Wear, 169 (1993) 69-75
Failure modes of pre-cracked contact
ceramic elements under rolling
M. Hadfield and T.A. Stolarski Bmnel University, Depatiment of mechanical ~ngi~ee~~~ Uxbridge, Middksex UB& 3PH (UKj
R.T. Cund~ll and S. Horton SKFEngineering & Research Centre Bk: (Received
20, 1992; accepted
Postbus 2350, 3430 DT ~~u~egei~
March 18, 1993)
Abstract A hybrid ceramic/steel angular contact ball-bearing was experimentally modelIed using a modified four-ball machine. Ceramic ball surfaces were artificially damaged with ring, lateral and radial pre-cracks. Rolling contact fatigue failure modes were studied under high contact stresses and speeds. The role of lubricant type in roiling contact fatigue failure mode was also assessed.
The use of ceramic materials applied to rolling element bearings shows some practical advantages over bearing steels [l, 21. Silicon nitride has been found to have the optimum combination of properties suitable for this application. Hybrid ball-bearings, i.e. precision angular contact ball-bearings with ceramic balls, are now used as standard components within the ballbearing industry. Full scale hybrid ball-bearing tests  have proved satisfactory and identify a ceramic bail spalling mode of fatigue failure. Quali~ control of the ceramic ball surfaces has reached a satisfactory position and h~gh”volume inspection is practical. Surface cracks were, however, occasionally found on ceramic balls , and hence research on their influence on fatigue failure modes was necessary. Ring cracks caused by manufactu~ng pressing faults or blunt impact loads are the most common type found on ceramic ball surfaces. These circular cracks have been studied in the past by Htiber [S] and more recently by Li et al. . Lateral and radial surface cracks are less common and occur due to poor sintering mixture or by sharp indention loads . Experimental work using a modified four-ball machine has produced steel/ceramic rolling contact fatigue faifures of pre-cracked ceramic elements. This paper presents and discusses results obtained during these studies.
2. Test specifications
A modified four-ball machine, shown in Fig. 1, was used to test ceramic balls. This machine was employed as it correctly models ball-bearing motions and precisely defines contact load.
Fig. 1. Schematic of the modified four-bait machine. 1: upper ball and collet; 2: lower balls; 3: spindle; 4: loading lever; 5: driving motor;
0 1993 - Elsevier
8: belt drive.
44. Hadfield et al. ! Failure modes of pre-cracked
undw mliirzg crtrit,tt?
The machine consists of an assembly which simulates an angular contact ball-bearing. The steel cup represents the bearing outer-race, three lower ceramic balls represent the balls within the hybrid bearing-race and the upper ball represents the hybrid bearing inner-race. The assembly was loaded via a piston below the steel cup, from a lever-arm load. The upper ball was connected to a drive shaft via a collet and was in contact with the three lower balls when the machine was stationary. The contacting positions between the upper ball and tower balls were immersed in lubricating oil. The machine has been used to examine rolling contact fatigue performance by various researchers. The Institute of Petroleum has gathered various papers [S] which described various test results, ball dynamics and kinematics. There are also descriptions of lubrication effects , and preliminary studies on ceramics [lo] have also been performed. 2.2. Test conditions The material tested was silicon nitride, manufactured by a hot isostatic pressing (HIP) method. Bail blanks
Fig. 3. Radial and lateral cracks (X ultraviolet light.
tb) Fig. 2. Typical ring crack light.
