This article was downloaded by: [University of Oklahoma Libraries]

On: 30 April 2014, At: 10:58

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Engineering Optimization
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/geno20

Exploring the geometry and material

space in gear design
a

Nagesh Kulkarni , B.P. Gautham , Pramod Zagade , Jitesh

b

Panchal , Janet K. Allen & Farrokh Mistree

Tata Research Development and Design Center, Tata Consultancy

Services, Pune, India
b

Department of Mechanical Engineering, Purdue University, West

Lafayette, Indiana, USA
c

The Systems Realization Laboratory @ OU, The University of

Oklahoma, Norman, Oklahoma, USA
Published online: 28 Apr 2014.

To cite this article: Nagesh Kulkarni, B.P. Gautham, Pramod Zagade, Jitesh Panchal, Janet K. Allen
& Farrokh Mistree (2014): Exploring the geometry and material space in gear design, Engineering
Optimization, DOI: 10.1080/0305215X.2014.908868
To link to this article: http://dx.doi.org/10.1080/0305215X.2014.908868

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
Content) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions

Engineering Optimization, 2014

http://dx.doi.org/10.1080/0305215X.2014.908868

Exploring the geometry and material space in gear design

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Nagesh Kulkarnia , B.P. Gauthama , Pramod Zagadea , Jitesh Panchalb , Janet K. Allenc and
Farrokh Mistreec
Research Development and Design Center, Tata Consultancy Services, Pune, India; b Department of
Mechanical Engineering, Purdue University, West Lafayette, Indiana, USA; c The Systems Realization
Laboratory @ OU, The University of Oklahoma, Norman, Oklahoma, USA

a Tata

(Received 24 May 2013; accepted 17 March 2014)

Gear design is a complex process. When new materials and manufacturing processes are introduced,
traditional empirical knowledge is unavailable and considerable effort is required to nd starting design
concepts. This forces gear designers to go beyond the traditional standards-based design methods. The
method presented here, the concept exploration method, is based on the compromise decision support
problem and is demonstrated for gear design which requires the simultaneous exploration of geometry,
material and manufacturing spaces to exploit synergies. The results obtained are in agreement with existing
knowledge. To further develop design guidelines, ternary contour plots of goal achievement are created
to show the interdependence among various design goals. These ternary charts help gear designers to
visualize and understand the complexity and compromises involved and help them to make trade-offs
among individual goals.
Keywords: gear design; concept exploration method; design and material space exploration

Nomenclature
b
C
d
d1+ , d2+ , d3+
d1 , d2 , d3
G
m
N
R
Rc

SU
SUMIN
SUMAX
bending
allowable
Corresponding

face width (mm)

cost (Indian rupees, INR)
centre distance (mm)
deviation variables to account for overachievement
deviation variables to account for underachievement
gear ratio
module of gear (mm)
number of teeth
reliability (%)
contact ratio
density (kg/m3 )
tensile strength
minimum tensile strength
maximum tensile strength
allowable bending stresses
author. Email: janet.allen@ou.edu

2014 Taylor & Francis

N. Kulkarni et al.
bending

induced
contact
allowable
contact
induced
T
W1 , W2 , W3
YZ

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

1.

induced bending stresses

allowable contact stresses
induced contact stresses
torque (Nm)
weights associated with the design goals
correction factor for reliability

Frame of reference

Traditionally, gears have been designed using design codes developed by professional bodies
such as the American Gear Manufacturers Association (AGMA) (ANSI/AGMA 1993, 2010) and
International Organization for Standardization (ISO) (ISO 2006). These design procedures are
similar and contain empirical design factors obtained from experiments and experience. These
methods are based on simple assumptions such as those on the geometry and they also offer
correction factors for requirements such as precision and quality. Some of these factors account
for uncertainties in operating conditions, material and manufacturing processes. Gears designed
using these design codes are usually safe and reliable but suboptimal as the conservative nature of
design codes results in overdesign and large margins of safety. For example, for a multispeed gear
box, the standards-based design priorities are usually given to strength, followed by compactness,
efciency and wear (MacAldener 2001). Apart from these demands, new demands such as higher
quality, low noise and high safety have emerged owing to increased competition, stringent legal
requirements, etc. Further, the manufacturers face additional demands for designing gears rapidly
and inexpensively. Currently, this is a US $45 billion a year industry. The gear industry vision
report (Gear Industry Vision 2004) highlights the vision, challenges, and strategic goals for the
gear industry in 2025. In these documents various goals are listed, as well as key technological
challenges and innovations required to achieve them, and it is clear that newer paradigms must
be explored for gear design.
With the many possible requirements, a designer should have the capability to explore the design
space and understand the appropriate compromises. The concept exploration method (CEM) is
proposed, which has the following characteristics:
the capability to concurrently explore coupled geometry and material spaces
the capability to explore a complex multidimensional space in presence of constraints and
conicting objectives
the capability to provide insights into the nal solution as well as the progression towards
solution with and understanding of the behaviour of design variables, constraints and design
objectives.
A vast amount of literature is available on gear design. The key articles are reviewed here and
the major ndings are summarized. Buckingham (1949) studies the kinematics of gear mesh and,
based on this, offers a design procedure for gears. Dudley (1962) and Shigley (1977) provide a
method of designing gears based on achieving desired factors of safety. Tucker (1980) and Estrin
(1980) study the effects of gear mesh parameters such as addendum ratio and pressure angle, and
present a procedure for modifying standard gear geometry to obtain more suitable gears. Wang,
Luo, and Wu (2010) study the effect of the sliding coefcient on gear design and outline a procedure
to design a gear tooth based on the desired sliding coefcients. Yeh, Yang, and Tong (2001) and
Tsai and Tsai (1998) develop design procedures for high load carrying capacity gears. Handschuh
and Zakrajsek (2011) study the effect of high values of pressure angle on the performance of
gears to determine the feasibility of designing gears for aeronautical and space applications. Lin

