Вы находитесь на странице: 1из 7

Design Against Tooth

Interior Fatigue Fracture

Magnus Mac!kAldener and Marten OlssoDi

Intredncnon TIFF has been observed in case hardened

10 a modern truck, the gear teeth are among idlers. A test series with idler wheels has been
the most stressed parts .. Failure of a tooth will run. Out of 75 specimens, 20 were classified as
damage the transmission severely. Throughout TIFF. Tooth root bending fatigue or contact
the years, gear design experience has beengained fatigue occurred in the other specimens.
and collected into standards such as DIN (Ref. 1) TIFF Mechanism
or AGMA (Ref. 2). Traditionally two types of in the interior of the tooth.
TIFF is initiated
failures are considered in gear design: tooth root Other known fatigue failures initiate at the sur-
bending fatigue, and contact fatigue. The face. A failure type called Zahnkopfbruch in
demands for llighter and more silent transmissions German. starting from spalling craters at the
have given birth to new failure types. One novel flank, has been reported by Shultz and Sauter
failure type, Tooth Interior Fatigue Fracture (Ref. 5). Hence Zahnkopfbruch and TIFF are not
(TIFF), has previously been described by the same type of failure, although they may
MackAldener and Olsson (Refs. 3 & 4) and is fur- appear similar. Alban (Reef. 6) described stress
ther explored in this paper .. ruptures, which have a fracture surface similar to
Observations of TWF TIFF. However, stress rupture is 001. a fatigue
TIFF is characterized by a failure, approximate- fracture since the crack is formed during the hard-
lymid-height on the tooth, which distinguishes it ening process.
from tooth root bending fatigue. Contact fatigue TIFF is the result of: 1) COlli stant residual ten-
andlor spall.ing craters are not a prerequisite at the sile stresses in the interior of the tooth due to case
flank of a TIFF. In Figures 1 and 2, photos and a hardening. and 2) alternating stresses due to the
schematic of a typical. TIFF can. be seen. idler usage of the gear wheel.
FE-model/or mesh cycle simfliaJiD.n. To ana-
lyze the crack initiation process in the gear, a 20.
FE-mesh was utilized, A plane strain 2D-mesh in
the normal plane (perpendicular to the flank) was
shown to give stress in a. cross-sec-
tion of the tooth that is virtually the same as that
of a 3D-mes'h. Hence a 2o.-mesh can be used for
stress analysis of a complete mesh cycle. The FE-
mesh is shown in Figure 3.
With a parametric model of t11eFE-mesh
Fig. I-A typical 'looth Interior Fatigue Fracture (TIFF) and a schematic of a
(developed in ANSYS, Ref. 7), an arb:itrary posi-
.- ---, I lion in the mesh cycle could be analyzed. The

gear calculation program LDP (Ref. 8) computed

the total force on one tooth as a function of posi-
tion in the mesh cycle. The tooth force was divid-
ed by the width of the tooth and applied as torque.
The residual stress due to case hardening was
found by applying an inelastic strain profile in
accordance with residual stress measurements.
The hardness penetration depth and ilts shape
were obtained by hardness measurements ..
Fig. 2-Close-up 01 a cross-sectional a TIFF. The wing crack indicates the
propagatio.n direction of the main crack (from the interior of the tooth towards Fatigue tests of the gear showed that shot peen-
the}ltink). . ing increased the fatigue endurance limit by 36%.
Residual. stress measurements of shot peened gear
wheels showed that the . hot peening increased the
compression stresses to a depth of 0.1-0.2 mm.
Crack inilialio.n criterion. In order to predict
crack initiation in high cycle fatigue during mul-
tiaxjall.oading, Findley (Ref. 9) suggested a.criti-
cal plane approach where the shear amplimde ft.)
and the maximum aormal stress (O'n.mnx)duriQg
one load cycle are used to form the criterion. The
criterion can be written
(1) Fig. 3-Two-dimtlnswnaljinile element mesh.
0'1' = '''!:til

