symmetric plane ∂u/∂y = 0, v = 0
Bottom Symmetric plane ∂u/∂y = 0, v = 0
Outlet Pressure outlet –
NACA 23,012 No-slip u, v = 0
airfoil (adiabatic)
Suction slit Velocity inlet Cs = 0.133
(Cs = u jet /U ) (suction)
mesher and prism layer mesher is used. Refined mesh region near the airfoil is used
to capture the dominant noise sources and a coarser mesh in the far-field region for
the computational purpose. A surface control is applied on the body surface. In these
simulations, the number of trimmer cells used is approximately 3.6M cells.
An unsteady implicit model with segregated flow model is used in the study.
RANS model with K-W turbulence model is used for turbulent modelling in the
simulation to capture the wake region accurately. The ideal gas equation is used as
the state equation. A second-order accurate temporal discretising scheme is used for
the transient term. The different boundary conditions used in the domain can be seen
in Table 1. In the active control study, control is provided to understand the effect
on vortices and far-field noise propagation. The computational domain dimension is
large enough to avoid the boundary wall effects. For the aero-acoustical study, FW
receivers have been positioned at different locations to capture the far-field noise.
2D Aero-acoustic simulation of two airfoils has been carried out to study the BVI
noise. The inlet Mach number is kept at 0.44 to simulate the flow near the rotor
tip. TEAS active control was used to diffuse the vortices on the upper surface of
the airfoil. In the active control case, the suction provided at the slit sucks fluid
into it resulting in the reduction in vorticity strength. To study the effect on vortex
strength, line probes are placed at different downstream locations. Time average
value of absolute vorticity (AAV) for each line probe is determined. Maximum AAV
is obtained near the trailing edge of the airfoil. The representational plot of percentage
reduction in vortex strength is shown in Fig. 3. From the plot, it is observed that the
maximum rate of reduction in vortex strength is seen until some distance from the
airfoil and then decreases gradually. The percentage reduction in vortex strength is
more as the suction coefficient is increased. To get a clearer picture, Fig. 4 shows the
instantaneous vorticity magnitude scalar function at the various downstream location
Numerical Analysis and Reduction of Blade-Vortex Interaction … 551
without control and with control. It can be seen from the figure that with the active
control, vorticity strength reduces significantly near the region of active control.
Using TEAS concept, significant suppression in vortex strength is achieved when
compared to the baseline case.
For the aero-acoustic study, frequency analysis is conducted on the captured data
by doing FFT. Figure 5 shows the comparison of SPL vs frequency of various FW-H
receivers for the baseline case and using TEAS concept. The first few peaks in the
figure are a result of the vortex generated acoustic noise and noise associated with
the BVI noise. From the plot, it is observed that the SPL values at lower frequencies
in the TEAS concept are lower than the baseline case. Hence a significant reduction
in noise level was achieved by the active control method.
4 Conclusion
References
A. Kumar
Indian Institute of Engineering Science and Technology, Shibpur, West Bengal 711103, India
V. Kumar (B) · S. Sen
CSIR-Central Mechanical Engineering Research Institute, Durgapur, West Bengal 713209, India
e-mail: [email protected]
S. Sen
e-mail: [email protected]
1 Introduction
neglecting the recognized nonsignificant effects. The present paper proposes a recur-
sive Lagrange formulation, an approach similar to [6], for computation of dynamics
and hydrodynamics (empirically in a rigid body dynamic framework) of an under-
water serial link manipulator. In the proposed Lagrangian formulation, the rigid body
mass and added mass contribute to the total kinetic energy of the system, while the
gravity and buoyancy effect get included in the total potential energy of the system.
The velocity dependent drag forces, being dissipative in nature, are included in the
Lagrangian formulation by considering Rayleigh dissipation potential. In order to
validate the formulation, a fully submerged underwater two-link pendulum is consid-
ered. The obtained equations of motion using the formulation are solved numerically
to find the response from a given initial conditions. A simple two-link pendulum of
specification same as the simulation is built, and experiments are carried out to find
response for underwater motion with the same initial conditions. The simulation and
experimental results are compared to establish the effectiveness of the formulation.
The paper is organized as in the following: Sect. 2 discusses the modeling of an
object moving through a fluid, in which major hydrodynamic forces are discussed.
Section 3 formulates the dynamics equation for underwater manipulators in a recur-
sive framework. Section 4 models the dynamic equation for two-link manipulators
and solves in MATLAB© using ODE solver to obtain the response of the manip-
ulator. Section 5 presents the experiment and discusses the results, followed by a
conclusion in Sect. 6.
The hydrodynamic forces are caused by the pressure distribution surrounding the
body. Empirically, the dominant effects are modeled through the concepts of added
mass/inertia, form or pressure drag and buoyancy, respectively [1, 8]. Other types of
drag like skin friction are neglected.
When a rigid body moves in fluid medium, a layer of surrounding fluid imposes
additional inertia on the body, which moves along with the body. In contrary to objects
moving in air, the added mass effect is necessary to be considered in under water
modeling due to the high density of water, which is approximately 800 times more
than air. A detailed derivation of the expression of the added mass of a submerged
body can be found in [8]. The kinetic energy of the surrounding fluid needs to be
considered in the model. This kinetic energy, computed according to [8], is given by:
˚
1
TW = ρ U 2 dV (1)
2
558 A. Kumar et al.
3
where U 2 = i=1 Ui2 is the square of the resultant relative fluid velocity such that
U1 = ∂ x , U2 = ∂ y , U3 = ∂φ
∂φ ∂φ
∂z
for velocity potential φ and ρ is the density of the fluid.
Author in [8] applied Green’s theorem to find surface integrals to express the kinetic
energy term in terms of surface area of the moving object and projected velocity on
the normal (n) to the surface area, A and eventually obtained the expression of the
energy as:
1
6 6
TW = λik Ui Uk (2)
2 1 1
˚ 2 2 2
1 ∂φ ∂φ ∂φ
TW = ρ + + dxdydz (3)
2 ∂x ∂y ∂z
˜ ∂φi
where λik = −ρ φ dA, i = 1 − 6 and k = 1 − 6 are the tensor notations.
∂n k
Term λik is called the added mass/inertia tensor of the body. Added masse is solely
dependent on the geometry of the body shape.
However, in all practical cases of objects with regular geometry, empirical expres-
sions are available in the literature, such as in [1, 5, 6]. Here in this paper also,
empirical recommendations available in the literature for estimation of added inertia
tensor of right circular cylinder are used within the scope of the paper.
A submerged object moving in a fluid medium is acted upon by two dominant forces,
namely drag and the lift, generated due to the relative motion. Drag force acts along
the line of net velocity and opposite to it, whereas lift is generated transversely,
depending upon the body geometry. For an axisymmetric body, net lift may become
zero, which is normally the case for the links of an underwater manipulator. Among
all the drag forces, only the dominant form drag is considered in the empirical
formulation. The form drag is seen as a resistance offered by the fluid medium,
which is directly proportional to the density of the medium, the square of magnitude
of relative velocities and the projected area normal to the net velocity direction of the
body. High density of water makes this as a dominating hydrodynamic force. The
empirical model of the form drag is characterized by the so called drag coefficient. It is
a dimensionless number and is defined by the a ratio Cd = ρU 2FD
2 A , where FD is the drag
force, U the relative flow velocity, and A is the projected area normal to the velocity
direction. The drag coefficient is a function of Reynolds number and the shape of the
body moving inside the fluid and is determined experimentally for different shapes
and flow conditions. Literature provides values of the drag coefficients for standard
bodies at different Reynolds numbers. Hence, the drag force is empirically expressed
as [1]:
Dynamics of Underwater Manipulator: A Recursive Lagrangian Formulation 559
1
FD = − ρCd U 2 A (4)
2
In this article, the manipulator is operated in still water. Links of the manipulator
are made of the shape of right circular cylinder. In the experiments, cylindrical
aluminum rods are used to make the links.
2.3 Buoyancy
Buoyancy is a hydrostatic effect which can have a dominant role on the motion of the
submerged object. It is an upward force which occurs due to the pressure difference
on the top and bottom of the body submerged in water. According to Archimedes’
principle, the upward thrust, or buoyance is equal to the weight of the displaced fluid
by the fully or partially submerged body.
Fb = ρgυ (5)
where υ is the volume of the body. In deriving the potential energy term in the
Lagrangian formulation, buoyancy effect plays a role.
This section briefly presents the derivation of the Lagrangian dynamics for the given
problem and obtains expressions suitable for recursive formulation (Fig. 1).
The vector of joint variables θ = [θ1 (t), . . . , θn (t)] is chosen as the set of generalized
coordinates to describe the configuration of the serial-link manipulator. Total kinetic
energy of the manipulator, fully submerged under water is given by T = Trob + TW ,
where,Trob is the kinetic energy of the manipulator and TW is the kinetic energy of
the surrounding water. Buoyancy gets included in the total potential energy term ρ.
The Lagrangian is formed as:
L θi , θ̇i = T θi , θ̇i − ρ(θi ) (6)
where potential energy being a function of generalized coordinates θi and the kinetic
energy is of generalized velocities θ̇i as well as generalized coordinates θi . Drag
force present in the system is of dissipative in nature and the consideration of all
dissipative force in the system is done using Raleigh dissipation potential, and the
Lagrangian equations of motion can be written as:
d ∂L ∂L ∂f
− + = τi ; i = 1, . . . , n (7)
dt ∂ θ̇i ∂θi ∂ θ̇i
1 1
Ti = m i UiT Ui + ωiT Γi ωi (8)
2 2
where m i is the mass of the ith link and Γi ∈ R3×3 is the moment of inertia of the
ith link about suitably assigned axes at the centre of mass.
1 m i I3×3 0
Ti = tr |Vi |ViT (9)
2 0 Γi
Ui
where I3×3 is an identity matrix and Vi = and tr (∗) is the trace operator for
ωi
a matrix.
Mass of the link is assumed to be concentrated at the link’s center of mass and Ui
and ωi are the linear and the angular velocity vectors of center of mass of ith link,
respectively. The angular and linear velocity of ith link can be calculated recursively
using (25) and (26), derived later.
Dynamics of Underwater Manipulator: A Recursive Lagrangian Formulation 561
The added mass components of the inertia tensor for the link are obtained from
(3). The expression for the mass matrix can be given as
⎡ ⎤
m i + λ11 0 0 0 0 0
⎢ m i + λ22 ⎥
⎢ 0 0 0 0 0 ⎥
⎢ ⎥
⎢ 0 0 m i + λ33 0 0 0 ⎥
Mi = ⎢ ⎥,
⎢ 0 0 0 Ii x x + λ44 Ii xY Ii x Z ⎥
⎢ ⎥
⎣ 0 0 0 IiY x Ii yy + λ55 IiY Z ⎦
0 0 0 Ii Z x Ii Z Y Ii zz + λ66
where for cylindrical link the added mass terms are evaluated as in the below [5]:
1 T
n n
T = Tj = tr V j V j Mi , (10)
j=1 j=1
2
According to the convention followed, the gravity is assumed to work along the
negative z-direction in the base frame.
Here the whole effect of buoyancy and gravity is taken under potential energy
which is equal to the resultant work required to raise the mass center of each link
from the horizontal reference plane and can be expressed as:
ρi = ρ̄i − (m i − ρυi )g T pi + i0 R ii c (11)
where υi is volume of water displace by ith link, g is acceleration due to gravity and
ρ̄i is a constant which depends on the particular reference plane. i0 R is the rotation
matrix of the ith frame with respect to base frame and pi is position vector of ith
562 A. Kumar et al.
joint with respect to base frame and ii c is the position vector of center of mass of ith
link with respect to ith joint in link frame.
Total potential energy of the system is
n
ρ= ρi (12)
i=1
Assuming link as right circular cylinder and fully submerged in homogenous fluid,
drag force on link is calculated by slicing the links into finite numbers of equal
elements. The drag force acting on the link is then expressed as:
1 s
li
FDi = − ρCdi di |Uie |Uie cos(ψi ) (13)
2 e=1
s
where
ρ = density of water
di = diameter of ith link
Cdi = drag coefficient
Uie = elemental translational velocity on ith link
e represents eth element along link-i.
li = length of ith link
ψi = angle of projection
s = number of elements (of equal size)
The angle between two vectors is computed as:
Uie · Uien̂
cos(ψi ) = ,
|Uie ||Uien̂ |
where Uien̂ is the vector perpendicular to x-axis of the frame attached on link-i.
Then the drag force is expressed as:
1
FDi = − ρCdi Ai |Ui |Ui (14)
2
where Ai is the projected area whose normal is along the velocity direction Ci =
1
2
ρ Cdi Ai
Dynamics of Underwater Manipulator: A Recursive Lagrangian Formulation 563
The virtual work (δW ) done by FDi due to virtual displacement δ p at time t is
given by
n
δW = − FDi .δpi (16)
i=1
N
δpi
N
δ ṗi
δpi = · δθe = · δθe (17)
e=1
δθe e=1
δ θ̇e
n
δF
δW = δθe (18)
e=1
δ θ̇e
n
where F = 13 i= Ci Ui3 is Rayleigh dissipation function for the drag force acting
on the under water manipulator.
j
∂L
N
1 ∂Vj T ∂ V jT N
∂ p j + 0j R j c
= tr V M j + Vj Mj + m j − ρυ j g T
∂θi j=i
2 ∂θi j ∂θi j=i
∂θi
(21)
Combining (20) and (21) and on simplification the Lagrange equation is obtained,
564 A. Kumar et al.
d ∂L ∂L n
∂ V j d(V j )
T
− = tr Mj
dt ∂ θ̇i ∂θi j=i
∂ θ̇i dt
j
n
∂ p j + 0j R j c
+ m j − ρυ j g T (22)
j=i
∂θi
j
n
∂ V j d(V j )
T n
∂ p j + 0j R j c ∂f
τi = tr Mj + m j − ρυ j g T + (24)
j=i
∂ θ̇i dt j=i
∂θi ∂ θ̇i
Ui+1 13×3 − K i,i+1 Ui z i+1 × ci+1
= + θ̇i+1 (27)
ωi+1 03×3 13×3 ωi z i+1
where
di = vector from center of mass of the ith link to the (i + 1)th joint,
ci = vector from (i − 1)th joint to the center of mass of the ith link,
di + ci+1 = K i,i+1 ,
13×3 = 3 ×3 matrix containg all element as 1,
03×3 = 3 ×3 matrix containg all element as 0,
d(V T )
Defining, Bi = dti Mi ,
The torque equation can be made recursive as:
∂ Vi n
∂Vj
τi = tr Bi + tr Bj (29)
∂ θ̇i j=i+1
∂ θ̇i
∂ V2 ∂f
τ2 = tr B2 + h 2 + (30)
∂ θ̇i ∂ θ̇i
∂ V1 ∂ V2 ∂f
τ1 = tr B1 + tr B2 + h 1 + (31)
∂ θ̇i ∂ θ̇1 ∂ θ̇1
1
2
h 1 = (m 1 − ρυ1 )01 R 11 c + (m 2 − ρυ2 )p1 + h 2 and f = Ci Ui3 .
3 i=1
Fig. 2 Evolution of configurations of the links at different time instant after releasing the pendulum
from straight and horizontal configuration at rest and falling under gravity in fully submerged
condition
condition and released from rest. The parameters for the simulation and experiment
are taken as: lengths of the links to be L 1 = 0.179 m and L 2 = 0.165 m, the mass of
the links being m 1 = 0.136 kg and m 2 = 0.127 kg, respectively. Radius of both the
links is taken same to be 0.01 m. Drag coefficients are tuned to be equal to Cd1 = 1.3
and Cd2 = 1.5 for link 1 and link 2, respectively, in line of [10].
The response (in terms of variations of joint angles and velocities) is obtained
with respect to time. The results are shown in Figs. 2, 3 and 4. Figure 2 shows the
evolution of manipulator configurations with time.
θ1 starts from 0 to finally settle at π/2 and starting from 0, θ2 returns to 0, i.e.,
the stable equilibrium configuration. Little oscillations are observed around final
equilibrium. The simulation ignores various factors (skin friction loss, joints frictions,
etc.).
A two-link arm is built with the links of aluminium material and of dimension and
mass mentioned in the previous section. The joints are passive and free to move on
antifriction bearings. The same experiment with identical initial conditions as the
simulation is carried out in a glass tank with transparent walls. Free fall motions of
the manipulator were captured using high-speed camera system (from Xcitex Inc.)
at 150 frames per second and post-processed in ProAnalyst® to generate the time
evolution of position of the tip and the joint. The experimental setup and snapshots
of the freely falling manipulator are shown in Fig. 5.
Dynamics of Underwater Manipulator: A Recursive Lagrangian Formulation 567
(a) (b)
(c) (d)
Fig. 5 Snapshots of freely falling two-link underwater arm (pendulum) under gravity. Traces of
tip of the arm (in green) and elbow joint (in red) with time are also shown
The experimental data are plotted on same Fig. 6 vis-à-vis the simulation results.
Only little difference in the experimental and the simulation result is observed, which
is inferred mainly due to unmodeled joint friction and dimensional uncertainties in
Fig. 6 Comparison of experimental and simulation results for a fully submerged free falling two-
link manipulator (pendulum)
Dynamics of Underwater Manipulator: A Recursive Lagrangian Formulation 569
the manufactured links and their assembly. The simulation exhibits more oscillation
compared to the experiment. This shows the presence of unmodeled friction in the
simulation model.
6 Conclusion
Acknowledgements Part of the work has been done under the activities of the CSIR project MLP-
223512, being carried out at CSIR—Central Mechanical Engineering Research Institute, Durgapur,
India.
References
1 Introduction
[2] and a relation between roundness cutting parameters like cutting speed and feed
was established for the dry turning process. Kilic et al. performed oblique turning of
Aluminium workpiece [3] using carbide cutting tools at a variable range of speeds
and found different dynamics of friction at different cutting conditions. Allamraju
et al. discussed about the generation of residual stresses [4] during lower cutting
speed due to higher heat dissipation at the plastic deformation in shear zones and the
friction at the tool-workpiece interface in case of turning operation.
When the round shaped products come into consideration, the characterization
and mathematical formulation of machining processes somehow become difficult.
In most of the above literature, the product geometry was found to be simple. The
dependency of the machining output and the quality of the product on the morphology
of the chips [5] generated during turning has not been discussed in the previous work.
Therefore an investigation has been carried out by turning an Aluminium 2024-T3
workpiece on a CNC lathe machine using an HSS tool with 5% Co (M-grade SI 6)
with zero rake angle and its effect on the morphology of the chips were studied using
SEM analysis.
The rods of Aluminium 2024-T3 having dimension ϕ25 × 100 mm were used as
workpiece. The constituents of the workpiece material as obtained from energy
dispersive X-ray spectroscopy (EDS/EDX) are depicted in Table 1. The alloy contains
91.06% of Al. As the pure Al is very soft and becomes sticky during machining so
other impurities are added to it to increase its strength, machinability, wear resistance,
melting point and to reduce corrosion [6].
The use of Mg in 2.19% in Al alloy make the alloy high strength to density ratio
(HSTR) and so the alloy can be more useful in automobile and aerospace industries
[7]. Cu percentage (0.7%) in the Al alloy helps the alloy to resist stress corrosion
cracking (SSC) [8]. The copper provides substantial increase in strength and facili-
tates precipitation hardening. Cu also helps to reduce ductility and reduce corrosion
resistance. The introduction of Fe (0.97%) in Al alloys increase the strength of the
alloy due to formation of inter-metallic which has brittle features and stress concen-
tration caused by needle-like morphology. Mn is also added to the alloy to compen-
sate the negative effects of inter-metallic compound. Zn in the alloy increases anode
efficiency of the alloy. So the material can be used in any electrochemical or electro-
thermal process easily. The Pb content of the material increases the machinability of
the alloy. The properties of the workpiece material are depicted in Table 2.
An HSS tool with 5% Co (M-grade SI 6) has been selected as a cutting tool for the
following work. The rake angle was kept zero intentionally to set the basis of the
mathematical formulation. The following tool geometry is assumed for the process.
The tool geometry is illustrated in Fig. 1.
To develop the mathematical model on the effect of tool geometry on the turning
process, the following assumptions are taken into account:
(a) The cutting is performed orthogonally, i.e. ϕ = 90º.
(b) The rake angle is taken to be zero as the basis for the analysis.
(c) The shape of the job is parabolic in nature.
(d) The material properties of the workpiece is not a priority over the shape of the
geometry, i.e. strength of the material is not important in the analysis.
(e) To generate the shape the depth of cut and feed rate are same in magnitude and
perpendicular to each other in direction.
574 K. Mondol et al.
It is known that
cos γo k1 a1 k 1
tan βo = = =
ξ − sin γo a2
a1
− k2 a2 − k 2 a1
or,
k1 f sin ϕ
tan βo = (1)
a2 − k2 f sin ϕ
where,
k1 = cos γo ,
k2 = sin γo ,
a2
ξ=
a1
a1 = f sin ϕ.
z = −Rx 2 (2)
Differentiating w. r. t. t.,
dz dx
= −2Rx ·
dt dt
Effect of Tool Geometry on Chip Morphology Formed During … 575
Or,
f = −2Rx · d (3)
where,
dx
d = depth of cut =
dt
dz
f = feed rate =
dt
Now from Eq. (2)
Ps df
∴ As = =
τs sin βo
Case I:
As f = −2Rd x.
Equation (5) holds good for the oblique cutting (ϕ = 90º) with γo = 0.
Case II: Oblique cutting with γo = 0 gives Eq. (6)
Case III: For orthogonal cutting (ϕ = 90º) and zero rake angles, the equation
becomes:
a2 = cot βo · f (7)
576 K. Mondol et al.
Case IV: For orthogonal cutting (γo = 0), the equation becomes:
So it has been observed from Eqs. (5) and (6) that for oblique cutting, chip thick-
ness varies with sinusoidal function of γ o and ϕ but linearly with feed rate. In case
of orthogonal cutting, the chip thickness varies with feed rate linearly, but with
rake angle, it varies with a sinusoidal function and Eqs. (7) and (8) supports the
observation.
4 Experimental Verification
In order to verify the above mathematical model, the experiments were performed on a
Sine wave CNC turning centre at Advanced Manufacturing Laboratory of University
of Engineering & Management, Kolkata. The experimental setup is shown in Figs. 2.
The generated chips were collected for three different positions (a. initial, b. middle
and c. final) of the round shaped job.
Figure 3 depicts the comparison between the predicted values and the experimental
values of the cut chip thickness under similar machining conditions and process
parameters. The predicted chip thickness values from the proposed mathematical
model for the above machining conditions are more precise and follow a specific
pattern similar to the experimental results. The experimental results are somewhat
scattered and this might be due to the instrumental or measurement error. So, if the
errors are ignored, the predicted values are well in agreement with the experimental
values.
0.18
0.16
0.12
0.1
chip thickness exp
0.08
0.02
0
0 10 20 30 40
EXPERIMENT NO.
5 Discussion
Figure 4 depicts SEM analyses of produced chips during machining at spindle speed
of 40 rpm, 20 mm/min feed rate and depth of cut of 0.5 mm, level settings have been
carried out. It was observed that the layers of the material are very rough with a clear
marking of tool nose on the machining side of the chip. The materials have been
removed in the form of layers. The portion of the chips is magnified which shows
that the strata are rough but unbroken which proves the material to be ductile.
The key inspiration of this research is to understand the depiction of chip formed
at the curved portion, i.e. the middle portion of the round shaped job. From Fig. 4b
it has been observed that material has also been removed layer by layer in this case
but the layers are very fine and also markings by the tool nose are comparatively
thinner. During machining by the cutting tool with zero rake angles, the better layers
are formed at the middle portion of the curve of the round shaped job.
Again Fig. 4c shows that lengthened curved chips with fissure have been formed
at the end position of the curved portion. This is because of the fact that at the end
position of the curved shape, normal turning has been occurred, and it has been
carried out with the cutting tool of zero rake angles. The other geometry of the tool
is also responsible for the generation of such shape.
578 K. Mondol et al.
a. b.
c.
Fig. 4 SEM analysis of the chip: a at initial, b at middle and c at end position
6 Conclusion
References
1. Bhattacharya A, Metal cutting—theory and practice. Book, New Central Book Agency, Kolkata
2. Marcos-Barcena M, Sebastian-Perez MA, Contreras-Samper JP, Sanchez-Carrilero M, Sanchez-
Lopez M, Sanchez-Sola JM (2005) Study of roundness on cylindrical bars turned of aluminium–
copper alloys UNS A92024. J Mater Process Technol 162–163:644–648
3. Kilic DS, Raman S (2007) Observations of the tool–chip boundary conditions in turning of
aluminum alloys. Wear 262:889–904
4. Rao KSS, Allamrajub KV (2017) Effect on micro-hardness and residual stress in CNC
turning of aluminium 7075 alloy. In: 5th international conference of materials processing and
characterization (ICMPC 2016). Mater Today Proc 4:975–981
5. Piispannen V (1948) Theory of formation of metal chips. J. Appl Phys 19(10):876
6. John M (Tim) Holt, Ho CY (eds) (1996) Structural Alloys Handbook. CINDAS/Purdue
University, West Lafayette, IN
7. Sony GL (2013) Corrosion prevention of magnesium alloys
8. Kannan MB, Shukla PK (2011)Stress corrosion cracking
Optimization of Equal Multi-square Cell
Crash Box for Enhanced Energy
Absorption
Abstract The crash box absorbs impact energy during a low speed crash. Multicell
square tubes are proven to absorb energy better than square or rectangular cross-
section. The present work is focused on optimization of the cell parameters like side
length, thickness and height of a symmetric multicell square cell crash box. The
optimization is carried out considering the volumetric space available in a passenger
car as constraint. Advanced process monitor is used for solving the optimization
problem. The performance of the crash box is evaluated using an analytical model and
validated with simulations (LS-DYNA) and experimental results (UTM). The results
obtained from three approaches are compared and percentage errors are reported.
1 Introduction
The crash box is a crumple zone located behind the front-end bumper. The crash
box absorbs the kinetic energy generated during a crash by undergoing defor-
mation. Different cross-sections of the crash box are developed to improve the
energy-absorbing capacity. Cylindrical tubes and square/rectangle cross-sectional
tubes find their application in use as a crash box. In order to improve the energy-
absorbing capacity multicell configurations are introduced into the existing cylinder
and square/rectangle cross-sectional crash boxes. The multicells can either be of same
dimensions and symmetric configuration or combination of different configuration
and dimensions. The present work considers symmetric multicell square tubes. Opti-
mization of this configuration is carried out in order to enhance the energy-absorbing
capacity.
2 Mathematical Model
Zhang et al. [2] generalized mean crushing load expression for multi-square cell of
equal cell size. Mean crushing load is given by,
Pm = σ y t (Nc + 4N O + 2N T )π L C t (1)
where
L C is the total flange length.
Nc , No , N T is the number of corner, orthogonal, T elements.
σ y is yield stress of the material.
The elements considered for the multicell tube is shown in Fig. 1. The
configuration of a multicell tube can be expressed as
m×n (2)
where
m, n is the number of square tubes length wise, width wise.
For a symmetrical square tubes, m = n and Eq. (2) n be rewritten as m × m.
Nc = 4 (3)
N O = (m − 1)2 (4)
N T = 4 (m − 1) (5)
Assuming the deformation of the multicell tubes to continue until 75% of the total
height the energy absorbed by the multicell tube is given by
Corner element
a
Optimization of Equal Multi-square Cell Crash Box for Enhanced … 583
E Abs = 0.75hσ y t 4 + 4(m − 1)2 + 8(m − 1) π (2m + 2) a t. (8)
The theoretical expressions derived for Pm and obtained in Eqs. (7) and (8) are
used to define the analytical model in Sects. 3 and 4.
Fig. 2 Configuration 2 × 2
(a) t = 0 s (b) t = 9 s
Fig. 3 Configuration 3 × 3
(a) t = 0 s (b) t = 9 s
Fig. 4 Configuration 4 × 4
(a) t = 0 s (b) t = 9 s
584 G. Prasad et al.
Force (kN)
60
2x2
40
20
0
0 20 40 60
Displacement (mm)
mentioned in Table 1 are meshed with quad elements of mesh size 0.5 mm. The
termination time is 10 s and the time step provided is 0.09 ms.
The force versus displacement curve obtained from LS-DYNA for the models
mentioned in Table 1 are shown in Fig. 8.
The area under the curve gives the energy absorbed by the multicell tube. The
average of the forces experienced by the model throughout the deformation cycle
gives the value of the mean force Pm .
Table 1 Configurations considered for simulation
Configuration a (mm) t (mm) h (mm) Simulation Analytical % Error
Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ)
2×2 18 1 60 14.2 0.64 15.8 0.70 10.13 9.78
3×3 18 1 60 26.2 1.04 27.3 1.22 4.03 14.90
4×4 18 1 60 55.6 1.88 58.3 1.98 4.55 5.05
Optimization of Equal Multi-square Cell Crash Box for Enhanced …
585
586 G. Prasad et al.
4 Experimentation
5 Parametric Study
The effect of variation in geometric parameters on the mean crushing load and energy
absorption is studied in this section. The analytical l was defined in MATLAB and
the study was carried out by iterating the geometric values. The variation of height
was found to have no effect on the mean force required to deform the crash box.
The variation of height had a positive influence on the E Abs by the crash box.
On an average, a 10 mm increase in height was found to increase the E Abs by 13%
(see Fig. 13). An increase in side length of the symmetric multicell tube has found
to increment both the Pm and E Abs of the crash box, but an increasing side length
meant an higher Pm was required to deform the crash box which is not a positive
attribute of a crash box. On an average, a 10 mm increase in side length was found
Table 3 Comparison of experimental results with analytical results
Sample Experimental Analytical Error % Simulation Error %
Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ) Pm (kN) E Abs (kJ)
1 7.4 0.3 7.1 7.1 1.35 6.67 1.35 6.67 4.05 6.67
2 16.8 0.72 14.2 14.2 5.95 2.78 5.95 2.78 15.48 11.11
3 29.55 1.11 26.2 26.2 7.61 9.02 7.61 9.02 11.34 6.31
Optimization of Equal Multi-square Cell Crash Box for Enhanced …
587
588 G. Prasad et al.
0.4
0.2
0
0 20 40 60 80 100 120
Height (mm)
by change in thickness
2x2
3x3
5 4x4
0
0 0.5 1 1.5 2 2.5 3
Thickness (mm)
to increase both Pm and E Abs by 23%. The parametric study yielded thickness as the
most sensitive variable. A 0.5 mm increment in thickness on an average caused a
twofold increase in energy absorbed (see Fig. 14). The optimization code allows to
set a priority for the design variables as per their contribution to their overall energy
absorption.
Optimization of Equal Multi-square Cell Crash Box for Enhanced … 589
60
Force (kN)
40
20
0
0 20 40 60 80
Displacement (mm)
Start
Generate Intermediates
True
False
Check mass constraint Evaluate Mass
True
Display optimized Parameters Terminate Iteration Stop
The optimizer is run, and the results obtained after achieving the objective criteria
are tabulated in Table 4.
Simulation of the multicell tube with the optimized geometrical parameters are
carried out and the simulation at different intervals are shown in Fig. 18. The Force
versus Displacement plot obtained from LS-DYNA is shown in Fig. 19. A comparison
of the analytical and simulation results is shown in Table. 4.
7 Conclusions
The maximization of the energy absorption for multi-square cell configuration under
dynamic axial crushing is conducted in this study. Expressions for Pm and E Abs
mentioned in Sect. 2 are modified for symmetric multicell tubes and the expression
is derived in terms of its configuration m. The nonlinear explicit finite element code
LS-DYNA is adopted to simulate the axial crushing process. Results from theoretical
formulae are validated with simulation model and experimental values with minimal
percentage error and were in good agreement with Zhangs model [2].AP Monitor
is used to obtain the optimum design based on the volumetric constraints and is
Optimization of Equal Multi-square Cell Crash Box for Enhanced … 591
shown to be efficient and effective in maximizing the energy absorption of the crash
box. The results show that for symmetric multi-square cell tubes for a given mass
constraint, an appropriate side length, thickness and configuration must be selected
to achieve the highest energy absorption A higher energy absorption is achieved
keeping in consideration the mean force encountered by the crash box during a low
velocity crash so as to fulfil its function as a crash box.
Reference
Abstract The productivity and the quality of machining depend on the dynamic
characteristics of spindle-holder joint interface. However, spindle-holder system is
one of the weak joint in the machining centres; therefore, it causes poor machining
performances. The main objective of machine tool industries is to improve the accu-
racy of the machined component, achieve good surface finish, increase tool life and
metal removal rate. To achieve these requirements, it becomes predominant to iden-
tify the joint stiffness for understanding the dynamic characteristics of spindle-holder
joint interface. In this paper, author discussed in detail on how the joint stiffness plays
an important role in the dynamics by considering the surface fractal topology theory.
Here, the 3D CAD model of BT-40 spindle and Tool holder has been modelled and
numerically analysed in FEA software by considering the typical Clamping forces,
then obtained natural frequency from FEA is correlated with the experimental modal
analysis results, this study will be more informative for understanding the influence
of joint stiffness behaviour on the dynamics of spindle-holder system.
1 Introduction
The CNC machines are the most complicated structure composed of more number
of substructures; these subsystems are usually assembled through different types
of joints like the bolt connections, the guideway connection, the welding connec-
tions etc. The presence of these joints directly influences the dynamic stiffness and
damping of the machine tool. Research shows that about 60% of overall dynamic
stiffness and about 90% of total damping are caused by the joints [1]. These joint
parameters directly influence the dynamic behaviour of the overall machine tool, it
affects the machine performance like surface finish, tool life, quality of the machined
component and metal removal rate etc. In order to improve the performance and accu-
rate modelling of machine tool dynamics is the knowledge of connection dynamics,
i.e. joint stiffness and damping at the interface [2]. In which the spindle-holder-tool
joint is one of the weakest joints in the machine tool, and the dynamic behaviour
of the spindle-holder-tool system changes as change in any one of the subsystem,
hence it consumes more time to develop the analytical dynamic model for analysis
and experimentation. In order to reduce experimentation Schmitz et al. [3] introduced
a reacceptance coupling technique based on the basic substructure analysis method.
Machine tool consists of more number of joints, presence of these joints decreases
the overall stiffness of the machine structure, this effect on the performance of the
machine tool, hence it is necessary to maintain the optimal joints for obtaining the
optimal stiffness so as to improve the overall stiffness of the structures. Similarly, the
joints also should possess damping effect under dynamic loading conditions. This
damping effect increases the overall damping of the structures. Thus, in general, the
amount of damping decreases with an increase in the static stiffness of the structure
[4]. Hence, it emphasises to design the machine tool to make best compromise
between damping capacity requirement and the normal static stiffness. The stiffness
of the fixed joint is defined as the ratio of normal force acting on the system to the
defection of the joints. The deflection is usually of two parts. 1. Deflection of the
solid material of the joint and deformation of surface asperities in contact at the joint
interface [4].
However, the most common approach to model the spindle-tool holder joint is
the substructure coupling analysis based on experimental method; in this method,
the spindle, tool holder and joints, joints are assumed to be combination of spring
and damper system. The frequency response function of individual substructures
and the assembly are obtained by experimental method and the numerical technique.
Then the stiffness and damping of the joints are calculated based on the reacceptance
coupling substructure analysis (RCSA) method [3]. Wang and Horng [6] identified
the spindle, tool holder joint stiffness and the effect of axial clamping forces. Mehdi
Namazi [7] modelled the spindle, tool holder interface by uniformly distributed
translational and rotation springs to identify the contact stiffness and damping at
the interface region. The spring stiffness is identified to reducing the errors between
experimental and estimated frequency response function. Agapiou [8] demonstrated
the methodology to determine the stiffness between tool holder and spindle interface,
but results show that the importance to establish the standardized experimental setup
in machine tool setup not just on the bench setup as it had been the case in the past.
Some researchers were worked on the simulation of contact stiffness and presenting
the elastic behaviour and stress distribution at the joints. In general, the real area of
contact between the two flat surface areas is the plastic deformation of the longer
asperities and it depends on the load acting between the joint interface and indepen-
dent of the apparent area of contact. Greenwood and Williamson [9] describe the
new theory of elastic contact it shows how the contact deformation depends on the
topology of the surfaces. These surfaces asperities are assumed to be spherical and
of uniform radius of curvatures and gauss distribution height. However the actual
contact surface area mostly multiscaled with non-uniform asperities. Topology of the
Joint Stiffness Estimation Between Spindle-Tool Holder … 595
machined steels by the measurement of roughness indicates the surface has multiscale
and random. Majumdar and Tien [10] and Majumdar and Bhushan [11] identified
fractal dimensions is the intrinsic property of the multiscaled structures and they
used Weierstrass-Mandelbrot (W-M) fractal function to introduce a new and simple
method of roughness characterization. These well-discussed researches provide new
insights into the study of joint stiffness between spindle-tool holder interfaces.
In this paper, author discussed the joint stiffness by fractal Topology theory at the
spindle-tool holder joint interface. This joint stiffness influences dynamic behaviour
of spindle by studying the Eigen frequency analysis. Here, the asperities in microscale
at the spindle-tool holder interface are considered to undergo elastic-plastic defor-
mation. Experiments are conducted to correlate the influence of clamping force on
contact stiffness with the theoretical calculated contact stiffness obtained by fractal
topology, and is useful to understand the dynamic characteristics of spindle system.
The spindle-holder system is one of the weakest joint in the machine tool structures;
hence the estimation of its dynamic inherent characteristics such as natural frequency
is having more importance. This is useful to understand the performance of the spindle
under high speed. Here, author considered one of the BT40 spindle shaft and tool
holders for the study is as shown in Figs. 1 and 2, respectively.
The BT-40 spindle shaft has the contact with taper angle 8.29°, whose spindle
and holder tapers are in tight contact based on the clamping force Fd . The major and
minor diameters of taper contacts are r0 and r1 , respectively, where ‘l’ is the axial
length of the taper.
The actual contact interface area of spindle and the tool holder is expressed as in
Eq. (1).
l
Aa = π (r0 + r1 ) (1)
cos α
The spindle and tool holders are grinded with high precession and subjected to
better heat treatment process, even though the surface has random, multiscaled and
disordered property. It is usually assumed that the real contact area between spindle-
tool holder substructures is very small and depends on the load induced between
that interface and independent of the apparent area of contact [9]. Developed the
new theory that shows how the contact deformation depends on the topology of the
surfaces and established the criterion which distinguishes the elastic deformational
contact of aspirates from the plastic deformation of the aspirates. So that surface
topology function should be continuous all over the surface but non-differentiable
at nowhere, it indicates that surface profiles can be simulated by using Weierstrass-
Mandelbrot function [10] which is given by Eq. (2).
∞
G 2π γ n x
γ (D−2)n cos ∅1n − cos
1
h(x) = L (ln γ ) 2 × − ∅1n (2)
L n=0
L
where h is the surface roughness height, x indicates the distance, L is the characteristic
length of the surface and the area could be derived by L 2 = Aa (nominal area of the
surface), G is the fractal roughness parameter (Characteristic length scale), where
γ = 1.5 is suitable value for the high spectral density and for phase randomization
[11]. In order to find out the surface roughness parameter (G) and fractal dimensions
(D) is calculated by the power spectrum of W-M function [11] is shown in Eq. (3).
G 2(D−1) 1
S(ω) = ∗ (5−2D) (3)
2 ln γ ω
where S(ω) is the power of spectrum and ω is the frequency that is reciprocal of the
roughness wavelength, S(ω) is plotted as a function of ω on log–log plot, then the
law shows the resultant line, then the slope of line would be related to the fractal
Joint Stiffness Estimation Between Spindle-Tool Holder … 597
dimension (D) and the Surface roughness parameter (G), and they are independent
of the frequency ω.
The spindle and tool holder contacting surfaces has an equivalent young’s modulus
can be calculated from Eq. (4).
−1
∗ 1 − ϑ12 1 − ϑ22
E = + (4)
E1 E2
where ϑ1 , ϑ2 and E 1 ,E 2 are the Poisson’s ratio and young’s Modulus of two
contact surface material, respectively. If two contact surfaces aspirates are in elastic
or plastic deformation depends on the truncated area, and the deformation depth of
the aspirates of peak-valley amplitude of W-M function is as shown in Eq. (5).
Here we assumed some parameters are negligible, those are 1. The surface rough-
ness is statistically isotropic during contact, 2. Work hardening due to elastic and
plastic transition, 3.Frictional force induced by the contact interface and 4. Change
of material hardness with depth from surfaces.
Greenwood Williamsons assumed that all aspirates are of equal radius of curva-
ture and the radius of curvature depends on the size of contact and is incorporated in
the fractal model. In this theory, the micro-contact aspirates are assumed to undergo
elastic deformations, the elastic deformation of the micro-contact can be determined
based on the hertz theory and micro-contact aspirates are obtained by Fractal dimen-
sions. Mujumdar and Bhushan [11] assume that the critical area is the region which
distinguishes the area by abrupt transition from elastic to plastic deformation. As
per Mujumdar and Bhushan [11] theory suggested that the smaller micro aspirates
undergo plastic deformation and the larger aspirates undergo elastic deformation, and
it is an intrinsic property of the material. Then elastic spherical aspirates compressed
by a rigid plane, the aspirate interface δ is given by hertz theory, the deformation of
spherical aspirates with the contact plane is as shown in Fig. 3.
According to Mandelbrot theory, the region which separates the elastic defor-
mation from the plastic deformation of the aspirates are expressed by Eq. (6)
[5].
2 1
E∗
(D−1)
where b = πq
2
2
and q = 0.454 + 0.41ϑ1 , ϑ1 is the poison’s ratio of the softer
surface, K is the factor relating to hardness to yield strength and its typical value K
= 2.8. Next section will discuss the two situations.
598 R. H. Aralaguppi et al.
The aspirates at the contact interface will undergo elastic deformation if the truncated
aspirates satisfy a > ac then the deformed contact area is as shown in Eq. (7) [14].
a
ae = (7)
2
The force which induced due to elastic deformation of the aspirates by hertz
spherical contact theory is given by Eq. (8) [14].
2( )
9−2D
2 (3−D)
(ln γ ) 2 G (D−1) E ∗ a
1
fe = 2 (8)
3π ( )
3−D
2
∂ fe
kn = (9)
∂δ
Substitute Eq. (8) and (5) in Eq. (9) the normal stiffness of the asperity can be
written as
√
2 2(3 − D) ∗ 1
kn = √ E (a) 2 (10)
3 π (2 − D)
Joint Stiffness Estimation Between Spindle-Tool Holder … 599
The plastic deformation of the aspirates whose truncated area ‘a’ satisfies, i.e.a < ac ,
the deformed contact area (a P ) is equal to the truncated area ‘a’ and the plastic contact
force f p are given in Eq. (11)
f p = K σy a (11)
D D/2 −(D+2)
n(a) = a a 2 (12)
2 l
where al is the surface’s largest truncated aspirates, and it varies with the change in
deformation depth of contact interface. The critical area is the intrinsic property of
the material and the fractal topology parameters and is as shown in Eq. (6).
The contact behaviour changes with the change in value of al and ac , i.e. when
the largest truncated area is less than the critical area (al < ac ), then all the aspirates
at the interface will undergo plastic deformation. Nevertheless in the case (al > ac )
small aspirates with truncated area (a < ac ) are plastically deforms and for (a > ac )
are undergoes elastic deformation. Therefore, it undergoes both elastic and plastic
deformation [5].
For the plastic deformations of the contacted aspirates, the stiffness induces at the
interface are almost equal to zero.
For elastic–plastic deformation of the contact interface aspirates, the total force
and the total real contact area consists of two parts, i.e. elastic and plastic deformation.
Hence, the total elastic–plastic force and area are as shown in Eqs. (13) and (14) [5].
D K σ y (2−D) c1 D (3−2D) D c1 D (3−D)
F= ac 2
− ac 2
al 2 + al 2 (13)
2− D 3 − 2D 3 − 2D
D 3D (2−D) D
Ar = al + ac 2 al 2 (14)
2(2 − D) 2(2 − D)
Kn = kn n(a)da (15)
ac
600 R. H. Aralaguppi et al.
Substitute Eq. (10) and (12) in Eq. (15) the final equation of the contact stiffness
is obtained as.
√ (2−D)
2 2D(3 − D) D 1
Kn = √ E ∗ ac 2 al 2 − al2 (16)
3 π (2 − D)(D − 1)
The study and understanding of the behaviour of joint parameter of spindle and the
tool holder interface is most significant task for the machine tool designers, because
the stability of the machining component greatly depends on the stiffness between
the spindle and the tool holder joint. The stiffness at that interface changes with the
change in the critical area of aspirates contact and largest truncated area of contact
aspirates and this behaviour depends on the tool holder clamping force. Hence,
here we consider the fractal topology study to calculate the interface stiffness using
roughness of the spindle and tool holder.
Here, author considers BT-40 spindle shaft and the tool holder whose dimensional
parameters are considered as per ISO standard and its taper angle is 8.29°. The yield
strength for the softer surface σ y of the spindle and tool holder and young’s modulus
E 1 and E 2 , respectively. Whose values are tabulated in Table 1.
The spindle and the tool holder substructure are well grinding with better heat
treatment process having the roughness value is around 0.2 μm, for this surface
roughness parameter, the characteristics length G and the Fractal dimensions D are
calculated using the power spectral density which was reported by Zhao [15]. These
values are tabulated in Table 2.
The spindle and the tool holders are connected by using the bolted mechanism
to induce the different clamping forces. One end of the tool holder is connected
to the long bolted screw passes through the hallow section of the spindle shaft,
and then other ends of the bolts are used to vary the different clamping force by
tightening the nut through flange (rear end of the spindle shaft). Here, author considers
different condition of clamping force to understand how the fractal surface behaves
as change in the contact area of the truncated aspirates, this means to say that the
largest truncated area of contact of the aspirates changes with the change in clamping
forces. Hence, the projected normal contact force induced on the contacted interface
for different clamping force can be calculated using the following Eq. (17)
F = Fd sin α (17)
From blue match technique, visual area of contact is around 90% of the apparent
area of contact. But in reality, the real area of contact is extremely small. However,
the actual contact area is Aa = 7.49×10−3 m2 from Eq. (1) and for 12 kN of drawbar
force the real area of contact Ar = 1.02 × 10−6 m2 (from Eq. (14)).
As per the above discussion of the bolted mechanism, author used the 16 mm
diameter steel bolt, and its co-efficient of friction is 0.2. By using this parameter the
clamping force can be altered for different torque and is tabulated in Table 3.
Where F d is the clamping force, F is the normal projected force at the contact
interface and P is the contact pressure. These forces influence the joint interface
behaviour this means to say that, as the force changes the joint stiffness also varies.
This is because of elastic–plastic deformation of the aspirate. Then the largest area
of the truncated aspirates can be calculated by using Eq. (13) similarly the critical
area of aspirates for the fractal parameters of roughness value 0.2 μm is ac = 7.37 ×
10−8 m2 calculated by using Eq. (6). The largest truncated are for different condition
of drawbar force is as tabulated in Table 4.
From the above calculation, the largest area of the truncated aspirates is greater
than the critical area of contact; this concludes that deformation of the aspirates at the
spindle-tool holder interface is elastic–plastic in nature. Then the total real area of
contact can be calculated by integrating the distribution function n(a) from smallest
contact area to largest truncated contact aspirates.
As per the fractal topology theory, the stiffness induced at the joint interfaces
depends on the fractal parameters and different clamping forces, for that the overall
stiffness induced at the interface is calculated by using Eq. (16). Similarly, for
different clamping forces, the overall stiffness induced is as tabulated in Table 5.
The above calculation indicates that the ratio of apparent area of contact to real
area of contact is extremely small (e.g. Ar /Aa = 0.013618% for 12 kN drawbar
forces).
In order to verify the efficiency of theoretically calculated contact stiffness value,
a finite element model of Eigen frequency study is established in the COMSOL
multiphysics software. Then obtained Eigen frequencies are correlated with the
frequencies of experimental modal analysis.
The dynamic inherent characteristics like natural frequencies and mode shapes of
the spindle and tool holder can be experimentally calculated using modal analysis
technique. These results will indicate the dynamic behaviour of spindle –tool holder
subsystem and how the structures will response to the dynamic loading.
To obtain the natural frequencies of spindle and tool holder substructure for
different clamping forces (or contact stiffness) is need to be subjected for experi-
mental modal analysis. The different clamping force can be induced by varying the
torque of tighten nut, the arrangement of spindle with bolted mechanism is as shown
in Fig. 4.
In experimental modal analysis, the free-free boundary condition is achieved by
hanging the spindle using optimum stiffen spring is as shown in Fig. 5.
Case 1
In the first case, we consider the spindle shaft and the tool holder is held by projected
normal contact force of 0.9 kN (i.e. for 20 Nm Torque and 6250 N Clamping force).
Spindle shaft and tool holder consists of 52 marked points. The frequency response
(FRF) obtained at all the marked point on the spindle shaft is as shown in Fig. 6.
The intermediate mode shape at frequency around 2300 Hz, mostly indicates the
maximum deformation at the spindle-tool holder joint section, and therefore, it can
conclude by the joint stiffness (3.35 × 108 N/m). The mode shape at that frequency
is as shown in Fig. 6 and the frequency response function is as shown in Fig. 7.
Case 2
In second case, author considers the typical condition of 12.5 KN clamping force,
by providing the torque rate around 40 Nm, whose joint stiffness, i.e. K n = 4.71 ×
108 N/m for that the frequency response function is taken from all marked 52 points
including both spindle and tool holder is as shown in Fig. 8. From this frequency
604 R. H. Aralaguppi et al.
analysis, the mode shape at frequency 3360 Hz indicates the maximum deformation
occurs at the spindle and tool holder interface, and therefore, it is mainly decided by
the contact stiffness.
The 3D CAD model of spindle shaft and tool holder is modelled, whose dimensions
are considered as per the ISO standards, the CAD model is as shown in Figs. 9 and
10.
Joint Stiffness Estimation Between Spindle-Tool Holder … 605
3208 Hz indicates the maximum deformation occurs at the spindle and tool holder
interface, and therefore, it can be decided by the joint stiffness at that interface, the
mode shape of that frequency is as shown in Fig. 11.
From the above analysis, the intimidated frequency obtained by experimental
result, i.e. 3360 Hz is predominantly influenced the maximum deformation at
the spindle and tool holder interface. Similarly, the numerical study indicates the
maximum deformation occurs at spindle and tool holder to 3208.8 Hz (for K n = 4.71
× 108 N/m joint stiffness). Hence, the theoretically calculated stiffness values from
fractal topology theory and their corresponding dynamic results from the numerical
analysis are almost correlated with the dynamic analysis experimental result.
5 Conclusion
In this paper, author considered the fractal topology theory to understand the effect of
joint stiffness by the dynamic behaviour of spindle shaft and the tool holder substruc-
ture assembly. The spindle and the tool holder substructure assembly is the delicate
joint in the machine tool, and hence, it is the most significant task to understand
these joint parameters. Many researchers were worked on this to understand the
joint parameters by using Reacceptance Coupling Substructure Analysis (RCSA)
method and Surface Fractal Topology Theory. Here, author considers the surface
fractal topology theory to understand the joint parameter. In this real contact area of
the spindle shaft and the tool holder surfaces are less than the apparent contact areas.
Joint Stiffness Estimation Between Spindle-Tool Holder … 607
This contacted area is calculated by integrating the aspirates present on the surface,
which is obtained based on the surface roughness parameters like fractal dimension
and the characteristics length. Then it is necessary to understand the relationship
between clamping force and joint stiffness. This can be derived based on the contact
stiffness model developed by surface fractal topology theory. The obtained theoret-
ical calculated joint stiffness value is used in Eigen frequency study to understand the
influence of joint stiffness on the dynamic behaviour of the spindle and tool- holder
substructure. Then obtained frequencies and mode shapes in FEA are correlated with
the frequencies obtained in experimental modal analysis. This experimental result
shows the efficiency of proposed fractal topology model in contact stiffness deter-
mination. Therefore, understanding the joint behaviour is most prominent for the
machine tool designers to optimize the joints.
References
1. Zhang GP, Huang YM, Shi WH, Fu WP (2003) Predicting dynamic behavior of a whole machine
tool structure based on computer aided engineering. Int J Mach Tool Manuf 43(7):699–706
2. Budka E, Erturk A, Ozguven HN (2006) A Modeling approach for analysis and improvement
of spindle-holder-tool assembly dynamics. CIRP Ann 55(1):369–372
3. Schmitz T, Smith L, Scott SK (2009) Machining dynamics-frequency response to improved
productivity. Springer. ISBN 978-0-387-09645-2
4. Chikate PP, Basu SK (1975) Contact stiffness of machine tool joints. Tribol Int 8(1):9–14
5. Zhao Y, Song X, Cai L, Liu Z Cheng Q (2014) Surface fractal topography-based contact
stiffness determination of spindle-tool holder joint. J Mech Eng Sci 0(0):I–9
6. Wang JH, Horng SB (1994) Investigation of tool holder system with a taper angle. Int J Mach
Tool Manuf 34(8):1163–1176
7. Namazi M, Altintas Y, Abe T, Rajapakse N (2006) Modelling and identification of tool-holder
spindle interface dynamics. Int J Mach Tool Manuf 47(9):1333–1341
8. Jhon S (2005) Agapiou a methodology to measure joint stiffness parameter for tool holder
spindle interface. J Manuf Syst 24(1):13–20
9. Greenwood JA, Williamson J (1996) Contact of nominally flat surface. Proc Royal Soc Mathe
Phys Eng Sci 295(1442). Online ISSN 2053–9169
10. Majumdar A, Tien CL (1990) Fractal characterization and simulation of rough surface. Wear
136(2):313–327
11. Majumdar A, Bhushan B (1991) Fractal model of elastic-plastic contact between rough surface.
American Soc Mech Eng J Tribol 113(1)
12. Johnson KL Contact mechanics. Cambridge University Press, Cambridge, UK, pp. 153–
155,175–176, 196–201,208–210 and 427–428.
13. Sackfield A, Hills DA Some useful results in the tangentially loaded hertzian contact problem.
J Strain Anal 18:107–110
14. Wang S, Komvopoulos K (1994) A fractal theory of the interfacial temperature distribution
in the slow sliding regime: Part II—multiple domains, elastoplastic contacts and applications.
American Soc Mech Eng J Tribol 116(4)
15. Zhao Y, Song X, Cai L (2015) Surface fractal topology based contact stiffness determination
of spindle holder joints 230(4)
Buckling Analysis of Nonlinear
First-Order Shear Deformation
Composite Plates
Abstract This paper developed the buckling analysis of nonlinear behavior of first-
order shear deformation laminated composite plates. The first-order displacement
functions are used based on some simplifying assumptions. This article further
simplifies the transverse displacement into bending and shear components at the
midplane have reduced the number of variables and governing differential equa-
tions. The von Karman assumptions are used to derive the nonlinear strains based
on the displacement functions. Hamilton’s principle is used to derive the equation
of motion and boundary conditions. The analytical solutions are developed from the
present theory and compared the results of the critical buckling loads with exact
and three-dimensional solutions. Result shows that the present theory has closely
associated with the exact and three-dimensional theory.
1 Introduction
A. Maji (B)
Asansol Engineering College, Asansol 713305, India
P. Kr. Mahato
Indian Institute of Technology (ISM), Dhanbad 826001, India
defeat the constraints of CLPT, Reissner [7] first developed the shear deformation
theory with the effect of transverse shear deformation and rotary inertia. Mindlin
[8] developed a plate theory based on first-order well known as Mindlin’s first-order
plate theory [9–11] where he considers the transverse shear effects in conjunction
with shear correction factors [12–15]. Many studies are presented based on the buck-
ling of laminated composites plates using first-order shear deformation plate theory
(FSDPT) can be found in Refs. [16–19].
This paper is to develop the buckling analysis of nonlinear behavior of orthotropic
type laminated composite plates by using FSDPT method. This paper considers Von
Karman nonlinear strain to develop the relations between strain and displacement.
The conventional first-order shear deformation theory has five functions and these
have been done by considered constant transverse normal strain and transverse shear
strains at the top and bottom surface of the laminates are zero by using the theory of
Lo et al. [20]. In this article, we simplifying transverse displacement w = wb + ws ,
where wb and ws are the bending and shear component of the midplane respectively
initially proposed by Refs. [21–23] and also φx = −∂ Wb /∂X and φ y = −∂ Wb /∂y
where φ is the rotation vector about x and y axis, respectively. Split of transverse
displacement in the FSDPT method the number of unknown functions and governing
differential equations has been reduced to four. This assumption makes in-plane
rotation is constant within the laminate plane and has little effect on the stability of
the orthotropic and composite laminate plates. Hamilton’s principle is used to derive
the equation of motion and boundary condition. A closed-form solution has been
developed and compared the results with the exact and three-dimensional theory.
2 Theoretical Formulations
Consider a laminated plate of thickness h and edge dimension a and b. The plate is
assumed to a Cartesian coordinate system x-y-z, where x, y plane is the middle plane
of the plate and z axis is normal to the middle surface of the plate. For laminated
composite plates, the basic assumption is taken as. (Fig. 1).
1. The displacement is small in comparison with the plate thickness ‘h’ and
therefore, consider plain stress problems.
2. The transverse displacement ‘w’ dividing three components extension wa ,
bending wb and shear ws . These components are functions of coordinates x
and y only.
2.2 Kinematics
where u0 , v0 and w0 are the in-plane and transverse displacement functions of the
corresponding coordinate points x and y on the reference surface and φ x and φ y are
the rotations about y and x axes respectively of the normal to the mid-surface in the
undeformed plates.
The transverse displacement ‘w’ making further assumptions that the extension
is very low as compare to bending and shear {i.e., w(x, y) = wb (x, y) + ws (x, y)}.
Also, φx = −∂wb /∂x and φ y = −∂wb /∂y, the displacement field of the FSDPT
can be rewritten as [21]:
∂wb
u = u0 − z
∂x
∂wb
v = v0 − z
∂y
w = wb (x, y) + ws (x, y) (3)
The Von Karman nonlinear strain associated with the displacement in Eq. (3) is
given by
∂u ∂u 0 1 ∂wb ∂ws 2 ∂ 2 wb
ε1 = εx = = + + −z
∂x ∂x 2 ∂x ∂x ∂x2
2
∂v ∂v0 1 ∂wb ∂ws ∂ 2 wb
ε2 = ε y = = + + −z
∂y ∂y 2 ∂y ∂y ∂ y2
612 A. Maji and P. K. Mahato
∂w
ε3 = εz = =0
∂z
∂ws
ε4 = 2ε yz =
∂y
∂ws
ε5 = 2εzx =
∂x
∂u 0 ∂v0 ∂wb ∂ws ∂wb ∂ws ∂ 2 wb
ε6 = 2εx y = + + + + − 2z (4)
∂y ∂x ∂x ∂x ∂y ∂y ∂ x∂ y
T
0= (δU + δW − δ K )dt (7)
0
where δU, δW and δK are the virtual strain energy, virtual work done by applied
external forces and virtual kinetic energy, respectively.
Buckling Analysis of NonLinear First-Order Shear Deformation … 613
a b +h/2
δU = σ1 δε1 + σ2 δε2 + σ6 δε6 + σ4 δε4 + σ5 δε5 dx d y dz
0 0 −h/2
⎧ ⎫
⎪ h ⎪
a b ⎪
⎪
⎨ 2 ⎪
⎪
⎬
= σ1 δε10 + zδε11 + σ2 δε20 + zδε21 + σ6 δε60 + zδε61 + σ4 δε40 + σ5 δε50 dz dx d y
⎪ ⎪ ⎪
⎪
⎪ h
0 0 ⎩ ⎪
⎭
−2
a b
∂δu 0 ∂ ∂wb ∂ws ∂ 2 δwb
= N1 + N1 + + M1 −
∂x ∂x ∂x ∂x ∂x 2
0 0
∂δv0 ∂ ∂wb ∂ws ∂ 2 δwb
+ N2 + N2 + + M2 −
∂y ∂y ∂y ∂y ∂y 2
∂δu 0 ∂δv0 ∂ ∂wb ∂ws
+ N6 + + N6 +
∂y ∂x ∂y ∂x ∂x
∂ ∂wb ∂ws
+ N6 +
∂x ∂y ∂y
∂ 2 δwb ∂δws ∂δws
+M6 −2
∂ x∂ y
+ Q2
∂y
+ Q1
∂x
dx d y (8)
We used the strain from Eq. (4) and the following stress resultants N, M and Q is
define by:
h
2
Ni, Mi = σi (1, z)dz (i = 1, 2, 6)
− h2
h
2
(Q 1 , Q 2 ) = (σ5 , σ4 )dz (9)
− h2
a b
δW = − qδ(wb + ws )dxdy (10)
0 0
a b +h/2
δK = ρ(u̇δ u̇ + v̇δ v̇ + ẇδ ẇ)dzdxdy
0 0 −h/2
614 A. Maji and P. K. Mahato
h
a b 2
∂ ẇb ∂δ ẇb ∂ ẇb ∂δ ẇb
= ρ u̇ 0 − z δ u̇ 0 − z + v̇0 − z δ v̇0 − z
∂x ∂x ∂y ∂y
0 0 −h
2
+(ẇb + ẇs )δ(ẇb + ẇs ) dzdxdy
a b
= I0 u̇ 0 δ u̇ 0 + v̇0 δ v̇0 + (ẇb + ẇs )δ(ẇb + ẇs )
0 0
⎫
∂δ ẇb ∂ ẇb ∂δ ẇb ∂ ẇb ⎪
−I1 u̇ 0 + δ u̇ 0 + v0 + δv0 ⎪
⎬
∂x ∂x ∂y ∂y
dxdy (11)
∂ ẇb ∂δ ẇb ∂ ẇb ∂δ ẇb ⎪
⎪
+I2 + ⎭
∂x ∂x ∂y ∂y
where dot-superscript indicates the differentiation with respect to time variable, the
mass density is given by ρ and (I 0 ,I 1 ,I 2 ) are the mass moment of inertias is given by:
+h/2
I0 , I1 , I2 = 1, z, z 2 ρdz (12)
−h/2
Substituting the appropriate energy expressions δU,δW and δK into the virtual
work statement in Eqs. (8), (10) and (11) and integrating by parts and collecting the
variables of δu0 ,δv0 ,δwb and δws to zero. The following equation of motion as:
∂ N1 ∂ N6 ∂ ẅb
+
δu 0 : = I0 ü 0 − I1
∂x ∂y ∂x
∂ N6 ∂ N2 ∂ ẅb
δv0 : + = I0 v̈0 − I1
∂x ∂y ∂y
2
∂ M1 ∂ 2 M6 ∂ 2 M2
δwb : + 2 + + q − N (w)
∂x2 ∂ x∂ y ∂ y2
∂ ü 0 ∂ v̈0
= I0 (ẅb + ẅs ) + I1 + − I2 ∇ 2 ẅb
∂x ∂y
∂ Q1 ∂ Q2
∂ws : + + q − N (w) = I0 (ẅb + ẅs ) (13)
∂x ∂y
∂ ∂wb ∂ws ∂ ∂wb ∂ws
N (w) = N1 + + N6 +
∂x ∂x ∂x ∂y ∂x ∂x
∂ ∂wb ∂ws ∂ ∂wb ∂ws
+ N6 + + N2 + (14)
∂x ∂y ∂y ∂y ∂y ∂y
δu 0 : N1 n x + N6 n y
δv0 : N6 n x + N2 n y
δwb : ∂ M1
∂x
+ ∂∂My6 − I1 u 0 + I2 ∂∂ẅxb nx + ∂ M6
∂x
+ ∂ M2
∂y
− I1 v̈0 + I2 ∂∂ẅyb n y
∂ Mns (15)
Nn ∂w
∂n
b
+ Nns ∂v∂s
b
+ ∂s
∂δwb
:Mn
∂n
δws : Q n + Nn ∂v
∂n
s
+ Nns ∂v
∂s
s
where
Nns = (N2 − N1 )n x n y + N6 n 2x − n 2y
Nn = N1 n 2x + 2N6 n x n y + N2 n 2y
Qn = Q1nx + Q2n y
∂ ∂ ∂ ∂ ∂ ∂
= nx + ny, = − ny + nx (16)
∂n ∂x ∂y ∂s ∂x ∂y
where nx and ny be the direction cosines of the outward unit normal to the mid-plane
boundary in the x and y directions. M n and M ns are the same forms as N n and N ns .
⎡ ⎤⎧ ∂ 2 wb ⎫
B11 B12 B16 ⎪ ⎨ ∂2x 2 ⎪ ⎬
⎣
− B12 B22 B26 ⎦ ∂ wb
⎪ ∂ y2 ⎪
2
where k be the shear correction factor and Aij , Bij, Dij are the stiffness defined as
h
2
Ai j , Bi j , Di j = Q i j 1, z, z 2 dz (i, j = 1, 2, 6) (19)
− h2
3 Analytical Solutions
A plate is considered with all edges simply supported of length a and width b
under external transverse load q that are based on Navier solution as:
∞ $
$ ∞
mπ x nπ y
wb (x, y, t) = Cbmn sin sin
m=1 n=1
a b
Buckling Analysis of NonLinear First-Order Shear Deformation … 617
∞ $
$ ∞
mπ x nπ y
ws (x, y, t) = Csmn sin sin (22)
m=1 n=1
a b
a b
4 mπ x nπ y
Q mn = q(x, y) sin sin dxdy (26)
ab a b
0 0
The constitutive boundary condition of Eqs. (22)–(24) put into the governing
Eq. (13) are obtained the following equation as:
Therefore, any fixed value of m and n in the element stiffness matrix is denoted
by [K] and element mass matrix is denoted by [M]. The coefficient { }T = {Au0mn ,
Bv0mn , C bmn , C smn }T and also the coefficient λ denoted the buckling load parameter.
The elastic constants for element stiffness matrices are used in this articles from
618 A. Maji and P. K. Mahato
4 Numerical Results
In this article, the critical buckling load of simply supported composite plates are
analyzed and compared the result with exact solution of Reddy and Phan [25], three-
dimensional result of Srinivas and Rao [24] and the result of Senthilnathan et al.
from Ref. [26]. The material constant used in this article are E 1 /E 2 = 40, G23 /E 2 =
0.5, G12 /E 2 = G13 /E 2 = 0.6 and ν = 0.25 where subscripts used 1, 2 and 3 are x, y,
and z axes of the laminated plates.
Table 1 compared the critical buckling load parameter λ developed by the present
theory with three-dimensional theory and theory by Senthilnathan et al. for homo-
geneous orthotropic plates with h/b = 0.05, 0.1 and 0.2. The observation can be
made from this table that the classical plate theory gives constant buckling load
even for thick homogeneous orthotropic plates. The accuracy of results in the three-
dimensional solutions and present theory are almost equally associated and difference
between the two results are minimal.
Figures 2 and 3 contains critical buckling loads vs. modulus ratios for both
the antisymmetric cross-ply and angle-ply laminated composite two-layer plates.
The following discussion may be made from the figures: (1) the bending-extension
coupling effects on the critical buckling load is greatest when degree of anisotropy
is increased (from Fig. 2). (2) The transverse shear stress on critical buckling load
directly depends on the modulus ratio of the materials. The ratio increases with the
increase in transverse shear on critical buckling load (from Fig. 3).
Table 2 and 3 discussed the critical buckling loads of antisymmetric cross-ply
and angle-ply laminated composite plates. From the tables observed that the numer-
ical results calculated from the present analysis and result obtained from the exact
and three-dimensional theory are closely associated. Therefore, for all plate side to
thickness ratio and aspect ratio are in excellent agreement with each other.
Table. 1 The critical buckling load of square homogeneous orthotropic plates where λ =
(12N 1 /π 2 E 1 )(b/h)2 , (k = 5/6)
h/b CPTa Exact theorya FSDPTb Present
0.05 3.039 2.966 2.9774 2.967
0.10 3.039 2.770 2.8074 2.778
0.20 3.039 2.210 2.2910 2.221
a using values in Ref. [24], b using values in Ref. [26]
Buckling Analysis of NonLinear First-Order Shear Deformation … 619
Fig. 2 The material anisotropy on the critical buckling load for antisymmetric Cross-ply orthotropic
laminates [0°/90°/…] (a/h = 10)
Fig. 3 The material anisotropy on the critical buckling load for antisymmetric Angle-ply orthotropic
laminates [45°/−45°/…] (a/h = 10)
5 Conclusions
First-order nonlinear shear deformation theory has been used to determine the critical
buckling load of laminated composite plates. The laminated composite plates are used
with simply supported boundary condition is considered. The present solutions are
obtained by using Von Karman nonlinear strain equations and compared the results
with the classical plate theory, exact solution and three-dimensional solutions. The
present observations suggest that the effect of critical buckling load of composite
plates have closely associated with other investigators and also have minimal effect
on the accuracy.
620 A. Maji and P. K. Mahato
Table. 2 The critical buckling loads for antisymmetric cross-ply square plates where λ =
(N 1 b2 /E 2 h3 ), (k = 5/6)
[0°/90°] [0°/90°/0°] [0°/90°/90°/0°]
(12.628)a (35.831) (35.831)
a/h FSDPTb FSDPTc Present FSDPTb FSDPT Present FSDPTb FSDPT Present
5 8.142 8.277 8.277 10.525 – 10.625 11.533 – 11.534
10 11.099 11.352 11.353 21.643 – 21.648 23.270 – 23.271
12.5 11.605 12.882 12.887 25.144 – 25.346 26.518 – 26.518
20 12.208 12.515 12.526 30.664 – 30.667 31.432 – 31.431
25 12.356 12.671 12.672 32.332 – 32.431 32.872 – 32.872
50 12.559 12.886 12.887 34.883 – 34.885 35.037 – 35.038
100 12.611 12.947 12.947 35.589 – 35.592 35.629 – 35.629
a CPT values, b using values in Ref. [25], c using values in Ref. [26]
References
Abstract This paper presents an algorithm for object avoidance while sustaining
the trajectory of the end effector of the robotic manipulators. Multiple filters are
developed that are based on classified constraints and applied to configuration space
to obtain a collision-free path. The recursive and iterative nature of the algorithm
makes it possible to acquire results under the permissible zone of error. The identifica-
tion of desired configurations from millions of choices is achieved to obtain smooth
and least deviated movements. MATLAB software is used to design, implement,
visualize, and test the proposed model.
1 Introduction
During the last decade, manipulators of diverse classes have appeared as an impera-
tive part of the automated industries. The flexible and configurable nature of most of
the manipulators makes it feasible to stretch their work domain. The very prevalent
objective of them is to move the tool located on the manipulator to some goal in
3D (Three Dimensional) space. The task becomes susceptible to failure when it is
implemented in a dynamic environment. In such environments, object avoidance and
real-time path planning become the fundamental task of the manipulator.
In some applications, it becomes crucial to trace the desired path [1] to reach the
goal such as autonomous grout filling of cracks on the dam. The imperialist Compet-
itive Algorithm was used in [2] to solve the path-planning problem. The generation
of C-Space (Configuration Space) is very computationally expensive because of the
exponential growth in the possible arrangements of serial chain manipulators. Arjang
Hourtash and Mahmoud Tarokh who presented their work on path planning decom-
position in [3] have solved this issue. The division of the serial chain into sub-chains
had been done to lessen the planning time. A* algorithm [4] is used to find the shortest
path between the pick and place locations of objects. Most of the approaches were
converged to reach the goal through any path between the initial and final state of the
manipulator. To solve the trajectory problem, different algorithms were discussed and
tested on a robotic tracking control testbed [5]. A nonlinear control design has been
proposed in [6] to incorporate the dynamics of the manipulator while performing
the tasks which require a trajectory control. Non-linear behavior of PD (Proportional
Derivative) controllers have been proposed in [7]. A robust and novel approach based
on multiple constraint filters is proposed in this paper to reduce the complexity and
incorporate the general behavior of robotic manipulators. Multiple constraints can
be added according to the environment, manipulator’s task, and the manufacturing
constraints.
The organization of the paper is as follows. Section 2 discusses the criteria of the
description of rigid bodies over 3-dimensional space. The proposed methodology,
its implementation, and the corresponding results are presented in detail in Sect. 3.
The approach is illustrated in the first four sub-sections that is followed by the
implementation and the simulation results. All simulations have been performed
on MATLAB and are supported by comparing the desired and actual trajectories.
Finally, a conclusion is made in Sect. 4, which also addresses future work.
corresponding geometric shapes of point clouds for points, line segments, and rect-
angles are the sphere, cylisphere, and box with rounded edges respectively. These
simple shapes reduce the computation cost while determining the characteristics of a
collision. A simple geometry shape is selected based on the shape of the object to be
avoided. Similarly, the individual links of the manipulator are considered cylinders
and superimposed to construct the simplified structure of the manipulator.
MATLAB software has been used to simulate the movement of manipulator
because of its versatility in representing objects as a surface or a mesh. Also, the
compatibility of the ROS (Robot Operating System) framework in MATLAB can be
further utilized to simulate the manipulator in more dedicated simulators like RViz
and Gazebo. The primary state, goal state, and the path traced are plotted in 3D space
in MATLAB to verify the algorithm. Object-Oriented Programming is employed to
store and handle data more efficiently. It reduces the complexity of the model and
makes it adaptive to all small constraints.
3 Methodology
In this paper, a novel approach has been suggested to avoid the objects and distur-
bances of surroundings while sustaining the desired trajectory. Trajectory represents
the path that is required to follow by the end effector. This approach is divided into
4 main steps. Step 1 deals with the generation of the C-Space and Workspace by
applying the initial parameters and mechanical constraints. Collision checking and
626 V. Garg and V. Rastogi
avoidance filters are introduced in step 2. Objects are tested for their interference with
the manipulator. Step 3 and step 4 further put more constraints on the collision-free
states to obtain a smooth and least deviated path.
⎡ ⎤
cos θi − sin θi 0 ai−1
⎢ sin θi cos αi−1 cos θi cos αi−1 − sin αi−1 −di sin αi−1 ⎥
Tii−1 =⎢
⎣ sin θi sin αi−1 cos θi sin αi−1 cos αi−1 di cos αi−1 ⎦
⎥ (1)
0 0 0 1
In the above equation, Tii−1 is the transformation matrix for two successive links,
θ i is the angle of ith joint, α i−1 is the twist of ith link with respect to (i−1)th link,
ai−1 is the length of ith link, and d i is the offset between links i and i−1.
⎡ ⎤
cos θ1 (5 cos θ2 + cos θ2+3 − 5 sin θ2+3 ) − 5 sin θ1
P = ⎣ cos θ1 (5 cos θ2 + cos θ2+3 − 5 sin θ2+3 ) − 5 cos θ1 ⎦ (2)
−5 cos θ2 − cos θ2+3 − 5 sin θ2+3
Object detection [14] and localization [15–17] are the primary tasks prior to the colli-
sion check. Localization of objects and mapping them to the simulated environment
is not required to discuss at this point. All the generated configurations are iterated to
measure the minimum distance from the object’s primitive. The minimum distance
is the perpendicular distance between any two geometric primitives. The configu-
rations, which failed to satisfy the condition (3) will be excluded from the original
C-space.
where, d(i, o) is the perpendicular distance between the ith link and the object o, r(i)
is the radius of the ith link, and r(o) is the radius of object and th is the minimum
required separation between the object and the manipulator.
This geometric image is simplistic and less computationally expensive in compar-
ison to superquadratics representation but ends in the exclusion of additional states
of the manipulator that are not in a collision. To circumvent this, another image of
the object can also be generated by creating a mesh of irregular shape in 3D space
and the constraint is modified to (4).
In the above equation, mesh(o) is the point cloud of object and d(i, mesh(o)) is
the distance of ith link from the point cloud. After passing from the collision filter,
the remaining configurations are collision-free and can be processed further.
The desired trajectory is presented as a set of points. The separation between consec-
utive points plays an important role in deciding the planning time of the algorithm
and the error between the theoretical and calculated trajectories. The former is nega-
tively correlated and the latter is positively correlated with the separation distance.
The greater the separation, the lesser will be the total number of points, planning time,
and more will be the error in tracing the trajectory. An optimum value of separation
can be decided on the basis of the computational speed and the toleration zone for
error. Figure 2 shows the relation of the planning time and the number of coordinates
i.e. points in 3D space used to represent the trajectory line. It can be inferred from
the figure that both of the parameters are directly related to each other. Once the set
of points is generated, each point is tested against the collision-free states to obtain
the configurations that take the wrist to the path.
The configurations responsible for crossing the toleration zone of error will be
excluded. These are further sorted to determine the least error state. The extracted
set of configurations is finally passed through a smoothing filter.
Fig. 2 Variation of planning time with the number of points on the trajectory line
Path Tracing and Object Avoidance Algorithm for Robotic … 629
The continuous motion of actuators is very important during the action of the manip-
ulator. This continuity enhances the life cycle of an actuator and diminishes the
impulses, which generally occur during jerks in motion. continuous motion can
be obtained by collecting the consecutive configurations of a manipulator, which
converges towards the task space. Initially, each state is given a common identity.
This identity is obtained from Eq. (5) by calculating the maximum movement of
each actuator to reach the goal state from the initial state.
IN = max θ f − [θi ] (5)
In the equation, θ f is the joint vector of goal state and θ i is the joint vector of the
initial state. This number is characterized by an integer value which lies between the
joint limits. It can either be positive or negative depending upon the reference state
i.e. initial state. IN (Identity Number) of the initial state is unique with a value of
zero.
A subset of configuration space is determined with IN ranging from IN of initial
state and goal state. Joint states with consecutive IN are extracted, which satisfies
the constraint (6) in which θ IN is the joint vector with IN value. The joint vector is a
column vector of dimension DoFx1 consisting of the joint states.
min θIN+1 − [θIN ] ≤ 1 (6)
values ranging from [INi , INf ]. If any type of discontinuity is detected then it returns
back to the nearest filter to pass another goal state.
Once the desired joint movements are achieved, the initial state and goal state are
updated. The previous goal state is updated to the initial state of the next iteration
and goal state is updated to the next point of the trajectory. Iterations are performed
until all points of trajectory are covered.
Path Tracing and Object Avoidance Algorithm for Robotic … 631
Fig. 4 a The desired trajectory of the end effector of the manipulator. b Manipulator under collision
with a spherical object. c The path traced during 1/4th of the period. d The path traced during 1/2nd
of the period. e The path traced during 3/4th of the period. f The collision-free path traced at the
end period
3.6 Results
The ideal solution is plotted to achieve the desired path (refer Fig. 4b) irrespective
of environmental interference. The links of the manipulator are considered to be
cylinders with an assumption that end effector doesn’t play any role in determining
its trajectory. 16% of the path is obstructed because of the object, it is required to be
avoided without disrupting the trajectory of the end effector.
Multiple filters based iterative algorithm is implemented to find a least deviated,
collision-free, and smooth path. Figure 4c–f visualize the motion of manipulator for
four progressive quarters of the time period. The period is defined by the time required
by the end effector to meet the entire trajectory. It is controlled by the characteristics
of the actuators such as the joint velocity, acceleration, and resolution. The object is
placed near to the reachable workspace of the manipulator and so no collision-free
joint configurations are left for the manipulator to overlap the desired position in the
obstructed region and hence, the manipulator started to deviate from its mean path
and looked for other nearest collision-free configurations at the end of the first quarter
(refer Fig. 4c). The other reason for such deviation is the lower resolution of actuators
which is negatively correlated with it. High resolution and a wide the range for joint
movements will give the best results at a cost of high computational power. In the
second quarter, when the manipulator is crossing the object, a maximum deviation
632 V. Garg and V. Rastogi
can be observed (refer Fig. 4d) and results into a bump in achieved trajectory as shown
in Fig. 5. The manipulator started to converge to the required trajectory at the end
of the quarter and the object is completely avoided while maintaining the smoothing
action. Figure 4e–f shows its continuous motion in the unrestricted environment
until the trajectory is completely traced. The required and reached trajectories are
compared in Fig. 5 to depict the deviation of the manipulator from the desired path.
4 Conclusion
A novel approach of path tracing and object avoidance algorithm has been presented
in this paper. A 6R open-chain manipulator is simulated on MATLAB to trace the
trajectory when an object is placed mid-between its path. Forward kinematics is used
to generate the configuration space. The manipulator is characterized by parameters
of D-H Notation and represented as a superimposition of multiple cylinders to cover
the rigid links. Similarly, nearby objects are also treated as a spherical shell. Multiple
filters are used to pass only those joint states, which show a smooth, collision-free
and minimum deviated path. These filters are robust to adapt to any environment.
This algorithm is highly configurable for either the planning time or the deviation. A
large number of simulations have been performed and results have shown the good
reliability of the proposed approach.
Future objectives of this work are to reduce the planning time such that the
algorithm can be implemented in the dynamic environment.
Path Tracing and Object Avoidance Algorithm for Robotic … 633
References
1. Tsai CS (2014) Online trajectory generation for robot manipulators in dynamic environment--an
optimization-based approach. Doctoral dissertation, UC Berkeley
2. Tabealhojeh H, Ghanbarzadeh A (2014) Two steps optimization path planning algorithm
for robot manipulators using imperialist competitive algorithm. In: 2014 second RSI/ISM
international conference on robotics and mechatronics (ICRoM), Tehran, pp 801–806.
3. Hourtash A, Tarokh M (2001) Manipulator path planning by decomposition: algorithm and
analysis. IEEE Trans Robot Autom 17(6):842–856. https://doi.org/10.1109/70.976006
4. e Silva JS, Costa P, Lima J (2013) Manipulator path planning for pick-and-place operations
with obstacles avoidance: an A* algorithm approach. In: International workshop on robotics
in smart manufacturing. Springer, Berlin, Heidelberg, pp 213–224
5. Nagy PV (1989) Trajectory tracking control for industrial robots. J Mech Working Technol
20:273–281
6. Oliver JPO, Domínguez-Ramirez OA, Quezada ESE (2008) Trajectory tracking control for
robotics manipulators based on passivity. In: 2008 electronics, robotics and automotive
mechanics conference (CERMA’08). IEEE, pp 484–489
7. Liu F, Er MJ Trajectory tracking of robot manipulators using linear and nonlinear PD-type
controllers
8. Hendarto HA, Munadi M, Setiawan JD (2014) ANFIS application for calculating inverse
kinematics of programmable universal machine for assembly (PUMA) robot. In: 2014 The
1st international conference on information technology, computer, and electrical engineering.
Semarang, pp 35–40
9. Craig JJ (1989) Introduction to robotics: mechanics and control, 2nd edn. Addison-Wesley
Longman Publishing Co., Inc, Boston, MA, USA
10. Silva ECE, Costa MF, Erlhagen W, Bicho E (2016) Superquadrics objects representation for
robot manipulation. In: AIP conference proceedings, vol. 1738(1). AIP Publishing, p 300004
11. Bosscher P, Hedman D (2009) Real-time collision avoidance algorithm for robotic manipula-
tors. In: 2009 IEEE international conference on technologies for practical robot applications.
Woburn, MA, pp 113–122
12. Lynch KM, Park FC (2017) Modern robotics: mechanics, planning, and control, 1st edn.
Cambridge University Press, New York, NY, USA
13. Denavit J, Hartenberg RS (1955) A kinematic notation for lower-pair mechanisms based on
matrices. Trans ASME E J Appl Mech 22:215–221
14. Cho JM, Kim K (2017) Precise object detection using local feature for robot manipulator.
In: 2017 14th international conference on ubiquitous robots and ambient intelligence (URAI),
Jeju, pp 497–499
15. Kim K, Cho J, Pyo J, Kang S, Kim J (2017) Dynamic object recognition using precise location
detection and ANN for robot manipulator. In: 2017 international conference on control, artificial
intelligence, robotics & optimization (ICCAIRO). Prague, pp 237–241
16. Kuehnle J et al (2009) 6D object localization and obstacle detection for collision-free manip-
ulation with a mobile service robot. In: 2009 international conference on advanced robotics.
Munich, pp 1–6
17. Yang Y, Cao Q-X (2012) Monocular vision based 6D object localization for service robot’s
intelligent grasping. Comput Math Appl 64:1235–1241. https://doi.org/10.1016/j.camwa.2012.
03.067
Characterisation of Composites Made
by Prepreg Waste
Abstract The current work deals with the utilisation of prepreg waste in making
composites. The leftover and out-of-shelf prepregs were chopped into chips of
different lengths to make the laminates. The fibre volume and void content tests
were performed to evaluate the quality of the manufacturing process. The prepreg
waste collected from the industries consists of bidirectional glass/epoxy material.
The mechanical characterisation of the laminates, made by the prepreg waste, was
performed. The effect of chip length on the tensile strength and modulus were
analysed.
Keywords Prepreg wastes · Chip length · Fibre volume · Void content · Strength
and stiffness
1 Introduction
Composites made from epoxy/carbon prepregs have broad applications in the field of
the aerospace industry due to their high performance [1]. Industries utilising prepregs
usually landfill the leftover prepreg. Composites made of this wastage can be utilised
in high stiffness applications. Feraboli et al. [2], have used discontinuous composites
made by chopped prepregs for making the window frame of an aircraft. Vishnu
et al. [3], have studied the size effect of chips on tensile properties of prepreg-based
discontinuous composites (PBDC) with chip lengths of 2, 4 and 8 mm with a constant
width of 2 mm and 6 × 24 mm chip size. Krishna Mohan et al. [4], have predicted
the stiffness of PBDC for chip aspect ratio of 2, 3, 4, 5 and 6 by developing the RVE
and studied the size effect of the RVE for a chip volume fraction of 22%. Anil et al.
[5], have studied the compression properties of PBDC. The primary objective of this
study is
(1) To develop a composite manufacturing technology using different sizes of the
chips (6 × 6, 6 × 12 and 6 × 18 mm) chopped from prepreg waste collected
from industry.
(2) To find the effect of chip length of the above-prepared laminates under tensile
load.
(3) To analyse the failure mechanism.
2 Material Preparation
The chips were derived from the leftover and out-of-shelf prepregs by manual chop-
ping. The width of the chip was maintained constant with a value of 6 mm and the
length of chips were varied from 6 to 18 mm in the steps of 6 mm to study the effect
of chip length at the different aspect ratio of 1, 2 and 3.
An epoxy resin system was used to make the resin -hardener mixture. A weight
ratio of 100: 80 was used for this purpose [6]. This mixture of epoxy was added to
the prepreg waste.
The chopped systems are compression moulded at 0.3 bar using a hot press. The
curing cycle adopted for this process was 75 ± 5 °C for 40 ± 5 min followed by a
post-curing cycle of 135 ± 5 °C for 60 ± 15 min. A typical composite panel made
by this procedure is shown in Fig. 1.
3 Experimentation
As the chips were chopped manually, there is a possibility of deviation in the size
of chopped chips. This manual error was calculated using Image-J software. The
pixel values are calibrated with the known reference, and the length and width of the
chopped chips were measured. The average and standard deviation of these chips are
calculated and are reported in Table 1. A typical image consisting of 50 chips, taken
for this analysis is shown in Fig. 2.
Adopting the procedure proposed in the work of Yee and Stephens (1996) [7], the
Thermo Gravimetric Analysis (TGA) test was performed using ‘SDT Q600 appa-
ratus. The fibre volume fraction was calculated by using Eq. (1). The weight ratios
638 P. R. Krishna Mohan et al.
are obtained from TGA test. By using Eq. (2), the theoretical density (ρ c ) was calcu-
lated. The density of the matrix (ρ m ) was obtained by the relative density test. In
this work, the density and mass/weight percentage are denoted by ‘ρ’ and ‘w’ with
subscripts ‘f ’ and ‘c’ representing the fibre and composite, respectively
ρc
νf = wf ∗ ∗ 100 (1)
ρf
1
ρc = wf wm (2)
ρf
+ ρm
Adopting the ASTM D792 standard, the experimental density (ρ e ) of the composite
was found. ASTM D2734 standard was used to calculate the percentage of the void
content and is shown in Eq. (3).
ρc − ρe
υv = ∗ 100 (3)
ρc
We have performed the uniaxial tensile tests of the samples obtained from the PBDC
material which are made from the leftover prepregs collected from the industry. The
samples of different chip lengths (6, 12 and 18) were made according to the ASTM
3039D standards. The experiments were carried using Instron 1195 Universal Testing
Machine (UTM) with 50 kN load cell and the test results are reported in Table 3.
The average and standard deviations of the chopped chips are reasonably good, and
they are very close to the desired size of the chip. The standard deviations of the chip
aspect ratio were between 0.14 and 0.26. This indicates that the chopping process
adopted is very satisfactory.
Characterisation of Composites Made by Prepreg Waste 639
The fibre content of the laminates made from BD glass/epoxy prepreg waste was
between 61.99 and 65.24%. The void volume fraction is between 0.81 and 0.85%.
The fibre volume fraction and void content of all the laminates made with different
aspect ratio are reported in Table 2.
Fig. 3 6 × 6 glass
fibre/epoxy (industry)
640 P. R. Krishna Mohan et al.
Fig. 4 6 × 12 glass
fibre/epoxy (industry)
Fig. 5 6 × 18 glass
fibre/epoxy (industry)
Among the aspect ratio of the tested samples, the maximum average ultimate strength
was obtained for 6 × 18-chip size. Further, the tensile modulus increases with the
increase in the chip size; this behaviour can be seen from Fig. 7. The failure mecha-
nism of the samples was observed using a Zeiss optical microscope. A typical failed
sample under tensile loading is shown in Fig. 8. The failures are mainly due to
debonding of the fibre and matrix. The crack initiation is observed by splitting of the
chips and gradually it progress to the surface by the chip delamination.
5 Conclusions
References
Abstract In the present research a novel cooling setup that can help the turning
process to be performed in various media, i.e., dry, liquid CO2 or compressed air,
cooling lubricant, and the mixture of the liquid CO2 or compressed air mixed with
the cooling lubricant. The conventional cooling oil is used to dissipate the heat
and increasing life of a cutting tool. So, the surface finish of the workpiece improves
leading to a better quality product. Customarily, the contaminated cutting fluid needs
to be disposed of after its use that involves a good amount of money of total manufac-
turing cost. Primarily this work is a turning operation on Al 3055 alloy under three
different cutting media namely dry, Liq. CO2 and Liq. CO2 assisted the minimum
quantity of liquid in order to get better quality of the product.
1 Introduction
machining might dominate the wet method if only economic and ecological perspec-
tives are concerned. But dry machining has some limitations, like during machining
of sticky material e.g. pure Aluminium. The material easily gets stuck to the cutting
tool face as the turning of the material is going on. Consequently the machining
rate decreases as well as the tool wear rate and cutting temperature increases. Thus
the tool life is hampered. The good surface finish of the product achieved by the
application of MQCL and liq. CO2 conditions compared to dry machining condi-
tion whereas MRR is greater in dry machining environment [2]. It has also been
observed that proper selection of process parameters such as feed rate, depth of cut,
cutting speed, and tool nose radius will produce good quality of surface roughness
under MQL with Nano coolant [3]. Good quality of surface roughness value can
be achievable by MQL compared with dry machining conditions as well as flooded
cooling conditions [4]. This is also noted that higher value of feed rate and cutting
speed under the MQL conditions can damage the surface quality [5]. This innovative
MQL technique can be used in drilling, turning, milling, and grinding process also
[6]. By means of the analytical model, it is noted that specific cutting energy reduces
when the flow rate and pressure of the MQL process increase [7]. The usage of MQL
technique with a cryogenically treated tool inserts shows improved results in terms
of minimum surface roughness [8]. In addition to the above literature, it becomes
customary to prepare sustainable machining setup in order to provide a multi-type of
cooling medium like dry, near dry, liq. CO2 and MQL/MQCL etc. The sustainable
machining setup followed by experimental validation has been necessary in order to
survive in the field of manufacturing.
2 Experimental Procedure
The experiments were performed in a SINEWAVE CNC turning center at the work-
shop of UEMK. The schematic diagram of the machine set up has been shown
in Fig. 1a. The whole control of the cutting fluid can be accomplished with the help
of a single mixing chamber as shown in Fig. 1a. The flow of CO2 gas and mixed
coolant spray is controlled by the mixing chamber by regulating the knob left side
and right side respectively. Dry machining is performed by adjusting the knob at the
mid position. When the value of the liquid CO2 is kept closed and the knob is at right
position then only coolant lubrication is emerged out of the mixing chamber. The
nozzle as shown in Fig. 1b supplies the cutting fluid at the junction of workpiece and
tip of the cutting tool.
Experimental Investigation on the Effect of Process Parameters … 645
Fig. 1 a Schematic diagram of the experimental setup and b its original photograph
Aluminium alloy 3055 (UNI3055) material was chosen to perform the turning oper-
ation in CNC machine under three different cutting environment (Dry, Liquid CO2 ,
Liquid CO2 assisted minimum quantity cooling conditions). Workpiece material
composition is shown in Table 1. The design of experiment (DOE) was set as per
response surface methodology to carry out the experiments. Three input parameters,
i.e., spindle speed in rpm, feed rate in mm/min, and depth of cut in mm were varied
and two sets of each experiment were chosen.
The experiment in the present research work has been set in such a way so as to analyze
the effects of the input parameters on the turning operation under various cooling
media. Input parameters are spindle speed (A) in rpm, depth of cut (B) in mm and feed
rate (C) in mm/min. Output responses are material removal rate (MRR) in g/min,
surface roughness values (Ra) in micron, Chip thickness (t) in mm in degree Celsius.
The experiment is designed based on response surface method especially central
composite design. The factors and levels for the current experiment are tabulated in
Table 2.
646 S. Mondal et al.
In the research work, the prime intention is to develop a system to supply dry, Liq.
CO2 and MQL machining whichever is required during machining. To validate the
performance of the setup the experiment was carried out in the setup as per the
RSM design of the experiment. The 20 number of experiments were carried out.
The responses that were measured are MRR, Ra value. The regression equations
as obtained from response surface methodology are shown in Table 3. In support
of these ANOVA tables are provided (Table 4) which illustrate that the quadratic
equations can correlate the experimental results entirely. The regression equations
as depicted in Table 3 are used to predict the responses and the predicted values are
compared with the actual values from experiments (as shown in Fig. 2).
4 Discussion
It has been observed from Fig. 2a that MRR hardly deviates when medium changes
from dry to CO2 and then MQL. Though in liq. CO2 it is little bit lower than dry
medium but as soon as the MQL medium is used then the MRR becomes almost the
same as that of dry medium. Figure 2b depicts that surface roughness is distinguish-
ably lower in MQL medium but it is maximum in dry medium. Liq. CO2 medium
can help to improve the surface quality as depicted in the same figure.
Experimental Investigation on the Effect of Process Parameters … 647
Fig. 3 Surface plot for MRR at different combination of input parametric levels in various medium
a dry, b liquid CO2 , and c MQCL
It implies that the process can produce better surface finish steadily when the
MQL medium is used. In comparison to MQL medium liq. CO2 have obtained the
‘runner up’ position. The surface plots for MRR (Fig. 3), and Surface roughness (Ra)
(Fig. 4) at various media and level setting are depicted. The surface plots are used to
explain the effect of the process parameters on different responses (Table 4).
Here Fig. 3a depicts that in the dry medium the MRR increases at first with the
spindle speed but after the 1500 rpm the MRR decease whereas it increases steeply
with the depth of cut. The causes of such observation are that with an increase of
spindle speed up to 1500 rpm along with the increase of depth of cut at first the chip
of greater thickness are coming out from the job. But as soon as the spindle speed is
increased then the finer chips are coming out of the machining zone. But it is clearly
understood that with the increase of feed rate MRR increases impressively as the
process goes to rough turning condition consequently chips of higher thickness are
released. The same kinds of graphs are also obtained in liq. CO2 and MQL medium as
depicted in Fig. 3b and c. So it is observed that cooling medium has a comparatively
lower effect on MRR.
Experimental Investigation on the Effect of Process Parameters … 649
Fig. 4 Surface plot for surface roughness (Ra) at different combination of input parametric levels
in various medium a dry, b liquid CO2 and c MQCL
As the surface roughness is coming into the picture, it has been observed from
Fig. 4a that with the increase of spindle speed at the dry condition the Ra value
decreases steeply. The same steepness of increase in Ra value is observed with the
increase of depth of cut and feed rate. So the dry medium is responsible for poorer
surface finish of the product. In line to the above discussion and supported by Fig. 4b
650 S. Mondal et al.
that Ra value is not much improved in liq. CO2 medium but if Fig. 4c is taken as a
reference it is clearly been seen that nature of the surface plots is convex. The convex
surface as shown in Fig. 4c explains that Ra value increases up to an optimum and
then decreases.
5 Conclusions
In the above-mentioned research work it has been illustrated that for the sustainability
issues it is time to manufacture an unusual setup which can be able to supply coolant
in any form like dry, liq. CO2 /compresses air and MQL/MQCL.
The authenticity of experimental results was analyzed and correlated by ANOVA
table which shows that experimental results are well in the agreement with the regres-
sion equations for material removal rate (MRR) and surface roughness (Ra value)
that are the important responses for getting better surface quality of the product.
The RSM-based regression equations confirm of experimental results and it ensures
fewer chances of error. The surface plot clearly depicts the optimum region for Ra
value can be obtainable from the level setting of the process parameters. Without the
delve into the optimization process, it can be seen from the analysis that though the
liq. CO2 and MQL media cannot improve the MRR much over dry medium but it can
be concluded from the above research work that surface roughness is lower in Liq.
CO2 and MQL conditions hence the Ra value decrease. So, to get a better quality
product from the UNI 3055 Al alloy the MQL conditions can be used sustainably.
References
1. Sharma VS, Dogra M, Suri NM (2009) Cooling techniques for improved productivity in turning
Int J Mach Tools Manuf 49:435–453
2. Mondal S, Paul G, Mondal SC (2019) Investigation into the application of liquid CO2 and MQL
for CNC turning of Al Alloy 3055. Research into design for a connected World proceedings of
ICoRD 2019, vol 1, pp 977–987
3. Patole PB, Kulkarni VV (2018) Optimization of Process parameters based on surface roughness
and cutting force in MQL turning of AISI 4340 using nano fluid. Mater Today: Proc 5:104–112
4. Joshi KK, Kumar R, Anurag (2019) An experimental investigation in turning of incoloy 800 in
dry MQL and flood cooling conditions. Procedia Manuf 20:350–357
5. Çakır A, Yağmur S, Kavak N, Küçüktürk G, Şeker U (2016) The effect of minimum quantity
lubrication under different parameters in the turning of AA7075 and AA2024 aluminium alloys.
Int J Adv Manuf Technol 84:2515–2521
6. Boswell B, Islam MN, Davies IJ, Ginting YR Ong AK (2017) A review identifying the effective-
ness of minimum quantity lubrication (MQL) during conventional machining. Int J Adv Manuf
Technol 92:321–340
7. Chetan, Ghosh S, Rao PV (2018) Specific cutting energy modelling for turning nickel-based
Nimonic 90 alloy under MQL condition, Int J Mech Sci https://doi.org/10.1016/j.ijmecsci.2018.
07.033
Experimental Investigation on the Effect of Process Parameters … 651
8. Deshpande YV, Andhare AB, Padole PM (2018) Experimental results on the performance of
cryogenic treatment of tool and minimum quantity lubrication for machinability improvement
in the turning of inconel 718. J Braz Soc Mech Sci Eng 40:6
Reciprocating Wear Behaviour of Al–SiC
Composite Processed with MWS
Abstract The present study involves the fabrication and characterization of Al–SiC
composites by means of varying content of SiC (i.e., 5, 10, 15 and 20 wt. %). The
material was synthesized with developed hybrid microwave sintering (MWS) tech-
nique incorporating powder metallurgy (PM) process. Findings are in good agree-
ment for the selection of reinforcement content in the matrix material for the material
development to be used in a lightweight material with improved properties. A system-
atic approach was made to validate the importance of reinforcement in the matrix
material for enhanced product response for various functional applications. The hard-
ness of the developed composite material increases significantly with maximum up to
38.1% for 20 wt.% reinforcement in aluminum. The higher reinforcement percentage
in the composite material reduces the wear loss up to 96% as observed in the Al +
20% wt.% SiC composites.
1 Introduction
In the last three-decade innovation and research in the field of processing and design
of new material grows exponentially. Sustainability of the system depends largely
on the selection of material so the basic and important step is to select a material
having low weight, low cost and environment-friendly. Development in the direc-
tion of lightweight material leads to the processing of material like aluminum metal
matrix composites (Al-MMCs) which are very light in weight as well as their func-
tionality which perform better in the desired environment [1–3]. However, the lower
wear resistance of Al-based material limits the applicability in the different func-
tional surface. To adlib, the wear resistance properties numerous hard ceramics,
oxide-based material is utilized [4]. However, the wear behaviour of the material
depends upon the type of application undergoing during the operating environment.
Present paper labels the reciprocating wear behaviour of the Al-based composite
material reinforced with a different weight percentage of silicon carbide [5, 6]. The
replacement of the conventional material demands lightweight and improved wear
resistance for different tribological applications [7]. In a similar context Samal et al.,
studied the wear resistance behaviour of AA5052 MMCs with the addition of TiC
during the stir casting process. They found that the addition of a hard carbide phase
reduces the volumetric wear loss by 24% compared to the monolithic phase of the
material [6]. The effect of heat treatment on fretting wear of A356 was evaluated by
Chen and their colleagues. It was observed that the 15 vol.% silicon carbide (SiC)
offers greater wear resistance at lower load. However, T6 treatment found to be a
good agreement for better wear resistance at higher load and higher working cycles
[8]. The use of fine reinforcement in the matrix leads to greater strength and wear
resistance compared to the coarser particle. As the finer particles hold better inter-
facial strength and load-bearing capacity of the matrix material [5]. However, the
choice of processing mode and reinforcement percentage holds a unique position
in the material response under extreme survival conditions. The present study uses
the innovative developed hybrid microwave sintering (MWS) approach for material
development. The use of MWS helps in improvement of densification, improved
mechanical and tribological response with shorter processing time, environmental
friendly with minimal wastage [2, 9].
From the literature, it is evident that the lesser properties have been given to inves-
tigate the reciprocating wear behaviour of the material. The Al/SiC composite mate-
rial is developed by incorporating powder metallurgy and hybrid MWS method. The
present study aims to develop the correlation between the reinforcement percentage
in the matrix material and wear resistance of the material under the reciprocating test
condition. The reciprocating test was carried out under constant sliding velocity of
0.3 m/s for a different travelled distance under dry environment. The test results
will help in the selection of the amount of reinforcement for the wear resis-
tance application in the field of automobile, aerospace and numerous structural
applications.
Reciprocating Wear Behaviour of Al–SiC Composite Processed … 655
2 Experimental Details
2.1 Materials
The pure form of Al powder (Assay>99.5%) with particle size in the range of 44–200
micron and SiC with an average particle size of 44 microns used as starting material
for the development and characterization of Al-MMCs. The material was supplied
by Alpha Chemika Pvt. Ltd. Mumbai, India. The material was used as supplied with
any further purification and test.
The powder metallurgy and microwave sintering approach used for the development
of Al/SiC composite material. For the development of Al/SiC composite material,
four different reinforcement percentage 5, 10, 15 and 20 wt.% of SiC were used
in the matrix material. In this process, the initial powder (i.e. matrix and reinforce-
ment) were blended by means of mechanical ball milling. The milling was carried
out for 2 hours at a rotational speed of 150 rpm at normal atmospheric pressure and
at room temperature. The blended powder material is cold compacted by the appli-
cation of uniaxial compaction pressure of 580 MPa [10]. The cold compacted part
is further sintered by innovative developed microwave sintering approach. Figure 1
depicts the schematic representation of the hybrid microwave sintering approach.
The sintering setup consists of silicon carbide susceptor material and alumina insu-
lation. The susceptor material absorbs the microwave energy and converts it into heat.
The susceptor assisted heating in microwave sintering approach helps in heating the
samples in bidirectional mode (i.e. susceptor heating from outside to inside and
Fig. 2 Pictorial
representation of
reciprocating wear test rig
microwave heating from the core of the material to the surface) [11]. The compacted
part is exposed to 11 minutes in the microwave radiation to reach the sintering
temperature of 640 °C (near the melting point of aluminium) with no holding time in
this temperature. The sample is allowed to cool inside the developed setup to room
temperature with natural cooling mode. The sintered samples are further used as
different characterization and end-use.
The microstructural observation of mechanically mixed powder material and
sintered specimens were done by the help of scanning electron microscope (SEM).
The quanta FEI 250 was for the examination of the microstructural images. The
hardness of the monolithic and composite sample was evaluated by applying 60 Kgf
load as per the ASTM standard E18-02. The load was applied for a dwell time of
5 s in the sample. For the repeatability of the test sample, an average of 5 successive
indentations were reported. The hardness measurement was done with the help of
Rockwell hardness test scale B. For the reciprocating wear analysis, in-house fabri-
cated reciprocating test setup was employed. The pictorial representation of the test
setup is given in Fig. 2. The test setup consists of an AC motor and gearbox which
operate at a different rotational speed. The test was carried out at constant recipro-
cating velocity of 0.3 m/s. The total stroke length was kept constant throughout the
test and kept as 0.6 m. During the test, the weight loss method was adopted for the
calculation of wear rate and moss loss of the Al/SiC composite material. The wear
test was carried out with a normal applied load of 20 N throughout the test. Prior to
the test sample was polished against fine abrasive paper and cleaned with the help
of acetone. The counter surface is made of hardened steel with HRC>55.
The sample weight was measured before and after the test with the help of a preci-
sion balance of 0.1mg. The specific wear rate (wear co-efficient) of the developed
composite material was calculated by the use of Archard model:
m
k= (1)
pl
Reciprocating Wear Behaviour of Al–SiC Composite Processed … 657
where k is the wear co-efficient, m is the loss in weight of the sample, p resembles
the normal applied load at the test sample, l is the total travelled distance of the test
sample.
The Al powder material is in nearly spherical morphology and SiC is in the form of
nearly flakes likes structure. The Fig. 3 shows the morphology of the milled powder
material after 2 h of milling. The morphology revealed that the powder material is
homogenously mixed and some part of the reinforcement is diffused into the matrix
material [12]. There are no substantial changes in the powder morphology is observed
due to the lower milling speed and time. However, uniform mixing of the matrix and
reinforcement is clearly seen from the SEM morphology of the Al/SiC powders.
Fig. 3 represent the milled powder sample of Al+5%SiC composite powder.
3.2 Hardness
The variation of the hardness value of the developed material is given in Fig. 4.
The higher content of the SiC was attributed to a higher hardness value. The hard-
ness of the sintered sample is found to be increasing with the higher content of SiC
percentage with 4.76%, 11.9%, 23.81% and 38.1% for reinforcement percentage of
5, 10, 15 and 20 wt.% in aluminum respectively. The formation of strong interfacial
bonding between the reinforcement and matrix resists the plastic deformation during
658 N. K. Bhoi et al.
50
Hardness (HRB)
40
30
20
10
0
Pure Al Al+5%SiC Al+10%SiC Al+15%SiC Al+20%SiC
Specimen
the application of load. Similar trends were reported by the researcher by the consid-
eration of several reinforcements such as TiC, TiB2 , B4 C, Y2 O3 in the Al-MMCs
[4, 6, 12]. The improvement in the material hardness can be related to homogeneous
particle distribution, uniform microwave heating of the samples and the restriction
of the grain movement to a greater extent. The well-known Hall-Petch relation and
Orowon strengthening well described the stated observation in a beautiful manner
for the property enhancement [11–14]. On a similar exploration by Samal et al.
[6] shows the highest achieved microhardness of 32% higher in caparison with base
matrix material [6]. The formation of intermetallic compound in the composite mate-
rial is sole responsible for the improved material response under similar operating
environment.
The reciprocating wear behaviour of the Al/SiC composite was assessed by means of
applying reciprocating wear conditions as stated in Fig. 2. For the assessment of wear
behaviour, a fixed amount of deadweight is applied over the sintered sample during
the reciprocating actions. The weight loss and specific wear rate of the specimen
are illustrated in Fig. 5. The presence of a hard-ceramic phase in the composite
material softens the contact zone by the generation of stable and hard phase in the
material. The higher reinforcement percentage in the material reduces the tendency
of crack propagation with strong interfacial bonding between the intermolecular
of the material during rubbing and sliding action [6]. The wear loss of pure Al is
higher compared to other reinforced material with similar operating conditions. The
variation of the specific wear rate is plotted in Fig. 5. The specific wear rate of the
material reduces with a higher content of SiC in the aluminum. This can be related
Reciprocating Wear Behaviour of Al–SiC Composite Processed … 659
0.000010
0.02
0.01 0.000005
0.00 0.000000
0 5 10 15 20
wt.% of SiC
to the hardening of the phase formed during the reinforcement in the material which
limits the worn-out stage of the material. In case Al/SiC composite the wear, the loss
is reducing with a minimum of 40% and a maximum of 96% in the case of 5 and
20 wt.% SiC respectively. Higher hardness is responsible for the reduction in the
wear loss of the material which lowers the overall particle fracturing during sliding
and rubbing action. On a similar contrary, the presence of TiC in the aerospace-
grade Al material exhibits better and higher sliding wear resistance compared to un-
reinforced material [6]. Similarly, the observations were made for the nanostructured
B4 C composite material under different loading condition with sliding conditions on
a pin on disc configuration. It was noted that presence of hard carbide particles and
formation of intermetallic phases in the material reduces the wear rate to much extent
[15].
4 Conclusion
References
1 Introduction
In the most recent couple of decades objective of copying human movement through
strolling biped robots has gotten huge consideration among scientists. Biped robots
can move in obscure landscapes, climb staircases which wheeled robots are unfit
to perform. They can be employed in hazardous works such as rescue operations
[5], disaster situation [1], or rehabilitation of disabled people, for example, dynam-
ically controlled prosthetics [10] and made humanoid and biped, to rescue injured
victims into safety; they can prevent human life from being put in danger. Due to the
complexity of human walking, which increases with the increase of no. of links and
degree of freedom (DOF), research progress in this area has been limited.
D. Kundu
Ashok Leyland, Kanpur, Uttar Pradesh 208002, India
A. Dan (B)
Indian Institute of Technology Delhi, New Delhi 110016, India
N. B. Hui
Department of Mechanical Engineering, National Institute of Technology Durgapur, Durgapur,
West Bengal 713209, India
are connected via knee joints and the hip joint connects the upper leg with the torso.
Therefore, one leg is having three joints and four links (torso is common to both the
legs). Torso, leg, and foot are modelled specifying geometric parameters and ground
contact is modeled using ‘Sphere to plane force’ block of SimMechanics [8] from
MATLAB contact library.
Torso is modelled as a rectangular block connected from right to left hip. Represen-
tation of torso in SimMechanics environment is shown in Fig. 2. Dimensions of the
torso are represented in Table 1.
Legs are modelled as rigid cylinders. Each leg contains three links named as foot,
lower leg, and upper leg (Simulink block diagram is shown in Fig. 3). They are
connected by revolute joints which are named as ankle joint (between foot and lower
leg), knee joint (between the lower leg and upper leg) and hip joint (between upper
leg and torso). All joints are actuated by input motions. Required parameters for legs
are represented in Table 2.
Here q is the actual ankle joint angle after application of torque whereas qr is the
reference ankle joint angle.
The contact force between foot and ground is unilateral, i.e., only repulsive force
exists between them when they come to the contact. To model this kind of contact,
Sphere to plane force [8] SimMechanics block has been used. This block implements
a contact force between a sphere and a plane. Feet are placed at the spherical end and
the ground is at the plane end. Block diagram of ground contact in SimMechanics
environment is shown in Fig. 5. All parameters of ground contact are represented in
Table 3.
Input gait is provided in the form of joint trajectories. In a walking cycle, all six
joint trajectories (two ankle, two knee, and two hip joints) are expressed as cubic
polynomials of time. A cubic polynomial requires four points at four given time
instants to specify all coefficients. To make joint trajectories position, velocity, and
acceleration continuous, additional three conditions are required. To take that into
consideration, three more joint angles at three given time instants are specified. So
seven joint angles have been specified at given time instants for generating a trajectory
for each joint.
All simulations are based on the ODE 15s solver which is used for solving stiff
problems, i.e., problems in which two or more solution components vary on dras-
tically different time scales [4]. Performance of PID controller also depends on the
selection of controller gains. Initially, simulation was performed with default gains
K p (proportional gain) = 1, K i (Integral gain) = 0 and K d (differential gain) = 2
(to maintain condition for critical damping). Later on, autotuning was applied with
the help of transfer function-based PID tuner app available in SimMechanics and
optimal PID controller gains are noted as K p (proportional gain) = 0.05, K i (Integral
gain) = 0.57 and K d (differential gain) = 0.05. Simulation results of walking due
to given trajectories and applied PID control at ankles are shown below. Bipedal
motion achieved with these control parameters shows that robot walking is contin-
uous, smooth, and dynamically stable, i.e., it does not fall off during locomotion. It
is observed that increasing K p and K d and reducing K i from optimal values make
biped walking non-human like gait, unstable. Details about dynamic stability are
discussed in Sect. 4.
Dynamics and Control of a 6-DOF Biped Robot … 667
Figure 6 shows that required torque at ankle joint is maximum when it leaves the
ground (changes from support phase to swing phase). Otherwise, torque requirement
at the hip joint is greater compared to the other joints. It can be validated from the
fact that in a serial chain system revolute joint that connects fixed base and first link
requires more torque than others as first joint needs to move the whole system. Here,
hip joint is serving the same purpose. Torque profile is similar in right leg and left
leg as shown in Figs. 6 and 7.
Angular speeds of three joints (ankle, knee, and hip) for both the legs are presented
in Figs. 8 and 9.
668 D. Kundu et al.
Maximum speed is observed for ankles and knee and hip joints are moving in
similar manner. Maximum speed of all the joints is not seen at the same time and
maximum torque is seen for hip joint when it moves with maximum hip speed. The
3D model of the robot during simulation is shown in Fig. 10.
Dynamics and Control of a 6-DOF Biped Robot … 669
4 Dynamic Stability
There are few methodologies for available in the literature for analyzing stability for
a biped robot model. Most used approaches are based on zero moment point (ZMP)
[11], limit cycle theory [9], and Lyapunov theory [3]. In this paper, limit cycle theory
is used to check and improve the dynamic stability of the biped model.
For a biped locomotion to be stable, phase portrait of controlled variable should
conform to a limit cycle as walking progress [9]. In biped robot system which is
discussed in this paper, PID controllers are applied at the right ankles and the left
ankles only. So phase portrait of associated joint angles and angular velocities for
last two walking cycles are studied and presented in Figs. 11 and 12. If K p and K d
are improved further, distance between last two walking cycles increases.
In Fig. 12 controller parameters used are K p (proportional gain) = 1, K i (Integral
gain) = 0 and K d (differential gain) = 2. In Fig. 11, distance between last two walking
cycle is lesser than that in Fig. 12. Varying K p , K i , and K d it can be stated that distance
between last two walking cycle is minimum for optimal controller gains as mentioned
in Sect. 2.3. It refers to the fact that with optimal controller gain values, biped robot
is dynamically most stable.
It is also observed that there are two areas in Fig. 12 where significant angular
velocity is changed with small change in angular displacement. This phenomenon
is observed at the time of impact on the ground with a high velocity. For walking of
a robot, this may damage the system. So this kind of situation should be avoided as
well.
670 D. Kundu et al.
5 Conclusion
This paper presented a way to model 6-DOF biped robot system in MATLAB/
SimMechanics. Simulation with and without PID tuning parameters makes notice-
able difference in walking pattern and dynamic stability of the system. With PID-
tuned controller, movement of each joint occurs in a more organized way and more
human like movement is obtained. A limitation of this system is that it produces flat
feet motion. To overcome this problem, toe joints can be implemented at each foot.
In turn it will incorporate more complexity to the system as more controllers will be
required and simulation of the system will require more time as degree of freedom of
the system will increase. Walking of model can be made more humanoid if contact
forces are fed as feedback to movement of the system.
Dynamics and Control of a 6-DOF Biped Robot … 671
References
1 Introduction
hinges. Wu et al. [6] proposed a typical 3-RRR type spherical parallel manipu-
lator. Author demonstrated a graphically different configuration for workspace size,
dexterity and singularity of the wrist manipulator and concluded that the reconfig-
urable wrist manipulator can meet various task requirements due to its real-time
performance enhancement. Literature surveys indicate numerous participations of
researcher in designing manipulator for various engineering application, but a limited
work in the area of vibration analysis of manipulators for dynamic performance.
The present paper considers a 3-RRR SPM to demonstrate the finite element
simulation modal analysis. Initially a 3D model is developed using Unigraphics Nx
10.0 and modal analysis is carried in the Ansys 18.1 environment. First three natural
frequency of the top plate is obtained. An experimental model of the same size
is fabricated and experimental modal analysis is carried out using LMS Test.Lab.
Results obtained by finite element simulation are having a very good agreement
with experimental results. The frequencies and mode shapes directly influence the
performance and stability of the attachment on the top plate.
The 3D SPM model is generated using Unigraphics Nx 10.0. Four basic elements,
viz., bottom circular base, three-curved guides, three-curved links and one mobile
top plate forms a simple SPM. A circular guide is considered to be the base. Three
side units are attached to the circular guide and rotate along the guide. Each side unit
is attached to a curved link through a revolute joint. The mobile platform is attached
to the side units through three curved links. The side of the top plate considered
according to the angular orientation of the curved link angle. Depending on the
orientation of the curved link, the top plate dimensions vary. The guides are placed at
120° apart to each other on the base plate, supports the top plate in horizontal position.
Change in the position of the guides, the top plate will get a different orientation other
than the horizontal position. The angular orientation of the curved link is considered
45°. Figure 1 shows the developed SPM model.
3 Simulation of SPM
Simulation model includes the circular guide, side units, curved links and top plate
made of aluminum alloy. The boundary condition considered is the fixed-free. The
base circular guide is considered as fixed support and the mobile platform is consid-
ered as free. The curved link acts as the stiffness and the top platform acts as the
mass. The finite element modal analysis of the model is carried out in Ansys 18.1.
Figures 2, 3 and 4 show the mode shapes of SPM model.
The first mode shape exhibits the structural deformation in lateral direction.
The second mode shape exhibits the structural deformation in transverse direction.
Modal Analysis of 3-RRR SPM Model 675
Fig. 2 First mode shape for SPM model isometric and front view
Fig. 3 Second mode shape for SPM model isometric and front view
676 V. Kulkarni et al.
Fig. 4. Third mode shape for SPM model isometric and front view
Table 1 Comparison of
Mode Natural frequency (Hz)
simulation and
experimentation results of Simulation Experimentation Error
SPM model (%)
1 335.64 361.85 7.24
2 394.77 427.18 7.58
3 395.16 428.50 7.70
The third mode shape exhibits the structural deformation in transverse direction.
The obtained natural frequencies through simulation are reported in Table 1 along
with the experimentation result in the Sect. 5. The first mode describes the lateral
deformation of the system. The second mode and third mode describe the transverse
deformation of the system. The second and third mode deform along the same plane
so there is no large variation in the natural frequencies. The mode shapes clearly
define the deformation in the three degrees of freedom.
4 Experimentation of SPM
5 Comparison of Results
The natural frequencies are obtained through the simulation using Ansys, and exper-
imental results are extracted using LMS setup. The simulation and experimental
results are compared with each other and listed in Table 1.
The natural frequencies values obtained from the simulation and experimentation
are in a good correlation and with a maximum error of 7.70% in the third mode.
6 Conclusion
In this paper, 3D SPM simulation model is developed. Modal analysis of the devel-
oped SPM is carried out using Ansys 18.1. The natural frequencies and mode shapes
of the system up to sixth mode is extracted and up to third mode is reported. The
mode shapes of the system clearly indicate the deformation of link and top plate
at each mode. Production drawing is developed and system get it fabricated using
commercially available aluminum. Fabricated SPM model is used to conduct the
experimental modal analysis. Proper boundary conditions are established. Using
LMS Test.Lab, experimental natural frequencies are extracted. Experimental results
obtained using LMS Test.Lab are in close agreement with the simulation results.
References
1. Patel D, George PM (2012) Parallel manipulator applications—a survey. Modern Mech Eng
2:57–64
2. Bai S, Hansen MR, Andersen TO (2009) Modelling of a special class of spherical parallel
manipulators with Euler parameters. Robotica 27:161–170
3. Wu G (2012) Multiobjective optimum design of a 3-RRR spherical parallel manipulator with
kinematic and dynamic dexterities, modelling, identification and control. 33(3):111–122
4. Castri C, Messina A (2011) Vibration analysis of multilink manipulators based on timoshenko
beam theory. J Robot 2011(890258)
5. Rong W, Luan Y, Qi L, Xie H, Sun L (2012) Stiffness analysis and modal analysis of preci-
sion parallel manipulator with flexure hinge. In: International conference on manipulation,
manufacturing and measurement on the nanoscale, 29 August-1 September
6. Wu G, Dong H, Wang D, Bai S, (2013) A 3–RRR spherical parallel manipulator reconfigured
with four-bar linkages, IEEE
Mode Based Crack Identification
of Rotor
Ridha Ali, T. Pooja Priya, V. Rashmi, C. V. Chandrashekara,
and Suneel Motru
1 Introduction
Rotating machines play an important role in almost every field of engineering, viz.
power generation, transportation and machine tools etc. Rotor is an important part
of the turbine that is subjected to bending and torsional vibrations during operating
conditions. The presence of crack in rotor, directly affects the dynamic characteristics
hindering the performance and productivity of turbine. The rotor failure accounts for
large proportions of turbine failure downtime. Identifying the crack would potentially
prevent serious damage and expensive repairs due to turbine failure.
A review on dynamics of cracked rotor by Wauer [1] summarizes, finding the most
appropriate procedure for detection of failure in the preliminary stages is crucial.
Studying the behavior of non-rotating rotor also provides a relevant basis for crack
modeling and detection of rotor to develop better diagnostic procedure. Darpe et al.
[2] studied the response of Jeffcott rotor with a central transverse crack. The bending
natural frequencies and side frequencies around impulse excitation frequency are
used for crack detection. Agarwalla et al. [3] analyzed the effect of open crack on
Crack d y
x
x
y v1 xt v2
x xd
θ1 θ2
l
Fig. 3 Cross-section of
rotor with crack
r4 θ ×π bh 3
Ic = × − sin θ + (1)
8 180 36
π d4
I = (2)
64
The equation of motion for free vibration of the system is given by Thompson,
[K ] − [M]ω2 {X i } = 0 (3)
While assembling the element matrix, care is taken appropriately to place the
crack element matrix in the global matrix.
Using Eq. (4), the eigenvector for both uncracked rotor, represented as [φ mn ] and for
cracked rotor as [φ mn ]c is extracted. Procedure to identify the crack and location is
demonstrated with the help of a flow chart shown in Fig. 8 (Annexure A).
682 R. Ali et al.
4 Illustrative Examples
The developed procedure is demonstrated with three numerical examples. The rotor
is discretized into 201 elements. First two examples demonstrate for single trans-
verse crack at two different positions. Next one example demonstrates the multiple
transverse cracks for different position. The material and geometrical properties of
the rotor used are listed in Table 1.
In this case, crack is introduced in the 111th element. Similar procedure is adopted
as described for case 1. The deviation in eigenvalues of uncracked and cracked rotor
is not shown for further case (Fig. 6).
5 Conclusion
In this paper, a new modal-based procedure to identify the location of the crack
precisely is demonstrated. As the procedure is based on the finite element method, it
is very simple and more suitable for computational-based dynamic analysis of rotor.
Extensive illustrative numerical examples are considered to check the robustness of
the procedure adopted. As the limitation of pages, only limited values are reported.
The same procedure can be extended to simulation as well as experimental validation.
Annexure A
See Fig. 8.
Mode Based Crack Identification of Rotor 685
References
1. Wauer J (1990) On the dynamics of cracked rotors: a literature survey. Appl Mech Rev 43(1):13–
17
2. Darpe AK, Chawla A, Gupta K (2002) Analysis of the response of cracked jeffcott rotor to axial
excitation. J Sound Vibr 249(3):429–445
3. Agarwalla DK, Parhi DR (2013) Effect of crack on modal parameters of a cantilever beam
subjected to vibration. Proc Eng 51:665–669
4. Chandrashekara CV, Pavan S, Dharani J, Himanshu A, Raj Arjun SI (2018) Formulation of
effective stiffness for predicting natural frequency of cracked beams. Vibroeng Proc 19:135–140
A Note on Implementation of
Raghavan–Roth Solution for
Wrist-Partitioned Robots
Abstract This paper focuses on providing a modification in the Raghavan and Roth
algorithm for solving inverse kinematics. We show that this algorithm fails for wrist-
partitioned robots. The causes of failure are explored and the underlying architecture
is modified to avoid the same. Other cases, where this algorithm does not provide
the solution, are also discussed. The modification is illustrated through an example
showing inverse kinematics of KUKA KR 5 robot.
1 Introduction
Inverse kinematics solutions are an important part of robot control. The solutions
serve as a map from the end-effector space to the joint space. Analytical expressions
for a general six-degree-of-freedom system were a challenge until 1990s. During
that time, an algorithm to compute the inverse kinematics solutions of a general
six degrees of freedom serial manipulator was presented by Raghavan and Roth in
[4]. Various algorithms based on the “Raghavan–Roth solution” were developed to
efficiently compute the inverse kinematics solutions, like the one presented in [2]. The
solution procedure is generalized and can be used in robotic simulation softwares.
The solution procedure developed in [4] is based on utilizing the matrix equations
as well as the power product equations and then, computing the determinant of a
12 × 12 matrix.
In the present work, we show that the methodology fails for some special cases.
This occurs when the 12 × 12 matrix is rank deficient. One of such cases is the
occurence of wrist partitioning robots. We analyze the cause of failure of the “Ragha-
van and Roth” solution and present a modification in the same while computing
solutions for the wrist-partitioned robots [1].
Section 2 presents a brief overview of the existing Raghavan–Roth method.
The modification in the equations for wrist-partitioned robots are given in Sect. 3.
Although this paper focuses on wrist-partitioned robots, other cases of failure are
also possible. Further insights are presented in Sect. 5.
where Ai is the transformation matrix relating the coordinate frame moving with the
i th joint to the coordinate frame fixed to the ground. In the solution algorithm, the
matrix A2 is partitioned into A2s and A2v . A2s contains fixed DH parameters while
A2v contains joint variables such that A2 = A2v × A2s , as done in [4]. Six equations
are formed after equating last 2 columns of T and U are shown in Eqs. 2 and 3.
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
cos θ2 sin θ2 0 1 0 0 cos θ3 sin θ3 0 a2
⎝ sin θ2 − cos θ2 0 ⎠ l̂ = ⎝ 0 −λ2 μ2 ⎠ ⎝ sin θ3 − cos θ3 0 ⎠ m̂ + ⎝ 0 ⎠ (2)
0 0 1 0 μ2 λ2 0 0 1 b2
γ1 β1
⎛ ⎞ ⎛ ⎞⎛ ⎞
cos θ2 sin θ2 0 1 0 0 cos θ3 sin θ3 0
⎝ sin θ2 − cos θ2 0 ⎠ ĉ = ⎝ 0 −λ2 μ2 ⎠ ⎝ sin θ3 − cos θ3 0 ⎠ d̂ (3)
0 0 1 0 μ2 λ2 0 0 1
γ2 β2
Expressions for different terms used in the above equations are given below.
A Note on Implementation of Raghavan–Roth Solution … 689
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
l1 m1 c1 d1
l̂ = ⎝ l2 ⎠ m̂ = ⎝ m 2 ⎠ ĉ = ⎝ c2 ⎠ d̂ = ⎝ d2 ⎠
l3 m3 c3 d3
m 1 = cos(θ4 ) j1 + sin(θ4 ) j2 + a2
m 2 = −λ3 (sin(θ4 ) j1 − cos(θ4 ) j2 + μ3 j3
m 3 = μ3 (sin(θ4 ) j1 − cos(θ4 ) j2 + λ3 j3 + b3
g1 = cos(θ4 )n 1 + sin(θ4 )n 2
g2 = −λ3 sin(θ4 )n 1 − cos(θ4 )n 2 + μ3 n 3
g3 = μ3 sin(θ4 )n 1 − cos(θ4 )n 2 + λ3 n 3
j1 = cos(θ5 )a5 + a4
j2 = − sin(θ5 )λ4 a5 + μ4 b5
j3 = sin(θ5 )μ4 a5 + λ4 b5 + b4
n 1 = sin(θ5 )μ5
n 2 = cos(θ5 )λ4 μ5 + μ4 λ5
n 3 = − cos(θ5 )μ4 μ5 + λ4 λ5
e = −l x a6 − (m x μ6 + n x λ6 )b6 + ρx
f = −l y a6 − (m y μ6 + n y λ6 )b6 + ρ y
g = −l z a6 − (m z μ6 + n z λ6 )b6 + ρz
x = m x μ6 + n x λ6
y = m y μ6 + n y λ6
z = m z μ6 + n z λ6
i 1 = cos(θ1 )e + sin(θ1 ) f − a1
i 2 = −λ1 (sin(θ1 )e − cos(θ1 ) f ) + μ1 (g − b1 )
i 3 = μ1 (sin(θ1 )e − cos(θ1 ) f ) + λ1 (g − b1 )
c1 = cos(θ1 )x + sin(θ1 )v
c2 = −λ1 (sin(θ1 )x − cos(θ1 )y) + μ1 z
c3 = μ1 (sin(θ1 )x − cos(θ1 )y) + λ1 z
where, λ = cos(α) and μ = sin(α). The six equations (Eqs. 2 and 3) are written as
two vector equations as presented in Eq. 4.
γ1 = β1 , γ2 = β2 (4)
The other eight equations arise from the equivalence of power product terms. Power
products are the terms from a polynomial excluding constant coefficients. (e.g., the
power products of the polynomial 2x y + x + x 2 y = 0 are x y and x, x 2 y). A detailed
explanation for the power products can be found in [3]. So, for the Raghavan–Roth
solution, eight additional equations obtained from the power product are given in
Eqs. 5 and 6.
γ1 T γ2 = β1 T β2 , γ1 T γ 1 = β1 T β1 , γ1 × γ2 = β1 × β2 (5)
So, the total no. of equations obtained is 14. The 14 equations can be written in
the form of Eq. 7. Any 8 of the 14 equations are taken from Eqs. 2, 3, 5 and 6 and
can be written in form of Eq. 8, where Aα is a 8 × 9 matrix and Bα is a 8 × 8 matrix.
⎛ ⎞
sin(θ4 ) sin(θ5 ) ⎛ ⎞
⎜ sin(θ4 ) cos(θ5 ) ⎟ sin(θ1 ) sin(θ2 )
⎜ ⎟ ⎜ sin(θ1 ) cos(θ2 ) ⎟
⎜ cos(θ4 ) sin(θ5 ) ⎟ ⎜ ⎟
⎜ ⎟ ⎜ cos(θ1 ) sin(θ2 ) ⎟
⎜ cos(θ4 ) cos(θ5 ) ⎟ ⎜ ⎟
⎜ ⎟ ⎜ cos(θ1 ) cos(θ2 ) ⎟
⎜
(A) ⎜ sin(θ4 ) ⎟ ⎜ ⎟
⎟ = (B) ⎜ sin(θ1 ) ⎟ (7)
⎜ cos(θ4 ) ⎟ ⎜ ⎟
⎜ ⎟ ⎜ cos(θ ) ⎟
⎜ sin(θ5 ) ⎟ ⎜ 1 ⎟
⎜ ⎟ ⎝ sin(θ ) ⎠
⎝ cos(θ5 ) ⎠ 2
cos(θ2 )
1
2xi
sin(θi ) =
1 + xi2
(10)
1 − xi2
cos(θi ) =
1 + xi2
⎛ ⎞
x42 x52
⎜ x 2 x5 ⎟
⎜ 42 ⎟
⎜ x ⎟
⎜ 42 ⎟
⎜ x4 x ⎟
⎜ 5 ⎟
(σ ) ⎜
⎜ x 4 x ⎟
5⎟=0 (11)
⎜ x4 ⎟
⎜ 2 ⎟
⎜ x ⎟
⎜ 5 ⎟
⎝ x5 ⎠
1
where σ has terms containing x3 only. Equation 11 can be rearranged [4] to form
Eq. 12 with determinant equal to zero.
σ 6×9 06×3
=0 (12)
06×3 σ 6×9
3 Modified Architecture
Wrist-partitioned robots have special characteristics such that the last three axes
intersect at a particular point (Fig. 1). A set of general DH parameters for a wrist-
partitioned 6R robot is given in Table 1. The DH parameters are defined as in [5].
By using generalized DH parameters for wrist-partitioned robots, we can prove
the failure of Raghavan–Roth solution.
Using the definitions of variables in Sect. 2 and [4], the subequations reduce to the
following set of Eqs. 13–18 for DH parameters presented in table 1.
j1 = 0 (13)
j2 = 0 (14)
692 R. Kumar et al.
j3 = b4 (15)
m 1 = a3 (16)
m 2 = μ3 b4 (17)
m 3 = λ3 b4 + b3 (18)
So, the initial set of six equations (Eqs. 2–3) (Sect. 2) reduces to the set of Eqs. 19–24
sin(θ3 )d1 − cos(θ3 )d2 = −λ2 (sin(θ2 )c1 − cos(θ2 )c2 ) + μ2 c3 (23)
The other eight equations are to be formed from the power product equations. It
should be observed that the left-hand side of Eqs. 19–21 is independent of any
trigonometric function of θ4 and θ5 . However, Eqs. 22–24 have terms on the left-hand
side containing coefficients (cos(θ4 ) sin(θ5 ), cos(θ5 ), sin(θ4 ) sin(θ5 )). So, the power
product terms have terms with coefficients (cos(θ4 ) sin(θ5 ), cos(θ5 ), sin(θ4 ) sin(θ5 ),
1). While synthesizing the matrix A, the rows corresponding to the coefficients
(sin(θ4 ) cos(θ5 ), cos(θ4 ) cos(θ5 ), sin(θ4 ), cos(θ4 ), sin(θ5 )) are null column vectors.
So, A matrix is a rank 4 system. If we partition matrix A into Aα with 8 rows and
Aβ with 6 rows, then the net matrix σ reduces to Eq. 25. It should be noted that wrist
partitioning does not affect any element of matrix B, so under normal circumstances,
the submatrix Bα is of full rank and remains invertible.
So, the matrix σ is a matrix of rank 4 for the case of wrist-partitioned robots. So, we
extract a σ6×4 from σ6×9 to formulate a new submatrix P of the form in Eq. 26 to
ensure that the determinant is not trivially zero and we are able to gain a characteristic
equation out of the same. However, in case, the submatrix σ consists of redundant
rows, then a submatrix with independent set of rows need to be selected (σ ) needs
to be selected appropriately. This might happen for some selected robots with some
special set of DH parameters making other equations redundant, like for the case of
KUKA KR 5 (presented in Sect. 4) which has equal b2 and b3 .
σ 6×4 02×4
=0 (26)
02×4 σ 6×4
P
The solution to Eq. 26 results in a polynomial which has inverse kinematics solution
of wrist-partitioned robots. The determinant of 8 × 8 matrix gives a characteris-
tic polynomial (degree 16) for wrist-partitioned robots. The methodology has been
applied for a case of KUKA KR 5 robot, which is wrist partitioned and also has
orthogonal axes. However, the proof presented does not ensure failure of Raghavan–
Roth solution when three intersecting axes are not terminal.
694 R. Kumar et al.
For the case of KUKA KR 5, the set of Eqs. 19–24 reduce to the set of Eqs. 28–33.
120 cos(θ3 ) − 620 sin(θ3 ) = cos(θ2 )(12.34 cos(θ1 ) + 100 sin(θ1 ) − 180 (28)
120 sin(θ3 ) + 620 cos(θ3 ) = sin(θ2 )(12.34 cos(θ1 ) + 100 sin(θ1 ) − 180) − 892.81
(29)
0 = 100 cos(θ1 ) − 12.3455 sin(θ1 ) (30)
Equations 28–30 do not contain θ5 and θ4 at all. The other eight equations need to be
created from the power product. The left-hand side of Eq. 30 does not contain any term
because KUKA KR 5 has equal b2 and b3 , which further reduces the rank of σ . This
later results in a row of σ to be a zero row vector. Hence, the we use a 5 × 4 matrix
of σ and a 3 × 4 zero matrix in Eq. 26. The determinant results in eight feasible
solutions of θ3 given by {−165.66◦ , −165.66◦ , −160.14◦ , −160.14◦ , 2.03◦ , 2.03◦ ,
7.55◦ , 7.55◦ }. The final solution is verified from the KUKA KR 5 inverse kinematics
presented in [5].
A Note on Implementation of Raghavan–Roth Solution … 695
5 Discussion
Fig. 2 Equivalence of forward kinematics of the parallel robot to the inverse kinematics of the
serial robot
696 R. Kumar et al.
6 Conclusion
This paper is based on modifying the Raghavan and Roth solution to implement on
wrist-partitioned robots of general geometry. A submatrix is detected and utilized
to formulate a polynomial whose roots lead to inverse kinematic solutions. This
method augmented with the original Raghavan and Roth solution can be used for a
larger class of general robots, including wrist-partitioned robots, which are prevalent
in industrial uses. There can be other cases of failure of the solution which can be
investigated using a similar way and is a subject of further study.
References
1. Hollerbach JM, Sahar G (1984) Wrist-partitioned, inverse kinematic accelerations and manipu-
lator dynamics 2:152–161. https://doi.org/10.1109/ROBOT.1984.1087172
2. Manocha D, Canny JF (1994) Efficient inverse kinematics for general 6R manipulators. IEEE
Trans Robot Autom 10(5):648–657. https://doi.org/10.1109/70.326569
3. Raghavan M, Roth B (1990) Kinematic analysis of the 6r manipulator of general geometry, pp
314–320
4. Raghavan M, Roth B (1993) Inverse kinematics of the general 6R manipulator and related
linkages. J Mech Des 115(3):502–508
5. Saha SK (2014) Introduction to robotics. Tata McGraw-Hill Education, New York
Experimental Identification of Residual
Unbalances for Two-Plane Balancing
in a Rigid Rotor System Integrated
with AMB
1 Introduction
The rotating machinery, such as generators, gas turbines, machine tools, and indus-
trial turbomachinery, is commonly used in industries. Vibration due to residual unbal-
ance is a major concern as it limits the performance and fatigue life of the rotating
components. The residual unbalances (inherent unbalances) in machineries occur
due to manufacturing errors, improper commissioning, wear and tear, and so on.
The amount of vibration induced by residual unbalance is proportional to square
of rotating speed which can cause serious damage to the rotor system passing the
critical speed. Therefore, the balancing of rotor system is mandatory to reduce the
vibration occurrence in the system at high speed. To reduce excessive vibration, many
balancing methods such as single-plane balancing, cradle balancing, and influence
coefficient method have been developed by practicing engineers and researchers. The
influence coefficient method is widely used approach for balancing of rotor systems
in industrial applications. The balancing procedure involved is easy to carry out on
a computer once measurement is available and requires very little knowledge of the
rotating machineries mathematical modeling [1, 2]. Kang et al. [3] formulated influ-
ence coefficient matrices for unsymmetrical rotors from equations of motion using
the finite element method. It was found that two trial operations for the measurement
of unbalance responses are required for determination of influence coefficients and
obtained the correction masses from influence coefficient, which reduced the residual
vibration. Pennacchi et al. [4] investigated several robust regression methods for
rotor balancing along with influence coefficients. They tried to develop an automatic
balancing procedure for estimation of balancing masses based on the above method.
Seve et al. [5] presented a balancing procedure of machines composed of a flexible
rotating part (rotor) and a non-rotating part (stator) mounted on suspensions. The
procedure was based on a numerical approach using rotor dynamics theory coupled
with the finite element and influence coefficient method. Lin et al. [6] used finite
element analysis to simulate the balancing of flexible rotor systems under various
arrangements of sensors and planes. They showed that accuracy of balancing can
be improved by selecting sensor locations and balancing planes which reduce the
condition number.
Lots of researches have been carried out to actively control the rotor system vibra-
tion, and it can be done effectively by the use of active magnetic bearing (AMB),
which provides contactless actuation and helps in reduction of wear and tear. AMB
reduces excessive vibration by applying magnetic force in the opposite direction to
unbalance forces. However, instead of applying this corrective force all the time, if
this corrective force from AMB can be used for finding the residual unbalance state
of the rotor that will save unnecessary power consumption of the AMB. Chougale
and Tiwari [7] proposed an algorithm for the estimation of residual unbalances and
dynamic parameters of active magnetic bearing for flexible rotor system. The algo-
rithm uses the AMB current and unbalances responses for estimating the dynamic
parameters (for each AMB) and unbalances in flexible rotor. Kim and Lee [8] devel-
oped an active balancing device which is of an electromagnetic type and proposed an
Experimental Identification of Residual Unbalances for Two-Plane … 699
active balancing method using the influence coefficient method. They also verified
the effectiveness of balancing method by experiments. Moon et al. applied an active
balancing program using the influence coefficient method and an active balancing
device of electromagnetic type with both simple and reliable structures to the devel-
oped high-speed spindle system. Gyan and Tiwari [9] described a generalized influ-
ence coefficient method (GICM) to estimate the unbalances present in the flexible
rotor-AMB system supported on conventional bearings.
However, it is beneficial to utilize AMB in the system while estimating the
balancing mass in order to suppress the excessive vibration reducing fatigue stress.
Also after balancing with reduction in unbalance, the requirement of controlling
current to suppress vibration reduces resulting in less power consumption. AMB
in the system after balancing can be utilized to avoid any unexpected increase in
response of the rotor system. So, in the present work, an experimental simulation is
described to perform the dynamic balancing of the rigid rotor system integrated with
an AMB.
The influence coefficient method is used to predict residual unbalances in a rigid rotor
system with the use of displacement values at two measuring planes. Displacements
values are obtained for different trial unbalances at balancing planes for estimation
of influence coefficients. Then, these are used in obtaining residual unbalances at
balancing planes. A typical rotor system consisting of a shaft, rigid disks (balancing
planes), and sensors (measuring plane) mounted on conventional bearings is shown
in Fig. 1.
In the ICM [2], two balancing planes m and n are selected for trial unbalances, and
two measurement planes a and b are considered for the measurement of displacements
of the rotor system. Apart from this, a reference signal is measured at a convenient
shaft location to have measurement of all phases with respect to this signal. Displace-
ments of the shaft are considered as r ia and r ib at the left and right measuring planes,
respectively, and are shown below as
Fig. 1 Schematic of a
rotor-bearing system
a b
CB-1 CB-2
m n
700 G. Ranjan et al.
with
where AL and AR are the amplitudes of measurement, √ and φ L and φ R are the phase
angles with respect to a reference signal, and j = −1. The subscript i (=1 to 3)
represents different rotor runs, i.e., for i = 1, no trial unbalance is taken, and for i =
2 and 3, trial unbalance is taken at right and left balancing plane, respectively. In the
second rotor run, the trial unbalance T R is taken at the right plane, and in the third
run, the trial unbalance T L is taken at the left plane. The displacements for different
trial runs are related as shown below
r1a αan αam Wn
= (2)
r1b αbn αbm Wm
r2a αan αam Wn + T R
= (3)
r2b αbn αbm Wm
r3a αan αam Wn
= (4)
r3b αbn αbm W m + TL
where W n and W m are the residual unbalances at the right and left planes, respectively;
α an , α bn , α am, and α bm are the influence coefficients.
The influence coefficients obtained are given as [2]
The correction unbalances are then added at the balancing planes to reduce the
vibrational response of the system.
Experimental Identification of Residual Unbalances for Two-Plane … 701
3 Working of AMB
An active magnetic bearing system supports a rotating shaft, without any physical
contact by suspending the rotor in the air, with the help of magnetic force. AMB
has unique characteristics of the absence of wear, lubrication, and seals due to free
of contact between the rotor and the stator. The component of an AMB is actuator,
amplifier, proximity sensor, and controller. Whenever air gap changes between the
rotor and the AMB, the change is sensed by the sensor, and subsequently, the current
is fed into the actuator through controller to apply magnetic force and maintain the
equilibrium position of the rotor. A schematic diagram of AMB has been shown
Fig. 2.
The AMB uses displacement stiffness and current stiffness in two orthogonal
directions. A proportional-derivative (PD) controller has been used in the system.
The controller uses the transverse displacement at the node to generate the controlling
current which is fed to the actuator to produce adequate magnetic force.
The controlling force is expressed as [2]
f mb = − K s ηm + K i i c (9)
With
k sx x k sx y ki x x ki x y
Ks = , Ki =
ks yx ks yy ki yx ki yy
ic COILS
N S
AMPLIFIER
fm
ROTOR
4 Experimental Simulation
The experimental setup for the balancing of flexible rotor using AMB is developed
in the present work. The experimental setup available in the laboratory is a standard
Bently Nevada Rotor Kit RK4 as shown in Fig. 3. The rotor kit consists of a 10 mm
diameter and 560 mm long shaft supported on two journal bearings. The shaft weighs
350 gm and has two detachable disks of 800 gm each mounted on it. The disks have
slots for the addition of unbalance in the system to predict the balance masses using
influence coefficient method. The additional components used are magnetic actuator,
PD controller (dSPACE), and amplifier.
The physical connection of the experimental setup for the vibration control problem
is shown in Fig. 4.
In the setup, change in position of the shaft is sensed by the eddy current sensors,
and the analog signal is sent to input panel of the dSPACE I/O (input–output) panel
as shown in Fig. 4. The analog signal is converted into a digital signal, and then,
controlling current is generated though the PD controller (Simulink model based).
The controlling current is transferred to the I/O panel to get an analog output for the
amplifier. The current from the amplifier is fed to a magnetic actuator to reduce the
vibration of the rotor system. In Fig. 5, the electrical circuit developed for working
of differential-type actuator is shown.
The circuit consists of a dSPACE platform, dc voltage source, a magnetic actuator,
eddy current sensor, and amplifier. The bias voltage is supplied to two opposite poles
in equal amount, i.e., eight-pole ib1 for upper and lower poles and ib2 for right and left
pole as shown in Fig. 5. The displacement signal measured through the eddy current
sensors from the rotor is utilized to generate the controlling current through a PD
controller in dSPACE platform. The controlling current, i.e., ic1 flowing through the
poles gets added in upper pole bias current, and at the same time, it gets subtracted
from lower pole. As the shaft moves downward, the overall current in upper pole,
i.e., ib1 + ic1 and that in lower pole, i.e., ib1 − ic1 allows the shaft to move back in its
original position. In similar manner, control currents are generated for both the right
and left poles.
Analog to
Rotor system Position sensor
Digital input
dSPACE
Controller
Digital to
Magnetic actuator Amplifier
Analog output
Fig. 4 Block diagram showing the physical connections for the vibration control problem
704 G. Ranjan et al.
Amplifier
Sensor
ic1
ib1 ic1
y
Amplifier
ic1 N S
Dspace
ib 2 ic 2 Sensor
Amplifier
ic 2
x
Amplifier
ib 2 ic 2
ic 2
ib1 ic1
Fig. 5 Block diagram of the electrical circuit for working of differential-type actuator
The following steps are involved in the installation and working of experiment setup
for the reduction of vibration due to unbalance.
1. The active magnetic bearing was mounted on the shaft, and the shaft was rotated
manually to check the proper alignment of bearing with the axis of shaft. The
bearing was mounted in between two balancing disks, as shown in Fig. 3.
2. The piezoelectric sensors were mounted and were maintained at a gap of approx-
imately 30 mils (0.8 mm) or −5.75 ± 0.5 Vdc on the voltmeter. The gap was
maintained to avoid nonlinearity arising in the displacement data received from
the sensors.
3. The controller was designed for the system, using Simulink and MATLAB and is
executed with the help of the dSPACE interface. The connection was established
between magnetic bearing, amplifier, and dSPACE, as shown in Fig. 5.
4. In the dSPACE control desk, the controlling parameters (k P and k D ) are changed
while the real-time simulation of the Simulink model. The real-time plot of
displacement of rotor system along vertical axis vs. time was seen using data
acquisition instruments available.
Experimental Identification of Residual Unbalances for Two-Plane … 705
5. The value of k P was varied from 0 to 14,000 A/m and that of k D from 0 to 15
A-s/m to see the gain limit of the controller. The fluctuation of the shaft was
seen and stored with the use of data acquisition toolbox.
6. The values of k P and k D are kept within the stable limit to see the reduction in
vibration of the shaft with a variation of controlling parameters than without the
use of magnetic bearing.
7. The system is applied with an impulsive force at the standstill condition, and
the changes in the damping and stiffness are calculated for the system with or
without the effect of magnetic bearing.
8. Also, the system was found to be stable for the value of k P between 5000 and
14,000 and k D between 3 and 15.
The system response is obtained after the application of the impulsive force on
the system at standstill condition without or with application of AMB as shown in
Fig. 6. The increase in proportional gain increases the stiffness and the derivative gain
increases the damping property of the magnetic bearing. As seen in Fig. 6d, with the
application of PD control through AMB, the vibration response gets eliminated in
fewer cycles than in case of without AMB (d). The change in stiffness and damping
property of the system is observed after the application of AMB in the system in a
vertical direction. The controller parameter is considered to be k P = 13,000 and k D
= 14. First, the FFT of the displacement response is carried out to obtain the natural
frequency of the system. With the help of the overall mass of the system and natural
frequency, the stiffness and damping of the system are calculated as shown in Table
1.
Fig. 6 In response to a impulsive force, b displacement of the rotor system without AMB,
c impulsive force, and d displacement of the rotor system with AMB
706 G. Ranjan et al.
a b
Coupling
Motor
CB-1 CB-2
AMB
m n
Fig. 7 Schematic of the experimental rotor-AMB setup
The dynamic balancing of the rigid rotor system is carried out, including AMB in the system.
The two disks (m and n) act as balancing planes, whereas eddy current sensors at two different
locations (a and b) act as measurement planes as shown in Fig. 7. The AMB suppresses the
vibration, and the vibration response is then utilized in the influence coefficient method to
predict the balance masses.
The following steps are involved in the dynamic balancing of the rotor system.
1. The controller parameter is considered to be k P = 13,000 and k D = 14.
2. The reduction in displacement with the application of AMB at 900 rpm is
checked as shown in Fig. 8. The half spectrum plots obtained after FFT for the
displacement data with and without AMB are also shown in Fig. 9.
3. The displacement amplitude without AMB at 1X harmonic is found out to be
9.5 × 10–5 m and with AMB is equivalent to 1.3 × 10–5 m, which gives a
percentage reduction of 86.2%.
4. The zero degree axis on the balance planes, i.e., disks is made coincident with
the key phasor on the coupling of the system.
5. The system is rotated to obtain the displacement along with both x and y
directions at the measurement planes due to residual unbalance in the system.
6. A reference signal is also obtained from the key phasor location to obtain
the desired displacement signal with phase compensation from the reference
signal.
7. A trial mass is then added to the right plane, and displacement is obtained at
the measurement planes.
Experimental Identification of Residual Unbalances for Two-Plane … 707
-4
x 10
1.5
With AMB
Without AMB
1
Displacement (m)
0.5
-0.5
-1
-1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)
Fig. 8 Displacement of the rotor system with and without consideration of AMB at 900 rpm
-4
x 10
1
With AMB
Without AMB
Displacement amplitude (m)
0.8
0.6
0.4
0.2
0
0 500 1000 1500 2000 2500 3000 3500 4000
Spin speed(rpm)
Fig. 9 Displacement amplitude of the rotor system with and without consideration of AMB versus
spin speed in rpm
8. Similarly, the next set of displacement is obtained for trial unbalance at the
left plane.
9. The corrected unbalances are obtained at the two balancing plane using ICM.
10. The obtained corrected unbalances are then added at the balancing planes,
and displacement responses are obtained without AMB to see the reduction in
vibration.
The identification of residual unbalances is done for different spin speed, i.e.,
900, 1000, and 1200 rpm considering AMB and for 900 rpm without considering
AMB. The balance masses obtained at left (m) and right (n) balancing planes are
708 G. Ranjan et al.
Table 2 Estimated balance masses at the left and right balancing planes for cases with or without
considering AMB
Balance mass at right plane Balance mass at left plane
Cases Speed Magnitude Phase Magnitude Phase
(RPM) (kg-m) (degree) (kg-m) (degree)
Without AMB 900 2.4 × 10–4 – 11.6 3.5 × 10–4 129
With AMB 900 2.6 × 10–4 – 3.8 3.1 × 10–4 134.8
1000 2.2 × 10–4 – 12.5 2.9 × 10–4 131.9
1200 1.5 × 10–4 – 23.9 2.1 × 10–4 125.7
shown in Table 2 for both the cases. To balance the system, the balance masses of
2 gm at a radial distance of 3 cm from the center are added at 0° at the left plane
and 135° at the right plane. After balancing, the displacement obtained is compared
with the initial displacement response (without balancing) at 900 rpm as shown in
Fig. 10. The FFT of both the displacement data is carried out to see the % reduction
of vibrational displacement. The displacement amplitude before balancing at 1X
harmonic is found out to be 9.5 × 10–5 m and after balancing equivalent to 8 × 10–5 m
which gives a percentage reduction of 15.78%. As the displacement response consists
of displacement due to shaft bow and residual unbalances, the slow run (250 rpm)
response of the system is also obtained. The slow run amplitude at 1X harmonics
is found out to be 6.8 × 10–5 m which is then subtracted from the amplitude of the
response before and after balancing as mentioned above. After slow run removal, the
percentage reduction in amplitude of vibration due to residual unbalance is found
out to be 55%.
-4
x 10
1.5
After balancing
Before balancing
1
Displacement (m)
0.5
-0.5
-1
-1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)
Fig. 10 Displacement of the rotor system before and after balancing at 900 rpm
Experimental Identification of Residual Unbalances for Two-Plane … 709
5 Conclusions
References
Abstract Articulated steering system of a load haul dump (LHD) machine is made
possible with the help of one/two hydraulic cylinders mounted between the two body
sections. The present article investigates the variables affecting the steering perfor-
mance like maximum possible steering angle (MPSA) and minimum steering force
at articulated joint. In this respect, the geometrical approach is adopted to study the
steering system of the LHD machine, and how the actuator mounting points influence
the steering performance. Based on the study, MATLAB® code has been developed
for finding the optimum mounting points of the steering actuators for MPSA. 3D
model of the LHD machine is developed based on the optimum points obtained
from the MATLAB® program. The force analysis of the 3D model is carried out in
ADAMS environment. Based on the results, it is found that the steering cylinders
with inward type offer MPSA and minimum steering force. Also, the optimization
technique illustrates that the maximum steering angle may be achieved up to 58°,
which leads to angle increase by 28% of the existing LHD machine.