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

et al. (1998) deal with the effects of dynamic analysis in compact spur gear design; they compare
their in-house method of calculation of dynamic factor to that of the AGMA (2010) recommended
method. They observe that the size of an optimal gear set is signicantly inuenced by dynamic
factors and that the peak dynamic factor estimated at system natural frequencies dominates the
design of gear sets operating over a wide range of speeds. They create design charts that can be
used either to design a gear for single speed or over a range of speeds. Hewitt (1992) discusses the
selection of gear materials and the estimation of reliability. He considers a statistical distribution
for material strength obtained under different processing as well as precision conditions.
Kapelevich (2007) and Goldfarb, Kapelevich, and Tkachev (2008) propose a method where the
tooth prole of gear teeth is optimized for desired performance. In this, unlike the traditional methods, the constraints on the geometry are not severely constrained by the manufacturing processes
and the manufacturing has only secondary effects. The method results in non-standard geometry,
which requires non-standard tooling and affects gear interchangeability. With the advances in
computational techniques, nite element analysis (FEA) has also been widely used in gear design
(Brauer 2004; Kawalec, Wiktor, and Ceglarek 2006; Wang and Howard 2006; Kirov 2011). Brauer
(2004) provides a general method to create a nite element model of involute gears. Kawalec,
Wiktor, and Ceglarek (2006) perform comparative analysis of tooth root strength obtained using
design standards such as AGMA (2010) and ISO (2006) and verifying them with FEA. They conduct analysis for spur and helical gears, for various key geometric (gear design), manufacturing
(racks and gear tools) and performance (load location) parameters. Wang and Howard (2006)
investigate large numbers of two- and three-dimensional gear models using FEA. They study
parameters such as torsional stiffness, tooth stresses and stress intensity factors, and suggest that
caution should be exercised when two-dimensional assumptions are used for modelling gears.
Kirov (2011) studies and compares gear design using AGMA and FEA. He argues that it is difcult to directly compare AGMA and FEA for gear design, and suggests that FEA should be
preferred and used whenever AGMA does not provide a design method.
While the literature provides equations and methods for use in design of gears for given performance requirements, gear design involves satisfying multiple conicting goals such as high
reliability, compactness and low cost. High reliability demands gears of larger size and/or highstrength material, which conicts with the compactness and/or low cost requirements. Smaller
gears are not necessarily low-cost gears as they may require expensive materials and manufacturing processes (e.g. advanced heat treatment or shot peening). Dudley (1964) and Tucker (1980)
observed that the cost trade-off generally favours small gears with high hardness. Savage, Coy, and
Townsend (1982) and Carroll and Johnson (1984) provide literature surveys as well as methods
for the optimal design of spur gear sets.
Optimization methods are widely used in gear design (Estrin 1980; Lin et al. 1998; Savage, Coy,
and Townsend 1982; Carroll and Johnson 1984; Thomson, Gupta, and Shukla 2000; Wang and
Wang 1994). A detailed literature review of these methods is provided by Papalambros and Wilde
(2000), Arora (2011) and Kulkarni et al. (2012). Studies by Estrin (1980), Lin et al. (1998), Savage,
Coy, and Townsend (1982) and Wang and Wang (1994) deal with various design optimization
methods for designing a compact gear set where the design must satisfy constraints on bending
strength, pitting resistance and scoring. Thompson et al. (2000) report on the trade-offs involved in
gear design and provide a method for preliminary design of a compact spur gear reducer. Wang and
Wang (1994) consider the selection of bearings and dimensioning of the shaft along with the gear
mesh parameters and have formulated and solved a multi-objective gear design optimization
problem.
Typically, a range of materials is available to a gear designer for preliminary material selection
(Ashby 2005). However, a designer uses these as look-up databases for shortlisting material
options for further exploration. The in-service performance of the material depends on the effect
of the manufacturing process in addition to the material composition. Integrated computational

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

N. Kulkarni et al.

materials engineering (ICME) is conceived as a way in which future materials development will
be done in close association with end-product design (Pollock et al. 2008; McDowell et al.
2010; Schmitz and Prahl 2011). It is based on the use of modelling and simulation tools to
facilitate exploration of the materials and product design space simultaneously and thus reduce
time-consuming and expensive experimentation. These tools are expected to map the composition,
processing, structure and properties of materials, and ultimately link them to product performance.
However, the current state of the art of materials science maps these only partially, with signicant
gaps (Pollock et al. 2008; McDowell et al. 2010; Schmitz and Prahl 2011; Fullwood et al. 2010).
Attempts to develop materials for specic needs of gears can be found in Schmitz and Prahl (2011)
and Wright et al. (2010). Schimitz and Prahl (2011) discuss development of microalloyed steels
for automotive gear applications and Wright et al. (2010) have designed and developed highperformance gear materials for aerospace applications. However, in these studies the material
space and the geometric design space are not explored simultaneously and are thus limited to the
development of materials and manufacturing processes for achieving specied material properties.
The simultaneous exploration of materials and products could lead to better products.
A vast number of parameters of composition and processing governs the properties of materials.
This makes the design space very large. Besides this, industry prefers to use standardized material
compositions and simplify the incorporation of material variables in design. In view of this, as
a starting point, the material space is restricted to be represented by equivalent material tensile
strength and to develop a method for concurrently exploring geometric design and material spaces
for gears. This involves large numbers of constraints and conicting goals and therefore has been
a difcult problem to solve. In this work an attempt has been made to ll this gap by offering a
method to explore the complex and coupled geometry and material space in the design of gears
using an example of the conceptual design of a spur gear. When seeking a conceptual design,
designers typically explore a large region of the design space to identify a good design, which
then can be used as a starting place for further evaluation and design.
For the conceptual design of a spur gear design, a designer is interested in a region of the
design space containing satiscing1 design solutions, rather than a single optimum solution. The
decision support problem (DSP) technique proposed by Mistree and co-authors (Bras and Mistree
1991; Kamal et al. 1987; Mistree et al. 1989, 1990), especially the compromise decision support
problem (cDSP) (Mistree, Hughes, and Bras 1993), has been developed specically to assist
designers in this context. The cDSP is useful when multiple conicting objectives are present
and a designer seeks a compromise among these objectives. The cDSP has been used in a wide
variety of applications for the design of components and complex systems (Jivan and Mistree
1985; Lyon and Mistree 1985; Nguyen and Mistree 1986; Bascaran, Mistree, and Bannerot 1987;
Marinopoulos et al. 1987; Shupe and Mistree 1987; Rolander et al. 2006; Kulkarni et al. 2013;
Kumar et al. 2013). The cDSP differs from standard optimization formulations in that it is a
hybrid formulation based on mathematical programming and goal programming. The cDSP is
implemented in the DSIDES (Decision Support In the Design of Engineering Systems) software.
The cDSP, along with DSIDES, has been used extensively to solve a variety of multi-objective,
nonlinear problems in engineering, such as in the design of turbulent convective systems (Rolander
et al. 2006), the preliminary design of ships (Lyon and Mistree 1985), the conceptual design of
an aircraft (Marinopoulos et al. 1987), damage-tolerant structural systems (Shupe and Mistree
1987), thermal design (Bascaran, Mistree, and Bannerot 1987), the design of horizontal pressure
vessels (Nguyen and Mistree 1986), the design of helical compression springs (Jivan and Mistree
1985), the design space exploration for continuous casting of steel (Kumar et al. 2013) and the
design of gears (Kulkarni et al. 2013).
In this work, the cDSP gear design formulation (Kulkarni et al. 2013) is extended to make it
possible to explore the combined design space of the material and the gear geometry and offer
the designer and the customer compromises or trade-offs among the various goals.

