See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328940444

Failures on the gears

Preprint · November 2018

DOI: 10.13140/RG.2.2.12425.29281

CITATIONS READS

0 219

3 authors, including:

Boris Abersek Srečko Glodež

University of Maribor University of Maribor
164 PUBLICATIONS 954 CITATIONS 162 PUBLICATIONS 2,093 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Boris Abersek on 14 November 2018.

The user has requested enhancement of the downloaded file.

Failures on the gears

Boris Aberšek1.2, Jože Flašker1 and Srečko Glodež1.2

Faculty of Mechanical Engineering1 and Faculty of Education2
University of Maribor, Maribor, Slovenia

Abstract
Modern gears play an important role in high performance engineering, Fit-for-purpose design
rules have emerged over generations of development. However, the rate of change has
accelerated with the advent of modern computers and more thorough analysis techniques,
coupled with a greater understanding of materials behaviour. Particular types of spur gears, as
used m lorry gearboxes, have been studied in great detail using these latest developments,
including the adverse effects on fatigue life due to non-designed characteristics. These
comprise stress-raising features on or near the gear surfaces, and are usually due to contacting
surfaces or stress concentrations which lead to crack-like failures. The paper reviews these
features and illustrates them with practical examples.

1 Introduction
More than 25 years ago, we started to research the applicability of numerical methods, initially
of the final element method (FEM) and then also of the boundary element method (BEM), to
gears, which are mechanical elements of considerable significance. We started with the
comparison of stresses in the tooth root, by means of the method of photoelasticity and of the
DIN method, which is the most frequently used in practice [1, 2, 3]. It has been discovered that
the numerical method of FEM is an excellent alternative to standard methods, e.g., to the ISO
method. With the development of new software and computer technologies, a great advantage
over the usual standard methods has also been shown regarding time consumption. It has been
proven that every change in the tooth geometry or in the tooth, which has influence on the
stress field in the radius of the tooth root of a gear, can be mastered quickly and simply. In the
initial phase, the results acquired had to be compared with those acquired by means of standard
procedures and theoretical starting points. The results were comparable, thus, research work in
the field of application of FEM could be continued. A few years ago, we started intensive
research on stresses in teeth flanks – Hertzian pressures (stresses) – which have a significant
influence on the damages in teeth flanks. Gears always appear in pairs when used, which
causes stresses on the surface or in the subsurface of teeth flanks. Sliding contact of the
mechanism of gear meshing results in micropitting, which leads to macropitting – this can
cause such damages to teeth flanks that gear(s) can no longer operate.
A number of matters are still open, some solutions are only indicated. It has to be stressed
that the operation mechanisms of a gear pair are still a very complex matter; one has to be
aware of that and to take that into consideration in order to prepare a practically applicable
model.

1.1 Why gears?

Gears are the most frequently used mechanical parts and familiarity with gears is essential for
any mechanical engineer. Gears have been and continue to be one of the most important means
of mechanical power transmission and this situation will undoubtedly continue in the future.
They offer highly efficient, precise and direct power transmission at relatively low cost and
maintenance.
These are the reasons for the interest of the research community in developing suitable
mathematical and numerical models. The main goal of any developed model or procedure is to
make it possible to model the behavior of gears under realistic working conditions as
accurately as possible, even prior to manufacturing of the gears, thus minimizing the
possibility of failure of gears in a real assembly.
Nowadays, the market for gears has become global and trade can be carried out without
problems only under the condition that there are no technical barriers. The achievement of
general interchangeability of gears, without any technical barriers, is the responsibility of the
International Standardisation Organization (ISO).

1.1.1 General description of the market

Power transmission gearing is characterized by its huge diversity and essential application into
almost all industries, from transportation, industrial manufacturing, and power generation,
through to such state-of-the-art technological applications as complex aerospace components.
Gears must be capable of operating in extremes of environment. They must be available in
every part of the world to support replacement demands. This requires gear products to be
designed, manufactured, and measured using standardized methods. Power transmission
gearing, in many forms, is a very mature product and standardization work has been taking
place for almost 100 years.
Nevertheless, design innovations, new materials, improved manufacturing technologies and
an increasing demand for solutions by new and developing industries, require an active
standardization agenda. The primary goal of ISO/TC 60 is to ensure that gear standards are
created, kept up-to-date with changing technologies, reflect the needs of the industries
concerned, and give assurance to the end users that gear products are durable and safe.

