Vous êtes sur la page 1sur 9

Wear 311 (2014) 31–39

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Calculation of gear tooth flank surface wear during an FZG


micropitting test
José A. Brandão a,n, Ramiro Martins a, Jorge H.O. Seabra b, Manuel J.D. Castro c
a
INEGI—Instituto de Engenharia Mecânica e Gestão Industrial, Universidade do Porto, rua Dr. Roberto Frias 400, 4200-465 Porto, Portugal
b
FEUP—Faculdade de Engenharia da Universidade do Porto, rua Dr. Roberto Frias, 4200-465 Porto, Portugal
c
ISEP—Instituto Politécnico do Porto, Rua Dr. António Bernardino de Almeida 431, 4200-072 Porto, Portugal

art ic l e i nf o a b s t r a c t

Article history: A numerical simulation of the wear sustained by the surface of an FZG pinion during a gear micropitting
Received 31 July 2013 test was performed. Underlying the simulation is a mixed lubrication model that takes into account both
Received in revised form surface roughness and lubricant properties, as well as a wear model based on Archard0 s wear equation.
18 December 2013
The adequacy of the models was ascertained by comparing simulated and measured roughness profiles
Accepted 31 December 2013
at the end of each stage.
Available online 9 January 2014
& 2014 Elsevier B.V. All rights reserved.
Keywords:
Gear
Wear
Micropitting

1. Introduction and wear on spur gear dynamics. They developed an elastic,


dynamic model with worn teeth having two degrees of freedom.
While the study of wear has been usually focused on the Bajpai et al. [7] proposed a surface wear prediction methodol-
contact between dry surfaces, there has been since the mid 1990s ogy for spur and helical gears, employing a finite elements-based
a flowering of interest on the wear of lubricated gear tooth flanks. gear contact mechanics model to predict wear of contacting tooth
Flodin and Andersson [1] developed a model for wear predic- surfaces. Interestingly, they used direct gear measurements as
tion of spur gears based on a generalized Archard0 s wear equation input to the model in order to analyze gears with actual manu-
[2] and modeling the contact by Winkler0 s elastic foundation factured surfaces with profile and lead modifications.
model. They later developed a computer model for the simulation Kahraman et al. [8] employed a surface wear prediction model
of the wear behaviour in helical gears [3], dealing with tooth for helical gears pairs to investigate the influence of tooth profile
modifications and different profiles. deviations in the form of intentional tooth profile modifications or
Kuang and Lin [4] studied the effect of tooth wear on the manufacturing errors on gear tooth surface wear.
vibration spectrum variation of a rotating spur gear pair by Ding and Kahraman [9] combined a finite elements-based
applying a mathematical model approximating the dynamic char- deformable-body model and a simplified discrete model to study
acteristics of an engaging spur gear pair, the load sharing alterna- the interaction between gear surface wear and gear dynamic
tion, position dependent mesh stiffness, damping factor and response. The influence of worn surface profiles on dynamic tooth
friction coefficient. Their numerical results indicated that the forces and transmission error as well as the influence of dynamic
dynamic load histogram of an engaging spur gear pair may change tooth forces on wear profiles were included.
greatly with tooth wear. Onishchenko [10] developed a wear model of tooth flank
Brauer and Andersson [5] conducted a theoretical study of wear surface, including an elastic dynamic model with four degrees of
in spur gears with interference, using a mixed finite element (FE) freedom and a tooth wear model for boundary lubrication.
and analytical approach. Their results showed that wear of gear Osman and Velex [11] set up a model which incorporates the
tooth flanks may eliminate interference. contribution of tooth wear in quasi-static and dynamic simula-
Wojnarowski and Onishchenko [6] carried out analytical and tions of wide-faced solid spur and helical gears. This model
experimental investigations of the influence of the deformation includes the influence of the lubrication regime on the wear
coefficient, the flank profile deviations with respect to its ideal
geometry and lead modifications.
Park et al. [12] proposed a method for the computation of the
n
Corresponding author. surface wear of hypoid gear pairs, combining Archard0 s wear
E-mail address: jbrandao@inegi.up.pt (J.A. Brandão). model with a semi-analytical hypoid gear contact model.

0043-1648/$ - see front matter & 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.wear.2013.12.025
32 J.A. Brandão et al. / Wear 311 (2014) 31–39

From the brief enumeration above, it can be seen that works During one full revolution of the pinion, a point situated at
dealing with the simulation of wear on lubricated gears have fallen coordinate x on its tooth will then have its height diminished by
into one or both of these main groups: one containing works that Z tE
study the interaction between wear and tooth flank shape, some- ΔhðxÞ ¼ κ pðx; tÞjU 2 ðtÞ  U 1 ðtÞj dt ð3Þ
tA
times including the effect of imperfection and misalignments; and
another containing works that study the interaction between wear where tA is the instant when the tooth first comes in contact with
and dynamic load. None took any account of surface roughness its counterpart on the wheel and tE is the instant when the tooth
and all used “dry” contact pressure distributions, thus essentially ceases contact. Consequently, the depth worn during N turns turns
disregarding the influence of overpressures caused by the rough- of the pinion will be
ness of the tooth flank surfaces.
The present work has a different focus, since it is a continuation hðxÞ ¼ Nturns ΔhðxÞ ð4Þ
of a previous publication that dealt with the simulation of Hence, the volume lost by wear ΔV on all pinion teeth during
micropitting on the flanks of gear teeth [13]. It was there Nturns turns of the pinion is
hypothesized that it is likely that wear plays an important part Z xA
in the evolution of the roughness profile even during FZG tests ΔV ¼ Z 1 b hðxÞ dx ð5Þ
designed specifically to ensure the development of micropitting. xE