(2) "

I an equation for effective fatigue stress (findley

critical plane stress}. The criterion states that crack
initiation will occur if 0'1' is greater than. 0' ,crlC' Here,
O'erll and acp are materia] constants that can be
determined by combining the result of two
fatigue tests. Here. 0'''';1 in the core was, deter-
Magnus IMackAld.ner
(M. Se; Lie. ElIg.) nct/vtll
mined 'W' be 'O'crit.mrc= 479.8MPa while for the lib degret! in J'dlck
case" O'cril.<:ast= W09O.0 MPa. The aep parameter Engillt,"n .,. 'llA,· Boy,td
F.ig. 4-Comparisoll of CIRF. 0) the twtied'gelU InstflUlt ,0/ Trehftologjl'
was determined to be 8ep,<llft = 037 andacp ••ase = wed as ,a singu stage gear, b) the studied gear used tcry) ,of Slodcho.lm"
1.00, respectively. Shot peenjo,g is c-onsidered in as an idler. SIfII!t/m, ill J 995. Since
the FE-analysis by increasing the fatigue liotit The region with high ClRF is located approxi- 1997.h~ hlls bet!1I /fll'
mately mid-.heigb~ of Ule tooth and sUghtJiy below in4J1SlTlaJ Ph.D. stulllnJ ~
(Gail) 36% LIl:the ood'es.·of the model within 0,2
Scoma, Ernope) third-
mm :from the surface ..The O'cril. and 'the acp para- the case-core boundary. Thu • if a crack. i devel- WTgf!.t hellll)l truck mll"u-
meter were taken to vary continuously between oped in the interior of the tooth, it will initiate filctlllle,. 4fttl at tilt dtpan-
the case and core in tile same maJIner as the hard- between the center-line and Ithe case layer .. The wunJ ,0/ MllcJdllt .Dalgn
til KTH. His r:t!Stflrell roc»J
ne s profile. CIR~ in the interior is increa ed by 20% if the is, Oil' tile u:plo,utio" of
NumericJlI analysis. The fE-analysis is carried gear is used as an idler instead of as 8. single stage jflliglltp,.t!nOIflt!1I1l in

