Professional Documents
Culture Documents
Spalling Formation Mechanism For Gears
Spalling Formation Mechanism For Gears
net/publication/222191455
CITATIONS READS
112 9,553
2 authors, including:
Yan Ding
RMIT University
40 PUBLICATIONS 353 CITATIONS
SEE PROFILE
All content following this page was uploaded by Yan Ding on 16 October 2018.
Abstract
Though the basic phenomenon of wear on gear tooth contact surfaces is the removal of a piece of material from the working surface, the
sizes of the wear debris may be different, due to the different physical causes in their formation processes. No common definitions have
been established to distinguish spalling from pitting in the literature. This is probably due to the fact that the physical causes of pitting
and spalling have not yet been established. In this paper, a brief literature review is presented with the intention to differentiate spalling
from pitting. Three types of wear phenomena are defined. Furthermore, the results of a recent experimental study of gear tooth spalling
formation in AISI 4340 gears in a test rig are also presented to demonstrate a possible process of spalling due to the development of cracks
beneath the tooth contact surfaces and crack linkages in plastically-collapsed metal ligament between the crack tip and the adjacent tooth
contact surface. These experimental results substantiated the ligament collapse spalling mechanism proposed by Ding et al. [Y. Ding, R.
Jones, B.T. Kuhnell, Elastic–plastic finite element analysis of spall formation in gears, Wear 197 (1996) 197–205; M. Heems, F. Lagarde,
R. Courtel, P. Sorin, C. R. Acad. Sci. 257 (3) (1963) 3293].
© 2003 Published by Elsevier Science B.V.
Keywords: Spalling; Subsurface cracks; Crack failure; Ligament collapse
to pitting, spalling results in more rapid deterioration of from an inclusion or a hard particle, as well as material
surface durability, and often induces early failure by se- defects such as flaws and pre-existing dislocations. After
vere secondary damage. It has been often reported as the a crack initiates, the cyclic contact stress produced by re-
most destructive surface failure mode of a gear (e.g. [4]). peated rolling or rolling–sliding contacts drives cracks to
A detailed understanding of spalling is not only essen- grow. Eventually, failure of cracks will take place and result
tial for detecting incipient gear failure, before it becomes in some kind of surface contact damage, such as spalling,
disastrous and causes expensive secondary damage, but pitting, or delamination. In general, the study of surface con-
is also central for planning preventative maintenance for tact damage reported in literature may be divided into two
gear components. Furthermore, it can also be used to deter- groups: surface-defect-origin contact fatigue and subsurface-
mine the critical parameters for designing gears that satisfy defect-origin contact fatigue, which are discussed in follow-
special requirements. One instance is the design of gears ing subsections.
with the maximum capacity and the minimum size and
weight, for high-speed automation or aerospace power trans- 2.1. Existing theories of surface-defect-origin
missions. contact fatigue
Although there has been considerable research devoted
to this area over the last half-century, there is still lack of For decades, there has been a dominant opinion in lit-
a thorough understanding of the spalling formation process erature regarding spalling/pitting formation, which consid-
on gears. Most of the relevant previous work concentrated ered that, as surface cracks driven by a liquid lubricant,
on symmetrical contact problems. The knowledge of sub- spalling/pitting was the result of surface crack propagation
surface crack behaviour with non-symmetrical contact ge- in the direction of contact travel. This opinion was based on
ometry is very limited. The majority of the previously re- Way’s hypothesis [7], which postulated that lubricating oil
ported research has focused on either metallurgical studies pre-filled in a surface crack was trapped between the crack
of the worn material, or on the theoretical analysis of the be- surfaces when the approaching contact reached the surface
haviour of subsurface cracks. There appears little published opening and pinched the crack closed. As a result, the crack
work aimed to achieve a mutual agreement from the results tip was extended by a hydraulic pressure of the lubricant oil
of these two aspects of the research. Neither much study sealed between the crack surfaces. The extensively quoted
was focused on the stress state change of the material be- Way’s hypothesis was not challenged until a half-century
tween the contact surface and subsurface crack tips, which later. Keer and Bryant [8] simulated Way’s experiments and
is of great practical interest in regard to the physical pro- found that the dominant mechanism for surface-breaking
cess of spalling formation. The objective of this research is crack growth was mode II (shear) propagation. Their con-
to investigate the spalling mechanism in gears by studying clusion is not consistent with Way’s assumption of Mode I
the behaviour of subsurface cracks. The behaviour of sur- (tension) crack propagation. However, they also suggested
face cracks under contact loading is controlled by a different that the phenomenon of crack curving, i.e. changing its di-
mechanism and needs to be studied separately. rection to the contact surface to form a crater or cavity, could
be due to the influence of tensile rather than shear stress.
