You are on page 1of 6

THE 7TH INTERNATIONAL CONFERENCE RESEARCH AND DEVELOPMENT OF MECHANICAL ELEMENTS

AND SYSTEMS

AN APPLICATION OF OPTIMAL SOLUTION CHOOSING METHODS IN PLANETARY


GEAR TRANSMISSION OPTIMIZATION
Jelena STEFANOVI-MARINOVI
Milo MILOVANCEVI

Abstract: In this paper, an original model for multicriteria optimization of planetary gear transmissions
is presented. The aim of this paper is the application of multicriteria optimization to planetary gear
transmissions limited to geared pairs. A mathematical model for optimization is defined by the variables,
objective functions and conditions required for the proper functioning of a system. The following is
adopted as optimization variables: teeth numbers, number of planetary gears, gear module and gear
width. Conditions required for the proper functioning of the system in the scope of geometry and strength
are expressed by the functional constraints. The basis of the model are analytical expressions for volume,
mass, efficiency and production costs, which represent objective functions. In this model, apart from the
determination of the set of Pareto optimal solutions, methods for choosing the optimal solution from this
set are included, too. There are: weighted coefficients method, lexicographic method, the - constraints
method and distance method. The complete optimization procedure is implemented in Plan Gears
Optimization software. Based on numerical examples obtained by application of this software, the
comparison of the optimization methods and program results analysis is presented.
Key words: planetary gear transmissions, multicriteria optimization, mathematical model, methods for
optimal solution choice

1. INTRODUCTION
The multi-objective optimization is a method which is
implemented in the development of many products and
processes in various ways. The number and actuality of
published researches indicate the importance and
contemporarity of optimization topics.
Multicriteria optimization problems are very common in
many scientific and technical problems and also in the
human life. In this type of optimization it is not possible
to express the solution quality by one criterion but several
unharmonized criteria must be discussed. The functions
which express these criteria cannot all have optimal
values at the same time. Such problems are called nontrivial multiple criteria (or multiple objective,
multicriteria) optimization problems. There are a lot of
different methods for solving such problems.
Optimization of mechanical systems supposes a very
complex mathematical model which has to describe the
operation of a real system in real circumstances.
Optimization tasks for mechanical systems are compound
processes of abundant theoretical researches which
integrate knowledge of mechanical systems design in
general, uniqueness of concrete mechanical system and
methods of mathematical optimization.
Gear trains as concrete mechanical systems can be also
subject of multicriteria optimization.
Planetary gear trains are important kind of gear
transmissions, and they can be subject of multicriteria

optimization application, too. It is impossible all types of


planetary gear transmissions include in the same paper,
specially taking into consideration that they have their
own geometrical conditions. Therefore, optimization is
applied to concrete type of planetary gear transmission. In
this paper, the optimization of the basic type of planetary
gearing is considered.

2. MATHEMATICAL MODEL FOR


PLANETARY GEAR TRANSMISSION
OPTIMIZATION
The basic type of planetary gearing (Fig.1), i.e. a design
which has a central sun wheel (external gearing - 1),
immobile central toothed rim (internal, annulus gear - 3),
and planetary gears (satellites - 2), is the subject of the
paper, limited to geared pairs. Satellites are in the same
time in mesh with external and internal gear. This type of
planetary gear transmission has the broadest use in
mechanics. It is often used as a single stage transmission
and thus as a building block for higher compound
planetary gear trains.
This gear transmission is characterized with high
efficiency, reliable work, small dimensions and mass etc.
The base of optimization process in mechanical systems is
comparison of mechanical systems with different
parameters in the same conditions and selection of the
best variant. Optimization process begins with generating
of solution groups for assigned starting conditions. Based
on established objective functions and constraints, an
1

optimal solution is selected, determined by designed


parameters.
The solution of optimization task finding begins by
mathematical model determination. Mathematical model
for optimization is defined by the variables, objective
(goal, target) functions and conditions required for the
proper functioning of a system expressed by the
functional constraints.

Selection of face width is usually assigned in regard to


reference diameter of pinion (b/d1), module (b/mn), or to
center distance (b/a) [6]. In this algorithm face width is
introduced in regard to reference diameter of pinion.
The optimization variables are of mixed type: numbers of
gear teeth ( z1 , z2 , z3 ) are integers, positive and negative,
number of planetary gears ( nw ) is a discrete value,
module ( mn ) is a discrete standard value (acc. to DIN
780), while gear width ( b ) is a continual variable.
Numbers of gear teeth and number of planetary gears are
non-dimensional values, while module and gear width are
given in millimeters.