gear. Moreover. the area (or volume) with high gllln, GS 11',0IlS robm
through in two rages: 1) calculation of the tate of
gelll' dt!51gn..A't SC4n1a,.hI!
tress hi tOly in dletooth during one load cycle. CIRF is larger in case of idler usage ..Allalysi of 1I'0rb ~ ,1M dquuLNJrnt /0'
including residual tresses, and 2) evaluation of the the risk of TIFF at differelIl. loads show that. as r-_mioft lJewlDplftDU,
wIJIIT·he is domg SlrllctrmiAl
risk of crack initiation. The engaging gear teeth are the tooth load is increased. it is more li.kely to have
analyzed in 19 different points of Ithe me h cycle, a crack initiating m the root than in the interior.
each. point representing a ''frozerf' moment in the Thus, TIFF is a presumptive failure mode occur- Marten Olsson
mesh cycle. The points included nine points with ring at medium load. i.e. the load level is between (M. Se., Plt.lJ., .iJocorl) is
A55o>ciIlI'ePrui-rssOI' Ilml
contact on each flank and one with the tooth the load at which contact fatigue .is achieved and
unloaded .. Results of the 19 analyses are combined the load at with tooth rool bending fatigue is .SllldiLS al till! Dtp4rtllttnt
to represent one load cycle (i.e, one revolution of obtained. This agrees weU with the experience of Solid Mt!dlllnir • BO)laJ
IlIXtiJIIl'e o/r«luwlogp.
the idler). The Findley critical plane stress is from the rig test of the idler gear.
,Stockholm, SlfII!dtlt., .His
fonned for every degree IQf plane inclination at Parameter Study re tt1ld ba&kgrolLM
each node of the cross-section of the loaded tooth. Crucial for a. successful gear design is that the incllldu IIl1a1p& ol,tOflglt-
enmg lItt!chani£1N lor
Th maximum Findley criticaJ plane stre at each designer understands how different design. para- ct'4miC$, tit_,ll ,rycllltg
node is stored (or evaluation purpose . meters :infill.ence the respcn e (cen:ain. propertie ), 01 cer:1ll1fk reUt/orrell IMII1l
Numericolresults. In order to judge th risk of 'Of the structure. Such knowledge can be gained by IJfllmx ctJlffJHJSile .: 1I't'1I1'
rflMIIJICt of FGJh iIlI4
crack initiation. the crack. initiation risk factor a factorial de jjgll. In contrastto a "one-factor-at-
Conlaa/fdillfle ,0/ CMt-
(CIRFJ must be calculated. This i done by divid- a-time" approach. the factorial. design also gives htl.rdent.dttt.&' ClITn_1IJ
i.ng the Fmdley crilica] plane tre by the O"cri,va1- information about. interaction 'effects. For a ITstarch ,COllur:m llIIi.glll!
inilJJJlloll t11HI pNfpllglllio1f,
ueat every material point. A contour plot of the detailed discussion on factorial. design. ee 60x et IIoth/fOllI 1II/IIJUl'tmtnnm'
CIRF of the gear can be seen m Figure 4b. It is also al. (Ref. 10) or Montgomery (Ref. 1I). pOlm ,o/vlew lIlld III' appU.
Here s 'a factorial design with five factors (A to E), CflIiD1I.$, Ih coo~rtI1~ ,willi'
interesting to eompare the emF for the studied
tire &4nJII tnuk CDIffJH1If,
gear used as an idler and in a single stage gear; The was conducted. The parameters areth two maieri- lil(ga.l'tiinll cOIl/lIcl '1!/lIwl
comparison can also be seen in Figure 4. a1 constants O'crit (A) and 8cp (B), the slender- f~gllt In trlUUmlsJ101l.$,
Fig. S The sle1lder (left) and not-slender (right) geometry compared "".itk the geometry olthe original tootl, (middle).
seated by an FE-anal.ysis determining the ClRF in
Aller:ageof factors and Interaction
the unerior of the tooth.
Result of the paramete,. study. The result of
oe -.l r .t
the factorial design is the CmF (response) in the
os. -\-.--..--L-----L-----..L--~----~~-.
~ j j interior of the tooth denoted Yi for the ith experi-
104 ment (Table 4). Emanating from the results it is
~ 03
possible to estimate the main effect. The main
C; 0.2
effect should be interpreted as the change in
response while changing the level of the factor
.... A+ C-C+ E-E+- N:;-A(J+ 0. D-o. A,E.I,E ... OE-CE+ from low to high. The main effect is given by
<T.,o Sh,"~$~ Too/Ji Carburi~jfJg
ntlio lOad d&pth . [J.:+ [Ie,
