Article

Journal of Vibration and Control

17(6) 953–958

Modeling and response analysis ! The Author(s) 2010

Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
of dynamic systems by using DOI: 10.1177/1077546310377913

ANSYSß and MATLABß

jvc.sagepub.com

SM Khot and Nitesh P Yelve

Abstract
In the last decade, many advances and changes have been seen in engineering theory and applications. The modern trend
is to combine modeling, theoretical analysis and computer simulation. The conventional techniques of mathematical
modeling of real life complicated structural systems for vibration analysis are tedious and time consuming. The finite
element (FE) method is becoming popular because of its simplicity and feasibility of using for real life applications. In the
present study, extraction of the system model from its FE model by using commercially available softwares such as
ANSYSß and MATLABß is demonstrated. A two-degree-of-freedom spring-mass-damper system is considered here. Its
state space mathematical model in principal coordinate system is obtained in MATLABß from the modal analysis results
of its FE model done in ANSYSß. For checking the validity of model, transfer function and state space model in physical
coordinates of the same system are derived mathematically by assuming proportional damping. Frequency responses of
all the system models are compared and found to be in good agreement. To demonstrate the application of this study,
system model of a cantilever beam is also extracted in MATLABß from the result of modal analysis of its FE model in
ANSYSß and its frequency response is plotted. Response of cantilever beam to unit impulse input is also obtained in both
ANSYSß and MATLABß and compared for validating the model. This study will definitely help in simplifying the vibration
analysis of real life systems.

Keywords
ANSYSß, MATLABß, structural system, system model
Received: 30 March 2010; accepted: 7 June 2010

Here an alternative method for obtaining a mathe-

1. Introduction matical model of mechanical systems is discussed. A
The objective of vibration analysis of a mechanical generalized two degree-of-freedom (DOF) spring-
system is prediction of its behavior under different mass-damper system is considered here. The state
dynamic conditions. Since real systems are usually space mathematical model of this system in principal
quite complicated when viewed in detail, an exact anal- coordinates is obtained in MATLABß from the results
ysis of any system is often impossible. Thus, simplifying of modal analysis of its finite element (FE) model in
assumptions must be made to reduce the system to an ANSYSß. Here ANSYSß is preferred to build the FE
idealized version whose behavior approximates that of model and performing its modal analysis as it is a well
the real system. The process by which a physical system
is simplified to obtain a mathematically tractable situ-
ation is called modeling (Vu and Esfandiari, 1997). The
Department of Mechanical Engineering, Fr. C. Rodrigues Institute of
resulting simplified version of the real system is called Technology, Vashi, Navi Mumbai, Maharashtra State, India
the mathematical model, or simply the model, of the
Corresponding author:
system. Analytically deriving a mathematical model of SM Khot, Department of Mechanical Engineering, Fr. C. Rodrigues
a complicated mechanical system for vibration analysis Institute of Technology, Vashi, Navi Mumbai, Maharashtra State, India
is not a simple task. Email: [email protected]