Description of the total market The world gear market has three major sectors of production
and consumption as well as several areas that are expected to be significant suppliers in the
coming years. According to data developed by the trade associations in Europe, Japan and the
United States, the world market for non-automotive gearing and related products was over 10
billion dollars in 1998 (the last year when such comprehensive data was available). The
comparable number in 1992 was $ 7.2 billion.
Table 1 details production, trade and consumption by sector in 1992 and 1998, in US
dollars. As is evident from the above mentioned facts, the market is extremely large. Thus,
everyone in the world makes a lot of effort to compete for market share, or to find a market
niche. This can be achieved only on the basis of innovation, in the following fields: material
selection, selection of manufacturing processes, heat treatment, new forms of profiles of gear
teeth flanks with the purpose to reduce weight, to decrease the costs and to increase the quality
of operation during the planned life time. Gears have a very wide range of applications – from
environments with extreme conditions (universe) to very large loads (transmitters in the field
of energy production) and to miniature products.

Table 1: production of the gears

1998 Total, $ Europe, $ Japan, $ United States, $

Production 10 032 000 5 410 000 1 910 000 2 711000
Imports 3 009 000 1 707 000 133 000 1 168 000
Exports 4 343 000 2 914 000 858 000 570 000
Consumption 8 697 000 4 203 000 1 185 000 3 309 000

1992 Total, $ Europe, $ Japan, $ United

States, $
Production 7 225 000 3 410 000 2 015 000 1 850 000
Imports 1 625 000 940 000 74 000 611 000
Exports 2 368 000 1 426 000 607 000 335 000
Consumption 6 482 000 2 924 000 1 482 000 2 126 000
This data does not include automotive industry captive production

Naturally, this means that there is an increasing requirement to prepare a generally

acceptable standard; a standard that can be developed within the ISO.
Our research work has, however, shown that numerical methods, finite element method
(FEM) in particular, are a good alternative, which is clearly presented in many works [1, 2, 11].
Every standard method is based on coefficients acquired on the basis of standard tests of a
standardised gear pair. Those coefficients are then, according to the similarity theory, applied
for other gear pairs, smaller and larger ones, for other gear ratios and for other geometrical
data. For the time being, however, gears and the processes occurring after the initiation of
cracks in tooth root, or in the surface of a flank are still not mastered by means of standard
procedures. It has been proven by research that FEM and other methods (e.g., boundary
element method; BEM) are, or will be, an excellent tool for mastering a gear pair or a gear
when the above mentioned anomalies occur.
Of course, further development of methods of diagnostics is required to make it possible to
establish the state of a gear pair in use with certainty and to find faults in time so as to plan
measures, which are to be taken, at an early stage. Some of the research work carried out
makes this already possible, however, in practice, the level of reliability of the method of
diagnostics is still not acceptable. Basic characteristics for major faults will have to be defined,
and on the basis of results, it will be necessary to develop reliable models for diagnostics,
which will be fairly simple, reliable and cheap. The model has to be reliable and practically
applicable, thus, practical tests will have to be carried out in order to improve the model.
Considering the possibilities of FEM, successfully tested in practice, we are of the opinion
that, by means of numerical methods, it will be possible to, with a high level of reliability,
predict the life time of a gear or of a gear pair, in view of the crack in the tooth root or the
surface/subsurface of a gear flank. This is not possible by means of the classical standard
method. With further development of software and hardware, this will become increasingly
simpler, cheaper and more reliable.
We are of the opinion that in the presented chapters, practical applicability of the method is
already well presented, it is also evident that the method is open and thus, constantly improved
with new findings and new cognitions. Furthermore, it is simple to use under the condition that
one is familiar with all rules and mechanisms of the operation of a gear pair and also with other
effects which have an important influence on changes (positive and negative), on operation and
thus, on the life time.
On the basis of this work, it is clear that an analysis of a mechanical element of such
significance as a gear can be carried out only by an expert who is familiar with the mechanisms
of the operation of a gear pair and with other influences (lubrication, selection of a material)
and also, in detail, with the rules of numerical methods so as to ensure the credibility of the
evaluation carried out with numerical methods.