1 Introduction
Articulated vehicles are having two or more body sections connected by a joint called
articulated joint. Wheel loaders and LHD machine are the examples of articulated
vehicles employed in mining industries as a trackless loading equipment to trans-
port the mineral from ore site to the dumping site in narrow constrained pathway.
Steering system of the LHD machine is provided by means of hydraulic steering
cylinders. They are mounted symmetrically between the two body sections of the
vehicle about the articulated joint of the LHD machine. Depending upon the size,
capacity of the machine and its working conditions, manufacturers developed articu-
lated steering system (ASS) using single or double hydraulic cylinders. In the current
scenario, the MPSA achieved by the LHD machine is in the range of 40–45° . The
mounting arrangements of the steering cylinders to achieve the maximum steering
angle, minimum steering force, better response and steering performance are the
wider field of research.
The literary works on the steering performance and optimization of steering
actuator mounting points of the articulated vehicle are discussed below:
Zhu et al. [1] adopted a multidisciplinary design optimization (MDO) method
for the steering system of the large-wheeled harvesting equipment. Almost, all the
design parameters were considered to find optimal objective function and validated
the results. Lei-yan et al. [2] applied genetic algorithm (GA) on steer-by-wire (SBW)-
type steering mechanism. Dynamic modeling of SBW steering mechanism was devel-
oped in MATLAB, and only vehicle parameters like tire relaxation length, front and
rear tire cornering stiffness are optimized through GA. Lei et al. [3] established
optimum design model of steering mechanism using mathematical equations. Using
MATLAB optimization toolbox, optimized pivoting points of the steering actuator
of a HT25J wheel loader is determined.
Thulsiraman et al. [4] discussed about the mounting points of steering actuators
using CAD package optimization tool and validated experimentally. The results show
that with optimized parameters of the mounting points and steering actuator dimen-
sions’, stresses at the mounting points are reduced, and steering force required at the
bore end side is also reduced. Li et al. [5] in his article discussed about a new method
to optimize the wheel loader digging force which combines sensitivity analysis with
genetic algorithms to reduce design variables and to improve the optimization effi-
ciency. The results show that approximately six percent increase in loader digging
force is observed after the optimization. Jin et al. [6] optimized the steering proper-
ties of a hydrostatic steering system in simulation environment using combination of
PID controller with GA. The required oil flowing through the cylinder is controlled
by electro-hydraulic proportional valve using PID controller, and to reduce pressure,
fluctuations constants of PID are optimized using GA. This resulted in rapid dynamic
response and good steering performance. Yin et al. [7] report based on kinematic and
dynamic analysis of the articulated vehicle suggest that length of the steering actua-
tors and steering actuators geometrical arrangement influences the maneuverability
and steering efficiency. Optimized results using analytic hierarchy process (AHP)
model suggest that reduction in steering cylinder length provides more compact
steering frame design and better steering performance of the vehicle. Zhou et al. [8]
proposed GA-based optimization of steering mechanism, and its effectiveness is veri-
fied using automated dynamic analysis of mechanical systems (ADAMS) software.
Cao et al. [9] worked on the depth optimization of hinged position of the articulated
steering system using GA.
None of the above articles discussed on the prediction of MPSA based on the
steering cylinder mounting points. This article focuses the maximization of steering
angle and minimization of steering force with respect to the geometric parameters of
the steering arrangements to obtain the optimum steering performances of the LHD
Determination of Steering Actuator Mounting Points … 713
machine. This type of study has not been carried out to the best of authors’ knowledge,
which is the novelty of the present article. This also facilitates the manufacturer
to estimate the optimized mounting points of the steering cylinders used for any
articulated vehicle.
Fig. 2 Geometric
arrangement of the steering
cylinders
The angle subtended with circle of radius ‘r’ between the points AA’ or BB’
with articulated joint as center of the circle is the steering angle. As shown in the
below figure, ‘α’ is the steering angle considered in this case. The angle ‘φ’ is the
inclination of the steering actuator with a line parallel to reference line along in ‘x’
direction. The angles ‘γ ’ and ‘ψ’ are the subtending angles with respect to the initial
inclination of the steering actuators when vehicle is taking turn.
Overall, the steering cylinder mounting arrangement can be expressed in terms
of four parameters, namely ‘L,’ ‘d,’ ‘r’ and ‘θ.’ Maximum extending length and
minimum retracting length will be same for a particular type of actuator. Dimensions
of the 912E LHD model is considered, and the possible range of values for the steering
cylinder mounting points are determined as shown in Table 1:
With the help of geometric relationships
Le r d
= = (6)
sin(180 − (θ + β + α)) sin(θ − ϕ − γ ) sin(β + ϕ + γ − α)
L r d
= = (7)
sin(180 − (θ + β)) sin(θ − ϕ) sin(β + ϕ)
Similarly,
Lc r d
= = (8)
sin(180 − (β + θ − α) sin(θ − ϕ + ψ) sin(β + ϕ − ψ − α)
sin(θ + β − α) sin(θ + β)
− = sin(β − α) − sin(α) (9)
sin(θ − ϕ − γ ) sin(θ − ϕ)
L +e sin(θ − ϕ + γ )
= (10)
L −c sin(θ − ϕ + ψ)
d∗ sin θ − r ∗ sin β
tan(ϕ) = (11)
d∗ cos θ + r ∗ cos β
2.1 Flowchart
The flowchart depicted in Fig. 3 explains the logic for the program code developed
in MATLAB platform. This program determines the parametric values of cylinder
arrangement for which the desired steering angle is obtained.
716 S. Rowduru et al.
Based on the flowchart mentioned in Fig. 3, the MATLAB code has been developed,
where the ranges for the steering arrangement parameters are varied as shown in
Table 1. The set of the possible steering arrangement with parametric values for the
desired range of MPSA (i.e., 58–60°) is shown in Table 2.
MATLAB program provides the results for only MPSA for the steering cylinder
mounting parameters. Only finding the MPSA is not sufficient for the optimum
Determination of Steering Actuator Mounting Points … 717
Table 2 Parametric values of steering arrangement (L, d, r and θ) for maximum range of MPSA
L (mm) d (mm) r (mm) θ (degrees) Steering angle (α), (degrees)
650 585 165 28 58.8067
660 575 175 28 58.4201
660 580 170 28 59.1150
660 585 165 28 59.8538
650 585 165 30 58.8067
660 575 175 30 58.4201
660 580 170 30 59.1150
660 585 165 30 59.8538
650 585 165 32 58.8067
660 575 175 32 58.4201
660 580 170 32 59.1150
660 585 165 32 59.8538
650 585 165 34 58.8067
660 575 175 34 58.4201
660 580 170 34 59.1150
660 585 165 34 59.8538
In this study, force experienced at the articulated joint is taken into account. This
is because of two reasons. First one is the load acting on the machine is assumed
constant, i.e., only machine self-weight is considered. So therefore, steering force
required will be constant. Secondly, articulated joint is the weakest point on the
machine, and therefore the dynamic force experienced at the articulated joint should
be minimum.
The force analysis has been carried out with the help of ADAMS environment,
and all the real ground parameters acting on the vehicle when employed will be same.
The following forces are considered for the simulation analysis as shown in Fig. 4:
• Force due to machine weight,
• Pressure acting across the cylinders according to the load (assumed constant load)
and
• Friction forces acting along with the tires (assumed constant friction force).
718 S. Rowduru et al.
Figure 5 explains the neutral and maximum steered position of the LHD vehicle
in ADAMS simulation environment. The variation in the articulation angle with
time and the force experienced at the articulated joint has been analyzed with each
configuration.
Combined results based on MATLAB program for MPSA and through ADAMS for
force analysis have been discussed for better readability and understanding purpose.
Mounting positions can be identified with the help of four parameters (L, d, r and θ ),
and hence, four cases have been developed for the analysis part. In each case, one
parameter is kept varying, and other three parameters are kept constant. The effect
of variation of MPSA and force experienced at articulated joint with each parameter
varying is discussed below in detail:
Case (i) With varying theta (‘θ ’)
(L = 650 mm, d = 585 mm and r = 165 mm are constant).
Discussion
Referring to Fig. 6, MPSA is independent of individual ‘θ ’ value, whereas the force
at the articulated joint is having increasing nature with the increase in the value of ‘θ.’
The range of forces is also increasing with increasing value of ‘θ.’ Hence, based on
the results, it can be concluded that the value of ‘θ ’ should be the possible minimum
value for the better steering performance. The minimum possible value of theta is
decided by the selection of ‘L,’ ‘d’ and ‘r’ values.
Determination of Steering Actuator Mounting Points … 719
Fig. 5 Simulation of LHD vehicle in ADAMS a neutral case and b steered to extreme right
Discussion
By keeping all the other parameters constant, from Fig. 7, it can be observed
that MPSA follows (approximately) second-order polynomial function by varying
distance ‘d.’ But, the sum of angles ‘θ +β’ value is increasing linearly with increase
720 S. Rowduru et al.
in distance ‘d.’ This is because the sum of ‘θ +β’ is dependent majorly on lengths
‘d’ and ‘r.’ This can be referred from Eq. 3. As the length of ‘d’ is increasing, sum
of ‘θ +β’ value is also increasing. No significant impact on the force is experienced
at the articulated joint with varying value of ‘d.’ Therefore, from the graph, it can be
observed that with increase in the value of ‘d’ there is only a slight variation in the
force and MPSA decreases with increasing value of ‘d.’
Case (iii) With varying ‘r’
(For L = 650 mm, d = 585 mm and θ = 28° are constant).
Discussion
By referring to Fig. 8, it can be observed that similar pattern is observed as that of the
case with varying ‘d.’ Value of MPSA is decreasing with increasing value of ‘r’ but
the sum of ‘θ +β’ value is increasing. But the slope of ‘sum (θ +β)’ is less compared
to the slope of ‘sum (θ +β)’ with the earlier case. In case of force experienced at the
articulated joint, it is decreasing with the increase in the value of ‘r.’ Higher value of
‘r’ is useful for experiencing least force, and lower value of ‘r’ provides the higher
MPSA value. Hence, there is a need to optimize the value of ‘r.’
Case (IV) With varying ‘L’
(For r = 165 mm, d = 585 mm and θ = 28° are constant).
Discussion
Referring to Fig. 9, sum of ‘(θ +β)’ value is decreasing with increase in value of
‘L.’ But the value of MPSA initially increases and achieves a maximum value, and
thereafter, MPSA value decreases with increase in value of ‘L,’ whereas force expe-
rienced at the articulated joint behaves differently when compared to MPSA. With
increasing value of ‘L,’ force range also increases which is not desirable. The value
of length ‘L’ at which maximum value of MPSA is obtained experiences the least
force. By observing the graph, the value of length L = 650 and at length L = 660,
MPSA is attaining maximum value. But force experienced at L = 650 is less than
L = 660. Further, by increasing the value of ‘L,’ force is increasing and MPSA is
also decreasing which is not desirable. Hence, by considering both parameters, i.e.,
MPSA and force experienced at articulated joint, L = 650 favors the most desirable
value of ‘L’.
5.1 Conclusion
which can provide higher MPSA value and lower value of the force experienced at
the articulated joint. Based on the results, it can be concluded that:
• The MPSA obtaining is independent of the value ‘θ ’ but the force experienced at
the articulated joint increases with increasing value of ‘θ ’.
• The value of MPSA obtained decreases with the increase in value of ‘d,’ whereas
there is only a slight change or no significant change on the force experienced at
the articulated joint with increase in value of ‘d.’
• The value of MPSA with varying ‘r’ follows the same pattern as that of the ‘d,’
i.e., the value of MPSA obtained decreases with increasing value of ‘r.’ However,
the force experienced at the articulated joint also decreases with the decreasing
value of ‘r’ which is desirable.
• With the varying value of ‘L,’ MPSA initially increases and reaches a maximum
value for certain value of ‘L,’ and thereafter, MPSA decreases with increasing
value of ‘L.’ However, force curve increases with increasing value of ‘L.’
• In the above-considered configuration, the MPSA value achieved is around 58°,
and force experienced at the articulated joint is minimum for L = 650, d = 585,
r = 165 and θ = 28°, respectively, the mentioned steering actuator parameter
values.
• The increase in MPSA is approximately 28% when compared with the existing
steering arrangement of the LHD machine.
• The results obtained from the dynamic analysis of the LHD vehicle need to be
validated analytically or experimentally.
References
Abstract The armament layout of every military helicopter is decided by the avia-
tion department based on a variety of factors. Amidst a variety of other constraints,
the structural behavior plays a vital role in deciding the layout on the helicopter
boom. It is necessary to avoid interaction of the rotor blade frequency with that of
the natural frequency of the weapon system layout. Vibration analysis was carried
out for various configurations, to determine the natural frequencies which are close
to that of the rotor blade and therefore avoid them during flight. The finite element
analysis followed by impact hammer tests was carried out to confirm the configura-
tions. A passive damping technique is suggested to be adopted for attenuation of the
vibrations due to resonance, if any.
1 Introduction
The dynamic behavior of any vehicle is a major design criterion of interest for
every structural designer. The vibrations of the structure have mostly led to failure
of the structure or its functionality during operation. A military helicopter is no
exception. Additionally, vibrations are even more critical owing to the frequency of
the rotor blades of the helicopter. The dynamics of the structure of the helicopter
is not sufficient enough to ensure the non-interaction of the structure with the rotor
blade frequency. The arrangement of weapons on the boom of the helicopter is also
a major concern, as it may exhibit a natural frequency close to that of the rotor blade.
The helicopter boom is a slender tubular structure attached to the helicopter body, on
to which the different weapon systems are mounted during flight. The helicopter’s
arsenal can be chosen from a wide variety of weapons including air-to-air missiles,
armed projectiles, anti-ground missiles, anti-tank missiles, anti-ship missiles, etc.
The functioning of the weapon’s detecting unit is largely dependent on the structural
dynamics of the helicopter. The visual output of an imaging device can be corrupted
due to vibrations of the boom during flight. It is therefore necessary to verify the
dynamic response of the helicopter for various weapon configurations on the boom.
Similar exercise may be trivial for an aircraft as it does not involve the frequency
of the rotor blade. The resonance behavior can be controlled using various damping
techniques, provided we have the liberty to change the design of the helicopter boom
or the helicopter itself.
The helicopter boom is a hollow tube with a length-to-outer diameter ratio (L/D) of
20 and outer diameter to thickness ratio (D/t) of 44. The boom is running through the
body of the helicopter and supported to the cockpit of the helicopter, and the distance
between the supports is 25% of the boom length. Additionally, two links connect
the boom to the external helicopter fuselage body on either side. The schematic
of the boom with the links and weapon stations is shown in Fig. 1. The helicopter
boom has four stations along its length, where the armament is attached, termed
as left-hand in-board (LH IB), left-hand out-board (LH OB), right-hand in-board
(RH IB), and right-hand out-board (RH OB). The helicopter arsenal includes air-
to-air missile (ATAM), anti-tank missile (ATM), projectiles with ammunition and
anti-radiation missiles (ARM). Each station can hold a launcher with two anti-tank
missile or a launcher with two ATAMs or an ARM or a round of armed projectiles.
The weapons can be arranged in the stations in any of the permissible configurations
as recommended by the aviation department. It is therefore necessary to determine
the natural frequency of all these permissible configurations, both analytically and
experimentally.
A finite element analysis (FEA) was carried out to determine the natural frequency
of various weapon configurations [1]. The weapon stations may hold empty ATM
launcher, launcher with 1 ATM, launcher with 2 ATMs, launcher with projectile or
empty projectile launcher in different combinations. A modal analysis was carried out
using a commercial FEA software (ANSYS 17), considering the armament at each
station as a lumped mass in a beam model of the helicopter boom. The attachment
points of the boom with the helicopter and that of the links with the helicopter are
idealized to be simply supported, as rotations of the boom and links are allowed at
these locations. A beam model of the boom with the links and supports is shown
in Fig. 2. The mass, center of gravity, and moment of inertias of different weapon
systems are given as input to the FE software. The results of modal analysis of
various configurations are given in Table 1. The first mode is of primary concern
as the higher modes are far away from the rotor blade frequency. Interestingly, few
of the configurations were found to have a natural frequency very close 20 Hz, the
frequency of the rotor blades. Figure 3 shows the mode shape of configuration 7,
which has empty in-boards, launcher with 2 ATMs at out-board LH and empty
projectile launcher at out-board RH. A harmonic analysis was carried out for a
frequency sweep of 0 25 Hz to verify resonance 20 Hz, on either side of the boom.
The response of the boom is captured at the origin of the boom as well as between
the armament points on either side of the origin. Figure 4 shows the response at the
origin and on the boom, respectively. It is seen that effect of resonance is prominent
on the stations between the armaments, while the origin does not experience high
amplitude vibrations. Impact hammer tests were carried out on various configurations
of weapons mounted on the helicopter boom, with the helicopter.
In order to verify the phenomenon of resonance, a captive flight trial of the heli-
copter with the weapon systems in configuration 7 was carried out. On-board triaxial
accelerometers were positioned on the port and star board sides to pick up the ampli-
tude during flight. The PSD data of the port side accelerometer in the vertical direction
captured a grams of 2.01 during the captive flight. Furthermore, the seekers of the
on-board missiles were not able to detect the target due to high grms levels. Con-
secutively, captive trials were also carried out in configuration 5 and the measured
maximum grms level was 0.65. Therefore, it is always advisable to fly with weapon
configurations with frequencies away 20 Hz.
6 Sensitivity Analysis
The easiest solution to avoid resonance during flight is to fly in configurations not
close 20 Hz. However, the weapon system comprises of various missiles and armed
projectiles and controlling the sequence of release at all times may not be feasible. It
is hence advisable to consider damping options, even if the helicopter boom enters
resonance with the main rotor blade. Therefore, the effect of structural damping was
studied by carrying out a damping analysis using finite element software. The effect
of amplitude of vibration on the LH side of the boom has been plotted in Fig. 5. It is
observed that a 0.8% structural damping causes an 80% reduction in the amplitude
of vibration. The damping can be achieved using passive or active techniques.
Stiffening the boom or the links to the helicopter shall help in altering the natural
frequency of the structure. However, owing to the constraints on the modifications
730 R. Gopikrishna et al.
of the existing helicopter design, this solution was not feasible to be carried out.
The alternate solution is to dampen the resonant vibrations. Viscoelastic materials
are commonly used as a passive damping solution for vibration and noise control
in structures. The structure is impregnated with layers of viscoelastic material to
achieve damping. The layer treatment enhances damping resulting in control of
excessive resonance vibrations. Layered damping may be constrained or free layer
damping. Mathematical models are available for characteristics of free, constrained,
and partial constrained layer damping [3, 4]. From experimental results, it was found
that constrained layer damping provides higher attenuation [5]. This technique can
be employed in the design of the helicopter boom and link rods, to dampen the high
amplitude vibrations, if any, during flight.
8 Conclusion
The configuration of armaments attached to the helicopter therefore plays a vital role
in the performance of the weapon system. It is necessary to consider the rotor blade
frequency in design of the helicopter structure as well as the attachment of weapon
systems to it. Passive damping techniques may also be considered during the design
of the helicopter structure.
Dynamic Analysis of Helicopter Boom with Different … 731
References
1. Heylen W, Lammens S, Saus P (1997) Model analysis theory and testing. Department of Mechan-
ical engineering, Ku Leuven
2. Avitable P (2018) Modal testing a practitioner’s guide. University of Massachusetts, Lowell
3. Khalfi B, Ross A (2013) Transient response of a plate with partial constrained viscoelastic layer
damping. Int J Mech Sci 68(1):304
4. Bangarubabu P, Kishore Kumar K, Krishna Y (2012) Damping effect of viscoelastic material on
sandwich beams. In: International conference on trends in industrial and mechanical engineering,
pp 171–173
5. Kishore Kumar K, Krishna Y, Bangarubabu P (2015) Damping in beam using viscoelastic layers.
J Mater: Des Appl 229(2):117–125
Characterization of Composites Made
with In-House Prepregs at Different
Curing Cycle
Abstract The main objective of this work is to improve the properties of 90°
carbon/epoxy laminates made with prepregs B-staged at different curing cycles.
The effect of prepreg curing at 80 °C for 15 min and 100 °C for 5 min on the
transverse modulus was analyzed for the epoxy system used. A series of tensile test
were conducted to study the strength of material. Simultaneous Differential scanning
calorimetry and Thermogravimetric analysis (SDT) is used to find the fiber volume
fraction of the laminate. Void volume fraction was calculated by using ASTM D2734.
1 Introduction
Composites are used in several industries because of its low weight and good mechan-
ical properties like high strength, high stiffness, no corrosion, and good aging. Apart
from that, optimizing weight and easy processing are another advantages of prepregs.
The major reason behind the selection of prepreg for any applications is its cost and
performance. Kulkarni [1] has worked in modifying the prepreg making machine
to improve the properties of the unidirectional prepreg laminate and in the estima-
tion of fracture toughness. The quality of the prepreg laminate has been checked by
performing the chemical and mechanical test by Ghiorse et al. [2]. It was reported that
the epoxy system used is highly sensitive to moisture, and the decrease in service
temperatures. Krishna Mohan et al. [3] have conducted a series of modification
and experimentation in the prepreg making machine to improve the quality of in-
house prepregs. Piyush et al. [4] have studied the effect of temperature on B-stage
2 Material Making
The prepreg making process involves matrix impregnation into the carbon fiber.
Diglycidyl ether of bisphenol-A (DGEBA) as epoxy resin (LY556), methyl tetra
hydrophthalic anhydride (MTPHA) as hardner, 2,4,5-tris [(dimethylamino)methyl]-
phenol (DMP-30) as accelerator and TC36s, 12 k grade carbon fiber are the
constituents of the prepreg. Few modifications on the in-house prepreg making
machine were done by Krishnamohan et al. [3], and prepreg sheets are made using
the same. Prepreg sheet can be seen in Fig. 1.
B-staging is introduced to improve the quality and property of the prepreg. B-stage
curing cycle has been chosen by considering drap and tacky as the parameters. So,
B-stage curing has been conducted at following curing cycles: (1) 80 °C for 15 min
and (2) 100 °C for 5 min. Temperature for B-staging is provided uniformly to the
prepreg sheets by the silicon thermal pads as shown in Fig. 2. Later, the prepreg is
stored in freezer which is maintained at −18 °C to preserve the properties of the
material.
The prepreg sheet was cut in a required dimensions, and hydraulic press machine is
used to make laminate. Laminate is made by adapting [3] the curing cycle of 85 °C
for 3 hr and followed by 140 °C for 12 hr under 0.7 MPa pressure as shown in Fig. 3.
Water jet machine is used, and tabs are applied to the specimen as shown in Fig. 4.
Tensile test of 90° specimens is performed according to ASTM 3039 [5] as shown
in Table 1.
3 Experimentation
Fiber and void volume fraction plays important role in mechanical behavior of lami-
nates. To find the volume fraction of the composite, Eqs. (1) and (2) are used, and
thermogravimetric analysis (TGA) ‘SDT Q600 apparatus is used by burning the
matrix for this purpose.
ρc
vf = wf × × 100 (1)
ρf
1
ρc = wf wm (2)
ρf
+ ρm
where vf is the total volume of fiber, ρ f and ρ m are the density of fiber and matrix,
ρ c is experimental density of composite, and wf and wm are weight fraction of the
fiber and matrix, respectively.
The composite manufactured with void content below 1% is treated as the well
fabricated composite [6]. To find void volume fraction, Eq. (3) is used, where ρ c is
theoretical density of the material, and ρ e is the experimental density of the material.
Weight versus temperature relation is obtained from TGA ‘SDT Q600 apparatus
and the same can be seen in Fig. 5. Fiber content and void volume are reported in
Table 2.
ρc − ρe
vv = × 100 (3)
ρc
from data provided by the machine. Elastic modulus and strength are obtained from
the stress–strain data as shown in Figs. 6 and 7.
Figures 6 and 7 shows the strength and modulus of the laminates made from different
B-staged prepreg material and the same are tabulated in Table 3. Three samples for
both the cases are tested to check the repeatability of the result.
Maximum and minimum value of the modulus of laminate with B-stage of
80 °C/15 min are found to be 24.75 GPa and 17.77 GPa, respectively, with very
low value of variation 3.59 GPa. Similarly, strength values of B-staged 80 °C/15 min
are observed in between 58.8 MPa and 39.4 MPa with variation of 10.97 MPa.
For B-staged 100 °C/5 min case, maximum and minimum modulus values are
observed to be 10.54 GPa and 9.39 GPa, respectively. Variation in modulus is 0.76
738 P. Sute et al.
a 80 oC/15min
100o C/5min
b
Fig. 5 Weight dropping versus temperature relation for a 80 °C/15 min and b100 °C/5 min
GPa, while minimum strength value of 16.95 MPa and maximum value of 36.56 MPa
is observed, with variation of 10.27 MPa.
From the work of Krishnamohan et al. [3] and Reddy et al. [7], the strength and
modulus of 90° carbon/epoxy laminate can be seen. However, the fiber content of
Characterization of Composites Made with In-House Prepregs … 739
Table 3 Strength and modulus values of 90° layup for different B-stages
Sample 80 °C/15 min 100 °C/5 min
Modulus Modulus
Strength (MPa) Modulus (GPa) Strength (MPa) Modulus (GPa)
1 40.2 17.77 36.56 9.105
2 58.8 24.75 21.45 9.39
3 39.4 19.77 16.95 10.54
Average 46.133 20.76 24.98 9.68
Standard deviation 10.97 3.59 10.27 0.76
these laminates was different. A tabular and graphical comparison of these values
with the current work has presented in Table 4 and shown in Fig. 8, respectively.
From Fig. 8, it can be seen that among all the cases, the laminate made from
80 °C/15 min B-staged prepreg has significantly good transverse strength and
modulus.
Table 4 Comparison of
Specimen Strength (MPa) Modulus (GPa)
strength and modulus of
laminate made of in-house 80 °C/15 min 46.133 ± 10.97 20.76 ± 3.59
prepreg with present literature 100 °C/5 min 24.98 ± 10.27 9.68 ± 0.76
Reddy et al. [7] 11.25 8.12
Krishnamohan et al. [3] 8.36 ± 1.64 –
740 P. Sute et al.
Using Zeiss optical microscope, the failure mechanism of the material is studied.
Failure mechanism is primarily due to the delamination and matrix breaking. A
typical image of failure is shown in Fig. 9.
5 Conclusions
The main objective of this work is to improve the properties of 90° carbon/epoxy lami-
nates made with in-house prepregs B-staged at two different curing cycles. The effect
of prepreg curing cycle on the transverse modulus of the laminate is investigated.
The main observation of this study is as follows:
(1) The fiber volume fraction of laminate was found to be 35.72% for 80 °C/15 min
case and 60.72% for 100 °C/min.
(2) The void volume fraction is between 0.86 and 0.92% which is considerably
good.
(3) B-stage curing of prepreg effects the mechanical properties of material. Among
all the investigated and literature reviewed cases of 90° layup, the laminate
made with 80 °C/15 min B-staged curing has better mechanical properties
with strength and modulus of 46.133 MPa and 20.76 GPa, respectively.
(4) The failure mechanism is primarily due to delamination and matrix cracking.
The qualitative and quantitative analysis was performed to reveal the goodness of
manufacturing process adopted.
Characterization of Composites Made with In-House Prepregs … 741
Fig. 8 Graphical comparison of laminate with present literature a Modulus and b Strength
742 P. Sute et al.
Fig. 9 90° failed specimen microscopy image a Front view and b Side view
References
Abstract This work deals with the problem of detection of isomorphism and inver-
sions among planar kinematic chains. To detect isomorphism and inversions a least
distance matrix (LDM) is formed using the values on the basis of the parameter.
Isomorphism is detected on the basis of chain string (CS) of least distance matrix
(LDM). An inversion is generated by grounding a different links in the kinematic
chain. Here, link string (LS) is generated for the different kinematic inversions of the
chains so that we can detect kinematic inversions with the help of link string (LS).
1 Introduction
The main objective of this paper is to propose and develop a general method for
detection of isomorphism and inversions among planar kinematic chains. Two planar
kinematic chains are said to be isomorphic to each other if their type of links, link
assortment and connectivity of links is similar. An inversion is created by grounding
a different link in the kinematic chain. However some of the inversions may be struc-
turally identical depending on the type of link fixed. The detection of distinct inver-
sions is as important as detection of distinct kinematic chains, since various distinct
inversions will produce so many structurally distinct mechanisms for use. Isomor-
phism in kinematic chains has been the subject of intensive research, but detection
of distinct inversions has received somewhat lesser attention. Isomorphism or topo-
logical equivalence of planar kinematic chains has been studied, by Zeng et al. [1]
uniquely represented kinematic chain by a graph, and a fast deterministic algorithm
called the Dividing and Matching Algorithm (DMA) is proposed. Rai [2] proposed
binary code method in which link labeling algorithm is used to detect isomorphism of
kinematic chains. This method is test for simple and multiple joints also in epicyclic
K. Dewangan (B)
Kalinga University, Naya Raipur 492101, Chhattisgarh, India
A. K. Shukla
Lakshmi Narain College of Technology, Bhopal 462021, Madhya Pradesh, India
gear trains for its efficiency. Kamesh et al. [3] presented graph theory to detect
isomorphism in planar and geared kinematic chains. Mruthyunjaya [4] started with
a multiple jointed binary chain and transformed them gradually in stages until all the
joints became simple joints. Rai and Panjabi [5] proposed work with different struc-
tural invariants, i.e., primary structural invariants for isomorphism identification and
secondary structural invariant provides power transmission and transmission effi-
ciency. Sun et al. [6] joint-joint matrix proposed to describe the kinematic chain,
which can uniquely represent the kinematic chain structure. The links and joints
information were extracted from joint-joint matrix and the link code and joint code
were use to represent the link and joint attributes isomorphism detection was done
by comparing links, joints and matrices. Sun et al. [7] proposed improved hamming
number method to detect isomorphism for multiple joints kinematic chains. Joint-
joint matrix is used to represent kinematic chain uniquely then improved hamming
number matrix is obtained by improved hamming number method. Joint and chain
hamming number and joint hamming string are used to determine isomorphism.
Shukla and sanyal [8] proposed a gradient concept used to identify isomorphism and
inversion of kinematic chains. Gradient analogy is used to prepare gradient matrix in
that first gradient chain string and second gradient chain string is formed is used for
identification of isomorphism and inversion. The closest any method has come so
far in terms of directness of approach is the contracted link adjacency matrix method
proposed by Hwang and Hwang [9].
2 Methodology
The method involves a parametric approach for detection of isomorphism and inver-
sions. The method is explained using Watt and Stephenson’s chain. A least distance
matrix (LDM) is formed using the values on the basis of the parameter.
The following terminologies have been taken from “prajapati et al.” [10]
(1) Node Value—A node value is defined as the ratio of the number of parameters
at particular node and the number nodes node at particular link.
Generalized equation for node value will be
(n − 1)
NV = (1)
n
n = number of node in each link (Fig. 1).
(2) Joint value (JV)—A joint value is defined as a sum of node connecting at a
particular joint in the planar kinematic chain (Fig. 2).
(3) Least distance matrix (LDM)—A least distance matrix is a square matrix and
it is formed on the basis of a sum of least distance between joint values of the
two links.
A Parametric Approach to Detect Isomorphism … 745
Fig. 1 Node values of all possible links, a Binary link, b Ternary link, c Quaternary link, d Pentnary
link, e Hex nary link
(4) Link value (LV)—A link value is the sum of the elements of each row of least
distance matrix (LDM). It will give a particular value and this value is termed
as link value (LV) for that particular link.
(5) Link string (LS)—A link string refers to elements of a particular row (or
column) of least distance matrix (LDM) taken in ascending order.
(6) Chain string (CS)—A chain string is the sum of all the link value (LV) of
a particular planar kinematic chain along with all the elements of the least
distance matrix will give the chain string (CS) for that particular planar
kinematic chain.
(7) Total chain value (TCV)—A total chain value is the sum of the all elements of
least distance matrix.
Watt chain is shown in Fig. 3 Link A & D are ternary link and link B, C, D & E
are binary link. Each node of binary link is assigned a node value 1/2 and for ternary
746 K. Dewangan and A. K. Shukla
link joint value 2/3 is assigned. Also quaternary link and pentnary link had assigned
the node values 3/4 and 4/5 respectively.
For simplicity of calculations, a non-fractional value is assigned to each joint
value on the basis of taking least common multiple of joint values by which can be
taken same as the denominator for all joint value for various links connectivity. A
least distance matrix (LDM) for Fig. 3 is given below:
As shown in the matrix, for Watt chain, ternary link A is connected to binary link
B and the node value is assigned “7”. For connectivity between link A and link C a
shortest distance is considered i.e., 7 + 6 = 13. Similarly for connectivity between
link A and link D the joint value will be “8” and so on. A 6*6 matrix will be formed
for a 6 link watt chain.
Link string (LS) for link A, B, C, D, E & F are—[0, 2 (7), 8, 2 (13)], [0, 6, 7,
13,14,20], [0, 6, 7, 13, 14, 20], [0, 2 (7), 8, 2 (13)], [0, 6, 7, 13, 14, 20] and [0, 6, 7,
13, 14, 20] respectively.
Watt chain as shown in Fig. 3 the link value (LV) for link A, B, C, D, E & F
are—48, 60, 60, 48, 60, and 60 respectively and chain string (CS) for Watt chain are
336, 2 [2 (6), 4 (7), 8, 6 (13), 4 (14), 2 (20)].
A Parametric Approach to Detect Isomorphism … 747
3 Isomorphism
As shown in the matrix, for Stephenson chain, ternary link A is connected to binary
link B and the node value is assigned “7”. For connectivity between link A and link
C a shortest distance is considered i.e., 7 + 7 = 14 & connectivity between link A
and link D a shortest distance is considered i.e., 7 + 6 = 13. Similarly connectivity
between link A and link E the joint value will be “7” and so on. A 6*6 matrix will
be formed for a 6 link Stephenson chain.
Link string (LS) for link A, B, C, D, E & F are—[0, 3(7), 13, 14], [0, 2(7), 3(14)],
[0, 3(7), 13, 14], [0, 6, 7, 13,2(14)], [0, 6, 7, 13,2(14)], [0, 2(7), 3(14)] respectively.
Stephenson chain as shown in Fig. 4 the link value (LV) for link A, B, C, D, E
& F are—48, 56, 48, 54, 54 and 56 respectively. Also the sum of all the link values
(LV’s) of a Stephenson chain will give the chain string (CS) is 316, 2[6, 6(7), 2(13),
6(14)].
Watt chain as shown in Fig. 3 the link value (LV) for link A, B, C, D, E & F
are—48, 60, 60, 48, 60, and 60 respectively and chain string (CS) for Watt chain are
336, 2[2(6), 4(7), 8, 6(13), 4(14), 2(20)]. Watt chain Fig. 3 and Stephenson chain
Fig. 4 have a different chain string (CS) hence they are non-isomorphic.
A least distance matrix (LDM) for Fig. 5 is given below:
Link value (LV) for link A, B, C, D, E, F, G, H & I are—176, 224, 224, 176,
176, 176, 224 & 176 respectively and chain string (CS) for eight link one degree of
freedom kinematic chain are 1600, 2[2(12), 4(14), 2(16), 4(26), 4(30), 2(32), 4(42),
2(44), 2(56)].
A least distance matrix (LDM) for Fig. 6 is given below:
A Parametric Approach to Detect Isomorphism … 749
Link value (LV) for link A, B, C, D, E, F, G, & H are: - 164, 164, 200, 200,
164, 164, 200 & 200 respectively and chain string (CS) for eight link one degree of
freedom kinematic chain are 1456, 2[2(12), 4(14), 4(16), 4(26), 8(30), 2(32), 4(44)].
Eight link one degree of freedom chain as show in Fig. 5 the link value (LV)
for link A, B, C, D, E, F, G, H & I are—176, 224, 224, 176, 176, 176, 224 & 176
respectively and chain string (CS) for eight link one degree of freedom kinematic
chain are 1600, 2[2(12), 4(14), 2(16), 4(26), 4(30), 2(32), 4(42), 2(44), 2(56)].
Eight link one degree of freedom chain Figs. 5 and 6 chain have different chain
string (CS) hence they are non isomorphic.
Ten link one degree of freedom kinematic chain is shown in Fig. 7. Link A, D, E,
H, I & J are ternary link and link B, C, F & G are binary link. Each node of binary
link is assigned a node value “1/2” and for ternary link joint value “2/3” is assigned.
Also quaternary link and pent nary link had assigned the node values 3/4 and 4/5
respectively.
Thus at a node joining a binary link and binary link,binary link and ternary link
and ternary link and ternary link a node value will be assigned as 60/60,70/60,80/60
(2/3, joint value of ternary link and 1/2, joint value of binary link) as shown in Fig. 7.
A least distance matrix (LDM) for Fig. 7 is given below:
750 K. Dewangan and A. K. Shukla
Chain string (CS) for Fig. 9 = 2[60, 6(70), 2(75), 2(80), 2(85), 2(130), 3(140),
145, 5(150), 3(155), 5(160), 2(200), 210, 3(220), 225, 5(230), 290].
Chain string (CS) for Fig. 10 = 2[60, 6(70), 2(75), 2(80), 2(85), 130, 135, 4(140),
145, 7(150), 2(155), 3(160), 170, 200, 205, 2(210), 4(220), 2(230), 240, 280].
A Parametric Approach to Detect Isomorphism … 753
Ten link one degree of freedom planar kinematic chain Figs. 9 and 10 have a
different chain string (CS) hence they are non isomorphic.
A least distance matrix (LDM) for Fig. 11 is given below:
Chain string (CS) for Fig. 11 is 2[8(70), 8(80), 7(140), 10(150), 9(160), 7(210),
5(220), 4(230), 3(240), 3(290), 2(300)].
Chain string (CS) for Fig. 12 is 2[8(70), 8(80), 7(140), 10(150), 9(160), 3(210),
9(220), 8(230), 3(290), (300)].
Chain string (CS) for Fig. 13 is 2[8(70), 2(75), 4(80), 2(85), 6(140), 2(145),
8(150), 2(155), 5(160), 1(170), 2(210), 10(220), 2(225), 5(230), 4(290), (295)].
Chain string (CS) for Fig. 14 is 2[8(70), 2(75), 4(80), 2(85), 6(140), 2(145),
9(150), 2(155), 5(160), 1(170), 6(210), 8(220), 4(230), 2(225), 4(290), (295)].
Twelve link one degree of freedom chain 11, 12, 13 and 14 chain have different
chain string (CS) hence they are non isomorphic. Twelve link one degree of freedom
planar kinematic chain Figs. 11, 12, 13 and 14 have a different chain string (CS)
hence they are non-isomorphic.
Nine link two degree of freedom chain is shown in Fig. 15. Link C, D, E & I
are ternary link and link A, B, F, G & H are binary link. Each node of binary link
756 K. Dewangan and A. K. Shukla
is assigned a node value “1/2” and for ternary link joint value “2/3” is assigned.
Also quaternary link and pent nary link had assigned the node values 3/4 and 4/5
respectively.
Thus at a node joining a binary link and binary link,binary link and ternary link
and ternary link and ternary link a node value will be assigned as 60/60,70/60,80/60
(2/3, joint value of ternary link and 1/2, joint value of binary link) as shown in Fig. 15.
A least distance matrix (LDM) for Fig. 15 is given below:
A Parametric Approach to Detect Isomorphism … 757
Link Value (LV) for link A, B, C, D, E, F, G, H & I are—1360, 1180, 1010, 1030,
1010,1180,1280,1280 and 1030 respectively and chain string (CS) for nine link two
degree of freedom kinematic chain are 10,360, 2[3(60), 4(70), 4(80), (120), 4(130),
8(150), 2(160), 2(190), 2(210), 4(220), 2(280)].
Nine link two degree of freedom chain is shown in Fig. 16. Link A, B, C & H
are ternary link and link D, E, F, G & I are binary link. Each node of binary link
is assigned a node value “1/2” and for ternary link joint value “2/3” is assigned.
Also quaternary link and pent nary link had assigned the node values 3/4 and 4/5
respectively.
For simplicity of calculations, a non fractional value is assigned to each joint value
on the basis of taking L.C.M. of joint values by which we can take same denominator
for all joint value for various links connectivity.
Thus at a node joining a binary link and binary link,binary link and ternary link
and ternary link and ternary link a node value will be assigned as 60/60,70/60,80/60
(2/3, joint value of ternary link and 1/2, joint value of binary link) as shown in Fig. 16.
A least distance matrix (LDM) for Fig. 16 is given:
758 K. Dewangan and A. K. Shukla
Nine link two degree of freedom chain as show in Fig. 16 the link value (LV)
for link A, B, C, D, E, F, G, H & I are—1030, 1020, 1070, 1360, 1180,1280, 1230,
990 and 1140 respectively and chain string (CS) for nine link two degree of freedom
kinematic chain are 10,300, 2[2(60), 6(70), 3(80), (120), 2(130), 4(140), 7(150),
(160), 2(190), (200), 3(210), 2(220), 2(280)].
Nine link two degree of freedom chain Figs. 15 and 16 chain have different chain
string (CS) hence they are non isomorphic.
4 Inversion
The process of fixing different links of a kinematic chain one at a time to produce
distinct mechanisms is called inversion. Row and column strings obtained from the
least distance matrix of a planar kinematic chain are useful to detect distinct inversions
of a planar kinematic chain. The total number of inversion that can be obtained from
a planar kinematic chain is equal to the number of the links in the planar kinematic
chain. However, some of the inversions may be identical. The fact that the inversions
are distinct or identical can be determined by comparing their link strings.
Link value (LV) and link string (LS) for Table 1 (Watt chain) (Table 2).
A comparative study of Watt chain & Stephenson chain shows that Watt chain
Table 3 has two inversions and Stephenson chain Table 4 has three inversions.
5 Results
The method is tested for six, eight, nine and ten links with simple joint. These results
are verified by many researchers. Our method also verifies the result.
The numbers of mechanisms are derived from the six link one degree of freedom
kinematic chains with simple joints shown Tables 3 and 4 is not isomorphic.
Eight link one degree of freedom kinematic chains with simple joints shown in
Table 5 (appendix) are not isomorphic.
Nine link two degree of freedom kinematic chains with simple joints shown in
Table 6 (appendix) are not isomorphic.
Ten link three degree of freedom kinematic chains with simple joints shown in
Table 7 (appendix) are not isomorphic.
760 K. Dewangan and A. K. Shukla
6 Conclusion
In this paper, a simple, efficient, and reliable method is proposed to identify isomor-
phism and inversions in planar kinematic chains. The method is simple and involves
less mathematical calculations. Further, it is directly applicable to kinematic chains
and does not require converting the kinematic chain to their graphs. This method
can be easily implemented whenever the number of links and degree of freedom
increases, as well as with some modification it can be suitable for n-number of link
and f- number of degree of freedom. Such method can also computerize and coding
can be developed to reduce time and acquire results faster.
Appendix
(continued)
761
Table 5 (continued)
762
Chain No Distinct link Kinematic chains Link string (LS) Chain string (CS) Distinct
inversion
4 (A) 2(14),16,26,28,30,32 2[12,6(14),3(16),2(26),4(28),6(30),32,40,2(42), 2(44)] 8
(B) 2(14),2(28),30,40,42
(C) 2(14),16,2(28),30,42
(D) 14,2(16),26,2(30),32
(E) 12,14,26,2(30),40,44
(F) 12,14,26,28,30,40,44
(G) 2(14),28,2(30),2(44)
(H) 14,2(16),28,2(30),42
5 (A,H) 14,2(16),28,2(30),42 2[12,6(14),3(16),2(26),4(28),6(30), 3(40),2(42),44] 4
(B,E) 2(14),16,26,28,30,40
(C,D) 12,14,26,28,30,40,42
(F,G) 14,2(16),28,2(30),42
(continued)
K. Dewangan and A. K. Shukla
Table 5 (continued)
Chain No Distinct link Kinematic chains Link string (LS) Chain string (CS) Distinct
inversion
7 (A,B,D,E) 2(14),16,28,2(30),44 2[8(14),2(16), 4(28),8(30), 6(44)] 2
(C,F,G,H) 2(14),28,2(30),2(44)
(continued)
763
Table 5 (continued)
764
Chain No Distinct link Kinematic chains Link string (LS) Chain string (CS) Distinct
inversion
10 (A) 2(15),2(17),2(27),31 2[2(12),4(14), 2(15),2(17), 2(26),2(27),3(28),30,31,2(32), 5
(B,H) 12,15,26,30,32,40,42 2(40),4(42),56]
(C,G) 12,14,27,28,2(42),56
(D,F) 2(14),17,26,28,32,42
(E) 2(14),2(28),31,2(40)
(continued)
K. Dewangan and A. K. Shukla
Table 5 (continued)
Chain No Distinct link Kinematic chains Link string (LS) Chain string (CS) Distinct
inversion
13 (A) 4(15),27,2(29) 2[12,4(14),4(15),16,26,27, 2(28),2(29), 9(30),3(42)] 7
(B) 12,15,26,3(30),42
(C) 12,14,27,28,30,2(42)
(D) 2(14),16,26,29,2(30)
(E) 2(14),16,29,2(30),42
(F,H) 14,15,28,3(30),42
(G) 14,15,28,4(30)
14 (A) 4(15),2(29),43 2[6(14),4(15), 7(28),2(29),4(30),4(42),43] 4
(B,F,G,H) 14,15,2(28),2(30),42
(C,E) 3(14),28,29,2(42)
(D) 2(14),4(28),43
A Parametric Approach to Detect Isomorphism …
(continued)
765
Table 5 (continued)
766
Chain No Distinct link Kinematic chains Link string (LS) Chain string (CS) Distinct
inversion
16 (A,D 4(15),2(27),30 2[2(12),8(15), 4(27),12(30), 2(42)] 3
(B,C,E,F) 12,15,27,3(30),42
(G,H) 2(15),5(30)
Total Inversion 71
K. Dewangan and A. K. Shukla
Table 6 Nine link two degree of freedom kinematic chains distinct link, link string, chain string and distinct inversions
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
1 (A) 2(60),2(130),2(210),2(280) 2[3(60), 4(70), 4(80), (120), (130), 5
(B,F) 60,70,120,2(150),190,2(220) 8(150), 2(160), 2(190), 2(210), 4(220),
(C,E) 70,2(80),130,2(150),160,190 2(280)]
(D,I) 70,2(80),130,2(150),160,210
(G,H) 60,70,130,2(150),2(200),280
(D) 2(70),140,2(150),220,2(280)
(E) 60,70,120,2(150),190,2(220)
(F) 2(60),2(130),200,2(210),280
(G) 60,70,120,140,150,200,210,280
(H) 2(70),80,130,140,150,160,190
(I) 2(70),2(140),2(150),200,220
(continued)
767
Table 6 (continued)
768
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
3 (A,C) 2(70),80,140,150,160,210,230 2[2(60), 6(70), 3(80), (120), 2(130), 5
(B,D) 2(70),140,2(150),2(220),280 2(140), 8(150), 2(160), 2(210), 4(220),
(E,I) 70,2(80),130,3(150),160 2(230), 2(280)]
(F,H) 60,70,120,2(150),2(220),230
(G) 2(60),2(130),2(210),2(280)
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
5 (A,D) 2(70),80,130,140,150,190,220 2[2(60), 6(70), 3(80), (120), 2(130), 5
(B,H) 2(70),140,2(150),200,2(220) 4(140), 6(150), 2(190), 2(200), 4(210),
(C,I) 70,2(80),140,2(150),2(210) 4(220)]
(E,G) 60,70,120,140,150,190,210,220
(F) 2(60),2(130),2(200),(230)
(F,H) 60,70,120,140,150,190,210,260
(G) 2(60),2(130),2(200),210,270
(I) 3(80),4(150),210
(continued)
769
Table 6 (continued)
770
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
7 (A) 2(70),80,2(140),150,2(210) 2[(60), 8(70), 2(80), 2(130), 8(140), 9
(B) 2(70),2(140),200,210,220 5(150), 4(200), (210), (220)]
(C) 2(70),80,130,140,150,200,210
(D) 60,70,130,140,150,2(200),210
(E) 60,70,130,2(140),200,2(210)
(F) 3(70),130,2(140),200,210
(G) 2(70),3(140),150,200,210
(H) 70,2(80),140,3(150),210
(I) 2(70),2(140),2(150),200,220
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
9 (A) 70,2(80),140,3(150),210 2[2(60), 6(70), 3(80), 4(130), 3(140), 9
(B) 70,2(80),130,2(150),160,210 7(150), (160), 4(200), 3(210), 3(220)]
(C) 2(70),80,130,140,150,200,210
(D) 2(70),140,2(150),200,2(220)
(E) 60,70,130,2(150),200,2(220)
(F), (I) 60,70,130,140,150,200,210,220
(G) 2(70),80,2(130),150,160,200
(H) 60,70,130,140,150,2(200),210
(continued)
771
Table 6 (continued)
772
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
11 (A,B) 70,2(80),130,2(150),160,210 2[2(60), 6(70), 3(80), 4(130), 3(140), 5
(C,F) 2(70),80,130,140,150,160,210 6(150), 2(160), 4(210), 4(220), 2(280)]
(D,E) 60,70,130,140,150,210,220,280
(G,H) 60,70,130,150,210,2(220),280
(I) 2(70),2(140),2(150),2(220)
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
13 (A,I) 2(70),2(140),150,210,220,280 2[(60), 8(70), 2(130), 6(140), 4(150), 8
(B) 3(70),2(140),210,220,270 (160), (200), 5(210), 3(220), (270),
(C) 2(70),3(140),150,200,210 2(280)]
(D) 2(70),80,130,140,160,2(210)
(E) 70,2(80),130,3(150),220
(F) 2(70),80,140,150,160,2(210)
(G) 60,70,130,140,2(210),2(280)
(H) 60,70,130,150,200,2(220),270
(E,H) 2(70),80,130,140,2(150),220
(F,G) 60,70,130,140,150, 2(220),210
(I) 2(70),2(140),2(150)2(220)
(continued)
773
Table 6 (continued)
774
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
15 (A,D) 2(70),80,130,140,2(150),200 2[(60), 8(70), 2(80), 2(130), 7(140), 5
(B,H) 2(70),2(140),2(150),200,220 8(150), 3(200), 2(210), 3(220)]
(C,I) 2(70),80,140,2(150),210
(E,F) 60,70,130,140,150,200,210,220
(G) 2(70),2(140),2(150),2(220)
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
17 (A,G) 70,2(80),130,3(150),160 2[2(60), 6(70), 3(80), 4(130), 3(140), 5
(B,H) 60,70,130,2(150),200,2(220) 8(150), 2(160), 2(200), 2(210), 3(220),
(C,I) 60,70,130,140,150,200,210,280 (280)]
(D,F) 2(70),80,130,140,150,160,210
(E) 2(70),2(140),2(150),2(200)
(E,G) 3(70),2(140),2(210),220
(F) 2(70),2(140),150,210,2(220)
(continued)
775
Table 6 (continued)
776
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
19 (A,C) 3(70),3(140),2(210) 2[10(70), (80), 13(140), 4(150), 4
(B) 2(70),4(140),2(210) 6(210), 2(220)]
(D,F,G,I) 2(70),3(140),150,210,220
(E,H) 2(70),80,2(140),2(150),210
(D) 70,75,140,4(150),200
(E) 4(75),135,2(145),195
(F) 60,70,120,140,150,195,2(220)
(G) 2(60),130,135,200,3(210)
(H) 60,75,120,3(150),190,220
(continued)
777
Table 6 (continued)
778
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
23 (A,D) 2(70),85,130,140,2(160),220 2[3(60), 4(70), 2(75), 2(85), (120), 6
(B,C) 60,70,130,140,155,2(230),290 2(130), (135), 3(140), 4(160), 2(220),
(E) 2(75),2(85),135,3(155) 6(230), 3(290)]
(F) 2(70),2(140),155,2(230),290
(G,I) 60,75,120,2(160),3(230)
(H) 2(60),135,2(220),3(290)
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
25 (A) 2(60),135„210,220,280,290,350 2[3(60), 3(70), 3(75),(80),(85), (120), 8
(B,G) 60,75,120,150,160,220,230,290 2(130), (135), (140), (145), 3(150),
(C) 3(75),85,135,145,155,215 (155), 2(160), (200), (210), (215),
(D) 70,75,140,3(150),200,210 3(220), 2(230), (280), 3(290), (350)]
(E) 2(70),80,130,145,2(220),280
(F) 70,80,85,130,150,2(160),220
(H) 60,70,130,140,215,2(290),350
(I) 60,70,130,155,200,2(230),290
(continued)
779
Table 6 (continued)
780
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
27 (A) 2(70),80,130,145,160,192,210 2[3(60), 3(70), 3(75),(80),(85), (120), 9
(B) 70,75,140,3(150),200,210 2(130), 2(135), (140), (145), 6(150),
(C) 3(75),85,2(135),145,195 (160), (190), (195), (200), 5(210),
(D) 60,75,130,2(150),2(210),270 (220), 2(270)]
(E) 60,70,135,160,2(210),220,270
(F) 70,80,85,130,2(150),160,210
(G) 60,75,120,2(150),160,190,210
(H) 2(60),130,135,200,2(210),270
(I) 60,70,120,140,150,195,220,270
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
29 (A) 3(75),85,2(135),145,155 2[2(60), 5(70), 3(75),(85),2(130), 9
(B) 2(70),85,130,140,2(160),210 2(135), 5(140), (145), 3(150), (155),
(C) 2(70),3(140),155,2(200) 2(160), 2(200), 5(210), (220), (280)]
(D) 3(70),130,140,145,210,220
(E) 60,70,135,2(140),2(210),280
(F) 60,75,130,2(150),160,200,210
(G) 60,70,135,140,3(210),280
(H) 60,75,130,2(150),200,210,220
(I) 70,75,2(140),2(150),160,210
(continued)
781
Table 6 (continued)
782
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
31 (A,F) 60,70,135,140,150,200,210,220 2[3(60), 3(70), 3(75),(80),(85), 3(130), 7
(B) 2(70),80,2(130),150,165,210 3(135), (140), 6(150), 2(160), (165),
(C) 70,80,85,130,2(150),2(160) 2(200), 5(210), 2(220)]
(D) 3(75),85,3(135),165
(E,I) 60,75,130,2(150),160,200,210
(G) 60,70,135,150,2(210),2(220)
(H) 60,75,130,2(150),3(210)
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
33 (A,I) 70,75,2(140),150,160,200,210 2[2(60), 5(70), 3(75),(85),3(130), 8
(B) 3(70),130,145,200,220,260 (135), 4(140), (145), 2(150), (155),
(C) 60,70,130,2(140),200,215,260 2(160), 5(200), 2(210), (215), (220),
(D) 60,70,130,140,155,3(200) (260), (270)]
(E) 2(70),85,2(130),2(160),200
(F) 60,70,135,140,2(210),200,260
(G) 60,75,130,2(150),200,225,260
(H) 3(75),85,135,145,155,215
(D,H) 60,70,135,140,150,200,2(210)
(E) 2(70),80,2(130),2(150),205
(F) 2(70),80,135,2(150),2(210)
(continued)
783
Table 6 (continued)
784
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
35 (A,B) 2(70),80,130,135,2(150),210 2[2(60), 4(70), 4(75),(80),2(130), 5
(C,G) 60,70,135,140,150,2(210),220 2(135), 2(140), 2(145), 10(150),
(D,F) 60,75,130,3(150),2(210) 6(210), (220)]
(E) 4(75),4(135)
(H,I) 70,75,140,4(150),210
(continued)
K. Dewangan and A. K. Shukla
Table 6 (continued)
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
37 (A,E,F,G,H,I) 60,75,135,3(150),2(210) 2[3(60), 8(75), 6(135), 13(150), 3
(B,D) 4(75),3(135),150 6(210)]
(C) 2(75),6(150)
(C,E) 3(75),90,3(135),165
(D) 60,75,135,2(150),3(210)
(continued)
785
Table 6 (continued)
786
Chain No. Distinct Link Kinematic chains Link String (LS) Chain String (CS) Distinct
Inversion
39 (A) 5(78),2(138),148 2[3(60), 2(70), 5(78), (120), (130), 7
(B,F) 70,78,2(140),3(156),216 2(138), (140), (148), 8(156), 5(216),
(C) 3(70),130,148,2(226),286 2(226), (276), (286)]
(D) 60,70,138,2(140),2(216),276
(E) 60,78,130,4(156),216
(G,I) 60,78,120,3(156),216,226
(H) 2(60),138,3(216),276,286
(continued)
787
Table 7 (continued)
788
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
5 (A,J) 2(60),120,130,190,210,270,280,340 2[4(60), 4(70),4(80),2(120), 4(130), 6(150), 5
(B,I) 60,70,120,2(150),180,220,230,280 2(160),(180),2(190), 4(210), 2(220), 2(230), 2(270), 4(280),
(C,H) 70,2(80),130,2(150),160,190,210 2(340)]
(D,G) 70,2(80),130,150,160,210,230,270
(E,F) 60,70,130,150,210,220,2(280),340
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
8 (A,H) 2(60),2(130),2(210),2(280),340 2[4(60), 4(70), 4(80),2(120),4(130), 4(140), 8(150), 2(160), 3
(B,F,G,I) 60,70,120,2(150),190,2(220),280 4(190), 4(210), 4(220), 4(280), (340)]
(C,D,E,J) 70,2(80),130,2(150),160,190,210
(continued)
Table 7 (continued)
790
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
12 (A) 2(70),140,2(150),200,3(220) 2[3(60), 6(70),3(80), (120),4(130),3(140), 10
(B) 70,2(80),140,2(150),3(210) 6(150),(160),2(190), 3(200), 5(210), 5(220), (230), 2(260)]
(C) 2(70),80,2(130),150,160,190,220
(D) 60,70,120,140,150,190,200,220,230
(E) 2(60),2(130),2(200),2(210),260
(F) 60,70,120,140,150,190,210,220,260
(G) 2(70),80,130,140,150,190,210,220
(H) 60,70,130,140,150,200,210,220,260
(I) 60,70,130,150,200,210,2(220),260
(J) 70,2(80),130,2(150),160,210,230
13 (A) 2(70),140,2(150),3(220),280 2[3(60), 6(70), 3(80),(120), 4(130), 3(140), 10
(B) 70,2(80),140,3(150),2(210) 7(150),(160),2(190), 3(200), 4(210), 4(220), 2(260),(270),
(C) 2(70),80,2(130),150,160,190,200 (280)]
(D) 60,70,130,140,150,200,210,220,260
(E) 60,70,130,140,150,190,210,220,270
(F) 2(70),80,130,140,150,200,210,270
(G) 60,70,120,140,150,190,200,220,270
(H) 2(60),2(130),200,2(210),260,280
(I) 60,70,120,2(150),190,2(220),260
(J) 70,2(80),130,2(150),160,190,210
14 (A,J) 2(70),140,2(150),2(220),2(280) 2[3(60), 6(70), 3(80),2(120),2(130), 2(140), 8(150), 2(160), 5
(B,I) 2(70),80,140,150,160,210,230,270 (180),2(190),2(210), 4(220),2(230), 2(270), 4(280)]
(C,H) 70,2(80),130,3(150),160,190
(D,G) 60,70,120,2(150),180,2(220),230
(E,F) 2(60),120,130,190,210,270,2(280)
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
15 (A) 2(60),2(130),200,2(210),2(280) 2[3(60), 6(70), 3(80),(120), 4(130), 3(140), 6(150), 2(160), 6
(B,I) 60,70,120,140,150,190,220,230,280 2(190),(200),4(210), 4(220), 2(230), 4(280)]
(C,H) 2(70),80,130,140,150,160,190,210
(D,G) 70,2(80),130,2(150),160,210,230
(E,F) 60,70,130,150,210,2(220),2(280)
(J) 2(70),2(140),2(150),200,2(220)
(G) 2(70),80,130,140,160,210,230,270
(H) 2(70),2(140),2(150),2(200),220
(I) 2(70),80,130,140,150,160,210,230
(J) 60,70,120,140,150,210,230,280,300
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
18 (A,I) 60,70,130,140,150,210,220,2(280) 2[3(60), 6(70),3(80), (120),4(130),3(140), 8(150), 2(160), 6
(B,H) 2(70),80,130,140,150,160,210,230 4(210), 4(220), 2(230), 5(280)]
(C,G) 70,2(80),130,3(150),160,210
(D,F) 60,70,120,2(150),2(220),230,280
(E) 2(60),2(130),2(210),3(280)
(J) 2(70),2(140),2(150),2(220),280
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
21 (A,H) 70,2(80),130,3(150),160,210 2[3(60), 6(70),3(80),6(130), 2(140), 8(150), 2(160), 5(200), 5
(B,I) 60,70.130,2(150),200,3(220) 4(210), 5(220), (280)]
(C,J) 60,70,130,140,150,200,210,220,280
(D,G) 2(70),80,2(130),150,160,200,210
(E,F) 60,70,130,140,150,2(200),210,220
(J) 2(70),2(140),2(150),200,2(220)
23 (A) 3(70),2(130),140,2(200),270 2[2(60), 8(70),2(80),4(130), 7(140), 5(150), 8(200), 4(210), 7
(B,H) 60,70,130,2(140),2(200),2(210) (220),(260), 3(270)]
(C,G) 60,70,130,140,150,2(200),210,260
(D,F) 2(70),80,130,140,150,200,210,270
(E) 2(70),2(140),150,2(200),220,270
(I) 70,2(80),140,3(150),2(210)
(J) 2(70),2(140),2(150),2(200),220
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
25 (A) 3(70),130,140,150,2(200),210 2[2(60), 8(70),2(80),3(130), 7(140), 7(150), 7(200), 7(210), 10
(B) 2(70),3(140),150,2(200),210 2(220), (280)]
(C) 2(70),80,2(140),150,3(210)
(D) 2(70),2(140),150,200,210,220,230
(E) 2(70),80,130,140,150,200,2(210)
(F) 60,2(70),140,2(150),2(200),210
(G) 60,70,130,2(140),2(200),2(210)
(H) 70,2(80),130,3(150),200,210
(I) 60,2(70),130,2(150),2(200),220
(J) 60,70,130,2(140),200,2(210),230
26 (A) 2(60),130,210,280,290,350,2(360) 2[2(60), 8(70),2(80), (120),(130),5(140), 6(150), 8
(B,I) 60,70,120,150,220,230,290,2(300) (160),(210), 3(220), 4(230),(280),3(290), 4(300),(350),
(C) 2(70),80,130,150,160,220,2(230) (360)]
(D) 70,2(80),140,4(150),210
(E) 2(70),80,140,150,160,2(230),290
(F,J) 2(70),2(140),150,230,2(300),360
(G) 3(70),2(140),220,2(290),350
(H) 2(70),2(140),2(150),2(220),280
(H) 60,70,130,2(140),200,2(210),260
(I) 3(70),130,140,200,220,270,290
(continued)
795
Table 7 (continued)
796
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
31 (A) 60,70,130,2(140),2(210),2(280) 2[2(60), 8(70), 2(80),4(130),7(140), 4(150), 10
(B) 60,70,130,140,150,200,220,230,280 (160),2(200),6(210), 3(220), (230),(270), 4(280)]
(C) 2(70),80,130,140,150,160,2(210)
(D) 70,2(80),130,3(150),210,220
(E) 60,70,130,150,200,2(220)270,280
(F) 60,70,130,140,2(210),3(280)
(G) 2(70),80,130,140,160,2(210),230
(H) 2(70),3(140),150,2(200),210
(I) 3(70),130,2(140),210,220,270
(J) 2(70),3(140),150,210,220,280
32 (A) 3(70),130,2(140),200,2(210) 2[2(60), 8(70),2(80),3(130), 7(140), 5(150), 7
(B,G) 2(70),3(140),150,2(200),210 (160),5(200),6(210), 2(220), 2(280)]
(C,F) 2(70),80,130,140,150,160,2(210)
(D,E) 60,70,130,140,150,200,210,220,280
(H) 70,2(80),130,3(150),200
(I) 60,70,130,2(150),2(200),2(220)
(J) 60,70,130,2(140),2(210),2(280)
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
34 (A,G,I,J) 60,70,130,140,150,200,210,2(220) 2[2(60), 8(70),2(80),4(130), 4(140), 8(150), 3(200), 4(210), 4
(B,F) 2(70),80,140,2(150),2(210),220 10(220)]
(C,E) 2(70),80,140,2(150),2(210),220
(D,H) 2(70),140,2(150),4(220)
(continued)
797
Table 7 (continued)
798
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
37 (A,D,E,H) 60,70,130,140,150,210,2(220),280 2[2(60), 8(70),2(80),4(130), 6(140), 8(150), 4(210), 9(220), 3
(B,C,F,G) 2(70),80,130,140,2(150),210,220 2(280)]
(I,J) 2(70),2(140),2(150),3(220)
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
40 (A) 2(70),4(140),200,2(210) 2[(60), 10(70),(80),2(130), 10(140), 4(150), 9
(B) 3(70),2(140),2(210),220,270 (200),7(210),6(220), (270), 2(280)]
(C,I) 2(70),2(140),150,210,2(220),280
(D) 2(70),80,140,2(150),2(210),220
(E) 2(70),80,130,140,2(150),210,220
(F) 60,70,130,140,150,200,2(220),270
(G) 60,70,130,2(140),2(210),2(280)
(H) 3(70),130,2(140),2(210),220
(J) 2(70),3(140),210,220,280
41 (A,J) 2(70),3(140),2(210),2(280) 2[(60), 10(70),(80),2(130), 10(140), 2(150), 2(200), 8(210), 5
(B,I) 3(70),2(140),2(210),220,270 3(220), 2(270), 4(280)]
(C,H) 2(70),3(140),150,200,210,220
(D,G) 3(70),130,140,150,2(210),220
(E,F) 60,70,130,140,200,210,270,2(280)
A Parametric Approach to Detect Isomorphism …
(continued)
799
Table 7 (continued)
800
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
44 (A) 3(70),130,2(140)200,2(210), 2[(60), 10(70),(80),2(130), 12(140), 4(150), 4(200), 8(210), 9
(B) 2(70),4(140),200,2(210) 2(220), (280)]
(C) 3(70),3(140),3(210)
(D,J) 2(70),80,130,140,2(150),200,210
(E) 60,70,130,140,150,2(200)210,220
(F) 60,70,130,2(140),200,2(210),280
(G) 2(70),3(140),150,210,220,280
(H) 2(70),80,2(140),2(150),2(210)
(I) 2(70),80,2(140),2(150),2(210)
45 (A,E) 2(70),4(140),2(210),280 2[12(70), 14(140), 12(210), 7(280)] 3
(B,D,F,H) 3(70),2(140),3(210),280
(C,G,I,J) 2(70),3(140),210,280
(continued)
801
Table 7 (continued)
802
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
51 (A) 2(60),130,215,2(290),300,2(350) 2[4(60), 4(70),2(75),2(85),(120),3(130), 2(135), 7
(B,J) 60,70,120,155,2(230),240,2(290) (140),(150), 2(155), 2(160),(170),2(200), (215),2(220)
(C) 2(70),85,130,2(160),170,2(220) 4(230), 2(240), 6(290),(300),2(350)]
(D) 2(75),2(85),2(135),2(155),215
(E,I) 60,75,130,150,160,200,2(230),290
(F,H) 60,70,135,140,200,220,2(290),350
(G) 2(70),85,2(130),170,2(240),300
(I) 60,75,120,2(150),160,180,210,220
(J) 70,75,140,2(150),160,200,2(210)
56 (A) 2(60),120,130,195,210,270,2(280) 2[4(60), 3(70),3(75), (80),(85), 2(120), 10
(B) 60,70,120,150,155,180,2(220),230 2(130),(135),(140),(145),6(150),(155),2(160), (180),(190),
(C) 70,80,85,130,2(150),2(160),190 (195),4(210),3(220), (230),(270), 3(280)]
(D) 2(70),80,130,145,150,210,220,280
(E) 60,70,135,140,150,210,220,270,280
(F) 60,75,130,2(150),160,210,230,270
(G) 3(75),85,2(135),145,155,195
(H) 60,75,120,2(150),160,180,220
(I) 2(60),120,135,190,2(210),270,280
(J) 70,75,140,3(150),220,280
(continued)
803
Table 7 (continued)
804
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
57 (A) 2(60),135,210,220,280,2(350) 2[4(60), 3(70),3(75), (80),(85), 2(120), 9
(B,J) 60,75,120,150,160,220,230,2(290) 2(130),(135),(140),(145),4(150),(155),3(160),
(C) 3(75),85,135,145,2(215) (200),(210),2(215), 3(220), 3(230), (280),5(290),2(350)]
(D) 70,75,140,2(150),160,200,210,230
(E) 2(70),80,130,145,150,2(220),280
(F) 60,70,120,140,150,2(290),350
(G) 2(60),2(130),200,215,2(290),350
(H) 60,70,120,150,155,3(230),290
(I) 70,80,85,150,3(160),220
58 (A,J) 2(60),120,135,210,220,280,350 2[4(60), 3(70),3(75), (80),(85), 2(120), 2(130), 8
(B,I) 60,75,120,2(150),220,230,290 2(135),(140),(145), 3(150),(155),3(160), (190),(195),
(C) 3(75),85,2(135),145,155,215 (200),2(210),(215), 4(220), 2(230), 2(280), 2(290), 2(350)]
(D) 70,75,140,150,160,200,2(210)
(E) 2(70),80,130,145,2(220),2(280)
(F) 60,70,130,140,215,2(290),2(350)
(G) 60,70,130,155,200,2(230),2(290)
(H) 70,80,85,130,3(160),2(220)
(H) 60,75,120,2(150),160,2(210),240
(continued)
805
Table 7 (continued)
806
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
63 (A) 2(60),120,130,195,200,3(270) 2[3(60), 5(70),3(75), (85),2(120), (130),(135), 9
(B) 60,70,120,140,155,180,210,2(230) 4(140),(145),3(150), 2(155), 3(160),(190),(195),(200),
(C) 2(70),85,130,140,3(160),190 3(210), 2(220), 3(230),(260),3(270), (280)]
(D) 2(70),3(140),155,200,230,260
(E) 3(70),140,145,210,220,270,280
(F,J) 70,75,140,2(150),160,210,230,270
(G) 3(75),85,135,145,2(155),195
(H) 60,75,120,2(150),160,180,220,230
(I) 2(60),120,135,190,2(210),260,280
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
66 (A) 2(60),130,135,200,2(210),260,280 2[3(60), 5(70),3(75), (85),(120), 3(130),(135),4(140), 9
(B) 60,75,120,2(150),160,220,230,290 (145),2(150),2(155), 3(160), 5(200), 2(210),(215),(220),
(C) 3(75),85,135,145,2(155),215 3(230),(260),(280), 2(290)]
(D,J) 70,75,140,2(150),200,210,230
(E) 3(70),130,145,200,220,280,290
(F) 60,70,130,2(140),200,215,260,290
(G) 60,70,130,140,155,3(200),230
(H) 2(70),85,2(130),3(160),200
(I) 60,70,120,140,155,200,2(230),290
(G) 70,75,2(140),2(150),160,2(210)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
69 (A) 60,70,120,140,155,210,2(230),280 2[3(60), 5(70),3(75), (85),(120), 3(130),(135),5(140), 10
(B) 2(70),85,130,140,3(160),220 (145),3(150),2(155), 3(160),(200),4(210), 2(220), 4(230),
(C) 2(70), 3(140), 155, 200, 2(230) 2(270), (280)]
(D) 3(70),130,140,150,210,220,270
(E) 60,70,130,2(140),210,220,270,280
(F) 60,75,130,2(150),160,210,2(230)
(G) 3(75),85,130,135,150,2(155)
(H) 60,75,120,2(150),160,210,220,230
(I) 2(60),130,135,200,2(210),2(270)
(J) 70,75,2(140),2(150),160,210,230
70 (A) 2(60),130,210,220,280,2(290),350 2[3(60), 5(70),3(75), (85),(120), 3(130), 5(140), 9
(B,I) 60,75,120,150,160,220,2(230),290 (145),2(150),2(155), 3(160),(200),(210),(215),3(220),
(C) 3(75),85,130,145,2(155),215 4(230), (280),4(290),(350)]
(D) 70,75,2(140),2(150),160,200,210
(E) 3(70),130,140,145,2(220),280
(F) 60,70,130,2(140),215,2(290),350
(G) 60,70,130,140,155,200,2(230),290
(H) 2(70),85,130,140,3(160),220
(J) 2(70),3(140),155,2(230),290
(G) 3(75),85,3(130),155,205
(H) 60,75,2(150),200,2(210),270
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
75 (A,J) 70,75,140,3(150),2(210),220 2[3(60), 4(70),4(75), (80),(120), 2(130), 2(135), 9
(B) 2(70),80,145,2(150),2(210),220 2(140),(145),9(150), (190),(195),2(200), 2(205), 8(210),
(C) 70(2),80,2(130),2(150),190,205 3(220)]
(D) 60,70,120,140,150,195,200,2(220)
(E) 2(60),130,135,200,4(210)
(F) 60,75,120,3(150),190,210,220
(G) 4(75),2(135),145,195,205
(H) 60,70,135,140,150,200,3(210)
(I) 60,75,130,3(150),200,2(210)
76 (A) 60,70,135,140,150,210,2(220),270 2[3(60), 4(70),4(75), (80),(120), 2(130), 2(135), 2(140), 10
(B) 2(70),80,130,145,2(150),210,220 2(145), 9(150),(190),(195),(200), 4(210), 5(220), 2(270)]
(C) 2(70),80,130,145,2(150),190,220
(D) 60,70,120,140,150,195,2(220),270
(E) 2(60),130,135,200,3(210),270
(F) 60,75,120,3(150),190,210,220
(G) 4(75),2(135),2(145),195
(H) 60,75,130,3(150),210,220,270
(I) 70,75,140,4(150),200,220
(J) 70,75,140,4(150),210,220
77 (A) 2(60),135,2(210),2(280),2(350) 2[3(60), 4(70),4(75), (80),(120), 2(130),(135),2(140), 6
(B,J) 60,75,120,2(150),2(220),2(290) 2(145), 7(150), 2(200), 2(210), 2(215), 4(220), 2(280),
(C) 4(75),130,2(145),2(215) 4(290), 2(350)]
(D,I) 70,75,140,4(150),200,210
(E,H) 2(70),80,130,145,150,2(220),280
(F,G) 60,70,130,140,200,215,2(290),350
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
78 (A) 60,70,135,140,150,2(210),2(220) 2[3(60), 4(70),4(75), (80),3(130),3(135), 8
(B) 2(70),80,130,145,2(150),2(210) 2(140),(145),10(150), 2(200),(205),9(210), 2(220)]
(C) 2(70),80,2(130),2(150),205,210
(D,H) 60,70,135,140,150,200,2(210),220
(E,I) 60,75,130,3(150),200,2(210)
(F) 4(75),3(135),145,205
(G) 60,75,130,3(150),3(210)
(J) 70,75,140,4(150),2(210)
79 (A,E) 3(70),130,145,190,2(220),260 2[2(60), 6(70),4(75), (120),(130),6(140), 2(145), 4(150), 5
(B,D) 60,70,120,2(140),190,215,2(260) 2(190), 4(200), 2(215), 4(220), 5(260), (275)]
(C) 2(60),2(130),4(200),275
(F,H,I,J) 70,75,2(140),2(150),200,220,260
(G) 4(75),2(145),2(215),275
A Parametric Approach to Detect Isomorphism …
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
81 (A) 3(70),130,145,200,2(220),270 2[2(60), 6(70),4(75),2(130),(135),6(140), 2(145), 5(150), 9
(B,J) 70,75,2(140),1(150),200,210,220 6(200), 2(210), 2(215), 4(220), (260),(270)]
(C) 4(75),135,2(145),2(215)
(D) 60,75,130,3(150),200,220,260
(E) 60,70,135,2(140),200,2(210),270
(F) 3(70),2(130),145,200,2(220)
(G) 60,70,130,2(140),3(200),210
(H) 60,70,130,2(140),2(200),210
(I) 70,75,2(140),3(150),200,220
82 (A) 3(70),140,145,2(210),2(220) 2[2(60), 6(70),4(75),2(130), 2(135), 6(140),(145),6(150), 7
(B,J) 70,75,140,3(150),3(210) 2(200),(205),10(210), (215), 2(220)]
(C) 4(75),2(135),145,205,215
(D,H) 60,75,130,3(150),200,210,220
(E,I) 60,70,135,2(140),4(210)
(F) 3(70),2(130),140,205,2(210)
(G) 2(70),4(140),2(200),205
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
84 (A,E) 60,70,135,2(140),3(210),270 2[2(60), 6(70),4(75),2(130), 2(135), 7(140), 2(145), 6(150), 6
(B,D) 60,75,130,3(150),200,210,220 2(200), 6(210),(215),4(220), (270)]
(C) 4(75),2(135),2(145),210
(F,H) 3(70),130,140,145,210,2(220)
(G) 2(70),4(140),2(200),210
(I,J) 70,75,2(140),3(150),210,220
(continued)
813
Table 7 (continued)
814
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
87 (A,E) 2(60),2(135),4(210),270 2[5(60), 6(75),(90), 2(120), 6(135), 6(150), 4(165), 12(210), 4
(B,D,F,H) 60,75,120,2(150),165,2(210) 2(240), (270)]
(C,G) 3(75),90,3(135),2(165)
(I,J) 60,75,135,2(150),4(210)
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
90 (A,E) 2(60),2(135),4(210),270 2[4(60), 8(75),2(120),4(135), 12(150), 4(195), 8(210), 4
(B,D,F,H) 60,75,120,3(150),195,210,270 3(270)]
(C,G) 4(75),2(135),150,2(195)
(I,J) 2(75),5(150),2(210)
(B,C,E,F,G,H,I,J) 60,75,135,3(150),3(210)
(continued)
815
Table 7 (continued)
816
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
93 (A,J) 2(60),120,138,2(216),226,2(276) 2[5(60), 2(70),4(78), (88),2(120),2(130), 6
(B,I) 60,78,120,3(156),166,2(216) 4(138),(140),6(156), 2(166), 2(200), 8(216), 2(226), 4(276)]
(C) 4(78),88,4(138)
(D,H) 60,78,130,3(156),200,2(216)
(E,G) 60,70,138,140,200,2(216),2(276)
(F) 2(70),88,2(130),2(166),226
(continued)
K. Dewangan and A. K. Shukla
Table 7 (continued)
Chain No Distinct Link Kinematic Chains Link String (LS) Chain String (CS) Distinct
Inversion
96 (A) 2(60),138,3(216),276,286,336 2[4(60), 3(70),5(78),2(120), (130),2(138),2(140), 8
(B,I) 60,78,120,3(156),216,226,276 (148),9(156),(190),(198),2(200), 5(216), 2(226),
(C) 5(78),2(138),148,198 3(276),(286), (336)]
(D) 60,78,120,4(156),190,216
(E) 2(60),130,138,2(200),2(216),276
(F) 60,70,120,2(140),198,2(276),336
(G) 3(70),130,148,190,2(226),286
(H,J) 70,78,140,4(156),200,216
References
1. Zeng K, Fan X, Dong M, Yang P (2014) A fast algorithm for kinematic chain isomorphism
identification based on dividing and matching vertices. Mech Mach Theory 72:25–38
2. Rai RK, Punjabi S (2019) A new algorithm of links labelling for the isomorphism detection of
various kinematic chains using binary code. Mech Mach Theory 131:1–32
3. Venkata Kamesh V, Mallikarjuna Rao K, Balaji Srinivasa Rao A (2017) An innovative approach
to detect isomorphism in planar and geared kinematic chains using graph theory. J Mech
Des 139(12)
4. Mruthyunjaya TS, Balasubramanian HR (1987) In quest of a reliable and efficient computa-
tional test for detection of isomorphism in kinematic chains. Mech Mach Theory 22(2):131–139
5. Rai RK, Punjabi S (2019) Isomorphism detection of planar kinematic chains with multiple
joints using information theory. J Mech Des 1–42
6. Sun W, Kong J, Sun L (2018) A joint–joint matrix representation of planar kinematic chains
with multiple joints and isomorphism identification. Adv Mech Eng 10(6):1687814018778404
7. Sun W, Kong J, Sun L (2017) The improved hamming number method to detect isomorphism
for kinematic chain with multiple joints. J Adv Mech Des Syst Manuf 11(5):JAMDSM0061-
JAMDSM0061
8. Shukla A, Sanyal S (2017) Gradient method for identification of isomorphism of planar
kinematic chains. Australian J Mech Eng 1–18
9. Hwang WM, Hwang YW (1992) Computer-aided structural synthesis of planar kinematic
chains with simple joints. Mech Mach Theory 27(2):189–199
10. Purushottamprajapati, Deshmukh PB, Shukla AK (2013) A novel approach for detection of
isomorphism of kinematic chains. Inacomm, December 2013, pp 525–532
Feasibility of Tensegrity-Based Walking
Robot
1 Introduction
Robots have been traditionally constructed with rigid links. Electrical and hydraulic
actuators are the traditional methods of actuation [1]. Designers of such robots have
traditionally struggled with the problem of reduction of mass of links while main-
taining structural integrity. In other hand, design of soft robots has been attempted
to address this issue. Tensegrity mechanisms have the potential to address these
drawbacks. Tensegrity mechanisms consist of tensile and compressive members as
discussed by Zhang [2], the tensile members are called strings or cables, and the
compressive members are called struts or bars. Change of shape of these mecha-
nisms can be achieved by changing the length of struts or changing the length of
cables. Change of length of a strut would require the strut to be replaced by a linear
actuator while change of length of a cable can be achieved by reeling in or out a cable
through a motor mounted at the end of a strut. Change of length in cables results
change in positions of nodes which can be used to generate a desired motion. It is
important to identify a sequence of cable actuations which create nodal displacement
such that the tensegrity mechanism undergoes walking locomotion. Tensegrity-based
locomotive robot was first introduced by Kanchanasaratool and Williamson [3, 4].
They provided a solution to the path tracking problems through a passive nonlinear
particle dynamics model. Jager and Skelton [5] have developed a nonlinear model and
linearized the nonlinear dynamics for a 2D tensegrity mechanism. But the method
fails for 3D tensegrity mechanisms because the nonlinearity cannot be handled by
their solution method. Aldrich et al. [6] have developed a dynamic model by consid-
ering the struts as fully rigid. But still, difficulty level is not reduced much due to
nonlinearity in dynamics of the mechanism.
To remove the above-mentioned difficulties, the proposed work considers the
walking process as quasi-static. A user-defined path has been created and divided
into small segments. Thereafter, a sequence of inputs (in terms of cable actuations)
has been found out so that the tensegrity mechanism travels through the end points
of each small segment step by step through quasi-static processes. This ensures that
the tensegrity mechanism will be in equilibrium at the end of each small segment in
following the path.
The sensitivity analysis is being conducted to have an estimation of how the actu-
ation(s) of a particular cable(s) affect the displacement of a particular ground node
keeping all other ground node’s positions unchanged. During actuation, the cables
are actuated up to 1.5% of unstretched length for both increase and decrease in step
sizes of 0.1% of the unstretched length. Walking in the forward direction without
turning requires the trajectory of the ground nodes to be only in the XZ plane. It is
seen that cable numbers 5, 6, 7, 8 and 10 are the most sensitive ones for movement
of node 6 in the XZ plane. Figure 2 shows the net magnitude of displacement of node
6 for actuation of the different cables. A similar analysis was done for other nodes.
Each step of the gait cycle is divided into two sub-steps. In the first sub-step, a
particular node is moved forward with a small upward movement (along Z-direction),
and in the second sub-step, it is moved forward with a small downward movement
Feasibility of Tensegrity-Based Walking Robot 823
so that the node will come to rest on the ground. A positive actuation value indicates
increase in length and negative means decrease in length. In the first sub-step, node
6 will be moved forward (along X-direction) with some upward movement (along
Z-direction) after giving the specified cable actuations as mentioned in Table 1.
In the second sub-step, the node 6 will be moving further in forward direction
and will rest on the ground after giving the cable actuations as mentioned in Table
2. The net translating motion after this step is 0.0406 unit in X-direction.
Similarly, the actuations of step-up and step-down are performed for nodes 8, 5
and 7. Tables 3, 4, 5, 6, 7 and 8 describe the actuation of nodes 8, 5 and 7, respectively.
From the above result, it is clear that the movement of all the nodes in X-direction
is same (within 2.5% range). At the end fourth step, the mechanism regains its
initial configuration but in a different position which is ahead of the initial position.
Thus, we can conclude that a gait cycle has been completed successfully, and the
mechanism has translated to a new position. These steps can be repeated to make
the mechanism follow a predefined path. Figures 4 and 5 show the plots of X, Y and
Z movements of each ground node vs normalized gait cycle time and gait cycle,
respectively. The figures clearly show that the movement of each node is very close
to the ideal movement. The small deviations seen in Y motion of node 5 and X motion
of node 7 do not create a major change in trajectory of the gait.
It was observed that none of gait cycles in tensegrities having bar 5, 6 and 8 are not
feasible because all of them led to significant unwanted movement of the stationary
ground nodes. We consider it to be a matter of great significance that a walking gait
Feasibility of Tensegrity-Based Walking Robot 825
3 Conclusion
This paper explores the feasibility of walking gait generation in simple tensegrity
mechanisms by cable actuation where the movement of node has been discretized into
small steps. Movement at every such step was assumed to be slow enough to avoid
826 P. K. Malik et al.
problems associated with vibration. It was found that a walking gait is possible only
in four-bar tensegrity. For all the other mechanisms—5, 6 and 8 bar—an attempt
to move one or more ground nodes led to significant unwanted movement. This
would lead to ground scraping in real life. Thus, one is forced to conclude that high
connectivity between nodes in complex tensegrity mechanisms makes walking gait
infeasible. However, a study involving simultaneous actuation of struts and cables is
necessary to prove or disprove this hypothesis, and such a study will be taken up in
future.
References
Abstract Harvesting fruits or nuts from tall trees have been a great challenge in
recent times. There is a shortage of skilled manpower required to harvest, and also, the
risk attached to this task is high (Punchihewa PG, Arancon RN (2006) Coconut: post-
harvest operations. Asian and Pacific Coconut Community). There is the absence of
standard devices in the market for climb-free harvesting from tall trees likes coconut
or areca nut. In this paper, a deployable mechanism concept capable of reaching tall
tree heights is discussed. This design was achieved after comparing and evaluating
continuous cable-type and discrete cable-type telescopic mechanisms. In addition
to the telescopic mechanism design, a novel support system which can follow the
morphology of the tree was also designed to avoid buckling of the telescopic struc-
tures. The overall mechanism was designed and prototyped as an efficient platform
to enable accessibility to tall tree heights to which an end effector may be attached.
1 Introduction
India is one of the leading producers of coconut [1] in the world producing around 13
billion nuts per annum [2]. The leading coconut producer countries are Philippines,
Indonesia, India, Sri Lanka, Thailand and Malaysia. Coconut trade is a major source
of export revenues for these countries [3]. The livelihood of about one crore people
in India is dependent on coconut [4].
The major problem with harvesting from tall trees (e.g. coconut trees) is that one
cannot carry safety equipment while climbing the tree as there is no such provision
2 Design Concept
The design consists of a climbing mechanism and an end effector. The end effector
depends upon the type and location of the fruit to be harvested. The paper describes
climbing mechanism only which consists of two parts—the telescopic mechanism
and a support system to prevent self-buckling [12] (Fig. 1).
Node 1
To
Winch Node 2
Fig. 2 Continuous
rigging-type telescopic
mechanism (a schematic of
its functionality, not the
actual construction of the
mechanism)
The major problem with this mechanism is that the wire or cable may lose contact
with the pulley when it retracts back at very high speed because of inertia effects.
To overcome such situation, separate cables for lifting and retracting can be used,
but this makes the system even more complex. One major drawback of this system
is that slots are needed to be cut out on the concentric tubes which will in turn affect
the stiffness or strength of the concentric tubes.
Static Analysis of Continuous Rigging-type Telescopic Mechanism. The different
concentric tubes can be termed as stages, viz. first stage, second stage and third
stage, respectively. Considering S 1 , S 2 and S 3 as the weights of the different stages.
The first stage with weight S 1 will be fixed to the ground. The cable is passed over to
830 B. Singh et al.
the pulleys present at various nodes, namely Node 1, Node 2 and Node 3, respectively.
As the cable is going to support the weight to be lifted W, Stage S 2 and Stage S 3 , the
tension in the whole cable would be the summation of W, S 2 and S 3 .
T = W + S2 + S3 (1)
S2 ẍ (W + S3 )
T = W + S2 + S3 + ( )+ ẍ (2)
g 2 g
T2
T2
S2
Node 1
T1
Node 2
T1
̈
S1
Design and Development of a Climb-Free Telescopic Mechanism … 831
Static Analysis of Cascade Rigging-type Telescopic Mechanism. Just like in the case
of continuous rigging-type telescopic mechanism, it has three stages, namely Stage
1, Stage 2 and Stage 3, respectively. Likewise, it may have more number of stages
depending upon the degree to which system must be made compact. Since individual
cables are used, every cable has its own tension, namely T 1 and T 2 depending upon
its static configuration.
T1 = 2(W + S3 ) + S2 (3)
T2 = W + S3 (4)
4(W + S3 ) S2
T1 = 2(W + S3 ) + (ẍ) + S2 + ẍ (5)
g g
(W + S3 ) 2(W + S3 )
T2 = (W + S3 ) + (2 ẍ) = (W + S3 ) + (ẍ) (6)
g g
These tensions are more than those in case of static tensions and are dependent
on the value of acceleration. So, the acceleration either has to be avoided or limited
against the strength of the cable.
x
Relative Displacement in between Stages = (8)
n−1
Time for Deployment. Due to greater amplification factor, the cascade rigging-type
telescopic mechanism will take lesser time to deploy for a particular speed of winch.
If τ is the time taken by the continuous rigging-type telescopic mechanism to deploy,
then cascade rigging-type telescopic mechanism would take the time which is the
time in case of continuous rigging-type telescopic mechanism divided by number of
x
working stages, i.e. n−1 .
Tensions. In continuous rigging-type telescopic mechanism, there is only one tension
which is uniform throughout the span of the cable. On the other hand, every
discrete cable has different tensions in case of cascade rigging-type telescopic mecha-
nism. The thing which is noticeable in the equations of tensions is that at an expense of
more displacement amplification, the tension T 1 seems much more than the tension T
in case of continuous rigging-type telescopic mechanism and tension T 2 is smaller in
magnitude than T. It can also be noticed that different cables have different tensions,
and different cables of different strengths can be used so as to optimize the material
requirement of the mechanism.
Tensions (Considering Pseudo Forces). Even though the fact that the system is safe
against static forces, still system can fail because pseudo-forces come into play when
mechanism is deployed with acceleration. So, acceleration either has to be avoided
or limited against the strength of the cable. The tensions considering pseudo-forces
are as under.
S2 ẍ (W + S3 )
T = W + S2 + S3 + + ẍ (11)
g 2 g
4(W + S3 ) S2
T1 = 2(W + S3 ) + (ẍ) + S2 + ẍ (12)
g g
(W + S3 ) 2(W + S3 )
T2 = (W + S3 ) + (2 ẍ) = (W + S3 ) + (ẍ) (13)
g g
Design and Development of a Climb-Free Telescopic Mechanism … 833
Strength of the Mast. The strength of the concentric masts is lesser in case of
continuous rigging-type telescopic mechanism because slots have to be cut out of
the masts so as to accommodate the continuous cable. Slots have adverse effect on
the stiffness of the masts. Moreover, slots reduce fatigue strength as well as they
introduce discontinuity in material. The strength of the concentric masts in case of
cascade rigging-type telescopic mechanism is more because no slots need to be cut
for cables.
From the equations, it is evident that there is a tradeoff between time of deployment
and tensions in cable. If the mechanism has to deploy faster for a particular winch
speed, then the suitable mechanism is cascade rigging-type telescopic mechanism.
But increased time for deployment will be on the expense of generation of higher
tension in the cable. Likewise, opting for continuous rigging-type telescopic mech-
anism increases the time for deployment for a given winch speed but the tension in
the cable will be lesser in magnitude.
Considering that there will be a tradeoff between time for deployment and tensions
in cable, cascade rigging-type telescopic mechanism was selected as the mechanism
to go up the tall tree, as the mechanism is supposed to be speedy while going up
for a particular winch speed. However, the mechanism should be designed carefully
because cables have to bear more tension and the slot cutouts to accommodate the
cables also have to be considered.
A very long and tall structure tends to buckle under its own weight whenever its
height is more than a certain threshold. The mechanism, when erected, may buckle
under its own weight plus the weight that has been supported over it. Therefore, to
avoid buckling, the telescopic mechanism can be provided a support of the tree.
The mechanism consists of linkage systems (L and L’) and hinges (H and H’)
to adjust according to the morphology of the tree. The support part is closed with
the help of chain C which enables the support to adjust according to the diameter of
the tree without opening up completely and leaving the contact with the tree. The
torsion springs S and S’ make sure that there is positive contact with the tree trunk.
The conceptual design and the top view of the mechanism are as shown in Fig. 4.
The mechanism can compensate for varying radial distance from tree, tree shape and
trunks going sideways as shown in Fig. 5.
834 B. Singh et al.
a. Radial Adjustment
5 Testing of Prototype
A model coconut tree was built using PVC pipes. The telescopic mechanism was
prototyped using pipes and pulleys. A weigh balance was used to measure the tension
in the strings. The pulleys were configured as per the type of mechanism to be tested.
The tests were conducted on the prototype with the following parameters (Table 1;
Fig. 6).
During the tests conducted, the theoretical and practical results were compared
(Table 2).
It was observed that cascade rigging-type telescopic mechanism is faster to
deploy to its complete length when compared to continuous rigging-type telescopic
mechanism.
6 Conclusion
In this paper, we have discussed the types of telescopic mechanism which are contin-
uous rigging-type and cascade rigging-type mechanism. From the tests, we came to
know that there is an inverse relation between speed of deployment and tension in the
cable. If we have to go for the speed, then the cable will have to bear more tension.
Likewise, with less speed of deployment, the tension in the cable would be lesser.
The criterion for selection of mechanism was speed. Therefore, we opted for cascade
rigging-type telescopic mechanism. Additionally, the concept of the support system
has been described which can adjust according to the shape and trajectory of the tree.
This simple, cost-effective and easily transportable system can form the platform for
harvesting system from tall trees. A suitable end-effector can be attached depending
on the type of fruit to be harvested.
Acknowledgements The research work was supported by SERB (a statutory body of the Depart-
ment of Science and Technology, Government of India) through grant EEQ/2017/000598, for
which the authors are immensely grateful. The support of CSIR-CSIO for providing laboratory
and prototyping facilities is also deeply acknowledged.
References
1. Punchihewa PG, Arancon RN (2006) Coconut: post-harvest operations. Asian and Pacific
Coconut Community
2. Coconut Development Board. https://coconutboard.nic.in. Last accessed 19 Jul 2019
3. Burton J The world leaders in coconut production. https://www.worldatlas.com/articles/the-
world-leaders-in-coconut-production.html
4. Singh RM India leading in coconut production and productivity in the world. https://pib.nic.
in/newsite/PrintRelease.aspx?relid=149492
5. Megalingam RK et al 2017 IOP Conf Ser: Mater Sci Eng 225012201
6. Tilbert G (2002) Deployable tensegrity structures for space applications
7. Lee D, Chang D, Shin Y, Son D, Kim T, Lee K, Kim J (2011) Design and application of
a wire-driven bidirectional telescopic mechanism for workspace expansion with a focus on
shipbuilding tasks. Adv Robot 25:699–715
8. Leavy W, Griffin C (1979) Antenna deployment mechanism for use with a spacecraft
Design and Development of a Climb-Free Telescopic Mechanism … 837
Abstract The safety and comfort of the locomotives depend on the basic suspen-
sion design against the dynamic load, to meet the desired dynamic characteristics.
The passenger’s comfort and goods safety in a locomotive are correlated with the
performance of suspension system subjected to external excitation. In the present
paper, a complex 37 degrees-of-freedom Integral Coach Factory (ICF) locomotive
model is developed using discretized lumped mass system approach. The equations
of motion are derived using Newton’s second law of motion. The scaled down simu-
lation model is developed using multi-body dynamics software ADAMS. The first
nine natural frequencies observed in simulation model are in good correlation with
the analytical results.
1 Introduction
The safety and comfort of the locomotives depend on the basic suspension
design against the dynamic load, to meet the desired dynamic characteristics. The
passenger’s comfort and goods safety in a locomotive are correlated with the perfor-
mance of suspension system subjected to external excitation. Wickens [1] discussed
the lateral stability of railway vehicles and compared with the experimental work.
Chang et al. [2] studied the behavior of the vertical irregularities for a six-axle loco-
motive. For the first time, a system with higher degrees-of-freedom is considered and
is used for determining different design parameters and dynamic response. Nishimura
et al. [3] discussed the suspension design for high-speed railway vehicles for a 10
degrees-of-freedom of system. The design strategy includes the development of a
lumped parameter vehicle model, formulating the design objective and to achieve
optimized key suspension parameters. Dikmen et al. [4] examined the vibrational
characteristics of a rail vehicle considering a 19 degrees-of-freedom system. Critical
speeds of the system are determined for yawing and pitching motions. Sharma [5]
formulated a 37 degree-of-freedom system using Lagrangian dynamics to evaluate
the influence of rail parameters on vertical and lateral ride behavior. Graa et al. [6]
investigated the effects of vehicle speed and rail irregularity on ride comfort through
numerical simulation. A seventeen degrees-of-freedom of system is developed using
Lagrangian dynamics. A parametric study is carried out in order to improve the ride
characteristics such as sterling index.
In the present paper, a complex coupled 37 degrees-of-freedom Integral Coach
Factory (ICF) locomotive model is developed using discretized lumped mass system
approach. The equations of motion are derived using Newton’s second law of motion.
The geometrical parameters for solid modeling of the components are obtained from
field visit to Carriage Repair Workshop Hubballi, Karnataka. The overall dimen-
sions of the coach are obtained from the open source. The scaled down solid model
is developed using Solid Edge software. The simulation of the model is carried out
in the ADAMS environment. The responses and natural frequencies are extracted
for the coupled system and being reported for the first time. The rail car parame-
ters taken into account for analysis are car body, bolster, bogie frame, primary and
secondary suspension, and wheel–axle sets. The effects of disturbances/excitation
and the relative motion of various parts on the model are studied. The first nine
natural frequencies observed in simulation model are in good correlation with the
analytical results.
2 Simulation Modeling
Thirty-seven degrees-of-freedom rail car body is considered for study. Scaled down
ICF coach system is modeled using Adams software. All masses are rigid in the
model. Realistic values are incorporated for masses of each body and the values
of stiffnesses and damping coefficients along different directions as listed in the
analytical model. The model is constructed with following assumptions:
• Friction between moving parts is neglected
• All masses are rigid and the system is symmetric about the longitudinal plane
• The wheel does not lose contact with the track
• Base excitations are considered as sinusoidal displacement–time functions that
reflect the irregularities of the track profile.
It includes various types of translational and rotational motions like pitch, roll,
and yaw for individual masses in the model. The track is made as ground and the
profile of the track is modeled in terms of individual motions about three axes (x, y
and z) given to the four wheel–axle assemblies as listed in Table 1. The motions are
Simulation Modeling of 37 Degrees-of-Freedom ICF Coach 841
defined as sinusoidal input. It is assumed that the wheel–axles are undamped with
respect to the track at all 8 points of contacts.
All masses have restricted translational motion along x-axis and allowed to freely
vibrate along y-axis and z-axis. There are 8 spring-dampers along vertical direction
(y-axis) and 8 spring-dampers along lateral direction (z-axis) connecting the 4 wheel–
axles to the 2 bogie frames. Each bogie frame is connected to a bolster via 2 spring-
dampers along vertical direction. There are 4 spring-dampers linking the 2 bolsters to
the car body along vertical direction and 4 spring-dampers along the lateral direction
linking the 2 bolsters to the car body.
The four wheel–axle assemblies constitute 16 degrees-of-freedom (4 vertical, 4
lateral, 4 yaw, and 4 roll). The 2 bogie frames account for 10 degrees-of-freedom (2
vertical, 2 lateral, 2 pitch, 2 roll, and 2 yaw). The 2 bolsters constitute 6 degrees-of-
freedom (2 vertical, 2 lateral, and 2 roll). The car body is having 5 degrees-of-freedom
(1 vertical, 1 lateral, 1 pitch, 1 roll, and 1 yaw). The total degrees-of-freedom for
the lumped mass system adds up to 37. The parameters of the simulation model
constructed are in reference to Sharma [5]. The model is shown in Fig. 1.
y y
z
x
Fig. 1 37 degrees-of-freedom rail car body side view and front view
842 B. B. Mahadikar et al.
3 Equations of Motion
The equations of motion for all 37 degree-of-freedom are derived using Newton’s
second law of motion considering the coupled motion for the first time. The equations
of motion are reduced to matrix form as shown in Eq. (1).
where [M], [K ], [C] are the matrices extracted from the 37 equations. Using Eq. (1),
natural frequencies and response of the system are evaluated.
4 Natural Frequencies
Fig. 2 Response curves plotted for vertical and lateral vibration of bogie frames
The geometrical and physical parameters considered for the system are the realistic
values for an ICF coach. To understand the behavior of car body as influenced by the
excitation at wheel–axles, the simulation is conducted. In the ADAMS interface, the
‘Run an Interactive Simulation’ option is selected under the ‘Simulation’ tab. The
‘End Time’ and the ‘Step Size’ are now set in the ‘Simulation Control’ pop-up dialog
box. These parameters are set for 20 s with a step size of 0.01. The responses are
extracted and are plotted in time domain using the ‘Plot output’ option.
Vertical and lateral responses are obtained for bogie frames as shown in Fig. 2.
The response curves for car body in lateral and rolling modes are shown in Fig. 3.
The response curves for car body in vertical and pitching modes are shown in
Fig. 4.
It is observed from the graphs that the responses are uniform and periodic for
the given set of kinematic parameters of the system. It is also evident from the plots
of the car body that the amplitude of vibration is reduced from that of the wheels.
Along the vertical direction, the amplitude of car body is around 3 mm and along
lateral direction it lies within 1 mm. The complex models considered in Sect. 4.8
considering higher degrees-of-freedom help in understanding of the behavior of
each part subjected to vibration along different directions. The dynamic analysis of
simulation reports shows that the vibration of the car body housing the passengers
lies within the comfort range.
6 Conclusion
natural frequency and the responses of the system. The responses obtained are in
good correlation with the mathematical modeling equations. The natural frequencies
and responses of the system are also in good correlation with both the simulation
and mathematical model for the basic model.
References
Abstract Polyamide, with its low friction coefficient, high heat resistance and good
moldability, is a promising polymer to meet the rising demand for lightweight,
durable gears. This paper considered injection-molded polyamide 66-polyamide 66
gear pair wear performance under unlubricated and wet lubricated condition. The
experiment was conducted using in-house developed power absorption gear test
rig. The torque of 1.8 Nm, rotational speed 800 rev/s and lubricant SAE 75W85
was considered during testing. During the test, gear tooth temperature under unlu-
bricated condition and lubricant temperature under wet lubricant condition was
measured and monitored. The net surface temperature on the gear surface and lubri-
cant was observed as 370 K and 303.4 K, respectively. The transmission efficiency
was increased by 5% compared to the unlubricated condition. Testing confirmed
that the gear exhibited surface wear at both the face and flank region in unlubricated
condition. While testing under wet lubricant condition, the wear was observed on the
face of the gear. Simulation result confirmed that the deflection of test gear is signif-
icantly higher in double tooth contact region shared by the face side as compared to
flank side of the driven (test) gear. That could be the possible reason for test tooth
wear in the face region.
1 Introduction
Polyamide gears are being extensively used for light-duty engineering applications
such as robotics, automated teller machine, food processing industry and office equip-
ments. Polymer material exhibits inferior wear and poor heat resistance, which affects
the transmission performance of gears. Usage of reinforcement would reduce the
amount of frictional and hysteresis heating [1, 8, 14]. Multiple works have been
carried out to understand the behavior of polymer–polymer and metal–polymer
gear pair under unlubricated condition. In polymer–polymer gear pair contact, gears
fail due to local thermal softening. Extensive investigation has been carried out on
metal–metal gear pair under wet lubricated condition. Hence, in the present research,
unreinforced polyamide 66 spur gear was run under the wet lubricated condition.
Breeds et al. [2] evaluated the wear characteristics of polyamide–polyamide and
acetal–acetal gear pair. Result revealed that gear exhibited wear failure at low loading
condition. In contrast, at higher loading, gear exhibited failure due to thermal failure.
Pogačnik and Tavčar [15] adopted an accelerated testing method on polyoxymethy-
lene (POM) and polyamide 6 gear pair. Gear failed mainly due to fatigue and sudden
melting. Hu and Mao [6] examined the misalignment effect on the acetal–acetal gear
pair under unlubricated condition. Mertnes and Senthilvelan [13] have attempted
to improve polymer gear performance by supplying compressed air on the contact
surface. Results confirmed that air cooling reduced the temperature and improved
the wear resistance. Bushimata et al. [3] investigated the wear characteristics of
unreinforced PA 66, polyacetal, PEEK and glass fiber-reinforced polyamide, glass-
reinforced polyacetal and glass-reinforced PPS gears. Results confirmed that when
similar material gear pair meshed, unreinforced PA 66-PA 66 gear pair exhibited
highest wear. In contrast, acetal and PA/PEEK gear pair exhibited the lowest level
of wear in comparison with other dissimilar material gear pair, due to less adhe-
sion. Results suggested that wear failure may be avoided by using an appropriate
combination of gear pair.
Senthilvelan and Gnanamoorthy [18] investigated the performance of the
polyamide 66 carbon fiber reinforced and polyamide 66 gear. Increase in gear tooth
temperature worsens the gear stiffness during testing which causes lower transmis-
sion efficiency. Dearn et al. [4] evaluated the efficiency of polyether ether ketone
(PEEK)–polyether ether ketone gear pair, and polyether ether ketone–steal gear pair
lubricated with automotive engine oil. Results revealed that PEEK–PEEK gear pair
exhibited higher efficiency compared PEEK–steel gear pair may be due to large tooth
deformation occurred on the PEEK gear. Walton et al. [20] examined the behavior
of the gear pair’s efficiency on speed and load. Polyamide composite- steel gear
pair efficiency at low speed found to be less compared to polyamide composite–
polyamide composite gear pair. On the other hand, at higher speed, both gear pairs
showed similar efficiency. Mao et al. [12] investigated the performance of wire cut
acetal gear pairs and compared it with injection-molded acetal gear pair. Results
confirmed that both gears failed similarly.
Kodeeswaran et al. [9] attempted to evaluate the performance of unidirectional and
bi-directional load polyamide gear. Results concluded that thermomechanical and
root crack were detected in the case of bi-directional loading. On the other hand, the
only root crack was observed in the case of unidirectional loading. Kodeeswaran et al.
[10] predicted the transmission error using finite element technique and compared
with experimental data. The static transmission error with the linear material models
was lower than that of the nonlinear material models, and it is further decreased with
Transmission Efficiency and Surface Damage of Polymer–Polymer … 849
increased strain rate, for both single tooth contact and double tooth contact. Senthil-
velan and Gnanamoorthy [16] investigated the effects of different filet radius on the
performance of polyamide 66 gear. Higher fillet radius 0.75 mm gear exhibited a
better performance compared to lower fillet radius 0.25 mm. Dearn et al. [5] studied
the different solid lubricant coated on polymer gear. Results found that the frictional
heating reduced significantly in case of PTFE coated gear. The failure of gear is
due to mainly delamination of coated and abrasive wear. Kirupasankar et al. [7]
investigated the polyamide 6 and clay-reinforced polyamide gear. Results concluded
that reinforcement is improved gear efficiency. Senthilvelan and Gnanamoorthy [18]
attempted to correlate the reinforcement on the profile of the polymer gear. Results
confirmed that reinforcement increases lead deviation. Sarita and Senthilvelan [16]
investigated the performance of polyamide 66–steel gear pair under lubricated condi-
tion. Results confirmed that the scuffing occurred on the flank region of test gear.
Pitting, tooth breakage and tooth root crack are the failure modes of the PEEK gear
run against steel under lubricated condition examined by Lu et al. [11].
Although from the above-mentioned literature, the gear performance of polyamide
against steel gear and polyamide against polymer gear has been studied extensively,
the amount of literature on polyamide–steel gear pair and polyamide–polyamide gear
pair behavior under lubricated condition is still limited.
In this paper, the effect of wet lubrication on the injection-molded polyamide 66–
polyamide 66 spur gear pair transmission efficiency and temperature is discussed
and compared with the unlubricated condition. Damage modes of gear observed
with different test conditions are reported.
2 Methodology
In-house developed power absorption gear test rig was used to evaluate the contact
load-bearing performance of injection-molded polyamide 66 spur gear. Gear tests
were carried out at a torque of 1.8 Nm and 800 revs/min rotational speed under unlu-
bricated and wet lubricated condition. Wet lubricated condition test was carried in
a splash lubricated system using SAE 75W85 grade gear oil (kinematic viscosity at
373 K: (11–15) cSt). 8 and 4 mm face width polyamide gears were used as driving and
driven gear, respectively. A personal computer-based data acquisition method was
used to continuously monitor gear’s surface temperature (only unlubricated condi-
tion) for driven gear and torque for both driving and driven gears. Lubricant temper-
ature was measured using J-type thermocouple. The detailed explanation of the test
rig is described elsewhere, Sarita and Senthilvelan [16]. Test gear the post-damage
morphology investigation was carried out by the digital microscope (celestron, digital
microscope pro). In-house developed power absorption gear experimental setup and
molded polyamide 66 gears are shown in Fig. 1a and b, respectively. Transmission
850 S. Bharti and S. Senthilvelan
Fig. 1 a In-house developed power absorption gear experimental setup and b Molded polyamide
66 gears
efficiency between the driver and the driven gear shaft was calculated using Eq. (1).
Here, for simplification in a calculation, the bearing loss and gear inaccuracies
were neglected (Table 1).
Fig. 2 a Multipair contact model of the spur gear and b meshed model
The net surface temperature of test gear tooth was measured under the unlubricated
condition measured using the non-contact temperature sensor (Raytek, MID10LT).
Under the wet lubricated condition, lubricant temperature was measured using a
thermocouple. Figure 3 shows that the gradual increase in temperature was observed
in the gear tooth surface during the initial period. The thermal balance was found
to be stable between heat generated and heat dissipated after approximately 1.2 ×
105 number of cycles. During the test, frictional heat was generated because the
pair of gears experienced both rolling and sliding contact due to surface interaction.
Since the thermoplastic material is viscoelastic in nature, material hysteresis due to
852 S. Bharti and S. Senthilvelan
Fig. 3 Shows the variation of gear tooth surface temperature and lubricant temperature with respect
to no. of cycles run
repeated gear tooth deflection results in heat generation. A similar trend in tempera-
ture has been reported by several researchers. Sarita and Senthilvelan [16] observed
around 360 K as stable temperature while testing PA 66 against steel gear at 1.8
Nm. Senthilvelan and Gnanamoorthy [17] also noticed stable temperature (340) K
when PA 66 runs at 2 Nm. The increase in the lubricant temperature is very low, and
this lubricant temperature plays a vital role in heat dissipation from the gear tooth
surface.
To quantify the effect of gear tooth wear performance, evaluation of gear transmission
efficiency was carried out. The sliding and rolling motion of gear tooth which leads
to frictional heating affect the power loss and transmission efficiency. In this work,
transmission efficiency was evaluated only considering the loss due to friction and
wear. Transmission efficiency under the unlubricated condition and wet lubricated
condition is shown in Fig. 4. Transmission efficiency under lubricated condition was
found to be higher as compared to the unlubricated condition. Under the unlubricated
condition, test gear exhibited more frictional heating on the contact surface which led
to less transmission efficiency. In the presence of lubricant, the generated frictional
heat between the contact surfaces is less due to the lubricant present between the
contact surfaces. This dissipates the heat generated and increases efficiency. The
percentage (%) variation in transmission efficiency for unlubricated condition and
wet lubricated condition is less than ± 5% for and ± 6%, respectively.
Transmission Efficiency and Surface Damage of Polymer–Polymer … 853
Fig. 4 Shows the gear efficiency under the unlubricated and wet lubricated conditions
Figure 5 shows the damage surface morphology of gear tooth surface subjected at
a torque of 1.8 Nm under unlubricated and lubricated condition. Under the unlu-
bricated condition, high surface wear was observed compared to the wet lubricated
condition. Since the thermal conductivity is low in the polymer material, the damage
in polymer gears is higher under unlubricated condition. In the presence of lubri-
cant, the direct contact of tooth surfaces is prevented by the lubricant. The adsorp-
tion of lubricant between the contact surfaces smoothens the rotation between the
surfaces and reduces the friction between them. The microscopic examination of
gear tooth damage in unlubricated condition revealed the occurrence of excessive
wear in both face and flank regions. In contrast, under lubricated condition, the wear
was higher near the face region compared to the flank region. Similar wear pattern
Fig. 5 Damage surface of gear subjected at 1.8 Nm, a under unlubricated condition and b under
lubricated condition
854 S. Bharti and S. Senthilvelan
was observed by Pogačnik and Tavčar [15] while testing the POM–POM polymer
gear under unlubricated condition.
Figure 6 depicts the variation of tooth deflection of driven gear simulated in
ABAQUS. The maximum deflection is at the highest point of the single tooth contact
region (HPSTC). In the double tooth contact region, tooth deflection progressively
increases in the addendum (face) region of the test gear tooth, whereas it decreased in
the dedendum (flank) region. Under the lubricated condition, the wear was observed
in the face region only, as the lubricant minimized the gear tooth wear.
4 Conclusion
In this work, polyamide 66–polyamide 66 spur gears were tested. Test gear exhibited
higher transmission efficiency under the wet lubricated condition at 1.8 Nm loading
condition. Under the unlubricated condition, the net gear tooth surface becomes
stable at 370 K. Lubricant temperature was limited at 303.4 K under lubricated
condition, and however, actual gear tooth surface temperature could be higher. Gear
surface wear was more under unlubricated condition compared to the wet lubricated
condition. Under wet lubricated condition, wear was observed at the gear tooth face
region due to higher gear deflection in this region.
Transmission Efficiency and Surface Damage of Polymer–Polymer … 855
References
Abstract In an application where a robotic arm hits an object, the effect of impact is
required to be assessed. A classical approach determines the effect of impact based on
the velocities of the objects. Magnitudes of acceleration are ignored in the analysis.
In the robotic application, velocity shall always be associated with acceleration as it
is a start and stop type of motion. This paper demonstrates the need of accounting
for the acceleration magnitudes also to determine the effect of impact.
Nomenclature
1 Introduction
The effect of acceleration on the momentum transfer is investigated using the experi-
mental method. Figure 2a demonstrates that in this situation as the vertical downward
gravitational force is still acting on a mass its velocity is increasing. Hence it is having
an acceleration in tangential direction and hence a component in horizontal direction.
860 P. V. Deosant et al.
Thus, at the moment of contact the swinging ball is having velocity and acceleration
both in the direction of impact. Velocity and acceleration in the direction normal to
the impact are ignored as justified earlier.
A conventional pendulum experiment is used for the experimentation since the
approximation of point mass is possible in such experiments and the relation between
the velocities, angle of oscillation, etc. are well established [1]. Figure 3 shows
oblique impact between two smooth masses. The velocity vector components along
x and y axes are resolved as u1x and u1y . The momentum of system is conserved
along the line of impact (x axis) only [1].
Experiments are performed by lifting the moving ball through angle (θ 1 ) and
releasing it. After collision, the stationary ball has maximum angle of oscillation of
(α). The angle of 50°, 60°, 70° are used as (θ 1 ) and for each value of (θ 1 ) multiple
experiments are performed. Average value of (α) for each (θ 1 ) is presented in Table
2.
The effect of retardation on the momentum transfer is investigated using the experi-
mental arrangement as shown in Fig. 4. It demonstrates a situation of contact where
the swinging ball hits the stationary ball after crossing the bottom position. As tangen-
tial velocity starts reducing after this point the velocity is associated with retardation
in this situation.
For equal masses (m1 = m2 = 0.04635 kg), coefficient of restitution (e) has been
considered as 0.906 as discuss earlier based on previous experiments. In case of
unequal masses (m1 = 0.04635 kg, m2 = 0.01213 kg), coefficient of restitution (e)
is consider as 0.7815. The length of strings L 1 = 0.465 m, L 2 = 0.4 m and the angle
of impact (β) = 31°.
Table 3 shows, different angles of swinging ball (θ 1 ), maximum oscillation angle
of moving ball (θ ) and maximum oscillation angle of stationary ball (α ) after
impact has been calculated for both equal and unequal masses. For each value of
(θ 1 ) multiple experiments are performed to obtain maximum oscillation angle of
stationary ball (α). Average value of (α) for each (θ 1 ) is presented in Table 3.
From Table 3, it has been observed that the effect of retardation on stationary ball
during oblique impact reduces the angle of oscillation of stationary ball. (α is always
862 P. V. Deosant et al.
less than α ) indicating that the retardation is decreasing the effect of the velocity.
The decrease is to the tune of 8%.
5 Conclusion
When there is an impact between two free bodies transfer of momentum is dependent
on both the velocity and acceleration of moving bodies. Acceleration increases the
momentum transferred and retardation reduces the momentum.
References
1. Hibbeler RC (2010) Dynamics engineering mechanics, 12th edn. Pearson Prentice Hall, New
Jersey
2. Hessel R, Perinotto AC, Alfaro RAM (2006) Force-versus-time curves during collisions between
two identical steel balls. Am J Phys 74(3):176–179
Effect of Acceleration of Moving Object During Collision … 863
3. Peraire J, Windnall S (2009) Linear impulse and momentum collision. Dynamics MIT Open
Source, pp 1–12
4. Kuwabara G (1987) Restitution coefficient in a collision between two spheres. Jpn J Appl Phys
26(8):1230–1233
5. Cross R (2013) The coefficient of restitution for collisions of happy balls, unhappy balls, and
tennis balls. Am J Phys 68(11):1025–1031
6. Dong H (2003) Measurement of impact behavior between balls and walls in grinding mills.
Miner Eng 16:543–550
Finite Element Modelling of the Human
Lumbar Vertebrae for Dynamic Analysis
1 Introduction
solid models are rather not suitable for all types of analysis. They usually consist of
many errors related to meshing due to geometrical discontinuities. Even few models
are dimensionally inaccurate when compared to the anatomical dimensions of the
vertebrae. These factors cause the limited availability of solid models of the vertebrae
that are suitable for finite element analysis.
Lavaste et al. [1] presented the development of a three-dimensional finite element
model of the lumbar vertebrae through dimensions obtained from two X-rays, i.e.
anteroposterior and lateral. The model is primarily developed using six limited
parameters digitized from the X-rays, and static analysis is carried out using finite
element method. Kasra et al. [2] presented the development of an axisymmetric
model of only a single L2–L3 lumbar disc-vertebra unit. The modelling is carried
out based on in vitro measurements of lumbar vertebrae specimens, assuming
symmetry about the sagittal plane. Lodygowski et al. [3] developed a simplified three-
dimensional model of a single vertebral segment (L4–L5) based on data produced
from computed tomographic (CT) scans and magnetic nuclear resonance (MNR).
The study concluded that it is often impossible to generate an accurate FE mesh for
models developed from CT data. Xu et al. [4] developed five nonlinear FE models
of the lumbar vertebrae from converting data obtained from CT scans of healthy
subjects. An additional smoothing process is performed to eliminate irregular surface
complexities such as spikes and holes, which could have been avoided by using an
appropriate modelling procedure.
In reported literatures, the most commonly used method of modelling bodies with
a complex geometry such as a vertebra is CT scans and converting them to generate the
three-dimensional model. During the process of conversion of CT data into surface
grids, many errors may arise in the geometrical reconstruction. This often leads to
errors in the process of finite element mesh generation as well. The complexity of
the geometry makes it impossible to generate a model with smooth surfaces, which
is critical in dynamic analysis. Models with uneven surfaces are suitable for limited
static analysis and visualization purposes only. Dynamic analysis of the vertebrae
requires a more refined geometrical model, allowing the error-free simulation of
relative motion between contact surfaces.
The present paper addresses the challenges in developing a solid model of the
human lumbar vertebrae. It presents the development of a more refined, simplified and
error-free model. The developed model has better scope for finite element analysis
and dynamic simulations. Section 2 outlines the modelling procedure of the human
lumbar vertebral column. Section 3 presents the modal analysis carried out to provide
evidence for the structural integrity of the developed model of the lumbar vertebrae.
2 Solid Modelling
The lumbar spine refers to the region of the vertebral column between the thorax
and sacrum. They are the largest segments of the vertebral column, consisting of
five vertebrae designated as L1–L5 from the top. A typical lumbar vertebra consists
Finite Element Modelling of the Human Lumbar Vertebrae … 867
of the vertebral body, pedicle, transverse processes, spinous process, vertebral arch
(lamina), superior and inferior articular processes and the vertebral foramen. The
upper and lower surfaces of the adjacent vertebral bodies are connected by means of
an intervertebral disc. The vertebral body is connected to the vertebral processes by
means of the pedicle. The superior articular process of a vertebra is attached to the
inferior articular process of an adjacent vertebra by means of a thin layer of articular
cartilage. The dimensions of each of these vertebral elements are considered for the
modelling of lumbar vertebrae.
A comprehensive study of the geometrical characteristics of the lumbar verte-
brae suggests that the dimensions of each vertebra are approximately equal and
does not show a significant variation in geometry. However, the dimensions of each
vertebra vary from person to person depending on various factors such as age, gender,
height, weight and country of origin. Therefore, to simplify the modelling procedure,
dimensions of the middle vertebra, L3 is considered to model all five vertebrae. The
dimensions of the L3 vertebra are borrowed from Zhou et al. [5], where the geomet-
rical dimensions are measured and averaged from 55 computed tomographic (CT)
images of male patients of mean age, 50 ± 13.60, and are shown in Table 1. The
reference notations for the L3 vertebra [6] are indicated in Fig. 1.
These dimensions only aid in the preliminary modelling of the vertebra and are
not sufficient to entirely model the vertebra to a higher level of accuracy. The required
supplementary dimensions of surface contours and posterior elements are measured
manually from the specimens of actual lumbar vertebrae, as they are not reported in
literatures. The required dimensions are measured with utmost precision by using
the appropriate scaling factor with respect to the dimensions available in [5]. Surface
contours and complex shapes are traced and measured using the three-point method
and approximated appropriately to obtain an accurate model. The manually measured
dimensions of the vertebra are shown in Fig. 2 and shown in Table 2.
The solid model of the L3 vertebra is developed in Solid Edge software. The
modelling process is initiated by developing the vertebral body. Subsequently, the
posterior elements are modelled using lofted protrusions between angles planes.
Several attempts and multiple iterations are made in developing irregular parts and
Table 1 Dimensions of L3
Dimension Notation Value (mm)
vertebra [5]
Upper vertebral width (UVW) a 46.1
Upper vertebral depth (UVD) b 34.1
Spinal canal width (SCW) c 25.2
Spinal canal depth (SCD) d 16.1
Pedicle width (PDW) e 10.7
Transverse process length (TPL) f 96.1
Vertebral body height (VBH) g 30.6
Intervertebral disc height (DH) h 12.4
Pedicle height (PDH) i 14.9
868 Raj Arjun S. I. et al.
f
g
d h
c e
i
a
b
Fig. 1 Dimensions of a lumbar vertebra [5]
m l
j
n
complex surface contours of the vertebra. The final model is well-refined and dimen-
sionally intact. Different views of the developed model of the vertebra are shown in
Fig. 3.
(a) Front view (anterior) (b) Rear view (posterior) (c) Top view
(a) Front view (b) Back view (c) Right view (d) Isometric view
Similarly, the solid models of the intervertebral disc and the articular cartilage
are also developed. The developed models of the vertebra, intervertebral disc and
articular cartilage are assembled together to form a single lumbar functional spinal
unit (FSU), as shown in Fig. 4.
Subsequently, the full model of the lumbar vertebrae is developed by assembling
five lumbar FSUs. The curvature of the lumbar vertebrae is taken into consideration
while assembling the models. A lumbar lordotic angle of 20° for a seated human in
upright position is considered [7]. Figure 5 shows the fully developed model of the
lumbar vertebrae in different views.
This model is ready to be imported into FE analysis for dynamic simulations.
The developed solid model shows scope for both static and dynamic simulation,
under any load conditions. For the present study, finite element modal analysis is
carried out to demonstrate the structural integrity and robustness of the developed
models. The models are imported into ANSYS finite element analysis package. The
appropriate material properties of each vertebral element are taken into consideration.
870 Raj Arjun S. I. et al.
(a) 1st Mode = 250.23 Hz (b) 2nd Mode = 281.51 Hz (c) 3rd Mode = 361.37 Hz
Flexion-extension Lateral flexion Rotation
Program controlled meshing with proximity and curvature size function is defined
to generate the finite element mesh. An additional tetrahedron meshing method is
defined to avoid errors in meshing complex faces and edges. No pre-stress conditions
are given. The top surface of the L1 vertebra and bottom surface of the L5 vertebra
are considered to have a fixed support.
The analysis is carried out; the total deformations and mode shapes are generated,
and results are compared. The three fundamental modes generated for the lumbar
vertebrae are shown in Fig. 6.
The obtained natural frequencies are in good correlation with analytical values.
The generated mode shapes also correspond to the actual modes of deformation, as
indicated in previously reported studies. A similar approach can be used to model
other biological systems with complex morphologies. This study shows scope for
further in-depth analysis of the lumbar vertebrae when subjected to complex loading
conditions.
References
Abstract Mobile robotics has received a lot of interest among the researchers in
recent years, particularly for applications in the remote areas. Six wheels multiterrain
robot (SW-MTR) is a robot with semicircular rocker and bogie mechanism developed
at PDPM IIITDM Jabalpur. The robot moves on different terrain including sandy area,
uneven rocky area, steps of 6 cm height, boulders of 8 cm height, inclined plane of
25°, etc. For steering of the SW-MTR, the hybrid steering system has been proposed
with a combination of differential steering mechanism with worm-sector steering
mechanism. This paper presents the dynamic model of the hybrid steering system
with specific consideration on the wheel skidding effects and load distribution on
each of the six wheels. A comparison of the turning radius, contact forces between
wheels and surface, torque and power consumption of the proposed hybrid steering
system with other individual steering systems is also presented through simulation.
1 Introduction
Various robotic systems are highly in demand for numerous purposes such as robotic
arms on assembly lines, aerial drones such as hexa-copter and quad-copter, and multi-
terrain robots. Out of these, the mobile robotics has emerged out as one of the most
popular interdisciplinary fields for the purpose of the planet and asteroid exploration,
rescue operations, and path finding or planning in complex areas where humans are
unable to reach and security surveillance. For accomplishing these tasks, the robot
A. Nakoriya (B)
Mechatronics Laboratory, PDPM Indian Institute of Information Technology Design and
Manufacturing, Jabalpur, India
e-mail: [email protected]
V. K. Gupta
Professor, PDPM Indian Institute of Information Technology Design and Manufacturing,
Jabalpur, India
e-mail: [email protected]
should have terrain adaptability and effective steering system. Terrain adaptability
for SW-MTR has already been discussed by Gupta and Gupta [1].
The steering system is a very important part of any wheeled robotic system.
It is observed that two basic steering systems, caterpillar differential steering or
worm-sector steering, have been used independently in many multiterrain robots.
The robots like ARES III [2] and shrimp [3] show the use of worm-sector steering
while caterpillar differential steering system has been used by robots like Packbot [4],
Talon [5], etc. In case of caterpillar differential steering mechanism due to skidding
of the wheels, the turning radius is smaller while friction between wheels and the
skidding surface is very high hence reduces wheel life. The worm-sector steering
mechanism provides the low frictional forces between wheel and surface but the
radius of steering is larger. In this paper, it is proposed to use a hybrid system for
steering to combine the advantages of both types of steering mechanisms.
A dynamic model of proposed steering for SW-MTR is necessary to understand
the analogy with the physical system. It is very helpful in the trajectory controlling
of the robot. A trajectory tracking control based on differential steering is discussed
by Caracciolo et al. [6]. Similar kind of work is presented by Kozlowski et al. [7],
Zhang et al. [8], and Wu et al. [9]. This paper presents an analytical formulation of
longitudinal rolling resistance, lateral forces on wheels, and moment of resistance on
each wheel to estimate skidding effect while taking a turn with the proposed steering
system. Simulations are carried out in MSC ADAMS® and results are compared.
The steering system is one of the important parts of the wheeled robotic system. Good
steering has the characteristics of a smaller turning radius, less power consumption for
turning and able to turn robot in a complex environment in a short time. In this work,
a hybrid steering system is proposed to combine the advantages of both differential
steering and worm-sector steering system. An individual multibody dynamic study
of differential steering, worm-sector steering, and proposed hybrid steering by using
MSC ADAMS® is presented in this paper.
problem. The coefficient of friction between rubber and flooring material is taken as
0.6–0.9 [10].
Willy problem. It is the upward movement of the rocker link (any one out of two)
depending on the differential speed of wheel while the robot in motion. Due to
3:5 rocker hinge pivoting ratio, the weight gets distributed unequally among the
first four wheels. Therefore, middle wheels are taking more weight as compared to
front wheels causing an upward movement of front wheel. If the robot is taking a
left turn then right-side rocker link gets subjected to Willy problem. To avoid this
problem, the differential speeds of the robot must not have a large difference in speed.
It is necessary to calculate the differential speed range for various environmental
conditions for which there is no Willy problem for SW-MTR. Hence, by successive
simulation in MSC ADAMS®, for different friction values, for high speed of 85
RPM on one side, corresponding lower speed is found out and presented in Table 1.
Due to the lower radius of steering, second parameters are taken for further simu-
lation purpose. Figure 2 shows the Willy problem in simulation. Steering radius,
contact forces, motion torque, and power consumption are calculated and compared
with other steering system in the results section.
The worm-sector steering consists of independent high torque motor to steer the
SW-MTR which is shown in Fig. 1c. The limitation of steering is set to 15° because
beyond this value the torque required on rear wheels motors increases as well as
(a) Inclined plane climbing by SW-MTR (b) Stair Climbing by SW-MTR (c) Hybrid steering of SW-MTR
mechanism gets stuck. In the simulation, the motion given to the motor is very well
described by the step function of STEP (time, 0, 0, 1, −0.75D). It returns an array
of y values, corresponding to x values [11].
The main objective of hybrid steering is to combine the advantages of both differ-
ential and worm-sector steering systems. The differential steering provides the low
steering radius. While in case of worm-sector steering, the frictional forces acting
on wheels are less hence the life of wheels is more. The hybrid steering system
Hybrid Steering System of Six-Wheel Multiterrain Robot (SW-MTR) 875
gives the advantage of low reducing steering radius with less amount of contact
forces acting on wheels. Hybrid steering also reduces steering torque requirement
and power consumption of high torque motor used in worm-sector steering to 55
kgf-cm and 0.65 kgf-cm/s, respectively, and is shown in Figs. 5 and 6.
A dynamic model of the hybrid steering system is developed by neglecting the effects
produced by the robot suspension system and tire deformation [6].
The following are the assumptions made during the development of the model.
(1) A rigid robot turning left on inclined terrain present toward left (angle 3°).
(2) Center of gravity of the robot body shifts laterally, longitudinally, and vertically.
(3) Speed of the robot is below 50 cm/s.
(4) Wheel slip along the length is neglected.
(5) Lateral force on wheel contact is the function of vertical load on corresponding
wheel contact.
The robots fixed reference frame is indicated by F(X, Y) and the local reference frame
is f(x, y). Since the robot turning on 3° plane toward left is considered, the center of
gravity of robot is located at a distance ‘a’, ‘b’ and ‘c’ from the front, middle, and rear
wheel axes, respectively. The height of the center of gravity is ‘h’ and it is laterally
positioned at a distance ‘t 1 ’ and ‘t 2 ’ from the right and left sides, respectively.
Let Ẋ , Ẏ , and θ̇ are longitudinal, lateral, and angular velocities of the robot in
local reference frame, i.e., f(x, y). Now we can write the velocities in the global
reference frame as given below.
Ẋ ẋ cos θ − ẏ sin θ ẋ
= = R(θ ) (1)
Ẏ ẋ sin θ + ẏ cos θ ẏ
By differentiating the above equation, we get the acceleration matrix of the robot.
Ẍ cos θ − sin θ ẍ − θ̇ ẏ
=
Ÿ sin θ cos θ ÿ + θ̇ ẋ
Ẍ a
= R(θ ) x (2)
Ÿ ay
where the terms ax and a y indicate the acceleration in local reference frame, i.e.,
f(x, y). For an infinitesimally small time, the robot moves in an exact circle with the
center of rotation as ‘i’. ‘i’ is called as the instantaneous center of rotation (ICR)
along which the robots longitudinal velocity in f(x, y) frame is almost zero. The
coordinates of ICR are given as follows.
xi − ẏ/θ̇
= (3)
yi ẋ/θ̇
878 A. Nakoriya and V. K. Gupta
When the robot moves in a straight-line path, the angular velocity θ̇ and lateral
velocity ẏ become zero. The position of ICR in this case is along the y-axis at the
infinity. When the robot moves along the curved path, the instantaneous center of
rotation moves forward by an amount |x i |. The loss of motion stability is observed
when the position of x i goes outside the wheel base of the robot. The longitudinal
and lateral velocities in the frame f(x, y) are expressed below.
Figure 7a shows the free body diagram of the robot in which all components of forces
as well as velocities are shown. The robot is having instantaneous positive velocity
components ẋ and θ̇ and negative velocity ẏ. When the power is transmitted through
the wheels, there is a generation of tractive forces F xi and rolling resistance force Rxi
(i = 1 … 6). For minimization of longitudinal slip, it is necessary to assume that the
wheel actuation of each wheel is the same. Because if each motor which is connected
to wheels is actuated at the same time there will be minimum possibility of Willy
problem. And hence, this will result in F x1 = F x4 = F x6 and F x2 = F x3 = F x5 . The
lateral skidding causes the F yi. There is a resistive moment M r caused by resistive
forces, i.e., Rxi and F yi . In the local reference frame f(x, y), the equations of motion
(a) Top view of free body (b) Side view of free body (c) Front view of free body
diagram of SW-MTR diagram of SW-MTR diagram of SW-MTR
for a robot of mass m and inertia I about its center of mass are written as:
ma y = −Fy (6)
The magnitude of the weight distribution of robot among the six wheels is neces-
sary. Since the magnitude on each wheel is used to find longitudinal rolling resistive
force Rx , lateral resistive force F y and resistive moment M r. Here we introduce a
Coulomb friction model for ground and wheel contact.
For calculation of reactions on wheels, the front four wheels are assumed to be
as a two-wheel system (one on each side). This assumption is valid since the front
four wheels are attached to the rocker system through a single axle. This converts the
six-wheel system into the four-wheel system. By referring to Fig. 7b, it can be stated
that the position of two hypothetical wheels is in line with front axle axis (Fig. 8).
Hence, front axle load is given as:
c h
FZf = mg − max (8)
c+d c+d
d h
Fzr = mg + max (9)
c+d c+d
The robot is taking a left turn on the steep surface which is 3° inclined. Due to
this, the CG shifts occur. Figure 7c illustrates a half-robot model for the front axle,
where the virtual mass of the front axle M f is calculated as follows:
880 A. Nakoriya and V. K. Gupta
The front axle left and right forces calculated above are further divided into 3:5
ratio. As the pivoting ratio in the rocker mechanism is maintained at 3:5. Fzfl is
divided into Fz1 and Fz4 as well as Fz f r is in Fz2 and Fz3 . They are written below.
3 c h t1 h.a y
Fz1 = mg − max × − (13)
8 c+d c+d t1 + t2 g.(t1 + t2 )
3 c h h.a y t2
Fz2 = mg − max × + (14)
8 c+d c+d g.(t1 + t2 ) t1 + t2
5 c h h.a y t2
Fz3 = mg − max × + (15)
8 c+d c+d g.(t1 + t2 ) t1 + t2
5 c h t1 h.a y
Fz4 = mg − max × − (16)
8 c+d c+d t1 + t2 g.(t1 + t2 )
For rear axle loading on each side, follow the same method as followed for the
front axle. But in this case, the load on each side will not get divided into 3:5 ratio.
The forces are written below.
d h h.a y t2
Fz5 = mg + max × + (17)
c+d c+d g.(t1 + t2 ) t1 + t2
d h t1 h.a y
Fz6 = mg + max × − (18)
c+d c+d t1 + t2 g.(t1 + t2 )
If the robot is moving on the curved path with lower speeds, its lateral load due
to centrifugal force can be neglected. The contact patch between ground and wheel
is assumed as rectangular. Hence, the uniform pressure distribution is produced by
tire vertical load. The rolling resistance equation is given by the following [6].
where fr is the coefficient of rolling resistance which is independent from the velocity
of the robot. The total longitudinal resistive force is given as
6
Rx = Rxi = fr Fz1 sgn(ẋ1 ) + fr Fz2 sgn(ẋ2 )
i=1
+ fr Fz3 sgn(ẋ3 ) + fr Fz4 sgn(ẋ4 ) + fr Fz5 sgn(ẋ5 ) + fr Fz6 sgn(ẋ6 ) (20)
The lateral force acting on wheels of the robot with μ as the coefficient of lateral
resistance is given below.
6
Fy = Fyi = μFz1 sgn( ẏ1 ) + μFz1 sgn( ẏ2 )
i=1
+ μFz1 sgn( ẏ3 ) + μFz4 sgn( ẏ4 ) + μFz5 sgn( ẏ5 ) + μFz1 sgn( ẏ6 ) (21)
The dynamic model can be written in global reference frame F(x, y), introducing
the generalized coordinates q = (X, Y, θ ) and matrix notation.
⎡ ⎤ ⎡ ⎤
m 0 0 Rx cos θ − Fy sin θ
where M = ⎣ 0 m 0 ⎦, c(q, q̇) = ⎣ Rx sinθ + Fy cos θ ⎦ and E(q) =
0 0 I Mr
⎡ ⎤
cos θ/r cosθ/r
⎣ sinθ/r sinθ/r ⎦, τi = 2r Fxi (i = 1, . . . , 4).
(t1 + t2 )/r −(t1 + t2 )/r
where r is the wheel radius of the robot and τ1 , τ2 , τ3 & τ4 are the torques produced
by left- and right-side motors at load side, respectively.
The parameters selected for comparison purpose are the radius of steering, contact
forces on wheels in all three directions, torque and power consumption of wheel
motors.
882 A. Nakoriya and V. K. Gupta
Table 2 Comparison of
Simulation Differential Hybrid Worm-sector
simulation radius
time steering steering steering
(s) (cm) (cm) (cm)
2.5 111 110 4608
5 110.3 105 1459
7.5 110 99 839
10 110 95 587
12.5 110 90 446
15 110 86 373
17.5 110 81 315
20 110 78 272
From Table 2, it is observed that the differential steering has a constant steering
radius throughout the simulation time. Worm-sector steering has reducing steering
radius with a large value of initial steering radius, while hybrid steering combines
both the steering has reducing steering radius with the lower value of initial steering
radius.
Contact forces between wheels and ground surface decide the life of the tire, power
consumption, and torque requirement. Tables 3, 4, and 5 show the contact forces in
Table 3 Comparison of
Wheel Differential Hybrid Worm-sector
lateral direction contact forces
steering F x steering F x steering F x
(kg-f) (kg-f) (kg-f)
Left front 3.3 1.9 −1.4
wheel
Left middle 3.3 −3.9 −1.5
wheel
Left rear wheel 4.7 3.35 −1.6
Right front −3.5 3.6 1.3
wheel
Right middle −1.9 −1.9 2
wheel
Right rear −1.9 −1.9 −1.25
wheel
Hybrid Steering System of Six-Wheel Multiterrain Robot (SW-MTR) 883
Table 4 Comparison of
Wheel Differential Hybrid Worm-sector
longitudinal direction contact
steering F y steering F y steering F y
forces
(kg-f) (kg-f) (kg-f)
Left front −3.3 −3 2
wheel
Left middle −4.0 −3 −2
wheel
Left rear wheel −4.0 −5 1.5
Right front −3.5 2.9 1
wheel
Right middle −2.0 −2.0 −3
wheel
Right rear −2.5 −2.5 2.5
wheel
Table 5 Comparison of
Wheel Differential Hybrid Worm-sector
reactive forces
steering F z steering F z steering F z
(kg-f) (kg-f) (kg-f)
Left front −8.0 −7 −7
wheel
Left middle −9.0 −10 −7
wheel
Left rear wheel −13.0 −10 −6
Right front −8.0 −6 −7
wheel
Right middle −6.0 −4 −7
wheel
Right rear −5.0 −5 −6
wheel
all three directions. The x-axis is taken along the lateral side, the y-axis is taken along
the longitudinal side, and the z-axis is taken along the height of the robot.
Table 3 shows the values of contact forces along lateral direction in hybrid steering.
Longitudinal direction contact forces are shown in Table 4. Most of the magnitude
of contact forces in case of the hybrid steering system is between the values of
differential and worm-sector steering system.
By observing Table 5, in hybrid steering, while taking the left turn, wheels present
on the right side experience less reactive force than other two steering systems. Hence,
the life of the tire increases.
884 A. Nakoriya and V. K. Gupta
Table 6 Comparison of
Wheel Differential Hybrid Worm-sector
motion torque
steering steering steering torque
torque torque (kgf-cm)
(kgf-cm) (kgf-cm)
Left front 19 13.5 12
wheel
Left middle 24 22.5 7
wheel
Left rear wheel 27 24.375 3
Right front 12.5 10.625 8
wheel
Right middle 12 8 12.5
wheel
Right rear 8 7 11
wheel
Table 6 presents the motion torque requirement for differential, worm-sector, and
hybrid steering system. The observation from the table is that though motion torque
requirement of worm-sector steering for each wheel are least, hybrid steering is
preferable because it gives reducing steering radius starting from the lower magnitude
of radius as compared to worm-sector steering.
Table 7 shows the power consumption for each of the six wheels in all three types of
steering studied. The magnitude of power consumption for each wheel in hybrid
steering falls between the magnitude of power for differential and worm-sector
steering.
All the comparison parameters tabled above are the basis of selection of hybrid
steering as a steering system for SW-MTR. Hybrid steering gives lower reducing
steering radius which is very useful for steering in very complex areas.
5 Conclusion
Through the work presented in this paper, a hybrid steering system is proposed
for SW-MTR. An individual multibody dynamic study using MSC ADAMS® of
differential steering, worm-sector steering, and hybrid steering is carried out. Various
parameters such as steering radius, wheel–surface contact forces, motion torque and
Hybrid Steering System of Six-Wheel Multiterrain Robot (SW-MTR) 885
Table 7 Comparison of
Wheel Differential Combined Worm-sector
power consumption
steering steering steering power
power power (kgf-cm/s)
(kgf-cm/s) (kgf-cm/s)
Left front 170 120 90
wheel
Left middle 215 210 50
wheel
Left rear wheel 240 217.5 30
Right front 65 50 75
wheel
Right middle 45 35 110
wheel
Right rear 38 40 100
wheel
power consumption on each wheel are compared for each of three steering systems.
Based on this, a hybrid steering system is selected. Dynamic modeling of proposed
hybrid steering has been presented.
References
1. Gupta AK, Gupta VK (2013) Design and development of six-wheeled multi-terrain robot.
In: CARE 2013—2013 IEEE international conference on control, automation, robotics and
embedded systems proceedings, pp 1–6. https://doi.org/10.1109/CARE.2013.6733751
2. Gliedes M, Santana P, Deusdado P, Mendonca R, Marques F, Henriques N, Lourenco A, Correia
L, Barata J, Flores L (2012) ARES-III: a versatile multi-purpose all-terrain robot. In: IEEE
international conference on emerging technology factory automation (ETFA 2012). https://doi.
org/10.1109/ETFA.2012.6489633
3. Dian S, Liu T, Liang Y, Liang M (2011) A novel shrimp rover-based mobile robot for moni-
toring tunnel power cables. In: Proceedings of the 2011 IEEE international conference on
mechatronics and automation. ISBN: 978-1-4244-8115
4. Yamauchi BM (2004) PackBot: a versatile platform for military robotics. Unmanned Gr Veh
Technol VI 5422:228–237. https://doi.org/10.1117/12.538328
5. Talon: Talon tracked mitary robot [Online]. Available: https://www.army-technology.com/pro
jects/talon-tracked-military-robot/
6. Luca C, Alessandro De Luca SI Trajectory tracking control of four wheel differentially driven
mobile robot. In: Proceedings of 1990 international conference on robotics and automation.
Detroid, Michigan
7. Kozłowski K, Pazderski D (2004) Modeling and control of a 4-wheel skid-steering mobile robot.
Int J Appl Math Comput Sci 14:477–496. https://doi.org/10.1016/j.compag.2011.12.009
8. Zhang Y, Hong D, Chung JH, Velinsky SA (1998) Dynamic model based robust tracking control
of a differentially steered wheeled mobile robot. Proc Am Control Conf 2:850–855. https://
doi.org/10.1109/ACC.1998.703528
9. Wu X, Xu M, Wang L (2013) Differential speed steering control for four-wheel independent
driving electric vehicle. In: IEEE international symposium on industrial electronics, p 1. https://
doi.org/10.1109/ISIE.2013.6563667
886 A. Nakoriya and V. K. Gupta
10. El-Sherbiny YM, Hasouna AT, Ali WY (2012) Friction coefficient of rubber sliding against
flooring materials. ARPN J Eng Appl Sci 7:121–126
11. Norton RL (2013) Adams tutorial kit for mechanical engineering courses, pp 1–352
Flexible Coupling—A Research Review
In many cases of industrial domain, two shafts are joined together with a coupling.
One of the situations is joining a motor shaft with centrifugal pump. This case is
considered to be complex as misalignment in shafts may hamper the performance of
the centrifugal pump. There are many cases which are confronting same situation. A
solution to such situation is the usage of flexible couplings, which are present in the
market. But it also allows the flexibility to some extent. There are some centrifugal
chillers which are directly placed on the motor shaft and thus there is no need of such
couplings [1]. Of course, it is a special case, but in most of the applications, one has
to couple the motor shaft with process unit. Study reveals that the maintenance cost
of any machine increases as misalignment in the shafts increases.
Misalignment in the coupling produces moment at the bearings which in turn
pulls the shaft away from its centerline. This action is periodic and cyclic; therefore
it produces vibrations in entire machine. Under these vibrations, the coupling and
its component are set into stress phenomenon which could otherwise be coined as
stress under vibrations. If this stress pattern is cyclic then localized area could be
formed which may culminate into its failure. Therefore it is inevitable to come out
with a proper solution by evolving a new design of Coupling to sustain the cyclic
stress and will give more flexibility in both the directions, i.e., angular and offset.
A rigorous Literature Review is executed for the present research. Study reveals
that misalignment is unavoidable condition. Due to misalignment, stresses induced
in components of a coupling are ascertained to be cyclic; therefore it compels to
localized stress area and some form of combined stresses. For the present research,
various papers are reviewed, but due to space limitations, few prominent research
papers and their reviews are discussed as below.
The research reports toward evaluation of vibration signatures induced by angular
misalignment. The experimental setup is made with two shafts four journal bearing
and one flexible disk coupling. Analysis is executed to find out the effect of misalign-
ment on displacement spectrum of the bearing. The effect of forces on the journal
bearing is also ascertained by considering the system dynamics and by the forces
and moment generated by the flexible coupling while accommodating misalignment.
In this paper, a structural analysis of disk type coupling is also reported by Finite
Element Method [2].
Through this research, a study is executed in which the power transmission
between the motor and load is performed by means of different types of coupling
mainly those most frequently used in the industry. Results show that the conclusion
drawn from a particular coupling is not necessarily applicable to others [3].
In this research, the author has tried to evaluate the accurate characterization of
torsional stiffness of flexible disk coupling. Torsional stiffness is a key factor while
studying torsional vibrations. Most of the literature reported the influence of coupling
type and design selection parameters on the operating limits of various couplings.
Flexible coupling while allowing the small amount of misalignment that may other-
wise lead to equipment failure. This paper reports about the performance analysis
of flexible disk coupling by comparing theoretical, experimental and computational
methods [4].
A special paper is published by “World Pumps” is reviewed. This paper focused
on the use of various types of flexible coupling for Industry. Their performance
comparison including transmitting torque capacity, the amount of misalignment to be
taken, bending moment, torsional stiffness, axial resisting force, lubrication required
are nicely discussed [5].
Shaft Misalignment and rotor unbalance are major concerns in rotating
machineries. To understand dynamic characteristics of these faults, a theoretical
model of a complete Motor-Flexible Coupling–Rotor System capable of describing
the mechanical vibration resulting from misalignment and unbalance is developed.
Flexible Coupling—A Research Review 889
The models of Kramer and Nelson and Crandall presented here, take into account
the inherent flexibility of the couplings. A parametric sensitivity analysis is of signif-
icant importance for the choice of the unbalance response and the d.o.f to be used in
the process [12].
From the above literature, it is ascertained that one may try a new type of coupling to
ascertain its performance characteristics followed by optimized design parameters.
Also one may look into the various patented flexible couplings which are available
in the market (few are listed below in the table) to evolve with a new type of Flexible
Coupling.
To ascertain the response of the proposed new type of Flexible Coupling we have
to test the coupling under various loading patterns with the help of an Experimental
Setup. The experimentation would be planned considering the geometrical aspects
of the coupling, the dynamic aspects of the experimental setup such as speed of the
motor, rated power delivered, and load torque imposed.
The main factor of dynamic condition is the load torque imposed. The load torque
pattern is considered to be in various phases. The phases are:
1. Random Pattern
2. Gradual Pattern
3. Instantaneous Pattern
4. Constant Pattern.
Under above loading pattern, the proposed coupling would be tested to ascertain
various responses like:
1. The stress induced in different components of the coupling
2. Vibration induced due to misalignment at the side of the bearing
3. Temperature induced at the outer body of the Coupling
4. Power consumption of the Motor.
The main intention of designing such coupling is to mitigate the stress under
vibration, vibration induced at the bearing, temperature induced at the outer body of
the Coupling, Power consumption of the motor.
This Aim could be accomplished by gathering the Experimental Data. A Math-
ematical Model between the Response Variables and Design Variables would be
formed. If one optimizes these models for the minimization of responses then the
best set of designed variables could be accomplished.
The data generated from the experimentation will now be useful to establish the
empirical relationship or models between the design variables of the Coupling. The
Quantitative and Qualitative Analysis of the established empirical models will be
Flexible Coupling—A Research Review 891
executed. The Quantitative Analysis will include sensitivity, reliability, and opti-
mization of the established models while Qualitative Analysis will ascertain the
influence of indices on response variables.
The probable and tentative variables for the proposed coupling are detailed out as
below:
Output Design Variables:
1. Rated Power.
2. Stress induced in the Swing Element of the Coupling.
3. Temperature induced in the outer body of the Coupling.
4. Vibration induced at the Bearings at the side of Coupling.
Input Design Variables:
1. Load Torque.
2. Speed of Shaft etc.
3. Atmospheric Variables.
4. Geometrical Variables of Coupling.
5. Time Dimensions and Gravitational Constants.
The forthcoming paper would be based on the proposed new design of a Flexible
Coupling considering the discussions made in Summary and Conclusion.
References
12. Tadeo AT, Cavalca KL (2003) A comparison of flexible coupling models for updating in rotating
machinery response. J Braz Soc Mech Sci Eng 25:235–246. https://doi.org/10.1590/s1678-587
82003000300004
Design of Post-Curing Inflator Using
Bistable Locking Mechanism
Abstract Bistable systems have two distinct stable states, where the system stays
without any application of external power. In this paper, stable states of the buckled
beam are used along with a four-bar mechanism to design a locking for the post-
curing inflator. Post-curing of tires is necessary to avoid deformation in nylon fabric
during cooling. Current setups, available in the market, use external force loop for
the application of pressure during the process causing significant bending stresses
in the frame, hence making the design bulky. Further, such conventional setup limits
the number of simultaneous operations on a shop floor. In this paper, we propose a
bistable structure-based locking mechanism, where, loading elements are in tension
instead of bending, which makes the whole setup compact. We also discussed the
scope of pseudo-bistability for design consideration.
1 Introduction
available in the market uses an external force loop, which loads the components in
bending. To design for this bending stress, lots of material is used, and the whole
setup requires lots of space to operate. In addition to this conventional setup operates
on hydraulics to move the sealing lids, which adds to the operational cost.
To avoid these issues and reduce the cost, external force loop can be shorted using
an internal force loop such that Instead of bending, the elements will be loaded in
tension. This would reduce the space, materials as well as the manufacturing cost. In
this paper, we propose a novel bistable mechanism-based locking design for creating
an internal force loop in the post-curing inflator (PCI) setup.
Section 2 of this paper discussed the design requirements for post-curing inflator.
Section 3 deals with the bistable system using a simple buckled beam example. It
further discussed pseudo-bistability, retractable mechanism, compliant mechanism
and four-bar linkage. In Sect. 4, the design of a new locking mechanism using a
bistable structure and four-bar linkage is proposed. At last, it discusses the prototyped
setup and its functionality.
2 Design Requirements
The primary design statement is to seal the axial ends of the tire without distorting the
rubber during the process. The assembly of setup with tire cannot be that of car axle,
and tire as the freshly cured hot tire will deform if it is mounted using tire changing
machine. For this, there are two sealing lids mounted over the axial opening of the
tire. One lid seals the bottom of the freshly cured tire, and another lid presses the tire
from the top. The chamber formed between the two covers and the tire is pressurized
using an external source. The cover at the top holds the pressed position up to 10 psi
pressure and moves up with 10 mm (to allow the tire to assume its final shape) as
the pressure builds up inside the tire. The top lid holds this position for the rest of
the process. The locking should be self-stabilizing, the more pressure builds up the
more lock should tighten up. The chamber should be Leak-proof for pressurization.
3 Bistable System
Bistable mechanisms have two stable equilibrium position with at least one unstable
equilibrium state in between [2, 3]. A simple buckled beam snaps from one stable
state to another by applying external forces. It is one of the most straightforward
bistable systems with limited motion range [4]. Similarly, a spherical shell can act
as a bistable system [5].
Design of Post-Curing Inflator Using Bistable Locking Mechanism 895
Fig. 1 a Buckled beam and its 4-link approximation b Potential energy diagram of bistable
A perfect bistable system upon actuation never returns to its original state unless
external actuation is provided. While a mono-stable structure returns instantly to
its original state once the force is removed. Pseudo-bistability lies in between the
two regions. Pseudo-bistable structures return to its original state without further
actuation within a finite time [6]. This finite time depends on various parameters like
elastic modulus, Poisson relations, boundary conditions, system dimensions etc. By
varying these properties one can go from mono-stability to perfect bistability [6].
