Optimized Bearing Ellipses of

Hypoid Gears1
J. Achtmann
For given machine tool settings of a universal hypoid gear generator, the tooth contact
G. Bär2 patterns are computed for the coast and drive side of a hypoid gear drive. Each contact
pattern is replaced by a determined tooth-bearing ellipse. The position, shape, and incli-
Dresden University of Technology, nation of each bearing ellipse is calculated. By the help of these data, an influence
Institute of Geometry, function is designed that describes the influence of supplemental kinematic flank correc-
D-01062 Dresden, tion motions (modified motions) on the gear-tooth contact. Examples show the influence of
Germany helical motion and modified roll. An evaluation function permits the calculation of modi-
fied motions which improve the tooth contact either at coast and drive side simulta-
neously, or only at one of the sides. For a given pair of start-bearing ellipses at coast and
drive side, and for given importance weights to the sides, we describe how modified
motions can be computed that best fit a given target pair of bearing ellipses.
关DOI: 10.1115/1.1625403兴

1 Introduction computation of the path of contact and transmission error of a

Computerized simulation of meshing and contact of hypoid
gear drives is state of art in industry and research. Corresponding
computer programs model the manufacture of the pinion and gear
关1兴, and also the meshing of the gear set 关2–7兴. Typical results of
such computations are the Ease-Off function, contact lines, path of
contact, contact pattern, and the transmission function. This data
describe the contact behavior of the gear set in dependence on the
machine tool settings. To improve the contact behavior of a gear
set, experts can use such Tooth Contact Analysis 共TCA兲 software
with trial design. Naturally, latest research contributions tackle the
challenging problem of gear optimization by the help of math-
ematical methods and corresponding software.
Gosselin 关8兴 uses the Surface Match computer algorithm to find
the machine settings producing a new theoretical tooth surface
equivalent to that of a measured master target surface, including
profile deviations, such that Corrective Machine Settings 共for in-
stance proportional changes兲 can be applied in reference to this
new theoretical surface generating requested profile deviations.
Litvin et al. 关5兴 propose an integrated TCA program for simu-
lation of meshing and contact of gear drives that enables to deter-
mine transmission errors and shift of bearing contact for mis-
aligned gear drives. To facilitate a gearing experts’ work, this
program allows integrated operations such as the development of
required algorithms, database handling, and design correction
from database.
Müller 关6兴 considers the designing of the Ease-Off Topography
to be the most important task in computing spiral bevel gears. An
algorithm is mentioned that changes the relative motion between
gear work and tool to achieve the desired Ease-Off by entering
numerical values of the crowning, pressure angle difference and
longitudinal twist.
Stadtfeld 关9兴 also describes a design method for the Ease-Off
Topography, namely the Universal Motion Concept. It means an
extension of the existing calculation methods for gearing by an
eight kinematic correction mechanism. Each of its parameters is
formulated by a fourth-order polynomial. A series of calculation
steps is mentioned that allow the calculation of updated basic
machine settings resulting in a desired Ease-Off.
Vogel et al. 关10兴 show a different approach to TCA and its
optimization using Automatic Differentiation. It allows the direct
1
The work was supported by Deutsche Forschungsgemeinschaft, Bonn.
2
Corresponding author: baer@math.tu-dresden.de
Contributed by the Power Transmission and Gearing Committee for publication in
the JOURNAL OF MECHANICAL DESIGN. Manuscript received July 2002; revised
April 2003. Associate Editor: A Kahraman.
gear set. At a point of contact between meshing gear tooth flanks,
contact properties of second order, for instance Litvin’s contact
ellipses 关11兴, are computable without computing the contacting
flanks first.
In this paper a numerical method is developed and tested that is
also based on the study of the Ease-Off. We use the software
BECAL 关12兴 developed at the Dresden University of Technology.
It permits the simulation of any hypoid gear cutting process by a
six-axis CNC hypoid generator and also the TCA within accept-
able calculation time. At the same time the 共global兲 Ease-Off func-
tion is calculated, a function that quantifies distances between
meshing gear-tooth flanks 关2,6,7,9兴 over a whole cycle of mesh-
ing. BECAL computes the appropriate Ease-Off values to a given
grid of flank points. Exploiting these data, the tooth contact pat-
terns are computed for the coast and drive side of a hypoid gear
drive. We developed a fitting procedure in order to replace each
contact pattern by its so-called bearing ellipse. Then, the position,
shape, and inclination of each bearing ellipse is available. These
values permit the gearing expert to evaluate the tooth contact
more precise than using the plot of the contact pattern alone.
Commonly, at this point an interactive design process follows.
The expert tries to improve the tooth contact either at coast and
drive side simultaneously, or only at one of the sides. For this
purpose, graphical and numerical output is used to evaluate con-
tact and transmission properties meanwhile altering several ma-
chine settings adequately. Our aim is to overcome this trial design
by a first step in direction of unerring computational optimization.
2 Tooth-contact Modeling by Bearing Ellipses
In order to visualize gear tooth contact, the spatial points of the
gear flank are often transformed to a planar coordinate system by
radial projection, i.e. each flank point is rotated about the gear axis
to its radial projection in a fixed picture plane which contains the
gear axis. Therefore, the position of the radial projection of a flank
point is determined by its radius ␰ and its width ␩. See Fig. 1.
Usually, the path of contact and contact lines are then represented
and studied 关7,11,12兴. Furthermore, the information on distances
between meshing gear flanks is given by the ease-off function
共EOF兲. This function of the two variables 共␰,␩兲 describes how far
every point of a gear flank is at least located from the opposite
flank when the flanks have finished a cycle of meshing. There are
many ways to visualize the EOF. Here, in Fig. 1 a gray-scale plot
is given. The density of color encodes the ease-off ␧ s at each gear
flank point in its radial projection.