1. 2 Scope of numerical methods

The finite element method (FEM) [4] and boundary element method (BEM) [5] are the well-
known products of research developed alongside the emerging digital computer. Their
applicability for analysing the states of solids and fluids under various physical conditions is
widely practised. For gear technology, the strength of these methods lies in their treatment of
various types of linear and non-linear stress and dynamics analyses. Two and three
dimensional models are accepted which allow realistic non-symmetric loads to be studied.
FEM allows virtually any degree of non-linearity to be followed, using element
subdivisions throughout the domain to represent geometries of any shape. The direct
evaluation of stiffness matrices at the element and domain level are central to this diversity.
BEM has the advantage that the elements are only defined over domain boundaries, so one
less dimension of input is required compared to FEM. This is the chief advantage of BEM over
FEM and is a great attraction to the user. A fuller comparison of the two techniques has been
given by Hellen [11]. However, the lack of a stiffness-type approach requires considerably
more complicated techniques to deal with general solutions to fracture and non-linear
mechanics.

2 Gear tooth breakage

Gear teeth can break off due to cracks emanating from stress concentrations in the root region.
The possibility of such fractures is investigated by combining fatigue and fracture principles.
First of all, it is necessary to evaluate the stress field in the tooth subject to the applied loading
and the way it is transferred into the tooth. This incorporates the effects of the contacting
surfaces and the influence of lubrication and friction, and, from a practical view-point, the
likely variation of load along the tooth flank, which could vary iii spur gears which are not
precisely aligned. Such an evaluation is possible using FEM and BEM.
From these results, the highest stress indicates the source of probable crack initiation and
the number of fatigue cycles required. Methods analogous to initiation of pitting cracks
described earlier are relevant.
Subsequent crack propagation can be followed by numerically integrating the Paris law, as
demonstrated by equation (8). As in the pitting long crack case, the incremental approach is
used, with values of ΔK calculated by FEM or BEM at, each instantaneous tip position, to give
the number of load cycles, N, required to grow the crack from a length aj-1 to aj, for each
increment.

2.1 Model of crack initiation

To be able to determine the properties of very short cracks, it is important in the analytical
models describing the S-N curves to consider the dominant properties of the fields of the crack
tip as well as the crack growth rate, which is high in the beginning but then decreases [1, 3].
This can be reached by considering the separate regimes in the Kitagawa - Takahashi diagram,
which gives a significant advance in the understanding of short crack behaviour. This diagram
shows the effect of defect size on the fatigue limit stress, see, for example, Figure 1. For large
defects, the allowable stress for infinite life must be low, within linear elastic fracture
mechanics regimes, and, therefore, the limiting condition is given by a straight line of slope
minus one half of the threshold stress intensity factor DKth, such that

DK th = Y(a/S) Ds p a th (1)

L SHORT FATIGUE CRACK

O
G
NOTCH
S EPFM CRACKS
T FATIGUE LIMIT
(high stress)
R
E COUNTOUR da/dN=0
S
S NON PROPAGATED
CRACK
R
MICROSTRUCTURALLY PHYSICALLY LEFM CRACK
A SHORT CRACK SHORT CRACK
N
G
E d1 d2 d3
0.01 0.05 0.5
LOG OF CRACK LENGTH ( mm )

Figure 1: Regimes of crack properties

where Y(a/S) is shape factor [3], Ds is stress ratio and ath is a threshold crack length. At the
other end of the spectrum, for vanishingly small defects, the allowable stress level must relate
to the uncracked specimen fatigue limit. The Kitagawa-Takahashi diagram shows the observed
behaviour between these two extremes.
It is known that the results are more accurate, if the number of data is great. By means of the
generator of random numbers the number of input data can be numerically increased and on the
basis of that new population it is possible to calculate the new mean value which can deviate
from the original value for not more than the specified standard deviation.
The linear elastic fracture mechanics is inapplicable for describing properties of very short
cracks, since elastic stress cannot accurately describe high deformation fields near the tip of
small defects in high-strength materials. Therefore, we used the theory of continuously
distributed dislocation for analysis of crystallographic slip deformations before the tip of short
fatigue crack in combination with crystal limits and on the basis of it we formed an appropriate
model [3]. The stress connected with dislocations can result in the slip of the inclusion or in
generation of vacation. Dislocations can unite on any obstacle of the slip which causes their
concentration in space. This results in nucleation of microcracks and/or crack growth with
uniting of remaining dislocations. The assumptions described above were used in our model
for initiation and propagation of microcracks.
 (a) (b)