Hence, a wear model is applied to simulate the same micropitting where b is the tooth width and Z1 the number of pinion teeth. This
test to try to ascertain the validity of the hypothesis. can be simplified, yielding the following expression:
Morales-Espejel and Brizmer have presented a numerical model Z tE
which includes both the effect of surface contact fatigue and wear
ΔV ¼ Nturns bZ 1 κ F N ðtÞjU 2 ðtÞ  U 1 ðtÞj t ð6Þ
[14], so as to account correctly for the competition between these tA
two phenomena. However, the authors were mostly interested in the
It can be deduced from the previous equation that the precise
application of this model to rolling element bearings, and selected
distribution of pressure in the contact between a pair of teeth has
their operating conditions accordingly: low roughness and low
no influence on overall wear mass loss. On the other hand, the
sliding, at least when compared to typical gear operating conditions.
distribution of load between simultaneously contacting pairs of
In the present case, the focus is on the meshing of real gears, which
teeth is important.
may behave differently from rolling element bearings.
The numerical model illustrated by Fig. 2 is suggested naturally
by Eqs. (2)–(6), provided that the following assumptions hold:

2. Model for wear on spur gear teeth  The problem is two-dimensional and the teeth are in plane
strain.
Archard [2] published in 1953 his famous wear law, which  One tooth is representative of all other teeth on the same gear.
describes the wear volume loss due to the sliding contact between  The pressure distribution history of one meshing can be used
flat surfaces: for all subsequent meshings.

ΔV K
¼ FN ð1Þ The simulation of one meshing between a pair of teeth is
S H
performed as follows:
where ΔV is the volume loss, S the sliding distance, K the At discrete time intervals, the surface pressure distribution
dimensionless wear coefficient, H the softer surface0 s hardness pðx; tÞ is computed by making use of the mixed lubrication model
and FN the normal contact load. described in Section 3. Using a discretized version of Eq. (2), the
To use Archard0 s wear law in the more complex case of contact increment in wear depth δh is then computed and added to a
between gear teeth, it must be written in a differential form: running total. After the disengagement of the pinion tooth under
dhðx; tÞ  study (instant tE), the wear depth distribution on one pinion tooth
¼ κ pðx; tÞU 2 ðtÞ  U 1 ðtÞj ð2Þ after one full turn of the pinion ΔhðxÞ has been obtained. Eq. (4) is
dt
then used to compute the total wear depth distribution h(x).
where h is the wear depth, p the contact pressure and κ the wear The wear volume ΔV can be obtained by two methods: by
coefficient (with units of Pa  1), which is presumed to be constant applying Eq. (6), in which case there is no need to perform the
in time and position. The coordinate x is the position on the discrete computations described in the previous paragraph, since
surface of the tooth as shown in Fig. 1 and t is the time coordinate.
U2 and U1 retain the meaning of tangential velocity respectively of
pinion and wheel tooth.

x
0

z
pitch cylinder

Fig. 1. Coordinates on the surface of the pinion tooth flank. Fig. 2. Diagram of the numerical wear model.
J.A. Brandão et al. / Wear 311 (2014) 31–39 33

the wear volume can be computed analytically, or by quadrature of


the wear depth distribution, as in Eq. (5).

3. Mixed lubrication model

The mixed lubrication model was used by Brandão et al. in their


simulation of micropitting on the flanks of gear teeth [13]. It is
described here briefly for the sake of completeness.
Authors often classify the lubrication regime according to the
specific film thickness [15,16], a well known Λ parameter devel-
oped by Tallian [17]:
h0 Fig. 3. Mixed film lubrication model: normal pressure.
Λ¼ ð7Þ
Rq
where h0 is the central film thickness and Rq the combined RMS
roughness of the contacting surfaces. Hence, it appears reasonable where pMIX is the normal contact pressure in mixed film lubrica-
to introduce the simplifying assumption that the fraction of the tion, pEHD the normal contact pressure of an ideally smooth EHL
normal contact load borne by a lubricant film depends mostly on Λ. contact with the same parameters and pBDR the normal contact
In the present case, Rq would be the combined RMS roughness pressure of the rough boundary lubrication problem.
of the pinion and wheel tooth flank profiles. However, some This procedure, illustrated in Fig. 3, and repeated at each time
difficulties present themselves when trying to apply this defini- step can be outlined as follows:
tion. The relative positions of the profiles are constantly changing,
so that it is difficult to accept a static value for Rq. Also, in
computing the RMS roughness of each profile, it is difficult to 1. The local composite roughness, the central film thickness
justify using the usual definition which calls for a sampling length (using Dowson and Higginson0 s formula [18], corrected for
of 4.8 mm: this is an order of magnitude wider than typical inlet shear heating as advised by Gohar [19]), the specific film
contact widths. Why should the roughness conditions far away thickness (Eq. (7)) and the value of the load sharing function
from the contact have any bearing on load sharing? f ðΛÞ are computed.
For these reasons was an alternative definition of the combined 2. Use the Hertzian pressure distribution as an approximation to
roughness developed, to be used only for load sharing computa- the normal contact pressure of an ideally smooth EHL contact
tions: a “local composite roughness”. At each moment, the limits pEHD.
xa and xb of the active contact zone, which must encompass the 3. Compute the pressure distribution of the rough boundary
Hertzian area as well as any point where pressures are nonnull, are lubrication problem pBDR as if there were dry contact by using
determined. A pseudo-RMS roughness Rq1 of the undeformed an algorithm devised by Polonsky and Keer [20].
pinion tooth profile is defined: 4. Combine these pressure distributions into the mixed lubrica-
Z xb tion pressure distribution by applying Eq. (13).
1
R2q1 ¼ ðy  yÞ2 dx ð8Þ 5. Determine the EHL part of the contact shear stress distribution
xb  xa xa
τEHD by applying the simplified thermo-viscoplastic model of
The pseudo-RMS roughness Rq2 of the wheel tooth is similarly full film lubrication described by Brandão et al. [21] assuming
defined. The “local composite roughness” Rq is then computed in the reduced pressure distribution f ðΛÞ  pEHD .
the usual way: 6. Determine the boundary part of the contact shear stress τBDR by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi applying
Rq ¼ R2q1 þ R2q2 ð9Þ
τBDR ¼ μBDR  ½1  f ðΛÞ  pBDR ð14Þ
The function of Λ that represents the fraction of the normal
contact load borne by the lubricant film receives the designation of where μ BDR
is the coefficient of the lubricant under boundary
load sharing function f ðΛÞ. The relations between full film EHL, lubrication, assumed to be independent of operating conditions
boundary lubrication and overall normal contact load can then be and constant in time.
expressed as follows: 7. Compute the mixed lubrication contact shear stress τMIX by
applying
F N ðtÞ ¼ F EHD ðtÞ þ F BDR ðtÞ ð10Þ
τMIX ¼ τEHD þ τBDR ð15Þ
F EHD ðtÞ ¼ f ðΛÞ  F N ðtÞ ð11Þ

