Beruflich Dokumente
Kultur Dokumente
Richard
Assistant Professor.
Computer Implementation of an
Dept. of Mechanical Engr.,
Laval University,
Quebec, Canada G1K7P4
Optimal Conformal Mapping for
D. Pare
Gear Tooth Stress Analysis
Design Engineer. This paper describes a computerized version of the complex potential approach
General Electric of Canada, which is a comprehensive mathematical model for the stress analysis of spur gear
795 First Ave., Lachine, teeth. The entire procedure is a basic application of Hirano's conformal mapping
Quebec, Canada H8S 2S8
theory in which laws of elasticity have been combined. The main concepts of the
method have been explained in previous publications but the work described herein
A. Cardou is an appreciable extension of this relatively new approach. The algorithm is
Professor.
eminently well-suited for computer-aided-design of gear teeth; it serves as the basis
Dept. of Mechanical Engr., for an interactive computer program which can model a gear tooth and can calculate
Laval University, the stresses and displacements within the tooth when subjected to a concentrated
Quebec, Canada G1K7P4 load. Results are compared with AGMA's and other published values.
Introduction
Computer-based analysis and display techniques are playing The paper shows how the trial and error process of the CPM
a rapidly expanding role in the engineering design process may be performed numerically by computer as accurately as
throughout industry. The availability of sophisticated soft- needed and simultaneously visualized by display techniques.
ware and increasingly powerful hardware is allowing timely, The treatment of the mechanical strength theory is inter-
cost effective, detailed analysis aiding in the design of more ef- calated into the complex potential formula giving birth to an
ficient and reliable transmission components such as gear interactive program. The method is very methodical and well-
teeth. The refinement of formulation techniques for gear suited for computer implementation, thus lending itself to a
tooth stress analysis has led to the development of efficient systematic use for gear profiles found in practice. The Com-
algorithms. Several have been described in the literature ex- plex Potential formalism is an ideal algorithm for a CAD
ploiting various techniques such as photo-elasticity [1, 2, 3, 4, system because of the pictorial relationship existing between
5], finite elements [6, 7, 8, 9, 10], boundary integral equations the physical and mathematical system. In fact, an interactive
[11, 12,13,14, 15] and naturally algorithms based on the com- computer program called MPCGRAPH (for Methode des
plex potential method [16, 17, 18]. Potentiels Complexes GRAPHique which stands for the
The complex potential method (CPM) is a mathematical French version of graphical complex potential method) has
model which calculates the stress in spur gear teeth using con- been written and tested.
formal mapping as its modeling framework. The first signifi-
cant application of complex potentials in stress analysis was Complex Potential Method
developed in 1951 by Hirano [19]. He applied the CPM to
plane elasticity problems in domains consisting of a half plane Stresses and displacements for a structure when subjected to
with a symmetric protruding hump or "bell" shape (Fig. 1). a normal concentrated load on its boundary have been
The original domain is transformed into a half plane via a developed in the field of plane elasticity by Kolossov [20],
proper conformal transformation. Once the transformation If one applies a concentrated load to a half plane at Z0 in
completed, Hirano exploited Kolossov [20] formulas to com- the Z=x+ iy plane, the well-known complex potentials, <j>{Z)
plete his analysis. Later, Aida and Terauchi [21] have attemp- and \j/(Z) can be expressed as,
ted to apply Hirano's formulation to spur gears with poor
results due to an inadequate conformal transformation. This <MZ) = - - ^ l o g ( Z - Z 0 ) + </>*(Z) (1)
2TT
method was also examined by Ustinenko [22] and Rubenchik
[23]. Baronet and Tordion [16], in the early 1970, have finally We~ We'J)3
developed a conformal mapping which was applied to spur * ( Z ) = 2TT log(Z-Z 0 ) + 27T Z — Za + r(Z) (2)
gears. The method was then clearly exposed in 1981 [17] for
various profiles. where a and a' are complex constants and <j>*(Z), yp*(Z) are
holomorphic functions in the whole domain. In our case,
however, we apply the load W as shown in Fig. 1 at an ar-
Contributed by the Power Transmission and Gearing Committee for publica-
bitrary point Z0 of the boundary of a "bell" shape.