2.2. Objective functions

Fig.1. Planetary gear transmission with immobile


internal gear

2.1. Variables
In the scope of the mathematical model definition, it is
necessary to determine the variables since each objective
function is the function of several parameters. By
applying optimization to planetary gear transmissions,
usually design parameters are adopted for variables.
In this paper, the following variables are considered: teeth
number of central sun wheel z1 , teeth number of
planetary gears (satellites) z 2 , teeth number of toothed
rim z3 , number of planetary gears nw , gear module mn
and gear width b .
Numbers of gear teeth are important parameters in
optimization in regard that the same or close gear ratio
can be achieved by different combinations of teeth
numbers. Most frequently, the choice of planetary
transmission gears teeth number starts with establishing a
relation between the central sun pinion and other gears
teeth number. The initial datum for teeth numbers
determination usually is central sun pinion teeth number
z1 .
Specifics of planetary gears imposes inclusion also the
number of satellites. Number of satellites is in relation
with condition of adjacency.
Gear module represents the basic parameter of gear,
because other dimensions are determined as module
function. Besides teeth number, this is the most frequently
used parameter for optimization of geared couples. For
limitation of tools number and control equipment of gear
transmissions, module values are standardized.
A basic criterion for selection of face width is possibility
for uniform distribution of load along the contact line.
2

In this model, the following characteristics are chosen for


objective functions of a planetary gear train: volume,
mass, efficiency and production cost of gear pairs.
Overall dimensions are the frequently used criteria for the
optimization of geared pairs in different shapes. Volume
mostly expresses this criterion. In this paper, the volume
of gear pairs is used.The approximation of gear volume
by cylinder volume with diameter equal to pitch diameter
and height equal to gear width, and having in mind that
satellites are inside the toothed rim, make possible for the
gear volume to be expressed by:

V=

m z
cos t
b n 3

4
cos cos wt 23

(1)

where t is the pressure angle at pitch cycle, wt 23 is the


working transverse pressure angle for the pair 2-3 and
is the helix angle at the pitch diameter.
Mass is one of the most used optimization criteria. Mass
is determined as sum of all gear masses in transmission.
Since the mass of a particular gear is determined as gear
volume multiplied by the density of gear material,
mz = Vz , this criterion receives the form

m = V1 + nw V2 + V3
The final expression of this function is:

m = 0.25 b
nw k2 z22

mn2
cos 2 t
[k1 z12
+
2
cos
cos 2 wt12

cos 2 t
cos 2 t
+ k3 z32
]
2
cos wt12
cos 2 wt 23

(2)

To determine the gear mass, the factor of deviation of real


gear shape from cylinder (k) has to be taken into account
also. For purposes of optimization, i.e. the comparison of
gears with different parameters, this factor does not have
a great significance, since it is given in advance due to the
shape of gear wheel hub and it is a constant in the process
of optimization.
Efficiency is one of the most important criteria for the
design and evaluation of construction quality. Power
losses in planetary transmissions consist of losses in gears
meshing, losses in bearings and losses due to oil mixing
and spraying. The calculation of gear transmissions
efficiency is generally confined to losses depending on
friction on tooth sides, i.e. on calculation of mesh power

losses. These losses can be determined with the highest


accuracy. At determination of losses in bearings
considerable errors can be occurred. Losses due to oil
mixing and spraying are insufficiently investigated. But,
losses in bearings and losses due to oil are minor part of
losses regarding losses at contacts and carriers. Bearing
those considerations in mind, we consider the following
expression for efficiency [1,5,7]:

=1

z3 0.15 0.35 0.20


+
+

z3 z1 z1
z2
z3

(3)

Economic demands must also be taken into consideration


when dealing with techno-economical optimization. First,
these demands are related to production costs. These costs
consist of costs for production material and costs of the
production process itself. The time needed for the
production of gears is taken as a measure of production
costs and as an economical factor. This function is then
determined as a sum of times needed for the production of
central sun wheel T1 , satellites T2 and toothed rim T3 , i.e.

FT = T1 + nw T2 + T3

(4)

Production times are determined according to the


technologies of Fette [8], Lorenc [9] and Hfler[10].