Engineering Optimization
Table 1.

Gear design goals and the associated conicts.

Goal

Requirements

Maximum reliability design

Compact design

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Minimum cost design

2.

Larger gear geometry (gear module, m,

number of teeth, N, and face width, b) and
high-strength material
Compact design, minimum cost design
Smaller m and N, and high-strength material
Smaller gear geometry (m, N and b) and
low-cost material

Conicts

High-reliability design, possible

weak conict with cost
High-reliability design

Geometry and material space exploration

Designing a gear that satises user-specied functional and performance requirements and performs satisfactorily in service is a challenging task. Gear designers have different design goals,
such as high reliability (which is important for the users safety), low cost (which is important
for the companys protability) and compactness (which is important from the geometry point of
view as the gearbox must t in the given space). These goals are conicting and couple tightly
the three aspects of gear design, namely, geometry design, material selection and manufacturing, and nal production cost. Thus, it is important that a designer considers geometry, material
and manufacturing processes in an integrated way to explore the synergy among them. An integrated gear design approach will overcome time-consuming design iterations and reduce product
development time and costs.
To show the advantages of concurrent exploration of design spaces, the scope of the current
study is limited to the concurrent exploration of the combined geometric and material space for
the design of a gear and will show the effectiveness of the proposed method. The details of the
various goals considered in this study and associated conicts are presented in Table 1.
The concurrent exploration of geometry design and material space also involves large numbers
of constraints and conicting goals. The cDSP, which is a mathematical construct capable of
modelling and addressing such problems, is used in this article (Mistree, Hughes, and Bras 1993;
Rolander et al. 2006). The cDSP enables the construction of different practical scenarios in a multiobjective formulation by offering the opportunity for assigning appropriate weights to different
goals and exploring the compromises among them. In the cDSP, the difference between the desired
(the target, Gi ) and the achieved (Ai (x )) value of a goal is minimized. The difference between
these values is modelled as two deviation variables, di+ and di , which represent overachievement
and underachievement, respectively, for each goal. A cDSP is constructed such that deviations
are always positive and simultaneous overachievement or underachievement is not allowed. The
cDSP template formulation is shown in Table 2. Further detail on the formulation and application
of the cDSP for the current design problem is given in Section 3.2.

3.

3.1.

Compromise decision support problem-based geometry and material space

exploration for spur gear design
Problem statement

The current analysis deals with a problem motivated by industry, the design of a pinion for the
rst gear reduction for a compact sized automobile. The problem statement is as follows:
Design a compact, highly reliable, compact and low-cost pinion of a commercial spur gear
system of AGMA precision number2 8 to carry a torque of 113 Nm at 4500 rpm with a gear ratio

N. Kulkarni et al.

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Table 2. General mathematical formulation of the compromise decision support problem (Mistree et al. 1990).
Given
An alternative to be
improved
Assumptions used to model
the domain
The system parameters
n number of system
variables
p number of inequality
constraints
q number of equality
constraints
m number of system goals
Find
Design variables
Deviation variables
Satisfy
Inequality constraint
Equality constraint
Goals
Bounds
Minimize
Deviation function:
Archimedean formulation

xi
di+ , di

i =1,,n
i =1,,n

gi (x ) 0
hi (x ) = 0
Ai (x ) di+ + di = Gi
xi,L xi xi,U
di+ 0, di 0, di+ di = 0

i =1,,p
i =1,,q
i =1,,m
i =1,,n
i =1,,m

Z=

m
1

Wi (di+ + di )

i =1,,m

of 3.5. The design must allow for moderate shock in the driving engine and moderate shock in the
driven machinery.3 The target for reliability should be 99.99% and should not less than 95%. The
gear teeth must have a pressure angle of 20 and they will be cut using a rack cutter. The expected
fatigue life is 109 cycles. The steel for the gear is available in the range of tensile strengths of
8001600 MPa. The maximum allowable centre distance is 300 mm. Standard gear geometry as
per AGMA standard is desired.
3.2.

Formulation

Following the example in Rolander et al. (2006) and the template described in Table 2, the cDSP
for the design of a gear is given below.
3.2.1.