tion in the JOURNAL OF MECHANISMS, TRANSMISSIONS, AND AUTOMATION IN Now that the potential equations <j>(Z) and \j/(Z) have been
DESIGN. Manuscript received August 1987. defined for the half plane and the "bell" shape domain, we
Computer Implementation
The mathematical substitutions theorized in the preceding
section are rarely suitable for manual solution mainly due to
the omission of any simplifying assumptions and to the com-
plex nature of stress analysis problems. Nevertheless, this
iterant procedure is very appealing when executed by a tireless
digital machine. The unified methodology set forth in this
work for evaluating stresses and displacements in spur gear
teeth was implemented into a comprehensive computer pro-
gram called MPCGRAPH. This program consists of a main
program and 117 subroutines. The main program essentially
plays a coordinating role when calling the diverse subroutines.
It has two main functions; first, it sets up the geometrical en-
vironment to suit the needs of the interactive algorithm and
second, it serves as a central command center for the program.
A simplified flow-chart of the main program in terms of com-
mand subroutines and visual displays is depicted in Fig. 2.
Fig. 1 Conformal mapping The logic sequence of the program is guided by the follow-
ing steps:
(/) Type of computer display
must establish a relationship between the "bell" shape and the (2) Tooth generation
half-plane coordinate system. This relation will allow us to ex- (5) Translation axis coefficient
ploit the half-plane potentials <j> (Z) and ^(Z) in order to solve (4) Conformal mapping transformation
for the complex potentials of the "bell" shape domain. (5) Position of the concentrated force
The conformal transformations to map a tooth-like profile (6) Superficial stresses
represented in Fig. 1 in the Z-pIane (Z=x+ iy) into the half f- (7) Internal stresses
plane (f=£ + i t]) is of the form: (8) Displacement of tooth symmetry axis
With the objective of rendering the whole process
z=W(r)=cr+2\ «* (3) transparent for a gear designer, the first step is to select the
type of computer display for his session. This operation bloc
where c, ak, bk are real coefficients. Hence, by defining a cur- allows the user to choose between a numerical or a graphical
mode. One can modify his selection at all times during the
vilinear coordinate t along the half-plane (£ = t & rj = 0) equa- analysis of the tooth.
tion (3) represents the equation of the bell profile. Applying
relation (3), one may write the potentials for a gear tooth as: The second step begins the actual tooth generation process.
The user must enter the technical characteristics (pressure
angle, number of teeth, tip radius of generating rack, offset,
etc.) of the gear and also, parameters of both mating gears.
h m i>'('bk) The analyst must enter these technical parameters from a
(4)
2TT •ibk s>'(ibk) single data file or directly from the keyboard. Once the input
We-if) a(f) is complete, the user selects the number of points P that will
*<f) = 2TT log(f-fo)- define the tooth profile (60<P<100) and the actual tooth
profile can be visualized as shown in the first display of Fig. 2.
Then, the translation axis coefficient must be calculated. The
given gear tooth profile is actually defined from the addendum
Ck 4>'{-ibk) circle to the trochoid's tangency point with the dedendum cir-
(5) cle by a number of points. The coordinates of these points are
& $+ibk <a'{-ibk)
taken with respect to some initial axes, Oxy. Because of sym-
where the calculations of <j>' (ibk) = Xk — i Yk andmetry, there is no ambiguity for the .y-axis. However, the x-
<t>' (—ibk)=Xk + i Yk are performed in [17]. Substitution of axis has to be translated to be compatible with the conformal
equations (4) and (5) into Kolossov formulas [20] would yield mapping (y—0@*—oo). This portion of the program provides
GENERATION
_^ MAPPING
TRANSFORMATION 2.25-
" N\ TOOTH
1.75-
\
COEFFICIENTS 1.25-
\
0.75-
1
IDENTIFICATION 0.25-
COMPLEX
POTENTIAL
y w
STRESSES 2.25-
CRITICAL
SECTION
COMPRESSION
1.75-
TENSION
DISPLACEMENTS 1.25-
0.75-
CALCULATIONS 0.25-
the analyst with the translation coefficient via the cross-hair as with the 30 teeth gear parameters). The search for more exact
shown in Fig. 3(a). parameters is then performed interactively or numerically. In
The fourth step deals with the conformal mapping transfor- the first case, an experienced analyst can converge towards a
mations. Two options are available for this operation bloc, an good profile rapidly (1-2-3 trials) as visualized in Fig. 3(b).
