See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/251989718

Gear pair design optimization by Genetic Algorithm and FEA

Article · November 2010

DOI: 10.1109/FAME.2010.5714820

CITATIONS READS

34 1,617

4 authors, including:

Padmanabhan Sambandam Ganesan Subbiah

Vel Tech Rangarajan Dr.Sagunthala R&D Institute of Science and Technology Sathyabama Institute of Science and Technology
103 PUBLICATIONS 988 CITATIONS 153 PUBLICATIONS 2,710 CITATIONS

SEE PROFILE SEE PROFILE

Dr.M. Chandrasekaran
Vels institute of Science Technology and Advanced Studies
207 PUBLICATIONS 2,050 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Investigation of Microstructural, Mechanical and Tribological Properties of Al8011-TiC Metal Matrix Nano-Composites View project

Strategic management of new product development View project

All content following this page was uploaded by Dr.M. Chandrasekaran on 22 August 2014.

The user has requested enhancement of the downloaded file.

Gear Pair Design Optimization by Genetic Algorithm
1
and FEA
S.Padmanabhan, 2S.Ganesan 3
M.Chandrasekaran
Department of Mechanical and Production Engineering, Department of Mechanical Engineering,
Sathyabama University, Rajalakshmi Engineering College,
Chennai, India Chennai, India
[email protected]
4
V.Srinivasa Raman
Department of Mechanical Engineering,
N.P.R College of Engineering and Technology,
Madurai, India

Abstract- Multiple, often conflicting objectives arise naturally in requirements, transmitted power, weight, tooth profile and
most real-world optimization. Gear is a mechanical device that material capabilities.
transfers the rotating motion and power from one part of a
machine to another. Searching for best gear is a very hard H Zarefar and S N Muthukrishnan [1] describes about
problem. Gear optimization can be divided into two categories, modified adaptive random-search algorithm for the design of
namely, single gear pair or Gear train optimization. The problem helical gears has been developed. The proposed methodology
of gear pairs design optimization is difficult to solve because it allows for the implementation of nonlinear design functions
involves multiple objectives and large number of variables. and constraints without the need for linearization. Bernd-
Therefore a reliable and robust optimization technique will be Robert [2] explains emphasizes on the two aspects sound
helpful in obtaining optimal solution for the problems. In this excitation and efficiency. The main reasons of sound excitation
paper an attempt has been made to optimize spur gear pair design in a gear stage are studied. Recommendations to minimise
using Genetic Algorithm (GA) and analytical tool MITCalc. A excitations and reduce the losses in the gear mesh are given.
combined objective function which maximizes the Power, Kalyanmoy Deb and Sachin Jain [3] describes about the use of
Efficiency and minimizes the overall Weight, Centre distance has a multi-objective evolutionary algorithm, which is capable of
been considered in this model. Finite Element Analysis (FEA) was solving the original problem involving mixed discrete and real-
carried out and results were compared with the allowable limit. valued parameters and more than one objectives, and is capable
of finding multiple no dominated solutions in a single
simulation run. On a number of instantiations of the gearbox
Keywords – FEA; Gear pair design; MITCalc; Spur Gear;Genetic
Algorithm. design problem having different complexities, the efficacy of
Non-Sorted Genetic Algorithm II for optimizing multi speed
gear box which consider multi objectives such as maximizing
I. INTRODUCTION the power and minimizing the total volume of the gear.
Mechanical design can be defined as the selection of Yong sheng Lian et al [4] this paper reviews about progress
materials and geometry, which satisfies, specified and implied in design optimization using evolutionary algorithms to solve
functional requirements. The method of design has to minimize real-world aero dynamic problems. Examples are given in the
the most significant undesirable effect and to maximize the design of turbo pump, compressor, and micro-air vehicles.
most significant desirable effect. The method of optimum Ramamurthi et al [5] have developed a methodology for fixing
design is arrived by overall considerations of many specific the leading dimensions of a two stage spur gear box
optimum design studies. transmitting a specific horse power for a given input speed and
reduction. Tsay and Tseng [6] have applied a multiple
Gear is a mechanical device that transfers the rotating optimization method to reduce the level of kinematical errors
motion and power from one part of a machine to another. The of helical gear train and investigated an optimal gear tooth
need for designing multistage gear drives has been increasing modifications. Caroll and Johnson [7] have presented an
with the increasing application of gear drives in high speed and optimal design technique to obtain compact and standard spur
small space. Gears are produced in a wide range of sizes, and gear meshes with an objective of minimizing centre distance.
they vary greatly in use. They range from the tiny gears that Savage et al [8] have described the optimal design of an
drive the hands of a watch to the huge gears that turn the enclosed parallel shaft spur gear reduction. The object of the
propeller of a military tanker. design was to determine a small, light weight transmission with
Single gear pair optimization have concentrated on a long service life.
investigating the optimal bending and crushing stresses,
displacement acting on the gear tooth with regard to space