F BDR ðtÞ ¼ ð1  f ðΛÞÞ  F N ðtÞ ð12Þ This procedure gives both the contact shear stress and pressure
distribution at a given times step, and hence the full history of
where FN is the normal contact load expressed as a force per unit contact stresses can be obtained by repeating these calculations
length, FEHD is the portion of FN borne by the lubricating film and over each time steps.
FBDR is the portion borne by direct contact between the surfaces.
From this stems the main idea of the model: that the contact
pressure distribution may be conceived to be an interpolation
between the individual solutions of the perfectly smooth EHL 4. Experimental procedure
problem and the rough boundary lubrication one (BDR) and that
each separate problem should contribute a fraction of the total Martins et al. have already given a full account of the tests
normal pressure distribution: simulated here [22]. For this reason, the micropitting test will only
be briefly summarized, retaining only the aspects relevant to the
pMIX ðx; tÞ ¼ f ðΛÞ  pEHD ðx; tÞ þ ð1  f ðΛÞÞ  pBDR ðx; tÞ ð13Þ present work.
34 J.A. Brandão et al. / Wear 311 (2014) 31–39

The tests were conducted on a back-to-back spur gear test rig An oil sample was collected and subsequently analyzed by
with power recirculation: the well known FZG gear test rig [23]. Direct Reading (DR III) and Analytical (FM III) Ferrography.
The tested gears were FZG type C14 spur gears, from DIN 20  The gears were dismounted from the rig, and cleaned by
MnCr5 carburized steel, lubricated with a highly saturated ester ultrasonic bath.
oil. That oil was studied by Brandão et al. as reference E2 with  The pinion was weighed (weighing precision 71 mg).
regard to full film EHL [21] and with regard to mixed and  The flank surfaces of the gears were visually inspected and
boundary lubrication [24]. photographed.
The fully formulated oil E2, some of whose properties are listed in  The gear tooth flank surface roughness and topography were
Table 1, was based on a biodegradable, highly saturated ester base oil. measured.
The additives used were selected for low toxicity and environmental  The gear were remounted into the testing rig.
compatibility without, however, sacrificing gear performance. A
further characteristic of oil E2 is that a very high viscosity ester oil Additionally, the lubricating oil was replaced with unused oil
(1000 cSt at 40 1C) was mixed into the base ester: the base ester oil at load stage transitions (when transitioning from K3 to K6,
totals 90% of the oil volume and the high viscosity ester, 5%. This is for example). When the interruptions occurred within a load
similar to the use of bright stock in mineral oils. stage, the test proceeded with the same used oil.
The operating conditions and testing programme are summar- Because the number of interruptions of each load stage was
ized in Table 2: the gears were submitted to four load stages: fairly modest, their influence was deemed negligible and no
running-in stage K3 and load stages K6, K8 and K9 of increasing comparative testing without intermediate interruptions was
severity. These were further subdivided into load sub-stages by performed.
additional interruptions that allowed the monitoring of gear
evolution during a load stage. As an example, load stage K3 was
divided into 3 unequal load sub-stages that lasted, respectively, 10, 5. Numerical simulation
30 and 50 thousands of revolutions of the driven gear. Load stage
K9, however, was not subdivided. The numerical simulation proceeded from the application of
At the end of each load sub-stage, the following operations the model of Section 2.
were performed: The properties of the oil used in the test were obtained from
previous measurements, already published by Brandão et al.
 [21,24], where the oil is marked as “E2”. Since the calculations
rely on the elastic half-plane simplification, the only mechanical
properties of the gear material relevant to these calculations are its
Table 1
Oil properties. Young Modulus (210 GPa) and its Poisson Ratio (0.3).
At each interruption of the test, a 6 mm long roughness profile
Chemical content (ppm) was taken from the roughness measurement of the pinion tooth
Zn 0 surface; since the tooth surface is approximately 8.5 mm long,
Ca 0
P 300
each roughness profile was extended to the whole tooth length by
S 5500 the simple expedient of performing a “mirror” operation at each
extremity of the profile, taking care that no spurious discontinu-
Biodegradability and toxicity (standards OECD 101, 202, 301 F)
Ready biodegradability (%) Z 60 ities were introduced. Care was taken to ensure that roughness
Aquatic toxicity with profiles in succeeding measurements were aligned with each
Daphnia EL50 (mg/l) 4 100 other. This was also performed for wheel gear tooth surfaces.
Algae EL50 (mg/l) 4 100 One full meshing cycle could in this way be calculated from first
Density at 15 1C ρ15 (kg/m3) 955 contact to disengagement of the pair of teeth.
Kinematic viscosity at 40 1C ν40 (cSt) 114.5 The active zone of the pinion tooth surface was discretized at
Kinematic viscosity at 100 1C ν100 (cSt)
intervals of 1 μm.
17.0
Viscosity index VI 162
A difficulty in applying Archard0 s wear law is in the definition
of the value of the wear coefficient κ. There is little agreement in