Fig.. r-The average of the facto.n with most .influence on the elRF in ,the inte-
mN+ "" -k+ L .Yi+ -T L »; (3)
rior: The ove.raUmean is marked .by the dashed line. i=l i=l
where k+ and k- are the number of times the fac-
tor appears at highand low level, respectively ..
The result of the factorial design is summarized
in Table 2.
In Figure 6 the average of the factors regarded!
as significant is plotted. The main effect is the dif-
D Carburizingl depth O.9mm L5mm ference between the low (-) and l1:igh(+) level for
each factor ..
E Tooth force 1238 Nm* 21164Nm*
The greatest ieffuence 00 the CIRF ill the
interior of the tooth comes from the factors A
ness ratio. C (defined as toe ratio between the (Gerlt). C (slenderness ratio), E (tooth load) and
height of the involute «da-dpf)/2) and the tooth B (acp) in that. order. The influence 011 the CIRF
thickness (Syn) half-way between the traasition in the interior from the factor D (the carburizing
from root fillet to involute (<iFf) and the addendum depth) is less than 5% but, somewhat unexpect-
diameter (da the carburizing depth (D) and the ed, it is negative. This means that the lower the
tooth load (E). For each factor. two levels were earburizing depth, the higher the risk of HFF . .It
assigned. one low (marked with.a minus sign) and is also worth noting that the influence from the
one high (marked with a plus sign), The low and carburizing depth (factor D) is less than the
high level of each factor is set to approximately interaction effect of O'eril (factor A) and the slen-
75% and 125%, respectively, of the value of the derness ratio (factor C).
original (previously analyzed) design of the stud- Since it is desirable to have as Iowa CIRFas pos-
ied gear. In Figure 5, the slender and not-slender sible, the optimal.settings of the factors are A+, B-,
gear design are compared with the geometry of C- and E-, meaning high O'eril in the interior of tile
the original gear wheel, tooth, low slenderness, low tool.hload and low acp'
For details about the tooth geometry, see Engineering Design Metb.od
Table 3. The levels of the factors are given in What so far has been shown in the present
Table r. paper is that it is possible to predict and analyze
Here, 32 experiments were conducted in TIFF by applying the FEM. However, this .istime
accordance with an ordinary L32 orthogonal consuming and not appropriate in the design stage
array. Each "experiment" in the array is repre- of gear development. Therefore. an "engineering
design method:' (EDM) was developed and imple-
mented in MatLab .
.Adesign method based on superposition of
elementary solutions. Two basic assumptions are
assumed for the EDM. These are:
i) The critical Findley plane is assumed to be per-
pendicular to 'the center-line of the tooth.
ii) Stresses itt the tooth interiora:re expressed 11.189
using elementary solutions of elasticity theory.
Normal stresses are estimated by beam theory.
Shear stresses with the half-space
are estimated
elution combined with beam theory,
Basically, three types of superimposed stress
states, acting on the pre-assumed criticalplane,
need to be considered. These are i) residual stress-
es due to case hardening ii) normal tension stress-
es due to bending and ill) shear stresses.
Residual stress. Abeam (or roo) with varying
em s-section can appmximate the tooth in the
residual. stressestimation .. In a surface layer wilh
thickness dCeqv. a constant volume expansion ~ is
prescribed. A sketch of the simplified tooth can be Fig. 7-Simplified tooth.
seen in Figure 7,.
The width of the tooth is b, the height is II and
the varying tooth thicknessas a function of the
diameter dy is given by syn(d,J The stress in 'the y-
direction in the interior oflhe tooth can be
neglected. By assuming that stresses are homoge-
nous in case and core, respectively. and combin-
ing equilibrium and! Hooke's law in the x- and z-
direction of the toollh. the folowing expression of
re idual stress is derived:

• (s . (dy) - 2dc +b)

O'x-core O'n:s<d,) ::: ZEEbdceqy )'!! .. S" (4)
Fig. 8-Tooth load 011 gear tooth.
Equation 4 is valid at the center line of the tooth, Shear stress. In estimating the shear sire s
but FE-calculation has shown thatlhe residual amplitude at point P due to the tooth. force, a pure
stress in the x-direction of the tooth is fairly con- beam analogy does not result in an acceptable
stant over a region on both sides of the center line. estimate. The sires di nribution, according to
Therefore, Equation 4 can be taken to hold~orthe Johnson (Ref. 12). in an elastic half-space, as in
interior of the tootll, not on}yal the center line .. Figure 9, due to a line load is given by
Nomud stress du,eto ,tvothload. Consider a
gear tooth as in Figure 8. Here, dey is the bottom (7)
diameter of the
active flank, da is the addendum
and dy is an arbitrary diameter. F(dy) isthe total However, the half-space solution will result. in
tooth force as a function of'the diameter ..The max- stresses on the tooth. boundaries (ill and il2) that

imum normal stress at a point P at diameter dy a are not present in the real case (:i.e. free surfaces
distance i" from the center line is given by: are always stress free).
Since linear elasticity is assumed, superposi-
tiOIl can be utilized andthe error of the non-stress
free boundaries can be compensated for. Tile
where method is to superimpose a stress state on the
elastic half-space solution that eliminates the
(6) stresses at the boundaries (i.e, subtract boundary
stresses). lIt order to implify the computation,
tile tapered tooth is approximated! by a. beam, with
I constant thickness, The superposition technique is
illustrated in Figuce 10.
\ ! F(dI
Consider the compensating solution (in Figure
x 10). The stresses on 01 and 02 in the compen-
sating solution are given by the half-space solu-
tion (n Figure 10). After subtraction. stress free
boundary conditions are fulfilled. Tile shear
stress due to the line load! on the half-space has
been presented. The shear stress due to the com-
pensation solution may be estimated! by beam
theory. The transverse force (T) on the studied
plane in which point P lays is given by integrat-
ing the compensating solution along the bound-
Fig. 9-Elastic half-space estimatio.n of the contact problem in the tooth.. aries ell and °
2, The latter integration should
only be performed from the tooth tip to the stud-
ied plane in which P lays. The contributions to T
from the two boundaries should be added to
make up the transverse force. The 'transverse
force at P is given by


- of(x=o.z)
'-+ T,.Vx= iJ,z)

r -,-,-,- Half-space
o'/(x.z=s,..! slrcsses on
Z 7 .. 1(x: .z=s,.,J f.} and 4
where Xdis1 = (dy-d)/2 and S = (da-dJ,)/2. The
shear stress due to the compensating solution,
P ~TXz
- x
estimated as bending shear stress due to thetrans.
verse force T, is then given by
) I
C (9)
tn = --I(dT")--=b--

Sires s free where S(zb) is the static moment given by

z boundaries - -

Fig. 10---S~perposilion of elastic half-space solution and compe.nsatingsolution.

Comparison between EDM andl'E·lullits

1.4 -,----------------------,
However, it should be noted that the compensating
1.2 +---..
-,I.. .•
•• _jj~ .... _, oW ....... !Ii _'.,lII •• ~- shear stress should only be subtracted if the tooth
••• !!f

force is p? itionedoutside the studied plane (i.e,

i," 0.8
~ > dy)' The !esulung shear stress at point P in a
planeat ely is thus given by
; 0.6 -ta.--flI+r------.;~!__-----I'-----~_IPtr+_~~___1H
u il·'tl' '" 1:
1- 9(d y - .d) 1:n c (l l)
I0.2 +------------"'--1

---tr-CIRF .EDM i
i where 9 is the Heaviside's step function. However, [CIRF.EDMV[C'RF -FE) I

O~~~~~~~rT"_.~,_rr~-rrT~_."~ i
i in the Findley critical. plane criterion, it is tile shear
- '" It) I"'" OJ - roLO ,.... (I) ....' '" It) r- en
N N N N N ro stress amplitude rather than the shear stress itself
..... '!""

that. is used. Therefore, the maximum and mini-

F~g. l1~Comp.~ri'son of the CIRF calculated with the engineering design mum shear stress at each point is sought n is not
method and the elRF according the .FE-calcu.lationsin the ptirameter,study; obvious when during the load cycle the maximum
and the minimum occurs, respectively .. By imple- risk of TIFF was decreased as the carburizing
menting the derived expres ions in a computer pro- depth was increased, This unexpectedrel.a:tion can
gram (here M_atLab is used), it i possible to deter- be understood if the bending stress in the tooth is
mine the maximum and minimum shear stress at all considered. The FE-analyses how that the region
points m the tootlil interior, and the shear stress in the tooth with the hlgnest CIRF is in the case-
amplirude can. be fonned by core boundary. Also, when the carburizing depth