interactive mode or a numerical mode. In either case, one Then, from this latest profile, a numerical search can be con-
must submit tentative initial coefficients. A table of conformal ducted. It can be stopped at any stage and the current approx-
mapping parameters is available which can be used as starting imate profile examined and compared visually with the exact
values for a new profile (e.g., for a 40 teeth gear, one can start one as shown in the second display of Fig. 2.
Journal of Mechanisms, Transmissions, and Automation in Design JUNE 1989, Vol. 111/299
Y ,
CONFORMAL
TOOTH MAPPING
2.00- \
\
1.50-
1.00-
0.50-
0.00-
YOUNG'S MODULUS=0.3»10 f
POISSON'S RATIO = 0 . 3 * 1 0 °
(10 EXP - 7
Y\
i ?. .. . W
2.25 -
\
1.75 -
\
1.25 -
\
0.75 -
I
0.25 -
V p»
-1.25 -0.75 -0.25 0.25 0.75 1.25 X -1.25 -0.75 -0.25 0.25 0.75
The use of an optimization algorithm is important during (via the cross-hair) a unit force per unit width on the
this reconstruction of the tooth since it will provide the best transformed tooth. This is depicted in the third display of Fig.
coefficients minimizing the distance between the two profiles 2. It is also possible to include a friction coefficient which in-
and, hence, realize a perfect mathematical model of the tooth fluences the angle of application of the unit load. Then, super-
under scrutiny. Powell's algorithm [24] was exploited for this ficial stresses can be visualized as reproduced in the fourth
minimization process. Since the distance between the two pro- display of Fig. 2. Coggins algorithm [25] is employed to find
files must be minimized, the distance u can be represented by a the maximum stress on the tooth as well as the curvilinear
Least-Square function of the type: coordinate t associated with this maximum value which will fix
the critical section. The algorithm starts once the stresses are
1 m known at the surface of the tooth. The stress calculation is
— EWJ(XJ-XJ)2< (6) realized in the following manner: two points are selected along
the tooth profile, ?start and rend, as shown in Fig. 5 in between
where Xj is the value of the profile while Xj is the value yielded which stresses are desired. After specifying the number of
by the transformation. The constant m represents the number stress points Npt along the profile, the step At is calculated as
of points selected along the profile and Wj is a weighing term follows:
which emphasizes the importance of specific sections of the
*end *star
profile (Fig. 4). Note that since the stress is high in the concave At = (7)
(lower) section of the tooth, a special effort to fit the curve in Np,-1
this region is attempted by introducing a weighing factor Wj The searching step is then defined as At\ = At/20 with a con-
equal to 2. vergence criteria of e = l x l 0 ~ 5 . Hence, this option also
Once the tooth profile is well represented, a concentrated locates automatically the critical section of the tooth.
force can be applied on the tooth and the stresses at the sur- Finally, the stress profile in the critical cross-section (or any
face can be calculated. Note that the conformal mapping has cross-section selected with the cross-hair) and the
no effect on the force itself but for its point of application f0. displacements from the tooth symmetry axis can be deter-
The force option allows to position numerically or graphically mined and traced directly on the screen as pictured in Figs.