Journal of Mechanical Design Copyright © 2003 by ASME DECEMBER 2003, Vol. 125 Õ 739

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/

 The contact point P ell is the center of the bearing ellipse. Com-
paring this point with the calculation point P F , the face width B,
and the profile depth H of the gear, we define

⌬l
the width de v iation y⫽
B
⌬p
and pro f ile de v iation z⫽ . (2.3)
H

Shape and inclination of the bearing ellipse with respect to the

face width B and profile depth H are given by the

Fig. 1 Gray-scale plot of an EOF a ell b ell

bearing width a⫽ , bearing height b⫽ ,
B H
and bearing rotation ␥ . (2.4)

In practice a marking paste of thickness ␴ t is used to visualize

the tooth contact pattern at both the coast side and the drive side
of a gearing. A point 共␰,␩兲 will belong to such a tooth contact
pattern if its corresponding EOF-value is at most ␴ t during the run
simulation. In mathematical terms the EOF is given by
␧ s ⫽␧ s 共 ␰ , ␩ 兲 (2.1)
and a contact pattern by
K s ⫽ 兵 共 ␰ , ␩ 兲 :␧ s 共 ␰ , ␩ 兲 ⭐ ␴ t 其 ; s苸 兵 c,d 其 . (2.2)
In order to distinguish the coast and drive side, the subscript s is
to be replaced by c and d, respectively.
Exploiting these data, by a fitting procedure, the best approxi-
mating ellipse for a given contact pattern K s is computed. Such an
ellipse is referred to as bearing ellipse3 for the marking paste
thickness ␴ t . The advantage of this concept is obvious. We can
use five characteristic parameters of such a bearing ellipse in order
to quantify gear tooth contact 共see Fig. 2兲.
3
It is an global feature in contrast to Litvin’s contact ellipse 关11兴 which is a local
feature defined by curvatures of gear tooth surfaces.
The set E⫽R⫻R⫻R⫹ ⫻R⫹ ⫻(⫺ ␲ /2 , ␲ /2兴 傺R5 contains all pos-
sible bearing ellipses because each one is determined by a 5-tuple
(y, z, a, b, ␥兲.
Next, the determination of a bearing ellipse is outlined 共for
mathematical details see 关13兴兲:
According to Eqs. 共2.1兲 and 共2.2兲 a contact pattern K with the
marking paste thickness ␴ t is given by K⫽ 兵 ( ␰ , ␩ ):␧( ␰ , ␩ )⭐ ␴ t 其 .
Here, subscripts s, c, and d are suppressed because coast and
drive side contact are treated analogously. At the computer, the set
K is represented by a discrete set K̃ of points ( ␰ i , ␩ i ) where i
belongs to some index set I. To find a best fitting ellipse to a set
K̃ of points in the plane we use two calculation steps:
1. Determination of the set K con v of vertices of the convex hull
of the set K̃, using inner normal vectors.
2. Approximation of K con v by an ellipse that is determined by
eigenvectors in the following way.
A second order curve—especially an ellipse—is described by a
symmetric matrix