't P=
max. {1:.YP h....
.....,..1c - mini t.,n
-J' (2)
is low, the layer with the compressive stresses is
thin and hence. the bending sires in the case-core
• 2 boundary is greater due to the greater distance
Findley str,ess. The total stress, in the radial from the center line. The conclusion is that the
direction of the toolil (or in the x-direction as in. effect 'of the bending stress is greater than the
Figure 7) is the sum of the residua] stress and the effect of the residual tensile stress in the interior of
bending stress. COllsequenlly, 'the Findley critical the tooth.
plane stress can estimated by The parameter study also showed that the two
factors having the greatest influence on the risk of
'0/ =tBP + acp;core (ores + oP olll1Dlll) (]3) TIFFareoCri1 in the interior of the tooth, and the
slenderness ratio .. This highlights a new problem
The bar over the OF indicates that this is the when it comes to gear design. When optimizing
EDM approximation of the Findley tress. By gears for noise, usually the slenderness ratio is
scanning through aU planediameters and! aU dis- increased, since slender gears allow hlgher contact
tances :from the center line, the point in the interi- ratio andpotentially more silent gears. Therefore,
or of the tooth with maximum Findley stress can TIFF has to be considered as a possible failure
be determined. mode in future gear design.
Comparison with FEM results 'Conctusions
The total tooth force as a function 'of contact [II this work, tlIe gear tooth failure mode1FF

force diameter is determined by the gear compu- is described and analyzed. In the analysi • FE-
tation program LDP (Ref. 8). F(dy) is then approx.- computatinnsare utilized in conjunction with the
imated by a fourth-order, least- quare fit to the Findley critical plane initiation criterion in order
calculated tooth force. to predict crack initiation. It is shown that the
In order to, verify the engineering design TIFF cracks are initiated in the interior of the
method. a comparison with an FE-analysis was tooth. Other results are:
conducted. Since an evaluation of the CIRF for • The region where the interior crack: will initiate
different gear geometries is already made ill the is located approximately mid-height of the tooth
parameter study, it i possible to make a compari- and slightly below the case-core boundary.
son for different geometries and parameter combi- • TIFF is a possibility at loads lower than. the load
nations, The previously conducted FE-analysis where tooth root bending fatigue is achieved and
has shown that the area inthe interior of the tooth at loads higher than the load where contact fatigue
with the highest Findley stresses is close to the occurs.
case-core boundary (i,e, at depth CD). ]n the • By usingthe gear wheel as an idler instead. of as
analysis, dceo:rv= 1.2 mm and .~ = 0.000833 are a single stage gear, the risk ofTlFF is increased by
used. In Figure ll,Bi compari on of the CIRF cal- 20%.
culated with. the engineering design method A parameter study was conductedin order to
(CIRF _EDM)and the CIRF according the FE- investigate which geometric and material parame-
computations (ClRF~FE) in the parameter study ters influenced! the risk of .WIFf. The parameter
is shown, study was performed as a factorial design. The key
Wtis clear fmm Figure I] 'that the overallcor- results from the study are:
relation between the CIRF calculated by the EDM • The parameters influencing the risk of TIFF
and by FEM i good. It i noticed that the EDM mostly are 0eril in the interior, the slenderness ratio
overe timates the CIRF. The average discrepancy and the tooth load. The lower the ,oeri!' the more
for all 32 investigated parameter configurations is slender the tooth and the higber the load, the
H % and never greater than 2'()%. greater the risk of TIFF.
Discussion • Tbeinfluence from the acp parameter in the inte-
The FE-analysis howedthat the hypothesis of rior is small. but positive, meaning the higher the
TIFF presented in this paper was strengthened. acp' the greater the risk of TIFF.
The parameter study surprisingly showed that the • The influence of the carburizing depth on HFF
Appendix-Gear data for gears in the factorial design. i.ssmall, and the risk of TIFF is lower for a high. car-
burizing depth than for a low catburizing depth.
Thelgear datal at the two, Igear designs in the, factorial designl are given
An engineering design method (EDM) for
intbe table below. fGr c'ompaliison.thegea r d'a.hllofthe 0 rigina'i design 'ofthe
gear is given in Table, 3. A sketch oftlte differem gear designs iis given in design against TIFF was developed, implemented
Figure, 5,.In Table 4. the' c~ack ,initiationlrisk factors ar,e presented as they in MatLab and compared with. the FE-calcl.llations
were computed by FEMin the 32 experiments in the factorial design. in the parameter study ..The follnwing results were
'. It is possible to estimate the risk of TIFF by the
EDM very quickly and with. acceptable accuracy .
• The EDM overestimates the CIRF in the interior
of the tooth compared to the FE-result by an aver-
age of 11%.
This work: was supporsed by the Swedish
Research Counei1 for Engineering Science (TFR).
Thanks are directed to Arne Larsson at Scania for
invaluable help in the field of gear design and gear
manufacturing. The authors thank: Gunnar
SttandeU at Scania for encouragement and for the
possibility to present this research, and they also
thank Prof. Soren Andessson at KTH for support
and guidance. 0
1. DIN (Deutsobe Instil!.!1filJ Normung) 3m Teil 3 (1987).
Tragfiihigkeitsberechnung \Ion Stimradem.
2. AGMA Standard 2003-A86 (1982). AGMA Standard for
Rating the Pining Resistance and Bending Strength of Spur
and Helical Involute Gear Teeth.
0.751 0.586 3. Mackaldener M. and M. Olsson, "Interior Fatigue
2 0.449 11.417 Fracture of Gear Teeth," Fatigue Fract. Engng MaJer.
0.752 0.646 Struct .• 23. pp, 283-292. 2000.
4. MackAldener M. and M. Olsson, "Interior Fatigue in
4 0.437 10.429
Gear Teeth," Proc, of Fa/igue 2{)()(),Eds. M.R. Bache etal.,
5 0.206 0,358 Cambridge, EMAS, 2000.
6 0..1115 10.259 5. Shulz, M., and J. Salller, Schadenuntersuchungen an
1 0.203 0.468 Zal:mriider, Konferenz. Einzelbericht: RaOOschichtenniidlll1g