The four-bar linkage is synthesized using four rigid members in a plane: the frame, the
input link, the output link, and the coupler link. These members are joined together
in a closed-loop kinematic chain with single-degree-of-freedom [9]. These members
can be joined via revolute pair or can be synthesized as a single component with
compliant flexible segments acting as joints. Compliant four-bar linkage will increase
reliability and reduce manufacturing and maintenance cost.
4 Design Setup
As once the external actuation is removed by the worker, this spherical shell goes
back to the lock state. Hence ensuring the safety of the shop floor.
To change the state of bistable beams, external actuation is required. This actuation
pressure is calculated using Eq. (1) [10].
Et 3 2
P= 3
k −1 (1)
12r
where,
P is the pressure (N/m),
E is the elastic modulus,
t is the thickness of the plate,
r is the radius of curvature,
2α is the angle subtended by the arc.
For 2α = 30◦ , t = 1mm, r = 0.7m, E = 200 ∗ 109 Pa. Pressure on each plate is
calculated to be 170 kPa which gives a total load of 2.54 kN. With total 3 such plates
in the structure required load will be 7.7 kN (Table 1).
898 Md Z. Anwar et al.
Buckled beam and Spherical shell setups are prototyped using 3D printing facility at
central Workshop, IIT Delhi to demonstrate the basic functionality of the designed
bistable locking mechanism. Of the two designs, the buckled beam system is perfect
bistable structure. Once actuated to a state, it remains there if no external load is
applied. In case of the spherical shell design, which is a “pseudo-bistable” [4] struc-
ture, once actuated it tends to snap back to its original shape without any further
actuation. The prototyped setup is shown in its configurations in Figs. 3 and 4.
6 Conclusion
In this paper, we discussed a locking system designed for post-curing inflator using
bistable structures coupled with four-bar mechanisms. Two stable states of the
bistable structure correspond to the locked and unlocked case. Two bistable locking
mechanism (Buckled beam and Spherical shell) based PCI had been prototyped using
rapid prototyping to check the functionalities. Buckled beam system is found to be
stable in both states while the spherical shell tends to snap back to its manufactured
shape without any application of external force. Different Bistable structure such as
retractable mechanism can also be tried to improve functionality. Further, a complete
working setup demonstrating all functionality is to be developed.
References
1. Lim WW (2002) Thermal properties of tire cords and their effects on post-curing inflation of
tires. Rubber Chem Technol 75(4):581
2. Zirbel SA, Tolman KA, Trease BP, Howell LL (2016) Bistable mechanisms for space
applications. PLoS ONE 12(11)
3. Howell LL (2001) Bistable mechanisms. In: Compliant mechanisms. Wiley, New York, pp
355–384, Wiley pp 355–357
4. Camescasse B, Fernandes A, Pouget J (2014) Bistable buckled beam and force actuation:
experimental validations. Int J Solids Struct 51:1750–1757
5. Taffetani M, Jiang X, Holmes DP, Vella D (2018) Static bistability of spherical caps. Proc Roy
Soc A 474(2213)
6. Brinkmeyera A, Santerb M, Pirreraa A, Weavera PM (2012) Pseudo-bistable self-actuated
domes for morphing applications. Int J Solids Struct 49(9):1077–1087
7. Howell LL, Magleby SP, Olsen BM (2013) Introduction to compliant mechanisms. In:
Handbook of compliant mechanisms. Wiley, pp 6–9
8. Opdahl PG, Jensen BD, Howell LL (1998) An investigation into compliant bistable mecha-
nisms. In: ASME design engineering technical conferences. Atlanta, GA
9. Pickarda JK, Carretero JA, Merlet J-P (2019) Appropriate analysis of the four-bar linkage.
Mech Mach Theory 139:237–250
10. Young WC, Budynas RG (1989) Elastic stability-formulas for elastic stability of plates and
shells. In: Roark’s formulas for stress and strain. McGraw-Hill, p 737
Stability Analysis of a Dual-Rate Haptics
Controller Using Discrete-Time
Root-Locus Method
1 Introduction
and damping coefficients that a haptic interface can stably implement is referred to
as Z-width [1]. Rendering stiff virtual walls while simultaneously maintaining the
stability of the interface has been recognized as main limitation of impedance-type
interfaces [1, 2].
High sampling rates are known to enhance the performance of haptic interfaces,
but the fundamental trade-off between performance and stability places a constraint
on the same [1, 2]. Nevertheless, several multi-rate control techniques have been
presented that overcome this performance-stability trade-off to a certain extent [3–
6]. In particular, Koul et al. [5] have proposed a dual-rate controller strategy for
improving the stability bounds of a one degree-of-freedom (1-DoF) haptic interface.
The stability analysis reported here pertains to such a class of controllers.
Haptic interfaces are desired to remain stable as they inevitably work in conjunc-
tion with a human operator. Stability of uniform-rate controllers has been studied
extensively in the literature [2, 7–12]; as opposed to this, multi-rate controllers have
received very limited attention [3–5]. Gil et al. [7], for instance, performed sta-
bility analysis of a 1-DoF haptic device by direct application of Routh’s criterion.
This work was later extended by Hulin et al. in [8] to consider the effect of the
human operator and by Gil et al. in [9] to study the effect of time delay and phys-
ical damping on stability, respectively. In [10] stability analysis of fractional-order
haptic controllers was presented. Mashayekhi et al. [11] carried out stability analysis
using frequency response function analysis; and more recently, the same group used
Lyapunov-Krazuvskii functional approach in [12]. Lee and Lee [3] and Dai et al.
[4] used Routh-Hurwitz method for obtaining stability bounds of their multi-rate
controllers, respectively.
Most of these works have however utilized analytical techniques that demand an
extensive mathematical treatment and in addition yield non-intuitive results. Fur-
thermore, based on the analytical conditions, evaluation of the stability limits is an
arduous task as a large number of iterations have to be carried out to find them.
Although, the iterative procedure could possibly be programmed, however, a knowl-
edge of the stable range of virtual stiffness and damping is still needed, which is
obviously not known beforehand. The classical discrete-time root-locus (DTRL)
method promises to be a viable solution to this issue. This method not only provides
a means for evaluating the stability bounds in a fast and reliable manner but also pro-
vides insight about the shapes of the stability curves obtained, owing to its graphical
nature.
In this work, we, therefore, carry out the stability analysis of a dual-rate hap-
tics controller, as proposed in [5], using the classical DTRL method. The controller
being a multi-rate and multi-variable sampled-data system, the direct application of
the DTRL method for analyzing its stability is accordingly not feasible. A sequence of
time-domain transformations and block-diagram manipulations are therefore applied
to formulate its characteristic equation in forms amenable to the DTRL analysis. Sta-
bility ranges of the virtual damping and virtual stiffness are subsequently obtained
and the Z-width contours are developed. DTRL method provides explanations con-
cerning the shapes of the Z-width contours.
Stability Analysis of a Dual-Rate Haptics Controller … 903
Dual-rate haptics controllers are characterized by the fact that they render virtual
wall damping and stiffness at two distinct (usually integer-related) sample rates, as
opposed to uniform-rate controllers that utilize a single sample rate for the purpose.
This decoupled sampling scheme enables rendering of stiff virtual walls and thereby
leads to performance enhancement [5].
The schematic of a typical 1-DoF haptics controller subjected to dual-rate sam-
pling scheme is depicted in Fig. 1. The rotational haptic device (plant) has been
represented by a second-order model, with moment of inertia I , and viscous damp-
ing b, respectively. The virtual environment, on the other hand, is modeled as a
spring-damper system in mechanical parallel with damping B (Nms/rad) and stiff-
ness K (Nm/rad). The virtual stiffness K is rendered at a sample period of T /N via
the stiffness loop, while the virtual damping B is rendered at T via the damping loop.
Although, N can theoretically take any value in the range [0, ∞], here we assume that
N ∈ Z+ . Furthermore, backward difference is used to obtain the velocity estimate.
The resultant torque τVE from the virtual environment is comprised of two compo-
nents; torque τK due to the virtual stiffness, and; the torque τB due to the virtual
damping, respectively. τVE is fed to the operator every T /N instants, after passing
through the zero-order-hold (ZOH) blocks.
Although an actual haptics controller contains several non-linear phenomena like
Coulomb friction, quantization, human operator dynamics, actuator saturation, and
unilateral constraint, the same are avoided here. The reason being: quantization
induced energy can be dissipated by Coulomb friction [13, 14]; and, in most cases
human operator adds stability to a haptic device by reinforcement of his own damping
[7, 8].
Haptic device
τH + θ
D(s)
Human - Stiffness
torque loop
τVE
T/N
+ τK 1-e-sT/N
s
K
+ Damping
loop
τB ZOHT/N
T
1-e-sT B(z2 - 1)
s Tz2
ZOHT Virtual wall
In this section, we employ the classical DTRL method for establishing the stability
limits of the dual-rate haptics controller (Fig. 1). For a given haptic device and a basic
sampling period of T , there are three variables; virtual damping B, virtual stiffness
K and ratio of sample periods N, that can be used as a feedback gain for the DTRL
method. However, due to the multi-rate nature of the controller it is not possible to
formulate its characteristic equation with N as a feedback gain. For each value of
N considered, two root-loci can thus be developed; one as a function of B and the
other as a function of K. In both the cases, prior to the application of the root-locus
method, the characteristic equation of the controller needs to be formulated in the
following form:
1 + αψ(z) = 0 (1)
where, α is the feedback gain, which in this case can be either B or K, and ψ(z) can
be thought of as the open-loop transfer function of the controller.
The presence of two sampling rates in the dual-rate haptics controller makes it imper-
ative to use two different discrete-time variables for the stiffness and damping loops.
Referring to Fig. 1, it can be seen that z1 and z2 have been used in this case, respec-
tively. To obtain the characteristic equation (with feedback gain B) that conforms
to Eq. 1, several time-domain transformations and some block-diagram algebra are
needed. A detailed derivation cannot be provided here owing to space constraints,
the procedure is however outlined below, and afterwards the main result is given.
The coefficients involved in Eq. 2, are the functions of the plant parameters and the
sampling rates of the controller, and the same are provided in the Appendix.
Based on Eq. 2, several root loci were obtained for different values of N at a fixed
value of T . Since the objective here was to determine the range of B for which a real
Stability Analysis of a Dual-Rate Haptics Controller … 905
−1.5 100
−1
B=4 900
−0.8
−1 800
80 −0.6
B = 7.5 700
Imaginary Axis
Imaginary Axis
−0.4
−0.5
60 −0.2 600
∞←B ∞←B
0 OX X 0O O X X 500
40 0.2 400
0.5 0.4 300
0.6 200
20
1
0.8
100
1
1.5 0 0
−15 −10 −5 0 Virtual −3 −2 −1 0 1 Virtual
Damping (B) Damping (B)
Real Axis Real Axis
(a) Root-locus plot for N = 1 (b) Root-locus plot for N = 2
−1 B = 18.75 −1
1800 B = 37.5 4500
−0.8
1600 4000
−0.6
1400 −0.5
Imaginary Axis
3500
Imaginary Axis
−0.4
1200 3000
−0.2
∞←B
0O O X X 1000 X 2500
0 O O X
0.2 800 2000
0.4 600 1500
0.5
0.6 400 1000
0.8
200 500
1 1
0 0
−1.5 −1 −0.5 0 0.5 1 Virtual −1.5 −1 −0.5 0 0.5 1 Virtual
Damping (B) Damping (B)
Real Axis Real Axis
(c) Root-locus plot for N = 5 (d) Root-locus plot for N = 10
positive value of K can be rendered, the value of K = 0.0001 Nm/rad was chosen.
The rlocus command of MATLAB was used to develop the root-loci, and the same
are depicted in Fig. 2. For all the cases, it can be seen that the system starts from being
stable as no pole lies outside the unit circle. However, as the value of B increases, the
poles start moving and at a certain critical value cross-over the unit-circle, as a result,
the system becomes unstable. Stability is never regained afterwards. This suggests
that for a particular value of N and T , virtual stiffness K can only be rendered for a
limited range of B. The value of B at which the poles cross-over, which is equivalently
the stable range of B, is higher for higher values of N , which is quite expected.
Following the similar steps as outlined in Sect. 3.1, the characteristic equation for
this case is given by:
906 S. Ganiny et al.
5 5
−1.5 x 10 −1.5 x 10
K = 4700 K = 6100
4.5 4.5
−1 −1
4 4
X
Imaginary Axis
Imaginary Axis
−0.5 3.5 −0.5 3.5
X 3 3
∞←K ∞←K ∞←K K→∞ ∞←K
0 O O O X 2.5 0 O O O X 2.5
X 2 2
0.5 0.5
1.5 X 1.5
1 1
1 1
0.5 0.5
1.5 0 1.5 0
−3 −2 −1 0 1 Virtual −3 −2 −1 0 1 Virtual
Stiffness (K) Stiffness (K)
Real Axis Real Axis
(a) Root-locus plot for B = 1 (b) Root-locus plot for B = 3
−1.5 5 −1.5 x 10
5
x 10
K = 2800
4.5 −1 X 4.5
−1
X 4
4
Imaginary Axis
Imaginary Axis
3 3
∞←K K→∞ ∞←K ∞←K K→∞ ∞←K
0 O O O X 0 O O O X
2.5 2.5
2 2
0.5 0.5
1.5 1.5
X 1 1
1 1 X
0.5 0.5
1.5 0 1.5 0
−3 −2 −1 0 1 Virtual −3 −2 −1 0 1 Virtual
Stiffness (K) Stiffness (K)
Real Axis Real Axis
(c) Root-locus plot for B = 6 (d) Root-locus plot for B = 8
In the previous section, the DTRL method was successfully applied to obtain the
stability bounds for a dual-rate haptics controller. The obtained results revealed that
for a particular sample period combinations B has a limited stable range, and for
each value of B, the stable value of K is different. It was also observed that upto a
particular value of B, the value of K has an increasing trend, but beyond that it starts
decreasing and eventually reduces to zero. These results suggest that the Z-width
boundary has the shape of a closed curve as there is a bound on the stable value
of both B and K. Figure 4 shows several such Z-width plots for different values of
T and N . A counterintuitive observation can be made based on these results; the
range of B is different for different values of N . This range should have otherwise
remained constant, as different values of N correspond to different sample rates of
the stiffness loop, and have no bearing on the sample rates of the damping loop. Upon
investigation, it was revealed that this behavior is not due to the root-locus method,
but is attributed to the sequence of transformation steps that were carried to obtain
the discrete-time characteristic equation in Sect. 3.1. In particular, we had employed
Tustin method for the discrete-to-discrete transformation (Step 3). A different choice
(say ZOH method) yields different results, but such a study is beyond the scope of
the present work. The DTRL method is extremely useful in the identification of the
most dominant poles of the controller and accordingly paves way for simplifying the
stability analysis via techniques like the model order reduction. This is particularly
helpful for haptics controllers that entail a temporal loop delay due to sampling effects
and other computational or communicational delays. An analysis of such controllers
will be the focus of our future work.
In this paper, an attempt has been made to study the stability characteristics of a
dual-rate haptics controller by extending the classical DTRL method. The method
empowers one to obtain the stability bounds of the controller quantitatively as well
4
10000 x 10
4
Virtual Stiffness − K (Nm/rad)
Virtual Stiffness − K (Nm/rad)
N=1
N=1
N=2 3.5 N=2
8000 N=5
N=5
N = 10 3 N = 10
6000 2.5
2
4000 1.5
1
2000
0.5
0 0
0 5 10 15 20 0 10 20 30 40
as to get a qualitative idea about the trends in the stability curves. DTRL is fast and
reliable and yields the stability boundaries with much ease as compared to the purely
analytical techniques like the Routh-Hurwitz or Jury’s method.
5 Appendix
The polynomial coefficients used in Eq. 2 (Section 3.1 ) are defined as:
p˜3 = Ih2 h3 ; p˜2 = I (h1 h3 + h2 h4 − h2 h3 ); p˜1 = I (h1 h4 − h1 h3 − h2 h4 )
q˜2 = T [2b2 (h1 − en h2 ) + IKh2 h3 ]; q˜1 = T [2b2 (h2 − h1 − en h1 + en h2 ) + IK(h1
h3 + h2 h4 )]; q˜0 = T [−2b2 (h2 − en h1 ) + IKh1 h4 ] where, h1 = N + 1; h2 = N − 1;
h3 = ẽh1 + êh2 ; h4 = ẽh2 + êh1 ;
−bT
ẽ = en − 1 + bT IN
; ê = 1 − (1 + bT
IN
)en ; en = e IN
The polynomial coefficients used in Eq. 3 (Section 3.2 s) are defined as:
pˆ2 = TIn2 ; pˆ1 = TIn1 ; pˆ0 = TIn0
qˆ3 = 2b2 Tm2 + BIn2 ; qˆ2 = 2b2 Tm1 + BI (n1 − n2 ); qˆ1 = 2b2 Tm0 + BI (n0 − n1 );
qˆ0 = −BIn0
where, m2 = N (1 − en ) + (1 + en ); m1 = −2(1 + en ); m0 = (1 + en ) − N (1 − en )
n2 = N 2 ẽ + N 2 ê − 2N ê − ẽ + ê; n1 = 2N 2 ẽ + 2N 2 ê + 2ẽ − 2ê; n0 = N 2 ẽ + N 2 ê +
2N ê − ẽ + ê.
References
1. Colgate JE, Brown JM (1994) Factors affecting the z-width of a haptic display. In: 1994 IEEE
international conference on robotics and automation, 1994. Proceedings. IEEE, New York, pp
3205–3210
2. Abbott JJ, Okamura AM (2005) Effects of position quantization and sampling rate on virtual-
wall passivity. IEEE Trans Robot 21(5):952–964
3. Lee K, Lee DY (2004) Multirate control of haptic interface for stability and high fidelity. In:
2004 IEEE international conference on systems, man and cybernetics, vol 3. IEEE, New York,
pp 2542–2547
4. Dai X, Zhang Y, Cao Y et al (2008) Stable multirate control algorithm for haptic dental training
system. In: International conference on intelligent robotics and applications. Springer, Berlin,
pp 27–35
5. Koul M, Manivannan M, Saha S (2017) Effect of dual-rate sampling on the stability of a haptic
interface. J Intell Robot Syst pp 1–13
6. Kim M, Lee DY (2019) Multirate haptic rendering using local stiffness matrix for stable and
transparent simulation involving interaction with deformable objects. IEEE Trans Ind Electron
7. Gil JJ, Avello A, Rubio A et al (2004) Stability analysis of a 1 DOF haptic interface using the
Routh-Hurwitz criterion. IEEE Trans Control Syst Technol 12(4):583–588
8. Hulin T, Preusche C, Hirzinger G (2008) Stability boundary for haptic rendering: influence
of human operator. In: IEEE, RSJ international conference on intelligent robots and systems
(2008) IROS 2008. IEEE, New York, pp 3483–3488
9. Gil JJ, Sanchez E, Hulin T et al (2009) Stability boundary for haptic rendering: influence of
damping and delay. J Comput Inform Sci Eng 9(1)
Stability Analysis of a Dual-Rate Haptics Controller … 909
10. Tokatli O, Patoglu V (2015) Stability of haptic systems with fractional order controllers. In:
2015 IEEE/RSJ international conference on intelligent robots and systems (IROS). IEEE, New
York, pp 1172–1177
11. Mashayekhi A, Behbahani S, Ficuciello F et al (2018) Analytical stability criterion in haptic
rendering: the role of damping. IEEE/ASME Trans Mechatron 23(2):596–603
12. Mashayekhi A, Behbahani S, Ficuciello F et al (2019) Delay-dependent stability analysis in
haptic rendering. J Intell Robot Syst pp 1–13
13. Diolaiti N, Niemeyer G, Barbagli F et al (2006) Stability of haptic rendering: discretization,
quantization, time delay, and coulomb effects. IEEE Trans Robot 22(2):256–268
14. Mashayekhi A, Boozarjomehry RB, Nahvi A et al (2014) Improved passivity criterion in haptic
rendering: influence of coulomb and viscous friction. Adv Robot 28(10):695–706
Design of Compliant Iris
Vinay Arora, Prakhar Kumar, Rajesh Kumar, and Jitendra Prasad Khatait
Abstract Stiffness-based approach has been used to design the compliant version
of mechanical iris. The rigid links of the mechanism are replaced by sheet flexures.
Iris has found use as grippers and also to carry out controlled opening of aperture.
This paper is based on formulating an estimate of the stiffness based on geometrical
parameters. Stiffness equation relates the input moment to the output deflection. The
equations are utilized to design based on the required use case.
1 Introduction
Mechanical iris has found use in many devices like iris diaphragm of a camera [2], iris
valves [4], etc. It is also used in optical setups to control the amount of light passing
through it [1]. Recently, iris structures have also found use in solar thermal receivers
where the amount of light is controller mechanically. It is inherently a single input–
output device, with many internal degrees of freedom. Many parts move relative to
one another, but the net output of interest is the radius of the central aperture [5].
However, there are some limitations of the mechanical iris as it is composed of many
plates moving relative to each other. So, a typical mechanical iris is composed of
various moving parts resulting in issues like friction and backlash. However, the com-
pliant iris is a single part mechanism which is based on large deflections of parts. The
compliant counterpart of the iris is composed of a single part. This paper is based on
prediction of bounds on stiffness of the iris system and a methodology to make design
decisions. Compliant iris has found use in gripping and holding the objects. It can be
used as an end effector gripper of a robotic arm. A closed-form expression of stiffness
is required to relate design the compliant iris. The novelty of the paper lies in provid-
Fig. 1 a Drawing, b 3D CAD model, and c 3D-printed prototype of the compliant iris
2 Iris Structure
3 Stiffness Analysis
In order to analyze the stiffness of the compliant iris arrangement in the compliant
direction, stiffness expression of curved beam has been derived in terms of geometric
dimensions and material properties and stiffness of multiple beams have been added
to obtain overall stiffness. For a general beam, stiffness is derived using Castigliano
theorem [3]. Castigliano’s theorem is one of the energy methods (based on strain
Design of Compliant Iris 913
energy) that can be used to solve a wide range of deflection problems. It relates the
deflection in an elastic body in any direction (‘a’) to the partial derivative of the strain
energy with respect to the input loads ‘Fi ’ in the direction ‘a’ (Eq. 1)
∂U
δi = (1)
∂Fi
M 2R
dU1 = dθ (2)
2EI
F 2R
dU2 = θ d θ (3)
2AE
MFθ
dU3 = − dθ (4)
AE
F 2R
dU4 = 1.2 n d θ (5)
2AG
914 V. Arora et al.
where ‘A’ is the cross-sectional area of the beam, ‘G’ is the shear modulus of the
material of the iris, ‘E’ is the elastic modulus of the material, and ‘I’ is the area
moment of inertia for the curved beam.
θ
U = dU1 + dU2 + dU3 + dU4 (6)
0
θ
δU R2 sin(θ )2 (sinθ )2 2(sinθ )2 1.2(cosθ )2
δ= = FR + − + d θ (7)
δF EI AE AE AG
0
bt 3
So, the stiffness of one beam of rectangular cross section A = bt and I = 12
; so, the
stiffness for each of the beams becomes Eq. 9.
bt 3
k= 2 2 (9)
3R(2θ − sin2θ ) R −tE /12 + t2
10G
For a compliant iris with six beams, we have overall torsional stiffness as 11 utilizing
the net deflection angle (α) and the net moment (Mt ). The net torsional stiffness is
given by Eq. 10.
Overall torsional stiffness (K) = 36 kriris
2
(10)
Design of Compliant Iris 915
where riris is the net radius of the whole iris (the bigger structure). The net analytical
expression of torsional stiffness of the iris ‘K’ is presented in Eq. 11.
2
12Eriris bt 3 θ 3
K= (11)
L3 (2θ − sin(2θ ))
However, the terms ‘R’ and θ are variables as the radius of curvature of the iris and
the angle spanned by the flexure keeps on changing. So, we need to express both
in terms of α which is the net rotation of the iris. The iris opens proportional to the
rotation angle α. We assume that the net change in the length of each of the flexures
is negligible. So, we have the geometric relation (Eq. 12).
Rθ = L (12)
Let αo be the initial rotation for the net iris output structure. Equation 12 is reduced
to Eq. 13.
α + αo θ
riris sin = Rsin (13)
2 2
4 Discussion
Although a closed-form expression for the net stiffness is not known, the bounds in
the stiffness can be used for design purposes. It should be known that the net stiffness
of the iris changes with change in the output angle. The upper bound and lower bound
for the expression of stiffness are expressed in Eqs. 15–16. The expression of stiffness
is increasing in θ for all values of L. So, the maximum stiffness occurs when each
of the flexures is at extreme values of θ . Let the value of θ is limited by θo . The
minimum stiffness occurs when θ = riris L
and R = riris . So, in case we have a required
stiffness during a complete operation of the opening of the aperture, then the various
parameters in maximum stiffness and minimum stiffness expressions can be varied.
2
12Eriris bt 3
Max stiffness = (15)
L3 (2θo
− sin2θo )
916 V. Arora et al.
12Ebt 3
Min stiffness = (16)
2L − riris sin r2L
iris
The variation of the stiffness values with respect to the length of each of the beams
as well as the variation of the analytical stiffness bounds with respect to the length of
the flexural beam and radius of the iris is presented in Figs. 4 and 5. With the rise in the
length of the flexural beam, both the upper bound and the lower bound of the stiffness
expression reduce. The bounds also tend to converge to a particular value. The graphs
presented in Figs. 4 and 5 correspond to the dimensions and material properties as
riris = 0.1 m, b = 0.005 m, t = 0.001 m, E = 3 Gpa, G = 800 Mpa. Figure 4 does
not have a constant iris radius, but the rest of the parameters are the same. With
the change in iris radius, the bounds on the stiffness widen, so the discrepancy in
design also raises. Hence, to ensure a proper design we might need to have large
length of beam flexure and small radius of the iris. So, the set of equations provide
a methodology to design compliant iris which is not yet available in the literature.
5 Conclusion
References
1 Introduction
recent days to achieve better machining performance. In this process, a very small
quantity of fine lubricant particles is sprayed between the tool–workpiece contact
surfaces. Das et al. [1] investigated on hard turning operation on EN 31 alloy steel,
and various output responses related to tool life, temperature generation and surface
finish were optimized. Upadhyay et al. [2] made a review of the MQL machining
to establish that the process is green and cost effective. A model on cutting force
for machining of Inconel 718 in MQL condition is proposed by Behera et al. [3].
Surface roughness of MQL nozzle and air pressure affects the lubricant droplet size
which is effectively reduced tool wear at the time of machining [4]. Effect of tool
wear rate of tungsten carbide tools and surface roughness of EN 31 alloy steel can
be predicted by developing models for getting best combination of design variables
[5]. In this current research, coated inserts have been used to turn EN 31 alloy steel
under two different cutting conditions, i.e. under dry and MQL conditions. Tool nose
radius has been varied.
2 Experimentation
The job selected for this research work is cylindrical EN 31 alloy steel of 16 ×
300 mm. The material is appropriate for the use in various fields of automobile,
e.g. axle, spindle, mandrels and also in moulding dies. These are being used by
turning operation mainly [6]. Coated carbide insert with 0.03 and 0.06 mm nose
radius has been used as a cutting tool. In this work, primary cutting parameters are
spindle speed (N), feed (f ), depth of cut (d). The levels and the values of cutting
parameters have been selected by prior experimentation within the range in the
machine tool. This experiment has been carried out on the lathe machine (maximum
spindle speed 1200 rpm) in dry condition and liquid CO2 -assisted minimum quantity
lubrication. Figure 1 shows the photographic view of experimental setup. On the basis
of preliminary experiments, the following process parameters with the level settings
are shown in Table 1.
Table 1 Experimental
Input parameters Level-1 Level-2 Level-3
process parameters and their
level settings Spindle speed (r.p.m.) 500 – –
Feed rate (mm/min) 20 40 60
Depth of cut (mm) 0.1 0.3 0.5
Total 18 numbers of experiments with two replicates for each nose radius were carried
out. The spindle speed was kept constant at 500 r.p.m., but feed rate and depth of
cut were varied using the above level settings. The graphs are drawn (as depicted in
Figs. 2, 3 and 4) based on the experimental results. The graphs not only illustrate the
result but also provide the best regression equations as shown in the graphs.
It has been observed from Fig. 2 that MRR is maximum at 40 mm/min feed rate in
MQL medium for both the tool nose radii, and it reduces with other feed rate settings.
This is due to the fact the generated heat when the feed rate is 40 mm/min get much
time to dissipate, and at the same time, the cool MQL medium helps to reduce the
temperature of cutting. In case of 60 mm/min feed rate, the generated heat does not
find much time to dissipate, and so, the MRR is also lower.
The temperature at the tool was measured by placing the thermocouple at two
different positions, i.e. at principal and auxiliary cutting edge, and it has been
observed that the temperature at principal cutting edge (Tp) is quite higher than
that of auxiliary cutting edge (Ta). This is due to the dissipation of heat through the
tool material. The temperature generation in both the edges for dry medium is greater
in comparison with MQL medium and well enough to decrease the tool life.
922 S. Misra et al.
a 50
45 MRR_MQL AT 20 MM/MIN FEED RATE
40
35 y = -1.49x2 + 12.66x - 4.57 MRR_MQL AT 40 MM/MIN FEED RATE
MRR (gm/min)
R² = .90
30
25 MRR_MQL AT 60 MM/MIN FEED RATE
20
y = 2.61x2 - 6.93x + 10.54
15 R² = .91 Poly. (MRR_MQL AT 20 MM/MIN FEED RAT
10
y = 4.005x2 - 14.245x+14.6
5 R² = .92
Poly. (MRR_MQL AT 40 MM/MIN FEED RAT
0
0.1 0.3 0.5 0.7 0.9
Poly. (MRR_MQL AT 60 MM/MIN FEED RAT
depth of cut (mm)
b 25 y = 1.7892x3 - 18.526x2 + 55.445x - 32.948 MRR _ MQL AT 20 MM/MIN FEED RATE
R² = 0.7469
20
MRR _ MQL AT 40 MM/MIN FEED RATE
MRR (gm/min)
50
MRR_DRY AT 60 MM/MIN
FEED RATE
40
30 Poly. (MRR_DRY AT 20
MM/MIN FEED RATE)
20 y= 3.3133x3 - 33.087x2 + 95.29x - 58.294 R² = 0.85
Poly. (MRR_DRY AT 40
10 MM/MIN FEED RATE)
Poly. (MRR_DRY AT 60
0 MM/MIN FEED RATE)
0.1 0.3 0.5 0.7 0.9
y = -3.32x2 + 20.59x - 16
depth of cut (mm) R² = 0.89
Fig. 2 Analysis for a MRR at 0.03 mm nose radius in MQL, b MRR at 0.06 mm nose radius in
MQL, c MRR at 0.03 mm nose radius in dry medium, d MRR at 0.06 mm nose radius in dry medium,
e comparison of MRR at dry and MQL conditions in 0.03 mm nose radius and f comparison of
MRR at dry and MQL conditions in 0.06 mm nose radius
Investigation on the Effects of Nose Radius … 923
25
MRR _ DRY AT 40 MM/MIN FEED
y = 1.6367x3 - 18.536x2 + 60.007x - 36.452 R² = 0.9267
MRR ( gm/min)
RATE
20 y = -1.2375x3 + 11.782x2 - 32.11x + 35.348 R² =
0.984
15 MRR _ DRY AT 60 MM/MIN FEED
RATE
10
Poly. (MRR _ DRY AT 20 MM/MIN
FEED RATE)
5
Poly. (MRR _ DRY AT 40 MM/MIN
0 FEED RATE)
0.1 0.3 0.5 0.7 0.9
y = 1.5325x3 - 15.407x2 + 44.7x - 28.126 R² = 0.8265
depth of cut (mm) Poly. (MRR _ DRY AT 60 MM/MIN
FEED RATE)
40 MRR_MQL AT 40 MM/MIN
FEED RATE
20 Poly. (MRR_DRY AT 40
MM/MIN FEED RATE)
y = -1.49x2 + 12.66x - 4.57
0
0.1 0.3 0.5 0.7 0.9 R² = 0.98
Poly. (MRR_MQL AT 40
depth of cut (mm)
MM/MIN FEED RATE)
f 25
y = -1.2375x3 + 11.782x2 - 32.11x + 35.348 R² MRR_DRY AT 40 MM/MIN
= 0.984 FEED RATE
20
15 MRR_MQL AT 40 MM/MIN
MRR(gm/min)
FEED RATE
10
Poly. (MRR_DRY AT 40
MM/MIN FEED RATE)
5
y = 2.450x4 - 27.61x3 + 102.9x2 - 144.0x +
72.91 R² = 0.98
0 Poly. (MRR_MQL AT 40
MM/MIN FEED RATE)
0.1 0.3 0.5 0.7 0.9
-5
depth of cut (mm)
Fig. 2 (continued)
924 S. Misra et al.
Tp_MQL AT 20 MM/MIN
a 300 y = 16.65x2 - 50.45x + 87.3 R² = 0.92
FEED RATE
50
Tp (oC)
Fig. 3 Analysis for a Tp at 0.03 mm nose radius in MQL, b Tp at 0.06 mm nose radius in MQL, c Tp
at 0.03 mm nose radius in dry medium, d Tp at 0.06 mm nose radius in dry medium, e comparison
of Tp at dry and MQL conditions in 0.03 mm nose radius and f comparison of Tp at dry and MQL
conditions in 0.06 mm nose radius
Investigation on the Effects of Nose Radius … 925
80
Tp_DRY AT 60 MM/MIN FEED RATE
Tp(oC)
200
Tp_MQL AT 40 MM/MIN FEE
Tp(oC)
150 RTAE
100
Poly. (Tp_DRY AT 40 MM/MIN
FEED RATE)
50
f 140
Tp_DRY AT 40 MM/MIN FEE
RATE
120
y = 10.18x4 - 120.7x3 + 483.0x2 - 746.1x + 445.6 R² =
100 0.86
Tp_MQL AT 40 MM/MIN FEE
80 RATE
Tp(oC)
60
Fig. 3 (continued)
926 S. Misra et al.
4 Conclusions
In the research work, the turning at dry and MQL conditions with varying tool nose
radius has been employed. The experiments have been set with full factorial design
of experiments, and the main objective is to investigate on the improvement on
machinability of EN 31 alloy steel at various machining conditions with variation
of tool geometry. The following observations were made from the above research
work:
1. MRR has dependency on the cooling medium, but in dry medium, the MRR
increases with D.O.C. and feed rate as far as EN 31 alloy steel is concerned.
2. MRR is higher in case of machining in MQL condition irrespective of nose
radius at 500 rpm, 40 mm/min feed rate and 0.5 mm D.O.C.
3. Temperature generation directly affects the tool life, and with increase of tool
nose radius, it is increased.
4. Generated heat is carried out by the cooling medium, so the temperature
generation is quite lower in MQL medium compared to that in dry medium.
The research work has lots of industrial use as far as EN 31alloy steel material is
concerned, and there is a scope to carry out analytical work on the material.
References
1. Anshuman D, Kumar PT, Hotta B, Bhushan B (2018) Statistical analysis of different machining
characteristics of EN-24 alloy steel during dry hard turning with multilayer coated cermet inserts.
Measurement 134:123–141
2. Vikas U, Jain PK, Mehta NK (2013) Machining with minimum quantity lubrication: a step
towards green manufacturing. J Mach Mach Mater 13(4):349–371
3. BarczakL M, Batako ADL, Morgan MN (2010) A study of plane surface grinding under
minimum quantity lubrication (MQL) conditions. Int J Mach Tools Manuf 50(11):977–985
4. Khan WA, Hoang NM, Bruce T, Hung Wayne NP (2018) Through-tool minimum quantity
lubrication and effect on machinability. J Manuf Process. https://doi.org/10.1016/j.jmapro.2018.
03.047
5. Laxman A, Hameedullahb M (2015) Simultaneous optimization of multiple quality character-
istics in turning EN-31 steel. Mater Today: Proc 2:2640–2647
6. Behera BC, Ghosh S, Rao PV (2017) Modelling of cutting force in MQL machining environment
considering chip tool contact friction. Tribol Int. https://doi.org/10.1016/j.triboint.2017.09.015
Battery Performance Analysis of Static
Temperature Variations for Medical
Environment
Abstract Improving battery life is very essential need in all battery-based applica-
tions to improvise the component operation for longer period of time. But, the battery
is easily influenced by surrounding temperature which leads to fast discharge and
reduction of charge capacity. To analyze these problems, thermography instrument is
used which provides better temperature measurement as well as heat flux variations
in the battery. The temperature is varied at 10 °C interval for each 10 successful
charging and discharging cycles. While increasing the temperature at 60 °C, the
charging period is reduced to 2 h 10 min and discharging time to 2 h 18 min. Beyond
this, continuous exposure of battery at different temperature makes the battery to get
damaged and broken. Finally, the scanning electron microscope (SEM) images are
analyzed to understand the temperature influence on anode and cathode plates.
1 Introduction
In medical field, the battery backup is very important to monitor the patient’s health
status continuously. Nowadays, batteries are playing a major role in medical field
to monitor the patient continuously from remote locations with the help of sensors
through sensor communications where a regular power supply is unable to meet the
demands and result in failure. Due to technology advancements, the doctors can able
to easily access the patient’s health status directly from the hospital. All remotely
operated components need power supply to monitor the health condition of patient.
Hence, increasing the lifetime of battery is essential since it is the main source of
B. Banuselvasaraswathy (B)
Department of Electronics and Communications Engineering, Sri Krishna College of Technology,
Coimbatore, India
R. Vimalathithan
Department of Electronics and Communications Engineering, Karpagam College of Engineering,
Coimbatore, India
power. So, maintaining the condition of battery is very essential to enhance the
device life and performance. But, the battery lifetime and performance varies based
on state of charge and environmental conditions. The major influence on the battery
performance is environmental temperature. Therefore, it is necessary to understand
the temperature influence on battery to predict the accurate state of charge and health
of the battery. Batteries are highly liable to temperature variations, which change
the effective charge capacity. In this circumstance, it is very difficult to estimate the
behavior of batteries for longer periods.
The overall performance of a battery depends on the individual cell performance.
As cell temperature varies from one to another, the charging and discharging ratios
also vary for each cell at every cycle. If the temperature difference prolongs, the cell
module becomes imbalance, thereby reducing the cell performance [1]. Hu et al.
[2] validated the performance of temperature on lead–acid battery and observed
that the charge acceptance rate gets reduced while decreasing the temperature at
sub-zero level. In embedded hardware, the operation of sensor node is significantly
influenced by the thermal effect; specifically, the batteries get affected at different
temperature variations [3, 4]. The electrochemical cells exhibit different effective
charge capacities based on temperature [5, 6]. Hence, it is challenging to determine its
behavior over time, especially the energy-aware algorithms such as voltage levels and
state of charge (SoC) [7–9]. Many battery models are used in wireless sensor networks
(WSNs), but most of the models do not consider the intrinsic behavior of battery such
as thermal effects which can deteriorate the lifetime of battery [10–13]. Dougal et al.
[14] introduced a dynamic model for Li-ion batteries which is widely utilized in
portable systems. Streza et al. [15] analyzed the amount of heat being dissipated in
electrode during discharge cycle by using IR camera with high resolution. In [16], a
small variation in thermal homogeneities is detected easily. Additionally, variations
in heat evolution are followed during discharging and charging processes.
Thus, in this paper, the effect of temperature variation on battery has been analyzed
by utilizing thermography imaging system. This method is very useful to identify
the heat flux of cell as well as the surroundings of battery. Additionally, SEM images
were taken to analyze the anode and cathode plate integrity.
2 Experimentation
The experimental setup is utilized to carry out the observations which are discussed
in the following section.
2.1 Implementation
The battery used in this study is lead–acid battery and dimension of battery is 151 ×
50 × 94 mm, battery capacity is 12 V, 7 Ah and cycle use of 14.5 V, 14.9 V at 25 °C.
Battery Performance Analysis of Static Temperature Variations … 931
Heating coils are used at both sides of the chamber to generate required amount of
heat. At this temperature condition, batteries are frequently charged and discharged.
K-type thermocouple sensors are used to measure the surrounding temperature of
the battery with an accuracy of ±2.2 °C. The results obtained from thermocouple
are processed with NI DAQ card. The final temperature is fed to LabView software
to measure the temperature of the chamber continuously as shown in Fig. 1. Infrared
thermography (IRT) is a non-contact-type surface utilizing gray scale (infrared
color images) to measure temperature. The IRT image with brighter area typically
represents a high-temperature zone, and darker area indicates a lower temperature
distribution [17].
2.2 Thermography
cardboard box was kept above the battery and stand. The camera was placed on
tripod situated outside the enclosure and focused in such a way that it can view only
the battery placed inside the box.
Field emission scanning electron microscope (FE-SEM) images are used to analyze
the change in alloy surface due to electrodeposition and corrosion during charging
and discharging [9]. In this study, FE-SEM, Thermo-Fisher FEI Quanta 250 FEG
is available in Sophisticated Instrumentation Center at VIT–Vellore. To prepare the
FE-SEM sample, the anode and cathode plates were removed from the battery. The
plates are washed with distilled water to naturalize the pH content. The samples are
dried under vacuum at 50 °C for one full day. The dried samples were collected and
made into triangular shape with a thickness of 2 mm.
The results obtained using thermography image analysis and scanning electron
microscope which were discussed in the following sections.
measured using thermography images at initial charging and fully charged condition.
Figure 2a shows the initiation of charging above room temperature (35.4 °C), and
Fig. 2b shows the maximum charging capacity of battery. The temperature difference
between the low charge capacity and high charge capacity is around 2 °C. Meanwhile,
the time taken to reach the maximum capacity is 3 h and 43 min. Once the battery is
completely charged, the battery is kept at resting state for about 30 min to reach the
stable condition.
After the resting period, the discharge is taken into consideration with constant
load. Figure 3a shows the temperature of battery at resting position. The temperature
gets reduced due to 30 min resting period. As the battery gets discharged, the temper-
ature starts to increase due to ion exchange and internal resistance of the battery. If
the battery’s internal resistance is high, the heat generation of the internal cell of
the battery is also observed at high range. Figure 3b shows the condition of battery
capacity at minimum level, and the temperature is measured at 39.6 °C. The heat
dissipation rate varies from cell to cell, and it mainly depends on the cell capacity,
cyclic period and temperature of the cell. This cycle of operation is carried out for
50 full charging and discharging cycles of battery.
(a) (b)
Fig. 2 Charging of battery at room temperature a temperature of 35.4 °C at low capacity of the
battery b temperature of 37.4 °C at full capacity of the battery
(a) (b)
Fig. 3 Discharging of battery at room temperature a 36.8 °C initial discharging at room temperature,
b 39.6 °C at high load discharging
934 B. Banuselvasaraswathy and R. Vimalathithan
(a) (b)
(a) (b)
Fig. 5 a Initial condition of discharge, the temperature range is 54 °C, b at fully discharged condition
of discharge and also reduces the battery capacity which is clearly depicted from the
IR images.
High-resolution SEM images were taken to analyze the integrity of anode and cathode
plates in the scale of 30 µm range. The samples for imaging purpose are taken in two
different areas of anode and cathode plates. The first sample is considered from the
portion which is close to the electrical connector, and the second sample is obtained
from the bottom side of each electrode. These kinds of samples help to identify the
plate integrity at two different places.
From Fig. 6, the orientations of anode plate at two different portions are clearly
examined. It is evident that the temperature profile has an influence on the plate
surface. Exposing the battery to higher temperature region will fasten the chemical
(a) (b)
Fig. 6 a Anode plate analysis at near the terminal region b cathode plate analysis at near the
terminal region
936 B. Banuselvasaraswathy and R. Vimalathithan
(a) (b)
Fig. 7 a Anode plate analysis—far from the terminal region b cathode plate analysis—far from
the terminal region
4 Conclusion
This paper discusses about the thermal influence on the battery’s anode and
cathode materials at different temperature conditions. IR thermography analysis was
performed to measure the cell and surrounding temperature of the battery. From the
thermography analysis, it was observed that charging and discharging of battery
at higher temperature cause a thermal effect which shortens the lifetime of the
Battery Performance Analysis of Static Temperature Variations … 937
battery. Additionally, the sensor network lifetime and transmission efficiency also
gets reduced. By maintaining at a room temperature, the battery’s thermal influence
was observed only for a minimal range. At this condition, battery can able to with-
stand for long period of time ensuring better data transmission in WSN. Likewise,
SEM analysis shows high-temperature influence on battery’s anode and cathode
plates. The material removal rate from anode plate is observed high. This nature of
material removal is varied at different areas of plates.
References
1. Pesaran AA, Vlahinos A, Burch SD (1997) Thermal performance of EV and HEV battery
modules and packs. National Renewable Energy Laboratory
2. Hu HY, Xie N, Wang C, Wu F, Pan M, Li HF, Wu P, Wang XD, Zeng Z, Deng S, Wu M (2019)
Enhancing the performance of motive power lead-acid batteries by high surface area carbon
black additives. Appl Sci 9(1):186
3. Bannister K, Giorgetti G, Gupta SK (2018) Wireless sensor networking for hot applications:
effects of temperature on signal strength, data collection and localization. In: Proceedings of
the 5th workshop on embedded networked sensors, pp 1–5
4. Boano CA, Tsiftes N, Voigt T, Brown J, Roedig U (2009) The impact of temperature on outdoor
industrial sensornet applications. IEEE Trans Ind Inf 6(3):451–459
5. Chen M, Rincon-Mora GA (2006) Accurate electrical battery model capable of predicting
runtime and IV performance. IEEE Trans Energy Convers 21(2):504–511
6. Jaguemont J, Boulon L, Venet P, Dubé Y, Sari A (2015) Lithium-ion battery aging experiments
at subzero temperatures and model development for capacity fade estimation. IEEE Trans Veh
Technol 65(6):4328–4343
7. Hörmann LB, Glatz PM, Hein KB, Weiss R (2012) State-of-charge measurement error simu-
lation for power-aware wireless sensor networks. In: 2012 IEEE wireless communications and
networking conference (WCNC). IEEE, pp 2209–2214
8. Xie D, Wei W, Wang Y, Zhu H (2013) Tradeoff between throughput and energy consumption
in multirate wireless sensor networks. IEEE Sens J 13(10):3667–3676
9. Wang CF, Shih JD, Pan BH, Wu TY (2014) A network lifetime enhancement method for sink
relocation and its analysis in wireless sensor networks. IEEE Sens 14(6):1932–1943
10. Haase J, Molina J, Dietrich D (2011) Power-aware system design of wireless sensor networks:
power estimation and power profiling strategies. Trans Ind Inf 7:601–613
11. Stehlík M (2011) Comparison of simulators for wireless sensor networks. M.S. thesis, Faculty
Information, Masaryk University, Brno, Czech Republic
12. Musznicki B, Zwierzykowski P (2012) Survey of simulators for wireless sensor networks. Int
J Grid Distrib Comput 5(3):23–50
13. Pereira RM, Ruiz LB, Ghizoni MLA (2015) MannaSim: A NS-2 extension to simulate wireless
sensor network. Proc Int Conf Netw 2015:95–101
14. Gao L, Liu S, Dougal RA (2002) Dynamic lithium-ion battery model for system simulation.
IEEE Trans Compon Packag Technol 25(3):495–505
15. Streza M, Nuţ C, Tudoran C, Bunea V, Calborean A, Morari C (2016) Distribution of current in
the electrodes of lead-acid batteries: a thermographic analysis approach. J Phys D Appl Phys
49(5):055503
16. Giess H (1997) Investigation of thermal phenomena in VRLA/AGM stationary lead/acid
batteries with a thermal video imaging system. J Power Sources 67(1–2):49–59
17. Balaras CA, Argiriou AA (2002) Infrared thermography for building diagnostics. Energy Build
34(2):171–183
18. Lin C, Tang A, Mu H, Wang W, Wang C (2015) Aging mechanisms of electrode materials in
lithium-ion batteries for electric vehicles. J Chem
Derivation of the Rotation Matrix for
an Axis-Angle Rotation Based
on an Intuitive Interpretation
of the Rotation Matrix
Abstract In this paper, we present the derivation of the rotation matrix for an axis-
angle representation of rotation. The problem is of finding out the rotation matrix
corresponding to the rotation of a reference frame, by a certain angle, about an arbi-
trary axis passing through its origin. The axis-angle representation is particularly
useful in computer graphics and rigid body motion. We have used an intuitive inter-
pretation of the rotation matrix for this derivation. The intuitive interpretation is as
follows: the columns of the rotation matrix are the coordinate axes unit vectors of
the rotated frame as seen from the fixed frame. This interpretation of the rotation
matrix helps in quick computation of the rotation matrices required to obtain the
required rotation matrix. The required rotation matrix can be computed from just
two rotation matrices which are easy to find and intuitive to understand. The deriva-
tion presented in simpler to understand than presented in most books and can be
grasped with minimum geometric visualization.
1 Introduction
solutions like Rodrigues’ formula [3] exist but its derivation requires the knowledge
of exponential maps. The derivation is not easy to understand at an undergraduate
level. In our work, we have proposed a method by which the rotation matrix can be
derived in the matrix domain without the knowledge of exponential map or other con-
cept. Very less or no graphical construction needed. The proposed method requires
lesser number of computations, in terms of matrix multiplications, as compared to
methods in standard textbooks [4, 5].
We denote this rotation matrix between frame A and A as AA R. This rotation matrix
transforms the expression of a vector in frame A to that in frame A. We go through
the following steps to obtain the required rotation:
1. Consider a frame K whose z-axis is along the vector K̂.
2. Consider that the frame K and the frame A are rigidly connected. So, both the
frames rotate together.
3. When the frame K is rotated about its z-axis, the frame A also rotates with it.
By this rotation, K changes to the new frame K’ and frame A changes to a new
frame which is our required frame A .
Let us denote the rotation matrix from frame K and A as KA R. A rotation about
the z-axis of frame K gives the new frame K . The net rotation matrix from frame
K to frame A will be:
K R = K R R z K (θ )
A A
(1)
Since we consider that the frame A and the frame K are rigidly connected, when
K gets rotated to K , frame A gets rotated to A . One can also say that the rotation
matrix from K to A remains the same for the rotation between K and A . Hence,
we can write:
A
K R =K R
A
(2)
Now using Eqs. 1 and 2, we can obtain our required rotation matrix AA R as:
A
A R = KA R KA RT (3)
Or
A
A R = KA R Rz K (θ ) KA RT (4)
In Eq. 4: ⎡ ⎤
cosθ −sinθ 0
Rz K (θ ) = ⎣ sin θ cosθ 0⎦ (5)
0 0 1
For computing KA R, we use the interpretation of the rotation matrix that a column of
the rotation matrix KA R represents a unit coordinate axes vector of frame K expressed
in terms of the unit coordinate axis vectors of frame A. If we say that î, ĵ, k̂ are unit
coordinate axes vectors and we use a subscript to denote the frame for which they
are the unit coordinate axes vectors, then î K , ĵK , k̂ K are unit vectors for frame K and
î A , ĵ A , k̂ A are unit vectors for frame A. Then the first column of the rotation matrix
A
K R represents î K expressed in î A , ĵ A , k̂ A . Similarly for second and third column.
942 R. K. Hota and C. S. Kumar
From the above discussion, we can see that for the rotation matrix KA R the last
column will simply be K̂ = [k x k y k z ]T which is k̂ K expressed in î A , ĵ A , k̂ A . Now
for the frame K we need to specify the x and y-axis. We know that the x-axis will
be perpendicular to k̂ K . For getting the x-axis, we chose an arbitrary vector (say
[0 0 1]T )and take its cross product with the z-axis:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
kx 0 0 −k z ky 0 ky
⎣k y ⎦ × ⎣0⎦ = ⎣ k z 0 −k x ⎦ ⎣0⎦ = ⎣−k x ⎦ (6)
kz 1 −k y k x 0 1 0
From Eq. 6, we obtain the unit vector along the x-axis of frame K expressed in the
frame A as:
⎡ ⎤
k y / k x2 + k 2y
⎢ ⎥
î K = ⎢ ⎥
⎣−k x / k x2 + k 2y ⎦ (7)
0
Now, if we put the values from Eqs. 9 and 5 into Eq. 4, we can obtain the terms
of the required rotation matrix AA R:
⎡ ⎤
r11 r12 r13
⎣
A R = r 21 r 22 r 23
A ⎦ (10)
r31 r32 r33
Derivation of the Rotation Matrix for an Axis-Angle … 943
We will show the detailed steps, for simplicity, for only the first term here.
(k 2y + k x2 k z2 )cosθ
r11 = + k x2
k x2 + k 2y
(k 2y + k x2 (1 − k x2 − k 2y ))cosθ
= + k x2
k x2 + k 2y
((k 2y + k x2 ) − k x2 (k x2 + k 2y ))cosθ
= + k x2
k x2 + k 2y
= cosθ − k x2 cosθ + k x2
= k x2 (1 − cosθ ) + cosθ (11)
−k x k y cosθ − k x2 k z sinθ − k 2y k z sinθ + k x k y k z2 cosθ
r12 = + kx k y
k x2 + k 2y
= k x k y (1 − cosθ ) − k z sinθ (12)
r13 = k x k z cosθ + k y sinθ (13)
(−k x cosθ + k y k z sinθ )k y + (k x sinθ + k y k z cosθ )k x k z
r21 = + kx k y
k x2 + k 2y
= k x k y (1 − cosθ ) + k z sinθ (14)
(−k x cosθ + k y k z sinθ )(−k x ) + (k x sinθ + k y k z cosθ )k y k z
r22 = + k 2y
k x2 + k 2y
= k 2y (1 − cosθ ) + cosθ (15)
(k x sinθ + k y k z cosθ )(−k x2 − k 2y )
r23 = + k y kz
k x2 + k 2y
= k y kz(1 − cosθ ) + k x sinθ (16)
r31 = k x k z (1 − cosθ ) − k y sinθ (17)
r32 = k x k y (1 − cosθ ) + k x sinθ (18)
r33 = k z2 (1 − cosθ ) + cosθ (19)
The terms of the required rotation matrix are given in Eqs. (11) through (19).
The complete rotation matrix is:
⎡ ⎤
k x2 νθ + cθ k x k y νθ − k z sθ k x k z cθ + k y sθ
A R =
A ⎣ k x k y νθ + k z sθ k 2y νθ + cθ k y kzνθ + k x sθ ⎦ (20)
k x k z νθ − kysθ k x k y νθ + kxsθ k z2 νθ + cθ
944 R. K. Hota and C. S. Kumar
3 An Example
Following is the example for computing the rotation matrix corresponding to axis-
angle representation of rotation. The expression for the rotation matrix in the simplest
form is:
A R = K R R z K (θ ) K R
A A A T
(21)
We need to compute two rotation matrix. One of them is the standard rotation
about Z-axis given by Eq. 5. The second is KA R which is given by Eq. (9).
√ √ √
Example Compute the rotation of a matrix about axis [2/ 14, 3 14, 1 14]T by
an angle π/6.
And, ⎡ ⎤
0.904 −0.076 0.420
A R =
A ⎣ 0.191 0.952 −0.238⎦ (24)
−0.381 0.296 0.875
4 Conclusions
In this work, we have derived the rotation matrix corresponding to axis-angle rep-
resentation of rotation. The method used employs an intuitive interpretation of the
rotation matrix which is that the columns of the rotation matrix represent the coor-
dinate unit vectors of the rotated frame expressed in the coordinate unit vectors of
the fixed frame. Our method requires knowledge of two rotation matrices. One of
them is the standard rotation matrix about the z-axis, and the other is the rotation
matrix between the initial frame and the frame attached to the axis of rotation. The
second rotation matrix is obtained by performing two cross products. So, effectively
the computations we are performing are two cross-products and two matrix multipli-
cations. Further, the rotation matrix is intuitively constructed directly from the data
given, that is, the unit vector of the axis of rotation.
Derivation of the Rotation Matrix for an Axis-Angle … 945
References
1. Euler L (1776) Formulae generales pro translatione quacunque corporum rigidorum. Novi Com-
mentarii academiae scientiarum Petropolitanae 189–207
2. Bob P, Richard P, Stephen R (2009) A disorienting look at Euler’s theorem on the axis of a
rotation. Am Math Monthly 116(10):892–909
3. Murray Richard M (2017) A mathematical introduction to robotic manipulation. CRC Press,
Boca Raton
4. Craig JJ (2009) Introduction to robotics: mechanics and control, 3rd ed. Pearson Education,
India (2009)
5. Fu KS, Gonzalez RC, George Lee CS (1987) Robotics: control sensing. Vis. Tata McGraw-Hill
Education
Resolving Hyper-Redundant Planar
Serial Robots to Ensure Grasp
1 Introduction
Caging and grasping planar objects have been a widely addressed problem in the
literature. Caging is referred as the process of restricting motions of an object using
a manipulator. A cage set [8] is defined as a set of contacts such that an object cannot
move out of the contact, though it still has some unrestricted motions. Process of
caging has been extended to grasping and then to manipulation of the objects [9]. The
methods to cage and then to manipulate the objects have been explored using various
methods like using multiple mobile robots [12] and [7]. Various other methods have
been studied in detail in the literature. Algorithms to search for the objects in order
to cage and grasp them have been developed. Optimal contact points are important
to ensure proper grasp. Two of such methodologies are presented in [1] and [2].
We intend first to cage the object and then grasp it using hyper-redundant serial
robots. We use a hyper-redundant serial robot to search for a single object within a
known boundary and grasp it. There should be only a single convex-shaped object
in the boundary to grasp it. Hyper-redundant serial planar manipulators have been
utilized to carry out operations imitating behavior of a snake. One such operation is
grasping objects [3]. It is known that a hyper redunant serial robot is comprised of
much more degrees of freedom in the joint space than in the end effector space. So, a
methodology needs to be developed to resolve the redundancy such that resolution is
used to grasp an object. One such methodology is utilization of “Tractrices.” Ghosal
et al. [11] showed that resolving redundancy using tractrix curves can be used to
avoid obstacles in a plane and also to carry out operations like tying a knot. Here, we
show that the method of resolution can be used to cage a rigid object in plane, under
knowledge of a bound containing the object boundary and the object location.
“Tractrices” are historical curves which are used to denote the motion of the non-
driven end of a geodesic curve on a surface when the other end of the geodesic
curve is driven along a known trajectory. Consider a particle moving on a frictional
surface. The surface need not be planar and can be a generalized one. A curve is
attached to the particle such that it represents a geodesic on the surface the particle
is moving. Consider that the surface is modeled using parameters (u(s), v(s)) with
the a function f(u,v) mapping the surface parameters to the Cartesian coordinate in
R3 . A geodesic curve is made by a particle on a surface when it starts at a constant
velocity and moves exactly under normal acceleration. Mathematically, the statement
is represented by Eq. 1. Consider the geodesic curve to be γ (s), which follows
a second-order differential equation. Detailed analysis of the rise of the geodesic
curve on surfaces is found in [5]. It is also the shortest curve between two points on
the surface.
d2 γ α β
α dγ dγ
μ
+ βμ =0 (1)
ds 2 ds ds
α
where βμ is written as per Einstein’s convention for brevity and γ ν represents the
ν parameter representing coordinates on the manifold. For a surface, the set γ
th
contains (u(s), v(s)) and α1,2 , β1,2 , μ1,2 are dummy variables with γ 1 = u(s) and
γ 2 = v(s). A detailed explanation can be found in [10]. The tractrix curve is the
locus of the endpoint of this curve when the initial point is moved along a curve. By
definition of geodesic, in case the surface happens to be a plane, then the geodesic
between any two points is the line joining the two points. Evolution of tractrices
on different submanifolds is explained in [6], but we will restrain only to planar
tractrices. Numerous equivalent definitions are presented in [4].
In this paper, we use the fact that when one end of a straight line is moving on a
frictional surface, the instantaneous velocity of the other end is along the link length.
The statement is mathematically represented in Eq. 2.
X head − X tail Ẋ tail
= (2)
Yhead − Ytail Ẏtail
Resolving Hyper-Redundant Planar … 949
Fig. 2 a Tractrix curve to a quarter circle tractor b Trajectory of a redundant robot trying to squeeze
an object with a known bound based on developed method
where (X head and Yhead ) are the start coordinates, and (X tail , Ytail ) represents the end
coordinates of the line. An example tractor–tractrix combination is shown in Fig. 2a.
An example of the tractrix curve on the plane surface (straight line geodesics) with
the head following a straight line is shown in Fig. 1.
3 Methodology
θ = (θ1 , . . . , θn−1 ) and the end point’s position (xend , yend ). The end effector needs to
move on a particular boundary curve resulting in two equations. So, the manipulator
has “n − 1” extra degrees of freedom in the joint space that needs to be determined
exactly. Consider the case when the head of the robot is following a closed convex
curve y(s) = 0, where “s” is the path parameter. The initial orientation of the first
link is assumed to be tangential to the curve. The initial velocity of the head is given
by v 1 = σ ẏ(s), where σ = ds dt
. We need to resolve the redundancy in such a way
that the other part (tail) of the manipulator starts making a convex shape to hold any
object within the bounded curve. We use inspiration from tractrix-based algorithm
to tying a knot to hold onto an object. So, the velocity of the tail of the first link is
given from Eq. 2. As the tail of the ith link and the head of the i + 1th link have the
same motion, so this continues till the head of the last link. Hence, we get “n − 1”
more equations for the complete trajectory. The velocity of the head of the second
link is given by Eq. 3.
i−1
Ẋ si −1 X s − X ei−1
= 1 (3)
Y˙si Ysi−1 − Yei−1
The first-order differential equation in Eq. 3 needs to be solved along with the con-
straint that the length of the link is constant.
The head tractor curve is a curve which bounds the object to be grasped. This is
based on the assumption that the object has a convex boundary. Consider a known
bound based on the prediction of the location and the boundary of the object, the
Resolving Hyper-Redundant Planar … 951
Fig. 4 Various instances of the snake-like robot trying to cage and then grasp an object kept near
the center of the bounded curve, using resolution methodology described
head of the manipulator follows a known curve, and the head moves on the bound
curve, whereas the rest of the curve tries to tie a knot within itself. The methodology
uses a circular-bounded curve as an input. The curve (circular here) should contain
the object. The true object boundary (though unknown) should be convex in nature
as the manipulator while interacting with the object should not get stuck within
concave-shaped regions. The end effector follows the bounded curve, whereas the
rest of the manipulator works under tractrix resolution. As shown in Fig. 4, the rest
of the manipulator moves helically toward the center of the bounded curve. The
“helical search” of the tail of the manipulator occurs until the links cannot move
the object. The object is free to move when one or two links hit it. The links of
the manipulator tend helically toward the center of the circular boundary, so if links
indeed hit the object, such that the set of links hitting the object can give motion
to it, then the manipulator tends to bring it toward the center and cage it. Then, all
the links apply a constant torque on the object to ensure grasp of the object and
its immobility. The methodology works because the tail of the manipulator moves
helically toward the center of the bounded curve while maintaining a convex curve
in its shape. The manipulator is stopped when at least three links hit a maximum
torque limit, which suggests the fact that at least three links have hit the object and
are rendered immovable, suggesting object grasp. In case, the object is soft then the
torque limit can be extended to squeeze the object as much as possible. For simulation
purposes, the motion of the object is assumed only along the instantaneous normal
of the contact region, and the manipulator is assumed to not interact with the object
tangentially. Only the component of the motion of the point in contact which is
instantaneously normal to the object is transferred to the object.
4.1 Example
Consider an object to be caged, and we know a bound to the object shape and position,
as shown in Fig. 6. The trajectory generated from the methodology is shown in
952 R. Kumar and S. Mukherjee
Fig. 5 Various instances of the snake-like robot trying to cage an object kept away from the center
of the bounded curve, using resolution methodology described
Fig. 2b. The tractrix curve was initially tangential to the boundary curve. The head
of the snake-like system follows the boundary, whereas the rest of the curve tries to
hold the object. If the object is not in the center of the curve, then one or two links
hit the object repeatedly until the object reaches to the center of the curve or a stage
where at least three links can cage the object. Figure 4 shows stages of an ellipse,
which was present at the center of the boundary curve to be caged by the manipulator.
The series of figures show how the object is caged and then grasped. Each link is of
length 0.2 units (except the last link) in the simulation example. In case the object to
be grasped is present in a non-central location, then the helically moving manipulator
tries to bring it to a point where at least three links can simultaneously come into
contact with the object, thereby caging it. The example manipulation for an elliptical
object is shown in Fig. 5. The links of the manipulator move the object to bring it
near the center where it can be properly grasped. The motion of the center of the
object is shown in Fig. 6.
When the object is near the boundary, the manipulator hits the object successively
in a near parabolic trajectory in the x − y plane. The near parabolic trajectory is
representative of the component of the motion of the contacting link normal to the
object to be held. The method works even for a larger object placed eccentrically
within the boundary, away from the center (Fig. 7). The manipulator first brings the
object near to the center and then tries to cage the object. In case the object is soft, the
robot’s links will squeeze the object. However, in case of a rigid object, the robot’s
links apply constant torque to the object.
Resolving Hyper-Redundant Planar … 953
Fig. 6 a x − y motion of the center of the object as the manipulator tries to cage the object with
time b x − y trajectory of the center of the object
Fig. 7 a x − y motion of the center of the object as the manipulator tries to cage the object with
time b x − y trajectory of the center of the object
954 R. Kumar and S. Mukherjee
5 Conclusion
This paper presents a methodology to cage and then grasp, planar smooth convex-
shaped objects with unknown shape and location. However, there needs to be a
knowledge of bound on the shape as well as the location of the object. A tractrix-
based method is used to resolve redundancy where the manipulator’s head moves
on a circular bound, and the rest of the manipulator performs a search for the object
and brings it to the center and then cages it. The methodology is useful when there
is a known loose bound on the position of the object, and we want to get hold of the
object without vision feedback.
References
1. Allen TF, Burdick JW, Rimon E (2015) Two-finger caging of polygonal objects using contact
space search. IEEE Trans Rob 31(5):1164–1179. https://doi.org/10.1109/TRO.2015.2463651
2. Allen TF, Rimon E, Burdick JW (2014) Two-finger caging of 3d polyhedra using contact
space search. In: 2014 IEEE international conference on robotics and automation (ICRA), pp
2005–2012. https://doi.org/10.1109/ICRA.2014.6907125
3. Chirikjian GS, Burdick JW (1993) Design and experiments with a 30 dof robot. In: Proceedings
IEEE international conference on robotics and automation, vol 3, pp 113–119. https://doi.org/
10.1109/ROBOT.1993.291862
4. Gander W, Bartoň S, Hřebíček J (2004) The tractrix and similar curves. In: Solving problems
in scientific computing using maple and MATLAB®. Springer, pp 1–25
5. Kumar GR, Srinivasan P, Holla VD, Shastry K, Prakash B (2003) Geodesic curve computations
on surfaces. Comput Aided Geomet Des 20(2):119–133
6. Madsen JJ, Markvorsen S (2017) Tractors and tractrices in riemannian manifolds. ArXiv
preprint arXiv:1707.09532
7. Pereira GA, Kumar V, Spletzer JR, Taylor CJ, Campos MF (2003) Cooperative transport of
planar objects by multiple mobile robots using object closure. In: Experimental robotics VIII.
Springer, pp 287–296
8. Rimon E, Blake A (1999) Caging planar bodies by one-parameter two-fingered gripping sys-
tems. Int J Rob Res 18(3):299–318
9. Rodriguez A, Mason MT, Ferry S (2012) From caging to grasping. Int J Rob Res 31(7):886–900
10. Sen D (1960) On geodesics of a modified riemannian manifold. Can Math Bull 3(3):255–261
11. Sreenivasan S, Goel P, Ghosal A (2007) Redundancy resolution using a tractrix and its appli-
cation to real-time simulations of hyper-redundant manipulators, snakes and tying of knots. In:
12th IFToMM world congress, Besancon, Paper. No. 524-1, Citeseer
12. Wang Z, Hirata Y, Kosuge K (2005) An algorithm for testing object caging condition by multiple
mobile robots. In: 2005 IEEE/RSJ international conference on intelligent robots and systems,
pp 3022–3027. https://doi.org/10.1109/IROS.2005.1545378
Boom Packaging with Yoshimura
Pattern: Geometrical and Deformation
Analysis
1 Introduction
Space structures equipped with large sized deployable reflectors and booms, often
known as gossamer structures, have made groundbreaking space missions possible
[1]. Deployable spaceborne structures offer the potential for exceptional pack-
aging efficiency, lightweight low-cost hardware over conventional space structures.
Foldable booms/masts are truly an essential part of any deployable space struc-
ture. A number of folding methodologies and mechanisms have been proposed
for spaceborne deployable booms/masts, such as Z-folding, spiral folding, origami
folding techniques, and telescopic folding [2]. The origami folding techniques
include Yoshimura pattern, bellows pattern, Miura-ori pattern, helically triangulated
folds, etc. Yoshimura pattern is an inextensional post-buckling solution for axially
compressed thin-walled cylinders [3]. A folding geometry similar to Yoshimura
pattern was obtained for the buckling pattern of a thin cylinder under torsional loading
[4]. Senda [5] analyzed several models of inflatable tubes folded with Yoshimura and
Miura-ori pattern.
The type of folding pattern and the number of fold vertexes/origami units repeated
around the circumference largely affects the packaging behavior and the deployment
deformations of the boom. The air entrapped in the vacant space of the folded structure
can cause an uncontrolled expansion of the structure, which led to numerous failures.
The proper management of residual gas is still a significant issue in the deployment of
inflatables. Till date, no study appears to have been reported which has analyzed the
geometry of the bellow folding pattern to obtain the residual space inside the folded
boom. The influence of different geometrical parameters of the bellows pattern on
the packaging and deployment behavior of the boom has not been investigated in the
literature. Moreover, the amount of deformation along the fold lines needed to be
quantified clearly with respect to different geometrical parameters.
In this study, the packaging and deformation behavior of a cylindrical boom
folded with Yoshimura pattern is investigated. This paper is organized as follows.
Section 2 explains the geometry of the Yoshimura folding pattern and mathemat-
ical expressions for various important parameters like radius of packaged cylinder,
packaging efficiency, and residual space after complete folding. In Sect. 3, the influ-
ence of number of origami units on packaging behavior of the boom is investigated.
Section 4 discusses the modeling and simulation of deployment of a single story
Yoshimura cylinder. A pin-jointed wireframe technique [7] is used to model folded
configuration of the boom. The expansion coefficient is defined and calculated for
the boom deployment. At last, Sect. 5 concludes the findings of this study.
mα = (m − 1)π (1)
a
b= sec β (3)
2
a
h= tan β (4)
2
The fold polygon of Yoshimura pattern is shown in Fig. 4. The radius of the circum-
scribed circle of the fold polygon is termed as the radius of packaged cylinder (Rc ).
This plays a vital role in the study of the packaging behavior of any cylindrical
folding pattern.
The radius of the packaged boom can be obtained as:
a
RC = α π (5)
4 sin 2
sin 2m
Vinitial − Vfolded
PE = × 100 (6)
Vinitial
Area of outer polygon (A). In Yoshimura folding pattern, the outer polygon is a regular
polygon, which consists of 2 m similar triangles ( AOD) as shown in Figs. 4 and
5 shows the details of AOD. The outer polygon area is obtained as
π
A = m RC2 sin (8)
m
Folded height (H f ). In s single story of the Yoshimura pattern, four layer overlaps;
hence, the total folded boom will be
Hf = 4 × t × S (9)
The total residual volume of the folded boom consists of inside residual volume
Rvol(inside) residual volume between layers Rvol(b/w the layers ) .
Inside residual volume Rvol(inside) . The inside residual volume in the folded boom
is calculated by multiplying the inside polygon area (A ) with the folded height H f .
Inside residual area (A ). It is the area of inner polygon, which consists of 2 m similar
triangles, POQ (see Fig. 5). The inner polygon area is obtained as
m 2 1
A = P (11)
2 tan(π/2m)
where
b sin(α − 90)
P= = PQ (12)
sin(α/2)
Rvol(inside) = Residual area A × Folded height H f (13)
1
A = Area of A P Q = P × AR (14)
2
3 Packaging Behavior
To study the packaging behavior of the boom, the influence of number of origami
units (m) on the packaged boom radius (RC ), packaging efficiency (PE), and residual
volume (Rvol ) is investigated using respective geometrical formulas. The perimeter
and length of the boom are taken as 600 mm and 800 mm, respectively. The thickness
of the material is taken as 0.2 mm in this section.
Boom Packaging with Yoshimura Pattern: Geometrical … 961
The perimeter (L) and length ratio (R) of the boom are kept constant for all values
of m, and all other dimensions vary according to Eqs. 1 and 2 with respect to m. The
variation in radius of packaged cylinder (RC ) and packaging efficiency (PE) with m
is shown in Figs. 6 and 7, respectively. Figure 8 shows the variation of the residual
volume with m; it can be seen that inside and total residual volume curves are nearly
equal; the variation of residual volume between layers with m is shown in Fig. 9.
polygon was made up of DC-DC, and the top polygon was made up of AB-AB fold
lines (see Figs. 2 and 3), and AD, AC, and BC fold lines acts as connection between
these polygons.
The nodes of the bottom polygon were restricted in z and θ directions and allowed
to move in radial direction only. The nodes of top polygon are given a displacement
(d = 0.75 × 2 h) in axial direction. Then, the axial strains of different fold lines
during the deployment were obtained numerically.
The deployment of the folded boom is shown in Fig. 10. Since the story height of
cylinders having different m will be different, so the applied displacement d (0.75 ×
2 h) will also be different. Hence, to compare the deformation behavior, the variation
of axial strains is shown in graphical form with respect to a normalized displacement,
which is (Fig. 11)
Instantaneous displacement
Normalized displacement = (17)
Total displacement (d)
Boom Packaging with Yoshimura Pattern: Geometrical … 963
5 Summary
The geometric design and packaging behavior of the Yoshimura pattern folded cylin-
drical boom were investigated. The fold line deformation analysis was carried out
using numerical simulations. The influence of the number of origami units on the
packaging behavior and fold line deformation during the deployment was studied.
The results indicate that:
1. As the number of origami unit increases, initially, the radius of packaged cylinder
decreases rapidly; later on, the decrement becomes very less. The packaging
efficiency decreases with the increment in the number of origami units around
the circumference. The residual volume inside the boom and between the layers
increases with the increases in number of origami units.
2. Under the action of a specific nodal displacement, the axial strains in the folds
increase with number of origami units (m). Also, for lesser value of m, as the
nodal displacement increases, the axial strains of the fold lines increase rapidly.
3. Expansion coefficient is negative for Yoshimura pattern folded boom, which
indicates the lateral contraction of boom during deployment. As the boom
expands, radius of packaged boom (RC ) decreases; for the same value of longi-
tudinal strain, expansion coefficient will be higher for the boom with lesser m
value.
In the case of inflatable booms, wireframe modeling approach can only predict
the initial deployment strains. For a better estimation of the deformation, the folded
boom should be modeled with a continuous surface using shell elements. This pattern
can be used as a folding mechanism to pack the inflatable/deployable booms for space
applications.
References
1. Jenkins CH (ed) (2001) Progress in astronautics and aeronautics: gossamer spacecraft: membrane
and inflatable structures technology for space applications, vol 191. Aiaa
2. Schenk M, Viquerat AD, Seffen KA, Guest SD (2014) Review of inflatable booms for deployable
space structures: packing and rigidization. J Spacecr Rockets 51(3):762–778
3. Yoshimura BY (1955) On the mechanism of buckling of a circular cylindrical shell under axial
compression. Natl Advis Commun Aeronaut
4. Hunt GW, Ario I (2005) Twist buckling and the foldable cylinder: an exercise in origami. Int J
Non Linear Mech 40(6):833–843
5. Senda K et al (2006) Deploy experiment of inflatable tube using work hardening, May 2006, pp
3–8
6. Tarnai T (1994) Folding of uniform plane tessellations. In: Origami science and art. Proceedings
of the second international meeting of origami science and scientific origami, pp 83–91
7. Cai J, Deng X, Feng J, Zhou Y (2015) Geometric design and mechanical behavior of a deployable
cylinder with Miura origami. Smart Mater Struct 24(12):
Multimodal Medical Image Fusion Based
on Interval-Valued Intuitionistic Fuzzy
Sets
Abstract Multimodal medical image fusion is the process of combining two multi-
modal medical images to increase the quality and to extract maximum information
from the output image for better treatment and precise diagnosis. The fused image
obtained from non-fuzzy sets lags with complementary information. Compared with
fuzzy set theory, intuitionistic fuzzy sets (IFS) are determined to be more suitable
for civilian and medical image processing as more uncertainties are measured. In
this paper, an algorithm based on an interval-valued intuitionistic fuzzy set (IVIFS)
is presented for efficiently fusing multimodal medical images and the final fused
image is passed through a median filter to remove noise. Simulations on few sets of
multimodal medical images are performed and compared with the existing fusion
methods, such as an intuitionistic fuzzy set and fuzzy transform. The superiority of
the proposed method is presented and is justified. Fused image quality is additionally
checked with different quality measurements, for example, entropy, spatial frequency
(SF), average gradient (AG), etc.
1 Introduction
With the latest developments in the field of technology, digital image processing
systems have turned into a reality in developing the number of fields, for example,
machine vision, medical imaging, and military applications. The consequence of the
utilization of these strategies is an awesome increase of the amount of data available.
To extract all the valuable information from the source images and to reduce the
increasing volume of data, a powerful method is used in image processing called
T. Tirupal (B)
Department of ECE, GPCET, Kurnool, Andhra Pradesh 518452, India
B. Chandra Mohan
Department of ECE, BEC, Bapatla, Andhra Pradesh 522101, India
S. Srinivas Kumar
Department of ECE, JNTUA, Ananthapuramu, Andhra Pradesh 515002, India
image fusion. The main aim of the image fusion is to produce new images that are
more appropriate for the purpose of human/machine perception and also to reduce
the amount of data.
Medical images provide different types of information: Computed tomography
(CT) images embed with less distortion and provide details regarding dense struc-
tures, such as bones, magnetic resonance imaging (MRI) provides information
about pathological soft tissues, magnetic resonance angiography (MRA) easily
detects abnormalities in the brain, X-ray detects fractures and abnormalities in
bone positions, vibro-acoustography (VA) provides the depth and thickness of the
disease object, and positron emission tomography (PET), and single photon emis-
sion computed tomography (SPECT) provides functional and metabolic data about
the human brain. By combining these in multimodal medical image pairs such as
CT-MRI [1], MRI-MRA, X-ray-VA [2], MRI-PET [3], and MRI-SPECT, extra clin-
ical data is extracted that is complimentary in nature. Accordingly, we can say that a
single image will give all applicable data, and henceforth multimodal medical image
fusion is fundamentally required to acquire all probable data in a single composite
image called a fused image [4].
2 Literature Review
Fuzzy set theory plays a vital role in the image processing proposed by Zadeh [5]
to remove the ambiguity and vagueness present in images. Fuzzy sets in image
processing increase contrast, flat the regions of interest, and refine the edges and
fine erections of the image. Fuzzy sets use the membership function to remove
the vagueness present in the image. While describing the membership function, an
uncertainty due to lack of knowledge or personal error is found, which primes to
another uncertainty called hesitation degree. The combination of membership, non-
membership, and hesitation degree defines the intuitionistic fuzzy set introduced by
Atanassov [6] in 1986.
Medical images are poorly illuminated, contain many uncertainties in the form
of noise, have vague boundaries, have overlapping gray levels, have invisible blood
vessels, and it is difficult to extract objects from the image. Many uncertainties exist
in every phase of image processing, and in using IFS [6–9] and fuzzy transforms
[10], these uncertainties can be removed. The most medicinal images are foggy for
two reasons: one is that the noise signal obscures the high frequency signal of an
image edge, the other one is the edge of the tumor with ordinary tissues cannot be
exceptionally all around characterized by the images, in this manner, it is troublesome
for radiology specialists to outline an image. There are a couple of utilizations of
fuzzy logic-based image fusion, for instance, image segmentation and integration,
deep brain stimulation, image retrieval, brain tumor segmentation, ovarian cancer
diagnosis, and gene expression.
Atanassov and Gargov [11] stretched out the intuitionistic fuzzy set to the interval-
valued intuitionistic fuzzy set (IVIFS), which is portrayed by a membership function
Multimodal Medical Image Fusion Based on Interval-Valued … 967
and a non-membership function whose values are intervals rather than real numbers.
This methodology not just extends the capacity of the intuitionistic fuzzy set to
deal with uncertain data yet in addition improves its capacity to tackle practical
decision-making problems. The notion of IVIFS is as follows:
Let μ Z (y) = [μ−Z (y), μ+Z (y)] and v Z (y) = [v − +
Z (y), v Z (y)] are intervals,
μ Z (y) = inf μ Z (y), μ Z (y) = sup μ Z (y), v Z (y) = inf v Z (y), v +
− + −
Z (y) = sup v Z (y),
then
1 α
m n
Eα = − p ln piαj , α ∈ T (1, 2, . . . , t) (4)
ln(mn) i=1 j=1 i j
πα
where piαj = m i jn
πiαj
and 0 ≤ E α ≤ 1.
i=1 j=1
3 Proposed Method
This section discourses a proposed system for multimodal medical image fusion
using IVIFS and contrast visibility.
1. The two input source medical images are initially read as I1 and I2 .
2. The first image I1 of size M × N is fuzzified by means of the formula
968 T. Tirupal et al.
Ii j1 − lmin
μ Z 1 (Ii j1 ) = (5)
lmax − lmin
where Ii j1 defines the gray level of the first image and ranges from 0 to L −1 (L
is the extreme gray-level value). lmin and lmax represent the least and extreme
gray-level values of the first image.
3. The optimum value of α is computed for the first image using the entropy
Eq. (4).
4. Find the fuzzified IVIFS image with the optimum value of α for the first input
medical image by using the Eq. (3) and denote as IY 1 .
5. The above procedure is repeated for from step 2 to step 4 for the second input
medical image to find fuzzified IVIFS, IY 2 .
6. The two fuzzified images IY 1 and IY 2 are decomposed into m ×n blocks—each
block size is considered as 5 × 5 in this paper—and then compute the contrast
visibility (CV) of each block separately using the below equation.
1 |I (i, j) − μk |
CV = (6)
p × q (i, j)∈B μk
k
where μk and p × q are the mean and magnitude of the block Bk , respectively.
7. Then, a decision map (DM) is constructed that determines the combination of
pixels of two images.
⎧ I I
⎨ 1 if C Vi Y 1 > C Vi Y 2
D M = −1 if C Vi < C Vi IY 2
IY 1
(7)
⎩
0 if C Vi IY 1 = C Vi IY 2
C Vi IY 1 and C Vi IY 2 are the contrast visibilities of the ith block of images IY 1 and
IY 2 , respectively. DM is the decision map built by taking the decision for each
coefficient using the contrast visibility of the respective block.
8. A new decision map (NDM) is created by refining DM with consistency
verification by using the majority filter introduced by Li et al. [13].
9. The fused image (FI) without uncertainty is attained based on the NDM as
⎧
⎨ IY 1 (i, j) if N D M(i, j) = 1
F I (i, j) = IY 2 (i, j) if N D M(i, j) = −1 (8)
⎩
[IY 1 (i, j) + IY 2 (i, j)]/2 if N D M(i, j) = 0
where lmin and lmax denote the lowest and extreme gray-level values of the
fused image, FI.
11. Finally, the fused image is filtered for noise using median filter.
4 Experimental Results
The first example shown in Fig. 1 of the first row discourses CT and MRI images
(www.metapix.de/toolbox.htm) that are complementary in nature. The CT image
gives data about bones and hard tissues while the MRI image gives soft tissue data.
Fusing these two images gives plenteous information in a single image, which better
analyzes a sickness. The fifth image of the first row of Fig. 1 is the fusion result of the
proposed method, which outwardly demonstrates that the image is of higher contrast
and luminance than the fused image of prevailing methods, intuitionistic fuzzy set
and fuzzy transform.
Further examinations of the outcomes are performed utilizing a couple of objec-
tive criteria utilized in [14], for example, average gradient (AG), entropy, spatial
frequency (SF), and edge strength preservation (QAB/F ), and the results are listed in
Table 1. It is seen from the table that the proposed technique gives better execution
with respect to contrast, luminance, and visibility of the fused image.
The second example addresses a T1-weighted MR image and the MRA image
with some illness as white structures that are shown in the second row of Fig. 1. The
CT MRI
MRI SPECT
Fig. 1 Fusion results for multimodal medical images. The first and second columns represent the
input source images. The third column represents the fused images by an intuitionistic fuzzy set.
The fourth column represents the fused images by fuzzy transform. The fifth column represents the
fused image by an interval-valued intuitionistic fuzzy set
970 T. Tirupal et al.
Table 1 Objective assessment of various image fusion methods with the proposed interval-valued
intuitionistic fuzzy set method for Fig. 1
Modality Fusion AG (change in ENTROPY SF QAB/F
method intensity/pixel) (bits/pixel) (cycles/millimeter)
CT-MRI IFS [9] 14.4342 5.3570 30.7839 0.7516
Fuzzy 14.4624 6.7823 31.5100 0.7324
transform
[10]
Proposed 18.5263 6.8026 35.1702 0.8451
IVIFS
MR-MRA IFS [9] 17.5735 5.1670 35.8483 0.6406
Fuzzy 16.7764 6.0653 33.5737 0.7042
transform
[10]
Proposed 18.701 7.1052 36.2217 0.7165
IVIFS
MRI-SPECT IFS [9] 5.6202 0.9789 32.5970 0.4743
Fuzzy 6.0501 1.4000 32.6430 0.4994
transform
[10]
Proposed 17.1271 5.9753 35.3234 0.7868
IVIFS
5 Conclusion
This paper grants a new method for fusing multimodal medical images using interval-
valued intuitionistic fuzzy sets. The proposed algorithm is tested on several pairs of
medical images, and it is found that the proposed method gives improved visual and
quantitative results, with high contrast and luminance. Better results are obtained
using IVIFS because it considers a greater number of uncertainties and incorporates
a hesitation degree. As medical images are a low contrast with vague regions and
boundaries, IFS aids in solving these problems. This algorithm is also included with
intuitionistic fuzzy entropy to optimize the best parameter for membership, non-
membership, and hesitation degree functions. Further work includes the usage of
neuro-fuzzy logic for improved image quality.
References
1. Shanker HO, Mishra, Bhatnagar S (2014) MRI and CT image fusion based on wavelet
transform. Int J Inf Comput Technol 4(1):47–52
2. Hosseini HG, Alizad A, Fatemi M (2007) Integration of Vibro-Acoustography imaging
modality with the traditional mammography. Int J Biomed Imaging 1–8
3. Tirupal T, Chandra Mohan B, Srinivas Kumar S (2018) Multimodal medical image fusion based
on fuzzy sets with orthogonal teaching–learning-based optimization. In: Verma N, Ghosh A
(eds) Computational intelligence: theories, applications and future directions, vol II. Advances
in Intelligent Systems and Computing, p 799
4. Tirupal T, Chandra Mohan B, Srinivas Kumar S (2015) Image fusion of natural, satellite, and
medical images using undecimated discrete wavelet transform and contrast visibility. In: IEEE
national conference on recent advances in electronics & computer engineering (RAECE), pp
11–16
5. Zadeh LA (1965) Fuzzy sets. Inf Control 8(3):338–353
6. Atanassov KT (1986) Intuitionistic fuzzy sets. Fuzzy Sets Syst 20(1):87–96
7. Tirupal T, Chandra Mohan B, Srinivas Kumar S (2017) Multimodal medical image fusion
based on Sugeno’s intuitionistic fuzzy sets. ETRI J 39(2):173–180
8. Tirupal T, Chandra Mohan B, Srinivas Kumar S (2019) Multimodal medical image fusion
based on Yager’s intuitionistic fuzzy sets. Iranian J Fuzzy Syst 16(1):33–48
9. Balasubramaniam P, Ananthi VP (2014) Image fusion using intuitionistic fuzzy sets. Inf Fusion
20:21–30
10. Meenu M, Rajiv S (2016) A novel method of multimodal medical image fusion using fuzzy
transform. J Vis Commun Image Repr 40:197–217
11. Atanassov K, Gargov G (1989) Interval valued intuitionistic fuzzy sets. Fuzzy Sets Syst
31(3):343–349
12. Yue C (2017) Entropy-based weights on decision makers in group decision-making setting
with hybrid preference representations. Appl Soft Comput 60:737–749
13. Li H, Manjunath BS, Mitra SK (1995) Multisensor image fusion using the wavelet transform.
Graph Models Image Process 57(3):235–245
14. Jagalingam P, Hegde AV (2015) A review of quality metrics for fused image. Aquatic Procedia.
Int Conf on Water Resources, Coastal and Ocean Engineering (ICWRCOE), March 12–14,
Mangalore, Karnataka, India, vol 4, pp 133–142
The Influence of Ultrasound
for the Protection of Animals
on Highways Through Electronic Circuits
Abstract While driving in regions where creatures are regularly present, it is not
unexpected to wind up in a mishap. Both wild and residential creatures might be
outside and can keep running into the street. Normally, a driver’s first nature is to
swerve to abstain from hitting the creature; however, that can have wrecking results,
such as losing control of the vehicle and enduring genuine wounds. Swerving can
deliver a domino impact, making the driver strike another vehicle or object, which can
prompt far more atrocious outcomes, similar to the vehicle moving over or genuine
damage to different drivers out on the road. It is essential to avoid such accidents
and protect animals as well. To overcome this, new method is proposed in this paper
which includes a circuit generating ultrasonic waves. It can be used as pest repellents.
For generating ultrasonic waves of high frequency, a generator using 555 timer can be
employed. These waves are designed to produce an extremely high-frequency sound
that is beyond what humans can hear. Ultrasound is used to bring about enough
irritation in animals and make them stay away from highways.
1 Introduction
There are different reasons why individuals must repulse creatures from territo-
ries where they can damage individuals or devastate important goods and further-
more stay away from mishaps. This objective can be accomplished in various ways
utilizing various techniques. We can recognize electrical, chemical, mechanical,
optical, reflective bags, acoustical strategies, and so on. The benefit of the acoustical
technique contrasting with others is: economical to utilize, not unsafe to creatures and
safe for individuals utilizing it. This is valid under the presumption that ultrasound is
utilized, which is indiscernible for individuals and does not cause any consultation
harm, notwithstanding when presented to sound weight levels up to 120 dB. At the
point when creatures hear these sounds they will just sit and gaze at the region where
it has all the earmarks of being radiating. Along these lines, a clever hardware frame-
work is essential which can be included vehicles to maintain a strategic distance from
the potential outcomes of mishaps.
Generally, creature–vehicle crashes have been tended to through signals
cautioning drivers of potential creature intersections. In different cases, natural life
cautioning reflectors, mirrors, or wildlife fences have been introduced to ward off
creatures from the street [1]. Nonetheless, regular cautioning signals seem to have
just constrained impact since drivers are probably going to habituate to them. Natural
life cautioning mirrors or reflectors may not be powerful. Wildlife fencing has been
joined with natural life intersection structures to address these impediments; at the
same time, fundamentally because of their moderately surprising expense, such
intersection structures are constrained in number and width.
2 Literature Review
so on bringing about mishap evasion [6]. Every creature has his very own capable of
being heard scope of frequency range and not all sort of creatures found all over the
place. Notwithstanding, we watched this reality with unmistakable fascination and
proposed the framework which can recognize the creature/vehicle and caution them
at their discernable range [7–9].
Table 2 demonstrates the scope of maximum point of confinement and the lower
breaking point hearing the frequency of the various species. Any endeavor to eval-
uate the impacts of sounds on creatures must consider species contrasts in hearing
capacities. In spite of the fact that the consultation scopes of most species cover to
an enormous degree, significant variety happens in high and low frequency hearing
just as in outright affectability. Accordingly, a sound that is effectively discernable
to one animal type might be less capable of being heard, or even unintelligible, to
another.
3 Proposed Method
Essential thought behind the framework is to create the sound signal which will be
indiscernible to human and irritating for creatures. The irritating frequency can be
evaluated by explicit rationale and can be chosen and balanced by an experimenta-
tion approach. The irritating frequency generator of a particular loudness is planned
furthermore with some intelligent tasks to detect the traffic of creature/vehicle. Intel-
ligent and energy productive framework is created with disturbing signal generation.
Such creature detecting and the ready framework can be mounted on the two sides
of streets. It will remain unusable until any vehicle is distinguished over the street.
976 T. Tirupal and S. Fowzia Sultana
Ultrasonic gadgets work by transmitting high frequency sound waves more note-
worthy than 20,000 Hz, while a few creatures, for example, dogs, bats, rodents,
feathered creatures, and creepy crawlies can hear well into the ultrasonic range. The
human ear comes up short on the ability to hear such sound. Ultrasonic gadgets are
structured and developed to radiate sound of this frequency, when focused at pests;
they make them awkward inside the zone of inclusion along these lines repulsing them
away from the zone without influencing nature and non-target life forms, including
man. Electromagnetic gadgets are fitted into home wirings and produce electromag-
netic waves which are unfriendly to bothers. The advantages of this strategy over
other bug control techniques incorporate the way that they are cheap, eco-framework
neighborly, earth amicable, and have no known hazard to people.
and
Choosing C1 , R1 and R2
R1 and R2 ought to be in the range 1 K ohm to 1 M ohm. It is ideal to pick C1 first
since capacitors are accessible in only a couple of qualities.
Choosing C1 to suit the frequency range according to necessity.
Choosing R2 to give the frequency (f ) which is required. Except that R1 is much
smaller than R2 , at that point utilizing:
4 Simulation Results
Using the above circuit, ultrasonic waves of required frequency can be obtained by
varying external capacitance and resistance values (Figs. 2 and 3).
5 Conclusion
Because of the sound signal, the creature can flee from the street and thus there
are least odds of mishaps. The fundamental thought behind the framework is to
generate the sound signal utilizing 555 timer which will be quiet to human and
irritating for creatures. Their irritating frequency can be assessed by explicit logic
and can be chosen and balanced by experimentation procedure. Irritating frequency
generator of a particular loudness is planned moreover with some shrewd tasks to
detect the traffic of creature/vehicle. Intelligent and energy efficient framework is
created with irritating signal production. Such creature detecting a ready framework
can be mounted on the two sides of streets. It will remain unusable until any vehicle
is distinguished over the street. It is another methodology in social perspectives for
wild creature passing shirking and mishaps anticipation. Creature explicit frequency
range signals are produced. The particular creatures are cautioned with these signal
of threat and effectively fled. The framework can be included in vehicles or trains as
opposed to mounting poles on the roadside.
The Influence of Ultrasound for the Protection of Animals … 979
References
1. Gunson KE, Chruszcz B, Clevenger AP (2003) Large animal-vehicle collisions in the central
Canadian rocky mountains: patterns and characteristics
2. Dodd N, Gagnon J, Schweinsburg R (2007) Evaluation of measures to minimize wildlife-
vehicle collisions and maintain wildlife permeability across highways in Arizona. USA
3. Sonone DJ, Patil DA, Rane KP (2014) Irritating and hearing frequency identification and
generation to avoid animals accident, July 2014
4. Saritha B, Elakiya P, Mathavi S, Monika M, Nivetha V (2017) To prevent the animals accident
and track crack detection system for railways, Mar 2017 Wildlife Protection Society of India
5. Valitzski SA (2004) Evaluation of sound as a deterrent for reducing deer-vehicle collisions.
University of Georgia
6. Tiwari DK, Ansari MA (2016) Electronic pest repellent: a review. In: International conference
on innovations in information embedded and communication systems (ICIIECS’16)
7. Ibrahim AG, Oyedum OD, Awojoyogbe OB, Okeke SSN (2013) Electronic pest control devices:
a review of their necessity, controversies and a submission of designal considerations. Int J Eng
Sci 2:2319–1805
8. Enayati A, Hemingway J, Garner P (2010) Electronic mosquito repellents for preventing
mosquito bites and malaria infection (review). Wiley, New York
9. Jhaver S, Singh R, Hiremani T (2009) Electronic pest repellent. EE318 Electronic Designal
Lab, Project Report, EE Dept, IIT Bombay
10. Yu DS, Zheng CY, Iu HHC, Fernando T (2015) A memristive astable multivibrator based on
555 timer. In: 2015 IEEE international symposium on circuits and systems (ISCAS)
11. Everest F (2000) The characteristics and use of the 555 timer. Electron Educ 2000(3):34–40
12. Camenzind HR (1997) Redesignaling the old 555 [timer circuit]. IEEE Spectr 34(9):80–85
Workspace Analysis of a 5-Axis Parallel
Kinematic Machine Tool
1 Introduction
Parallel mechanisms offer a number of advantages over their serial counterparts like
higher stiffness, better rigidity, higher payload/weight ratio and higher accuracy.
After NC machines, parallel kinematic machines have opened a new domain in the
machine tool industry. They are different from the conventional machines, since they
are based on variable geometry truss structures [1]. 6-DOF parallel manipulator like
Stewart platform requires six linear actuators for controlling; however, only three
linear actuators are required to control 3-RPS mechanism. Also, 6-DOF parallel
manipulator has complex kinematic analysis and mechanical design. So, the paper
presents a 5-DOF parallel kinematics machine tool architecture based on a 3-DOF, 3-
RPS parallel manipulator and two additional degrees of freedom of the movement of
the base in X- and Y-direction. The parallel manipulator slides on the frame structure
of machine tool and translates along X- and Y-direction in order to increase the
workspace in both the directions. Along with this, its rigidity also increases because
of the adaptation of gantry structure [1].
3 Position Analysis
Before doing workspace analysis of PKM, inverse and forward kinematics of the
3-RPS parallel manipulator have been carried out. In IK the position vector is given
and the problem is to find the actuated variables [2]. In another sense, it is focused
on deducing the joint motions when the position of End-Effector is known [3]. Here
IKis carried out to calculate the limb lengths (d 1 , d 2 , d 3 ) from the equation
and from the given location of the moving platform, while in FK limb lengths are
known to us and we have to find the location of the moving platform by using the
equation
T
qi −qi+1 qi −qi+1 − 3h 2 = 0 (2)
where, qi is the position vector of points, Bi with respect to the fixed coordinate
system, ai is the position vector of points, Ai in the coordinate system A, and i = 1,
2, 3 is used for the corresponding limb parameters, respectively. See Fig. 1.
As a sample at random, we have selected the values as Pz = 965 mm, ψ x = 10°
and ψ y = 0°. Radius of base platform, g = 500 mm and radius of moving platform,
984 A. Jain and H. P. Jawale
h = 255 mm. The values of limb lengths obtained in IK are d 1 = 995.14 mm, d 2 =
1033.8 mm and d 3 = 959.49 mm, respectively.
Now, in FK, considering limb lengths as d 1 = 995.14 mm, d 2 = 1033.8 mm, d 3
= 959.49 mm and g = 500 mm, h = 255 mm, we obtained total 16 values of limb
angle ϕ 1 , in total 8 pairs, mirror image of each other.
Correspondingly, other limb angles ϕ 2 and ϕ 3 can be obtained from ϕ 1 by back
substitution. Out of those values of ϕ 1 , we take our initial guess value of ϕ 1 = 74.83°,
ϕ 2 = 74.52° and ϕ 3 = 74.34° from the geometric constraints.
It has been found that the results obtained in forward kinematics for the task space
parameters are matched with those obtained in inverse kinematics.
4 Workspace Analysis
The reachable workspace of a manipulator is defined as the space that can be reached
by the reference point by at least one orientation [4]. Two cases have been considered
for workspace analysis to show the reach of the moving platform as follows:
Case (I) Workspace of point P (centroid of B1 B2 B3 ) of the moving platform is
plotted when one limb is increasing in dimensions while keeping the other two
constants. Let the increment in the respective limbs are of 1 mm, 3 mm, 5 mm,
7 mm and 10 mm. Limb lengths are d 1 = 995.14 mm, d 2 = 1033.8 mm and d 3 =
959.49 mm.
Figure 3 shows the workspace of point P of moving platform by keeping d 2 and
d 3 constant and increasing d 1 by 1 mm, 3 mm, 5 mm, 7 mm and 10 mm, respectively.
Figure 4 shows the workspace of point P of moving platform by keeping d 1 and
d 3 constant and increasing d 2 by 1 mm, 3 mm, 5 mm, 7 mm and 10 mm, respectively.
5 Results
Complete work on the position analysis (inverse and forward kinematics) and the
workspace analysis is carried out and quantitative improvement over reachable work
988 A. Jain and H. P. Jawale
References
1. Chen W, Zhao L, Zhang J (2011) A 3-DOF parallel manipulator based 5-axis parallel kinematics
machine tool. Adv Mater Res 317–319:698–702
2. Zhang D, Gao Z, Su X, Li J (2011) A comparison study of three degree of freedom parallel
robotic machine tools with/without actuation redundancy. Int J Comput Integr Manuf 230–247
3. Gao Z, Zhang D, Ge Y (2010) Design optimization of a spatial six degree of freedom parallel
manipulator based on artificial intelligence approaches. Robot Comput Integr Manuf 26:180–189
4. Li Y, Xu Q (2007) Kinematic analysis of a 3-PRS parallel manipulator. Robot Comput Integr
Manuf 23:395–408
Reaction Solvability Analysis Using
Natural Coordinates
1 Introduction
Over-constrained mechanisms have actual degree of freedom (DOF) more than the
number computed using the well-known Grubler–Kutzbach criterion [1]. This hap-
pens because of the presence of multiple joints, called redundant joints, constraining
the same degrees of freedom. These redundant constraints cause linear dependency in
constraint equations, leading to singular Jacobian matrices. The redundant constraints
can be removed arbitrarily without changing its kinematics [2], and the kinematics
of the system can be solved. However, one cannot solve for joint reactions uniquely
in such a system solely on the basis of rigid body constraint equations, and one needs
to include flexibility/the material constitutive equations to obtain the joint reactions.
This issue of “joint reaction indeterminacy” is the topic of this work.
The usual approach for handling redundant constraints is to solve for the system
by eliminating any arbitrary set of dependent constraints equations [3, 4], or by
using algorithms capable of dealing with dependent equations (e.g., minimum-norm
solution and augmented Lagrangian) [5–7], or by using penalty-based and weighing
factor-based methods [8, 9]. However, these approaches are good for kinematic
and certain dynamic analyses and do not yield correct results for joint reactions. For
calculating joint reactions, one needs to discard the rigid body assumption, introduce
material constitutive relations to have the complete set of equations, and solve the
system using a finite element approach (FEA) or analytically [10–12].
A new strategy was suggested based on the observation that in an over-constrained
mechanism, some joint reactions may be solvable without considering the material
constitutive equations [10]. Hence, if our joint reactions of interest lie in this set of
“solvable joint reactions,” then there is no need to use an FEA solver. This can save
tremendously in computation for complex mechanisms. In the subsequent works
[2, 5, 10], algorithms were developed to find such “solvable joint reactions” in
over-constrained mechanisms. This analysis for solvable joint reactions is termed as
“reaction solvability analysis (RSA)”.
In existing works, the coordinate system of choice for RSA has been absolute
coordinates. In this work, we show that using natural coordinates can lead to sub-
stantial benefits in terms of computational simplicity and efficiency. The main reason
is that the constraint equations using absolute coordinates are transcendental, while
the constraint equations in natural coordinates are maximally quadratic, leading to a
Jacobian matrix with linear terms. However, the application of natural coordinates
for RSA is not straightforward, and we had to modify the natural coordinates to
make them usable for RSA. We demonstrate the RSA application using our modified
natural coordinates with a planar over-constrained mechanism.
In this section, we discuss the necessary equations and representation that we have
used in this paper, and the existing RSA algorithms that have been developed in prior
works.
We will be using the standard representation used in [13] and [14] to represent the
dynamics and constraint equations of multi-body systems.
The equations of motion of a multi-body system can be written in a compact
form as
Mq̈ + f = Q (1)
where Mnxn is the mass/inertia matrix, qnx1 = (q1 , q2 , . . . , qn )T is the vector of gen-
eralized coordinates, and Qnx1 is the vector of external forces and other inertia terms
Reaction Solvability Analysis Using Natural Coordinates 993
such as the Coriolis and centripetal terms. The constraint equations of a multi-body
system can be written as
⎡ ⎤
1 (q)
⎢ 2 (q) ⎥
⎢ ⎥
(q) = ⎢ . ⎥ = 0 (2)
⎣ .. ⎦
m (q)
f = q T λ (4)
where λm×1 is the vector of Lagrange multipliers. Equation (4) is used to obtain
constraint reaction forces from the given constraint equations.
q T λ = Q − Mq̈ = f (5)
The algorithm states that the constraints forces for joints can be solved for uniquely
for which the following relation holds:
r = ri + r−i (6)
To implement the RSA algorithm discussed in the previous section, one needs a
formulation which has constraint equations for each joint. This is the key idea
differentiating which coordinates are “usable” for RSA and which are not.
The two commonly used types of coordinates are the absolute coordinates (or
reference point coordinates) and natural coordinates (or fully Cartesian coordinates).
In absolute coordinates, a coordinate frame is attached to every link, and joints
are defined by the constraint equations between these frames. Since the constraint
equations are for joints, absolute coordinates are directly usable for RSA and have
been used in prior works. In natural coordinates, one defines points and unit vectors
(which usually denote the links) and then constrain these points and vectors. In this
formulation, the joints are encapsulated in the formulation itself and the constraint
equations refer to the rigidity of the links. Hence, there are usually no constraint
equations for the joints, making them unusable for RSA. In the next section, we
show that natural coordinates can be made usable for RSA by introducing extra
points and unit vectors. Natural coordinates offer the benefit that all the constraints are
maximally quadratic [15, 16], while in absolute coordinates, we have transcendental
terms.
We illustrate the absolute and natural coordinates with a simple planar four-bar
mechanism, as shown in Fig. 1.
The constraint equations for four-bar modeled using natural coordinates, as shown
in Fig. 2, are:
Reaction Solvability Analysis Using Natural Coordinates 995
In these equations, the first corresponds to the rigid body constraint for the first
link, the second one for the rigidity of second link, and so on. The revolute joints
in the mechanism are captured by the sharing of the “basic” points; hence, no addi-
tional constraint equations are required for them. We propose a modified formulation
of natural coordinates which help us write constraint equations for other kinds of
joints. This formulation requires more “basic” points (and “unit vectors” for spatial
mechanisms).
tions, we now have equations constraining the joints. Note that we have dismantled
link ends in the figure to show clearly how the points between different links are not
shared as before, rather we share them using extra constraint equations.
We now have extra constraints that correspond to the revolute joints. They are
captured by four vector equations—hence, a total eight scalar equations—given as
rA = r1 , r2 = r3
(7)
r4 = r5 , r6 = rB
4.1 Assumptions
We follow the reference [4] to obtain the constraints with natural coordinates. The
prismatic (P) joint constraints are of two types: one, collinear constraint (making cross
product of the axes-defining vectors zero), and two, constraint for fixing the angle
between the two connected bodies. Hence, for each prismatic joint, the constraint
equations are
r1 × r3 = 0
P1: Prismatic Joint 1 (8)
(r3 − r10 ).r1 = l1l6 cos(φ1 )
r7 × r8 = 0
P2: Prismatic Joint 2 (9)
(r9 − r7 ).r8 = l5l4 cos(φ2 )
(r10 − r3 ) × (r3 − r9 ) = 0
P3: Prismatic Joint 3 (10)
(r10 − r3 ).(r9 − r7 ) = l5l6
For the revolute (R) joint, the points lying on the two bodies are coincident and
we get the constraint equations as:
R1: r2 = r3 Revolute Joint 1 (11)
998 S. Sharma and A. Ghosal
R2: r4 = r5 Revolute Joint 2 (12)
R3: r7 = r6 Revolute Joint 3 (13)
The RSA algorithm as discussed in Sect. 2.2 was applied on this mechanism using
mathematica® . When using natural coordinates, the rank of Jacobian matrix is r =
rank(q ) = 19 and the ranks ri and r−i are shown in Table 1. For all the prismatic
joints, ri + r−i > r = 19. Hence, for none of the prismatic joints, the reactions can
be found uniquely. However, for all revolute joints, ri + r−i = r = 19 and all the
revolute joint reactions can be found uniquely.
When using absolute coordinates, as was done in [10], the rank r of the Jacobian
matrix is 11. Here, for all prismatic joints, ri + r−i > r = 11 and for all revolute
joints ri + r−i = r = 11. Hence, all revolute joint reactions can be found uniquely,
but none of the prismatic joints can be. This is the same result as when using the
natural coordinates. Results from both analyses are shown in Table 1 for comparison.
4.4 Discussion
As can be seen from Table 1, both approaches of RSA analysis— using modified
natural coordinates and using absolute coordinates (as done by [10]) yielded the same
results. Further, it can be observed that natural coordinates had larger matrices (with
greater rank). However, this did not lead to increased computational complexity, as
Jacobian matrix obtained using natural coordinates had no transcendental terms as
seen when absolute coordinates are used. This led to a great increase in speed of
the RSA algorithm when using natural coordinates. Three more mechanisms were
analyzed apart from this planar over-constrained mechanism, including one spatial
mechanism (Bennett mechanism), in which the computational advantage was found
to be even more profound.
5 Conclusions
References
1. Gogu G (2008) [Part-1] Structural synthesis of parallel robots Pt. 1: methodology. Springer,
Dordrecht
2. Marek W (2009) Joint reactions in rigid body mechanisms with dependent constraints. Mechan
Mach Theo 44(12):2265–2278
3. Bayo E, Ledesma R (1996) Augmented lagrangian and mass-orthogonal projection methods
for constrained multibody dynamics. Nonlinear Dyn 9(1):113–130
4. Garcıa J, de Jalón Bayo E (1994) Kinematic and dynamic simulation of multibody systems.
Mechanical Engineering Series, Springer, New York
5. Marek W, Janusz F (2013) Solvability of reactions in rigid multibody systems with redundant
nonholonomic constraints. Multibody Syst Dyn 30(2):153–171
6. Xu Y, Wenlan L, Jiantao Y, Yongsheng Z (2015) A method for force analysis of the overcon-
strained lower mobility parallel mechanism. Mech Mach Theo 88:31–48
7. Wenlan L, Xu Y, Jiantao Y, Yongsheng Z (2017) The weighted Moore-Penrose generalized
inverse and the force analysis of overconstrained parallel mechanisms. Multibody Syst Dyn
39(4):363–383
8. Francisco G, József K (2013) Use of penalty formulations in dynamic simulation and analysis
of redundantly constrained multibody systems. Multibody Syst Dyn 29(1):57–76
1000 S. Sharma and A. Ghosal
9. Bilal R, József K (2011) A penalty formulation for dynamics analysis of redundant mechanical
systems. J Comput Nonlinear Dyn 6(2)
10. Marek W (2005) Joint reaction forces in multibody systems with redundant constraints. Multi-
body Syst Dyn 14(1):23–46
11. Bi ZM, Kang B (2014) An inverse dynamic model of over-constrained parallel kinematic
machine based on newton–euler formulation. J Dyn Syst Meas Cont 136(4):041001–041001-9
12. Zahariev E, Cuadrado J (2007) Dynamics of over-constrained rigid and flexible multibody
systems. In: 12th IFToMM world congress, Besançon, France
13. Nikravesh PE (1988) Computer-aided analysis of mechanical systems. Prentice-Hall, Engle-
wood Cliffs, N.J
14. Negrut D, Dyer A (2004) ADAMS/Solver Primer., Ann Arbor
15. Jalón JG (2007) Twenty-five years of natural coordinates. Multibody Syst Dyn 18(1):15–33
16. Uchida T, Callejo A, de Jalón JG, McPhee J (2014) On the Gröbner basis triangularization of
constraint equations in natural coordinates. Multibody Syst Dyn 31(3):371–392
Design, Analysis and Development
of Sweep Arm Scanner for Scanning Fast
Breeder Reactor Core
Abstract Fast breeder reactors (FBRs) utilize liquid metals as coolant. However, in-
service inspection of components using conventional optical methods is impossible
in sodium due to its opacity. Ultrasonic technique is therefore employed for this
purpose which requires the deployment of ultrasonic transducers (UTs) inside the
reactor using a carrier mechanism. The carrier mechanism should have a proper
kinematic design to access critical locations of reactor internals and ensure that
transducers are positioned within close tolerances. Sweep arm scanner (SAS), as
a carrier mechanism, is being developed to inspect and map the complete core of
FBRs. Kinematic and dynamic analysis of motions is carried out to estimate capacity
of drive motors and linear actuators.
1 Introduction
There are two access openings for insertion of carrier mechanism into the reactor,
one is in-vessel transfer port (IVTP) access opening in large rotatable plug (LRP),
and other is observation port in the small rotatable plug (SRP), as shown in Fig. 1
along with dimensions. In order to scan the reactor core, SRP and LRP have to be
rotated, but even after maximum possible rotations, a gap of around 840 mm exists
between the centre of observation port and central sub-assembly (SA) as shown in
Fig. 2. This means that there should be some provision in the carrier mechanism
with the help of which UTs can reach the central subassembly. This is achieved by
providing an extendable part (sweep arm), which can be extended once it is inserted
inside the reactor. Hence, this mechanism has a closed configuration in which overall
dimension of SAS is less than 450 mm, so that it can be inserted and retracted through
observation port and an open configuration to reach till the central SA. Also, in order
to get continuous signal, UTs are required to be translated in radial direction of the
reactor core [2, 3], or else only some points can be scanned on the SA top. These
points need interpolation to obtain the approximate orientation of the SA which may
lead to errors.
SAS consists of L, R, Z and θ-motions. It has a foldable arm called as sweep arm
with an array of 6 equi-spaced down viewing transducers (DVTs) attached to it.
For scanning the core, SAS has to be lowered in to the reactor through the access
openings, with sweep arm in closed configuration. Inside the reactor, sweep arm can
be opened and closed; this motion is called as L-motion. After opening, to and fro
motion of the sweep arm in radial direction is called as R-motion. θ-motion imparts
rotation of SAS which is limited to a sector of 270 deg to avoid any mechanical
interaction with control plug components. The purpose of Z-motion is to adjust
the height of the entire mechanism above the top of the general level of SA. SAS
consists of two parts, viz. upper part and lower part. Total height of upper part is
around 3 m and enveloping diameter is 600 mm. All components of upper part are
located well above the roof slab and it mainly consists of support structures and
Z, θ, L and R-motion drive systems. Entire load of SAS will be taken by support
structures. Geared motors are provided for Z, R and θ-motion while a linear actuator
is proved for L-motion. In addition to geared motor, ball screw nut mechanism
and slewing bearing are also provided to accomplish the task of Z-motion and θ-
motion, respectively. Lower part of SAS consists of sweep arm, spinner tube and
shield plug subassemblies. Sweep arm subassembly (Fig. 3) consists of sweep arm,
holder, frame, ball screw and nut, guide rods and bush, gear shaft, bevel gears,
universal joint, connecting link and anti-friction bearings. Spinner tube subassembly
is provided with translational shaft, rotational shaft, mechanical stopper and spinner
tube. Sweep arm subassembly is assembled below spinner tube subassembly in such
a way that during L-motion whole sweep arm subassembly rotates about the hinged
connection provided between gear shaft and the rotational shaft which is rotated by
the up and down motion of translation shaft. While a mechanical stopper prevents
1004 A. Kumar et al.
3 Analysis of L-Motion
L-motion of SAS is executed with the help of frame and gear shaft (link 2), connecting
link (link 3), translational shaft (link 4) and rotational shaft (link 1) as shown in
Figs. 4 and 5. The rotational shaft remains stationary during L-motion. To carry out
L-motion, translational shaft is actuated by a linear actuator, placed in the upper part
Design, Analysis and Development of Sweep Arm Scanner … 1005
of the SAS. The capacity of this actuator has to be estimated for which dynamic
analysis has been carried out preceded by kinematic analysis. As forces required for
closing the sweep arm is higher than opening, these analyses have been done for
closing. The position, angular velocity and angular acceleration of the frame (link 2)
and connecting link (link 3) during closing are shown in Figs. 6, 7, 8, 9, 10 and 11.
4 Analysis of R-Motion
For R-motion, it is required to transmit rotary motion from vertical rotational shaft
to horizontal ball screw. For this, bevel gears will be used as shown in Fig. 13. The
rotational velocity of rotary shaft is kept at 4 RPM while the gear ratio of bevel gear
is 1.5. The total traverse of sweep arm (nut) is 140 mm. Maximum torque required
for R-motion using bevel gears was estimated to be around 105 N-mm.
Design, Analysis and Development of Sweep Arm Scanner … 1009
Use of bevel gear in sodium has not yet been fully unveiled, and hence, there is a need
to have other options available for use. Literature survey [5] suggests the possibility
of cylindrical joints being used inside sodium. A concept using only cylindrical joint
to transmit rotary motion from the rotary shaft to the ball screw is by the use of
three elbows arranged symmetrically within 2 housings, one of which is connected
to rotary shaft while the other is attached to the ball screw as shown in Fig. 14. For
simulating this concept in air, the coefficient of friction of 0.2 was taken, while for
sodium operation, it was chosen as 0.7. The corresponding torque versus time plot
for execution of R-motion using elbow joint is shown in Fig. 15. Cyclic fluctuation is
observed in torque required to carry out R-motion using elbow transmission which
may be due to additional torque requirement to overcome geometric constraints
which occur in each cycle in a way similar to toggle positions in a four-bar linkage.
The details of analysis of variation of torque with using elbow joint need to be studied
and analysed further including the possibility of carrying out such transmission under
practical conditions.
Torque required for executing various motions were estimated using simple hand
calculations and they were found to be comparable with the analytical results.
5 Summary
References
1. Barrett LM, McKnight JA, Fothergill JR (1984) Ultrasonic viewing in fast reactors. Phys Techno1
15
2. Patankar VH, Lalwani SK, Agashe AA et al (2014) Under sodium ultrasonic imaging system
for PFBR, Technology Development Article, issue number 336
3. Patankar VH et al (2011) Instrumentation & control system for under sodium ultrasonic scanner
of PFBR, NDE
4. RCC-MRx 2012 Section III—Tome 1—subsection K: examination, handling or drive mecha-
nisms
5. Nagaraja Bhat YV, Asokane C et al (2016) Literature survey on sweep arm scanner, an internal
report of Indira Gandhi Centre for Atomic Research
Kinematics of Three Segment Continuum
Robot for Surgical Application
1 Introduction
Continuum robots have proved to be very beneficial in the field of medical robotics.
They have made Minimaly Invasive Surgery (MIS) in surgical space safer and
more efficient. It is possible due to their inherent compliance, hyper-flexibility, and
dexterity. The platforms used for robotic surgical operation can be broadly catego-
rized into two groups, based on the applied surgical tools: laparoscopy and flex-
ible endoscopy [1–3]. Instruments used in laparoscopic surgery are rigid, long, thin
tube instruments, with inbuilt high-resolution light and cameras for visual feedback.
However, the flexible endoscopy makes use of a steerable endoscope with a camera
and narrow passage, which allows flexible gripper or forceps in order to access the
human anatomy via natural orifices or a small incision.
Forward kinematics of the continuum robot is based on the constant curvature
model, which can be analyzed by the Euler Bernoulli beam model [4]. Many instances
2 Preliminary Design
Most surgical treatments require reaching deep surgical space through a natural
orifice and uneven pathway without disturbing the peripheral tissues. Surgeons gener-
ally use straight or bendable endoscopes, surgical forceps, and additional instruments
entering through the natural orifices like the anus, nostril, mouth, ear canal, or a
small incision. However, those instruments only provide inadequate dexterity, due
to which some regions stay inaccessible or hardly reachable. In order to perform the
treatment of disease in those regions, conventional surgery (open surgery) needs to
be performed. Continuum robots and manipulators have the capability to provide the
missing dexterity and required stiffness. This paper presents a continuum robot design
for surgical applications based on pre-assumed robot parameters, and kinematic
assumptions are discussed below.
Current medical endoscopes are fairly large-sized with large diameters, and also they
are of passive-type. So it is difficult to insert into a narrow space like a sinus cavity. It
requires a flexible endoscope with a smaller diameter with some required dexterity
Kinematics of Three Segment Continuum Robot … 1013
in order to perform the surgical procedure. Thus, this paper proposes a prototype of
a continuum robot with a spring backbone. The proposed continuum robot has three-
segments; each segment has 50 mm length and 8 mm diameter of spacer discs. The
presented continuum segment is four tendon-driven and has two degrees of freedom.
Where tendons 1, 3, and 2, 4 have relative actuation. The CAD model of the designed
continuum robot is shown in Fig. 1.
The special features of the design are flexibility with required stiffness and better
bending capability of the whole body by using a compliant spring backbone.
The constant curvature bending assumption has been taken to derive the kinematics
for the continuum robot segment. Spring compression has not been considered until
now. By discretizing the curvature bending movement in 3D space with four inde-
pendent motions in order to find the modified Denavit–Hartenberg (D-H) parameters
[7]. A schematic diagram of the continuum segment with local coordinate frames is
shown in Fig. 2.
Based on the above-modified DH parameters, a conventional homogenous
transformation matrix approach can be utilized to derive the forward kinematics.
Continuum robot kinematics has been sub-categorized into two stages kinematic
mapping: robot-specific mapping (first stage) and robot-independent mapping
(second stage).
1014 S. Bamoriya and C. S. Kumar
Forward kinematics is the kinematic relationship to calculate the end tip position and
orientation for a given set of the tendon (actuator) lengths as input. Forward kine-
matics mapping is subdivided into two-stage kinematic mapping [4]. First stage kine-
matic mapping is between joint space parameters to the configuration space parame-
ters called as the robot-specific kinematic mapping. Second stage kinematic mapping
is defined between configuration space parameters to task space (3D workspace)
parameters called robot-independent kinematic mapping.
Robot-Specific Forward Kinematics. Continuum robots are basically tendon or
cable-driven and pneumatically actuated. The robot-specific forward kinematics
leads to calculating the configuration parameters of a continuum robot segment
when corresponds to a single segment robot when tendon lengths are given as the
input parameters. Hence these are robot-specific. The continuum robot model is
tendon-driven.
Any variation in the tendon actuators length i.e. l1 , l2 , l3 , and, l4 will yield the
change in endpoint position and orientation of the segment. For any joint actuation,
it deforms in a circular curvature with ‘S’ curvature length, bending plane angle
∅(from the +x axis), bending angle of curvature θ and a radius of curvature r .
q = [l1 , l2 , l3 , l4 ]T (1)
where d is the distance from the center of a section of the robot to the center of the
tendon actuator (which is the same for all tendon actuators) and ∅i specifies the angle
between the robot’s bending direction and the location of the i th actuator as shown
in Fig. 3.
Assumption: The spacer discs are supporting the backbone of the continuum robot,
with minimal disc thickness. Hence the length of each robot segment is represented
as:
l1 + l2 + l3 + l4 l1 + l2 + l3 + l4
S= ; Lc = (4)
4 4n
(i − 1)
∅i = π − ∅, where i = 1, 2, 3 and 4 (5)
2
l4 − l2
−1
∅(q) = tan (6)
l3 − l1
(l1 − 3l2 + l3 + l4 ) (l4 − l2 )2 + (l3 − l1 )2
θ (q) = S (7)
d(l1 + l2 + l3 + l4 )(l4 − l2 )
where T kk−1 represents the transformation consists Rkk−1 and P kk−1 as the orientation
matrix and position vector of end tip of the k th continuum segment, respectively (see
Eq. 8). whereas ∅k , θk and Sk are the configuration parameters of the k th segment
of the continuum robot. It has been further extended for three segment continuum
robot by successive multiplication of the transformation matrices corresponding to
each section for (k = 1, 2, 3).
The total transformation matrix T03 contains the position vector and the orientation
matrix of the end tip of the distal segment concerning the reference frame, of the
three-segment continuum robot.
Most of the surgical procedures like single port surgeries, Natural Orifice Translu-
minal Endoscopic Surgery (NOTES), and surgeries in deep anatomical space require
high dexterity across the surgical field. Continuum robot-based architecture is used in
the surgical domain due to its flexibility and dexterity. The workspace and dexterity
are the most important characteristics in order to design the robot for specific surgical
procedures.
1018 S. Bamoriya and C. S. Kumar
Table 2 Simulation criteria followed for four tendons-driven continuum robot workspace plot
No. of segments Segment length S θ range (for each ∅ range (for each Steps
(mm) segment) segment)
3 50, 50 and 50 [−π, π ] [0, 0] (planar case) 40
Workspace analysis is the basic criteria to design the robot manipulators, which
decides the robot’s maximum and minimum reach and joint actuation range to restrict
the link collision. In the case of the surgical robot, surgical space is the workspace
where the robot needs to be operated with required dexterity. Hence, it is the required
workspace for the designed robot. “Monte Carlo method” is used to get the workspace
with defined simulation criteria’s as shown in Table 2.
Three-segment continuum robot can bend uniformly in all the direction. Hence,
its workspace is symmetric about the z-axis. A planar workspace for three segment
continuum robot has been plotted in MATLAB software with simulation parameters
(see Table 2) shown in Fig. 4.
Robot dexterity can be measured by approaches, like condition number of the Jaco-
bian matrix, multiplication of the singular value of Jacobian matrix [15], etc. In
contrast, a more effective way is to use kinematic flexibility to define dexterity
[16]. Dexterity is basically the measure of the possible number of configurations
Kinematics of Three Segment Continuum Robot … 1019
with which the robot end effector (tip of the distal segment) can reach a specific
position within the workspace or, in other words, the number of possible solutions
of inverse kinematics [11]. Absolute and relative kinematic flexibility is measured
in order to find the dexterity for a specific position and workspace, respectively
[11]. Hence in this paper, we analyze the workspace dexterity distribution based
on relative kinematic flexibility by utilizing the workspace positional data density
approach. Workspace is a set of all unique robot configurations; however, many
specific positions can also be achieved by many configurations that can be visualized
in the presented work. Following criteria, as shown in Table 3, has been followed to
compute the workspace for dexterity analysis.
Figures 5 and 6 are representing the relative dexterity distribution across the
workspace for three segments and two-segment continuum robots, respectively.
Dexterity across the workspace is color mapped (see Figs. 5 and 6) on the basis
of the relative data density (normalized), which signifies the kinematic flexibility
across the workspace in color bounded region.
Table 3 Simulation criteria followed for four tendons driven continuum robot
No. of segments Segment length S θ range (rad) (for each ∅ range (for each Steps
(mm) segment) segment) (planar case)
3 50, 50 and 50 [−π, π ] [0, 0] 60
2 75 and 75 [−π, π ] [0, 0] 465
5 Conclusion
This paper has presented a comparative analysis of the workspace and its dexterity
distribution for two-segment and three-segment continuum robots, respectively. Here
we have found that the two-segment and three-segment continuum robot can be used
in regions where dexterity is required. Still, shape optimization will be required to
get higher dexterity in the minimum workspace during the robotic surgical operation
in confined surgical spaces. A high dexterous region (high data density region) will
be suitable for dual-arm motion coordination and operations, which requires high
steerability like contact therapy (biopsy, suturing, etc.).
Kinematics of Three Segment Continuum Robot … 1021
References
1. Beasley RA (2012) Medical robots: current systems and research directions. J Robot
2. Hashizume M et al (2008) New real-time MR image-guided surgical robotic system for
minimally invasive precision surgery. Int J Comput Assist Radiol Surg 2(6):317–325
3. Song HS, Chung JH, Kim KY, Lee JJ (2006) The development of human-arm like manipu-
lator for laparoscopic surgery with force Sensing. In: Proceedings of the IEEE international
conference on industrial technology
4. Webster RJ, Jones BA (2010) Design and kinematic modeling of constant curvature continuum
robots: a review. Int J Rob Res 29(13):1661–1683
5. Hannan MW, Walker ID (2003) Kinematics and the implementation of an elephant’s trunk
manipulator and other continuum style robots. J Robot Syst 20(2):45–63
6. Hannan MW, Walker ID (2011) Novel kinematics for continuum robots. Adv Robot Kinemat
227–238
7. Jones BA, Walker ID (2006) Kinematics for multisection continuum robots. IEEE Trans Robot
8. Garriga-Casanovas A, Rodriguez F, Baena (2018) Kinematics of continuum robots with
constant curvature bending and extension capabilities. J Mech Robot 11(1):1–12
9. Wu L, Crawford R, Roberts J (2017) Dexterity analysis of three 6-DOF continuum robots
combining concentric tube mechanisms and cable-driven mechanisms. IEEE Robot Autom
Lett 2(2):514–521
10. Li Z, Wu L, Ren H, Yu H (2017) Kinematic comparison of surgical tendon-driven manipulators
and concentric tube manipulators. Mech Mach Theory 107(September 2016):148–165
11. Li Z, Ren H, Chiu PWY, Du R, Yu H (2016) A novel constrained wire-driven flexible mechanism
and its kinematic analysis. Mech Mach Theory 95:59–75
12. Aristidou A, Lasenby J (2011) FABRIK: a fast, iterative solver for the inverse kinematics
problem. Graph Models 73(5):243–260
13. Zhang W, Yang Z, Dong T, Xu K (2018) FABRIKc: an efficient iterative inverse kinematics
solver for continuum robots. In: IEEE/ASME international conference on advanced intelligent
mechatronics, AIM, vol 2018-July, pp 346–352
14. Kolpashchikov D, Laptev N, Danilov V, Skirnevskiy I, Manakov R, Gerget O (2018) FABRIK-
based inverse kinematics for multi-section continuum robots. In: Proceedings of the 2018 18th
international conference on mechatronics
15. Yoshikawa T (1985) Manipulability of robotic mechanisms. Int J Rob Res 4(2):3–9
16. Lenarcic J, Bajd T (2013) Robot mechanisms. Springer, New York
Automatic Seed Cum Fertilizer Sowing
Machine with Water Dripping on Seeds
1 Introduction
The wheel period of the storm is antagonistically influencing the nation’s precipita-
tion portion between June to July. Poor or no precipitation has furthermore influenced
the nation’s precipitation share during the long stretch of June. In India, late monsoon
is observed in certain parts of the country like; the west parts of Punjab, Haryana,
Delhi, Rajasthan, Gujarat, Uttar Pradesh, south parts of Karnataka, Tamil Nadu,
Andhra Pradesh, parts of Telangana, Madhya Pradesh, Chhattisgarh and the leeward
side of Maharashtra. The late arrival of monsoon affects the yielding capability of
the crop. In this regard, farmers should be trained to cultivate in the right season
time using modern technologies and with the minimum usage of water. Groundnut,
cotton, maize, jowar, bajra, sesamum, and dal are the major rainfed crops in the
Rayalaseema region including Anantapur, Chittoor and parts of Kadapa and Kurnool
districts, while the dry land legume is cultivated in large areas and in other areas also
in the country. Farmers often suffer a heavy loss due to crop failure because of inad-
equate or no rains while lack of the best practices in cultivation is also considered
to be one of the reasons that are depriving farmers getting good revenue from the
commercial crop.
In the developing technique, much of the time used customary seeding movement
takes extra time and more work. The seed feed rate is even more anyway the time
required for the complete expense is more and the total cost is extended in view of the
hiring of equipment, labour. The traditional seed planting machine is less proficient
and tedious. The present period is walking towards the quick development of all
areas including the agriculture sector. To fulfil the future nourishment needs, farmers
need to actualize the new procedures which would not influence the soil surface yet
will increase the total crop production. Agriculture in India has a noteworthy history.
Today, India positions second worldwide in farm yield.
Agriculture is demographically the broadest budgetary portion and plays a basic
occupation in the general economy of India. For the development of Indian economy,
automation is important. The primary reason for automation in agribusiness is to
improve the general efficiency and production. Planting is ordinarily done physically,
which includes both invigorate (people and draught creatures), and this results in a
greater expense of development and deferral in planting.
This paper talks about a programmed seed cum manure planting machine with
water dribbling on seeds at the season of planting. This will assist the farmers with
cultivating the land regardless of whether the entry of the storm is late. The essential
goal of this proposed technique is to put the seed, required amount of water, compost
in lines at an ideal depth and spread the seeds with soil and give suitable compaction
over the seed. The endorsed amount of water, line to line dispersing, seed rate, seed
to seed spacing and profundity of seed position change from harvest to harvest and
for various agro-climatic conditions to accomplish optimum yields.
2 Literature Review
Agriculture has consistently been the backbone of India’s sustained growth. As the
quantity of occupants in India continues building up, the interest for produce grows
as well. Accordingly, there is a progressively significant necessity for different crops
in the farms and this hence requires profitable and productive machines.
Roshan et al. [1] gave a technique managing different planting strategies utilized
in India for seed planting and manure situation. The author likewise looked at the
customary planting strategy and the new proposed machine as far as execution tasks
and preferences.
Amol et al. [2] exhibited a technique to contrast customary planting strategies and
new proposed machines. The proposed machine gives detail depiction about line to
line separation, the seed rate, seed to seed space and composts situation which shifts
from harvest to harvest. This machine lessens the planting time, human endeavours
Automatic Seed Cum Fertilizer Sowing Machine with Water … 1025
and labour cost. Ramesh et al. [3] exhibited a concise survey of data about the different
kinds of advancements done in seed planting equipment.
As day by day, the labour accessibility turns into the incredible worry for the
farmers and labour cost is expanding, this machine decreases the endeavours and total
expense of planting the seeds and compost arrangement. Pranil et al. [4] proposed a
technique which manages the different planting strategies utilized in India for seed
planting and manure position. Robotized sunlight-based controlled seed planting
machine has been proposed by Sateesh et al. [5].
Kunal et al. [6] proposed a technique where he interchanged the confounded
gear framework by the Hall Effect sensor for simple seed planting and furthermore
diminish the need of labour. The Hall Effect sensor changes over the pivot into separa-
tion for which seed planting at a specific distance. Additionally, there is customizable
framework for planting at various distances. By utilizing this machine, the planting
should be possible line by line and separation is kept up. So this machine lessens the
endeavours and diminishes the expense of the seed planting process with incredible
efficiency and precision with reduce in the labour prerequisite.
Thorat et al. [7] proposed a machine that can plant different types and different
sizes of seeds and it can also automatically vary the space between two seeds while
planting. This increased planting efficiency and accuracy. Nageswara Rao et al. [8]
made a design and fabrication of agriculture robot (AGROBOT). It consists of fire
bird V robot and seed sowing mechanism. The seed sowing mechanism is operated
by the rotor with four blades and rotor is driven with the help of motor. Krunal
et al. [9] reviewed about seed sowing the robot and explained the advantages and
applications of robots in agriculture.
Thorat et al. [7] proposed a machine that can plant various sorts and various sizes
of seeds and it can likewise consequently fluctuate the space between two seeds while
planting. This expanded planting proficiency and precision. Nageswara Rao et al. [8]
made a structure and fabrication of agriculture robot (AGROBOT). It comprises of
fire bird V robot and seed planting component. The seed planting component is
operated by rotor with four sharp edges, and the rotor is driven with the assistance of
an engine. Krunal et al. [9] investigated about seed planting robot and clarified the
points of interest and uses of robots in agribusiness.
The existing methods say that there is an automatic seed sowing machine for all
the variety of seeds, but there is no mechanism for water dripping on the seeds. Our
research paper discusses an automatic seed cum fertilizer sowing machine with water
dripping on seeds at the time of sowing.
3 Proposed Method
The proposed strategy talks about a programmed seed cum manure planting machine
with water dribbling on seeds at the season of planting. This will push the farmers to
cultivate the land regardless of whether entry of the rainstorm is late. The essential
goal of this proposed technique is to put the seed, the required amount of water,
1026 T. Tirupal and D. Rajasekhar
and manure in lines at wanted profundity and spread the seeds with soil and give
appropriate compaction over the seed. The suggested amount of water, line to line
space, seed rate, seed to seed dispersing and profundity of the seed position shift
from harvest to harvest and for various agro-climatic conditions to accomplish ideal
yields.
This programmed seed cum compost planting machine with water trickling on
seeds basically works on ‘vertical-discontinuous work principle’ which alludes to
the vertical development which can be trailed by an individual body in an agricultural
field and actualizes its broken activity in connection to the flat profession. According
to name, this machine is utilized for planting seed, compost and dribbling water on
the seed. The schematic diagram of the proposed technique is shown in Fig. 1 and
the proposed structure with water tank is shown in Fig. 2. The proposed machine
includes:
• Drill the ground
• Sow seed inside the drilled hole
• Put fertilizer in the hole
• Drip the water on the seed
• Cover the gap with the assistance of adjuster.
Firstly, a hole is drilled with the assistance of a 4 inch land boring apparatus having
shaft distance across of 7 mm and diameter of edges—25 mm with a depth of cut of—
76.2 mm. This is kept running with the assistance of an engine of 300 rpm and 12 kg-
cm torque. This is associated with 12 V and 7 A DC battery. This is straightforwardly
associated with the solar panel-based board through which it gets charged. This
Digging
Charging Con-
Mechanism
troller
Seed Sowing
Mechanism
DC MOTOR
Battery Fertilizer Insert-
ing Mechanism
Water Dripping
Mechanism
4 Conclusion
The paper discusses an automatic seed cum fertilizer sowing machine with water
dripping on seeds at the time of sowing. This will help the farmers to cultivate the
land even if the arrival of the monsoon is late. The essential target of this proposed
strategy is to put the seed, the required amount of water, compost in rows at wanted
profundity and spread the seeds with soil and give legitimate compaction over the
1028 T. Tirupal and D. Rajasekhar
seed. The prescribed amount of water, line to line separation, seed rate, seed to seed
spacing and profundity of the seed situation shift from harvest to harvest and for
various agro-climatic conditions to accomplish ideal yields. The present proposed
method project can be used for multiple applications in agriculture like seed sowing,
inserting fertilizer and dripping water on to the seed at the time of sowing. This type
of mechanism helps the farmers in some parts of our country to grow the crop with in
time even if the late monsoon is observed. This type of system makes the crop to yield
to a maximum extent. This type of mechanism is also an extension to the existing
system where an existing system contains automatic seed and fertilizer sowing except
the dripping water mechanism. If we include this water dripping mechanism, then
we can try to help the cultivators to a maximum extent in and around our country.
References
1. Roshan VM, Gajanan PT, Swapnil KA (2013) Design and implementation of multi seed sowing
machine. Int J Mech Eng Robot Res 2(4):422–429
2. Amol BR, Pavan DS, Sumit BP, Keshav KS (2014) A review on multi-seed sowing machine. Int
J Mech Eng Technol 5(2):180–186
3. Ramesh D, Girishkumar HP (2014) Agriculture seed sowing equipments: a review. Int J Sci Eng
Technol Res 3(7):1987–1992
4. Pranil VS, Amit W, Ashish S, Bhushan P, Rakesh B, Saurabh K (2015) Solar powered seed
sowing machine. Glob J Adv Res 2(4):712–717
5. Sateesh M, Mahantesh T, Sujata H (2016) Automated solar powered seed sowing machine. J
Adv Sci Technol 12(25):478–485
6. Kunal AD, Omkar RS, Megha SB, Achal AJ, Priyanka SC (2017) Design and development
of automatic operated seeds sowing machine. Int J Recent Innov Trends Comput Commun
5(2):277–279
7. Thorat SV, Madhu LK, Patil GV, Patil R (2017) Design and fabrication of seed sowing machine.
Int Res J Eng Technol 4(9):704–707
8. Nageswara Rao A, Pichi Reddy S, Raju N (2018) Design and development of seed sowing
AGROBOT. J Emerg Technol Innov Res 5(5):783–787
9. Krunal G, Shubham P, Mahima C, Hetal P, Bhushan P (2019) Seed sowing robot—a review. Int
J Adv Eng Res Dev 6(2):79–81
Automatic Drip Irrigation Control
System for Paddy Fields in Depleting
Water Resource Areas
1 Introduction
Over decades, agriculture is one of the most primitive professions around the world.
More than half of the Indian population resides on agriculture as their primary source
of employment. Even in Indian economy, agriculture contributes a major part in calcu-
lating the GDP [1]. Water is the main substance in the world which is responsible for
the health and wealth of all living things. Almost 80% of water present in earth need
to be used for farming. The need for water increases with rapid growth of popula-
tion. Overwatering and underwatering both the terms lead to ineffective farming [2].
Efficient usage of water for agriculture is very much important at present because of
constantly decreasing rain fall. Drip irrigation is a very successful method of water
usage for agricultural needs. Water and nutrient use efficiencies were found to be
maximum in drip irrigated paddy with 52% water savings than traditional flood-
irrigated paddy [3]. Paddy fields are very common in south India which consumes
2 Proposed System
into neural network backpropagation algorithm to analyze to find out the optimum
irrigation conditions (Figs. 1 and 2).
The system continuously monitors soil moisture using moisture sensor model FC-
28. Figure 3 shows the soil moisture sensor used in the automatic irrigation system. It
takes input voltage of 3.3–5 V and generates output voltage of 0–4.2 V of both analog
and digital signals. FC-28 gives the value from 0 to 1023 which can be measured in
percentage from 0 to 100, respectively.
Light intensity of the surrounding environment is measured using light-dependent
resistor (LDR) module as shown in Fig. 4. The photo resistor of model 5528 is used
for the qualitative detection of ambient light intensity. Maximum voltage (Vcc) of
5 V and maximum current of 6 mA.
Humidity of the surrounding environment is measured using DHT11 humidity
sensor. This sensor can measure relative humidity of 20–90% RH of accuracy ±5RH
in the temperature range of 0–50 °C. Figure 5 shows the humidity sensor DHT11
used in this proposed irrigation system.
Figure 6 shows the moisture sensors and solenoid valves placements in the paddy
field. Humidity and light intensity sensors can be placed anywhere in the field,
1032 K. Saravanakumar et al.
FARMER’S
MOBILE
PHONE MANUAL
CONTROL
MOISTURE SENSOR
ARDUINO UNO
SOLENOID
LIGHT INTENSITY ATmega328P VALVE 1
SENSOR MICROCONTROLLER
SOLENOID
VALVE 2
SOLENOID
HUMIDITY SENSOR
VALVE 3
SOLENOID
VALVE 4
because the humidity and light intensity will not change much in a particular zone.
But moisture content in a field which has to be analyzed throughout the field. Due to
ups and downs in land area, soil of the paddy crop will not be equally moisturized.
So moisture sensor should be placed in proper places after carefully surveying the
field.
Table 1 explains the scheduling of irrigation for different varieties of paddy. In
our proposed automatic irrigation system, we are taking short and medium duration
variety for water management analysis.
Automatic Drip Irrigation Control System for Paddy Fields … 1033
MOISTURE
SENSOR 2
SOLENOID VALVES
IN IRRIGATION LINE
HUMIDITY AND LIGHT
INTENSITY SENSOR
MOISTURE
SENSOR 1
Table 2 indicates the moisture level variations to be maintained in the soil during
the growth period of paddy.
MATLAB neural networks toolbox (nntool) is used as the simulation tool in this
paper. Sample data are collected from a normalized field without any pre-installation
of higher end logics or neural networks. We are tabulating the various levels of
1034 K. Saravanakumar et al.
Table 2 Depth of
Stages of crop growth Depth of submergence (cm)
submergence chart for low
and medium duration paddy At transplanting 2
variety After transplanting for 3 days 5
Three days after transplanting 2
upto maximum tillering
At maximum tillering (in fertile Drain water for three days
fields only)
Maximum tillering to panicle 2
initiation
Panicle initiation to 21 days 5
after flowering
Twenty one days after flowering Withhold irrigation
moisture, light intensity, and relative humidity. Corresponding flow rate changes
also measured in terms of liters per hour (Figs. 7 and 8; Table 3).
By this proposed method, total amount of water required for the paddy crop is
limited to 600 mm.
Automatic Drip Irrigation Control System for Paddy Fields … 1035
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
Data Data
7000 Fit 7000 Fit
Y=T Y=T
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
Validation 6
10
Test
Best
Mean Squared Error (mse)
5
10
5
Mu = 10000, at epoch 6
10
0
mu
6
10 10
-5
10
5 0
10 0 1 2 3 4 5 6
0 1 2 3 4 5 6 6 Epochs
6 Epochs
Table 3 Stages of cultivation and various sensor data and flow rate variations
Stages of Moisture % Light intensity Relative Average flow Average
cultivation (Lux) humidity (%) rate (l/h) irrigation
depth (mm)
Field 90.7 613 66 8258 198
preparation
Planting 87.85 652.75 64.95 6169 297
Flowering 88.75 653.175 64.925 1008 97
Maturity 89.1 639.6417 65.29167 190 10
References
1. Subathra MSP, Blessing CJ, Thomas George S, Thomas A, Dhibak Raj A, Ewards V (2016)
Automated intelligent wireless drip irrigation using ANN techniques. IEEE, pp 1–14
2. Dela Cruz JR, Baldovino RG, Bandala AA, Dadios EP (2017) Water usage optimization of
smart farm automated irrigation system using artificial neural networks. In: Fifth International
conference on information and communication technology (ICoICT), pp 1–6
3. Bhardwaj AK, Pandiaraj T, Chaturvedi S, Singh TC, Soman P, Bhardwaj RK, Labh B (2018)
Growth, production potential and inputs use efficiency of rice under different planting methods
in drip irrigation. Curr J Appl Sci Technol 1–9
4. Barkunan SR, Bhanumathi V, Sethuram J (2019) Smart sensor for automatic drip irrigation
system for paddy cultivation. Comput Electr Eng 73:180–193
5. Ananthi N, Divya J, Divya M, Janani V (2017) IoT based smart soil monitoring system for
agricultural production. In: International conference on technological innovations in ICT for
agricultural and rural development (TIAR2017). IEEE, pp 1–6
6. Rajalakshmi P, Mahalakshmi SD (2016) IOT based crop-field monitoring and irrigation automa-
tion. In: 2016 10th international conference on intelligent systems and control (ISCO). IEEE,
pp 1–6
7. Kang MS, Kim SM, Park SW, Lee JJ, Yoo KH (2007) Assessment of reclaimed wastewater
irrigation impacts on water quality, soil, and rice cultivation in paddy fields. J Environ Sci
Health Part A 42(4):439–445
8. Gao Z, Ni J, Zhu Y, Jiang Q, Cao W (2018) Water-efficient sensing method for soil profiling in
the paddy field. Int J Agric Biol Eng 11(4):207–216
9. Bhardwaj AK, Pandiaraj T, Soman P, Bhardwaj RK, Singh TC (2018) Drip irrigation scheduling
for higher growth, productivity and input use efficiency of direct seeded basmati rice in Indo-
Gangetic plains for climate resilient. Int J Environ Clim Change 332–340
Investigating the Ambient Thermal
Loading Failure of Lead–Acid Battery
Based on Thermal Analysis
1 Introduction
In all aspect of electrical components, electrical energy is the main source of power for
continuous operation. In case of power failure, the secondary power source is neces-
sary to ensure that the component operation remains successful. Thus batteries are
fulfilling the necessary demands. However, the performance of the battery is varied
in different environmental and design factors. But the major influence on the battery
depends on temperature. So, the analysis of temperature influence is very essential to
improve the performance and lifetime of battery. The battery performance reduction
depends on the plate performance as the number of cycling process increases. The
main cause for the fast capacity decay is due to shedding and softening in positive
active material, resultant of which the positive electrode’s cycling performance was
significantly compromised. The author had performed the experiment to improve the
performance of positive electrode with the help of additives (Sb2 O3 and SnSO4 ). As
a result, the performance of plate was improved at 40 and 60% discharge [1]. Spanos
et al. [2] analyzed the performance of plates at reverse charging. From the obtained
result, it was inferred that reverse charging results in growth of discharge capacity
on negative electrodes, whereas Pb–Ca–Sn positive capacity tends to decrease. Rand
[3] mainly focused on anode and cathode plates of lead–acid battery and worked on
different methods to lengthen the life time of battery. In addition, anode and cathode
plate characteristics are also determined on rate of change of temperature. Catherino
[4, 5] worked on thermal runaway effect on lead–acid battery. In brief, the observed
effects are found to be similar to the electrolyte distribution in the separator. Thus,
modifying the properties of AGM separator can yield a better method for control-
ling thermal runaway and decreases the failure mode. From all these analysis, it was
inferred that battery efficiency and plate characteristics gets easily affected due to
the thermal effect. This made the researchers to consider the effects of temperature
influence. Thus, few researchers analyzed the thermal influence (DSC and TGA)
on lead–acid battery materials. From this, it was noticed that the temperature rise
will minimize the plate material, thereby deteriorates the total lifetime of the battery
[6, 7].
In this study, the differential scanning calorimetry (DSC) method was used to
measure the differences in energy inputs of a sample as a function of temperature.
While the sample and reference materials were exposed to controlled temperature
[7], the change of energy inputs was observed between the materials. Additionally,
thermogravimetric analysis (TGA) was used to measure the mass of substances in a
sample as a function of temperature variations [7].
2 Experimentation
Lead–acid 12 V/7.2 Ah battery was utilized for this analysis. For heating purpose,
two Ni–Cr heating coils were used inside the wooden chamber. The chamber was
fully closed and equipped with fan to spread the generated heat uniformly in all four
directions of the battery. To measure the temperature, three K-type thermocouples
were used. It was connected with DAQ card and processed using LabView software
as shown in Fig. 1. In this condition, the battery was charging and discharging for
10 full cycles and the temperature was increased by 10 °C till it reaches 60 °C.
The battery gets heated due to internal resistance and surrounding environmental
temperature destroy the battery completely and at the same time discharging the
battery capacity with less time. Finally, the battery cover was removed, and both
anode and cathode plates were separated and subjected to thermal analysis as shown
in Fig. 2. Thermogravimetry analysis machine NETZCH (STA 449 F3 Jupiter) is
available at PSG iTech Coimbatore. This equipment is capable of measuring the
Investigating the Ambient Thermal Loading Failure of Lead–Acid … 1039
Fig. 1 Schematic of
experimental setup
mass changes and thermal effect between room temperature (27 °C) to 1550 °C. The
equipment can able to measure the materials including inhomogeneous substances
with good flexibility.
For the analysis of anode and cathode plates with respect to temperature, it provides
a deep knowledge about plate heat resistance and stability. Also, the battery behavior
1040 T. Chinnadurai et al.
From Fig. 3, the thermal loading for anode plate sample is increased slowly and
corresponding glass transition temperature is measured. During the initial thermal
loading, material exhibits exothermic nature up to 150 °C. Further increasing the
temperature up to 375 °C, the material changes in steady state but once the tempera-
ture reach 400 °C again, material starts changing into endothermic peak. The complex
peak is observed at 424 °C as an onset value. However, the DSC measurement takes
only about 10 min. While charging and discharging, there may be chance for the
material to react with atmospheric gases (CO2 ); as a result, the material gets turned
into carbonated. When the percentage of the carbonates is increased, the battery
performance is reduced for every 15%.
The cathode material (Fig. 4) exhibits very sharp endothermic peak due to crys-
talline behavior. Like anode analysis, cathode plate also initially shows exothermic
nature then turns into endothermic behavior. Beyond onset temperature, the material
TGA analysis is used to identify the material mass loss with respect to temperature
influence. The analysis is carried out in two different areas of both anode and cathode
1042 T. Chinnadurai et al.
Fig. 5 Thermal loading influence on anode analysis, the test portion is taken at bottom of the plate
Fig. 6 Thermal loading influence on cathode, the test portion is taken at bottom of the plate
materials. This will help us to understand and identify the nature of degradation of
material due to temperature loading. Figures 7 and 8 show the mass loss of anode
and cathode plate at the terminal lead position. The losses are observed at 450 °C due
to chemical corrosion and ion exchange, which leads to the material loss. Another
phenomenon for material loss is growth of grids and corrosion which also degrades
the life time of battery. The results depict that the rate of grid corrosion is influenced by
Investigating the Ambient Thermal Loading Failure of Lead–Acid … 1043
material structural properties. The internal resistance of the material is also influenced
by the temperature.
Investigating the Ambient Thermal Loading Failure of Lead–Acid … 1045
4 Conclusion
The best efficient thermal analyzing methods to find the basics of material nature are
DSC and TGA. Thermal analytical methods give a new information that the different
temperature influence is observed at single plate. This nature will lead to changes in
material composition and physical properties in a single plate. TGA results clearly
show the material degradation behavior. As like previous analysis, this also achieves
different material loss value at single plate.
References
1. Yang S, Li R, Cai X, Xue K, Yang B, Hu X, Dai C (2018) Enhanced cycle performance and
lifetime estimation of lead-acid batteries. New J Chem 42(11):8900–8904
2. Spanos C, Berlinger SA, Jayan A, West AC (2016) Inverse charging techniques for sulfation
reversal in flooded lead-acid batteries. J Electrochem Soc 163(8):A1612–A1618
3. Rand DAJ (1996) Energy storage in lead-acid batteries: the Faraday way to sustainability. Philos
Trans R Soc Lond Ser A Math Phys Eng Sci 354(1712):1529–1544
4. Catherino HA (2006) Complexity in battery systems: Thermal runaway in VRLA batteries. J
Power Sources 158(2):977–986
5. Catherino HA (2007) Secondary batteries: lead acid battery thermal runaway (No. TARDEC-
18509-RC). Tacom Research Development and Engineering Center, Warren, MI
6. Zhu H, Tan JJ, Xu ZL, Xu JS (2009) Experimental research on charging characteristics of a
pressure-controlled VRLA battery in high-temperature environments. J Zhejiang Univ-Sci A
10(3):418–422
7. Matrakova M, Pavlov D (2006) Thermal analysis of lead-acid battery pastes and active materials.
J Power Sources 158(2):1004–1011
Distance Operated Manipulator: A Case
Study for Rose Plucking
Abstract Few applications such as cutting shrubs and trees with prickles lead to
damage to the hands while cutting it conventionally. The research has been under-
taken for ease of rose plucking. A distance operated manipulator is designed and
manufactured to overcome the issue of cutting roses by conventional way. The
working of the tool is controlled by a lever operated mechanism. By a Bowden
cable, the lever is connected to a brake caliper. The force applied by fingers at the
lever is amplified at the point of application due to mechanical advantage. The end-
effector consists of a gripper-cum-cutter. When the lever is operated, the stem of
the rose is first gripped and then shearing of stem takes place. This unique feature
of mechanism results into safe harvesting. Due to its simplest construction, it is a
beneficial method. Trials have demonstrated the effective use of manipulator for rose
plucking.