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015

954 Journal of Vibration and Control 17(6)

proven software for this purpose. This became evident of a particular mode of vibration of the system. Setting
after referring to the literature which makes use of up the state space form of the uncoupled equations of
ANSYSß for FE modeling to achieve its desired motion in principal coordinates can be performed by
objective. inspection and solution can be obtained in MATLABß
Xu and Koko (2004) reported results on active with significant ease of programming.
vibration control of cantilever beam type of structures The construction of a mathematical model analyti-
by using ANSYSß. Karagulle et al. (2004) extended the cally for complicated real life dynamic structures is very
work of Xu and Koko (2004) and proposed the proce- difficult and time consuming. Therefore in the present
dure for simulation of active vibration control in study extraction of the mathematical model of a
ANSYSß, for cantilever beam and clamped plate type dynamic system from the results of modal analysis of
of structures. its FE model done in ANSYSß, is illustrated.
To check the validity of the model generated by
using ANSYSß and MATLABß, transfer function
2.1. Modal analysis
and state space model in physical coordinates of the
same system are obtained analytically while damping The equations of motion of a multi DOF system under
is assumed to be proportional and their frequency external forces are given by
responses are compared with that of the earlier one.
This study is further extended to demonstrate its appli- ½mx€ þ ½cx_ þ ½kx ¼ F: ð1Þ
cation for analyzing the frequency response of the can-
tilever beam by extracting its mathematical model in Since [m], [c] and [k] are non-diagonal matrices, the
MATLABß from its FE model in ANSYSß. To check above expression leads to n coupled second order dif-
the validity of the obtained model of cantilever beam its ferential equations. To uncouple these equations, first
response is obtained in both ANSYSß and MATLABß the eigenvalue problem is solved for equation (1) and
to unit impulse input and compared. the eigenvectors are obtained as x(1), x(2), . . . x(n) .
The modal matrix for the system is defined as
2. Extraction of system model ½xn  ¼ ½xð1Þ xð2Þ    xðnÞ: ð2Þ
from its FE model
In lightly-damped structures, where the damping comes For a multi DOF system with the assumption of pro-
from the losses at the joints and materials losses, modal portional damping, the solution of equation (1) can
analysis can be used to restructure the problem in terms be expressed as a linear combination of the normal
of individual modes of vibration with a particular type modes as
of damping called proportional damping (Hatch, 2001;
Meirovitch, 2001). For systems which have significant xðtÞ ¼ ½xn xp ðtÞ, ð3Þ
damping, as in systems with a specific damper element,
original coupled differential equations are required to where xp ðtÞ is known as the displacement in principal
be used for a solution. coordinates. If the normal modes are normalized with
This paper mainly deals with analyzing lightly- respect to mass, the equations of motion in the princi-
damped structures using normal modes. In ANSYSß pal coordinates become of the form
the FE model of such structures can be solved for
modal analysis to get eigenvalues (natural frequencies) x€ pi ðtÞ þ 2i !i x_ pi þ !2i xpi ðtÞ ¼ Fpi ðtÞ, ð4Þ
and eigenvectors (mode shapes). There are as many
eigenvalues and eigenvectors as DOF for a model. where i ¼ 1, 2, . . ., n and
This is the most intuitive part of the analysis and
gives one considerable insight into the dynamics of Fp ðtÞ ¼ ½xn T FðtÞ ð5Þ
the structure by understanding the mode shapes and
natural frequencies. To solve for frequency and time is the vector of forces in principal coordinates. Thus, a
domain responses, it is necessary to transform the set of n uncoupled differential equations of the second
model from the original physical coordinate system to order is obtained from the set of n coupled differential
a new coordinate system, the modal or principal coor- equations of the second order. These n uncoupled dif-
dinate system, by operating on original equations with ferential equations of the second order are converted
the eigenvector matrix. In the principal coordinate into the state space form as 2n differential equations of
system the original coupled equations of motion are the first order. A general algorithm for analyzing any
transformed to the same number of uncoupled equa- vibrating structure using ANSYSß and MATLABß is
tions. Each uncoupled equation represents the motion summarized in following section.

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015

Khot and Yelve 955

is performed. The values assumed for masses, stiffness

2.2. General theory of model extraction of springs and damping constants are m1 ¼ 1.2 kg,
The state space model of the system may be constructed m2 ¼ 1.0 kg, k1 ¼ 350 N/m, k2 ¼ 300 N/m, c1 ¼ 4.10 N.s/m,
by using eigenvalues and eigenvectors normalized with c2 ¼ 2.86 N.s/m. First, the modal analysis of the system
respect to mass. The matrices A, B, C and D of the state is done in ANSYSß.
space equation for a dynamic system with n modes
can be written as follows (Hatch, 2001). 2.3.1. Modal analysis in ANSYSß. MASS21 and
2 3 COMBIN14 elements are used for the analysis. The
0 1 ... ... ... system shown in Figure 1 is modeled in ANSYSß
6 !2 21 !1 ... ... ... 7 which is shown in Figure 2. Modal analysis of the
6 1 7
6 7 system is then performed. The Block Lanczos method
A¼6
6 . . . ... ... ... ... 7,
7 ð6Þ
6 7 is used to find the eigenvalues and eigenvectors normal-
4 ... ... ... 0 1 5 ized with respect to mass. Eigenvectors corresponding
... ... . . . !2n 2n !n to the UY DOF are extracted.
The damping coefficients of the mass and stiffness
2 3
0 matrices are assumed to be  ¼ 0.1364 and  ¼ 0.01
6 Fp1 7
6 7
6 7
6 0 7
B¼6
6 F 7,
7 ð7Þ
6 p2 7
6 7
4 ... 5
Fpn

where Fp ¼ [Fp1 Fp2 . . . Fpn]T is the force vector in prin-

cipal coordinates. Value of C depends upon the output
of interest. Since the values of displacement of the
nodes are desired, C is given by
2 3
xn11 0 xn12 0 ...
6 0 0 0 0 ...7
6 7
C¼6 7, ð8Þ
4 xn21 0 xn22 0 ...5
... ... ... ... ...

where Xn11, Xn12, . . . are the elements of Xn, the modal

matrix normalized with respect to mass. D is the direct Figure 1. Two DOF spring-mass-damper system.
transmission matrix. Here,

D ¼ ½0: ð9Þ

In this way the system matrices A, B, C and D may

be written by inspection once the eigenvalues and eigen-
vectors are known. The eigenvalues and eigenvectors
normalized with respect to mass are obtained by per-
forming modal analysis in ANSYSß. Then by import-
ing them in MATLABß matrices A, B, C and D can be
constructed. This method of model extraction is illus-
trated with the help of general two DOF system in the
following section.