- ?} C.P.M. -
4.0
•max
3.0 - 2) F.E.M. //A "
2.0
I 1.0 -
•w W
</UJ>
w^-W^-^^
rfp^-
CL
m -1.0
-
\
((f
I
-2.0
CxY
"
I
-3.0
'mm.
I _
-4.0
-5.0
~
Fig. 4 Importance of specific sections of the profile
.
-6.0
.2 .4 .6 .8 1.0 in.
GENERATION TYPES
PARAMETERS
1 2 3
PRESSURE
20° 20° 25 =
ANGLE
H O 1.25
DEDENDUM
P P p
Journal of Mechanisms, Transmissions, and Automation in Design JUNE 1989, Vol. 111/301
Table 2 Maximum tensile stress for a type 2 profile .45
AGMA (graphic method)
.35-
First, let us compare the maximum tensile stress <rmax for a NUMBER OF T E E T H ( N )
type 2 tooth with 20 teeth as calculated by MPCGRAPH with Fig. 7(a) Load applied at tip of tooth (profile #1)
that of Rubenchik [11] using the Boundary Equations In-
tegrals (which can be considered as an exact solution [11]) and A AGMA (graphic method)
that of earlier publication by Cardou and Tordion [17] using A AGMA (numerical method)
the CPM. For the case of a unit load per unit thickness applied D MPCGRAPH
on the addendum circle of the tooth (? = 0.004, /3= —150.5 3
deg) with W= 175 N/m (1 lb/in) and P=39.4 m" 1 (1 in." 1 ), rr
.6
WtP
(8)
where W, is the tangential load between gears, F is the face 20 40 60 80 100 120 140
width and / , is a tension geometry factor including stress con- NUMBER OF T E E T H ( N )
centration for a given load position. Since W,/F= W cos </>, Fig. 7(B) Load applied at highest point of single tooth contact (profile
equation (8) can be rewritten as: #1)
C O S <j>
J,= MPCGRAPH is always smaller than AGMA's estimates.
(9)
However, the spread between AGMA and MPCGRAPH fac-
This geometry factor J, will be our comparison criteria be- tors is small for a low number of teeth (N< 30) which increases
tween the AGMA methods (numerical and graphical) and the for a higher number of teeth (N> 60) to a maximum of 15 per-
Complex Potential Methods. cent. The same trend can be observed from Fig. 1(b) where in
The geometry factor J, is calculated by MPCGRAPH for a this case a 20 percent difference can be registered.
force W= 175 N/m (1 lb/in) applied to the addendum circle Let us now establish comparisons shown in Fig. 8(a) and
and for a force applied to the highest simple contact point. 8(b) for a type 2 profile between the factors Jt obtained from
These factors will be compared with the AGMA factors as well MPCGRAPH, AGMA's numerical method as well as Com-
as the factors obtained by different authors using the CPM: plex Potential Methods applied in earlier work. Again, an ex-
Cardou-Tordion [18] and Baronet-Tordion [16]. To identify cellent fit is attained by using MPCGRAPH where the
the AGMA geometry factors, the graphical sheets 226.01 [27] minimization coefficient u fluctuates between 10""6 and 10~"7.
are used. In the case where the type of tooth (type 2 and 3 in A better, faster and more consistent profile is generated by
our case) do not appear in these sheets, the AGMA numerical MPCGRAPH compared with all the previous complex poten-
method will be exploited [28]. tial algorithms. The variation between the methods increases
Let us examine the type 1 tooth where the geometry factors as the number of teeth becomes larger and the complex poten-
J, provided by MPCGRAPH are compared with the factors tial results are still lower than those suggested by AGMA.
obtained from the AGMA graphical and numerical methods. Finally, one can repeat the process for a type 3 tooth and
These results are plotted in Figs. 7(a) and 1(b). Note that for observe the same trends as before in Figs. 9(a) and 9(b).
both types of load the AGMA numerical method Note that in previous works, a perfect fit was difficult to
overestimates the / , factors by approximately 3 percent over achieve for this type of tooth. Nevertheless, MPCGRAPH
the AGMA graphical method. Hence, this spread should be modeled precisely the tooth with a u coefficient of 10 ~5. Even-
kept in mind when examining the other types of tooth though MPCGRAPH provided a better profile, a small
analysis. The modeling of the type 1 teeth provided an ac- variance of 3 percent was obtained between MPCGRAPH and
curate representation with a u value varying between 10 ~5 and Cardou-Tordion results.