1 Introduction
Floriculture means flower cultivation is being practiced in India since times immemo-
rial, and floriculture has blossomed into a viable business only in recent years. Consid-
ering the potential, this sector has helped in generating income and employment
opportunities. Commercial cultivation of cut flowers such as rose, orchids, gladi-
olus, carnation, anthurium, gerbera and lilies has also become popular. Rose flower
is one of the most important flowers commercially. Roses are dicot herbaceous often
cultivated as a garden or potted annual flower. The required working hours to produce
a rose are 3.770–5.240 h/ha, and cutting production accounts for 11.8–19.3% of this
time. Therefore, the mechanization of cutting production has been strongly desired
with rose industries. The acceptability of roses in trading particularly for exporting
to European countries depends on adequate post-harvest longevity [1]. Harvesting
of flowers is an important operation in floricultural practice. Economics as related
to timeliness, competition for labor and good market prices have pushed for mecha-
nization of almost all agricultural operations. Mechanization of small farms is uneco-
nomical as per Indian scenario. In small farms, harvesting of roses is carried out by
conventional methods, such as traditional tool-scissors, pruner, lopper, etc., where
labors work either with bare hands or wearing glows which may hamper human body
them due to prickles on rose. To avoid damage, drudgery and to increase the output,
a simple mechanism is needed with precise shearing action. This paper presents a
simple mechanism for precise shearing and gripping action.
2 Literature Review
Many conventional methods are available for plucking and pruning the roses. Consid-
ering the most traditional tool-scissor, this can pluck the branches by shearing. The
pruner, more advanced over scissors, is having the ability of plucking with quite
ease [2]. Again for distance actuation, the tools like lopper can be used which may
have telescopic rod which will increase its reaching area [3]. Loppers are a type of
scissors used for pruning twigs and small branches, like secateurs with very long
handles. Figure 1 shows the harvesting of roses by lopper and scissor. Loppers are
the largest type of manual garden cutting tool. This special kind of tool is operated
by both the hands [4]. The tools mentioned above should not be used with bare
hands, as pickles can harm the hands of worker, and hence it becomes necessary to
wear gloves. As both the hands are engaged while operating the tools like loppers,
etc., the problem may be faced for handling the plucked roses which may fallen off
the ground. It will cause damage to the roses reducing value added to it. We have
observed some issues of human factors (ergonomically discomfort) associated with
all these conventional hand tools. The angle made between thumb and first finger
while using conventional scissors is 30°, which develops stresses on hand. The manip-
ulator described in this paper will be cost effective and able to resolve problems faced
during plucking the roses. The simpler construction and cost-effectiveness make it
affordable to small-scale farmers.
As the roses have prickles, it is quite harmful to cut the roses manually. Our design
makes it easier and user friendly. The design uses links and a gripper with extended
attachment of cutter. The device is small and portable.
4 General Specifications
A robot with pneumatic actuator, sensors and picking mechanism can be used to
automate the process of rose harvesting. But for small-scale rose harvesting, it is
not feasible due to high installation cost of a robot. To provide an alternative for
1050 U. K. Mehta et al.
The basic criteria considered in the design of this device are included:
• Easy to use and handle
• Lightweight
• Height adjustable and flexible
• Cheap and simple design
• Fabrication with locally available materials.
Brake lever: Carbon fiber (CFRP) or aluminum is commonly used for brake lever
material. Aluminum is used where high strength and toughness are required which
considerably contributes to increase in weight of brake lever. Thus, carbon fiber
(CFRP) is used in this design due to its good rigidity, resistance to corrosion and
high strength to weight ratio.
Telescopic rod: Carbon fiber or aluminum alloy is used for telescopic rod due to its
lightweight, intermediate modulus which often provides high resistance to bending.
Grippers: Single-pivot side pull brake caliper is used. EPE foam sheet is attached
to the brake shoes, which grips the rose stem.
Distance Operated Manipulator: A Case Study for Rose Plucking 1051
Blades: Rectangular blades of carbon steel 75Cr1 are used to shear rose stem as it
allows high hardness and durability than the stainless steels.
Bowden Cable: It is a flexible type of cable, which is used to transmit mechanical
force.
Handle grip: Foam, plastic, or matte rubber-based anti-skid grip.
5.5 Dimensions
The pull force at the point of application is calculated by the principle of moment.
Figure 3 shows a schematic diagram of the device.
P ×L = F ×b (1)
where
F = Force at point of application (fulcrum).
b = Distance between wire holes to nut and bolts contact point (fulcrum).
P = Hand force acting to the lever surface.
L = Distance between wire holes to the lever surface.
The induced bending stresses are calculated for both materials (aluminum and
carbon fiber) using primary equations and stresses induced are within permissible
limits.
Induced bending stress (aluminum material): 28.51 N/mm2 .
Induced bending stress (carbon fiber material): 26.85 N/mm2 .
1052 U. K. Mehta et al.
While designing hand tools, it is essential to take into account the hand and finger
force capabilities. Hand tools require diverse and sometimes extreme levels of force
exertion depending on the movements involved; hence, in order to minimize discom-
fort and injuries of the upper extremities, finger force capabilities needs to be consid-
ered. Movements and exertions of the upper extremities, such as reaching, grip-
ping and pinching, combined with repetition in a forceful and/or awkward manner
are known contributing factors to the precipitation and aggravation of CTDs [6].
The range of motion (ROM) of the hand is the most commonly used functional
measurement variable. Investigations concerning the strength capabilities of the hand
began with explorations of grip strength. Average grip strength for males is stated
as 504.2 N and that of females as 311.0 N [7]. However, strength measurements
can vary depending on the instructions provided to the individual and the postures
maintained during the force exertions. Positioning of the arm can significantly affect
the maximum force exertions produced during grip. The Altezza level angle should
be 100°-150° while operating the manipulator for comfortable holding.
In general, a hand holding a manipulator could be abstracted as a ‘first’ type of
lever as shown in Fig. 4a, and in this case, the fulcrum is located between the load
and effort. The effort arm is usually kept more than the load arm to get mechanical
advantage.
Distance Operated Manipulator: A Case Study for Rose Plucking 1053
If we extend our view to the elbow, the whole system could be abstracted as a
‘third’ type of lever as shown in Fig. 4b, the effort is located between the load and
the fulcrum. In this case, the load arm is always greater than the effort arm and the
mechanical advantage is less than 1. Longer arm and a heavier mass could induce
fatigue on the biceps. As a result, using a heavy handle for better handling could
further increase arm fatigue, so the weight of handle is kept as low as possible [8].
6 Conclusion
References
1. Fard SHH et al (2008) Determining the shear strength and picking force of rose flower (Rosa
hybrida L.). In: 10th international congress on mechanization and energy in agriculture, 14–17
Oct 2008, Antalya-Turkiye
2. https://en.wikipedia.org/wiki/Pruning_shears
3. https://growerstools.com/blog/everything-need-know-about-loppers
4. https://en.wikipedia.org/wiki/Loppers
5. https://www.google.com/search?q=rose+harvesting+images&oq=rose+harvesting+images&
aqs=chrome..69i57.8760j0j7&sourceid=chrome&ie=UTF-8
6. Palanisami et al (1994) The effect of sitting on peak pinch strength. In: Aghazadeh F (ed)
Advances in industrial ergonomics and safety VI. Taylor & Francis, New York, pp 587–594
7. Schmidt et al (1970) Grip strength as measured by the Jamar dynamometer. Arch Phys Med
Rehabil 321–327
8. Bhandari VB (2010) Design of machine elements, 4th edn., pp 121–122
Topology Structure Design of Fish-Based
Propulsive Mechanisms
Abstract The research presented here has its centre point on imitation of fishes
and the kinematics of propulsive mechanisms to be used in efficient and versa-
tile autonomous underwater vehicles (AUVs). Today’s AUVs are based on subma-
rine. The paper started with the analysis of topological structural attributes of 4–11
link mechanisms with 1–6 DOFs. Using Hong Sen Yan’s creative design theory for
mechanical devices, all the possible combinations of propulsive mechanism were
synthesized to generate an atlas of 1149 new propulsive mechanisms subject to
isomorphism, which will provide more inputs in the design and fabrication of the
AUV models.
1 Introduction
Biomimetics is the study of shape of naturally existing biological living beings and
their locomotion techniques employed especially for synthesizing similar things by
the use of artificial mechanisms which can imitate actual mechanisms Sfakiotakis [1]
and Salazar [2]. Fish along with other species like birds are ideal species to mimic.
The different types of propulsion methods employed by fishes (see Fig. 1).
The fish swimming propulsion types mentioned above is being used by vari-
eties of fishes in their daily locomotion. Available fish-based swimming propulsion
is classified into different types, such as Thunniform, Anguilliform, Carangiform,
Ostraciform, Labiform, Amiiform, Diodontiform, Gymnotiform, Rajiform, Balisti-
form and Tetradontiform. These locomotion techniques are employed by fishes which
depend upon their body shape and also in accordance with the habitat in which they
are found. Some of the species of fishes also make use of their appendages specif-
ically fins (dorsal, anal, pectoral, caudal) alone and also in combination of two or
more type of fins in order to achieve a near perfect propulsive mechanism which will
help the fishes to swim swiftly in water.
Based on the measures used in the design of existing mechanisms, Dr. Hong Sen
Yan presented creative design procedures for the creation of all the feasible topolog-
ical layouts of mechanisms by Yan [3]. Flowchart (see Fig. 2) shows the proposed
followed for the creative design of mechanisms. The mains steps involved here can
be inferred from the book of creative design methodology by Yan [3].
The creative design theory proposed by Dr. Hong Sen Yan has to be followed
successively, viz. Generalization, Number Synthesis, Specialization, Particulariza-
tion and finally Atlas of new propulsive mechanisms.
The main steps that are there in creative design methodology are as follows:
1. Identifying existing designs with required design enumeration which the
designers would like to have and achieve the topological features of the said
propulsive mechanism designs so developed.
2. Selecting existing designs arbitrarily and transforming it into equivalent
kinematic chain.
Topology Structure Design of Fish-Based Propulsive Mechanisms 1057
3 Procedure Followed
In this paper, the study started with exploring different types of propulsive mech-
anisms that are already there in different patents and research papers, which have
1058 G. P. Manik and P. Dorlikar
Fig. 3 A propulsive
mechanism designed using
Epicyclic Gear Train by
Jiangsu University of
Science and Technology [5]
the ability to simulate the fin-based motion of swimming beings. Some of equivalent
kinematic diagrams (see Figs. 3, 4, 5 and 6). After going by the existing mechanisms,
the next step is to make equivalent mechanisms using different kinematics links and
thereafter by number synthesis technique for the number of links, joints whether the
said kinematic mechanism follows kutzbach–grubler criteria (mobility equation) for
the DOF’s available in mechanism so as to find out the number of inputs required,
when actually realizing the mechanisms.
Fig. 4 A propulsive
mechanisms designed using
simple links by Dumas [6]
Fig. 5 A propulsive
mechanisms designed using
simple joints by Pablo
Valdivia y Alvarado [7]
Topology Structure Design of Fish-Based Propulsive Mechanisms 1059
4 Equations
1. The basic equation used in the synthesis of mechanism is the mobility equation
by Grubler as follows
F = (3N − 1) − 2L (4.1)
(N + 1) − F
L= = J −N +2 (4.2)
2
3. Maximum no. of loops are given by
N L = k C1 + k C2 + k C3 + · · · + k Cn (4.3)
3N − (F + 3)
L= (4.6)
2
1060 G. P. Manik and P. Dorlikar
N
K = when F is odd (4.7)
2
N +1
K = when F is even (4.8)
2
8. In the synthesis of generalized kinematic chains, the selection of links is a crucial
step is to generate chains. It is generally expressed as
A L = N L2 /N L3 /N L4 /N L5 . . . (4.9)
where N L2 means the no. of binary links, N L3 means no. of tertiary links and so
on. The link selection of chains with N L links and N J joints can be obtained by
solving the below two equations:
N L2 + N L3 + N L4 + · · · + N Lm = N L (4.10)
2N L2 + 3N L3 + 4N L4 + · · · + m N Lm = 2N L (4.11)
m = N J − N L + 2 when N L ≤ N J ≤ 2N L − 3 (4.12)
N L ≤ N J ≤ N L (N L − 1)/2 (4.14)
propulsive mechanism and only taking the mechanisms that have unique propulsive
mechanisms.
Also from the computer program used by the research scholars mentioned in the
acknowledgement and for the purpose of justification of the results, kinematic chains
are drawn for mechanism with 9 links and having 2 degrees of freedom. Similarly,
the same procedure is followed to realize other mechanisms with different number
of links at various degrees of freedom.
For drawing the chains, Joint-Loop Adjacency Matrix (JLAM) and Joint-Joint
Adjacency Matrix (JJAM) as given in Eqs. (4.4) and (4.5) play a vital role. In case
of the JLAM and JJAM matrices, the order of the matrices is decided by the no.
of links and the matrix is a square and the elements of the matrix are determined
by the if a link or a particular joint is connected to a particular joint and loop, then
in that particular row and column number 1 is assigned and if the said pair is not
connected, then the entry in that particular row and column in 0. Depending upon
the above-mentioned theory matrix for a specific arrangement (see Figs. 7 and 8).
5 Results
After taking into consideration, the data compiled in the link assortment table above
following result table is generated and the results obtained for 9 links and 2 degrees
of freedom mechanisms in order to compare with the results obtained in the previous
research conducted are as shown in Table 2.
6 Conclusion
could be found easily out along with the coupler curves as shown in the software to
envisage the actual mechanism when in motion. Thereafter, the best suitable mech-
anism which perfectly imitates the actual fish motion is selected for usage in the fish
inspired propulsive device. The mechanisms presented in (see Fig. 9) examples of
fin undulation and oscillation mechanisms.
Note: that the mechanism presented below represent the propulsive mechanism of
fishes and the actual length of the mechanism rely upon the real size of the fish and
the prototype size into which the mechanism is to be installed.
Topology Structure Design of Fish-Based Propulsive Mechanisms 1063
(i) (i i) (iii )
Fig. 9 Unique mechanisms realized using commercially available SAM 7.0™ software
1064 G. P. Manik and P. Dorlikar
Fig. 9 (continued)
Acknowledgements We beholden towards all those who had helped us in gratification of this
paper. We do oblige to the following research scholars who have helped us while completing the
kinematic section of the paper.
Dr. Gurusharan Singh Bedi (Director Technical Education, Chhattisgarh).
Dr. Shubhashis Sanyal (Professor NIT Raipur, Chhattisgarh).
Dr. Hong Sen Yan (National Cheng Kung University, Taiwan, China).
Dr. Yii Wen Hwang (National Chung-Cheng University, Taiwan, China).
References
1. Sfakiotakis M, Lane DM, Bruce J, Davies C (1999) Review of fish based swimming modes for
aquatic locomotion. IEEE J Oceanic Eng 24(2)
2. Salazar R, Fuentes V, Abdelkefi A (2018) Classification of biological and bio-inspired aquatic
systems: a review. IEEE J Oceanic Eng 148:75–114
3. Yan HS (2008) Creative design of mechanical devices. Springer-Verilog, Singapore
4. Zhang T, Zhou C, Wang C, Zhang X (2011) Flapping wing mechanism design based on mechan-
ical creative design theory. In: International conference of mechatronics science, electric
engineering and computer
5. Shuzheng W, Zhu WX (2014) Tailfin imitating propelling device. China Patent,
CN104260864A
6. Dumas G (2007) Mechanism for control of an oscillating wing. WIPO Patent,
WO2008144938A1
7. Alvardo PV, Toumi KY (2004) Mechanised fish Robot exploiting vibration modes for
locomotion. U.S. Patent, US7865268
8. Weidong S (2016) Bionic fish movement mechanism. China Patent, CN106275536A
9. Bedi GS, Sanyal S (2013) Loop based algorithm for autonomous sketching of planar kinematic
chains. In: Proceedings of 1st and 16th national conference on mechanism and machines
(iNaCOMM 2013), IIT Roorkee, India 18–20 Dec 2013
10. Hsieh CF, Hwang WY, Yan HS (1998) Generation and sketching of generalized kinematic
chains. In: Proceedings of the ASME 2008 international design engineering technological
conference and computers and information in engineering conference, 3–6 Aug, New York
11. Torgal S (2016) Application of maxcode algorithm for the enumeration of kinematic chains of
9 link and 2 Degree of Freedom. SmartCom 2016, CCIS 628, pp 581–589
12. Mruthunjaya TS (1984) A computerized methodology for structural synthesis of kinematic
chains: Part 1 formulation. Mech Mach Theory 19(6):487–495
Topology Structure Design of Fish-Based Propulsive Mechanisms 1065
13. Mruthunjaya TS (1979) Structural synthesis by transformation of binary chains. Mech Mach
Theory 14:221–231
14. Mruthunjaya TS (1984) A computerized methodology for structural synthesis of kinematic
chains: Part 2-Application to several fully or partially known cases. Mech Mach Theory
19(6):497–505
15. Hwang WM, Hwang YW (1992) Computer aided structural synthesis of planar kinematic
chains with simple joints. Mech Mach Theory 27(2):189–199
16. Rao AC, Deshmukh PB (2001) Computer aided structural synthesis of planar kinematic chains
obviating the test for isomorphism. Mech Mach Theory 36:489–506
Impact of SOC Estimation on Primary
Frequency Regulation for Sustainable
Grid Energy Storage System
J. Dhanaselvam (B)
Srikrishna College of Technology, Coimbatore, Tamil Nadu 641025, India
V. Rukkumani
Sri Ramakrishna Engineering College, Coimbatore, Tamil Nadu 641022, India
1 Introduction
1.1 Introduction
AC BAR
DC to AC
PV ARRAY
Converter
Electric Grid
AC to AC
Wind System
Converter
2 State of Charge
where
• OCV SOC —Open-circuit voltage at specific SOC point
• Vo —Base voltage
• β—Slope
• SOC—State of charge with variation from 0 to 1
• M—SOC Correction factor (Fig. 3).
1070 J. Dhanaselvam and V. Rukkumani
In this model, the current at a particular time is calculated based on the available
capacity of the battery. The battery SOC is calculated by ampere hour integral method
which is fused into the model-based method to calculate available SOC.
t k
S O C(K ) = Y 0 − τ I L (t)d t/ Q
t0
where
• SOC(K)—SOC att K
• Y 0 —SOC att 0
• τ —Coulomb efficiency
• Q—maximum available capacity.
The advantage of model-based method is that the internal states such as the
chemical composition concentration, over-potential, and degradation of the battery
life can be easily estimated [5]. Like conventional generation units, battery energy
storage system provides frequency regulation as upwards and downwards regulation
by discharging and charging, respectively (Fig. 4)..
It is very important to maintain the smart grid network state within the operational
limits by balancing or re-balancing the supply and demand. Without initiating insta-
bilities and operational constraints, how to integrate load participation for many
applications is not clear. So, battery energy storage system plays an important role
here to compensate and regulate the frequency in terms of primary frequency control
[6, 7] (Fig. 5)..
Impact of SOC Estimation on Primary Frequency Regulation … 1071
The following flowchart illustrates how the frequency is maintained during up-
regulation and down-regulation by charging and discharging. BESS maintains it
consistently in this regard (Fig. 6).
Lithium-ion battery energy storage system is re-established every time when the
frequency reaches ±2 Hz. In smart grid applications, the PFR must be established
1072 J. Dhanaselvam and V. Rukkumani
START
Measurement of PFR
NO NO
49.8 49.98 HZ 50.02 50.2 Hz
YES YES
SOC YES
Reestablishment
YES 50%
NO NO
END
within 15 min, and when it is not, the PFR service will be completely interrupted
[8–10]. At this time, 50% of the SOC will be re-established.
time of the battery will be increased. Finally, to have a proper primary frequency
regulation an SOC must be maintained and correctly established. In future, a further
analysis can be made to reduce time duration for establishing an SOC below to the
minimum.
References
Abstract This paper presents a framework based on dual quaternions for obtain-
ing the forward and inverse kinematic models for a 6-dof (degrees of freedom)
UR5 cobot. A forward kinematics model based on dual quaternions is developed for
the cobot. Dual quaternion differential kinematics involving Jacobian transpose and
damped least squares methods along with pose error feedback are used for determin-
ing the inverse kinematics model. Implementation of inverse kinematics methods for
a given trajectory shows that Jacobian transpose method is faster, but gives more
jerky motions and is less immune to singularity compared to damped least squares.
1 Introduction
lends well to screw transforms [6]. Dual quaternions have been used for kinematic
modeling and pose control of serial manipulators [7, 8].
Differential kinematics is a widely-used approach for inverse kinematics (IK)
modelling. DQs and quaternion exponential maps have been used to solve IK prob-
lems by considering joint limits [6]. Dealing with Jacobian matrix is well discussed
in literature, unlike the dual quaternion differential operator.
The flow of the paper is as follows: Dual quaternion preliminaries are introduced
in Sect. 2. The forward and inverse kinematic models for a serial manipulator are
developed in Sect. 3 and applied to UR5 cobot in Sect. 4. In Sect. 5, the results are
discussed and the paper is concluded in Sect. 6.
2 Mathematical Preliminaries
a = a0 + a1 i + a2 j + a3 k = [a0 , a T ]T a0 , a1 , a2 , a3 ∈ R, a ∈ R3 , a ∈ H (1)
The unit quaternion r̂ = [cos θ2 , ûT sin θ2 ]T represents rotation by angle θ about unit
vector û. Given a rotational quaternion r̂ = [r0 , r T ]T , the rotation parameters θ and û
are obtained as θ = 2 cos−1 (r0 ) and û = r/ sin θ2 . In DQ form, rotation is represented
as r = r̂ + 0.
Let translation t = [tx , t y , tx ]T be depicted in quaternion form as t = [0, t T ]T and
in DQ form as t = 1 + t/2. Then, the combined transformation q consisting of
rotation r followed by translation t is given in DQ form as
q = t r = r̂ + 21 t r̂
= [cos θ2 , sin θ2 ûT ]T + 21 [−sin θ2 (t· û), cos θ2 t T + sin θ2 (t× û)T ]T (4)
Link i-1
i’
Screw(Z , , )
i−1
q i = q iZ q iX
= [η1 , η2 , η3 , η4 , −Dη4 − Aη2 , −Dη3 + Aη1 , Dη2 + Aη4 , Dη1 − Aη3 ]T (5)
∂ 0q n
q̇ = J θ̇, wherei-th column of Jis J i = (7)
∂θi
Kinematic Modelling of UR5 Cobot Using Dual … 1081
For obtaining the inverse kinematics (IK) model of the serial manipulator, inverse
differential kinematics is used. However, for a non-square Jacobian matrix, the
pseudo-inverse is found by pre-multiplying both sides of Eq. 7 by J T and rearranging
the terms so as to obtain
−1 T
θ̇ = J T J J q̇ (8)
During singularity condition, when two joint axes get coincident such that actu-
ating either joint produces the same differential movement of the end-effector, then,
the two corresponding columns of J become equal, and matrices J as well as ( J T J)
become singular. One way to deal with this is to use the Jacobian transpose (J T )
method [10], where ( J T J)−1 is approximated to α , given by
q̇ T β
α= , where, β = J J T q̇ (9)
βT β
To further reduce the chances of J J T losing rank, the damped least squares
method (DLS) is used [10]. Here, a damping constant δ ≈ 0.001 in the diagonal
elements modifies Eq. 8 to
−1
θ̇ = J T J − δ 2 I J T q̇ (10)
1 1
q̇ = ξq = [0, ω T ]T + [0, (v + t × ω)T ]T q (11)
2 2
This error is fed back to the next differential change in pose q̇ k+1 .
1082 M. Dalvi et al.
The UR5 cobot (or collaborative robot) from Universal Robots, as seen in Fig. 2a, is
a 6R serial manipulator having 0.85 m reach and 5 kg payload capacity. The cobot
architecture is derived using classical DH conventions [9] as seen in Fig. 2b and
Table 2.
For implementing the framework, a library for dual quaternions is developed in
Python 3.7, and run on a Windows 8 workstation with Intel i5-4570 3.2 GHz processor
and 8 GB DDR3 RAM. The DQ outputs of the FK model are tallied with results from
RoboAnalyzer software. Two IK models, one using J T method, and other using DLS
method, are built, and are applied to the trajectory shown in Fig. 3. The red, green,
and blue arrows correspond to the end-effector roll, pitch, and yaw axes, respectively.
X2
J3
Z2
J4 X3
X4
Z3
J5 X5
Z4
X6
X1 J6
Z1 Z5
Z0 TCP
J2 Z6
J1
X0
The results of the DQ FK model matched perfectly with those obtained from Robo-
Analyzer software as well as the FK model derived using homogeneous transforma-
tion matrices (HTM). Compared to 234 additions and 356 multiplications needed in
HTM FK model, the DQ FK model required 224 additions and 312 multiplications.
While DQs represented end-effector pose using 8 parameters, HTMs required 12
parameters for the same task. The pose specified by DQs is not intuitive to humans.
However, using DQs over HTMs led to reduction in computational time by up to half
in some cases.
Applying the two IK models on the trajectory for 50 times each showed that
J T method gave solutions faster, but also gave more jerky motions. For J T , mean
solution time ranged from 12ms to 39ms, whereas for DLS, the same ranged from
33ms to 87ms. The graphs of quaternion components in Fig. 4 show that fluctuations
in DLS method are lesser than in J T method. It is seen that the trajectory loop did not
get closed in either method, which means pose error did not get removed even when
simple feedback was used. This makes a case for exploring more feedback control
methods.
1084 M. Dalvi et al.
Fig. 4 Using end-effector DQ pose elements qd0 and qd1 to compare curves traced by J T and DLS
IK models
6 Conclusion
In this work, dual quaternions (DQs) were used to develop the forward kinematics
(FK) and inverse kinematics (IK) models for a serial manipulator. The DQ differential
operator as well as the J T and DLS methods of solving inverse differential kinematics
in DQ space were discussed. This approach was implemented for the UR5 cobot and
errors obtained during simulation over a closed trajectory were studied. Though DQs
seem unintuitive to humans, they are computationally efficient and lend themselves
well to working with elaborate trajectories. Future works include DQ implementation
of fast IK algorithms such as FABRIK [10] and DQ spline interpolation for path-
planning problems.
References
1. Corke P (2017) Robotics, vision and control: fundamental algorithms In MATLAB® , Springer
Tracts in Advanced Robotics, vol 118. Springer, 2nd edn
2. Shoemake K (1985) Animating rotation with quaternion curves. In: ACM SIGGRAPH com-
puter graphics, vol 19. ACM, pp 245–254
3. Sommer H, Gilitschenski I, Bloesch M, Weiss S, Siegwart R, Nieto JI (2018) Why and how to
avoid the flipped quaternion multiplication. CoRR
4. Allmendinger F, Eddine SC, Corves B (2018) Coordinate-invariant rigid-body interpolation on
a parametric C1 dual quaternion curve. Mech Mach Theor 121:731–744
5. Jia YB (2013) Dual quaternions. Iowa State University, Tech. rep
6. Kenwright B (2013) Inverse kinematics with dual-quaternions, exponential-maps, and joint
limits. Int J Adv Intelligent Syst 6(1 & 2):53–65
7. Radavelli L, Simoni R, De Pieri E, Martins D (2012) A comparative study of the kinematics
of robots manipulators by Denavit-Hartenberg and dual quaternion. Mecánica Computacional
Multi-Body Syst 31(15):2833–2848
8. Pham HL, Perdereau V, Adorno BV, Fraisse P (2010) Position and orientation control of robot
manipulators using dual quaternion feedback. In: 2010 IEEE/RSJ international conference on
intelligent robots and systems, pp 658–663
9. Siciliano B, Sciavicco L, Villani L, Oriolo G (2009) Robotics: modelling, planning and control.
Springer, London
Kinematic Modelling of UR5 Cobot Using Dual … 1085
10. Aristidou A, Lasenby J (2009) Inverse kinematics: a review of existing techniques and intro-
duction of a new fast iterative solver. University of Cambridge, Tech. rep
11. Wang X, Han D, Yu C, Zheng Z (2012) The geometric structure of unit dual quaternion with
application in kinematic control. J Math Anal Appl 389(2):1352–1364
Design of XY Air Bearing Stage
for Ultra-Precision
1 Introduction
2 Literature Review
Precision stages being an important field from industry perspective, there have been
lots of work in this direction. Various kinds of configurations have been designed
involving guideways, preloads, actuators, architecture, sensor and control. Zschaeck
et al. [3] developed a double-H-type XY positioning system supported by two linear
v-groove high precision guideways and driven by two ironless linear motors. Shinno
et al. [4] developed voice coil motors actuated aerostatic XY planar stage for nano-
machining. Tomita et al. [5] developed XY positioning stage composed of four linear
motors while guiding the stage by linear motion ball bearings. Chung et al. [6]
developed an XY positioning stage supported by magnetically preloaded air bearings.
Shinno et al. [7] built XY-theta table system supported with four porous vacuum
preloaded bearings. Gorniak [8] developed a long range air bearing stage with a T-type
configuration. Lu and Usman [9] developed a novel 6D direct drive technology for
planar stages having long range with positioning resolution of 1um. It uses magnetic
levitation as support and actuation from linear magnet arrays and stereo-vision as
position sensor.
The earlier trend can be seen as stages with linear guide driven by ball screw
which had inherent limitation of hysteresis and backlash. Then, hybrid drive stages
were developed which used ball screw for coarse and long range positioning while
using flexure or PZT or VCM for fine positioning. These structures had, however,
problems due to unpredictable friction in coarse stage and mutual interdependencies
among stages. After this linear motor-driven stages with air bearings reduced their
complexity while allowing long motion range. Then, a new approach like magnetic
levitation further improved the performance of these precision stages but currently
limited by short strokes and higher complexity. There are very few ultra-precision
table systems which enable simultaneous realization of high-speed feed drive as
well as nanometer positioning accuracy for over long motion range with simple
mechanical structure and control. Hence, the proposed motion stage design is directed
to meet these requirements.
3 XY Stage Design
The XY stage has modular design with series architecture shown in Fig. 1. Three
identical modules of linear stages are assembled into planar stage using flexural
Design of XY Air Bearing Stage for Ultra-Precision 1089
joints. This makes design general in nature, applicable for single to multi-degree of
freedom applications.
The positioning stage has an H-type configuration with two linear aerostatic guide-
ways in Y-axis and one in X-axis. Unlike stacked linear stages design in “plus” shape
which has inherent dynamics imbalance, it has dual y-axis H-type design, giving it
better dynamics.
Ten air bearings on each moving carriage have been used to provide linear guid-
ance. These porous air bearings are preloaded by opposite air bearings which provide
bidirectional load capacity and increases bearing stiffness. Three bearings making
a plane and two defining a line constrains total of 5 degree of freedom. Since each
air bearing is preloaded by another air bearing, total of 10 air bearing configuration
provides higher stiffness though being over constrained and non-kinematic. The given
configuration of air bearings is based on recommendations by one of air bearings
manufacturers’ design guide [10].
1090 R. Kumar and J. P. Khatait
The x-axis carriage is connected with flexural joints such that it constrains six
DoFs of the x-axis carriage and also allows yaw motion which is possible due to
non-synchronous behavior of two y-axis actuators. The yaw motion can then be
controlled from two position sensors on y-axis.
Two actuators on y-axis provide high acceleration and velocity to the stage
improving its turnaround performance. It is also tried to make it a coplanar stage
having location of center of force and center of mass in the same plane to reduce
moments, improving the dynamics of the system. The carriages are end supported
over a precision ground rectangular section granite beam. Whole stage is based on
granite base to minimize error due to thermal expansion.
Positioning feedback is provided by high resolution optical encoder attached on
support beam of each axis, especially dual encoders on y-axis which provides better
orthogonality/yaw control. Each encoder is mounted close to the worktable plane to
reduce Abbe error. Air compressor and filters are used to provide clean and dry air to
air bearings. Further positioning error is reduced by motion controllers using closed
loop feedback mechanism. Payload is mounted on top of moving carriage.
The maximum carriage load capacity is determined by the maximum load capacity of
bearings (which is a function of air gap between bearing and the surface), the preload
applied (higher preload reduces the carrying capacity), size of the air bearing (larger
size, higher capacity) and air pressure supply. Major factor to be considered is that
in working condition carriage should maintain minimum allowable gap between
bearing and the surface. Recommended fly height is around 5–6 μm. For correct
determination of load carrying capacity, load-lift chart supplied by the bearing vendor
is used.
Each bearing is modeled as spring of constant stiffness. The stiffness values of
each air bearing are obtained from manufacturer’s catalogue. A stiffness matrix can
be calculated for the stage as given in Table 1.
Natural frequency can then be calculated as follows (includes compliance due to
air bearing only) (Table 2).
Major sources of error are quantified and presented in the form of error budget
predicting the overall accuracy of the machine. Critical areas can then be determined
and solutions can be developed to minimize individual errors [11, 12].
The linear positional error for single axis consists of encoder scale error, the
resolution of the encoders and the uncertainty of the encoder signal. The straightness
error is due to the geometry of the guideways. Angular error are also called yaw,
pitch and roll errors. These errors are created from the contours of the guideways.
Dynamic errors occur from stiffness of machine components during acceleration and
from external loading. Thermal error consists of the expansion of encoder scale.
An error budget is prepared including above factors. Uncompensated and
compensated positioning accuracy of 6.41 μm and 1.33 μm has been predicted,
respectively.
Apart from the bridge axis being exactly constrained, it requires yaw compliance
to accommodate small non-synchronous behavior of Y-axis actuators and deviation
in parallelism between two Y axes. Flexural couplings attached to crossbeam can
provide yaw compliance thus allowing active yaw and straightness compensation
and improving orthogonality control. A combination of flexural couplings is added
to the XY stage design as shown in Fig. 3.
In XY assembly, cross flexure (Fig. 3a) provides yaw compliance at one end
constraining 5 DoFs while parallel flexure (Fig. 3b) constrains 3 DoFs (two in-plane
translation and one in-plane rotation). Thus, the total of 8 DoFs are constrained, over-
constraining once in roll direction (by cross flexure) and twice in pitch direction (by
parallel flexure). Though this arrangement is over constrained, it will contribute to
higher stiffness.
Monolithic flexures made from single metal plate are also viable options. Though
such flexures appear simple, they have comparatively higher stiffness and bending
stress (occurring at connections) but can be good alternatives. Two such flexures are
proposed in Fig. 4.
4 FEA
4.1 FE Model
Figure 5 shows the meshed model of XY stage. The system contains a total of 346,426
elements with element size of 5 mm. The simplest element for the modeling of the
stage is the hexahedron element.
Design of XY Air Bearing Stage for Ultra-Precision 1093
a) b)
c)
Fig. 3 a Cross flexure, b parallel flexure, c assembly
The structure of the bolt connections are modeled using rigid element since their
local deformation has no impact on the overall performance of the machine. Further
bracket, beam and magnet tracks are connected using 1D weld elements as they are
assumed to be rigidly connected members. Air bearings are modeled as linear spring
element of constant stiffness. Multiple-node end connection with the spring using
1D rigid elements have been used to avoid stress concentration.
Modal analysis has been performed for three cases of assembly. In the first case,
structure has been assumed of very high stiffness to capture stiffness due to air
bearing only. So that it can be correlated with analytical calculation. In the second
case, crossbeam is mounted rigidly with gantry structure as shown in Fig. 6a. 1D rigid
elements have been used for connection on both ends simulating bolted connections.
In the third case, assembly is done using flexural couplings as shown in Fig. 6b. In
last two cases, structural compliance has been taken into account. Tables 3 and 4
show results.
In Table 3, FEA results are in good correlation with analytical ones In Table 4, the
first two modes are translational modes in X and Y direction whose frequency will
be determined from the servo bandwidth. Since, here, it is not taken into account,
it appears as rigid modes. The dominant mode in the above simulation is the yaw
mode which is 100 Hz for bolted assembly. In case of flexural coupling assembly,
yaw mode has very low frequency showing yaw compliance which can be controlled
by Y-axis actuators. 46 Hz is next dominant frequency (Fig. 7).
5 Control
Dynamic modeling and prediction of the motion system forms a major portion of
the design process. In this section, dynamic modeling of XY air bearing stage is
presented. Similar analysis is done by García-Herreros et al. for rail guide-based
dual drive gantry stage [14].
The XY stage can be considered as three-degree of freedom mechanism which
can be described by the schematic in Fig. 8. Main physical parameters are given in
Table 5. x 1 , x 2 and Y are measured quantities using linear encoders in each stage
module. These quantities can be further mapped to X, θ and Y which can give global
location of center of mass of work table.
A Lagrangian-based model is derived. The equation of motion of the system can
be written as
M q̈ + H q̇ + C q̇ + K q = f (1)
where M, C and K are the inertia, viscous damping and stiffness matrices, H is the
coriolis and centripetal acceleration matrix, f is the vector of forces, and q is the
vector of coordinates (X, θ, Y ). The equation of motion is further simplified based on
assumptions given very small yaw angle θ, negligible centripetal and coriolis forces
as compared to inertia, damping and spring forces and negligible influence over the
motion of worktable due to linear and angular accelerations of the cross-arm and
vice versa except due to angular acceleration of worktable on cross-arm.
Simplified Model
where
m1 + m2 + m3 + m4 (m 1 − m 2 ) L2 − m 4 Y
Ms = 2 (3)
(m 1 − m 2 ) L2 − m 4 Y J3 + J4 + (m 1 + m 2 ) L4 + m 4 d 2 + m 4 Y 2
cv1 + cv2 (cv1 − cv2 ) L2
Cs = 2 (4)
(cv1 − cv2 ) L2 cv1 + cv2 + (cv1 + cv2 ) L4
0 0
Ks = (5)
0 k1 + k2
1098 R. Kumar and J. P. Khatait
FX
fs = (6)
Fθ + m 4 d Ÿ
T
qs = X θ (7)
m 4 Ÿ + c y Ẏ = Fy (8)
where M S , C S and K S are the simplified inertia (3), damping (4) and stiffness (5)
matrices, f S is the vector of simplified forces (6), and qS is the vector of coordinates
X and θ (7).
5.1 Results
6 Conclusion
In this paper, a novel mechanical design of an XY air bearing stage for ultra-precision
has been proposed. The serial architecture with H-type configuration of the stage
provides better dynamics for high-speed applications. The design being modular
makes it possible to assemble many such linear stages into higher DoF stages. Using
air bearing pads for support over granite guideway eliminates error due to friction like
hysteresis and backlash. Non-contact and direct drive features of the motion stage
give better positioning performance than traditional ball screw and sliding support
stage.
An error budget is prepared to check theoretical performance of the system. It is
predicted to be maximum uncompensated positioning error of 6.4 μm. A flexural
coupling is designed to provide yaw compliance to relax stringent criteria of parallel
Design of XY Air Bearing Stage for Ultra-Precision 1099
dual axis stage and compensate for dynamic differences between parallel actuators.
Design performance of XY stage is studied using FEA. Fundamental frequency of
vibration for XY stage is observed to be vertical motion at 46 Hz. A PID controller
is applied; tracking error is found to be below 2 μm at low speed.
1100 R. Kumar and J. P. Khatait
References
1. Wagner C, Harned N (2010) EUV lithography: lithography gets extreme. Nat Photonics
4(1):24–26
2. Jain K, Klosner MARC, Zemel MARC, Raghunandan S (2005) Flexible electronics and
displays: high-resolution, roll-to-roll, projection lithography and photoablation processing
technologies for high-throughput production. Proc IEEE 93(8):1500–1510
3. Zschaeck S, Amthor A, Ament C (2011) Decentralized high precision motion control for
nanopositioning and nanomeasuring machines. In: IECON 2011—37th annual conference on
IEEE industrial electronics society. IEEE, pp 546–551
4. Shinno H, Yoshioka H, Taniguchi K (2007) A newly developed linear motor-driven aerostatic
XY planar motion table system for nano-machining. CIRP Ann Manuf Technol 56(1):369–372
5. Tomita Y, Makino K, Sugimine M, Taniguchi N (1996) High-response XY stage system driven
by in-parallel linear motors. CIRP Ann Manuf Technol 45(1):359–362
6. Chung TT, Chu CH, Chian HF, Huang C, Fan KC, Yen JY, Szu KI (2011) Structural design
and analysis of a nano-positioning planar motion stage. IEEE, pp 833–838
7. Shinno H, Hashizume H, Yoshioka H, Komatsu K, Shinshi T, Sato K (2004) XY-θ nano-
positioning table system for a mother machine. CIRP Ann Manuf Technol 53(1):337–340
8. Gorniak JM (2010) Design and metrology of a precision XY planar stage. Master’s thesis,
University of Waterloo
9. Lu X, Usman I (2012) 6D direct-drive technology for planar motion stages. CIRP Ann Manuf
Technol 61(1):359–362
10. Bearings NWA (2006) Air bearing application and design guide
11. Donaldson RR (1980) Error budgets. Technology of machine tools 5
12. Slocum AH (1992) Precision machine design. Society of Manufacturing Engineers, Dearborn
MI
13. Soemers HMJR (2001) The design of high performance manipulators. In: 2001 IEEE/ASME
international conference on advanced intelligent mechatronics, proceedings, vol 1. IEEE
14. García-Herreros I, Kestelyn X, Gomand J, Coleman R, Barre PJ (2013) Model-based decou-
pling control method for dual-drive gantry stages: a case study with experimental validations.
Control Eng Pract 21(3):298–307
Design and Fabrication of a Bio-inspired
Soft Robotic Gripper
1 Introduction
Design and fabrication of soft gripper are required for robots used in health care and
food industry. Bio-inspired soft gripper has several advantages over existing rigid and
complex robotic grippers such as better control of gripping force, adaptability toward
the shape of the object, and feel of human touch. [1, 2]. Human fingers are the best
example of such gripper made of soft muscle and hard skeleton. While muscles act as
actuator, skeleton provides strength and act as a passive support defining the degree
of freedom of the finger. Currently, there are many actuation mechanism such as
pneumatic structure [3], shape memory alloys (SMA) [4], fluidic elastomeric actuator
(FEA) [5], shape morphing polymers (SMP) [6], and dielectric elastomer actuators
(DEAs) [7–9] which are being studied extensively for soft robotics. DEAs are very
promising due to their lightweight, large deformation, fast response time, and cost-
effective characteristics. DEAs are soft and highly stretchable resembling mechanical
2 Design of Gripper
Figure 1 shows the proposed design of three-fingered gripper developed in this work.
Acrylonitrile butadiene styrene (ABS) was used as the hard core for 3D printed
fingers. These fingers were fixed at 120° from each other onto a rigid fixture forming
the gripper hand as shown in Fig. 1a. Each finger consists of three segments which
were connected by pin joints. This gives a rotational degree of freedom to segments
with respect to each other. The tip of finger has smallest length of 18 mm and was
coated by a thin film of PDMS layer to provide soft contact to the gripped objects. In
order to ensure higher rotation of the tip, the length of subsequent segments increases
as we move away from the fingertip. Moreover, the segments of a finger were designed
in such a way that their rotation relative to each other would be constrained in 0 < θ
< 90 limit. The backward rotation of fingers is constrained by providing an extended
feature near the pin joint as shown in the side view of Fig. 1b. The actuation of
skeleton was enabled by bonding a 30% pre-stretched PDMS film to each finger
(Fig. 1c). After removing the constrained, the pre-stretched film deforms in-plane
forming a curved finger configuration of Fig. 1d.
In order to optimize the gripper design, different configurations were tested and
summarized in Table 1. In design A, carbon grease coated pre-stretched PDMS sheet
was glued only on one side of the finger (Fig. 1c). In a rather different arrangement of
design B, while one side is bonded with pre-stretched soft layer, an unstretched layer
is glued on the other side of skeleton (Fig. 1d). In this configuration, the equilibrium
is achieved between two opposing forces of stretched and unstretched PDMS sheets.
The actuation of both, design A and design B, is in opening mode. By changing the
actuating layer of design B from side 1 to side 2, the same configuration is extended
Design and Fabrication of a Bio-inspired Soft Robotic Gripper 1107
Fig. 1 Design of a gripper hand and b single finger. The skeleton configuration having PDMS
sheet on c one side and d both sides in design A and design B
to achieve closing mode actuation in design C. Later on, the thickness of PDMS layer
used in design B is doubled to get design D.
Fabrication of PDMS-based DE layer is the next step to process the soft gripper as
shown in Fig. 1. To this end, a sacrificial layer of PVA was coated on glass substrate
1108 A. Agarwal et al.
Fig. 2 a Fabrication process of PDMS thin film and b removal of PDMS thin sheet from the glass
substrate
and a cavity of desired dimensions is formed with the help of acrylic tape. PDMS
was poured into this cavity, and excess mixture was scrapped with the help of a sharp
blade as shown in Fig. 2a. The uncured PDMS layer was subjected to vacuum for
30 min to remove any air bubble trapped before curing it at 80 °C for 1 h. The glass
substrate was then dipped in water to dissolve the PVA layer and collect the PDMS
sheet (Fig. 2b).
Fig. 3 a Actuation state of on application of voltage, b plot of actuator displacement against applied
voltage, c arrangement for blocking force measurement, d plot of blocking force against applied
voltage
Maximum blocking force achieved for design B is 0.87 mN at 4 kV (Fig. 3c) which
is comparable to the reported data for similar configuration [15].
As the fingers of design B opens on the application of voltage, no voltage is
required during the time when object is gripped. Consequently, lesser energy is
required for operation of this type of actuator. To quantify the opening of claw type
arrangement on application of voltage, we can assume a circle passing through three
points of fingertips. At V = 0, the diameter of this circle is 25 mm which increases
to 32 mm at 4 kV. A lightweight object is clamped between the three fingers. When
voltage is applied, due to the opening motion of fingers, the object drops down at
3 kV as shown in Fig. 3d. Using the data shown in Fig. 3a, c, the efficiency η, which
is defined as the ratio of mechanical energy stored to electrical energy supplied
1110 A. Agarwal et al.
m
Um = Fi (V ) · x
i=0
where F i is the blocking force at the tip as function of applied voltage V at every
0.5 kV step, and x is displacement at same voltage step. The number of steps is
represented by m. The values of F i and Δx are given in Fig. 3a, c. From this, the
calculated mechanical energy is 3.3 × 10−6 J. On the other hand, the electrical energy
U c supplied to griper is calculated as,
1
Uc = CV 2
2
where C is the capacitance of DEA and V is maximum voltage applied. The capaci-
tance is obtained from C = ε0 ε (A/d) where ε0 is the permittivity of the free space,
ε is dielectric constant of PDMS, A is the area of electrode and d is the thickness of
PDMS sheet. While ε0 = 8.85 × 10−12 F m−1 , the dielectric constant (ε) of PDMS
is given in supplier data sheet as 2.6 [16]. The area (A) and thickness (d) of PDMS
sheet are continuously changing throughout experiment so dimensions of stretched
PDMS sheet A = 80 × 14 mm2 and d = 0.125 mm are used. Putting these values with
maximum applied voltage of V = 4 kV from Fig. 3c, the electrical energy supplied
to actuator comes out to be U c = 1.4 × 10−3 J. Using the calculated values of U m
and U c , the electromechanical efficiency for each finger found to be around 0.23%.
The obtained efficiency is relatively lower in comparison with reported values [15]
which may be due to the rigid skeleton used to represent bones in human fingers.
Improving the actuator efficiency using some modified designs and better actuator
material will be the subject of further study.
4 Conclusion
References
1. Suzumori K (1996) Elastic materials producing compliant robots. Rob Auton Syst 18:135–140
2. Manti M, Hassan T, Passetti G, D’Elia N, Lasachi C, Cianchetti M (2015) A bioinspired soft
robotic gripper for adaptable and effective grasping. Soft Rob 2(3):107–116
3. Guoliang Z, Yangdong H, Weiqiang D (2019) A soft pneumatic dexterous gripper with
convertible grasping modes. Int J Mech Sci 153–154:445–456
4. Riad A, Alhamany A, Benzohra M (2017) The shape memory alloy actuator controlled by the
Sun’s radiation. Mater Res Express 4
5. Barreiros J et al (2019) Fluidic elastomer actuators for haptic interactions in virtual reality.
IEEE Robot Autom Lett 4:277–284
6. Bowman CN (2014) Smart shape changing and shape morphing polymeric materials. Polymer
55(23)
7. Kofod G, Paajanen M, Bauer S (2006) Self-organized minimum-energy structures for dielectric
elastomer actuators. Appl Phys A 85:141–143
8. Ahmadi S, Mattos AC, Barbazza A, Soleimani M, Boscariol P, Menon C (2012) Fabrication
and performance analysis of a DEA cuff designed for dry-suit applications. Smart Mater Struct
22
9. Qiu Y, Zhang E, Plamthottam R, Pei Q (2019) Dielectric elastomer artificial muscle: materials
innovations and device explorations. Acc Chem Res 52:316–325
10. Gu G, Zhu J, Zhu L, Zhu X (2017) A survey on dielectric elastomer actuators for soft robots.
Bioinspiration Biomimetics 12
11. Suo Z (2010) Theory of dielectric elastomers. Acta Mech Solida Sin 23(6):549–578
12. Romasanta LJ, Lopez-Manchado MA, Verdejo R (2015) Increasing the performance of
dielectric elastomer actuators: a review from the materials perspective. Prog Polym Sci 8
13. Kofod G, Paajanen M, Bauer S (2006) Self-organized minimum-energy structures for dielectric
elastomer actuators. Appl Phys A Mater Sci Process 85(2):141–143
14. Petralia MT, Wood RJ (2010) Fabrication and analysis of dielectric-elastomer minimum-
energy structures for highly-deformable soft robotic systems. In: Proceedings of IEEE/RSJ
international conference on intelligent robots and systems, pp 2357–2363
15. Shintake J, Rosset S, Floreano D, Shea HR (2013) Effect of mechanical parameters on dielectric
elastomer minimum energy structures. In: Proceedings of SPIE electroactive polymer actuators
and devices (EAPAD)
16. Technical data sheet SYLGARD™ 184 silicone elastomer
Experimental and Simulation Study
of Haptically Enabled Robotic
Teleoperation for NOTES
1 Introduction
Inclusion of Robots in the medical field has the potential for every step of treatment
like diagnosis, therapeutics, pre-operative planning, post-operative care, hospital car-
riage and scheduling, quality check, and rehabilitation, etc. [1]. Specifically, the
inclusion of robots in surgery has turned out to be a great advantage. For each medi-
cal procedure, there is a requirement of some specific structure, intelligent and user
friendly interface [2]. Robot-assisted laparoscopy has significantly reduced recovery
time and hospital stay [3]. But laparoscopy has its drawback say operating tools
with fewer degrees of freedom. The first robotic surgery was introduced in 1985 to
Forward kinematic analysis is obtained for the master device (Phantom Omni haptic
device), as shown in Fig. 1. Phantom Omni is an electromechanical device that can
get kinesthetic feedback. It has six DOF stylus with three revolute and three gimbal
joints. The encoder placed in the base of the robot with three joints are active joints
and three joints are passive joints. Its D-H (Denavit-Hartenberg) parameters are given
in Table 1, are obtained from literature [12]. Where a is link length,α is twist angle, d
shows link offset and θ is joint angle. Using D-H parameter notation method, a 4 × 4
homogenous transformation matrix was used to relate (i − 1)th coordinate frame to
(i)th coordinate frame as given in (1).
2.2 The Relation Between Joint Angle and Tip Position of the
Slave Miniature Robot
⎡ ⎤
cφc1234 −cφs1234 −sφ cφ(L 1 c1 + L 2 c12 + L 3 c123 + L 4 c1234 )
⎢ sφc1234 −sφs1234 cφ sφ(L 1 c1 + L 2 c12 + Ls3 c123 + L 4 c1234 ) ⎥
⎢ ⎥
6 T = ⎣ −s
0
1234 −c1234 0 d − (L 1 s1 + L 2 s12 + L 3 s123 + L 4 s1234 ) ⎦
0 0 0 1
3 Experiement
For the fabricated model, link lengths are equal to 1.1 cm. Figure 3a, b respectively
shows the initial and final configuration of the links and the desired position. We take,
l1 = L 1 + L 2 , l2 = L 3 + L 4 , d = 0 (no linear motion) and φ=0 (planar motion case)
1118 S. Saini and P. M. Pathak
to simplify the analysis. Joint angles for the slave robot are calculated using inverse
kinematic algorithm for the in-vivo robot as obtained in [14]. Equations (5)–(7) can
be solved as the case of a two-link manipulator as given by (8) and (9).
x = l1 c1 + l2 c12 (8)
y = l1 s1 + l2 s12 (9)
Here c12 = cos(δ1 + δ2 ) and s12 = sin(δ1 + δ2 ). δ1 and δ2 are the joint angles for the
two links of the manipulator. The values of joint angles are obtained using analytical
method as mentioned below [15]. Using (8) and (9), we can write,
x 2 + y 2 − l12 − l22
cos δ2 = (11)
2l1l2
δ2 = atan(s2 , c2 ) (13)
y = k1 s1 − k2 c1 (15)
k2 = rsinα (17)
Experimental and Simulation Study of Haptically … 1119
• Force required for the tissue manipulation is depends upon the material properties
of the stomach model, i.e., pring constant and damping factor in our model.
• Force applied in the y-direction is more than the forces used in x and z-direction
as tissue manipulation take place is in the y-direction.
• Variation in the force exerted on the stomach model and the force felt in the master
device as haptic feedback is negligible.
Now,
(y/r )
tan(α + δ1 ) = (20)
(x/r )
The plot given in Fig. 4 shows the joint angles of the slave robot for different tip
position of slave robot corresponding to the tip positions of the master device during
a typical interaction. Here, dotted line shows the δ1 and solid line shows δ2 for
achieving different tip positions during an interaction. It can also be inferred from
Fig. 3b. Variation in δ1 is more compare to δ2 . It shows δ1 is more responsible for
reach compared to δ2 and δ2 is more responsible for orientation compared to δ1 .
For creating a virtual environment, a 3D model of the stomach and the four DOF
slave tools were developed in SolidWorks commercial software. Then these models
are converted into any of 3DS or PLY or OBJ for compatibility with Visual Studio
C++ 2010. The QHHeaderWin32 is the main header file. It includes sub-libraries like
global, device space and QHwin32class, that are utilized in the programming. Steps
for developing the haptic experiment of the stomach and slave cursor tool (concerning
origin) are shown in Fig. 5. The slave robot and stomach model concerning origin
Fig. 6a shows that of virtual space, and deformations in the stomach model are
shown in Fig. 6b. Force exerted on the stomach model and haptic/master device
stylus will follow the mathematical model F = kx − c ẋ (spring-mass-damper) the
system with appropriate spring and damping constants [16]. In our experiment, the
considered values of mass, spring stiffness and damping constant are 0.1, 0.2 and 0.05
1120 S. Saini and P. M. Pathak
respectively. Corresponding forces in the stomach model and stylus of the master
device are shown in Fig. 7a and b, respectively. The haptic feedback force is slightly
less than the interaction force between the stomach and stylus. This difference in
the force may be due to friction and signal attenuation in teleoperation. It can be
improved by using a better processor and by minimizing joint friction in the haptic
device.
Overall it is seen that the force in the virtual environment and the one felt in the
stylus match each other to a large extent, thus indicating a fair implementation of
haptic feedback in the virtual environment.
Fig. 6 a Virtual haptic environment showing in-vivo cursor and stomach model with respect to
origin. b Deformation of stomach on pressing with slave robot tip/cursor
• Force required for the tissue manipulation is depends upon the material properties
of stomach model, i.e., spring constant and damping factor in our model.
• Force applied in y-direction is more than the forces applied in x and z-direction as
tissue manipulation take place is in y-direction.
• Variation in interaction force and stylus haptic feedback force is is negligible.
1122 S. Saini and P. M. Pathak
Fig. 7 a Variation of absolute force and forces in different direction forces by virtual cursor exerted
on stomach model. b Variation of absolute force and its components in master/haptic stylus
Acknowledgements The financial assistance to Mr. Sarvesh Saini in the form of a Research Fel-
lowship (SRF)(sanction file no. is 09/143(0895)/2017-EMR-I) of the Council of Scientific and
Industrial Research (CSIR), New Delhi is greatly acknowledged.
References
5. Wang Z, Phee SJ, Wong J, Ho KY (2012) Development of a robotic platform for natural orifice
transluminal endoscopic surgery. Gastrointest Interv 1(1):40–42. https://doi.org/10.1016/j.gii.
2012.08.010
6. Sutar MK, Pathak PM, Sharma A, Mehta N, Gupta V (2013) Forward kinematic analysis of
in-vivo robot for stomach biopsy. J Rob Surg 7(3):281–287. https://doi.org/10.1007/s11701-
012-0375-y
7. Reiter A, Allen PK (2010) An online learning approach to in-vivo tracking using synergis-
tic features. In: 2010 IEEE/RSJ international conference 2010 intelligent robots and systems
(IROS), IEEE (2010), pp 3441–3446. https://doi.org/10.1109/IROS.2010.5650852
8. Maniya JB, Pathak PM, Mishra BK (2010) Design and development of virtual objects to be
used with haptic device for motor rehabilitation. J Softw Eng Appl 3(10):990–997. https://doi.
org/10.1109/IROS.2010.5650852
9. Saini S, Pathak PM, Orlando MF (2019) Bondgraph modelling for the master-slave robotic
teleoperation system. In: 2019 28th IEEE international conference on robot and human inter-
active communication (RO-MAN), New Delhi, India, pp 1–6 (2019). https://doi.org/10.1109/
RO-MAN46459.2019.8956409
10. McDermott SD (1999) A haptic assembly and disassembly simulation environment and associ-
ated computational load optimization techniques. J. Comput Inf Sci Eng. 1(1):113–122. https://
doi.org/10.1115/1.1389085
11. Salamanca MLP, Navarro JMS, Esmeral JS (2011) Analysing collision detection in a virtual
environment for haptic applications in surgery. Ingeniera e Investigacin 31(1):204–212. https://
doi.org/10.1115/1.1389085
12. Silva AJ, Ramirez OAD, Vega VP, Oliver JPO (2009) Phantom omni haptic device: kine-
matic and manipulability. In: Electronics, robotics and automotive mechanics conference. IEEE
(2009), pp 193–198. https://doi.org/10.1109/CERMA.2009.55
13. Sansanayuth T, Nilkhamhang I, Tungpimolrat K (2012) Teleoperation with inverse dynamics
control for phantom omni haptic device. In: SICE annual conference (SICE), 2012. IEEE, pp
2121–2126
14. Ahmmad SM, Khan M, Rahman M, Billah M (2013) Position control of a four link hyper
redundant robotic manipulator. Asian J Scient Res 6(1):67–77
15. Craig JJ (2005) Introduction to robotics: mechanics and control. Pearson Prentice Hall Upper
Saddle River (2005)
16. Toolkit O (2005) The SenSable technologies Inc., NJ, USA
Design of Robust Backstepping
Controller for Four-Wheeled Mecanum
Mobile Robot
Abstract The use of mobile and portable robots is increasing in structured and
unstructured environments. A robot with high dexterity and maneuverability in con-
fined spaces, such as omniwheeled robots, is required for such tasks. This paper
presents the trajectory tracking capability of four-wheeled mecanum mobile robot in
the presence of external uncertainties and disturbances. Firstly, the dynamic equation
of the robot is derived; then, the efficacy of a robust backstepping (RBS) controller
was tested on a reference trajectory. Simulation results prove that the proposed con-
troller tracks trajectory with greater accuracy than other standard controllers.
1 Introduction
Omnidirectional mobile robots have the added advantage that they can perform trans-
latory and rotatory motion independently and synchronously. This provides high
maneuverability well suited for factory floors, warehouses and rescue operations.
One such kind of mobile robot is four-wheeled mecanum mobile robot (FWMMR).
The wheels have rollers aligned at an angle of ±45◦ with the axis of the wheel,
helping it slide in the lateral direction. Our platform has one pair of diagonal wheels
having rollers at an angle of 45◦ and the other pair at −45◦
The kinematic [1] modeling of mecanum-wheeled mobile robot is a difficult task
which becomes more challenging with increased number of unactuated and actuated
joints of the wheels. The dynamic [2] modeling of a FWMMR becomes difficult with
2 Kinematic Modeling
The FWMMR is shown here in Fig. 1 in the top view, and there are a total of six
frames in the system. The frames associated with the wheels are Owi Xwi Ywi , where
i represents the wheel number. The robot frame of motion is Om Xm Ym Zm , and the
world frame is On Xn Yn Zn . γi represents the angle of roller of the it h wheel, which is
+45◦ for wheels first and third and −45◦ for the second and fourth wheel.
Let the position vector Pm = [xm ym φm ]T represents the position of the robot in
the Om Xm Ym Zm frame, where xm and ym are the positions of the mobile robot, and
φm is the orientation of the mobile robot with respect to the Xm axis. In terms of the
wheel’s rotational velocity around the wheel’s axis, i.e., θ̇i and wheel’s radius R the
value of velocity vector, Ṗm is given as [ẋm ẏm φ̇m ].
⎡ ⎤
⎡ ⎤ ⎡ ⎤ θ̇1
ẋm −1 1 −1 1 ⎢θ̇2 ⎥
⎣ ẏm ⎦ = R ⎣ 1 1 1 1 ⎦⎢ ⎥ (1)
4 −1 −1
⎣θ̇3 ⎦
φ̇m 1
d1 +d2 d1 +d2 d1 +d2
1
d1 +d2 θ̇4
⎡ ⎤
⎡ ⎤ ⎡ ⎤ θ̇
ẋn −C(φ) − S(φ) C(φ) − S(φ) −C(φ) − S(φ) C(φ) − S(φ) ⎢ 1 ⎥
⎣ ẏn ⎦ = R ⎣−S(φ) + C(φ) S(φ) + C(φ) −S(φ) + C(φ) S(φ) + C(φ)⎦ ⎢θ̇2 ⎥
4 −1 −1
⎣θ̇3 ⎦
φ̇n 1
d1 +d2 d1 +d2 d1 +d2
1
d1 +d2 θ̇4
(2)
where C(φ) = cos(φ) and S(φ) = sin(φ)
3 Dynamic Modeling
In Fig. 2, the forces acting on the mobile robot platform have been shown. The
forces due to actuator of the ith wheel in X and Y direction are given as Fxi and
Fyi , respectively. The net torque acting on the entire platform by the driving forces
is represented by τ . The external force acting on the system is represented by Fe
acting at an angle of δ with Ym and at a distance of b from the front wheel’s center.
Now, applying the Newton’s second law of translation and rotation, i.e., F = ma,
and τ = I α in the frame Om Xm Ym Zm and using rotation transformation matrix, we
have Ṗn = Rnm (φ)Ṗm and Fn = Rnm (φ)Fm . Upon solving and rearranging [7], we have
where
⎡ ⎤
2M
1
(βx S(2φ)ẏn − βy S(2φ)ẏn + 2βy ẋn + 2C 2 (φ)(βx − βy )ẋn + 4K1 ẋn
f (ẋ) = ⎣ (−βx S(2φ)ẋn + βy S(2φ)ẋn + 2βx ẏn + 2C 2 (φ)(βx − βy )ẏn + 4K1 ẏn
1 ⎦
2M
1
2In
(−2(d1 + d2 )2 K1 − βz )φ̇)
⎡ 1 ⎤
− (Sφ + Cφ) M1 (−Sφ + Cφ) − M1 (Sφ + Cφ) 1
(−Sφ + Cφ)
K2 ⎣ 1 M M
g(x) = √ M
(−Sφ + Cφ) M1 (Sφ + Cφ) M1 (−Sφ + Cφ) M
1
(Sφ + Cφ) ⎦
2 1
(d + d2 ) − I1n (d1 + d2 ) − I1n (d1 + d2 ) 1
(d + d2 )
In 1 In 1
⎡ ⎤
2M
1
(2Fe S(φ)S(δ) + 2Fe C(φ)C(δ))
h(Fe , φ, δ) = ⎣ (−2Fe S(φ)C(δ) + 2Fe C(φ)S(δ))
1 ⎦ ξ(t) = [ξx ξy ξφ ]T
2M
1
(d F C(δ) − d2 Fe S(δ) + bFe S(δ))
2In 1 e
where C(φ) = cos(φ), S(φ) = sin(φ), C(δ) = cos(δ) and S(δ) = sin(δ) and K1 and
K2 are motor coefficients in the driving force by motor given as Fdr = K1 Rθ̇ + K2 u
[8] and ξ(t) is uncertainty in the environment.
Design of Robust Backstepping Controller for Four … 1129
4 Controller Design
In backstepping(BS) control design [9, 10], the system is firstly broken down into
smaller subsystems, and input of each subsystem is considered as virtual input. All
the virtual control laws are integrated by stepping back to generate the actual control
law, and its stability is taken care of by Lyapunov function. Let the position vector
be defined by x1 = [xn yn φn ]T , velocity vector by x2 = [ẋn ẏn φ̇n ] and control inputs
by u(t) = [u1 u2 u3 u4 ]T . The two subsystems are given as
ẋ1 = x2 (4)
where x2 acts as the virtual control input for subsystem 1. The errors are defined as
where x1d and x2d represent desirable position and velocity, respectively. Taking the
Lyapunov function as
1 1
V (e1 , e2 ) = [e1 ][e1 ]T + [e2 ][e2 ]T (8)
2 2
To have V̇ (e1 , e2 ) ≤ 0, we take [ė1 ] = −Ka [e1 ] and [ė2 ] = −Kb [e2 ] where Ka and
Kb are 3X 3 positive definite matrix. Thus, we have
From Eqs. (4, 5) and (10, 11) and the Lyapunov design conditions, the backstepping
controller is given as in Eq. (12), and the discontinuous switching controller is given
as in Eq. (13).
where s = λ1 e1 + ė1 and is called weighted average error, and sat is defined as
sat(x) = x, if |x| < δ and sat(x) = Sign(x), if |x| ≥ δ, and Sign(x) = 1, ifx ≥ 0 and
1130 Z. U. Islam et al.
Sign(x) = −1, ifx < 0. Here, ξ(t)max ≤ |Kc | and λ are 3X 3 matrix The robust back-
stepping controller is finally given as
5 Simulation Results
To verify the effectiveness of the proposed control method, a reference nonlinear tra-
jectory tracking is done with external uncertainties and disturbances. The values of the
parameters of the robot [7] are taken as M = 6 kg, In = 0.0945 kgm2 , d1 = 11 cm
d2 = 18 cm b = 10 cm βx = βy = βz = 0.02 K1 = 0.087 N/V K2 = −11.4 kg/s,
and the boundary layer thickness is chosen as δx = 0.5 m δy = 1 m and δφ = 0.25 rad.
Results of error in trajectory tracing have been calculated in the form of integral
square error (ISE), integral average error(IAE) and integral time average error (ITAE).
The equation of trajectory ∀ t in (0 s, 30 s)
∀ t in (0 s, 30 s)
The external disturbing force is given as Fe = 10 N for 4 ≤ t ≤ 6 and δ = 1 rad
for 4 ≤ t ≤ 6 The parameters of the RBS controllers are obtained after over 250
iterations of parameter estimation toolbox in MATLAB R2016a.
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
64.06 0 0 128.08 0 0 5.34 0 0
Ka = ⎣ 0 70.29 0 ⎦ Kb = ⎣ 0 140.04 0 ⎦ λ = ⎣ 0 17.14 0 ⎦
0 0 60.04 0 0 140.63 0 0 2.6
and Kc is a diagonal 3X 3 matrix with each value = 20 The initial position and
orientation of the FWMMR is [xn yn φn ] = [3 0 0]T .
From Fig. 3, it can be observed that the RBS controller tracks the trajectory with
accuracy greater than PID and backstepping. PID undergoes a lot of chatter in the
last few seconds. However, RBS follows the reference path smoothly. Figure 4 shows
the control input voltage supplied to actuators, and RBS follows a sinusoidal curve
smoothly but PID undergoes rapid fluctuation for the entire interval. In practice, this
will damage the actuator. Hence, for practical purposes, RBS is a better choice. The
errors of RBS, PID and backstepping are compared in Table 1. Figures 5 6 and 7
show the magnitude of weighted error in all the three degrees of freedom with respect
Design of Robust Backstepping Controller for Four … 1131
4
Desired
PID
3 RBS
Backstepping
2
metre(m)
-1
-2
-3
-4
-4 -3 -2 -1 0 1 2 3 4
metre(m)
to the boundary layer for the RBS controller. It can be clearly seen that the system
is constrained within the boundary layer, except for certain time instants in Y axes,
which maybe perhaps due to slightly higher uncertainty in Y direction.
In this paper, a robust backstepping controller for FWMMR is developed. The robot
platform is given external disturbances and nonlinear uncertainties. The dynamic
equation of the FWMMR is developed in input affine form. Using the backstepping
technique of integrating all the virtual inputs of the subsystem, the robust backstep-
ping control law is developed. The stability of the control law is verified using the
Lyapunov method. Lastly, simulations were performed to check the efficacy of the
proposed controller on a reference path, and it keeps the system safely within bound-
ary limits. The proposed controller outperforms PID and backstepping in terms of
trajectory tracking, error magnitude and practical usability.
In the future, the controller would be developed in a way to have a dynamically
adjusting boundary layer that would provide two-fold benefits of reducing the con-
troller bandwidth and also to improve accuracy of the trajectory tracing. The control
policy would be implemented on the hardware system in order to determine the
practical feasibility and efficacy of the controller.
1132
4
5 #10
PID
RBS
4
-1
-3
-4
-5
0 5 10 15 20 25 30
time(s)
1
Weighted Error
0.8 Limits of Boundary Layer
Weighted Error in X(m)
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
0 5 10 15 20 25 30
time(s)
2
Weighted Error
1.5 Limits of Boundary Layer
Weighted Error in Y (m)
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
-3
0 5 10 15 20 25 30
time(s)
0.5
Weighted Error
Weighted Error in phi(rad)
0.4 Limits of Boundary Layer
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
0 5 10 15 20 25 30
time(s)
References
1 Introduction
Magneto hydrodynamic bearings first evoked interest at the beginning of the second
half of the last century. Initial studies by Kuzma [1, 2], and Sasada et al. [3] considered
a fully flooded bearing. Kamiyama [4] introduced Reynolds boundary condition thus
considering a partially flooded bearing. Similar studies were also carried out by
Shvarts [5, 6] and Dudzinsky et al. [7]. Malik and Singh [8] extended such analyses
to the case of a finite bearing. Stability behavior of finite magnetohydrodynamic
journal bearings was reported by Kulkarni and Rao [9]. In the literature studied,
there were no instances of studies of dynamic parameters of such bearings or their
stability in a rotating coordinate frame. The present study is an effort to fill up this
void. In the process of attempting the same perturbed forms of the electromagnetic
boundary conditions and the consequent flow rate and electric field, relationships
have been derived and introduced in the formulations which will be discussed next.
2 Formulations
2.1 Assumptions
The bearing is assumed to be infinitely long and of circular cross section as shown in
Fig. 1. The radius of the shaft is R, the center-to-center distance between the bearing
and the shaft is e, with the maximum being the clearance C, while the attitude angle
is φ. W is the weight of the shaft. The x, y coordinate system is as described in the
figure. The film thickness h is a function of x. The shaft rotates with a speed of ω and
processes with a speed of ωp . The external magnetic field is radial and is given by the
relation B = −B0 R/r , where r is the radial distance of a point within the bearing
from the origin. The viscosity of the lubricant is μ, the electrical conductivity is σ,
the pressure of the fluid film is p, and ez is the electric field in the z-direction.
θ θ
F2Im1 (θ) = M 2 Mg0−1 + θc dθ F2Re1 (θ) = 0 (M E 0 − Dθ P0 )g0−1 g0 sin θdθ
0
θ θ
F1Re2 (θ) = 0 0.5M 2 g0−1 f 0 cos θdθ F1Im2 (θ) = M 2 Mg0−1 sin θ + (1 − cos θc ) dθ
0
θ 2 −1 θ
F2Re2 (θ) = 0 0.5M g0 f 0 sin θdθ F2Im2 (θ) = M 3 g0−1 cos θ + sin θc − 2 f 0 ε0−1 dθ
0
In keeping with Ref. [1], but with the following velocity, boundary conditions
commensurate with Couette flow of a rotating coordinate frame, namely
ωp ωp
U (η = 0) = 1 − = 1 − Dτ φ, U (η = 1) = − = −Dτ φ (1)
ω ω
The complete solution for the velocity profile is in any radial section of the fluid
film which is given in non-dimensional form by
1
Q= U H dη = M −3 (Dθ P − M E)g + M −1 (1 − 0.5Dτ φ) f (3)
0
Following Ref. [1], for an insulated bearing in open circuit condition, we have
1138 D. Tripathy and K. Bhattacharyya
θc θc
E H (θ )dθ = M Qdθ θc : cavitation boundary (5)
0 0
A new form of perturbation of the flow rate is introduced in order to handle the
electric field variables and the boundary conditions in a convenient fashion, namely
The perturbed flow rates and corresponding equations are obtained from (3) and
(4)
Q 1Re = M −3 (Dθ P1Re − M E 1Re )g0 + 0.5M 2 f 0 + (Dθ P0 − M E 0 )g0 cos θ ,
Dθ Q 1Re = 0 (9)
Q 1Im0 + Q 1Im sin θ = M −3 (Dθ P1Im − M E 1Im )g0 , Dθ [Q 1Im0 + Q 1Im sin θ ]
+ λ R cos θ = 0 (10)
Q 2Re = M −3 (Dθ P2Re − M E 2Re )g0 + 0.5M 2 f 0 + (Dθ P0 − M E 0 )g0 sin θ ,
Dθ Q 2Re = 0 (11)
∂ P0
Applying Reynolds boundary conditions P0 (0) = P0 (θc ) = ∂θ (θc )
= 0, we get
M f 0 (θc ) − E 0 g0 (θc )
Q0 = , P0 (θ ) = M 3 Q 0 F01 (θ ) − M 2 F02 (θ ) + M E 0 θ (14)
M2
Applying boundary conditions Pi (0) = Pi (θc ) = 0 for the perturbed pressures,
we get
θc θc
S R R = −2 P1Re (θ ) cos θ dθ , Sφ R = −2 P1Re (θ ) sin θ dθ ,
0 0
1140 D. Tripathy and K. Bhattacharyya
θc θc
S Rφ = −2 P2Re (θ ) cos θ dθ , Sφφ = −2 P2Re (θ ) sin θ dθ (18)
0 0
θc θc
2 2
DR R = − P1Im (θ ) cos θ dθ , Dφ R = − P1Im (θ ) sin θ dθ
λR λR
0 0
θc θc
2 2
D Rφ =− P2Im (θ ) cos θ dθ , Dφφ =− P2Im (θ ) sin θ dθ (19)
λR λR
0 0
The pressure profiles for a bearing operating at ε = 0.8 with variation in Hartmann
number (M) are shown in Fig. 2 The corresponding variation of non-dimensional
stiffnesses and dampings (for λR = 1) are shown in Fig. 3. The pressure profiles
corroborate results of [4] and indicate the effectiveness of the magnetic field in
bolstering the load-bearing capabilities of a fluid film bearing (Fig. 2). However, the
fact that a strong magnetic field is a mixed blessing is brought out by the variation of
stiffness and damping characteristics with increasing M. While S RR and S Rφ show a
decreasing trend with M, S φφ increases with M, and S φ R increases initially followed
by a decrease. Similarly DRR and DRφ show an increasing trend with M, while Dφφ
and Dφ R decrease. Hence, we may conclude that simply increasing M can result in
instability of the bearing as the effect of M on the different stiffness and damping
coefficients is not always favorable and a stability analysis is needed to design a
bearing with optimal M.
References
1. Kuzma DC (1963) The magnetohydrodynamic journal bearing. J Basic Eng Trans ASME Ser
D 85(3):424–427
2. Kuzma DC (1964) The finite magnetohydrodynamic journal bearing. J Basic Eng Trans ASME
Ser D 86(4):445–448
3. Sasada T, Kurosaki Y, Honda K, Kamijo K (1974) MHD journal bearing in a magnetic
field perpendicular to its axis 1st report, analysis of an infinitely long bearing. Bull JSME
17(114):1645–1651
4. Kamiyama S (1969) Magnetohydrodynamic journal bearing (report 1). Trans ASME J Lubr
Technol 91:380–386
5. Shvarts IA (1966) Magnetohydrodynamic lubrication theory for a cylindrical bearing.
Mekhanika Zhidkosti i Gaza 1(1):9–15
6. Shvarts IA (1966) Investigation of the basic characteristics of the MHD bearing, Izv. AN SS13R.
Meklaanika Zhidkosti i Gaza 1(4):189–191
7. Dudzinsky SJ, Young FJ, Hughes WF (1968) On the load capacity of the MHD journal bearing.
Trans ASME J Lubr Technol 90(1):139–144
8. Malik M, Singh DV (1980) Analysis of finite magnetohydrodynamic journal bearings. Wear
64:273–280
9. Kulkarni PA, Rao BVA (1977) Stability behaviour of finite MHD journal bearings. Mech Mach
Theory 12:293–302
Synergistic Effect of Pocket and Bionic
Texture on the Performance Behaviours
of Thrust Pad Bearing
Nomenclature
p = ph 22 /(μ0 R12 )
r Radial coordinate (m)
r = r/R1
r initial = (0.5 × R − r z )
rz Width of pocket in r direction (m)
R1 Inner radius of bearing (m)
R2 Bearing outer radius (m)
R (or B) Bearing width (m)
θf Circumferential extents of pocket/texture
θ0 Pad angle (degree)
θp Circumferential extents of pocket in case of pocketed-fish textured pad
μ0 Viscosity 40 °C temperature (N-s/m2 )
Speed (s− 1 )
φ Cavity or void fraction (1 − ρ/ρ0 )
1 Introduction
Due to growing awareness towards the energy and resource conservations, world-
wide efforts are being made by the investigators to build up energy-efficient pad thrust
bearings. A hydrodynamic thrust pad bearing is used for guiding and supporting the
axially (thrust) loaded rotors efficiently. It is essential to mention here that fluid film
pad thrust bearing involves smearing of the oil at the pad and runner interface, which
produces significant frictional losses. In addition to this, it has been found that 25–
50% of total fluid film bearing losses occur in feeding the oil to the bearings. This
loss arises due to power accounted in accelerating the oil in feed ports and grooves
and against the friction drag on the wetted shaft and thrust collar. This demon-
strates that it is an important task to decrease the power losses in bearings. Presently
different surface textures (involving different shapes of nano and micro geome-
tries) and macropockets are being employed by the researchers for exploring and
improving the tribo-performances of the different bearings [1–4]. It has been demon-
strated that for low film thickness ratios (in the range of 1.0001–1.5), the textures
provided on the sliders/pads towards the oil entry side yield the substantial-high load-
carrying capacity and reduced friction coefficient [5–9]. Employing microgrooves
of the square, trapezoidal, circular and triangular cross-sectional shapes on pads’
surfaces, numerical investigations have been conducted by the authors to find the
best performing cross-sectional shape of groove among these [6, 7]. This investiga-
tion revealed the best results with the microgrooves of square cross-sectional shape
positioned towards the entry region on the pad.
From the presented literature review, it is noticed that the design of surface textures
with tiny geometries and its positioning on the pad surface are very important from
the performance perspective of fluid film bearing. Moreover, it is also observed
that conceiving the new textures for exploring to improve further the performance
Synergistic Effect of Pocket and Bionic Texture on the … 1145
behaviour of thrust pad bearings is an important research task. Hence, the main
objective of this paper is set to conceive new textures on the thrust pad for numerical
investigation and, accordingly, to achieve the best performing texture ever reported
in the literature.
The partial differential equation employed in the numerical simulation and the
computational approach adopted is presented in this section. Figure 1a, b shows
the CAD model of sector-shaped thrust pads and coordinate system used in this
paper. However, Fig. 1c–e illustrate through the CAD model the geometries of the
pocket, the pad having fish texture and pocket with fish textured pads for visualiza-
tion point of view. Two views of these pockets/textures are shown in Fig. 1f–h with
major variable symbols.
The film thickness variation between thrust collar and the conventional pad is taken
in the numerical model using the following relation:
h = h 1 − (h 1 − h 2 )(θ/θ0 ) (1)
In the presence of pocket/texture on the surface of the pad, the film thickness
relation is modified as follows:
Fig. 1 a CAD model of sector-shaped pad thrust bearing; b employed coordinate system; c CAD
model of pocketed pad d CAD model of the fish textured pad; e CAD model of the pocketed-fish
textured pad; f two views of the pocketed pad; g two views of the fish textured pad; h two views of
the pocketed-fish textured pad
1146 J. C. Atwal and R. K. Pandey
h T = h + hd (2)
Using Eq. (3), the film thickness expressions for pocketed pad are as follows:
h; if θ ≤ θ f ; (R/2 − r z ) ≤ r ≤ (R/2 + r z )
hd = (4)
0; else
where k = 1, 2, . . . , Mr .
If ‘N f ’ is an even number, taking all values of ii for a value of k yields the
following expressions:
⎛ ⎞ ⎫
h ; if(ii − 1)θc ≤ θ ≤ iiθc ; rl ≤ r ≤ ru ; ii = 1, 3, 5, . . . , (N f − 1) ⎬
h d = ⎝ h ⎠ ; if(−1 + ii)θc ≤ θ ≤ θc ii; rl1 ≤ r ≤ ru1 ; ii = 2, 4, 6, . . . , N f
⎭
0 ; else
(6)
where the lubricant’s density ‘ρ’ is constant and is equal to ‘ρ 0 ’ in the full film zone;
φ is cavity fraction, which is equal to zero in the full film zone, and it varies between
0 and 1 in the cavitation zone.
The solution of Eq. (8) provides the values of p and φ. The constraint pertaining
to mass conservation is modelled using Fischer–Burmeister–Newton–Schur (FBNS)
algorithm [11]. FBNS approach leads to the following equation:
p+φ− p2 + φ 2 = 0 (9)
Equation (8) has been discretized employing the FEM. In this method, Eq. (8) is
multiplied with weight function N, which is assumed to be differentiable once for
r and θ, followed by integration of the resulting equation over the element domain
yields:
1 ∂ ∂ ∂φh ∂h
N (F1 ) + (F2 ) − + r dr dθ = 0 (10)
r ∂r ∂θ ∂θ ∂θ
e
3 3
where F1 = h
r ∂p,
6μ ∂r
F2 = 1 h ∂p
r 2 6μ ∂θ
.
∂ ∂N ∂ ∂ F1 ∂N ∂
(N F1 ) = F1 + N F1 or − N = F1 − (N F1 ) (11)
∂r ∂r ∂r ∂r ∂r ∂r
∂ ∂ F1 ∂N ∂ F2 ∂N ∂
(F1 N ) = N + F1 or − N = F2 − (N F2 ) (12)
∂θ ∂θ ∂θ ∂θ ∂θ ∂θ
Using Eqs. (11) and (12) in Eq. (10) results:
1 ∂ ∂N ∂ ∂N ∂φh ∂h
(F N1 ) − F1 + (F2 N ) − F2 − N + N r dr dθ = 0 (13)
r ∂r ∂r ∂θ ∂θ ∂θ ∂θ
e
3 3
Assuming qn = n r 6μ
h
r ∂∂rp + 1 h ∂p
n ,
r 6μ ∂θ θ
Eq. (16) becomes:
3 3
h ∂N ∂ p 1 h ∂N ∂ p ∂φh ∂h
+ 2 +N −N r dr dθ − N qn ds = 0 (17)
6μ ∂r ∂r r 6μ ∂θ ∂θ ∂θ ∂θ
e e
n
n
p e (r , θ ) = p ej N j , φ e h(r , θ ) = φ ej h j N j (18)
j=1 j=1
The ‘2n’ independent e e algebraic equations are needed to find the values
of 2n
e e unknowns p
1 2, p , . . . , p e
n . Therefore, n independent shape functions
N1 , N2 , . . . , Nne are used for generating n equations and other n equations are
generated found complimentary constraints.
Hence, the ith equation is obtained by substituting N = Nie in Eq. (19) as follows:
Synergistic Effect of Pocket and Bionic Texture on the … 1149
⎧ ⎫
n ⎨
3 3 ⎬
h ∂ Ni ∂ N j 1 h ∂ Ni ∂ N j
+ 2 r dr dθ p ej
⎩ 6μ ∂r ∂r r 6μ ∂θ ∂θ ⎭
j=1 e
⎧ ⎫
n ⎨ ⎬
∂Nj ∂h
+ Ni h j r dr dθ φ j + Ni −
e
r dr dθ − Ni qn ds = 0
⎩ ∂θ ⎭ ∂θ
j=1 e e e
(20)
n
n
K iej p ej + Biej φ ej − f ie − Q ie = 0 (21)
j=1 j=1
where
3 3
h ∂ Ni ∂ N j 1 h ∂ Ni ∂ N j . ∂Nj
K iej = + 2 r dr dθ, ..Biej = Ni h j r dr dθ,
6μ ∂r ∂r r 6μ ∂θ ∂θ ∂θ
e e
∂h
f ie = Ni r dr dθ , Q ie = Ni qn ds
∂θ
e e
The boundary pressure values are known; thus, the column vector {Q} is omitted
in the calculation of pressure and fractional film content. Numerical results have been
generated using elements 160(Nθ ) × 160(Nr ) in the computational domain. These
numbers of elements are chosen based on mesh independent tests. The following
convergence criteria have been employed during the numerical computation of
variables:
Nθ +1 Nr +1
| p(i, j)new − p(i, j)old |
≤ 10−6 (23)
i=1 j=1
| p(i, j)old |
Nθ +1 Nr +1
|φ(i, j)new − φ(i, j)old |
≤ 10−6 (24)
i=1 j=1
|φ(i, j)old |
1150 J. C. Atwal and R. K. Pandey
The pressure result obtained based on the proposed model has been compared with
the results of the authors [5] for developing confidence in the present investigation.
Excellent matching between both the findings is found in Fig. 2; thus, the proposed
model is considered validated. Input data of Table 1 are used for producing the results
presented in this paper.
Figure 3 shows the comparative pressure profiles at the mid-width of the pad for
the data mentioned in Table 1 and given in the caption. It can be seen in this figure that
for constant minimum film thickness and film thickness ratio, the pressure values are
maximum with the pocketed-textured pad. High-pressure values are obtained with
a pocketed-textured pad due to the largest restriction of flow in the circumferential
direction. Variations of enhancement of film load capacity and reduction in coefficient
of friction as compared to the conventional pad are presented in Fig. 4. The maximum
increase in the load capacity occurs at 50 µm of texture and pocket depths.
Fig. 3 Variation of pressure profiles with pocket/textured pads (U = 11 ms−1 , sh = 16.20 µm, h2
= 30.0 µm, θ f /θ 0 = 0.7, θ p /θ 0 = 0.6, b/a = 0.5, b/R = 0.01, hd = 50.0 µm, 2r z /R = 0.9)
Fig. 4 Performance parameters variation with depth of the pocket and texture (U = 11 m/s, sh =
16.20 µm, h2 = 30.0 µm, θ f /θ 0 = 0.7, θ p /θ 0 = 0.6, b/a = 0.50, b/R = 0.01, hd = 50.0 µm,
2r z /R = 0.90)
Figure 5 illustrates the variation of film load capacity and friction coefficient with
fish texture spacing. The maximum load capacity and minimum coefficient of friction
with fish textured pad occur at a dimensionless spacing value of 0.01. However, Fig. 6
represents the performance parameters of the bearing with a dimensionless circumfer-
ential extent for pocketed with a fish texture pad. The dimensionless circumferential
extent of 0.6 for pocket then followed by a dimensionless circumferential extent
of 0.1 for fish texture yield the highest film load capacity and minimum friction
coefficient.
1152 J. C. Atwal and R. K. Pandey
Fig. 5 Variation of performance parameters with b/R for fish texture pad (U = 11 m/s, sh =
16.2 µm, h2 = 30 µm, θ f /θ 0 = 0.7, hd = 50 µm, 2r z /R = 0.9)
Fig. 6 Performance parameters variation with θp/θ 0 for pocket with fish textured pad (U = 11 m/s,
sh = 16.20 µm, h2 = 30.0 µm, θ f /θ 0 = 0.7, b/a = 0.50, hd = 50.0 µm, 2r z /R = 0.90)
From Fig. 7, it can be noted that with a rise in dimensionless circumferential extent,
the load-carrying capacity first increases up to a point (dimensionless circumferential
extent = 0.7), and thereafter, it diminishes. Moreover, the pad having pockets shows
greater load capacity compared to a fish textured pad. Enhanced pressure magnitude
in the pocket pad gives a higher value of load-carrying capacity than the fish textured
pad. Figure 8 depicts the variation of film load capacity and friction coefficient of
the fish textured pad with the parameter ‘b/a’. It can be seen that as the value of
‘b/a’ increases, the film load capacity decreases and friction coefficient increases.
The film load capacity is maximum at ‘b/a = 0.5’, and it is minimum at ‘b/a = 2’.
Synergistic Effect of Pocket and Bionic Texture on the … 1153
Fig. 7 Performance parameters variation with θ f /θ 0 for pocket and fish textured pad (U = 11 m/s,
sh = 16.20 µm, h2 = 30.0 µm, b/a = 0.50, hd = 50.0 µm, 2r z /R = 0.90)
Fig. 8 Performance parameters variation with b/a for fish textured pad (U = 11 m/s, sh = 16.20 µm,
h2 = 30.0 µm, θ f /θ 0 = 0.70, hd = 50.0 µm, 2r z /R = 0.90)
1154 J. C. Atwal and R. K. Pandey
4 Conclusions
Based on the investigations presented herein, the following points are concluded:
1. The pocket with fish texture, pocket and fish texture pads have yielded perfor-
mance improvement in the bearings in comparison with the conventional thrust
pad.
2. The pocket with fish texture, pocket and fish texture pads have yielded enhance-
ment in the film load capacity in the range of 85–93%, 83–89% and 62–71%,
respectively, in comparison to the conventional pad.
3. Friction coefficient has reduced with pocket/texture pads. The pocketed-fish
textured pad, pocketed pad and fish textured pad have yielded a reduction in
the friction coefficient falling in the range of 45–47%, 43–46% and 39–41%,
respectively, in comparison to the conventional pad.
4. Texture depth played a vital role in improving the performance behaviour of
thrust pads. Texture depth varies with operating parameters.
5. Pocketed-fish textured pad possessing the pocket and texture up to the
dimensionless circumferential length of 0.7 produced optimal performance
behaviours.
References
9. Atwal JC, Pandey RK (2020) Film thickness and friction investigations in a fluid film thrust
bearing employing new conceived micro-texture on pads. J Tribol. https://doi.org/10.1115/1.
4048500
10. Giacopini M, Fowell MT, Dini D, Strozzi A (2010) A mass-conserving complementarity
formulation to study lubricant films in the presence of cavitation. J Tribol 132:041702
11. Woloszynski T, Podsiadlo P, Stachowiak GW (2015) Efficient solution to the cavitation problem
in hydrodynamic lubrication. Tribol Lett 58:18
Analysis of a Soil-Moisture Sensor
for Potential Failure Modes and Mass
Manufacturing
1 Introduction
(0.5 mm diameter) and enclosed in a steel tube (0.8 mm ID and 1.5 mm OD). It
works on the principle of Joule heating by applying a certain voltage pulse (3.3 V)
across the nichrome wire for predetermined time. This heat travels through the soil
and peak temperature is detected by the thermocouple which is at a distance from
the heater. Based on different moisture levels of the soil, the volumetric specific heat
of the moist soil varies, which is computed using the temperature rise recorded by
the thermocouple. Thus, the moisture level is detected using the relation between
volumetric specific heat and temperature rise [3]:
q
Tmax = (1)
eπr 2 C
where Tmax is the maximum rise in temperature at a radial distance r from the
heater, q the heat input, and C the volumetric heat capacity of the soil.
2 Reliability Analysis
A design for failure mode and effect analysis (DFMEA) [4] was performed on the
envisioned sensor by listing the potential failure modes and prioritizing them to
mitigate the undesired effects sequentially. The design changes suggested by DFMEA
were checked for automated manufacturing and overall reliability before finalizing
the design and process.
By considering the repeatability and precision of sensors, changes in design were
made. Hence, a reliability study was carried out to check for the long-term effect of
the design changes. There are many tools and techniques available for improving the
Analysis of a Soil-Moisture Sensor for Potential Failure Modes … 1159
reliability of the system; we have used design for failure modes and effect analysis
(DFMEA), which involves identifying the potential failure modes and improving the
design accordingly, thus making the sensor robust.
DFMEA involves assigning an individual number (out of 10) to a failure mode
based on its severity (SEV), occurrence (OCC), and detectability (DET). These scores
are multiplied to get the risk priority number (RPN). The RPN is used to prioritize
the failure modes so that corrective actions can be taken to reduce the frequency,
severity, and/or improve the detectability of the failure mode. The objective is to
design the sensor such that RPN value is minimized to the target value (or less).
DFMEA can be conducted at a component, subsystem, or system level. At the
subsystem level, the heater probe is a critical component of the sensor, and DFMEA
for the same was carried out as can be seen in Table 1.
Based on RPN of the initial design, the following modifications were made:
1. Thin-film Parylene coating is replaced with powder-coating technique. Powder
coating has better adhesive properties as compared to Parylene and does not
affect the heating of the nichrome. Hence, it is capable of serving the primary
function of electrically insulating the nichrome wire in a much more efficient
manner as compared to Parylene.
2. A single thicker copper wire is used instead of folding a thin copper wire. The
thick wire could be aligned in the center of the steel tube with the support from
casing, thus ensuring uniform distribution of heat in all directions. This is in
contrast to folded copper wire, which could not be aligned centrally and had the
disadvantage of localized heating at certain points where the wire could touch
the steel tube (Fig. 2).
3. Pulse-arc welding is used to seal the tube from one end. But due to excessive heat
released during welding, the surface of steel can oxidize, thus exposing the free
iron ions to moisture, which can corrode the steel surface. This was overcome
using a standard process of passivation, which removes the oxide layer and
forms a thin and strong corrosion-resistive chromium oxide layer [5, 6].
4. Parylene coating could be damaged easily due to the friction between the
nichrome wire and steel tube. Although thermal grease did not provide a
significant contribution to heat conduction, due to its electrically insulating
and thermally conducting nature, it was used to prevent shorting of the circuit
between the nichrome and steel tube. Due to the good adhesive strength of
powder coating, thermal grease can be eliminated. This led to an elimination of
many problems from thermal grease (e.g., uniformly filling in steel tube) while
bringing down the cost of the sensor.
5. The voltage capacity of the nichrome wire was checked by increasing the voltage
at a fixed rate for a certain time period. It was observed that powder coated
nichrome can sustain voltage up to 13 V, which suits our application since the
maximum operating voltage for the sensor is 5 V.
The DFMEA chart was updated as shown in Table 2. It can be observed that the
RPN values have been reduced significantly. Additional electrical testing will be
carried out to ensure the reliability of the heater element in extreme conditions as
1160
Table 1 DFMEA chart for the initial design of the heater probe
System Soil-moisture sensor
Subsystem Heater probe
Component
Functional Potential failure Potential effects of SEV Potential OCC Current DET RPN Recommended
parameter/design modes failure causes/mechanism design/process actions
parameter of failure controls
SS corrosion Corrosion of SS Damages the heating 8 Welding of one end 5 4 160 Passivation of steel
resistance tube in presence of element of SS tube leads to
moisture corrosion
SS strength Bending/buckling Damages the heating 8 Material properties 3 3 72 Standard tests to be
of SS tube element conducted
Nichrome tensile Nichrome wire Breaks the circuit 8 Weak material 4 Sensor might 4 128 Check with
properties breaks properties stop working automation
prototype
Parylene coating Can come off Shorts the circuit 8 Weak adhesive 8 Erroneous 7 448 Alternate coating
of nichrome wire easily coating reading required
Folded copper Can touch the SS Parylene can scrape 7 SS tube is 8 Erroneous 5 280 Passivation of steel
wire tube leading to off due to friction electrically reading
localized heating conducting in nature
Thermal grease Non-uniform Uneven conduction 6 High viscosity of 5 Erroneous 6 150
(TG) filling of SS tube TG reading
Voltage capacity Wire breaks from Breaks the circuit 8 Surge in voltage 3 Sensor might 3 72 Check failure
of nichrome overheating stop working modes of nichrome
with coating for
working voltage
M. Mogra et al.
Analysis of a Soil-Moisture Sensor for Potential Failure Modes … 1161
Fig. 2 a Initial design of heater element of sensor with nichrome wire wounded around folded
copper wire. b Modified design of heater element of sensor with nichrome wire wound and re-wound
around thick copper wire
well. Similarly, DFMEA will be constructed for other subsystems and components
to ensure the overall reliability of the sensor.
Table 2 DFMEA chart for the modified design of the heater probe
System Soil-moisture sensor
subsystem Heater probe
component
Functional Potential failure Potential effects SEV Potential OCC Current DET RPN Recommended
parameter/design modes of failure causes/mechanism of design/process actions
parameter failure controls
SS corrosion Corrosion of SS Damages the 8 Welding of one end 2 4 64
resistance tube in presence of heating element of SS tube leads to
moisture corrosion
SS strength Bending/buckling of Damages the 8 Material properties 3 3 72 Standard
SS tube heating element strength tests to
be conducted
Nichrome tensile Nichrome wire Shorts the circuit 8 Material properties 4 Erroneous 4 128 Check with
properties breaks reading automation
prototype
Powder coating of Some spots on wire Shorts the circuit 8 Improper spraying 3 Erroneous 3 72 Conduct
nichrome wire without coating reading experiments to
check adhesive
strength
Voltage capacity Wire breaks from Breaks the circuit 8 Surge in voltage 1 Sensor might 3 24
of nichrome overheating stop working
M. Mogra et al.
Analysis of a Soil-Moisture Sensor for Potential Failure Modes … 1163
are connected. The intersection of the wires is glued to the hollow shaft connected
to the driven bevel gear.
The initial distance between the gripper arms is equal to the length of nichrome to
be wound around copper. The nichrome wire is fed to the gripper arms and is cut after
the gripper has held the two ends firmly. The nichrome wire is glued to the copper
wire at a single point before the process starts. When the arms start rotating, the
copper is fed along the axial direction. Since the arms rotate in the opposite direction
1164 M. Mogra et al.
around the copper wire, nichrome starts winding from both sides of the copper wire.
As the nichrome starts to wind, the distance between the gripper must decrease since
the length of the nichrome wire between the grippers is constant. Since the nichrome
does not have enough strength to pull the gripper arms, the stronger wire (which also
starts winding around the hollow shaft simultaneously as nichrome starts to wind
around copper) pulls the arm inwards. As the stronger wire winds around the hollow
shaft, it will start drawing wire from the respective spools connected to the hinges.
This tension generated in the stronger wire while drawing from spool is responsible
for pulling the arms inwards toward the axis. Thus, the required length is wound
around the copper. Once the winding process completes, the torsion spring inside
the spool pulls the strong wire back, which rotates the hollow shaft in the opposite
direction, causing the arms moves back to their original position, thus making it
ready for the next iteration of winding.
For the mechanism described in the previous section to work properly, the feed rate
of copper wire and the rotational speed of the arms have to be worked out.
As seen from Fig. 5a, the nichrome wire of length ln (87 mm) has to be wound
around copper of length lc (15 mm). The diameter of the copper wire is dc , x is the
length set aside on either side of the nichrome wire for soldering onto the connecting
board. Hence, the length that needs to be wound is (ln − 2x). We have considered x
Analysis of a Soil-Moisture Sensor for Potential Failure Modes … 1165
to be 10 mm. Using this data, we can calculate the number of turns in winding and
spacing between each winding.
From Fig. 5b, it can be inferred that the length of single winding is (π dc )2 + s 2 ,
where s is the spacing between each winding. Since winding takes place from both
sides of the copper wire simultaneously, the number of turns in winding (n) and the
spacing between each winding (s) can be found using
ln − 2x
n (π dc )2 + s 2 = (2)
2
ns = lc (3)
Assume that the time taken to wound nichrome around copper is t. Then, in this
time t, the copper wire needs to be fed by lc . Therefore, the feed rate of copper =
lc /t. Since the lengths of arm 1 and arm 2 (r1 and r2 ) are different, the rotational
velocities of each arm shall be different. However, each arm has to wound the same
length in time t so, the tangential velocity of each arm will be equal. Thus,
ln − 2x
ω1r1 = ω2 r2 =
2t
Nichrome wire wound around copper forms the heater element. To form the complete
heater, the heater element is placed inside a stainless steel (SS) tube. The length of
the copper wire is equal to the length of the SS tube, and the inner diameter of the
SS tube has a value close to the diameter of copper wire (with the nichrome wound
around it). Hence, the heater element fits perfectly inside the SS tube. To ensure it
is placed centrally inside the SS tube and to prevent any loose assembly, the heater
1166 M. Mogra et al.
element is secured inside the SS tube by applying a drop of glue on the top of the
heater element after it is assembled inside the tube.
To obtain the desired assembly, the process shown in Fig. 6 is proposed. When
the construction of heater element from the winding process is completed, it is sent
to the inclined plane, which funnels it into the nozzle. A plunger is then actuated to
push the heater element into the steel tube beneath it, which is carried by the conveyer
belt. The speed of the conveyor can be determined based on the supply rate of the
heater element from the winding process. While pushing the heater element into the
steel tube, the plunger also deposits a drop of glue on top of it (similar to the working
of a syringe). Hence, it becomes a continuous process and the heater is assembled
at regular intervals. This completes the assembly of the heater which is then sent to
other processes as depicted in Fig. 3.
4 Closure
The automation setup will bring down the cost of making a sensor by cutting down the
labor costs involved in making the critical parts. The setup improves the reliability of
the system by bringing uniformity in manufacturing. A reliability study was carried
out for the DPHP soil-moisture sensor, which resulted in systematic improvement
in the performance of the sensor. Further, assembly processes can also be automated
similarly. This would enable affordable and quick mass manufacturing of sensors,
thus enabling the objective of deployment of DPHP soil-moisture sensors in large
quantities in the field.
Analysis of a Soil-Moisture Sensor for Potential Failure Modes … 1167
Acknowledgements This work was supported by the NNeTRA project of the Indian Ministry of
Electronics and Information Technology.
References
Abstract Welds are most susceptible to the vibration loading. When subjected to
loading, weld failures are observed at the throat or at the toe. Vibration fatigue evalua-
tion is different from that of static fatigue evaluation as dynamic characteristics of the
considered system play important part in stress generation. Weld fatigue under vibra-
tion needs different methodologies and one of these is explained in this paper. The
paper contains details of the application of notch stress approach and its comparison
with the two other approaches, namely nominal and structural stress approach.
1 Introduction
Welded joints are permanent type of joints which have wide variety of applications.
Lightweight design is the advantage, welded joints provide over bolted and riveted
joints. With the large-scale applications, there comes the necessity to design and
analyze these joints with accuracy. Due to inherent complex damage mechanisms
which include material irregularities, microcracks in the weld bid, and the local notch
at the weld toe, simulating weld joint is a difficult task [1].
Weld failures are usually observed at the weld root and toe. There are multiple
theories which are proposed to evaluate weld fatigue: Nominal stress theory, Struc-
tural stress theory, Notch stress theory, and Linear Elastic Fracture Mechanics
(LEFM) [2]. These theories are classified into two categories as global and local
theories. Local theories are more precise, but they require more computational time
as compared to the global theories. Hence, it is advisable to choose the theory based
on the accuracy needed and the amount of time available for the computation. It
has been observed that various theories have various limitations too. Structural stress
theory is useful for the failures originating from the toe while notch stress approach is
useful for the both type of failures: originating from toe and root [3]. Nominal stress
approach can be used for evaluating welds in a very quick time but with limited
accuracy [4].
Nominal stress approach uses the S–N curves to represent the standard weld
configurations using FAT numbers. International Institute of Welding (IIW) has
provided set of S–N curves which provide S–N curve for each configuration [5].
Though most of the common configurations are covered under nominal stress
approach, there are product design-specific welds that are not available under this
approach.
Structural stress theory can be utilized when observed failure is originating from
the toe of the weld. This theory works for hand in hand with Finite Element Analysis
(FEA) to identify the high-stress area which are also called Hot Spots. Hence this
theory is also called as Hot spot stress theory [6]. In this approach, stress at the weld
is identified using the extrapolation method. Stresses are obtained at the distance 0.4t
and 1.0t and extrapolated using the following formula [7].
According to IIW, notch stress is the total stress at the root of the notch. The decrement
in the stress at the notch, for a certain depth, can be replicated using the enlarged
notch radius at the weld—parent metal interface. This method was developed by
Radaj.
Figure 1 shows the difference between structural stress and the notch stress at the
similar location for the same loading conditions. It is evident from the figure that
structural stress does contain only membrane and the bending stresses but not the
nonlinear stresses due to local stress raisers. Hence, notch stress includes both the
type of stresses: due to component geometry and due to local stress raisers. Notch
stress for the complex geometries can be found out with FEM modeling. Care must
be taken to mesh the notched area of the weld properly. IIW recommends element
size less than 1/6th of the notch radius for the linear elements and element size
less than 1/4th of the notch radius for higher order elements [9]. Notch needs to
be meshed with proper sized elements and to be connected with other parts with
smooth transition mesh. For the huge models, sub modeling needs to be done as this
Evaluation and Validation of Weld Joint Fatigue in Vibration … 1171
method is mesh sensitive. Due to high mesh density, this approach requires higher
computational time as compared to other approaches. Based on the desired higher
accuracy of the results and availability of computational resources, this approach can
be implemented.
3.1 Modeling
Notch stress approach has been implemented on the tube to plate weld to predict
the structural integrity of the structure under vibration loading in FEA. The structure
consists of a tube which carries a load at the one end while the other end is fixed to
the plate via welding. This whole structure is subjected to the vibration loading at
the base. Vibration test is performed in two steps:
1. Sweep from 0 to 30 Hz to identify the natural frequency of the structure.
2. Dwell at the obtained frequency in the sweep for the output response of 3G.
For qualification, the required number of test cycles the structure should withstand
is 1.0 × 105 . Based on the frequency identified in the sweep, the test duration in
seconds is calculated by multiplying frequency with test cycles, for which structure
should sustain the vibration.
The detailed simulation procedure and test procedure is as below:
The geometrical representation of the structure assessed and the details about the
welded geometry are as shown in Fig. 2. According to Notch stress approach, the
notch details/geometry needs to be introduced in the toe of definite dimensions. For
the tube thickness more than 5 mm, notch radius should be of 1mm while for the tube
thickness less than 5 mm, notch radius reduces to 0.05 mm [1]. In this application,
0.05 mm of notch radius has been used.
1172 A. Patil and S. Bhende
Fig. 2 Details about the assembly understudy and position of externally introduced notch
All different parts of the assembly are connected through bonded contacts at the
required locations. The meshing was performed in accordance with IIW recommen-
dations mentioned in the introduction section. The precaution was taken to mesh the
notch and area around notch with high density mesh to capture stresses appropriately.
Figure 3 shows the meshing used in the modelling for the current assembly.
Input acceleration is applied at the base (at the constraint node locations shown
in Fig. 4) in such a way that 3G response is recorded at the Accelerometer mounted
near center of gravity.
Frequencies obtained from the frequency analysis with the above boundary condi-
tions are 7 Hz (Dominant in Horizontal plane), 10 Hz (Dominant in Vertical plane),
59 Hz and 60 Hz for the first four modes. Out of these four, 7 and 10 Hz, being within
the excitation frequency range till 30 Hz, are of interest. The mode shapes for these
frequencies can be interpreted from Fig. 5.
Evaluation and Validation of Weld Joint Fatigue in Vibration … 1173
After the frequency analysis, harmonic analysis was performed to predict the
stresses. It was found that 10 Hz mode was most damaging and resultant stresses
obtained are as shown in Fig. 6.
From the obtained stresses, the next step was to obtain life by utilizing an appro-
priate S–N curve. According to IIW standards, the required S–N curve for Aluminum
is FAT 71. Based on the universal curve mentioned above, calculated life is less than
1.0 × 105 cycles and hence, analysis predicted failure for the structure in the test
(Table 1).
Test was performed with the help of vibration shaker to obtain 3G response at the
CG. Sine sweep performed on the structure showed a resonance frequency at 9.15 Hz
in a vertical plane (Fig. 7).
The obtained frequency from the analysis is 10 Hz which shows that test validates
the simulation. When structure was vibrated (Dwell test) at the obtained frequency
from the sweep, crack was observed at the weld toe which further validates the theory
as well as the failure predictions in the considered case. Figure 8 shows the cracked
region of the tube.
Table 2 Comparison table for the approaches used for the evaluation of weld fatigue in vibration
Approach Applicability to the Meshing Geometry S–N curve
current problem modification
Nominal stress NA Mesh insensitive NA NA
approach
Structural stress Applicable Mesh sensitive NA FAT 40
approach
Notch stress Applicable Mesh sensitive Introduction of FAT 71
approach notch
4 Conclusions
Notch stress approach can be effectively used to evaluate weld fatigue through simu-
lations. Notch stress approach is sensitive to mesh density near the notch. Hence,
sufficient solid 3D mesh needs to be modeled near the entire area of interest which
increases the time for solution. Nominal stress approach is applicable for standard
set of weld configuration while structural stress approach is applicable for weld toe
failures. Notch stress can be used to evaluate weld failures of all kinds. Availability
of Simulation time and amount of required accuracy are the deciding factors while
employing this approach.
Table 2 depicts the actual comparison between all the three approaches [7].
References
1. Fricke W (2012) IIW guideline for the assessment of weld root fatigue. IIW Guideline for the
Assessment of Weld Root Fatigue, Revision 3
2. Zamzami I, Susmel L (2016) On the accuracy of nominal, structural, and local stress-based
approaches in designing aluminium welded joints against fatigue. Int J Fatigue. ISSN 0142-1123
1176 A. Patil and S. Bhende
3. Malikoutsakis M, Savaidis G (2009) An approach to the effective notch stress concept to complex
geometry welds focusing on the Fe modeling of weld ends. In: 3rd ANSA & μETA international
conference
4. Wægter J (2009) Fatigue design based on S-N data
5. Hobbacher A (2016) Recommendations for fatigue design of welded joints and components.
IIW Document IIW-2259-15
6. Fuštar B, Lukacević I, Dujmovi D (2018) Review of fatigue assessment methods for welded
steel structures
7. Patil A, Bhende S (2018) Evaluation and comparison of frequency domain fatigue life approaches
for weld joints. In: F2018-NVB-019, SAE FISITA conference
Trajectory Control and Force Control
of Biomimetic Fingers by Tendon-Based
Actuation System Using Bond Graph
Vijay Saini, Simran Pal Singh, Neeraj Mishra, and Anand Vaz
1 Introduction
The human hand is a vital organ in our human body. It can perform a variety of tasks
very easily and efficiently due to its complex internal structure and arrangement.
Humans easily learn how to move and control the movements of the hands and the
fingers to grasp and hold objects. Yet how are these movements controlled [1]? It is
easy for humans to perform prehensile tasks. What is the control mechanism involved
in performing these tasks? A need is therefore felt to develop an experimental setup to
answer these questions. In addition, if there is partial hand impairment, it is not easy to
perform the above tasks. Partial human hand impairment considered here is when the
hand has lost one or more fingers but retains the ability of the remaining finger. The
experimental setup should be such that it cannot only explain the complex actuation
system using muscles and tendons but also the concept of the prosthesis and how to
actuate prosthetic fingers using the remaining natural fingers [2, 3]. The proposed
experimental setup additionally addresses the issues of trajectory control and force
control of fingers using a tendon-based actuation system. It also demonstrates the
implementation of the concept of ‘opposition space’ [4].
An experimental setup has been designed and developed to address these issues.
It consists of various essential components such as two prosthetic single link fingers
(developed through rapid prototyping), slider-nut and lead-screw mechanism, fishing
wires, string-tube mechanism, 12 V DC motor, Motor Driver (L298N), Arduino
UNO microcontroller (ATmega328P), Optical potentiometer (COPAL JT22-320-
500), force sensor, two power supplies, etc. Fishing wires act like tendons, slider-nut
and lead-screw mechanism acts like muscles, Arduino microcontroller acts as the
brain of the human, which generates the PWM signals. These control signals, which
are based on the PID algorithm, are used to drive the motor. The motor is coupled
with the slider-nut and lead-screw mechanism. The optical potentiometer is used
as angle feedback at the joint of the finger, and force sensor is used for feedback of
force at the tip of the finger. The experimental setup has been modelled and simulated
through the bond graph approach [5].
2 Methodology
The designing and development of the experiment setup are explained in detail. An
experimental setup is prepared for replicating the complex mechanism of the index
finger and the thumb and their actuation through tendons and muscles.
Trajectory Control and Force Control of Biomimetic Fingers by … 1179
Opcal potenometer
Fixed link
Proximal Link
link 2
G1
Passive Finger
Finger holder G2
H2
H1
Distal link
G3
H3
part 1 G4
H4
In this work, the two prosthetic fingers each contain a single link. Each link acts
as a distal part of the respective fingers. It is sufficient to perform the objective of
implementing the trajectory control and force control of biomimetic finger using
tendon-based actuation system.
The setup is composed of two single link fingers: link 1 of finger 1 and link 2 of
finger 2. Finger 1 acts like an index finger, and finger 2 acts like a thumb. These two
links are hinged on a third fixed link as shown in Fig. 1. In the actual human finger,
finger joints are generally spherical. However, in this model, the natural spherical
joints are replaced by the hinged joints containing a pulley type of arrangement [6].
Each joint contains two ball bearings to reduce the friction. This pulley arrangement
is the part of distal link. It is designed in such a way that tendons could be wrapped
easily around the pulleys and provides actuation torque to the joint. The torque helps
in the rotation of the links with respect to the fixed link. The holes are used for
fixing one end of the tendons so that the links get the rotational effect. There is a
potentiometer holder to hold the optical potentiometer. The potentiometer holder is
bolted on to the fixed link. Designing and drafting of the different part of 3D printed
prosthetic fingers are carried out in SOLIDWORKS, and their assembly is as shown
in Fig. 1.
Flexible
Coupling
Nut
DC Motor
(12V)
Gear Box
Lead-Screw
Frame
Ball
Bearing
When the motor rotates, power is transmitted to the screw through the coupling.
This rotation of the lead-screw results in the translational movement of the nut. As
the nut slides on a frame, to reduce its friction, a lubricant is applied underneath
the nut. The direction of the nut movement is decided by the direction of rotation
of motor either in a clockwise or anticlockwise direction. One end of the tendons is
connected to the nut so it could generate a force in the tendon likewise in the actual
muscle–tendon arrangement.
The string-tube mechanism is a great mechanism for transforming the motion from
one joint to the other joint. A major benefit of the string-tube mechanism is that the
joints need not be parallel and need not be fixed. This mechanism is very helpful in
the case where the distance between the two joints is changing with time:
Active connection. In the setup, Link 1 is an active link as a drive from the motor
through the tendons is given to this link. The link acts like an index finger which
is the remaining working finger of the patient. A string-tube mechanism is required
to transmit the motion from the nut of the lead-screw and slider-nut mechanism to
the active joint of link 1. The string-tube mechanism is used for the actuation of the
link as shown in Fig. 3. Two tendons per link joint are needed to actuate the link
as tendons work only in tension. One tendon is for flexion, and the other is for the
extension. One end of the tube is fixed on a fixed link while the corresponding string
end is connected to the next link 1 of the same joint. Each string winds around the
pulley of distal link. The other ends of the tubes are fixed on the slider-nut frame
while the corresponding string ends are connected to the common slider-nut in an
alternate configuration. In this configuration, one string is connected at the head and
the other on the tail side of the nut. This configuration is used to reduce the number
of actuators as both the strings of a link are actuated using a common slider-nut.
Trajectory Control and Force Control of Biomimetic Fingers by … 1181
link 2
Passive Finger
Fixed link
Proximal Link
String-tube 1 String 1
Nut String-tube 2 Link 1
String 1 Acve Finger
String 2 Distal link
DC motor String
ends fixed
Tube ends fixed String ends String 2 Tube ends on distal
on the frame of fixed on nut fixed on link
lead screw proximal link
mechanism
Passive connection. In the setup, link 2 is a passive link as no direct drive is given to
the link 2. As it is coupled to the finger of link 1 in an unlike configuration through
strings, the motion of link 1 will drive link 2 through passive connection. The passive
finger acts like a prosthetic finger say thumb which will get the drive from the active
finger. In an unlike configuration, the prosthetic thumb will rotate in the same sense
as that of the active finger but in opposite direction. In the proposed setup, for the
passive connection as shown in Fig. 4, it is not required to use the string and tube
mechanism as both the fingers joints of link 1 and link 2 are parallel and fixed. So,
only strings are used to actuate the passive finger. Two strings are required for passive
connections as well. One end of the first string is connected to the link 1 and then
winds around the pulley of link 1 and to the pulley of link 2 and then the other end is
String 2
Fixed
link
String 1
Proximal
Link
Link 1
Acve Finger
Distal link
String ends
Acve joint
fixed on
link 1
1182 V. Saini et al.
ForceSensor
Sensor Optical Force Passive
Potentiometer Force
Circuit
Circuit Potentiometer Sensor Finger
Circuit
Motor
Driver
L298N
Arduino UNO
Microcontroller
Geared
DC Motor
(12V)
Lead Screw and String and Passive Fixed Link Tendon Active Finger
Slider-Nut Tube Connection
Mechanism Mechanism
connected to the link 2. Likewise, the second string is required as string works only
in tension. Rotation of link 1 will exert a torque in the active joint. This torque will
exert a force in the strings. This force will transmit tangential to the joint of link 2
and then generates torque in the joint of link 2. This torque will result in the rotation
of link 2. During this rotation, the second string does not play any role as it is in
a loose state. Its role will come into the picture when link 1 rotates in the opposite
direction. At that time, the first string will come into a loose state.
A fully assembled experimental setup for trajectory and force control of tendons of
a biomimetic finger is shown in Fig. 5.
In this figure, the interconnection between different components and workspace
of the setup is as shown.
The human hand has the capability of applying different forces depending upon the
shape, size and weight of the object to grasp and lift the object. Therefore, there is a
need to control the fingertip forces so that the prosthetic finger can apply the exact
force which the patient should require to handle an object with precision. So, a force
sensor is required to get the feedback from the fingertip.
In the proposed work, a force sensor using strain gauges has designed and devel-
oped as shown in Fig. 6, which can even measure the minute forces even the touch
on the fingertip of the prosthetic finger.
Trajectory Control and Force Control of Biomimetic Fingers by … 1183
VIN +5V
RF
+5V -5 V
RG
Schematic of the actual force sensor circuit for the force sensor. Actual circuit
diagram for the force sensor is as shown in the Fig. 7.
Calibration of force sensor. At the known values of weights, the values of voltages
are recorded and then a graph is plotted between the voltage and the force (in Newton).
Then a linear equation is obtained which represents the force at the fingertip as a
function of voltage coming from the force sensor circuit.
Fig. 8 Bond graph model for single joint actuation system for trajectory control
slider-nut, string-tube mechanism, active joint of the link and PID algorithm with
the motor driver.
The desired angular position of the link is tracked using a PID control algorithm as
shown in the PID WBGO in Figs. 8 and 9. The concept of Word Bond Graph Objects
(WBGO) is implemented to represent the components subsystems as Objects. PID
stands for proportional, integral and derivative control. It is a closed-loop feedback
control system. The input to the PID algorithm is a flow, in terms of joint angle error.
The information of actual joint velocity is taken using a signal bond from the link
angular velocity junction. The output of the PID algorithm is torque corresponding
to the design error. The information of torque is sent to the motor controller, and it
Fig. 9 Bond graph model for single joint actuation system for force control
Trajectory Control and Force Control of Biomimetic Fingers by … 1185
produces the required current. The motor controller is modelled as a source of flow
for the motor.
Bond graph model for single joint actuation system for force control. Modelling
of the force control is also necessary for gripping the object precisely or for applying
the power of the fingers. So, it plays a requisite role in grasping the finger. Figure 9
shows the modelling of a single joint actuation system for force control of the
fingertip.
Modelling of the actuation for the force control is the same up to the revolute joint
of the finger as that in trajectory control after then it needs some modification. As the
force is applied at the fingertip, so it is necessary to convert the torque at the finger
joint to the force at the tip of the finger using a transformer. Linear inertia is taken
into account. As the finger is assumed to apply a force on the fixed surface, then it
is modelled with a source of flow zero with a coupling representing compliance and
resistance of force sensor. Force signal is taken from the ‘0’ junction with a negative
sign and on computing error by comparing it with the desired force. The error on
differentiating routes on to the PID algorithm as a source of flow signal for computing
the required torque. The information of the torque is sent to the motor controller, and
it produces the required current, which is given to the motor as a source of a flow.
For controlling purpose, program codes are written in Arduino IDE and processing
software with reference to PID front end code by Brett Beauregard (2009). The
Arduino code runs the main code which includes the PID algorithm and implemen-
tation of a timer interrupts using TimerOne.h library. Processing software is used
to plot the graph of the processed variables from Arduino in real time. Finally, the
design of the graphic user interface and the serial communication functions/helpers
between Arduino and processing have been done [7].
The proper interpretation of the results for step input, sinusoidal input is being done.
Various system responses are perceived and some of them are presented for trajectory
control and force control of the biomimetic finger.
1186 V. Saini et al.
For step input. A tuned response for step input should be necessary to get the
best response. By changing the values of gains by trial and error method, the tuned
response is obtained as shown in Fig. 10.
For step input. As controlling of force is done by the force sensor, which is very
sensitive to small pressure changes at the fingertip. So, it is troublesome to control
the fluctuation generated during the force control in the system response. Figure 11
shows the tuned response for force control at sampling time is equal to 0.010 s.
4 Conclusion
In this work, the study of human fingers and their different motions have been anal-
ysed from the prosthesis point of view. With the help of this educational experi-
mental setup, the biomechanics of the finger actuation is clearly explained. Instead
Trajectory Control and Force Control of Biomimetic Fingers by … 1187
References
1. Fowler NK, Nicol AC, Condon B, Hadley D (2001) Method of determination of three dimen-
sional index finger moment arms and tendon lines of action using high resolution MRI scans. J
Biomech 34(6):791–797
2. Vaz A, Hirai S (2007) A bond graph approach to the analysis of prosthesis for a partially impaired
hand. J Dyn Syst Meas Contr 129(1):105
3. Vaz A, Hirai S (2003) Actuation of a thumb prosthesis using remaining natural fingers. In:
Proceedings of IEEE/RSJ international conference on intelligent robots and systems (IROS
2003), Las Vegas, USA, 1998–2003 Oct
4. Iberall T (1997) Human prehension and dexterous robot hands. Int J Robot. Res. 16
5. Kalra S, Vaz J, Mishra N (2013) Development of a test rig for the study of musculoskeletal
actuation of human finger. In: ACM International conference proceeding series
6. Armstrong TJ, Chaffin DB (1978) An investigation of the relationship between displacements
of the finger and wrist joints and the extrinsic finger flexor tendons. J Biomech 11(3):119–128
7. Libraries homepage. https://processing.org/reference/libraries/. Last accessed 18 June 2019
Design of a Two Degrees of Freedom
Actuator for Rehabilitation Robotic
Applications
Abstract Rehabilitation robots and many other bio-inspired applications deal with
multiple degrees of freedom joints. Generally, each degree of freedom of such a
joint is actuated by “an equal” number of single degree of freedom actuators which
occupy a lot of space surrounding the joint and is a major hurdle in the development
of orthotics and exoskeletons. This paper attempts to model a robotic system where
the number of actuators used is less than the degrees of freedom of the robot, while
maintaining full actuation. The model of a two degrees of freedom robotic system
using only a single actuator is considered. Pneumatic artificial muscles (PAMs) are
used as these actuators are flexible and safe, have high power-to-weight ratio and
have resemblance with human muscles. This model would be greatly beneficial in
the development of rehabilitation robots as less number of PAMs would be used for
actuating the degrees of freedom about a joint, while maintaining full actuation.
1 Introduction
One of the major challenges behind the development of any prosthetic arm or leg
is to incorporate in it as many degrees of freedom (like extension, flexion, adduc-
tion, abduction, pronation, supination, etc.) as is found in a human arm or leg. Also,
many joints in a human body like shoulder, elbow and wrist have multiple degrees of
freedom. One of the common ways of developing such a joint is to cascade as many
single degree of freedom actuators as is the number of degrees of freedom in the
joint. For example, Salazar et al. [1] have developed a human-inspired mechatronic
finger. They used four pairs of pneumatic muscles in antagonistic arrangement to
actuate the four degrees of freedom of the mechatronic finger. Therefore, the devel-
opment of a prosthetic hand or arm with 27 and 32 degrees of freedom, respectively,
would be a difficult task considering the space constraint around the joints in a hand.
Moreover, due to the excellent power-to-weight ratio and flexibility as compared to
the conventional actuators, the pneumatic muscles serve as soft actuators and hence
a natural choice as substitutes for human muscles in prosthetics, exoskeletons and
human–machine interactions as in haptics [2–5]. Due to these features, PAMs are
used not just as actuators but also for flapping of wings or flaps in aerospace appli-
cations [6]. A detailed information on PAMs and its different types can be found in
the work of Daerden and Lefeber [7]. In the literature, the researchers have mostly
used a pair of PAMs in antagonistic arrangement to actuate a single degree of free-
dom joint in rehabilitative devices like in the works of Tsagarakis and Cladwell [3]
and Ferris et al. [8]. Due to these difficulties, the present paper comes up with the
conceptual design of a two degrees of freedom joint actuation using only a single
PAM. The paper has six sections. Section 2 discusses a single degree of freedom
actuator with a pair of PAMs and a single PAM, Sect. 3 presents the model of the
two-degrees-of-freedom actuator using a single PAM, Sect. 4 presents the model of
the PAM in the proposed actuator, Sect. 5 presents the mathematical modelling of
the proposed actuator, and Sect. 6 concludes the paper.
(a) (b)
Fig. 1 A single degree of freedom actuator using a the antagonistic arrangement of a pair of PAMs
and b a single PAM
Design of a Two Degrees of Freedom Actuator for Rehabilitation … 1191
a hinge. A small solid block with a hole in it is attached to the casing of the hinged
joint. The solid block is fixed at its place and does not move. The thread at the upper
end of the muscle, through the hole, is attached to the link directly. When the muscle
contracts, the link moves from its initial position and makes an angular displacement
of θ . When the muscle relaxes, the link goes back to its initial position due to gravity.
In Fig. 1b, the PAM with the solid line indicates the contraction of the PAM and the
PAM with the dashed line indicates the same PAM under relaxation. Therefore, it
can be concluded that the model of Fig. 1b also works exactly similar to the model
of Fig. 1a. But the model of Fig. 1b has an edge over the model of Fig. 1a as the link
is executing the same bidirectional rotation with a single PAM.
It is to be noted that in the absence of the gravitational force, the link would only
execute a unidirectional motion. The reason being when the muscle relaxes, the wire
becomes loose and the link will not be able to go back to its initial position by its
own.
3 A Two-Degrees-of-Freedom Actuator
(a) (b)
Fig. 2 A two-degrees-of-freedom actuator using a the traditional model and b the proposed model
1192 Saurav Kumar Dutta et al.
In the last section, it is observed that by removing the pulley and attaching the
thread directly to the link helped in doing away with one out of the two PAMs.
Therefore, the proposed model of the two-degrees-of-freedom actuator would also
have no pulleys. Figure 2b shows the model of the two-degrees-of-freedom actuator
using one PAM. In this case, the link is attached to the platform by a spherical joint.
A solid block with two holes is attached to the casing of the spherical joint. The axes
of these two holes are perpendicular to each other, as can be seen in Fig. 2b. Further,
in this model, the PAM has one inlet for the compressed air and has two outlets,
which are sealed. Since the PAM has two outlets, two threads can be attached to the
PAM. The pink-coloured thread, through the hole, is attached at the top of the link.
The pink-coloured thread is responsible for the actuation of the link about e3 -axis.
The orange-coloured thread, through the other hole, is attached to the side of the
link. The orange-coloured thread is responsible for the actuation of the link about
e2 -axis. When the muscle contracts, the link moves in a three-dimensional space
with its actuation both about the e2 - and e3 -axes. When the muscle relaxes, the wires
become loose and the link is free to move in the three-dimensional space subjected
to any arbitrary external force. If the system is in vacuum, the link would continue to
be in its displaced position and will not return to its initial position by its own. The
muscle in Fig. 2b can also be actuated selectively such that when the pink-coloured
wire is pulled, the orange-coloured wire remains loose and vice versa. This ensures
that the link has independent motions both about e2 - and e3 -axes.
The traditional McKibben muscles work in a way that when the muscle contracts,
its length decreases and when the muscle relaxes, its length increases. There is little
deformation along the lateral directions during both contraction and relaxation of
the muscles. The models of the actuators shown in Figs. 1 and 2a use PAMs of the
McKibben type. Also, the McKibben-type PAMs can carry load only when they
contract. These PAMs are unable to carry load when they relax. Hence, these PAMs
are often used antagonistically.
The single degree of freedom actuator with a single McKibben PAM shown in
Fig. 1b can only have unidirectional motion in the absence of gravitational force.
However, if an extensor contractor pneumatic artificial muscle (ECPAM) [9] is
directly connected to the link, as shown in Fig. 5, the single degree of freedom actua-
tor with a single ECPAM would exhibit bidirectional motion and its motion would be
similar to that of the actuator shown in Fig. 1a. In Fig. 5, the ECPAM drawn by solid
line indicates the ECPAM under contraction, the same ECPAM drawn by dotted line
indicates the ECPAM under extension and the same ECPAM drawn by dashed line
indicates the ECPAM under relaxation.
In the usual McKibben-type PAMs, the tube or the bladder has one inlet and one
outlet for the flow of compressed air. The bladder gets inflated when the outlet is
Design of a Two Degrees of Freedom Actuator for Rehabilitation … 1193
sealed. The sealed outlet of the tube provides the attachment for joining the wires or
threads to the joint or the link, as shown in Fig. 3a.
In the PAM for the proposed two-degrees-of-freedom actuator, two bladders are
taken inside a braided mesh. As can be seen in Fig. 3b, the inlet of the two bladders
is same but their outlets are different. In other words, it can be said that the PAM has
one inlet and two attachments for connection of wires or threads coming out of the
two outlets of the two tubes. The PAM used in the actuator of Fig. 2b is basically the
PAM of Fig. 3b. When air gets filled into both the bladders of the PAM, it results in
actuation of the link about both e2 - and e3 -axes simultaneously. However, by using a
5/2-way directional control valve, as shown in Fig. 4, both the bladders of the PAM,
for the proposed two-degrees-of-freedom actuator, can be selectively filled which
would lead to the independent actuation of the link about e2 - and e3 -axes, as shown
in Fig. 5b.
(a)
(b)
Fig. 5 a A single degree of freedom actuator using a single extensor contractor PAM and b the
proposed two-degrees-of-freedom actuator using the proposed PAM and a directional control valve
where q is the vector of joint variables (α, β), q̇ and q̈ represent the joint velocities
and accelerations, respectively, and [M(q)], [h(q, q̇)] and [c(q)] represent the mass
matrix, coriolis vector and gravity vector, respectively. It may be noted that in Eq. 1,
the force vector, [(q)]2×1 , is also a function of the joint variables. This is because
the torque exerted at the joint is a cross product of the force exerted by the proposed
PAM and the distance between the PAM and the point on the link at which the wire is
attached. Also, the force–contraction characteristics of most of the PAMs generally
exhibit a nonlinear relationship [11]. Hence, empirical force–contraction relations
obtained from experimental data points are also used to characterize a PAM [12].
Design of a Two Degrees of Freedom Actuator for Rehabilitation … 1195
6 1
0
4
α (rad)
β (rad)
−1
2
−2
0 −3
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time (t) Time (t)
(a) (b)
6 Conclusion
In this paper, the model and working of a single degree of freedom and two-degrees-
of-freedom actuators using a single PAM are presented. The single degree of freedom
actuator uses an extensor contractor PAM, whereas the proposed two-degrees-of-
freedom actuator uses a PAM with two bladders inside a braided mesh. A simplified
mathematical model of the proposed two-degrees-of-freedom actuator is used to
obtain the time response plots of the two degrees of freedom. One of the limitations of
the proposed two-degrees-of-freedom actuator is that it can have only unidirectional
motion and not bidirectional motion. Future work will be to fabricate the PAM of
the proposed actuator and also to improve the mathematical model considering all
the nonlinearities relating to PAM.
1196 Saurav Kumar Dutta et al.
References
Vikash Kumar
Abstract This paper presents kinematic and dynamic analysis of a 3-R robot arm
with ADAMS/MATLAB co-simulation, and its control with PID and PID-based
fuzzy logic controller; finally, simulated results are compared with numerical method
in MATLAB, for kinematics/dynamic analysis, a three degree of freedom robot
arm with all revolute joint is designed in SolidWorks, designed model is exported
in ADAMS, and exported plant motion is controlled with the MATLAB/Simulink
software. ADAMS software is now widely applied to most of the areas such as
robotics, automobile, mechanism and so on for an automatic dynamic analysis of
the mechanical system. The virtual model plant created by ADAMS software is
approaching the real model dynamics analysis in MATLAB with numerical methods,
and it can provide a credible result for the real model simulation with ode45 numerical
method test and analysis.
1 Introduction
Robotics a field that is grooving day by day, and many researchers still trying to
develop an effective algorithm (position and force control algorithms) to develop a
robot that is cost effective and utilized in most of the field for the maximum task.
Nowadays, robots used in industries have a wide range of applications, mostly for
welding and painting robots in car plants [1], electronic board assembly in electronics
and computer fields, repairing nuclear installations in nuclear plant [2], etc. Even
though all these tasks can a robot do accurately and precisely and also having more
advantages, development of a robot is required a lot of time and amount that involves
for testing and validation of robot before installation is done for the final task which
it is made for. With the help of ADAMS /MATLAB simulation robot can be analysed
V. Kumar (B)
Department of Mechanical Engineering, Thapar Institute of Engineering and Technology, Patiala,
Punjab 147004, India
Fig. 1 SolidWorks designed three-links robot manipulator, and same designed model is exported
ADAMS
Table 2 Manipulator links mass and inertia parameters data taken from ADAMS
Links Mass I xx (kg mm2 ) I yy (kg mm2 ) I zz (kg mm2 )
Link 1 7.173866 kg 1.3414708741E+05 1.34149E+05 2231.75143
Link 2 3.754276 kg 5.1823273E+04 5.145724E+04 822.1331641
Link 3 1.063915 kg 6020.2525946 5985.052737 98.2997074
⎡ ⎤
cθi −sθi 0 ai−1
⎢ sθi · cαi−1 cθi · cαi−1 −sαi−1 −sαi−1 · di ⎥
⎢ ⎥
1 T =⎣
i−1
(1)
sθi · sαi−1 cθi · sαi−1 cαi−1 cai−1 · di ⎦
0 0 0 1
After using the D-H parameter table, we get a transformation matrix for each
frame, like putting first frame D-H parameters in the transformation table, we get
the transformation of frame 1 to frame 0. It is represented as 0 T1 and given in the
following way
⎡ ⎤
cθ1 −sθ1 0 0
⎢ sθ1 cθ1 0 0 ⎥
0
T1 = ⎢
⎣ 0 0 1 L1 ⎦
⎥ (2)
0 0 0 1
Similarly, we get the transformation matrix for each frame, from frame 4 to frame
0 and multiply all transformation matrices to get the relation between third link end
point P (X 4 , Y 4 , Z 4 ) and base frame B0 and get 0 T4 transformation matrix as given
below
0
T4 = 0 T1 1 T2 2 T3 3 T4 (3)
The last column of the 0 T4 matrix gives the relation between the X, Y and Z
coordinate of the end point P with respect to base frame B0 , which are
⎡ ⎤ ⎡ ⎤
px cos θ1 [L 2 cos θ1 + L 3 cos(θ2 + θ3 )]
P(X 4 ,Y4 ,Z 4 ) = 0 P4 = ⎣ p y ⎦ = ⎣ sin θ1 [L 2 cos θ1 + L 3 cos(θ2 + θ3 )] ⎦ (4)
pz L 1 + L 2 sin θ1 + L 3 sin(θ2 + θ3 )
Robot kinematics, dynamics and control analysis, with the help of ADAMS and
MATLAB, make easy and accurate, so it is the best tool for applying control command
to examine system behaviour. With co-simulation of MATLAB/ADAMS that makes
a combined program, it is a more efficient package to simulate and analyses a system
like robot manipulator. In order to make co-simulation workable, a common format of
input and output is needed that helps to make changes in one program automatically
which is made by user in other programs. To recognize MATLAB exported data by
ADAMS, ADAMS requires the activation of its control plug-in manager that exports
ADAMS control plant model to MATLAB/Simulink software.
Kinematics/Dynamics Analysis with ADAMS/MATLAB Co-simulation … 1201
Analysis of results in Matlab model Simulate the model Build the control system in Matlab Simulation
using block diagram
Dynamics of Manipulator
For analysing robot arm dynamics, here, we used transformation matrix 0 T 4, which is
fourth column as given in Eq. (4), and these X, Y and Z coordinates give a relationship
between endpoint and base of the manipulator and set dynamic parameters (Fig. 2),
where the base coordinate system coincides with one end of the link 1. Basic dynamic
parameters of links are defined as follows: m1, m2 and m3 are, respectively, masses
of the links 1, 2 and 3; I 1 , I 2 and I 3 are, respectively, rotational inertia around Z-axis;
θ 1 , θ 2 and θ 3 are, respectively, angle variable; L 1 , L 2 and L 3 are, respectively, lengths
1202 V. Kumar
of the three links, and L c1 , L c2 and L c3 are, respectively, lengths between centre of
mass of the links and joint axes.
Euler–Lagrange method: Total kinetic energy of robot manipulator is given as:
1
TK = I1 + I2 + I3 + m 2 L 22 cos θ2 + m 3 {L 22 cos θ2 + L 23 cos(θ2 + θ3 )
2
1
+2L 2 L 3 cos θ2 cos(θ2 + θ3 )}]θ̇12 + [I2 + I3 + m 2 L 22 ]θ̇22
2
1 1
+ [I3 + m 3 L 2 ]θ̇3 + m 3 [L 3 (θ̇2 + θ̇3 )2 + 2L 2 L 3 θ̇2 (θ̇2 + θ̇3 ) cos θ3 ]
2 2 2
(5)
2 2
Total potential energy is given as:
L = TK − U P (7)
And with above Lagrange, we get the system dynamic equation as:
∂L ∂L
d
dt ∂ θ̇i
− ∂θi
= τi (8)
With the use of Eq. (8), we can calculate each joint torque as
m 11 m 12 m 13
M = m 21 m 22 m 23 (10)
m 31 m 32 m 33
where
m 11 = I1 + I2 + I3 + m 2 lc2
2
cos2 θ2 + m 3 (l2 cos θ2 + lc3 cos(θ2 + θ2 )]2 (11)
m 33 = I3 + m 3lc3
2
(12)
Similarly, centrifugal force, Coriolis force and gravity force and with obtained
torque equations we can analysis robot manipulator system numerically with different
numerical method.
Control System
In robotics, control of a robot manipulator is considered as a spine of the system; with
the help of controller, a robot can do the assigned task automatically and precisely.
So, the controller is very important part in today’s robotics; a controller is nothing
but it is a form of equation and written is in form of algorithms that can be understood
by the computer programs.
The objective of controller implementation in robotics is that the end point of
a robot called end effector follows the desired path and orientation accurately and
complete the assigned task automatic with higher precision [9]. The motion control
of a manipulator either serial or parallel manipulator is done with wide control
scheme like feed-forward [2], optimal feedback linearization [10], PID, adaptive
backstepping tracking, PID with sliding mode control [11], inverse dynamics control
[9, 10, 12], kinematics and dynamics-based PID controllers [13] sliding mode control
[5], output integral sliding mode controllers [12] and smooth integral sliding mode
controllers [14].
Most of these control schemes require mathematical models of the system to
implement a control law, so these controllers need the kinematic and dynamic analysis
of the model before implementing controller; however, in ADAMS/MATLAB co-
simulation, ADAMS takes care of kinematic and dynamic part and MATLAB sees
control part of the system. In below-given two-section, PID and PID-based fuzzy
controller implementation on manipulator is discussed.
PID Controller
Today most of the automatic system used a combination of proportional, integral and
derivative (PID) control. This combination is widely applied in the automatic system
1204 V. Kumar
de(t)
u = K P e(t) + K D + KI e(t)d(t) (13)
dt
If we ignore dynamic properties of joint driver, then according to PID control law,
the driven torque becomes
τ = K P (θdesired − θ ) + K V θ̇desired − θ̇ + K I (θdesired − θ ) (14)
θ desired and θ̇desired are denote desired signal of joint angle and angular joint velocity
respectively. K P , K V and K I = Respectively related gain coefficients.
In the above equation, K P , K D and K I are gain of the error signal in the control loop,
and their value remains fix after assigning according to desired response pattern of
the output results of the system. These gains modified the main system equation, and
according to that damping coefficient and frequency of the system are changed. For
any system to see its performance, its rise time, settling time, percentage overshoot
and steady-state error for the input signal are used. Each gain (K P , K D and K I ) value
has different effect on these performance parameters. For getting desired performance
of the system, a very fine-tuning of the gain parameter is required, and it is an
iterative process and done in a way like if we want to make system response fast,
we increase proportional element, and if we want to make system slow, we increase
damp that reduces system overshoot, and with increment of differential element,
we can analyse each element effect on system behaviour. With this process, we get
PID position control. MATLAB also has its inbuilt PID controller block in MATLAB
control toolbox (MCT); however, for this paper, we have used our own calculated gain
value that is computed by analysing transfer function output on different gain values
with step input signal. In this co-simulation work, a simple PID control system was
implemented for each individual angle outputs. Implementation of PID controller is
shown in Fig. 4, and its tuned gain parameter and corresponding gains output results
are discussed in results section.
robustness of PID controller, and its nonlinear effect of the arm is handle by fuzzy
logic controller. It improves the system performance in steady and transient state, in
this paper, a PID controller is tuned with fuzzy logic controller, and for this work,
MATLAB control toolbox fuzzy controller is used.
The fuzzy system basic configuration is shown in Fig. 5, and it has four
components [11, 16, 18–22]:
1. Fuzzifier: Maps crisp points into fuzzy sets.
2. Fuzzy inference engine: Mapping between the fuzzy sets in the input space
U ⊂ Rn to the fuzzy sets in the output space V ⊂ Rm .
3. Fuzzy rule-based system: It is the heart of the whole system in the sense which
comprises fuzzy rules describing how the fuzzy system performs.
4. Defuzzifier: Maps fuzzy sets of output space into crisp points.
There are various forms for fuzzy controller P, PD, PI, controller and PD-type
fuzzy controller. In the present work, a PID-type fuzzy controller is applied. In the
fuzzy PID controller, the control variables are the error (e) and the change of the
error (e).
The rule base is given as:
If error (e) is negative large (NL) and derivative of error (e) is positive large
(PL), then u is zero.
If error (e) is zero and derivative of error (e) is negative large (NL), then u is
negative large.
If error is zero (ZE) and derivative of error (e) is negative small (NS), then u is
negative small.
If error is zero and derivative of error (e) is zero, then u is zero.
If error is zero and derivative of error (e) is positive small (PS), then u is positive
small.
If error is zero and derivative of error (e) is positive large, then u is positive
large.
If error is positive large and derivative of error (e) is negative large, then u is
zero.
If error is positive large and derivative of error (e) is zero, then u is positive
large.
1206 V. Kumar
In the present work, the product inference rule and based on the above rule, the
membership function used is given in Fig. 6 and Table 4.
Membership functions for error input and output gains.
Implementation of PID-based fuzzy controller is represented by block diagram
as shown in Fig. 7, in this gain of the controller signal is decided by fuzzy controller,
it’s not remain fixed as like PID controller. In this gain is varying according to some
defined limit and controller output signal vary according to gains. Selected controller
gains and its output results is discussed in results section.
Fig. 6 Membership functions for inputs (e) and (e). and control signal memberships function for
outputs (Kp, Ki and Kd)
To run the simulation, ADAMS exported plant name assigned during plant export
time. In MATLAB, command window is called by entering “Adams sys” command,
and a control plant is created that contains dynamic model information. For simula-
tion, MATLAB Simulink tool is used in which a complete simulation control model is
created, and output results are analysed with different gains value. Gains are selected
as shown in Table 5. After some trail end error method for PID controller and for
fuzzy-based controller, limits of the gains are extended that vary according to the
input error signal, and output results are shown in Fig. 8. It shows joint angle in
degree versus time graph for each joint with PID and PID-based fuzzy controller,
and its output is compared with the desired signal.
The end point of the link three taken as an end effector and its coordinate X, Y
and Z variation with input signal is shown below in Fig. 9.
Fig. 8 Joint 1 and joint 2 angle variation with time for square and sine were taken as an input
1208 V. Kumar
4 Conclusions
References
1. Park KT, Shin YJ, Park CH, Mo YS, Jeong DC (2008)Robot application for the assembly
process of an engine part. ICCAS
2. Ray DD, Singh M (2010) Development of a force reflecting tele-robot for remote handling in
nuclear installations. CARPI
3. Craig JJ (2003) Introduction to robotics: mechanics and control, 3rd edn. Prentice Hall, United
Kingdom
4. Sosa-Méndez D, Lugo-González E, Arias-Montiel M, García-García RA (2017) ADAMS-
MATLAB co-simulation for kinematics, dynamics, and control of the Stewart–Gough platform.
Int J Adv Robot Syst 1–10
Kinematics/Dynamics Analysis with ADAMS/MATLAB Co-simulation … 1209
5. Kisir S, Bingul Z (2012) Position control and trajectory tracking of the Stewart platform.
In: Kucuk S (ed) Serial and parallel robot manipulators—kinematics, dynamics, control and
optimization. InTech, Croatia, Rijeka, pp 179–202
6. Zhu DL, Qin JY, Zhang Y (2010) Research on co-simulation using ADAMS and MATLAB
for active vibration isolation system. IEEE ICICTA 2:1126–1129
7. Affi Z, Romdhane L (2006) ADAMS/SlMULINK interface for dynamic modeling and control
of closed-loop mechanisms. In: Proceedings of the 7th WSEAS international conference on
automatic control, modelling and simulation, Prague, Czech Republic, pp 353–356
8. Naing ZZ, AI-Mamun MP (2009) Integrated ADAMS+MATLAB environment for the design
of an autonomous single wheel robot. In: Industrial electronics, IECON 35th annual conference
of IEEE 2009, pp 2253–2258
9. Taghirad HD (2013) Parallel robots: mechanics and control, 1st edn. CRC Press, Boca Raton,
FL, USA
10. Tourajizadeh H, Yousefzadeh M, Tajik A (2016) Closed-loop optimal control of a Stewart
platform using an optimal feedback linearization method. Int J Adv Robot Syst 13(144):1–9
11. Mathworks. https://in.mathworks.com/help/fuzzy/what-is-sugeno-type-fuzzy-inference.html
12. Fraguela SH, Fridman L, Alexandrov VV (2012) Output integral sliding mode control to
stabilize the position of a Stewart platform. J Franklin Inst 349:1526–1542
13. Ding B, Cazzolato BS, Stanley RM et al (2014) Stiffness analysis and control of a Stewart
platform-based manipulator with decoupled sensor-actuator locations for ultrahigh accuracy
positioning under large external loads. J Dyn Sys Meas Control 136(6):1–12
14. Kumar P, Chalamga A, Bandyopadhyay B (2015) Smooth integral sliding mode controller for
the position control of Stewart platform. ISA Trans 58:543–551
15. Huang SJ, Lee JS (2000) A stable self-organizing fuzzy controller for robotic motion control.
IEEE Trans Ind Electron 47(2):421–428
16. Yoo BK, Ham WC (2000) Adaptive control of robot manipulator using fuzzy compensator.
IEEE Trans Fuzzy Syst. 8(2):186–199
17. Lewis FL, Yesildirek A, Liu K (1996) Multilayer neural-net robot controller guaranteed tracking
performance. IEEE Trans Neural Networks 7(2):388–398
18. Li W (1998) Design of a hybrid fuzzy logic proportional plus conventional integral derivative
controller. IEEE Trans Fuzzy Syst 6(4):449–463
19. Li W et al (2001) Tracking control of a manipulator under uncertainty by fuzzy P + ID
controller. Fuzzy Sets Syst 122:125–137
20. Chen G, Joo YH (2001) Introduction to fuzzy control systems. Department of Electrical and
Computer Engineering, University of Houston, Houston, Texas 77204-4793, USA
21. Fuzzy logic controllers and methodology, advantages and drawbacks. Antonio Sala, Pedro
Albertos and Manuel Olivares
22. Mathworks. https://in.mathworks.com/help/fuzzy/types-of-fuzzy-inference-systems.html
Joining Aluminum Open Cell Sponge
by Friction Stir Welding
Abstract Metallic sponges delivered from aluminum 6XXX series are a standout
among the most as of late created ultra-lightweight cellular metals. Metallic sponge
features exceptionally high-energy ingestion limits with low density and are conse-
quently utilized for boundless applications in the production of ultra-lightweight
cellular parts. The manufacture of cellular component parts, for example, high-
temperature air or liquid filters and implants, is potential requirement zones relying
upon the cell structure size and type of the metal sponge. Metal sponge of closed cell
type is ordinarily appropriate for structural uses though open cell sponge; in general,
it will be favored for utilitarian applications. Advancement of satisfactory joining
innovations for these materials is a basic advance for their far-reaching modern
use. The present paper portrays a strong solid-state welding technique that is fit for
giving sound joints between aluminum 6XXX macro-porous sponge arrangements.
The joining is accomplished by friction stir welding technique abbreviated as FSW
through heat generation, softening of material and plastic deformation following the
going of non-consumable rotating high-speed steel [HSS] cylindrical tool through
the gap between the two solid edges of the sponge to be joined. Having optimized the
welding conditions and without utilizing filler metal strong edges were created. The
subsequent welds had comparable mechanical properties to that of solid FSW-welded
aluminum 6XXX series.
1 Introduction
Lightweight structural and auxiliary parts have been utilized progressively to lessen
the heaviness of different vehicles, railcars, defense vehicles, marine cruisers, and
airplane. Amid these cellular segments, multiple layer composite, high vitality
retention limits, great damping impacts, and high strength-to-weight proportions
are exceptionally compelling for mobility businesses and models. Metal sponge is
classified as innovative metals for assembling multiple layers composite, preferred
in lightweight boards, basic platform structures, and energy absorption [1]. Metal
sponge applications are explored to manufacture vitality safeguards, heat exchanger,
sound and magnetic shield, high-temperature filters, fuel cells, and restorative
medical implants. Metal sponge as of now is not generally utilized on a modern indus-
trial sector due to different aspects [2]. An expansion of information in their mechan-
ical properties and advancement in cellular welding methods will urge researchers
for investigation of these metals for modernized application along with the poten-
tial outcomes for their business exploitation. Open cell metallic sponge casted from
aluminum termed as ‘aluminum sponge’ in the present paper can be viewed as one of
the late created metal sponges equipped for use in load-bearing basic applications. To
explore the maximum capacity of sponge metals, a need is to create techniques to weld
sponge to sponge or solid metals. Recent investigations suggesting novel methods
were explored for manufacturing metal sponge, prevalently aluminum sponge [1–3].
Some attempts have been made to the joining of these aluminum sponges just
as aluminum sponge in vehicle applications as sandwich part; however, past experi-
mentation on joining or welding of open cell macro-porous sponge has been found
in the literature [4, 5]. Two late articles have investigated the weld ability and
mechanical properties of aluminum sandwich foams of closed cell type [5]. Due
to the cellular structure of the sponge, the utilization of customary welding proce-
dures—for example, tungsten inert gas (TIG) welding—leads to crumbling of the
cellular structure. Few encouraging outcomes were obtained during laser welding
of aluminum sponge [6]. Aluminum precursor bonding was adapted for welding
aluminum sponge as the temperature needed for bonding is around 500–580 °C of
the melting temperature of the metals being welded.
The FSW of aluminum sponge would be practical in light of the fact that the
bonding pressure required building up joints having sound qualities is well over the
Table 1 Data of density, porosity, and relative density of casted aluminum sponge
Sponge specimen Sponge density (kg/m3 )-ρ f Porosity (%) Relative density-ρ
Specimen 1 1729 35.887 0.64
Specimen 2 1492 43.698 0.55
Specimen 3 1227 53.698 0.45
Specimen 4 1117 57.894 0.41
Specimen 5 1300 50.943 0.48
Joining Aluminum Open Cell Sponge by Friction Stir Welding 1213
Fig. 1 Schematic representation of FSW. a Aluminum. b Aluminum sponge with solid edge
compressive strength of aluminum sponge during joining temperature. In this way, all
practical techniques to join aluminum sponge should ideally encounter two primary
requirements. To start with, the temperature for bonding ought to be underneath the
melting point of aluminum sponge to counteract basic crumbling of cellular structure
to be welded. Secondly, process of bonding ought to be related with development
of semi-solid phase at weld interface, to remunerate the need to high pressure appli-
cation and strength to the sponge amid the welding procedure. These conditions
are met by friction stir welding [FSW] as indictaed in Fig. 1a; a capable welding
method for aluminum sponge. This procedure is likewise equipped for delivering
joints having great quality. The present paper investigates a FSW method with solid
edge retention as shown in Fig. 1b, to deliver high-quality similar/dissimilar joints
of aluminum 6XXX series foam.
2 Experimental Procedure
The aluminum metal selected for the current work is commercial aluminum 6XXX
series. The solid aluminum 6XXX series ingots were melted in electric induction
furnace around 750–800 °C in graphite crucible. The molten aluminum is poured
into metal mold containing porous die [7]. Sponge casting can be made from sand
or metal mold. The gating elements such as runner and riser are separated by using
band saw. Solidified sponge structure is subjected to air blast to remove the sand
residue.
Aluminum casted sponge billet was later machined using vertical milling
machining as FSW needs the welding edges to be straight and has certain dimension
constraints for clamping the workpiece in welding fixtures. The dimensions of the
sample were 160 × 110 × 8 mm. The solid edges of the casted sponge are the faying
1214 A. Chandru and S. V. Satish
surfaces and are cleaned by emery paper to prepare the surfaces for welding. The
thickness of the edge is 8 mm as the pin length of the tool is 5.7 mm. The specimens
are placed in such manner such that solid edges are adjacent. Aluminum sponge
which has been casted and machined as shown in Fig. 2, was joined together using
FSW, a solid-state metal welding process that utilizes the heat generated due to fric-
tion between the non-consumable rotating tool and the workpieces along with the
compressive force due to the tool to join the components together. To retain the posi-
tion of the tool at or below the metal surface, a vertical downward force is needed. To
prevent breakdown associated machinery, minimize excessive wear and tear on the
tool, and prevent tool fracture, the FSW cycle should be regulated so that the forces
on the tool are reduced and sudden modifications are avoided. As observed, the load
acting during FSW ranged between 0.3 and 0.5 tons. The design of the tool and the
pin profile is major element as a sound tool improvises weld quality and enables to
operate at the maximum possible welding speed. It is desirable that the material used
in tool to be sufficiently tough, strong, and hard wearing, at the joining temperature.