2.3. Modeling of two DOF System

Now according to the general algorithm formulated
for vibration analysis, the complete analysis of a two
DOF spring-mass-damper system shown in Figure 1 Figure 2. Model of two DOF system in ANSYSß.

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015

956 Journal of Vibration and Control 17(6)

(Karagulle et al., 2004; Xu and Koko, 2004). The

results of the analysis are written to a file with the
extension .eig. The eigenvalues and eigenvectors are
then extracted as discussed in the following section.
Subsequent calculations are performed in MATLABß.

2.3.2. State space modeling in MATLABß. The

eigenvalues and eigenvectors of the two DOF system
obtained from above are extracted in MATLABß using
the code ‘‘ext56uy.m’’ (Hatch, 2001). The eigenvalues are
found to be 1.7543 and 4.2711 and the eigenvectors as
0 0
0:4985 0:7648
0:8378 0:5460:
Mass m2 is defined as both the point of application
of the input force as well as the point where the output
(displacement) is measured. The state space parameters, Figure 3. Frequency response of two DOF system.
i.e., the A, B, C and D matrices of the system are for-
mulated in principal coordinates for the full model as
follows, Taking Laplace transform of equations (10) and
(11), substituting the values and assuming initial condi-
2 3 tions to be zero which gives transfer function model as,
0 1 0 0
6 121:5 1:35 0 0 7  
6 7 X2 ðSÞ 1:2S2 þ 6:96S þ 650
A¼6 7, ¼ :
4 0 0 0 1 5 F2 ðSÞ ð1:2S4 þ 10:39S3 þ 1021:73S2 þ 2231S þ 105000Þ
0 0 720:18 7:3 ð12Þ
2 3
0
6 From equations (10) and (11), the state variables
6 0:84 7
7
B¼6 7 C ¼ ½ 0:838 0 0:55 0 for the two DOF system shown in Figure 1 can be
4 0 5 defined as,
0:55
and D ¼ ½0: X1 ¼ x1 , X2 ¼ x_ 1 , X3 ¼ x2 , X4 ¼ x€ 2 : ð13Þ

Then a state space system is defined using the ‘‘ss’’ After making appropriate substitutions of state var-
function in MATLABß with A, B, C and D as param- iables in equations (10) and (11) and arranging the
eters. By using the ‘‘bode’’ function in MATLABß, the terms in state space form (Jairath, 2008), matrices A,
frequency response for the system is plotted as shown in B, C, and D are obtained as,
Figure 3 in the following section. To prove the validity 2 3
0 1 0 0
of the model obtained from the FE model into 6 541:67 5:8 250 2:3833 7
MATLABß, its frequency response is compared with 6 7
A¼6 7,
those obtained by using an analytical transfer function 4 0 0 0 1 5
and state space approach. 300 2:86 300 2:86
2 3
0
3. Analytical transfer function 607  
6 7
B ¼ 6 7, C ¼ 0 0 1 0 and D¼ ½0:
and state space analysis 405
The equations of motion for the two DOF system 1
shown in Figure 1 can be written as State space system is defined using the ‘‘ss’’ function
in MATLABß with A, B, C and D as parameters.
m1 x€ 1 þ ðc1 þ c2 Þx_ 1  c2 x_ 2 þ ðk1 þ k2 Þx1  k2 x2 ¼ 0, Frequency responses of the two DOF system are then
ð10Þ obtained in MATLABß by using the ‘‘bode’’ function
for both transfer function and state space models as
m2 x€ 2  c2 x_ 1 þ c2 x_ 2  k2 x1 þ k2 x2 ¼ f2 : ð11Þ shown in Figure 3.