the literature about the value to be used in cases of lubricated
Table 2 contact, with authors using wear coefficients ranging from 10  18
Micropitting testing programme (rotational speeds are constant across load stages:
[5] to 10  12 [9].
2250 RPM for the pinion and 1500 RPM for the wheel).
The issue was sidestepped in the present case because we are
Load Oil Contact Hertzian Load kcycles mainly interested in the distribution of volume loss along the
stage temperature load (N) in driven profile more than in the rigorous prediction of the volume loss.
(1C) gear
Since the mass of the pinion was measured after each interruption
Pressure at Half-width at sub-
pitch point pitch point stage of the test, the loss of mass Ml during each sub-stage is also
(MPa) (μm) known. If it is accepted that the density ρ ¼ 7850 kg=m3 of the
gear steel remains constant during the tests, the volume loss ΔV is
K3 80 851 514 75.0 K3-1 10 easily computed:
K3-2 30
K3-3 50 ΔV ¼ Ml =ρ ð16Þ
K6 90 2924 953 139.0 K6-1 100 On the other hand, Eq. (6) gives us a relation between κ and
K6-2 400
K6-3 940
ΔV which can be written, after appropriate manipulation of the
expressions:
K8 90 5073 1256 183.1 K8-1 100
K8-2 400 Ml
κ¼ R tE ð17Þ
K8-3 940 ρN turns bZ 1 tA F N ðtÞjU 2 ðtÞ  U 1 ðtÞj dt:
K9 90 6373 1408 205.2 K9-1 1440
The right-hand side of the expression can be computed
J.A. Brandão et al. / Wear 311 (2014) 31–39 35

analytically before any simulation has taken place, which means  Fig. 6(a) shows the prediction of wear depth at the end of load
that an estimate of κ becomes available at the beginning of the sub-stage K3-3 using as a departure point the roughness at the
simulation. beginning of load sub-stage K3-1 and the corresponding
The method is used for the simulation of load stage K9 as an pressure distribution history.
example of the method and to render the explanation clearer.  Fig. 6(b) shows the prediction of wear depth at the end of load
By weighing of the pinion, it was determined that it had lost sub-stage K6-3 using as a departure point the roughness at the
between 19 and 21 mg in the course of load stage K9. Assuming beginning of load sub-stage K6-1.
that all of the mass loss can be attributed to wear, the wear coefficient  Fig. 6(c) shows the prediction of wear depth at the end of load
is obtained from Eq. (17), so that 4:36  10  18 o κ o4:82 sub-stage K8-3 using as a departure point the roughness at the
10  18 m2 =N. beginning of load sub-stage K8-1.
Two wear simulations are then performed: one for the lower  Fig. 6(d) shows the prediction of wear depth at the end of load
bound of κ and another for the upper bound. At the close of the stage K9 using as a departure point the roughness at the end of
simulation, the wear depth along the tooth flank surface has been load stage K9.
determined, as shown in Fig. 4.
Notice how a basic wear depth curve serves as baseline for Similarly, Fig. 7 contains four sub-figures:
peaks of wear. The base curve corresponds to the EHL (smooth)
portion of the contact pressure and the wear peaks to the pressure  Fig. 7(a) shows in red the prediction of the roughness profile at
spikes of the boundary lubrication part, coinciding with strong the end of load sub-stage K3-3 from the measured roughness
interactions between roughness features on the surface of the profile at the beginning of load sub-stage K3-1, shown in black,
pinion and wheel tooth. Interestingly, Osman and Velex [11], who and compares it to the measured roughness profile at the end
performed similar simulations, except for omitting altogether the of load sub-stage K3-3, shown in blue.
influence of roughness and adding the influence of dynamic loads,  Fig. 7(b) shows in red the prediction of the roughness profile at
obtained wear depth curves whose shape corresponds to that of the end of load sub-stage K6-3 from the measured roughness
the baseline curve in Fig. 4. profile at the beginning of load sub-stage K6-1, shown in black,
The roughness profile of the pinion tooth was measured and compares it to the measured roughness profile at the end
immediately before load stage K9. It is hence possible to predict of load sub-stage K6-3, shown in blue.
the final profile by subtracting the wear depth distribution from  Fig. 7(c) shows in red the prediction of the roughness profile at
the initial roughness profile. Since the roughness profile was also the end of load sub-stage K8-3 from the measured roughness
measured at the end of load stage K9, it is possible to compare the profile at the beginning of load sub-stage K8-1, shown in black,
prediction to the actual measurement. This is illustrated in Fig. 5, and compares it to the measured roughness profile at the end
where three roughness profiles are plotted against the coordinate of load sub-stage K8-3, shown in blue.
x on the surface of the pinion tooth: the black profile is the  Fig. 7(d) shows in red the prediction of the roughness profile at
roughness profile at the beginning of load stage K9; the blue the end of load stage K9 from the measured roughness profile
profile is the roughness profile at the end of load stage K9; and the at the end of load stage K9, shown in black, and compares it to
red profile is the prediction of final roughness profile. the measured roughness profile at the end of load stage K9,
shown in blue.