2.3. Functional constraints


Planetary gears represent a specific group of gear
transmissions. Therefore, there are numerous exceptions
that need to be taken into account for these transmissions
to function correctly compared with classical gear
transmissions. The exceptions considered in this model
are related to mounting conditions, geometrical conditions
and strength conditions.
The mounting conditions comprise the condition of
coaxiality, the condition of adjacency and the condition of
conjunction [3].
Geometrical conditions relate to undercutting and profile
interference, ratio of pressure angle to working transverse
pressure angle, tooth thickness and space width,
transverse contact ratio value, sliding speeds, ratio of gear
facewidth to reference diameter of the driving gear, etc.
These conditions are ensured in accordance with the
actual standards (ISO TC 60 list of standards 090915) [4].
As strength conditions, safety factors for bending strength
and surface durability of each gear are provided [4,11].

3. OPTIMIZATION PROCEDURE
In the formulated problem, six decision variables exist,
corresponding to the basic design parameters, thus vector
of decision variables receives the form:

x = ( x1 , x2 , x3 , x4 , x5 , x6 ) = ( z1 , z2 , z3 , nw , mn , b )

(5)

This vector is a solution of optimization task.


Also, four objective functions equal to the volume V ( x) ,
mass m( x) , efficiency ( x) and production costs T ( x )
exist. Since mass, volume and production costs should be
minimized, and efficiency should be maximized, the
following is denoted in this model:

f1 ( x) = V ( x), f 2 ( x) = m( x),
f3 ( x) = p ( x), f 4 ( x) = T ( x)

(6)

The mathematical model of this nonlinear multicriteria


problem can be formulated as follows:

max { f1 ( x ), f 2 ( x ), f 3 ( x ), f 4 ( x )}
subject to x S

(7)

The variables must satisfy given constraints which are


expressed as inclusion x S where S is the set of
feasible solutions (or feasible set). The set S is defined
as the set of all 6-tuples of design parameters such that
functional constraints are satisfied. According to the
structure of the set S , there exist discrete and continuous
multicriteria optimization problems, depending on
whether the set S is finite or continuous. The
optimization problem in this paper is discrete multicriteria
optimization problem.
Every point x S is mapped to the point
( f1 ( x ), f 2 ( x ),..., f k ( x )) in k - dimensional objective
space. Therefore it can be introduced as the objective set
in this mathematical model:

F = {( f1 ( x), f 2 ( x), f 3 ( x), f 4 ( x)) | x S }

(8)

As it can be seen from the definition, multicriteria


optimization problems are mathematically ill-defined.
This means that they have a set of mathematically
"equally good" optimal solutions in the objective space.
The most important criterion for selecting these "equally
good" solutions is Pareto optimality concept: Solution
x S is Pareto optimal if there is no solution y S
such that holds f i ( x ) f i ( y ) for all i = 1,K , n and for
at least one index i holds strict inequality, i.e.
f i ( x ) < f i ( y ) . Determination of the Pareto optimal
solutions set is a first step in optimal solution finding.
Next step is optimal solution choice from Pareto
solutions.
Thus, some additional information is needed in order to
be able to select one of them as a final solution. This final
decision is usually made either by decision maker or by
the corresponding scalarized problem. In the latter case,
one or more single criterion optimization scalarized
problems have to be constructed and solved.
The following methods based on the construction of
scalarized problem are suggested for usage in this
mathematical model: weighted coefficients method,
lexicographic method, the - constraints method and
distance method [2].

3.1. Weighted coefficients method


In the weighted coefficients method the following
scalarized problem is constructed:

max f M ( x ) = w1 f10 ( x) + K + wm f m0 ( x)
s.t. x S

(9)

Here, weighted coefficients (or weights) wi are positive


real numbers i = 1, 2,3, 4 . All solutions obtained by this

method are Pareto optimal [4]. The input parameters for


this method are values of weights.

3.2. Lexicographic method

optimal [4].
The shortened algorithm for the complete optimization
procedure is shown in Fig. 2.

In application of lexicographic method objective


functions need to be sorted by the given priorities. It can
be said that f k1 ( x ) has maximum priority, then f k2 ( x ) ,
etc. and f km ( x ) has the least priority. Then the following
list of scalarized problems for i = 1, K , m can be solved:

f kopt
= max f ki ( x)
i
s.t. x S
f k j ( x) = f

(10)
opt
kj

, for j = 1,K, i 1

In other words, objective functions are maximized


sequentially, but in each iteration the feasible set is
reduced to the set of optimal solutions in previous
iteration. A solution given by this method is also Pareto
optimal [4].

3.3. The constraints method


In the constraints method, one objective function
f q ( x ) which has to be maximized under the starting and
additional conditions is chosen. These additional
conditions are of the form f i ( x ) i , where i , i q
are given thresholds. Therefore, the following scalarized
problem is solved:

max f q ( x )
s.t. x S

(11)

f i ( x ) i , for i = 1,K , n, i q
In this case, a systematic variation of i yields a set of
Pareto optimal solutions. This method is very commonly
used because it is possible to control the values of all
objectives exactly, which is also very important in
practical applications [4].