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015

Khot and Yelve 957

It can be seen that frequency responses obtained for SOLID45 element is used for the analysis. Modal
the two DOF system shown in Figure 1 by the three analysis of the beam is performed using the Block
methods, namely by given algorithm, transfer function Lanczos method. The range of frequency is defined as
and state space methods are in good agreement. Hence 0 to 50000 Hz. The damping coefficients of the mass
the method of extracting a mathematical model of a and stiffness matrices are assumed to be  ¼ 0.687 and
dynamic system from the results of the modal analysis  ¼ 2.89e-4 (Karagulle et al., 2004; Xu and Koko,
of its FE model done in ANSYSß, is found to be valid 2004). Ten modes are extracted. From  &  the
and can be implemented for getting mathematical modal damping ratios i corresponding to ith mode
models of complicated real life dynamic systems. can be obtained by using the relation (Xu and Koko,
After observing the success of the method of extract- 2004)
ing a system model from the FE model for the two
DOF system, the same method is implemented for  þ !2i
extracting a system model of cantilever beam, which i ¼ : ð14Þ
2!i
is discussed in the following section.
The eigenvalues obtained for beam are as follows:
3.196, 18.830, 49.662, 92.774, 150.95, 227.070,
4. Cantilever beam 321.110, 431.090, 554.370, 690.650
An aluminium beam of dimensions (508  25.4  0.8) The results of the analysis are written to a file with
mm3 is used for the analysis. The first step in obtaining the extension ‘‘.eig’’. This file is imported in
the mathematical model of the cantilever beam is to MATLABß. The state space model may now be
find its eigenvalues and eigenvectors normalized with formed as described in the following section.
respect to mass. This can be done by performing
modal analysis in ANSYSß.
4.2. State space modeling in MATLABß
For the cantilever beam, the input is an impulse force
4.1. Modal analysis in ANSYSß applied at the tip and the output is the tip displacement.
The FE model of cantilever beam in ANSYSß is shown The tip is chosen because here, there is no node for any
in Figure 4. mode (where node is a point of zero displacement).

Figure 4. Modeling of cantilever beam in ANSYSß.

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015

958 Journal of Vibration and Control 17(6)

0 x 10–3
Full model 5
–20 Matlab model
4 Ansys model
–40
–60 3

Displacement (m)
–80 2
20logM

–100 1
–120 0
–140
–1
–160
–2
–180
–3
–200
101 102 103 104 10
–4
log w 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)

Figure 5. Frequency response of full model.

Figure 6. Transient responses in MATLABß and ANSYSß.

Hence all the modes can be excited or observed at algorithm, transfer function and state space methods,
the tip. validates the approach followed in this paper i.e. form-
Both input and output are along the Z axis. Hence ing the state space model from the eigenvalues and nor-
the eigenvectors pertaining to the UZ displacement malized eigenvectors obtained from ANSYSß. The
are required. These eigenvalues are extracted from given algorithm is also implemented for obtaining a
the ‘‘.eig’’ file exported from ANSYSß using the file system model of cantilever beam type structure and
‘‘ext56uz.m’’ (Hatch, 2001). This code forms the found successful as the transient responses of the can-
vector of eigenvalues and the modal matrix of the sys- tilever beam to the unit impulse input obtained in
tem consisting of the eigenvectors pertaining to the UZ ANSYSß and in MATLABß from the state space
displacement of all the nodes of the FE model. model formulated found to be in good agreement.
The state space parameters, i.e., the A, B, C and D This approach may be generalized and applied to
matrices of the system are formulated for the full model more complex structures.
of cantilever beam. These system matrices are then used
to form a state space system in MATLABß using the
‘‘ss’’ function. The frequency response of the system is References
plotted using the ‘‘bode’’ function in MATLABß as Hatch MR (2001) Vibration Simulation Using MATLAB
shown in Figure 5. The contribution of each individual and ANSYS, 1st ed. Boca Raton, Florida: Chapman &
mode is also shown. Hall/CRC.
The transient responses of the cantilever beam Jairath AK (2008) Control Systems: The State Variable
obtained in ANSYSß and MATLABß to unit impulse Approach. New Delhi, India: Ane Books India.
input are also plotted and compared for validating the Karagulle H, Malgaca L and Oktem HF (2004) Analysis of
accuracy of the model as shown in Figure 6. active vibration control in smart structures by ANSYS.
Smart Materials and Structures 13: 661–667.
The transient responses obtained to unit impulse
Meirovitch L (2001) Fundamentals of Vibrations. New Delhi,
input in both ANSYSß and MATLABß are found to
India: McGraw Hill International Edition, Mechanical
be in close agreement. Engineering Series.
Vu HV and Esfandiari RS (1997) Dynamic Systems: Modeling
5. Conclusions and Analysis. New York: McGraw-Hill.
Xu SX and Koko TS (2004) Finite element analysis and
The close agreement between the frequency responses design of actively controlled piezoelectric smart structures.
of the two DOF system obtained by the given Finite Element Analysis Designs 40: 241–262.

Downloaded from jvc.sagepub.com at UNIV PRINCE EDWARD ISLAND on March 2, 2015