Figure 2: Edge (a) and spiral (b) dislocation

The border between the area where the slip occurred and area where the slip did not occur
is called the edge dislocation, see Figure 2. The presence of the edge dislocation increases the
free energy of crystal. In order to calculate, let us consider a cylindrical crystal of length l with
spiral dislocation on which the Burger's vector is located along its axis. The Burger's vector b
describes the size of shift and is the entire distance between atoms in case of crystal grating.
The energy on the volume unit of this cylindrical Crystal of radius r and thickness dr:

dE 1
= G γ ro2 (2)
dV 2
where gro is the elastic stress in thin coil. From equation (2) it is possible to calculate the
deformation energy by integration:

R
1 2
E= ò 2 G γro dV (3)
r0

The equations (3) show that the energy is proportional to the length of dislocation. Such a
deformed dislocation has linear stress T which is a vector and is calculated as follows:

δE
T= (4)
δl

In this way it is possible to calculate the relevant shear stress:

f (xo ) dx
T(xo ) = A (5)
( x - xo )

and for equilibrium the integral equation is obtained [4]:

f (x) T (xo )
ò dx = (6)
D x - xo A

Gb Gb
where A = applies to edge dislocation, A = for spiral dislocation and G
2 p (1 - n ) 2π
is the shear modules.
The solution of this integral equation was shown aheadly by Muskhelishvili in 1946 [6].
Thus the general notation of the distribution function in a slightly simplified compact form can
be written:

σfé -1 æ 1 - dξ ö æ 1 + dξ ö ù
f(ξ(ξ 2ê cosh çç ÷÷ - cosh -1 çç ÷÷ ú+
π A êë è d-ξ ø è d+ξ ø úû
(7)
σf ξ æ æ öö
+ 2 ç 2 sin -1 d + π ç T - 1÷ ÷
( )
π A 1 - ξ 2 1 / 2 çè çσf
è
÷÷
øø

where x is dimensionless co-ordinate of crack tip.

2.2 Model of crack propagation of gear tooth

Statistical analysis of experimental results shows that the material parameters in the equations
of crack growth are random variables. As an example let as take the modified Paris - Erdogan
equation:

da
= C ( DZ )m (8)
dt
or in stochastical form:
da
= f( DZ) A(t) = C f( DZ) [ µ + Y(t)] (9)
dt

Where µ = E(A(t)) and A(t) is stochastic time process, and Y(t) is random process with mean
µ. The two parameters C and m can be considered to be random function. A good approximation is
obtained if it is assumed that both functions, i.e., C and m are normal stationary random field with
the zero mean and independent values (white noise) [7]. We obtain the conditional probability
function according to [6, 1]
t
f(u) = ò f(u | t) f T (t) dt (10)
0

The probability function f(u) for crack size a(t) in the time t for u>a can be calculated from the
conditional probability density function fa(u|t), and the probability function ft (t0) from the time of
crack initiation. By integrating the equation 29 we obtain the distribution of the crack size for a
great number of cracks initiated at any possible moment between 0 and t with density
corresponding to fT (t).
To conclude, we would like to mention that equation (10) represents the probability function of
crack size in the time t for long or LEFM cracks longer than the initial crack a0. By using the
integration cycle-by-cycle of equation (10) we obtain the so- called "master curve" for the ratio
crack size/time, i.e., the curve of service life. The crack growth ratio with certain crack size can be
calculated by deriving that "master curve".
 On the basis of the model presented above it is possible to calculate the service life of the
individual gear tooth. The service life of gear can be obtained from equation.
1
-
Li = f(u) N e (11)

2.3 Numerical validation

In order to verify correctness of functioning of the presented model, we have compared the
numerical results with experimental results [8]. The service life of the gear tooth without defects
is:

f(u) pinion = 2.0467 106 cycles = = 568.5 hours

f(u)gear = 2.0467 106 cycles = 2785.65 hours

so the service life of the gear for the final step is:

Lpinion = f(u) NT-1/e = 568.5 21-1/2.5 = 186.2 hours

and Lgear = 436.3 hours

Figure 3: Lifetime curve for gear tooth

Into the calculations we introduce the reliability factor. By using the statistics and reliability
calculation it is possible to calculate the service life of the gear pair according to [1,3] with
SGD = 95% reliability.