10~6. Also, note that the tooth generated with N= 18 had a The increase in precision of the u function and the visualiza-
curved flank profile while that ofN= 150 had a quasi-straight tion process of the algorithm allow us to confirm the efficien-
flank profile. cy of the Least-Square formula (6) as our objective function.
Regarding the variation between the results obtained from Now, the top or bottom (fillet) portion of the tooth profile is
MPCGRAPH and AGMA for a load applied on the adden- just as well represented with this new computerized algorithm.
dum circle, it can be observed that the factor J, calculated by However, it appears that the conformal mapping has more
§ .«•
LU
° .225
.2 —I— ~~I
0 60 80 100 120 40 60 80
80 100 120
NUMBER OF T E E T H ( N ) NUMBER OF T E E T H ( N )
Fig. 8(a) Load applied at tip of tooth (profile #2) Fig. 9(a) Load applied at tip of tooth (profile #3)
.55-
O 45-
i
O 2
LU O
a
~i i i i i i i i i i — i — i — i — i — r
0 ^o 40 60 80 100 120 140 160
~~1—
NUMBER OF T E E T H ( N ) 20 40 60 80 100 120
Fig. 8(b) Load applied at highest point of single tooth contact (profile NUMBER OF T E E T H ( N )
#2) Fig. 9(b) Load applied at highest point of single tooth contact (profile
#3)
Acknowledgments
. 1 5 - ] — r — | — | — | — | — , — | — r — | — | — | — , — | — | — | — The authors wish to acknowledge the contribution of G. V.
-40 -30 -20 -10 0 10 20 30 40
Tordion for his initial guidance, A. Gakwaya for his expertise
FRICTION COEFFICIENT ( |i X Mf2) in the field of FEM and BEM, and the assistance of the
Fig. 10 Influence of friction coefficients on geometric factors Natural Sciences and Engineering Research Council of
Canada which supported this work.
Papers in all areas of Mechanical Design Theory and Methodology are invited. Suggested topic areas are:
• Design Representations • Cognitive Design Theories
and Languages • Design for Manufacturing
8
Design Management • Constraint Understanding and
8
Design Work Spaces Management
9 8
Design Histories Analytical Decision Making for Design
8 8
Concurrent Design Design Tool Development
8 8
Life-Cycle Design Design Education
8 8
Design for Quality Computational Methods of Design
8
• Prescriptive Theories Integration of Design Education with
and Models of Design Research and Practice
Contributions in the form of full-length papers should be submitted directly to the Program Chairman. Accepted papers will
be published in bound conference proceedings by the ASME. Papers of special note will be further reviewed after the con-
ference for publication in ASME transactions.
Conference Chairman Program Chairman
Dr. David G. Ullman Dr. James R. Rinderle
Department of Mechanical Mechanical Engineering
Engineering Department
Oregon State University Carnegie Mellon University
Corvallis, OR 97331 Pittsburgh, PA 15213
ullman @cs.orst.edu rinderle@edrc.cmu.edu
(503) 754-2336 (412) 268-3677
Deadline Dates for Technical Papers
Contributions (full-length papers) 4 copies to be submitted
to Program Chairman no later than March 1, 1990
(Non-North American submissions Feb. 1, 1990)
Review process completed, notification and materials mailed May 1, 1990
Final paper(s) on author-prepared materials due to ASME June 1, 1990
Journal of Mechanisms, Transmissions, and Automation in Design JUNE 1989, Vol. 111/305