Given

A spur gear pinion is required to be designed based on AGMA (1993, 2010) standards based on
the 20 standard pressure angle system. The gear is required to carry 113 Nm torque. The required
gear ratio, G, is 3.5.
The standard geometry of a gear is dened by module, m, number of teeth, N, and face width, b
(ANSI/AGMA 1993, 2010). Descriptions of gear geometry in terms of module, number of teeth
and face width, material strength and reliability can be found in ANSI/AGMA (1993, 2010) and
Shigley (1977). The material space is determined by the ultimate tensile strength of the steel, SU,
and is used as a design variable for materials design. Reliability R also has a range of minimum
permissible to maximum targeted values and there is scope for exploration in the design. Thus,
there are ve design variables for the design problem.
The physical limitations on design are represented as bounds and system constraints. The bounds
on these design variables are selected based on the knowledge available to the designer. Based on

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

the problem statement and the AGMA requirements (ANSI/AGMA 1993, 2010), seven constraints
are selected. Two constraints are specied on the factor of safety in bending and contact. The space
constraint is incorporated using the maximum allowable centre distance. Considering uniform
load application on the entire face of the gear, AGMA (1993, 2010) recommends restricting
the minimum and maximum allowable face width as a function of module and these become two
additional constraints. The contact ratio is an important factor for smooth running of the gear drive
and two additional constraints are specied on the minimum and maximum allowable values for
contact ratio based on AGMA (1993, 2010) recommendations.
Three design goals are selected for investigation. These are maximization of reliability, maximization of compactness (i.e. minimization of centre distance between gear and pinion) and
minimization of cost. These lead to six deviation variables, one each corresponding to the
overachievement and underachievement of three design goals. The cDSP allows designers to
incorporate given objectives as constraints as well as goals with desirable target values. In this
study, compactness limit is modelled as constraint with a desirable goal.
Assumptions: Helical gears are preferred for the application described above. However, this
study is restricted to spur gears for the purposes of illustration of the method. The material space
is represented by considering only the ultimate tensile strength (SU ) of the material as many of
the material properties, such as hardness and fatigue endurance strength, can be related to the
ultimate material strength. In practice, modules for the gear have discrete values due to tooling
limitations. However, for simplicity, the module is assumed to be a continuous variable.
3.2.2.

Find

The standard gear geometry is expressed using AGMA standards (ANSI/AGMA 2010) in terms
of:
module (m) in mm: the module is the ratio of pitch circle diameter to the number of teeth; it
has units in millimetres and it is the index of tooth size in SI
number of teeth (N)
face width (b) in mm: the face width of a gear refers to the thickness of the gear teeth.
These geometry variables are used along with
material strength (SU ) in MPa
reliability (R).
Find design variables m, N, b, SU and R and the deviation variables, di+ , di .
3.2.3.

Satisfy

Design constraints
C1 : Minimum face-width : b 3m

(1)

C2 : Maximum face-width : b 5m

(2)

C3 : Maximum limit on centre distance# : d = m(1 + G)N/2 300 mm

(3)

C4 : Bending stress induced# : allowable induced 0

(4)

bending

bending

N. Kulkarni et al.
contact
contact
C5 : Contact stress induced# : allowable
induced
0

C6: Minimum contact ratio : Rc 1.4

C7 : Maximum contact ratio : Rc 1.8

(5)
(6)
(7)

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

as per AGMA guidelines (ANSI/AGMA 1993, 2010); # as per problem statement.

These constraints are normalized based on the method given by Mistree, Hughes, and Bras
(1993).
Goals
Based on the problem statement, maximization of reliability, minimization of centre distance
and minimization of cost are selected as the three design goals. The target value for the three
design goals is chosen (based on the guidelines in Mistree, Hughes, and Bras 1993) such that
it is greater than or equal to the maximum expected value for maximization goals
it is less than or equal to the minimum expected value for minimization goals.
The goals are dened as given below:
G1: Achieve the maximum desired reliability R.
The goal for maximum reliability is taken as 99.99% (which is the maximum achievable reliability given in AGMA). To model the greatest achievement of the goal (Mistree, Hughes, and
Bras 1993), the reliability goal is expressed as:
R
+ d1 d1+ = 1
(8)
0.9999
In AGMA-based design, the inuence of reliability is brought in by a correction factor (Yz ) in the
allowable stresses during gear design through constraints C4 and C5. The details are available in
Shigley (1977).
G2: Minimize centre distance d (makes the design compact).
The centre distance (d) is dened as
d=

m(1 + G)N
2

(9)

The target centre distance is taken a value corresponding to the minimum m and minimum N,
which is calculated as d = 162 mm. The centre distance goal, with a target of 162 mm, is expressed
as
162
+ d2 d2+ = 1
(10)
d
G3: Minimize cost C.
The total cost consists of the material cost and the manufacturing cost and is calculated using an
empirical relationship as assumed below:



1.5 
SU (SU )min
C = W a0 + b0 1 +
(11)
(SU )max (SU )min
where W is the weight of a component, a0 is the manufacturing cost [taken as Indian rupees (INR)
5/kg], and b0 is the material cost (taken as INR 40/kg for 800 MPa SU steel). The increase in
material cost with increase in tensile strength is given in Equation (11). The above equation for

Engineering Optimization

cost is valid between (SU )min and (SU )max , which are the minimum and maximum tensile strengths
of available materials. The target value of cost corresponds to a gear with a minimum volume and
lowest material strength, which is calculated to be C = INR 57.00. The cost goal is expressed as
57
+ d3 d3+ = 1
C

(12)

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Bounds on design and deviation variables are given by:

B1: 4 m 8 (mm)
B2: 18 N 40
B3: 40 b 80 (mm)
B4: 800 SU 1600 (MPa)
B5: 0.95 R 0.9999
B6B11: restriction on deviation variables, di+ di = 0 for i = 1 . . . 3 and di+ , di 0 for
i = 1...3

3.2.4.

Minimize

The goals are normalized as suggested by Mistree, Hughes, and Bras (1993). The Archimedean
approach is used to construct the effective deviation function and it is solved using DSIDES
software.
Z(d , d + ) = W1 (d1 d1+ ) + W2 (d2 d2+ ) + W3 (d3 d3+ )
where

3


Wi = 1

and Wi 0

(13)

i=1

3.3.