LGD = ( Lpinion -e + Lgear -e )- 1/e = 178 hours

If the respective service lives of bearings were additionally taken into account, very low
values of service lives were obtained. If we want to increase the service life, it is necessary to
use the high-reliable components.
 Figure 3 show a comparison of the results of crack propagation from the known initial of size
a = 0,5 mm. The figure shows that by means of the presented model we have reached good
accordance with experiment particularly in the area, which is most interesting for technical
practice, i.e., in the area of permissible subcritical cracks.

3 Pitting resistance of gears

Tooth wear failure occurs as pitting, a gradual deterioration of the contacting surfaces, as
shown in figure 1. Failure occurs when either the surface has deteriorated so much that the
component no longer functions as designed, or when the damage becomes severe enough to
fail by another mechanism, such as breakage. The process of surface pitting can be visualised
as the formation of small surface or sub-surface cracks which can grow under repeated contact
loading. They eventually become long enough to grow unstably, causing the material surface
layer to break away and leaving a void in the form of a pit.
Gear tooth pitting may be surface or sub-surface initiated. The former is observed in gears
with rough surfaces and poor lubrication, since it is strongly influenced by surface roughness
and damage-like machining marks, notches, etc. Sub-surface pitting initiation is common in
gears with smooth contact surfaces and good lubrication, so that sub-surface stress raisers
initiate cracking. This second type is the more relevant since modern gears are manufactured to
high quality.
Gear metals are polycrystalline and consist of grains, each of which has a particular crystal
orientation. Crack initiation can occur because of travelling and build-up of dislocations along
persistent slip bands in the crystal grains. This is strongly increased in high stress regions and
is the most common cause of crack initiation in metals [10].
Fatigue development in metals can be sequenced m 4 stages [1]:
• travelling and build-up of dislocations
• internal micro-crack initiation
• crack growth
• pitting, when the crack reaches the surface.
Approximate classical procedures for determining the pitting resistance of gear teeth flanks
have been developed (DIN, AGMA, ISO, etc). They are based on comparing the maximum
contact stress CTH, at the pitch point or at the inner point of a single tooth pair engagement, to
the maximum permissible value of O'HP- It is required that CTH < CTHP- Actual working
conditions have to be taken into account, such as additional internal and external dynamic
forces, the actual contact area, the material, and surface roughness. The most important
parameter when detennining the permissible contact stress is the reference contact stress of the
gear material, (THiim, which is defined as the maximum stress that can be applied to the gear
teeth flanks for a certain number of cycles before pitting occurs. It is a material parameter
obtained from tests. The classical procedures are exclusively based on the experimental testing
of reference gears and consider only the final stage of the fatigue process, the occurrence of
pitting.

3.1 Determination of pitting resistance

A numerical model incorporating all the above 4 stages has been proposed [1]. This
provides a more reliable determination of the pitting resistance, and makes full use of modern
numerical methods. In this model, the total number of stress cycles N for pitting of a gear tooth
flank to occur is obtained from
N= No+N,+Ni (3)
where Ny is the number of cycles for an iuitial crack to appear, Ns is Ihe iiuinl)(ii of cycles for
the ensuing short crack growth, and NI is the number of cycles for the final long crack growth.
Because the gears of interest have smooth surfaces and good lubricants, the maximum
shear and von Mises equivalent stresses due to contact of the meshing gears lie away from the
surface, underneath the point of contact. This is the likely site of crack propagation. For low
coefficients of friction, p, < 0.05, the position of the maximum shear stress is at a depth H ==
0.78b under the surface, where 26 is the contact width [11]. Thus, boundary effects can be
ignored. Mura and Nakasone [12] proposed an analytical model for determining the number of
cycles for fatigue crack initiation, based on the theory of dislocations motion on persistent slip
bands, and described the changes in the material microstructure due to tho applied load.They
used the Gibb's free energy required for the change from the state of dislocation dipole
accumulation along persistent slip bands to the crack initiation along this layer. Minimising the
Gibb's free energy with respect to the number of cycles gives [12]
(4)