6. Simulation results Table 3 compiles mass loss and wear coefficient data for the
simulation cases corresponding to Figs. 6 and 7. As an example, the
The process explained in the previous section for load stage K9 first line corresponds to the simulation of the case pictured in
was applied to each load stage. Only a subset of the simulation Fig. 7(a), running-in load stage K3 in its entirety, using a mass loss
results is presented here, but the remaining simulations showed corresponding to the lowest boundary to the weighing precision.
similar results that support the arguments presented here. The second column gives the mass loss for the whole tooth flank
Fig. 6 contains four sub-figures, each corresponding to a surface in mg. The third column gives the mass loss per unit of
different simulation: tooth surface in 10  3 mg=mm2 averaged over the whole tooth

Fig. 4. Depth of wear along a pinion tooth flank during load stage K9.

Fig. 5. Roughness profiles of load stage K9: measured initial roughness profile (black), measured final roughness profile (blue) and predicted final roughness profile (red).
(For interpretation of the references to color in this figure caption, the reader is referred to the web version of this article.)
36 J.A. Brandão et al. / Wear 311 (2014) 31–39

Fig. 6. Depth of wear along a pinion tooth flank: (a) from beginning of K3-1 to end of K3-3; (b) from beginning of K6-1 to end of K6-3; (c) from beginning of K8-1 to end of
K8-3; and (d) from beginning of K9 to end of K9.

Fig. 7. Variation of roughness profile of a pinion tooth flank: (a) from beginning of K3-1 to end of K3-3; (b) from beginning of K6-1 to end of K6-3; (c) from beginning of K8-1
to end of K8-3; and (d) from beginning of K9 to end of K9. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this
article.)
J.A. Brandão et al. / Wear 311 (2014) 31–39 37

surface. The fourth column gives the mass loss per unit of tooth In a similar way, Table 4 gives information on the evolution of
surface in 10  3 mg=mm2 averaged over the portion of the tooth the roughness of a pinion tooth profile for simulations of complete
surface situated between the dedendum and the pitch line. The load stages by listing the roughness parameters RZdin, Ra and Rq.
fifth column gives the mass loss per unit of tooth surface in The first line gives roughness parameters of the roughness profile
10  3 mg=mm2 averaged over the portion of the tooth surface measured at the beginning of load stage K3, which corresponds to
situated between the addendum and the pitch line. The sixth the beginning of load sub-stage K3-1. The second line corresponds
column gives the wear coefficient κ computed from Eq. (17). to the roughness profile measured at the end of load stage K3,
The second line of the table gives the same information for load which is also the end of load sub-stage K3-3. The third line
stage K3 but using the upper boundary to the weighing precision. corresponds to the prediction of the roughness profile at the end
Subsequent lines repeat this pattern for load stages K6, K8 and K9. of K3 using the lower bound of the mass loss range. The fourth line
corresponds to the prediction of the roughness profile at the end
of K3 using the upper bound of the mass loss range. Subsequent
Table 3
lines follow the same pattern, but applied to load stages K6, K8
Mass loss and wear coefficient. and K9.