Gear design scenarios

The cDSP allows a designer to explore compromises among different goals by generating various
design scenarios by giving different weights to each of the goals. A schematic of the different
design scenarios considered in this study is shown in Figure 1.
At each vertex, an individual design goal has been given the full weight of 1.0, e.g. Scenario
S1 corresponds to a highly reliable design. Hence, the reliability goal is given a weight of 1.0
and the other two goals have weights of 0.0 each. Other scenarios are generated by considering
appropriate contributions from each of these goals, e.g. Scenario S4 corresponds to a reliable
and compact design, with weights of 0.5 each for reliability and compactness, and cost is given
a weight of 0.0. Scenario S7 corresponds to a reliable, low-cost and compact design, with equal
weights of 0.33 for each of the goals.

4.

Results and discussion

DSIDES provides many control options, namely, opportunities to control exploration of the
design space, specication of the maximum number of iterations and criteria for convergence
of design/deviation variables. In this work, the maximum number of iterations is set to 100 and
the stationarity (which is part of the stopping criteria and denes the convergence tolerance) of
deviation and design variables to 2%.
The results from DSIDES of the seven different scenarios illustrated in Figure 1 are analysed
in detail. While analysing the results of different scenarios, the following observations are made.

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

10

N. Kulkarni et al.

Figure 1. Schematic representation of different design scenarios. Studying these scenarios offers an understanding and
exploration of the design space. At each scenario location, the goals are weighted to assign their importance. In S1, S2
and S3 (reliability, compactness and cost goals, respectively) are weighted 1 and the other goals are assigned zero weight.
In scenarios S4, S5 and S6, the goals indicated are each assigned a value of 0.5 and the remaining goal is assigned a value
of 0. In S7, all goals are assigned a value of 0.33.

In scenarios such as compact design (S2) or high-reliability design (S1), oscillations in the convergence behaviour of design variables and deviation variables are observed. The convergence
behaviour of design variables for Scenario S2 is shown in Figure 2(a). The variables dening
compact designs (i.e. minimum m and minimum N) converge within the rst few iterations. The
other three design variables oscillate among three possible states. In all of these states the contact
stress constraint is active and violated within acceptable limits. This convergence behaviour is
expected as consideration of the compact design goal alone leads to an underconstrained system
with no constraints on face width, reliability or tensile strength, and these three parameters can
take theoretically innite feasible combinations. Similarly, in other scenarios such as S1, the same
reliability can be achieved by different combinations of geometry (m, N and b) and material (SU ).
For Scenario S4, which is a combination of scenarios S1 and S2, oscillation is observed in SU and
b in a few initial iterations and nally convergence is obtained.
The cost goal involves all of the design variables and hence it is well constrained. This is
observed from the convergence behaviour of the design variable (Figure 2b) for the design scenarios S3, S5, S6 and S7. The output design variables and achieved goal values for all these
scenarios are listed in Table 3.
A comparison of normalized goals achieved for different scenarios is shown in Figure 3. These
results are in line with the known properties of gear design; for example, when a compact design
is required, the material strength needs to be high, and high reliability requires an increase in the
size of gear as well as the strength of the material.
Observations from the results show that the minimum cost design is simultaneously a compact
design and a minimum face-width design. However, a compact design is not necessarily a minimum cost design as it is a design with larger face width compared to the minimum cost design.
The requirement for high reliability conicts with compactness or cost, as can be expected. Scenarios S4, S6 and S7 are the best gear design scenarios, which have high reliability and exploit
compromises among various goals.
To observe the achievability of goals visually, additional scenarios are considered with different
weights of goals, formulated with the corresponding cDSP and solved using DSIDES. All possible

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

11

Figure 2. Convergence behaviour of design variables for (a) compact design, S2: this convergence behaviour suggests
that there are multiple acceptable designs; and (b) minimum cost design, S3: this convergence behaviour suggests that
there is only one acceptable design.

Table 3.

Scenario #
S1
S2
S3
S4
S5
S6
S7

Results: design variables and goals for different scenarios.

Module,m
(mm)

Face width,
b (mm)

Number of
teeth, N

Tensile strength,
SU (MPa)

Reliability,
R (%)

Centre distance,
d (mm)

Cost,
C (INR)

6.59
4.00
4.00
4.99
4.00
5.69
4.99

67.29
57.32
40.26
76.72
40.26
57.59
76.70

18
18
18
18
18
18
18

1221
1444
1464
1550
1464
1550
1550

99.99
99.30
95.40
99.99
95.40
99.99
99.99

266.94
162.00
162.00
201.95
162.00
230.53
210.96

349.48
134.55
96.14
307.71
96.14
300.96
307.67

Note: In S1, S2 and S3 respectively, the reliability, compactness and cost goals are weighted 1 and the other goals are assigned zero weight.
In scenarios S4, S5 and S6, the goals indicated are assigned values of 0.5 and the remaining goal is assigned a value of 0. In S7, all goals
are assigned a value of 0.33.

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

12

Figure 3.

N. Kulkarni et al.

Comparison of calculated goal achievement for different scenarios.