Hence, high-speed steel with 10% cobalt having a shoulder length of 20 mm and pin
length of 5.7 mm was selected.
The weld quality depends majorly on the tool rotational speed and feed rate of the
tool. The optimum tool rotational speeds for welding of aluminum foams are 700,
800, and 900 rpm. And as per literature review, the feed rate was fixed at 30 mm/min.
The mechanical properties and microstructure will fluctuate a lot depending upon
these factors. Via optical microscopy, microstructures were examined and were used
to explore the composition at the weld interface. Mechanical tests were conducted
on few welded samples to investigate the weld strength.
Joining Aluminum Open Cell Sponge by Friction Stir Welding 1215
Initial attempts targeted at welding the aluminum sponge, by FSW, proved ineffec-
tive. Due to porosity content at the higher end and compressive load acting on the
foam structure leads to collapsibility of weld foam region. Therefore, the solid edge
retention on casted sponge was considered toward the use of welding techniques,
like FSW, to eradicate or minimize any cellular structure damage to the sponge being
welded. Figure 3a shows the surface of FSW welded aluminum foam at 900 rpm,
in which a smooth bond is obtained at the faying surfaces without any structural
damage. The weld joint can be obtained without the aid of filler metal and free from
defects such as no kissing bonds, cavity, or groove-like defects as observed from the
X-ray image shown in Fig. 3b.
Porosity refers to the voids, i.e., the empty spaces on or within what should be
a solid metal casting. It is a defect, but in case of metal sponge, it is an impor-
tant characteristic and relative density (ρ) which is ratio of foam density to solid
density (ρ f /ρ s ) which is deciding factor for exploring macroscopic properties like
energy absorption, stiffness, thermal, electrical, and acoustic properties. Relative
density influences the mechanical properties much stronger and complex compared
to microstructure features.
Figure 4 shows typical images of the microstructure of aluminum 6XXX sponge
metal. Figure 4a represents parent metal section featuring fine but long needs of
unmodified eutectic silicon in an aluminum alloy matrix. Angular particles of primary
silicon are also seen. Figure 4b shows thermomechanical affected zone (TMAZ)
features’ dendritic structure. Fine eutectic silicon particles are seen at the inter-
dendritic regions. A few undissolved primary silicon particles are also seen. Figure 4c
Fig. 3 a Macro-structure of aluminum sponge welded by FSW. b X-ray image of aluminum sponge
welded by FSW
1216 A. Chandru and S. V. Satish
welded section shows metal flow. Fine eutectic silicon particles in the matrix of
aluminum solid solution and fine particles of eutectic silicon are also seen.
Compressive tests of uniaxial type according to ASTM E9 standards were
conducted on the aluminum sponge with different porosity and relative densities
at similar strain rate as indicated in Table 1. The compressive loads were applied
along directions parallel to the direction of pouring of molten metal.
Figure 5 shows the stress–strain curves of the aluminum sponge with varying
levels of porosity and relative density. The aluminum sponge reveals stress–strain
behavior compared with other metal sponge [8]. The stress–strain curve for metal
sponge is characterized by linear increase of stress with strain at lower stress inputs
(elastic deformation), succeeded with a long deformation plateau as the cell walls
buckle and collapse, and finally cell walls get interconnected to each other resulting
in flow stress increase causing a densification regime.
Tensile tests were performed on the samples at room temperature on universal
testing machine based on ASTM E8M-15a. Tensile samples were cut according
to standards using electric discharge machining (EDM). Tensile tests showed that
all welded tensile specimens were fractured near the welded region. However, the
tensile specimens with low relative density were fractured at the base metal due to
the presence of voids. Figure 6 shows stress–strain variation of friction stir welded
samples at different tool rotation speeds. The results are equivalent to the strength of
friction stir welded solid aluminum metal [9] because of the solid edge which results
in mechanical properties of the weld to be unaffected by sponge porosity.
4 Conclusions
The use FSW method for joining aluminum metal sponge with retention of solid edge
to commercial aluminum 6XXX series proved effective. By FSW approach, mechan-
ical strength of the aluminum sponge joints which are equivalent to that of the solid
aluminum were reliably produced, without resulting in any cellular integrity damage
or modification to the sponge during the welding. It is concluded that weld joints of
the sponge with the solid edge have enhanced mechanical properties, due to absence
of any precursor metal at the faying surface. Hence, FSW process may be consid-
ered for welding aluminum similar/dissimilar sponge. The novel method of solid
edge retention by gravity casting technique for sponge manufacturing and welding
of sponge by FSW technique proved very successful and defect free. Furthermore,
the weld ability of various metals/alloys sponges could be explored by FSW.
Acknowledgements The technical support extended to this research work from PESIT Banga-
lore South Campus affiliated to Visvesvaraya Technological University, Belagavi, is gratefully
acknowledged by authors.
1218 A. Chandru and S. V. Satish
References
1. Peng P et al (2018) High-performance aluminium foam sandwich prepared through friction stir
welding. Mater Lett 236
2. Banhart J (2000) Manufacturing routes for metallic foams. JOM 52(12):22–27
3. Lefebvre L-P et al (2008) Porous metals and metallic foams: current status and recent
developments. Special issue on metal foams, 2000, September, 2008, pp 775–787
4. Pogibenko AG et al (2015) The weldability of aluminium based foam materials. Weld Int 15
5. Kramer I et al (2006) Friction stir welding of foamable materials & foam core sandwiches. In:
Conference on materials, processes, friction and wear MATRIB
6. Haferkamp H et al (2004) Laser based welding of cellular aluminium. Sci Technol Weld Joining
9(1)
7. Chandru A et al (2019) Casting of macro porous aluminum sponge with controlled cellular
structure. Mater Today Proc 19:715–720
8. Koza E et al (2003) Compressive strength of aluminium foams. Mater Lett 58:132–135
9. Chandru A et al (2017) Evaluation of joint properties of aluminium alloy after friction stir
welding. Int J Sci Res 88–93
Harmonic Response Analysis
of Photovoltaic Module Using Finite
Element Method
1 Introduction
To match with the exponential growth of energy demand in today’s world, renewable
energy sources are becoming a popular choice. Photovoltaic systems are one of the
most famous renewable energy sources. They can be installed where solar radiation
are available. Earlier, they were installed in static conditions such as on roof tops, on
land, etc. But nowadays, PV modules are installed above cars, trains, metros, metro
stations, ships, etc. where they are subjected to dynamic conditions. These dynamic
Fig. 1 Schematic PV module in the air flow (black) with negative pressure (Blue) and positive
pressure (red) and flow separation
conditions occur generally due to base excitation and wind loads [1, 2]. Vibration
gets induced into such modules and affect its performance as well as life.
PV modules are affected by its surrounding environment which may affect its
effectiveness and will indirectly lead to decrease in productivity. From fluid engi-
neering point of view, the PV panel acts as a barrier to wind flow when it is in inclined
position; hence, wind load is an important aspect of analysis. As illustrated in Fig. 1
[3] due to flow of wind, low pressure and high pressure regions are formed, below and
above the surface of PV module, respectively. Oscillations occur due to the variation
of pressure on both sides of PV module, hence inducing vibration [3–5].
Most important factors while analysing operational efficiency and life cycle
of photovoltaic cell or a complete system are wind and temperature fluctuation
[6, 7]. Karl-Anders Weiss et al. [8] studied correlation between PV module deflection
corresponding to wind velocity acting on its surface. Outdoor experimental measure-
ments show the relation between deflection and wind speed and hence the dynamic
behaviour of PV module. Correlation between pressure p acting on module surface
and deflection y at the centre of surface is shown in Eq. 1. To analyse dynamic
loading, indoor facility can be set up in which this model will be very helpful.
Rain, snow, winds, etc. are some of the external mechanical loads to which PV
modules are exposed. Strong dynamic conditions occur when rain or wind loads
induce vibrations or otherwise only static conditions are created by steady wind and
deposited snow. Connecting wires and PV cells are more likely to fail under dynamic
loading conditions as fatigue failure. Cracks and micro-cracks may be formed due
to induced vibration decreasing performance of photovoltaic module [9–11]. Due to
Harmonic Response Analysis of Photovoltaic Module … 1221
found were 17.47, 35.15, 39.18, 51.03, 53.56 and 63.32 Hz. This validated FEM
model was used for determination of harmonic response of PV module to externally
induced forced vibration.
3 Results
Taking into consideration the weight of PV module and wind speed, load acting on
the module was calculated. Four responses are found: two angular position of PV
module at 21◦ and 90◦ , and two load condition considering pure wind load and base
excitation of same magnitude as wind load.
Figure 4 shows finite element model of wind load acting on PV module when
module was kept at 21◦ angle from horizontal. Figure 5 shows the harmonic response
spectrum of PV module when purely wind load is acting and module is inclined at
21◦ with respect to horizontal.
Figure 6 shows the case of PV module for 90◦ inclination with horizontal, which
makes the incident wind velocity direction perpendicular to module front surface.
Maximum amplitudes are observed when excitation frequencies match with various
natural frequencies of Pv module.
Figure 7 shows the harmonic response spectrum of PV module when base exci-
tation loads are acting and module is inclined at 21◦ with respect to horizontal.
Figure 8 shows the case of PV module for 90◦ inclination with horizontal. Maximum
amplitudes are observed when excitation frequencies match with various natural
frequencies of PV module in case of base excitation loads also.
Table 3 shows peak amplitudes and corresponding natural frequencies for different
cases of inclination angles and load type as shown in above harmonic responses
(Figs. 5, 6, 7 and 8).
Harmonic Response Analysis of Photovoltaic Module … 1225
Table 3 Peak amplitude and its corresponding frequency for different cases of response spectrum
Case Frequency (Hz) Amplitude (mm)
WL (21◦ ) 63.32 3.49e−5
WL (90◦ ) 17.47 8.30e−6
BE (21◦ ) 62.78 4.49e−6
BE (90◦ ) 17.16 2.91e−9
Wl wind load, BE base excitation)
4 Conclusions
But, when the inclination was decreased to 21◦ with horizontal, it was observed
that peak amplitudes occurred at sixth mode, that is highest natural frequency of PV
module.
From above observation, we can conclude that as we go on increasing the inclina-
tion angle with horizontal in dynamic conditions created by wind load and base exci-
tation, peak deflection occurrence moves from highest to lowest natural frequency
modes.
This harmonic response analysis enables us to quantify nature of vibration
response for different dynamic conditions. With the help of dynamic response
results, one can design suitable structures and mounting conditions which can prevent
unwanted effects of externally induced vibration.
Further study can be performed for different wind speeds at different inclination
and with different support conditions.
References
1. Danyel R, Mischa B (2004) Policy differences in the promotion of renewable energies in the
eu member states. Energy Policy 32(7):843–849
2. Parida V, Iniyan S, Goic R (2011) A review of solar photovoltaic technologies. Renew Sustain
Energy Rev 15(3):1625–1636
3. Artūras K, Audrius Č, Kristina K (2016) Analysis of external dynamic loads influence to
photovoltaic module structural performance. Eng Fail Anal 66:445–454
4. Menoufi K, Chemisana D, Rosell JI (2013) Life cycle assessment of a building integrated
concentrated photovoltaic scheme. Appl Energy 111:505–514
5. Sanghoon Y, Sehyun T, Jinsoo K, Yongseok J, Kisuk K, Jiyoung P (2011) Application of
transparent dye-sensitized solar cells to building integrated photovoltaic systems. Build Environ
46(10):1899–1904
6. Skoplaki E, Palyvos JA (2009) On the temperature dependence of photovoltaic module elec-
trical performance: a review of efficiency/power correlations. Solar Energy 83(5):614–624
7. Dirk G, Emmanuel VK (1999) Aeolian dust deposition on photovoltaic solar cells: the effects
of wind velocity and airborne dust concentration on cell performance. Solar Energy 66(4):277–
289
8. Weiss K-A, Assmus M, Jack S, Koehl M (2009) Measurement and simulation of dynamic
mechanical loads on pv-modules. In: Reliability of photovoltaic cells, modules, components,
and systems II, vol 7412. International Society for Optics and Photonics, pp 741203
9. Yih-Chih C, Jian-Zong L, Yu-Teng L (2011) Micro crack detection of multi-crystalline silicon
solar wafer using machine vision techniques. Sens Rev 31(2):154–165
10. Köntges M, Kunze I, Kajari-Schröder S, Breitenmoser X, Bjørneklett B (2011) The risk of
power loss in crystalline silicon based photovoltaic modules due to micro-cracks. Solar Energy
Mater Solar Cells 95(4):1131–1137
11. Dallas W, Polupan O, Ostapenko S (2007) Resonance ultrasonic vibrations for crack detection
in photovoltaic silicon wafers. Meas Sci Technol 18(3):852
12. Jørgensen M, Norrman K, Krebs FC (2008) Stability/degradation of polymer solar cells. Solar
Energy Mater Solar Cells 92(7):686–714
Development of a Micro-forming System
for Micro-extrusion Process of Micro-pin
in AZ80 Alloy
1 Introduction
becomes important in future technology because these materials have higher strength
and wear resistance than the polymers that are being used most widely.
However, micro-forming technology is still at laboratory research status, and few
industrial micro-forming process chains have been realized due to the size effects
in micro-forming, when the component feature sizes are reduced to a few hundreds
or tens of microns. Machines, forming tools and handling of micro-parts are crit-
ical elements which significantly determine industrial applications of micro-forming.
The development of micro-parts by micro-forming, however, cannot totally be based
on the traditional macro-forming knowledge. In micro-forming, the material defor-
mation behavior is characterized by a few grains in the deformation zone. Different
properties of grains make the deformation behavior inhomogeneous and difficult to
predict. Further, the size effect-related deformation phenomena further affect the
performance and product quality in terms of defect formation, dimensional accuracy
and surface finish of the micro-formed parts. In this work, a novel micro-forming
system (Fig. 1) is developed and equipped with a tool. Micro-extrusion process is
investigated with the micro-forming system, and the optimum forming parameters
are obtained. Based on the results, micro-pins are successfully fabricated with high-
dimensional accuracy. This work also aims at discussing the size effect-related defor-
mation behaviors which will help to understand the mechanisms and fundamentals
of the size effects in micro-forming process.
3 Summary
References
Abstract This proposal is for carrying out topological analysis of epicyclic gear
trains (EGTs) starting from the graph theory based model. With respect to a given
EGT (graph), the topological analysis is a phase of conceptual design where the
designer determines all the possible combinations of input, output, and fixed links
in an EGT. In this phase, symmetry analysis of an EGT is vital, as it becomes easy
to discard kinematically equivalent input-output-fixed link combinations. Further-
more, it is essential to evaluate those input-output combinations causing one or more
redundant links. Addressing these twin objectives, graph theory based methods are
presented and elucidated with examples.
1 Introduction
Epicyclic gear trains (EGTs) represent a class of geared kinematic chains that are
widely used as a major part of mechanical transmission systems, finding applications
in automobiles, machine tools, robotic manipulators, etc. During the conceptual stage
of design, it is common to represent the kinematic structure of an EGT as a simple
graph where vertices and edges denote links and joints respectively [1]. The graph
representation enables development of graph theory based algorithms which when
applied upon a set of parent graphs, enumerates EGT chains of a certain complexity
in the form of candidate graphs.
Once all the non-isomorphic graphs (i.e. concepts) are enumerated, the next stage
of conceptual design is topological analysis involving determination of all the distinct
V. R. Shanmukhasundaram (B)
Madanapalle Institute of Technology & Science, Madanapalle, Andhra Pradesh, India
Y. V. D. Rao · S. P. Regalla
Department of Mechanical Engineering, BITS PILANI, Hyderabad Campus, Secunderabad,
Telangana 500078, India
ways of assigning fixed, input, and output links, for a given EGT [2]. For a set of
combinations of fixed, input and output links, the following need to be evaluated:
(1) Topologically identical possibilities owing to the existence of symmetry.
(2) Those possibilities that render certain link/s as redundant.
As first effort in this regard, Olson et al. [2] analyzed all the distinct 5-link EGTs
for combinations of an input, output and a fixed link and discarded those containing
kinematically superfluous link/s. Hsu and Lin [3] devised a method based on the
fundamental circuits of the EGT, for identifying redundant gears. Liu and Chen [4]
introduced the concept of kinematic fractionation and related rules following which
redundant links can be prevented while assigning different input and output links of
the EGT chain.
Del Castillo [5] derived functional constraints to be satisfied by an EGT graph to
avoid idle links. These constraints when applied result in specific combinations of
sun, planet, and carrier links of an EGT. All such EGT graphs with these combinations
are then enumerated. However, Del Castillo [5] also mentions that the conditions
imposed in the enumeration are necessary but not sufficient for guaranteeing all the
inversions with driver and driven link combinations, to be free from idle links.
Rao and Rao [6] and more recently Rajasri et al. [7] have used the Hamming
number matrix in a procedure for partitioning the links of EGT into different classes
based on symmetry. Their endeavor [6, 7] was to determine symmetrical links
corresponding to topologically identical vertices in the EGT graph.
The topological analysis of an EGT chain, being a basic phase of conceptual
design, ultimately paves way for analyzing various kinematic configurations for
speeds, torques, and in turn efficiency. A similar problem to topological analysis
is that of enumerating the clutching sequences for an EGT graph to be used in
multi-speed transmissions with sometimes multiple operating DOFs [8–12].
From the literature survey, it is evident that different works on EGTs were aimed at
obtaining kinematic configurations of EGTs and understand the kinematic structure
of an EGT for any symmetry as well as driver-driven link combinations that result
in certain redundant link/s. Though graph theory based synthesis methods are well
established for catering to the enumeration of distinct possibilities of EGTs, methods
for topological analysis in the literature are not comprehensive enough and are very
much needed for an EGT designer. The present work aims to address this aspect and
is organized in the following manner.
In Sect. 2, graph representation of an EGT is briefly dealt with. Determination
of symmetry properties of an EGT are explained in Sect. 3 with suitable examples.
Information of a given rotation graph is input to Nauty and Traces package [13, 14]
and orbits and automorphism group are obtained as output, which are in turn used
to characterize the symmetry inherent in the rotation graph [6, 7] and in the EGT.
Section 4 introduces a flow graph model for visualizing the internal torques of
the EGT. This flow graph can be constructed using the information from the rotation
graph. Using connectivity properties of the flow graph, it is shown that, condition
for the presence of redundant link/s can be expressed [15].
Topological Analysis of Epicyclic Gear Trains—Symmetry … 1233
In the graph of an EGT, vertices, and edges respectively imply the links and joints
(pair connections). In this manner, through the topology existent in the graph, the
kinematic structure inherent in the EGT is revealed. There exist two kinds of pairs
in any EGT, namely, turning pairs and gear pairs. In the EGT graph, the edges are
colored in order to differentiate turning and gear pairs. The graph of an N-link and
F-DOF EGT consists of N vertices, (N − 1) turning pairs and (N − 1 − F) gear
pairs [1].
In the graph of N-vertex EGT, all the (N − 1) turning pairs are connected (i.e. not
disjoint). Therefore, the set of turning pair edges (TPE) constitutes a spanning tree
and as a consequence, every gear pair edge along with path of corresponding TPEs
gives rise to a fundamental circuit (FC). The TPEs are labeled with levels specifying
the position of axis in space. Further, all those edges having same label are connected
and constitute a tree subgraph with respect to the spanning tree graph of all TPEs [16].
In every FC, each TPE can possess one of the two different edge labels prevailing
in the circuit. Recalling that, since the same labeled TPEs are not disjoint, therefore
as a consequence, in every FC there is a transfer vertex (TV) that separates TPEs
possessing one common label from those TPEs having the other common label.
A rotation graph has the information of all the gear pair edges (GPEs) and its TVs
(i.e. as in the FCs). Displacement graph/s can be generated by labeling the TPEs in
different ways such that TVs of all FCs remain the same as given in the rotation graph.
In Fig. 1a, a 5-link EGT is shown in its closed rotation graph representation [17] with
TV marked as vertex-1. The three possible displacement graphs with respect to the
rotation graph are shown in Fig. 1b–d.
Traditionally, the structural synthesis of EGTs, with a given number of links and
DOF, focuses on enumerating all the non-isomorphic rotation graphs and all the non-
isomorphic displacement graphs with respect to every rotation graph [18]. Using a
displacement graph, an EGT schematic can be sketched, thereby giving ideas to an
EGT designer during the conceptual stage.
(a)
permuting the group elements. All such exhaustive permutations (mappings) belong
to symmetric group whose size given by:
group elements
Group size of symmetric group = i. (1)
i=1
group elements
where, i=1 i. denotes factorial function. For the purpose of symmetry anal-
ysis of a graph, permutations of group elements of the graph can be used to convey
those compositions (group actions) that preserve the topology of the graph and there-
fore that of the EGT chain. All those structure-preserving permutations constitute
the automorphism group, which is also a subgroup of the symmetric group.
The group elements can be partitioned into equivalent classes or orbits. Those
group elements belonging to the same equivalent class (vertex orbit or edge orbit),
occupy similar topological positions in the corresponding graph. For a given group
object (EGT graph), equivalence classes and automorphism group are related by the
Burnside lemma [18]. Nauty and Traces program [14] can be used to obtain the orbits
and automorphism group with respect to a rotation graph of EGT, as input. In our
previous work [20] symmetry analysis results with up to 6 links were tabulated.
Topological Analysis of Epicyclic Gear Trains—Symmetry … 1235
The knowledge of vertex orbits with respect to an EGT graph can be applied to the
topological analysis of same. The EGT schematic of a displacement graph in Fig. 1c
is given in Fig. 2 which is commonly known as “Ferguson paradox”.
It is basically a compound EGT with link-1 as carrier link along with its associated
compound planet being link-2 which meshes with suns: link-3 and 4 (external gear)
and link-5 (internal gear). Generally, planet carrier link-1 is given an input. One
among the three sun links is fixed and output power is derived from the remaining
two suns.
The symmetry analysis with respect to the EGT graph of Fig. 1 revealed the orbits
as: {(1); (2); (3 4 5)}. Therefore, the sun links (3, 4, and 5) are all topologically similar
since they all fall in the same vertex orbit (indicated in blue color in Table 1). As a
result, the 3 suns can be interchanged for the choice as fixed link thereby leading to
different inversions which are all kinematically equivalent as mentioned in Table 1.
An EGT graph can be decomposed into kinematic units (KUs) by splitting them
at the TVs [16]. In the previous example of a 5-link EGT, it was seen that similar
links were associated with FCs belonging to a single carrier link (TV) and therefore
only a single KU exists. In the case of multi-carrier EGT graphs (more than one KU),
symmetrical links can exist across KUs if and only if the TV of respective KUs is
topologically similar. As an example, a 6-link EGT is shown in Fig. 3a along with
Table 1 The kinematic equivalence of different operating modes owing to the graph symmetry
Operating
Input link Output links Fixed link Comments
Mode
Mode-1 1 4, 5 3
All the three modes are
Mode-2 1 3, 5 4
kinematically equivalent.
Mode-3 1 3, 4 5
1236 V. R. Shanmukhasundaram et al.
(a)
(b)
the KUs in Fig. 3b. The vertex orbits are found to be: {(1 2); (3 4); (5 6)}. This 6-link
EGT has TVs link-1 and link-2 being topologically similar and as a consequence
all associated links from the respective KUs, are also similar. Link-3 in KU of TV-1
is similar to link-4 in KU of TV-2. Likewise, link-5 and link-6 are also similar. A
possible schematic representation of this EGT is given in Fig. 4.
There exist four central links for this 6-link EGT excluding the two planet links:
the carrier links 1 and 2 as well as sun links 5 and 6. If we consider giving input
to a link, fixing a link and deriving outputs from the remaining two links then there
are many such possibilities. Owing to symmetry as apparent from vertex orbits, a
link corresponding to a vertex orbit can be interchanged for another link falling in
the same orbit thereby giving rise to similar possibilities as listed in Table 2. For
example, Mode-1 and Mode-2 are similar because they are mutually convertible by
Table 2 The sets of topologically similar possibilities for input-output-fixed links for Fig. 4
Operating Mode Input link Output links Fixed link Comments
Mode-1 1 5, 6 2 Mode-1 and Mode-2 are
Mode-2 2 5, 6 1 topologically identical
Mode-3 6 1, 2 5 Mode-3 and Mode-4 are
Mode-4 5 1, 2 6 topologically identical
Mode-5 2 1, 6 5
Mode-5, Mode-6, Mode-7
Mode-6 2 1, 5 6
& Mode-8 are topologically
Mode-7 1 2, 5 6
identical
Mode-8 1 2, 6 5
interchanging link-1 and link-2 since they belong to same vertex orbit (marked in
blue color); the other vertex orbit consists of links 5 and 6 (marked in green color)
and they are outputs.
4.1 Background
across KUs of the EGT, such that there is not any KU which does not contain at least
one actuating link or an output link.
A flow graph model is introduced, whose vertex set consists of all the links and gear
pair connections. The edges of the flow graph represent internal torques.
All the links in an EGT have rotary motion and their shafts naturally possess
external torques. A rotary motion from one link to another is transmitted by means
of a gear element (keyed to the respective shafts of links), resulting in an internal
torque. The algebraic sum of all the internal torques at a link is equal to its external
torque. Now, a distinction can be made regarding the nature of internal torques
existing in an EGT namely: meshing torques and planet-Carrier torques. A meshing
torque at a link is due to a gear pair connection with another link and it is numerically
given by:
Planet − Carrier torque = Arm length of carrier ∗ Meshing force of planet (3)
set consists of the 6 links and the four gear pairs: G (1, 2), G (1, 5), G (2, 3) and G
(4, 5). The edges of the flow graph along with their internal torques are tabulated in
Table 3.
Given the input and output links, the direction of internal torques is not obvious
in a flow graph. This is because the direction of torques depends on the magnitudes
of meshing powers which is determined by the values of angular velocities of the
links. However, by employing the flow graph, it is possible to determine the presence
of otherwise redundant link/s in the EGT with respect to the input and output links.
The approach used here relies on the connectivity of the proposed flow graph model
and it is explained with the help of an example given by Hsu and Lin [3].
With respect case-1 of Table 4, inputs are link-1 and link-3 and output link is 4.
All those edges belonging to the possible paths from link-1 and link-3 leading to
output link-4, are marked as red in Fig. 7a. It is evident that edge 5–6 is not marked
red because whatever torque the planet link-5 transmits to the carrier link-6 does
not contribute to the overall external torque at output link-4. Therefore, link-6 is
redundant. In this operating mode, the torques transmitted to planet link-5 from link-
4 and link-1 is futile. This is because: under normal operating conditions (without any
Table 3 Details of nature of internal torques with respect to flow graph of Fig. 6
Edge in the flow graph Type of internal torque
2–4 Planet-carrier torque between links 2 and 4
5–6 Planet-carrier torque between links 5 and 6
2 − G(2, 3) Meshing torque of link-2 owing to gear pair 2–3
3 − G(2, 3) Meshing torque of link-3 owing to gear pair 2–3
1 − G(1, 2) Meshing torque of link-1 owing to gear pair 1–2
2 − G(1, 2) Meshing torque of link-2 owing to gear pair 1–2
5 − G(1, 5) Meshing torque of link-5 owing to gear pair 1–5
1 − G(1, 5) Meshing torque of link-1 owing to gear pair 1–5
4 − G(4, 5) Meshing torque of link-4 owing to gear pair 4–5
5 − G(4, 5) Meshing torque of link-5 owing to gear pair 4–5
1240 V. R. Shanmukhasundaram et al.
(a) (b)
Fig. 7 a Flow graph of case-1 with those edges marked red which constitute the paths between
input links (1 and 3) and output link 4; b the two components of the flow graph obtained by splitting
the graph at link-4 and link-1
circulating power), the torque supplied to link-5 takes away useful torque developed at
link-4. Hence link-5 is also redundant. Another reasoning is also offered to justify that
link-5 and link-6 are redundant: angular velocity of output link-4 can be calculated
using the input angular velocities of link-1 and link-3 and by solving the Willis
equations of FCs: 4(1 2) and 4(2 3). Both the FCs, which include the constituent
links (1, 2, 3, and 4), belong to same kinematic unit. Another KU consists of FCs:
6(1 5) and 6(4 5). Link-1 and 4 are common to both the KUs (Fig. 7b). Since all the
input and output links belong to same KU, they are kinematically fractionated with
respect to the other KU [4]. Therefore, link-5 and link-6, being in a different KU
than the input and output links, are redundant.
Regarding the case-3 of Table 4; inputs are given to link-1 and link-4 and output
is taken from link-6. As evident from the flow graph with respect to this mode in
Fig. 8a, gear pair (2, 3) and link-3 are redundant. For cases 2 and 4, none of the
links are redundant as apparent from Fig. 8b. This is because: all the edges in the
flow graph are traversed by the possible paths between input and output links (all the
edges belonging such paths are marked in red).
Topological Analysis of Epicyclic Gear Trains—Symmetry … 1241
(a) (b)
Fig. 8 a Flow graph with respect to case-3; b flow graph for cases 2 and 4
5 Conclusions
Given the wealth of numerous EGT graph solutions as design alternatives, there
is always a need for rational guidelines to assist a designer in selecting a trans-
mission for a particular application. The developed graph theoretic framework for
topological analysis can help the EGT designer in evaluating various EGT trans-
mission architectures and thereby synthesizing different operating modes, as per the
requirements.
If symmetry is present in the EGT graph, then some of the input-output link
combinations (i.e. those concerning the topologically similar links), will turn out to
be kinematically equivalent.
The presence of redundant link/s in an operating mode will have a detrimental
effect on mechanical efficiency. This is because: the meshing power due to the redun-
dant link/s does not contribute to the power derivable at the output link. Using the
connectivity properties of a flow graph, the presence or otherwise of redundant link/s
can be found with respect to specified input and output links of the EGT. In summary,
the topological analysis gives deeper insight about the kinematic structure and the
kinematic interactions within the EGT. In order to answer the question as to which
EGT concept is best for an intended application and what being the most efficient
operating mode, it necessitates a detailed study of internal power flows. The topolog-
ical analysis methods can be perhaps extended, in order to facilitate such a thorough
power flow study; and it is planned as our future work in EGTs.
References
3. Hsu CH, Lin YL (1994) Automatic identification of redundant gears in planetary gear trains.
Math Comput Model 20(7):99–109
4. Liu CP, Chen DZ (2001) On the application of kinematic units to the topological analysis of
geared mechanisms. J Mech Des 123(2):240–246
5. Del Castillo JM (2002) Enumeration of 1-DOF planetary gear train graphs based on functional
constraints. J Mech Des 124(4):723–732
6. Rao AC, Rao YVD (2002) Symmetry in planetary gear trains. Indian J Eng Mater Sci 9:311–314
7. Rajasri I, Gupta AVSSKS, Rao YVD (2014) Symmetry and its effects on structures of planetary
gear trains. J Inst Eng (India): Ser C 95(1):30–38
8. Hsieh HI, Tsai LW (1998) The selection of a most efficient clutching sequence associated with
automatic transmission mechanisms. J Mech Des 120(4):514–519
9. Hwang WM, Huang YL (2011) Connecting clutch elements to planetary gear trains for
automotive automatic transmissions via coded sketches. Mech Mach Theory 46(1):44–52
10. Peng ZX, Hu JB, Xie TL, Liu CW (2015) Design of multiple operating degrees-of-freedom
planetary gear trains with variable structure. J Mech Des 137(9). https://doi.org/10.1115/1.403
0856
11. Gao MF, Hu JB (2018) Kinematic analysis of planetary gear trains based on topology. J Mech
Des 140(1). https://doi.org/10.1115/1.4038072
12. Xu X, Sun H, Liu Y, Dong P (2019) Automatic enumeration of feasible configuration for the
dedicated hybrid transmission with multi-Degree-of-Freedom and multiplanetary gear set. J
Mech Des 141(9). https://doi.org/10.1115/1.4042846
13. Mckay BD, Piperno A (2018) Nauty and traces user’s guide (version 2.7). Technical report,
Departimento di Informatica, Sapienza Universita di Roma
14. The nauty and Traces page. http://pallini.di.uniroma1.it/
15. Harary F (2001) Graph theory. Narosa Publishing House
16. Shanmukhasundaram VR, Rao YVD, Regalla SP (2019): Enumeration of displacement graphs
of epicyclic gear train from a given rotation graph using concept of building of kinematic units.
Mech Mach Theory 134:393–424
17. Ravisankar R, Mruthyunjaya TS (1985) Computerized synthesis of the structure of geared
kinematic chains. Mech Mach Theory 20(5):367–387
18. Yang W, Ding H (2018) The complete set of 1-DOF planetary gear trains with up to 9 links. J
Mech Des. https://doi.org/10.1115/1.4041482
19. Duzhin SV, Chebotarevsky BD (2011) Transformation groups for beginners. Paperback.
Universities Press, Hyderabad
20. Shanmukhasundaram VR, Rao YVD, Regalla SP (2019) Analysis of symmetry in epicyclic
gear trains. In: Uhl T (eds) Advances in mechanism and machine science. IFToMM WC 2019,
vol 73. Springer, Zurich, pp 1079–1090
Condition Monitoring and Identification
of Misalignment with Initial Unbalance
of Flexible Rotor-Bearing System
Abstract Shaft misalignment, rotor unbalance, rubbing, bearing faults and cracks
are the commonly observed faults in the rotor-bearing systems. Vibration signa-
ture has been considered to be one of the best ways to detect and diagnosis subse-
quent condition monitoring of rotor-bearing faults. A flexible shaft with a rigid disc
supported by flexible bearings at both the ends has been considered for the present
critical observations. The bearing is considered as a spring with linear stiffness
and damping. Further, Timoshenko beam theory is considered to describe math-
ematically the model of rotor element. We used finite element analysis to obtain
the critical whirling speeds of rotor-bearing system for different modes at various
system parameters with the illustration of Campbell diagram and Fourier spectrum.
Vibration spectrums in the transverse direction are illustrated for different modes of
vibration using successive approaches of Hamilton principle, Galerkin’s method and
Newmark’s integral method. Condition monitoring of vibration spectrums consid-
ering rotor unbalance and shaft misalignment as well as their identification has been
analysed with FFT, time responses, phase portrait and Poincare’s map, numerically
and experimentally.
1 Introduction
The most important configurations in rotor dynamics, i.e. shaft or shaft-disc assem-
blies are generally used to transmit power. Shaft misalignment, rotor unbalance,
rubbing, bearing faults and crack are the commonly observed faults in the rotor-
dynamic systems. Detecting vibration signature is one of most efficient ways to
diagnosis the presence of faults in rotor-bearing system. Some researchers have
worked on analysis of the rotating machines such as conditioning monitoring and
diagnosis along with mathematical modelling of the system. Sekhar et al. [1] and
Gibbons [2] studied the forces act on the coupling by parallel misalignment and
angular misalignment, respectively. Mohanty et al. [3] studied vibration signature
of shafting system to determine misalignments. In his book, Mahanty [4] described
conditioning and health monitoring of the machineries. Sudhakar et al. [5] used
equivalent loads minimization and vibration minimization method to determine the
position and amount of unbalance in the system. Nelson et al. [6] formulated the
rotor-bearing system considering the effect of rotatory inertia, gyroscopic moments,
axial load, bearing stiffness and damping by finite element method in fixed coordinate
and rotating coordinate system. Zorzi and Nelson [7] included all the parameters of
[6] with the effect of viscous damping and hysteresis damping. Nelson [8] included
all the parameter of [6] with the effect of shear deformation. He concluded that
the accuracy of the finite element modal improves as the number of an element is
increased. Isermann et al. [9] described evolution in the application of model-based
fault detection and diagnosis. Xu and Marangoni [10] derived theoretical model
and performed an experiment to the analysis of misalignment and unbalance. Patel
and Darpe [11] conducted experiment for analysis of parallel and angular misalign-
ment signature by different amount of misalignment and different running speed of
the system. He explained the misalignment by orbit plot, time waveform and FFT.
Very recently, Phadatare et al. [12, 13] studied nonlinear dynamics of flexible rotor-
bearing system using time response, fast Fourier transform, Campbell diagram and
Poincare’s map. Karpenko et al. [14] modelled Jeffcott rotor model with preloaded
snubber ring to analyse the interactions between the whirling rotor and the massless
snubber ring. However, the experiment and the theory results are compared using
bifurcation diagrams, Poincare’ maps and phase plane diagrams. Mohanty et al.
[15] widely described in detail about the methods basically used in conditioning
monitoring of machineries. Schwarz and Richardson [16] explained how to deter-
mine the modal parameter (modal frequency, damping ratio and modes shape) by
shaker and hammer test and explained the parameter to determine them like about
frequency response function, operating deflection shape, pretrigger delay, windows
type, leakage and curve fitting methods, etc. Pratiher et al. [12, 13, 17] studied
behaviour of the rotating system and used method of multiple scales to approximate
solution of the mathematical model. He also presented free vibration analysis and
behaviour of the rotating system under harmonic base excitation [14]. David Hutton
[18] presented fundamental theory regarding finite element analysis. John Vance
et al. [19] described analysis and principles of rotating machines. Piotrowski [20]
has given that the basic steps required for alignment of the machine and types of
maintenance. He explained about validity rule and different method to measure the
amount of misalignment. Chandra et al. [21] used time frequency techniques to detect
misalignment fault in rotating system and the detection performance were evaluated
by comparing CWT, HHT and STFT.
Most of rotating machines are subjected to major defects such as unbalance and
misalignment as malfunctions experienced in either at manufacturing level or at
assembly level which produce vibration. Thus, fault detection and diagnosis in
Condition Monitoring and Identification of Misalignment … 1245
rotating machine during operation are very important by both ways of econom-
ical and safety purpose. Hence, condition monitoring of the dynamic behaviour of
the rotor-bearing system with initial unbalance and shaft misalignment is studied
here. A mathematical model has been derived considering Hamilton’s principle and
Galerkin’s method using finite element approach. Study of modal characteristics as
well as detecting and diagnosis of the vibration caused by misalignment with a mass
unbalance has been performed. The theoretical and experimental results have been
verified. The results have been presented using Campbell diagram, time response,
FFT, phase portrait plot and Poincare’ map. This also enables us to understand the
system behaviour under mutual effect of unbalance and misalignment.
2 Mathematical Modelling
Two translation motion u(s, t) = u b (s, t) and v(s, t) = vb (s, t) are considered for
the disc. Two small rotations are α(s, t) and β(s, t) about x-axis and y-axis. The
shear effect is neglected in the analysis of the disc.
The frames of references for the disc mounted on a rotating shaft are shown as
F(XYZ) in Fig. 1. It is an inertial frame and R(x yz) is local coordinate which is
fixed to the disc. In order to find the orientation of the disc, following three consec-
utive rotations of the disc are considered: (1) β about Y axis defines x1 y1 z 1 (2) α
about x1 axis defines x2 y2 z 2 (3) ϕ about z 2 axis defines x yz. The angular rate of
relative to F(X Y Z ) can be obtained by the Euler angles as described in [6]. In rotor
configuration, translation displacement has been considered since axial displace-
ment is negligible as compared to the translation motion. The linear and rotational
displacement of the element can be approximated by the relation [8].
l
1
US = E I α 2 + β 2 + k AG (u − β)2 + (v + α)2 ds. (2)
2
0
The kinetic energy of a shaft element (T s ) and the disc (T d ) rotating at a constant
speed (Ω), including the translation and rotational forms, are given by
l l
1 2 1
TS = ρ A u̇ + v̇ 2 + ρ I α̇ 2 + β̇ 2 − 2I p α β̇ ds + I p 2 ds. (3)
2 2
0 0
1 1 1 1
Td = m d u̇ 2 + m d v̇ 2 + Id (α̇ 2 + β̇ 2 ) + I p ϕ̇ 2 − Idp α β̇ ϕ̇. (4)
2 2 2 2
Using Hamilton principle and then applying Galerkin’s method, the equation of
motion can be expressed in following form as [1].
( MTs + M Rs ) q̈es + (cv [Cvd ] − G s ) q̇es + K s qes = Q s . (5)
This model considers the flexible coupling because it permits some amount of parallel
or angular misalignment between two connecting shafts. Due to misalignment in
rotating machinery, reaction forces and moments are developed in the coupling [1,
7]. These reaction forces are acting as periodic loads on the rotating shaft with a
Condition Monitoring and Identification of Misalignment … 1247
periodic function of half-sinusoidal having a time period of π/. This model has
considered 1 × and 2 × components of the reaction forces in finite element
formulation. The Z1 is an axis of driving machine and Z2 is an axis of driven machine.
The reaction forces and moment on both sides of the coupling are given below. For
parallel misalignment:-
M X 1 = Tq sin θ1 + K b φ1 ,
M X 2 = Tq sin θ2 − K b φ2 ,
MY 1 = Tq sin φ1 − K b θ1 ,
MY 2 = Tq sin φ2 + K b θ2 ,
Fx1 = −M y1 − M y2 Z3,
Fx2 = −Fx1 Fy1 = (Mx1 + Mx2 ) Z 3 ,
Fy1 = (Mx1 + Mx2 ) Z 3 ,
Fy2 = −Fy1 . (7)
The nodal force vector Q 1c and Q 2c corresponds to the left and right nodes of
the coupling element (Fig. 2).
⎧ ⎫
⎪
⎪ Fx1 sin(t) + Fx1 sin(2t) ⎪
⎪
1 ⎨ Fy1 cos(t) + Fy1 cos(2t) ⎬
Qc = ,
⎪
⎪ 0 ⎪
⎪
⎩ ⎭
0
⎧ ⎫
⎪
⎪ Fx2 sin(t) + Fx2 sin(2t) ⎪
⎪
2 ⎨ Fy2 cos(t) + Fy2 cos(2t) ⎬
Qc = (8)
⎪
⎪ 0 ⎪
⎪
⎩ ⎭
0
Effect of geometrical parameters like disc mass, disc position and shaft length on
the critical whirling speed is analysed with help of Campbell diagram for different
modes. Similarly, eigenspectrums are also portrayed for the different disc mass. As
well as, effect of the misalignment with initial imbalance is also discussed in this
section.
Experiment Set-up and Procedures: Schematic diagram of the rotating system is
prepared as shown in Fig. 1. The same setup is considered as a experimental setup.
It comprises a flexible shaft with a rigid disk and the shaft is supported by bearings.
One of the end of the shafts is coupled with a motor using the flexible coupling. The
motor can be controlled to attain different speeds of the rotor using a control panel
or computer interface. As well as the motor can be made accelerate/decelerate with
desired rate till the specified speed using the control interface. Two proximity sensors
are used to measure horizontal and vertical deflection of the shaft. The sensors are
located using L-shaped holder as shown in Fig. 1. SpectraQuest DAQ system with a
computer software interface is used to acquire and process the vibrations signals from
the sensors for better interpretation of the vibrational behavior of the system. Here,
specially manufactured papers with a thickness are used to vary the misalignment
value by accommodating beneath the bearings attachment base. The misalignment
value is measured near the coupling.
Modal analysis has been studied to evaluate the eigenfrequencies and eigenspectrums
to understand critical whirling speeds, i.e. forward and backward natural frequencies
with spin speed when the disc is located at L/3 distance for different disc mass by
keeping other parameter constant. From Fig. 3, the increase and decrease in forward
(N f )/backward (N b ) natural frequency of the rotating system can be observed with
increase in the shaft spin speed (). This phenomenon of splitting the frequency
in two branches is due to action of gyroscopic couple which is also spin speed-
dependent.
For most of the applications, the rotating systems whirl with forward natural
frequency (N f ). Therefore, it can be concluded that the critical speed of the shaft
gets increased with the spin speed. It is also observed that the natural frequencies
related to 3rd mode of vibration are greater than that of 2nd mode of vibration and 1st
mode of vibration. It has been found that natural frequency in 2nd mode of vibration
is nearly equal to four times as compared with 1st mode of vibration while natural
frequency is find to be nearly eight times as compared with fundamental natural
frequency. Similarly, the eigenspectrums corresponding 1st, 2nd and 3rd mode of
forward natural frequency have been represented in Fig. 4, when the disc is at the
mid-position. It helps to understand behavioural displacement patterns of the shaft
Condition Monitoring and Identification of Misalignment … 1249
Fig. 3 Campbell diagram and respective characteristic mode shapes for different disc mass
for different running speed. Based on the analysis of these eigenspectrums, proper
positioning of the disc can be selected to vary the effect of gyroscopic couple on the
system.
It is expected from the shown eigenspectrums that larger amplitude of the eigen-
spectrums for a heavy disc can be observed as compared to that of smaller value of
disc mass. Therefore, understanding the modal characteristics and its development
over the various mode of vibration is required to monitor their condition of safe
operating zone.
Hammer test has been conducted to understand the modal characteristics of exper-
imental set-up, and it is especially the eigenfrequency. Here, the experiment is
performed at stand still position of the shaft. The calibrated hammer is used to apply
initial excitation to the rotor system by striking at mid-point of the shaft. Resulting
displacement behaviour of the system is captured and stored in the computer using
data acquisition system. The data acquisition system includes proximity sensors,
tachometer, Spectra-Pad signal processing unit, motor control and Spectra-Quest
software module. The resulting data is stored in excel file and then converted into
displacement form using calibration sheet of proximity sensors for further graph-
ical presentation and analysis. The time response, fast Fourier, phase portrait and
1250 S. Singh et al.
Poincare’s map are portrayed using this data as shown in Fig. 6. Free vibration anal-
ysis of the rotor-bearing system at zero spin speed has been demonstrated numerically
and experimentally. Signal samples of 10 s duration are acquired using data acqui-
sition system and better parts of the signals after hammer impact have been consid-
ered for further signal conditioning and analysis purpose. The corresponding time
response has shown attenuation of the amplitude per oscillating cycle till the system
come to rest position. This phenomenon is due to presence of inherent damping in the
system that causes response to set at rest after some oscillations. This behaviour can
be used to detect characteristics of the system at stand still position such as damping
coefficient and natural frequency. A peak in FFT depicts the presence of frequency
of free vibration.
The system has lost energy per cycle of oscillations, and this is also observed
from phase portrait and Poincare map. It has been depicted that the presence of the
positive damping caused the equilibrium point to act as a stable attractor. Similar
system parameters with the experimentally obtained damping coefficient have been
used to verify proposed mathematical model. The numerical results are portrayed in
Condition Monitoring and Identification of Misalignment … 1251
Fig. 5 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare map with a
disc located at L/3 distance from the bearing; h = 0.0161 m Theoretically
Fig. 5. Then, these results have been compared with experimental results (Fig. 6).
In Figs. 5b and 6b, it has been observed that the error in prediction of the frequency
obtained by mathematical model has been found less than 5%. As well as, Figs. 5a
and 6a have depicted that the rate of attenuation of the signal amplitude observed to
be same. So, the both results are in pretty much compliance and proposed model can
be considered for further analysis. Slightly more amplitude of peak in FFT of the
experimental result is observed compared to the theoretical; it may be result of more
than one frequency because corresponding time response shows beat phenomenon.
But it is neglected due to no detection of more than one in FFT plot.
Alignment is very necessary for any system, if the system has misalignment; it
produces vibration and results into loss of the power and suffer coupling wear, etc.
1252 S. Singh et al.
Fig. 6 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare map with a
disc located at L/3 distance from the bearing; h = 0.0161 m Experimentally
Any system like turbine, compressor and pump, etc., has work of power transmission.
Those systems become vulnerable towards vibration due to the misalignment in
the system. Results obtained numerically considering shaft misalignment have been
depicted in Figs. 7 and 8 and corresponding experimental findings which have been
obtained by considering parallel misalignment with initial unbalance are portrayed
in Figs. 9 and 10.
In Figs. 7 and 8, the rotating system behaviour has been portrayed by varying
magnitude of the parallel misalignment (i.e. 1.08 and 2.16 mm). The corresponding
time response history has depicted periodic behaviour with presence of more than
one frequency (Figs. 7b and 8b). Increase in magnitude of the 2X frequency has been
detected with increase of parallel misalignment magnitude from 1.08 to 2.16 mm.
As well as, looping can be observed in the phase portrait diagram due to increasing
magnitude of the 2*X component. To develop Poincare’s map, the response signal
has been sampled with 70 Hz frequency. The resultant Poincare maps have shown
two dots, and it has depicted double periodic behaviour of the system.
Condition Monitoring and Identification of Misalignment … 1253
Fig. 7 a. Time response b. Fast Fourier Transform plot c. Phase portrait d. Poincare’ map with a
disc located at L/3 distance from the bearing; parallel misalignment 1.08 mm. Theroetically
Experimentally obtained results have been plotted in Figs. 9 and 8. In the time
response, the beat phenomena have been observed. In FFT (Figs. 9b and 10b), the
vibration has also shown peak at 1-X, 2-X, 3-X and 4-X and consequently it signi-
fies presence of the misalignment in the experimental set-up. Same behaviour as
described for that of theoretical results has been observed in these experimental
results also. More no. of peaks have been observed in experimental results (Figs. 9b
and 10b) but 3X and 4X have very small effect on the system behaviour. This can be
depicted by comparing Figs. 7c and 8c with Figs. 9c and 10c, respectively, and the
corresponding trajectories have similar behaviour on phase portrait.
Parallel misalignment without an unbalance (m u )
It is always difficult to completely balance the rotating system. Defective manu-
facturing of the rotor or defective assembly are main cause of presence of an unbal-
ance in the system. To analyse the misalignment effect, mathematical model of the
rotating system with parallel misalignment by neglecting initial unbalance has been
presented in Figs. 11 and 12. When there is no initial unbalance in the system, only
the misalignment in the system at the 2*X frequency is dominated with respect to
1254 S. Singh et al.
Fig. 8 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare’ map with a
disc located at L/3 distance from the bearing; = 35 Hz, mu = 0.0006 kg and parallel misalignment
2.16 mm. Theoretically
1*X frequency, this has been observed from FFT plot. In the time wave form graph,
two frequencies are seen clearly and looping of trajectory has been observed in phase
portrait which is due to dominance of the 2*X component. Comparison of FFT plots
of Figs. 8 and 12 depicts attenuation of the misalignment effect with presence of
the unbalance mass. So, 1-X component is more dominated with respect to the 2-X
component due to the unbalance mass. Thus, the misalignment effect gets attenuated
comparatively but at the expenses of increased unbalance force. Same behaviour of
the system also observed in Figs. 7 and 11.
Thus, condition monitoring of the rotating system can be understood well with
help of above analysis. These observations are helpful to detect the health condition
of the rotating system prior to catastrophic failure of the system and as well as it will
help to determine type, exact source and vibrational behaviour.
Condition Monitoring and Identification of Misalignment … 1255
Fig. 9 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare’ map with a
disc located at L/3 distance from the bearing; = 35 Hz, mu =0.0006 kg and parallel misalignment
1.08 mm. Experimentally
4 Conclusion
The free vibration analysis helps to identify the characteristics of the system, and the
characteristics of the system are much in compliance for experimental and theoretical
model. Experimentally and theoretically analysis is done for the mass unbalance of
the system, and it is observed that when the system has only mass unbalance then
only 1-X peak observed in FFT. When the system has only parallel misalignment
then the 2-X peak gets dominated in the system. When the system has both initial
unbalance and parallel misalignment but the initial unbalance amount is high, then
the 1-X frequency is dominated and effect of 2-X frequency becomes less. This
method may be applicable for the large systems like a turbine, compressor, gearbox,
etc., to investigate the unbalance and misalignment in the system. The misalignment
effect can be monitored or controlled at the expenses of increasing unbalance force
as per applications need.
1256 S. Singh et al.
Fig. 10 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare’ map with a
disc located at L/3 distance from the bearing; = 35 Hz, mu = 0.0006 kg and parallel misalignment
2.16 mm. Experimentally
Condition Monitoring and Identification of Misalignment … 1257
Fig. 11 a. Time response b. Fast Fourier Transform plot c. Phase portrait d. Poincare’ map with
a disc located at L/3 distance from the bearing; = 35 Hz and parallel misalignment 1.08 mm.
(Theoretically)
1258 S. Singh et al.
Fig. 12 a. Time response b. Fast Fourier transform plot c. Phase portrait d. Poincare’ map with
a disc located at a L/3 distance from the bearing; = 35 Hz and parallel misalignment 2.16 mm.
Theoretically
References
6. Nelson HD, McVaugh JM (1976) The dynamics of rotor-bearing systems using finite elements.
J Eng Ind 19:75
7. Zorzi ES, Nelson HD (1977) Finite element simulation of rotor-bearings systems with internal
damping. J Eng Power 8:76
8. Nelson HD (1980) A finite rotating shaft element using timoshenko beam theory. J Mech Desi
102:799
9. Isermann R, Balle P (1997) Trends in the application of model based fault detection and
diagnosis of technical processes. Control Eng Pr 5(5):709–719
10. Xu M, Marangoni RD (1993) Vibration analysis of a motor-flexible coupling-rotor system
subject to misalignment and unbalance, Part I: theoretical model and analysis. J Sound Vib
176(5):663–679
11. Patel T, Darpe A (2009) Experimental investigation on vibration response of misaligned rotors.
Mech Syst Sig Proc 23:2236–2252
12. Phadatare HP, Maheshwari V, Vaidya KS, Pratiher B (2017) Large deflection model for
nonlinear flexural vibration analysis of a highly flexible rotor-bearing system. Int J Mech
Sci 134:532–544
13. Phadatare HP, Pratiher B (2016) Nonlinear frequencies and unbalanced response analysis of
high speed rotor-bearing systems. Procedia Eng 144:801–809
14. Phadatare H, Choudhary B, Pratiher B (2017) Evaluation of nonlinear responses and bifurcation
of a rotor-bearing system mounted on moving platform. Non-lin Dyn 90:493–511
15. Karpenko EV, Wiercigroch M, Pavlovskaia EE, Neilson RD (2006) Experimental verification
of Jeffcott rotor model with preloaded snubber ring. J Sound Vib 298:907–917
16. Schwarz BJ, Richardson MH (1999) Experimental modal analysis. CSI Reliability Week,
Orlando, FL
17. Choudhary B, Pratiher B (2015) Numerical studies of a nonlinear flexible rotating system
under harmonic ground motion. In: Proceedings of 9th IFToMM international conference rotor
dynamics, vol 21, pp 1677–1687 (2015)
18. Hutton D (2004) Fundamental of finite element analysis. McGraw-Hill
19. Vance J, Zeidan F, Murphy B (2010) Machinery vibration and rotordynamics. Wiley (2010)
20. Piotrowski J (2007) Shaft alignment hand book. Third Edition by CRC Press
21. Chandra HN, Sekhar AS (2015) Fault detection in rotor bearing systems using time frequency
techniques. Mech Syst Sig Proc 72(73):105–133
Effect of Unbalance with Bearing
Flexibility on Vibration Phenomenon
of Geometrically Nonlinear Rotating
Shaft with Ball Bearing
1 Introduction
Free vibration analysis has the most important role in designing vibrating systems,
because high energy vibrations mostly occur at critical speed of the system, and
it affects its regular functionality and sometime may likely to face possibility of
catastrophic failure or weaken the system. There are number of applications of rotor-
bearing system such as turbomachinery, airplanes, pump, compressor, ships, and
Limited researchers studied the rotor dynamics considering flexible shaft with rigid
disk supported by a flexible bearing. The characteristics of the bearings have a promi-
nent role not only in the dynamics of modal parameters but also system performance
characteristics. The primary concern of this paper is to understand the effect of bear-
ings flexibility along with mechanical unbalance on the flexural vibration character-
istics of the rotor-bearing system. The mutual contributions of different parameters
under unbalance condition have also been studied. Similarly, the study the effect of
damping and bearing stiffness to its behavior in forcing condition has been performed
experimentally and theoretically. We here study the vibration aspects of geometri-
cally nonlinear rotating shaft combined with rigid disk. The shaft has been modeled
and Euler–Bernoulli beam element accounting of the effects of rotary inertia, gyro-
scopic effect, higher-order large deformations and rotor mass unbalance. In next
section, a mathematical model of rotor-bearing system consisting of a flexible shaft
with a disk simply supported by bearings has been formulated using Euler’s beam
theory. As well as, stretching nonlinearity, gyroscopic effect, and rotary inertia effect
have been taken into consideration.
The system is consisting of a flexible shaft, a rigid disk, and flexible bearings. The
shaft has length L and the disk is located along the shaft length at a distance L d from
an end of the shaft. The shaft is pinned at its both end and supported by the flexible
bearings at same distance lb from both ends of the shaft. The bearing is characterized
as a flexible linear spring.
The bearing is characterized as a flexible linear spring. Expressions for kinetic
energy of balanced rotor-bearing system can be expressed as [5]
L L L
ρA 2 ρI 2
Tt = v̇ + ẇ 2 dx + α̇ + θ̇ dx + (2ρ I )α̇ θ dx
2
2 2
0 0 0
L
1 2
+ (m) v̇ + ẇ 2 δ(x − L d ) dx
2
0
L
1 2
+ I α̇ + I2 θ̇ 2 + β̇ 2 δ(x − L d ) dx. (1)
2
0
Here, ρ, A, I, I d are the mass density, cross-sectional area, polar and diametrical
moment of inertia of the shaft, respectively, and m is mass of the disk α̇, θ̇ and β̇
are the time derivatives of the angular displacement about the axes of Z, Y, and X.
1264 H. P. Phadatare et al.
Dirac delta function has been incorporated in order to represent the disk effect in the
distributed system. The kinetic energy due to unbalance mass has been expressed as
[10, 11].
mu
Tu = m u d(v̇ cos t − ẇ sin t) + d2 . (2)
2
Here, mu , and d are the unbalance mass and its location from geometric center
of shaft. Strain energy of the shaft due to bending in rotating shaft is developed by
neglecting the effect of term of time functioning in the consequence of displacement
in x, y, z directions as the concept of u x = −zθ + yα, u y = v, u z = w [11]. Strain
energy of the flexible shaft can be expressed as
L L
E I 2 2 E A 4 4 2 2
U= v + w dx + v + w + 2 v w dx
2 8
0 0
L L
2 2 2 2
+ v + w dx v + w dx (3)
0 0
Here, (·) and over the letters represent time and space derivatives, respectively.
EA and EI are considered as axial rigidity and flexural rigidity, respectively. The
bearings are considered to be 2D linear spring. The bearings are located at a distance
L b from ends of the shaft. So, the strain energy stored in single bearing can be
expressed as
L
1
Ub = [(K l v)v + (K l w)w]δ(x − L b ) dx (4)
2
0
L L
2
= Md f (x) x=L d + Idy f (x) x=L d + ρA f (x) dx + ρ I
2
f 2 (x) dx,
0 0
L
Id x
2 2ρ A 2
κ1 = f (x) x=L d
+ f (x) dx,
0
L
EI 2 2K l
κ2 = f (x) dx + f 2 (x) x=L b ,
0
⎛ L ⎞
L L
EA 4 EA 4 mu
μ1 = f (x) dx, μ2 = ⎝ f (x) dx ⎠ dx, m u = .
l
0 0 0
(7)
Equations (5) and (6) can be expressed as below for free vibration analysis of the
system after neglecting effect of the unbalance terms,
1
V̈ − κ1 Ẇ + κ2 V + μ1 + μ2 V 3 + V W 2 + c V̇ = 0 (8)
2
1
Ẅ + κ1 V̇ + κ2 W + μ1 + μ2 W 3 + V 2 W + c Ẇ = 0 (9)
2
Since the governing equation of motion contains the nonlinear terms, the exact
solutions are not available as obtaining closed form is somehow difficult. Thus, the
perturbation approach (i.e., method of multiple scales) [12, 13] is used to get solution
of the Eqs. (5), (6) and Eqs. (8), (9). One may also go for a numerical method for
solution also. The MATLAB function ode45 is used to obtain numerical solution of
these equations.
1266 H. P. Phadatare et al.
Experimental setup of the rotating system is prepared as shown in Fig. 1. The setup
comprises a flexible shaft with a rigid disk and the shaft is supported by bearing
at a distance L b from the both ends. One of the end of the shafts is coupled with
a motor using the flexible coupling. The motor can be controlled to attain different
speeds of the rotor using a control panel or computer interface. As well as the motor
can be made accelerate/decelerate with desired rate till the specified speed using the
control interface. Two proximity sensors are used to measure horizontal and vertical
deflection of the shaft. The sensors are located using L-shaped holder as shown in
Fig. 1. SpectraQuest DAQ system with a computer software interface is used to
acquire and process the vibrations signals from the sensors for better interpretation
of the vibrational behavior of the system.
Here, a rotating flexible shaft has 0.93 m length (L) which is supported by the
bearings at a distance of 0.05 m from its both ends. The shaft has 7800 kg/mm3
mass density (ρ), 200 GPa modulus of elasticity (E) and 0.0152 m diameter (d). A
rigid disk is located along the shaft at a distance L d . The disk has 2340 kg/ mm3
mass density (ρ), 0.12 m diameter (D), and 0.03 m thickness. Equivalent damping
coefficient is taken as 30 Ns/m. Numerical results are found by varying geometrical
parameter of the system such as location of the disk, thickness of the disk, spinning
speed of the shaft, bearing stiffness, and viscous damping.
Free vibration analysis of the rotor-bearing system has been performed at zero spin-
ning speed by giving small initial excitation and received the output in the terms of
time response. The results have been obtained by varying geometrical parameters of
the system such as the location of the disk and the disk thickness. An experimental
setup to represent a theoretical model of rotor bearing has been considered. The
numerically obtained results have been compared with the experimental results in
terms of the system characteristic parameters (i.e., natural frequency and damping)
to verify the proposed theoretical model. The fundamental natural frequency of the
experimental setup having zero spin speed has been obtained by applying the small
impact at the middle of the shaft using the calibrated hammer. The resultant responses
of the system have been obtained in time domain then which has been converted into
frequency domain with use of fast Fourier transform.
The results of free vibration analysis in Figs. 2, 3, 4 and 5 are in terms of time
responses and fast Fourier transforms. The numerical and experimental results have
been presented in solid and dotted lines, respectively, in these figures for better
comparison. Figures 2 and 3 have been portrayed to show change in characteristic
behavior of the system due to change in disk position (L d = L/3 and L/6) with disk
Fig. 2 a Time response, b fast Fourier transform: with a disk located at L/3 distance from the
bearing
Fig. 3 a Time response, b fast Fourier transform with a disk located at L/6 distance from the bearing
1268 H. P. Phadatare et al.
Fig. 4 a Time response, b fast Fourier transform plot with a disk located at L/3 distance from the
bearing. h = 0.032 m
Fig. 5 a Time response, b fast Fourier transform plot with a disk located at L/6 distance from the
bearing; h = 0.032 m
thickness (h = 0.016 m). In Fig. 2b, frequency peaks can be observed at 57 and
57.21 Hz for numerical and experimental investigation, respectively. Such peaks can
also be observed at 65 Hz and 66.9 Hz in Fig. 3b, 48.47 Hz and 49 Hz in Fig. 4b,
57.51 Hz and 60 Hz in Fig. 5b for the numerical and the experimental models,
respectively. Figures 2 and 3 show the increase of the fundamental frequency of the
system as the disk location changes from L/3 to L/6. From these observations, we can
depict that the numerical results are in close compliance with experimental results
as the error is less than 5%.
Similar frequency change can be observed in Figs. 4 and 5 for another configura-
tion of the system. As well as, the decrease in fundamental frequency of the system
Effect of Unbalance with Bearing Flexibility on Vibration … 1269
Fig. 6 a Phase portrait, b Poincare map corresponding to Fig. 2 with a disk located at L/3 distance
from the bearing
with increase in the thickness of the disk can be also detected by observing Figs. 2,
3, 4 and 5.
As the system has viscous damping, the initially excited amplitude of the system
gets damped out and the system response comes to rest with time.
For better understanding of this dynamic behavior the system, phase portrait and
Poincare maps have also been plotted (Fig. 6). The fundamental natural frequency
has been considered as a sampling frequency to plot the Poincare maps. We can
observe in Fig. 6 that the trajectory has been attracted toward the origin and remains
at the same position for infinite time. When the system was perturbed infinitesimally
from the equilibrium point, it caused the system to attract toward the initial state. So,
the equilibrium point behavior looks similar to an attractor (Fig. 6a). When Poincare
plane cut the trajectory in phase space, the points have been observed approximately
in a straight line and pointing toward the origin as well as the cluster of points at the
origin has also been observed in Fig. 6b. Thus, it can be depicted that the motion
would remain at rest after finite oscillations, which has shown periodical behavior.
The corresponding damping coefficient has been obtained using time response data
of the hammer test.
As the experimental system are difficult to fully balance, a small initial imbalance
mass (mu = 0.0005 kg) in the system is determined using procedure as described in
[22].
The experimental results have been obtained by varying system parameters such
as spin speed and imbalance mass. These have been presented in Figs. 7, 8, 9 and
10 with consideration of steady-state part of the response. Due to presence of the
1270 H. P. Phadatare et al.
Fig. 7 a Fast Fourier transform plot, b phase portrait, c Poincare’ map with a disk located at L/3
distance from the bearing; Ω = 35 Hz (Experimental)
Fig. 8 a Fast Fourier transform plot, b phase portrait, c Poincare’ map with a disk located at L/3
distance from the bearing: Ω = 45 Hz (Experimental)
Fig. 9 a Fast Fourier transform plot, b phase portrait, c Poincare’ map with a disk located at L/3
distance from the bearing; Ω = 35 Hz, mu = 0.0055 kg. (Experimental)
imbalance in the system, the unbalance force acts on the rotor with frequency of
spinning speed (Ω). Figure 7a shows the peak at 35 Hz when spinning speed of the
shaft has been set at 35 Hz and same type of behavior has also been observed for the
spinning speed 45 Hz in Fig. 8.
Effect of Unbalance with Bearing Flexibility on Vibration … 1271
Fig. 10 a Fast Fourier transform plot, b phase portrait, c Poincare’ map with a disk located at L/3
distance from the bearing; Ω = 35 Hz, mu = 0.0095 kg. (Experimental)
From Figs. 7 and 8, it has been also observed that the magnitude of forcing
frequency peak increased with increase in the spin speed (Ω). Further, it increases
till critical speed and then may get decreased. The frequencies which are other than
forcing frequency have comparably very small magnitude. This causes very small
effect of these frequencies on vibration behavior of the system. Presence of damping
in the system has led the transient term of the response to vanish at initial stage and
resulted into steady-state response which has been dominated only by the forcing
term.
The phase portrait and Poincare’ maps have been plotted only for 1 s span of time
response of the steady state. The spin speed (Ω) has been considered as sampling
frequency to plot Poincare’ maps. From Figs. 7, 8 and 9, phase portrait diagrams have
depicted presence of the two frequencies at the lower spinning speed and the low
imbalance mass, as it shows looping of the trajectories with irregular closed loop. But
at higher imbalance mass (Fig. 10), the phase portrait plot has shown closed loop with
approximately oval shape. Consequently, the influence of frequencies other than that
of the unbalance has been observed almost negligible, so, it has been depicted that
the irregular behavior of the system has been eliminated due to increase of unbalance
forcing term.
The cluster of points at single location has been observed in Figs. 7, 8, 9, and 10
(Poincare’ map). It has been depicted that the system vibrated periodically with the
sampling frequency (i.e., forcing frequency) and the system has dominance of the
unbalance force.
The theoretical results (Figs. 11 and 12) have not shown heart shape trajectory
in phase portrait plots as compared to that of experimental results. This behavior is
due to presence of more than one frequency and this can also be depicted from the
FFT plots also. Source of the undefined frequency in the experimental model may
be due to misalignment or lose assembly of the system, etc. But, Fig. 10 shows that
this behavior has been attenuated by increasing unbalance magnitude in the system.
This unbalance magnitude can be increased by varying the spin speed (Ω) or the
unbalance mass (mu ). Theoretical analysis has shown dominance of the unbalance
force in the response of the system as same as the experimental analysis.
1272 H. P. Phadatare et al.
Fig. 11 a Fast Fourier transform plot, b phase portrait, c Poincare’ map with a disk located at L/3
distance from the bearing; Ω = 35 Hz, h = 0.016 m, mu = 0.0005 kg (Theoretical)
Fig. 12 a Fast Fourier transform plot, c phase portrait, d Poincare’ map with a disk located at L/3
distance from the bearing; Ω = 35 Hz, mu = 0.0095 kg. (Theoretical)
Further, the investigation of the rotor-bearing system has been extended theoret-
ically to analyze effect of the damping and bearing stiffness with inclusion of the
stretching nonlinearities. The obtained results have been portrayed in Figs. 13, 14. In
Fig. 13, the results are related to different damping values such as c = 0.0, 2.5, and
30 Ns/m obviously, and the comparison of different damping values has shown early
diminishing of transient term at high viscous damping value than the low damping
value. With no damping (c = 0.0 Ns/m), the time response has shown beating behavior
for infinite time due to presence of two frequencies as shown in Fig. 13. The pres-
ence of two frequencies in the vibration of the system has been depicted from corre-
sponding fast Fourier transform plot (FFT) also. Initial five seconds data have been
used to process the FFT plot with thousand samples per second sampling rate. The
peaks have been detected at 40 and 72.4 Hz, which are spin speed of the shaft and
natural frequency of the system, respectively. In the phase portrait plot, torus shape
trajectory has been observed and it has also depicted presence of two frequencies in
the vibration data. The Poincare’s map has been plotted by using plane cut the phase
space at spin speed frequency. The Poincare’ map has shown intersection of plane
as points distributed in an approximately circular closed-loop shape.
Effect of Unbalance with Bearing Flexibility on Vibration … 1273
Fig. 13 Time response, fast Fourier transform plot, phase portrait, and Poincare’ map: Ω = 35 Hz,
a c = 0.0 Ns/m, b c = 2.5 Ns/m, c c = 30 Ns/m. (Theoretical)
1274 H. P. Phadatare et al.
Fig. 14 Time response, fast Fourier transform plot, and phase portrait: Ω = 40 Hz, mu = 0.0095 kg
a K l = 2.1151 × 103 N/m b K l = 2.1151 × 105 Ns/m c K l = 2.1151 × 107 N/m (Theoretical)
So, it can be concluded that these frequencies are not commensurate and the
system has quasi-periodic behavior. There is possibility that the frequencies are
commensurate at very long period. So, it is always hard to differentiate periodic and
quasi-periodic behavior. This analysis has been limited to 30 s time history. Results
with damping c = 2.5 Ns/m have been portrayed in Fig. 13b., time response has shown
diminishing of transient term during initial 10–12 s and prevailing only steady-state
response afterward. The FFT has been represented for initial five second samples. It
has shown influence of natural frequency in the system behavior due to consideration
of the transient part. The diagram of FFT for the steady state has shown one peak
only and has depicted dominance of only one frequency, i.e., forcing frequency (Ω).
Effect of Unbalance with Bearing Flexibility on Vibration … 1275
As the phase portrait has been plotted by considering transient part of the response, it
has shown torus type shape trajectory. But for the steady state, it has shown circular
closed loop. One point has been observed on the Poincare map and it has been
depicted that the steady-state behavior of the system is a periodical. For damping
value c = 30 Ns/m, transient part of the response diminished in early stage so the
system showed a periodical behavior of the steady state.
Figure 14 has been plotted to analyze effect of change in stiffness of the bearings.
Increase in the natural frequency of the system has been observed in the FFT plot
due to increase in stiffness of the bearings, and some movement away from the
forcing frequency has also been detected. Consequently, the decrease in amplitude
of the system response has also been observed. The phase portrait plots have shown
reduction of the closed loop almost to a point with increase in the stiffness of the
bearing. At this configuration, decrease in transient part span in time domain has
been observed with increase in the bearing stiffness and almost negligible at its high
value.
4 Conclusion
The nonlinear behavior of rotating shaft coupled with disk has been analyzed for the
first mode of vibration accounting large elastic deformation. It has been observed that
the significant influences of gyroscopic, rotary inertia, and mass unbalance effects on
the dynamics of the rotor bearing. The present perspective of this work is to demon-
strate vibration characteristics of a flexible rotor-bearing system with the experi-
mental validation. We here illustrate the overall dynamics with time histories, FFT,
Poincare’s section, and phase plane trajectories.
The imbalance is already present in the system, thus 1× frequency can be seen in
FFT and this 1× frequency dominates in response of the rotor-bearing system. As the
transient response of the system damps out with time, only steady response can be
observed in time response. The damping and stiffness of the bearings are also play an
important role to decide the system qualitative behavior of the system. The damping
s property of the bearing can be used to change the behavior of the system from the
quasi-periodic to a periodic behavior. Moreover, a bearing with large stiffness value
can be used to suppress the vibration amplitude.
References
1. Sturla FA, Argento A (1996) Free and forced vibrations of a spinning viscoelastic beam. J Vib
Acoust 118(3):463–468
2. Karunendiran S, Zu JW (1999) Free vibration analysis of shafts on resilient bearings using
Timoshenko beam theory. J Vib Acoust 121(2):256–258
3. Melanson J, Zu JW (1998) Free vibration and stability analysis of internally damped rotating
shafts with general boundary conditions. J Vib Acoust 120(3):776–783
4. Sheu GJ, Yang SM (2005) Dynamic analysis of a spinning Rayleigh beam. Int J Mech Sci
47(2):157–169
5. Hosseini SAA, Khadem SE (2005) Free vibration analysis of rotating beams with random
properties. Struct Eng Mech 20:293–312
6. Hosseini SAA, Khadem SE (2009) Free vibrations analysis of a rotating shaft with nonlinear-
ities in curvature and inertia. Mech Mach Theo 44(1):272–288
7. Shabaneh NH, Zu JW (2000) Dynamic analysis of rotor–shaft systems with viscoelastically
supported bearings. Mech Mach Theory 35(9):1313–1330
8. Chang-Jian CW, Chen CK (2006) Nonlinear dynamic analysis of a flexible rotor supported by
micropolar fluid film journal bearings. Int J Eng Sci 44(15):1050–1070
9. Grybos R (1991) The effect of shear and rotary inertia of a rotor at its critical speeds. Arch
Appl Mech 61(2):104–109
10. Jei YG, Lee CW (1992) Modal analysis of continuous asymmetrical rotor-bearing systems. J
Sound Vib 152(2):245–262
11. Kim W, Argento A, Scott RA (1999) Free vibration of a rotating tapered composite Timoshenko
shaft. J Sound Vib 226(1):125–147
12. Phadatare HP, Maheshwari V, Vaidya KS, Pratiher B (2017) Large deflection model for
nonlinear flexural vibration analysis of a highly flexible rotor-bearing system. Int J Mech
Sci 134:532–544
13. Phadatare H, Choudhary B, Pratiher B (2017) Evaluation of nonlinear responses and bifurcation
of a rotor-bearing system mounted on moving platform. Nonlinear Dyn 90:493–511
14. Karpenko EV, Wiercigroch M, Pavlovskaia EE, Neilson RD (2006) Experimental verification
of Jeffcott rotor model with preloaded snubber ring. J Sound Vib 298:907–917
15. Phadatare HP, Pratiher B (2016) Nonlinear frequencies and unbalanced response analysis of
high speed rotor-bearing systems. Procedia Eng 144:801–809
16. Shad MR, Michon G, Berlioz A (2011) Modeling and analysis of nonlinear rotordynamics due
to higher order deformations in bending. Appl Math Modell 35:2145–2159
17. Yamamoto GK, Costa, C, Sousa JSS (2016) A smart experimental setup for vibration measure-
ment and imbalance fault detection in rotating machinery. Case Studies in Mech Sys Sig Proc
4:8–18 (2016)
18. Sudhakar AS, Sekhar AS (2011) Identification of unbalance in a rotor bearing system. J Sound
Vib 330:2299–2313
19. Shekhar AS, Prabhu BS (1994) Effects of coupling misalignment on vibration of rotating
machinery. J Sound Vib 185(4):655–671
20. Isermann R, Balle P (1997) Trends in the application of model based fault detection and
diagnosis of technical processes. Control Eng Pr 5(5):709–719
21. Amiya R (2015) Mohanty: machinery condition monitoring. CRC Press, Principles and
Practices
22. Nayfeh AH (1981) Introduction to perturbation techniques. Wiley-Interscience, New York
Analysis of Parametric Influence
on Control of a Two-Link Flexible
Manipulator Incorporating a Payload
Abstract The control problem for the trajectory tracking of a flexible two-link
manipulator incorporating a payload described by a nonlinear model is addressed in
the present work. The influence of parametric variation of system attributes on the
design of PD inversion-based nonlinear control of a two-link manipulator has been
demonstrated. The kinetic energy expression for the system is obtained by using the
position vector in generalized coordinate system. The Euler–Lagrange’s approach
in conjunction with the assumed mode method is utilized to obtain the dynamic
model of the system composed of four nonlinear ordinary differential equations
which are simulated, and the results have been graphically illustrated. The effect of
system parameters such as payload mass, joint mass, system inertia, and physical
and geometric properties of the links are explored by comparing the simulation
outcomes. Moreover, the presented results exhibit that the system parameters have
a significant effect on the input–output characteristics and should be accounted for
while designing such systems.
1 Introduction
The traditional robot manipulators have been designed to minimize the vibration
during and after the operation of manipulator and to achieve higher position accu-
racy by maximizing their stiffness. This design in turn has resulted in the higher
power consumption and lower productivity due to their bulky design. Moreover, the
payload carrying capacity of such manipulator is lower as compared to the manip-
ulator system weight. On the contrary, the flexible manipulators have higher speed,
higher payload capacity, larger maneuverability, smaller actuators, and lesser capital
cost. The mechanical flexibilities have been inducted in the links as well as actuator to
exploit the advantages of flexible robot manipulators. However, the flexibilities have
resulted in the challenging problem of increased vibrations and positioning inaccu-
racies of manipulators employed in precise applications in medical, space, military,
and nuclear industries. Therefore, an appropriate manipulator dynamic model and
control mechanism accounting of the rigid and flexural motions of the system is
necessitated. Moreover, it is also important to understand the influence of system
attributes on the control parameters which may help the design engineer in manip-
ulating the system for accurate positioning. A brief literature regarding the flexible
manipulators is presented in next to discern the recent developments in control and
vibration suppression of robots.
The vibration suppression by developing boundary and adaptive control laws of
a three-dimensional Euler–Bernoulli beam with a point payload has been demon-
strated by Al-Solihat et al. [1]. The estimation of variables representing the rigid and
flexible motions of single link flexible manipulator by a robust nonlinear observer
has been carried out by Chalhoub and Kfoury [2]. A smart structure including piezo-
electric actuator being modeled as an Euler–Bernoulli beam having a point payload
has been analyzed by Choi [3]. The active vibration control of the manipulator under
actuator hysteresis and natural frequency variation through a robust control using
quantitative feedback theory has been proposed. A control strategy for moving a
torque-driven planar single-link flexible manipulator to a specified angle with simul-
taneous vibration suppression has been developed by Meng et al. [4]. A nonlinear
state-dependent Riccati equation-based control strategy has been used for control-
ling the tip position of a single-link flexible manipulator by Shawky eta al. [5]. An
attempt to reduce the lateral oscillatory vibrations due to longitudinal motion of a
flexible link carrying a payload has been made by Shin and Rhim [6]. In order to
reduce the lateral vibrations, a controller based on input-shaping technique has been
developed with further discussion on the effects of payload mass and damping on
the dynamic behavior of the system. An Euler–Bernoulli beam with a hub and tip
mass payload undergoing rigid and elastic motions has been dynamically modeled by
Tavasoli [7] using Hamilton’s principle. A model-based control law for asymptotic
stabilization using Lyapunov-based boundary control for achieving the desired set
point through rigid body motion with reduced vibrations has been demonstrated. A
rigid link model with flexibility in the joint has been considered by Patel et al. [8] and
developed a feedback linearization controller for the trajectory tracking. The results
have been compared with those of the PD controller, and it has been illustrated that
the proposed controller exhibits better performance in presence of joint flexibility.
The trajectory tracking problem of a two-link manipulator with intermediatory
lubricated revolute joint modeled as a short journal bearing for the purpose of
reducing the friction and wear has been addressed by Sun [9]. A fuzzy self-tuning
proportion-integration-derivative controller has been used to achieve the desired tip
position and the influence of clearance and lubrication of journal on the precision of
the manipulator has been demonstrated. A rigid-flexible two-link manipulator incor-
porating a point mass payload undergoing large deformation has been considered for
vibration suppression using an optimal trajectory planning technique by Abe [10].
Abe and Hashimobo [11] considered the point-to-point motion of two flexible links
Analysis of Parametric Influence on Control … 1279
The position vector of general point and end point on the link is given as:
(xi ,yi ) (xi+1 ,yi+1 )
x x
Ri = , and Ri+1 = Ri (L , t) + . (2)
(vi + y) vi+1 (L , t)
The kinetic energy (T ) and potential energy of the system (U) are expressed as:
Fig. 1 A planar two-link flexible robotic manipulator connected with revolute pair
Analysis of Parametric Influence on Control … 1281
⎡ L ⎤
i
n
T = 0.5 ⎣ ρi RitT Rit dx (x,t) + m i RitT Rit (L ,t) + Ji δit + v(i−1)xt (L , t) ⎦.
2
i=1 0
(3)
⎡ L ⎤
n i
U = 0.5 ⎣ E i Ii (vi x x )2 dx ⎦. (4)
i=1 0
The external torques τ1,2 applied
at the joints and the non-conservative work
done by structural damping c1,2 in the links are:
2
δWnc = 0.5 c1 v1t + c2 v2t
2
(5)
L 1
ρ1 A1 xv1tt + x 2 δ1tt + v12 δ1tt + 2v1 v1t δ1t dx
0
+ (m 1 + ρ2 A2 L 2 ) L 1 v1tt + L 21 δ1tt + v12 δ1tt + 2v1L v1t δ1t (L , t)
L 2
2v2 v2t δ1t + 2v2 v2t δ2t + 2v2 v2t v1xt + v22 δ1tt + v22 δ2tt
+ ρ2 A 2 dx
+v2 δ1t v1xt + xv2tt + x δ1tt + x δ2tt + x v1xtt
2 2 2 2
0
2v1 v1t δ1t + v12 δ1tt + L 1 v1tt + L 21 δ1tt + 2v2 v2t δ1t + 2v2 v2t δ2t
+ mP
+2v2L v2t v1xt + v22 δ1tt + v22 δ2tt + u 22 v1xtt + L 2 v2tt + L 22 δ1tt + L 22 δ2tt + L 22 v1xtt (L ,t)
+ J1 δ1tt + J2 {δ1tt + δ2tt + v1xtt (L , t)} = τ1 . (7)
L 2
2v2 v2t δ1t + 2v2 v2t δ2t + 2v2 v2t v1xt + v22 δ1tt + v22 δ2tt
ρ2 A2 dx
+v22 δ1t v1xt + xv2tt + x 2 δ1tt + x 2 δ2tt + x 2 v1xtt
0
2v2 v2t δ1t + 2v2 v2t δ2t + 2v2 v2t v1xt + v22 δ1tt + v22 δ2tt
+ mP
+v22 v1xtt + L 2 v2tt + L 22 δ1tt + L 22 δ2tt + L 22 u 1xtt (L ,t)
1282 P. Kumar and B. Pratiher
The governing equations of motion for the links are expressed as:
ρ1 A1 v1tt + xδ1tt − v1 δ1t
2
+ c1 v1t + E 1 I1 v1x x x x = 0. (9)
v2tt + xv1xtt (L , t) + xδ1tt + xδ2tt − v2 δ1t2
− v2 δ2t
2
− v2 v1xt
2
(L , t)
ρ2 A2
−2v2 δ1t δ2t − 2v2 δ2t v1xt (L , t) − 2v2 δ1t v1xt (L , t)
+ c1 v1t + E 2 I2 v2x x x x = 0. (10)
In this section, the derivation of the output and state-equations composed of rigid
body rotation of joints and flexible motion of the links of the manipulator is realized
to demonstrate the influence of important system variables on the control parameters
specifically on the angular
tip positions, modal displacements, and input torques.
Now, the deflection u 1,2 of a point located at a distance x along the links of the
manipulator using assumed mode method is expressed here as:
Here, q1 (t), and q2 (t) are the modal displacements of first and second link, respec-
tively; φ1 (x), and φ2 (x) are the eigenfunction of first and second link for first mode of
vibration, respectively, derived by Kumar and Pratiher [13] for rigid joint conditions
and expressed here as:
φ1 (x̄) = G 11 cos γ̄ 1 x̄ − cosh γ̄ 1 x̄ + G 12 sin γ̄ 1 x̄ + G 14 sinh γ̄ 1 x̄ . (12)
φ2 (x̄) = H11 cos ξ γ̄ 1 x̄ + H21 sin ξ γ̄ 1 x̄ + H31 cosh ξ γ̄ 1 x̄ + H41 sinh ξ γ̄ 1 x̄ .
(13)
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
q1tt q1t ζ1 q1,2 , δ1,2t 0
⎢ q2tt ⎥ ⎢ q2t ⎥ ⎢ ζ2 q1,2 , δ1,2t ⎥ ⎢ 0 ⎥
M q1,2 ⎢ ⎥ ⎢ ⎥ ⎢
⎣ δ1tt ⎦ + η q1,2t , δ1,2t ⎣ δ1t ⎦ + ⎣
⎥ = ⎢ ⎥.
⎦ ⎣ τ1 ⎦ (14)
0
δ2tt δ2t 0 τ2
Here, M q1,2 is the mass matrix consisting of rigid and flexible motion
components,
η q1,2t , δ1,2t is the Coriolis and centrifugal component matrix,
ζ q1,2 , δ1,2t is the stiffness matrix, and last matrix column in Eq. (14) repre-
sents the force matrix containing the torque exerted at the joints of the manipulator.
Now, the open input loop torques are calculated, and those are capable of replicating
the given trajectories for both the joints (δ1d , δ2d ) of the manipulator. Firstly for the
required given trajectories, the resultant modal deflections (q1d , q2d ) are calculated
by numerical integration of governing equations of flexible manipulator dynamics
from of Eq. (14) for zero initial conditions. Now, the required input torques (τ1d , τ2d )
of the system are obtained for the rigid body motion of the manipulator from the
inverse dynamics of Eq. (14) by using the desired trajectory (δ1d , δ2d ) and computed
modal deflections (q1d , q2d ). In order to add robustness to the system, a linear feed-
back proportional-derivative (PD) controller on the joint trajectory error is used and
expressed as:
τinput = τd (qd , q̇d , ϕ̇d , ϕ̈d ) + 1 D K p (ϕd − ϕ) + K D (ϕ̇d − ϕ̇) (15)
Here, K p and K D are the proportional and derivative gains, respectively, and can
be selected suitably so that the poles of the linearized system
of Eq. (14) are in left
half of s plane. When hubs are driven by the torques τinput given in Eq. (15), the
desired trajectories of the joints are obtained.
The procedure of inverse dynamics of the system is exploited to design and deter-
mine the open-loop torque accounting of the rigid body motion of the joints for the
accurate positioning of the links of manipulator. Here to conduct the simulations,
the geometrical and physical characteristics for both the links have been considered
as b1,2 = 0.05 m, h 1,2 = 0.0055 m, ρ1,2 = 7800 kg/m3 , E 1,2 = 2.1 × 1011 N/m2 ,
m 1 = 0.1 kg, m 2 = 0.1 kg, J1,2 = 0.02 kgm2 , and L 1,2 = 0.5 m. The K p and K D
are taken as (100, 150) and (500, 600), respectively. A sinusoidal torque profile is
used to position the first link at 60° and second link at 30°, respectively, in 4 s. The
influence of payload on the angular position, input torque, and modal displacement
is shown in Fig. 2. It can be observed that both the links achieve the desired posi-
tions and the error (of order 0.001) reduces with the increase of payload mass from
no payload condition to 0.5 kg. The required input torque and the amplitude of the
modal displacements also increase significantly with the increase in payload mass.
1284 P. Kumar and B. Pratiher
Fig. 2 Effect of variation of payload mass on the input torques, angular positions, and modal
displacements
It can also be observed that the tip have small amount of residual vibrations even
after the required torque is removed. Hence, while lifting, an increased amount of
payload will require increased amount of torque and the end point vibrations have
to be suppressed for accurate positioning. A negligible effect on the input torque of
the second hub is noticed with the increase in second joint mass as shown in Fig. 3;
however, the torque input of the first joint increases due to the increase in overall
inertia of the system driven by first joint.
The increase joint inertias increase the power consumption of the actuators due to
increased input torques which is shown in Fig. 4, and it also increases the amplitude
of residual vibrations of the end effector which may result in increased error trajec-
tory tracking of the manipulator. The end effector vibrates at significantly higher
amplitudes for larger joint inertias after reaching the desired set angular positions.
The increase in length of the first link increases the required amount of torque for
both the joints; however, the rigid motion of the end point of the second link increases
Analysis of Parametric Influence on Control … 1285
Fig. 4 Effect of variation of joint inertia on the input torques and modal deflections
significantly which is evident from Fig. 5. While the end point vibrations of the first
link damp down after some time for the larger link length, the tip of the second link
keeps on oscillating for a larger duration. It is observed from Fig. 6 that, as the mate-
rial of the links is changed from steel to aluminum, the input torque reduces, but the
tip of the manipulator links vibrates with large amplitudes. The residual vibrations
are prominent in case of the aluminum. Thus, the aluminum shall reduce the power
consumption of the manipulator system significantly, but the design engineer will
have to reduce the end point vibrations for accurate trajectory tracking.
1286 P. Kumar and B. Pratiher
Fig. 5 Effect of variation of first link length on the input torques and modal deflections
Fig. 6 Effect of variation of material of the links on the input torques and modal displacements
Analysis of Parametric Influence on Control … 1287
5 Conclusions
References
10. Abe A (2009) Trajectory planning for residual vibration suppression of a two-link rigid-flexible
manipulator considering large deformation. Mech Mach Theo 44(9):1627–1639
11. Abe A, Hashimobo K (2015) A novel feedforward control technique for a flexible dual
manipulator. Robot Comput-Integr Manuf 35:169–177
12. Ashayeri A, Eghtead M, Farid M (2008) Trajectory tracking for two-link flexible arm via
two-time scale and boundary control methods. ASME. International mechanical engineering
congress and exposition, October 31 to November 6 (2008)
13. Bai M, Zhou DH, Schwarz H (1998) Adaptive augmented state feedback control for an
experimental planar two-link flexible manipulator. IEEE Trans Robot Autom 14(6)
14. Kumar P, Pratiher B (2018) Nonlinear modeling and vibration analysis of a two-link flexible
manipulator coupled with harmonically driven flexible joints. Mech Mach Theo 131:278–299
15. Lochan K, Roy BK (2015) Position control of two-link flexible manipulator using low chattering
SMC techniques. IJCTA 8(3):1137–1145
16. Oh S, Kong K (2015) Two-degree-of-freedom control of a two-link manipulator in the rotating
coordinate system. IEEE Trans Ind Electron 62(9)
17. Ozer A, Semercigil SE, Kumar RP, Yowat P (2013) Delaying tool chatter in turning with a
two-link robotic arm. J Sound Vib 332(6):1405–1417
18. Perez PJ, Perez JP, Soto R, Flores A, Rodriguez F, Meza JL (2012) Trajectory tracking error
using PID control law for two-link robot manipulator via adaptive neural networks. Proced
Technol 3:139–146
An Assistive Chair Using a Series-Elastic
Actuator
Abstract Series-elastic actuators are mechatronic devices used for force control.
They consist of an actuator in series with a spring in conjunction with a sensor
for measuring the actuator displacement and another for spring displacement. A
controller helps deliver required force profile by driving the actuator in a feedback
mode as per the measured displacements and the applied load. We present design,
prototyping, and testing of a series-elastic actuator integrated with a slider-rocker
linkage. This device is retrofitted to a chair such that the user is assisted while sitting
and rising. Multi-body dynamics using Simscape and control system design are
discussed.
1 Introduction
The elderly often have difficulty in the routine activity of sitting in and rising from a
chair. This could be attributed to their inability to bear excessive loads on the knees
and hip. It is, therefore, beneficial to have an assistive appendage in the chair to
provide support and actuation, which is otherwise provided by a caretaker. An all-
mechanical device of that kind was developed by our group [1, 2]. It had an inclined
seat that can rotate about a virtual axis using a compliant hinge mechanism with the
Y. Agrawal · G. K. Ananthasuresh
Department of Mechanical Engineering, IISc Bengaluru, Bengaluru, Karnataka, India
e-mail: [email protected]
G. K. Ananthasuresh
e-mail: [email protected]
A. Kushwaha (B) · S. Khandai · K. V. S. Hari
Department of Electrical and Communication Engineering, IISc Bengaluru, Bengaluru,
Karnataka, India
e-mail: [email protected]
K. V. S. Hari
e-mail: [email protected]
© Springer Nature Singapore Pte Ltd. 2022 1289
R. Kumar et al. (eds.), Machines, Mechanism and Robotics, Lecture Notes in Mechanical
Engineering, https://doi.org/10.1007/978-981-16-0550-5_122
1290 A. Kushwaha et al.
Fig. 1 a Prototype of actuated assistive chair with series-elastic actuator (SEA), b Torque profiles
to be customized for 60 kg user, and c SEA device built in this work
help of open-section shells that bend and twist simultaneously to provide required
torque-angle profiles. A pair of cams, wherein one decides the torque-angle profile
for sitting in and another for rising, needs to be customized for individual users. To
enable such customization, we have designed an actuated chair and a multi-body
dynamics simulation programme. The actuated chair uses a series-elastic actuator
(SEA) in a slider-rocker linkage. This is the focus of this paper (Fig. 1).
The slider-rocker linkage, which connects the SEA to the chair frame and the
seat, is designed such that the SEA does not rotate much w.r.t. the frame. This
ensures that the dynamics of SEA are decoupled from the dynamics of the linkage,
by neglecting the rotations altogether. The stroke, dead length, and overall size of the
SEA are designed accordingly. The torque-profile, is converted to a displacement-
profile using relations pertaining to the linkage.
The SEA comprises an actuator element (motor and lead screw) assembled in
series with springs, employing two feedback sensors. This is used to provide a vari-
able force output at each stroke value, which, via a single slider-rocker linkage, is
translated to the chair seat. As a concept, the SEA has existed since 1995 [3], finding
applications in various natural tasks of mechatronic systems as well as addressing
the problems involving failure of power transmission components due to shocks. It is
amenable to force control, as shown by [4–6]. The device works by reducing a force
control problem, (where you move the actuator when force exceeds some value),
to a position control problem, (where you move the actuator until some position
is reached). The springs are assembled in series with the actuator, such that force
felt at the end is commensurate with the deformation in the springs. We control this
deformation by controlling the position of the actuator, such that at each stroke, this
deformation is equal to a reference value generated from the requisite torque-profile.
Feedback of the actuator stroke and device stroke is taken from a twin optical-
incremental encoder set, positioned at the seat and the motor. The difference between
reference value and measured value is fed as error to the PID controller, imple-
mented in the micro-controller. The nominal model of the system is obtained from
An Assistive Chair Using a Series-Elastic Actuator 1291
experiments on a manufactured prototype and used to find the PID parameters. This
completes the design stage.
The multi-body dynamic simulation programme models the linkage and the user
limbs whose parameters (weight, lengths of limbs) are measured. Customization of
our mechanical chair comprises three steps: (i) simulation with user parameters to
see what torque-angle profile helps in suitable sitting and rising; (ii) real experience
of the user with profile using the actuated chair and tuning of the profile based on
that; and (iii) designing the pair of cams as per the profile that user likes (see [1] for
details).
The remainder of the paper is organized as follows. The mechanical design aspects
of the device are discussed in Sect. 2. The relations needed for motor selection and
dynamics equation for the device are presented in Sect. 3. Control system design is
presented in Sect. 4. Electronics implementation is described in Sect. 5. Multi-body
dynamic simulations are presented in Sect. 6. After completing tasks of all sections
a protoype is operated for a test user and the results are analyzed in Sect. 7. Section 8
is closure.
2 Series-Elastic Actuator
Amplification of force from the actuator is of essence because the device has to
support users in wide range of weights. We choose to go with a power screw since they
are easy to obtain and by nature provide force amplification. This is in conjunction
with compression springs attached in series and two rotary encoders, one at the motor
end and one at the seat. The device has one revolute joint with the seat and one with
the chair frame in a slider-rocker mechanism.
Ball screws are the ideal choice as power screws since they provide low frictional
resistance to motion and low noise with minimal backlash. Because of the manner in
which the SEA-device is mounted on the seat, the ball screw will be in compressive
load at all times of operation. When the seat is being raised, the torque applied by the
motor will be in the same direction as the motion, extracting power from the power
supply. When the seat is lowered, i.e., when the occupant is sitting, the torque applied
on the motor is in the direction of motion. This makes the motor run in a generative
1292 A. Kushwaha et al.
Fig. 2 SEA—Components
modes, with higher than ordinary currents which are not desirable. To counter this,
we use a lead screw, which relies on the sliding contact between surfaces of the
screw and the nut to transmit motion and forces, having a self-locking feature. This
happens when the lead angle is lower than the friction angle, making it necessary
to apply torque on the lead screw, in the direction of motion. Consequently, the
generative mode of motor is never activated. This comes with a trade-off in the form
of requirement of higher power from the motor, due to lower mechanical efficiency
of the transmission of a lead screw as compared to a ball screw. Some relations and
parameters that help in the selection of the motor are presented in the next section.
Figure 2 shows the manufactured prototype with major components labeled. We
have used a DC motor, which is coupled to an optical incremental rotary encoder.
Using a 4:1 timing belt-pulley drive, the motion is transferred to a lead screw, with
kinematical accuracy and high transmission efficiency. This lead screw is supported
by angular contact ball bearings on both sides. A lead nut moves on the lead screw,
guided on two guide rods on either side. This is connected on another part using con-
necting plates. The end-effector and compression springs are assembled on additional
guide pins. This is all mounted appropriately on a base plate.
Figure 3 shows the SEA device, mounted on the chair frame along with its geom-
etry. We have two feedback encoders, one measuring the seat angle, from which
the device stroke could be derived, and one measuring the stroke of the lead nut.
The output from these can be processed to give the instantaneous compression in
the spring, which is proportional to force felt at the seat, provided that we neglect
An Assistive Chair Using a Series-Elastic Actuator 1293
friction, backlash, and inertial forces. We convert the required torque profile to a
required spring deformation profile as a specified function of angle of the seat. Using
this force-control could be reduced to a position-control problem, i.e., maintaining
the required spring compression at measured seat angle. This position-control can be
achieved by implementing a proportional-integral-derivative (PID) control scheme.
This section will serve as the derivation for open-loop dynamics governing equations.
If not stated otherwise, the symbols have their usual meanings.
3.1 Kinematics
Figure 3 shows the geometry at a general angle of the chair seat, θ . Here, s is the
length of the device, l, h and r are the indicated lengths. Using geometrical relations
we can write
s = l 2 + h 2 + r 2 + 2r (h sin θ − l cos θ ) (1)
Tr eq s
xr = (3)
K r (h cos θ + l sin θ )
1294 A. Kushwaha et al.
We can find the torque exerted on the seat by a force equal to K x, acting along the
device . Since we wish to control this torque, if we equate the two we can find the
spring compression to be maintained at that θ , denoted by xr . In this expression, Tr eq
is the requisite torque felt at the seat, which is a function of θ .
Treq s θ̇ + Tr eq ṡ K r θ̇(h sin θ − l cos θ ) 2
x˙r = xr + xr (4)
Tr eq s Tr eq s
Here, () represents differentiation with respect to θ . Using these relations, we find
the angular speed of the motor(N ) as,
|(x˙r + ṡ)|
N =γ = φ̇ (5)
p
Preceding relations determine the motor speed as well as necessary variables for the
calculation of errors.
3.2 Dynamics
3.2.1 Leadscrew
Lead screws have different impedance for motion in both the directions. Also, they
could be designed to have a self-locking feature such that any amount of axial force
is unable to move the lead nut, which is used in our application. The expressions for
torque required to turn the lead screw during extension and retraction are:
dm
T = (K x + Meq (ẍ + s̈)) β (6)
2
μ cos λ + sin λ
β= f or extension (7)
cos λ − μ sin λ
μ cos λ − sin λ
β= f or r etraction (8)
cos λ + μ sin λ
It is important to note that β > 0 for both the cases and greater for the extension
case.
At this stage, we have enough data to select the motor. These relations were
simulated on an Excel sheet where the dimensions, torque and seat RPM, etc. were
input parameters and motor torque, angular velocity, and power were the output
parameters. The spring stiffness was taken such that 20 mm of compression produces
1000 N of force, enough to produce approximately 175 Nm of torque at the seat hinge.
We take a safety factor of 1.3 to take into account inertial effects and unwanted friction
An Assistive Chair Using a Series-Elastic Actuator 1295
at other locations. The motor selected and procured has a power output of 350 W at
rated RPM of 2750 RPM and provides 1.1 Nm rated torque. This motor, with a 4:1
ratio timing pulley-belt transmission, drives the lead screw. The torque applied by
the motor comes out to be
T
Text = (9)
γη
where η is the transmission efficiency, which is 95% to 98% for timing belt drive
systems. Note that we do not consider the inertial forces due to the slow rotation
(≈ 0.01 rad/s) of the device w.r.t. the pivot on the chair frame.
3.2.2 DC Motor
Using preceding relations, we can find the governing equations for open loop dynam-
ics. We start with the model of the motor.
The second term in the preceding equation needs elaboration. Since the value of L
comes out to be very small, it is possible to completely neglect the term, hence we do
˙ . Further manipulation yields a second order equation
not need an expression for Text
in x,
α1 ẍ + α2 ẋ + α3 x = Vin − (α2 ṡ + α1 s̈) (13)
where
R Jeq γ Rγ Meq dm β
α1 = + (14)
κp 2κη
Beq R γ
α2 = +κ (15)
κ p
β Rγ dm K
α3 = (16)
2κη
1296 A. Kushwaha et al.
4 Control System
The variable, Vin , is the input variable while x is our output variable and ṡ is a
disturbance. We have an additional constraint that x has to follow a profile w.r.t. s,
i.e., it is a function of s. The nature of s depends on physical properties of the user
like mass, limb lengths, weight distribution, and the effort applied from joints, which
change from person to person. The aim of this analysis is to provide a robust control
system such that effects of s are filtered out using PID control.
The relation for reference signal xr for a selected profile is given by Eq. (3). Following
relations are needed to calculate the actual values of the output variable and its
derivative:
pφm 2
xm = − l + h 2 + r 2 + 2r (h sin θm − l cos θm ) (17)
2π γ
The equations for the transfer function model for the system are given by
KC α2
α1
(τ s + 1) s2 + α1
s
X (s) = X r (s) − M(s) (19)
(s) (s)
KD 2 α3 + K C KC
(s) = s 3 + α2 + s + s+ (20)
α1 α1 α1 τ
where reference signal X r (s) is calculated from Eq. (3). The control system block
diagram is shown in Fig. 4. We use a PID control scheme with inner loop rate feedback
to filter out effects of sudden jumps in the reference signal. K C , K D , τ are the PID
An Assistive Chair Using a Series-Elastic Actuator 1297
parameters that need to be tuned to achieve a robust control. M(s) is the instantaneous
velocity of the device, which is to be treated as an unknown disturbance.
The characteristic polynomial ( (s)) dictates dynamics of the system along with
many performance parameters such as steady-state performance, settling time, and
peak overshoot. The aim of PID control is to mould the characteristic polynomial to
the following form by obtaining the appropriate PID parameters from the nominal
model of the system.
K C = α1 (ωn2 + 2ζ ωn a) − α3 (22)
K D = α1 (a + 2ζ ωn ) − α2 (23)
KC
τ= (24)
α1 ωn2 a
Although we have an estimate for the nominal model of the system from the constants,
we conducted experiments where the system was excited by step input of voltage and
the angular velocity was measured. The system model was estimated from that to get
experimental values of α1 and α2 . Next, we tested the system for various values of ωn
to find the one with the best performance. The results are discussed in the following
sections.
1298 A. Kushwaha et al.
The electronics sub-system is a critical part of the device and its architecture is
shown in Fig. 5. It has (a) microcontroller, (b) encoders, (c) motor driver, (d) SMPS,
(e) motor, and a (f) limit switch.
The encoders are the most critical parts. The DC motors have commutator brushes
that produce sparks everytime they lose contact and the environment is exuded with
the electromagnetic interference caused by this phenomena. Encoders, via the wires
that carry signals get affected by this noise. So we need to ensure their fidelity.
The twin incremental optical encoder set are used to measure instantaneous stroke,
compression, and velocities given by equations (1) and (2). They are procured from
Autonics and have a resolution rating of 2500 and 360 pulse per revolution respec-
tively. They are provided with four channels, A, A.B, B. Here A and B provide
square-pulse form output. A and B are 90◦ out of phase from each other. A and B
are the inverted forms of the respective signals, used for noise immunity. The four
form a differential line driver signal with IC AM26C32CN used as the reciever at
the microcontroller end.
The microcontroller (TM4C123GH6PM, Texas Instruments, 80 MHz clock fre-
quency, 32 KB RAM) interfaces with encoders, limit switches, and the motor, imple-
menting the control algorithm. This microcontroller has a quadrature encoder module
which is used to decode the signals from a differential line reciever. The direction of
rotation is determined by current and previous states of channel A and B, done within
the microcontroller. A counter is present to either increment or decrement on every
rising and falling edge of each channel. In clockwise rotation the counter increments
and vice versa in counter-clockwise rotation. Since four edges are present per pulse,
the resolution is four times that of the resolution rating. Thus, for the motor encoder,
each rotation is quantised in 1440 parts and provides us with a 1.4 μm resolution.
Similarly, for the seat encoder, we have 0.036◦ resolution. They provide instantaneous
angular velocities calculated from the time duration between subsequent edges.
We use a 350 W brushed-permanent magnet DC motor with rated current of 19.2 A,
rated torque of 1.1 Nm, rated voltage of 24 V, and rated angular speed of 2750 RPM.
This is a close fit for the motor required for the purpose of our device, as discussed
in Sect. 3.2.1. H-bridge (Cytron) is used to control the motor with rating of 24 V
and 20 A, which matches the output from the SMPS, ensuring that current remains
limited to the continuous operating range of the motor. The motor is driven by 24 V
Pulse Width Modulation (PWM) signals with frequency of 10 kHz while the voltage
is varied by modulating the duty cycle.
We have isolated the user part from the device such that the user interacts with torque
profiles as a passive nonlinear spring. These torque profiles need to be optimized for
classes of people such that user comfort is maximized. We can categorize people on
the basis of their overall mass and their Body Mass Index and test different torque
profiles to identify which profiles cater to different categories of users. Primarily, the
effort spent by a person in sitting can be characterized by the torque at the knee joint.
If the user’s weight is low, during sitting, some torque profiles would be too stiff
and he/she would stop at the point of static equilibrium, unable to reach the sitting
posture. Similar condition could occur for an over-weight person with a torque profile
intended for lower weight class while rising. In this section, we present the use of
multi-body dynamics simulations to ensure that appropriate profile is operated for a
user.
The multi-body dynamic (MBD) analysis is done using Simscape toolbox on
Matlab. For simplicity, the human body is divided into three parts: shank with the
foot, thigh, and torso including neck and head, all connected with revolute joints,
in a 2-D model. The proportions of lengths and weights w.r.t. to the total height
and mass are taken from [8], given in Table 1. In a three-link link model, the center
of mass of each link is considered at its center. The seat is taken as a ternary link
with dimensions identical with the actual one. The interaction between the seat and
human, through the thigh, is defined by a sub-system consisting of transform sensor
block, which uses a contact model. The foot is fixed to the ground.
MBD analysis is the first step to ensure that torque profiles serve their intended
purpose. In the simulation experiments, we perform a test to ensure that the person
would be able to completely sit using the bespoke profile and similarly if the person
is able to stand without using any of his/her own muscles.
In the first case, while sitting, we allow the user to sit due to gravity alone, with
torque-free conditions at the knee joint. While rising, the occupant rises in similar
1300 A. Kushwaha et al.
Table 1 Mean limb lengths and weight distribution. H is the total height and W is the total weight
Limb Description Limb Length (H) Limb Weight (W)
Lower Leg 0.285 0.11
Thigh 0.245 0.22
Upper Body 0.470 0.67
Fig. 6 Different configurations the user takes while operation. While sitting the sequence is forward
and while rising the sequence is backward
conditions. In the second case, we specify a finite movement in both hip and knee
joints while sitting and compare the results with and without external torque profiles.
We wish to emulate the first case with real users (Fig. 6).
7 Device Operation
We have implemented the discussed control scheme of the prototype. Using MBD
analysis, we can get a suitable torque profile for a person. This is important since we
have a limit of current and voltage the motor could be safely subjected to. At any θ ,
as the force applied exceeds the reference value, the motor adjusts the lead nut by
applying some voltage according to the PID control law. Also, when a higher force is
applied, the motor will run in a saturated state, the maximum voltage will be applied.
Further, this device could be programmed to follow different force profiles for sitting
in and rising. This device is tested on a person weighing 60 kg and 5 feet 10 inches
tall. The reference torque profile versus actual torque profile obtained from the test
is plotted in subsequent figures, one for rising and one for sitting in. As expected,
during rising, actual value remains below the reference value and during sitting, the
reference value is below actual value; it follows with acceptable accuracy. Now, we
are able to provide the subjects with a variety of torque profiles to choose from and
thereby identify the most comfortable one for her/him (Fig. 7).
An Assistive Chair Using a Series-Elastic Actuator 1301
8 Closure
Acknowledgements This project was supported by the Technology Initiative for the Disabled and
Elderly (TIDE) programme of the Department of Science and Technology, Government of India.
We would like to acknowledge contributions from Dhananjay Yadav in making the Simscape model
for MBD analysis and construction of a previous prototype which served as an inspiration for this
idea. We also thank Navin Engineering, Bengaluru for assistance in manufacturing of the prototype.
References
1. Hampali S, Pai AS, Ananthasuresh GK (2019) An open-section shell designed for customized
bending and twisting to ease sitting and rising in a chair. In: Badodkar D, Dwarakanath T (eds)
iNaCoMM 2017, Machines, Mechanism and Robotics. Lecture Notes in Mechanical Engineer-
ing. Springer, Singapore, pp 427–439
2. Sarojini D, Lassche TJ, Herder JL, Ananthasuresh GK (2016) Statically balanced compliant
two-port bistable mechanism. Mech Mach Theor 102:1–13
3. Pratt GA, Williamson MM (1995) Series elastic actuators. In: Proceedings 1995 IEEE/RSJ inter-
national conference on intelligent robots and systems. Human Robot Interaction and Cooperative
Robots, IEEE
1302 A. Kushwaha et al.
4. Paine N, Oh S, Sentis L (2014) Design and control considerations for high-performance series
elastic actuators. IEEE/ASME Trans Mechatronics 19(3):1080–1091 (IEEE)
5. Kong K, Bae J, Masayushi T (2009) Control of rotary series elastic actuator for ideal force-mode
actuation in human–robot interaction applications. IEEE/ASME Trans Mechatronics 14(1):105–
118 (IEEE). https://doi.org/10.1109/TMECH.2008.2004561
6. Kong K, Bae J, Masayushi T (2011) A compact rotary series elastic actuator for human assis-
tive systems. IEEE/ASME Trans Mechatronics 17(2):288–297 (IEEE). https://doi.org/10.1109/
TMECH.2010.2100046
7. Ogata K (2010) Modern control theory, 5th edn. Prentice Hall, One Lake Street, Upper Saddle
River, New Jersey 07458
8. Anthropometric Data. https://www.ele.uri.edu/faculty/vetter/BME207/anthropometric-data.
pdf. Last Accessed 19 Jul 2019
Natural Control of Virtual Models
of Mechanisms Using Leap Motion
for Interactive Learning
1 Introduction
The study of planar and spatial mechanisms is essential in every mechanical engi-
neering curriculum due to their extensive use in the industry. A prime example
is the slider–crank mechanism that forms the basis for all reciprocating engines.
Throughout the history of mechanical engineering, institutions have continually
relied on diagrams in textbooks to teach complex mechanisms to students. However,
this method of instruction is often ineffective in inculcating a thorough understanding
of mechanisms [1]. Now, with the advent of technology, video lectures and tutorials
have gained popularity in educating students in the workings of intricate mechanisms.
Nonetheless, both of these methods create a one-way communication environment,
wherein the students simply consume information presented to them, without a prac-
tical understanding. Research shows that interactive learning (i.e., based on two-way
communication) environments bolster learning in students [2].
to fully define the configuration of the system. Hence, one independent variable is
present that can be controlled. In both the cases, the orientation of the crank link,
i.e., crank angle, can be considered as the controlled variable. Inventor allows user
to grasp a link through mouse and move it, and its “Assembly Constraint Manager”
allows the corresponding movement of the other links, in an interactive fashion.
However, if one has to control a higher DOF mechanism, like a five-bar mechanism,
mouse grasp does not deliver the required behaviour. To overcome this inability, one
can use “Angle Constraint” as explained next.
For a virtual mechanism (say given in Fig. 2), the crank angle can be constrained
using “Angle Constraint” in Inventor. Changing the angle of the constraint updates
the angle in the assembly, and the “Assembly Constraint Manager” of the software
moves the other links as per the kinematic constraints. The “Angle Constraint” can
play a major role in the simulation of mechanisms with higher DOF. For example, in a
five-bar mechanism, two revolute joints can be assigned with “Angle Constraint” and
updating their value, the mechanism gets updated. Figure 2 illustrates two config-
urations of a five-bar mechanism that has different joint angles in its two Angle
Constraints. Thus, the user does not have to solve the kinematic equations of a
complicated mechanism, which need not have closed-form solutions, to know the
position and orientation of the other links in the given mechanism. However, this
Autodesk Inventor exposes its Application Programming Interface (API) that can be
accessed from other software applications. Through this, the user can access different
functions that allow them to create parts and assemblies, constrain them, change their
position and orientation, etc. Almost all operations that can be performed with the
Graphical User Interface (GUI) of the software can be executed by other applications,
known as plugins or addins.
The angles of the “Angle Constraints” in an assembly can also be modified by an
addin. One such application was developed by Chittawadigi et al. to determine the
Denavit–Hartenberg (DH) parameters from the CAD model of serial robot [7], where
the Angle Constraints of a robot were modified, one joint at a time, and the position of
the end-effector (last link) was determined using the application, programmatically.
It acted as a validation for the research methodology proposed in that reported work.
In this work, a similar approach is used, as explained next.
In this paper, Leap Motion sensor has been integrated with Autodesk Inventor soft-
ware. A client application, referred to as addin, has been developed using Visual C#.
It has a reference to Leap Motion Controller, an interface to read the data received
from the Leap Motion device. The addin tracks the index finger of the user, as illus-
trated in Fig. 3a. The coordinates of the fingertip are measured for every frame of
image capture by the device. These measured values are with respect to the coor-
dinate system at centre of the Leap Motion device. However, to obtain meaningful
coordinates with regard to controlling the crank angle of a virtual mechanism, an alter-
nate coordinate system is assumed 200 mm vertically above the original coordinate
system, as shown in Fig. 3b. Therefore, the tracking of finger is done with respect
to the alternate coordinate system and the coordinates of the fingertip are deter-
mined. Based on the type of simulation, as explained next, the Angle Constraint(s)
value(s) is/are updated in Autodesk Inventor through the API. Thus, the integration
of Leap Motion and Autodesk Inventor is completed for an effective visualization of
mechanism motion.
Natural Control of Virtual Models of Mechanisms Using Leap … 1307
(a) Leap Motion tracking user’s fingertip (b) Alternate coordinate system
(a) Position of the finger tip (b) Rotation of crank angle due to finger movement
mechanism which has one Angle Constraint, whose angle can be modified based on
the user input over Leap Motion.
Five-bar mechanism has two DOFs, and hence two Angle Constraints can be defined
at the joints that connect the two links connected to the grounded link. Since it has
two DOFs in a plane, its end-effector (position of the joint located farthest from the
grounded link) can be used to trace any shape on a 2D plane. Inverse kinematics
deals with the determination of input joint angles required to reach to the desired
end-effector position.
The user moves his/her fingertip over the Leap Motion device, and the coordinates
in the alternate coordinate system are measured. The X and Y coordinates measured
are supplied as input to an inverse kinematics solver, developed based on the formu-
lations discussed in [3]. The solver returns back the corresponding joint angles to
be used in the active joints, thus moving the mechanism such that the end-effector
point reaches the desired point measured by the device. The end-effector’s position
in the Inventor environment is recorded over time, and a sketch is made to connect
each consecutive point. An illustration of the positions of the fingertip and the corre-
sponding configurations of the five-bar mechanism in Inventor is shown in Fig. 5a,
b, respectively.
Leap Motion can also be used to control physical prototype of a mechanism. A
similar attempt is made by the second author in [8], where a 3 DOF translational
DOF Delta Manipulator is controlled using Leap Motion as input.
(a) Position of the finger tip (b) End-effector following the path traced by finger
4 Conclusions
References
Abstract Calibration of material model parameters for the validation of spine finite
element (FE) model involves the tuning of parameters in multiple spinal compo-
nents. A method to automate calibration of spine kinematics was developed in
this study, which will find application in subject-specific model development and
biomimetic mechanisms. Downhill simplex method was used to find the optimum
value of twenty-four material parameters in a single motion segment FE model of
cervical spine. The model consisted of two vertebrae and the disc and ligaments in
between them. Flexion and extension loading cases and combined flexion-extension
loading were considered for the automated model calibration. It was seen from the
study that a combined consideration of the loading cases is required to obtain a
model calibrated in multiple loading directions as spinal components have different
contributions in different loading conditions.
1 Introduction
Spinal column consists of alternating hard and soft tissue structure that facilitates
movement and transfer of loads in the human body. The complex-shaped hard tissues
(vertebrae) articulate with each other through intervertebral soft tissues (disc and
ligaments) and contact surfaces to produce distinct kinematics in different directions
like flexion-extension, lateral bending and axial rotation as shown in Fig. 1. Finite
element (FE) modelling has become an indispensable tool for understanding the
mechanics of spine structure [1]. Not only does FE models allow researchers to
study complex loading conditions that are not easily feasible in in vitro experiments,
but also is useful for the design of medical devices [2, 3].
One of the main steps of FE model development is the validation of the model
responses with experimental responses. In the case of spine model, this step is partic-
ularly challenging because of the number of components in the spinal structure [4].
Most of the spine model developments utilize manual tuning of the material parame-
ters to obtain validated kinematic responses. This is an arduous and time-consuming
task [5, 6]. An automated method for calibration will be required for rapid develop-
ment of spine models. The objective of this study was to develop an optimization
framework to automate material parameter selection, with a focus on tuning the spine
motion segment kinematics.
2 Methods
A validated FE model of single cervical spine motion segment was adapted for
geometry and mesh size in this study [7]. The model consisted of two vertebrae and
the soft tissues (discs and ligaments) corresponding to C5–C6 spine level (Fig. 2).
The trabecular bone in the vertebrae was defined as linear elastic material (E =
400 MPa, ν = 0.29). The intervertebral disc (IVD) is an important component that
influences the spine kinematics along with the ligaments and thus a detailed model
of IVD as described in was used [8]. The disc was modelled as two separate regions:
annulus and nucleus. Annulus fibres were modelled as rebar layers within the annulus
elements. Because of the differences in orientation of annulus fibres between the
anterior (forward) region and posterior (rearward) region, the annulus was modelled
as two separate regions. The annulus fibres in the anterior region were oriented at an
angle of 45 degrees and modelled in a criss-cross manner. The fibres in the posterior
Automated Calibration of Cervical Spine Motion Segment Finite … 1313
region were oriented in the vertical direction. The annulus ground and nucleus were
modelled with linear properties, while the annulus fibres were defined as hypoelastic
material [9].
Five cervical spine ligaments were defined in the model: anterior longitudinal
ligament (ALL), posterior longitudinal ligament (PLL), capsular ligament (CL), liga-
mentum flavum (LF) and interspinous ligament (ISL). The ligaments of the spine
act only when they are loaded in tension and hence were modelled as truss elements
that do not provide resistance in compression. The nonlinear material behaviour of
the ligaments was modelled using hypoelastic definitions. The model was simu-
lated in moment-loading conditions: flexion and extension. The model was loaded
to moments of 2 Nm in flexion–extension. The lower nodes of C6 vertebrae were
fixed in all degrees of freedom. The moment was applied on the upper nodes of C5
vertebra using a multipoint constraint.
1314 D. Natarajan et al.
parameters. The optimized parameters for the individual load cases (flexion and
extension) and the combined flexion–extension load case showed considerable differ-
ences. The responses of the optimized model and the experimental corridor for sepa-
rate flexion and extension load cases are shown in Fig. 4. The response in the case of
combined flexion–extension load case is shown in Fig. 5. The optimized parameter
values are given in Table 1. The difference in optimized parameters between separate
and combined load cases can be attributed to the dissimilar load-bearing functions
of the spinal components in different loading directions. For example, most of the
1316 D. Natarajan et al.
a) Flexion Moment
b) Extension Moment
Fig. 4 Response of the optimized FE model in the individual loading cases, flexion (top) and
extension (bottom). The shaded region shows the experimental responses
Automated Calibration of Cervical Spine Motion Segment Finite … 1317
Fig. 5 Response of the optimized FE model in the combined flexion–extension loading case
Table 2 Details of downhill simplex optimization procedure in the three loading cases
Flexion Extension Flexion–extension
Initial simplex size 0.5 0.5 0.5
Initial value of objective function 4.88 0.11 4.99
Final value of objective function 0.0388 0.008 0.14
Time taken (h) 16 4 20
No. of iterations 450 120 240
1318 D. Natarajan et al.
lateral bending and axial rotation cases need to be further integrated in the automated
calibration process, which will be future study.
4 Conclusions
In summary, a method for automated calibration of the spine FE models was devel-
oped in this study. The kinematics of a single motion segment model was calibrated
with experimental response with a good match in flexion and extension loading. The
model can be improved by considering other loading cases namely lateral bending
and axial rotation along with additional spine model elements like facet cartilages
and muscles which have not been considered in this study.
References
1. Noailly J, Lacroix D (2012) Finite element modelling of the spine. Biomater Spinal Surg
144–234e
2. Oxland TR (2016) Fundamental biomechanics of the spine-what we have learned in the past
25 years and future directions. J Biomech 49(6):817–832
3. Zafarparandeh I, Lazoglu I (2012) Application of the finite element method in spinal implant
design and manufacture. The Des Manuf Med Devices 153–183
4. Jones AC, Wilcox RK (2008) Finite element analysis of the spine: towards a framework of
verification, validation and sensitivity analysis. Med Eng Phys 30(10):1287–1304
5. Kallemeyn N, Gandhi A, Kode S, Shivanna K, Smucker J, Grosland N (2010) Validation of a
C2–C7 cervical spine finite element model using specimen-specific flexibility data. Med Eng
Phys 32(5):482–489
6. Fagan MJ, Julian S, Siddall DJ, Mohsen AM (2002) Patient-specific spine models. Part 1:
Finite element analysis of the lumbar intervertebral disc—a material sensitivity study. Proc
Inst Mech Eng, Part H: J Eng Med 216(July 2016):299–314
7. John JD, Saravana Kumar G, Yoganandan N (2019) Cervical spine morphology and ligament
property variations: a finite element study of their influence on sagittal bending characteristics.
J Biomech 85:18–26
8. John JD, Arun MWJ, Saravana Kumar G, Yoganandan N (2017) Cervical spine finite element
model with anatomically accurate asymmetric intervertebral discs. In: Summer biomechanics,
bioengineering, and biotransport conference. Tucson, USA, pp 153–154
9. Erbulut DU, Zafarparandeh I, Lazoglu I, Ozer AF (2014) Application of an asymmetric finite
element model of the C2-T1 cervical spine for evaluating the role of soft tissues in stability.
Med Eng Phys 36(7):915–921
10. Wheeldon JA, Pintar FA, Knowles S, Yoganandan N (2006) Experimental flexion/extension
data corridors for validation of finite element models of the young, normal cervical spine. J
Biomech 39(2):375–380
11. Zander T, Dreischarf M, Timm A-K, Baumann WW, Schmidt H (2017) Impact of material and
morphological parameters on the mechanical response of the lumbar spine—a finite element
sensitivity study. J Biomech 53:185–190
12. Mattucci SFE, Cronin DS (2015) A method to characterize average cervical spine ligament
response based on raw data sets for implementation into injury biomechanics models. J Mech
Behav Biomed Mater 41:251–260
Automated Calibration of Cervical Spine Motion Segment Finite … 1319
13. Zanjani-Pour S, Winlove CP, Smith CW, Meakin JR (2016) Image driven subject-specific finite
element models of spinal biomechanics. J Biomech 49(6):919–925
14. Lavecchia CE, Espino DM et al (2018) Lumbar model generator: a tool for the auto-
mated generation of a parametric scalable model of the lumbar spine. J R Soc Interface
15(138):20170829
Identification of Inertial Parameters
and Friction Coefficients for One-Link
Manipulator
1 Introduction
There is always a disparity between the CAD model and actual installed robot, due to
the manufacturing errors, and some parts such as washers, wiring, transmissions, and
joint friction which are not modeled with required precision. The identification of
inertial parameters and joint friction coefficients is essential mainly for the model-
based control of the multibody systems. Inertial parameters of a rigid link which
is the part of a multibody system consist of ten standard inertial parameters (SIP),
The identification of inertial and frictional parameters is required for precise simu-
lation and control [3]. The philosophy of parameter identification is based on the
input–output data of a manipulator under study. The linear-in-parameter (LIP) form
of the Newton–Euler equation was used for the identification of inertial and joint
friction parameters of the manipulator. The detailed derivation to formulate the LIP
form using the DeNOC matrices [9, 10] is presented in [3] for the multibody systems.
The LIP form for one-link manipulator is discussed next.
Identification of Inertial Parameters and Friction Coefficients … 1323
Yχ = τ (1)
T
where Y ≡ θ̈ sgn θ̇ θ̇ is the regressor matrix, χ ≡ I yy f c f v is the vector of
unknown parameters for the horizontal motion of the manipulator, I yy is the moment-
of-inertia about Y-axis of the body-fixed frame, and f c and f v are the coefficients of
Coulomb and viscous friction, respectively. Because of the minimal normal force due
to gravity in the horizontal plane, the Coulomb friction coefficient was considered
as constant, whereas the viscous friction coefficient as the function of joint rate.
Two different joint trajectories were considered for dynamic identification and torque
reconstruction, namely the trapezoidal (at velocity level) and sinusoidal trajectory.
The sine wave trajectory as given in the EPOS studio manual was considered, which
is written below:
θ (t) = a sin 2π nt − π 2 + a (2a)
θ̇ (t) = 2π na cos 2π nt − π 2 (2b)
θ̈ (t) = −(2π n)2 a sin 2π nt − π 2 (2c)
1324 A. K. Sharma et al.
where t is the instantaneous time in seconds, a and n are the amplitude (radian) and
frequency (Hz) of sine wave, respectively. The trapezoidal trajectory was divided in
three phases, i.e., constant acceleration, constant velocity, and constant deceleration.
The mathematical formulation for the trapezoidal trajectory is given below:
θ (t) = a0 + a1 t + a2 t 2 (3a)
Here, tf is the final/total time of trajectory. The polynomial form for the constant
velocity can be given as
θ (t) = b0 + b1 t (5a)
θ̇ (t) = b1 (5b)
Because of continuity, the joint angle and joint rate at the end of constant accel-
eration phase must be equal to the joint angle and joint rate in the starting of the
constant velocity phase, i.e.,
b1 = a1 + 2a2 t1 (6a)
Here, td is the time duration of deceleration phase. The polynomial form for the
constant deceleration phase can be expressed as
θ (t) = c0 + c1 t + c2 t 2 (7a)
Identification of Inertial Parameters and Friction Coefficients … 1325
The experimental setup of one-link manipulator is shown in Fig. 2. The input param-
eters for the trapezoidal trajectories are given in Table 1. The joint trajectory ‘Trape-
zoidal 1’ was used to excite the manipulator for parameter identification. The joint
angle, joint rate, and current taken from the EPOS studio software were in quadrature
count, rpm, and mA, respectively. Further, joint angle, joint rate, and current were
converted to rad, rad/s, and Nm, using the conversion factors given in the Maxon
motor manual. The joint acceleration was found using the numerical differentiation
of the joint rate data. The noise in joint rate, joint acceleration, and joint torque
data was removed through zero-phase digital filtering in both forward and reverse
direction using the ‘filtfilt(b, a, x)’ command of MATLAB.
The mass moment-of-inertia about the joint axis was identified as 0.1321 kg m2 ,
whereas Coulomb friction coefficient as 0.2. The identified viscous friction coeffi-
cient as the function of joint rate is shown in Fig. 3. An eleventh-order polynomial
was used to fit the identified data using ‘polyfit(x, y, n)’ command of MATLAB.
The polynomial form of viscous friction coefficient is valid for a range of joint rate
from 0 to 2.55 rad/s. For the joint rate beyond 2.55 rad/s, the polynomial may lead
to a negative viscous friction coefficient, which is incorrect. The validation of the
identified parameters was done using joint trajectories ‘Trapezoidal 2’ and a sine
wave, Eq. (2a), with amplitude of π 6 and frequency of 0.5 Hz.
The results for joint torque reconstruction are shown in Figs. 4, 5, and 6.
Fig. 3 Variation of
identified viscous friction
coefficient with respect to
joint rate
Identification of Inertial Parameters and Friction Coefficients … 1327
The dynamic and friction coefficients of a one-link planar manipulator were identified
through experimental data. A constant Coulomb friction coefficient was proposed for
the horizontal motion of the manipulator and found in good agreement with the exper-
imental results. Moreover, the viscous friction coefficient was identified as a function
of joint rate. The identified parameters were validated for different joint trajectories
and found that the proposed identification strategy for friction parameters is very
close to the experimental results. The identification results can be improved using
the more precise measuring instruments for joint acceleration and motor current. The
numerical differentiation of the joint rate leads to noisy joint acceleration data. The
accelerometer mounted on the link can improve the accuracy of joint acceleration,
and its numerical integration would be less noisy to get the joint rate and joint angle
data. The proposed friction coefficient identification strategy will be implemented to
get the variable Coulomb friction coefficient with respect to the joint angle.
References
Abstract This paper deals with the human-robot compatibility issue of upper-limb
exoskeleton through a dimensional synthesis problem. The work is a contribution
to solving misalignment. In this paper, the objective is the task-based dimensional
synthesis of a wearable upper-limb exoskeleton for emulating natural human motion.
A planar hybrid architecture is used for the purpose, with a four-bar connected to
another four-bar in series. The task is selected based upon the standard rehabili-
tation exercises, only for the planar motion (parallel to sagittal plane). To achieve
the proposed objective, the work has consisted of the formulation and solving of a
constrained optimization problem, with reachability, design limits and solution con-
tinuity as constraints. Genetic algorithm is used for problem-solving. The results
are detailed for proposed manipulator for the upper-limb exoskeleton, showcasing
variation in design limits and constraints.
Stroke is a leading cause of disability in the society in which a person deficits his/her
motor-function required to do the daily activities of life (ADL). Neurocognitive reha-
bilitation is considered the prominent technique to resolve the situation [1, 2] and
subacute stage, i.e., within six months of stroke is the best stage for better recov-
ery [3]. Several therapies have been designed to target this critical period. However,
due to the repetition of exercises, the manual techniques are monotonous and time
consuming for a physiotherapist. Robotic assistance in the direction may help in over-
coming the raised issues, and exoskeletons are worked upon for the purpose. Such
designs of exoskeleton encounter many challenges, including ergonomically accept-
able wearing technology, architectural design, human-motion compatibility, human
robot interaction, etc. [4]. Motivation of this work is handling one of the challenges
of misalignment. To emulate natural human motion, two aspects had been introduced
in the problem formulation. One is task-based dimensional synthesis which involves
the data taken from arm motion for standard rehabilitation exercises. The second is
the planar architecture with closed loops. A four-bar over a four-bar configuration
is used for this purpose which is expected to minimize the misalignment between
human and robotic motion. The selection of the four-bar linkage is majorly due to
possesses of varying instantaneous centers of the coupler [5, 6]. The work includes
exercise-based data collection, kinematic modeling of double four-bar configuration
and dimensional synthesis of the mechanism for the selected rehabilitation exercise
which is considered as the task.
Human motion data extraction had been done for conventional and unconventional
techniques, targeting standard rehabilitation exercises as shown in Fig. 1. These exer-
cises have been chosen on the basis of experts recommendations (courtesy: Stroke
Rehabilitation Centre Indian Spinal Injury Centre, Delhi) and available anthropo-
metric exercise-based data for rehabilitation purpose in the rehabilitation centers.
The selected exercises for presented work, as high-lighted, are assumed to be planar
movements.
The flexion-extension movement of shoulder and elbow, which is parallel to the
sagittal plane, has been retrieved through human motion camera. In the initial phase
of the work, this data has been collected for emulating natural human motion. As
extracted the selected exercises, the targeted motion tasks are lifting empty hand
(liftrighthand, liftlefthand, liftbothhand) and lifting 5 kg load(lift5kgright, lift5kgleft,
Fig. 1 Standard rehabilitation exercises (courtesy: Indian Spinal Injury Centre, Delhi)
A Task-Based Dimensional Synthesis of an Upper-Limb … 1331
lift5kgboth). The subject data has defined from 25 to 33 years age people and height
from 1.6 to 1.9 m. Both male and female gender data have been taken. A sample
human motion data is shown in Table 1.
To accomplish the selected tasks, the basic architecture is decided as a hybrid archi-
tecture. The work has selected double four-bar connected in series, i.e., a four-bar
over another four-bar configuration connected in series which is harness at the wrist
as shown in Fig. 2a. The idea to choose four-bar linkage is due to the property of
varying instantaneous centers of the coupler[5, 6]. The purpose of the configura-
tion is to provide the assistance flexion-extension movement at shoulder and elbow,
i.e., two-DoF, which is parallel to the sagittal plane. The details of the proposed
configuration are shown in Fig. 2b.
Fig. 2 Double four-bar mechanism connected with series: a concept design, b configuration
1332 S. Gupta et al.
The lengths of the configuration are named as l1 to l7 , and all the joint angles are
considered in anti-clockwise direction from x-axis marked with θ1 to θ7 as shown in
Fig. 2b. The DoF of the system is two, and in this problem, θ1 and θ5 are considered
as active joints. To evaluate the performance of the double four-bar configuration,
inactive joints θ2 , θ4 , θ6 , θ7 have been expressed in terms of θ1 and θ5 .
The location of the end-effector, considered at point P, can be computed using
closed-loop equations as.
F1 = l1 C1 + l2 C2 − l3 C3 − l4 C4 = 0, (1)
F2 = l1 S1 + l2 S2 − l3 S3 − l4 S4 = 0,
F3 = l5 C5 + l6 C6 − l2 C2 − l7 C7 = 0,
F4 = l5 S5 + l6 S6 − l2 S2 − l7 S7 = 0.
Px = l1 C1 + l2 C2 + l7 C7 , (2)
Py = l1 S1 + l2 S2 + l7 S7 .
l1 S(θ1 −θ2 ) l5 S(θ5 −θ6 )
S(θ2 −θ4 )
S(θ4 ) S(θ6 −θ7 )
S(θ7 )
J= S(θ1 −θ2 ) S(θ5 −θ6 ) . (4)
− l1S(θ 2 −θ4 )
C(θ4 ) − l5S(θ 6 −θ7 )
C(θ7 )
To emulate the natural human motion and avoid elbow misalignment, the following
designing criteria and assumption have been made:
A Task-Based Dimensional Synthesis of an Upper-Limb … 1333
Assumptions:
For the initial analysis, the lower bound and upper bound constraints for the link
lengths are set as 2 and 40 cm and joint angles are set as 0.01 and 360◦ , respectively.
The illustration of results with the design limits for task-based dimensional synthesis
of double four-bar configuration is connected in series, as shown in Fig. 3. The optimal
configuration has 0.14, 0.45, 0.28, 0.41 and 0.062 condition index with respect to
each TSLs. The results are further improved by modifying the design limits in terms
to achieve ergonomically and aesthetically compatible configuration. The modified
design lower bound and upper bound constraints for the link lengths are set as lower
bound = [2 6 2 2 10 17 2] and Upper bound = [40 12 6 40 20 25 45]. The joint
angles are set as 0.01 and 360◦ , respectively, as illustrated in Fig. 4. Moreover, in
this iteration introduced joint
n,mangle i+1continuity as another objective of the problem.
Joint angle movement = i=1, j=1 ( θ j −i θ j ). Therefore, optimal results for the
modified problem are 0.30, 0.45, 0.56, 0.8 and 0.38 with respect to each TSLs. It is
observed that the modified optimal results have a better Jacobian conditioning index
than previous optimal problem results.
Fig. 3 Optimal results of double four-bar mechanism for the formulated problem
Fig. 4 Optimal results of double four-bar mechanism for the modified design limits
A Task-Based Dimensional Synthesis of an Upper-Limb … 1335
Here, [xk ,yk ] represents kth coupler position, ∀k = 1 to n and [xe ,ye ] represent
elbow position.
It has been noticed that the obtained Jacobian conditioning index for each TSLs are
0.12, 0.02, 0.13, 0.35 and 0.45, respectively, as shown in Fig. 5 which is not so good
as an earlier case but misalignment issue minimized due to target the dimensional
synthesis of the base four-bar corresponding to elbow movement.
6 Conclusions
A four-bar over another four-bar configuration, connected in series, has been given
the flexion-extension movement of shoulder and elbow, i.e., a planar movement
parallel to the sagittal plane. This paper presents the kinematic model formulation for
task-based synthesis of a two-DoF wearable exoskeleton for rehabilitation exercises.
Objective of the problem formulation is to emulate natural human motion within
some desirable conditions such as good kinematic condition, least misalignment
and joint limits. An optimization problem is formulated and solved for reachability
at the collected working points of forearm while minimizing joint movements and
misalignment at elbow joint.
1336 S. Gupta et al.
Acknowledgements The authors sincerely acknowledge the grant from Indo-UK industrial project
(Department of Science and Technology— Global Innovation and Technology Alliance) for finan-
cial support of this work.
References
1 Introduction
The topology optimization method helps define connections and holes in the initial
material domain subject to boundary conditions. In this way, a novel shape and size
of the components are obtained [1]. A significant study on the application of topology
optimization was started in 2006; few members of the humanoid were optimized for
energy saving [2, 3]. Links were optimized topologically, considering static loading
conditions [4, 5]. Few attempts are made on five-bar mechanism, welding robot,
and filtration techniques-related optimization [6–8]. Srinivas and Javed provided a
comprehensive review of the optimization techniques for industrial manipulators [9].
In the present work, three-DOF articulated manipulator was considered for
topology optimization, which is subjected to dynamic loading conditions. In order to
apply topology optimization for a rotating mechanical member of an articulated arm,
parameters such as initial design space, material properties, and loading conditions
need to be identified. For a rotating link, loading conditions are the most complicated
and sensitive parameter to define. However, the obtained optimal topology shows a
thin structure with tiny holes, which increases the stress values on the overall link
and also offers the complexity for the manufacturing process. In the present work, an
image processing-based filter is developed and validated through the stress values.
The topology is generated using MATLAB as well as ANSYS software for better
illustration. The topology optimization process is detailed in Sect. 2. The method-
ology for filter is provided in Sect. 3, and the validation of the simplified topology
is discussed in Sect. 4.
2 Topology Optimization
The topology optimization helps to optimize design space by SIMP approach and
optimality criteria method [10]. Two different zones are defined in the initial material
(design space) domain, specifying the elements that can undergo the topology opti-
mization process, and specifying material zone with a fixed value of density param-
eters to accommodate assembly and motion constraints. With these considerations,
the topology optimization problem is defined, as shown in Eq. (1).
⎫
min : C(xe ) = DT KD ⎪ ⎪
⎬
Subjected to : V (xe ) − V × f = 0
(1)
KD − F= 0 ⎪ ⎪
⎭
0 < xe ≤ 1
where ‘f ’ is the volume fraction, ‘C’ is the compliance, and ‘D’, ‘F,’ and ‘K’ are the
global displacement, force vectors, and global stiffness matrix. ‘V (x e )’ and ‘V ’ are
the material volume and design domain volume, respectively, and ‘x e ’ is the element
design variable.
Topology Refinement from Design to Manufacturing Using Image … 1339
A three-DOF articulator manipulator with three revolute joints that is an RRR arm
configuration with joint variables and free body diagram (FBD) is shown in Fig. 1.
Each link has its own self-weight (W) which always acts vertically downwards at
center of gravity of the design space. The mass of the motor (M) and normal reaction
(N) affect on each link due to contact at the joints. The moment or torque (T) is
exerted by the motor at three joints of the manipulator-link. Here, the central link is
subjected to the topology optimization, which has length and height, 400 mm and
100 mm, respectively, with thickness 4 mm. The material chosen for manipulator
link is mild steel, which has mechanical properties, Young’s modulus of elasticity:
200 GPa, Poisson’s ratio: 0.33, density: 7700 Kg/m3 .
The manipulator links one end is subjected to a fixed support, and another end
is bearing 30 N payload. The constraint of the volume fraction was selected as 0.5.
A MATLAB code is written to adopt the varying boundary condition, element size,
dynamic change in center of gravity, and self-weight, with each iteration. The code
is made to capture performance values as maximum deflection and von Mises stress.
It can update the design variables of topology up to the convergence of the objective
function. The final optimum topology along performance values are presented in
Table 1 based on stated boundary conditions and constraints after mesh independency
test [11].
Maximum von
Mises stress
1342 G. Lakshmi Srinivas and A. Javed
3 Methodology
In the obtained topology, various thin members can be observed, as shown in Table 1
and Fig. 3. These members offer high-stress values and are impractical to manu-
facture. To make it feasible at the manufacturing phase, a filtration method is
proposed.
where I 1 , I 2 , and I 3 are values carrying the various discrete level (n) of individual
processing levels for edge contrast, a separation between the darkest and lightest
area, reduction of each pixel values, respectively. The effect of process parameters
on image processing is shown in Table 3.
These three processes are tested manually to obtain this sequence, which is suitable
to reduce the thin pixel elements, whether dark or bright, with the surrounding domain
of opposite pixel values. Each of these processes is labeled with effect parameters
P’s.
The obtained images for each ‘P’ value are further filtered with the volume fraction
and von Mises stress values. From these topologies, the stepwise simplification of
the topology can be seen with the reduced value of von Mises stress and a marginal
increase in the desired volume fraction, which was originally 0.5, as shown in Table 4.
The image processing technique based on filtration of process parameters is useful
to design to manufacturing phase with tolerant topology optimization [12]. The filtra-
tion process eliminates the thin structures and tiny holes present in the initial material
domain or space.
From results, as volume fraction increases, topology simplified with the elimi-
nation of tiny holes and thin members by adding material to the design space. Von
Mises stress is reduced to 3.82 MPa for volume fraction 0.52, which is 26.2% less
than the initial optimum topology.
Topology Refinement from Design to Manufacturing Using Image … 1343
Initial to-
pology
80
70
60
50
5 Conclusion
Acknowledgements The present work was funded by the “Department of Science and Technology
and Science and Engineering research board, India” (Grant Number: ECR/2017/000799).
References
1. M.P. Besndsoe, O.S.: Topology optimization theory, methods and applications. Springer
Singapore (2003)
2. Albers, A., Brudniok, S., Ottnad, J., Sauter, C., Sedchaicharn, K.: Upper body of a new
humanoid robot - The design of ARMAR III. Proc. 2006 6th IEEE-RAS Int. Conf. Humanoid
Robot. HUMANOIDS. 308–313 (2006). https://doi.org/10.1109/ICHR.2006.321289
3. Lohmeier, S., Buschmann, T., Schwienbacher, M., Ulbrich, H., Pfeiffer, F.: Leg Design for a
Humanoid Walking Robot. 536–541 (2006)
4. Denkena B, Bergmann B, Lepper T (2017) Design and optimization of a machining robot.
Procedia Manuf. 14:89–96. https://doi.org/10.1016/j.promfg.2017.11.010
5. Li, X., Shao, H., Li, G., Liu, W., Liu, C.: Static simulation and structure optimization of key
parts of joint welding robots. Proc. 2018 IEEE Int. Conf. Mechatronics Autom. ICMA 2018.
282–287 (2018). https://doi.org/10.1109/ICMA2018.8484469
6. Briot S, Goldsztejn A (2018) Topology optimization of industrial robots: Application to a
five-bar mechanism. 120:30–56
Topology Refinement from Design to Manufacturing Using Image … 1345
7. Luo H, Fu J, Wang P, Liu J, Zhou W (2019) Design optimization of the ram structure of friction
stir welding robot. Mech. Adv. Mater. Struct. 01:1–11. https://doi.org/10.1080/15376494.2018.
1471758
8. Lakshmi Srinivas G, Javed A (2020) Topology optimization of rigid-links for industrial
manipulator considering dynamic loading conditions. Mech. Mach. Theory. 153:
9. Lakshmi Srinivas G, Javed A (2020) Optimization approaches of industrial serial manipulators
to improve energy efficiency: A review. IOP Conf. Ser. Mater. Sci. Eng. 912: https://doi.org/
10.1088/1757-899x/912/3/032058
10. Srinivas GL, Javed A (2020) Numerical evaluation of topologically optimized ribs for
mechanical components. Mater. Today Proc. https://doi.org/10.1016/j.matpr.2019.12.292
11. G. Lakshmi Srinivas and Arshad Javed: Numerical Simulation and Experimental Study on
Lightweight Mechanical Member. In: Advanced Engineering Optimization Through Intelligent
Techniques. pp. 631–641. Advances in Intelligent Systems and Computing, Springer Singapore
(2020)
12. Sigmund O (2009) Manufacturing tolerant topology optimization. Acta Mech. Sin. Xuebao.
25:227–239. https://doi.org/10.1007/s10409-009-0240-z
Soft Robotic Gripper for Agricultural
Harvesting
Abstract Historically robots have been a compilation of rigid parts or objects which
move relative to one another to transfer motion. One of the main drawbacks of this
classical method has been the physical restrictions and rigidity of the structure of the
components. Conventional robots are made of rigid materials that limit their ability
to elastically deform and adapt their shape to external constraints and obstacles. It
is also harder to use these robots in any industry which require a certain degree of
sensitivity and delicacy when interacting with the environment. Soft Robotics is a
relatively new subfield of Robotics with incredible potential to change the industry
due to their construction from highly compliant materials. Soft robots have increased
flexibility and adaptability as well as improved safety when working around humans.
In this work, we have described our attempts at the design and fabrication of few soft
robotic grippers. We have experimented with three types of liquid silicone rubber
materials with different properties and tried fabricating one-fingered, three-fingered
and four-fingered soft pneumatic actuators. We have demonstrated grasping in lifting
few objects of different shapes and sizes. The gripper developed has the potential to
be used for harvesting in agricultural fields.
1 Introduction
Robots can be programmed to do the same task repeatedly with minimal error which is
something very difficult for humans. But if these robots are taken out of the factories,
where the environments are not perfectly known, then the robots struggle and even
fail to do even a simple task which does not require much precision. For many years,
robots have been designed to emphasize speed and precision, and this translates into
a very specific architecture. In this robotic structure, one has to perfectly measure
the working environment and one has to perfectly program every movement of the
robot joints as even a small error can generate a very large faulty outcome, which
can damage the robot or its surroundings or both.
Soft robotics is bio-inspired from animals, the way they move and adapt to their
surroundings [1]. There is a wide range of medical applications of soft robotics
like prosthetic robots and also in food handling equipment. Soft robotics Inc. [2]
is a company that deals with food and beverage, advanced manufacturing and e-
commerce, and retail. The grippers of these manufacturers are very flexible and do
not damage the object. Soft robotic gripper has been developed to pick uneven and soft
items such as food packs very effectively [3]. Soft robotics toolkit is a platform that
provides people to learn about the design, fabrication, modeling, characterization,
and control of soft robotic devices [4]. They hold competitions where researchers
and enthusiasts are allowed to use soft robotics in their respective fields, thus making
them aware of soft robotics and learn its potential uses. A recent article [5] compares
the development and testing of soft gripper developed using soft silicone material
and being actuated by different methods, such as pneumatic pressure, SMA (shape
memory allow) wire, and electromagnets.
One of the verticals or applications where soft robots are useful is in agricul-
ture harvesting. Hiring manual labor for harvesting is one of the significant costs
borne by an agriculturist. During peak seasons, there can be a significant demand for
labor which results in shooting up of costs. Hence, any automation in harvesting is
beneficial.
Due to the fragility of the fruits or vegetables, the conventional robots have mostly
been out of the scope for harvesting purposes. To achieve the gentle handling which
is usually required for harvesting using the conventional rigid robots, force feedback
during gripping is a solution, but it is also complicated and expensive. Hence, a soft
robot gripper might be a feasible alternative.
In this paper, we have made an attempt on developing a few soft robotic actuated
grippers using different materials available. Though various methods are available
for the development of soft actuator, first we considered fiber-reinforced method,
which is explained in Sect. 2. This method had major challenges and hence was not
completely successful. Another method based on PneuNets (pneumatic networks)
was attempted and has been reported in Sect. 3. This method has a better success
rate, considering the materials available with us. Grippers with different number of
fingers were developed and they were able to grasp few objects. Section 4 discusses
the further scope and conclusions of the work.
Liquid silicone rubber when mixed with suitable agents can be poured into molds.
After a certain curing time, the liquid solidifies into an elastic material which has
Soft Robotic Gripper for Agricultural Harvesting 1349
many desirable properties for a soft actuator. Solidified elastic material with a hollow
space can be injected with compressed air, like a balloon, and the material tries to
expand in all possible directions, i.e., in the longitudinal and the lateral direction.
This can be explained by considering Fig. 1a, where an elastic material in the form
of a finger is subjected to compressed air. The expansion is uniform in all directions.
If some form of restraint are placed so that the radial or transverse expansion is
restricted, the volume is allowed to expand only longitudinal direction, as shown
in Fig. 1b. These restraints are known as fiber-reinforcements. In addition to these
reinforcements, if one of the face/side of the finger is stuck with a non-extensible
material (bottom side in Fig. 1c), the material on the above part has a provision to
expand its volume whereas the bottom side is restricted. Hence, the finger curls down,
as shown in Fig. 1c. There are many different ways of obtaining the reinforcement, as
explained in [6]. By carefully selecting the dimensions of the mold and the thickness
of the wall of the finger, we can obtain certain characteristic relationship between
the pressure of the input compressed air and the angle of curl, the maximum contact
force the gripper can apply on an object, etc.
Soft Robotics Toolkits [4] website and other online resources have a good range of
tutorials and DIY (do it yourselves) video tutorials on creating soft robotic actuators.
The material used in most of the projects in the above resource is Eco-Flex 00-30.
However, when the project was initiated, its availability in Indian market had a longer
period for delivery. Hence, before an order was placed, we wanted to explore other
materials that were readily available in the market.
The first material was “Dow Corning MS-2002”. The silicone rubber and its
curing agent were procured and when mixed in the ratio of 1:10, respectively, the
mixture was poured in a mold made up of cardboard, as shown in Fig. 2a. By this
process, three of the four walls of each finger are formed. After a curing period of
24–30 h, paper is used as the reinforcement or restraining material and another layer
of mixture is allowed to cure for another 24–30 h, as shown in Fig. 2b, which forms
the fourth and missing wall of each finger. During this process, a provision is made to
insert a pipe at the center of the gripper to allow compressed air to pass through and
actuate the gripper. After many trials, we were able to get the gripper to curl for lower
pressure. At pressure, more than 0.2 bar, the connection between the reinforcement
and the main part was getting punctured, as shown in Fig. 2c, hence making this
material unusable.
(a) No restraints
(a) Cardboard mold (b) Paper as restraint (c) Puncture in (d) Prototype using
with mixture for base Prototype using first material second material
An alternative material with no formal name was procured from another vendor
which had a curing period of 4–6 h. Similar set of operations were carried out and
a gripper was molded as shown in Fig. 2d. Here, instead of paper, cloth was used
as a restraining material. As this material was less elastic than the first material, it
was not able to curl to hold any object. However, we learnt that this material was
also not suitable for application in soft robotics. Thereafter, we placed order for
Eco-Flex 00-30, which was at least 3 times as expensive as the first two materials.
As the volume ordered was lower, we attempted to fabricate the soft gripper using
PneuNets method, which consumes lesser material than the reinforcement method.
This is explained next.
PneuNets are a type of soft robotic gripper. A finger has several chambers and passage
for compressed air through them, as shown in Fig. 3a. When compressed air is sent
through the chambers, the air expands the chambers. If a restraining material is
stuck to one of the faces, that face is not allowed to expand, thus resulting in a
curl, illustrated in Fig. 3b. This methodology requires accurate mold preparation, as
compared to the method explained in the previous section. A thorough overview of
(a) 3D CAD model of mold in Autodesk Inventor (b) 3D printed mold (c) Cured griper finger
the mold preparation using a 3D CAD modeling software and subsequent 3D printing
of the molds is discussed in [7], which was further implemented as a prototype in
Fig. 3c. In this paper, few variants of PneuNets fingers and grippers were fabricated
and tested.
Single-Finger Gripper: A 3D CAD model as mold was developed using
Autodesk Inventor software considering suitable dimensions of the air passage and
chambers, as shown in Fig. 4a. The mold was later manufactured using a 3D printer
(Fig. 4b). Unlike the first two materials, Eco-Flex 00-30 material is to be mixed in
1:1 ratio for the two solutions, A and B. The mixture was then poured into the mold
and is cured to become an elastomer (elastic material). After the first round of curing,
restraining material (cloth) was used to cover up the base with additional round of
liquid mixture, and allowed to cure. The resulting single finger (Fig. 4c) was flexible
and was operational as expected. As compressed air was supplied, the curling of the
finger was observed, as illustrated in Fig. 5a. The finger was able to grasp fruits and
vegetables such as lemon and tomato, which were of the order of 50 g, illustrated in
Fig. 5b, c.
Multi-Fingered Gripper: Similar to single-fingered gripper, PneuNets method-
ology was used to create multi-fingered gripper. First, a four-finger gripper was
designed and printed using a 3D printer (Fig. 6a). Note that the fingers are connected
at a hub at the center, through which compressed air is sent into individual fingers’
chambers. Hence, all fingers are bound to actuate/curl synchronously. The proto-
type developed was successful in holding cylindrical-shaped objects, as illustrated
in Fig. 6b, c. One problem faced was that the mold and prototype had sharp edges
at the joining of neighboring fingers and hence was at times a cause of leakage.
To overcome that, fillet and rounding were added at the common edge in a three-
fingered gripper, as shown in Fig. 6d. Here, the gripper performed better with regard
to potential puncture at the common edges.