978-1-4244-9083-7/10/$26.00 ©2010 IEEE

396
 Alexander L. Kapelevich [9] The Direct Gear Design Maximize, f1 = P (1)
method defines parameters of the gear mesh to provide
π π
complete geometry of the involute profile of the teeth, Minimize, f2 = [[ × d 1 2 ×b×ρ] + [ × d 2 2×b×ρ]] (2)
including the base diameter, form diameter, outside diameter, 4 4
tooth thickness, tip radii, etc. The fillet profile initially is Maximize, f3 = 100 – P L = η
defined as a trace of the tip of the mating gear tooth. This kind (3)
of fillet profile is used for plastic molded gears. Prayoonrat et 50f (H s2 + H 2t )
al [10] have presented an algorithm to design and optimize PL = ×
multi spindle gear trains for the non speed-change type in CosΦ (H s + H t )
which the designer may choose minimum overall centre (4)
distance, minimum overall size, minimum gear volume, or Where,
other desirable criteria, such as maximum contact or overlap f = 0.08, Φ = 20o
ratios to optimize gear trains. Yoon and Rao [11] have
(i + 1)  r 2 
×   0  − cos 2 Φ  – sin Φ
presented a novel method to minimize the static transmission
Ht =
error using cubic splines for gear tooth profiles. Innocenti [12] i  r  
has proposed a new approach to the efficiency evaluation of  
single or multidegree freedom of gear trains. Rosic [13] has (5)
proposed an analytical and computer aided procedure for the
  R 2 
multi criteria design of gear train transmission system. Monte H s = (i + 1) ×  0 − cos 2 Φ  – sin Φ
Carlo method was implemented in the program. Abersek et al   R  
 
[14] have developed the expert system for design and
manufacture of a gear box. Marcelin [15, 16] has applied GA (6)
for the design optimization of the gears and also developed a
meta-model using neural network and GA for an integrated R 0 = R + one addendum
optimal design of mechanism. Addendum for 20o full depth involutes system = one module
ro = r + m
Based on the above observation, this paper has made an Ro = R +m
attempt to use the potential GA and MITCalc to solve the gear
pair design problem. m
Minimize, f 4 = (Z 1 +Z 2 )
II. DESIGN OF SPUR GEAR 2
(7)
This section describes about the design objectives,
constraints, considered in this work. This work uses a C. Design Constraints for Spur Gear Pair
combined objective function, which minimizes the volume and The tooth breakage is caused by fatigue due to repeated
centre distance and maximizes power and efficiency. bending stresses. To safeguard the tooth against the breakage,
A. Design of Spur Gear Pair the gear should have adequate bending strength. i.e., the
induced bending stress when transmitting a torque should be
Design a spur gear drive to transmit 30kw at 1440rpm with lesser than the allowable bending stress. Eqn. (8) represents
a speed reduction of 2.5. this constraint. Eqns. (9), (11), (14) and (17) have been adopted
from [18]:
The objective functions considered in this work are given
below: σ b ≤ [σ b ]
• Maximization of power delivered by the gear pair (f 1 ) (8)

• Minimization of the over all weight – which is The induced bending stress is calculated by eqn. (9):
indirectly related to the volume of the gears (f 2 )
• Maximization of the efficiency of the gear pair (f 3 ) σb =
(i + 1) ×[M t ]
(a m b y )
• Minimization of the center distance between the input (9)
and output shafts (f 4 ).
B. Objective functions for Spur Gear Pair To avoid the surface failures of the tooth profile like
Pitting, Surface Abrasion, Seizure etc., and to have the
The following Objective functions have been considered in satisfactory life, gear should have the wear resistance. i.e., the
this model. The equation 1, 2, 3 and 7 represent the induced crushing stress should be lesser than the allowable
maximization of Power, minimization of Weight, maximization crushing stress. This constraint is represented in eqn. (10):
of Efficiency and minimization of Centre distance. The
efficiency equation (3) has been adopted from Dudley [17].