Bower [9] performed a fracture mechanics analysis of crack
2. Previous research on crack propagation and failure propagation in the presence of lubricating oil. His results
did not appear to support Way’s hypothesis, either. Further-
Three stages are involved in the formation process of more, the experimental results obtained by Cheng et al. [10]
a spall: crack initiation, crack propagation and failure. It showed that the surface crack growth was very slow and
has been well documented that crack initiation is resulted the depth of the slowly growing surface cracks was in the
Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317 1309
range of 5–10 m. Therefore, it is almost impossible to form 2.3.1. Surface crack propagation is most likely
a spall cavity from surface crack growth, unless there is a to result in pitting or delamination only
near-surface inclusion to act as a bridge to significantly ac- Despite the controversy over whether Mode I or II is the
celerate the propagation of the surface crack into the sub- dominant mode for surface crack propagation, it can be ar-
surface region. gued that, if Mode I growth of a surface crack is the domi-
nant mode, the crack would extend along the direction of the
2.2. Existing theories of subsurface-defect-origin maximum shear stress at the crack tip. This direction could
contact fatigue be determined according to the initial inclined angle of the
surface crack. When the initial angle of the crack is equal
Many researchers have also studied the behaviour of to or less than 45◦ , the surface crack would propagate either
subsurface cracks under contact loads for the purpose of un- parallel to the surface or at an angle towards the contact sur-
derstanding of spalling/pitting mechanisms. Among them, face. As a result, a very shallow pit would form. When the
Fleming and Suh [11] were the first ones who utilised initial inclined angle of the surface crack is greater than 45◦ ,
fracture mechanics as a method to analyse the propagation the direction of the surface crack would be downwards into
of subsurface cracks parallel to the contact surface. Their the bulk of the material. On the other hand, if the Mode II
results showed that the stress intensity factors (SIFs) for growth were the dominant mode, a surface crack would ex-
Modes I and II were very low. Kaneta et al. [12] studied tend along the direction of the surface crack line itself. Both
the growth mechanism of subsurface cracks by numerically analytical and experimental work discussed in Section 2.1
analysing the behaviour of a three-dimensional subsurface have demonstrated that a surface crack alone is unable to
crack parallel to the contact surface. They concluded that penetrate into the bulk of material by propagation [10,15,16],
the propagation of subsurface cracks is mainly by Mode II. regardless which mode dominates the growth. Furthermore,
They also found that the direction of the Mode II propa- when surface roughness is considered, surface crack growth
gation was different for the trailing and leading tips of the may end up by cutting off asperities, resulting in delamina-
crack with respect to the direction of the surface traction. tion. Thus, it can be concluded that surface crack growth in
The tip at the trailing side of the traction extends along the both Modes I and II only results in pitting or delamination.