feasible designs obtained during the analysis are collated. Three ternary charts are created (as
shown in Figure 4), considering the averaged value of the normalized achievement values of the
three goals to study the design scenarios, for each of the goals.
In Figure 4 (a) the visual representation of the design space where the feasible designs are
collated is shown, and a ternary plot is created for the reliability goal achievement. The ternary
charts shown in Figure 4(b) and (c) are created in similar way for achieved cost and achieved
compactness goals, respectively. Maximum and minimum values of the achieved goals are shown
in Table 4.
These achieved goal values are normalized between 0 and 1 such that 1 (blue) represents maximum achievement of the given goal and 0 (red) represents minimum achievement. To illustrate,
the achievement of maximum reliability (99.99%) is represented by blue and that of minimum
reliability (95.4%) is represented by red (Figure 4a); in the case of cost the objective is to minimize
it, so the achieved minimum cost (INR 96) is represented by blue and the achieved maximum
cost (INR 350) is represented by red (Figure 4b). For compactness, the objective is maximization, which corresponds to minimization of the centre distance, so the achieved minimum centre
distance (162 mm) is represented by blue and the maximum achieved centre distance (267 mm)
is represented by red.
These ternary charts can be used to compare and observe the implications of a designers
decision on the selection of goals for gear design. To illustrate, if the designer is interested in high
reliability design (the blue region marked by the green dashed line in Figure 4a), its implication
(i.e. maximum achievable values for compactness and cost goals) on the design space of other
goals (shown by the green dashed line in the corresponding region in Figure 4b and c) can be
seen. Similarly, the implications of the designers emphasis on low cost (the region shown by the
black dashed line in Figure 4b) and compactness (the region shown by the purple dashed line in
Figure 4c) on the other goals can be seen. The reliability goal is in direct conict with the other two
(i.e. low cost and compactness) goals. The low-cost design is simultaneously a compact design,
but a compact design is not necessarily a low-cost design. This can be explained as follows.
A low-cost design corresponds to a design having lower weight and lower material cost. A
design with lower weight corresponds to a design having minimum centre distance (compact

13

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

Figure 4. Visualization of the design space for different goals. The colour bars indicate the degree to which the goal is
maximized in the solution; blue indicates the maximum achievement of the given goal while red indicates the minimum
achievement of the goal. (a) Solution space of scenario where the goal is maximization of reliability; (b) solution space
of scenario where the goal is minimization of cost; (c) solution space of scenario where the goal is maximization of
compactness. In all the scenarios (a, b and c), the desirable values are represented in blue. The green dashed line shows
the boundary of the region (blue region) in (a) where the normalized reliability achievement is more than 80%, while this
same line in (b) and (c) shows the corresponding region where reliability achievement is more than 80% in (a). Similarly,
the black dashed line is the boundary of the region where minimum cost achievement is more than 80% (in b and the
corresponding region in a and c) and the purple dashed line is the boundary of the region where maximum compactness
achievement is more than 80% (in c and the corresponding region in a and b).

14

N. Kulkarni et al.
Table 4. The maximum and minimum values achieved
for each goal obtained in any of the scenarios.
Goal

Achieved value
Minimum

Maximum

95.4
96
162

99.99
350
267

Reliability (%)
Cost (INR)
Centre distance (mm)

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Table 5. Validation of results obtained using DSIDES with

AGMA design.
Face width b (mm)
Using AGMA
Using DSIDES
% Error

Design #1

Design #2

Design #3

77.81
78.73
1.18

39.23
40.00
1.96

58.40
61.34
5.03

design) as well as having lower face width. Lower material cost is obtained by using a material
of lower strength.
On the contrary, in the compact design no restriction is placed on material strength (which
affects the material cost) or on the face width. Therefore, a compact design corresponds only
to a design having minimum centre distance.
Overall, the central region shows the region of compromise. Such design space visualization
charts (as shown in Figure 4) are useful for a gear designer to make informed decisions on the
selection of design goals and the weights for these goals.

5. Verication
Various designs for different scenarios are obtained using DSIDES, which linearizes constraints,
and three of these gear designs are selected randomly for validation. To validate the designs, the
design values obtained for m, N, SU and R are substituted in the AGMA gear design formulae
and the face width for the selected cases is computed using appropriate factors of safety. The face
widths computed using AGMA and those given by DSIDES are given in Table 5.
From Table 5, it can be observed that the results obtained using DSIDES agree well with those
obtained using AGMA. The difference in the value of face width for design #3 is due to the
presence of active constraint C5 in the formulation.

6.

Closing comments

Designing a gear to meet specied functional, non-functional and performance requirements is

a challenging task. However, having the option of exploring different combinations of materials and geometry allows a designer to understand trade-offs and explore different options. It is
advantageous to explore geometry and material space simultaneously as this can reduce costly
design iterations and save design time as well as reduce the nal product cost. In this article,
a novel method for the design of gears using the concurrent exploration of the geometry and
material space, using the CEM, is demonstrated. This method will have signicant utility for
industry, where the total cost of the product must be optimized with trade-offs among the cost of

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

15

material, weight of the transmission system, and its durability and reliability. All of these must be
accomplished within the constraints of available space.
Conicting gear design goals such as compactness, high reliability and low cost are considered.
Inspection of the results from simulations provides insights into the behaviour of design variables,
deviation variables and constraints. The geometry and material strength requirements for different
scenarios that have been obtained are in agreement with existing knowledge of gear design and
the method used provides the means for a systematic exploration of the design space.
In this case, a conceptual design is sought for a gear rather than a nal design. This work is
intended to be used to rapidly explore the design space and provide a starting place for a designer,
who then will introduce various considerations of manufacturing, materials availability and purity,
etc. Therefore, there are some approximations and limitations in the accuracy of the model. For
instance, only the ultimate tensile strength (SU ) of the material is used to represent the material
space, as many of the material properties, such as hardness and fatigue endurance strength, can be
related to the ultimate material strength. However, a more complete and complex representation
of the material space could be included in the formulation. This does not detract from the method
presented here for conceptual design.
In the longer term, this work is intended to form a module in a demonstration information
technology platform for ICME (Pollock et al. 2008) in which it will be possible to simultaneously design both the material and the product, instead of relying on the current procedure of
selecting materials to build products. Current work (Bhat et al. 2013; Gautham et al. 2013) at Tata
Consultancy Services on a Platform for the Realization of Engineered Materials and Products
(PREMP) (Bhat et al. 2013; Gautham et al. 2013) is enriched with various tools for modelling
and simulation, supported by tools for enabling knowledge engineering, collaboration, robust
design, decision support and data-mining, along with appropriate databases and knowledge bases
that will facilitate the simultaneous exploration of the material and product design spaces. Apart
from the proposed method for gear design, this work is useful because of the ternary design charts,
which are helpful in the conceptual design phase for gear designers to visualize the complex and
coupled design space, and to make informed decisions on weights to be given to individual design
goals. This will aid the understanding of a variety of engineers and managers who wish to use
information about gear design but do not have the time or resources to calculate this information.
Further, the CEM is generalizable to other problems that require trade-offs.
Acknowledgements
The authors gratefully acknowledge the management of Tata Consultancy Services for providing support for carrying out
this research.