Fig. 2 Parameters of bearing ellipses at coast and drive side of the gear

740 Õ Vol. 125, DECEMBER 2003 Transactions of the ASME

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/

 A⫽ 冉 a 00
.
.
a 01
a 11
.
a 02
a 12
a 22
冊 with the normalizing condition
H 共 t 兲 ⫽H 0 ⫹H 1 t⫹H 2 t 2 ⫹ . . . ⫹H 7 t 7 .
Modified roll is a kinematic correction of the ratio of roll R 0
which is defined by
R 共 t 兲 ⫽R 0 ⫹R 1 t⫹R 2 t 2 ⫹ . . . ⫹R 7 t 7
a 211⫹2a 212⫹a 222⫽1.
where t is the roll angle of the cradle in both cases.
All points x̃⫽(1,␰ , ␩ ) T that fulfil the equation There are eight modified motions 共named in Fig. 1兲 which are
x̃ T Ax̃⫽0 (2.5) analogously defined 关7兴. Generally, a modified motion is given by
the coefficients of the describing polynomial. For simplification,
form this second order curve. In order to determine a best fit, it is we consider only the coefficients of the linear terms as optimiza-
necessary to define a metric concept. The Euclidean metric is tion parameters because they dominate the modification of flanks.
useless here since the perpendicular distance from a point to a In the above example, the linear coefficients are
second order curve is not uniquely determined in general. There-
fore, the power p of a point x̃ with respect to a second order curve x 1 ªH 1 and x 2 ªR 1 .
共2.5兲 is used: Generally, these linear coefficients x 1 , . . . ,x n are called modified
p 共 x̃ 兲 ⫽x̃ T Ax̃⫽x̃ T ã⫹x̄ T ā (2.6) motion parameters and are arranged together in the so-called

冉冊 冉 冊 冉 冊
‘‘mm-vector’’ xគ ⫽(x 1 , . . . ,x n ). With this argument we define the
1 a 00 x 21 influence function
where x̃⫽ x 1 , ã⫽ 2a 01 , x̄⫽ x 22 , and ␨ :D ␨ 傺Rn →E⫻E, (3.1)
x2 2a 02 &x 1 x 2