crack line, and the tip at the leading side extended towards
the surface at an angle less than 6.2◦ . Sin and Suh [13] stud- 2.3.2. Though subsurface cracks propagate in Mode II,
ied the mechanism of subsurface crack propagation in slid- they do not appear to fail by crack propagation
ing elastic–plastic solids using the finite element analysis. As the general perception shown in the literature, spalling
They suggested that crack tip sliding displacement (CTSD) and pitting are assumed to be the result of propagation of
be taken as the propagation rate of subsurface cracks. They a crack until it reaches a critical size. O’regan et al. [17]
also suggested that the direction of the propagation be the defined the critical size as a/d = 1, where a is the half
direction of the maximum shear stress in front of the crack crack length and d the depth of the crack from the con-
tip and along an angle between −5 and 5◦ to the crack line, tact surface. However, the literature discussed previously has
implying that the propagation was almost parallel to the demonstrated that subsurface crack growth is very slow. Ac-
contact surface. They claimed that, based on the CTSDs cording to the concept of brittle fracture, the SIFs for Modes
calculated using FEA, their predicted wear rates com- I and II, obtained by Fleming and Suh [11], were too small
pared favourably with the experimentally determined wear to be the driving force for subsurface crack growth. The re-
rates. sults of Kaneta et al. [12] showed that, for the case of low
More recently, Ding et al. [14] studied the behaviour of coefficients of friction between contact surfaces, i.e. 0 <
subsurface cracks beneath the pitch line of a gear tooth with a f < 0.1, the values of SIFs in Mode II, KII , were just in the
view to develop a fundamental understanding of the spalling threshold stress intensity range: KII,th ∼ = 1.5 MPa m1/2 .
mechanism in gears. Using the finite element method, they These results indicated that the growth of a subsurface crack
analysed the potential modes of crack propagation and fail- would be very slow, particularly for the case of low fric-
ure, and found that the values of the stress intensity factors tion contact, such as gears. The results of CTSD obtained
of the subsurface cracks were below the critical SIF, Kc . Sin and Suh [13] also implied a very slow growth of sub-
Consequently, ligament collapse at crack tips as the cause surface crack propagation. Besides, the CTSD criterion for
of spalling from subsurface cracks was hypothesised. Ding crack propagation in elastic–plastic solids proposed by Sin
et al. [1] also performed an elastic–plastic finite element and Suh [13] and Bastias et al [18] lacks support accord-
analysis that further evaluated the hypothesis as the failure ing to available data for such complex loading conditions.
mechanism of spalling in gears. Moreover, their work only suites the cases of high surface
traction (f ≥ 0.25), such as in abrasive or adhesive sliding
2.3. Discussion on crack propagation theories wear, but not for the case of gear tooth contact in which
f ∼= 0.06, since the influence of surface traction on subsur-
The literature presented above inspired the discussion on face crack propagation is detrimental [19]. Furthermore, the
the following observations. direction of the crack propagation reported in the literature
1310 Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317
Table 1
The maximum values of SIF ( KII ) at subsurface crack tips [1]
d/c a/c = 0.11 a/c = 0.33 a/c = 0.5 a/c = 0.67 a/c = 0.89
Trailing tip Lead tip Trailing tip Lead tip Trailing tip Lead tip Trailing tip Lead tip Trailing tip Lead tip
0.177 3.9677 3.925 6.950 7.199 13.402 18.050 8.863 5.818 13.922 10.789
0.237 3.790 3.624 10.657 7.670 11.198 10.763 8.608 7.042 12.009 15.711
0.296 3.823 4.469 6.810 4.314 15.124 8.516 16.174 12.602 11.345 12.880
0.355 4.571 4.236 6.494 1.853 19.926 11.309 7.549 5.141 8.796 7.372
0.414 1.427 1.595 5.840 1.643 15.669 9.793 11.707 7.284 9.602 6.526
0.532 1.239 1.177 4.085 2.504 13.541 9.954 – – 9.727 6.992