Funding
F. Mistree and J.K. Allen gratefully acknowledge support from NSF [grant CMMI 1258439] and the L.A. Comp and John
and Mary Moore Chairs, respectively. J.H. Panchal acknowledges the nancial support from NSF [grants 1265622 and
1201114].

Notes
1.

Satiscing is a decision-making strategy or cognitive heuristic that entails searching through the available
alternatives until an acceptability threshold is met. http://en.wikipedia.org/wiki/Satiscing
2. AGMA 2000-A88. AGMA has dened a set of quality numbers which dene the tolerances for gears of various
sizes manufactured to a specied accuracy. Quality numbers 37 include commercial-quality gears. Quality numbers
812 are precision quality.

16

N. Kulkarni et al.

3. AGMA (1993, 2010) recommends the overload factor, which is intended to make allowance for all externally applied
loads in excess of the nominal tangential load, to be selected appropriately based on operating conditions.

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

References
ANSI/AGMA. 1993. Design Guide for Vehicle Spur and Helical Gears. ANSI/AGMA 6002-B93, Revision of AGMA
170.01.
ANSI/AGMA. 2010. Fundamental Rating Factors and Calculation Methods for Involute Spur and Helical Gear Teeth.
ANSI/AGMA 2001-D04, Revision of ANSI/AGMA 2001-C95, Reafrmed Jan. 2010.
Arora, J. S. 2011. Introduction to Optimum Design. 2nd ed. London: Academic Press-Elsevier.
Ashby, M. F. 2005. Materials Selection in Mechanical Design. Oxford: Butterworth-Heinemann.
Bascaran, E., F. Mistree, and R. B. Bannerot. 1987. Compromise: An Effective Approach for Solving Multiobjective
Thermal Design Problems. Engineering Optimization 12 (3): 175189.
Bhat, M., S. Shah, P. Das, P. Kumar, N. Kulkarni, S. S. Ghaisas, and S. S. Reddy. 2013. PREMP: Knowledge Driven
Design of Materials and Engineering Process. ICoRD13, Lecture Notes in Mechanical Engineering, A. Chakrabarti
and R. V. Prakash, eds., 13151329. Springer India.
Bras, B. A., and F. Mistree. 1991. Designing Design Processes in Decision-based Concurrent Engineering. SAE Transactions, Journal of Materials and Manufacturing (SAE Paper 912209), Warrendale, Pennsylvania: SAE International.
451458.
Brauer, J. 2004. A General Finite Element Model of Involute Gears. Finite Elements in Analysis and Design 40 (1314):
18571872.
Buckingham, E. 1949. Analytical Mechanics of Gears. New York: McGraw-Hill.
Carroll, R. K., and G. E. Johnson. 1984. Optimal Design of Compact Spur Gear Sets. Journal of Mechanisms,
Transmissions and Automation in Design 106 (1): 95101.
Dudley, D. W. 1962. Gear Handbook. New York: McGraw-Hill.
Dudley, D. W. 1964. How Increased Hardness Reduces the Size of Gear Sets. Product Engineering Nov. 6: 92102.
Estrin, M. 1980. Optimization of Tooth Proportions for a Gear Mesh. New York: American Society of Mechanical
Engineers. Paper 80-C2/DET-101.
Fullwood, D. T., S. R. Niezgoda, B. L. Adams, S. R. Kalidindi. 2010. Microstructure Sensitive Design for Performance
Optimization. Progress in Materials Science 55: 477562.
Gautham, B. P., A. K. Singh, S. S. Ghaisas, S. S. Reddy, and F. Mistree. 2013. PREMP: A Platform for the Realization
of Engineered Materials and Products. ICoRD13, Lecture Notes in Mechanical Engineering, A. Chakrabarti and
R. V. Prakash eds., 13011313. Springer: India.
Gear Industry Vision: A Vision for the Gear Industry in 2025. 2004. Developed by the Gear Community. Accessed May
27, 2012. http://agma.server294.com/images/uploads/ gearvision.pdf
Goldfarb, V. I., A. Kapelevich, and A. A. Tkachev. 2008. An Advanced Approach to Optimal Gear Design. Gear Solutions
Aug: 4051.
Handschuh, R. F., and A. J. Zakrajsek. 2011. High-Pressure Angle Gears: Comparisons to Typical Gear Designs."Journal
of Mechanical Design 133 (11): 114501.
Hewitt, J. 1992. Design and Materials Selection for Power-Transmitting Gears. Materials and Design 13 (4): 230238.
ISO. 2006. Calculation of Load Capacity of Spur and Helical Gears Parts 1 to 3. ISO 6336-1 to 3.
Jivan, A., and F. Mistree. 1985. A Computer-Based Method for Designing Statically Loaded Helical Compression
Springs. ASME Design Automation Conference, Cincinnati, ASME Paper No.85-DET-75.
Kamal, S. Z., H. M. Karandikar, F. Mistree, and D. Muster. 1987. Knowledge Representation for Discipline-independent
Decision Making. Expert Systems in Computer-Aided Design: Proceedings of the Ip5.2 Working Group on Expert
Systems on Computer Aided Design. J. Gero, ed. 289321.
Kapelevich, A. 2007. Direct Design for High Performance Gear Transmissions. Gear Solutions Dec. 2231.
Kawalec, A., J. Wiktor, and D. Ceglarek. 2006. Comparative Analysis of Tooth-Root Strength Using ISO and AGMA
Standards in Spur and Helical Gears with FEM-Based Verication. Journal of Mechanical Design 128: 11411158.
Kirov, V. 2011. Comparing AGMA and FEA Calculations. Gear Solutions Feb. 3845.
Kulkarni, N. H., B. P. Gautham, N. Singhal, and P. R. Zagade. 2012. Ordinal Optimization Based Framework for Optimization of Cooling Conditions for Reducing Distortion in Hot Rolled Asymmetric Sections. Materials Performance
and Characterization 1 (1), Paper ID MPC104430.
Kulkarni, N., P. R. Zagade, B. P. Gautham, J. H. Panchal, J. K. Allen, and F. Mistree. 2013. PREMP: Exploring the
Design and Material Space for Gear. ICoRD13, Lecture Notes in Mechanical Engineering, A. Chakrabarti and R.
V. Prakash, eds. 745757. Springer India.
Kumar, P., S. Goyal, A. K. Singh, J. K. Allen, J. H. Panchal, and F. Mistree, F. 2013. PREMP: Exploring the Design
Space for Continuous Casting of Steel. ICoRD13, Lecture Notes in Mechanical Engineering, A. Chakrabarti and
R. V. Prakash, eds., 759772. Springer India.
Lin, P. H., H. H. Lin, F. B. Oswald, and D. P. Townsend. 1998. Using Dynamic Analysis for Compact Gear Design.
DETC98, New York: American Society of Mechanical Engineers. Paper DETC98/PTG-5785.
Lyon, T. D., and F. Mistree. 1985. A Computer-Based Method for the Preliminary Design of Ships. Journal of Ship
Research 29 (4): 251289.