3.4. Distance method


The main idea in distance method is the minimization of
distance between one given (infeasible) reference point
f z and the objective set F . The following scalarized
problem can be formulated in this way:

min d ( f ( x ), f z )

(12)

s.t. x S

Here d ( x, y ) can be any metric function. In this


mathematical model Euclidean distance is used, thus
distance obtains the shape:

d ( x, y ) =

w (x y )
i =1

(13)

Here wi are given positive real numbers, weight


coefficients. Under the condition wi > 0 for all
i = 1, K , n it can be proven that a solution obtained by
this method and using Euclidean metric is Pareto
4

Fig.2. Shortened algorithm for optimization procedure

This model is basis of computer program for optimal


solution choice development.

The program made on this algorithm could provide


effective optimal solution choice for given input data:
transmission ratio, input number of revolutions, input
torque, selected materials of all gears, accuracy grade,
allowed deviation of gear ratio, range of z1 variation,
working conditions factor and minimal safety degree for
sides and root.

4. NUMERICAL EXAMPLES
The complete optimization procedure is implemented in
the Plan Gears Optimization software. Numerical
examples obtained using this software are presented in
this section.
The input data are:
i = 6.353 , nin = 3000 min-1, Tin = 39.787 Nm,

T = 3000 h, K A = 1.1 , IT 6 for all gears, S H min = 1.2 ,


S F min = 1.4 ,
material z1 /material z 2 /material z 3 =18CrNi8/18CrNi8/34
CrNiMo6, i = 3% , z1 = 15 36
These input data are in accordance with a realized
planetary transmission.
Set of feasible solutions consists of 33 elements. The
number of Pareto solutions is 6. Ideal values of functions
are determined and shown in Table 1.
Table 1: Ideal point coordinates

f 1id

in mm3
293205.628

f 2id

f 3id

f 4id

in kg
1.417

0.987

in min
82.967

Euclidean distance method with ideal point as a reference


point gives the solution, from Pareto set, shown in Table
2, with set of objective functions values shown in Table 3.
Table 2: Solution obtained by Euclidean distance method
z1

z2

16

33

Variable values
z3
nw
-83

mn

17

Table 3: Objective function for solution shown in table 2

f1

in mm3
367921.34

f2
in kg
1.747

f3

f4

0.985

in min
96.07

Taking into consideration the fact that it is needed to


determine the priority of functions for the next methods
anticipated here, the following examples are noted.
Function f1 is adopted as a priority function.
Weighted coefficients method demands application of
weighted coefficients values first, which has significant
influence on optimal solution selection. Thus, much
bigger weighted coefficient is given to the first function,
i.e. w1 = 0.7 , w2 = w3 = w4 = 0.1 . The obtained
solution is given in Table 4, with the set of objective
functions in Table 5. Other variations of weighted

coefficients
example w1

w1 >> w2 > w3 > w4 ,

with

for

= 0.8 , w2 = 0.1 , w3 = w4 = 0.05 , etc.,

also direct on the same solution.


Table 4: Solution obtained by weighted coefficients
method
z1

z2

15

30

Variable values
z3
nw
-78

mn

16

Table 5: Objective functions for solution shown in table 4

f1

in mm
293205.63

f2

f3

f4

in kg
1.417

0.984

in min
82.967

Application of Lexicographic method is on the same


example, with coefficients of tolerances 0% (this is
practically the basic Lexicographic method, without
relaxation) and priority of functions 1,2,3,4. Mentioned
solution given in table 4 is also obtained.
The - constraints method demands choosing the priority
function and others are translated into constraints by
giving the lower or upper bounds. In this example,
priority is given to the first function. The following values
are taken for the - constraints: 1.7 for second function,
0.98 for the third and 100 for the fourth. These values are
generally taken arbitrary. Values for constraints are in
this case chosen according to the ideal values of
functions, by increasing or decreasing of these values for
approximately 20% . Moreover, this method also points
to the same solution, table 4.
Based on the solutions in this example if the first function
has priority, we choose the solution shown in table 4.
Reason for this is in fact that the majority of methods
point to this solution. Also, values for first, second and
fourth function of this solution are values of ideal solution
components, while difference between the third function
components is 0.3% .
Solution obtained by Euclidean distance method is
characterized by a larger number of teeth on all gears than
other anticipated methods. This is due to the nature of the
third function. The increase in efficiency is followed by
the increase in gear transmission dimensions (diameter
and width). Euclidean distance method does not take into
consideration preferences of particular criteria.
In this way optimal solution regarding design parameters
is obtained. Before adopting solution for next step in
product development, technological demands should be
considered.
Optimal solution by prioritizing some other objective
function can be obtained by using the same procedure.
Based on all of them, the relations between particular
objective functions can be observed.
It can be concluded that, in the case of priority functions
existence, it is suitable to give precedence to the weighted
coefficients method due to very clear physical meaning
and experience in application on technical systems
optimization [4]. This method is possible to apply also in
the case of equal priority functions.
5