Load sub-stage Mass loss (mg) Mass loss rate κ


10  3 mg=mm2 10  18 m2 =N
7. Discussion of the wear simulation results
tot bpp app
Throughout this work, the part that plastic deformation has
K3 predicted-low 3 1.66 2.53 1.24 82.5 taken in shaping the gear tooth surface profiles has been dis-
K3 predicted-high 5 2.77 4.21 2.07 137 regarded. This may seem untenable since the elastic calculations
K6 predicted-low 13 7.18 10.9 5.38 6.50 performed here give contact pressures that sometimes reach
K6 predicted-high 15 8.28 12.6 6.21 7.50 values so high as 7 GPa. However, there are reasons to believe
K8 predicted-low 10 5.50 8.33 4.14 2.88 that little of the profile changes result from plastic deformation in
K8 predicted-high 12 6.60 10.0 4.96 3.46 the present case.
K9 predicted-low 19 10.4 15.8 7.85 4.36 The surface hardness of the case hardened gears is approxi-
K9 predicted-high 21 11.5 17.5 8.68 4.82 mately 700 HV. According to Pavlina and Van Tyne [25], yield
strength can be evaluated from hardness using the formula:
sY ðMPaÞ ¼  90:7 þ 2:876 HV ¼ 1920 MPa ð18Þ
Table 4
The Von Mises yield stress can then be estimated as k ¼
Roughness.
0.577  1920 ¼1100 MPa.
Load sub-stage RZdin (μm) Ra (μm) Rq (μm) Johnson [26] studied surfaces submitted to a moving Hertzian
stress distribution in linear contact. He found that the surface
K3 initial 7.65 1.60 1.98
remained elastic for maximum Hertzian pressures p0 o 3k; and
K3 final 6.57 1.38 1.73
K3 predicted-low 7.40 1.55 1.90
that the surface would shakedown elastically for 3k o p0 o 4k. For
K3 predicted-high 7.49 1.51 1.87 the current case, the range of elastic shakedown is estimated as
3:3 GPa o p0 o 4:4 GPa.
K6 initial 6.57 1.38 1.73
K6 final 7.03 1.79 2.24 If we accept that a pressure spike acts more or less like a very
K6 predicted-low 6.47 1.44 1.87 narrow, linear Hertzian pressure distribution over a very small part
K6 predicted-high 6.67 1.48 1.92 of the surface, it is seen that only pressure spikes above 3.3 GPa
K8 initial 7.03 1.79 2.24 should cause any plastic deformation. Moreover, the influence of
K8 final 6.06 2.93 3.24 the pressure spike should only extend to a depth approximately
K8 predicted-low 6.69 2.15 2.56 equal to its width.
K8 predicted-high 6.66 2.23 2.63
Pressure spikes above the 3.3 GPa limit are indeed calculated
K9 initial 6.06 2.93 3.24 in some instances, although they are not very widespread. An
K9 final 4.56 4.06 4.20
example of an instantaneous contact pressure distribution calcu-
K9 predicted-low 5.65 3.79 4.02
K9 predicted-high 5.64 3.89 4.11 lated for load stage K9 in a particularly battered area of the surface
is shown in Fig. 8. The whole contact pressure distribution is

Fig. 8. (a) Snapshot of the contact pressure distribution; and (b) detail of the preceding pressure distribution around its highest pressure peak.
38 J.A. Brandão et al. / Wear 311 (2014) 31–39

shown in Fig. 8(a) and a portion of the pressure distribution have appeared to be safely separated in the simulation. This means
around the highest peak is shown in Fig. 8(b) with a different that it would be excessively optimistic to expect the predicted
horizontal scale. The pressure peaks appear jagged because their roughness profiles to coincide precisely with measurements.
width is only a few micrometers and the distance between two However, one would expect the overall shape of the predicted
discretization grid nodes is 1 μm. In particular, the highest roughness profile to be similar to that of the measured one.
pressure peak, which reaches 7 GPa, only has a width of approxi- With that in mind, there seems to be remarkable agreement
mately 5 μm. Its influence on the stress fields therefore only between the predicted and the measured profiles of Fig. 7. How-
extends to a depth of 5 μm. Remember that this particular ever, it should also be pointed out that the wear model fails to
pressure spike occurs only on this particular spot and for a limited predict micropitting, which can be observed in the profiles. As an
time: an instant after the time at which this “snapshot” is taken, example, Fig. 7(b) shows that a micropit appeared near position
the pressure distribution will have changed because the opposing x¼ 0.5 mm and another at x ¼  0.5 mm, micropits that the wear
roughness features that came together to create the pressure spike simulation failed to predict.
will have separated again. Fig. 6 show that wear is significantly more severe in the portion
Fig. 7 shows that the variation in tooth profile is in many cases of the tooth surface below the pitch line of the pinion tooth, which
greater than 1 μm in depth. It is difficult to believe that the correspond to positive values of the x coordinate, than above it.
occasional pressure peaks above 3.3 GPa could account exclusively This is in agreement with observed facts.
for this change. Taking the example of Fig. 8(b), this would mean In Table 3, it can be observed that the values of the wear