冉 冊
a 11
ā⫽ a 22 .
&a 12
The set S⫽ 兵 (ã,ā)苸R ⫻R :ā ⫽1 其 contains all second order
3 3 2
curves. For a point set K con v ⫽ 兵 x̃ 1 ,x̃ 2 , . . . ,x̃ N 其 傺R2 the best
fitting ellipse is determined by the solution of the optimization
problem
f 共 ã,ā 兲 ⫽ã T X 00ã⫹ā T X 11ā⫹2ã T X 01ā→min!
with the abbreviations
N N
X 00⫽ 兺
i⫽1
x̃ i x̃ iT , X 11⫽ 兺
i⫽1
x̄ i x̄ iT , and
The necessary condition
⳵ to be continuous in a vicinity of the starting vector. With the
f 共 ã,ā 兲 ⫽2X 00ã⫹2X 01ā⫽0
⳵ ã
allows to eliminate ã. So, the extremal problem ā T Zā→min! with
the matrix Z⫽X 11⫺X T01X ⫺100 X 01 remains. Because Z is symmetric
and positive semi-definite, the least eigenvalue of Z exists. A cor-
responding eigenvector ā * of Z along with the solution ã * of Eq.
共2.9兲 determines the optimal ellipse.
3 Modeling the Influence of Modified Motions on
Tooth Contact
The software BECAL 关12兴 permits the simulation of any hypoid
gear cutting process by a six-axis CNC hypoid generator and also
the procedures of the Tooth Contact Analysis 共TCA兲 within ac-
ceptable calculation time. In order to improve the tooth contact of
a gear set, the current way of ‘‘optimization’’ is trial design, i.e.,
the machine tool settings for pinion and gear cutting are altered as
long as chosen features match most design specifications. How-
ever, considering the great number of machine tool parameters,
always uncertainty remains whether the real optimum was found.
To overcome this problem by computation, in this paper we de-
velop an optimization strategy using supplemental kinematic flank
correction motions 共simply called modified motions兲 as the vari-
ables of optimization.
Modified motions are commonly used in the pinion cutting pro-
cess on the six-axis CNC hypoid generator in order to modify or
correct the pinion tooth flanks with the result of improved contact
properties of the drive, as shown in 关6,7,9兴. For example, helical
motion is a kinematic correction of the work base setting H 0
which is defined by
xគ ⫽ 共 x 1 , . . . ,x n 兲 哫 共 E c ,E d 兲
⫽ 共共 y c ,z c ,a c ,b c , ␥ c 兲 , 共 y d ,z d ,a d ,b d , ␥ d 兲兲
which determines the pair of bearing ellipses (E c ,E d ) at the coast
and drive side of the considered tooth for each mm-vector in a
domain D ␨ 共according to Section 2兲.
The numerical determination of the function ␨ is challenging
because it depends on eight variables as found in practice. Our
aim is to interpolate this function from a set of base points, each
(2.7) one calculated by two full runs of the BECAL-TCA program. In
principle, the number of these base points is indefinite but each
base point is paid by considerable calculation expense. To reduce
N these difficulty, for practical reasons of implementation we con-
X 01⫽ 兺 x̃ x̄
i⫽1
T
i i . fine our problem on two modified motions x 1 , x 2 (n⫽2).
In order to determine a set of base points an estimation of the
(2.8) domain D ␨ is needed. For a basic gear drive design a starting
vector xគ 0 ⫽(x 01 ,x 02 ) of modified motions is known. Often, the cal-
culation starts with xគ 0 ⫽(0,0). The influence function is assumed
(2.9)
formulation
D ␨ ⫽I 1 ⫻I 2 , I k ⫽ 关 x ak ,x bk 兴 , k⫽1,2, (3.2)
the end points for each interval I k are looked for. Sometimes the
end points are known by the gearing expert’s practical experience.
Otherwise, beginning with the starting vector xគ 0 of modified mo-
tions, step-by-step the influence function is calculated as long as
there are computable function values. This procedure can take a
lot of computing time. On the other hand, the computed function
values can be stored and later used for interpolation. After this
procedure, the domain D ␨ is not characterized by all computable
bearing ellipses 共coast and drive side兲 but estimated as given by
Eq. 共3.2兲.
To design the influence function ␨ we follow a method shown
by Nielson 关14兴 and refined by Pottmann 关15兴, a scattered-data
interpolation method for functions depending on two variables.
The essential element of the refinement is that some interpolation
base points can have unknown function values. Since the method
is based upon a curvature minimizing condition the ‘‘no-value-
points’’ have to be within the convex hull of ‘‘real-value-points.’’
The acquirement of base points is carried out by BECAL which
handles a list of random-generated mm-vectors equally spread on
the rectangular domain D ␨ . Some of the listed mm-vectors result
in useful bearing ellipse pairs and some will not. First, according
to 关14兴, all points without acceptable ellipses are canceled. Sec-
ond, as well all points lying within a convex hull of points with
valid ellipses are erased according to 关15兴. A triangulation of the
cleared mm-vector set follows, identifying it with a point set in
the Euclidean plane. We recommend to use the Delauney triangu-