The -constraints method can be important if significant


constraints are known. This method is especially
applicable in situations when one function has the most
importance (i.e. mass due to building in some objects or
volume due to overall dimensions etc.), while others must
be in the range of some allowed limits. Setting the limits
for functions has significant influence on the result.
Reassigning the limits, more or less, influence the
problem solution. Euclidean distance method is very
suitable in case that there is some referent solution which
is usually infeasible. By applying this method, the
solution closest to this referent solution is obtained. A
good choice for the referent solution is an ideal point.

5. CONCLUSION
In this paper, an original model for multicriteria
optimization of planetary gear transmissions is presented.
Furthermore, this approach indicates a possibility for
mathematical methods application in planetary gear
transmission optimization. The mathematical model
consists of objective functions, variables and functional
constraints.
Besides the determination of the set of Pareto optimal
solutions, the presented original approach includes
methods which select an optimal solution from the Pareto
solutions set: weighted coefficients method, lexicographic
method, the - constraints method and distance method.
According mentioned examples it can be concluded that
these methods although start from different prepositions
and have different mathematical basis have very strong
correlation and they lead to the same results in numerical
examples, which provides them a physical meaning. It is
shown that the established optimization model gives good
results and can be used for this type of planetary gear
transmission.
Procedure applied here would be the same for other types
of planetary gear transmissions, only mathematical
expressions would be different.

REFERENCES
[1] DEL CASTILLO, J.M., (2002) The Analytical
Expression of the Efficiency of Planetary Gear
Trains, Mechanism and Machine Theory 37, pp 197214
[2] MIETTINEN K., KIRILOV, L. (2004) Interactive
reference direction approach using implicit
parametrization
for
nonlinear
multiobjective
optimization, MCDM 2004, Whistler, Canada
[3] NIEMANN
G.,
WINTER
H.
(1989):
Maschinenelemente, Band II, Springer-Verlag Berlin,
[4] STEFANOVI-MARINOVI
J
(2008)
Viekriterijumska optimizacija zupastih parova
planetarnih prenosnika, doktorska disertacija,
Mainski fakultet Univerziteta u Niu, Ni.
[5] VOLMER, J.(1990):Getriebetechnik,
Umlaufrdergetriebe, Verlag Technik, Berlin.
[6] VULI A., STEFANOVI -MARINOVI J. (2006)
Design
Parameters
for
Planetary
Gear
Transmissions Optimization, The 2nd International

Conference "Power Transmissions 2006", Novi Sad,


2006, Proceedings, pp 137-142
[7] VULI A., STEFANOVI-MARINOVI J. (2008)
Objective
Functions
for Techno-Economical
Planetary Gear Transmissions Optimization, The
Fifth International Symposium about Forming and
Design in Mechanical Engineering "KOD 2008",
Novi Sad 2008, pp 111-116
[8] FETTE-Technologie-Ihr Profit, Die
Schnittbedingungen beim Wlzfrsen, WILHELM
FETTE, GMB
[9] Verzahnwerkzeuge, Ein Handbuch fr Konstruktion
und Betrieb 3 (1977). Auflage, LORENZ GmbH&Co
Ettlingen
[10] BHS HFLER operating instructions, 6301000/2./125630-1000/2./166, 630-1000/4./016301000/2./67
[11] ISO 6336 (1996) Calculation of load capacity of spur
and helical gears: ISO 6336-1, ISO 6336-2, ISO
6336-3.

CORRESPONDANCE
Jelena
STEFANOVI-MARINOVI, Ph.D.
assistant prof.
University of Ni,
Faculty of Mechanical Engineering,
Aleksandra Medvedeva 14,
18000 Ni, Serbia
e-mail: jelenas@masfak.ni.ac.rs
Milo MILOVANEVI, Ph.D.
assistant prof.
University of Ni,
Faculty of Mechanical Engineering,
Aleksandra Medvedeva 14,
18000 Ni, Serbia
e-mail: milovancevic@masfak.ni.ac.rs

You might also like