that accumulated residual elastic deformation and plastic defor- coefficient κ must vary significantly if the predicted mass loss is to
mation causing 1 μm of surface displacement would have to fit the measured mass loss. It is not certain that κ must remain
develop over 5 μm of depth. constant under varying operating conditions; indeed, considering
Popescu et al. [27] measured the residual indentation left by a the range of values reported in the literature, this seems unlikely.
6.35 mm ceramic ball pressed repeatedly against a hardened SAE In addition, normal wear must not account for the totality of mass
52100 plate. The circular Hertzian pressure distribution that loss: there must be a balance between the predictions of the wear
resulted reached a maximum of 5.1 GPa. Under these conditions, model and those of the micropitting model. In any case, it is
the residual indentation left was 0:7 μm. This loading is more obvious that similar wear coefficients are computed for load stages
severe than the 7 GPa of Fig. 8(b) pressure spike with regard to K6, K8 and K9; while the wear coefficient computed for the
plasticity: the pressure above the elastic limit is applied on a much running-in load stage K3 is much bigger. This shows the need
larger area, which results in a more widespread and deeper for a running-in stage.
volume affected by plastic deformation. However, less than 1 μm Another interesting aspect is that the roughness parameters of
of residual indentation is left after unloading. Table 4 seem to fail to capture similarities that become evident
Notice that plastic deformation is acting concurrently to wear when observing graphical representation of the profiles. A case in
damage in the experiment studied here: the same material which point is that of Fig. 7(d), where there is remarkable similarity
could deform plastically is being removed by wear. All of this gives between prediction and measurement of the roughness profile, a
some assurance that, while there must certainly have been some similarity that is not reflected in the roughness parameters of the
plastic deformation caused by high contact pressures, its influence corresponding four last lines of Table 4.
on the evolution of the profiles in the present test must be small At this point, it is possible to return to the issue mentioned in
when compared to that of wear. However, this points to an the introduction to the present work. Brandão et al. presented a
interesting subject for future research. simulation of surface rolling contact fatigue (RCF) of this same
It could also be argued that the high pressure spikes might in micropitting test [13], assuming that the mass lost by the pinion
some instances also lead to roughness cavitation. The present would correspond to the removal of the gear material affected by
mixed lubrication model is quite incapable of reproducing this surface RCF. The evolution of the pinion tooth flank surface profile
phenomenon. However, this is not seen as a serious problem could then be estimated by removing the affected areas. Fig. 9
because the wear model employed here is a phenomenological shows partially the results of this approach: the roughness profile
model which accommodates many basic types of wear mechan- evolution of a small portion of the pinion tooth profile near the
ism, including erosion by cavitation. root of the tooth. It was commented [13] that the roughness
It is useful at this point to discuss briefly the measured rough- evolution predicted by RCF simulation alone is unlike the one
ness profiles. The profiles were measured with a stylus profil-
ometer and it must be understood that the pinion and wheel were
positioned manually with regard to the measurement device, and
that they were measured at the end of each load sub-stage.
Consequently, the orientation and position of the measured device
with regard to the profilometer changed from measurement to
measurement. This is significant because a gear tooth surface has a
large curvature that must be filtered out by means of a Gaussian
filter: if the tooth is not measured in rigorously the same way,
profiles taken from successive measurements will appear slightly
deformed versions of the previous one. Additionally, the axial
position of the profile cannot be guaranteed, so that some slight
variation of the roughness must be expected.
Another aspect that accrues from the uncertainty in the
positioning of the roughness profiles is that it is nearly impossible
to know with certainty the relative position of the roughness
profiles on a pair of tooth which will mesh during the test. A
relative translation of 50 μm may cause the pressure history to be
significantly different, since some roughness features will appear Fig. 9. Roughness profile evolution predicted by a surface rolling contact fatigue
to collide when they did not, while others that did collide may simulation.
J.A. Brandão et al. / Wear 311 (2014) 31–39 39