1352 S. M. G. Vidwath et al.
(a) Right to Left: With increase in the input pressure, the curl angle increased
(b) Grasping and lifting of a lemon (c) Grasping and lifting a tomato
(a) Mold for four (b) Holding a jar (c) Holding cylindrical (d) Three fingered
fingered gripper object gripper
4 Conclusions
Soft robotics deals with usage of elastic material to grasp or manipulate objects. One
of the potential applications is in agricultural robotics, particularly for harvesting
fruits and vegetables. In this paper, an attempt was made to develop single-finger
and multi-finger gripper using available liquid silicone rubber material. Based on
the tests and observations, we deduced that Eco-Flex 00-30 gives the best results
and should be used in such applications. As the two materials, other than Eco-Flex
00-30, were not able to hold object properly, quantitative comparative tests between
the three materials could not be conducted. Tasks such as grasping lemon, tomato and
few cylindrical-shaped objects were conducted successfully. In future, the grippers
Soft Robotic Gripper for Agricultural Harvesting 1353
would be designed to grasp uneven shaped objects of varying sizes and forms. Also,
other alternative materials/household objects that can be used as soft robotic gripper
finger shall be explored and reported in the future.
References
1. Laschi C, Cianchetti M, Mazzolai B, Margheri L, Follador M, Dario P (2012) Soft robot arm
inspired by the octopus. Adv Robot 26(7):709–727
2. Soft Robotics Inc. www.softroboticsinc.com
3. Wang Z, Torigoe Y, Hirai S (2017) A prestressed soft gripper: design, modeling, fabrication,
and tests for food handling. IEEE Robot Autom Lett 2(4):1909–1916
4. Soft Robotics Toolkits. www.softroboticstoolkit.com
5. Karmarkar S, Sarkar A (2019) Design and implementation of bio-inspired soft robotic grippers.
In: Advances in robotics: 4th international conference of The Robotics Society
6. Polygerinos P, Wang Z, Overvelde JT, Galloway KC, Wood RJ, Bertoldi K, Walsh CJ (2015)
Modeling of soft fiber-reinforced bending actuators. IEEE Trans Rob 31(3):778–789
7. Sun Y, Song YS, Paik J (2013) Characterization of silicone rubber based soft pneumatic actuators.
In: IEEE/RSJ international conference on intelligent robots and systems, pp 4446–4453
Surface Profile Accuracy of Deployable
Mesh Reflectors Based on Focal Offset
1 Introduction
In this paper, we consider a deployable mesh reflector antenna whose curved reflective
surface is an offset paraboloid. This type of surface helps in focusing electromagnetic
rays falling on the reflector from a distant object onto the receiver, which is located
at the focus. Such an antenna deployed using a tension truss was first presented by
Miura and Miyazaki [1]. It consisted of five components: two cable networks in
the front and rear, tension ties connecting corresponding nodes in the front net to
those in the rear net, a reflective mesh attached to the front net, and a ring truss. The
cable network (front and rear net) is used to approximate the curved surface of the
reflector [2], whose shape is an offset paraboloid. Morterolle et al. [3] presented a
design technique for generating the offset paraboloid. The offset reflector surface
is obtained by the intersection of the paraboloid and a cylinder, and then the facets
are generated by successive division of this surface into segments at the aperture
circle. The lines joining these are connected to the center of the reflector. By doing
so, triangular segments are obtained, which are further subdivided. These are then
mapped on to the surface using a suitable origin of the coordinate system to obtain
the coordinates of vertices of the facets. These flat facets collectively approximate the
reflector surface, whose vertices lie on the curved reflector surface and illustrated in
Fig. 1. Such an approximation necessitates determination of surface accuracy, which
is measured by the deviation between the curved working surface, and the facets are
formed by connecting these nodes. This deviation is quantified as the facet error and
is a widely adopted approach found in the literature.
Existing approaches try to minimize the root mean square error between the
faceted surface and the local quadratic approximation of the paraboloid for designing
the cable net. An alternate approach for computing the surface accuracy of the faceted
reflector surface is proposed here based on the proximity of the reflected rays at the
receiver (focus) of the antenna. Agarwal [4] discussed a method to choose the initial
nominal facet length. Yuan [5] discussed an optimization procedure for finding the
optimum position for each node by minimizing the error in the reflector surface,
which is different from the offset error presented in this paper.
The parameters from Table 1 were used in the design of our reflector with 1176
facets. The faceted reflector is shown in Fig. 2, and the facet error distribution is shown
in Fig. 3. The facet error computed for this reflector is shown in Fig. 4 graphically
along with a color bar indicating the error (in m) across the reflector. It can be
observed that the facet error decreases along the x-axis. This is because the curvature
of the reflector decreases gradually, resulting in flatter surface being more accurately
approximated by flat facets. Also, the facet error value is symmetric about the y-axis.
The RMS value of the facet error is 0.35 mm.
Ideally, the objective of any antenna is to focus all the incoming electromagnetic rays
onto the receiver. But due to the aforementioned faceting of the surface, some of the
facets (or parts of the facet) do not reflect the rays to the receiver. The offset distance
Surface Profile Accuracy of Deployable Mesh Reflectors … 1359
by which these rays miss the target point on the receiver (focus) is measured by a
factor we call the focal offset error. The offset distance of reflected rays from the
focus is a more geometrically relevant indicator for determining the accuracy of the
faceted surface than the closeness of the faceted surface to the original surface.
A diagrammatic illustration of the offset error is shown in Fig. 5 with an inset
illustration of the receiver disk. Incoming parallel rays fall on the reflector surface
along the z-direction. Three such rays are denoted by O, O1, and O2 .
Consider the ray from O. In the absence of faceted surface, this ray would get
reflected from point O* on the surface of the paraboloid, and hence, it would reach
the focus as a result of true surface reflection. This is shown by the red colored ray.
Similar is the case with point Q lying on the paraboloid except that it happens to be
that point at which the tangent plane is parallel to the facet and this will be discussed
later to derive offset error.
Now, considering the faceted surface, the same ray from O would get reflected
from point G, which is the centroid of the facet, and reach a different point on the
receiver disk, say J. This is shown by the blue colored ray. Hence, the same incoming
ray would reach different points on the focus depending upon the surface of reflection.
This distance (FJ) on the receiver disk, by which the ray misses the focus, is called
as the focus offset error or simply the offset error. The offset vector is denoted by wp
(see Fig. 5).
Derivation of the offset error The offset error wp is obtained by the projection of
w on the receiver disk shown in Fig. 6, where w is the distance between the reflected
rays QF and GJ. Hence, w is calculated first. We obtain w by calculating the distance
from the point G to the reflected ray nr passing through point Q as shown in Fig. 7.
Q is to find the only point on the paraboloid that reflects the ray through the focus in
the same direction as the rays reflected from the facet.
(i) The equation of the paraboloid surface is written as
H (x, y, z) = 4 f z − x 2 + y 2 = 0 (2)
∇H −x î − y ĵ + 2 f k̂
n(x, y) = = (3)
∇ H x 2 + y2 + 4 f 2
(R2 − R1 )(R3 − R1 )
nf = (4)
(R2 − R1 )(R3 − R1 )
n(x Q , y Q ) = n f (5)
H (x Q , y Q , z Q ) = 0 (6)
f −q −xq î − yq ĵ + ( f − z q )k̂
nr = = (7)
f − q xq2 + yq2 + ( f − z q )2
where f is the position vector of the focus, F. Thus, w can be calculated as the shortest
distance between point G from line QF, which is given by
w − (w · n p )n p
wp = + (w · n p )n p (9)
nh · nr
where
nr × nh
np = (10)
nr × nh
1362 S. S. Siddesh et al.
and nh is the normal to the focal disk which determines its orientation. The assumed
orientation of this disk is such that its normal nh is in the same direction as that of a
ray reflected from the center of the paraboloid, to ensure maximum amount of signal
received. The magnitude of wp , is the offset error.
The rays reflected from O and O2 are parallel to each other as can be seen in
Fig. 5. Hence, the rays falling on any point on the facet surface would travel parallel
to ray reflected from the centroid of the facet, G. Therefore, the offset error of these
rays can be calculated by substituting its position vector for g in Eq. 8.
Figure 8 shows the offset error over the entire reflector graphically, wherein the
error is indicated by the color. The error is symmetrical about the XZ plane. The rays
reflected from the centroids of three facets are shown. They are indicated by their
facet numbers 1 (central facet), 600 (intermediate facet), and 1046 (peripheral facet).
Figure 9 shows the offset error in the top view. In comparison with the facet error
plot in Fig. 4, similar features of symmetry and distribution of error can be observed.
Most of the peripheral facets have more offset than others due to their steep change
in orientation as they are attached to the ring truss through the anchoring cables.
Figure 10 shows the offset error distribution. The average offset error is found to
be 0.544 cm. The RMS value of the offset error computed over the entire reflector is
3.59 cm. The ratio of effective area to total area of reflector is 0.91.
Effective reflector area Fig. 11 shows how much of the facet surface participates
in reflecting the incoming rays to the focus. The orange colored portion of the facets
represents the points of reflection from where the reflected rays reach the focus and
the white portion represents those points where the rays miss the focus. This portrayal
of the focal offset error helps to see how the faceted surface and the receiver size can
be adjusted to ensure that all of it is capable of reflecting rays to the receiver. Thus,
the new measure of error emphasizes the efficacy of the faceted surface rather than
its accuracy relative to the paraboloidal surface.
3 Closure
In contrast to the computation of the facet error used widely in the literature, the focus
offset error offers a practical approach to evaluate the performance of the antenna.
How much of the reflected rays are actually able to reach the receiver is given by the
focus offset error presented here. The faceting of the reflector surface and the size of
the receiver can be optimized using the new error measure.
Acknowledgements This work was supported in part by Electronics and Radar Development
Establishment, Bengaluru, a laboratory of Defence Research and Development Organization
(LRDE-DRDO) India.
Surface Profile Accuracy of Deployable Mesh Reflectors … 1365
References
1. Miura K, Miyazaki Y (1986) Concept of the tension truss antenna. In: 37th Congress of the
International Astronautical Federation 28(6)
2. Tibert AG (2003) Optimal design of tension truss antennas. In: Structures, structural dynamics,
and materials conference, Virginia, AIAA 2001-1629
3. Morterolle S, Maurin B, Quirant J, Dupuy C (2012) Numerical form-finding of geotensoid
tension truss for mesh reflector. Acta Astronaut 76:154–163
4. Agarwal PK (1981) Preliminary design of large reflector with flat facets. IEEE Trans. Antennas
Propag AP-29(4)
5. Yuan S, Yang B Improvement of surface accuracy for large deployable mesh reflectors. In:
AIAA/AAS Astrodynamics specialist conference, AIAA 2016-5571. https://doi.org/10.2514/6.
2016-5571
Design and Control of a Low-Cost
EMG-Based Soft Robotic Ankle-Foot
Orthosis for Foot Drop Rehabilitation
Abstract Stroke patients often suffer from foot drop, a gait abnormality caused due
to the paralysis of the anterior portion muscles of the lower leg, causing an inability or
impaired ability to raise the foot at the ankle joint. This condition leads to extremities
of the foot being dragged along the ground while walking, causing tripping and other
accidents. Braces or splints that fit into shoes are prescribed to help hold the foot in
a normal position. For rehabilitation, most patients are trained to walk with canes,
and therapists prescribe physiotherapy for a series of short, intensive sessions. These
solutions are expensive and slow processes as they require the presence of skilled
personnel. In this paper, we present a novel design and control methodology for a
1-DoF Soft Active Ankle-Foot Orthosis (AFO) to address these issues. The AFO is
designed to augment the human musculoskeletal system. The AFO is actuated using
McKibben muscles (pneumatic artificial muscles), which are driven pneumatically.
They are cost-effective and lightweight, offering a significant advantage over motor-
driven orthoses. The orthosis is controlled using electromyography (EMG) signals
from the muscles involved in the motion of the ankle. The use of EMG for control is
found to be a better option than existing methods due to its non-invasive nature.
1 Introduction
According to the World Health Organization, 15 million people suffer stroke world-
wide each year. Of these, 5 million people die, and another 5 million are permanently
disabled [1]. A common ailment for stroke survivors is foot drop, i.e., the inability
to actively perform Dorsiflexion of the foot. This leads to the occurrence of steppage
Supported by Institute grants through the Robotics and Machine Intelligence Club, NIT Tiruchi-
rappalli, INDIA.
Fig. 1 Hypothetical pattern of recovery after stroke with the timing of intervention strategies.
Source [4]
gait, where the foot drags along, requiring the patient to raise their knee higher than
usual in the swing phase to counter this [2]. The patient’s foot slaps the ground during
the foot strike phase of the gait cycle. Studies have shown that the entire kinematic
chain of the lower body is disturbed, resulting in the functioning of the motor system
under abnormal load [3]. Patients generally undergo therapy to regain motor control,
and the required recovery can take from months to years to complete as shown in
Fig. 1 [4]. The available number of medical professionals is far less than the require-
ment, and they may not be accessible to all those who need it [5]. To this end, robotic
devices can significantly decrease the burden on therapists and caregivers and can
be used to provide intensive task-oriented practice.
Considerable research has gone into developing such robotic devices, especially
for upper-body rehabilitation [6–8]. Most of these devices, however, are rigid, heavy
and unwieldy. In recent years, the interest in a compliant robotic system has increased
exponentially, due to the significant advantages they provide. McKibben muscles or
pneumatic artificial muscles (PAM) are a possible actuation method with properties
that make them uniquely suited for the application presented. These properties are
explored in depth in Sect. 4.
Design and Control of a Low-Cost EMG-Based Soft … 1369
In this paper, we describe the design and control of a single DOF orthosis, actu-
ated with PAMs and controlled with electromyography (EMG) signals. The EMG
sensors are non-invasive, and the design of the orthosis is compliant, lightweight and
wearable. Section 2 explains the practices and techniques involved in the rehabilita-
tion of the foot and Sect. 3 explains the existing orthoses and exoskeletons for foot
rehabilitation. Section 4 discusses the construction and working of the prototype,
and the results are presented in Sect. 5. Some concluding remarks and future scope
are mentioned in Sect. 6.
2 Rehabilitation Techniques
2.1 Physiotherapy
2.2 Technology-Assisted
actuation devices are actuated using compliant methods which generally include
pneumatic or using shape memory alloys (SMA) [10–13]. The AFO described in
this paper falls under this category.
Fig. 2 AFO’s Layout. Components labelled from (1) to (11): (1) Arduino Uno; (2) motor drivers;
(3) solenoid valves; (4) Li-Po; (5) panel for electronic circuitry; (6) knee braces; (7) PAM mount;
(8) PAM; (9) foam padding; (10) adjustable strings; (11) orthotic sole
• Simple setup
• Minimal cost.
Pneumatic Artificial Muscle Design Each PAM consists of a silicone rubber tube,
placed within a nylon mesh sleeve. One end is pneumatically sealed, and the other
end is used as an inlet for pressurized air. When air flows into the tube, it expands
volumetrically. Since the sleeve’s structure only allows it to expand radially while
contracting longitudinally, the PAM contracts when pressurized.
Multiple studies have been performed on these PAMs, and the force-length char-
acteristics are found to be similar to that of a muscle [16]. The materials used in the
PAM are compliant, lightweight and low-cost, making them well suited for the AFO.
The analysis of various characteristics of the PAM is presented in Sect. 5. The
required length, diameter and pressure of the PAM were decided upon through exper-
iments with various configurations, presented in Sect. 5.
Orthosis Design
The computer-aided design of the prototype is presented in Fig. 2. The orthosis
consists of thigh and shin supports connected by a knee brace or a hinge joint which
offers full range of motion of the knee. This framework along with the shin of the
leg acts as the ground link for the actuating mechanism. The sole attached to the foot
is the end effector which can be modelled as hinged at the shin.
The knee brace and the sole are connected using two PAMs in an agonist-
antagonist configuration, each one providing the necessary dorsiflexion and plan-
tarflexion forces, respectively. The point of the PAM hinged to the sole is dependant
1372 N. Gudapati et al.
on the required actuation torque and range of motion. Since PAMs are shown to pro-
vide a good amount of force [16], but only limited contraction, the choice was made
to place them very close to the joint and the length of the force arm was determined
experimentally as 3 cm. The corresponding data is presented in Sect. 5. This allows
the AFO to achieve the entire Range of Motion which is roughly 20◦ for dorsiflexion
and 50◦ for plantarflexion [17].
Some similar devices [18, 19] have been constructed with the ground at the shin,
connected to the ankle end effector in a similar fashion. We found that in this config-
uration, the transverse reaction forces during PAM contraction resulted in skin defor-
mation at the skin-AFO interface causing user discomfort and restriction in the range
of motion of the AFO. By placing additional support above the knee, the reaction force
is made almost normal to the skin-AFO interface, thereby providing good support
with no loss in user comfort. Figure 3 presents the kinematic model of the designed
orthosis along with the degrees of freedom of actuation that the orthosis provides.
Electronics and Pneumatic Circuitry Each PAM is controlled by a pair of 2-
position 2-way solenoid valves, where one acts as an inlet and the other acts as an
outlet. Each pair of valves is connected to an MCU and an L298N Driver for power
Design and Control of a Low-Cost EMG-Based Soft … 1373
and logic. Two MyoWare EMG sensors are connected to the same MCU. The entire
setup is powered by a 12V lithium-polymer battery and placed above the knee to
optimize weight distribution throughout the AFO.
Both the inlet valves are connected to an air tank with pressurized air at 45 PSI, and
the exhaust valves are left open to the atmosphere as shown in Fig. 4. All pneumatic
connections were made with 8 mm OD × 6 mm OD polyurethane pneumatic tubes.
The control methodologies for the PAMs are described in Sect. 4.2.
Signal Acquisition In order to detect the intent of the user, the AFO uses MyoWare
muscle sensors and EMG electrodes. Each McKibben muscle is placed so as to assist
the functioning of a specific muscle. The targeted muscles are:
• Tibialis anterior for dorsiflexion
• Soleus for plantarflexion.
The positions of these muscles are depicted in Fig. 5.
1374 N. Gudapati et al.
Each muscle requires a non-invasive EMG sensor and each sensor requires the
placement of three electrodes to capture the differential component of the signal
travelling across the muscle. As per EMG electrode placement standards [21], one
electrode is placed near the centre of a muscle, one electrode near the end of a
muscle and a ground electrode is placed near a bone for reference. The rectified and
integrated signal is sampled by the 10-bit ADC of the MCU at a rate 200 Hz.
EMG Analysis The rectified and integrated signals (signal envelopes) from the
MyoWare muscle sensors produce output with added noise. Studies show that the
noise from sEMG signal includes both high frequency components and low frequency
components arising from movement artefacts [22]. To overcome this, the signals from
both EMG sensors were passed to two independent 1-dimensional Kalman filters.
Due to the difficulty involved in creating a dynamic model, a linear predictive model
was used, where the filtered output from the previous iteration was taken as the
initial estimate for the next iteration. It was found that a value of 0.001 for process
noise covariance gave the best results for both sensors while inducing only negligible
latency in the filtered output as shown in Fig. 6.
Design and Control of a Low-Cost EMG-Based Soft … 1375
EMG signals for a muscle are highly dependent on a multitude of factors such
as electrode placement, fatigue, and the physical and mental state of the user and
other EMG artefacts [23, 24]. Hence, real-time calibration is done before every
session. This happens automatically on initialization of the AFO. This calibration
procedure sets the threshold for each muscle to enable intent estimation, which is
described below. This also addresses the issues due to EMG degradation by adjusting
the threshold.
The rest state characteristics of each muscle are obtained by recording the unfil-
tered EMG envelope for 10 s. The mean (μrest ) and standard deviation (σrest ) of this
data are extracted from this data. Using the collected data, a threshold is calculated
for each target muscle using Eq. (1) for dorsiflexion and Eq. (2) for plantarflexion.
This threshold is then used to detect activation of each muscle. The muscle is found
to be actively contracting when the filtered signal exceeds the threshold.
For tibialis anterior muscle, the threshold is determined by Eq. (1),
Studies have shown that a threshold equal to the mean added to a multiple of
the standard deviation gives accurate results when applied to the EMG envelope
[25]. The values of the constant multiplier for standard deviation were determined
experimentally. The accuracy of intent detection with these values was tested and
the results are described in Sect. 5.
Intent Detection Logic Using the described method, we determine the state of a
muscle in real time. It was determined that the soleus muscle was active during plan-
1376 N. Gudapati et al.
tarflexion, and the initial and final stages of dorsiflexion. This is visualized in Fig. 10,
which shows plantarflexion, followed by dorsiflexion. Using this information, user
intent is detected by the method specified in Fig. 7.
Control of PAM with EMG
The user closes the feedback loop by attempting to move their ankle by muscle
contraction. This eliminates the need of external feedback and allows real-time intent
detection.
In order to contract a PAM, pressurized air is inflated into it by opening the inlet
valve and by closing the outlet valve. In order to relax (expand) a PAM, the air
inside it is exhausted by closing the inlet valve and opening the outlet valve. Table 1
summarizes the valve positions for different PAM configurations.
In order to ensure smooth movement of the foot, the speed of inflation or deflation
of the PAMs is reduced by applying a pulse width modulated (PWM) signal to the
valves. Through experiments, it was found that a PWM signal with a time period
of half a second and a duty cycle of 10% for 5 s was ideal. The duty cycle can be
modified to provide varying angular velocities. Figure 8 depicts the operation of the
AFO. The AFO is an open-loop system; however, the user closes the control-loop
while using the orthosis. This make the device simple, efficient and cost-effective.
Design and Control of a Low-Cost EMG-Based Soft … 1377
After detecting user intent, control signals from the MCU are sent to the LM298
drivers that control the pneumatic valves. Table 2 describes the state of each PAM
with respect to various user inputs.
The PAMs were tested on an experimental setup consisting of a rigid support to attach
the actuator. The other ends of the PAMs were attached to a pan to test contraction
under varying loads. The PAMs were connected to an air compressor maintained at
a constant pressure while testing. A vernier scale was fixed along the length of the
PAM.
Initially, PAMs of different lengths and diameters were tested to determine the
combination that provides maximum percentage contraction under a minimal load
of 1 kg and 45 PSI pressure. These results are presented in Table 3. PAM of 30 cm
length with an inner diameter of 1/3 in. was found to be optimal due to its percentage
contraction and satisfying the length constraints of the average leg.
1378 N. Gudapati et al.
Once the PAM characteristics were decided, the length of the force arm was
determined by testing the contraction of the PAM under various pressures and loads
in the abovementioned experimental setup. This data is presented in Fig. 9. The length
of force arm was selected as 3 cm by taking into account the maximum pressure of
the tank and the load borne by the actuator, and the required range of motion which
is 20◦ for dorsiflexion, 50◦ for plantarflexion [17].
The AFO was tested by three of the authors as the particants for the study. The
participants flexed their legs over the entire range of motion for both dorsiflexion
and plantarflexion, and the percentage accuracy of intent detection in every case
Design and Control of a Low-Cost EMG-Based Soft … 1379
was calculated and presented. Table 4 presents the data when they were relaxed and
Table 5 after a while when fatigued.
The control scheme achieves a mean dorsiflexor accuracy of 97.22% and inaccu-
racies being 1.11% false-positive and 1.67% false-negative. The mean plantarflexion
accuracy is 93.33%, and the inaccuracies are 1.11% false-positive and 5.56% false-
negative.
The control scheme achieves a mean dorsiflexor accuracy of 92.22% and inaccu-
racies being 2.22% false-positive and 5.56% false-negative. The mean plantarflexion
accuracy is 90.56% and the inaccuracies are 3.88% false-positive and 5.56% false-
negative.
From the results mentioned in Tables 4 and 5, it is found that the simple con-
trol scheme presented is able to achieve a mean dorsiflexion accuracy of 94.72%,
and inaccuracies being 1.66% false-positive and 3.62% false-negative. The mean
plantarflexion accuracy is 91.94% and the inaccuracies are 2.5% false-positive and
5.56% false-negative. It is worth noting that the accuracy decreases minimally when
the participants are fatigued.
Performance The orthosis exhibited a good response time to the changes in the EMG-
based muscle activation signal with minimal lag due to the simple control logic. The
absence of a feedback control signal also eliminates the time delay introduced by
sensors and error detectors. The activation function’s output for both the muscles is
shown in Fig. 10. An ergonomic and compliant prototype of the orthosis was made
as a proof of concept. The completed prototype is shown in Fig. 11.
1380 N. Gudapati et al.
Fig. 10 Voltage-time graph depicting filtered sensor data and the corresponding thresholds
6 Conclusions
The design and control of a 1-DOF Soft Ankle-Foot Orthosis presented in this paper
shows promise to be a therapeutic alternative for foot drop rehabilitation. The AFO
uses EMG signals which are reliable and easy to calibrate. The main purpose of the
work is to present the scope of soft-robotic devices as an alternative for rigid and
heavy hardware previously used for robotic rehabilitation. The AFO uses PAMs,
which are pneumatically actuated, low-cost and lightweight. The torque provided
by the AFO can be increased by enhancing air pressure driving techniques, and
this could allow the use of the device as an exoskeleton. Also, machine learning
algorithms could be used to estimate the intent of the user with greater accuracy. The
device could also be further developed by incorporating more degrees of freedom.
This would allow the AFO to perform inversion and eversion of the leg.
References
1. World Health Report—2002 from the World Health Organization Retrieved on: 2019 from
http://www.who.int/whr/2002/en/
2. Walking abnormalities MedlinePlus Retrieved on: 2019 from https://medlineplus.gov/ency/
article/003199.htm
3. Wiszomirska I, Błażkiewicz M, Kaczmarczyk K, Brzuszkiewicz-Kuźmicka G, Wit A (217)
Effect of drop foot on spatiotemporal, kinematic, and kinetic parameters during gait. Appl
Bionics Biomech. 2017:3595461. https://doi.org/10.1155/2017/3595461
4. Langhorne P, Bernhardt J, Kwakkel G (2011 ) Stroke rehabilitation, stroke care 2, The lancet,
vol 377
5. Khor KX, Rahman HA, Fu SK et al (2014) A novel hybrid rehabilitation robot for upper and
lower limbs rehabilitation training. In: Internatioinal conference on robot pride (ConfPRIDE)
2013–2014, procedia computer science, vol 42, pp 293–300
6. Asokan A, Vigneshwar M, Design and control of an EMG-based low-cost exoskeleton for
stroke rehabilitation. In: 2019 fifth Indian control conference (ICC), New Delhi, India, pp
478–483. https://doi.org/10.1109/INDIANCC.2019.8715555
7. Kim B, Deshpande AD (2017) An upper-body rehabilitation exoskeleton harmony with an
anatomical shoulder mechanism: design, modeling, control, and performance evaluation. Int J
Rob Res (IJRR) 36(4):414–435
8. Perry JC, Rosen J, (26) Design of a 7 degree-of-freedom upper-limb powered exoskeleton. The
First IEEE, RAS-EMBS Interenational conference on biomedical robotics and biomechatron-
ics, (2006) BioRob 2006. Pisa 2006:805–810
9. Stroke rehabilitation: What to expect as you recover (2017) Retrieved on 2
June 2019, from https://www.mayoclinic.org/diseasesconditions/stroke/in-depth/stroke-
rehabilitation/art-20045172
10. Ikuta K (1990) Micro/miniature shape memory alloy actuator. In: Proceedings IEEE interna-
tional conference on robotics and automation, vol 3. Cincinnati, OH, USA, pp 2156–2161.
https://doi.org/10.1109/ROBOT.1990.126323
11. Park Y-L, Chen B, Pérez-Arancibia NO, Young D, Stirling L, Wood RJ, Goldfield EC, Nagpal
R (2014) Design and control of a bio-inspired soft wearable robotic device for ankle-foot reha-
bilitation. Bioinspiration Biomimetics 9(1). https://doi.org/10.1088/1748-3182/9/1/016007
12. Gordon KE, Sawicki GS, Ferris DP (2006) Mechanical performance of artificial pneumatic
muscles to power an ankle-foot orthosis. J Biomech 39(10):1832–1841. https://doi.org/10.
1016/j.jbiomech.2005.05.018
1382 N. Gudapati et al.
1 Introduction
The mechanical systems having less control inputs as compared to degree of freedom
(DOF) of the system comes under the underactuated system family. These systems
have diverse applications in the field of aerospace, marine engineering, flexible
manipulators, mobile robots, etc. The use of such kind of systems is not limited
up to the cost reduction of the system but they play significant role in fault-tolerant
design, lower power consumption, lower environment impact, etc. Although, these
systems have several advantages, but these systems are difficult to control due to
G. Singh (B)
School of Mechanical and Materials Engineering, University College Dublin, Dublin, Ireland
A. Singla
Mechanical Engineering Department, Thapar Institute of Engineering and Technology, Patiala,
India
2 Experimental Setup
Fig. 1 a Actual setup and b schematic diagram of Googoltech Linear Inverted Pendlum (GLIP)
The rotary encoder was placed on the pendulum for the rotating feedback consid-
ering pendulum downward position as zero reference and left/right motion as posi-
tive/negative magnitude. As it is a closed loop system, first the initial input to the
motor was provided with the help of Simulink interface with the help of motion
controller card. Further, signal data transfers to the AC motor drive which amplifies
the data and gives command to the motor. In the same manner, the feedback from
both encoders reverts to the Simulink interface with help of driver and controller
card. The conversion factors were used to convert the signal into angular and linear
data. The parameter used for GLIP was shown in Table 1.
The Euler-Lagrange is used to drive the analytical model of the system. Figure 2b
depicts the schematic line diagram of the two DOF system. The cart movement is
restricted to a particular plane to move in horizontal direction and pendulum is free
to move in vertical and horizontal direction in a plane. Consider that M and m are the
mass of cart and pendulum, l is the distance between the center of pendulum and the
pendulum cart joint, b is the friction coefficient for the cart. Moreover, x(t) and φ(t)
are horizontal and angular position of the cart and inverted pendulum from stable
equilibrium position and f (t) is the horizontal force acting on the cart. The state
vector and its derivative of the system are represented in Eqs. (1) and (2). Further,
the total kinetic energy is calculated given as Eq. (3).
xp x − l sin φ
z(t) = = , (1)
yp l cos φ
ẋ p ẋ − l φ̇ cos φ
ż(t) = = . (2)
ẏ p −l φ̇ sin φ
K = K Cart + K Pend
= 21 (M + m)ẋ 2 + 21 I φ̇ 2 + 21 ml 2 φ̇ 2 − ml cos φ̇ ẋ (3)
The potential energy of the system only depicts from the pendulum movement
and contains only the strain energy stored in the pendulum and given as Eq. (4).
P = Pcart + PPend
= 0 + (−mg(l − l cos φ) (4)
1 1 1
L= (M + m)ẋ 2 + I φ̇ 2 + ml 2 φ̇ 2 − ml cos φ̇ ẋ − mg (5)
2 2 2
The damping factor due to friction between cart and shaft is calculated by
Rayleigh’s dissipation function and given as Eq. (6)
R = 21 b ẋ 2 . (6)
I + ml 2 φ̈ − ml cos φ ẍ − mgl sin φ = 0. (8)
The dynamic model of the AC servo motor is coupled with the derived
mathematical model of the system to develop a realistic dynamic model of the system.
Im Bm Km Kb Km
M +m+ ẍ − ml cos φ φ̈ + ml sin φ φ̇ 2
+ b + + ẋ = Vc .
r2 r2 Rm r 2 Rm r
(9)
4 Controller Design
PPC method is the classical and cheap approach to stabilize the unstable closed loop
system. In this method, a feedback controller is designed to place the eigenvalues
of the closed loop system at the required place. The controller is designed to satisfy
the transient and steady-state configuration of the system. Pole Placement control
design consists of multiple steps. First, the coefficient vector a = [a1 , a2 , . . . , an ] for
the polynomial of state matrix A is calculated as given in Eq. (11). The state-space
equation transforms into the controller companion form by using T given in Eq. (12).
In addition, the desired polynomial is calculated by using desired eigenvalues from
Eq. (13) and determine the coefficient vector α = [α1 , α2 , . . . , αn ]. Further, the state
feedback gain matrix K can be calculated from Eq. (14).
T = PW (12)
⎡ ⎤
an−1 an−2 . . . a1
1
⎢ an−2 an−3 1 0 0⎥
⎢ ⎥
⎢ .. ⎥
W = ⎢ ... ..
.
.. ..
. . .⎥ (13)
⎢ ⎥
⎣ a1 1 ... 0 0⎦
1 0 ... 0 0
Comparison of PPC and LQR Controller for Stabilization … 1389
(s − μ1 )(s − μ2 ) . . . (s − μn )
= s n + α1 s n−1 + · · · + αn−1 s + αn. = 0 (14)
To design optimal controller for a closed loop system, Linear Quadratic Regulator
(LQR) controller is used. However, the given controller is difficult to customized and
required heavy calculations. The LQR controller based on the minimization of the
integration of the sum of state weighting (Q) and control cost matrix (R) of the
system. The object function used in LQR control design is elucidated in Eq. (16).
Further, the optimal feedback control matrix (K LQRC ) can be calculated as Eq. (17).
The control matrix depends upon the algebraic Riccati Equation given by Eq. (18).
∞
J∞ (t) = (z T (τ )Qz(τ ) + u T (τ )Ru(τ )dτ, (16)
0
K LQRC = R −1 B T M0 , (17)
M0 A + A T M0 − M0 B R −1 B T M0 + Q = 0. (18)
It is observed that the feedback control matrix directly depends upon Q and R.
These matrices have to be determined by manual tuning, which increase the time
cost of the controller.
Simulation and experimental studies are demonstrated in the present section to check
the efficacy of both controllers. The calculated eigenvalues for the state matrix A are
[0, −250.86, −5.64, 5.64]. The present system is unstable having one of the positive
eigenvalue. Thus, a closed loop controller is required to design to shift the eigenvalue
to the negative direction. The step response of the system is calculated to find out
the gains of the PPC controller. The settling time of the step response is constraint
√to
be within√in 2 s. Therefore, the desired poles are selected as [−10, −10, −2 +j 3,
−2 − j 3]. PPC controller gains are found to be KPPC = [−54.4218, −24.4898,
93.2739, 16.1633]. PPC with designed configurations is used to stabilize the inverted
pendulum and both experimental and simulation are compared. To decide the initial
position for φ for stabilization, the pendulum lifted in anti-clock direction manually
to area of unstable equilibrium (±20◦ from the upper reference point). It is found
that the controller starts work at −19.5 and stabilize the inverted cart pendulum
at upper equilibrium point in 2 s. The overshoot value for the experimental results
found to be 5° (Ref. Fig. 2a). Figure 2b depicts the displacement of cart to stabilize
1390 G. Singh and A. Singla
the pendulum with PPC. It has found that the 0.15 m overshoot distance is taken
by cart from reference point to stabilize the system. The whole system stabilizes
within 2 s. In analytical results pendulum and cart overshoots to 11° and 0.3 m for
the stabilization of system, which is higher as compared to experimental results. The
gap between analytical and experimental results is due to the real-time activation of
linear controller.
Figure 2c depicts the control input comparison between analytical and experi-
mental results. The control input magnitude found to be 50 units for analytical study
and 48 units for experiment. The maximum input magnitude of both analytical and
experimental results is found to be near, which validate the mathematical modeling of
the system. To develop LQR controller for the present system, the Q and R matrices
are decided manually by hit and trail in order to settle the system with 2 s.
The gain matrix for LQRC is calculated as K LQRC = [−31.628, −20.1507, 72.718,
13.152] using Eqs. (16)–(18). To stabilize the pendulum at upright position, the next
test run is executed with LQR controller, in the same manner, as explained earlier
with Pole Placement controller. The pendulum stabilizes at its inverted position with
a settling time of 2 s and a marginal overshoot difference of 6° between experimental
and analytical results, as shown in Fig. 3a. It can be observed from Fig. 3b that the
maximum overshoot value for cart displacement is 0.17 m for analytical results and
0.12 m for experimental results. LQR controller is able to stabilize the cart to its
desired position within 2 s. The comparison of control input is shown in Fig. 3b,
where the maximum control input magnitude is found as 25 for the experimental run
and 20 for the simulation run.
The comparison between two controllers is discussed. Both the controllers can
stabilize the cart-pole system. From the experimental and analytical results, it is
observed that the settling time for both controllers is near 2 s for stabilization. For
Fig. 4 Comparison between PPC and LQRC for a control inputs and b pendulum position at
stabilize equilibrium
the case studied, the control input magnitude by both the controllers from experi-
mental results is compared in Fig. 4a. From the results, it is found that the maximum
control input magnitude for stabilization of inverted pendulum using Pole Placement
controller is 48 units, whereas, it is found as 25 units with the LQR controller. Hence,
LQR controller performs better than the Pole Placement controller, as it has resulted
in 47.9% reduction in maximum control input. Further, experimental results show
that the pole is oscillating marginally around its upward position. It can be seen
from Fig. 4b that the pendulum oscillates in the range of −0.6° to 0.6° using LQR
controller and between −1.35° and 0.6° using Pole Placement controller. There-
fore, low oscillations motion was observed in case of LQR controller as comparison
with Pole Placement controller. In other words, the LQR controller results in 55.5%
reduction in the oscillation of the inverted pendulum.
6 Conclusions
This paper has investigated the real-time control of cart-pole system, which is under-
actuated in nature. Two linear controllers (PPC and LQRC) has been used to solve this
problem. Both the controllers have been designed to stabilize the system within 2 s.
The actuator dynamics has been incorporated in the mathematical model to ensure
the accuracy of the analytical model, which is validated through the close agreement
of the simulation results with the experimental results. Control input magnitude by
both controllers to the system is compared, which results in the LQRC outperforming
the PPC as it results in 47.9% reduction in maximum control input magnitude and
55.5% reduction in the oscillations of the inverted pendulum at unstable equilibrium
position.
References
1. Mason P, Broucke M, Piccoli B (2008) Time optimal swing-up of the planar pendulum. IEEE
Trans Autom Control 53(8):1876–1886
1392 G. Singh and A. Singla
2. Yang J, Shim S, Seo J, Lee Y (2009) Swing-up control for an inverted pendulum with restricted
cart rail length. Int J Control Autom Syst 7(4):674–680
3. Mus N, Tovornik B (2006) Swinging up and stabilization of a real inverted pendulum. IEEE
Trans Industr Electron 53(2):631–639
4. Solihin MI, Akmeliawati R (2010) Particle swam optimization for stabilizing controller of a
self-erecting linear inverted pendulum. Int J Electr Electron Syst Res 3(June):13–23
5. Chaturvedi NA, Mcclamroch NH, Bernstein DS (2008) Automatica stabilization of a 3D axially
symmetric pendulum. Automatica 44:2258–2265
6. Googoltech (2016) Inverted pendulum experimental manual suitable for GLIP Series
7. Song Z, Song X, Liu C, Zhao Y (2013) Research on real-time simulation and control of linear
1-stage inverted pendulum. J Comput 8(4):896–903
8. Roshdy AA, Wang T (2012) Stabilization of real inverted pendulum using pole separation
factor. In: Proceedings of 2012 international conference on mechanical engineering and material
science (MEMS). (Mems), pp 711–715
9. Singla A, Singh G, Virk GS (2016) Matlab/simMechanics based control of four-bar passive
lower-body mechanism for rehabilitation. Perspect Sci 8:351–354
10. Singh G, Singla A, Virk GS (2016) Modeling and simulation of a passive lower-body
mechanism for rehabilitation. In: Conference on mechanical engineering and technology
(COMET-2016), IIT (BHU), Varanasi, India
11. Singla A, Singh G (2017) Real-time swing-up and stabilization control of a cart-pendulum
system with constrained cart movement. Int J Nonlinear Sci Numer Simul 18(6):525–539
12. Singh G, Singla A (2017) Modeling, analysis and control of a single stage linear inverted
pendulum. In: 2017 IEEE international conference on power, control, signals and instrumen-
tation engineering (ICPCSI), pp 2728–2733 (2017)
Effective Education Using a 2-DOF
Five-Bar Mechanism
Abstract Innovative methods to teach kindergarten students are on the rise which
use virtual environment like a mobile phone or a tablet. However, the students who
get great joy in using these gadgets are moving away from actual writing on a
paper or slate. In this paper, the authors have proposed a simple mechanism which
can be controlled to draw a shape, a numeral, or a letter. It consists of a five-bar
mechanism which has 2 degrees-of-freedom (DOF) and can control the position of
the end-effector (EE). The novelty of the proposed mechanism is that it has an iris
mechanism at the EE. The student has to hold a pen/pencil and start following the
motion of the EE, which draws as per the command from a computer. The pen/pencil
has a contact patch and if it touches the inner circle of the iris mechanism, the
computer notes it down and the student loses a point. The iris mechanism can be
opened and closed to have different difficulty levels for the students, thus making its
usage more challenging. A working prototype was tested with few children, and based
on their feedback, a newer version is being developed and has also been reported in
this paper.
1 Introduction
In earlier days, teachers used to hold the wrist of the students in the entry level (now
kindergarten students) and guide them on writing shapes, numerals, or alphabets
on slates by using chalks. Though this tradition still exists in rural India, most of
the schools in urban areas have resorted to books which have dotted lines on which
students have to write and practice. Though it is convenient for teachers, the students
tend to take more time to master the art of tracing on dotted line. Also with the
exposure of smartphones and tablets, a plethora of mobile applications are available
which let students trace letters on the screen of the gadgets using their finger. Some of
these applications also have a game element to it to motivate the students to perform
better. The authors have tried to bring together the aspects of writing on a paper and
a game environment in the proposed mechanism.
Mechanisms to write on a paper/surface have been developed by a number of
researchers. Since the end-effector of a mechanism has to trace a curve in 2D space,
2 degree-of-freedom (DOF) mechanisms can achieve this. They can be broadly clas-
sified as open-loop and closed-loop mechanisms. Open-loop mechanisms can be in
the form of 2R (R: revolute) or 2P (P: prismatic) planar mechanisms. The forward
kinematics in such mechanisms is straight forward but the inverse kinematics usually
has multiple solutions. Though the construction of these mechanisms is simple, they
are prone to inaccuracies due to lesser rigidity and bending of links with time. An
example of a serial mechanism used for writing is reported in [1] and a commercially
available 2P plotter is AxiDraw [2]. Closed-loop mechanisms can also be used to
write on a paper. These have unique inverse kinematics but might have multiple solu-
tions for forward kinematics. Though they have better rigidity, they have lesser range
or workspace as compared to a serial chain of similar dimensions. A widely used
example of a closed-loop mechanism for writing is a 2-DOF pantograph or five-bar
mechanism. Collaborative Haptics and Robotics in Medicine (CHARM) Laboratory
at Stanford University has a dedicated Web resource on the kinematic analysis of the
five-bar mechanism [3]. The five-bar mechanism was used in [4] to write letters by
taking input from the user, by creating a human–machine interface (HMI). Another
category of mechanisms uses cables for driving the motion of the end-effector link.
H-Man cable-driven mechanism is one such variant [5].
In this paper, the authors have used a 2-DOF five-bar mechanism with some modi-
fications at the end-effector. These modifications include using an iris mechanism
and letting the student trace the shape/letter/numeral by keeping a pen/pencil within
the inner circle of the iris mechanism. The details of the mechanism, the integration
of the system, and results are discussed in the following sections.
A five-bar mechanism consists of five links of which one is grounded. The other four
moving links are connected as a closed loop and have five revolute joints, as shown
in Fig. 1a. The degree-of-freedom (DOF) of the mechanism can be determined as 2
using the Grubler–Kutzbach equation as
where N = number of links, i.e., 5; P1 = number of joints with 1 relative DOF, i.e.,
5; and P2 = number of joints with 2 relative DOF, i.e., 0. Though any two joints can
be chosen to be driven or active joints, the joints on either side of the fixed link, i.e.,
Effective Education Using a 2-DOF Five-Bar Mechanism 1395
S O
X
Y
R
P
Q
T
(a) Five-bar mechanism (b) Iris mechanism
J1 and J5 , are used as it enables the placement of the motors on the fixed link, thus
reducing the moving mass of the mechanism.
The link length of links is a1 through a5 . The Link 3 is extended by a3 beyond
Q to allow fixing of an iris mechanism at point T. A mathematical model of the
mechanism, known as the kinematic model, exists to correlate the joint angles at
joint J1 and J5 , i.e., θ1 and θ5 , and the position of the point T, as explained next.
Iris mechanism on the other hand has one link that can be fixed and several curved
links connected to it through revolute joints. The curved links have an extrusion which
go inside slots of another floating part. The geometry is such that floating part forms
a virtual revolute joint with the fixed link. This is due to the synchronous motion of
the curved links with respect to the central axis. Hence, by controlling the movement
of the floating component, the radius of the circle (polygon) formed by the curved
links can be changed. In Fig. 1b, the circular disk is the fixed link and the curved
links are colored. The floating part is not shown for better visualization. In this paper,
the iris mechanism is used in a novel way, as explained later.
Forward Kinematics: For the given joint angles θ1 and θ5 , the determination of the
coordinates of point Q and thereafter point T is known as forward kinematics. Two
solutions of the point Q exist for given values of θ1 and θ5 . This can be visualized as
the intersection of two circles with radii a2 and a3 , centered at P and R, respectively.
One of the intersection points is the Q as shown in Fig. 1a. The other solution is the
other intersection points of the two circles. The formulations are explained in [6].
Inverse Kinematics: Determination of the angles θ1 and θ5 for joints J1 and J5 for
a given location of point Q (or T ) is known as inverse kinematics. If the coordinates
of T are known, one can solve the inverse kinematics by drawing a circle centered
at T with (a3 + a3 ) as radius and another circle with a4 as radius and S as its center.
These two circles intersect at two points (hence two solutions of θ5 ). Of these two
solutions, one of them would be point R in Fig. 1a. Q point can be determined using
the values of a3 and a3 . To determine point P, draw a circle with radius a2 centered
1396 S. S. Grama et al.
3 Prototype I
The links of the five-bar mechanisms were created using aluminum sheet metal,
as shown in Fig. 2a. The joints J1 and J5 were powered using two servomotors.
These motors were rigidly fixed to the ground and the distance between the axes of
the motors acted as dimension a5 . A readymade iris mechanism was procured and
attached at the end of the Link 3. The iris mechanism had a lever, rotation of which
caused the opening and closing of the inner curved metallic links, thus changing the
radius of the circle inside the iris mechanism.
The servomotors were connected to an Arduino Uno controller board, as shown
in Fig. 2b. The inner links of the iris mechanism were connected to one of the digital
Fig. 3 Numerals drawn using a pen clamped in iris mechanism to check the motion trace
pin of Arduino board and a pen with a metallic strip on its body was connected to a
digital input. Whenever a contact is made between the metallic strip and the iris links,
the circuit is completed and an LED is connected to show the output as a contact.
The Arduino is connected to a computer through a USB cable. On the computer, a
client application was developed using Visual C#, which would send out the angles
(θ1 and θ5 ) to the Arduino board through serial port communication.
Using the inverse kinematics formulations discussed earlier, motion planning was
performed in the client application for each of the numerals (0–9). To test the working,
a pen was grasped tightly in the iris mechanism and the mechanism was controlled
from the computer. The drawing of the pen for few of the numerals is shown in Fig. 3.
The kids enjoy to play games and through the proposed mechanisms, and the authors
are trying to integrate learning with a gaming element. For any given numeral
(decided by the computer), the end-effector/iris mechanism moves to trace the
selected numeral. The user (a student) has to hold the pen (with a metallic strip
on its body) and has to follow the iris mechanism. If there is a contact between the
pen with the inner circle of the iris, it is noted as a penalty or error by the user.
In the initial trials, the user has a tendency to make more mistakes, and for the
repeated trials, the number of errors should reduce. In addition to the above feature,
the iris mechanism also allows the user to select different levels of complexity. If
the inner radius of the iris is more, the user has more tolerance to make an error and
with smaller radius, and the error tolerance is strict. Hence, there are multiple ways
in which the performance of a user can be modified.
The prototype developed was tested over 5 students in the age group of 7–12 years.
The number of errors made by students for fully open (big ring) and semi-open (small
ring) conditions for five trials was noted down. It was observed that for the big ring, the
number of errors with usage was reducing. However, with the small ring, the number
of errors did not have a direct correlation with the usage. This could be attributed
to the inherent issue of the servomotors lacking good resolution. Sample results of
two students, Person 1 and Person 2 (both 8-year-old female writing numeral 5), are
shown in Fig. 4a, b, respectively.
1398 S. S. Grama et al.
Fig. 4 Game results: left column is for big ring and right column is for small ring
4 Prototype II
The prototype explained in the previous section had servomotors with limited reso-
lution. Also, the workspace available for writing was restricted to a small region.
A new software application was developed in Visual C# to simulate the writing of
different numerals and letters, and this is illustrated in Fig. 5a. Also, in the newer
version of the prototype, high resolution DC stepper motors are being used. To have
an increased area of the workspace [8], the length of the fixed link was made zero
such that the axis of the joints J1 and J5 are collinear, as shown in Fig. 5b. The links
and connectors available in MAKIT toolkit were used to develop the prototype. The
integration of the software application (developed using Visual C#) and the hardware
(mechanisms) are underway. The pedagogy results with more users shall be reported
in the due course.
5 Conclusions
In this paper, five-bar and iris mechanisms have been used to develop a novel method-
ology to provide effective education. The letters and numerals to be learnt are fed
into the mechanism to trace and the user has to follow the iris mechanism attached
on one of the links of the mechanism. Whenever the user’s pen touches the inner
circle of the iris mechanism, error is noted, thus bringing in the elements of gaming
and education together. A prototype was developed and was used by a few students.
Based on the feedback and observations, a new prototype is being developed which
shall be evaluated and reported in the due course.
References
1. Tresset P, Leymarie FF (2013) Portrait drawing by Paul the robot. Comput Graph 37(5):348–363
2. AxiDraw: Website accessed in Jan 2018. https://www.axidraw.com/
3. CHARM Lab Project Page: Website accessed in Jan 2018. https://charm.stanford.edu/ME327/
JaredAndSam
4. Campolo D, Tommasino P, Gamage K, Klein J, Hughes CM, Masia L (2014) H-Man: a planar,
H-shape cabled differential robotic manipulandum for experiments on human motor control. J
Neurosci Methods 235(1):285–297
5. Can FC, Sen H (2017) Real time controlled two dof five bar robot manipulator. In: Proceedings
of the AzC-IFToMM international symposium of mechanism and machine science
6. Hampali S, Chittawadigi RG, Saha SK (2015) MechAnalyzer: 3D model based mechanism
learning software. In: 14th World Congress in Mechanism and Machine Science
7. Zhao L, Yen AKW, Coulombe J, Bigras P, Bonev IA (2014) Kinematic analyses of a new medical
robot for 3D vascular ultrasound examination. Trans Canad Soc Mech Eng 38(2):227–239
8. Koul MH, Rabinowitz D, Saha SK, Manivannan M (2013) Synthesis and design of a 2-DOF
haptic device for simulating epidural injection. In: 13th World Congress in Mechanism and
Machine Science
Renewable Energy System Using
Thermoelectric Generator (RESTEC)
Abstract Continuous variable transmission (CVT) systems are widely used in all-
terrain vehicles (ATV). Over long runs, the sheaves of CVT generate sufficient
amount of heat energy and get damaged due to friction between the sheaves and
belt mainly due to clutch slippage during acceleration and deceleration. This further
deteriorates the condition and performance of the vehicle, and breakdown is also
discovered as a worst-case scenario. A novel method has been developed and imple-
mented by creating a renewable energy system to maintain the temperature of CVT
and improve longevity. The renewable energy system makes use of thermoelectric
generators (TEG) working on the principle of Seebeck effect to utilize the heat dissi-
pated from the exhaust chamber of engine which comprises of 80% of the waste
energy of the ATV and is used for cooling the CVT.
1 Introduction
2 Previous works
Dhongde et al. [3] performed an experiment to map the temperature from CVT
by placing thermocouple sensors at various locations inside the CVT housing and
measuring CVT components temperatures with the help of IR sensor as shown in
Fig. 1. This would help in better understanding the airflow [4] and also help in
Table 1 Placement
IR Temperature sensor
information
1. Primary sheave
2. Secondary sheave
Thermocouple sensor
3. Driver region
4. Central region
5. Driven region
cause in restricting their use in electrical power generation to specialized fields with
extensive applications where reliability is a major consideration and cost is not.
An easy of improving efficiency and longevity of thermoelectric generators is
to keep the temperature difference between the two sides as low as possible which
can be done with the help of heatsinks [9]. Figure 3 shows the heatsink used on the
thermoelectric generators during competition to improve cooling.
3 RESTEC
3.1 Design
at 1.5 A is to be maintained which provided sufficient results after trial runs with
variations of V max and I max of modules over a span of 1 h.
The electricity generated is used as power source for cooling of CVT but it also
has another application, i.e., recharging of lithium polymer (LiPo) battery which is
used in the all-terrain vehicle as power source of brake light as well as other sensors.
A battery isolator is used to reduce battery exhaustion with time.
STAGE 3: The main goals of the design are completed within the first two stages
as explained above. The third stage deals with improving the system to adapt to the
condition of sensors as well as system to increase or decrease efficiency depending
upon the dynamic state. The feedback loop is defined as shown in Fig. 4, which makes
use of a microcontroller unit that reads the incoming sensor readings to analyze the
current state, i.e., the temperature of CVT components and LiPo battery level.
A microcontroller unit is used which regulates the power supply input given to
Stage 2 thermoelectric generator modules. The contactless IR temperature sensor
is used for measuring the CVT component temperature at the primary sheave and
secondary sheave. A voltage detector is used as feedback from LiPo battery to the
microcontroller unit. This voltage detector sensor feeds when the battery is charged
above 70% to microcontroller unit, which on receiving this data stops the charging
and all the electricity generated is focused on CVT cooling. Similarly, the contactless
IR sensor at sheaves of CVT continuously monitors the temperature of sheaves in
real time, and if the CVT is running in the proper temperature range, the electricity
generated is shifted toward battery recharging only. Although during the test runs,
this case would never occur as the temperature of CVT needed to be constantly
maintained within the limited temperature range as it would heat up instantly after
long runs.
3.2 Implementation
The RESTEC design makes use of TEC1-12706 Peltier modules for thermoelectric
generation which has a maximum operating temperature of 138 °C. The temperature
of various parts of the engine was checked using contactless IR temperature sensor,
and the top face of exhaust chamber was chosen as suitable option for placement as it
did not exceed beyond 120 °C at maximum. Along with help from data acquired from
Fig. 2, the placement of TEC on the CVT was on spread over the top of primary sheave
and central region as it gave comparatively better results by hit and trial method. An
Arduino Uno was used as the microcontroller unit for making the system adaptive.
Figure 5 shows the basic setup on the all-terrain vehicle.
1406 R. Dhar et al.
4 Observation
RESTEC was successfully tested and gave decent results in test runs. The changes
in hardware with respect to the feedback loop were initially slow for which changes
in the algorithm were made to provide faster refreshing and response rate.
A comparative analysis of the all-terrain vehicle in two conditions, i.e., with the
model and without the model mounted was considered to determine the efficiency
of the system at three positions. The three positions taken for readings were the
temperature at primary sheave, secondary sheave using MLX906014 sensor, and
region between the sheaves using thermocouple sensor as noted by region 4 (central
region) in Fig. 1. Table 2 shows the temperature difference and output efficiency of
Renewable Energy System Using Thermoelectric Generator (RESTEC) 1407
the system taken over a continuous period of 1 h of testing and noted at four intervals
of 15 min.
1408
5 Conclusion
The system was considered successful after assessing the difference in tempera-
ture without its implementation. The all-terrain vehicle performed at annual BAJA
SAEINDIA competition with the system incorporated and no CVT-related break-
downs occurred during the competition. The system was also presented at the Inno-
vation Design Competition at BAJA SAEINDIA and was awarded the first prize out
of 120 teams performing from all over the country.
Table 2 gives the test readings of the system. Although the system is successful
in accomplishing its main objective, it can be seen that the system can be seen dete-
riorating linearly with time. This may be due to the continuous working of semicon-
ductors under harsh conditions. Further works to improve the overall system would
involve working on custom thermoelectric generators with different semiconductor
materials to improve maximum operating temperature without affecting longevity
and also a more packed transmission housing to maintain a distinct temperature
enclosure.
References
1. Aaen O (2011) Olav Aaen’s clutch tuning handbook. Racine, WI, AAEN performance
2. Wurm J, Fitl M, Gumpesberger M, Väisänen E, Hochenauer C (2017) Advanced heat transfer
analysis of continuously variable transmissions (CVT). Appl Thermal Eng 114:545–553. ISSN
1359-4311
3. Dhongde SV, Chandran V (2014) Experimental study of cooling of continuously variable
transmission (CVT) in scooter. https://doi.org/10.4271/2014-01-2003
4. Vaishya A, Phadnis S (2013) Experimental investigations of forced air cooling for continuously
variable transmission (CVT). SAE Technical Paper 2013-32-9073
5. ElCosnier W, Gilles M, Lingai (2008) An experimental and numerical study of a thermoelectric
air-cooling and air-heating system. Int J Refrig 31:1051–1062
6. Jugsujinda S, Vora-ud A, Seetawan T (2000) Analyzing of thermoelectric refrigerator perfor-
mance. In: Proceedings of the 2nd international science, social-science, engineering and energy
conference, vol 25, pp 154–159
7. Yadav N, Meha N (2013) Review on thermoelectric materials and applications. Int J Sci Res
Dev 1:413–417
8. Ismail BI, Ahmed WI (2010) Thermoelectric power generation using waste-heat energy as an
alternative green technology. Recent Patents Electr Eng 2. https://doi.org/10.2174/187447611
0902010027
9. Chein R, Chen Y (2005) Performances of thermoelectric cooler integrated with microchannel
heat sinks. Int J Refrig 28:828–839
Bending and Free Vibration Analysis
of Exponential Graded FG Plate Using
Closed-Form Solution
Abstract This article is intended to study the bending and vibration response of
the exponentially graded FGM plate using the closed-form solution. The governing
equations of motions are formulated on the basis of Hamilton’s principle, whereas the
Navier method has been used to solve the governing equation with simply supported
boundary conditions. The results have been validated for non-dimensional deflection
values of a square plate subjected to transverse loading. Apart from these, the expo-
nential gradation model has been considered for calculating the effective material
properties of a functionally graded plate and its mechanical properties are assumed
to vary in the thickness direction. Bending analysis and non-dimensional frequency
has been evaluated for various modes of vibration of functionally graded thin and
thick plates.
1 Introduction
Functionally graded materials (FGMs) are the materials in which the properties of
the materials vary exponentially about the thickness coordinates. Extensive literature
has been published in the field of functionally graded material structures. Koizumi
[1] introduced the concept of FGMs. Reddy [2] presented the closed-form solution
to investigate the dynamic response of FGM plates based on the third-order shear
deformation theory (TSDT). Abrate [3] proposed that the functionally graded (FG)
plates can be treated like homogeneous plates after considering the suitable reference
surfaces. Mantari and Guedes [4] investigated the flexural response of exponential
graded plates using trigonometric higher-order shear deformation theory (HSDT).
Thai and Kim [5] examined the bending and vibration characteristics of FG plates
using HSDT. Sburlati [6] presented the structural response of FG multi-layered struc-
tures. Chakraverty and Pradhan [7] examined the vibration characteristics of expo-
nential FG plates with various boundary conditions. Zenkour [8] investigated the
effect of thermo-mechanical loading on structural response of exponentially graded
thick plates. Ebrahimi and Mokhtari [9] analyzed exponentially graded Timoshenko
beams. Gupta and Talha [10] provided an extensive literature review on the recent
development on the structural responses of FG structures. Akavci [11] studied the
mechanical behavior of FG sandwich plates resting on elastic foundations. Li et al.
[12] investigated the thermal analysis of exponentially graded plates. Pandey and
Pradyumna [13] analyzed FG sandwich plates using higher-order layer-wise theory.
Zenkour and Alghanmi [14] employed refined quasi-3D shear normal and defor-
mation theory to investigate the structural characteristics of functionally graded
plates. Zenkour and Alghanmi [15] analyzed exponentially graded plates with piezo-
electric composite actuators on elastic foundations. Mantari and Guedes [16] have
investigated the exponentially graded plate using trigonometric higher-order shear
deformation theory.
2 Mathematical Formulation
3
4z
f (z) =
3h 2
where h is the thickness of the plate, ũ and ṽ represents the displacements in the x
and y direction in midplane of functionally graded plate. w̃b and w̃s represent the
bending and shear component of transverse displacements [5]
∂ ũ 1 ∂ ũ ∂ 2 w̃b 4z 3 ∂ 2 w̃s
ε̃x = = −z − 2
∂x ∂x ∂x2 3h ∂x2
∂ ũ 2 ∂ ṽ ∂ w̃b
2
4z ∂ 2 w̃s
3
ε̃ y = = −z − 2
∂y ∂y ∂ y2 3h ∂ y2
∂ ũ 1 ∂ ũ 2 ∂ ũ ∂ ṽ ∂ 2 w̃b 8z 3 ∂ 2 w̃s
γ̃x y = + = + − 2z − 2
∂y ∂x ∂y ∂x ∂ x∂ y 3h ∂ x∂ y
2
∂ ũ 1 ∂ ũ 3 4z ∂ w̃s
γ̃x z = + = 1− 2
∂z ∂x h ∂x
2
∂ ũ 2 ∂ ũ 3 4z ∂ w̃s
γ̃ yz = + = 1− 2 (2)
∂z ∂y h ∂y
Equation (2) shows the transverse shear strains (γ̃x z , γ̃ yz ) are zero at the top (z =
+h/2) and bottom (z = −h/2) surfaces of the plate.
1414 D. Singh et al.
+ 2
h
+ 2
h
(ñ x , ñ y , ñ x y ) = σ̃x , σ̃ y , σ̃x y dz
− h2
+ 2
h
(m̃ bx, m̃ by, m̃ bx y, ) = σ̃x , σ̃ y , σ̃x y zdz
− h2
+ 2
h
(m̃ sx, m̃ sy, m̃ sx y, ) = σ̃x , σ̃ y , σ̃x y f (z)dz
− h2
+ h2
4z 2
P̃x y = 1− σ̃x y dz
h2
− h2
+ 2
h
4z 2
P̃yz = 1− σ̃ yx dz (4)
h2
− h2
+ 2
h
(I0, I1, I2, I3, I4, I6 ) = 1, z, z 2 , z 3 , z 4 , z 6 ρ(z)dz (7)
− h2
T
0= (δU +δV − δ K )dt (8)
0
where
1 Ec 1 ρc
δ1 = ln , δ2 = ln .
2 Em 2 ρm
⎧ ⎫ ⎡ ⎤⎛ ⎧ ∂ ũ ⎫ ⎧ ∂ 2 w̃ ⎫ ⎧ ∂ 2 w̃ ⎫⎞
⎨ ñ x ⎬ 1v 0 ⎨ ∂x ⎬ ⎪
⎨ − ∂2x 2
b
⎪
⎬ ⎪
⎨ − ∂2x 2
s
⎪
⎬
⎜ ⎟
ñ y = ⎣ v 1 0 ⎦⎝d0 ∂∂ ṽx + d1 − ∂∂ yw̃2b + cd3 − ∂∂ yw̃2s ⎠
⎩ ⎭ ⎩ ∂ ũ ∂ ṽ ⎭ ⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
+ −2 ∂∂ x∂w̃yb −2 ∂∂ x∂w̃ys
2 2
ñ x y 0 0 1−v
2 ∂x ∂x
(13)
⎧ ⎫ ⎡ ⎤⎛ ⎧ ∂ ũ ⎫ ⎧ ∂ 2 w̃ ⎫ ⎧ ∂ 2 w̃ ⎫⎞
⎪
⎨ m̃ x ⎪
b
⎬ 1v 0 ⎨ ∂x ⎬ ⎪
⎨ − ∂2x 2
b
⎪
⎬ ⎪
⎨ − ∂2x 2
s
⎪
⎬
⎜ ⎟
m̃ by = ⎣ v 1 0 ⎦⎝d1 ∂∂ ṽx ∂ w̃b
+ d2 − ∂ y 2 ∂ w̃s
+ cd4 − ∂ y 2 ⎠
⎪
⎩ m̃ b ⎪⎭ ⎩ ∂ ũ ∂ ṽ ⎭ ⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
+ −2 ∂∂ x∂w̃yb −2 ∂∂ x∂w̃ys
2 2
xy 0 0 1−v
2 ∂x ∂x
(14)
⎧ ⎫ ⎡ ⎤⎛ ⎧ ∂ ũ ⎫ ⎧ ∂ 2 w̃ ⎫ ⎧ ∂ 2 w̃ ⎫⎞
⎪
⎨ m̃ x ⎪
s
⎬ 1v 0 ⎨ ∂x ⎬ ⎪
⎨ − ∂2x 2
b
⎪
⎬ ⎪
⎨ − ∂2x 2
s
⎪
⎬
⎜ ⎟
m̃ sy = ⎣ v 1 0 ⎦⎝cd3 ∂∂ ṽx ∂ w̃b
+ cd4 − ∂ y 2 ∂ w̃s
+ c d5 − ∂ y 2
2
⎠
⎪
⎩ m̃ s ⎪⎭ ⎩ ∂ ũ ∂ ṽ ⎭ ⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
+ −2 ∂∂ x∂w̃yb −2 ∂∂ x∂w̃ys
2 2
xy 0 0 1−v
2 ∂x ∂x
(15)
( ) ∂ w̃s
P̃x z 10
P̃yz
= As ∂x
∂ w̃s (16)
01 ∂y
+ 2
h
E(z)
(d0 , d1 , d2 , d3 , d4 , d5 ) = 1, z, z 2 , z 3 , z 4 , z 6 dz (17a)
1 − v2
− h2
+ 2
h
4z 2 E(z)
A =
s
1− 2 dz (17b)
h 2(1 − v)
− h2
Navier solution has been used to find the bending response of functionally graded
plate, and initial solutions can be assumed as shown in Eq. (19).
∞ *
* ∞
ũ(x, y, t) = Ũmn eiωt cos αx sin βy
m=1 n=1
∞ *
* ∞
ṽ(x, y, t) = Ṽmn eiωt sin αx cos βy
m=1 n=1
*∞ * ∞
w̃b (x, y, t) = W̃bmn eiωt sin αx sin βy
m=1 n=1
∞ *
* ∞
w̃s (x, y, t) = W̃smn eiωt sin αx sin βy (19)
m=1 n=1
√
where i = −1, α = mπ a
and β = nπ
b
, angular frequency (ω). (Ũmn , Ṽmn , W̃bmn and
W̃smn ) all are coefficients. Transverse load P has been expanded into double Fourier
sine series as shown in Eq. (20) and in Eq. (21), P̃mn represents the load coefficient.
∞ *
* ∞
p(x, y) = P̃mn sin αx sin βy (20)
m=1 n=1
The coefficients P̃mn are given below for the some typical loads
a b
4
P̃mn = p(x, y) sin αx sin βy (21)
ab
0 0
Bending and Free Vibration Analysis of Exponential Graded … 1419
Substitute Eqs. (20), (21) and (22) in motion Eq. (18), the analytical solution can
be obtained from:
K̃ − ω2 M̃ Q̃ mn = P̃mn (22)
4×4 4×4 4×1 4×1
Q̃ mn is the displacement coefficient matrix where (Ũmn ,Ṽmn , W̃bmn and W̃smn ) are
coefficients of this matrix. Stiffness matrix coefficients and mass matrix coefficients
of FG plates are calculated from the above Eq. (22).
1−v 1+v
k11 = d0 α 2 + d0 β 2 , k12 = d0 αβ, k13 = −d1 α(α 2 + β 2 ),
2 2
1−v
k14 = −cd3 α(α 2 + β 2 ), k22 = d0 α 2 + d0 β 2 , k23 = −d1 β(α 2 + β 2 )
2
k24 = −cd3 β(α 2 + β 2 ), k33 = d2 (α 2 + β 2 )2 , k34 = cd3 (α 2 + β 2 )2 ,
k44 = c2 d5 (α 2 + β 2 )2 + As (α 2 + β 2 ).
m 11 = m 22 = I0 , m 12 = 0, m 13 = −α I1 , m 14 = −αcI3 , m 23 = −β I1 ,
m 24 = −βcI3 , m 33 = I0 + I1 α 2 + β 2 , m 34 = I0 + cI4 α 2 + β 2 ,
m 44 = I0 + c2 I6 α 2 + β 2 .
Firstly, the validation study has been carried out to confirm the accuracy of the present
methodology. Table 2 shows the non-dimensional (ND) frequency of exponentially
graded Al/Al2 O3 plates. The obtained results have been compared with the results
given by Chakraverty and Pradhan [7]. Results obtained for verification purpose have
been found to be in good agreement with previously published works.
4 Conclusions
In this study, the bending and free vibration analysis of functionally graded square
plate have been analyzed using HSDT. The study provided the valuable results in
terms of vibration characteristics and effects of Young’s modulus on displacement in
the thickness direction. The results showed that the change in Young’s modulus ratio
Bending and Free Vibration Analysis of Exponential Graded … 1423
at constant densities of metal and the ceramics has pronounced effects on frequencies
of the used materials. Significant reductions in frequencies have been reported in
results. Similarly, the change in density ratio has also led to the significant increment
in frequency of the materials.
References
Abstract This article deals with the structural analysis of geometrically imper-
fect functionally graded plates (FGM) with microstructural defects (porosity) under
a hygrothermal environment. The mathematical formulation is based on non-
polynomial-based quasi-3D higher-order structural kinematics developed by the
authors. The comparative studies have been carried out by comparing the obtained
results with the published results. Finally, the influence of various parameters such
as moisture contents, thermal environment, and various defects on the vibration and
bending of FGM plates has been studied.
1 Introduction
Functionally gradient materials (FGMs) are a heterogeneous material which has been
progressively applied for ultra-modern engineering structures subjected to extreme
temperature gradient [1]. Since the last few decades, remarkable studies have been
performed on the structural response of FGM structures. Nguyen-Xuan et al. [2] used
an isogeometric formulation with the use of NURBS to investigate the structural
response of the FGM plate. Zenkour [3] investigated the vibration response of FGM
plates using a non-polynomial-based deformation plate theory. Gupta and Talha [4–
6] analyzed the linear/nonlinear structural response of gradient plates with geometric
and microstructural imperfection. Wu et al. [7] studied the structural response of the
FGM plate subjected to dynamic load using the modified Pagano method.
In the present study, the objective is to illustrate the vibration and the bending
response of imperfect porous FGM plates under a hygrothermal environment. The
A. Gupta (B)
Mechanical Engineering Department, Shiv Nadar University, Chennai, India
e-mail: [email protected]
M. Talha
School of Engineering, Indian Institute of Technology, Mandi, India
geometrical imperfection has been induced in the plate using imperfection function,
which is capable to model the plate with various types of geometrical imperfections.
The material property of FGM plates with microstructural defects (porosity) has been
estimated using the modified rule of mixture. The mathematical formulation has been
done using previously developed displacement field by the authors. The influence
of the various parameters like geometric imperfection, porosity inclusion, moisture
contented, and temperature gradient on the vibration and the bending response of
FGM plates have been investigated in detail.
2 Mathematical Formulation
Structural kinematics developed by the authors’ [8] has been employed in the present
study as given below,
⎧
⎪ U i (x1 , x2 , x3 ; t) = u i (x1 , x2 ; t) − x3 αxi
⎪
⎪
⎨ + f (x )β (x , x ; t) + g(x )θ (x , x ; t)
3 xi 1 2 3 xi 1 2
⎪ Where, for i = 1, 2; {θi } = 0
⎪
⎪
⎩
for i = 3; {αxi , βxi } = 0
⎧
⎪
⎪ hcosh2 (/2) x3
⎪
⎪ f (x3 ) = − −1 sinh−1 − x3
⎨ h h
1 + 2 /4 −1
(1)
⎪
⎪ x x
⎪
⎪
⎩ g(x3 ) = cosh2
h
The field variables are represented as {D} = u i αxi βxi θxi .
2
εxi = εxi
0
+ zεxi
1
− C sinh−1 xh 3 − h x3 εxi , for i = 1, 2
εxi = εxi
0
for i = 3
Structural Responses of Geometrically Imperfect Functionally … 1427
⎧
⎪ x3
⎪
⎪
⎪
0
γx1x2 = γx1x2 1
+ x3 γx1x2 − C sinh−1 − 2
x3 γx1x2
⎪
⎪ h h
⎪
⎪ ⎛ ⎞
⎨
1 2 x 3 β (x, y, t)γ 2 ,
0
γxi x3 = γxi x3 + C h ⎝ ⎠ 1
⎪
⎪
⎪ 2 − 1 γxi x3 + cosh h
z xi x3
⎪
⎪ 1 − x3 h
⎪
⎪
⎪
⎩
for i = 1, 2
⎧
⎪ 2 x 3 x3 nl/2
⎪ nl nl/0
⎨ εxi = εxi + cosh
nl/1
εxi + 2 cosh4 εxi for i = 1, 2
h h
⎪
⎪ x3 x3
⎩ γx1x2
nl nl/0
= γx1x2 + cosh2
nl/1
γx1x2 + 2 cosh4
nl/2
γx1x2
h h
where
⎧
⎪ ∂u i
⎪
⎪ εxi = ∂ xi
0
⎪
⎪
⎪
⎨ ∂αxi C ∂βxi
εxi
1
=− − for i = 1, 2
⎪
⎪ ∂ xi h ∂ xi
⎪
⎪
⎪
⎩ ε2 = C ∂βxi
⎪
xi
∂ xi
The properties of the porous FGM plate are presumed to vary gradually through the
thickness as given in Eq. 2.
2z + h n 2|z|
E(z, T ) = [E c − E m ] − (λ/2)[E c + E m ] 1 − + Em
2z h
2z + h n 2|z|
ρ(z) = [ρc − ρm ] − (λ/2)[ρc + ρm ] 1 − + ρm
2z h
2z + h n 2|z|
α(z, T ) = [αc − αm ] − [αc + αm ] 1 − + αm
2z h
2z + h n
kt (z) = [ktc − ktm ] + ktm (2)
2z
{σ}6x1 = Qi j {ε} 6x1 − {α}6x1 T − {β}6x1 C (3)
1428 A. Gupta and M. Talha
⎧ ⎫ ⎡ ⎤
⎪
⎪ σx x ⎪
⎪ Q 11 Q 12 Q 13 0 0 0
⎪
⎪ ⎪
⎪
⎪
⎪ σ yy ⎪
⎪
⎪
⎪
⎢Q Q Q
⎢ 21 22 23 0 0 0 ⎥
⎥
⎨ ⎬ ⎢ ⎥
σzz ⎢Q Q Q 0 0 0 ⎥
= ⎢ 31 32 33 ⎥
⎪ τx y ⎪
⎪ ⎪ ⎢ 0 0 0 Q 44 0 0 ⎥
⎪
⎪ ⎪ ⎢ ⎥
⎪
⎪
⎪ τ yz ⎪
⎪
⎪
⎪
⎣ 0 0 0 0 Q 55 0 ⎦
⎩ ⎭
τx z 0 0 0 0 0 Q 66
⎛⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎞
⎪
⎪ εx x ⎪
⎪ ⎪
⎪ 1⎪⎪ ⎪
⎪ 1⎪⎪
⎜⎪
⎪ε ⎪ ⎪ ⎪ ⎪ 1⎪
⎪ ⎪
⎪ 1⎪
⎪ ⎟
⎜⎪
⎪
⎪ yy ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ⎟
⎜⎨ ⎬ ⎨ ⎬ ⎨ ⎪ ⎬ ⎟
⎜ εzz 0 0 ⎟
×⎜ − α(z, T ) − β(z, C)⎟
⎜⎪ γ ⎪ ⎪ 1 ⎪ ⎪ 1 ⎪ ⎟
⎜⎪
⎪ xy ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎪ ⎪
⎪ ⎟
⎝⎪
⎪
⎪ γ ⎪
⎪
yz ⎪
⎪
⎪
⎪ ⎪0 ⎪
⎪ ⎪
⎪
⎪ 0 ⎪
⎪
⎪ ⎠
⎪
⎩ ⎭ ⎪
⎪ ⎩ ⎪ ⎭ ⎩ ⎪
⎪ ⎭
γx z 0 0
E(z, T ) 1 − υ 2
Q 11 = Q 22 = Q 33 =
1 − 2υ 2 − 2υ 3
E(z, T )υ(1 + υ)
Q 12 = Q 21 = Q 13 = Q 31 = Q 23 = Q 32 =
1 − 2υ 2 − 2υ 3
E(z, T )
Q 44 = Q 55 = Q 66 =
2(1 + υ)
In the present analysis, FE formulation has been employed using C0 isoparametric element
with nine DOFs per node. The shape functions of the selected element are as follows:
⎧ 1
⎪
⎪ ξ 2 + ξi ξ η2 + ηi η , for i = 1, 2, 3, 4
⎪
⎪ 4
⎪
⎪
⎪
⎪ 1
⎪
⎨ 1 − ξ 2 η2 + ηi η , for i = 5, 7
Ni = 2 (4)
⎪
⎪ 1
⎪ ξ 2 + ξi ξ 1 − η2 , for i = 6, 8
⎪
⎪
⎪ 2
⎪
⎪
⎩ 1 1 − ξ 2 1 − η2 , for i = 9
⎪
2
Structural Responses of Geometrically Imperfect Functionally … 1429
The characteristics equation for vibration, and bending response of the FGM plate
is obtained using Lagrange equation as follows:
$ %
d ∂ ∂ ∂V
+ + = 0 for i = 1, 2, . . .
dt ∂ Ḋi ∂{Di } ∂{Di }
To validate the present solution, some illustrations have been given in the first part.
The initial geometric imperfection has been included using the following function
[9].
X1 X1
ζ = hς sech δ1 − ψ1 cos μ1 π − ψ1
a a
X2 X2
× sech δ2 − ψ2 cos μ2 π − ψ2
b b
In the first section, the comparison study has been performed. The ND frequency has
been calculated and compared with reported results given by Zhu and Liew [10] as
shown in Fig. 2.
1430 A. Gupta and M. Talha
In this segment, the influence of different parameters like moisture contents, porosity
inclusions, and thermal environment on the deflection and vibration response have
been explored. Ti–6AL–4V/Si3 N4 FGM plate has been considered for the analysis.
The following non-dimensional (ND) parameters are used in the present study.
Structural Responses of Geometrically Imperfect Functionally … 1431
ND frequency
ω=ω 12(1 − ν 2 )ρc a 2 b2 /π 4 E c h 2
ND deflection
w = 10wE c h 3 Qa 4
is also an important parameter while dealing with the FGM structures subjected to
the hygrothermal environment.
In Fig. 6, the ND deflection of FGM plate under the hygrothermal environment
has been computed. The T has been considered as 200 K, whereas the moisture
content (C) is considered as 1. The central deflection of the FGM plate has been
plotted against ‘ς ’ for different various values of ‘n’. It is evident from Fig. 6 that
Structural Responses of Geometrically Imperfect Functionally … 1433
Fig. 5 Influence of moisture contents (C) and ‘n’ on ND frequency of FGM plate
as ‘ς ’ increases from 0 to 0.4, the central deflection decreases. It is also found that
the ND deflection decreases as ‘n’ increases. Therefore, it is again verified that the
geometric imperfection will lead to make the plate stiffer.
In Fig. 7, the ND central deflection has been plotted for the different values of
‘n’ and ‘λ’. Again the T and C are considered as 200 K and 1, respectively. It
is found that the ND deflection increases as ‘n’ and ‘λ’ increases. Therefore, it is
1434 A. Gupta and M. Talha
concluded that the porosity inclusion in the structures will result in the increment of
the flexibility of the plate.
Figure 8 depicts the change of ND deflection with the moisture contents (C). T
is taken as 200 K. It is obvious from the figure that as the moisture content increases,
the ND deflection increase for all values of ‘n’.
5 Conclusion
In this study, a non-polynomial HSNDT is implemented for the bending and vibra-
tion response of geometrically imperfect porous FGM plate under a hygrothermal
environment. It is concluded that the consideration of the moisture contents as well
as temperature gradient during the structural analysis of the FGM plate is important
because it influences the mechanical behavior of the plates significantly.
References
1 Introduction
The perfection of bird flight is always a source of attraction to the researcher. The
physics of bird flight is a complex phenomenon and still a hot area of research. A bird
has large lungs to flap the wings at higher frequency for longer time and hollow bones
to get the high buoyant mass. The wing’s structure of a bird is like a human hand and
mainly contributes to the dynamic control of the bird flight. The wings can be divided
into primary and secondary feathers. During the upstroke, primary feathers are rotated
and separated, and therefore allowing air to pass through them. This phenomenon
reduces the air resistance on the wing and provides some thrust during the upstroke of
the flapping flight. However, during the downstroke, the primary feathers are closed
and hit the surrounding air to produce the lift and thrust forces. The secondary feathers
provide lift force by creating the airfoil structure of the wings. The aerodynamic
characteristics of bird wings were studied at Reynolds number of 1 − 5 × 104 , and a
correlation with morphological parameters such as nose radius, camber, and aspect
ratio was established in [1]. The researchers from zoology and biology have examined
many complicated aerodynamics phenomena that happens during the flapping wing
flight of birds and insects [2–4]. The maneuverability and ability of bird flight and
inertial power required for wing beat are related to the mass moment-of-inertia of the
wings [5]. In [5], strip theory was used to estimate the moment-of-inertia of the wings
of 29 birds and three bats. A study on upward stroke of wingbeat was presented in [6].
A computational simulation of a robotic bird and its prototype was developed in [7,
8] for different flight modes such as take-off, landing, and cruising. In [9], modified
strip theory was used to formulate the aerodynamic model of a semi-elliptical wing,
and parametric study was done to describe the effect of different parameters on
lift, thrust, and drag force. A nonparametric mathematical model of a flying bird was
developed in [10] from flight test data using the time domain identification technique.
The singular value decomposition (SVD) and modal analysis were performed to get
the minimal order system identification model. The flapping wing aerodynamics
for the three-class of species, i.e., birds, bats, and insects were studied in [11]. The
motion of the flapping wing is quite complex and is characterized by unsteady, three-
dimensional flow, where the state of the wings change every instant of flapping flight
[12]. A detailed study on aerodynamics of a bird flight, such as induced drag, lift,
and thrust forces due to wing-generated vortices, airfoil shape of the secondary wing,
spread out of the primaries at the tip, etc., was discussed in [12]. A comprehensive
sizing method for flapping wing micro-air vehicle (FWMAV) based on statistical
and theoretical analyses was proposed in [13], which was verified with the help of a
FWMAV, namely, Thunder I. The design and operation of a microrobot capable of
swimming, flying, and transitioning between water and air was presented in [14]. The
multibody dynamic approach for the mathematical modeling of space vehicle was
proposed in [15, 16] using the DeNOC matrices [17, 18]. In this paper, the kinematic
configuration of the bird model is represented using the modified DH parameters and
dynamic modeling using the DeNOC approach. A simplified aerodynamic modeling
of a robotic bird is presented, where variable drag and lift coefficients were used
to perform the dynamic simulations. The joint torques for the desired sinusoidal
trajectory of the wingbeat were generated through PD controller.
The multibody dynamic approach is used to model a bird flight. The dynamic model
of a flying bird is required for precise simulation and control of bird flight [8]. Here,
Simplified Aerodynamic Modeling of a Bird Robot … 1439
a robotic bird was divided into three modules, namely main body, right wing, and
left wing. The kinematics and dynamics of the bird model are presented next.
Z9
X9 , X10
Left wing
Z6
Z
X3 Z X4
3
Z0 , Z1 X2 Z4 , X5 , X6
b3 X7 , X8
Z2 Z5
b2 Z8 Main body of
b1 robotic bird
X1
Y0
Right wing
X0 Z7
0 1 O
where pi ≡ , and Ai,i−1 ≡ , in which 0 is three-dimensional
ei ai,i−1 × 1 1
vector of zeros, ei is the unit vector along ith joint axis, 1 and O are 3 × 3 identity
and null matrices, respectively. The three-dimensional vector ai,i−1 represents the
position vector between two consecutive origins. The set of generalized coordinates
T
q ≡ b1 b2 b3 θ4 θ5 θ6 θ7 θ8 θ9 θ10 and its time derivative are represented as q̇.
The twist vector for the complete system can be expressed as
t = Nl Nd q̇ or t = Nq̇ (2)
The matrices Nl and Nd are known as the decoupled natural orthogonal comple-
ment matrices for the bird model under study, whose combined form was originally
proposed for the serial-chain system in [19] as natural orthogonal complement matrix
N.
The dynamic equations of motion for an ith rigid body using the Newton–Euler
equations can written as
where Iio is the inertia tensor about the origin, m i is mass, ωi is angular velocity,
oi is position vector of origin with respect to the inertial frame, di is the position
Simplified Aerodynamic Modeling of a Bird Robot … 1441
vector from origin to the center-of-mass (CoM). The three-dimensional vector nio
and fio are the moment about and force at the origin of the ith body, which are the
combination of gravity and buoyancy, drag and lift, and external moments and forces.
The matrix–vector form of the above equations can be represented as
where
Iio mi d × 1 ωi × 1 O 1 O
Mi ≡ ; Wi ≡ and Ei ≡ .
−m i d × 1 m i 1 O ωi × 1 OO
The wrench vector wi is the combination of gravity and buoyancy wrench wib , drag
and lift wrench wid , and external wrench wie . The Newton–Euler equations for the
complete system can be expressed as
M ṫ + W MEt = w (5)
Also, the 6n-dimensional generalized wrench vectors due to gravity and buoyancy,
wb , drag, wd , and external moments and forces, we , are given by
T T
T T T T
wb ≡ w1b · · · wnb ; wd ≡ w1d · · · wnd ; and
T
(7)
T T
we ≡ w1e · · · wne
The reduced-order equations of motion are obtained by pre-multiplying Eq. (5) with
the transpose of the natural orthogonal matrix N ≡ Nl Nd given by Eq. (2). The
resulting equations are as follows:
NT M ṫ + W ME t = NT w (8)
The dynamic simulation of the ten-DoF robotic bird was performed by integrating
Eq. (10) numerically in MATLAB. The physical parameters of the bird model are
summarized in Table 2. The initial state vector for the dynamic simulation is given
below:
T
y = y0T ẏ0T ; y0 = 0 0 0 0 3π 2 0 π 2 0 3π 2 0
ẏ0 = 2 2 0 0 0 0 0 0 0 0 (11)
The desired trajectory for the flapping was considered as sinusoidal, which is
written below:
θ8d = π 6 sin(2π f t); θ̇8d = π 6 (2π f ) cos(2π f t) (12)
The joint torque for the flapping wing can be calculated through proportional-
derivative controller as
The same joint torque input was given to the flapping joint of the left wing, i.e.,θ10 .
The value of k p and kd were taken as 5 and 2, respectively. The results for the lift
and forward motion are shown in Figs. 2 and 3.
The lift force generated through flapping of the wing is shown in Fig. 4.
0.4
Lift of the bird (m)
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4
Time (s)
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4
Time (s)
1444 A. K. Sharma et al.
10
-5
0 0.1 0.2 0.3 0.4
Time (s)
The kinematic configuration and dynamic modeling of a robotic bird were presented
using the modified DH parameters and DeNOC approach, respectively. The sinu-
soidal joint trajectories were used to get the joint torques using PD controller. The
proposed dynamic and aerodynamic model was able to provide enough lift and
forward thrust to move the robotic bird. There is a little drop of the bird during the
upward stroke of the wings as shown in Fig. 2. Figure 3 shows that there is minimal
loss of kinetic energy in the forward direction due to streamline characteristics of
main body of the bird. The lift is also increasing in each cycle of wingbeat, and the
mean of the lift force is very close to the weight of the robotic bird. The proposed
aerodynamic model is used to generate preliminary simulation results and later will
be used for the simulation and control of a more complex robotic bird.
References
1. Withers PC (1981) An aerodynamic analysis of bird wings as fixed aerofoils. J Exp Biol
90:143–162
2. Norberg UM (1990) Vertebrate flight: mechanics, physiology, morphology, ecology and
evolution. Springer, New York. ISBN 13:978-3-642-83850-7
3. Spedding GR (1992) The aerodynamics of flight. Mech Animal Locomot 11:52–111
4. Tobalske BW, Dial KP (1996) Flight kinematics of black- billed magpies and pigeons over a
wide range of speeds. J Exp Biol 199:263–280
5. Berg C, Rayner J (1995) The moment of inertia of bird wings and the inertial power requirement
for flapping flight. J Exp Biol 198(8):1655–1664
6. Poore SO, Sanchez-Haiman A, Goslow GE Jr (1997) Wing upstroke and the evolution of
flapping flight. Nature 387(6635):799
7. Wu JC, Popović Z (2003) Realistic modeling of bird flight animations. In: ACM Trans Graph
(TOG) 3:888–895
Simplified Aerodynamic Modeling of a Bird Robot … 1445
8. Couceiro MS, Figueiredo CM, Ferreira NF, Machado JT (2009) The dynamic modeling of a bird
robot. In: Proceedings of the 9th conference on autonomous robot systems and competitions
(Robotica’09)
9. Malik MA, Ahmad F (2010) Effect of different design parameters on lift, thrust, and drag of
an ornithopter. In: Proceedings of the World Congress on Engineering, vol 2, pp 1460–1465
10. Shepherd S, Valasek J (2012) Modeling and analysis of eagle flight mechanics from experi-
mental flight data. In 50th AIAA Aerospace Sciences Meeting including the New Horizons
Forum and Aerospace Exposition 2012, p 27
11. Chin DD, Lentink D (2016) Flapping wing aerodynamics: from insects to vertebrates. J Exp
Biol 219(7):920–932
12. Dvořák R (2016) Aerodynamics of bird flight. In: EPJ Web of Conferences, vol 114, p 01001,
EDP Sciences
13. Hassanalian M, Abdelkefi A, Wei M, Ziaei-Rad S (2017) A novel methodology for wing
sizing of bio-inspired flapping wing micro air vehicles: theory and prototype. Acta Mech
228(3):1097–1113
14. Chen Y, Wang H, Helbling EF, Jafferis NT, Zufferey R, Ong A, Ma K, Gravish N,
Chirarattananon P, Kovac M, Wood RJ (2017) A biologically inspired, flapping-wing, hybrid
aerial-aquatic microrobot. Sci Robot 2(11) eaao5619
15. Saha SK (1993) Modeling and simulation of space robots. In: Proceedings of 1993 IEEE/RSJ
International conference on intelligent robots and systems (IROS’93), vol 3, pp 2033–2040
16. Saha SK (1996) A unified approach to space robot kinematics. IEEE Trans Robot Autom
12(3):401–405
17. Saha SK (1999) Analytical expression for the inverted inertia matrix of serial robots. Int J
Robot Res 18(1):20–36
18. Saha SK (2014) Introduction to robotics, 2nd edn. Tata McGraw-Hill Education, New Delhi
19. Angeles J, Lee S (1988) The formulation of dynamical equations of holonomic mechanical
systems using a natural orthogonal complement. Trans ASME J Appl Mech 55:243–244
Parallel Mechanism-Based Master–Slave
Manipulation
1 Introduction
There exist two types of the manipulators by architecture design. (1) Serial manip-
ulator (SM), (2) parallel manipulator (PM). The links are connected in articulated
chain with end-effector at the tip in SMs, whereas PM is a closed-loop kinematic
mechanism whose end-effector (platform) is connected to the base by independent
serial kinematic chains. These chains are similar and possess one controllable DoF.
SMs have large workspace but they are very sensitive to kinematic positioning errors
because error of each joint adds up in sequence. The load handling capability of SMs
is less in comparison with PMs because the load is distributed among several parallel
links [1]. PMs can provide higher accuracy and can be faster in manipulation; there-
fore, they are preferred for the tasks which require precision and agility. This paper
deals with kinematic analysis and construction of a 6 DoF parallel mechanism-based
master–slave (M-S) framework. Both the master and slave are based on parallel kine-
matic configuration. The operator through master serves to provide joints trajectory
and control. The direct kinematics problem in serial chain will yield a unique solu-
tion at the task space of the slave manipulator. The simpler direct kinematic problem
apart from their higher workspace is also the reason for the popularity of the serial
mechanism-based servo M-S manipulators which is one of the earliest architecture.
However, low payload to weight ratio, low stiffness, accumulation and amplification
of positioning errors leading to low accuracies are certain weak factors that led to an
alternative framework for M-S arrangements. The parallel architecture-based mech-
anism which exhibits reciprocal properties to serial chain is the obvious choice. The
limited working volume, complexity in singularity identification and highly complex
direct kinematic problem are the factors for its absence in M-S arrangement. The
paper deals with the ways of addressing the above factors and proposes methods to
overcome the issues.
Fig. 1 Parallel
mechanism-based 3 DoF
M-S arrangement
Parallel Mechanism-Based Master–Slave Manipulation 1449
correspondence of master and the slave connectors along the active axis for identical
motion response at the slave for the given motion input at the master is shown.
competitive merits suitable for applications where active and real-time control is
required.
Serial mechanism-based master–slave manipulators are very popular and widely
used for various applications. But parallel mechanism-based master–slave manipu-
lation is yet to be established because of the unavailability of closed-form solution of
FKP of 6 DoF parallel mechanism and limited workspace. But there are many appli-
cations where the workspace is not the constraint like neurosurgery, etc. We have
developed an approach for master–slave manipulation with dissimilar structures of
master and slave arm as shown in Fig. 4.
We have developed the 6 DoF passive master manipulator. The magnetic linear
encoders are attached with each leg for feedback of 6 leg lengths. We compute the
leg lengths of 6 links using FPGA-based encoder interface. After the computation
of leg lengths, the forward kinematics algorithm computes the task space which is
transmitted to the slave manipulator after proper scaling the task space required for
dissimilar structures.
We have developed and tested many algorithms of the forward kinematics for fast
and accurate convergence, out of which one algorithm is used in this master–slave
manipulation.
The forward and the inverse kinematic equations of serial and parallel manipula-
tors show duality. Whereas the forward kinematics of SM is comparatively easy to
1452 R. Kumar et al.
Table 1 Computation time and avg. no. of iterations for FKP algorithm
1 Average computation time for FKP algorithm 50 ms per pose
2 Average no. of iterations (real-time implementation) 20 per pose
compute, the same for PM is challenging issue to solve. But the scenerio is vice-
versa in case of inverse kinematics. The forward kinematic equations of parallel
manipulators are highly nonlinear, coupled and exhibit multiple solutions for the
same input. A closed-form solution is not available generally. Many schemes require
further simplification and reformulation of equations. Numerical methods with fast
computations are used to get time optimal accurate solution [5]. The kinematic equa-
tions for the 6DoF UPS-based parallel manipulator are formulated and converted to
reduce canonical form to implement a Newton–Raphson (N-R) solver. The output
from the N-R solver is not necessarily the desired solution. Due to the highly coupled
and nonlinear nature of the equations, the output may be a pose that is significantly
far away (in terms of Euclidean distance) from the actual pose of the manipulator.
This happens if the input trajectory is complex and/or when the seed-point given to
the solver falls outside the sphere of influence of the desired solution. To check for
correct convergence, the previous converged pose is checked against the pose that the
N-R solver converges to. In the case of incorrect convergence, i.e., a considerably
large Euclidean distance between the previous converged pose and the converged
pose, the resolution of the trajectory is doubled by re-formulating the input argu-
ments for the N-R solver. This was done by taking the mid-point of the leg lengths
for previous pose (correctly converged) and current leg lengths. The resolution is
doubled till a correct convergence is achieved. These intermediate values are stored
separately. Once any particular convergence occurs correctly (as per the set target leg
lengths), the target leg lengths are then set to be the actual leg lengths. This process is
continued till the output from the N-R solver correctly converges to the actual pose.
The intermediate values inserted in between the original pose values in the trajectory
forms the modified trajectory. This modified trajectory ensures a smoother motion of
the manipulator and correct convergence of the forward kinematics during operation
by creating the required resolution in the trajectory. The following observances made
have been shown in Table 1.
The computer used for FKP computation has a 3.30 GHz i3-3220 processor by
Intel and 4 GB RAM.
3 Communication Channel
Currently, Ethernet has been used as the channel to establish communication between
the master and the slave node [6].
Parallel Mechanism-Based Master–Slave Manipulation 1453
4 Slave Manipulator
5 Conclusion
Fig. 5 Prototype of 6 DoF master and 6 DoF slave manipulator in M-S arrangement
References
1. Patel YD, George PM (2012) Parallel manipulators applications—a survey. Mod Mech Eng
2:57–64
2. Lagoo KD, Dwarakanath TA, Badodkar DN (2015) Single actuator shaker design to generate
infinite spatial signatures. In: 2nd International and 17th National Conference on Machines and
Mechanisms iNaCoMM, p 55
3. Dwarakanath TA, Lagoo KD, Badodkar DN (2017) 6-PSS based parallel manipulators. In:
Corves B, Lovasz EC, Hüsing M, Maniu I, Gruescu C (eds) New advances in mechanisms,
mechanical transmissions and robotics. mechanisms and machine science, vol 46. Springer,
Berlin
4. Dwarakanath TA, Bhutani G (2011) Beam type hexapod structure based six component force-
torque sensor. Mechatronics 21(8):1279–1287
5. Merlet J-P (2004) Solving the forward kinematics of Gough-type parallel manipulator with
interval analysis, vol 23, pp 221–236
6. Advanced Master Slave Manipulators. https://barc.gov.in/technologies/msm/msm.html
7. Bhutani G (2015) Modeling design and development of frameless stereotaxy in robot assisted
neurosurgery. HBNI, Mumbai
Comparative Stiffness and Damping
Analysis for Various Flow Controlling
Devices of Hole Entry Worn Hybrid
Conical Journal Bearing Under
the Variation of Speed
Abstract The choice of proper flow controlling device plays a significant part in
determining the performance parameters of hybrid conical journal bearings. During
the operation, they are subjected to variation from low to high speed to meet the need
of today’s industrial demand. The objective of this paper is to study the comparative
dynamic performance analysis of capillary, CFV and orifice compensated, hole entry,
worn hybrid conical journal bearing under varying speed by analytical method. The
finite element analysis technique is used to solve the modified Reynolds equation.
The numerically simulated results show the appreciable change in the performance
characteristics due to change in the compensating device with variation in speed with
worn/unworn condition of bearing.
1 Introduction
Hybrid bearing finds wide spread application in modern machineries. The perfor-
mance of these hybrid bearing system is greatly depends upon the type of flow
controlling device used in the system [1, 2]. In the beginning, Raimondi and Boyd
[3] as well as Shaw and Mack’s [4] have studied the analytical and experimental
performance characteristics of hydrostatic bearing compensated with capillary and
orifice restrictor. The characteristics of membrane-type variable restrictor compen-
sated externally pressurized bearings were studied by Cusano [5] He reported that
the maximum pressure generated and stiffness coefficients are strongly dependent
on the membrane compliance, the pressure ratio, and the eccentricity ratio. He with
Conry [6] also analyzed pocketed externally pressurized bearings from minimum
total power loss, point of view. Donoghue [7] solved the problems of earlier system
by using parallel feed system. Pande and Somasundaram [8] reported that the vari-
able restrictors with position sensing arrangement are superior than conventional
restrictors.
2 Mathematical Formulations
4
p= Nj P j (2)
j=1
Comparative Stiffness and Damping Analysis … 1457
The fluid film thickness h in non-dimensional form for a conical journal bearing is
given as
h = 1 − X j cos α − Z j sin α cos γ (3)
Q R = Q sp (4c)
3 Solution Procedure
A mathematical model is established to find the result for the characteristics of hole
entry worn hybrid conical journal bearing and FORTRON 77 coding language is
used. Based on analysis, due to unavailability of literature, the obtained results are
compared with the outcomes of Stout of Rowe [13]. The comparative performance
parameters have been shown for the double rows of 12 holes symmetrically arranged
on worn hybrid conical journal bearing in terms of stiffness and damping coefficients.
1458 V. M. Phalle and S. R. Pawar
The computed results are obtained using the following operating and geometrical
parameters of non-recessed hole entry worn hybrid conical journal bearing.
λ = 1, θ = 30°, IR = 0, 1 and 2ab = 0.25, β* = 0.5, U = 10o (Fig. 1).
Figure 1 represents the performance of various restrictors in worn and unworn condi-
tion under the variation of speed in the range of 0–2. Very small or negligible amount
of change is observed in the stiffness coefficient S 11 for CFV and capillary restrictor,
when they are subjected to variation in the speed parameter. However, it is reducing
when the bearing is subjected wear. For orifice restrictor, the stiffness coefficient
remains the same in unworn condition and reduces progressively when speed param-
eter varies in the range of 0–2. CFV restrictor is found to be better for acquiring the
good stiffness coefficients followed by orifice and capillary restrictor.
The graphical 2 plot describes the stiffness behavior of various flow controlling
devices in worn/unworn condition under the variation of speed. Stiffness coefficient
S 22 is found to be more for CFV restrictor than orifice and capillary in the worn
out and new of non-recessed hybrid conical journal bearing. The reduction in the
stiffness coefficient S 22 due to wear is also more in case of CFV restrictor followed
by orifice and capillary restrictor.
Figure 3 depicts the variation of damping coefficients C 11 with respect to the change
in the speed parameter ( = 0–2) for various flow controlling devices in worn
and unworn conditions of the non-recessed hybrid conical journal bearing. Very
small amount of change in the damping coefficient with change in the speed is
observed for all the three types of flow controlling devices. CFV restrictor gives better
performance as compared to the orifice restrictor followed by capillary restrictor. So,
the comparative performance C 11 for non-recessed hole entry hybrid conical journal
bearing may be noted as C 11 CFV > C 11 orifice > C 11 capillary.
Figure 4 indicates the plot of the damping coefficient C 22 verses the speed param-
eter for 100 semi-cone angle hybrid conical bearing using different flow controlling
devices in worn and unworn condition.
5 Conclusion
So, it has been observed that flow control devices, when properly selected and
tuned, can deliver very good bearing performance.
References
1. Malanoski SB, Loeb AM (1961) The effect of method of compensation on hydrostatic bearing
stiffness. Trans. ASME, J. Basic Eng. 83:179–187
2. Ling TS (1962) On the optimization of the stiffness of externally pressurized bearings. Trans.
ASME J Basic Eng 84:119–22
3. Raimondi AA, Boyd J (1954) An analysis of orifice and capillary compensated hydrostatic
bearings. In: Proceedings of ASME/ASLE Joint Conference on Lubrication, 18–19 Oct 1954,
Baltimore, Md
Comparative Stiffness and Damping Analysis … 1461
4. Mayer JE, Shaw MC (1963) Characteristics of externally pressurized bearing having variable
external flow restrictors. ASEM J Basic Eng 85:291
5. Cusano C (1974) Characteristics of externally pressurized journal bearings with membrane
type variable flow restrictor as compensating elements. Proc IMechE 188:527–536
6. Cusano C, Conry TF (1974) Design of multirecess hydrostatic journal bearing for minimum
total power loss. Trans ASME J Lubr Technol 94(1):226–232
7. O’Donoghue JP (1972) Parallel orifice and capillary control for hydrostatic journal bearings.
Tribology 5:81–82
8. Pande SS, Somasundaram S (1979) Analysis of a four-pocket hydrostatic journal bearing with
a position-sensing variable restrictor. Wear 54:331–341
9. Yoshimoto S, Anno Y, Amari K (1990) Static characteristics of hydrostatic journal bearings
with a self-controlled restrictor employing floating disk. Trans JSLE 56:3360–3367
10. Palzewicz A (1992) Hydrostatic journal bearing without pressure chambers between bearing
surfaces. Wear 159:31–38
11. Jain SC, Sinhasan R, Sharma SC (1992) Analytical study of a flexible hybrid journal bearing
system using different flow control devices. Trib Int 25:387–395
12. Rippel HC (1965) Cast bronze hydrostatic bearing design manual, 2nd edn. Cast Bronze Bearing
Institute, Cleveland, U.S.A., Apr 1965
13. Stout KJ, Rowe WB (1974) Externally pressurised bearings—design for manufacture Part-1(a).
J Bearing Sel Tribol Int 7/3 98–106. Part-3(b) Design for liquid externally pressurized bearings
for manufacture including tolerancing procedure. Tribol Int 7(3):195–212
A Comparative Study of Three Methods
for the Computation of Determinants
of Univariate Polynomial Matrices
Abstract This paper compares three different methods for computing the determi-
nant of a univariate polynomial matrix. From a systematic empirical study involving
100 trials of each numerical experiment, it appears that the method of FFT-IFT-based
interpolation and evaluation performs the best in terms of both speed and accuracy
of computation. The utility of this method is further demonstrated by a successful
application to the classical problem of forward kinematics (FKP) of the general
Stewart-Gough platform manipulator (SPM).
1 Introduction
2 Mathematical Preliminaries
The three methods used in this paper for the computation of the determinant of a
matrix with (univariate) polynomial entries are described in brief below.
A Comparative Study of Three Methods for the Computation ... 1465
Most CASs offer in-built routines for the symbolic expansion of determinants. In
this work, the routine Det available in the CAS Mathematica has been used
primarily to establish the level of complexity of the calculations involved, which, in
turn, indicates the necessity of customised algorithms, such as the following ones.
This method relies upon the fact that the required determinant of a matrix A, whose
elements are polynomials in a single variable (say, x) is also a polynomial in x.
Moreover, the degree of the determinant polynomial (say, m) can be computed with
the knowledge of the degrees of its entries alone. To compute the m + 1 coefficients
of the determinant polynomial, it is evaluated at m + 1 distinct values of x. This
results in a system of linear equations in the unknown coefficients, which is solved
subsequently.
The key appeal of this method lies in its simplicity. It is indeed trivial to choose
m +1 distinct, random values of x, and evaluate the determinant at these values. Addi-
tionally, Horner’s rule can be used to speed up such computations [3]. However, a
practical difficulty encountered frequently in the implementation of this method is
that the Vandermonde matrix appearing as the coefficients of the unknown vector in
the linear system described above is often numerically ill-conditioned. A solution to
this problem has been presented in [5], in which the concepts of fast Fourier transfor-
mation (FFT) and inverse Fourier transformation (IFT) have been used to compute
the unknown coefficients (details may be found in [5], which are not reproduced here
due to lack of space).
1466 V. Safar et al.
3 Numerical Experiments
Each method
underwent
100 trials for every value of n. However, for Method 1,
7 being of O 106 , experiments for larger values of n are conducted only for
Methods 2 and 3. The results are discussed in greater details below.
Figure 1 depicts the erosion of accuracy with increasing n, for small values of n. As
can be seen clearly, Method 1 degenerates unacceptably, and hence it is not studied
further for larger matrices.
The remaining two methods show comparable trends (see Fig. 2) in accuracy
till n = 15, after which errors in Method 2 grow at a much faster rate, while those in
Method 3 remain relatively unaffected by the increase in n. These observations may
be explained by the fact that the number of arithmetic operations increases as O(n 4 )
in Method 2, leading to greater accrual of truncation errors.
A Comparative Study of Three Methods for the Computation ... 1467
Methods 2 and 3 are implemented in C++ and compiled using the -O2 optimiser of
GCC compiler(Version 7.4.0) on a PC with an Intel Core i7-4930 CPU running at
3.40 GHz. The execution times shown in Fig. 3 are the CPU times consumed for the
computation of det( An ) only, and not the previous or subsequent computations.
As expected from the above discussions, up till n ≈ 15, the time taken by the
two methods are comparable, but afterwards, Method 2 becomes much slower in
1468 V. Safar et al.
comparison with Method 3. In particular, for n = 28, the time taken by Method 2 is
almost six times as that of Method 3.
It appears from the above empirical study that Method 3 is likely to be the best
method for applications in kinematics, e.g., in solving the FKP of the general SPM
following [6] where n = 28. The details are skipped for the want of space, but it may
be noted that when the problem is solved for the data given in [6], the determinant
expansion takes on an average of 1.45 ms, with a standard deviation of 40.0 µs. The
mean and standard deviation of n are 2.35 ×10−4 and 1.26 ×10−6 , respectively.
For the same data, method 2 takes an average of 15.4 ms (over 100-trials), with a
standard deviation of 39.6 µs. The mean and standard deviation of n are 6.89 ×10−1
and 5.21 ×10−3 , respectively.
5 Conclusions
References
Abstract Identification of the safe working zone (SWZ) of a parallel manipulator can
be an important part of its path-planning process or design. Computationally efficient
and accurate methods are required to make these demanding calculations practically
feasible. This paper looks at the Cartesian and polar variants of a 2-D numerical scan-
ner which are applicable for 3-degrees-of-freedom (DoF) planar or spatial parallel
manipulators and presents a detailed analysis of their relative speeds, advantages
and disadvantages, based on a case study involving the spatial 3-RRS manipulator.
Results of such a study provide the analysts with a deeper understanding of the func-
tioning of the scanner, which in turn, helps the computation of the SWZ of other
manipulators in a fast and reliable manner.
1 Introduction
Safe working zone of a parallel manipulator implies a subset of its workspace in which
the manipulator is free to move without encountering either the gain- or the loss-type
singularities, interferences among the links, or physical limits on the motions of the
passive joints [1, 6]. Identifying such regions inside the workspace of a manipulator
renders the task of path-planning (inside the SWZ) rather trivial.1 Also, the concept
provides a mathematical basis for the design of parallel manipulator for certain a
user-specified SWZ [2, 7].
Computation of the SWZ needs to be performed only once for an existing manipu-
lator with a fixed geometry. However, it is a non-trivial task, as the various boundary
functions (denoted by Si ), which demarcate the regions free of each of the above-
mentioned issues, are typically not available in analytical forms—these can only be
evaluated numerically at the points of interest. Therefore, as explained in [1, 2, 6],
one needs to scan (a subset of) the ambient space of the workspace of the manipulator
to identify the regions free of each of the issues, and eventually leading to the identi-
fication of the SWZ. Such a scanning involves the overlaying of a grid on the region
of interest and identifying the blocks of the grid which sees the boundary surfaces
(defined by Si = 0), pass through them (an odd number of times for a given side of the
block, to be precise). Several variations are possible, based on the chosen geometry
of the grid, and the logic behind the expansion and contraction of their regions of
coverage. These details may appear to be trivial on the prima facie; however, indeed,
these determine the feasibility of the scanning scheme as a component of a design
process, such as the ones described in [2, 7]. The computational requirements of
such a design can be very high, as the SWZ needs to be computed typically for tens
of thousands of candidate designs, so as to identify the optimal ones.
In this context, this paper presents a comparative study between two scanning
schemes for the plane: the Cartesian, used in [1], and the polar, used in [5]. The
example of the spatial 3-RRS manipulator in [5] is re-worked in both the methods,
and the results are compared for computational efficiency. Detailed analysis of the
relative advantages and disadvantages is presented in a quantitative manner. Such
a study is hoped to be highly useful for researchers in the field while choosing
the appropriate scanner for their work, as per the architecture of the manipulator is
concerned.
The rest of the paper is organised as follows: Sect. 2 presents a brief description of
the numerical scanning techniques studied in this paper. The case study on a spatial
3-RRS manipulator is described in Sect. 3. Finally, the paper is concluded in Sect. 4.
The scanners mentioned above operate on a simple strategy: they identify the zero-
level sets2 of the boundary functions, Si , one at a time (see [1] for the details; in
particular, see Fig. 4, where the progressive refinement of the scanning grid is shown).
The discrete nature of the grid allows the identification of only a finite number of
points of these boundary curves (in the case of 2-D scanners, which is the case
in point here). The point closest to the origin3 among the said points is identified
next, which lies on the largest circle enclosing the points inside the boundary curve.
Such circles are stacked up at a desired resolution along the axis orthogonal to the
2-D slices considered, and together, they define the 3-D region is free of the issue
modelled by the associated boundary function. The process is then repeated for the
next such issue, inside this region, and so on.
2 The zero-level set of a function in a plane refers to a region which is a subset of the workspace of
the manipulator.
3 Following the reason presented in [1], the SWZ is considered to be centred at the origin in this
paper.
A Comparative Study of Different Numerical Scanning … 1473
In the following, two different scanners are described and compared, namely, the
Cartesian and the polar, which follow the same general ideas as described above, but
differ in the geometry of the grids used.
In this case, the chosen search region is divided into a number of rectangular (typically
square) blocks, as per a chosen initial resolution. If a boundary curve crosses (for
an odd number of times) any side of a rectangle (or a grid line, equivalently), then
the signs of the corresponding boundary function, when evaluated at the vertices
delimiting the said grid line would differ. This indicates that the block is of interest,
and hence it is subdivided, and the process is repeated in a recursive manner. This is
done for all the blocks in the grid, till sizes of the smallest blocks in the grid reduce
to a desired limit, at which point the geometric centre of the each of the smallest
blocks is taken to be a point on the boundary curve.
The grid, in this case, consists of concentric circles and radial lines, enclosing blocks
which are of the form of pieces of an annulus. The peripheral dimension of the blocks
would increase linearly as the distance of the blocks from the origin increase; hence,
the number of divisions in each of the annuli is increased progressively with the
distance from the origin, so as to keep the peripheral dimensions nearly constant in
all the blocks, depicted in Fig. 1. The computational strategy is identical with the
Cartesian scanner in all other aspects.
The two scanners described above are employed in this section to compute the SWZ of
a spatial 3-RRS manipulator shown in Fig. 2, having the same geometry as described
in [3]. The three legs have RRS architecture with identical dimensions. The fixed
and moving platforms are in the form of equilateral triangles having a circumradius
of b and a, respectively.
1474 B. Patra et al.
The manipulator has 3-DoF, which can be considered to manifest as the roll (denoted
by α), pitch (denoted by β), and heave (denoted by z) of the moving platform. The
region of interest in the heave direction is z ∈ [0.1 m, 1.475 m]. A total of 2751
instances of the α-β planes are considered, orthogonal to the Z -axis, and spaced
equally.
The ranges of scanning in the 2-D slices mentioned above differ as per the geom-
etry of the scanner. In the Cartesian scanner, only half of the desired rectangles in
the α-β plane are scanned, since there is a symmetry in all the boundary functions
about β = 0◦ ; hence, scanning for the region {α, β} ∈ [−180◦ , 0◦ ] × [−180◦ , 180◦ ]
is sufficient. The initial resolution in α, β for the scanner is set at 0.5◦ , and after two
stages of refinement, the final resolution becomes 0.125◦ . On the other hand, there
A Comparative Study of Different Numerical Scanning … 1475
The boundary functions are the same as those described in [5]. In brief, S1 = 0
demarcates the boundaries of the workspace; S2 = 0 defines the set of points where
the manipulator suffers from gain-type singularities; S3 ≥ 0 identifies the regions
in which at least two of the links interfere. The limitations on the motions of the
passive joints are ignored for the sake of consistency in comparison, as these were
not considered in [5], either. However, the definition of the function S3 is different
in this case, as both the proximal and distal links in each limb are considered to
be enclosed in bounding boxes of cylindrical shape (for the ease of detection of
interference between these) in this paper, as opposed to the cuboidal bounding boxes
used for the distal links in [5]. This approximation leads to much faster computations
(following the approach presented in [4]) of interferences between the links, albeit
being a bit more conservative.
4 The manipulator has a three-way symmetry because of the fixed and moving platforms are equi-
lateral triangles and the legs are identical in their architecture.
1476 B. Patra et al.
The analysis of the computational efficiency of the two scanners needs to be done
carefully, for two reasons: (a) the extent of the search domains are different in the
two scanners for the differences in their inherent symmetries (see Sect. 3.1 for the
details) and (b) the total time taken in each scanner is a sum of the time taken for
the generation of the initial grid, the refinements thereof and the consequent calls
to the different boundary functions. In terms of the total time, the Cartesian scanner
takes 214.22 s, while the polar variant consumes 309.86 s, both the numbers being
averaged over 50 runs each (see Fig. 4d). This is in sharp contrast with Fig. 4a, where
it can be seen that the polar scanner makes a much lesser number of function calls
than the Cartesian for the first two boundary functions, and comparable number for
the third. Further, Fig. 4b shows that the times per the function call are comparable
between the two scanners in all three functions. Therefore, as seen in Figs. 4c, d,
the total time taken in the calls to the respective functions Si consumed comparable
times in the two scanners. This discrepancy is resolved in Fig. 4d, where it can be
seen that the polar scanner takes a lot more time in the generation and refinement
A Comparative Study of Different Numerical Scanning … 1477
of the grid; hence, the Cartesian scanner emerges as more economical in the overall
sense. This observation is significant, since from the relative sizes of the scan regions
alone, one may be lead to believe otherwise.
4 Conclusions
This paper has presented a comparison of the Cartesian and polar scanners for
the purpose of computing the SWZ of spatial parallel manipulators. The spatial
3-RRS manipulator is used as an example for a detailed case study. Various aspects
of the scanners, e.g., certain symmetries in their grid structures and the manipulator’s
architecture, geometric features, such as uniform versus non-uniform grid size, are
brought out clearly. The comparisons show that for the resolutions and scan dimen-
sions used in the paper, the Cartesian scanner requires more number of function calls
in the initial stages of the scan, i.e. for the function S1 . However, since the run-time
consumed by this function is very small, the overall time taken for the computation of
1478 B. Patra et al.
all the boundary functions is comparable between the two scanners. The polar scan-
ner, on the other hand, takes a longer time for the generation and the refinement of
its grid. On the whole, the Cartesian scanner seems to be a better choice in this case,
which is somewhat non-intuitive, as the Cartesian scanner utilises only a two-way
symmetry as opposed to the three-way symmetry of the polar scanner.
References
1. Karnam MK, Baskar A, Srivatsan RA, Bandyopadhyay S (2020) Computation of the safe work-
ing zones of planar and spatial parallel manipulators. Robotica 38(5):861–885. https://doi.org/
10.1017/S0263574719001139
2. Kilaru J, Karnam MK, Agarwal S, Bandyopadhyay S (2015) Optimal design of parallel manipu-
lators based on their dynamic performance. In: Proceedings of the 14th IFToMM world congress.
Taiwan, pp. 406–412. https://doi.org/10.6567/IFToMM.14TH.WC.OS13.129
3. Muralidharan V, Bose A, Chatra K, Bandyopadhyay S (2020) Methods for dimensional design
of parallel manipulators for optimal dynamic performance over a given safe working zone. Mech
Mach Theory 147. https://doi.org/10.1016/j.mechmachtheory.2019.103721
4. Nag A, Bandyopadhyay S (2019) Analytical determination of a sphere inside which the stewart
platform translates without suffering any leg interference. In: Lenarcic J, Parenti-Castelli V (eds)
Advances in robot kinematics 2018. Springer International Publishing, Cham, pp 74–82.https://
doi.org/10.1007/978-3-319-93188-3_9
5. Patel D, Kalla R, Tetik H, Kiper G, Bandyopadhyay S (2017) Computing the safe working zone
of a 3-RRS parallel manipulator. In: Wenger P, Flores P (eds) New trends in mechanism and
machine science, vol 43. Springer International Publishing, Cham, pp 113–120. https://doi.org/
10.1007/978-3-319-44156-6_12
6. Srivatsan RA, Bandyopadhyay S (2014) Determination of the safe working zone of a parallel
manipulator. In: Thomas F, Perez Gracia A (eds) Computational kinematics, vol 15. Springer
Netherlands, Dordrecht, pp 201–208. https://doi.org/10.1007/978-94-007-7214-4_23
7. Tamilmani E, Bandyopadhyay S (2019) Computation of singularity free region and design opti-
misation of a four-degrees-of-freedom hybrid arm. In: Proceedings of the advances in robotics
2019. AIR 2019, Association for computing machinery, New York. https://doi.org/10.1145/
3352593.3352673
Motion Control of a Phalange Using
Tendon-Based Actuation System: A Bond
Graph Approach
Abstract Daily interactions with the environment by the hands such as grasping,
manipulating, writing, pull/push, carry, and playing music require specific forces
and some precise movements provided by the various elements of the human hand.
Skeletal muscles play a significant role in locomotion of upper as well as lower limbs.
The muscle tendon unit involves nonlinear characteristics. A biomechanical model
has been proposed to control the motion of a phalange, actuated through the muscle
tendon unit by a virtual phalange, a part of the control by the central nervous system.
The authors propose a bond graph based mathematical model to control the motion
of a phalange through a virtual system using tendon actuation. Simulation results
validate the proposed bond graph model.
1 Introduction
We perform most of our daily interactions with the environment by the hands such as
grasping, manipulating, writing, pull/push, carry, playing music, and expressing. All
these interactions require specific forces and some precise movements provided by
the various elements of the human hand. Biomechanics of the human hand comprises
the rigid bones termed as phalanges and muscle tendon structure. It is important to
understand the musculoskeletal actuation system of the human hand to analyze its
motion. It is difficult to understand the behavior of the muscle tendon unit (MTU)
in-vivo as well as in-vitro. Most of the studies have been carried out on specific static
posture aspects without considering the dynamics of the system. This work presents
a bond graph [1, 2] -based mathematical model to control the motion of a phalange
through muscle tendon actuation by taking inputs from a virtual system considered as
a part of the control by the central nervous system (CNS). The CNS itself determines
the required effort to be applied by the MTU at the respective insertion point of the
tendon on the real phalange.
2 Proposed Model
A system analogous to the real biomechanical system [3] has been proposed here for
the development of the model. In real-life biomechanical system, every locomotion
action which is to be performed by the body whether we talk of upper limbs or lower
limbs movements requires some initiation at the level of brain. It is the brain which
decides upon what action is needed for a specific motion. The brain then stimulates
the CNS, which further send the specific controlled signals through the network
of nerves to the corresponding skeletal muscles for the required actuation at their
end. Then after receiving the controlled signal from the CNS, the skeletal muscles
generate specific efforts to bring the body into motion.
In order to describe the concept developed, we have considered one of the
phalanges of the hand which is to bring in motion through muscle tendon actu-
ation for the desired motion. The phalange has been considered as a cylindrical
body for simplicity. In order to control the motion of this phalange considered as real
phalange, another phalange with same specifications has been considered as a virtual
phalange. There are no MTU in the virtual system but the insertion locations of the
tendons on the virtual phalange have been considered and taken at the same locations
as that in the real phalange. Figure 1 shows the system considered. The frame {1}
is fixed at the bottom center of phalange, and the frame {0} also initially coincides
with the frame {1}.
In order to actuate the real phalange, four muscle tendons have been considered
on the real phalange at symmetrical locations as shown in figure. These MTUs need
to be actuated to provide the desired motion to the real phalange after getting the
input stimuli from the virtual system.
Corresponding to each MTU, there is a hook point in the system which is fixed
with respect to the inertial frame, through which the muscle tendon freely glides like
a pulley system. Tendons are attached to the insertion locations in the real phalange
and are made to pass through these hook locations to provide specific direction of
effort application and moment generation.
Major part of the body is composed of skeletal muscles which is the prime mover
for locomotion [4]. Skeletal muscles are viscoelastic in nature having a nonlinear
behavior. These muscles possess very high strength to hold high tensile loads. A
well-developed and extensively used concept based on Hill’s model [5] has been
Motion Control of a Phalange Using Tendon-Based Actuation … 1481
Hook
Hook
Location
Location X0, X1
X0, X1
Z0, Z1
Z0, Z1
Front View Of
Front View Of Real Phalange
Virtual Phalange
Tendons
X0, X1 X0, X1
Z0, Z1 Z0, Z1
Top View Of Top View Of
Virtual Phalange Real Phalange
utilized to model the biomechanics of the skeletal muscles, referred here as MTU,
i.e., muscle tendon unit, is shown in Fig. 2.
Bond graph model of Hill’s model is shown in Fig. 2. Here T CE is the contractile
element which develops the effort during active movement. There is no contribu-
tion from this element while in passive motion or in resting position. K PE is the
parallel elastic member which resist passive stretching in the muscle and has a
strain-dependent nonlinear behavior as represented by Eq. (1) below.
Fig. 3 Force versus strain plot for nonlinear element (K PE ) of Hill’s model
as desired. Multibond graph approach has been used to model the biomechanical
system.
A detailed multibond graph model [6] based on rigid body dynamics has been
demonstrated for real and virtual phalange system as shown in Fig. 4. The figure
below shows the real phalange connected with single MTU, and similarly, the real
phalange is connected to other MTUs. The same muscle tendon structure has been
used for the other MTUs. Figure 4 clearly demonstrates the virtual and real phalange
rigid body mechanics and the controller part to control
the motion
of real phalange
through muscle tendon actuation. MTF modulus C0 1 r O11v × is the skew symmetric
T
matrix obtained from position vector C0 1 r O11v = C0 1 x O11v C0 1 y O11v C1
0
z O11v . Simi-
larly, other MTF moduli have been evaluated. MTF modulus 0p r̄ˆH is the unit vector
describing the direction along which the effort is applied. The desired flow is provided
to the virtual phalange through bond 91v. Flow information from the virtual and real
phalange is taken from bond 107v and 107, respectively. The controller compares
these flows and generates an effort at bond 12 through the C and R elements (PD
Control) connected to it at bond 10 and 11, respectively. This controlled effort signal
is then passed on to the MTU through the small mass δm connected at free end of the
muscle tendon. MTU then develops an effort based upon the nonlinear behavior of
the muscle tendon, to be applied at the insertion location on the real phalange through
bond 107. By this effort, the real phalange moves and tracks the virtual phalange as
desired.
The biomechanical behavior [4] of the phalange motion using tendon-based actua-
tion, with the help of proposed model, has been simulated. Simulation code has been
written from the bond graph model itself using MATLAB. The mass of the virtual
and real phalange is taken as 0.0055 kg. The phalanges have been considered as
cylindrical bodies having radius of 0.005 m. The behavior of the model is studied
under two conditions using MATLAB coding.
Case I: The virtual phalange is dragged in −ve X-direction in inertial frame in
a trajectory given by polynomial equation 2. The motion in rotational domain is
permitted only about Z-axis, that has been considered as flexural axis of phalange
and the rotation in other two axes is constrained. Corresponding to movement of
virtual phalange, the real phalange tracks the trajectory of virtual phalange through
muscle tendon actuation. Figure 5 shows the position trajectory of the center of mass
of the virtual and real phalange over the time period w.r.t inertial frame.
1484 S. K. Uppal and A. Vaz
Fig. 4 Multibond graph model of virtual and real phalange with controller and MTU
Motion Control of a Phalange Using Tendon-Based Actuation … 1485
10 -9
0.02
0
Virtual x-cord. Real x-cord.
Virtual y-cord. Real y-cord. SE1
0.01 Virtual z-cord. Real z-cord. SE2
-5
SE3
SE4
0
-10
-0.01 -15
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
-4 -4
10 10
1 1.5
Tendon 1 Tendon 3
0.8 Tendon 2 Tendon 4
0.6 1
0.4
0.5
0.2
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Case II: Initially, the virtual phalange is dragged in −ve X-direction in inertial
frame in flexure motion by using polynomial trajectory mentioned in equation 2
above and then is rotated in inertial X–Z plane about the initial axis of symmetry
of phalange with a radius of 10 mm. The real phalange follows the trajectory of the
virtual phalange. Figure 6 shows the position trajectory of the center of mass of the
virtual and real phalange over the time period w.r.t inertial frame, and Fig. 7 shows
the forces applied to the real phalange by the muscle tendon unit.
-7
10
0.02 2
Virtual x-cord. Real x-cord.
Virtual y-cord. Real y-cord. 1
0.01 Virtual z-cord. Real z-cord.
0
-1 SE1
0 SE2
-2 SE3
SE4
-0.01 -3
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
-3 -3
10 10
1 4
0.8
3
0.6
2
0.4
1
0.2
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
0.03 0.03
0.02 0.02
0.01 0.01
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
4 Conclusion
References
1. Karnopp DC, Margolis D, Rosenberg RC (2000) System dynamics: modeling and simulation of
mechatronic systems. Wiley-Interscience, New York
2. Mukherjee A, Karmakar R (2000) Modeling and simulation of engineering systems through
Bondgraphs. Narosa Publishing House, New Delhi
3. Valero-Cuevas FJ, Elise Johanson M, Towles JD (2003) Towards a realistic biomechanical model
of the thumb: the choice of kinematic description may be more critical than the solution method
or the variability/uncertainty of musculoskeletal parameters. J Biomech 36:1019–30
4. Fung YC (1993) Biomechanics, mechanical properties of living tissues. Springer
5. Vaz A, Singh K, Dauphin-Tanguy G (2015) Bond graph model of extensor mechanism of finger
based on Hook—string mechanism. MAMT 91:187–208
6. Mishra N, Vaz A (2017) Bond graph modeling of a 3-joint string-tube actuated finger prosthesis.
Mech Mach Theory 117:1–20
Taguchi Optimization for Wear
Behaviour of Drum Brake Shoe Interface
Abstract The effect of parameters on wear behaviour of drum brake shoe is opti-
mizing using Taguchi method. Pin samples made from brake shoe material is tested
on pin-on-disc test rig against disc of same material as of drum. The effects of vari-
ation in load (40, 60 and 80 N), sliding speed (300, 600, 900, 1200 rpm) and track
diameter (40, 60, 80 mm) on wear rate of shoe surface were examined. From signal-
to-noise ratio (S/N) and analysis of variance (ANOVA) of experimental results, the
major parameter responsible for wear of drum brake shoe is found and different
conditions are suggested for improvement in wear life and reliability of brake.
1 Introduction
The most substantial safety aspect of a vehicle is its braking system, which retard
the motion of vehicle quickly and reliably under varying operating conditions.
Brake shoe of an automotive drum brake system indicates one of the most complex
composite materials, since they have numerous constituents that are diverse in
different properties. The effect of various operating parameters causes wear of shoe
material. Different researchers studied the Taguchi optimization for the influence of
operating parameters on wear behaviour of materials. Deuis et al. studied the impact
of load, sliding conditions like distance, speed, time and volume percentage on wear
of different composites under dry sliding condition [1]. Xu et al. examined that the
sliding motion had major influence by contact of parts and behaviour of material
under different tribological condition. The formulation of suitable composite for
improving wear-resistant properties, evaluation of relation between various param-
eter and wear rate have been necessary [2]. Montgomery suggested that the Taguchi
method has been used for optimizing designs and to achieve different manufacturing
parameters among the various techniques of experimental design [3]. This technique
was specially designed to minimize the number of experiments and to investigate the
2 Experimental Design
2.1 Materials
The pin samples used for wear test, which was made up of commercial automotive
brake liner material. The friction materials of brake liner were attached to mild
steel pin with the help of strong epoxy adhesive. The pin samples were machined
hemisphere shape at end as per ASTM G-99 standards, such that each pin was of
30 mm in length and 12 mm circular in cross section. The cast iron disc was casted
with expendable mould in dry sand casting of dimension of 162 mm in diameter
and 8 mm in width as that of same material of drum of automotive vehicles. In wear
test, the pin samples rotating against cast iron disc to examine the wear behaviour of
brake samples at different operating conditions.
The drum brake performance was strongly influenced by applied load, sliding speed
and interface radius, and the performance was affected by these design parameters.
The optimum value of each parameter is necessary for increasing the wear behaviour
and performance of braking system, and the below parameters are considered for
the experiments, as listed in Table 1. The main parameters used are load, the sliding
speed and the track radius. Parameters and their levels are shown in Table 1.
Taguchi Optimization for Wear Behaviour of Drum Brake … 1489
The wear experiments were carried out on the pin-on-disc tribometer for studying
the dry sliding behaviour of brake friction material as shown in Fig. 1. The wear tests
were conducted by varying the load from 40 to 80 N, speed from 300 to 1500 rpm
and track radius from 20 to 60 mm, while keeping the sliding distance constant at
1000 m. Wear volume of pin samples was calculated by using the formula,
πh 3d 2
Wear Volume mm3 = + h2 (1)
6 4
2 1/2
where h = r − r 2 − d4
d is the wear scare diameter (mm) and r is the pin end radius (mm).
There are different orthogonal arrays available for the design of experiments; each of
the arrays is indicating the definite number of independent variables which has to be
design and their levels. In this study, experimental test is designed for the three and
for five levels, then total number of experiment is 243. The total number of results
provide by the system which is about 170 exact results. In association with the above
method, Taguchi method provides the list 25 experimental combinations based on
the orthogonal array in a specific order which cover all factors. Those twenty-five
experimental test results will give 99.96% accurate result. By using Taguchi method
of optimization, number of experiments reduced to 25 instead of 243 with almost
similar accuracy. Statistical analysis of data collected from experimental data was
done using Minitab V.17.1.0. Analysis of test results is done in five stages. The rela-
tion between response variable in the form of mathematical formulation is developed
in terms of wear volume as a control variables using regression analysis in the first
stage. In second stage, ANOVA is done for investigating the significant parameters
affecting the wear volume of brake sample. In the last stage, optimal values for
control variables for best response factor are recognized using Taguchi technique. It
is applied to the 25 experimental test results conducted as per L25 orthogonal array.
Table 2 endows orthogonal array along with experimental test results and S/N values
for the response variable. Equation (1) predicts this relationship with an adjusted R2
of 98.21%. The R2 value is a measure of goodness of fit between observed data and
modelled data. According to the model developed, 98.21% variability in the depen-
dent variables is predictable from the independent variable. Hence, it ensures that
predicted model gives fairly good clarification of correlation between input variables
and the response.
Table 3 gives the results of ANOVA test for wear volume of brake sample at 95%
confidence level. The results illustrate that load and sliding speed are significant
factors as their p-value is less than 0.05, whereas interface radius is not signifi-
cantly influencing response variable. Further, it is observed that the applied load has
maximum influence (35.40%) on wear volume. The other factors like sliding speed
(33.88%) and interface radius (29.67%) have less influence on response variable
(Figs. 2 and 3).
Taguchi Optimization for Wear Behaviour of Drum Brake … 1491
15.0
Mean of SN ratios
14.5
14.0
13.5
13.0
12.5
Fig. 2 Effect plots for SN ratios on wear behaviour of brake friction materials
0.24
0.23
Mean of Means
0.22
0.21
0.20
0.19
0.18
0.17
40 60 80 300 600 900 1200 1500 20 30 40 50 60
Fig. 3 Effect plots for means on wear behaviour of brake friction materials
4 Conclusions
References
1 Introduction
2 Formulations
2.1 Assumptions
The bearing is assumed to be infinitely long and of circular cross section as shown in
Fig. 1. The radius of the shaft is R, the center to center distance between the bearing
and the shaft is e, with the maximum being the clearance C, while the attitude angle
is φ. W is the weight of the shaft. The x, y coordinate system is as described in
the figure. The film thickness h is a function of x. The shaft rotates with a speed
of ω and precesses with a speed of ωp . The external magnetic field is radial and is
given by the relation B = −B0 Rr −1 , where r is the radial distance of a point within
the bearing from the origin. The viscosity coefficients of the lubricant are μ and χ
electrical conductivity is σ, pressure of the fluid film is p, and ez is the electric field
in z direction.
2.2 Non-dimensionalization
x y h u pC 2
θ= ,η = , H = ,U = ,P = ,
R h C ωR μω R 2
σ σ C μ 1 γ
M= B0 C, E = ez , N = ,L =
μ μ ωR 2μ + χ C 4μ
Analysis of a Hydrodynamic Journal Bearing of Circular … 1497
1
Q= U H dη = M −2 (Dθ P − M E){2FB (H ) − M H } + FB (H )(1 − 2Dτ φ)
0
(4)
Dθ Q + Dτ H
= Dθ M −2 (Dθ P − M E){2FB (H ) − M H } + FB (H )(1 − 2Dτ φ) + Dτ H = 0
(5)
θc θc
E H (θ )dθ = M Q dθ θc : cavitation boundary (6)
0 0
We get the perturbed flow rates and the corresponding equations from (3) and (4)
1
Q 2Im0 + Q 2Im sin θ = (Dθ P2Im − M E 2Im )g0
M3
2 λR
− f 0 , Dθ [Q 2Im0 + Q 2Im sin θ ] + λ R sin θ = 0 (11)
M ε0
M f 0 (θc ) − E 0 g0 (θc )
Q0 = ,
M2
P0 (θ ) = M 3 Q 0 F01 (θ ) − M 2 F02 (θ ) + M E 0 θ (13)
Table 2 Variables and functions related to flow equations and pressure profiles
θ
= (θc + ε0 sin θc )−1 , f i = FB (Hi ), gi = 2FB (Hi ) − M Hi , F01 (θ) = g0−1 dθ, F02 (θ) =
0
θ
f 0 g0−1 dθ
0
θ θ
F1Re1 (θ) = − (Dθ P0 − M E 0 )g0−1 g0 cos θdθ F1Re2 (θ) = 0.5M 2 g0−1 f 0 cos θdθ
0 0
θ θ
F1Im2 (θ) = 0 M 2 Mg0−1 sin θ + (1 − cos θc ) dθ F1Im1 (θ) = 0 M 2 Mg0−1 + θc dθ
θ θ
F2Re1 (θ) = − 0 (Dθ P0 − M E 0 )g0−1 g0 sin θdθ F2Re2 (θ) = 0 0.5M 2 go−1 f 0 sin θdθ
θ
F (θ) = F2Im1 (θ) = M 2 Mg0−1 + θc dθ
2Im2
θ 3 g −1 cos θ + sin θ − 2 f ε −1 g −1 dθ
0
0 M 0 c 0 0 0
1500 D. Tripathy and K. Bhattacharyya
Fig. 2 Effect of variation of coupling number on pressure P0 , stiffness (S ij ) and damping (Dij )
Analysis of a Hydrodynamic Journal Bearing of Circular … 1501
Fig. 3 Effect of variation of length parameter on pressure P0 , stiffness (S ij ) and damping (Dij )
and causes a significant drop in peak pressure. Additionally, it may be noted that
with increase in N, S Rφ and both DφR and Dφφ which are already negative become
drastically more so after N = 0.7. This indicates that while increasing N beyond 0.7
may apparently lead to a significant increase of load bearing capacity, the bearing
may become unstable.
Same variations with L (for M = 8, ε = 0.8, λ R = 1, N = 0.8) are presented in
Fig. 2. Here, the critical value of L is 0.1. S φR , DφR and Dφφ which are already negative
decrease sharply till this value. Pressure profile also remains largely unaltered beyond
this value indicating that increasing L beyond this value may not be needed to improve
load bearing capacity and stability will also not be affected if the bearing is stable
up to this point. Hence, choosing L beyond this value does not help much (Fig. 3).
The presence and development of negative stiffnesses and dampings indicate that
simply opting for higher values of N and L while apparently indicating a benefit in
terms of increased load bearing capacity may lead to instability. Hence for optimal
bearing design, a detailed stability analysis considering the effects of M, L, N is
required to obtain the optimal values of these parameters for the lubricant to get
optimal performance with assured stable operation.
References
1. Kuzma DC (1963) The magnetohydrodynamic journal bearing. J Basic Eng TRANS ASME
Ser D 85(3):424–427
2. Kuzma DC (1964) The finite magnetohydrodynamic journal bearing. J Basic Eng TRANS
ASME Ser D 86(4):445–448
3. Sasada T, Kurosaki Y, Honda K, Kamijo K (1974) MHD journal bearing in a magnetic
field perpendicular to its axis 1st report, analysis of an infinitely long bearing. Bull JSME
17(114):1645–1651
4. Kamiyama S (1969) Magnetohydrodynamic journal bearing (Report 1). Trans ASME J Lub
Tech 91:380–386
5. Malik M, Singh DV (1980) Analysis of finite magnetohydrodynamic journal bearings. Wear
64:273–280
1502 D. Tripathy and K. Bhattacharyya
6. Kulkarni PA, Rao BVA (1977) Stability behaviour of finite MHD journal bearings. Mech Mach
Theory 12:293–302
7. Prakash J, Sinha P (1975) Lubrication theory for micropolar fluids and its application to a
journal bearing, ht. J Eng Sci 13:217–232
8. Huang TW, Weng CI (1990) Dynamic characteristics of finite-width journal bearings
withmicropolar fluids. Wear 141:23–33
9. Das S, Guha SK, Chattopadhyay AK (2005) Linear stability analysis of hydrodynamic journal
bearings under micropolar lubrication. Tribol Int 38:500–507
10. Tanahashi T, Nakai T (1995) Fundamental nonlinear theory for micropolar electrically
conducting fluids. JSME Int J Ser B 38(2):273–279
11. Ido Y (2005) Basic equations and constitutive equations of micropolar magnetic fluids with
E-B analogy and the Abraham expression of electromagnetic momentum. JSME Int J Ser B
48(3):273–279
Twin-Plate Turbine Using Parallel
Four-Bar Mechanisms
Abstract Turbines are used for the generation of electric power using hydel, air
flow or wind. In a conventional wind turbine, the angle of attack of the wind changes
as the cups or plates rotate. However, if the angle of attack remains perpendicular
to the plate, the effort of the wind on the turbine rotation would be higher. In this
paper, a novel concept of a turbine is proposed that ensures that the plates always
remain perpendicular to the direction of wind flow. This is achieved by using parallel
four-bar mechanisms. A simplified mathematical model is presented along with that
of Persian turbine, of similar form. The analyses were compared, and the proposed
mechanism was found to generate more power than Persian turbine. Physical proto-
types of both the turbines were fabricated and tested, and the proposed turbine was
found to support the theoretical calculations.
1 Introduction
Renewable energy generation is the need of the hour. Wind or air is one such source
of energy. Over the years, wind energy is being used for domestic and small-scale
applications. The Persian turbine, one of the earliest of its kind, was used for grinding
grains and pumping water depending on the necessity. Wind turbines are broadly
classified as horizontal axis wind turbine (HAWT) and vertical axis wind turbine
(VAWT).
Persian turbine is a VAWT that consists of plates or blades that are in the radial
direction of a rotating column [1, 2], as shown in Fig. 1a. Half of the width is exposed
to the wind, and therefore, wind applies force on each plate for half of the revolution.
Its plates receive the wind at distinct angles of attacks on the plate areas, which results
in two different components of forces. The component that is tangential to the plate of
the turbine is responsible to create turning moment, and the radial component results
Fig. 1 Vertical axis wind turbines and paddling using four-bar mechanisms
in the reaction force on the vertical column. Many researchers and organizations
have worked on means of improving the efficiency in the form of different shapes
of plates or blades and plates that orient themselves during each revolution to have
larger turning moment. The former adopts airfoils instead of flat plates to produce
pressure difference which drives the rotor, e.g. Savonius, Darrieus, etc. The latter
deals with the concepts of mechanisms and has been explored in this paper.
To get higher turning moment, the plates should be aligned such that the angle of
attack should be such that the direction of wind is perpendicular to the plate surface.
Few commercial products such as aerogenerator exawind [3] exist which have means
of making the plates orient using some mechanisms, as shown in Fig. 1b. Thang [4]
has proposed wind turbine with flipping airfoils, where the airfoils/blades orient
themselves based on the rotation angle. However, there is a grove in which the airfoil
links move, thus making the system susceptible for frictional losses. It is shown in
Fig. 1c.
The authors conceptualized a mechanism that used parallel four-bar mechanism,
and in their literature survey, they found a similar mechanism used in paddling of
steamers or boats [5], shown in Fig. 1d. When the plate is at the bottom region, within
water, it should remain vertical for maximum thrust. The plates while entering water
or exiting should be aligned such that there is minimal splash and hence saving of
energy. The paddle wheel though uses four-bar mechanisms to orient the plates, they
do not ensure that the plate remain parallel to one of the directions, which is what
the authors have proposed in their novel design named ‘Twin-plate Turbine’, which
is discussed next. The novel design uses several parallel four-bar mechanisms whose
coupler links act as plates or blades of the turbine. The parallel four-bar ensures that
the coupler’s orientation always remain parallel to the fixed link, as shown in Fig. 2.
Twin-Plate Turbine Using Parallel Four-Bar Mechanisms 1505
B A
B
A A
C B
D
D C
Fig. 2 Coupler (BC) link in parallel four-bar mechanism remains parallel to fixed link (AD)
2 Twin-Plate Turbine
The proposed turbine consists of two circular discs, vertical and parallel to each other.
The discs are constrained with a support link such that the axes of the circular discs
have an offset in the vertical direction and have revolute joint with it. Side view of
the CAD model is shown in Fig. 3a. Plates with Z shape are used to connect between
the two discs, each end of the plate forming a revolute joint with the corresponding
circular disc. The distance between the revolute joint axes of the plate is kept same
as the vertical offset between the centres of the circular discs. If the support link is
considered as fixed link, the circular discs form the crank and the follower of a four-
bar mechanism and the plate link forms the coupler. Note that the coupler link would
always remain vertical due to the fact that all these links form a parallel four-bar
mechanism. Similarly, holes on the circular discs can be created and identical plates
can be inserted between them, forming a series of parallel four-bar mechanisms, as
shown in Fig. 3b.
The lower half of the disc (i.e. relatively higher, referred to as ‘Higher disc’) is
covered on the sides and ensures that if wind is flowing, it interacts only with the
Support
Lower disc
Base Cover
(a) Side view (b) Isometric view with more plates
plates that are instantaneously above the cover. The wind pushes the plates that are
in its line of action (above the cover) and the plates then rotate the circular discs. The
axis of the ‘Higher disc’ is considered as output, and a generator can be connected to
it to extract electrical energy. The support link and the cover can in turn be mounted
on another base link through a vertical revolute joint, which would let the turbine
orient itself about the vertical direction such that wind’s direction is perpendicular
to the surface of the plates. These are shown in Fig. 3a.
Working of the turbine mainly depends on various physical parameters such as
radius of rotation, number of plates and plate area. Optimizing these physical param-
eters also plays a major role in performance. Radius of the turbine is the crank rotation
radius since all the points have the same radius of rotation. Number of plates can be
increased to generate more power and to reduce the fluctuation in the system. Area
of plates or dimensions of the plates should be designed geometrically to get the
maximum output.
The proposed setup has horizontal axis of power generation. However, if the
circular discs are considered horizontal, the output shaft would be vertical and
hence would fall under VAWT category. For easier fabrication, horizontal version is
selected.
3 Mathematical Model
The simplified mathematical model of the Persian turbine and the proposed turbine
are covered in this section. The wind direction is assumed to be parallel to the
horizontal plane. As the area of plate perpendicular to the wind direction may be
different, pressure is considered and the effective force on the plate is considered as
pressure upon projected area of the plate. The striking force acting on the plate has
been assumed to be acting at the centre of the plate, which would make the analyses
simpler. Consider the Persian turbine in Fig. 4a. The turning moment on the output
axis is due to the tangential component of the force. At an instant when the plate
makes θ with respect to the horizontal, the turning moment on the output shaft can
be determined as
b
T = Pbl r + sin2 θ (1)
2
where P is the pressure, b and l are the breadth and length of the plate, respectively,
r is the radius of the crank or link. The force applied is considered as pressure ×
area, i.e. P × (b × l). The turning moment is a function of θ and has maximum value
when θ is 90°.
For the proposed turbine, static force analysis method is used to determine the
turning moment on the output shaft, connected to the ‘Higher disc’, as shown in
Fig. 4b. The free-body diagrams of each link of a parallel four-bar mechanism are
shown in Fig. 4c, for the crank making θ with horizontal. The internal forces acting on
the links are considered such that follower link (CD) is a two-force member, coupler
(BC) is a three-force member and the crank (AB) is a two-force and one moment
member. For the coupler link, since geometric symmetry exists, the force polygon
is an isosceles triangle, and hence, the solution is obtained analytically. Hence, the
direction and the magnitude of the forces and hence the moment due to these at the
output shaft can be determined by knowing the perpendicular distance between the
forces acting on the crank link. Due to space constraint, the brief derivation is shown
in Fig. 4c. The turning moment is determined as
The static force analysis is a simpler method where inertia effects of the moving
links are not considered. The formulations derived for the Persian turbine and the
proposed turbine were implemented as a MATLAB program. The physical param-
eters considered are P = 1 MPa, l = 0.15 m, b = 0.1 m, r = 0.15 m. The turning
moment due to one plate in each of the turbine for the range of motion of crank from
0° to 180° is compared in Fig. 5a. It can be observed that the turning moment in the
proposed turbine is larger than that in Persian turbine of same size. When six plates
are considered equally distributed in an angular fashion, the effect of wind acting on
Fig. 5 Comparison of turning moment plots for Persian and twin-plate turbines
1508 B. V. Chennoju et al.
the plates is only considered for the range of θ between 0° to 90°. This is because
for the plates beyond 90°, some other plate maybe obstructing the direct impact of
the wind on the plate under consideration. The comparative plots for six plates are
shown in Fig. 5b. Similarly, Fig. 5c compares the analysis for eighteen plates. The
fluctuations observed can be reduced by using higher number of plates or with by
using a flywheel on the output shaft. It can be observed that the proposed method has
larger area under its curve, compared to Persian turbine and hence can be concluded
to be more efficient.
4 Physical Prototype
Physical prototype of the Persian and the proposed twin-plate turbines were devel-
oped using circular discs made up of plywood, plates in the form of plastic boxes and
PVC plastic pipes for connecting the plates with circular discs. The dimensions used
in the previous section were used for the fabrication, and six plates were used in the
prototype. The twin-plate turbine in action is shown in Fig. 6. Air blower was used
to generate the action of wind on the plates and for similar conditions. A dynamo
(used in a cycle) was connected to the output shaft, so that the electrical output can
be measured. The physical prototypes are shown in Fig. 7a, b.
A voltmeter was connected across the dynamo, and an ammeter (multimeter) was
connected in series. The air blower was held close to the plates and the turbines rotated
due to the force acting on the plates. The current and voltage measured for both the
prototypes are reported in Fig. 7c. The output power in the proposed ‘Twin-plate’
turbine was found to be around 2.5 times more that of Persian turbine. The results
compliment the analyses obtained by the mathematical model proposed earlier.
The physical prototypes were developed using common household items, and
the joint axes were not aligned exactly as per the design. Hence, the rotation of the
turbines caused vibration of the support. An exact model of the proposed turbine is
being developed, and the results shall be reported in the due course.
Fig. 6 Physical prototype of twin-plate turbine with rotating plates (numbered 1–6)
Twin-Plate Turbine Using Parallel Four-Bar Mechanisms 1509
Dynamo
(a) Persian turbine (b) Twin-plate turbine (c) Power generated by the dynamo
Fig. 7 Physical prototypes of turbines and the electrical output generated as results
5 Conclusion
In this paper, a novel turbine design is proposed which uses a set of parallel four-
bar mechanisms to keep the plates or blades of the turbine always perpendicular
to the direction of the wind. Static force analysis was performed to determine the
turning moment produced by the mechanism, and it was found to be higher than
that obtained by Persian turbine of similar size. Physical prototypes of both variants
were developed, and the generated power was measured in the two cases. The power
generated in the proposed method was found to be higher than that obtained in Persian
turbine. In future, more accurate model of the prototype is being planned and realistic
formulation considering the effect of drag and power coefficient of wind would be
considered.
Reference
1. Gasch R, Twele J (eds) (2011) Wind power plants: fundamentals, design, construction and
operation. Springer
2. Bennert W, Werner UJ (1989) Windenergie (Wind Energy), VEB Verlag Technik
3. Exawind: https://www.bastantecnologies.com. Last accessed 2018/12/01
4. Wind turbine with Flipping airfoils: https://www.youtube.com/watch?v=zDW7jGOZGo0. Last
accessed 2018/12/10
5. The Feathering Sidewheel: https://nautarch.tamu.edu/PROJECTS/denbigh/WHEEL.HTM. Last
accessed 2019/01/16
Intuitive Manipulation of Delta Robot
Using Leap Motion
1 Introduction
Robots in industries can be broadly classified as serial and parallel robots or manipula-
tors. The former have links connected from the base link to the last link (end-effector)
in series, generally using single degree-of-freedom (DOF) joints. The latter have the
base and the end-effector connected through a set of parallel links or arms. The serial
robots have larger workspace and are simpler to design and develop. However, they
are prone to inaccuracies due to the hanging of one link with respect to its previous
link. On the other hand, the parallel robots have better rigidity and accuracy but have
poor workspace coverage and are bound to get into singularity.
Serial robots are used to perform operations that require more movement of the
robot arm link pick and place, palletizing, painting, welding, etc. These robots are
generally taught positions and the corresponding motion between the positions using
P. Giridharan · G. Udupa
Department of Mechanical Engineering, Amrita Vishwa Vidyapeetham, Amritapuri, Clappana,
India
R. G. Chittawadigi (B)
Department of Mechanical Engineering, Amrita School of Engineering, Amrita Vishwa
Vidyapeetham, Bengaluru, India
e-mail: [email protected]
a teach pendant, which is a handheld device with few buttons and display unit.
Some simple robot programs can also be developed using a teach pendant, and
for complicated robot programs, proprietary environments are provided by the robot
manufacturers to simulate, test and feed the program to the robot controller. However,
it requires thorough training of robot technicians before they can attempt to teach
and control robot using a teach pendant.
Delta robots or manipulators are one of the commonly used parallel robots in
industries. Primarily, they are used to sort objects moving on a conveyor belt, where
an image processing unit identifies the objects moving and accordingly controls the
robot. Delta robots are also used to pick objects from one place and place onto another.
In such applications, again teach pendants are used for teaching or demonstration
purpose by a technician. The robot has three DOFs, and its architecture (discussed
later in the paper) allows only translation of the end-effector along the three principal
axes. Hence, to control using a teach pendant, which just has buttons (X+ , X−, Y +
,Y −, Z+ and Z−), it may require clear understanding of the Cartesian frame with
respect to which the motion is defined.
For intuitive control of robots, many alternative methods have been proposed.
Master–slave manipulation is one such scheme where the user wears an exoskeleton
to control a physical robot, acting as a slave. One such work is reported in [1].
Another intuitive method is to have identical kinematic chains for master and slave
manipulators. Joint encoders on master send out data to the slave manipulator to copy,
as reported in [2]. Vision-based sensors such as Microsoft Kinect, as reported in [3],
are also used to intuitively control a robot arm. Leap Motion is another vision-based
sensor that has very good accuracy of tracking human hands. Few researchers have
also successfully used it to control serial robots, as reported in [4]. The authors have
proposed to use Leap Motion to intuitively control a delta robot which is reported in
this paper. Recently, it was found out that a related work was also published at [5].
Delta robot is one of the commonly used parallel manipulators, whose architecture
has been used in commercial robots like ABB FlexPicker. The robot consists of a base
platform in the shape of an equilateral triangle, which is fixed to the ground and has
three arms. A revolute joint connects the base with bicep link of each arm. The bicep
link is connected to a pair of identical links through universal joints. The identical
links are individually connected to an equilateral triangle (of lower dimension than the
base) and referred to as the moving platform, through another set of universal joints.