Journal of Mechanical Design DECEMBER 2003, Vol. 125 Õ 741

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/

 Table 1 Blank data and machine settings for the hypoid gear drive

Pinion Gear
Design Parameters convex concave convex concave
Blank data
Number of teeth 11 34
Mean normal module 5.636
Shaft angle 共deg.兲 90.00
Hypoid offset 共mm兲 30.00
Mean spiral angle 共deg.兲 ⫺49.035 35.365
Pressure angle 共deg.兲 26.14482 18.85518 18.85518 26.14482
Pitch cone angle 共deg.兲 21.72249 67.70491
Face angle 共deg.兲 25.37165 69.01350
Root angle 共deg.兲 20.45007 63.95204
Addendum modification factor 0.48292 ⫺0.48292
共mm兲
Parameters and machine settings
Blade pressure angle 共deg.兲 29.00000 16.00000 29.00000 16.00000
Cutter tilt angle 共deg.兲 1.94273 2.34697 7.67192 7.67192
Cutter swivel angle 共deg.兲 4.41856 0.39604 69.33995 69.33995
Sliding base 共mm兲 0.00000 0.00000 0.00000 0.00000
Blank offset 共mm兲 ⫺3.36543 ⫺8.07882 ⫺17.10351 ⫺17.10351
Radial distance 共mm兲 111.16384 105.41210 105.17088 105.17088
Machine root angle 共deg.兲 ⫺108.45000 ⫺108.45000 ⫺153.95204 ⫺153.95204
Cradle offset 共mm兲 ⫺26.61200 ⫺26.65800 ⫺5.84804 ⫺5.84804
Tailstock 共mm兲 5.25000 ⫺7.06000 ⫺2.56629 ⫺2.56629
Cradle angle 共deg.兲 290.02899 288.91183 56.19975 56.19975
Horizontal shift 共mm兲 0.00000 0.00000 0.00000 0.00000
Ratio of roll 0.29955 0.32031 0.97155 0.97155

lation in order to avoid obtuse angled triangles. Furthermore it A iyj w i j

well harmonizes with the tendency of both methods to spread the
function curvature as equally as possible. This special triangula-
兺
i j苸K i L i3j
共 3 共 S j ⫺S i 兲 ⫺2A ixj S ix ⫺2A iyj S iy ⫹A xji S xj ⫹A yji S yj 兲 ⫽0,

tion is dual to the Voronoi diagram of this point set which can be
found by a standard divide-and-conquer algorithm. i⫽1 . . . M ,
The main result reads: Let V N ⫽ 兵 v 1 . . . v N 其 ⫽ 兵 (x i ,y i ):i
⫽1 . . . N 其 傺U傺R2 be a set of points whose function values wij
S(x i ,y i ) are known for the interpolating function S:U→R sought 兺 L i3j
共 2 共 S j ⫺S i 兲 ⫺A ixj S ix ⫺A iyj S iy ⫹A xji S xj ⫹A yji S yj 兲 ⫽0,
for.4 The points of the set V M ⫽ 兵 v N⫹1 . . . v M 其 傺U are added with i j苸K i

unknown function values. The edge indices of the triangulation T

are collected in the set K⫽ 兵 i j:e i j is an edge 其 . Each edge e i j i⫽N⫹1 . . . M ,
with arc length l i j is mapped by the function S on a space curve
c i j . The reduced function S T ⫽S 兩 T is to find such that the where
weighted total curvature

冕冉 冊
A ixj ⫽ 共 1,0兲 T 共 v j ⫺ v i 兲 , A iyj ⫽ 共 0,1兲 T 共 v j ⫺ v i 兲 , and
d 2c i j 2
␬共 ST兲⫽ 兺wK
ij
ci j dl i j
dl i j , w i j ⬎0 (3.3)
K i ⫽ 兵 i j 兩 e i j is an edge with end v i 其 .
becomes minimized. Here, w i j denotes the weight of the curve
c i j . The minimization is carried out with all continuous curves L i j denotes the length of e i j .
defined on T and match S restricted to V N . As shown in 关15兴, it An explicit specification of S is only possible within one tri-
follows: angle of T. Hence we transform such a triangle 关v i 兴 ⫽ 关 (x i ,y i ):i
If V N will contain at least three non-collinear points, the mini- ⫽1,2,3 兴 into barycentrical coordinates b i ⫽b i (x,y) which are de-
mization condition with Eq. 共3.3兲 determines S T uniquely. The termined by the conditions
partial curves c i j (l i j ) are cubic and are fixed by their start and end
points and their derivatives at these points into the direction of x⫽b 1 x 1 ⫹b 2 x 2 ⫹b 3 x 3 ,
edge e i j .
For the 3M ⫺N unknowns y⫽b 1 y 1 ⫹b 2 y 2 ⫹b 3 y 3 , (3.6)
⳵S ⳵S
S ix ⫽ 共 v 兲, S iy ⫽ 共 v 兲 共 i⫽1 . . . M 兲 and 1⫽b 1 ⫹b 2 ⫹b 3 .
⳵x i ⳵y i
Thus, we can define the function
S j ⫽S T 共 v j 兲 共 j⫽N⫹1 . . . M 兲 (3.4)
the following 3M ⫺N linear equations are provided:
A ixj w i j
S [ v i ] 共 x,y 兲 ⫽ 兺 S T 共 v i 兲关 b i2 共 3⫺2b i 兲 ⫹6wb i 共 b k a i j ⫹b j a ik 兲兴