where AT = Tmax ~ Tm«n is the applied stress amplitude on the slip layer, r/ is the frictional
stress of the material, h is the width of a slip band, 21 is the length of the dislocations pileup, /
is the irreversibility factor of the dislocations pileup, and 7 is the surface energy of the crack.
Experimentally determined values of the irreversibility factor / are small and range from
10~4 for small plastic strains to 10~1 for large plastic strains [10]. The value of Ar can be
determined from the contact stresses.
The short crack growth, with N, cycles, has also been described by Glodez et al [1], and is
assumed to progress only tlirough a small number ofgrains. Because the crack is short, the
local stresses are high enough to generate elastic-plastic behaviour. The crack growth takes
into account the successive blocking of the plastic zone by slip barriers, such as grain
boundaries, and the subsequent slip band initiation in the next grain. This gives a discontinuous
progress of crack growth. In each grain, the crack growth rate decreases as the crack
approaches the grain boundary. However, as the stress concentration ahead of the crack tip is
sufficient to initiate the slip band (plastic displacement) in the next grain, the crack growth rate
increases. The initial period of short crack growth is characterised by large variations in the
parameter n = o/c, where a is the half crack length aud c equals a plus the length of plastic zone
extension ahead of the crack, which always extends to the grain boundary. When n reaches a
critical value n = ric, the plastic zone extends into the next grain.
The Glodez et al [1] model is based on the modified Bilby-Cottrell-Swinden model [13] of
short crack growth. Here, the growth rate is proportional to the plastic displacement A.Spi at
the crack tip as
(5)
The plastic displacenient A5p( is then expressed in terms of the stress intensity factor AA',
written in this form to indicate that it is the value of K achieved per cycle, in a formula due to
Navarro and Rios [14]:
M
"^'^^ t6)
where G is the shear modulus and K = 1 or 1 — v depending on whether screw or
edge dislocations are being considered, v being Poisson's ratio. The critical parameter rtc is
given by [14] as
^)] (7)
Here, o'y is the yield stress and Kth is the threshold stress intensity factor. Thls equation is
solved iteratively by Newton's method in a few iterations, using Hc = 1 as an initial value.
Using equation (6) for the plastic displacement enables the treatment of both short and
long cracks in a similar fashion. If AA' is known, the number of cycles required to propagate to
a crack length a from ao is determined as follows [15]:
1. The number of grains, z, traversed by the crack is calculated by dividing the crack length
a by the typical grain diameter D, so z = a/D.
2. Analogous to integrating the Paris equation, the number of cycles required to propagate
through one grain is obtained by integrating equation (5) to give:
(8'
where j ranges over each grain and the crack advance is from a,_i to dj, determined on the basis
of the critical parameter •n^ since the plastic zone exteDds to the next grain only when the
parameter n = ric.
3. By addition, the total number of short crack cycles is
z
Ns=Y,Nj (9)
j=l
The long crack growth cycles, NI, is obtained by direct integration of the Paris law, as
demonstrated by equation (2). For both the short and long crack cycles, the above
developments assume that AJf is known at each successive crack tip position. Such values are
calculated by FEM or BEM using the incremental approach as described earlier. Thus, changes
in crack direction can be followed using the VCE technique, which in turn uses the maximum
potential energy release rate criterion.

3.2 Practical demonstration of pitting resistance

An actual spur gear pair has been tested for pitting occurrences, against which the above
model has been compared [1]. Equivalent cylinders were used with the

sanie radii as the pair ofteeth at the point ofcontact, to simplify the analysiK. The calculations
were made by FEM with a very fine mesh. Using the known iiiat.nrial properties ofthe flame
hardened steel AISI 4130, the comparison considercd crack initiation to determme No and
subsequent propagation to determine 7V,,. It was discovered that the crack growth was too
short to venture into the long crack stage, since only 6 grains were traversed when the crack
lcngtli became critical.
The experiments were tested using a FZG pitting testing machinc. The tests were
interrupted at certain load cycle intervals so an inspection for pittiiiR could be made. When pits
of diameter 0.5mm to l.Omm were detected on anv tooth flank, the test was stopped with the
corresponding number of cycles recordcd.
The total number of cycles N obtained for thc occurrence of pittiiif; froin the numerical
model was N = 3.454 * 106 to 3.751 * 106 cycles. The resnlt.s uf •I experiments showed N =
3.001 * 100 to 3.915* 106 cycles. This shows a re.Lsuiiablc agreement between the two
approaches. The pit sliapes on tlic contact surface.s from the 4 experiments are seen in figure 1
to be well prcdicted by thc FEM fracture predictions.