397
σ c ≤ [σ c ]
m = 1.26 ×
[M t ]
(10)
(y σ b Ψm Z1 )
min 3

The induced crushing stress is calculated by eqn. (11) (17)

Ψ m = b/m (taken from recommended values as 10)
 i +1  i + 1  
σ c = 0.74   ×  ib  × E × [ M t ] The module ‘m’ obtained through the optimization process
 a     should be greater than the minimum module to assure the
(11) proper transmission of rotational motion. This constraint is
represented by eqn. (18):
Where, [Mt] = M t × k × k d
k - Concentration factor m≥m min
k d - Dynamic load factor (18)

The gear ratio should be a constant and it should be equal to Variable Bounds
the ratio between the number of teeth in gear and the number of  Module – varied from 4 to 5mm in the range of 0.001.
teeth in pinion. The eqn. (12) represents this constraint:  Thickness of gear pair – varied from 0 to 50mm in the
Z2 d range of 0.001.
i = 2.5 = (or) 2  Number of teeth in pinion - 14, 16,18,20,22,24,26,28.
Z1 d1
(12)  Power – varied from 30 to 32 kW in the range of 0.001.
The centre distance between the pinion and gear is calculated
D. Proposed Design Objective function
by eqn. (13):

a=
(d1 + d 2 ) = m [Z 1 + Z 2 ]
The gear pair design problem has four different parameters
in the objectives considered in this work. i.e., power, weight of
2 2 material, efficiency and center distance. Since all these
(13) parameters are on different scales, these factors are to be
normalized to the same scale [19]. For maximizing criterion
The minimum centre distance is calculated by eqn. (14): value, the values are normalized by dividing its value with the
normalizing factor, max i, which is the maximum value of this
 0.74  2  E [M ]  criterion obtained from the solutions that have been explored
a = (i + 1)   × t 

 [σ c ]   iΨ 
min 3 so far and for a minimizing criterion value, it is normalized by
  dividing the normalizing factor, min i , with its value. The
(14) maximum and minimum value of the criterion will be updated
Centre distance between the shafts ‘a’ should be greater whenever the proposed algorithm finds another feasible
than the minimum centre distance ‘a min ‘ to assure the solution. In addition, to ensure the overall objective value to
required clearance between the tip of the pinion tooth and the fall between 0 and 1, the weight of each criterion is also
root of the gear tooth and vice versa. normalized. The normalized objective function is obtained as
follows:
The eqn. (15) represents this constraint: n
a≥a min (15) COF = Σ NW i * N (X i )
i =1
b Where,
Ψ= , (taken from recommended values, as 0.3)
a COF = Combined objective function
W i = pre normalized weight of criterion i.
The number of teeth must be integer and within the lower and NW i = normalized weight of criterion i.
higher limits. This constraint is represented by eqn. (16): Wi
Where NWi =
Z i ε I, for i = 14, 16,18,20,22,24,26,28  n

(16)
 iΣ=1Wi 
 
The minimum module ‘m min ’ is calculated by eqn. (17):
(19)
N (X i ) = normalized value of criterion i of solution X.
Where,

398
N (X i ) =
Xi
for maximizing criterion.
(1) Choose a random number ui , ∈ [0,1]
max i
(2) Calculate β qi as given in the following equation.
(20)
min i
N (X i ) = for minimizing criterion. 
1
ηc +1
Xi
(2ui ) , ui ≤ 0.5
(21) 
β qi = (23)
1
ηc +1
Xi = pre normalized value of criterion X.
 1 
 2(1 − u ) 
max i = pre normalized maximum value of criterion i among , otherwise
all solutions explored so far.  i 
min i = pre normalized minimum value of criterion i among
all solutions explored so far .
N = number of criteria.
Where β qi is the spread factor and is defined as the ratio of the
Hence the COF for this problem is, absolute difference in offspring values to that of the parents. η c
is the crossover index.
 power

  min .weight
xNW1  + 

xNW2  + 

(3) Then compute the offspring xi(1,t +1) & xi(2,t +1) as,
COF =    weight 
max . power 
  
+1)
xi(1,t= 0.5 (1 + β qi ) xi(1,t ) + (1 − β qi ) xi(2,t ) 
 efficiency xNW3  +  min .cent.dist xNW4 
 max .efficiency 
  cent.dist 
(22)
+1)
xi(2,t= 0.5 (1 − β qi ) xi(1,t ) + (1 + β qi ) xi(2,t ) 
Where NW 1, NW 2, NW 3 and NW 4 = 0.25. (24)
III.METHODOLOGY D Polynomial Mutation
Newly generated offspring undergo polynomial mutation. Like
Here non-traditional optimization technique Genetic Algorithm in the SBX operator, the probability distribution can also be a
result was compared with MITCalc analytical tool. The better polynomial function, instead of a normal distribution. The new
result was validated by ANSYS tool.
offspring yi(1,t +1) is determined as follows,
A. Genetic Algorithm
Representation
Initially the specified population is generated randomly.
yi(1,t +1) = xi(1,t +1) + ( xiU − xiL )δ i (25)
The control string is represented as follows. x iU and x iL are the upper and lower limit values.Where, the