0.887 1.393 0.996 – – – – – – – –
was almost parallel to the contact surface. This does not ex- surface had failed. This was done using the following equa-
plain the orientations of the walls of a spall cavity. Although tion:
ysurface
Miller [20] recognized the incapability of Mode II propaga- ytip σe dy
tion resulting in spalling and attempted to explain such fail- σm = (1)
ysurface − ytip
ure as subsurface crack branching in Mode I, the tendency
of crack branching shown in Miller’s results did not appear where σ m is the mean stress in a ligament region, σ e the
to confirm this theory. Using a realistic asymmetrical gear von Mises stress caused by the contact load, the ligament
tooth FE model, Ding et al. [1,14] investigated the behaviour length: (ysurface −ytip ) is the shortest distance from the crack
of subsurface cracks under a gear tooth in contact with a tip to the contact surface. Comparing σ m to the material
heavy contact load and low surface traction. They calculated ultimate strength, σult = 1207 MPa, it was found that the
the stress intensity factors for both Modes I and II of the sub- material in the ligament (ysurface − ytip ) failed during a load
surface cracks beneath a gear tooth surface with normalised cycle for most of the crack cases investigated, and that the
size: a/c = 0.89, 0.67, 0.50, 0.11, at normalised depth: crack with size of a/c = 0.5 at the depth of d/c = 0.355
d/c = 0.118, 0.177, 0.237, 0.296, 0.355, 0.414, 0.532, 0.887 was the most severe case. Based on these results, ligament
(where a is the half crack length, c the half contact width, and collapse spalling mechanism for gears was hypothesised as
d the depth of the crack from the contact surface). The high- the failure mode of subsurface cracks resulting in spalling.
est value of KI obtained was 4.54 MPa m1/2 , which did
not exceed the threshold value. On the other hand, the stress 3.2. Experimental apparatus
intensity factor range in Mode II, KII , was more appeal-
ing. As shown in Table 1, the values of KII are generally In order to verify Ding’s hypothesis, an experimental in-
greater than the threshold value KII,th . However, the high- vestigation was conducted in the Department of Mechan-
est value of KII is 19.9 MPa m1/2 , less than the suggested ical Engineering, Monash University, Australia. The aims
critical stress intensity of crack failure Kc ∼ = 20 MP m1/2 of the experimental work were to investigate the important
by Blake and Cheng [21]. It should be noted here that this phenomenon and interdependencies, as well as to examine
value of Kc is conservative. For gear material, the value of experimental evidence to test the proposed gear ligament
Kc should be significantly greater. collapse spalling mechanism. Fig. 2 shows the schematic di-
In summarising the findings from literature, it can be con- agram of the experimental apparatus of the single stage spur
cluded that (1) surface crack propagation only results in gear test rig.
pitting or delamination; (2) spalls are developed from subsur- As indicated in Fig. 2, the hydraulic gear pump provided
face cracks; (3) the growth of a subsurface crack is in Mode input load for the gear set. The load control system included
II; and (4) a subsurface crack beneath a gear tooth contact a pressure control valve, a pressure relief valve, two pressure
surface does not fail by crack propagation. In another word, transducers, and amplifiers. The input load was controlled
spalling is not caused by the failure of subsurface cracks. by adjusting the pressure control valve, and monitored con-
stantly by the two pressure transducers. The pressure re-
lief valve installed between the pressure transducers and the
control valve for safety.
3. Ligament collapse spalling mechanism The calibration for the reading of the amplifier meter Re ,
in terms of the input hydraulic pressure pinput of the pump,
3.1. The hypothesis was carefully conducted, and the relation between pinput and
Re was obtained using the straight line of best fit:
In searching for the possible cause of spalling, Ding et al.