Downloaded by [University of Oklahoma Libraries] at 10:58 30 April 2014

Engineering Optimization

17

MacAldener, M. 2001. Tooth Interior Fatigue Fracture and Robustness of Gears. diss., Department of Machine Design,
KTH Royal Institute of Technology, Sweden.
Marinopoulos, S., D. Jackson, J. Shupe, and F. Mistree. 1987. Compromise: An Effective Approach for Conceptual
Aircraft Design. AIAA/AHS/ASEE Aircraft Design, Systems and Operations Meeting, Reston, Virginia: AIAA
Paper No. AIAA-87-2965.
McDowell, D., J. H. Panchal, H-J. Choi, C. Seepersad, J. K. Allen, and F. Mistree. 2010. Integrated Design of Multi-scale,
Multifunctional Materials and Products. New York: Elsevier.
Mistree, F., O. F. Hughes, and B. A. Bras. 1993. The Compromise Decision Support Problem and Adaptive Linear
Programming Algorithm. In Structural Optimization: Status and Promise, edited by M. P. Kamat, 247286. Reston,
VA: AIAA.
Mistree, F., D. Muster, J. A. Shupe, and J. K. Allen. 1989. A Decision-Based Perspective for the Design of Methods for
Systems Design. In Recent Experiences in Multidisciplinary Analysis and Optimization. Hampton, VA: NASA. CP
3031, 11111136.
Mistree, F., W. F. Smith, B. A. Bras, J. K. Allen, and D. Muster. 1990. Decision-Based Design: A Contemporary Paradigm
in Ship Design. Transactions, Society of Naval Architects and Marine Engineers 98: 565597.
Nguyen, N., and F. Mistree. 1986. A Computer-Based Method for the Rational Design of Horizontal Pressure Vessels.
Journal of Mechanisms, Transmissions and Automation in Design 108 (June): 203210.
Papalambros, P. Y., and D. J. Wilde. 2000. Principles of Optimal Design: Modeling and Computation. 2nd ed. Cambridge:
Cambridge University Press.
Pollock, T., J. E. Allison, D. F. Backman, M. C. Boyce, M. Gersh, E. A. Holm, R. LaSar, et al., eds. 2008. Integrated
Computational Materials Engineering: A Transformational Discipline for Improved Competitiveness and National
Security. Washington, DC: National Academies Press.
Rolander, N., J. Rambo,Y. Joshi, J. K. Allen, and F. Mistree. 2006. An Approach to Robust Design of Turbulent Convective
Systems. Journal of Mechanical Design 128: 844855.
Savage, M., J. J. Coy, and D. P. Townsend. 1982. Optimal Tooth Numbers for Compact Standard Spur Gear Sets. Journal
of Mechanical Design 104 (3): 749758.
Schmitz, G. J., and U. Prahl, eds. 2011. Integrative Computational Materials engineering: Concepts and Applications of
a Modular Simulation Platform. Wiley-VCH.
Shigley, J. E. 1977. Mechanical Engineering Design. 3rd ed. New York: McGraw-Hill.
Shupe, J. A., and F. Mistree. 1987. Compromise: An Effective Approach for the Design of Damage Tolerant Structural
Systems. Computers and Structures 27 (3): 407415.
Thomson, D. F., S. Gupta, and A. Shukla. 2000. Tradeoff Analysis in Minimum Volume Design of Multi-Stage Spur
Gear Reduction Units. Mechanism and Machine Theory 35 (5): 609627.
Tsai, M. H. and Y. C. Tsai. 1998. Design of High-Contact-Ratio Spur Gears Using Quadratic Parametric Tooth Prole.
Mechanism and Machine Theory 33 (5): 551564.
Tucker, A. L. 1980. The Gear Design Process. New York: American Society of Mechanical Engineers. ASME Paper in
Design 106 (1): 95101.
Wang, J. D., and I. M. Howard. 2006. Error Analysis on Finite Element Modeling of Involute Spur Gears. Journal of
Mechanical Design 128: 9096.
Wang, J., S. Luo, and Y. Wu. 2010. A Method for the Preliminary Geometric Design of Gear Tooth Proles with Small
Sliding Coefcients. Journal of Mechanical Design 132 (5), 054501. doi:10.1115/1.4001410.
Wang, H. L., and H. P. Wang. 1994. Optimal Engineering Design of Spur Gear Sets. Mechanism and Machine Theory
29 (7): 10711080.
Wright, J., J. Sebastian, C. Kern, and R. Kooy. 2010. Design, Development and Application of New, HighPerformance
Gear Steels. Gear Technology Jan: 4653.
Yeh, T., D. C. H. Yang, and S. H. Tong. 2001. Design of New Tooth Prole for High-Load Capacity Gears. Mechanism
and Machine Theory 36 (10): 11051120.