δi
X = [m, b, Z 1 , P]
parameter is calculated from the polynomial probability
The design variables within limits have been represented as
discrete variables. The objective functions and the COF have distribution.
been computed for all the strings. ) 0.5(ηm + 1) (1 − δ )ηm
P (δ =
B. Tournament Selection 1/(ηm +1)
The tournament selection has been carried out by creating a 
(2ri ) − 1, if ri < 0.5
δi =  (26)
1 − [ 2(1 − ri ) ]
tournament competition among the individual strings. The best 1/(ηm +1)
individual (the winner) from this group is selected as parent.  , if ri ≥ 0.5
This process has been repeated until the mating pool for new
off spring is filled. Tournament selection is used as selection
mechanism in order to avoid premature convergence. To form Where, η m is the mutation index. In this operator the shape
new parents for the next generation the following cross over of the probability distribution is directly controlled by the
and mutation operation have been carried out. external parameter η m and distribution is not dynamically
changed with generations.
C. Simulated Binary Crossover
The Simulated Binary Crossover (SBX) creates children E .MITCalc
solutions in proportion to the difference in parent solutions. It is an open system designed in Microsoft Excel which
allows not only easy user-defined modifications and user
The following steps have been followed to create two children
extensions without any programming skills, but also mutual
solutions from two parents:
interconnection of the calculations, which is unique in the
development of tailor-made complex calculations.

399
 IV. RESULTS AND DISCUSSIONS
A. FEA Analysis
The Mathematical models for the test problem have been
formulated in terms of design variable m, b, z and P. Initially, After optimizing with GA and MITCalc, the Table II shows
The input values are generated randomly with their variable the optimum results of the Spur Gear for the module 5mm.
bounds. If the generated values satisfy the design constraints, From the Table II, MITCalc tool shows maximum efficiency
then the values of objective functions f 1 , f 2 , f 3 and f 4 are than GA. Hence FEA analysis was conducted for the Spur gear
computed along with COF. The optimum values of objective (Pinion) of thickness of 30mm.
function and design variables corresponding to the minimum
COF value obtained by the proposed algorithms for the test
problem are shown in Table II. The input values of the test
problem are shown in Table I.
TABLE I
THE INPUT VALUES WITH THEIR BOUNDS
Values for
Parameter/Constraint
spur gear pair
Gear material 34CrNiMo6
Lower Limit 5 mm
Gear Thickness
Upper Limit 50 mm
Minimum 30 kW
Power delivered
Maximum 32 kW
Gear-ratio 2.5
Lower Limit 14
Number of Teeth
Upper Limit 28 Fig.1. Geometry model of Pinion and Gear using Pro/E
2
Allowable crushing stress 1100N/mm PCD for Pinion : 140 mm
Allowable bending stress 400N/mm2 PCD for Gear : 350 mm
Young’s modulus 2.15x105 N/mm2
Centre distance : 245 mm

Input speed 1440 rpm

Minimum 4
Module
Maximum 5
Poisson’s ratio 3

TABLE II
OPTIMUM RESULTS FOR THE MODULE 5 MM
FOR SPUR GEAR PAIR

Tool MITCalc GA
Thickness(mm) 30.0000 28.7856
No. of Teeth in Pinion 28 28
Power (kW) 30.0000 30.1121
Weight (kg) 29.4739 28.3961
Efficiency (%) 99.1254 98.9136 Figure.2. Meshing of Pinion and Gear using ANSYS
Centre Distance(mm) 245 245

400
 Fig.5. Hoop Displacement Plot (Uy) of Pinion

Figure 3. Enlargef View - Meshing of Pinion and Gear

PLANE42 - Plane Stress with thickness option is used for

Fig.6. Von Mises Stress Plot of Pinion
analysis.