I im Wiilzkontakt, AWT-Tagung,. Arbeitsgemeinscbatt WIinn-
...... ebehandlung und Werkstofftechnik, Subl, D, 06.-07. 1992.

110 0,449 0.353 i 6. Alban, L., E., Systematic A/Wlysis of Gear Failures,
American Society for Metals, 1993.
11 0.148 0.550 7. ANSYS 5.5 Users manual (1998) ANSYS Inc.
0.372 8, LDP, Load Distribution Program v 10.01. (1997) Ohio-
13 0.200 0.317 Stale University.
14 0.105 0.224 9. Findley W. N., "A Theory for the Effect of Mean.Stress
'5 0.203 0.3'1'1, on Fati,gue of Metals Under Combined Torsion and Axial
16 11.100 01.251 Load or Bending," J. Ellgn.gfor Industry, November, 1959,
11 11,720 01.916 pp. 301-306.
10. E. P. Box, W. G. HU!lter and J. S, Hunter (1978),
18 11.380 0.'695 Statistics for Expenmenters, John Wiley and Sons.
1'91 0.180 1.()66 II. D. C. Montgomery (1997.), Design and Analysis of
Experiments, Fourth Edition, John Wiley and SOilS.
21 0.206 0.596, 1.2.K. L. Johnson (1993). Contact Mechanics, Cambridge
22 0.105 0.432 University Press.
23 0.203 0.681
24 0.100 0.476
25 0.'690 0.839
26 0.4301 0.589
21 0.750
291 0.206
30 0,105
31 0.203 0.'61'9
32 O,TOO 0.428