4 Concluding remarks
It can be seen above that for accurate calculations of service life of gearing the use of deterministic
methods according to different standards is not sufficient. Therefore, for calculations we have
developed different numerical procedures that takes into account not only the stress of
homogeneous material in critical cross section but also the stress concentration due to defects
occurring there.
 Numerical methods are central to many aspects of gear technology. As well as aiding design
by providing such key items as the basic stress states, they play an important part in the
analysis of detailed operational behaviour. Their roles in two important aspects of this, surface
pitting due to wear and the possibility of tooth root fracture, have been described. Both aspects
involve combined fatigue and fracture principles, by considering sites of maximum contact
stress and the ensuing crack initiation and propagation. Fracture mechanics laws have been
utilised with progressive degrees of sophistication, solved using finite element techniques. The
number of cycles required for the occurrence of pitting has been shown to correlate well with
experiments. Crack growth from the highest stressed point near the base of a gear tooth has
been studied.
With our models we have reached the deviation of 5 - 12 % from experimental results, which
is a very good accuracy in such calculations.

5 References
[1] Aberšek, B. and Flašker, J., How Gears Break, WITPress, Southampton, 2004
[2] Flašker, J. and Aberšek, B., Calculation of gears by finite element method considering the
gear tooth root errors, Proc. Numeta-90, Swansea, UK, pp. 160-169, 1990.
[3] Aberšek, B. and Flašker, J., Numerical methods for evaluation of service life of gear,
International Journal for Numerical Methods in Engineering, Vol. 38, 2531-2545, 1995.
[4] Zienkiewicz, O.C., The Finite Element Method, McGraw-Hill, London, 1977.
[5] Brebbia, C. A., and Dominguez, J., Boundary element method for potential problems, Journal of
Applied Mathematical Modelling, 1, pp. 372-378, 1977.
[6] Muskhelishvili, N.I., Some Basic Problems of the Mathematical Theory of Elasticity, Noordhoff,
Groningen, Holland, 1953.
[7] Provan, J. W., Probabilistic Fracture Mechanics and Reliability, Martinus Nijhoff Publishers,
Dordrecht, Boston, Lancaster, 1987
[8] Aberšek, B. and Flašker, J., Experimental Analysis of Propagation of Fatigue Crack on Gears,
Proceeding of ICM7, Haag, Netherland, 1995
[9] Glodez, S., Flašker, J. and Ren, Z., A New Model for the Numerical Determination of
Pitting Resistance of Gear Teeth Flanks, Fatigue Fract. Engng. Mater. Struct., 20, pp 71-
83, 1997.
[10] Pehan, S., Hellen, T.K. and Flasker, J., Applying Numerical Methods for Determining the
Service Life of Gears, Fatigue Fract. Engng. Mater. Struct., 18, pp 971-979, 1995.
[11] Hellen, T.K., A Comparison of Finite and Boundary Element Methods in CAD-CAM,
Proceedings 2rd Int. Conf. Design to Manufacture in Modern Industry, Bled, Slovenia,
1995.
[12] Hellen, T.K., On the Method of Virtual Crack Extensions, Int. J. Num. Meths. Engng., 9,
1975, 187-202.
[13] Johnson, K.L., Contact Mechanics, Cambridge University Press, Cambridge, 1985.
[14] Bilby, B.A., Cottrell, A.H. and Swinden, K.H., The Spread of Plastic Yield from a Notch,
Proc. R. Soc., London, A-272, pp 304-314, 1963.
[15] Navarro, A. and de los Rios, E.R., Short and Long Fatigue Crack Growth – a Unified
Model, Philosophical Magazine A, 57, pp 15-36, 1988.

View publication stats