pinput (psi) = 2.43 + 420.29Re (2)
[1] used a general fracture mechanics based approach to
investigate the subsurface crack failure by checking if the Consequently, the relation between the input hydraulic pres-
remaining ligament between the crack tip and the contact sure pinput of the pump and the contact load F on gear tooth
Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317 1311
surfaces was derived: of heavily loaded contacts, as occur in hypoid gear drives,
EP (extreme pressure) additives are adopted. However, for
F (N) = 43.1 + 0.76pinput (3)
this study the lubricant was chosen to be adequate in that it
The lubricating system consisted of an oil pump, an oil filter would provide a lubrication film in the elasto-hydrodynamic
(3 m pore size), an oil relief valve, an oil jet feeder, and a regime, but in view of the need in minimising the test time,
large oil tank. The oil circulated by the oil pump was first it was not chosen to provide significant increase in life of the
pumped through the oil filter, then injected directly onto the gears. The test conditions were deemed to be fair as far as the
meshing area of the gear teeth through the oil feeder situated tooth spalling phenomenon was concerned and subsequent
on the top of the gear box as shown in Fig. 3. The oil, then, observation proved that the spalling of the test gear teeth was
flowed back to the large oil tank where it was cooled down typical of that found in heavily loaded in-service gear drives.
and then re-circulated. In order to keep the oil temperature
within the room temperature range, the ratio of the oil vol- 3.3. Gear sets
ume in the gearbox to that in the tank was 1:20. The brand of
the lubricating oil was Spirax hd80w/90, which was additive Three pairs of spur gear sets, identical in dimensions,
free. It should be noted that gear lubricants often contain sev- manufacturing process and material, were used in this in-
eral additives to improve gear life. For example, in the cases vestigation. Fig. 4 shows one pair of them. The dimension
Fig. 3. The gear transmission and lubrication system with the oil jet feeder.
1312 Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317
Table 2
Dimension parameters of the tested gear sets
Gears Pitch diameter (mm) Teeth num. (z) Diametral pitch (z/D) Base Diameter (mm) Outer diameter (mm)
parameters of the gears are given in Table 2. They were a filtergram analysis indicated that there were no particles
generated from AISI 4340, a typical through-hardening, an- larger than 3 m in the lubricant. In order to examine the
nealed steel for gears, of which the material property is spalling progress on the same teeth at the end of each op-
given in Table 3. The surface finish of the gear teeth was be- eration step, three teeth of each gear were chosen randomly
low 1.6 m roughness, which is the typical smoothness of and labelled A, B, and C, respectively, as shown in Fig. 4.
tooth surface for in-service applications. In order to avoid The test procedure of the first gear set was as follows:
the phenomenon of case crushing, no case-hardened gears
were used. (1) Initially, the gears were run for 1 h at a pinion rotation
speed of 300 rpm and a low contact load (10 N), in order
3.4. Experimental procedure to smooth the original machine-finished teeth surfaces.
(2) Then, the gears were run for 8 h with the operation con-
Three sets of the gear pairs were tested with different ditions as the loading stage 1 shown in Table 4, and then
loading conditions and different operating periods. Before stopped.
the test, the lubricating oil was filtered through a 3 m fil- (3) After being removed from the rig, the gears were washed
ter, then, circulated in the lubricating system of the rig until in acetone first, then soaked in Shell X55 solvent in an
Table 3
Material properties of the gears
Material Approximate Hardness BHN Modulus of Yield strength Tensile strength
carbon (%) elasticity (GPa) (MPa) (MPa)
AISI 4340 0.4 350 197 1145 1207
Table 4
Operating and loading conditions of the first gear set
Load stage Pinion speed Contact Input pressure of Contact Maximum Hertzian Operating Total test
(RPM) ratio the pump (psi) load (N) pressure (MPa) hours hours
1 1200 1.6 250 233.1 656 8 8
2 1200 1.6 500 437.7 898 8 16
3 1200 1.6 500 437.7 898 16 32
4 1200 1.6 500 437.7 898 32 64
Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317 1313
Table 5
Operation and loading conditions of the second and third gear sets
Gear set Pinion speed Contact Input pressure of Contact Maximum Hertzian Operating Total test
no. (RPM) ratio the pump (psi) load (N) pressure (MPa) hours hours
2 1200 1.6 1500 1226.8 1503 4 4
3 1200 1.6 2500 1929.7 1885 2 2
1314 Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317
Fig. 6. Scanning electron micrographs of the wheel in gear set 1 showing the spalling phenomenon after 4.32 × 105 contact cycles under the maximum
Hertzian pressure of 898 MPa.