F
igure.4. Radial Displacement Plot ( Ux) of Pinion
Fig.7. Von Mises Stress Plot of One tooth

Fig.8. Enlarge View -Von Mises Stress Plot of One tooth

401
 It is observed that from Table II, the Genetic algorithm [5] Ramamurthi.V, Gautam.P, Kothari. A, “Computer-aided design of a two-
stage gearbox”, Advances in Engineering Software, Vol. 28, 1997,
performs better than MITCalc tool in weight reduction and pp.73-82.
power transmitted. It is noted that the MITCalc has shown [6] Tsay.C-B, Tseng.C-H, “Modified Helical Gear Train”, Journal of
better efficiency than GA. Mechanical Design, Vol. 119, 1997, pp.307-314. Carrol. R.K.,
Johnson. G.E., “Optimal design of compact spur gear sets”, ASME
Further analysis has shown Maximum stress is 6.525 MPa , Journal of Mechanisms, Transmissions and automation in design,
Vol.106, 1984, pp.95-101.
which is comparatively less than theoretical value. [7] Savage. M, Lattime.S.B., Kimmel,J.A., “Optimal Design of Compact
Spur Gear Reductions”, Transactions of the ASME, 690 / Vol. 116.
V. CONCLUSION 1994, pp.690-696.
[8] Alexander L. Kapelevich et al, “Direct Gear 1Design: Bending Stress
Minimization”, Gear 1Technology , September/October 2003.
In this paper an attempt has been made to obtain an optimal
[9] 1Prayoonrat,S, Walton,D, “Practical approach to 1optimum gear train
design solution for a spur gear pair and further analysed with design”, Mechanical 1Engineering,Department, University of
FEA. Within the various design variables available for a gear 1Birmingham, 1Birmingham B15 2TT, UK, Vol. 20, 11988, pp.83-192.
pair design, power, weight, efficiency and centre distance have [10] 1Yoon,K.Y, Rao.S.S, “Dynamic Load Analysis of 1Spur Gears Using a
been considered as objective functions. Bending stress and New Tooth Profile”, Journal of 1Mechanical Design, Vol. 118, 1996,
pp.1-6.
crushing stress have been considered as vital constraints to get [11] Innocenti.C. “A Framework for Efficiency Evaluation of Multi-
an efficient, compact and high power transmitting gear pair. Degree-of-Freedom Gear 1Trains”, Transactions of the ASME, Vol.
118, 1996, 1pp. 556-560.
[12] Rosic.B, “Multicriterion optimization of multistage 1gear train
By FEA analysis the maximum stress was determined as transmission”, Mechanical Engineering, 1Vol. 8, 2002, pp. 1107-1115.
6.525 MPa , which is comparatively less than the theoretical [13] 1Abersek.B, Flasker.J, Balic.J. “Expert system for 1designing and
value.As a future work, minimization of vibration and manufacturing of a Gearbox”, Expert 1systems with applications, Vol.
minimization of noise can also be included in the objective 11(3), 1996, 1pp.397-405.
function to obtain a more reliable gear pair design. [14] 1Marcelin.J.L, “Genetic optimization of gears”, 1International Journal of
Advanced Manufacturing 1Technology, Vol. 17(12), 2001, pp. 910-915.
[15] 1Marcelin.J.L, “A meta model using neural networks 1and Genetic
REFERENCES Algorithms for an Integrated Optimal 1design of mechanism”,
International Journal of 1Advanced Manufacturing Technology, Vol.
24, 12004, pp. 708-714.
[1] H Zarefar and S N Muthukrishnan, “Computer-aided optimal design via [16] Dudley.D.W, “Handbook of Practical Gear 1Design”, McGraw –Hill
modified adaptive random-search algorithm” , Journal of Computer- Book Co: New York, 11994.
aided design, Volume 25 no 4 april 1993. [17] 1“Design Data Book”, Faculty of Mechanical 1Engineering, PSG College
[2] Bernd-Robert Höhn , “Improvements on noise reduction and efficiency of of Technology, 1Coimbatore-641004, India.
gears”, Meccanica (2010) 45: pp. 425–437. [18] 1Ho.Y.C, Moodie.C.L, “Machine Layout with a 1linear Single-Row
[3] Deb. K, Jain, S, “Multi-Speed Gearbox Design Using Multi-Objective Flow Path in an Automated 1Manufacturing System”, Journal of
Evolutionary Algorithms”, Journal of Mechanical design, Vol.125, 2003, Manufacturing 1Systems, Vol 17, 1998, pp. 1 – 22.
pp.609-619. [19] 1Saravanan.R, Sachithanandam.M, “Genetic 1Algorithm for multi
[4] YongshengLian et al , “Progress in design optimization using evolutionary variable surface grinding 1process optimization using a multi
algorithms for aerodynamic problems”, Progress in Aerospace Sciences”, objective 1function model”, International Journal of 1Advanced
pp. 199–223, 2010. Manufacturing Technology, Vol. 17, 12001, pp. 330 - 338.

402

View publication stats