兺
(i, j,k)苸J
共 3 共 S j ⫺S i 兲 ⫺2A ixj S ix ⫺2A iyj S iy ⫹A xji S xj ⫹A yji S yj 兲 ⫽0,
i j苸K i L i3j ⫹S ⬘k 共 v i 兲关 b i2 b k ⫹wb i 共 3b k a i j ⫹b j ⫺b k 兲兴
(3.5)
⫹S ⬘j 共 v i 兲关 b i2 b j ⫹wb i 共 3b j a ik ⫹b k ⫺b j 兲兴 (3.7)
4
In our application S is individually sought for each considered component of the
influence function ␨. with

742 Õ Vol. 125, DECEMBER 2003 Transactions of the ASME

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/

 Table 2 Representation of components of the influence function „coast side…

⳵S ⳵S 4 Numerical Results for the Influence Function

S ⬘j 共 v i 兲 ⫽ 共 x j ⫺x i 兲 共 v 兲 ⫹ 共 y j ⫺y i 兲 共 v 兲,
⳵x i ⳵y i The impact of the considered modified motion parameters on
the several components of the influence function is differently
b 1b 2b 3 strong. In the following examples, helical motion x 1 and modified
w⫽ ,
b 1 b 2 ⫹b 1 b 3 ⫹b 2 b 3 roll x 2 are chosen and studied within the variation intervals
⫺0.7⭐x 1 ⭐0.7 and ⫺0.07⭐x 2 ⭐0.07.
L 2jk ⫹L ik
2
⫺L i2j The influence function is computed for a bevel gear drive and
ai j⫽ 2 , and b 1 ,b 2 ,b 3 ⭓0.
2L ik for a hypoid gear drive as usual in automotive industry. The de-
sign parameters and machine settings of the hypoid gear drive are
This interpolation method was applied for each of the ten compo- given in Table 1. The blank data and the essential machine data
nents y c ,z c ,a c ,b c , ␥ c ,y d ,z d ,a d ,b d , ␥ d of the influence function for the bevel gear drive are equivalent. For the bearing ellipse of
␨ according to Eq. 共3.1兲. So, for instance the bearing width is now the two gear drives on the coast side, the computed dimensions
an interpolated function, a c ⫽a c (x 1 ,x 2 ). To this end, the whole like bearing width a c , bearing height b c , and bearing rotation ␥ c
influence function are shown in Table 2 in dependance of x 1 and x 2 . For the case of
␨ 共 x 1 ,x 2 兲 ⫽ 共 E c ,E d 兲 ⫽ 共共 y c ,z c ,a c ,b c , ␥ c 兲 , 共 y d ,z d ,a d ,b d , ␥ d 兲兲 the bevel gear we observe a relative maximum of the bearing
width nearby the point (x 1 ,x 2 )⫽(0.2,⫺0.03). For the case of the
is determined. hypoid gear, at least two relative maxima of the bearing width