observed experimentally, it was then hypothesized that wear [2] J. Archard, Contact and rubbing of flat surfaces, J. Appl. Phys. 24 (8) (1953)
might explain the discrepancy. 981–988.
[3] A. Flodin, S. Andersson, Simulation of mild wear in helical gears, Wear 241 (2)
Contrasting Fig. 9 with Fig. 7(b), both dealing with the simula- (2000) 123–128 http://dx.doi.org/10.1016/S0043-1648(00)00384-7.
tion of the same load stage K6, it is obvious that the RCF [4] J.H. Kuang, A.D. Lin, The effect of tooth wear on the vibration spectrum of a
simulation seems to predict too many micropits, giving rise to an spur gear pair, J. Vib. Acoust. 123 (3) (2001) 311–317 http://dx.doi.org/10.1115/
1.1379371.
unrealistic evolution of the surface profile; while, on the other
[5] J. Brauer, S. Andersson, Simulation of wear in gears with flank interference a
hand, the wear simulation produces a mostly correct overall mixed fe and analytical approach, Wear 254 (11) (2003) 1216–1232.
evolution of the roughness profile without, however, being cap- [6] J. Wojnarowski, V. Onishchenko, Tooth wear effects on spur gear dynamics,
able of predicting micropits. Mech. Mach. Theory 38 (2) (2003) 161–178 http://dx.doi.org/10.1016/
S0094-114X(02)00091-5.
This seems to support the hypothesis that most mass loss (and [7] P. Bajpai, A. Kahraman, N.E. Anderson, A surface wear prediction methodology
roughness alteration) is caused by mild wear, but a wear model for parallel-axis gear pairs, J. Tribol. 126 (3) (2004) 597–605 http://dx.doi.org/
alone is not capable of accounting for the appearance of micropits: 10.1115/1.1691433.
[8] A. Kahraman, P. Bajpai, N.E. Anderson, Influence of tooth profile deviations on
an integration of RCF simulation and wear simulation must be the helical gear wear, J. Mech. Des. 127 (4) (2005) 656–663 http://dx.doi.org/10.
way to correctly predict surface evolution. 1115/1.1899688.
[9] H. Ding, A. Kahraman, Interactions between nonlinear spur gear dynamics and
surface wear, J. Sound Vib. 307 (3–5) (2007) 662–679 http://dx.doi.org/10.
8. Conclusion 1016/j.jsv.2007.06.030.
[10] V. Onishchenko, Tooth wear modeling and prognostication parameters of
engagement of spur gear power transmissions, Mech. Mach. Theory 43 (12)
A simulation of wear in a real micropitting test was performed. (2008) 1639–1664 http://dx.doi.org/10.1016/j.mechmachtheory.2007.12.005.
The test was composed of four load stages of increasing load, the [11] T. Osman, P. Velex, Static and dynamic simulations of mild abrasive wear in
wide-faced solid spur and helical gears, Mech. Mach. Theory 45 (6) (2010)
first three of which were also periodically interrupted to monitor
911–924 http://dx.doi.org/10.1016/j.mechmachtheory.2010.01.003.
the gear tooth flank surfaces and lubricating oil alterations during [12] D. Park, M. Kolivand, A. Kahraman, Prediction of surface wear of hypoid gears
load stages, thus dividing each load stage into sub-stages. using a semi-analytical contact model, Mech. Mach. Theory 52 (0) (2012)
The meshing of one pinion tooth flank with one wheel tooth 180–194 http://dx.doi.org/10.1016/j.mechmachtheory.2012.01.019.
[13] J. Brandão, R. Martins, J. Seabra, M. Castro, Surface damage prediction during
flank was simulated for each load sub-stage. The simulation an FZG gear micropitting test, in: Proceedings of the Institution of Mechanical
consisted in applying a model that includes a mixed film lubrica- Engineers, Part J: J. Eng. Tribol., vol. 226 (12) (2012) 1051–1073. http://dx.doi.
tion model and in using a differential form of Archard0 s wear law. org/10.1177/1350650112461879.
[14] G.E. Morales-Espejel, V. Brizmer, Micropitting modelling in rolling sliding
The analysis of the load sub-stages has shown that much of the contacts: application to rolling bearings, Tribol. Trans. 54 (4) (2011) 625–643
mass loss in the test could be attributed to wear, but that a fatigue arxiv:www.tandfonline.com/doi/pdf/10.1080/10402004.2011.587633, http://
model is still needed to account for the micropits, which did not dx.doi.org/10.1080/10402004.2011.587633.
[15] H.A. Spikes, Mixed lubrication an overview, Lubr. Sci. 9 (3) (1997) 221–253
appear in the simulation. Hence, it is necessary to use both a wear
http://dx.doi.org/10.1002/ls.3010090302.
and a fatigue model to account fully for the alterations of a pinion [16] P. Vergne, Super low traction under EHD and mixed lubrication regimes, in:
tooth profile during a micropitting test. A. Erdemir, J.-M. Martin (Eds.), Superlubricity, Elsevier BV, 2007, pp. 429–445.
While the simulation of a micropitting test has given valuable [17] T.E. Tallian, On competing failure modes in rolling contact, ASLE Trans. 10 (4)
(1967) 418–439.
insights into the influence of wear on tooth profile alterations, [18] D. Dowson, G.R. Higginson, A numerical solution to the elastohydrodynamic
more tests would be needed to ascertain the validity of the present problem, J. Mech. Eng. Sci. 1 (1959) 6–15.
wear model, tests with better measurement precision, particularly [19] R. Gohar, Elastohydrodynamics, Ellis Horwood, Chichester, 1988.
[20] I.A. Polonsky, L.M. Keer, A numerical method for solving rough contact
in weighing the lost mass. problems based on the multi-level multi-summation and conjugate gradient
techniques, Wear 231 (2) (1999) 206–219.
[21] J.A. Brandão, M. Meheux, J.H.O. Seabra, F. Ville, M.J.D. Castro, Traction curves
Acknowledgments and rheological parameters of fully formulated gear oils, in: Proceedings of the
Institution of Mechanical Engineers, Part J: J. Eng. Tribol., vol. 225 (7) (2011)
577–593. http://dx.doi.org/10.1177/1350650111405111.
The present work is funded by the European Regional Devel- [22] R. Martins, C. Locatelli, J. Seabra, Evolution of tooth flank roughness during
opment Fund (ERDF) through the ‘COMPETE – Competitive Factors gear micropitting tests, Ind. Lubr. Tribol. 63 (1) (2011) 34–45 http://dx.doi.org/
Operational Program’ and by Portuguese Government Funds 10.1108/00368791111101821.
[23] H. Winter, K. Michaelis, FZG gear test rig—description and possibilities, in:
through ‘FCT – Fundação para a Ciência e Tecnologia’ as part of Coordinate European Council Second International Symposium on the Perfor-
mance Evaluation of Automotive Fuels and Lubricants, 1985.
 project ‘High efficiency gears and lubricants for wind mill [24] J.A. Brandão, M. Meheux, F. Ville, J.H. Seabra, J. Castro, Comparative overview
of five gear oils in mixed and boundary film lubrication, Tribol. Int. 47 (0)
planetary gearboxes’, reference number ‘PTDC/EME-PME/
(2012) 50–61 http://dx.doi.org/10.1016/j.triboint.2011.10.007.
100808/2008’, for the experimental work; [25] E. Pavlina, C. Van Tyne, Correlation of yield strength and tensile strength with
 project ‘Projecto Estratégico – LA 22 – 2011–2012’, reference hardness for steels, J. Mater. Eng. Perform. 17 (6) (2008) 888–893.
number ‘Pest-OE/EME/LA0022/2011’, for the rolling contact [26] K. Johnson, A shakedown limit in rolling contact, in: Proceedings of the Fourth
National Congress of Applied Mechanics, American Society of Mechanical
fatigue simulation; Engineers, Berkeley, California, 1962, pp. 971–975.
 project ‘Gear transmissions of high tribological efficiency and [27] G. Popescu, G.E. Morales-Espejel, B. Wemekamp, A. Gabelli, An engineering
reliability’, reference number ‘EXCL/EMS-PRO/0103/2012’, for model for three-dimensional elastic-plastic rolling contact analyses, Tribol.
Trans. 49 (3) (2006) 387–399 http://dx.doi.org/10.1080/05698190600678739.
the wear simulation.

References

[1] A. Flodin, S. Andersson, Simulation of mild wear in spur gears, Wear 207 (1)
(1997) 16–23.

Vous aimerez peut-être aussi