The connection of base link till the moving platform exists through three parallel
chains, and hence the robot is classified as a parallel manipulator. The architecture of
the robot ensures that the identical links between a bicep arm and moving platform
always remain parallel. Overall, the moving platform is constrained to only have
translation motion with respect to the base link and hence the moving platform
always remains parallel to the base link. Thus, this robot architecture is ideal for
Intuitive Manipulation of Delta Robot Using Leap Motion 1513
Platform
Forearm
Base
3
Bicep
1 2
Fig. 1 Delta manipulator with 3 DOFs for translation along X, Y and Z directions
Forward Kinematics: For given values of θi , i.e., individual joint angles, deter-
mination of the position (X, Y and Z) of the end-effector or platform is known as
forward kinematics. Parallel robots generally have multiple solutions for the forward
kinematics as well, and they are complicated to solve. It is useful for simulation
purposes to check collision between links, to check for singularity, to determine the
position of the end-effector so that effective control strategies can be implemented.
3 Physical Prototype
The base link was made in a triangular shape using plywood. The bicep links were
fabricated using 3D printing, where the extensions on the bicep arm had spherical
shaped ends, for the connection with forearm links. The spherical joints were created
using tie-end rods, which were connected to the either ends of forearm links. The
platform was fabricated using 3D printing and again had extensions with spherical
shape for the tie-end rods to be connected. The assembled robot is shown in Fig. 1b.
The inverse kinematics of the robot was implemented as a Python program on a
system running Windows 8.1 operating system. The program made available online
[7] takes the X, Y and Z coordinates of the end-effector and determines the joint
angles. These joint angles are then sent to Arduino Mega 2560 Controller, connected
through USB in the form of serial port communication. The Arduino has a program to
read the incoming data and convert to corresponding step value to be fed to TB6560
motor driver, which further sends data to DC stepper motors (of NEMA 23 make) to
rotate the bicep arm with respect to the base link. A 24V:3A external power supply
is connected to the stepper motor driver to actuate the motors. The circuit diagram
depicting the connections between the components is given in Fig. 2.
Several test motions like incremental jogging along X, Y and Z and circular motion
of the end point in a plane parallel to XY plane were developed as Python programs
and executed. The robot was found to move as expected. The integration of robot
with Leap Motion is covered next.
Leap Motion is a vision-based sensor available for a low cost (around 80 USD) and
has revolutionized the gaming industry. It has 3 LEDs and two cameras, and the
technology has very good accuracy to track the human hands in its hemispherical
view. The device comes with a proprietary software to be installed on a computer.
The device is connected to the computer through USB. The software gets the data of
the position of human hands (10 fingers) and interpolates position of all important
joints and bones in human hand. Though it has found good acceptance in the gaming
and virtual reality-based applications, many researchers have used it in their research.
Intuitive Manipulation of Delta Robot Using Leap Motion 1515
The second author has proposed its application to interact with virtual mechanisms,
as reported in [8].
In this work, Python program has been developed to track the human hand and
the position information of the index finger of the right hand is recorded for every
time frame. A coordinate system is attached at the center of Leap Motion device,
and coordinates measured are in this frame. The incremental motion of the fingertip
is provided as input to the inverse kinematics module/method of the program, which
determines the joint angles to be supplied to the delta robot. The new set of joint
angles are sent through Arduino to the drivers, and motors move correspondingly.
The integration of the Leap Motion with the computer and then the delta robot was
successfully completed. Few tests conducted by a user, as illustrated in Fig. 3, are
reported next.
1516 P. Giridharan et al.
Leap Motion
Computer
Fig. 3 Delta manipulator with 3 DOFs for translation along X, Y and Z directions
The intent of this work was to intuitively control a delta robot. To achieve this,
several trials of human interaction were performed. The user moved his hand approx-
imately along the X, Y and Z directions, with respect to the frame attached on the
sensor. The position of the fingertip for few of these tests and the corresponding joint
trajectories, after inverse kinematics for each of the position, are shown in Fig. 4. The
resultant motion of the delta robot was observed to follow the same trend as the input
motion. However, due to lack of encoders (at joints or at end-effector), the accuracy
of the prototype was unable to be determined. In the future, suitable sensors would
(a) Finger-tip along Z (c) Finger-tip along Y (e) Finger-tip tracing a circle
(b) Joint trajectories for ‘a’ (d) Joint trajectories for ‘c’ (f) Joint trajectories for ‘e’
be used for accurate motion through feedback control. Finally, it was observed that
the control of the robot was intuitive as compared to conventional teach pendants.
5 Conclusions
In this paper, a physical prototype of delta parallel robot developed by the authors
has been integrated with Leap Motion sensor. The sensor tracks a fingertip of a user,
and the incremental motion of the fingertip is mapped onto the corresponding motion
of the moving platform of the robot. The proposed methodology was found to be
intuitive for the user to effectively control the robot. In the future, encoders at joint
and for the end-effector would be used for achieving better accuracy.
References
1. Arun K, Nair AP, Chittawadigi RG, Venkatesh K, Patnaik S (2018) An efficient methodology
to determine the usability of exoskeleton to control a serial manipulator. In: Proceedings of 5th
IFToMM Asian conference on mechanism and machine science
2. Ajith A, Nambiar NM, Akshay VP, Ajit A, Ramachandran R (2018) SAKSHA-self automated
kinematic smart haptic arm. Procedia Comput Sci 133:711–717
3. Du G, Zhang P, Mai J, Li Z (2012) Markerless kinect-based hand tracking for robot teleoperation.
Int J Adv Robot Syst 9(36)
4. Bassily D, Georgoulas C, Guettler J, Linner T, Bock T (2014) Intuitive and adaptive robotic arm
manipulation using the Leap Motion controller. In: Proceedings of 41st international symposium
on robotics
5. Zhang X, Zhang R, Chen L, Zhang X (2019) Natural gesture control of a delta robot using Leap
Motion. In IOP J Phys Conf Ser 1187(3)
6. Williams II RL (2016) The delta parallel robot: kinematics solutions, internet publication, www.
ohio.edu/people/williar4/html/pdf/DeltaKin.pdf
7. Code: https://github.com/giridharanponnuvel/Delta-Robot-Inverse-Kinematics (2019)
8. Pullil S, Chittawadigi RG (2019) Natural control of virtual models of mechanisms using Leap
Motion for interactive learning. In: 4th international and 19th national conference on machines
and mechanisms
A Computation Model of Contact
Interaction Between the Scaphoid and Its
Neighboring Bones Using Bond Graph
Approach
1 Introduction
The human wrist represents the most complex joint system in the human body.
The wrist joint consists of eight small, intricately, and unique shaped carpal bones,
interposed between the distal end of radius and ulna and base of the five metacarpals.
The extrinsic and intrinsic ligaments provide stability to wrist joint. The morphologic
variations of the carpal bones and their contacts with soft cartilage layer have its own
significance. In regard to biomechanical modeling of natural joint systems, there
are two typical ways to create a computational model of a biological joint system:
finite element analysis (FEM) [1] and multi-body simulation (MSB) [2]. However,
the existing models of the wrist had not considered the physiological path of the
implemented ligaments. The cartilage layers of the different wrist bones and wrist
joint systems, respectively, have also not been considered. Also, the order of the
system modeled based on FEM depends on the number of nodes and turns out to be
extremely large in cases of mating bones with intervening soft cartilage. However,
the large order lays a severe burden on the computation involved [3].
Alternatively, modeling of the biological joints requires a simpler model which
requires less computational time. The objective of this work is to present one such
alternative model of cartilage layer at a synovial joint. It is of interest to understand
the effects of change in the contact area between the bone and the cartilage layer, the
deformation of the soft cartilage between the bones, and the force distribution as the
contact area changes, during dynamic interactions.
To demonstrate the simplified conceptual model and its implementation, a bond
graph model of contact interaction of the scaphoid, one of the carpal bones, with its
neighboring bones is presented. The surface geometry of each bone is represented by
point cloud. The intervening cartilage layer between bones has nonlinear characteris-
tics, and its constitutive relations have been logically and systematically derived. The
ligaments connected between the scaphoid and its neighboring bones are modeled as
linear, tension only spring element in combination with a linear damper. The equa-
tions governing the dynamics of the system have been derived from the bond graph
model [4].
This paper is organized as follows. Section 2 elaborates the bones geometry,
the proposed nonlinear model of the cartilage layer, and the ligament modeling.
The multibond graph model for the proposed system is discussed in Sect. 3. The
simulation results are analyzed in Sect. 4. Concluding remarks and potential for
future work are discussed in Sect. 5.
2 Proposed Model
In the present study, the bones are considered to be rigid in nature. The surface
geometries of the mating bones are represented by 3D point clouds. The point cloud,
representing the irregular surface geometry of all bones, is taken from a digital
database of carpal anatomy developed and published by Moore et al. [5]. Three-
dimensional Voronoi and Delaunay triangulation-based surface reconstruction algo-
rithm have been used to generate the planar triangulated patches over 3D point cloud
[6]. Surface normals are computed at respective centroids of these patches. With
reference to the frame fixed in the respective bone, the directions of these surface
normals are always fixed. Multibond graph sub-models are used to represent their
rigid body mechanics of all bones [7].
Instead of modeling the soft material using the conventional finite element method
(FEM), which can prove to be an elaborate task, a nonlinear stiffness field is used to
characterize the behavior of the intervening soft cartilage layer in the normal direction
at the patch. The detailed bond graph model for the same had been presented by Pathak
and Vaz [8]. In this study, the contact force model does not account the contribution
of the contact area between the bone and the cartilage lyaer. In the present study, the
1522 A. K. Pathak and A. Vaz
areas of patches lying in the contact region between the bone and cartilage layer are
considered. A suitable candidate function, representing the cartilage layer stiffness,
is proposed as,
F̄B1i K
B2 j
⎧ ⎛ ⎞ ⎫
⎪
⎪ ⎪
⎪
⎨ 2
B1i q B2 j ⎜ A B2 j ⎟ ⎬
= n̂ B1i K
T
⎜ ⎟ A B1 B2 r̄ n̂ B1i .
⎪ 2 ⎝ ⎠ i j B1i
⎪
⎪
⎩ B1i q B2 j A B2 j + A B1i ⎪
⎭
B1 q B2max 1 − B1 q
2
B2max j=1 i=1
(1)
where R is positive constant and B2 j ṙ B1i n̂ is a component of relative velocity
B1i
2.3 Ligament
Ligaments are passive structures connecting bones, stabilizing joint systems, guiding
and constraining the joint motion. In the present study, the behavior of each ligament
is modeled using linear, tension only spring element in combination with a linear
damper [9], as shown in Fig. 2. The information of anatomical directions and loca-
tions of insertion points on the bone surfaces, for each ligament, is identified and
A Computation Model of Contact Interaction Between … 1523
RLig
approximated based on anatomical atlases and literature data. The ligament offers
constraints to bones only when it is in tension.
The force offered by the ligament is,
F Lig = K Lig qLig B1L r̂ B1L + R Lig B1 L ṙ B1 L B1 L r̂ B1 L
(3)
A bond graph model is developed for the interaction between a pair of two mating
bones separated by intervening cartilage layer and a ligament connected between
them and is shown in Fig. 3. Similarly, all the bones interacting with the scaphoid
and respective ligaments are modeled using bond graph. The model is initiated using
flow mapping which represents the kinematics of the system under consideration.
The
0 two bones are considered to be rigid and modeled based on rigid body dynamics.
C B2 r B2 L × is the skew symmetric matrix, represented as MTF, obtained from posi-
T
tion vector C0 B2 r B2L = C0 B2 x B2L C0 B2 y B2L C0 B2 z B2L . Similarly, all the MTF has been
formulated. The stiffness and damping characteristics of intermittent cartilage layer
are same as discussed in the Sect. (2.2). The junctions 0 F B1 and 0 F B2 are repre-
i j
senting the force offered by the cartilage layer at the ith and jth patch, respectively.
The junctions 0 F B1 and 0 F B2 are representing the ligament insertion points on the
L L
bone 1 and bone 2, respectively. The force offered by the ligament is the effort on the
junction 0 B2 F B1 . This effort is shared by all bonds connected to this junction. Direc-
L L
tions of power bonds clearly tell that the two bones experience the same magnitudes
of forces but in opposite directions.
1524 A. K. Pathak and A. Vaz
Fig. 3 Bond graph model for contact interaction between two mating bones and a ligament
connected between them
The dynamics of scaphoid bone while interacting with its neighboring bones has
been simulated in this work, based on the model shown in Fig. 3. The model is
general and is valid for motion in three dimensions. The simulation is for a duration
of 1 s. A force due to acceleration due to gravity is acting on the scaphoid bone.
The articulation between the scaphoid and its neighboring bones and the minimum
distance between them are mentioned in Table 1.
An initial movement of 2 mm is given to a point PS on the surface of scaphoid bone
along the positive y-direction of inertial frame {0}. The desired smooth trajectory of
that point is given in (4).
r PS (t) = 0, t<0
= a0 + a1 t + a2 t 2 + a3 t 3 + a4 t 4 + a5 t 5 + a6 t 6 + a7 t 7 0 < t<t f (4)
= r PS t f t > t f
A Computation Model of Contact Interaction Between … 1525
Table 1 Articulation
Articulation between bones Minimum distance
between the scaphoid and its
Bone 1 Bone 2 between bones (mm)
neighboring bones
Scaphoid Radius 0.86
Scaphoid Capitate 0.57
Scaphoid Lunate 1.13
Scaphoid Trapezium 0.74
Scaphoid Trapezoid 0.87
0.0248
0.0247
0.0246
0.0245
0.0244
0.074
0.0738 0.0816
0.0814
0.0736
0.0812
0.0734
0.081
Fig. 4 Position of COM of the scaphoid bone observed and expressed in the inertial frame
30
Radioscaphoid (RS)
4
Scaphotrapezium (STM)
20 Scaphotrapeziod (STD)
Scaphocapitate (SC)
2
Scapholunate (SL)
10
0
0.02 0.04 0.06
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
40
Colletral Radioscaphoid Palmar Scaphotrapezoidal
30 Palmar Radio - scapho - capitate Palmar Scaphocapitate
Dorsal Radioscaphoid Palmar Scapholunate IO
Dorsal Radio-scapho-lunate Dorsal Scapholunate IO
20
Radiopalmar Scaphotrapezial Proximal Scapholunate IO
Ulnopalmar Scaphotrapezial
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 5 (a) Contact force offered by the cartilage layer, (b) Ligament forces
5 Conclusion
The effect of soft cartilage between the scaphoid and its neighboring bones is
emulated.
Instead of modeling the cartilage using the conventional Finite Element Method
(FEM), which can prove to be an elaborate task, a nonlinear stiffness and damping
field is used to represent the behavior of the intervening soft cartilage layer in the
normal direction at the contact area. The thickness of the cartilage layer between the
mating bones does not deform below a minimum specified thickness. The stiffness
of the soft cartilage is nonlinear. The force offered by the soft cartilage layer between
the two bones, increases asymptotically as the distance between the surface of bone 1
and bone 2 moves closer to minimum thickness of soft cartilage layer. The ligaments
are modeled as a linear, tension only spring element in combination with the linear
damper. They are guiding and constraining the joint motion. Future work may include
A Computation Model of Contact Interaction Between … 1527
the studies on the effects on human wrist while considering all bones and ligaments
of the human wrist based on the proposed model.
References
1. Wilson W, Van Donkelaar CC, Van Rietbergen R, Huiskes R (2005) The role of computational
models in the search for the mechanical behavior and damage mechanisms of articular cartilage.
Med Eng Phys 27(10):810–826
2. Majors BJ, Wayne JS (2011) Development and validation of a computational model for
investigation of wrist biomechanics. Ann Biomed Eng 39:2807–2815
3. Pérez-González A, Fenollosa-Esteve C, Sancho-Bru JL, Sánchez-Marín FT, Vergara M,
Rodríguez-Cervantes PJ (2008) A modified elastic foundation contact model for application in
3D models of the prosthetic knee. Med Eng Phys 30(3):387–398
4. Karnopp DC, Margolis DL, Rosenberg RC (2000) System dynamics: modeling and simulation
of mechatronic systems, 3rd edn. Wiley-Interscience, New York
5. Moore DC, Crisco JJ, Trafton TG, Leventhal EL (2007) A digital database of wrist bone
anatomy and carpal kinematics. J Biomech 40(11):2537–2542
6. Amenta N, Bern M, Kamvysselis M (2005) A new Voronoi-based surface reconstruction
algorithm, pp 415–421
7. Mukherjee A, Karmakar R (2000) Modelling and simulation of engineering systems through
bondgraphs. Narosa Publishing House, New Delhi
8. Pathak AK, Vaz A ()2019 A simplified model for contact mechanics of articular cartilage
and mating bones using bond graph. In: Badodkar DN, Dwarakanath TA, (eds) Machines,
mechanism and robotics, lecture notes in mechanical engineering, 1st edn. Springer Nature
Singapore Pte Ltd. 2019, pp 549–559
9. Carrigan SD, Whiteside RA, Pichora DR, Small CF (2003) Development of a three-dimensional
finite element model for carpal load transmission in a static neutral posture. Ann Biomed Eng
31(6):718–725
10. Eschweiler J et al (2016) Development of a biomechanical model of the wrist joint for patient-
specific model guided surgical therapy planning: Part 1. In: Proceedings of Institute Mechanical
Engineering Part. H J Eng Med 230(4):310–325
Mathematical Model of SMA Spring
Actuator in a Miniature Flexible Tube
Robot
Abstract Shape memory alloy (SMA) has the potential to be used as an actuator
in miniaturized and flexible models. In this paper, an existing theoretical model is
used to predict the force capabilities of an SMA spring in challenging applications
like a miniature flexible tube robot for tumor irradiation applications. A heat transfer
model based on Joule heating is used to predict the temperature variations with time.
A Brinson model-based phase kinetics together with constitutive model predicts the
variation of martensite fraction and recovery force with temperature. The results help
to estimate the force capabilities of an SMA spring when used as an actuator.
1 Introduction
The use of robots in medical applications has gained momentum in recent years and
making a huge impact in healthcare system. The initial idea of incorporating robotics
in health care was to relieve medical personnel from their routine tasks, but now with
ever growing technology, it has been extended to bring flexibility to the operating
tasks like surgery, endoscopy, laparoscopy, etc. Many of these applications require
miniature robots that can move inside human body and perform the assigned tasks.
One of such applications is to irradiate a tumor which is either inside or outside
a human body, and for that application, a flexible tube like optical fiber tube can
be used. There are many actuation methods that have been identified to provide
required maneuverability to flexible tubes. In the literature on designs of flexible
tube robots, mechanical joints like single revolute, multi-revolute and other multi-
DOF joints were found which provided high maneuverability to tube [1]. Despite
the constant use of mechanical actuators in designs of flexible tubes, cannulas and
continuum robots; friction, inertia and requirement of auxiliary transmission system
to convert motion, have limited their use in microactuation with minimum hardware
requirements.
Another popular type of actuation is cable-driven systems, where these cables
are wrapped around a wire binder to keep in tension. These cables are driven via a
transmission system which pulls one cable and release another to bend the flexible
tube [2]. Even a set of bellows arranged in parallel as the vertices of an equilateral
triangle on a moving platform were used for a continuum endoscope configuration
[3]. Though the performance is good with pneumatic actuators, the problem is need
of auxiliary components that makes the system heavy. In the category of actuation
systems, shape memory alloys have been extensively used for flexible tube configura-
tions like endoscopes [4, 5]. A prototype of laparoscopic tool by University of Patras
has been developed, which utilizes SMA wires in antagonist tendon configuration
[6]. Another recent design includes neurosurgical robot which uses SMA antagonist
springs in combination with tendon sheath [7].
The extent of flexibility required in medical operations could be different in each
scenario, and SMAs have been preferred due to their good power to weight ratio.
These actuators require less space and less accessories, and as a result, they are
popular in minimal invasive surgeries or medical applications. It can be concluded
that SMA materials in one form or another have been used in flexible robotics and
when compared to conventional as well as non-conventional actuators, they have
shown great potential in the application area.
In the present work, a study is conducted on force capabilities of an SMA spring
which is typically used as a prime actuator in miniature flexible tube robots. The
mathematical model for the same is discussed below in subsequent section.
2 Mathematical Model
Any product development needs a mathematical model which comprises of all the
functional bodies that form a final design of the product. The proposed block diagram
of mathematical model of flexible manipulator is shown in Fig. 1. The focus of this
paper is on the highlighted part which includes heat transfer, phase kinetics and
constitutive model.
As shape memory alloys actuate through electric heating, the temperature of
material is indirectly controlled by electric current. The equation relating the heat
transfer process with Joule heating is as follows [8]:
dT
mCv = V I − h Ac (T − Tamb ) (1)
dt
Mathematical Model of SMA Spring Actuator in a Miniature … 1531
VI
1 − e− U + To + (To − Tamb )e− U
t t
T = (2)
h Ac
ρvolCv
U= (3)
h Ac
shows temperature variation (both heating and cooling) with time as shown in Fig. 2.
The variation of temperature with time in an exponential function as given by (2).
During heating cycle, the initial temperature is equal to the ambient temperature.
Thus, the extreme right term in (2) vanishes and solution becomes
VI
1 − e− U + To
t
T = (4)
h Ac
A steep line for temperature in Fig. 2 during heating cycle indicates that the heat
transfer to ambient is very small as compared to resistive heating when a constant
power is supplied to SMA spring. The maximum temperature attainable is 91 °C
when supplied with a constant current of 1.6 A up to 2.2 s. During cooling, the
initial temperature is the final temperature of the heating cycle and the variation of
temperature is given as per (2). In this cycle, the decrease in temperature is slow due
to free convection in air. Thus, the cooling time is found around 46.8 s which cools
down the heated SMA spring below M f , as shown in Fig. 2.
1532 N. Bhatt et al.
Once the temperature is known for the time interval for which current is applied
to the spring, the next task is to determine the force exerted on SMA spring while
heating. In Fig. 3, an SMA spring is stretched by an amount δ o and both ends of it
are fixed such that the displacement is zero, i.e., δ = δ o . The initial deflection results
in 2.98 N force. The constitutive equation of shear stress and shear strain of SMA
where τ , γ , G, Ω s and θ are shear stress, shear strain, modulus of rigidity, phase
transformation coefficient for spring and coefficient of thermal expansion, respec-
tively. The change in material configuration is denoted by martenistic fraction ξ. As
material exists in two phases: low temperature phase (martensite) and high temper-
ature phase (austenite). The material fraction ξ denotes the amount of martensite
present in material. The martensite can again found in two forms: stress induced (ξ s )
and stress free (ξ t ). When the spring is subjected to an external force F, the maximum
shear stress developed is given as per (6)
8K FC
τ= (6)
π d2
where K, C, d are Wahl correction factor, spring index and diameter of SMA spring
wire, respectively. To simplify the relation between force and shear stress, it is
assumed that the maximum shear stress developed is the average shear stress acting
on spring. When substituting (6) in (5) and neglecting effect of thermal expansion,
(9) gives the relation of force and axial deflection as:
C1 (Fi − Fo ) = C2 G(ξ )(δi − δo ) − C2 δl G(ξ ) ξsi − ξso (7)
8K F D d
C1 = C2 = (8)
πd 3 π N D2
G(ξ ) = G a + ξ (G m − G a ) (9)
Now, when the current is applied to spring, the temperature rises as per (2) and
when it reaches to austenite transformation temperature As , the springs try to contract,
but since both of its ends are fixed, a recovery force develops which can be measured
by a force sensor. The amount of recovery force gives us an idea of using SMA
springs as an actuator for our robot. The details of material properties are shown in
Table 1. The simulation results of spring are discussed in results and conclusions
section.
3 Results
Consider the spring as shown in Fig. 3, since both ends are fixed so initial
displacement is equal to final displacement, which makes (7) as
1534 N. Bhatt et al.
C1 (Fi − Fo ) = −C2 δl G(ξ ) ξsi − ξso (10)
Once, we know the temperature, martensite fraction can be obtained from Brinson
phase transformation equations [10] while considering the relation of shear stress
and normal stress as per the von Mises theory
√
ξi−1 3τi
ξi = cos aa Ti − As − +1 (11)
2 CA
ξsi−1
ξsi = ξsi−1 − (ξi−1 − ξi ) (12)
ξi−1
ξti−1
ξti = ξti−1 − (ξi−1 − ξi ) (13)
ξi−1
Here, i and i − 1 denote current and previous state, respectively. The change in
martensite fraction and stress-induced martensite fraction with temperature is shown
in Fig. 4 with their initial values as 1 and 0.5, respectively. The variation shows
a gradual decrease in both the fractions, and as predicated, it becomes zero after
reaching Af .
The recovery force can be calculated by following Newton–Raphson’s method
as mentioned by Ho [9]. The variation of recovery force with temperature is shown
in Fig. 5. As depicted, the force increases with temperature and as it reaches the
Mathematical Model of SMA Spring Actuator in a Miniature … 1535
Af , it becomes maximum and the further increase in temperature does not affect
the magnitude of force as martensite fraction becomes 0. The maximum value of
recovery force is found 3 N.
1536 N. Bhatt et al.
4 Conclusion
The results for recovery force show that with rise in temperature, recovery force
increases and becomes maximum to 3 N for mentioned parameters of spring. Such
spring when stretched within detwinned martensite limits, could lift up to 306 gm
of load with recovery of all plastic deformation. The force to weight ratio of spring
is found around 253. Also, the variation in martensite fraction with temperature is
found in accordance with the available literature. The results show a great potential
of SMA spring to be used as an actuator where flexibility is desired in miniature
models.
References
1. Faraz A, Payandeh S (1997) Synthesis and workspace study of endoscopic extenders with
flexible stem. J Mech Des 119:412–414
2. Chen Y, Tanaka S, Hunter IW (2010) Disposable endoscope tip actuation design and robotic
platform. In: Proceedings of the 32nd annual international conference of the IEEE EMBS.
IEEE, Argentina, pp 2279–2282
3. Garbin N, Wang L, Chandler JH, Obstein KL, Simaan N, Valdastri P (2018) A disposable
continuum endoscope using piston-driven parallel bellow actuator. In: Proceedings of the
international symposium on medical robotics (ISMR). Atlanta, pp 1–6
4. Ikuta K, Tsukumoto M, Hirose S (1998) Shape memory alloy servo actuator system with electric
resistance feedback and application for active endoscope. In: Proceedings of the international
conference on robotics and automation. PA, USA, pp. 427–430
5. Maeda S, Abe K, Yamamoto K, Tohyamaand O, Ito H (1996) Active endoscope with SMA
(Shape Memory Alloy) coil springs. In: Proceedings of the ninth international workshop on
micro electromechanical systems. San Diego, USA, pp 290–295
6. Giataganas P, Evangeliou N, Koveos Y, Kelasidi E, Tzes A (2011) Design and experimental
evaluation of an innovative SMA based tendon driven redundant endoscopic robotic surgical
tool. In: Proceedings of the 19th mediterranean conference on control and automation. Corfu,
Greece, pp 1071–1075
7. Ho M, Kim Y, Cheng S, Gullapalli R, Desai JP (2015) Design, development and evaluation of
an MRI-guided SMA spring-actuated neurosurgical robot. Int J Robot Res 38(8):1147–1163
8. Rao A, Srinivasa AR, Reddy JN (2015) Design of shape memory alloy actuators. Springer
9. Ho M, Desai JP (2012) Characterization of antagonistic SMA spring actuators for use in a
MRI-compatible intracranial robot. In: Proceedings of DSCC2012-MOVIC2012. ASME, FL,
USA, pp 1–5
10. Brinson LC (1993) One dimensional constitutive behavior of shape memory alloys: thermo
mechanical derivation with non constant material functions and re-defined martensite internal
variable. J Intell Mater Syst Struct 4(2):229–242
Analysis of Inner Block in a Roller Chain
Using Glass Reinforced Composites
Abstract Roller chains are used in power transmission; the materials used in a roller
chains are generally steel. In the advancement of composite materials in various fields
for its high strength-to-weight ratio, its high tensile strength and its lower density
reduce mass as well. This paper does an finite element analysis of inner block which
comprises two plates, bushes and roller with fiberglass composites such as E–S Glass
and compares them with the material that is more prominently used—medium carbon
steel
1 Introduction
Roller chains are one of the most important form of power transmission systems that
have wider application where the center distance is usually small to medium with
higher efficiencies in comparison to the other drive systems [1]. In usual conditions,
a roller chain material is medium carbon steel, whereas this work is done using a
fiberglass composites due to its higher tensile strength and lower density which in
turns provide lower mass.
The parts of a roller chain include roller, bushes, inner and outer plates and pins. The
block which holds two inner plates, bushes and rollers are collectively called as an
inner block, and the part containing two outer plates and pins is called as an outer
link [2]. The materials usually used are steel, with medium carbon steel being the
most predominant one and stainless steel (austenitic) is used at non-corrosion and
special environments.
The chain material is also selected with the force that is used. The amount of
braking load [3] also has an impact on the load of the chain as it directly manages
with the application and material that should be used for the chain.
3 Fiberglass
Fig. 1 Tensile strength versus temperature for E&S Glass temperatures between −196 and 548 °C
[5]
Analysis of Inner Block in a Roller Chain Using Glass … 1539
The tensile strength at room temperature of E&S Glass according to ASM [6] lies
at 3800 MPa, and 4590 MPa respectively whereas that of AISI 1055 [7] (ASTM) is
660 MPa.
4 Working Conditions
The conditions are taken from the chain model of ISO Standard 16b, wherein the
braking load is 60 kN [3]. Having the safety factor as 15, the load applied on the block
is 4 kN with bushes and rollers are given with fixed support. The mesh provided is
tetragonal mesh with triangular element on the plate as it is a solid material and the
meshes provided on the bushes and roller are hexahedrons with quadratic elements
(Fig. 2).
5 Results
The above graph shows that the total deformation increases with increase in force,
which proves that the composites follow Hooke’s law although the deformation of
E-Glass is 50% more than that of medium carbon steel (Fig. 3).
1540 V. Gyaneshwar and V. Sivasankaran
5.2 Strain
The strain increases with increase in force, though E-Glass’s strain is a bit higher,
S-Glass and medium carbon steel have almost equal strains (Fig. 4).
5.3 Stress
It is shown that the stress increases linearly with increase in force, also that the stress
is almost equal for all the three materials (Fig. 5).
Analysis of Inner Block in a Roller Chain Using Glass … 1541
Fig. 4 Strain
6 Conclusion
Although in the cases of stress, strain and total deformation, the existing medium
carbon steel looked better on working, and these are the following conditions that
will be considered as well.
• Other than the plate, bushes and rollers were kept as a fixed support, an this was
done to understand the relationship between the stress, strain and the deformation
of the bushes under such circumstances. In actual conditions, though bushes are
kept in press fit, the rollers are not fixed, and thus the friction is reduced provided
that the tolerance is also reduced.
• Deformation is compared with the theoretical values. That gave us the idea whether
the deformation is under control or not (Fig. 6).
• Strain is almost same for both S and medium carbon steel, and for E-Glass it is
higher because of higher deformation. Since the deformations are lesser than the
actual, the strain is also under control
• Stress follows Hooke’s law, and the stresses are identical with both the glasses
and slightly higher with steel. The effectiveness is the part where the stresses of
the material show the advantage of the composites (Fig. 7).
1542 V. Gyaneshwar and V. Sivasankaran
Fig. 5 Stress
• Mass plays a vital role in the choice of composites. As density of the fiberglass
is only one-third of carbon steel, the mass automatically gets reduced. With its
high weight-to-strength ratio, greater working conditions at lower mass can be
attained.
• Lubrication, the necessity of a lubricant, is removed as fiberglass has a layer of
SiO2 which acts as a self-lubricant [5]. Increasing and introduction of boron also
have an impact on the lubrication of the material.
Analysis of Inner Block in a Roller Chain Using Glass … 1543
Fig. 6 Deformation, with comparison between the allowable and the actual. We could see that the
actual values are negligible of the allowable deformation
References
1. Dekker M (2005) Chains for power transmission and material handling, design and applications
handbook, 2nd edn. CRC Press, Michigan
2. Jackson & Moreland (1958) Design manual for roller and silent chain drives, 3rd edn. Association
of Roller Silent Chain Manufacturers, Detroit
3. Indian Standard (2006) Short pitch transmission precision roller chains and chain wheels, 2nd
rev. Bureau of Indian Standards, New Delhi
4. Bhosale P, Patil AR (2018) Weight optimisation of chain link using glass fiber composite as
alternative material. Int Res J Eng Technol 5:6
5. Hartman D, Greenwood ME, Miller DM (2006) High strength glass fibers. AGY, South Carolina
6. Wallenberger FT, Watson JC, Li H (2001) Glass fibers. ASM International, Ohio
7. ASTM (2006) Standard specification for steel bars, carbon and alloy, hot-wrought, general
requirements. ASTM A29. ASTM International, Philadelphia
Study on the Effect of Process
Parameters on Machinability
Performance of AA7050/B4C Metal
Matrix Composite on Wire Cut EDM
Arvind Kumar , Subhan Pandey, Virendra Singh, and Ram Naresh Rai
Abstract The present investigation has been initiated based on the wire electric
discharge machining of Al7050/7.5 B4C stir cast metal matrix composite. Taguchi
L9 orthogonal array has been used to analyses the concurrent effect pulse current
(I p ), pulse-on time (T on ) and servo voltage (S v ) on material removal rate (MRR) and
surface roughness. From Taguchi analysis, it is observed that the optimal setting of
process parameters for maximum MRR is I p 3T on 3S v 3, whereas for minimum surface
roughness, the optimal configuration of process parameters is I p 3T on 2S v 1. The error
estimated between predicted MRR and surface roughness with experimental MRR
and Surface roughness at the optimal setting of process parameters is within ±5%.
Analysis of variance shows that I p contributes maximum towards MRR and T on
contributes maximum towards surface roughness.
1 Introduction
Metal matrix composite is a new engineered material which possesses the inherent
properties of matrix and reinforcements [1]. Metal matrix composites have high
specific strength most suitable for aerospace and automobile components [2]. There-
fore, machining of those components with close tolerance and in desired shape is
the major industrial concern especially through the conventional machining process
[3, 4]. Wire cut EDM serves this purpose up to some extent. Some of the previous
findings on wire cut EDM are as follows: Velmurugan et al. [5] investigated on
nickel-titanium shape memory alloy machining in WEDM. They used servo voltage,
pulse-on time, pulse-off time, current and wire-speed as an input parameter. Servo
voltage is the major influencing factor for MRR and surface roughness. Choudhuri
et al. [6] investigated on surface quality of WEDM machined stainless steel 316.
T on , T off , I p , S v , WT were the input parameters for machining 25 sets of experiment.
From ANOVA result, Ton found to be the most controlling factor for roughness as
well as recast layer. Jain et al. [7] have used ANN for the evaluation machining
performances with the simultaneous effect of T on , T off , I p and bed speed on surface
roughness and AE signals. The result shows that better R values obtained when
network trained with 70% of the data compared to 50 and 60% of data. Nain et al.
[8] have evaluation of surface roughness and waviness of the WEDM of aeronautic
superalloy with the various process parameters. The results concluded that apart
from T on T off and S v wire tension has significant effect on surface roughness. Also,
the machining performance of aeronautics superalloy can be efficiently evaluated by
BP-ANN model as compared to fuzzy logic method. Devarasiddappa et al. [9] have
predicted the surface roughness of Inconel 825 aerospace alloy machined through
WEDM through ANN model. The parametric study shows that the lower SR can
be obtained at low levels of T on and S v . ANN model accuracy recorded as 93.62%
and average predicted error recorded as 6.38% at ANN architecture 4-16-1. This
ANN architecture found optimum, which were statistically validated by conducting
hypothesis tests. ANOVA showed that T on is the most affecting factor for SR with
76.12% contribution, followed by S v and T off , respectively, with 7.18 and 5.3% contri-
butions. Das et al. [10] have used Taguchi L16 OA to conduct WEDM experiment on
Al6061/0.5% SiC/0.5%B4C hybrid nano-composite to evaluate the effect of I p , S v ,
T on and T off on surface roughness. The experimental data have 96.32% accuracy with
predicted values from RSM modelling. Also from the ANOVA Ton is the most influ-
ential factor for surface roughness. Garg et al. [11] have machined the ZrSiO4 /6063
aluminium MMC using CNC WEDM. In this study, the author has developed the
quadratic model for dimensional deviation to evaluate the contemporaneous effect of
machining parameters, namely I p , S v , T on and T off . Experimental results show that
the dimensional deviation (DD) is directly proportional to the pulse-on time and peak
current. The objective of the present study is to synthesis AA7050/B4C composite
through stir casting method and performs machining through CNC wire cut EDM.
2 Experimental Setup
Orthogonal array (OA) designed by Dr. Genichi Taguchi, the primary purpose was
to reduce the number of the experiment by considering whole domain of process
parameters [13]. For the present study, three process parameters are selected which
varies at three-level each, so as per the full factorial experimental design 27 experi-
ments proposed which is not only time taking rather costly as well. Nine tests can also
make a similar study. Therefore, Taguchi L9 orthogonal array experimental design
has used as shown in Tables 2 and 3.
The material removal rate (MRR) has been calculated using Eq. (1). Machining
time has been noted using stopwatch with least count of 1 s. The volume of material
removed is calculated by (length of machining × thickness of machining × average
kerf width).
Length and thickness of machining for the present study considered as 5 mm and
7 mm, respectively, and Kerf widths were obtained from the optical microscope for
entire machining length of 5 mm, with optical magnification at 5×. Taylor Hobson
Profilometer was used to measure surface roughness of the machined surface. Surface
roughness has been measured considering the value of average surface roughness (Ra )
of the machined surface.
Table 3 Experimental
Run order Factor 1 Factor 2 Factor 3
arrangement as per Taguchi
L9 orthogonal array 1 1 1 1
2 1 2 2
3 1 3 3
4 2 1 2
5 2 2 3
6 2 3 1
7 3 1 3
8 3 2 1
9 3 3 2
Study on the Effect of Process Parameters on Machinability Performance … 1549
The experimental results in terms of MRR and surface roughness obtained with
different sets of process parameters are listed in Table 4.
where n and yi are the total number of experiments and values of MRR or surface
roughness for ith experiments. The S/N ratio of each response parameters listed in
Fig. 2 Main effect plot for mean SN ratio of MRR and surface roughness
Table 4. In order to get maximum MRR, the S/N ratio plot is shown in Fig. 2. The
similar graph has obtained from mean effect plot Fig. 3.
Minimum MRR has derived from lower I p T on and S v . Since MRR in EDM
process is a function of spark energy and the spark energy is a function of I p T on and
S v as shown in Eq. (4).
E= Ip T onSv (4)
Fig. 3 Main effect plot for the mean of MRR and surface roughness
Study on the Effect of Process Parameters on Machinability Performance … 1551
Table 5 shows that although all the process parameters have significant contribu-
tion towards achieving maximum MRR pulses current have maximum contribution
compare to other two parameters. For surface roughness evaluation, the pulse-on
time and servo voltages are the highly significant factors than pulse current. The
similar trend shows in Table 6.
Taguchi method can be used to predict the S/N ratio, using the optimal level of the
process parameters can be calculated as Eq. (5) [16].
n
χ̃ = χm + (χi − χm ) (5)
i
where χ̃ is the mean of S/N ratio χ m is the total mean of S/N ratio and n is the number
of process parameters. Table 7 compared the predicted and experimental values of
MRR and surface roughness at optimal setting of process parameters. For MRR, the
optimal configuration of process parameters is I p 3T on 3S v 3 that is pulse current 230
A, pulse-on time 110 µs and servo voltage of 40 V. The predicted MRR estimated
as 31.0086 mm3 /s, whereas the experimental value of MRR at optimal setting of
process parameters evaluated as 30.1242 mm3 /s. The error calculated from predicted
1552 A. Kumar et al.
and experimental MRR calculated as 2.85% which is within the acceptable range.
Similarly, the optimal setting of process parameters for surface rough obtained from
the Taguchi method is I p 3T on 2S v 1, i.e. pulse current 230 A, pulse-on time 108 µs and
servo volt 20 V. The predicted surface roughness estimated as 2.1543 µm. Whereas
the experimental values of surface roughness at optimal setting of process parameters
evaluated as 2.18129 µm. The error calculated from predicted and experimental MRR
calculated as 1.24% which is within the acceptable range.
To verify the consistency of the proposed Taguchi method, the predicted MRR
and surface roughness obtained at the initial set of process parameters (I p 2T on 2S v 2)
compared with experimental one. The errors estimated are within ±5%. Hence, the
proposed Taguchi method is consistent (Table 8).
4 Conclusions
2. Optimal setting of process parameters obtained from the Taguchi analysis for
MRR is I p 3T on 3S v 3, whereas for surface roughness optimal setting of process
parameters are I p 3T on 2S v 1.
3. The error estimated between predicted MRR and surface roughness with experi-
mental MRR and surface roughness at the optimal setting of process parameters
is within ±5%.
4. Proposed Taguchi analysis is consistent as the error estimated between predicted
and experimental values at an initial set of process parameters are within ±5%.
References
1. Kumar A, Rai RN, (2018) Fabrication, microstructure and mechanical properties of boron
carbide (B4Cp) reinforced aluminum metal matrix composite—a review. In: IOP conference
series: materials science and engineering, vol 377, no 1. IOP Publishing, p 012092
2. Kalaiselvan K, Murugan N, Parameswaran S (2011) Production and characterisation of
AA6061–B4C stir cast composite. Mater Des 32(7):4004–4009
3. Bobbili R, Madhu V, Gogia AK (2015) Modelling and analysis of material removal rate and
surface roughness in wire-cut EDM of armor materials. Eng Sci Technol Int J 18(4):664–668
4. Kumar A, Sayed AIH, Rai RN (2019) Optimization by AHP-ARAS of EDM process param-
eters on machining AA7050–10%B4C composite. In: Advances in industrial and production
engineering. Springer, Singapore, pp 285–296
5. Velmurugan C, Senthilkumar V, Dinesh S, Kirubakaran DA (2018) Artificial neural network
prediction of wire electrical discharge machining properties on sintered porous NiTi shape
memory alloy. Mater Today Proc 5:8382–8390
6. Choudhuri B, Sen R, Ghosh SK, Saha SC (2018) Study of surface integrity and recast surface
machined by wire electrical discharge machining. Mater Today: Proc 5:7515–7524
7. Jain PS, Ravindra HV, Ugrasen G, Prakash N, Rammohan YS (2017) Study of surface roughness
and AE signals while machining titanium grade-2 material using ANN in WEDM. Mater Today
Proc 4:9557–9560
8. Nain SS, Sihag P, Luthra S (2018) Performance evaluation of fuzzy-logic and BP-ANN methods
for WEDM of aeronautics super alloy. Methods X 5:890–908
9. Devarasiddappa D, George J, Chandrasekaran M, Teyi N (2016) Application of artificial intel-
ligence approach in modeling surface quality of aerospace alloys in WEDM process. Proc
Technol 25:1199–1208
10. Das S, Vaiphei SL, Chandrasekaran M, Samanta S (2018) Wire cut EDM of Al6061 hybrid
nano composites: experimental investigations and RSM modeling of surface roughness. Mater
Today Proc 5:8206–8215
11. Garga MP, Sharma A (2017) Examination of accuracy aspect in machining of ZrSiO4p /6063
aluminium MMC using CNC wire electrical discharge machining. Compos Commun 6:1–6
12. Kumar A, Rai RN (2019) Evaluation of wear properties of stir cast AA7050–10% B4C ex situ
composite through fuzzy-TOPSIS MCDM method. Solid State Phenom 291:1–12
13. Taguchi G (1990) Introduction to quality engineering. Asian Productivity Organization (APO),
Tokyo, Japan
14. Mandal N, Doloi B, Mondal B, Das R (2011) Optimization of flank wear using Zirconia Tough-
ened Alumina (ZTA) cutting tool: Taguchi method and Regression analysis. Measurement
44(10):2149–2155
1554 A. Kumar et al.
15. Kumar A, Rai RN (2019) Optimisation of EDM process parameters for AA7050-10 (WT)
% B4C composite through ARAS, grey and Taguchi technique. Int J Mater Product Technol
59(2):102–120
16. Singh AK, Dey V, Rai RN (2017) Study on the effect of high-temperature ceramic fibre
insulating board to weld grade P-91 steel. J Manufact Process 25:1–7
Biomechanical Response of Seated
Human Body Subjected to Vertical
Vibrations Using Coupled Matrix Model
1 Introduction
Occupational vehicle drivers often suffer from back disorders for sufferings [1].
Study of drivers exposed to whole-body vibrations requires biodynamic models of
a suitable level of complexity and associated constraints. The vibrations transmitted
from road tyre interaction to the human body depend on many factors such as road
profile, vehicle and seat suspension cushion properties. These mathematical models
will help to understand the biodynamic response of the human body in an extensive
way, consequently, contribute towards better seating and vehicle dynamics. Human
body models developed in literature classified as lumped parameter models [2],
finite element models [3], multibody models [4], and matrix models [5]. Most of
the developed models are one dimensional and constraint due to backrest support is
usually ignored.
Seated human body subjected to vertical direction vibrations at the base, head
exhibit motion in vertical and horizontal directions. These vertical and horizontal
motions are called as direct vertical (DV) and cross-axis fore-aft (CAFA). Coupled
matrix model is developed to represent the human body seated with backrest allowing
two-dimensional motion. The model parameters are optimized using a genetic algo-
rithm and subsequently road optimization technique to align the model simulated
results with the experimental findings. The model is then subsequently integrated with
quarter car vehicle model to obtain optimum suspension parameters corresponding
to random road profile.
2 Biomechanical Modelling
Above Eq. (5) gives rise to eight equations of motion representing the motion
of masses in x–z directions. The equations for masses 1–4 can be written in the
formulated as [M]ü + [C]u̇ + [K ]u = F
Here,
[M] Mass matrix (size 8 × 8)
[K] Spring stiffness matrix (size 8 × 8)
[C] Damping coefficient matrix (size 8 × 8)
[F] Input excitation force vector (size 8 × 1).
u, u̇ and ü represent the output displacement, velocity, and acceleration vectors
of each masses (size 8 × 1). The equations of motion are transfigured from the time
domain to frequency domain using Laplace transform in order to represent frequency
domain measured data.
The seat-to-head transmissibility (STHT) transfer functions, namely direct
vertical (DV), cross-axis fore-Aft (CAFA), are defined as follows:
z4 ż 4 z̈ 4
DV STHT = or or (6)
z0 ż 0 z̈ 0
x4 ẋ4 ẍ4
CAFA STHT = or or (7)
z0 ż 0 z̈ 0
3 Parameters Optimization
N
J1 = (|DV STHT M z ( f i )| − |DV STHT E z ( f i )|)2
i=1
(8)
N
J2 = (|CFA STHT M z ( f i )| − |CFA STHT E z ( f i )|) 2
i=1
2 ∗
Ji − Ji 2
JModel = (9)
i=1
Ji∗
The mass and inertia values for different human body parts are obtained from the
literature [8–10] and presented in Table 1.
The methodology for parameter optimization is shown in Fig. 2
The direct vertical and cross-axis fore-aft seat-to-head transmissibility for model-
simulated and experimental results are shown in Fig. 3. It can be observed that model
results are in line with experimental findings.
The stiffness and damping coefficient values obtained after optimization are
presented in Table 2.
Fig. 3 Model versus experimental seat-to-head transmissibility a Direct vertical. b Cross axis
fore-aft
The developed 8-DoF is integrated with a vehicle model to obtain seat suspension
parameters. The human–vehicle model integration is presented in Fig. 4.
The suspension performance is evaluated using seat effective amplitude transmis-
sibility (SEAT) factor. It’s the ratio of weighted RMS acceleration of output to the
input.
z̈ f,w RMS
SEAT = (10)
z̈ s,w RMS
f = SEAT
To minimize:
Subject to: z s − z f max − z s − z f min ≤ z sf,allow .
The maximum allowable seat travel z rs,allow is taken as 45 mm [11]. Frequency
weighting function is used to express differences in discomfort at different vibration
frequency [12]. The vehicle is subjected to random road profile moving at a speed
of 60 km/h using random noise signal depicted in Fig. 5.
The vehicle model [13] consists of unsprung mass (m us ), sprung mass (m s ), seat
frame mass (m s ), over which developed driver model is mounted.
m us z̈ us + kt (z us − zr ) + ct (ż us − żr )
+ ks (z us − z s ) + cs (ż us − ż s ) = 0 (11)
m s z̈ s + ks (z s − z cu ) + cs (ż s − ż cu )
+ kse z s − z f + cse ż s − ż f = 0 (12)
m f z̈ f + kcu z cu − z f + ccu ż cu − ż f
+ k P4 z cu + c P4 ż cu + k B4 z cu + c B4 ż cu = 0 (13)
m cu z̈ cu + kcu z cu − z f + ccu ż cu − ż f
+ c p4 + cb4 ż cu + k p4 + kb4 z cu − c p4 ż 1 − k p4 z 1
− c p3 ẋ1 − k p3 x1 − cb3 ẋ2 − kb3 x2 − cb4 ż 2 − kb4 z 2 = 0 (14)
5 Conclusion
A detailed coupled human body model is developed and optimized using a genetic
algorithm to represent direct and cross-axis motion of the human head, subjected
to vertical axis seat excitations. Equations of motion are transferred from time
to frequency domain to obtain model direct vertical and cross-axis head motion
frequency response. Model parameters were optimized using a genetic algorithm to
represent experimental direct and cross-axis motion of the human head of the seated
human body subjected to vertical axis seat disturbances. The model is integrated with
a vehicle model to obtain seat suspension parameters to improve drivers’ comfort
level.
References
1. Lings S, Leboeuf-Yde C (1998) Whole body vibrations and low back pain. Ugeskr Laeger
160(29):4298–4301
2. Wan Y, Schimmels JM (1995) A simple model that captures the essential dynamics of a seated
human exposed to whole body vibration. Adv Bioeng 31:333–334. https://epublications.mar
quette.edu/mechengin_fac/84/
3. Zhang X, Qiu Y, Griffin MJ (2015) Developing a simplified finite element model of a car seat
with occupant for predicting vibration transmissibility in the vertical direction. Ergonomics
58(7):1220–1231
4. Zheng G, Qiu Y, Griffin MJ (2011) An analytic model of the in-line and cross-axis apparent
mass of the seated human body exposed to vertical vibration with and without a backrest. J
Sound Vib 330(26):6509–6525
5. Marzbanrad J, Afkar A (2013) A biomechanical model as a seated human body for calculation of
vertical vibration transmissibility using a genetic algorithm. J Mech Med Biol 13(04):1350053
6. Wang W, Rakheja S, Boileau P-É (2006) Effect of back support condition on seat to head
transmissibilities of seated occupants under vertical vibration. J Low Freq Noise Vib Act
Control 25(4):239–259
7. Rao SS (2009) Engineering optimization: theory and practice. Wiley
8. Gagorowski A (2010) Simulation study on stiffness of suspension seat in the aspect of the
vibration assessment affecting a vehicle driver. Logist Transp 11:55–62
9. Winter DA (2009) Biomechanics and motor control of human movement. Wiley
10. Chandler RF, Clauser CE, McConville JT, Reynolds HM, Young JW (1975) Investigation of
inertial properties of the human body. Air Force Aero Med Res Lab Wright-Patterson AFB
OH. https://apps.dtic.mil/sti/citations/ADA016485
11. Wang YH, Shih MC (2011) Design of a genetic-algorithm-based self-tuning sliding fuzzy
controller for an active suspension system. Proc Inst Mech Eng Part I J Syst Control Eng
225(3):367–383
12. Wu X (1998) Study of driver-seat interactions and enhancement of vehicular ride vibration
environment. Concordia University
13. Du H, Li W, Zhang N (2012) Integrated seat and suspension control for a quarter car with driver
model. IEEE Trans Veh Technol 61(9):3893–3908
Categorization of the Indian Males’ Foot
Data for Age 18–25 Years Based
on Plantar Footprints
Abstract This study aims out the purpose and outcomes of research performed to
evolve a reasoned categorization method of the elderly foot assortment, and it is
rooted on framework derived from the plantar footprint. DIERS Pedoscan pressure
plate and the corresponding system including DIERS Statico-3D systems were used
to obtain plantar footprints from 30 men, aged between 18 and 25 years. Hallux-
Valgus (HV) Angle and Chippaux-Simark Index (CSI) have been utilized, and foot
assortment has been categorized. Mainly five disciplines were separated into groups,
disciplines with Hallux-Varus Foot, Normal Foot, HV Foot, High Arched Foot,
Normal foot and Flat Foot, grounded on the acquired result. For a more precise
calculation, being able to record it in the identical method, for all disciplines, the
authors have evolved a process to assess the mid-plantar footprint middle area for
a more precise calculation. Distinctions between different age batches have been
controlled by having defined age group and revealing the unavoidable modelling
with crafting shoe shapes, protective constituents and footwear accordingly. The
superior apt of the shoe on foot in young age can be attained under definite levels and
requires reforming of shoe shape. The customization of the product is rationalized by
incorporating having biomechanics-based foot data and distinguishes categorization
of this, decreased tests numbers and expanding a footwear technological operation
effectiveness.
1 Introduction
Beneath definite criterions, a decent apt of the shoe on the foot may well be earned by
fixing the last. The customization of the merchandise gets streamlined by having the
biomechanical data of the foot and reduces the quantity of tests and raises the footwear
technological procedure effectiveness. The acquaintance of each foot measurements
and features is essential for generating modified footwear [1–6]. The identification
of comparable features and their sorting by clusters of topics have a high influence
on the shoe last’s design [7–10], soles and insoles design as well as in the treat-
ment and hindrance of foot disorders [11–18]. Foot type is a common term generally
used, to label several architectural features of the foot in demand to acquire clues
of its vibrant functioning [19]. Comprising peculiar injury from running and plantar
ache, particularly on the heel [20], foot dynamics is observed as being related to a
diversity of musculoskeletal indications. With or without pathologies under exami-
nation [21], the biomechanical studies have advanced by comparing the existence,
lack or dimension of the foot typological constraints between groups [22, 23]. It
was documented that despite the extensive use of these constraints, the unbiased and
computable study of foot typologies rests subtly. It has led to a significant distinction
in selecting the dimensions to decide the foot typology [24, 25] due to the lack of an
utter measurement of the foot type. This disagrees with the proposal that to classify
the relations between foot typology and pathology, and it is essential to practice an
exclusive classification system to permit precise recognition in each condition [26].
2 Method
Disciplines remained shoeless the minute their foot pressures occupied. Disciplines
stood by both feet equally arranged on the compression plate in a balance loca-
tion meant for arithmetical dimensions. Coating a dimension aimed at the right
foot then one intended for the left foot meant for dynamic dimensions, disciplines
were requested to walk ordinarily on the plate. Meant for separate disciplines to be
examined, the whole technique was recurrent three times. A dynamic footmark is
attained distinct to old-style procedures of enchanting plantar footprints. The soft-
ware mechanically relates pictures in entire periods of walking utilizing this arrange-
ment. Five sizes were studied [27, 28] to govern the chief constraints of the foot-
mark; also, pictures were transmitted to scale 1:1. The foot arrangement could stand
measured employing this simple technique of dynamic plantar footmark inspection.
This investigation is quick and non-invasive as well.
Categorization of the Indian Males’ Foot Data for Age … 1567
DIERS Pedoscan pressure plate [29] and the linked systems together with DIERS
Statico-3D system were used to acquire plantar footmarks.
2.2 Subjects
The plantar footmarks were occupied from 30 men, 18–25 years old and weighing
45–70 kg. A well-informed and written consent was signed from all the disciplines
before testing. All the subjects were checked to ensure to have no comorbidity and
pre existing laxity, which might affect the result of the study.
3 Result
The plantar footmarks have manually been processed by constructing the reference
lines [30]. Point Q (extreme interior point of the heel) will join point M (the intersec-
tion of toe line with longitudinal axis) to obtain the MP segment. Point R (extreme
point of the metatarsal phalangeal joint) will join Q (extreme exterior point of the
heel) to obtain the RQ segment. The segment VW is drawn parallel to RS, where MP
and RQ intersect (Fig. 1).
Following the analysis if Chippaux-Simark index, the result shows that 4% of the
subjects have high arched foot in both left and right foot, 55% have flatness in right
foot and 64% in left foot. The remaining subjects have lower arched foot (23% in
left foot and 18% in right foot); normal foot (5% in left foot and 9% in right foot),
intermediate arched (4% in left foot and 14% in right foot), high arched (4% in right
foot as well as left foot) as shown in Fig. 2 and in Table 1.
1568 P. Mishra et al.
Fig. 1 Footprint for plantar data. UT foot length, PQ heel width, VW distance between the
intersection of QR with MP, RS toe width, β angle of I toe deviation, Hallux-Valgus Angle (HVA)
The Hallux-Valgus angle (HVA) measures the existence over foot of both Hallux-
Valgus, the misalignment of first toe as well as Hallux-Varus which may be defined
as misalignment of first toe on exterior side. The obtained data of angle β the outcome
exhibits 5% participants have Hallux-Valgus condition, 5% in left foot and 9% in
right foot have the potential risk to occur it (Table 2).
4 Conclusion
Fig. 3 Foot types based on HVA index for the entire group, in %
Ethical Approval “All procedures performed in studies involving human participants were in
accordance with the ethical standards of the institutional research committee and with the 1964
Helsinki declaration and its later amendments or comparable ethical standards”.
Informed Consent All the subjects participated gave a well-informed and written consent prior
to the study. They have been explained before the study trials, procedures and the outcome as a
research article to be published.
References
1. Asanka SR, Goonetilleke RS, Witana CP (2012) Model-based foot shape classification using
2D foot outlines. Comput Aided Des 44:48–55. Available at www.elsevier.com/locate/cad
2. Wunderlich RE, Cavanagh PR (1999) External foot shape differences between males and
females and among races. In: Abstract book, 23rd Annual meeting of the American Society of
Biomechanics. ASB 1999, pp 68–69
3. Herghiligiu LV, Mihai A, Sarghie B, Souto Bizarro R, Arias C (2016) Framework of the e-
learning training program on corporate social responsibility. In: The 12th international scientific
conference e-learning and software for education. Bucharest, p 526. https://doi.org/10.12753/
2066-026X-16-255
4. Sarghie B, Mihai A, Herghiligiu IV (2016) E-learning application for 3D modelling of custom
shoe lasts using templates. In: The 12th international scientific conference elearning and
software for education. Bucharest, p 553. https://doi.org/10.12753/2066-026X-16-260
5. Ionesi D, Ciobanu L, Sarghie B (2014) E-learning application for a better understanding of
shoes 3D modeling. In: 10th international scientific conference ELSE, p 196. https://doi.org/
10.12753/2066026X14285. ISSN 2360-2198
6. Deselnicu DC, Vasilescu AM, Mihai A, Purcarca AA, Militaru G (2016) New products devel-
opment through customized design based on customers’ needs. Part 2: foot pathology manu-
facturing parameters. In: The 9th international conference interdisciplinarity in engineering,
INTER-ENG 2015. Procedia technology, vol 22, pp 1059–1065. Available at www.sciencedi
rcct.com
7. Kouchi M (1995) An analysis of foot shape variation based on the medial axis of foot outline.
Ergonomics 38:1911–1920
8. Mochimanu M, Kouchi M, Dohi M (2000) Analysis of 3D human foot forms using the free form
deformation method and its application in grading shoe lasts. Ergonomics 43(9):1301–1313
Categorization of the Indian Males’ Foot Data for Age … 1573
9. Wunderlich RE, Cavanagh PR (2001) Gender differences in adult foot shape: implications for
shoe design. Med Sci Sports Exerc 33(4):605–611
10. Aştina MP, Milhai A, Bilalis N (2012) Finite element analysis for insole-sole prototypes. In:
Proceedings of “the 4th international conference on advanced materials and systems”. ICAMS
2012. Bucuresti, pp 359–364. ISSN 2068-0783
11. Hwang TJ, Le K, Oh HY, Jeong JH (2005) Derivation of template shoe-lasts for efficient
fabrication of custom-ordered shoe-lasts. Comput Aided Des 37(12):1241–1250
12. Hawes MR, Sovak D, Miyashita M, Kang SJ, Yoshihuku Y, Tanaka S (1994) Ethnic differences
in fore foot shape and the determination of shoe comfort. Ergon Omics 37(1):187–196
13. Nigg BM, Nurse MA, Stefanyshyn DJ (1999) Shoe inserts and orthotics for sport and physical
activities. Med Sci Sports Exerc 31(7):S421–S428
14. Goonetilleke RS, Luximon A, Tsui KL (2000) The quality of footwear fit: what we know don’t
know and should know human factors and ergonomics society conference. San Diego, CA, pp
515–518
15. Dohi M, Mochimaru M, Kouchi M (2001) Foot shape and shoe fitting comfort for elderly
Japanese women. Jpn J Ergon 37(5):228–237
16. Goonetilleke RS, Luximon A (2001) Designing for comfort: a footwear application. In:
Proceedings of computer-aided ergonomics and safety conference. Maui (Hawaii, Plenary
Session)
17. Witana CP, Goonetilleke RS, Feng J (2004) Dimensional differences for evaluating the quality
of footwear fit. Ergonomics 47(12):1301–1317
18. Rathleff MS, Nielsen RG, Simonsen O, Olesen CG, Kersting UG (2010) Perspectives for
clinical measures of dynamic fool function reference data and methodological considerations.
Gait Posture 31(2):191–196
19. Lee H, Grosse R, Ranganath R, Ng AY (2009) Convolutional deep belief networks for scalable
unsupervised learning of hierarchical representation. In: International conference on machine
learning, pp 77–85
20. Sivapalan S, Chen D, Denman S, Sridharan S, Fookes C (2011) Gait energy volumes and frontal
gait recognition using depth images. In: IEEE international joint conference on biometrics, pp
501–506
21. Zheng S, Huang K, Tan T (2011) Evaluation framework on translation-invariant representa-
tion for cumulative foot pressure image. The 18th IEEE international conference on image
processing (ICIP), pp 201–204
22. Huang ZK, Tan T, Tao D (2012) A cascade fusion scheme for gait and cumulative foot pressure
image recognition. Pattern Recogn. 45:3603–3610. Available at www.elsevier.com/locate/pr
23. Felzenszwalb PF, Girshick RB, McAllester D, Ramanan D (2010) Object detection with
discriminatively trained part-based models. IEEE Trans Pattern Anal Mach Intell 32:1627–1645
24. Bouchrika I, Nixon M (2007) Model-based feature extraction for gait analysis and recognition.
In: International conference on computer vision/computer graphics collaboration techniques,
pp 150–160
25. Zhang T, Li X, Tao D, Yang J (2008) Multimodal biometrics using geometry preserving
projections. Pattern Recogn 41:805–813
26. Mathieson I, Upton D, Birchenough A (1999) Comparation of footprint parameters calculated
from static and dynamic footprints. Foot 9:145–149
27. Shan C, Gong S, McOwan PW (2008) Fusing gait and face cues for human gender recognition.
Neurocomputing 71:1931–1938
28. Costea M, Vasilescu AM, Hortal G, Mihai A (2014) Plantar footprints analysis case study (part
2). Leather Footwear J 14(4):243–250 (2014). Certex Publishing House. ISSN 15834433
29. https://www.rsscan.co.uk/. Accessed at 03 Oct 2016
30. Mihai A, Pastina M (2012) Classification of foot types, based on plantar footprint. In: Proceed-
ings of the 4th international conference on advanced materials and systems. ICAMS 2012,
Bucuresti. 2012, pp 347–352. ISSN 2068-0783
Mechanization of Peppermint Oil
Extraction Plant of Rural India
Abstract Peppermint leaves have wide medical applications. However, the tradi-
tional methods of peppermint oil extraction require high labour and large number
of workers. It is least profitable, and post-harvesting method adopted is not safe
for workers and farmers. Several surveys are conducted in various districts of Uttar
Pradesh States(India) like Bahraich, Barabanki, Ramnagar, Sitapur and Lakhimpur,
and the existing problems are compiled in this paper. The complaints of the workers
and complications that occur post-harvesting of peppermint leaves to oil extraction
are considered. The major problem are the long time taken to fill the drum (boiler)
(with the peppermint leaves), the uneasy method of handling the drum, compres-
sion of leaves and the hazardous removal of burnt leaves. The paper contains the
detailed list of problems and its classification based on the findings during the survey.
The old and traditional method is being analysed, and economical suggestions, for
farmers of rural India, are given through various proposed models. With the proposed
mechanized model, the efficiency of plant can be improved effectively.
1 Introduction
Peppermint essential oils are extracted from Mentha piperita. The main ingredients of
peppermint oil are menthol (45% approx.) and menthone (22% approx.). Peppermint
oil is commonly used as a fragrance in soaps, cosmetics and as spice since it possesses
a fresh, minty and cooling effect due to the presence of menthol [1]. Due to this
flavouring property, peppermint is often found in chewing gums, toothpastes, etc.
Farmers prefer to cultivate peppermint plant because its cultivation gives a high
profit in least time of cultivation. The crop takes 3.5 months of time and is ready
by May for oil extraction. There is no major labour for cultivation except proper
irrigation. The oil when sold gives considerable amount. Its one of the product is
synthetic menthol. It is sold at very high rates by retailers.
At present, Japanese mint is cultivated in India on about 60,000 ha of land with
estimated production of 12,000 tonnes of mint oil which accounts for about 75% of
total menthol mint production in the world. It is grown as an annual in subtropical
parts of north India. It yields an essential oil, containing high (75–80%) menthol
content. Table 1 gives estimated area and production of various mint species in India
vis-à-vis world production [2].
The traditional methods, adopted for peppermint oil extraction, are complex, least
profitable and not safe for workers and farmers. Several visits/surveys were conducted
in various districts of Uttar Pradesh like Bahraich, Barabanki, Ramnagar, Sitapur and
Lakhimpur. The workers have complained about the complications that occur during
peppermint oil extraction in Sect. 1.1. The major problem are the long time taken to
fill the drum (boiler) (with the peppermint leaves), the uneasy method of handling
the drum, compression of leaves and the hazardous removal of burnt leaves.
In most part of Uttar Pradesh, the cultivation of Peppermint leaves is carried on.
The plant is cut, leaves are collected and boiled in a drum, with water, to extract the
peppermint oil via condensation. After the survey, the details of the plant are found
to be as follows:
The plant has a drum (boiler) of 182 cm height 152 cm (available data, obtained
from survey at various sites), a condenser drum (open) of diameter 182 cm and
height 35 cm. Usually, there is no electricity in the field. The furnace and exhaust are
added to the plant. The working area for entire plant is 5.5 × 5.5 m2 . Raw material
preparation time is 2 h. The boiler runs for 4.5 h. Usually, four labour are there. 02
tanks can only be used in a day. The earning from the plant is Rs. 2000/day. Figures 1,
2, 3, 4 and 5 depict the traditionally existing plant and boilers/ vessels along with
Mechanization of Peppermint Oil Extraction Plant of Rural India 1577
iv. Rod is used to push dry leaves/ bagasse for burning at the underneath of the
drum.
v. The steam is collected after it is passed through condenser drum.
vi. The required oil separates out due to difference of density.
Mechanization of Peppermint Oil Extraction Plant of Rural India 1579
Fig. 5 Oil collection area (at the exterior) connected with condenser unit
The peppermint plant in rural India is in very pathetic conditions. Following is the
list of problems generally found in method adopted for the extraction of peppermint
oil.
inside the drum removes the boiled leaves. There are chances of getting the hot steam
on face and falling from the edge of the drum.
Technical Problems
Too much time consume to compress the dry peppermint leaves in drum. A good
mechanism can save human effort and minimize time of preparing stacks of pepper-
mint plants or leaves. Several persons put effort to cover the drum with its lid, which
hardly fits accurately at first attempt. A mechanism can make the work more accu-
rately and save time. There is no continuous flow of cold water for condensation in
condenser unit.
Ergonomic Problems
Workers have complained about the pain in wrist and in effective use of hooked rod
for collecting leaves. The person who puts fuel in furnace has to continuously bend
his head to see the status of the flames and uniform heating of the base of the boiler
(Fig. 7).
Management Problems
One drum takes 5 h to boil for extraction. Large time is spent in compressing the
leaves, filling the drum and removal of leaves. Preheat system with extra drum will
save time. All such steps need to be managed in order to optimize efficiency.
Economical Problem
One of the major problems is that the plant owner is unable to earn because of slow
process of leave preparation, stack preparation, lid closure time, slow burned leave
Fig. 7 A child with a metal stick, putting fuel (dry leaves or straw) to heat the boiler
Mechanization of Peppermint Oil Extraction Plant of Rural India 1581
removal and time taking process of preparing dry leaves. Hence, modern technologies
will save the time, which will increase overall efficiency of the plant.
Environment Problems
Usually, a furnace is made of mud at lower level to floor. During rain, water gets
collected. Removal of collected water is a big problem. It stops functioning of the
plant.
2 Literature Survey
Braimah, S. R., Andoh, P. Y., have designed a mechanical system to extract and sepa-
rate lemon grass oil, used in medicine, perfumery industry, vitamin A manufacturing
and pharmaceuticals. This paper has directed to design a mechanical system that will
be used to extract and separate lemon grass oil. Design analysis was performed for
the fabrication. Tests were performed to find machine performance and quality of
the oil produced. It was concluded that the prototype machine developed can be used
to extract lemon grass oil from the leaves [3].
Several techniques have been presented and analysed by various authors [4].
Hesham H. A. Rassem et al. have worked with a set-up comprises of condenser
and a decanter to collect the condensate and to separate essential oils from water,
respectively. The principle of extraction is based on the isotropic distillation [5].
Other authors have also worked on lemon grass [6, 7]. However, they are not found
to be economical for Indian farmers. The basic need for Indian subcontinent is to
design a model that would work without electricity as most of the rural areas do not
have power supply.
K. Satish Kumar in 2010 has laid focus on adoption of methods and mechanisms
for easy extraction of essential oils. He has suggested that method of extraction
depends on the extraction method employed. Distillation-based recovery processes
are preferred for the extraction of essential oils from plant materials [8].
Virendra P. S. in 2006 has mentioned that the way in which oils are extracted
from plants is important because some processes use solvents that can destroy the
therapeutic properties. There are wide number of ways to extract the Essential oil
1582 M. Anas and Abusad
but the quality never remains the same. “Steam Distillation” method for extraction is
the cheapest way for the extraction of Oils from the different parts of the plants [9].
Some authors have worked on modern methods of extraction. Seid Yimer has
mentioned in his research work, in which eucalyptus leaves are extracted for essen-
tial oil by steam distillation process. Such process was complex [10]. Alina Kunika in
2014 has worked on ultrasound-assisted machine for essential oil from waste carrot
seeds (Daucus carota) was investigated. The Taguchi method was applied to deter-
mine the levels of input factors. Essential oil yield increased by approx. 33% [11].
Gomez has used microwave protocol to extract lipophilic substances from tissue was
modified to extract essential oils (EOs) from plant tissue and insect feculae. For dried
material, mechanical means are utilized first to break up tissue. Statistical analysis
shows that the reproducibility of the modified method is high. Several samples can
be run within an hour with this method. However, such methods are very expensive
and require technical skills [12].
No definite research is available about peppermint plant industries of rural India.
The mechanized models so far made are not accepted by our farmers because of
complex mechanism of handling, high installation cost and least profit. As a result,
we see the models have not replaced the old traditional methods of extraction of
peppermint oil. The present mechanized model has very less installation cost and
can be easily handled, and besides that double the existing efficiency will be obtained
when implemented. The mechanization of plant will not only make the process faster
but also will be very safe to operate.
As mentioned above, the survey of peppermint oil extraction plants was done in
different villages and towns of Uttar Pradesh (India). People working in those plants
were quarried, and the existing traditional system was analysed. It was found that the
plants were run without electricity. Models are also proposed keeping in view the rural
areas not having power supply. Three-dimensional modern models are proposed to
equip the traditional system with high profit modern technical systems. The different
problems are listed in Sect. 1.2. The mechanized models proposed are given below:
a. Non-electric Mechanized in Line System (Two-Dimensional)
In this system the boiler, stacks of leaves and drier units will be in line. The crane
operated by chain and pulley arrangement will perform necessary task. Since all units
will be in a line, lot of extra movement is avoided in this type of mechanized model.
It will be preferred by farmers of rural area because no power supply is required in
this model.
b. Non-electric Flexible System
Mechanization of Peppermint Oil Extraction Plant of Rural India 1583
The structure will be based on assembly of hollow pipes which will be fastened or
screwed by bolts during the season of peppermint harvesting. When the work is over,
the structure can be disassembled. The space can be used for cultivation of crops.
c. Non-electric Mechanized System (Three-Dimensional)
Here the drum boiler will be able to move in three directions. It also has a mechanism
of compressing the leaves and fastening it in wires. Two chain pulley arrangements
are used to carry on entire process. It works in all dimensions. The movement is not
limited. Large space helps to dry used peppermint leaves. Installation cost and profit
is high. The model does not require power supply.
d. Electrically Mechanized System (Three-Dimensional)
It will run on heavy motors. The different motions will be obtained through belts
or chains run by heavy motors. Skilled labour will be required for this. This highly
mechanized system requires skilled and technical labours. The installation cost is
considerably high. It will require large space. It is easy to work and safer to operate.
e. Fully Automated Mechanism
This system will have conveyor belts, sophisticated system of leave compression and
its removal. Modern techniques such as microwave-assisted methods can be used for
oil extraction. There will be least wastage of oil. Technicians are preferred for the
operation. The installation cost is very high. It is very effective mode of extraction.
Highly efficient and sophisticated equipments will be used.
4 Conclusions
Various mechanized models are proposed and presented in Sect. 3. These mechanized
models are helpful to overcome the uneasy, hazardous, non profitable, complex and
highly laborious methods of our old traditions. The proposed mechanized models
are easy to operate and work with safety. The major conclusions of various models
are as follows:
i. In non-electric mechanized in line system all units are in a line. Thus, lot of
extra movement is minimized. However, there is a limit of using extra boiler.
Hence, profit is limited. No power supply is required.
ii. In non-electric flexible system is very helpful to dismantle the structure and
keep the entire system in a store so that the space can be used for cultivation.
iii. The non-electric mechanized 3D system works in all dimensions. The move-
ment is not limited. Large space is required. Installation cost is high. It is highly
profitable. No power supply is required.
iv. The electrically mechanized system (three-dimensional) is highly mechanized
system. It requires skilled labours. The installation cost is very high. Power
1584 M. Anas and Abusad
supply is must. It requires large space. It is easy and bears safety to the workers.
Use of multiple boilers saves lot of time.
v. The fully automated mechanism is highly profitable. There is least wastage of
oil. It is highly mechanized system. It requires skilled labours. The installation
cost is very high. Power supply is must. It requires large space. Highly efficient
equipments are used. It bears safety to the workers.
References
1 Introduction
throughout the structures and significantly influence the structural response of the
skewed plates as well.
Many researchers dealt with static behavior and a few with the dynamic behavior of
functionally graded imperfect plates. In the sequence, Yang and Huang [3] performed
a nonlinear transient analysis of initially imperfect FGM plates in a thermal environ-
ment. Shariyat [4] explored the vibration and dynamic buckling behavior of FGM
plates that are subjected to thermo-electro-mechanical loadings. Gupta and Talha [5]
performed nonlinear flexural and vibration analyses of imperfect FG plates. Gupta
and Talha [6] investigated the static and stability response of geometrically imperfect
FGM flat plate resting on the elastic foundation while considering porosity. Tomar and
Talha [7] analyzed the flexural and free vibration responses of FG skewed sandwich
plates with initial geometric imperfection. Shakir and Talha [8] studied the dynamic
behavior of FG flat plates with the combined effect of geometric imperfection and
material uncertainty.
In the present study, the dynamic behavior of skewed FG flat plates with initial
geometric imperfection has been considered. Reddy’s HSDT is used as the plate kine-
matics comprising geometric imperfection. Numerical results for different studies
have been obtained using the finite element method along with Newmark’s integration
technique.
2 Theoretical Formulation
Consider a skewed FG flat plate (as shown in Fig. 1) with length, width, uniform
thickness, and skew angle as a, b, h and θ, respectively. Since FG plates are the
special case of composite plates in which the mechanical properties (i.e., Young’s
modulus and density) are assumed to be graded along the thickness direction [1].
In this study, the gradation is accomplished between two surfaces of the plate in
terms of volume fraction of the constituents using a simple power-law distribution
function. In this way, the ceramic in rich at the top surface of the plate and metal in
rich at the bottom surface can be achieved. Poisson’s ratio is to be considered as 0.3.
The material properties are varied smoothly in the thickness direction (z-direction)
as [1],
Young’s modulus,
Density,
where “c” and “m” used for ceramic and metal, respectively. V c is volume fraction
of the ceramic and given as,
Imperfection Sensitivity of Skewed FG Flat Plates Under … 1587
n
2z + h
Vc = (3)
2h
u(x, y, z, t) = u 0 + f 1 (z)ψx + f 2 φx
v(x, y, z, t) = v0 + f 1 (z)ψ y + f 2 φ y
w(x, y, z, t) = w0 + w (4)
where f 1 (z) = z − c1 z 3 ; f 2 (z) = −c1 z 3 ; c1 = 4/3h 2 and w denotes initial
geometric imperfection given by [2],
where κ and h are the imperfection amplitude and thickness of the skewed flat plate,
respectively; δ 1 and δ 2 : Degrees of localization; μ1 and μ2 : Half-wave numbers.
The continuum displacements are written as follows:
T
{∀} = u 0 , v0 , w0 , ψx , ψ y , θx , θ y (6)
The mathematical formulation is carried out using the finite element method to
obtain strain and kinetic energies. In the sequence, a C 0 continuous Lagrange’s
isoparametric element with nine nodes has been adopted to discretize the plate
domain.
The total strain energy is therefore given by the expression as [8],
where k (i) and m (i) are elemental stiffness and mass matrices of the ith element,
respectively.
For the skew plates, the edges of the boundaries are not parallel to the coordinate
axis. Therefore, the transformation is required to make a relation between local and
global coordinate systems. The transformation can be expressed as [7],
∀ j = Tg ∀ j (9)
where ∀ j and ∀ j are the displacement vectors for the jth node in global and local
coordinate systems, respectively, and defined as,
T
∀ j = u 0 , v0 , w0 , ψx , ψ y , θx , θ y j
T (10)
∀ j = u 0 , v0 , w0 , ψx , ψ y , θx , θ y
j
Moreover, the nodes are also not lying on the edges; therefore, the final
transformation matrix is expressed as,
T f = Tg [I ] (11)
where [I ] is the identity matrix with a size equal to the number of nodes in the
element.
Imperfection Sensitivity of Skewed FG Flat Plates Under … 1589
And then, Eq. (12) is solved for the dynamic response using Newmark’s integration
technique.
The results and discussion comprise convergence and validation studies along with
the parametric study for skewed FG ZrO2 /Al flat plates. Here, the dynamic response
is presented in terms of the non-dimensional parameter which is given by [10],
Non-dimensional center deflection: w ∗ = wE mh
.
q0 a 2
Figure 2a illustrates the convergence study for ZrO2 /Al skew FG plate (side to thick-
ness ratio, a/h = 10; aspect ratio, b/a = 1; skew angle, θ = 30°) with initial geometric
imperfection (imperfection amplitude, κ = 0.5). The sine-type imperfection mode is
taken into account. All edges of the plate are considered as simply supported. For the
present dynamic model, the convergence criterion is met at 8 × 8 mesh size which,
therefore, is employed throughout the analysis.
Figure 2b presents the validation study for ZrO2 /Al FG perfect plate with different
power-law indices (n = 0.2, 0.5, and 1). The plate is considered with thickness ratio,
a/h = 20, and aspect ratio, b/a = 1. The validation is performed by comparing the
present results with the exiting results [10]. The solution accuracy of the present
Table 1 Material
Material Young’s modulus, E Density, ρ (kg/m3 )
specification of ceramic and
(GPa)
metal
Zirconia, ZrO2 151 3000
Aluminum, Al 70 2707
1590 M. Shakir and M. Talha
Fig. 1 Geometric configuration of skewed FG flat plate in x–y and y–z planes
Fig. 2 Convergence and validation studies of the dynamic response FG flat plates
model shows a very good agreement with published results which are based on the
first-order shear deformation theory (Fig. 2b).
Here, the behavior of skewed FG ZrO2 /Al flat plates for different parameters, like
volume fraction index, imperfection amplitude, and skew angle, has been examined.
The response is obtained for sine-type imperfection mode because it shows a signified
effect on the structural response [8].
Figure 3a shows the dynamic response of skewed FG flat plate with simply
supported boundary condition (SSSS), thickness ratio (a/h) = 10, aspect ratio (a/b)
= 1, skew angle (θ ) = 30°, and imperfection amplitude (κ) = 0.5 for different volume
Imperfection Sensitivity of Skewed FG Flat Plates Under … 1591
Fig. 3 Dynamic response of skewed FG flat plates for a different volume fraction indices,
n b different imperfection amplitudes, κ
fraction indices (n). The response increases with an increase in the value of “n”, and
the highest and lowest responses have been observed for metal (n → ∞) and ceramic
(n = 0), respectively, as expected.
Figure 3b explores the dynamic behavior of skewed functionally graded plate
[thickness ratio (a/h) = 10, aspect ratio (a/b) = 1, skew angle (θ ) = 30° and volume
fraction index (n) = 2] for different imperfection amplitudes (κ). The response
increases with a decrease in the value of “κ” and the highest and lowest responses have
been noted for perfect (κ = 0) and imperfect (κ = 1) plates, respectively, as expected.
Moreover, the dynamic behavior changes with increment in non-dimensional time,
and peaks of the response shift toward the left.
In Fig. 4a, the behavior of skewed simply supported functionally graded flat plate
is shown. The square (a/b = 1) plate is considered with a/h = 10, n = 2 and κ = 0.5.
The dynamic response has been explored for different skew angles (θ ). As shown in
Fig. 4a, the response increases when the skew angle (θ ) decreases.
In Fig. 4b, the time-dependent response of skewed functionally graded plate (a/h
= 10, a/b = 1, θ = 30° and n = 2) for different boundary conditions (i.e., SSSS,
CCCC, SSCC, CFCF, and SCSC) has been noted. As shown in Fig. 4b, the response
for CFCF highly affects the structures, while it is least for CCCC and SCSC.
4 Conclusion
Fig. 4 Dynamic response of skewed FG flat plates for a different skew angles, θ b different
boundary conditions
occur before the peaks as occurred in the case of the perfect plate; however, the
imperfect plates possess lesser dynamic response. Moreover, the skewed imperfect
plates are less sensitive, while it is highly sensitive for CFCF boundary condition.
References
1. Talha M, Singh BN (2010) Static response and free vibration analysis of FGM plates using
higher order shear deformation theory. Appl Math Model 34(12):3991–4011
2. Kitipornchai S, Yang J, Liew KM (2004) Semi-analytical solution for nonlinear vibration of
laminated FGM plates with geometric imperfections. Int J Solids Struct 41(9–10):2235–2257
3. Yang J, Huang XL (2007) Nonlinear transient response of functionally graded plates with
general imperfections in thermal environments. Comput Methods Appl Mech Eng 196(25–
28):2619–2630
4. Shariyat M (2009) Dynamic buckling of imperfect laminated plates with piezoelectric sensors
and actuators subjected to thermo-electro-mechanical loadings, considering the temperature-
dependency of the material properties. Compos Struct 88(2):228–239
5. Gupta A, Talha M (2017) Nonlinear flexural and vibration response of geometrically imperfect
gradient plates using hyperbolic higher-order shear and normal deformation theory. Compos
B Eng 123:241–261
6. Gupta A, Talha M (2018) Static and stability characteristics of geometrically imperfect FGM
plates resting on Pasternak elastic foundation with microstructural defect. Arab J Sci Eng
43(9):4931–4947
7. Tomar SS, Talha M (2019) On the flexural and vibration behavior of imperfection sensitive
higher order functionally graded material skew sandwich plates in thermal environment. Proc
Inst Mech Eng Part C J Mech Eng Sci 233(4):1271–1288
Imperfection Sensitivity of Skewed FG Flat Plates Under … 1593
8. Shakir M, Talha M (2019) On the dynamic response of imperfection sensitive higher order
functionally graded plates with random system parameters. Int J Appl Mech 11(03):1950025
9. Reddy JN (1984) A simple higher-order theory for laminated composite plates. J Appl Mech
51(4):745–752
10. Praveen GN, Reddy JN (1998) Nonlinear transient thermoelastic analysis of functionally graded
ceramic-metal plates. Int J Solids Struct 35(33):4457–4476
A Bond Graph Model for the Estimation
of Torque Requirements at the Knee
Joint During Sit-to-Stand
and Stand-to-Sit Motions
Abstract Sit-to-stand (SiTSt) and stand-to-sit (StTSi) are some of the frequent but
crucial motions executed by the human body during daily life activities. During
these motions, information of torque required at the knee joint is necessary for the
development of knee assistance devices. In this work, a bond graph model has been
developed to estimate torque requirements at the knee joint during SiTSt and StTSi
motions. The bond graph offers graphical representation of a system, causality and
power transactions between the subsystems, and algorithmic derivation of system
equations. Word Bond Graph Objects (WBGOs) have been used for their compact
representations of large systems. The model comprises two seven-link sub-models,
one is virtual and the other is actual. The virtual sub-model is considered to be a part
of the central nervous system (CNS). It generates the required joint angles for SiTSt
and StTSi motions for desired movement trajectories. Based on these joint angles
required, joint torques are provided to the actual sub-model through proportional–
derivative (PD) controllers. Variations in the torque requirements have been estimated
during SiTSt and StTSi motions, and the simulation results are presented.
1 Introduction
According to Lutz et al. [1], the median age of the world’s population is increasing
from 26.6 years in 2000 to 37.3 years in 2050 and further projected to 45.6 years in
2100. With the increase in age, human beings start to have physical and cognitive
issues, and when these become extreme, it would become difficult to do basic living
activities, and assistance is required. Therefore, there is a growing need for assistive
devices that can help old individuals in the execution of daily living tasks indepen-
dently [2]. Further, SiTSt and StTSi are some of the frequent but crucial motions
executed by the human body during daily life activities. For elderly people who are
affected by knee osteoarthritis, or any other musculoskeletal problem, it becomes
difficult and painful to execute basic SiTSt and StTSi motions. In this context, many
of knee assistance devices have been developed either for gait rehabilitation or for
power augmentation. Development of assistance devices to assist a particular motion
of the human body requires a dynamic model of that motion. This work focuses on
the modeling of SiTSt and StTSi motions for the development of a knee assistance
device for elderly people. Different models of SiTSt motion have been developed by
various researchers which are either based on the experimental acquisition of data by
optoelectronic systems, force plates, etc., or optimization of cost functions subjected
to some kinematic constraints or energy-based formulations [3]. In comparison with
the SiTSt motion, fewer studies have been carried out for StTSi motion. Only a few
computational models are available which have considered both SiTSt and StTSi
motions.
In this work, the bond graph technique has been used for the modeling of both
SiTSt and StTSi motions of the human body. The bond graph is a graphical technique
to model the physical systems in multi-energy domains. Pop Cristian et al. used the
bond graph approach to model human locomotion [4]. The information of the change
in the center of pressure and ground reaction forces was predicted using the model in
close accordance with experimental results. Ghafari et al. [5] prepared a bond graph
model for the human lower extremity musculoskeletal structure and analyzed the gait
pattern. Mishra and Vaz used bond graph for the modeling of a 3-joint string tube
actuated prosthesis. They used WBGOs for the modeling of different subsystems of
a prosthetic finger mechanism and analyzed their dynamic behavior [6].
In this work, a bond graph model has been developed for the SiTSt and StTSi
motions of the human body. It comprises two seven-rigid-link sub-models. One of
the sub-models is considered to be a part of the CNS and is referred to as a virtual
sub-model and the other works as the actual sub-model for SiTSt and StTSi motions
of human beings. The virtual sub-model generates the required joint angles for SiTSt
and StTSi motions for desired movement trajectories. Based on these joint angles,
the required joint torques are provided at joints of the actual sub-model through PD
controllers. In the subsequent section, the details about the physical system for the
SiTSt and StTSi motions and its bond graph model are discussed.
2 Physical System
Seven rigid links of the model include left and right feet, shanks, thighs and the
trunk. The geometry of all links other than the trunk is considered as a cylinder while
that of the trunk is considered as cuboid as shown in Fig. 1. The total height of the
body is considered equal to 180 cm. The total mass of the body is considered equal
to 80 kg. The length and the mass of the individual links are taken as a fraction of
A Bond Graph Model for the Estimation of Torque Requirements … 1597
Fig. 1 Line diagram of the physical system and geometry of the links
the total height of the body and total mass of the body, respectively [7]. Figure 1
shows the line diagram of the physical system along with body frames at different
joints. Table 1 shows the properties of different links of the physical system. The
radius of the cylinder is taken equal to the approximate average radius corresponding
to the actual body part. The width and the thickness of the trunk are taken equal to
approximate dimensions of a body of height 180 cm. The trunk length is taken equal
to the distance from the hip to the top of the head.
The mass of the trunk includes the mass of the torso, both arms and the mass of
the head and neck.
Figure 2 shows the bond graph model of the physical system shown in Fig. 1. Vector
bonds are represented by thick power bonds while the scalar ones are represented
by thinner power bonds. The links that are feet, shanks, thighs and the torso have
been modeled as rigid bodies. The detailed bond graph of the right foot modeled as
a rigid body is represented as WBGO of Right foot. The detailed bond graphs of the
translational and the rotational coupling between the right foot and the ground are
Fig. 2 Bond graph model of a seven-rigid-link physical system of SiTSt and StTSi motions of
the human body where ‘cRC’ represents the conditional rotational coupling, ‘RC’ represents the
rotational coupling and ‘TC’ represents the translational coupling
A Bond Graph Model for the Estimation of Torque Requirements … 1599
The range of motion of the ankle, knee and hip joint in the human body is limited and
has natural constraints. This has been realized in the modeling through a conditional
rotational coupling (cRC).
Figure 3a shows rigid link 1 which is fixed, and link 2 is moving w.r.t. to link 1
through a revolute joint between them. The range of rotation of link 2 is constrained
and has θmin and θmax as the minimum and the maximum limits, respectively. Link
2 is free to move between the minimum and maximum limits. Figure 3b shows the
bond graph of the cRC between link 1 and link 2. The joint internal damping due to
the muscles and ligaments is modeled through an R element in the bond graph as
Rpassive . The working principle of cRC is as follows.
From Fig. 3 Let,
εa = θ − θmax ; (1)
Fig. 3 a Link 1 moving w.r.t. link 2 through a revolute joint. b Bond graph of the conditional
rotational coupling (cRC) between link 1 and link 2
1600 V. Soni and A. Vaz
εb = θ − θmin ; (2)
If εa > 0, then
e14 = K c qc + Rc f 14 ; K c = 0, Rc = 0 (3)
e14 = K c qc + Rc f 14 ; K c = 0, Rc = 0 (4)
Else,
e14 = 0; K c = 0, Rc = 0 (5)
In (3) and (4) it can be noticed that the effort e14 will act only in case the link
2 tries to go beyond the minimum or maximum limits. When it moves between
the limits effort e14 will be zero (5) as the values of K c and Rc are modulated to
zero. In this case, only effort e17 will act which is due to the internal damping of
the joint. All the WBGOs of cRC in Fig. 2 work on the same principle, and for
each joint, corresponding joint limits of motion have been taken for simulation. The
internal damping of the joint has been taken into account in both virtual and the
actual sub-models.
Based on the information of the joint angle from the virtual sub-model, joint torque
is provided on the joint of the actual sub-model through PD controller as shown in
Fig. 4. One of the WBGOs of the cRC of virtual and that of the actual sub-model are
Fig. 4 PD controller between virtual and actual sub-models. a WBGO of a cRC of the virtual
sub-model. b WBGO of a cRC of the actual sub-model
A Bond Graph Model for the Estimation of Torque Requirements … 1601
5 Results
This section presents the MATLAB simulation results of the developed bond graph
model. As shown in Fig. 1, initially, physical system is considered in a sitting posture
on the chair. For SiTSt and StTSi motions, points 4e and 4f of the virtual sub-model
are moved along a desired circular arc. As discussed earlier, the actual sub-model
will track the motion of the virtual through PD controllers and therefore, moves along
the same desired trajectory. In the simulation, SiTSt motion starts at 0 s and ends at
2.5 s. From 2.5 s to 3.5 s, the system remains in the standing position. Further, StTSi
motion starts at 3.5 s and ends at 6.5 s.
Figure 5 shows the ankle, knee and hip joint angles of the virtual and that of the
Fig. 6 Joint torques applied by the PD controller at the joints of the actual sub-model
actual sub-model. It can be observed that the ankle and knee joints of the actual sub-
model almost track the corresponding joints of the virtual sub-model. The hip joint
of the actual sub-model does not track that of the virtual sub-model completely and
a small error can be observed. The error can be further removed by fine-tuning of PD
gains of the controller between the virtual and the actual sub-model. Figure 6 shows
the joint torques applied through the PD controllers at the joints of the actual sub-
model. It can be noticed that the torque profile during the SiTSt motion is smooth, and
there are no transients. At the start of the StTSi motion, transients can be observed
which are shown as encircled portions. These transients are caused due to disconti-
nuities in the desired trajectory profile and can be omitted by tuning the PD gains of
the controller between the virtual and actual system.
6 Conclusion
In this work, a seven-rigid-link bond graph model has been developed to model the
SiTSt and StTSi motions of the human body. For a desired trajectory, information of
joint angles has been taken from the virtual sub-model which is considered to be a
part of the CNS. The internal damping of the natural joints has been modeled using
an R element. Conditional rotational couplings (cRC) have been used to account for
the natural constraints in the motion of the ankle, knee and hip joints. The actual
sub-model, through PD controllers, tracks the desired movement trajectory given to
the virtual sub-model. Transients have been observed in the torque profiles and are
caused due to discontinuity in the desired trajectory profile. Tuning of PD gains of
the controller can be used to control these transients. Estimated torque requirements
A Bond Graph Model for the Estimation of Torque Requirements … 1603
at the joints would be helpful for the development of assistive devices for human
beings. Future scope of the present work would include, simulating the developed
bond graph model for the exact trajectory of natural SiTSt and StTSi motions and
fine-tuning of the PD.
References
1. Lutz W, Sanderson W, Scherbov S (2008) The coming acceleration of global population ageing.
Nature 451:716–719
2. Rupal BS, Rafique S, Singla A, Singla E (2017) Lower-limb exoskeletons : Research trends and
regulatory guidelines in medical and non-medical applications. Int J Adv Robot Syst 14(6):1–27
3. Taghvaei S, Tavasoli A, Feizi N, Rajestari Z, Abdi M (2017) A control-oriented dynamic
model for sit-to-stand motion with fixed support. Proc Inst Mech Eng Part K J Multi-body
Dyn 232(2):265–273
4. Pop C, Khajepour A, Huissoon JP, Patla AE (2003) Experimental/analytical analysis of human
locomotion using bondgraphs. J Biomech Eng 125(4):490–498
5. Selk Ghafari A, Meghdari A, Vossoughi GR (2007) Modeling of human lower extremity
Musculo-Skeletal structure using bond graph approach. in Volume 9: Mechanical Systems and
Control, Parts A, B, and C:1489–1498
6. Mishra N, Vaz A (2017) Bond graph modeling of a 3-joint string-tube actuated finger prosthesis.
Mech Mach Theory 117:1–20
7. Plagenhoef S, Evans FG, Abdelnour T (1983) Anatomical data for analyzing human motion
University of Massachusetts -Amherst. Res Q Exercise Sport 54(2):169–178
Detecting Cancerous Cells Using Data
Augmentation In Deep Cascaded
Networks
Abstract In this article, an approach has been introduced for detecting cancerous
cells. Image processing techniques have been used, based on cancer cell area using
CNNs. A very intriguing aspect of this experiment was that from a very small image
dataset, a large number of images were generated using information augmentation
which was then taken as the training set data. The suggested scheme detects can-
cer behaviors through a convolutional neural network in images of celled samples.
Previously, the same attempts failed to stay away from the database dependencies,
which were somewhat proportional to the number of images in datasets, so we used
a method called data augmentation on smaller sets of images. The scheme prepro-
cesses the input image by grayscale, binarization, inversion, median filtering, and
flood-fill procedures. Depending on the sort of feature to be identified, the prepro-
cessed image is then cancerous cell detected. This methodology was used for several
sets of pictures, and the system was optimized with the feedback from those tests. For
independent cancer cell detection with narrower datasets, the suggested technique
can be efficiently used, which will greatly accelerate the study of cancer and open
greater dimensions.
1 Introduction
Cancer is a collection of associated diseases in which the cells of the body begin
dividing without stopping and spread into surrounding tissue and form growths called
tumors [1]. It is one of the leading causes of deaths worldwide, and the number of
new cases is expected to rise to 23.6 million by 2030 [2]. Computed tomography
(CT) and magnetic resonance imaging (MRI) are two popular methods used in the
diagnosis of cancer.Through detection, the tumors are classified into two categories:
2 Proposed Methodology
CNN is used in this paper to identify cancer based on CT images. This study’s
prime objective is to assess whether the tumor is malignant or benign. We also need
to increase pictures along with the goal in order to achieve better outcomes when
training. We will first use basic training for the potential comparison and then use
the bigger enhanced information set to detect cancer cells.
To execute a CNN image classification, we will train our own tiny dataset. We will
conduct the tumor classification with a narrower set of information. We will then use
typical methods for information augmentation and retrain our models.
Data augmentation usually has two distinct methods. The first strategy is to achieve
added data before the classifier is trained. The second strategy tries to learn to augment
through a neural net that is prepended. This neural network takes two random pictures
from the set of training data at the moment of training and delivers a single picture
Detecting Cancerous Cells Using Data Augmentation . . . 1607
so that this picture fits either in fashion or in context with a defined picture from
the training set. In addition to the initial training data, this output, which represents
an improved picture created by the network, is fed into the second classification
network. The training loss is then back-propagated to train the network’s expanding
layers as well as the network’s classification layers [5].
Traditional Transformations Traditional transformations consist of manipulating
the training data by using a mixture of affine transformations. We produce a “dupli-
cate” picture for each input picture by shifting, zooming in/out, rotating, flipping,
distorting, or shading it with a hue. The neural net is supplied with both picture and
duplicate. We create a 2N size dataset for a size N dataset [6].
Generative Adversarial Networks We can choose a style picture from a subset of
six different styles for each input picture: Cezanne, Monet, Ukiyo-e and Van Gogh.
A styled transformation is generated of the original image. In order to train the net,
both the original and the styled image are fed [7].
2.3 Dataset
A small amount of tiny pathological pictures were given with the dataset used to
classify from Kaggle Platform. A binary label showing the existence of cancer cells
is annotated for each picture. A favorable label shows that a patch’s core region
of 32 × 32px includes at least one pixel of tumor tissue and thus makes the cell
malignant, whereas if the label was negative, there were no cancer cells present and
therefore a benign one. The dataset used has been taken from https://www.kaggle.
com/c/histopathologic-cancer-detection/data [8].
CNN shares weights in the convolutional layer reducing the memory footprint and
increases the performance of the network. The important features of CNN lie with
the 3D volumes of neurons, local connectivity, and shared weights. By combining
different sub-areas of the input image with a kernel, a feature map is produced by
the convolution layer. The ReLu layer then uses the anonymous activation feature
to increase the convergence characteristics when the error is small. A pixel with
the maximum value or average values is chosen as a representative pixel for the
picture/feature map area, and a grid of 2 × 2 or 3 × 3 is reduced to a single scalar
value in the pooling layer. The sample size is thus greatly reduced. In combination
with the convolutionary layers toward, the output phase traditional fully connected
(FC) layer is sometimes used. The convolution layer and the pool layer are generally
used in some conjunction in CNN architecture. Two kinds of activities generally
take place in the pooling layer—max pooling and means pooling. In mean pooling,
the average area is calculated within the characteristics and calculated within the
maximum of feature points in the maximum pooling. Mean pooling decreases the
mistake created by the size of the area and maintains background data. Max pooling
decreases the estimated error of the convolution layer parameter induced by medium
deviation [11] (Fig. 1).
Convolution Layer The key construction block of the CNN is the convolution layer.
It holds the primary computational load of the network. This layer produces a dot
product between two matrices in which one matrix is a set of learning parameters
known as a kernel and the other matrix is the limited part of the receptive area.
The kernel is smaller, but in-depth than an image. This implies the kernel width and
height, but the depth, when the picture is made up of three channels (RGB), is spatially
low. The kernel slides across the height and width of the picture and represents the
receptive region on the front. This results in a two- dimensional representation of the
Detecting Cancerous Cells Using Data Augmentation . . . 1609
picture known as an activation map, which shows the kernel response in each image
space position. The kernel’s sliding size is known as a stride [10].
Pooling Layer The pooling layer replaces the network output at certain places by
obtaining an overview of the outputs in the vicinity. This helps to reduce the spatial
dimension of the image, which reduces the necessary calculations and weights. The
pooling procedure is carried out separately on each section of the representation.
Some pooling functions are such as the rectangular district average, the L2 rectangular
district norm, and the weighted average from the core pixel distance. The most
common method, however, is max pooling, which reports the highest output [10]
(Figs. 2 and 3).
Fully Connected Layer In this layer, fully connected layer neurons have complete
connectivity to all neurons in the layer above and below. For this reason, a matrix
multiplication and a bias impact can be calculated as normal. The FC layers help to
map the image from input to output [10].
Non-linearity Layer Because convolution is linear and images are far from linear,
nonlinearity layers can often be positioned right after the convolution layer so that
nonlinearity appears on the activation map. Nonlinear activities have various kinds,
most of which are sigmoid, Tanh, ReLU, etc. [10].
3 Experiment
4 Results
The neural network based on convolutional and watershed segmentation has been
implemented in Jupyter Notebook using Python and the system is trained with sample
datasets for the model to understand and familiarize the cancer cell (Fig. 4).
A sample image has been fed as an input to the trained model, and the model
at this stage is able to tell the presence of cancer. In case malignancy is present, a
message indicating 1 will be displayed which indicates the presence of cancer cells.
While the major focus of the experiment was to show how even with smaller sets
of images, we can achieve great accuracy. The following table shows the highest
accuracy or the best score in all the 40 epochs (Table 1).
Table 1 Results
Experiment No. Best score Final score
1 0.674 0.629
2 0.9071 0.8776
1612 A. Jain et al.
5 Future Work
References
11. Sasikala S, Bharathi M, Sowmiya BR (2018) Lung cancer detection and classification using
deep CNN. Int J Innov Technol Explor Eng (IJITEE) 8(2S)
12. Chaturvedi P, Gupta L (2018) Study and detection of eye blink artifacts in EEG signals,. In:
2018 IEEE international students’ conference on electrical, electronics and computer science
(SCEECS), Bhopal, India.https://doi.org/10.1109/SCEECS.2018.8546907
13. Makwana G, Gupta L (2018) The unreasonable effectiveness of data. IEEE Intell Syst
7(2.16):29–32
Free Vibration Analysis of the Sandwich
Curved Panels with the Gradient
Metallic Cellular Core
Abstract In this study, free vibration analysis of the sandwich curved panels with
the gradient metallic cellular core is presented. The present finite element model is
established on higher-order shear deformation theory using curvilinear coordinate
system. The sandwich curved panel consists of two isotropic face sheets and an FG
metallic cellular core layer. The internal pores in the core layer follow different types
of distributions. The material properties of the gradient metallic cellular core layer
of the sandwich curved panel vary in the thickness direction as a function of porosity
coefficient and mass density. The present model is validated with the limited results
available in the open literature, and few new results are also discussed that can be
utilized as a benchmark solution. The influence of porosity coefficient (e0 ) and pore
distribution types on the free vibration characteristics of the sandwich curved panel
with the gradient metallic cellular core are also analyzed.
1 Introduction
The curved panels are very commonly used in several engineering applications like
civil engineering, aerospace, ship engineering, high-pressure vessels, and chemical
industry and so on [1]. The sandwich curved panels with gradient metallic cellular
core have exceptional mechanical properties of the lightweight structures which
makes it more suitable to use it in aircraft, spacecraft and space structures. Therefore,
for designing such types of structures, the vibration analysis of these structures is
required.
Zhao et al. [2] scrutinized the free vibration behavior of the gradient metallic
cellular shallow curved panels with general boundary conditions. Amir and Talha [3]
investigated the nonlinear vibration behavior of the functionally graded (FG) porous
curved panels. Pandey and Pradyumna [4] presented a layer-wise finite element
formulation for studying the vibration responses of the functionally graded sand-
wich curved panels. MagnuckaBlandzi [5] proposed mathematical modeling of the
rectangular sandwich porous plate based on the poroelasticity. He considered the
core as gradient metallic cellular with properties varying across its thickness.
Liu et al. [6] scrutinized the blast resistance and dynamic response of hollow sand-
wich cylinders with gradient metallic cellular cores using the finite element method
and found that the sandwich structures with thinner inner face sheets are superior
to the sandwich structure with thicker face sheets of the cylinders. Jing and Zhao
[7] studied the blast resistance, dynamic response and energy absorption capabili-
ties of the sandwich curved panel with a gradient metallic cellular core, under the
air-blast loading. Xiang et al. [8] studied the impulsive response of the clamped–
clamped sandwich plate with FG core under blast impulsive loadings. Magnucka-
Blandzi and Magnucki [9] calculated the effective design of the optimal dimension-
less parameters of a porous sandwich beam using a theory of minimum total potential
energy. Amir and Talha [10] studied the vibration characteristics of the porous curved
beams/arches.
Free vibration analysis of the sandwich curved panels with gradient metallic
cellular core has not been reported so far. So, the present study focuses on the
influence of porosity coefficient (e0 ) and pore distribution types on free vibration
characteristics of sandwich curved panels with the gradient metallic cellular core.
The internal pores in the core layer follow different types of distributions. The mate-
rial properties of the gradient metallic cellular core layer of the sandwich curved
panel vary in the thickness direction as a function of porosity coefficient and mass
density.
2 Theoretical Formulations
Consider a sandwich curved panel having gradient metallic cellular core, and the
upper and the lower faces are made of isotropic metals. The total thickness h and the
dimensions along the orthogonal curvilinear axes (ξ, ζ ) are a and b, respectively, and
the principal radii of the curvature are denoted by Rξ and Rζ , respectively (Fig. 1).
The displacement field based on higher-order shear deformation theory is taken
as:
z
u = 1+ u + zφ1 + z 2 ϕ1 + z 3 ψ1
Rξ
z (1)
v = 1+ v + zφ2 + z 2 ϕ2 + z 3 ψ2
Rζ
w=w
Free Vibration Analysis of the Sandwich Curved Panels … 1617
where (u, v, w) are displacements of a random point along with the orthogonal
curvilinear coordinates (ξ, ζ, z) and (u, v, w) are the corresponding displacements
at the mid-plane and (φi , ϕi , ψi ) are higher-order rotation parameters of a point. The
linear strain–displacement relationship for the sandwich curved panel is given as in
[3]
⎧ ⎫
⎧ ⎫ ⎪ ⎪
⎪
1
(1+z/R )
∂u
∂ξ
+ w ⎪
⎪
⎪
⎪ εξ ξ ⎪ ⎪
⎪ ξ R ξ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ∂v ⎪
⎪
⎪
⎪ ⎪ ⎪ 1
+ w
⎪
⎨ εζ ζ ⎪ ⎬ ⎪ ⎨ (1+z/Rζ ) ∂ζ Rζ ⎪
⎬
∂u ∂v
γξ ζ = (1+z/Rζ ) ∂ζ + (1+z/Rξ ) ∂ξ
1 1
(2)
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
γξ z ⎪ ⎪ ⎪ ∂w ⎪
⎪
⎪
⎩ γ =⎪ ⎭ ⎪ ⎪
1
(1+z/Rξ ) ∂ξ
− Ruξ + ∂u ∂z ⎪
⎪
⎪
⎪ ⎪
⎪
ζz ⎪
⎩ ∂w ∂v ⎪
⎭
1
(1+z/Rζ ) ∂ζ
−
Rζ
v
∂z
+
The sandwich curved panel with the upper and the lower faces having thickness
hf is made of isotropic metal. The metallic cellular core of the curved panel has a
thickness hc with varying mechanical properties (Young’s modulus E c , density ρc ,
and Poisson’s ratio νc ) due to the distribution of the pores. Two types (type 1 and
type 2) of pore distributions are taken in the modeling of the core layer. Figure 2a
represents the type 1 pore distribution, in which mid-plane is highly porous compared
to the top and bottom surface. The mechanical properties (i.e., density and Young’s
modulus) have a minimum value (P0 ) at the mid-plane and maximum value P1 at the
upper and lower surfaces. Figure 2b illustrates type 2 pore distribution in which the
top is less porous and the bottom surface has the highest porosity. The mechanical
1618 M. Amir and M. Talha
Fig. 2 Cross sections and patterns of pore distributions: a Type 1 and b Type 2
properties P(z) vary with the thickness and have minimum value P0 at the bottom
surface and maximum value P1 at the top.
The material properties of the functionally gradient metallic cellular core with
different pore distributions are expressed as in [2]
For distribution type 1
E c (z) = E c1 [1 − e0 cos(π ζ )]
(3)
ρc (z) = ρc1 [1 − em cos(π ζ )]
where
ζ = z/ h c
e0 = 1 − E c0 /E c1 (0 < e0 < 1) and
em = 1 − ρc0 /ρc1 (0 < em < 1) (5)
where e0 is the porosity coefficient, em is the porosity coefficient for mass density,
E c1 and ρc1 are the maximum values of Young’s modulus and density, E c0 and ρc0
are the minimum values of Young’s modulus and density, respectively.
Free Vibration Analysis of the Sandwich Curved Panels … 1619
To ensure the accuracy and efficiency of the present finite element model, two numer-
ical results are compared. First of all,√the convergence and the comparison of the first
three frequency parameters (λ = ωh ρ1 /E 1 ) of the simply supported (SSSS) spher-
ical panel with those results given by Zhao et al. [2] are mentioned in Table 1. The
geometric parameter of the spherical panel, a = b = 1 m, h = 0.02 m, Rζ = 2 m,
Rξ = 2 m, and material properties, E 1 = 70 Gpa, ρ 1 = 2702 kg/m3 , ν = 0.3, and
pore distribution type 2 with e0 = 0.5 is considered. From the table, it is found that
there are a good convergence and excellent agreement between the present and the
reference results.
Due to a lack of results available for the present type of formulation, we compared
the results of the frequency parameters of the sandwich cylindrical panels with FG
(Al/Al2 O3 ) core at the a/h = 10, Rζ /a = 0.5, n = 0.2 and thickness proportions 1-8-
1. The geometry and material properties are taken from Pandey and Pradyumna √ [3].
The non-dimensional frequency parameter is expressed as: λ = ωL 2 / h ρ0 /E 0 ,
where E 0 = 1 GPa and ρ0 = 1 Kg/m3 . The convergence and comparison of the
present results are given in Table 2, and the present results are compared to the
results calculated by Pandey and Pradyumna [3] using different shell theories. From
the table, it is observed that there is a good agreement with reference results.
Now, consider a sandwich spherical panel with gradient metallic cellular core
and isotropic face layers having a proportion of the thickness 1:10:1. The material
properties are E 1 = 200 GPa, ρ1 = 7800 Kg/m3 , ν = 1/3 and geometric properties
of the spherical sandwich panel are a/b = 1, a/h = 100. The two distinct types of
pore distributions and clamped–clamped (CCCC) √ are considered. The dimensionless
frequency parameter is defined as:λ = ωh ρ1 /E 1 .
Table 3 presents the frequency parameters of the clamped sandwich panels with
curvature ratio (R/a = 20, 50, 100) and porosity coefficient (e0 = 0 and 0.7), for
both types of pore distributions (type 1 and type 2). The effect of curvature ratio is
Table 2 Comparison of the frequency parameter of the sandwich panel (1-8-1) at the a/h = 10,
Rζ /a = 0.5, and n = 0.2
Mesh size Frequency
2×2 2.7686
3×3 2.7918
4×4 2.8036
5×5 2.8100
6×6 2.8139
7×7 2.8163
Sendor’s theory [2] 2.7746
Love’s theory [2] 2.7797
Donnell’s theory [2] 2.8709
observed that on increasing the values of R/a, the frequency parameters decrease for
both types of pore distributions.
Table 4 shows the effect of the porosity coefficient on frequency parameters
of the clamped sandwich panels at the curvature ratio (R/a = 20, 50). From the
table, it is observed that the effect of the porosity coefficient depends on the type
of pore distribution taken. For distribution type 1, on increasing the value of e0 , the
frequency parameters increase, while for distribution type 2, the frequency parame-
ters decrease with the increase in porosity coefficient. For a better understanding, the
effect of porosity coefficients for both distribution types on frequency parameters
Fig. 3 Effect of porosity coefficients for both distribution types on frequency parameters
of the clamped sandwich panels (at R/a = 100) is displayed in Fig. 3. The same
conclusions can be drawn from the figure as it is observed from Table 4.
4 Conclusions
Acknowledgements The authors would like to acknowledge financial support by the DST-Science
and Engineering Research Board, Govt. of India, under Project No. YSS/2015/001290, dated-09th
Nov 2015.
References
1. Pang F, Li H, Cui J, Du Y, Gao C (2019) Application of flügge thin shell theory to the solution
of free vibration behaviors for spherical-cylindrical-spherical shell: a unified formulation. Eur
1622 M. Amir and M. Talha
Abstract Coal was, is, and will continue to be the backbone of global electricity
generation. This is because coal as a fuel is one of the most cost-effective ways to
provide affordable, safe, and reliable electricity at the scale that is needed to achieve
genuine access to modern electricity services world-wide. However, the conven-
tional processes involved in mining are associated with quite few accidents rendering
injuries to the workers. In addition, a huge amount of greenhouse gases (GHG) is also
emitted during the process. Thus, the process of mining coal must undergo a paradigm
shift in terms of the machinery and equipment employed which will not only bring
about a safe working environment for the workers, but also an efficient, economic,
and eco-friendly method. In this regard, the present article analyses the initiatives
adopted by an Indian coal mining organization located in Odisha to promote work-
place safety, environment conservation, and economic viability which are also the
three major aspects of sustainable development. A comparative study between the
conventional coal mining process and the cleaner way of coal mining expressed in
the form of the energy consumption has also been presented by deploying the surface
miner technology. The subsequent savings in the form of energy, cost, and reduction
of carbon emissions have been highlighted.
1 Introduction
At present, population explosion and per capita energy consumption are the two
important aspects in the globe. Coal is the major fossil fuel in the planet earth. Coal
as a fossil fuel is a major player of energy security of world. It is considered as a
S. S. Panigrahi (B)
Operations Management, International Management Institute, Bhubaneswar, India
e-mail: [email protected]
P. C. Panigrahi
Mahanadi Coalfields Limited, A Subsidiary of Coal India Limited, Sambalpur, India
significant source and material for electricity generation, especially for the industries
which are energy-intensive like steel or cement industry [2]. It supplies energy for
the development of the society. Hence, coal mining cannot be avoided. A number
of greenhouse gases (GHGs) are emitted in the processes of coal mining including
carbon dioxide. Coal mining also has a significant contribution towards greenhouse
effect which leads to climate change and global warming. In 2013, more than 83%
of the total carbon dioxide emission originated from the combustion of coal [3]. An
amount of almost 79% of sulphur dioxide (SO2 ), 57% of nitrogen oxides (NOx ), and
44% of particulate matter (PM) originated out of the direct coal combustion of coal.
Also, 93% of SO2 , 70% of NOx , and 67% of PM emissions originated from all kinds
of usage of coal constituting direct combustion emission and emission from coke
stoves and other industrial furnaces [2]. Owing to which, an effective means for the
reduction of carbon emissions is of extreme importance in the current scenario. And
in order to facilitate coal mining in ecologically and economically viable way, clean
coal technologies (CCTs) are essential. Hence, application of CCTs in mining, be it
coal or non-coal mining, is the need of the hour. The biggest coal reserves are found
in the USA, Russia, China, and India. As per the BP Statistical Review of World
Energy [7], the top coal producers in 2016 are indicated in Table 1.
Clean coal technologies involve a wide range of coal production and utilization
related technologies, including green mining, surface miners, coal purification, high-
efficiency power generation, advanced coal conversion, pollution control, and carbon
capture, utilization, and storage.
As the present article is concerned with the coal mining organization located in
Odisha, the following are some figures about the availability of the same in the state.
The total mineable coal reserve available in Odisha is 75.8 BTe, with two coalfields,
i.e. one is Talcher coalfield and the other is Ib coalfield. Mineable coal reserves of
Talcher coalfield and Ib coalfield are 50.968 BTe and 24.830 BTe, respectively.
The coal producing company under study is a pioneer in the introduction of clean
and new technologies. It was also the first company to introduce surface miners for
coal production in India. The organization was one of the early birds to
augment its clean coal
production programme. It has been the trendsetter in introducing blast-free tech-
nology for winning coal in open cast mine by surface miner. There also exists a
SILO operating at one of the mines owned by this organization. The organization
has transformed itself from one with the smallest capacity (23 MTY) in 1992 to
the one with the largest capacity (100MTY) in 2010–11. With the growing demand
of clean and high-quality coal as the measures of energy security, this organization
has always been looked as viable, productive, and growing coal company since its
inception.
Surface miners were first employed in the Indian context on a limestone deposit in
the year. The surface miner was introduced for the first time in coal mining in the year
1999 in Lakhanpur OCP which is one of the mines of this company. Currently, at
least more than 100 surface miners have been employed in the Indian mining sector
owing to the benefits of higher productivity as well as eco-friendliness associated
with it [5]. The surface mining technology was introduced in 1970s which was
inspired from the road milling machines which is based on cutting the old road surface
for road construction [4]. Surface miner finds its natural applications in opencast
projects where operations such as drilling, blasting, and crushing are prohibited. The
surface miner produces coal with desired size as demanded by the consumers. It also
segregates the parting/dirt band which is responsible for reduction in the quality of
coal. This help in providing high-quality coal generating huge revenues. Apart from
the above, one most important aspect is the safety feature which comes with the
surface miner as it eliminates drilling, blasting, and crushing operations. Figure 1a
represents a surface miner and (b) represents the cutting of coal by the surface miner.
One of the reasons for deploying surface miners in the Indian context was strategic
advantage. In conventional mining, owing to the depletion of surface deposits,
the surface mines have now become deeper. Moreover, the operating surface area
required is becoming larger. India is a highly populated country and the land avail-
able for mining is less than the nation’s mineral demand. Now, the operating surface
mines have gotten closer to the human settlements where conventional drilling and
blasting technique is prohibited in anticipation of ground vibration [6]. Hence, in the
Indian context, employing the surface miners has become popular. The following
sub-sections below discuss the advantages obtained by using the surface miner for
coal mining:
1. Reduction in dependency on human factor and reduction in stress/ strain / health
/ hygiene / safety of manpower:
Fig. 1 a Surface miner and b cutting of coal by surface miner at the coal face
A Study on Clean Coal Technology in the Indian Context 1627
• Surface miners have combined different operations earlier which were part
of mining activities resulting in high productivity and reduced idle time.
• Surface miner produces coal with less coal loss and dilution, raising the
quality and quantity of coal produced.
• Surface miner also segregates the parting/dirt band which decreases the
quality of coal, thereby providing quality coal generating huge revenues
and increasing productivity and profitability of the company.
• Improved coal recovery especially in areas sensitive to blasting results in
increased quantity of coal produced.
It can be observed that the total energy conserved per Te by deployment of surface
miner instead of conventional methods (drilling, blasting, and crushing) for coal
production is given by:
1630 S. S. Panigrahi and P. C. Panigrahi
A study has been made by analysing the cost incurred in different operations involved
in the conventional method of mining in opencast coal mines and that of in the
surface miner. Table 3 below represents the cost savings per tonne of coal made by
the organization by using surface miner compared to the conventional method of
mining.
The coal deposit available in the company in focus is of thermal grade. The raw coal
ash as per the test results of CMPDI is 36.3%, and the likely average monthly raw coal
ash considering dilution has been arrived as 40.1%. Hence, the main washing process
A Study on Clean Coal Technology in the Indian Context 1631
for the washery has been selected on the basis of likely average monthly raw coal ash,
i.e. 40.1% and qualitative requirement of washed coal of the thermal power stations.
The ash percentage of washed coal has been kept around 33.5 ± 0.5%. Further, the
study of test results revealed that washing of 50–13 mm coal and mixing of untreated
−13 mm size fraction are suitable for achieving the targeted washed coal ash. The
process of the proposed washery for coal processing is based on closed water circuit
system. All the water fed into the system will be collected after use and re-circulated
after treatment in various units, and no effluent will be allowed to escape into the
natural drainage system. The coal processing through washery will reduce the ash
contents to the desired level, and at the same time, it will reduce the weight of the
yield by 7–10%. This will result in decrease in transportation cost and in turn reduces
the fuel consumption. Thus, the coal processing will reduce carbon emission. The
company is planning to set up four washeries of 10 MTY capacity each, shortly in
different phases; the details of time line for commencement are mentioned in Table
4.
4 Silo
In order to save energy, control pollution, and prevent accidents by eliminating expo-
sure of workforce in loading and transporting operation, it is desirable to transport
1632 S. S. Panigrahi and P. C. Panigrahi
coal and over burdens through conveyor system. This will also lead to substantial
amount of investment which will result in long-term savings on transportation cost
both in departmental and outsourced areas. The SILO with rapid loading system
(RLS) was commissioned in the year 1989, and coal is being dispatched to NALCO
through the RLS from Bharatpur OCP as shown in Fig. 2. SILO with RLS expedites
the transporting and loading operations which saves huge amount of energy elimi-
nating a large fleet of vehicles/trucks and loading machines. This quantity of energy
saved is manifested in reduction of carbon emission which is the need of the day.
5 Conclusion
References
Sanjay Sharma
Nomenclature
Dimensional Parameters
c Geometrical clearance, Mm
D Diameter of journal, Mm
e Eccentricity of journal, Mm
S. Sharma (B)
School of Mechanical Engineering, Shri Mata Vasihno Devi University, Katra 182320, India
e-mail: [email protected]
Non-dimensional Parameters
c̄ c/RJ
F̄, F̄0 (F, F0 ps RJ2 )
F̄x , F̄z (Fx , Fz ps RJ2 )
h̄ h/c
h̄ min h min c
p̄, p̄c ( p, pc ) ps
p̄L , p̄max ( pL , pmax
) ps
t¯ t c2 ps μr RJ2
W0
W̄0 RJ2
ps
μ̄ μ μr
ωJ μr RJ2 c2 ps , Speed parameter
1 Introduction
bearing. However, it would also result in an increase in friction losses and leads
to more power requirement. In order to keep losses at a minimum level, the loca-
tion, numbers and depth of texture is an important parameters to be optimal. Several
researchers have studied different parameters and configuration of surface textures
in different bearings in order to improve its performance. Qiu et al. [1] studied six
different texture shapes to evaluate the friction and stiffness coefficient of gas lubri-
cant parallel slider bearings. They found and finalized ellipsoidal shape to yield the
maximum performance of bearing. Dadouche and Conlon [2] experimentally studied
the effect of dimple shape texture on the steady-state performance characteristics
of highly loaded journal bearing contaminated lubricant. They found that lightly
textures bearing showed exceptional performance as compared to highly textured
journal bearing. Su et al. [3] conducted parametric analysis by using finite differ-
ence method (FDM) to obtain the optimum surface textures parameters in terms of
load carrying capacity of slider bearing with elastic deformation. Zheng et al. [4]
studied the effect of hexagonal texture size on tribological properties under mixed
lubrication. They found that friction coefficient decreases with increase in width
within examined range. Feng and Peng [5] numerically studied the effect of rotary
speed, texture depth, width and number on the static performance of water-lubricated
thrust bearing by using JFO (Jakobsson–Floberg–Olson) cavitation model and found
that well-designed groove texture can improve the performance of bearing. Shinde
and Pawar [6] studied numerically by using COMSOL multiphysic 5.0 software
to find static performance analysis of square-shape microtextured hydrodynamic
journal bearing by placing texture in 90°–180° region results in improvement in
static performance. Liang et al. [7] used CFD technique to investigate the effect of
partial texture location on the performance of hydrodynamic bearing. They reported
that texture located in inlet area and having shallow size can improve the performance
of bearing. Yu et al. [8] used CFD technique to investigate the effect of surface texture
on grease-lubricated journal bearing. They reported that texture shape has a more
significant effect on its static performance parameters. Tala-Ighil et al. [9] studied
textured hydrodynamic journal bearing with spherical dimple by using FDM to inves-
tigate the static performance and found that these values increase with the change in
texture geometrical parameters. Cupillard et al. [10] studied circular dimple shape
texture hydrodynamic journal bearing by using computational fluid dynamics to
investigate the static performance parameters and found low friction coefficient for
deep dimples than swallow dimples. Uddin et al. [11]studied texture hydrodynamic
slider bearing with star-like structure by using finite difference method (FDM) to
investigate the static performance parameters and found friction coefficient lower
than the ellipse-, circle- and chevron-shaped texture. Yadav and Sharma [12] studied
dimple textured hydrodynamic thrust bearing by FEM and JFO mass conservation
algorithm to investigate the static performance parameters and found that for better
performance, an optimum location of the lubricant supply hole and depth of dimple is
required. Brizmer and Kligermam [13] studied microdimple texture hydrodynamic
journal bearing by using laser surface texturing (LST) to investigate the load capacity
and attitude angle and found that maximum load capacity for partial textures. Shen
and Khonsari [14] studied different texture shapes in hydrodynamic slider bearing
1638 S. Sharma
2 Methodology
The triangular-shape textured journal bearing has been shown in Fig. 1. The non-
dimensional governing Reynold’s equation for a hydrodynamic bearing is given by
Sharma et al. [23]:
Lubrication Characteristics of Newtonian-Lubricated Hydrodynamic Bearing … 1639
Fig. 1 a Cross-sectional view bearing having textured surface; b Triangular-shape texture views;
c Full triangular-shape textured distribution on the bearing surface
∂ 3 ∂p ∂ 3 ∂p ∂ F1 ∂h
h F2 + h F2 = 1− h + (1)
∂α ∂α ∂β ∂β ∂α F0 ∂t
where F 0 , F 1 and F 2 are viscosity functions and are given by the following
expressions:
1 1 1
1 z z F1
F0 = dz, F 1 = dz and F 2 = z− dz
μ μ μ F0
0 0 0
6 25
δt=0.0 δt=0.5 δt=0.0 δt=0.5
δt=1.0 δt=1.5 δt=1.0 δt=1.5
5
LOAD CARRYING CAPACITY
COEFFICIENT OF FRICTION
δt=2.0 20 δt=2.0
4
15
3
10
2
1 5
0 0
0.2 0.4 0.6 0.2 0.4 0.6
ECCENTRICITY RATIO ECCENTRICITY RATIO
(a) (b)
Fig. 2 a, b Variation of load carrying capacity and coefficient of friction with eccentricity ratios
of textured hydrodynamic journal bearings having full texture region (0°–360°)
Lubrication Characteristics of Newtonian-Lubricated Hydrodynamic Bearing … 1641
6 14
δt=0.0 δt=0.5 δt=0.0 δt=0.5
δt=1.0 δt=1.5
5 δt=1.0 δt=1.5 12
LOAD CARRYING CAPACITY
δt=2.0
COEFFICIENT OF FRICTION
δt=2.0
4 10
3 8
2 6
1 4
0 2
0.2 0.4 0.6 0.2 0.4 0.6
ECCENTRICITY RATIO ECCENTRICITY RATIO
(a) (b)
Fig. 3 a, b Variation of load carrying capacity and coefficient of friction with eccentricity ratios
of textured hydrodynamic journal bearings having pressure developing region (144°–288°)
In order to find the optimal values of texture depth and location, the performance
enhancement ratio defined by Sharma et al. [23]:
LCC of textured bearing
LCC of plain bearing
PER =
COF of textured bearing
COF of plain bearing
From Fig. 4 it is clear that at texture depth of 1.0 the performance enhancement
ratio is maximum in pressure increasing region only, at eccentricity ratio of 0.2.
Performance enhancement ra o
2
1.6
1.2
0.8
0.4
0
0-360 0-180 180-360 144-288
Texture region
ε = 0.2 ε = 0.4 ε = 0.6
Fig. 4 Performance enhancement ratio of textured hydrodynamic journal bearings at texture depth
of 1.0
1642 S. Sharma
References
1. Qiu M, Delic A, Raeymaekers B (2012) The effect of texture shape on the load-carrying capacity
of gas-lubricated parallel slider bearings. Tribol Lett 48:315–327
2. Dadouche A, Conlon MJ (2016) Operational performance of textured journal bearings
lubricated with a contaminated fluid. Tribol Int 93:377–389
3. Su B, Huang L, Huang W, Wang X (2017) The load carrying capacity of textured sliding
bearings with elastic deformation. Tribol Int 109:86–96
4. Zheng L, Gao Y, Zhong Y Lu GO Liu Z, Ren L (2018) The size effect of hexagonal texture on
tribological properties under mixed lubrication. Ind Lubr Tribol 70:1798–1805
5. Feng H, Peng L (2018) Numerical analysis of water-lubricated thrust bearing with groove
texture considering turbulence and cavitation. Ind Lubr Tribol 70:1127–1136
6. Shinde AB, Pawar PM (2017) Effect of partial grooving on the performance of hydrodynamic
journal bearing. Ind Lubr Tribol 69: 574–584
7. Liang X, Liu Z, Wang H, Zhou X (2016) Hydrodynamic lubrication of partial textured sliding
journal bearing based on three-dimensional CFD. Ind Lubr Tribol 68:106–115
8. Yu R, Li P, Chen W (2016) Study of grease lubricated journal bearing with partial surface
texture. Ind Lubr Tribol 68:149–157
9. Tala-Ighil N, Maspeyrot P, Fillon M, Bounif A (2007) Effects of surface texture on journal-
bearing characteristics under steady-state operating conditions. Proc Inst Mech Eng Part J Eng
Tribol 221:623–633
10. Cupillard S, Glavatskih S, Cervantes MJ (2008) Computational fluid dynamics analysis of a
journal bearing with surface texturing. Proc Inst Mech Eng Part J Eng Tribol 222:97–107
11. Uddin.MS, Ibatan, Shankar TS (2016) In fluence of surface texture shape , geometry and
orientation on hydrodynamic lubrication performance of plane-to-plane slider surfaces
12. Yadav SK, Sharma SC (2016) Performance of hydrostatic textured thrust bearing with supply
holes operating with non-Newtonian lubricant. Tribol Trans 59:408–420
13. Brizmer V, Kligerman Y (2012) A laser surface textured journal bearing. J Tribol 134:031702
14. Shen C, Khonsari MM (2015) Tribology international numerical optimization of texture shape
for parallel surfaces under unidirectional and bidirectional sliding. 82:1–11
15. Tala-Ighil N, Fillon M, Maspeyrot P, Effect of textured area on the performances of a
hydrodynamic journal bearing. Tribol Int 44:211–219
16. Sharma SC, Yadav SK (2014) Tribology International performance analysis of a fully textured
hybrid circular thrust pad bearing system operating with non-Newtonian lubricant. 77:50–64
17. Grützmacher PG, Rosenkranz A, Szurdak A, Gachot C, Hirt G, Mücklich F (2017) Effects
of multi-scale patterning on the run-in behavior of steel—Alumina pairings under lubricated
conditions. 1700521:1–8
18. Gropper D, Harvey TJ, Wang L (2018) Tribology international numerical analysis and
optimization of surface textures for a tilting pad thrust bearing. Tribiol Int 124:134–144
19. Gropper D, Harvey TJ, Wang L (2018) A numerical model for design and optimization of
surface textures for tilting pad thrust bearings. Tribol Int 119:190–207
Lubrication Characteristics of Newtonian-Lubricated Hydrodynamic Bearing … 1643
1 Introduction
2 Mathematical Formulation
The synchronous generator is analyzed using dq0 model. The set of six coupled
equations is used to calculate the currents, voltages and magnetic fluxes on d-q-0
reference frame as shown in differential eqs. (1)–(6). Wherein u, i, ψ denote voltages,
currents and flux linkages, respectively. The subscripts d, q, 0 are associated with the
d-axis components, q-axis components and zero sequence components, respectively,
of voltages, currents and flux linkages as given in Eq. (7). The detailed dq0 model
for synchronous generator modeling is given in [9, 10].
d[ψ]dq0
[u]dq0 = −[R][i]dq0 − + ω[ψ]dq0 (1–6)
dt
where
T
[u]dq0 = u d u q u 0 u f d u D u Q
T
[i]dq0 = i d i q i 0 i f d i D i Q
T
[ψ]dq0 = ψd ψq ψ0 ψ f d ψ D ψ Q
xa + xb + xc
x0 = ,x → u, i, ψ (7)
3
Big Turbo-Generator Shaft Vibrations Control … 1647
J dω
= −Tem + Tdrive (8)
p dt
3
Tem = p ψq i d − ψd i q (9)
2
dγ
= −ω (10)
dt
where
T drive is torque from turbines
J is polar moment of inertia
p is number of pole pairs.
The above set of Eqs. (1)–(10) is used to model the synchronous generator. The
Park’s transformation as given in Eq. (11) is used to express parameters u, i, ψ from
a-b-c reference frame to d-q-0 reference frame.
⎡ ⎤ ⎡
4π ⎤⎡ ⎤ ⎡ ⎤
xd cos γ cos γ + 2π cos γ + x a xa
⎣ xq ⎦ = 2 ⎣ sin γ sin γ + 2π sin γ + 4π ⎦⎣ xb ⎦=[ξ ]⎣ xb ⎦
3 3
3 3
(11)
3 1 1 1
x0 2 2 2
xc xc
where, x → u, i, ψ.
The generator is connected to the infinite load through grid as shown in Fig. 2. ua , ub ,
uc are voltages at the terminal of the generator and u∞a , u∞b , u∞c are voltages at the
point of fault on the transmission line. The generator is considered to transfer power
through line with resistance Re and inductance L e for each phase. During occurrence
of fault, the phase voltages at generator terminals are different in loaded condition
[10] in contrary to unloaded case [9] as given by Eqs. (12) and (13).
d
[u abc ] = [u ∞abc ] + Re [I ][i abc ] + L e [I ] [i abc ] (12)
dt
d
u dq0 = [ξ ][u ∞abc ] + Re [I ][ξ ][i abc ] + L e [I ] [ξ ][i abc ] (13)
dt
u ∞a = 0, i b = i c = 0
di f d di D
L f dσ − L Dσ = −R f d i f d + R D i D + u f d (14)
dt dt
di d di f d di D
L md + L md + LD = −R D i D (15)
dt dt dt
di q di Q
L mq + LQ = −R Q i Q (16)
dt dt
⎡1 ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤
L
3 0
ia −R ia
⎢ L ⎥ ⎢ i ⎥ ⎢ −ωL ⎥ ⎢ i ⎥
⎢ d1 ⎥ ⎢ d ⎥ ⎢ d2 ⎥ ⎢ d ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ L q2 ⎥ d ⎢ i q ⎥ ⎢ ωL q1 ⎥ ⎢ i q ⎥
⎢ ⎥ ⎢ ⎥=⎢ ⎥⎢ ⎥ − ua (17)
⎢ L md1 ⎥ dt ⎢ i f d ⎥ ⎢ −ωL md2 ⎥ ⎢ i f d ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ L md1 ⎦ ⎣ i D ⎦ ⎣ −ωL ⎦ ⎣ i D ⎦
L mq2 iQ ωL mq1 iQ
u ∞b = u ∞c , i a = 0, i b = −i c
⎡ ⎤T ⎡ ⎤
⎡ ⎤T ⎡ ⎤ √2 R ib
3
L d2 id ⎢ ⎥ ⎢i ⎥
⎢ −L ⎥ ⎢ i ⎥ ⎢ ωL d1 ⎥ ⎢ d ⎥
⎢ q1 ⎥ ⎢ q ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ d⎢ ⎥ ⎢ ωL q2 ⎥ ⎢ iq ⎥
⎢ L md2 ⎥ . ⎢ i f d ⎥ = ⎢ ⎥ ⎢ ⎥ − ub + uc (19)
⎢ ⎥ dt ⎢ ⎥ ⎢ ωL md1 ⎥ ⎢i fd ⎥
⎣ L md2 ⎦ ⎣ iD ⎦ ⎢ ⎥ ⎢ ⎥
⎣ ωL md1 ⎦ ⎣ iD ⎦
−L mq1 iQ
ωL mq2 iQ
u ∞a = u ∞b = u ∞c = 0
3
Tem = p( − L md i md i q + L mq i d i mq ) (21)
2
The description of various parameters is given in [10].
Phase Synchronization Fault
In mal-synchronization fault, the power delivered by the synchronous generator is
different from required power at grid. For power plant to operate continuously the
1650 T. Kumar et al.
The complete dynamic model for line-to-ground and line-to-line faults are repre-
sented by set of coupled Eqs. (22), (24), (25) and for three-phase and mal-
synchronization faults are represented by set of coupled Eqs. (23)–(25). The coupled
turbine-generator equations are solved numerically using fourth-order Adams
predictor–corrector scheme with startup by fourth-order Runge–Kutta method in
MATLAB [10]. The elements of L and X matrices can be evaluated using equations
describing the different faults.
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
i phase i phase u fd
d ⎢ i fd ⎥ ⎢ ⎥ ⎢
⎥ + X⎢ i f d ⎥ = ⎢ 0 ⎥
⎥
L· ⎢ ⎣ ⎦ ⎣ ⎦ ⎣ (22)
dt iD iD 0 ⎦
iQ iQ 0
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
id id ud
⎢i ⎥ ⎢i ⎥ ⎢u ⎥
d⎢⎢
q ⎥
⎥
⎢ q ⎥ ⎢ q ⎥
⎢ ⎥ ⎢ ⎥
L · ⎢ i f d ⎥ + X⎢ i f d ⎥ = ⎢ u f d ⎥ (23)
dt ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ iD ⎦ ⎣ iD ⎦ ⎣ 0 ⎦
iQ iQ 0
J dω
= −Tem + Tdrive (24)
p dt
dγ
= −ω (25)
dt
The turbo-generator shaft is modeled using finite element method. The rotor is
divided into number of elements with two rotational dofs (single dof at each node).
The Hamilton’s principal is used to drive elemental stiffness and mass matrices.
Three types of sections, solid circular, hollow circular and tapered, are used to model
Big Turbo-Generator Shaft Vibrations Control … 1651
Fig. 3 Modified Bouc–Wen model for MR fluid damper (Left) and force–velocity plot [25]
the rotor elements. The elemental matrices are combined to form the global matrices,
and complete FEM equation is represented by Eq. (26).
[I ]g θ̈ g + [C]g θ̇ g + [K ]g {θ }g = T p g + [Tuniform ]g (26)
F = c1 ẏ + k1 (x − x0 ) (27)
where
1
ẏ = [αz + c0 ẋ + k0 (x − y)]
c0 + c1
ż = −γ |ẋ − ẏ|z|z|n−1 − β(ẋ − ẏ)|z|n + δ(ẋ − ẏ)
3 Validation
The dynamic model described in Sect. 2.3 is already validated by the author in ref
[10]. The values of inductance L e and resistance Re are assumed to be zero for
validation purpose only.
1652 T. Kumar et al.
Table 1 Comparison of first three natural frequencies from MATLAB and Ref. [19]
S. No. No. of elements First 3 natural First 3 natural % Error Avg. % error
considered in freqs. from freqs. of Ref. [19]
MATLAB MATLAB (×103 (×103 Hz)
Hz)
1 5 1.682 1.534 9.647 174.46
7.737 4.911 57.544
46.247 8.315 456.187
2 6 1.576 1.534 2.737 14.294
5.553 4.911 13.072
10.566 8.315 27.071
3 7 1.557 1.534 1.499 6.047
5.045 4.911 2.728
9.472 8.315 13.91
4 8 1.551 1.534 1.108 1.904
4.859 4.911 −1.058
8.786 8.315 5.664
5 9 1.547 1.534 0.847 −0.302
4.774 4.911 −2.789
8.401 8.315 1.034
6 10 1.547 1.534 0.847 0.0353
4.773 4.911 -2.810
8.487 8.315 2.068
Further, the finite element model for the rotor is validated using reference [19]. The
first three natural frequencies are calculated to validate the global mass and stiffness
matrices. The first three natural frequencies are compared with the reference [19], and
convergence of the code is established in Table 1. The first three natural frequencies
of the tube as given in reference [19] show close resemblance to the simulated results.
The MR fluid damper is validated for vehicle model given in ref [25]. Modified
Bouc–Wen Model is used to model the damper. Various parameters are given in
Reference [25]. From Fig. 4a, b, it is clear that there is close resemblance in results
of ref [25] and simulated results.
The generator is modeled using dq0 formulation. The complete system is connected
with the infinite load through grid with resistance Re and inductance, L e . Per
Big Turbo-Generator Shaft Vibrations Control … 1653
Fig. 4 a Comparison of mass m1 position as obtained from MATLAB simulations and refer-
ence [25] results. b Comparison of mass m2 position as obtained from MATLAB simulations and
reference [25] results
unit system is used to calculate the peak values of torque. The coupled turbine-
generator equations are solved numerically using fourth-order Adams predictor-
corrector scheme with startup by fourth order Runge–Kutta method in MATLAB.
The peak values of per unit torque for line to ground, line to line, three-phase and
mal-synchronization faults are 12 pu, 7.9 pu, 17.1 pu and 8.1 pu, respectively. The
actual value of electromagnetic torque is calculated by multiplying per unit value by
torque base 3 pUn In ωn−1 [9].
The rotor is modeled using finite element method. Complete rotor is divided in to
187 numbers of elements and 188 numbers of nodes. Each element has two degrees
of freedom, i.e., one rotational degree of freedom at each node. Since torsional
vibrations are independent from translational and axial vibrations so other DOFs are
neglected to keep the formulation simple. The whole rotor is modeled using three type
of elements (solid cylinder, hollow cylinder and tapered), and stiffness is calculated
accordingly. The generator torque is applied on nodes associated with the generator
position on rotor length. The complete dynamic system is solved using ODE15s in
1654 T. Kumar et al.
MATLAB. The diameter of various sections of coupled rotor along length is shown
in Fig. 5.
Further, MR fluid damper is used to damp the torsional vibration of turbo-
generator shaft. The damper is placed at element No. 111. The vibrations at element
No. 39 are simulated for various different faults. The various parameters of the
MR fluid damper are shown in Table 2. The comparative results for control and
uncontrolled system vibrations at element No 39 are shown in Fig. 6a–d.
The peak torque, valley torque and average amplitude of torque for first cycle
in control and uncontrolled condition during various electrical faults are given in
Table 3. The amplitude reduction for controlled system during line to ground faults
is 17.32%, for line-to-line fault is 8.57%, for three-phase fault is 4.57% and for mal-
synchronization fault is 40% at element No. 39. Further peak value of torque remains
same for line-to-ground fault and three-phase fault for controlled and uncontrolled
Fig. 6 a Torque at element 39 under one-phase fault. b Torque at element 39 under two-phase fault.
c Torque at element 39 under three-phase fault. d Torque at element 39 under mal-synchronization-
phase fault
Table 3 Various values of torque at element No 39 under controlled and uncontrolled conditions
Fault type Torque (×105 Nm) Uncontrolled MR fluid damper
Line to gound Peak 3.38 3.24
Valley −3.20 −2.19
Avg. torque amplitude 3.29 2.72
% Amplitude reduction – 17.32
Line to line Peak 1.77 0.557
Valley – 1.73 −2.65
Avg. torque amplitude 1.75 1.60
% Amplitude reduction – 8.57
Three phase Peak 4.79 4.89
Valley −4.39 −3.88
Avg. torque amplitude 4.59 4.38
% Amplitude reduction − 4.57
Mal-synchronization Peak 0.119 0.465
Valley −2.29 −0.989
Avg. torque amplitude 1.20 0.72
% Amplitude reduction – 40
1656 T. Kumar et al.
systems, but for other faults the peak value of torque increases extensively as shown
in Fig. 6d.
5 Conclusions
The dynamic models of loaded synchronous generator under four types of electric
faults are analyzed in MATLAB. The rotor is modeled using FEM, and complete
dynamic system is numerically simulated using ode15s in MATLAB. Further,
magnetic rheological fluid damper is employed at element No. 111, and torsional
vibrations at element 39 are simulated and compared with the uncontrolled system.
The MR fluid damper leads to different reduction of torsional vibration for different
faults. Passive MR fluid control effectively control line-to-ground and three-phase
faults. In line-to-ground fault, effect of damper is minimum. In case of mal-
synchronization fault, the average amplitude of the vibrations remains same, but
vibration peak values increase tremendously. Further, active MR fluid damper can
be studied to control the vibrations come during various electrical faults on the line.
References
1. Wachel JC, Szenasi FR (1993) Analysis of torsional vibrations in rotating machinery. In 22nd
Turbomachinery symposium, Dallas, TX, Sept, pp 14–16
2. Pasca N, Marsavina L, Muntean S, Negru R (2012) Failure analysis of a storage pump shaft
3. Shafts, transmitting. Design of power-transmitting shafts (1984)
4. Khangar VS, Jaju SB (2012) A review of various methodologies used for shaft failure analysis.
Int J Emerging Technol Adv Eng ISSN 2250-2459
5. Asi O (2006) Fatigue failure of a rear axle shaft of an automobile. Eng Fail Anal 13(8):1293–
1302
6. Reid MJ (1988) Analysis of the causes of recent roll shaft failures in natal sugar mills. In:
Proceedings of the annual congress-South African Sugar Technologists’ Association
7. Mitrovic, D. Stojanovic D. Petrovic N. “Analysis of Torsional Torques of Big Turbine-Generator
Shafts.”
8. Cudworth CJ, Smith JR (1990) Steam turbine generator shaft torque transients: a comparison
of simulated and test results. In: Generation, transmission and distribution, IEE Proceedings
C, vol 137, no 5. IET, pp 327–334
9. Lupsa-Tataru L (2009) Comparative simulation study on synchronous generators sudden short
circuits. Model Simul Eng 2009:8
10. Kumar T, Bangunde A, Kumar R, Jain SC (2018). The torque simulation of synchronous gener-
ator under loaded condition for different types of electrical disturbances. In: 2018 2nd Inter-
national conference on power, energy and environment: towards smart technology (ICEPE).
IEEE, pp 1–9. https://doi.org/10.1109/EPETSG.2018.8659271
11. Jackson MC, Umans SD (1980) Turbine-generator shaft torques and fatigue: Part III-
refinements to fatigue model and test results. IEEE Trans Power Apparatus Syst 3:1259–1268
12. Szenasi FR, Von Nimitz W (1978) Transient analyses of synchronous motor trains. In: 7th
Turbomachinery symposium, Houston, TX, vol 7, pp 111–117
13. Szolc T (2000) On the discrete-continuous modeling of rotor systems for the analysis of coupled
lateral torsional vibrations. Int J Rotating Mach 6(2):135–149
Big Turbo-Generator Shaft Vibrations Control … 1657
14. Taplak H, Parlak M (2012) Evaluation of gas turbine rotor dynamic analysis using the finite
element method. Measurement 45(5):1089–1097
15. Sakamoto S, Hirata T, Kobayashi T, Kajiwara K (1999) Vibration analysis considering higher
harmonics of electromagnetic forces for rotating electric machines. IEEE Trans Magnet
35(3):1662–1665
16. He Q, Du D (2010) Modeling and calculation analysis of torsional vibration for turbine
generator shafts. J Inf Comput Sci 7(10):2174–2182.
17. Chan DSH (1996) The transient torsional vibration behaviour of a turbine-generator system
under short circuit excitation. In: ASME 1996 international gas turbine and aeroengine congress
and exhibition. American Society of Mechanical Engineers, pp V003T07A001-V003T07A001
18. Vaziri A, Nayeb-Hashemi H (2006) A theoretical investigation on the vibrational characteristics
and torsional dynamic response of circumferentially cracked turbo-generator shafts. Int J Solids
Struct 43(14):4063–4081
19. Sung C-C, Varadan VV, Bao X-Q, Varadan VK (1994) Active torsional vibration control
experiments using shear-type piezoceramic sensors and actuators. J Intell Mater Syst Struct
5(3):436–442
20. Liu C, Jiang D, Chen J (2014) Coupled torsional vibration and fatigue damage of turbine
generator due to grid disturbance. J Eng Gas Turbines Power 136(6):062501
21. Oliquino R Jr, Islam S, Eren H (2003) Effects of types of faults on generator vibration signatures.
Curtin University of Technology, Western Australia
22. Tsai J-I, Lin C-H, Tsao T-P (2004) Assessment of long-term life expenditure for steam turbine
shafts due to noncharacteristic subharmonic currents in asynchronous links. IEEE Trans Power
Syst 19(1):507–516
23. Chan KS, Enright MP, Golden PJ, Naboulsi S, Chandra R, Pentz AC (2012) Probabilistic high-
cycle fretting fatigue assessment of gas turbine engine components. J Eng Gas Turbines Power
134(6):062502
24. Dyke SJ, Spencer BF Jr, Sain MK, Carlson JD (1997) On the efficacy of magnetorheological
dampers for seismic response reduction. In: Proceedings of the ASME design engineering
technical conferences
25. Butz T, Von Stryk O (2002) Modelling and simulation of electro-and magnetorheological fluid
dampers. ZAMM-J Appl Math Mech/Zeitschrift Für Angewandte Mathematik Und Mechanik:
Appl Math Mech 82(1):3–20
Antagonistic Actuation of Pneumatic
Artificial Muscle (PAM)
with Chain-Sprocket Mechanism
Abstract In this work, a system has been designed with the help of a chain-sprocket
mechanism to lift a load for particular applications with the help of a pneumatic
artificial muscle (PAM) actuator. The system will mimic the human arm where one
PAM and spring are in antagonistic arrangement just like the human biceps and
triceps. The PAM behaves as the biceps and the spring as the triceps muscle to
provide additional support to the system. The PAM will contract when the air pressure
is supplied and actuate the system to reach the desired position. With the help of
this system, the experiment can be performed to obtain various system parameters
for a wide range of the muscle. The system is designed focusing on its flexibility
for multiple types of experiments with different dimensions of pneumatic muscles.
Finally, the hysteresis present in the muscle has been plotted and compared to verify
the developed system. The developed system with the chain-sprocket mechanism is
very useful to understand the dynamics of the PAM, and as well as the designer and
researcher can use this for various applications in the field of medical and robotic
industries.
1 Introduction
Pneumatic artificial muscle has large applications in the field of medical robotics as
well as in industries over the years. Due to the operational similarity with the natural
muscles, PAM has attracted many researchers and designers for various applica-
tions which requires high force-to-weight ratios, high fatigue life and compliance [1,
2]. These artificial muscles are generally unidirectional which generate the pulling
force by inducing compressed air pressure into it. PAM consists of a core of elastic
tube cover with spiral network braid which limits the expansion of the muscle with
an increase in the air pressure. Generally, the materials used for the core of the
muscle are latex and silicon rubber where for the braid, fibers like nylon are used [3].
The operation of such kind of muscle basically depends on the shape, contraction
behavior, generated tensile force along with the materials and geometry of the core
and the braid of the muscle [4]. These factors of the artificial muscles for various
applications have motivated an investigation for optimizing the performance of the
operating principles of these PAMs.
Most of the medical, robotic and industrial applications need the bidirectional
linear or rotational motion. So, to obtain the required motion, there is a need for
an antagonistic arrangement of two artificial muscles similar to the biceps–triceps
muscle arrangement of the human arm. These biceps and triceps muscle can be
classified as flexor and extensor muscles to move the human hand. Biceps is included
in the flexor group, and it will bend the arm by decreasing the angle between the
forearm and upper arm, whereas triceps muscle is included in the extensor muscle. In
flexion condition, biceps muscle will contract while triceps muscle will relax, while
in extension condition, the biceps will relax and triceps contracted. The length of
muscle is shorter, and the contracting force becomes smaller [5]. PAMs consist of
the core and the braid mimics the same principle like human muscles and the air
pressure as its source of actuating force. This antagonistic arrangement is mostly
employed with conventional actuators like hydraulic or pneumatic cylinders along
with newer smart materials, such as shape-memory alloys [6] and electrostrictive
materials [7]. Due to the advantages of PAM over the other actuators, in various
applications like aerospace [8], 3D printing [9], rehabilitation [10] where the artificial
muscle is used to obtain the antagonistic actuation. Balara and Tóthová [11] described
about the static and dynamic properties of the PAM with the help of antagonistic
arrangement experimental rig. Vocke et. al. [12] designed a mechanism with the
help of the PAM to achieve the bidirectional motion. But due to the presence of the
highly nonlinear force-contraction characteristic and compliance in PAM, there is a
challenge to achieve the antagonistic actuation.
Many researches have been done on various arrangements attached with the PAM
to achieve antagonistic actuation with a wide range of actuator motion along with the
control of the stiffness of the actuator. These arrangements involve simple pulleys
[13, 14], variable complex radius pulleys [15] and offset lever mechanisms [16, 17] to
modify the system kinematics. Due to the advantage of this antagonistic arrangement
to achieve a wide range motion of the muscle actuator, a system has been designed
and developed in this work. The system has been attached with a chain-sprocket
mechanism with two artificial muscles which provided the required rotational as
well as the linear motion for a particular application. A detailed study about the idea
and flexibility of the system with its attachments has been done in the next section
followed by the justification with typical hysteresis plot for different loads with the
developed system.
Antagonistic Actuation of Pneumatic Artificial Muscle (PAM) … 1661
A system has been designed to enable the antagonistic arrangement of PAM mimic
like the biceps and triceps of the human arm as shown in Fig. 1. The system can be
efficiently used to test the PAM with different lengths and types. Figure 2 describes
the schematic diagram of a system where one PAM which mimics like the triceps
of the human arm can be replaced with a spring to simplify the system. In the
developed system, these both types of experiment can be performed to understand
the working of the PAM and its effect on the application. Aluminum slotted extrusions
are used primarily to build a lightweight and stable frame. To fix the muscles as the
arrangement of a human arm as shown in Fig. 1, four snap hooks (6) have been
attached as shown in Fig. 3. A chain-sprocket mechanism is designed as shown in
Fig. 4 which is adjustable with the required dimension of the PAM. To actuate the
muscles mounted on the frame, compressed air is supplied through the 4 mm tube
(9). The frame also has two manual airflow regulators (3, 4) which enables controlled
air supply to individual muscles.