Journal of Mechanical Design DECEMBER 2003, Vol. 125 Õ 743

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/

 Fig. 3 User Interface to control fitting and optimization
occur in the vicinity of the points (⫺0.3, 0.06兲 and 共0.2, 0.05兲. In
general, the influence of helical motion and modified roll is highly
different for bevel and hypoid gear drives, respectively. Note that
the influence function is not defined at the plane regions on the
front left side for the case of the bevel gear drive.
5 Determination of Optimized Bearing Ellipses
In connection with Eq. 共3.2兲 a starting vector xគ 0 ⫽(x 01 ,x 02 ) of
modified motions is known. Let (E c ,E d ) be the starting pair of
bearing ellipses that belongs to this vector xគ 0 . Evaluating a
graphical representation of (E c ,E d ) corresponding to Fig. 2 or 3,
the gearing expert is able to predict position, shape, and inclina-
tion of an improved pair of bearing ellipses. Therefore, the com-
putational problem is now to find an optimal mm-vector xគ * which
determines a pair of bearing ellipses (E * c ,E *
d ) best fitting a given
pair (E tc ,E td ) of bearing ellipses called the target pair.
By this formulation we partially avoid that our problem is not
solvable. Multiple solutions are neglected. Nevertheless we have
mathematically to answer the question which of two bearing el-
lipse pairs is to prefer in above terms.
For this reason we design an evaluation function
␮ :E⫻E→ 关 0,1 兴 ,
兩 Ē c 艚Ē tc 兩
共 E c ,E d 兲 哫 ␮ 共 E c ,E d 兲 ⫽w c ⫹ 共 1⫺w c 兲
兩 Ē c 艛Ē tc 兩
that quantifies the deviation of a bearing ellipse pair (E c ,E d )
from a target pair (E tc ,E td ) by the value ␮ (E c ,E d ). Thereby, the
following notation is used:
Ē s denotes the set of all inner grid points of a bearing ellipse E s
with s苸 兵 c,d 其 . As usual, intersection and union of sets is de-
scribed by 艚 and 艛, respectively. The number of points 共cardinal
number兲 of a point set X is given by 兩 X 兩 .
744 Õ Vol. 125, DECEMBER 2003
Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/
Finally, in order to calculate the value ␮ (E c ,E d ), a weight
w c with 0⭐w c ⭐1 is used. This real number w c weights the im-
portance of the coast side contact in relation to the drive side
contact. For instance, the weight w c ⫽0.8⫽80% means that coast
side contact is given 80% importance and therefore, drive side
contact is given (1⫺0.8)⫽20% of importance in the course of
optimization.
Note that by w c ⫽0 only drive side contact is treated and by
w c ⫽1 only the coast side is evaluated. Therefore, the given evalu-
ation function allows to model simultaneous gear cutting tech-
niques as well as single flank cutting.
By the help of the defined functions, our problem reads as
follows:
Given a target pair (E tc ,E td )⫽((y tc ,z tc ,a tc ,b tc , ␥ tc ),
(y d ,z d ,a d ,b d , ␥ d )) of bearing ellipses the optimal mm-vector xគ *
t t t t t
is the solution of the optimization problem
xគ * ⫽ max ␮ 共 ␨ 共 xគ 兲兲 . (5.2)
xគ 苸D ␨
6 Implementation
A Delphi program is used to test the above given methods. The
user interface of the program is shown in Fig. 3. Let us continue
the example of the hypoid gear drive 共Table 1兲 to explain an
optimization step. For instance, the calculation begins with the
start vector xគ 0 ⫽(0,0) for helical motion x 1 ⫽0 and modified roll
x 2 ⫽0.
兩 Ē d 艚Ē td 兩 Following the algorithm given in Section 2, the corresponding
start bearing ellipses (E c ,E d ) are computed. According to the
兩 Ē d 艛Ē td 兩
(5.1) definition 共2.3兲 and 共2.4兲 the width and profile deviation, the bear-
ing width, height, and rotation is calculated and printed on the left
and right hand side of the output screen 共Fig. 3兲 for the coast side
and drive side, respectively.
Now, the gearing expert evaluates the gear contact and formu-
lates a target ellipse pair (E tc ,E td ) by filling the target data win-
dows. The corresponding target ellipses are dotted shown in the