Fig. 7. Scanning electron micrographs of the pinion in gear set 2 showing the spalling phenomenon after 2.88 × 105 contact cycles (4 h engagement)
under the maximum Hertzian pressure of 1503 MPa.
the spall cavities have zigzag surfaces. The only difference cess of a spall cavity, in terms of whether a spall starts to
in spalling phenomenon is that the depth of spalls appeared develop from the surface or subsurface, also needs further
to become deeper and the size of them be larger as the load study. In order to observe these phenomena, some samples
increased. of gear tooth cross-sections prepared and examined in SEM.
Figs. 9–11 present some of the SEM results.
3.5.2. Spalling features observed from sections of From the SEM results, the features of spalling can be
gear teeth summarised as: (1) spalls are concentrated in the dedendum
Though spalling features have been clearly revealed from vicinity of the pitch line (DVP); (2) a spalling cavity consists
the worn surfaces, certain aspects of them still need to be of a shallow wall and a steep wall; (3) the slope of the
further established, such as the depth of a cavity and the shallow wall is about 20–30◦ from the bottom crack line at
slopes of the cavity walls. Furthermore, the formation pro- the trailing tip towards the contact surface, and the slope
Fig. 8. Scanning electron micrographs of the pinion in gear set 3 showing the spalling phenomenon after 1.44 × 105 contact cycles (2 h engagement)
under the maximum Hertzian pressure of 1885 MPa.
Y. Ding, N.F. Rieger / Wear 254 (2003) 1307–1317 1315
Fig. 9. Scanning electron micrographs taken from the gear teeth sections of gear set 1 showing profiles of spall cavities and different stages of spall
formation process. The maximum Hertzian pressure for the gear set 1 was 898 MPa.
Fig. 10. Scanning electron micrographs taken from the gear teeth sections of gear set 2 showing profiles of spall cavities and different stages of spall
formation process. The maximum Hertzian pressure for the gear set 2 was 1503 MPa.
Fig. 11. Scanning electron micrographs taken from the gear teeth sections of gear set 3 showing profiles of spall cavities and different stages of spall
formation process. The maximum Hertzian pressure for the gear set 3 was 1885 MPa.
[18] P.C. Bastias, G.T. Hahn, C.A. Rubin, Finite element modelling of [22] H.E. Knechtel, W.F. Kindle, J.L. McCall, R.D. Buchheit, Metal-
subsurface Mode II cracks under contact loads, Eng. Fracture Mech. lographic Methods, ASM Metal Handbook, eighth ed., vol. 8,
33 (1) (1989) 143–152. American Society of Metals, Metals Park, OH, 1973, pp. 1–14.
[19] L.M. Keer, M.D. Bryant, G.K. Haritos, Subsurface and surface [23] M.L. Atkin, R.A. Cummins, G.P. Sharp, A Study of Gear Wear in a
cracking due to Hertzian contact, J. Lubr. Technol. Trans. ASME Model Gearbox, Mechanical Engineering Transactions, The Institute
104 (1982) 347–351. of Engineers, Australia, 1979, pp. 40–52.
[20] G.R. Miller, A preliminary analysis of subsurface crack branching [24] X.Z. Jin, N.Z. Kang, A study on rolling bearing contact fatigue
under a surface compressive load, J. Tribol. Trans. ASME 110 (1988) failure by macro-observation and micro-analysis, in: Proceedings of
292–296. the International Conference on Wear of Materials, Denver, CO,
[21] J.W. Blake, H.S. Cheng, A surface pitting life model for spur gears. USA, 1989, pp. 205–213.
Part I. Life prediction, J. Tribol. Trans. ASME 113 (1991) 712– [25] D.C. Stouffer, L.T. Dame, Inelastic Deformation of Metals: Models,
718. Mechanical Properties, and Metallurgy, Wiley, New York, 1996.