Fig. 2 An arrangement of
spring-muscle system
(24) made of acrylic sheet. An idle gear is also mounted for smooth movement of
the rack. For easy mounting of the muscles and springs, four snap hooks (6) of steel
material are used. Two hooks are mounted on the top portion of the frame with two
M5 Allen bolts (5) as shown in Fig. 3. Other two hooks are tied to two ends of a chain
(16). The chain roles over a sprocket (17) of teeth = 36 nos. which is mounted on the
shaft (13) and fixed with grub screw. To measure angular deflection while inflating
the muscles, the value can be recorded from the protractor (15) as shown in Fig. 3.
The rear end of the frame also has a scale (22) with which the linear movement of
the rack (25) can be observed and recorded simultaneously. This linear displacement
corresponds to the change in muscle length while extension/contraction. The CAD
models are shown in Fig. 5a where only one PAM is in the antagonistic arrangement
with a spring, and Fig. 5b depicts that two PAMs are attached in the system.
In the case of using only one muscle, a nonlinear tension spring is used which will
play role of a secondary muscle as shown in Fig. 5a. In Fig. 6a, the PAM of length
L 1 is mounted between the hooks of the frame and the chain. To the other end of
the chain, a nonlinear spring of length S 1 is mounted. Then, the movable platform is
adjusted by distance, H x and set to an initial tensioning value T 1 . A mass of weight,
W, is loaded to the arm and placed at fix distance, R, with the help of wing nuts.
Now, while supplying the compressed air to the muscle L 1 , it will make the muscle
to inflate to a length L 2 which in turn rotates the sprocket by an angle, δ and increase
the length of spring to length, S 2 as shown in Fig. 6b. The new tension, T 2 , on the
chain will depend on the value of weight, W, radius, R, and the deflection angle, δ.
Antagonistic Actuation of Pneumatic Artificial Muscle (PAM) … 1663
The pressure on the muscle while deflecting can be recorded from the dial indicator
corresponding to the muscle. The same motion can be achieved in both directions
with the help of two PAMs attached to the system as shown in Fig. 5b for the required
application.
The developed system has been shown in Fig. 7 with front and rear view to
conducting various experiments to understand the behavior of the PAM. The typical
hysteresis for the PAM has been observed which is provided proper justification
to the system. The hysteresis present in the muscle has been plotted with the help
of contraction in the muscle with the variation in pressure for a particular load. In
1664 B. Kalita et al.
Fig. 8a, b, the typical hysteresis has been shown in 15 N and 30 N load ,respectively,
subjected to the system. These plots are almost similar to the hysteresis plots which
is found in the previous literature [18, 19]. With the help of these plots, different
system parameters present in the PAM can be obtained and used for the study of
nonlinear nature exhibiting in the muscle. The safe operating range of different
system parameters can be obtained from the mathematical work proposed by Kalita
and Dwivedy [20, 21] for various resonance conditions.
Antagonistic Actuation of Pneumatic Artificial Muscle (PAM) … 1665
Fig. 5 CAD model of the system with a one PAM and a spring b two PAMs in the antagonistic
arrangement
3 Conclusion
The developed system with the chain-sprocket mechanism can be used for testing
a wide range of muscle available in the market before use in the applications. The
system has been designed keeping in mind some major factors like the flexibility,
robustness, weight and uses. The principle of the antagonistic arrangement can be
easily understood with the help of this design. Various nonlinear behaviors present
in the muscle can be studied by performing experiments on this system. The system
can provide the angular motion along with the linear motion simultaneously by
attaching an arrangement of rack and pinion in the system. The configuration of the
designed chain-sprocket mechanism with the PAM requires less control to achieve
the demands like desired load and position along with the fewer implementation
costs. The behavior of the muscles is completely different for different values of the
1666 B. Kalita et al.
Fig. 6 Experimental observation of the antagonistic arrangement with one PAM and and the spring
system load. So, to justify the developed system, hysteresis present in the muscle has
been plotted for different loads and shows the typical behavior. Hence, the system is
very useful for various applications in the field of health care and industries.
Antagonistic Actuation of Pneumatic Artificial Muscle (PAM) … 1667
Fig. 7 Actual developed system with a front view and b rear view
(a) (b)
12 10
10 8
8
6
6
4
4
2
2
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Fig. 8 Typical hysteresis present in the used PAM in the developed system for a 15 N and b 30 N
References
1. Tondu B, Lopez P (1997) The McKibben muscle and its use in actuating robot-arms showing
similarities with human arm behaviour. Ind Robot 24(6):432–439
1668 B. Kalita et al.
2. Klute G, Czerniecki J, Hannaford B (2002) Artificial muscles: actuators for biorobotic systems.
Int J Robot Res 21(4):295–309
3. Daerden F, Lefeber D (2002) Pneumatic artificial muscles: actuators for robotics and
automation. Eur J Mech Environ Eng 47(1):11–21
4. Tóthová M, Pitel J (2013) Dynamic model of pneumatic actuator based on advanced geometric
muscle model. In: 9th International conference on computational cybernetics (ICCC). IEEE,
Tihany, Hungary, pp 8–87
5. Lightner S, Lincoln R (2002) The fluidic muscle: a ‘new’ development. Int J Mod Eng 2(2):1–8
6. Hirose S, Ikuta K, Sato K (1989) Development of shape memory alloy actuator. improvement
of output performance by the introduction of a σ-mechanism. J Ad Robot 3(2):89108
7. Szufnarowski F, Schneider A (2010) Compliant piezo-flexdrives for muscle-like, antago-
nistic actuation of robot joints. In: International conference on biomedical robotics and
biomechatronics. IEEE, Tokyo, Japan, pp 381–388
8. Wereley N, Kothera C, Bubert E, Woods B, Gentry M, Vocke R (2009) Pneumatic artificial
muscles for aerospace applications. In: 50th structures, structural dynamics, and materials
conference 17th adaptive structures conference, p. 2140. AIAA, Palm Springs, California
9. Peele BN, Wallin TJ, Zhao H, Shepherd RF (2015) 3D printing antagonistic systems of artificial
muscle using projection stereolithography. Bioinspir Biomim 10(5):055003
10. Knestel M, Hofer EP, Barillas SK, Rupp R (2008) The artificial muscle as an innovative actuator
in rehabilitation robotics. IFAC Proc Vol 41(2):773–778
11. Balara M, Tóthová M (2012) Static and dynamic properties of the pneumatic actuator with arti-
ficial muscles. In: 10th Jubilee international symposium on intelligent systems and informatics.
IEEE, Subotica, Serbia, pp 577–581
12. Vocke RD III, Kothera CS, Wereley NM (2014) Mechanism and bias considerations for design
of a bi-directional pneumatic artificial muscle actuator. Smart Mater Struct 23(12):125039
13. Tondu B, Ippolito S, Guiochet J, Daidie A (2005) A Seven-degrees-of-freedom robot-arm
driven by pneumatic artificial muscles for humanoid robots. Int J Robot Res 24(4):257–274
14. Shin D, Sardellitti I, Khatib O (2008) A hybrid actuation approach for human-friendly robot
design. In: International conference on robotics and automation. IEEE, Kobe, Japan, pp 1747–
1752
15. Shin D, Yeh X, Khatib O (2011) Variable radius pulley design methodology for pneumatic arti-
ficial muscle-based antagonistic actuation systems. In: International conference on intelligent
robots and systems. IEEE, San Francisco, CA, USA, pp 1830–1835
16. Woods BKS, Kothera CS, Wereley NM (2011) Wind tunnel testing of a helicopter rotor trailing
edge flap actuated via pneumatic artificial muscles. J Intell Mater Syst Struct 22(13):1513–1528
17. Vocke III RD, Kothera CS, Chaudhuri A, Woods BKS, Wereley NM Design and testing of a
high-specific work actuator using miniature pneumatic artificial muscles. J Intell Mater Syst
Struct 23(3):365–378
18. Zang X, Liu Y, Heng S, Lin Z, Zhao J (2017) Position control of a single pneumatic artificial
muscle with hysteresis compesnsation based on modified Prandtl-Ishlinskii model. Bio-Med
Mater Eng 28(2):131–140
19. Li H, Kawashima K, Tadano K, Ganguly S, Nakano S (2011) Achieving haptic perception
in forceps’ manipulator using pneumatic artificial muscle. IEEE/ASME Trans Mechatron
18(1):74–85
20. Kalita B, Dwivedy SK (2019) Nonlinear dynamics of a parametrically excited pneumatic
artificial muscle (PAM) actuator with simultaneous resonance condition. Mech Mach Theory
135:281–297
21. Kalita B, Dwivedy SK (2019) Dynamic analysis of pneumatic artificial muscle (PAM) actuator
for rehabilitation with principal parametric resonance condition. Nonlinear Dyn 97(4):2271–
2289
Deep Neural Network Approach
for the Prediction of Journal Bearing
Static Performance Characteristics
1 Introduction
Deep neural network is used to solve complex problems and has been proved as a
powerful tool for most precise and accurate predictions of system behaviour [1, 2].
This approach can be used in journal bearing performance predictions. There is no
available literature which shows the mapping of FEM results with neural network
model to predict the performance of journal bearings. Theoretical investigation for
analysing performance characteristics of journal bearing is more time-consuming
process. Many calculations are required to obtain the desired characteristics. By
using the literature data, predictions can be made for the performance characteristics
by neural network approach.
Sinanoğlu investigated the pressure variations at the circumference of journal
bearing experimentally. The results obtained by neural network model were found
very close to experimental values [3]. Neural network model was successfully used
for the fault detection of bearings and their prevention [4, 5]. Only one hidden layer
was considered by researchers in the previous work for the predictions of journal
bearing performance using neural network [1, 3]. But, in this work, multiple hidden
layers along with multiple input and output data sets are considered for more accurate
predictions.
Hybrid journal bearing is considered for FEM analysis in the present work due to
its vast applications in heavy load and high-speed machines [6, 7]. Static and dynamic
performance characteristics of journal bearings using different restrictors can be
analysed theoretically and experimentally [8–10]. Their results can be used as target
output for the performance predictions using neural network. Orifice, capillary and
constant flow valve restrictors are used to control the flow through journal bearings.
Many researchers analysed the hybrid journal bearings using orifice restrictor as a
flow control device [11, 12]. FEM was used for theoretical analysis [13, 14].
In the present work, finite element method (FEM) is used
to obtain journal bearing
performance characteristics. Maximum pressure p max and minimum fluid film
thickness h min are used as target output for feedforward backpropagation neural
network model. Hidden layers and number of neurons in these layers are decided
heuristically. Logistic activation function is used in hidden and output layers. Predic-
tions for journal bearing performance are made within and out of the prescribed range
of input parameters. The use of multiple hidden layers for journal bearing perfor-
mance predictions and multiple data sets for input and output neurons is the novelty
of the present work.
Journal bearings have been analysed by many researchers to determine their static
and dynamic performance characteristics [8, 9]. In the present FEM analysis, the
flow field is discretised as shown in Fig. 1. The number of nodes is taken as 48, and
the number of four-noded elements is 36. The number of holes is considered as 24
(12 per row) in hole-entry hybrid journal bearing in the present work.
The governing Reynolds equation in non-dimensional form for lubricant flow
field in journal bearing is given as Eq. (1) [15–17];
∂ 3 ∂p ∂ 3 ∂p ∂ F1 ∂h
h F2 + h F2 = 1− h + (1)
∂α ∂α ∂β ∂β ∂α F0 ∂t
Deep Neural Network Approach for the Prediction … 1671
nle
p= pj Nj (2)
j=1
Fluid film thickness can be obtained by the expression given in Eq. (3);
Table 1 Dimensionless
Parameters Value/range
operating and geometric
parameters for FEM analysis External load (w) 0.2–1.4
Speed parameter () 1.0, 1.2, 1.4
Restrictor design parameter C s2 0.1
L/D ratio 1.0
Land width ratio (ab ) 0.25
Restrictor Orifice
⎡ ⎤⎧ ⎫ ⎧ ⎫ ⎧ ⎫
F 11 F 12 ··· F1j ··· F 1n ⎪ ⎪ p1 ⎪⎪ ⎪
⎪ Q1 ⎪ ⎪ ⎪
⎪ R H1 ⎪ ⎪
⎢ . ⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎢ .. .. .. .. .. .. ⎥ ⎥
⎪
⎪
⎪ .. ⎪
⎪
⎪
⎪
⎪
⎪ .. ⎪ ⎪
⎪
⎪
⎪
⎪ .. ⎪ ⎪
⎪
⎢ . . . . . ⎥⎪ ⎪ . ⎪ ⎪ ⎪
⎪ . ⎪
⎪ ⎪
⎪ . ⎪
⎪
⎢ ⎥⎪⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎢ F i1 F i2 ··· Fi j ··· F in ⎥⎪ ⎪ pi ⎪⎪ ⎪
⎪ Q ⎪
⎪ ⎪
⎪ R ⎪
⎪
⎢ . ⎥ ⎨ ⎬ ⎨ i ⎬ ⎨ H i ⎬
⎢ . .. .. .. .. .. ⎥ .. = .
. + .
.
⎢ . . . . . . ⎥⎪ . ⎪ ⎪ . ⎪ ⎪ . ⎪
⎢ ⎥⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎢ F j1 ··· ··· F jn ⎥⎪ ⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ Hj ⎪
⎪
⎥⎪ j⎪ ⎪ ⎪ ⎪ ⎪
F j2 F jj p Q R
⎢ ⎪
⎪ ⎪
⎪ ⎪
⎪
j ⎪
⎪ ⎪
⎪ ⎪
⎢ .. .. .. .. .. .. ⎥⎪⎪ .. ⎪⎪ ⎪ ⎪ .. ⎪ ⎪ ⎪ .. ⎪
⎪ ⎪
⎪
⎣ . . . . . . ⎦ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎪ . ⎪
⎪
⎩ ⎪ ⎪
⎭ ⎪ . ⎪
⎩ ⎪
⎭ ⎪ . ⎪
⎪
⎩ ⎪
⎭
F n1 F n2 · · · F n j · · · F nn pn Qn R Hn
⎧ ⎫ ⎧ ⎫
⎪
⎪ R x1 ⎪⎪ ⎪
⎪ R z1 ⎪ ⎪
⎪
⎪ ⎪ ⎪ ⎪
⎪
⎪ .. ⎪ ⎪
⎪
⎪
⎪
⎪ .. ⎪ ⎪
⎪
⎪
⎪ . ⎪ ⎪
⎪
⎪
⎪
⎪ . ⎪ ⎪
⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ R
⎨ i⎬ x ⎪ ⎪
⎨ i⎪
R z ⎬
+ Ẋ J .. + Ż .. (6)
⎪ . ⎪ J
⎪ . ⎪
⎪
⎪ ⎪ ⎪ ⎪
⎪ Rx j ⎪
⎪ ⎪
⎪
⎪
⎪ Rz j ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪ ⎪ ⎪
⎪ .. ⎪
⎪ ⎪
⎪
⎪
⎪ .. ⎪
⎪ ⎪
⎪
⎪
⎪ . ⎪
⎪ ⎪
⎪ . ⎪
⎪
⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
R xn R zn
The boundary conditions for the elastic deformation flow field are as follows:
i. There is an ambient pressure at bearing edge.
ii. At internal nodes, flow is observed to be zero.
iii. Flow is nonzero on holes and external boundaries.
iv. Restrictor flow and bearing input flow are identical.
v. At trailing edge of the positive region, p = ∂∂αp = 0.
Non-dimensional operating and geometric parameters used in FEM analysis are
presented in Table 1:
The journal bearing solution procedure using FEM is shown as a flowchart in Fig. 2.
Deep Neural Network Approach for the Prediction … 1673
The input data is computed to achieve fluid film thickness for initial journal centre
coordinate position. Fluidity matrix is generated and solved for the Newtonian lubri-
cant. Pressure values are obtained using boundary conditions for lubricant flow field.
Iterative procedure is done to achieve journal equilibrium for applied radial load (w).
Increments in journal centre coordinates are computed to achieve the convergence
criteria, i.e. PERR < 0.001. Finally, static performance characteristics are obtained
for the journal bearing. Scilab software is used in the present work for FEM anal-
ysis and DNN predictions. It provides the computational environment for matrices
manipulation and algorithm implementation.
The values of p max and h min are obtained using FEM and considered as target output
for training and testing a feedforward neural network model. Backpropagation algo-
rithm is used for minimising the error and updating the weights [1, 3]. Weights are
selected initially as random numbers.
The deep neural network is multilayer neural network which contains more than
one hidden layers. This approach is used for the solution of complex problems. Large
number of input data is required for the processing. The number of neurons in each
hidden layer can be decided heuristically whereas the neurons in input and output
layers are fixed depending upon the problem. Feedforward deep neural network
1674 S. Kumar et al.
1
vi1 = (8)
1 + e−h i1
1
vi2 = (10)
1 + e−h i2
Input to the output layer h i3 is given
as Eq. (11). Using logistic activation
function, the response of output layer yin is obtained and given as Eq. (12):
n2
h i3 = win i2 vi2 (11)
i 2 =1
1
yin = (12)
1 + e−h i3
The associated weights between layers are updated using the expression (13) to
minimise the error. Mean square error (E) is obtained by using Eq. (14):
∂E
w(new) = w(old) − η + αw (13)
∂w
n2
1 2
E= yiTn − yin (14)
2i
n =1
where yiTn is the target output and yin is the neural network model output. (η)
is the learning rate, and its value should be in the range 0–1. (α) is the momentum
term used to improve training speed and accuracy. The value should be in the range
0–0.9999. A bias which is a positive value can be added in hidden and output layer
neurons.
Deep neural network for journal bearing analysis is shown in Fig. 4. In this figure,
radial load (w) and rotational
speed (Ω) are considered as inputparameters,
whereas
maximum pressure p max and minimum fluid film thickness h min are the output
parameters. A bias (+1) is connected with all the neurons of hidden layers to get
the nonzero value during the iteration process. Both weights and bias are updated in
each run.
The parameters for the neural network predictor are shown in Table 2. In this table,
(η) is the learning rate which shows the speed of learning for the error convergence.
(n i ), (n h1 ), (n h2 ) and (n o ) represent neurons in input layer, first hidden layer, second
hidden layer and output layer, respectively. (α) shows the momentum term and (N)
shows the number of epochs. Epochs are the training cycles performed in the program
1676 S. Kumar et al.
to achieve the convergence criteria. The initial weights are chosen randomly. Hence,
the number of iterations to converge the error varies each time when running the
programme because, initially, the weights are chosen as random values. One pattern
each time is computed during iteration process.
It is observed that p max values linearly increase with increase in w, but due to the
increase in Ω, the p max values decrease for the same values of w.
The graph for minimum fluid film thickness h min versus radial load (w) is shown
in Fig. 7. The range for radial load (w) is 0.2–1.4, and the values of speed parameter
(Ω) are 1, 1.2 and 1.4. It is observed that h min values linearly decrease with increase
in w, but due to the increase in Ω, the h min values increase for the same values of w.
1678 S. Kumar et al.
Fifteen sets of input parameters (radial load w and speed Ω) are considered for
deep neural network model training as listed in Table 3.The ranges
for w and Ω
are 0.2–1.4 and 1.0–1.4, respectively. Maximum pressure p max and minimum fluid
film thickness h min are considered as output parameters. The target output and
neural network model output are compared, and error is calculated. The calculated
sum of squares error is obtained as 0.0000934. The percentage error observed for
p max is −0.51 to 0.39 and for h min is −0.47 to 0.62 during training of the model.
Normalisation of input data set and output target values is not required because all
the values are already in the range of 0–1.
Neural network model is tested for six input data sets, and the results are presented
in Table 4. The error observed for maximum pressure as −0.27 to 0.26 and for
minimum fluid film thickness as −0.40 to 0.20 during testing the neural network
model.
Predictions are made for the journal bearing performance within the prescribed
range of load (0.2–1.4), and the percentage error is presented in Table 5. The error
is calculated and found in the range of −0.87 to 0.03 for maximum pressure and −
0.94 to 0.23 for minimum fluid film thickness. The neural network predictions out
of prescribed range of load are presented in Table 6.
1680 S. Kumar et al.
This deep neural network model is suitable for predictions of journal bearing
performance within and out of the prescribed range. By using feedforward back-
propagation neural network model, very precise and accurate results are obtained for
all the predictions.
DNN approach was used extensively in recent years due to its excellence in
detecting the faults and improving the life span of journal bearings [5, 19, 20]. It
gives more accurate results compared to ANN. The number of iterations used to
converge the error should be less to save the convergence time. In the past, 100,000–
500,000 number of iterations were required to achieve the convergence criteria [1,
3, 21], but now, the same problem can be solved within 20,000 iterations for the
same value of percentage error due to the advancement in computational technology.
Hence, the time saving means getting the solution converged in comparatively less
number of iterations.
6 Conclusion
Mapping of the patterns in deep neural network found to be excellent for journal
bearing performance predictions. Feedforward backpropagation algorithm handles
the complexity of the problem in 15,000 iterations and gives the more accurate and
precise results. The neural network model predictions are found to be very near
to target values. The error is calculated for deep neural network and found to be
very small (−1.0% < error < 1.0%) during training and testing of the model. The
predictions are made within and out of the prescribed range with very high accuracy.
This deep neural network approach is perfect for performance predictions within and
out of the prescribed range of input parameters. It is concluded that lot of time is
saved to achieve the convergence criteria using DNN compared to the mathematical
analysis of the journal bearing performance.
Deep Neural Network Approach for the Prediction … 1681
Appendix
References
1. Sinanoğlu C (2006) A neural predictor to analyse the effects of metal matrix composite structure
(6063 Al/SiCp MMC) on journal bearing. Industr Lubr Tribol 58(2):95–109
2. Patel PM, Prajapati JM (2011) A review on artificial intelligent system for bearing condition
monitoring. Int J Eng Sci Technol 3(2):1520–1525
3. Sinanoğlu C, Kurban AO, Yildirim Ş (2004) Analysis of pressure variations on journal bearing
system using artificial neural network. Industr Lubr Tribol 56(2):74–87
4. Samanta B, Al-Balushi KR, Al-Araimi SA (2003) Artificial neural networks and support vector
machines with genetic algorithm for bearing fault detection. Eng Appl Artif Intell 16(7–8):657–
665
5. Liu R, Yang B, Zio E, Chen X (2018) Artificial intelligence for fault diagnosis of rotating
machinery: a review. Mech Syst Sign Process 108:33–47
6. Garg HC, Sharda HB, Kumar V (2006) On the design and development of hybrid journal
bearings: a review. Tribotest 12(1):1–19
7. Garg HC, Kumar V (2013) Static performance characteristics of hybrid journal bearings with
plugged entry holes. Industr Lubr Tribol 65(5):333–340
8. Dhawan R, Verma S (2014) Analyzing micropolar lubrication in noncircular hybrid journal
bearings. Tribol Trans 57(2):182–189
9. Kumar R, Verma S (2016) Effect of micropolar lubrication in non-circular hole-entry hybrid
journal bearing with constant flow valve restrictor. Industr Lubr Tribol 68(6):737–751
1682 S. Kumar et al.
10. Khatak P, Garg HC (2018) Performance comparison of hole-entry and slot entry hybrid journal
bearings considering combined influence of thermal effects and micropolar lubricant. Industr
Lubr Tribol 70(6):1037–1050
11. Nicodemus ER, Sharma SC (2011) Orifice compensated multirecess hydrostatic/hybrid journal
bearing system of various geometric shapes of recess operating with micropolar lubricant.
Tribol Int 44(3):284–296
12. Ram N, Sharma SC (2012) Analysis of orifice compensated non-recessed hole-entry hybrid
journal bearing operating with micropolar lubricants. Tribol Int 52:132–143
13. Kucinschi BR, Fillon M, Fre J, Pascovici MD (2000) A transient thermo elasto hydrodynamic
study of steadily loaded plain journal bearings using finite element method analysis. J Tribol
122(1):219–226
14. Khatri CB, Sharma SC (2017) Behaviour of two-lobe hole-entry hybrid journal bearing system
under the combined influence of textured surface and micropolar lubricant. Industr Lubr Tribol
69(6):844–862
15. Dowson D (1962) A generalized Reynolds equation for fluid-film lubrication. Int J Mech Sci
4(2):159–170
16. Fowles PE (1970) A simpler form of the general Reynolds equation. J Lubr Technol 92(4):661–
662
17. Crosby WA, Chetti B (2009) The static and dynamic characteristics of a two-lobe journal
bearing lubricated with couple-stress fluid. Tribol Trans 53(2):262–268
18. Sharma SC, Kumar V, Jain SC, Nagaraju T (2003) Study of hole-entry hybrid journal bearing
system considering combined influence of thermal and elastic effects. Tribol Int 36(12):903–
920
19. Thamba NB, Aravind A, Rakesh A, Jahzan M, Duraiswamy RP, Mangalaraja RV (2018)
Automatic fault classification for journal bearings using ANN and DNN. Arch Acoust
43(4):727–738
20. Ren L, Sun Y, Cui J, Zhang L (2018) Bearing remaining useful life prediction based on deep
autoencoder and deep neural networks. J Manuf Syst 48:71–77
21. Sinanoğlu C (2009) Design of neural model for analysing journal bearings considering effects
of transverse and longitudinal profile. Industr Lubr Tribol 61(3):132–139
Kinematics and Foldability Analysis
of Bennett Mechanisms and Its Networks
1 Introduction
In 1903, G. T. Bennett invented a spatial mechanism with four links and four joints
[1]. All four links are twisted to make it a spatial mechanism. It has a closed-loop kine-
matic structure. Due to its special geometrical properties, researchers have tried using
it in applications like deployable structures, foldable tends, liquid mixing machines,
etc.
Two types of Bennett 4R mechanisms were identified. In this research, networks
were made using these two types of mechanisms using MATLAB SimMechanics
Type 1 Bennett 4R mechanism has four links such that their opposite links are of
equal length. It also has four joints. Its graphic diagram is shown in Fig. 1 [2].
As shown in Fig. 1, ABCD is a type 1 Bennett 4R mechanism having non-parallel
and non-intersecting revolute joint axes. A, B, C and D are the revolute joints in the
mechanism. The two opposite links of type 1 Bennett 4R mechanism have same twist
angle and length [2].
DC = l3 = AB = l1 , AD = l4 = BC = l2 ,
DC = AB = α, AD = BC = β,
where l1 , l 2 , l 3, l 4 represent lengths of the links AB, BC, CD and AD. ‘α’ represents
twist angles in the links AB and DC and ‘β’ represents twist angles in the links AD
and BC. Type 1 Bennett 4R mechanism exists only if one more condition is satisfied.
That is, the opposite link lengths and twist angles should satisfy Eq. (1) [2]
sin α sin β
= (1)
l1 l2
It is also known by the name of equilateral Bennett mechanism. Equations for its
existence are given by Eq. (2) [3]
α+β =π
l1 = l2 = l3 = l4 (2)
4 Coupler Curves
The link which connects input link (AD) and output link (BC) of the Bennett 4R
mechanism is known as coupler link (CD). The path traced by a point in the coupler
link during its motion is known as coupler curve. It is very important to find this path,
because this path denotes the output for the application used by the mechanism. It also
helps to solve rigid-body guidance and path generation problems [4]. The revolute
joint coordinates for the Bennett 4R mechanisms are obtained as [2].
To plot the coupler curve for type 1 and type 2 Bennett 4R mechanisms, the
revolute joint coordinates for joint C are considered. Now, the input angle ‘θ ’ and
the output angle ‘η’ are obtained using Eq. (3) [2].
50
z (m)
-50
100
50 300
0 250
200
y (m) -50 150
-100 50 100
x (m)
50
z (m)
0
-50
100
50 200
0 150
-50 100
y (m) -100 0 50
x (m)
⎧ ⎫
⎪ θ ⎪
⎨ 2π + 2 arctan K tan 2 , θ ∈ [0, π ]⎪
⎪ ⎬
η= (3)
⎪
⎪ θ ⎪
⎪
⎩ 2 arctan K tan , θ ∈ [π, 2π ] ⎭
2
where
sin(β ± α)
K =− (4)
sin(α) − sin(β)
For plotting the above coupler curve (Fig. 2) for type 1 Bennett 4R mechanism, the
following values are considered. α = 30◦ ∗3.14/180 rad, β = 17.46◦ ∗3.14/180 rad,
K = +3.68 (K+ve taken), l1 = 167.7 mm and l2 = 100 mm.
Coupler curve for type 2 Bennett 4R mechanism is shown in Fig. 3.
For plotting the above coupler curve for type 2 Bennett 4R mechanism, the
following values were considered α = 30◦ ∗3.14/180 rad, β = 150◦ ∗3.14/180 rad,
K = 1.15 (K+ve taken), l1 = 100 mm and l2 = 100 mm.
5 Bennett Network
A single Bennett 4R mechanism was taken, and its links were extended to connect
eight other Bennett 4R mechanisms to form a network as shown in Fig. 4.
Black colored lines in Fig. 4 represent extended links. Red color represents the
links of eight Bennett 4R mechanisms which are connected to it. The arrangement
of network is made in such a manner that actuation needs to be given to the input
link of the Bennett 4R mechanism placed at the center (numbered 1 in Fig. 4) so that
the entire network gets actuated. The four extended links of the mechanism 1 divide
the whole network into four regions, I, II, III and IV. Type 1 and type 2 Bennett 4R
mechanisms are used to construct two different networks.
Kinematics and Foldability Analysis of Bennett Mechanisms … 1687
7 Results
Foldability test was conducted on type 1 and type 2 Bennett 4R mechanisms. The
coordinates of Bennett 4R mechanism at various input angles (θ ) were calculated. The
configuration of the type 1 Bennett 4R mechanism at θ = 12◦ and θ = 131◦ is given
below in Fig. 5a, b. There is singular configuration before and after the specified
minimum and maximum input angle. The configuration at which the mechanism
losses its degree-of-freedom, and there is no motion possible to the next position.
Fig. 7 a Input angle (θ) at 12° for type 1 Bennett RRRS mechanism, b input angle (θ) at 131° for
type 1 Bennett RRRS mechanism
Fig. 8 a Input angle (θ) at 1° for type 2 Bennett RRRS mechanism, b input angle (θ) at 147° for
type 2 Bennett RRRS mechanism
Kinematics and Foldability Analysis of Bennett Mechanisms … 1689
Fig. 9 a Network at θ = 30◦ for type 1 Bennett 4R mechanism, b network at θ = 130◦ for type 1
Bennett 4R mechanism
type 1 Bennett RRRS mechanism after folding and expanding. There is singular
configuration before and after the specified minimum and maximum input angle.
The configuration at which the mechanism losses its degree-of-freedom, there is no
motion possible to the next position.
It is to be noted that the Bennett mechanism with spherical joints (RRRS) has
the same foldability like the Bennett 4R mechanism. Hence, type 2 Bennett RRRS
mechanisms can also be used in place of type 2 Bennett 4R mechanism where compact
folding is required. The results are same because singularity happens on revolute joint
which is not replaced by spherical joint. Once this is also replaced, a better result
can be obtained due to three degree-of-freedom motion of spherical joint. This also
indicates that the Bennett RRRS mechanism has the same properties like Bennett
4R mechanism as mentioned by Morgan V. Brown and Paul Milenkovic [5].The
coordinates of all the joints in the network are calculated for type 1 Bennett 4R
mechanisms for two different input angles of θ (θ = 30◦ , 130°). The network was
modeled for these configurations in MATLAB SimMechanics, and different shapes
of the network are obtained as shown in Fig. 9a, b. There is singular configuration
before and after the specified minimum and maximum input angle. The configuration
at which the mechanism losses its degree-of-freedom, there is no motion possible to
the next position.
The coordinates of all the joints are calculated for different input angles of θ
(θ = 30◦ , 121°) for type 2 Bennett mechanism. The network was modeled for these
configurations in MATLAB SimMechanics, and different shapes of the network are
obtained as shown in Fig. 10a, b.
From Figs. 9 and 10, it is found that the input angle working range differs for
both type 1 and type 2 Bennett 4R mechanisms. Input angle working range for type
1 is from 30° to 130°, and the input angle working range for type 2 is from 30° to
121°. Even though type 1 and type 2 have same minimum folding angle, type 2 is
better when considering its shorter links. Therefore, working with type 2 is more
compatible than working with type 1 when it comes to network. Singularity occurs
Fig. 10 a Network at θ = 30◦ for type 2 Bennett 4R mechanism, b network at θ = 121◦ for type
2 Bennett 4R mechanism
1690 T. Punnoose Valayil
at θ = 29◦ and θ = 122◦ for the type 2 network and at θ = 29◦ and θ = 131◦ for
type 1 network.
Twist angles present in the links of a Bennett mechanism play an important role in
kinematic analysis. Twist angles are the main reasons for the formation of saddle
surfaces in a deployable Bennett network. Selection of twist angles is based on actual
needs [3]. By changing the twist angle, the shape of the network is analyzed for type
1 and type 2 Bennett 4R mechanisms.
Case 1: With α = 20◦ , l1 = 167.7 mm and l2 = 100 mm.
Case 2: With α = 30◦ , l1 = 167.7 mm and l1 = 100 mm.
From above Fig. 11a, b, it can be observed that when twist angle changes, the
shape of the networks changes. When twist angle is increased from 20° to 30°, more
saddle like shape is formed. Thus, twist angle is a governing factor for formation of
saddle surface in Bennett mechanism. By choosing the appropriate link length and
twist angle, a perfect and expected saddle shape can be obtained. Saddle surfaces
are most commonly used in building roofs in structures. These saddle roofs have a
hyperbolic paraboloid shape. This is a doubly curved surface having a concave surface
on one side and a convex on the other. Unlike other structures, these roofs gain their
strength through their shape and not by their mass. These roofs are lightweight and
have high stiffness. They can withstand unequal loading, both dead load and live load.
An example of saddle roof is shown in Fig. 12. It is a railway station at Warszawa
Ochota, Poland. Such type of roofs if constructed using Bennett mechanisms can be
folded.
A physical model of Bennett 4R mechanism which was constructed using PVC
pipe is shown in Fig. 13. Their links were connected using nut and bolt, acting as
revolute joint. In Fig. 13 (left), a single Bennett 4R mechanism can be observed, and
to the right, a network made of Bennett 4R mechanism can be observed. A saddle
shape can be observed in the network. It can also be constructed using steel for roofs
in buildings.
9 Conclusion
network, a saddle shape was observed. It was observed that twist angles present in
the links of the Bennett mechanism were the reason for obtaining perfect saddle
surface. The saddle structures gain their strength through their shape. Thus, Bennett
networks provide a simple way for constructing deployable saddle roofs.
References
Abstract Air, a crucial requirement of human life, has been increasingly polluted
day by day with rapid growth of industries, especially fossil fuel-driven ones, defor-
estation and urbanization. It has been severely affecting our lives and environment.
One active as well as passive source of air pollution is cigarette smoke which has
various health concerns. This smoke on the one hand releases complex aerosols like
vapor oxides, tar and particulate pollutants in the surroundings, and on the other
hand, cigarette buds get accumulated in land and ocean, thus affecting the ecological
aspects of the environment, which demands for the requirement of efficient smoke
filtration with proper disposal of used filters. This study analyzed the use of PAN
nanofibers in cigarette smoke filtration with reusability in energy applications. PAN
nanofibers showed 42.61% porosity with increased tortuosity; yet permeability was at
par with commercial cigarette filters. The nanofiber filters gained weight up to 115%
because of absorbed pollutants after a complete cigarette smoking test. Positive elec-
trical conductivity upon carbonization and catalytic contribution in electrochemical
tests was observed for the nanofiber filters.
1 Introduction
Ashish Kakoria and Sahil Verma contributed equally for this chapter.
are found to be cacogenic [1–6]. Particulate pollutants can cause diseases like lung
cancer, chronic obstructive pulmonary disease and asthma [7]. PM2.5 is small enough
to enter lungs and interfere with oxygen exchange in the arteries [7]. Cigarette buds
when released in atmosphere can get accumulated in water bodies and severely
affect aquatic ecology. Thus, efficient filtration of smoke constituents and proper
disposal/utilization of filter buds is necessary.
Air filters made up of fiber materials are used for most of the smoke filtration
applications. Filtration of PM in the flow stream is subjected to combined effect
of diffusion, inertial impaction, interception, electrostatics and earth’s gravitational
force [8, 9]. Filtration contribution of 40% and 60% by diffusion and interception was
observed by Overton, showing the contribution of inertial impaction to be negligible
and interception to be important process for the filtration of sub-micron particles
[10, 11]. Traditional filtration media including spun-bonded, melt blown and glass
fibers has low particulate matter filtration efficiency because of large pore size; on
the other hand, preparing thicker media causes high pressure drop and energy costs
[6]. Thus, electrospinning process was carried out for the formation of nanofibers.
2 Methodology
2.1 Materials
Polyacrylonitrile (PAN) (mol. wt. = 150 kDa) was purchased from Sigma-Aldrich.
N, N-Dimethylformamide (DMF) was purchased from Alfa Aesar.
9 wt. % solution of PAN in DMF was prepared by stirring the mixture at 40 °C and
700 rpm, for 14 h. The solution was electrospun with a conventional electrospinning
apparatus whose parameters are detailed in Table 1. The schematic of electrospinning
setup is shown in Fig. 1.
Table 1 Electrospinning
S. No. Parameter Value
parameters
1 Syringe capacity 10 mL
2 Syringe diameter 18 mm
3 Needle size 18 gauge
4 Needle to collector distance 12 cm
5 Solution flow rate 1.6 mL/h
6 High voltage 10 kV
Nanofibers for Sustainable Filtration: A Waste to Energy Approach 1695
Fig. 1 Schematics of electrospinning setup (left) and PAN fiber mat prepared by electrospinning
(right)
Thus, the produced fiber mat was compressed in polymer press at 50 °C and
392 kPa for 60 s. These mats were rolled into cigarette filters of 7.6 mm diameter
and 15 mm length.
Scanning electron microscopy (SEM) and ImageJ software were used for architec-
tural characterization of the nanofiber mat. Pressure and permeability were deter-
mined with the help of manometer and anemometer. Cigarette smoking test was
performed on a laboratory-made bottled setup that mimics human smoking process
with its squeezing and expanding action (Fig. 2).
2.4 Carbonization
First, the nanofiber mat was stabilized in air at 300 °C followed by carbonization
at 900 °C in N2 atmosphere. A ramp rate of 10 °C/min. and soaking time of 1 h.
were used for the purpose. Cooling to room temperature was also performed in N2
atmosphere [12].
2.5 Electrochemistry
Fig. 2 Pressure drop and flow measurement (left), schematics of setup for cigarette smoking test
(right)
To prevent the formation of bubbles, rotating disk electrode (RDE, PINE Instruments)
setup was used with rotation ranging from 100 to 1600 rpm.
The electrochemical experiments were performed in three-electrode setup, using
Pt wire as a counter electrode and Ag/AgCl (3M) as reference electrode. The potential
against Ag/AgCl was converted into potential against reversible hydrogen electrode
(RHE) by using the equation
3 Results
Fiber mat with average fiber diameter of 430 nm was produced with the electrospin-
ning process. Porosity of the fiber mat was determined to be 42.61%, with a mean
pore diameter of 2.60 µm (Fig. 3).
The diameter was reduced to 335 nm after stabilization and to 275 nm after
carbonization.
Nanofibers for Sustainable Filtration: A Waste to Energy Approach 1697
Fig. 3 SEM of PAN nanofibers mat (left) and carbonized mat (right)
1
τk = (2)
1 − α(1 − ∅)
where εp is the volumetric porosity, dp is the pore diameter, R is the specific gas
constant, T is temperature in absolute scale and M is molecular wt. of the fluid used.
The permeability was determined as [16–18]
−2μ RTLN
k= × − DKeff (4)
Pa + Pb Pb − Pa
where μ is the dynamic viscosity of the flowing medium (air), Pa and Pb are pressures
at inlet and outlet, L is the length of filter medium and N is the molar flux through
the medium.
Measuring the required values of porosity, pore diameter, temperature, pressure
drop and size of the filter, the value of Knudsen diffusivity was calculated to be
159.98 × 106 µm2 /s and permeability to be 0.0349 µm2 .
The normal time taken by smoke to pass through the filter section would be
L 15 × 10−3
t= = = 0.0115 s (5)
v 1.3
1698 P. Giri et al.
Therefore, the apparent length would be extended to 1.82 m, which is nearly 121
times more than the filter length. These results show good filtration capabilities with
increased hydraulic retention time or increased effective length of the filter, allowing
for efficient diffusion across the filter section.
PAN nanofiber filters were used for cigarette smoke filtration where the filter gained
up to 115% of its initial weight. Compressed polymer mat on simple rolling gained
15% of initial wt., and the absorption capacity was seen to increase with fluffiness
of the filter. A section of the filter was washed in ethanol and the absorbed parti-
cles leeched out in the solution. A similar study is demonstrated in Ref [19]. This
elucidates the reusability of the filters.
PAN nanofibers are porous and rich in carbon content. Their carbonization showed
a positive electrical conductivity. The conductivity of the specimen was measured to
be 98.47 −1 m−1 (c.f. Fig. 4).
For CV analysis, the carbonized sample was purged with oxygen in 0.1M KOH
electrolyte. The presence of reduction peak in CV confirmed that the sample was
oxygen reduction reaction (ORR) active. Furthermore, LSV studies carried out in
0.1M KOH in the potential range of 1.164 to −0.036 V (vs. RHE) under the scan
rate of 10 mV/sec showed an onset potential of 0.78 V (vs. RHE) with maximum
current density of 0.94 mA/cm2 , at 1600 rpm (Fig. 5).
Fig. 5 a CV, b OER, c HER, d ORR of carbonized sample in 0.1M KOH under scan rate of
10 mV/s. using three-electrode setup
After ORR analysis, electrochemical cell was purged with N2 , and LSV proce-
dures were carried out between 0 and −1.4 V (vs. RHE) in 0.1M KOH concentration
at room temperature which confirmed the presence of hydrogen evolution reaction
(HER) activity in the sample, revealing an onset potential of 0.6 V (vs. RHE) with
the current density of 25.7 mA/cm2 .
Furthermore, LSV test was again performed for the sample in the potential range
of 0.9–1.9 V (vs. RHE) under scan rate of 10 mV/s with rotation ranging from 0 to
1600 rpm in 0.1M KOH, and it was observed that the carbonized sample was oxygen
evolution reaction (OER) active, showing an onset potential of 1.58 V (vs. RHE)
with the current density of 6.5 mA/cm2 , which is just 350 mV additional compared
to IrO2 .
4 Conclusion
Nanofibers of PAN in DMF made with electrospinning process showed good filtra-
tion capabilities and excellent reusability applications. Besides, commercial cigarette
1700 P. Giri et al.
filters can also be tested for replacement with alternative carbon rich fiber filters,
which have electrochemical reusability because of their porosity and carbon content.
This experiment has already demonstrated that the fibers such as of PAN can have
their application as a current collector, electrode material, catalyst support material
or a catalyst itself. Moreover, this research opens further analysis about the effect
of heavy metal deposition from cigarette smoke on these properties. This research
provides good insight about filtration capabilities of PAN and its reusability applica-
tions. These fibers can be excellent alternatives for commercially available air/smoke
filters, preventing themselves to be exposed in our surroundings by finding their
reusability in energy applications.
References
1. Chiba M, Masironi R (1992) Toxic and trace elements in tobacco and tobacco smoke. Bull
World Health Organ 70(2):269
2. Kensler CJ, Battista S (1963) Components of cigarette smoke with ciliary-depressant activity:
their selective removal by filters containing activated charcoal granules. N Engl J Med
269(22):1161–1166
3. Patrianakos C, Hoffmann D (1979) Chemical studies on tobacco smoke LXIV. On the analysis
of aromatic amines in cigarette smoke. J Anal Toxicol 3(4):150–154
4. Hossain MT, Hassi U, Huq SI (2018) Assessment of concentration and toxicological (Cancer)
risk of lead, cadmium and chromium in tobacco products commonly available in Bangladesh.
Toxicol Rep 1(5):897–902
5. Burnett RT, Smith-Doiron M, Stieb D, Cakmak S, Brook JR (1999) Effects of particulate
and gaseous air pollution on cardiorespiratory hospitalizations. Arch Environ Health Int J
54(2):130–139
6. Morabet RE (2018) Effects of outdoor air pollution on human health. Elsevier Inc.
7. Mannucci PM, Harari S, Martinelli I, Franchini M (2015) Effects on health of air pollution: a
narrative review. Intern Emerg Med 10(6):657–662
8. Zhu M, Han J, Wang F, Shao W, Xiong R, Zhang Q, Pan H, Yang Y, Samal SK, Zhang F, Huang
C (2017) Electrospun nanofibers membranes for effective air filtration. Macromol Mater Eng
302(1):1600353
9. Wen D, Jianhui W, Bing P, Xiaobin Z, Fuwei X, Huimin L, Kejun Z (2015) An improved
theoretical model of cigarette smoke filtration across mono-segment cellulose acetate filters.
Beiträge zur Tabakforschung/Contrib Tobacco Res 26(5):232–240
10. Siddiqi K, Shah S, Abbas SM, Vidyasagaran A, Jawad M, Dogar O, Sheikh A (2015) Global
burden of disease due to smokeless tobacco consumption in adults: analysis of data from 113
countries. BMC Med 13(1):194
11. Overton JR (1973) Filtration of cigarette smoke: relative contributions of inertial impaction,
diffusional deposition, and direct interception. Beiträge zur Tabakforschung/Contrib Tobacco
Res 7(3):117–120
12. Rahaman MS, Ismail AF, Mustafa A (2007) A review of heat treatment on polyacrylonitrile
fiber. Polym Degrad Stab 92(8):1421–1432
Nanofibers for Sustainable Filtration: A Waste to Energy Approach 1701
13. Pisani L (2011) Simple expression for the tortuosity of porous media. Transp Porous Media
88(2):193–203
14. Arya G, Chang HC, Maginn EJ (2003) Knudsen diffusivity of a hard sphere in a rough slit
pore. Phys Rev Lett 91(2):026102
15. Knudsen M (1909) Effusion and the molecular flow of gases through openings. Ann Phys
29:179
16. Cunningham RE, Williams RJ (1980) Diffusion in gases and porous media. Plenum Press, New
York
17. Carrigy NB, Pant LM, Mitra S, Secanell M (2013) Knudsen diffusivity and permeability of
PEMFC microporous coated gas diffusion layers for different polytetrafluoroethylene loadings.
J Electrochem Soc 160(2):F81–F89
18. Kerkhof PJ (1996) A modified Maxwell-Stefan model for transport through inert membranes:
the binary friction model. Chem Eng J Biochem Eng J 64(3):319–343
19. Giri P, Kakoria A, Verma S, Sinha-Ray S (2021) Reuse of cigarette filters for energy
applications. Function Text Cloth 2020:161–168. Springer, Singapore
Effect of Heat Treatment on Wear
Behaviour of Austenitic Stainless Steel
Abstract Stainless steel is one of the most widely used steels for structural appli-
cations. This paper quantifies the effect of heat treatment on the wear resistance
behaviour of stainless steel 304 grades using pin on disc method. Ageing at different
temperature ranges imparts hardness to the SS304 specimen which in turn affects
the wear resistance characteristic. Different wear parameters have been calculated
using Archard’s equation and wear scar diameter method. Microstructure has been
studied using optical microscopy. Scanning electron microscopy has been used to
investigate the wear mechanism.
1 Introduction
Austenitic stainless steel with its superior mechanical and tribological properties
is now used extensively in strategic applications. Stainless steel 304 grade is a
preferred material for structures with high structural integrity and aggressive service
environments. It cannot be usually heat treated, but can be annealed at different
temperature ranges to bring it into application of refrigeration, chemical, paper and
food processing industry. Some applications such as conveyor belt, screws and bolts
have relative sliding motion against different components. This leads to wear of the
mating surface and can ultimately lead to component’s failure [1]. Stainless steel 304
possesses adequate toughness and sufficient ductility to resist failure for a consid-
erable long duration of time. They can also be readily welded with similar and
dissimilar metals for various applications [2]. 304 grade austenitic stainless steel is
a high alloy Cr–Ni–Mo steel. It has high pitting resistance equivalent number which
makes it significantly corrosion resistant. The literature reveals that the previous
researchers have worked in the field of estimating wear characteristic of austenitic
stainless family [3–7].
But very limited literature is available regarding the effect of heat treatment on
austenitic stainless steel 304 grade’s wear characteristics [8]. This paper aims to
investigate the effect of heat treatment on hardness, its correlation with microstructure
and related wear behaviour. Pin on disc technique has been used to study the wear
behaviour in laboratory. Various wear property parameters have also been calculated
using Archard’s equation and scar diameter method. Wear mechanism has been
studied using scanning electron microscopy (SEM).
2 Material
Stainless steel 304 plate was received in the dimension of 250 × 250 mm. The
chemical composition of stainless steel 304 is given in Table 1.
The pins made out of SS304 were made to slide against the disc made of EN 31
bearing steel. Table 2 gives the chemical composition of EN31 bearing steel used.
3 Experiment
Pins were prepared from the SS304 plate in the dimensions shown in Fig. 1.
Effect of Heat Treatment on Wear Behaviour … 1705
The pin specimens were subjected to the following heat treatment as shown in Table
3.
The pin on disc test was conducted on Ducom wear testing machine with a load of
5 N at a rotating speed of 500 rpm for 15 min duration. The volume loss in material
was calculated using two methods: (i) change in scar diameter (ii) weight loss of pin
due to material removal. Figure 2 represents the schematic of wear testing machine
used in this experiment.
The specimens in as-received and heat-treated condition were polished with emery
paper up to a grade of 2000 and then with a diamond paste. The mirror polished
1706 W. N. Khan et al.
specimens were then etched with Villela reagent to reveal the microstructure. Optical
microscope was used to observe the developed microstructure owing to the heat
treatment. The microstructure images in Fig. 3 clearly indicate that heat treatment
has lead to grain coarsening. This in turn will affect the hardness of the specimen,
thereby influencing the material loss due to wear.
Microhardness measurement carried out with a load of 10gf with a dwell time of 10 s
is shown in Fig. 4.
Dry sliding wear test was performed for specimens with a load of 5 N. The wear
track radius was kept 60 mm with a rotational speed of 500 rpm for 15 min. Sliding
distance is represented mathematically as per Eq. 1.
π DNT
SD = (1)
1000
Effect of Heat Treatment on Wear Behaviour … 1707
Fig. 4 Hardness
measurement
1708 W. N. Khan et al.
where SD is the sliding distance in metre, D is the wear track diameter in mm, N is
the rotation of disc measured in rpm, and T is the time of pin sliding over disc in
minutes. Here, D is 60 mm, N is 500 and T is 15. This makes the sliding distance
to be 1413 m in total. Before and after each run of 15 min, the pin specimens were
weighed on an electronic weigh balance with and accuracy of ±0.1 mg. Table 4
represents the initial and final weight of pin specimens. This weight loss data is used
to calculate the volume loss as per Eq. 2. Volume loss is further used for calculating
the volume loss per unit sliding distance which is generally referred as average wear
in mm3 /m.
W
V = (2)
ρ
where V is the volume loss in mm3 , W is the weight loss in grams occurring due
to wear and ρ is the density of pin material in g/cm3 .
V
Q= (3)
SD
Here, Q is the average wear occurring in mm3 /m, SD is the sliding distance in
metres which is 1413 m in this case.
A comparative plot of all specimens for wear occurring measured mm3 /m is shown
in Fig. 5. The data in Table 4 and Fig. 5 clearly indicate that the highest loss in weight
due to wear occurs in non-heat-treated specimen. Heat treatment improves the wear
resistance characteristics, specifically in HT 1 condition where weight loss reduces
multiple folds as compared to others.
Wear coefficient and wear resistance are another important parameters which
are used to define the wear resistance characteristics of a material. Equations 4
and 5 represent the mathematical expression for these parameters as per Archard’s
equation.
Q∗H
K = (4)
Fn
1
Rw = (5)
K
Effect of Heat Treatment on Wear Behaviour … 1709
where K is the wear coefficient, Q is the average wear in mm3 /m, H is the hardness
of pin specimen and Rw is the wear resistance. Table 5 represents the wear coefficient
and wear resistance value for all the specimens calculated using the above-mentioned
equations.
Figure 6 shows comparative plot of wear coefficient and wear resistance.
Each pin specimen after undergoing the rotationary sliding motion on disc developed
scar on its surface. This scar is mainly due to the material loss at the interface of pin
and disc. The scar is measured precisely using vernier calipers, and then based on
mathemtical formular given in Eq. 6, the volume loss is calculated.
Volume Loss = π ∗ (wear scar diameter in mm)4 /64 ∗ (sphere radius in mm)
(6)
1710 W. N. Khan et al.
Table 6 represents the data of wear performance calculated as per the wear scar
diameter method.
Scanning electron microscopy was used to estimate the wear mechanism occurring.
SEM images shown in Fig. 7 were taken at the face of pin’s sliding surface. The
5 Conclusion
• Austenitic stainless steel of 304 grade has been given suitable heat treatments.
• Pin on disc technique has been used to estimate the dry sliding wear behaviour in
three heat-treated and one non-heat-treated as-received condition of specimens.
• Wear characteristic parameters namely; loss in weight, volume loss, average
wear, wear coefficient and wear resistance have been calculated using Archard’s
equation.
• Wear scar diameter has also been used to calculate the volume loss occurring.
• Heat treatments improve the wear resistance of SS 304 steel.
• Heat treatment 1 specimen exhibits the best wear resistance, with its resistance
value turning out to be almost 6.5 times better than that of material in as-received
condition.
• Heat treatment 2 and Heat treatment 3 show an improvement of marginal 1.13
and 1.1 times in wear resistance as compared to as-received non-heat-treated
specimen.
• Applications involving rigorous wear due to sliding, SS304 can be suggested to
be used in heat treatment 1 condition for better performance.
References
1. Kumar S, Mukhopadhyay A (2016) Effect of microstructure on the wear behavior of heat treated
SS-304 stainless steel. Tribol Ind 38(4):445–453
2. Abudaia FB, Bull SJ, Oila A (2012) Surface wear resistance of austenitic stainless steels modified
by colossal carbon supersaturation and TiN coating. J Mater Sci Eng B, 2(2):103–111
3. Hoier P, Malakizadi A, Friebe S, Klement U, Krajnik P (2019) Microstructural variations in
316L austenitic stainless steel and their influence on tool wear in machining. Wear 428:315–327
4. Mello CB, Ueda M, Lapienski CM, Reuther H (2009) Tribological changes on SS304 stainless
steel induced by nitrogen plasma immersion ion implantation with and without auxiliary heating.
Appl Surf Sci 256:1461–1465
5. Yang ZY, Naylor MGS, Rigney DA (1985) Sliding wear of 304 and 310 stainless steels. Wear
105(1):73–86
6. Bregliozzi G, Ahmed SIU, Schino AD, Kenny JM, Haefke H (2004) Friction and wear behavior
of austenitic stainless steel: influence of atmospheric humidity, load range, and grain size. Tribol
Lett 17(4):697–704
7. Straffelini G, Trabucco D, Molinari A (2002) Sliding wear of austenitic and austenitic-ferritic
stainless steels. Metal Mater Trans A 33A:613–624
8. Bressan JD, Daros DP, Sokolowski A, Mesquita RA, Barbosa CA (2008) Influence of hardness
on the wear resistance of 17–4 PH stainless steel evaluated by the pin on disc testing. J Mater
Process Technol 205:353–359
Design and Development of Intelligent
Moving Machine Using LabVIEW
Abstract The intelligent moving machine is designed for fast response, cost effec-
tiveness and it avoids the obstacle in its path so it is called safe intelligent moving
machine (SIMM). An introduction of intelligent behavior of moving machine that
can adapt the change as like environment. Also describes the application of intelligent
machines. This work presents design and development of an automatic intelligent
system for industrial environment using LabVIEW and its real-time hardware imple-
mentation. The machine moves forward, backward, left and right through the signals
supplied by the control station. LabVIEW programming has been proved to be more
efficient than microcontroller to track this intelligent moving machine with the help
of DAQ. The additional feature is that our vehicle is also identifying the obstacle in
its path when it moving and generates necessary control commands to avoid it.
1 Introduction
The intelligent moving machine which may avoid the obstacle move intelligently and
reach to the destination safely, accurately in minimum time and minimum cost. Earlier
many moving machines were developed using different techniques for a real-time
path finder system. Man intelligent moving machine has become an important topic
in the robots community. An advanced intelligent moving machine must integrate
capabilities to detect human’s presence in their vicinity and interpret their motion for
an active interaction; the machine must also be able to track the static and dynamic
obstacle in its path. The humans and moving machines coexist, closely interact and
perform a certain amount of work. This research work is divided into two parts. The
first part deals with the development of safe intelligent moving machine (SIMM),
which maintains a certain relative positional relationship between the human and
moving machine. The moving machine can carry loads that are required by people
working in hospitals, airports, and industries. These machines are considered as an
intelligent moving machine because they are able to interact with its environment and
learn from the surroundings automatically. Interacting with nature and indoor situ-
ation involves learning and adaptation from changing environment. The intelligent
machine can work as an assistant for humans in various situations. The second part
of work related to the intelligent moving machine is obstacle avoidance in real-time
situation. The real-time obstacle avoidance system of intelligent moving machines
will be developed with help of soft computing techniques.
1.2 Motivation
Human intelligence, speed of thinking, capability of adaptation and the way obstacle
avoidance in real-time dynamic environment inspired the researcher to develop an
intelligent moving vehicle which is safe, intelligent and optimally find its path in the
static and dynamic, real-time environment.
First we have to design a block diagram panel for giving the commands to our
automatic vehicle. The commands will be forward, backward, left and right. The
command will be transmit through the DAQ, card to the receiver station. The
functional block is shown below in Fig. 1.
In this section, our receiver antenna gets the information from the DAQ card and
starts doing task. The vehicle moves forward, backward, left and right. And also
avoids obstacle in his path. The UV sensor is used to detect the objects and gives
commands to control motor and speed of front wheels to take turns of automatic
vehicle. The speed of automatic vehicle is 10 km/h [1]. It takes turn very fast in
motion. The intelligent moving machine is safe running motor which takes action
according to the situations. The control programming plays very important role in
this type situation which delivers the correct action to the motor drives (Figs. 2 and
3).
Wireless Intelligent
DAQ Signal System Machine
Start
Is Ob- Y
stacle Detect
N
Vehicle goes on forward
Stop
2 Objective of Paper
The primary objective of the work to develop a real-time obstacle avoidance machine
to complete the task that requires distributed functionalities. In this context, the core
research objectives of the thesis are given by:
1. To design and develop the machine in lab.
2. To design and develop an obstacle avoidance algorithm for moving machine.
3. To implement in laboratory a framework for vehicle and the obstacle avoidance
system.
2.1 Methodology
3 Results
Intelligent moving machine has been developed which is moving forward, back-
ward, left and right. This is a smooth running machine controlled through software-
based controlling technique. This is a highly useful intelligent machine for industrial
purpose which can work more efficiently in real time industrial circumstances. This is
completely controlled through a control station. This moving machine also effectively
avoids the obstacles coming in its path, sensing them through the ultrasonic sensor. It
changes its movement very fast. When we provide a signal, it changes its directions
1718 A. Yadav et al.
in 0.3 s. The trajectory angle with obstacle is kept as 45° when machine is moving
forward and backward [2]. At the same time, the direction preference for turning has
been also taken as left turn for both, forward and backward movement.
4 Conclusion
The intelligent moving machine is developed in this project is smooth running and
response is also fast enough for utlization in lab. The developed automatic intelli-
gent machine also senses and avoids the static obstacles in its path. So, the devel-
oped moving machine is very useful in industrial and airport as a load vehicle with
necessary adoptation for the respective field of operation.
Design and Development of Intelligent Moving Machine … 1719
References
Jitendra Adhikari, Rajeev Kumar, Vikas Narain, and Satish Chandra Jain
Abstract The present study proposes enhancement of harvested power and voltage
by tuning the poling orientation in piezoelectric materials. The dependency of
piezoelectric strain coefficients on performance is presented mathematically and to
demonstrate the effect, a cantilever-based energy harvester having platinum substrate
is considered with seven different materials. It is observed that PZT-2 shows an
improvement of 598% in harvested power and 165% in voltage by poling tuning to
45°. Similar poling tuning helps PZT-7A to improve 325 and 106% in power and
voltage generation, respectively. Huge improvement of 1425% in power and 290%
for voltage is observed for PMN-0.35PT. PbTi03 shows a minimal improvement at
poling angle of 30°. The performance of materials like Ba2 NaNb5 O15 and PVDF gets
deteriorated with an increase in poling orientation. The peak values of power and
voltage are observed at different poling angles for different piezoelectric materials.
The least magnitudes of power and voltage generation occur at poling angle of 90°
for any material system.
1 Introduction
Piezoelectric materials are high potential smart materials which gained importance
because of its development in future self-powered microsystems. Technological
advancement increases researchers’ interest in improving the efficiency and perfor-
mance of piezoelectric energy harvesting devices. Studies in mechanical aspect of
the piezoelectric energy harvester focus on materials, geometrical entities (cantilever,
diaphragm, and spherical, etc.) and working mode of energy harvester. The working
mode deals with piezoelectric strain coefficients: d 31 (transverse), d 33 (longitudinal)
and d 15 (shear). Ng et al. [1] used simple geometrical configuration suitable for
harvesting the energy using transverse mode (d 31 mode) when mechanical vibra-
tional frequencies are low. Shen et al. [2] harvested energy based on longitudinal
mode (d 33 mode). But this configuration is less effective due to polarization issues,
and large mechanical force is required to produce the strain as compared to the trans-
verse mode. Wang et al. [3] investigated energy harvesting for shear mode (d 15 mode)
configuration. This arrangement appears to be most effective for output power, but
it involves a complex fabrication process. Even though piezoelectric constant for
d 31 is low as compared to the d 15 and d 33 modes still d 31 is most accepted due to
its real-time applications when compared against its counterparts. Performance in
the d 31 mode for a piezoelectric material can be escalated using dipole engineering
and through alteration of crystallographic orientation [4–6]. Davis et al. analyzed the
eff
response of poling change on d31 for various piezoelectrics using dipole engineering
[7].
The study primarily focusses on enhancement of performance of energy harvesters
using various piezoelectric materials in cantilever configuration by changing the
poling direction operating in d 31 mode.
This research deals with the effect of poling in the performance of energy harvesters
of the piezoelectric materials. Different piezoelectric materials are considered for
analysis including PMN-0.35PT(Pb(Mg1/3 Nb2/3 )O3 –PbTiO3 ) [8], Barium sodium
niobate (Ba2 NaNb5 O15 ) [9], PZT-5A and PZT-7A (Pb[ZrxTi1-x]O3 ), [10] Lead
Titanate (PbTiO3 ) [11], PCR-1 (Piezoelectric Ceramics of Rostov) [8], polyvinyli-
dene difluoride (PVDF) [9]. Table 1 presents the material properties of piezoelectric
materials. Here, we have mathematically illustrated the dependency of alteration of
poling direction over performance of piezoelectric materials.
Suppose stress (T ) be a second-order tensor, d is the coupling piezoelectric
coefficient and polarization (P) is a vector related by these two variables as
PZT2 11.6 −3.3 −4.9 14.8 45 29.9 −60 440 152 990 450
Ba2 NaNb5 O15 5.3 −1.98 −1.2 8.33 15.4 13.2 −7 52 37 235 51
PMN-0.35PT 13.2 −3.96 −6.05 14.7 33.4 34.32 −133 936 270 4610 3270
PCR-1 12.5 −4.4 −5.8 15.9 38.8 33.8 −95 420 220 1130 650
PZT-7A 10.7 −3.22 −4.62 13.9 39.5 27.84 −60.2 364 151 843 427
PVDF 4.43 −9.91 −3.1 1.14 1.82 1.45 1.36 1.93 2.97 6.8 7.3
PbTiO3 7.7 −1.7 −1.2 8.2 19 31.4 −6.8 68 56 240 190
Effect of Poling Orientation in Performance of Piezoelectric … 1723
Pi = d i jk T jk (1)
Polarization vector (P) is transformed from old co-ordinate system to new co-
ordinate system which can be written as:
P = a P (2)
where P and P are the polarization vectors in old and new co-ordinate system
respectively and a is the direction cosines matrix for rotation about any axis that is
given as:
⎡ ⎤
l1 m 1 n 1
a = ⎣ l2 m 2 n 2 ⎦ (3)
l3 m 23 n 3
l1 = cos x , x , m1 = cos y , x ,n 1 = cos z , x , l2 = cos x , y , m 2 =
where
cos y , y , n 2 = cos z , y , l3 = cos x , z , m 3 = cos y , z , n 3 = cos z , z .
Substitution of Eqs. (1) into (2) leads to:
Pi = α d i jk T jk (4)
T jk = αT jk (5)
Matrix α is derived from the direction cosines of x, y, z co-ordinate axes given as:
⎡ ⎤
l12 m 21 n 21 2m 1 n 1 2n 1l1 2l1 m 1
⎢ l2 m 22 n 22 2l2 m 2 ⎥
⎢ 2 2m 2 n 2 2n 2 l2 ⎥
⎢ 2 ⎥
⎢ l m 23 n3 2
2m 3 n 3 2n 3l3 2l3 m 3 ⎥
α=⎢ 3 ⎥ (6)
⎢ l2 l3 m2m3 n 2 n 3 m 2 n 3 + n 2 m 3 l2 n 3 + n 2 l3 m 2 l3 + l2 m 3 ⎥
⎢ ⎥
⎣ l3 l1 m3m1 n 3 n 1 m 1 n 3 + n 1 m 3 l2 n 3 + n 1 l3 m 1 l3 + l1 m 3 ⎦
l1 l2 m1m2 n 1 n 2 m 1 n 2 + n 1 m 2 l1 n 2 + n 1 l2 m 1 l2 + l1 m 2
Here, poling vector is considered to be rotated with angle θ along 2nd direction
eff
as shown in Fig. 1. Upon rotation, it is noted that d31 depends on d31 , d33 , d15 and
angle θ as mentioned below:
eff
d31 = cos θ d31 cos2 θ + d33 sin θ − d15 sin2 θ (8)
1724 J. Adhikari et al.
eff
It is noticed from Eq. (8) that magnitude of d31 increases with surge in magnitudes
of d31 and d33 . It also gets negatively affected with an increase in d15 .
The voltage in d31 mode is related to piezoelectric coefficients and electrical
permittivity as
eff
d31
V = C (9)
ε31
eff
where V and C are the voltage and numerical constant that depends on stress and
thickness of the piezoelectric layer, respectively:
On further expansion [12]
cos θ d31 cos2 θ + d33 sin θ − d15 sin2 θ
V =C (10)
ε11 sin2 θ + ε33 cos2 θ
eff
The effect of poling orientation on voltage and power due to change in d31 is
explored for different piezoelectric materials. It is observed that different materials
have discrete response for change in poling angle. It is also noted from Eq. 8 that
eff
to have higher values of d31 , a significant difference has to be there in between the
magnitudes of d31 and d15 . Sensing voltage has a proportional relation with piezo-
electric strain coefficients and shares an inverse relation with dielectric constants
as mentioned in Eq. 20 [12]. PZT-2 shows an improvement of 598% in harvested
power and 165% in voltage when poling orientation changes from 0° to 45°. Huge
improvement in power of around 1425% was observed for PMN-0.35PT along with
a voltage gain of 290%. As the poling angle of PCR-1 increases, the harvested power
also surges reaching the maximum value at 30° and further improvement leads to
fall with least value at 90°. Here, power improvement of 33% and voltage increment
of 17% were observed. PZT-7A indicates rise in power and voltage with increase in
poling orientation touching the peak at 45° along with showing voltage and power
enhancement of 106% and 325%, respectively. Material system like PbTiO3 shows
minimal increase of 3.5 and 2.12% in power and voltage with peak orientation at
30°. On the contrary, a material system such as Ba2 NaNb5 O15 and PVDF experiences
decrease in performance with increase in poling orientation. It is noticed that for any
material system (Figs. 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 and 15) the least value
of harvested power and voltage occurs at 90°.
4 Conclusion
for PMN-0.35PT, PZT-2 and PZT-7A, harvested power get enhanced by 1425%,
598% and 325% due to large difference in magnitudes of piezoelectric coefficients.
On the contrary, Ba2 NaNb5 O15 and PVDF show a negative trend with increase in
poling angle. An interesting observation was that the poling angle for peak sensing
voltage and generated power for different materials need not to be same.
1728 J. Adhikari et al.
References
1. Ng TH, Liao WH (2005) Sensitivity analysis and energy harvesting for a self-powered
piezoelectric sensor. J Intell Mater Syst Struct 16(10):785–797
2. Shen Z, Liu S, Miao J, Woh LS, Wang Z (2013) Proof mass effects on spiral electrode D33
mode piezoelectric diaphragm-based energy harvester. In: 26th international conference on
micro electro mechanical systems (MEMS), pp 821–824
3. Wang DA, Liu NZ (2011) A shear mode piezoelectric energy harvester based on a pressurized
water flow. Sens Actuators A Phys 167(2):449–458
4. Yin J, Cao W (2000) Domain configurations in domain engineered 0.955 Pb (Zn 1/3 Nb 2/3)
O 3–0.045 PbTiO3 single crystals. J Appl Phys 87(10):7438–7441
5. Zhang R, Cao W (2004) Transformed material coefficients for single-domain 0.67 Pb (Mg 1/ 3
Nb 2/ 3) O 3–0.33 PbTiO3 single crystals under differently defined coordinate systems. Appl
Phys Lett 85(26):6380–2
6. Nakanishi R, Kanda K, Fujita T, Kanno I, Maenaka K (2019) Multilayer piezoelectric MEMS
energy harvester based on longitudinal effect. J Phys Conf 1052(1)
7. Davis M, Damjanovic D, Hayem D, Setter N (2005) Domain engineering of the transverse
piezoelectric coefficient in perovskite ferroelectrics. J Appl Phys 98(1):014102
8. Dantsiger AY, Razumovskaya ON, Reznitchenko LA, Grineva LD, Devlikanova RU, Dudkina
SI, Gavrilyatchenko SV, Dergunova NV, Klevtsov AN (1994) Highly effective piezoceramic
materials (Handbook). Kniga, Rostov-on-Don. (in Russian)
9. Comsol Multiphysics, Materials Database v4.3 in Materials Database (2012)
10. Dunn ML, Taya M (1993) Electromechanical properties of porous piezoelectric ceramics. J
Am Ceram Soc 76:1697–1706
11. Xu Y (1991) Ferroelectric materials and their applications. North-Holland, Amsterdam
12. Kiran R, Kumar A, Kumar R, Vaish R (2018) Poling direction driven large enhancement in
piezoelectric performance. Scripta Materialia 151:76–81
Parametric Analysis of Vertical Contact
Mode Triboelectric Energy Harvester
Satish Kumar, Rajeev Kumar, Vikas Narain, and Satish Chandra Jain
1 Introduction
power. These small electronics need a battery to operate, but sometimes it is diffi-
cult to replace/charge the battery. Energy storage systems have emerged as one of
the most promising technologies for meeting smaller energy needs, such as running
electronic devices with low power and wireless sensors. While several other methods
based on various mechanisms for harvesting mechanical energy have been demon-
strated, such as piezoelectric harvesters, electro-magnetic harvesters, and electro-
static harvesters, triboelectric energy harvesters (TEHs) are more influential in power
output and energy conversion efficiency. That makes TEH the most common source
of energy for portable electronics and also possibly to use for large-scale power
production in the near future.
The triboelectric mechanism is developed on the basis of contact electrification
and electrostatic induction. Fan and Wang [1] published the first paper on triboelectric
energy harvesting. Through mounting two polymer sheets made of materials with
significantly different triboelectric features, they produced the TEH with metal films
coated on upper and lower layers of the assembled structure. When subjected to
mechanical deformation, due to the nano-scale surface roughness, a friction between
the two films produces equal but opposite charging signs on two sides. Therefore, at
the interface zone, a triboelectric potential layer is created, which serves as a charge
“pump” to drive the electrons flow in the external load when there is a difference in
the device capacitance. Such a versatile polymer TEH provides up to 3.3 V of output
voltage at a output power density of 313 W/m2 [2].
Surface topography plays a significant role in deciding the effective area of contact
between the triboelectric layers. Fan et al. [3] showed that patterned films exhibit
substantially better performance than unpattern films. Triboelectric layer patterning
is known to increase the efficiency of the triboelectric energy harvester. Zhu et al.
[4] developed a triboelectric energy harvester that not only has quite a-simplified
structure but also significantly higher output power allowed by nanoparticular surface
treatments. Niu et al. [5] developed a mathematical model for TEH contact-mode and
conducted experiments to validate those theoretical findings. Kia et al. [6] performed
on systematic simulations of the macro/nanostructure adhesive contact behavior at the
TEH interface; an interaction potential was used to describe the adhesive interactions
while using Green’s half-space function, surface deformations are coupled. Niu et al.
[7] demonstrated the first equivalent circuit model and the first integrated simulator
for TEH system and validated through comparison with the analytical results. Rasel
et al. [8] exhibited a human-skin triboelectric generator based on human skin, a micro-
structured surface PDMS sheet, a ground-connected Cu electrode, and polyurethane
(PU) as substrate material. In this paper, a triboelectric tile is proposed to harvest the
energy from walking. A simplified model of tile has been analyzed, and mathematical
model of the tile has been developed to predict the harvested power.
Parametric Analysis of Vertical Contact Mode Triboelectric … 1735
2 Mathematical Modeling
Physical model of TEH is shown in Fig. 1. When a force is applied on the TEH,
then it vibrates. During the vibration, PDMS layer makes intermittent contact with
a bottom copper layer that results in an electric voltage difference between both
layers. A simplified model of the TEH (see Fig. 2). As a result of contact elec-
trification, the inner surfaces of top and bottom layers should have opposite static
charges (tribocharges) of equal quantity and opposite charging density (σ ). As the
two layers begin to separate from each other, a potential difference (V ) between the
two electrodes would be triggered with increasing separation.
It is appropriate for the insulator (layer 2) to presume that the tribocharges are
distributed uniformly with marginal decay at the surface. So there are two sections
to the total charge in layer1: One is the triboelectric (Sσ ) charge, another is the
transferred charge between the two electrodes (−Q). This charge flows through an
external resistance (R). Thus, the total charge in layer 1 is (Sσ − Q).
From the Gauss theorem, the electric field strength in the dielectric layer and the
air gap are given by:
Inside dielectric:
Q
E air = − (1)
Sε0 εr 1
− QS + σ (t)
E air = (2)
ε0
Q σ y(t)
V =− (d0 + y(t)) + (4)
Sε0 ε0
Combining Eq. (4) with Ohm’s law will calculate the performance of output
properties:
dQ
V = IR = R (5)
dt
Merging Eqs. (4) and (5)
dQ Q σ y(t)
R =− (d0 + y(t)) + (6)
dt Sε0 ε0
Q = 0 at t = 0
A sinusoidal motion of the top electrode, defined in Eq. (7), was selected because
this type of motion was fitted with the experimental testing rig (Keyboard Life Tester
by Shenzhen ZXD Testing Equipment Co. Ltd.) [9]. Also a motion of constant
velocity was found. Once t = 0, the top electrode’s two contact surfaces and the
dielectric layer have been undergoing contact for a sufficiently long time to achieve
their saturation by the contact electrification. The top electrode then began to isolate
itself from the dielectric.
Parametric Analysis of Vertical Contact Mode Triboelectric … 1737
where A0 and ω are the amplitude and the angular velocity of the motion, respectively.
θ is the initial phase angle. The parameters used in the simulation matched those used
in the experiments which are described in Table 1.
4 Parametric Study
It is of high significance to increase TEH output power and to reduce device size. Here,
implement a numerical analysis of the different parameters under which the power
produced is increased and miniaturizes the size. Especially, the first scheme uses a
predefined geometry with the exact value of the resistance circuit. The second scheme
is to define TEH design parameters when the power output is either unchanged or
increased.
or Nylon are positive tribomaterials and tend to lose electrons when coming into
contact with negative-charge tendency materials [e.g., PTFE] that have a tendency
to gain electrons. For this study, choose PTFE and nylon as the two triboelectrical
materials, due to their opposite position in the triboelectric series. PTFE is a widely
used polymer, due to its flexibility, manufacturing ease, transparency, biocompati-
bility, and super-hydrophobicity. It is also widely used as triboelectric material in the
construction of TEH for a broad range of applications. On the other hand, nylon has
1740 S. Kumar et al.
good mechanical properties (strength and stiffness), high impact resistance, which
is easy to fabricate and maintains its properties over a large temperature range.
Aiming the improvement of the triboelectric effect, we studied the influence of the
PDMS layer thickness on the generated electrical outputs. PDMS was the triboelectric
material chosen to vary the thickness (between 10 and 250 µm), while Cu was used
in film form. Here, the relation between the thickness of the friction layer and the
process of transporting triboelectric charges in a PDMS film is investigated in this
study. The power output is inversely proportional to the dielectric layer thickness
Parametric Analysis of Vertical Contact Mode Triboelectric … 1741
Fig. 7 Relationship of
output power with dielectric
layer thickness (d)
(d), as shown in Fig. 7. They decrease rapidly as the dielectric layer thickness (d)
increases from 0.1 to 0.2 mm which is usually used in experiments for TEH.
Here the density of the surface charge, geometries including area, dielectric thickness
and frequency are the same as in the previous section. The goal is to calculate the
load resistance for different dielectric layer thicknesses in order to obtain maximum
value of average power output. The simulation results are presented in (see Fig. 8),
from which it is clear that the output power meets its limit at a certain load resistance.
Fig. 8 Relationship of
output power with load
resistance (d)
1742 S. Kumar et al.
Fig. 9 Relationship of
output power with motion
parameter (frequency)
The nature of the profile of input motion influences the magnitude and rate of
induced TEH outputs. Sinusoidal and constant velocity periodic motion profiles
are commonly used to describe TEH, of which key parameters are the magnitude of
the motion (amplitude) and the motion rate (frequency). Increasing the frequency of
movement of the TEH layer increases the power output (Fig. 9). The overall peak
power increases directly proportional to frequency.
The output power is dependent on the size of a TEH surface area (S). This study used
a pair of square-shaped TEH layers consisting of PDMS and Cu surface contacts, and
power output was evaluated at various surface areas. As the surface area increases, the
output power is increased (see Fig. 10). As a result, the power output was higher for
the larger surface area unit, which was observed at a relatively lower load resistance,
both favorable for practical TEH applications in terms of outputs [10].
Fig. 10 Relationship of
output power with surface
area
Fig. 11 Relationship of
output power on surface
density
where device parameters are the same as described in Sect. 3. As shown in Fig. 11,
output power increases exponentially as the surface charge density (σ ) increase.
5 Conclusion
In summary, a simplified model for TEH for sinusoidal motion is presented in this
work and the influence of various design parameters on the triboelectric energy
harvester (TEH) has been studied. TEH has been modeled under the vertical contact
mode of the triboelectric harvester. Simulink model is developed for explicit expres-
sions, and charge, current, voltage, and power have been calculated for an externally
1744 S. Kumar et al.
References
1. Fan FR, Tian ZQ, Wang ZL (2012) Flexible triboelectric generator. Nano Energy 1(2):328–334
2. Jiang T, Zhang LM, Chen XY et al (2016) Structural optimization of triboelectric nanogenerator
for harvesting water wave energy. ACS Nano 9:12562–12572
3. Fan FR, Lin L, Zhu G, Wu W, Zhang R, Wang ZL (2012) Transparent triboelectric nanogen-
erators and self-powered pressure sensors based on micropatterned plastic films. Nano Lett
12(6):3109–3114
4. Niu S, Wang S, Lin L, Liu Y, Zhou YS, Hu Y, Wang ZL (2013) Theoretical study of contact-mode
triboelectric nanogenerators as an effective power source. Energy Environ Sci 6(12):3576–3583
5. Yang Y, Zhou YS, Zhang HL et al (2013) A single-electrode based triboelectric nanogenerator
as self-powered tracking system. Adv Mater 25:6594–6601
6. Kia DS, Towfighian S, Jin C (2016) Predicting the output of a triboelectric energy harvester
undergoing mechanical pressure. In: ASME 2016 conference on smart materials, adap-
tive structures and intelligent systems. American Society of Mechanical Engineers, pp
V002T07A007–V002T07A007
7. Rasel MS, Halim MA, Park JY (2015) A PDMS based triboelectric energy harvester as self-
powered, active tactile sensor system for human skin. In: Sensors. IEEE pp 1–4 (2015)
8. Niu S, Zhou YS, Wang S, Liu Y, Lin L, Bando Y, Wang ZL (2014) A simulation method
for optimizing the performance of an integrated triboelectric nanogenerator energy harvesting
system. Nano Energy 8:150–156
9. Yang B, Zeng W, Peng Z, Liu S, Chen K, Tao X (2016) A fully verified theoretical analysis
of contact-mode triboelectric nanogenerators as a wearable power source. Adv Energy Mater
1600505
10. Wang J, Li SM, Yi F et al (2017) Sustainably powering wearable electronics solely by
biomechanical energy. Nat Commun 7:12744
Mathematical Model of Sliding Mode
Triboelectric Energy Harvester
1 Introduction
harvester is much easier for packaging and more effective for static charge generation
[5]. Basic structure of contact mode and sliding mode energy harvester is same, basic
difference between these two is direction of motion of top plate, in sliding it moves
in lateral direction, whereas in case of contact mode, it moves in vertical direction.
Triboelectric effect is known from decades, but first triboelectric nanogenerator
(TENG) was invented by Wang group in 2012 [6]. The first TENG device consists
of Kapton film and polyester (PET) as triboelectric materials, and working principle
of this device was on vertical contact separation mode. Due to triboelectric effect,
charges are generated on the surface of two different triboelectric materials and
because of the electrostatic induction charge is transferred from triboelectric material
to electrode. A continuous AC output was achieved if it repeated.
Initially, most of the TENG were performed on the vertically contact mode. Sliding
mode triboelectric nanogenerator was demonstrated by Wang et al. by using PTFE
(Polytetrafluroethylene) and nylon (Polyamide6, 6) as triboelectric materials, glass
as substrate and Cu as electrode [7]. This TENG has same working principle as
vertical contact separation mode. Maximum output obtained as V OC = ∼1300 V and
short-circuit current density of 4.1 mA/m2 with a peak power density of 5.3 W/m2 .
Zhu et al. fabricated a TENG and obtained output by lightening two rows of 80
LED’s [8]. To harvest energy from distilled water, Wang group fabricated a sliding
mode TENG [9]. A U-tube as triboelectric material and distilled water slides over it.
This TENG works on triboelectric effect and Pascal’s law. This TENG is also used
as multifunctional sensor which could measure displacement, pressure, torsion, etc.
In sliding mode, TENG along with triboelectric effect, other effects are consid-
ering for energy harvesting. Han et al. designed a TENG based upon printed circuit
board (PCB) technology by using PTFE and Kapton as triboelectric materials [7].
It is found that as the frequency changes from 600 Hz to 3 kHz, the power output
changes from 1.3 to 25.7 W. This TENG had been used for charging a phone by
using a small transformer.
Park et al. fabricated a small-size rotating TENG. PTFE and fiber are taken as
triboelectric material. Wind energy was used for providing rotation motion. As the
wind speed increased from 10 to 30 m/s, the voltage output also increases from 5.1
to 23.8 V [10].
In this paper, a sliding mode triboelectric energy harvester is modeled to predict
the harvested energy and numerical results were compared with the experimental
result.
Basic mechanism of sliding mode consists sliding motion of one triboelectric material
over other; generally, this motion can be achieved by two ways—to and fro motion
of slider over the base and rotating motion of one plate over another plate.
Mathematical Model of Sliding Mode Triboelectric Energy Harvester 1747
Fig. 1 a Initial position of TENG, b sliding between plates, c maximum sliding position, d sliding
back to initial position
As shown in Fig. 1a, it has two dielectric layers and two metal electrodes. In this
position, electric charge generates because of electronegativity difference between
two triboelectric materials, but in this position there is no electron flow between them
(I = 0).
Let L is the Length of the dielectric material and x is the displacement from initial
position. Bottom plate is fixed, and upper layer is moving with a constant velocity
in parallel direction of the bottom plate as shown in Fig. 1b. To balance the charge
generated in the dielectric, charges start generating on the electrode and due to
the potential difference between two electrode electrons start moving from bottom
electrode to upper electrode and current flows from top electrode to bottom electrode.
1748 T. P. Singh et al.
This is the position when upper plates slide to the maximum position of sliding. In
this situation, there is no current flow (I = 0) because both the electrodes and the
dielectric are balanced as shown in Fig. 1c.
In this position, top electrode starts sliding back to initial position, now upper plate
has more electron so to balance these electrons, electrons start moving from top
plate to bottom plate, so current direction is reverse, and it flows from bottom to top
electrode as shown in Fig. 1d.
3 Mathematical Modelling
1
V =− × Q + VOC (x) (1)
C(x)
Fig. 2
Dielectric-to-dielectric
sliding mode energy
harvester
Mathematical Model of Sliding Mode Triboelectric Energy Harvester 1749
−σ x
For bottom electrode =
(l − x)
σx
For top electrode =
(l − x)
VOC = E 1 d1 + E 2 d2
σx d1 d2
VOC = + (2)
ε0 (l − x) εr1 εr2
As dielectric thickness is very less compared to length, the capacitor between the
overlapped region will dominate the total capacitance. The total capacitance C can
be calculated as parallel plate capacitor model
ε0 w(l − x)
C= (3)
d1
εr1
+ εdr22
dQ
V =R (5)
dt
From Eqs. (4) and (5), generalized equation can be obtained-
dQ 1 d1 d2 σx d1 d2
R =− + Q+ + (6)
dt wε0 (l − x) εr1 εr2 ε0 (l − x) εr1 εr2
4 Numerical Results
Fig. 3 Charge–time relationship comparison between present result and published result
Mathematical Model of Sliding Mode Triboelectric Energy Harvester 1751
1 2 xmax xmax xmax
x = at for t < ; For time range of ≤t <2
2 a a a
2
1 xmax xmax
x = xmax − a 2 − t ; For t ≥ 2 , x = xmax
2 a a
As shown in Fig. 5, these curves are plotted for R = 100 M. The maximum
value of current or voltage is achieved for 0.08 s. For a value of x, number of charges
transferred increases, but after that its value get stabilized but current and voltage
values decrease. Maximum output obtained as current about 8 µA and voltage about
800 V as shown in Fig. 5b, c.
5 Experimental Results
Voltage(V)
0.4
2
0.2
0 1
-0.2 0 20 40 60
0
-0.4
0 20 40 60
-0.6 -1
Time(s) Time(s)
6 Conclusions
In summary, mathematical model for the sliding mode triboelectric energy harvester
is presented in this work. Based upon result obtained from mathematical model,
voltage, charge, and current are studied for sliding mode triboelectric energy
harvester. It is also finding out that as the surface charge density increases, the output
increases, so method like nanocoating may help to achieve a higher output. Based on
the output achieved from the mathematical model, these types of energy harvester
can be utilized to power small electronic devices like electronic sensors, GPS system,
pedometer, hand watch.
1754 T. P. Singh et al.
References
1. Horn RG, Smith DT (1992) Contact electrification and adhesion between dissimilar materials.
Science 256(5055):362–364
2. Horn RG, Smith DT, Grabbe A (1993) Contact electrification induced by monolayer
modification of a surface and relation to acid–base interactions. Nature 366(6454):442
3. Wang ZL, Lin L, Chen J, Niu S, Zi Y (2016) Triboelectric nanogenerators. Springer
International Publishing, Berlin, Germany
4. Yang B, Zeng W, Peng ZH, Liu SR, Chen K, Tao XM (2016) A fully verified theoretical analysis
of contact-mode triboelectric nanogenerators as a wearable power source. Adv Energy Mater
6(16):1600505
5. Niu S, Liu Y, Wang S, Lin L, Zhou YS, Hu Y, Wang ZL (2013) Theory of sliding-mode
triboelectric nanogenerators. Adv Mater 25(43):6184–6193
6. Fan FR, Tian ZQ, Wang ZL (2012) Flexible triboelectric generator. Nano Energy 1(2):328–334
7. Wang S, Lin L, Xie Y, Jing Q, Niu S, Wang ZL (2013) Sliding-triboelectric nanogenerators
based on in-plane charge-separation mechanism. Nano Lett 13(5):2226–2233
8. Zhu G, Chen J, Liu Y, Bai P, Zhou YS, Jing Q, Pan C, Wang ZL (2013) Linear-grating
triboelectric generator based on sliding electrification. Nano Lett 13(5):2282–2289
9. Pan L, Wang J, Wang P, Gao R, Wang YC, Zhang X, Zou JJ, Wang ZL (2018) Liquid-FEP-based
U-tube triboelectric nanogenerator for harvesting water-wave energy. Nano Res 11(8):4062–
4073
10. Park S, Ryu H, Park S, Hong H, Jung HY, Park JJ (2017) Rotating triboelectric generator using
sliding contact and noncontact from 1D fiber friction. Nano Energy 33:184–194
Free and Forced Flexural Vibration
Responses of the Laminated Composite
Plates
Abstract In this paper, the investigation of the free and forced vibration response of
the laminated composite plate is carried out. This analysis is based on the first-order
shear deformation theory (FSDT) in conjunction with the finite element method.
The steady-state response has been obtained in the time domain using the Newmark
method. Free vibration responses of laminated composite plates obtained have been
compared with the results available in the literature and by modeling in ANSYS
showing excellent agreement. The parametric study has been accomplished to inves-
tigate the effect of geometric parameters thickness ratio, load amplitude, and damping
factor on the free and forced vibration behavior of the laminated composite plate.
1 Introduction
N. Ahmad · A. H. Khan
Mechanical Engineering Department, AMU, Aligarh 202002, Uttar Pradesh, India
M. Amir (B)
School of Engineering, Indian Institute of Technology Mandi, Suran 175005, Himachal Pradesh,
India
[2] obtained the numerical solution of static, dynamic, free vibration, and buck-
ling behavior by applying the finite element method and various plate theories.
Somashekar et al. [3] introduced a field consistent four-noded (5 DoF per node)
anisotropic plate/shell element to analyze the degree of anisotropy and their effect
on in-extensional bending of corresponding shapes. Reddy [4] investigated the vibra-
tion behavior of the anti-symmetric angle ply laminated plates by taking transverse
shear deformation into account using the finite element model. Ganapathi et al. [5]
study the free vibration analysis of simply supported composite laminated panels in
the framework of FSDT and obtained governing equations using the energy method.
Bhar et al. [6] analyzed the significance of the higher-order shear deformation theory
(HSDT) over the FSDT for the laminated composite plates with the help of the finite
element method. Recently, Amir and Talha [7] presented a nonlinear finite element
model based on HSDT for the functionally graded porous composite curved panels.
Khan and Patel [8] studied the nonlinear vibrational behavior of laminated
composite plates under the periodic excitation by considering the geometric nonlin-
earity in von Kármán sense. Kapuria and Yasin [9] developed a coupled efficient
layer-wise (zigzag) theory for piezoelectric laminated composite and sandwich
plates. Patel et al. [10] projected a constitutive model for the analysis of bimodular
and laminated composite structures. Kapuria and Alam [11] developed a 1D beam
element based on the layer-wise (zigzag) theory for the dynamic analysis of the
composite beam with piezoelectric layers using the Newmark direct time integration
method.
The free and forced vibration behavior of the laminated composite plates is the
main concern of this study. The 8-noded finite element model based on the FSDT
has been considered here. An additional drilling degree of freedom (DoF) is added
with the usual 5 DoF per node to use the transformation. In order to study the forced
vibration response, the laminated composite plate is excited by harmonic forces with
the forcing frequency varying in the neighborhood of first fundamental mode. The
forced vibration responses have been obtained by solving governing equations in the
time domain using an implicit integration technique based on the Newmark method.
The results obtained have been validated with those available in the literature. The
influence of the thickness ratio, load amplitude, and damping factor on the dynamic
responses of the laminated composite plates has been obtained.
2 Theoretical Formulations
The displacement model adopted for the plates is based on FSDT includes the in-plane
displacement and at mid-plane, an out of plane displacement, w0 , of the mid-plane
and two rotation θx and θ y . The displacement field is assumed as follow:
Free and Forced Flexural Vibration Responses … 1757
where U , V, and W are the displacements along the x, y, and z-directions and θx andθ y
are the linear rotations of initially vertical planes on the x–z and y–z planes after
distortion, at the location (x, y) on the mid-plane. The mid-plane displacements in
the x, y, and z-direction are u 0 , v0, and w0 . The strain–displacement relation for the
composite plate is as:
∂u ∂u 0 ∂θx
εx = = +z
∂x ∂x ∂x
∂v ∂v0 ∂θ y
εy = = +z
∂y ∂y ∂y
∂u ∂v ∂u 0 ∂θx ∂v0 ∂θ y
γx y = + = +z + +z
∂y ∂x ∂y ∂y ∂x ∂x
∂w ∂u ∂w0
γx z = + = + θx
∂x ∂z ∂x
∂w ∂u ∂w0
γ yz = + = + θy (2)
∂x ∂z ∂y
{σ } = [Q]{ε} (3)
The Newmark β method is a numerical integration method used for solving the
differential equations as:
kˆ xi = Pi ˆ (12)
γ 1
kˆ = K + c+ m (13)
β t β t2
m γc m γ
Pˆ = Pi + + ẋi + + t − 1 c ẍ1 (14)
β t β 2β 2β
Free and Forced Flexural Vibration Responses … 1759
Pi
xi = (16)
kˆ
1 1 x˙i 1x¨i
x¨i = xi − − (17)
β t2 β t 2β
γ x˙i γ γ
x˙i = − xi˙+ t (1 − )x¨i (18)
β t β 2β
The free vibration analysis has been achieved using the standard eigenvalue
approach by treating external forces and damping to be zero. The forced vibra-
tional responses of the composite plates under the harmonic excitations are obtained
by employing the Newmark time integration scheme and solving the equation of
motion in the time domain.
To make sure the accuracy of the present model, convergence and validation are
− √
performed. The convergence study of frequencies, ω=ωa 2 / h ρ/E 2 , for the four-
layered [0°/90°/90°/0°] square plate with simply supported (SSSS) boundary condi-
tions. The square (a/b = 1) composite plates with b/h = 100, and material properties
E 1 /E 2 = 40, G 23 = 0.5E 2 , G 12 = G 13 = 0.6E 2 , ν12 = 0.25 are considered. It
is evident from Table 1 that 10 × 10 mesh sizes give fairly good converged results
in terms of the first four natural frequencies. The present results are also compared
with the results acquired from different theories as given by Reddy and Phan [12]
and found a good agreement between the results.
Table 2 presents the fundamental frequency of the symmetric four-layered cross-
ply square plate with SSSS boundary conditions. The material properties of the
composite plate E 1 /E 2 = 40, G 12 = G 13 = 0.6E 2 , G 23 = 0.5E 2 , ν12 = 0.25 are
considered. The present results are compared with those of Reddy and Phan [12] and
those obtained from FEM-based software ANSYS and found an excellent agreement
between the results.
In Table 3, the effect of the thickness ratio on the free vibration of laminated
composite plates with simply supported boundary conditions is investigated. A square
composite plate with four layers having orientation [0°/90°/90°/0°] and material
properties of carbon epoxy: E 1 = 180× 109 Pa,E 2 =E 3 = 10 × 109 Pa, G 12 =
7.3 × 109 Pa, G 13 =G 23 = 6.1 × 109 Pa, ν12 = ν23 = 0.28, ν13 = 0.35 and ρ =
1600 kg/m3 is analyzed for different thickness ratios. Table 3 depicts the influence
1760 N. Ahmad et al.
Table 1 Convergence of frequencies for four-layered plate with a/b = 1, b/h = 100
Mesh size Mode sequence number
1 2 3 4
4×4 18.87 34.46 68.92 71.83
5×5 18.85 33.91 67.70 68.22
6×6 18.84 33.79 66.53 67.98
7×7 18.84 33.74 66.08 67.89
8×8 18.84 33.72 65.87 67.84
9×9 18.84 33.71 65.75 67.81
10 × 10 18.84 33.70 65.69 67.80
11 × 11 18.84 33.70 65.69 67.80
CPT [12] 18.804 – – –
FSDT [12] 18.751 – – –
HSDT [12] 18.755 – – –
Table 3 Effect of thickness ratio on the natural frequencies (in Hz) of composite plates
b/h Mode sequence number
1 2 3 4 5
50 107.15 201.37 366.7 388.24 422.38
75 71.639 134.84 247.2 261.09 284.74
100 53.78 101.28 186.1 196.43 214.41
125 43.046 81.086 149.1 157.38 171.85
of thickness ratio on the natural frequency. It can be seen that there is a decrease in
natural frequency on increasing the thickness ratio.
Now, consider a plate is subjected to uniformly distributed harmonic force as
shown in Fig. 1. For the forced vibration, the square four-layered composite plates
having orientation [0°/90°/90°/0°] and material properties of carbon epoxy: E 1 =180
× 109 Pa,E 2 = E 3 = 10 × 109 Pa, G 12 = 7.3 × 109 Pa, G 13 = G 23 = 6.1 × 109 Pa,
ν12 = ν23 = 0.28, ν13 = 0.35 and ρ = 1600 kg/m3 and simply supported boundary
Free and Forced Flexural Vibration Responses … 1761
conditions are taken. The influence of the thickness ratio on the forced vibration
response of laminated square composite plates subjected to uniformly distributed
harmonic force is displayed in Fig. 2. It is clear from Fig. 2 that with the increase of
thickness ratio, the peak amplitude increases. The non-dimensional peak amplitude
corresponding to thickness ratio b/h = 75 being 135.75 compared to 326.1 for b/h
= 100.
Figure 3 shows the effect of the damping factor on the forced vibration responses of
laminated square composite plates under the uniformly distributed harmonic force.
It can be observed from Fig. 3 that with the increase of damping factor, the peak
amplitude decreases. The non-dimensional peak amplitude corresponding to ζ =
0.001 being 485.8 compared to 86.3 for ζ = 0.05.
The influence of load amplitude on the forced vibration behavior has been depicted
in Fig. 4. It is evident from the figure that the peak amplitude decreases with the
increase in load amplitude. The non-dimensional peak amplitude corresponding to
150
75
0
0.8 0.9 1.0 1.1 1.2 1.3
/ n
1762 N. Ahmad et al.
W/h
200
100
0
0.8 0.9 1.0 1.1 1.2
n
45
40 F 0
= 100 Pa
F 0
= 500 Pa
35 F = 1000 Pa
0
30
25
w/h
20
15
10
0
0 .8 0 .9 1 .0 1 .1 1 .2
n
Fig. 4 Effect of frequency ratio and forces applied on the central deflection of a clamped laminated
composite plate with layer arranged at 0°/90°
F 0 = 1 kPa being 40.507 compared to 4.0507 for F 0 = 100 Pa. This increase in the
displacement amplitude is proportional to increase in the load amplitude.
4 Conclusions
In this paper, the free and forced vibration responses of the laminated composite
plates have been analyzed. The present finite element model is based on FSDT
Free and Forced Flexural Vibration Responses … 1763
and the steady-state responses have been obtained in the time domain using the
Newmark time integration scheme. The effect of geometric parameters, load ampli-
tude, and damping factor on the free as well as forced vibration characteristics has
been obtained. The following conclusions can be drawn from the parametric study
are:
1. The fundamental frequency of the laminated composite plate increases with the
decrease in the thickness ratios.
2. The peak amplitude decreases with the increase of damping factor.
3. The peak amplitude increases with the increase in thickness ratio of composite
plate.
References
1. Reddy JN (2004) Mechanics of laminated composite plates and shells: theory and analysis.
CRC Press, Boca Raton
2. Vuksanovic DJ (2000) Linear analysis of laminated composite plates using single layer higher-
order discrete models. Compos Struct 48:205–211
3. Somashekar B, Prathap G, Ramesh B (1987) A field consistent four-node laminated anisotropic
plate/shell element. Comput Struct 25:345–353
4. Reddy JN (1979) Free vibration of anti-symmetric, angle ply laminated plates including
transverse shear deformation by the finite element method. J Sound Vib 66:565–576
5. Ganapathi M, Kalyani A, Mondal B, Prakash T (2009) Free vibration analysis of simply
supported composite laminated panels. Compos Struct 90:100–103
6. Bhar A, Phoenix SS, Satsangi SK (2010) Finite element analysis of laminated composite
stiffened plates using FSDT and HSDT: a comparative perspective. Compos Struct 92:312–321
7. Amir M, Talha M (2019) Nonlinear vibration characteristics of shear deformable function-
ally graded curved panels with porosity including temperature effects. Int J Press Vessels Pip
172:28–41
8. Khan AH, Patel BP (2014) Nonlinear forced vibration response of bimodular laminated
composite plates. Compos Struct 108:524–537
9. Kapuria S, Yasin MY (2013) A nonlinear efficient layerwise finite element model for smart
piezolaminated composites under strong applied electric field. Smart Mater Struct 22:055021
10. Patel BP, Khan K, Nath Y (2014) A new constitutive model for bimodular laminated structures:
application to free vibration of conical/cylindrical panels. Compos Struct 110:183–191
11. Kapuria S, Alam N (2006) Efficient layerwise finite element model for dynamic analysis of
laminated piezoelectric beams. Comput Methods Appl Mech Eng 195:2742–2760
12. Reddy JN, Phan ND (1985) Stability and vibration of isotropic, orthotropic and laminated
plates according to a higher order shear deformation theory. J Sound Vib 98(2):157–170
Shape Control of Piezolaminated
Structure Using Poling Tuned
Piezoelectric Actuators
1 Introduction
Sullivan et al. [13]. Static shape control of smart beams and plates laminated with
sinusoidally shaped sensors and actuators was studied by Lin and Hsu [14]. Lee and
Saravanos [15] presented a mixed multi-field FE formulation for shape control of
thermopiezoelectric composite shells. The genetic algorithm (GA) was applied to
obtain the optimal design and control of beams and plates by Ribeiro et al. [16] using
finite-element method. Optimal shape control of FG structures was studied by Liew
et al. [17] using different optimal algorithms.
Although good as smart structures, piezoelectric materials, have not been focused
upon by the researchers for enhancement in their inherent properties. One such
attempt was made by Kiran et al. [18], introducing the idea of altering the poling direc-
tion of piezoelectric materials to enhance their electromechanical properties. This
study is focused on utilizing the method of poling direction tuning in shape control
of smart structures. Different piezoelectric materials, which includes PMN-0.35PT,
Pb[Zrx Ti1-x ]O3 (PZT-7A, PZT-5A, Pb(Zr0.5 4Ti0.46 )O3 ,), x[Ba(Zr0.2 Ti0.8 )O3 ]-(1-
x)[(Ba0.7 Ca0.3 )TiO3 ] (BCT-0.48BZT) and (K0.44 Na0.52 Li0.04 )(Nb0.8 4Ta0.10 Sb0.06 )
(KNLNTS), are used one by one and their control performance is presented. Next,
piezoelectric materials are simulated for active vibration control of the same structure,
with poling tuning.
2 Methodology
K φu {q} − K φφ φ p = Fq (2)
where,
[K ] Elastic stiffness matrix,
L T
K uφ = K φu Elastic–electric coupling stiffness matrix,
K φφ Electric stiffness matrix,
[K uθ ] Elastic–thermal coupling stiffness matrix,
{F
m } Mechanical force vector,
Fq Electric charge vector,
{q} Nodal displacement vector,
{φ} Electric charge vector.
Equation (1) can be written as,
[K ]{q} = {Fm } − K uφ {φ} (3)
where the term K uφ {φ} is known as actuator force, which is generated in response
to the actuator voltage {φ}.
Piezoelectric and Elastic Constitutive Relations
The converse piezoelectric effect is the development of mechanical strain due to
applied electric field.
The constitutive equations for direct and converse piezoelectric effect are given
as [19],
where {D}, {e} and [b] are reduced elastic constants’ matrix, piezoelectric constants’
matrix and permittivity matrix, respectively. {P}, {σ }, {}, {E} are electric displace-
ment, stress vector, infinitesimal strain vector and electric field, respectively.
Equations (3) and (4) serve as the governing law of piezoelectricity for sensor and
actuator patches. Further, to define the mechanical behavior of the structure under
external stimuli, elastic constitutive law, also known as the stress–strain relationship,
is required. In the local coordinate system, the stress–strain relationship is given as,
σ = [D] ε (6)
Shape Control of Piezolaminated Structure … 1769
where
σ Stress components in the local coordinate system,
Strain components in the local coordinate system, and.
[D] Material stiffness matrix in terms of engineering constants.
Using transformations to write Eq. (5) in the global coordinate system as
where
{σ } Stress components in the global coordinate system,
{} Strain components in the global coordinate system, and.
[DTr ] Transformed material stiffness matrix in terms of engineering constants.
Fig. 2 Effect of poling tuning on piezoelectric materials’ performance in direct piezoelectric effect
1770 S. Sharma et al.
{P} = [a][d]T [α]−1 {σ } + p [E] (9)
where [α] and [a] are transformation matrices obtained from direction cosines of the
rotation, written as;
⎡ ⎤
l12 m 21 n 21 2m 1 n 1 2n 1l 1 2l1 m 1
⎢ l 2 m 2 n2 2l2 m 2 ⎥
⎢ 2 2 2 2m 2 n 2 2n 2 l2 ⎥
⎢ 2 ⎥
⎢ l3 m 23 n 23 2m 3 n 3 2n 3l3 2l3 m 3 ⎥
α=⎢ ⎥ (10)
⎢ l2 l3 m 2 m 3 n 2 n 3 m 2 n 3 + n 2 m 3 l2 n 3 + n 2 l3 m 2 l3 + l2 m 3 ⎥
⎢ ⎥
⎣ l3 l1 m 3 m 1 n 3 n 1 m 1 n 3 + n 1 m 3 l1 n 3 + n 1 l3 m 1 l3 + l1 m 3 ⎦
l1 l2 m 1 m 2 n 1 n 2 m 1 n 2 + n 1 m 2 l1 n 2 + n 1 l2 m 1 l2 + l1 m 2
and,
⎡ ⎤
l1 m 1 n 1
a = ⎣ l2 m 2 n 2 ⎦. (11)
l3 m 3 n 3
and,
T
p = [a] p [a]−1 . (13)
A bimorph beam is considered for analyzing its actuation capabilities and the results
are validated against the studies conducted by Nestorovic et al. [20]. Two different
materials are considered for validating the actuation capabilities of the bimorph beam.
In first case, polyvinylidene fluoride (PVDF) is used as the material of the beam and
in second case a piezoelectric ceramic PIC151 is utilized for modeling the same
beam. The material properties of the beam are given in Table 1.
A five-element mesh in the length direction is used to discretize the beam. A
voltage difference of 1 V is applied across the thickness of the beam. The comparison
of the results obtained with those of Nestorovic et al. is shown in Tables 2 and 3 for
PVDF and PIC 151, and a very good agreement is observed (Figs. 3and 4).
Shape Control of Piezolaminated Structure … 1771
Fig. 4 Comparison of the actuation results of bimorph beam for PVDF and PIC 151
A piezolaminated square plate, simply supported (as shown in figure) and subjected
to a point load of 50 N at center, is numerically simulated for shape control. The plate
has a length of 0.2 m and thickness 1 mm. The host structure under consideration
consists of a square plate of aluminum (Young’s modulus E = 70 GPa, Poisson’s
ratio ν = 0.3 and density ρ = 2707 Kg/m3 ). Layers of piezoelectric material of
thickness 0.5 mm are bounded on the top and bottom surfaces of this plate. A 16 ×
16 FEM mesh is used to calculate the static deflection of the plate (Fig. 5).
Three different materials, PMN-0.35PT, Pb(Zr0.54 Ti0.46 )O3 ,, and PZT-7A, are used
as actuators for shape control of the plate structure. The properties of the materials are
Host layer
Shape Control of Piezolaminated Structure … 1773
Fig. 6 Center line deflection of SSSS plate under a center point load of 50 N with actuators of
a PMN-0.35PT, b Pb(Zr0.54 Ti0.46 )O3 , and c PZT-7A
used as given in the reference Fig. 6 shows the comparison of actuation performance
of these materials in shape control of the plate.
To take advantage of higher values of d15 and d33 while still operating in d31 mode,
the poling direction is changed, and it is observed that the value of the coefficient d31
changes as a function of coefficients d31 , d15 , and d33 . This way transverse, shear
and longitudinal (i.e., d31 , d15 , and d33 ) modes are activated simultaneously. The
method of controlling the effective values of piezoelectric constants by changing
poling direction is termed as poling tuning. The effective values of these coefficients
with the poling angle are given by Eqs. (12) and (13). The variation of properties
with the poling angle of piezoelectric materials are plotted in Fig. 7. PMN-0.35PT is
found to have a maximum increase in the value of d31 as shown in Fig. 7a, whereas Pb
(Zr0.54 Ti0.46 ) O3 and PZT-7A have comparatively less enhancement in d31, whereas
KNLNTS has a negligible enhancement in d31 . By expanding Eq. (12), it can be
observed that the effective value of d31 depends on the difference between d31 and
d15 of a material, i.e., higher the difference between d31 and d15 , higher will be the
effect of poling angle on the value of effective d31 . However, it is observed that the
other two materials, i.e., BCT-0.48BZT, PZT-5A, do not show an improvement but
a decrease in the value of d31 , after the change in poling direction. Thus, the concept
of poling tuning can be vital in improving the control performance of piezoelectric
1774 S. Sharma et al.
4 Conclusion
Fig. 8 Effect of poling tuning on shape control performance of a PMN-0.35PT, b Pb (Zr0.54 Ti0.46 )
O3 , and c PZT-7A actuator patches
conducting shape control studies on the plate and are compared for their effective-
ness in controlling the deformations of the structure. PMN-0.35PT is found to have
superior control performance than other materials under consideration. This because
of its high magnitude of the piezoelectric coupling coefficient d31 . Next, poling tuning
is incorporated to enhance the performance of piezoelectric patches and its impact
on the control behavior is examined. Three out of six materials are found to show an
improvement in performance at optimum poling angle. A significant improvement in
control performance of PMN-0.35PT is observed. However, other two materials, i.e.,
Pb (Zr0.54 Ti0.46 ) O3 and PZT-7A, show a comparatively low improvement in perfor-
mance. This behavior of performance enhancement can be attributed to the varying
difference in d15 and d31 for different materials. In general, more the difference in
the values of two piezoelectric coupling coefficient, more will be the variation in the
effective value of d31 with the change in poling direction.
1776 S. Sharma et al.
References
1. Haftka RT, Adelman HM (1985) An analytical investigation of shape control of large space
structures by applied temperatures. AIAA J 23.3(1985):450–457
2. Agrawal SK, Daqun T, Nagaraja K (1994) Modeling and shape control of piezoelectric actuator
embedded elastic plates. J Intell Mater Syst Struct 5.4(1994):514–521
3. Koconis DB, Kollar LP, Springer GS (1994) Shape control of composite plates and shells with
embedded actuators 1. voltages specified. J Compos Mater 28.5(1994):415–458
4. Irschik H, Heuer R, Ziegler F (1997) Static shape control of redundant beams and trusses by
thermal strains without stress. In: Hetnarski RB, Noda N (eds) Proceedings of 2nd international
congress on thermal stresses, pp 469–472
5. Donthireddy P, Chandrashekhara K (1996) Modeling and shape control of composite beams
with embedded piezoelectric actuators. Compos Struct 35.2(1996):237–244
6. Koconis DB, Kollar LP, Springer GS (1994) Shape control of composite plates and shells with
embedded actuators. II. Desired shape specified. J Compos Mater 28(3):262–285
7. Agrawal BN, Treanor KE (1999) Shape control of a beam using piezoelectric actuators. Smart
Mater Struct 8.6(1999):729
8. Agrawal BN, Elshafei MA, Song G (1997) Adaptive antenna shape control using piezoelectric
actuators. Acta Astronaut 40.11 (1997): 821–826
9. Yoon H-S, Washington G, Theunissen WH (2000) Analysis and design of doubly curved
piezoelectric strip-actuated aperture antennas. IEEE Trans Anten Propag 48(5):755–763
10. Kapuria S, Sengupta S, Dumir PC (1997) Three-dimensional piezothermoelastic solution for
shape control of cylindrical panel. J Therm Stress 20.1(1997):67–85
11. Chandrashekhara K, Varadarajan S (1997) Adaptive shape control of composite beams with
piezoelectric actuators. J Intell Mater Syst Struct 8.2(1997):112–124
12. Choi, J.S., Ashida, F. and Noda, N. (1997) Control of thermally induced elastic displace-
ment of an isotropic structural plate bonded to a piezoelectric ceramic plate. Acta mechanica
122.1(1997):49–63
13. Sullivan JM, Hubbard JE Jr, Burke SE (1997) Distributed sensor/actuator design for plates:
spatial shape and shading as design parameters. J Sound Vib 203(3):473–493
14. Lin CC, Hsu CY (1999) Static shape control of smart beam plates laminated with sine sensors
and actuators. Smart Mater Struct 8.5(1999):519
15. Varelis D, Saravanos DA (2006) Coupled mechanics and finite element for non-linear laminated
piezoelectric shallow shells undergoing large displacements and rotations. Int J Numer Methods
Eng 66.8(2006):1211–1233
16. Ribeiro R, da Mota Silva S, Rodrigues JD, Vaz MAP (2000) Genetic algorithms for optimal
design and control of adaptive structures. In: Smart structures and materials 2000: Mathematics
and control in smart structures, vol 3984, pp 268–278
17. Liew KM, He XQ, Meguid SA (2004) Optimal shape control of functionally graded smart
plates using genetic algorithms. Comput Mech 33.4(2004):245–253
18. Kiran R, Kumar A, Kumar R, Vaish R (2018) Poling direction driven large enhancement in
piezoelectric performance. Scr Mater 151:76–81
19. Dosch JJ, Inman DJ, Garcia E (1992) A self-sensing piezoelectric actuator for collocated
control. J Intell Mater Syst Struct 3(1):166–185
20. Nestorović T, Marinković D, Chandrashekar G, Marinković Z, Trajkov M (2012) Implementa-
tion of a user defined piezoelectric shell element for analysis of active structures. Finite Elem
Anal Des 52:11–22
Power Optimization of a Wind Turbine
Using Genetic Algorithm
Prakash Poudel, Rajeev Kumar, Vikas Narain, and Satish Chandra Jain
Abstract Modeling and optimization of the wind turbine for better power and perfor-
mance have become the demand in the development of renewable energy. Power
coefficient (Cp) is an important parameter which determines the efficiency of the
wind turbine and it depends on the velocity of the wind, blade pitch angle, and tip
speed ratio of the turbine. Selection of the appropriate value of these parameters
while designing a wind turbine will provide the optimum value of coefficient of
performance. In this paper, the optimum value of the power coefficient is obtained
by using the genetic algorithm optimization technique. The optimum value of the
power coefficient is found to be 0.46 which is increased by 0.05 than that of the value
obtained from blade element momentum theory.
1 Introduction
Wind turbine is a device that converts the kinetic energy of wind into mechanical
energy by rotation of shaft, which is then converted to electrical energy by the use
of generator. Research on modeling and optimizing of various design parameters
of wind turbine to obtain the optimum performance has been carried out a long
time before. Sedaghat et al. [1] have performed analysis of compact blade element
momentum theory to design blades of continuously varying speed horizontal axis
wind turbine for optimum performance. Jain et al. [2] have presented a optimization
technique based on genetic algorithm and estimated the optimum power of the wind
turbine on the basis of parameters like blade rotating speed blade radius tip speed
ratio, electrical power, etc. Mahmuddin [3] has developed a computational model
based on blade element momentum theory by dividing the blade into several segments
and analyzed different parameters required for optimal blade design process. In the
analysis, tip and root losses are analyzed and the computational results are vali-
dated by comparing with the results obtained from QBlade software. Shen et al.
[4] described an optimized model of horizontal axis wind turbine that is based on
lifting surface method and genetic algorithm. In this model, chord length of airfoil
and twist distribution of the blade is the decision variables that are analyzed by using
micro-genetic algorithm in order to attain the best trade-off between the maximum
annual energy production and minimum blade loads.
Designing of wind turbine blade using blade element momentum method and
optimizing the model with the use of non-classical optimization techniques will
allow the wind energy farms to produce the best energy under different aerodynamic
and loading conditions. In this paper, first, a wind turbine model is designed using
the blade element momentum theory, and the model is then optimized for better
performance using genetic algorithm. The velocity data required for the operation
of wind turbine is collected from a selected site, and other required parameters
are determined to form blade element momentum theory analysis. The optimized
parameters of the wind turbine presented in this study are the wind velocity and the
radius of the turbine blade. The optimization of these parameters provides better
power performance of the wind turbine.
The objective of an aerodynamic design of the wind turbine is to maximize the effi-
ciency for given airfoil of the blade. Blade element momentum (BEM) also referred
to as strip theory combines both momentum and element theory to provide useful
relations for determining the parameters for wind turbine design [3].
In momentum theory, analysis of the forces acting at the blade is performed that is
based on the conservation of linear and angular momentum. The momentum theory
analysis of wind turbine can be observed inside a stream tube in which the wind
turbine is placed inside the stream tube and the wind is allowed to flow through one
end as shown in Fig. 1.
The thrust force acting on a wind turbine blade inside a stream tube as can be
obtained as [5]:
where ‘v1 ’ is the downstream wind velocity, ‘ρ’ is the air density, ‘r’ is the distance
of the element from hub, and ‘a’ is the axial induction factor which could be written
as:
v1 − v2
a= (2)
v1
If a rotating stream tube is introduced, then the torque that acts on the blade is
given by:
dT = 4a (1 − a)ρv1 r 3 π dr (3)
where ‘w’ is angular velocity, ‘’ is the blade rotation speed, and ‘a ’ is angular
induction factor and is given by:
w
a = (4)
2
The forces on the blades of a wind turbine can also be expressed as a function of lift
and drag coefficients and the angle of attack. For blade element theory analysis, the
whole blade is divided into smaller elements of infinitesimal thickness and assumed
that there is no aerodynamic interactions between different blade elements as shown
in Fig. 2.
The geometry of the blade along with different forces acting on the blade profile
is shown in Fig. 3. The forces that act on the blade elements can be determined by
the lift and drag coefficients as given in the equations below [5]:
1780 P. Poudel et al.
1
dFT = BρW 2 (C L sin β + C D cos β)cdr (5)
2
1
dT = BρW 2 (C L sin β − C D cos β)cr dr (6)
2
where ‘C L ’ is the lift coefficient, ‘C D ’ is the drag coefficient, ‘B’ is the number of
blades, ‘β’ is the relative inflow angle, ‘c’ is the chord length of an airfoil, and ‘W ’
is the resultant velocity. The relative angle ‘β’ and chord length ‘c’ of the airfoil at
different sections of the blade can be calculated as:
2 −1 1
β = tan (7)
3 λr
8πr
c= (1 − cos β) (8)
BCL
(1 − a)
Resultant velocity, W = (9)
cos β
Power Optimization of a Wind Turbine Using Genetic Algorithm 1781
Blade element momentum theory combines both the momentum and blade element
theory to derive the equations that is very useful to design the wind turbine. The
thrust force and the torque obtained from the blade element momentum theory can
be expressed in equations shown below [5]:
(1 − a)2
dFT = σ πρv2 (C L cos β + CD sin β)r dr (10)
cos2 β
(1 − a)2
dT = σ πρv2 (CL cos β − CD sin β)r 2 dr (11)
cos2 β
Bc
σ = (12)
2πr
In the wind turbine, there occur some losses in the tip of the turbine blade which is to
be accounted in blade element momentum theory as a correction factor. Correction
factor indicates the loss reduction of forces along the blades and its value varies from
0 to 1 given as [5]:
2 B R −r
Q= cos−1 exp − (13)
π 2 r sin β
where ‘R’ is the maximum radius of the rotor. The force and the torque equation after
accounting the tip loss factor can be expressed as:
a σ (C L sin β + C D cos β)
= (14)
1−a 4Q cos2 β
a σ (C L cos β + C D sin β)
= (15)
1 − a 4Qλr cos2 β
where ‘λr ’ is the local tip speed ratio and is given as:
r
λr = (16)
v
1782 P. Poudel et al.
The total power output that can be obtained from the rotor is a function of parameters
like velocity of wind, density of air, radius of the rotor, and coefficient of performance.
The total power of the rotor can be expressed as:
R
P= dT dr (17)
rh
P
Cp = (18)
Pwind
where ‘Pwind ’ is the power contained in the wind and is given as:
1
Pwind = ρπ R 2 v3 (19)
2
On solving the above three equations, we can obtain the Cp as a function of various
parameters required the design of wind turbine as [5]:
λ
8 CD
Cp = Qλ3r a (1 − a) 1 − tan β dλr (20)
λ2 CL
λh
where ‘λr ’ is tip speed ratio and is defined as the ratio between the tangential speed
of the tip of a blade and the actual speed of the wind. Tip speed ratio is expressed as:
R
λ= (21)
v
3 Genetic Algorithm
Genetic algorithm is a random search ant optimization technique which works on the
principle of genetics and natural selection. Genetic algorithm requires initialization
and representation of the population and appropriate fitness function. The need of
appropriate fitness function is to lead the algorithm to provide optimal solution.
Power Optimization of a Wind Turbine Using Genetic Algorithm 1783
The selection operator usually selects or reproduces the individuals that have better
fitness values as that of parents. The crossover operator creates the new population
by combining genes of the parents that is in the current population. The mutation
operator creates the entirely new population by changing the genes of individual
parents. At every step, the algorithm uses individuals of current generation to generate
new population, and fitness evaluation is carried out until the best population is
determined. The genetic algorithm can be stopped to achieve the desired value by
defining the limit of generation required for iteration. The flow diagram of genetic
algorithm showing different steps is shown in Fig. 4.
4 Computational Model
The computational model used for the study is an airfoil of NACA 4412 and the airfoil
that is obtained from the QBlade software is shown in Fig. 5. The graph between C L
versus C D and C L versus alpha is also obtained from the QBlade software as shown
in Figs. 6 and 7, respectively. The QBlade results shown below in the graph are quite
similar to the results obtained from the computational study of NACA 2415 [3]. From
the graph, it is seen that the value of C D and C L are 0.01 and 1, respectively. These
values of C D and C L are used for the calculation of coefficient of performance of the
wind turbine model presented in this study.
The different steps that are involved in the design of wind turbine blade using
blade element momentum theory are listed below.
1. Determine the radius of rotor for 4 KW power using other known parameters
such as velocity of wind and density of air.
2. Choose the design tip speed ratio value and number of blades for wind turbine.
3. Select the type of airfoil used in blade design (NACA 4412) and obtain the lift
and drag coefficients value using angle of attack.
4. Divide the blade into 10 elements and calculate the chord length, blade solidity,
axial and angular induction factors, and relative inflow angle for each sections.
5. Recompute the values of chord length, blade solidity, axial, and angular induc-
tion factors, and relative inflow angle and compare the values obtained from the
previous iteration.
6. Continue the iteration until the values get converged.
7. Calculate the tip loss correction factor.
8. Compute the value of coefficient of performance.
The velocity data of wind required for this study is collected from a site, and the
moving average of the collected data is taken. Different input parameters that are
required to design the wind turbine model for this study are listed in Table 1.
Fig. 6 C L versus C D
Power Optimization of a Wind Turbine Using Genetic Algorithm 1785
λ
82 4 CD
Objective function: C p = Qr a (1 − a) 1 − tan β dr
v2 R 2 CL (22)
λh
The control parameters that are required in genetic algorithm are chosen as:
1786 P. Poudel et al.
• Population size: 50
• Number of generations: 300
• Crossover fraction: 0.8
• Elite count: 2
• Pareto fraction: 0.035
• Penalty factor: 100
• Migration interval: 20
• Migration fraction: 0.2.
The genetic algorithm is terminated if any of the below specified criteria are met.
• Number of generations limit: 300
• Stall generations limit: 100
• Function tolerance: 1e-50
• Constraint tolerance: 1e-12.
The input parameters mentioned in Table 1 are used to calculate other important
parameters using the concept of blade element momentum theory mentioned in
Sect. 3. The blade is divide into 10 elements and the local tip speed ratio ‘λr ’,
relative inflow angle ‘β’, and the chord length ‘c’ at different sections of the blade
are calculated using MATLAB and tabulated in Table 2.
Tip loss correction factor is calculated to be 0.98. All the parameters obtained are
used to calculate the coefficient of performance ‘C p ’ value as 0.41 in MATLAB. The
plot obtained by plotting different functions in genetic algorithm code is shown in
Fig. 8. The plot consists of the subplots of best individual, score diversity, stopping,
and selection function. It can be seen that algorithm is converged to the best functional
value at 0.45 at radius of the blade 4.05 m and velocity of wind 6.6 m/s shown by
the current best individual subplot. The score histogram shows the score value of
individuals at each generation, whereas the selection histogram shows the parents
chosen for each generation. The optimization process is terminated because average
change in the fitness value is less than that of defined function tolerance value. The
number of generations used is 111 and the number of function evaluation is 5600.
The plots of C p with design variables show the increase in power coefficient for
the optimized blade by 0.05. The graph between power produced by the wind turbine
and radius of the blade in Fig. 11 shows the increase in power by 0.49 KW. Power
versus wind speed curve also shows the increase in power produced of the optimized
blade by 0.5 KW (Figs. 9, 10, and 12).
7 Conclusion
Wind energy is a very important source of renewable energy and proper modeling
of wind turbine is required for generating optimum power under various conditions.
Genetic algorithm is an important optimization tool that can be used to design an
appropriate wind turbine for specific site. The algorithm developed in the MATLAB
is based on different assumptions in which iterations are carried out until the desired
1788 P. Poudel et al.
result is not achieved. The present study is based on blade element momentum
theory and various wind parameters are analyzed to provide an optimum solution to
the problem. The optimized power is 500 W more than that is obtained from blade
element momentum theory calculations. Prediction of appropriate wind speed and
radius of the turbine blade gives the best performance of the wind turbine. We can
implement this design to promote the wind energy for best turbine efficiency in areas
where it is difficult to predict the parameters required to design wind turbine.
References
1 Introduction
1.1 Cams
A cam is an element of the cam-follower mechanical system that compels the move-
ment of the follower by direct contact. The motion of the follower is the result of a
programme. Just as a computer is programmed, so is a cam. Thus, the system can be
thought of as a mechanical information device. Accordingly, the goal of the designers
is to build a programme, establish the locus of the contact points between the cam and
follower, produce the cam profile coordinates system, and fabricate the cam within
an acceptable accuracy. After all the parts are assembled, the performance of the
cam-follower system is observed.
Before the start of CAD, the drawings are done in 2D design systems which are very
difficult to understand. It was very hard to create, modify, and identify interference
and errors. Even for a small mistake, the entire drawing should be recreated. 3D
CAD software uses an intuitive and concrete way to present objects to users, which
fundamentally avoids above-mentioned problems. By the evolution of time and tech-
nology, even CAD drawings need to be done at a faster pace in order to decrease the
lead time in designing. The design engineers are subjected to do a lot of design iter-
ations before finalizing a new product design. Since the iterations include repetitive
tasks, these tasks should be automated in order to save a lot of time. By automating
the repetitive tasks, it is observed that 90% of the time is saved in designing a product.
So, it is observed that design automation is inevitable in any manufacturing indus-
tries. In this paper, we are suggesting a method in order to automate the tasks while
designing in the commercial CAD package CATIA V5 with the help of VBA macro
programming. The automated design is also integrated with user interface so that
any layperson could be able to design the part/product skills in CAD drawing.
Previous study by [1] shows us a study of macros programming as parametric
method of the commercial aircraft design. The developed macro programming was
used to compare the two different models of Boeing 737–400 and Boeing 777. It
is found that advantages of macros programming allowing flexibility for design
exploration and increase the usability of the design solution.
Instead of repeatedly creating the similar models and nearly 80% of design activ-
ities have a routine nature, engineers should be able to create automation programme
so that it automatically creates the new CAD models or to generate the further
Automated Design for Cam Profile Using CATIA V5 … 1793
modification of the same model. This automation process helped in reducing the
development time of a product especially in mass production industry [2].
The shift from manual modelling method to CAD automation process can be
done by using VB programming in CATIA V5. The automation and GUI creation
make file handling easy and reduce the time for designing tremendously [3, 4]. The
automation was illustrated using 3D modelling of helical spring [5].
A component under a stress level that exceeds the material tensile strength will fail.
This often calls for the application of factor of safety to many engineering applica-
tions. There is, however, another factor to take into account when mechanical compo-
nents undergo repeated or cyclic stresses. It has been found that such components
will fail far below the tensile strength. In cam-follower motion, the cam is under-
going a repeated cyclic load due to the follower action and load. So, it is subjected
to dynamic loads and it may fail due to fatigue. Along with designing, a design is
concerned over the life of the designed part. Determining the life of a part will enable
the company to obtain an accurate forecast in the production management. It will
also help in analysing the PLM. So, determining the life of the designed part is quite
important in any manufacturing industry. Therefore, the analysis must be performed
to estimate the fatigue life of the component [6]. In this paper, the fatigue life of the
cam in cam and follower mechanism is simulated and analysed using ANSYS 18.
2 Automation Methodology
We have to automate the 3D model of the cam profile in CATIA V5. Here, we are
automating two types of cams, i.e. radial cam and offset cam according to the given
parameters which is to be given s input in the developed user interface in CATIA V5
itself.
Parameters for Input
For radial cam: base radius, type of follower movement, type of displacement motion
for rise and type of displacement motion for return, rise angle, return angle, and dwell
angles.
For offset cam: base radius, offset radius type of follower movement, type of displace-
ment motion for rise and type of displacement motion for return, rise angle, return
angle, and dwell angles.
1794 K. Theril et al.
x = (s + r ) ∗ cos(θ )
y = (s + r ) ∗ sin(θ )
where; so = r 2 − ro 2
Fig. 1 User interface for selecting the type of cam required to model (offset and radial)
3.1 Results
Offset cams and radial cams are the two major cams automated in the system. For each
type of cam, there three types of follower motions available (D-R-D-R-D, D-R-R-D,
R-D-R), and for every cam, there are three types of displacement motions available
(SHM, cycloidal, uniform velocity) for the rise and return of cam. In the beginning,
the user form to select which type of cam (radial or offset) will appear. It will then
be directed to another user interface to input the parameter values. Consequently,
the 3D CAD geometry of cam can be automatically created within few seconds in
CATIA V5. It could be generated by any user even if he/she is not skilled enough to
work on CATIA V5 (Figs. 1 and 2).
3.2 Observations
Following Tables 2 and 3 give the average time required for manual modelling done
by any proficient expert in CATIA V5 and automated modelling done by using system
developed in this work. It indicates that the required time for automated modelling
is very less than the manual modelling with the similar accuracy in both approaches.
1796 K. Theril et al.
Fig. 2 User interface for cam profile modelling (left image) and 3D CAD model of the cam
developed using automated method (right image)
Table 3 Average time taken for change in parameter manual modelling and automatic modelling
Type of cams Type of Change in Manual Automated
displacement parameter modelling (h) modelling (s)
motion
Radial cam SHM for both rise Base circle radius = 1 3–5
(R-D-R) and return 35 mm
Radial cam SHM for rise and Rise angle = 150° 2–2.5 5–10
(R-D-R) cycloidal for return
Radial cam SHM for rise and Rise angle = 135° 4–5 5–10
(D-R-D-R-D) cycloidal for return Return angle = 30°
Offset cam SHM for rise and Rise angle = 150° 4–6 5–10
(D-R-D-R-D) cycloidal for return Offset circle (r o ) =
20 mm
Automated Design for Cam Profile Using CATIA V5 … 1797
The life assessment is done in the ANSYS workbench. A flat faced reciprocating
follower is created along with the cam geometry and opened in static structural
workbench. The parameters of cam selected for analysis are offset D-R-D-R-D cam
with SHM rise and cycloidal return. The contact is taken as frictional with friction
coefficient 0.01 and other contact definitions are given appropriately. A longitudinal
spring is connected at one end of the follower with spring constant 5 N/m and with
a load of 5 N acting perpendicularly on it.
Static Structural Analysis: The follower is given a cylindrical joint at its one end
so that the follower can rotate according to cam movement in an axis perpendicular
to its plane of motion. The spring is given a load of 5 N on its top. The cam is rotated
at a speed of 1 revolution in 10 s. With the above boundary conditions and with the
given speed, the cam and follower motion are simulated.
• Contact: frictional
• Friction coefficient: 0.01
• Formulation: augmented Lagrange
• Normal stiffness: 0.1
• Longitudinal stiffness: 5 N/m (body to ground)
• Pinball region: 0.0002 m (Fig. 3).
• Element size: default
• Total no. of elements: 376
• Total no. of nodes: 2937
• Materials: both structural steel.
Fig. 3 Meshing
1798 K. Theril et al.
Fatigue life assessment: S–N curve approach and calculated equivalent alternating
stresses are used to obtain fatigue life of cam profile. In order to predict the fatigue
life of a cam profile using S–N curve method, we must obtain the equivalent stress
that occurs due to nominal alternating stresses and mean stresses. The equivalent
stress is then compared with the fatigue strength of the material, and the fatigue life
is found using the S–N curve. Equivalent stress equation is given by:
σa
σe = ,
1 − ( σσmx )n
5 Conclusions
The novelty of this paper is that a method is proposed to automate the repetitive tasks
in designing any mechanical part in CATIA V5 using VBA macros by taking the
example of cam profile. This automated design tool helps in filling the gap between
design engineers and computational experts when analysing product development
Automated Design for Cam Profile Using CATIA V5 … 1799
process. Also, the fatigue life of the cam in cam-follower mechanism is predicted
using ANSYS 18.
References
1. Ramly MR et al (2013) Automated a complex computer aided design concept generated using
macros programming. In: 2nd international conference on mechanical engineering research
(ICMER2013), IOP conference series: materials science and engineering 50-012041
2. Zibiciak M et al (2015) An automation of design and modelling tasks in NX Seimens with
original software - generator module. In: IOP Conf. Series: Materials Science and Engineering,
vol 95, p 012117
3. Ali MB et al (2018) Implementation of automation process in generating CAD model for rim
wheel. ARPN J Eng Appl Sci 13(2). ISSN 1819-6608
4. Kumbhar R, Jadhav D, Pawar S, Dhanrale N (2014) Customization of Catia V5 for creating
different types of holes on disc wheel. Int J Eng Res Technol 3(5):506–509
5. Siddesh S, Suresh BS (2015) (2015) Automation of generating CAD models. J Mechan Eng
Autom 5(3B):55–58
6. Shenoy PS (2004) Dynamic load analysis and optimization of connecting rod. Master’s thesis,
University of Toledo
7. Kumar A, Kadam V, Nimbalkar UM (2015) Automatic assembly modelling for product variants
using parametric modelling. Int J Eng Res Technol 4(04):79–89; 9th international proceedings
on proceedings, pp 1–2 (2010)
8. Mokhede Y, Chikshe S, Ghavate R (2017) Automation of bearing modelling and drafting
using CATIA–VBA International conference on ideas, impact and innovation in mechanical
engineering (ICIIIME 2017), vol 5, no 6, pp 471–477
9. Sawant YH, Nimbalkar UM(2015) Automated modelling of screw conveyor components in
CATIA. Int J Eng Tech Res (IJETR) 3(4):112–116
10. Joshi H, Ghole P, Shirse S, Madke S (2016) Automated CAD modelling of mechanical compo-
nents. In: International conference on communication and signal processing 2016 (ICCASP
2016)
11. Raotole ML et al (2013) Prediction of fatigue life of crank shaft using S-N approach. Int J
Emerg Technol Adv Eng 3(2)
12. . Romanowicz PJ, Szybiński B (2019) Fatigue life assessment of rolling bearings made from
AISI 52100 bearing steel. Institute of Machine Design, Cracow University of Technology, ul.
Warszawska vol 24, pp 31–155, Cracow, Poland
13. Dupare YB et al (2014) Fatigue anlaysis in connecting rod using ANSYS. Int J Modern Trends
Eng Res (IJMTER) 2(2). e-ISSN: 2349-9745, p-ISSN: 2393-8161
14. Rothbart HA (2003) Cam design handbook, 1st edn. McGraw-Hill Professional, pp 27–55
15. Norton RL (2002) Cam design and manufacturing handbook—2nd edn. Industrial Press, Inc.,
pp 17–212
16. Rattan SS (2014) Theory of machines, 4th edn. McGraw Hill Education (India) Pvt. Ltd., New
Delhi, pp 232–277
17. CATIA V5 automation Version 5 Release 19 (2009) Instructor Manual. https://www.academia.
edu/35569799/CATIA_V5_Automation_CATIA_V5_Automation_CATIA_V5_Training_
Foils
18. Dhavale AS et al (2012) Study of modelling and fracture analysis of camshaft. IJERA 2(6):835–
842
19. Hendrickson D, Fatigue failure due to variable loading. Department of Computer Science,
Physics and Engineering, University of Michigan-Flint, Flint
Vibration Control of Smart Cantilever
Beam Using Fuzzy Logic Controller
Abstract This paper presents active vibration control of smart cantilever beam
using collocated piezoelectric sensor and actuator. The vibrating response of piezo-
laminated cantilever beam is modeled using lumped mass approach. Fuzzy logic
controller is used to control the vibration, and 49 rules have been established to
develop the controller. Input sensor voltage and rate of change of sensor voltage are
considered as inputs while actuator voltage is considered as output. Eight combina-
tions of different membership functions have been considered. Finally, it has been
observed that Gaussian-type membership function controls the vibration fast.
1 Introduction
Structural vibration is one of the key issues in most of the industries such as mechan-
ical, defense, aerospace and communication. Surface precision requirements of struc-
tures used in aforementioned industries are rigorous, and structural vibrations make
them suitable for a long period of time. Active and passive vibration controls are
the two strategies implemented to attenuate the structural vibrations [1]. The passive
vibration control strategy involves the application of mass, spring and damper treat-
ments to the vibrating structures [2]. In active vibration control, the host structure is
sandwiched between two piezoelectric layers. One piezoelectric layer acts as a sensor
while the other piezoelectric layer acts as an actuator. As the structure gets external
disturbance, the sensor layer senses the disturbance and sends signal in the form of
voltage to controller [3]. The controller sends the voltage signal to the actuator which
after amplification is supplied to the actuator layer resulting in control force. In the
last few decades, the piezoelectric ceramics of lead zirconate titanate Pb(Zr1-xTix)O3
(PZT) family have been extensively studied by various research groups for active
vibration control. Kumar et al. [4] and Umeda et al. [5] were among the pioneers to
study the PZT generator and proposed an electrical equivalent model being converted
from mechanical lumped models of a mass, a spring and a damper that describe a
transformation of the mechanical impact energy into electrical energy in the PZT
material. Kasyap et al. [6] formulated a lumped element model that represents the
dynamic behavior of the PZT device in multiple energy domains and replaces them
with electric circuit components. The model has been experimentally verified by
using a one-dimensional beam structure. Huang et al. [7] and DeVoe et al. [8] did
the displacement and tip-deflection analysis along the beam and made a comparison
with the experimental results. However, both proposals were limited to the actuator
mode. Williams et al. [9] analyzed a PZT structure by using a single degree of freedom
mechanical model. However, the model did not extend to a bimorph multilayer struc-
ture. Lu et al. [10] improved the electrical model by adding an electromechanical
coupling that represents a dynamic behavior of the beam vibrating under a single
degree of freedom.
The fuzzy set theory was established by Zadeh, and it has been extensively
researched in various fields of engineering. The main advantage of fuzzy logic
controller (FLC) over conventional control approaches is that the FLC is consid-
ered artificial intelligence where control laws are designed by human intelligence
based on expert’s experience, not by a deterministic numerical calculation. FLC
does not require the accurate mathematical model of the controlled object, and it
can represent almost any deterministic controller. Therefore, FLC method has been
applied widely for active vibration control of flexible structures Sharma et al. [11],
Marinaki et al. [12] and Wei et al. [13]. In 1983, Brown and Yao [14] used the
fuzzy theory to the engineering structures at first time. In 1986, Juang and Elton [15]
adopted fuzzy logic to estimate the density of earthquake on the extent of damage
for the constructions. Park [16] established the approximate model of the driver, the
sensor and the fuzzy logic controller to solve the problems of vibrations for flexible
structure. The result indicated that the fuzzy logic control had the stronger robust
and self-adaptive for the linear and nonlinear system. Zeinoun and Khorrami [17]
presented an adaptive control scheme based on fuzzy logic theory for active vibra-
tion control of smart beam structure. Experimental results were compared with the
analytical results. Sharma et al. [11] used fuzzy logic controller to perform vibration
control of cantilever beam using piezoelectric sensors and actuators and validated
the same by performing experiments. Lin and Liu [18] experimentally controlled
the active vibration of cantilever beam using simplified fuzzy logic controller and
compared results with PD controller. Wei et al. [19] experimentally performed the
active control of vibration of cantilever beam embedded collocated piezoelectric
sensor/actuator pair using fuzzy logic controller. In this paper, smart cantilever beam
is simplified into lumped parameter model, and fuzzy logic controller is used for
vibration control.
Vibration Control of Smart Cantilever Beam … 1803
2 Mathematical Modeling
Figure 1a shows a three-layered smart cantilever beam with two piezoelectric layers.
The upper layer acts as sensor, and bottom layer acts as an actuator. The middle layer
is substrate layer, made of stainless steel (SUS-304). Both sensor and actuator layer
are made of PZT-PZNM ceramic. Figure 1b shows equivalent lumped parameter
model of a smart cantilever beam to predict the vibratory response of free end, and
Fig. 1c shows free body diagram of a lumped parameter model to derive the mathe-
matical equation. The lumped parameter model simplifies the behavior of spatially
distributed physical system into discrete entities that approximate the behavior of the
distributed system under certain assumptions. Sensor, actuator and substrate mate-
rial are parallel; therefore, equivalent stiffness of a smart cantilever beam can be
calculated as:
k = K S + Ka + K (1)
where K s , K a and K are the stiffness of sensor layer, actuator layer and substrate layer,
θs and θa are the electrically induced damping coefficient of sensor and actuator layer,
and c is the damping coefficient of smart cantilever beam.
Let z(t) is the displacement; then, the equation of motion of the mechanical system
is given by:
F
Sensor
layer
Actuator
layer
Substrate
layer
(b) (c)
(a)
Fig. 1 a Smart cantilever beam, b lumped parameter model of a smart cantilever beam and c free
body diagram of a lumped parameter model
1804 K. Singh et al.
⎛ ⎞
w ⎝
n1 n2
k= 3 n p E p h 3p + n s E s h 3s ⎠ (3)
4l p=1 s=1
where l is the length, w is the width of the beam, and np and ns are number of
layers in piezoelectric and substrate material. E p and E s are the Young’s modulus
of piezoelectric and substrate layer, respectively, and hp and hs are the respective
heights of piezoelectric and substrate layer.
Equivalent mass of the smart cantilever beam is given by [20]:
⎛ ⎞
n1
n2
m = 0.235wl ⎝ n pρph p + n s ρs h s ⎠ (4)
p=1 s=1
k
wn = (5)
m
c = 2ξ wn m (6)
D3 = d31 σ + ε33
s
E3 (7)
d31 kzlh ps εs V
D3 = − 33 (11)
2I hp
Vibration Control of Smart Cantilever Beam … 1805
ε33
s
(bl) d31 klh ps (bl)
where C P = hp
and θ = 2I
q + CP V = θ z (15)
Since there is no charge in the sensor layer, therefore {q = 0} in Eq. (15). Sensor
voltage and its rate of change with time are calculated as
θS z
VS = (16)
Cp
dVs θs dz
= (17)
dt C p dt
When the beam is deformed, the sensor layer produces sensor voltage and rate
of change of sensor voltage both the input is fed to the fuzzy logic controller. After
applying control rules, the controller sends a control signal to the actuator layer
in form of voltage called actuator voltage. A coupled equation of mechanical and
electrical system is used to control the vibration of the smart cantilever beam.
m z̈ a + cż a + kz a + θa Va = F − f a (18)
(a)
Deflection(mm)
Time (sec)
(b)
Fig. 3 a Different combinations for fuzzy membership functions. b Performance evaluation of
different combinations of membership functions
Fuzzy Rule Base Set of 49 rules is used in present fuzzy logic controller. To define
these membership functions and fuzzy rules input and output variables are described
by seven linguistic terms, positive large (P++), positive medium (P+), positive small
(P), zero (ZE), negative small (N), negative medium (N) and negative large (N++)
(Table 1).
1808 K. Singh et al.
Fig. 4 Fuzzy membership functions for input (sensor voltage and rate of change of sensor voltage)
and output (actuator voltage) variables
Material and geometric properties of the smart cantilever beam are given in Table
2. Equations (2), (16) and (17) are used to develop a Simulink model as shown in
Fig. 5 and used to calculate the value of deflection, sensor voltage and rate of change
Vibration Control of Smart Cantilever Beam … 1809
Fig. 5 Simulink model of smart cantilever beam to calculate deflection, sensor voltage and rate of
change of sensor voltage
of sensor voltage. In this Simulink model, step force of 0.1 N is applied for 0.1 s on
smart cantilever beam because of this force a structural deflection of ±5 mm, sensor
voltage ±1.2 V, rate of change of sensor voltage ± 30 V/s is calculated.
Material and geometric properties used are shown in Table 2 [20]:
The simulation output of the smart cantilever beam is shown below in Fig. 6 where
Fig. 6a shows deflection vs. time response, Fig. 6b shows voltage vs. time response,
and Fig. 6c shows rate of change of sensor voltage vs. time response.
The value of sensor voltage and rate of change of sensor voltage acts as an input for
the fuzzy logic controller to calculate actuator voltage. The value of actuator voltage
is substituted in Eq. (18) to control the vibration of smart structure. Figure 7 shows
Simulink model of smart cantilever beam using fuzzy logic controller. Figure 8 shows
comparisons between controlled and uncontrolled response with respect to time.
Further study will be performed to compare the controlled response from different
type of controller.
1810 K. Singh et al.
Fig. 7 Simulink model of smart cantilever beam using fuzzy logic controller
Vibration Control of Smart Cantilever Beam … 1811
Fig. 8 Simulink output of smart cantilever beam using fuzzy logic controller
5 Simulation Output
See Fig. 8.
6 Conclusion
In this paper, smart cantilever beam is simplified into lumped parameter model,
and fuzzy logic controller is used for vibration control. The sensor voltage and its
derivative are given as inputs to the controller which computes actuator voltage using
49 if–then type rules. It has been find out that combination of Gaussian membership
function at ±3 V of actuator voltage give, the best controlled response. This type
of simplified model would be a very powerful tool to guide the design of the device
structure and can be applied in real environment.
References
1. Fahy FJ (1998) Fundamentals of noise and vibration control. In: Fundamentals of noise and
vibration. E & FN Spon, London, vol 12, pp 225–309
2. Jalili N (2010) Piezoelectric-based vibration control. From macro to micro/nano scale systems.
Springer, New York
3. Gupta V, Sharma M, Thakur N et al (2011) Active vibration control of a smart plate using a
piezoelectric sensor–actuator pair at elevated temperatures. Smart Mater Struct 20(10):105023
4. Kumar R, Mishra B, Jain S (2008) thermally induced vibration control of cylindrical shell using
piezoelectric sensor and actuator. Int J Adv Manuf Technol 38(5–6):551–562
1812 K. Singh et al.
Abstract A planar four-bar mechanism is the simplest and the smallest movable
closed chain linkage. It consists of four revolute pairs and is therefore often referred
to as 4R mechanism. This paper presents the kinematic and inverse dynamic force
analysis of a planar four-bar mechanism (in crank rocker configuration), assuming the
motion of one crank as input. The equations for the forces and moments of respective
links are obtained in the matrix form by combining the dynamic equilibrium equa-
tions and the kinematic relationships of each element by using Newton’s second law
of motion and Euler’s equation. The numerical computation was performed using
MATLAB. Alternatively, RecurDyn was used to perform multibody dynamic anal-
ysis, and the results of two are compared and found in good agreement. The planar
four-bar mechanism can be seen in many everyday appliances, like in car wipers,
bicycles, door frames (opening and closing), etc., and therefore, has several real-life
applications.
1 Introduction
In the spectrum of planar mechanisms, the simplest category of lower pair mecha-
nisms are four-bar linkages. The four-bar mechanism is a single degree of freedom
device and can also be used to form various other useful mechanisms through
marginal variations, such as altering the character of the pairs or the proportions
of the links, etc. [1].
A four-bar mechanism consists of four rigid members: the frame or fixed member,
to which pivoted the crank and follower, connected through the coupler. These
members are connected using four revolute pairs [2].
There are two major categories of four-bar mechanisms depending upon the
lengths of the links, namely Grashof and non-Grashof mechanisms.
A Grashof mechanism in which the drive crank is the shortest acts as a crank-
rocker mechanism and the criterion to be met by the system for this configuration
is:
where link lengths are measured between bearing centers, L max is the length of the
longest link, L min of the shortest link, L a and L b are lengths of the remaining links
[2].
2 Notations
The following set of notations was used while deriving the equations for analysis:
l1 : Length of link 1 (Frame)
l2 : Length of link 2 (Crank)
m 2 : Mass of link 2
θ2 : Angular position of link 2 w.r.t horizontal
w2 : Angular velocity of link 2
α2 : Angular acceleration of link 2
I2 : Moment of inertia of link 2
G 2 : Center of gravity of link 2
F2 : External force on link 2
M2 : Driving torque for link 2
l3 : Length of link 3 (Coupler)
m 3 : Mass of link 3
θ3 : Angular position of link 3 w.r.t horizontal
w3 : Angular velocity of link 3
α3 : Angular acceleration of link 3
I3 : Moment of inertia of link 3
G 3 : Center of gravity of link 3
F3 : External force on link 3
Multi-body Analysis for a Four-Bar Mechanism … 1815
3 Theoretical Formulation
In the analysis, the laws of vector addition were used to derive the equations for the
position, velocity, and acceleration analysis of the bodies.
From Fig. 1.
l1 = l2 + l3 − l4 (2)
where
Since l1 is a fixed vector; therefore, differentiating Eq. (2) w.r.t. time gives:
While θ4 and θ3 are calculated using Freudenstein’s equation and Eq. (2)
respectively, taking θ2 and lengths of links as inputs.
Taking the values of θ2 , w2 , and α2 of crank(driver) as inputs, the corresponding
values of w3 , w4 and α3 , α4 can be obtained by solving Eqs. (3), (4) and (5), (6)
respectively.
Multi-body Analysis for a Four-Bar Mechanism … 1817
The reaction forces on links and the driving torque are calculated by applying
Newton’s second law of motion and balancing the moments about the center of
mass of links as shown in Fig. 2.
Since link 2 rotates about ground, thus the acceleration of its COM can be written
as:
Using Eq. (7), the acceleration of COM of link 3 can be expressed as:
aG3x = −[l2 cos θ2 .(w2 )2 + l2 sin θ2 .α2 ]+[ − d3 cos θ3 . w3 )2 + d3 sin θ3 .α3
1818 N. Chaudhary and A. Gupta
aG3y = [l2 cos θ2 .α2 − l2 sin θ2 .(w2 )2 ]+[d3 sin θ3 . w3 )2 + d3 cos θ3 .α3
For link 3,
For link 4,
In the above equations, external forces on all links in the x-direction are assumed
to be zero.
Therefore, F2x = F3x = F4x = 0.
And external forces in the y-direction are due to link’s own masses; thus,
Multi-body Analysis for a Four-Bar Mechanism … 1819
1 0 1 0 0 0 0 0 0
0 1 0 1 0 0 0 0 0
2 2 - 2 2 - 2 2 2 2 0 0 0 0 1
0 0 -1 0 1 0 0 0 0
[A] 0 0 0 -1 0 1 0 0 0
0 0 - 3 3 3 3 - 3 3 3 3 0 0 0
0 0 0 0 -1 0 1 0 0
0 0 0 0 0 -1 0 1 0
0 0 0 0 - 4 4 4 4 4 4 - 4 4 0
PAx I2.α2
PBy I3.α3 - M3
M2 I4.α4 - M4
The above set of equations (in matrix form) depicts nine equations which were
solved using MATLAB to determine the values of nine unknowns in the {X} matrix
that represent the reaction forces on all four revolute joints and the driving torque to
be applied on the crank.
1820 N. Chaudhary and A. Gupta
A similar model of the four-bar mechanism was built using RecurDyn (simulation
software). The links were taken as cylindrical rods, and the values of the input
variables were taken as follows (Fig. 3):
l1 = 0.867 m
l2 = 0.120 m, m 2 = 0.991 kg, l2 = 0.0012 kg m2
l3 = 1.018 m, m 3 = 9.028 kg,l3 = 0.7797 kg m2
l4 = 0.600 m, m 4 = 5.357kg, l4 = 0.1607 kg m2
The kinematic and dynamic analysis of the proposed model was then performed
on RecurDyn for one time period (T) of the crank, keeping the angular velocity and
acceleration of the crank fixed at w2 = 2π , α2 = 0.
For w2 = 2π ,
T = 2π/w2 = 1 s
The computed values of all the variables obtained after the analysis of the mechanism
using both MATLAB and RecurDyn were plotted against each other to compare the
results (Figs. 4, 5, 6, 7, 8, 9, and 10).
Multi-body Analysis for a Four-Bar Mechanism … 1821
From the above graphs, we can infer that the values of angular velocities and
accelerations obtained by analysis through the two, MATLAB and RecurDyn, are
found to be in good agreement with each other for both the Rocker and the Coupler.
On the other hand, it can also be observed that there is a little discord in the
values of the reaction forces and the turning moment from the two sources. These
difference in the values from MATLAB and RecurDyn can be ascribed to the unlike
models in both the software, as in MATLAB, the four-bar mechanism is a 2-D model
Multi-body Analysis for a Four-Bar Mechanism … 1823
comprising of links that are defined as 2-D lines. While in RecurDyn, the simulation
is more exhaustive because of a 3-D model which comprises of links in the form of
cylinders having considerable width as well unlike straight lines.
Thus, these differences can be attributed to dissimilarity in the type of simulation
model; one being two dimensional and the other being three dimensional.
6 Conclusion
The study underlines the reactions forces acting on individual joints and the links;
thus, it can help one choose the suitable materials and the dimensions of the respective
links. Also, it draws special attention to the amount of driving torque required to
sustain the mechanism at a given velocity and acceleration of the crank, and thus, it
provides clear insights about the type and the capacity of actuator required to power
the system.
Moreover, the study shows the dependence of kinematics and dynamics of four-
bar mechanism on the type of model (2-D or 3-D); thus, it can assist a designer
while working with the four-bar mechanisms to estimate the design parameters as
per his/her design criteria.
References
Y Yadav, Saurabh, 1
Yadav, Amit, 1713
Yadav, Dharvendra P., 1713
Yadav, Harishankar Singh, 491 Z
Yadav, Jai Narayan, 479 Zodpe, D. B., 513