figure. Here, they have been chosen equally. Furthermore, a start
weight w c ⫽80% and a marking paste thickness 0.006 mm are
given. After clicking the start button the program performs a stan-
Transactions of the ASME
 Table 3 Dimensions of the optimized bearing ellipses References
关1兴 Fong, Z.-H., 2000, ‘‘Mathematical Model of Universal Hypoid Generator With
Measure Coast side Drive side
Supplemental Kinematic Flank Correction Motions,’’ ASME J. Mech. Des.,
Width deviation ⫺3% 6% 122, pp. 136 –142.
Profile deviation ⫺14% ⫺10% 关2兴 Bär, G., and Iotchev, V., 1999, ‘‘Accurate Tooth Contact Determination for
Bearing width 33% 33% Hypoid Bevel Gears using Automatic Differentiation,’’ 4th World Congress on
Bearing height 38% 33% Gearing and Power Transmission, M.C.I. Paris, Vol. 1, pp. 519–529.
Bearing rotation 9° ⫺22° 关3兴 Hünecke, C., 2001, ‘‘Manual ARCOTEC,’’ SU-Hurth-Modul GmbH, Chem-
nitz.
关4兴 Hutschenreiter, U., 1996, ‘‘A New Method for Bevel Gear Tooth Flank Com-
putation,’’ Proc. 2nd. Int. Workshop on Comput. Differentiation, Santa Fe
共N.M.兲, Feb. 12–14.
关5兴 Litvin, F. L., De Donno, M., Peng, A., Vorontsov, A., and Handschuh, R. F.,
dard non-convex optimization routine to solve the optimization 2000, ‘‘Integrated Computer Program for Simulation of Meshing and Contact
of Gear Drives,’’ Comput. Methods Appl. Mech. Eng., 181, pp. 71– 85.
problem 共5.2兲. Here we obtain the optimal mm-vector xគ * ⫽(0.4, 关6兴 Müller, H., 1999, ‘‘A Modular Approach Computing Spiral Bevel Gears and
⫺0.05). The data of the corresponding optimized bearing ellipses Curvic Couplings,’’ 4th World Congress on Gearing and Power Transmission,
(E *
c ,E *
d ) are given in Table 3. The graphical representation of
M.C.I. Paris, Vol. 1, pp. 531–541.
关7兴 Stadtfeld, J., 1995, Gleason—Bevel Gear Technology, The Gleason Works,
(E *
c ,E *
d ) is bold printed in Fig. 3. Rochester, N.Y.
This optimum is achieved with respect to function 共5.1兲. In 关8兴 Gosselin, C., 1999, ‘‘Corrective Machine Settings of Spiral-Bevel and Hypoid
order to come closer to a wanted result, the program can be started Gears With Profile Deviations,’’ 4th World Congress on Gearing and Power
once more introducing a new target pair and a new weight. Transmission, M.C.I. Paris, Vol. 1, pp. 543–554.
关9兴 Stadtfeld, J., 1999, ‘‘The Universal Motion Concept for Bevel Gear Produc-
tion,’’ 4th World Congress on Gearing and Power Transmission, M.C.I. Paris,
Vol. 1, pp. 597– 607.
关10兴 Vogel, O., Griewank, A., and Bär, G., 2002, ‘‘Direct Gear Tooth Analysis for
Hypoid Bevel Gears,’’ Comput. Methods Appl. Mech. Eng., 191, pp. 3965–
7 Conclusion 3982.
关11兴 Litvin, F. L., 1994, Gear Geometry and Applied Theory, Prentice Hall, Engle-
Based on the optimization method in the paper, for a given pair wood Cliffs, N.J.
of starting bearing ellipses at coast and drive side, and for given 关12兴 Baumann, V., 1995, ‘‘Programm BECAL,’’ FVA-Forschungsheft 429.
importance weights to the sides, two chosen modified motions can 关13兴 Geise, G., 1977, ‘‘Einige Ausgleichsprobleme und ihr Zusammenhang mit so-
be found which will produce a new pair of bearing ellipses. This genannten Übertragungsprinzipien,’’ Journal of Geometry, 9共1/2兲, pp. 57– 64.
关14兴 Nielson, G. M., 1983, ‘‘A Method for Interpolating Scattered Data Based Upon
pair is optimally fitting the given target pair of bearing ellipses. a Minimum Norm Network,’’ Math. Comput., 40共161兲, pp. 253–271.
Further work will include the transmission error into the evalua- 关15兴 Pottmann, H., 1992, ‘‘Interpolation on Surface Using Minimum Norm Net-
tion function 共5.1兲 in order to avoid its increase. works,’’ Computer Aided Geometric Design, 9, pp. 51– 67.

Journal of Mechanical Design DECEMBER 2003, Vol. 125 Õ 745

Downloaded From: http://mechanicaldesign.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jmdedb/27766/ on 02/14/2017 Terms of Use: http://www.asme.org/