70): (a) normal light, (b) ultraviolet
(a) normal light, (b)
were ground and polished to half-inch diameter, using standardized procedures to ensure a consistently high quality of material and geometry. The average roughness (R,) of ceramic ball surfaces was 0.008 pm and ball roundness was within standard ball-bearing tolerances. Ring or cone cracks were generated on the balls’ surfaces by blunt impact loads. A finished ceramic ball was placed in a mild steel recess, with a tungsten carbide ball positioned on top of the ceramic ball. A flat load was dropped under gravity through an afignment tube to impact on the ball assembly. A typical ring crack is shown in Fig. 2. This crack was produced by a 160 g impact load, dropped from a height of 500 mm. Figures 2(a) and (b) show the microscopic ring crack image under normal and ultraviolet light. The ring crack diameter was 0.65 mm; this dimension corresponds to the theoretical Hertzian circle diameter for this load condition. Radial and lateral cracks were propagated on the ball surface by a Vickers hardness test, using a diamond pyramid indentor. The standard hardness test procedure
et al. / Failure modes of pre-cracked
according to IS0 146 was used, in which a 5 kg load was applied for 5 s. The indentations cause both radial and lateral cracks. Figure 3 shows typical optical analysis of radial and lateral cracks caused by a 5.0 kg indentor load. Figure 3(a) is a microscopic photograph of the impression under normal light conditions, while Fig. 3(b) illustrates the same pre-crack after treatment with a dye penetrant. Radial crack length and hardness impression width were measured using a calibrated viewing scale; in this case the crack length is 0.1 mm and the impression width is 0.053 mm. All four-ball tests were conducted at 6.4 GPa maximum and spindle speed of 5000 r.p.m. A standard ball-bearing steel upper ball contacts with three precracked ceramic balls. Each ceramic lower ball was pre-cracked with nine equally spaced ring cracks (tests A, B and C) or six clusters of four hardness indentations (tests D, E and F). Lubricants were chosen to assess the effects of elastohydrodynamic film thickness and penetration ability on the surface pre-cracks. Highviscosity lubricant, B.P. HiTec 174, low-viscosity turbine lubricant, Exxon 2389, and high penetration ability lubricant, standard kerosene, were used.
elements under rolling contact
2.3, Test results The test results (Table 1) describe six typical precracked ball experiments. Cycles to failure are recorded from the steel upper ball. The test is stopped at high vibration levels. Final bulk oil temperature is a function of test cycles and lubricant type. (b) Fig. 4. First stages section ( x 80).
of spa11 (test
3. Fatigue failure modes 3.1. Ring crack spa11
Ring-type pre-cracks failed by a spalling mode during all tests. Examples of the various spa11 stages (Figs. 4-6) explain the mechanisms involved. The first stages of failure are shown in Fig. 4. Figure 4(a) shows the TABLE
A B C D E F
Ring Ring Ring Radial lateral Radial lateral Radial lateral
Stress cycles to upper ball failure (millions)
Final oil temperature
HiTec 174 Exxon 2389 Kerosene HiTec 174
1.05 0.25 0.10 14.0
63 44 40 75
beginning of material removal and Fig. 4(b) shows the crack front within the ceramic material. Stage two is incipient spall, shown in Fig. 5(a), which indicates further material removal adjacent to the original ring cracks. The initial third stage spa11 (Fig. 5(b)) indicated severe material removal as lubricant hydrostatic pressure produces bending moments within the material. A fully developed spa11(Fig. 6(a)) indicates the smooth conical brittle fracture surface of the original ring crack. This post-test example is propagated from the ring crack of Fig. 2. The cliff fracture edge indicates a combination of brittle failure and markings of beachlike fatigue failure. Secondary damage may occur after the initial ring crack spa11(Fig. 6(b)). Additional surface cracks propagate from the damaged area, which results in secondary spalling. 3.2. Lateral crack spa11 Lateral crack spa11 failure mode is produced with tests using the Exxon 2389 lubricant. The failure example shown in Fig. 7 illustrates a lateral crack spa11produced
et al. I Failure modes of pre-cracked
elements under rolling conruo
(b) Fig. 5. Second and third stage, (b) third stage.
stages of spa11 (test B): (a) second
Fig. 6. Fully developed and secondary spall damage: developed (test B), (b) secondary damage (test c‘).
from the pre-crack of Fig. 3. Measurement of radial crack length before and after testing reveals zero propagation. Lateral crack propagation is evident from Figs. 7(a) and (b) and is confirmed by comparing microscopic images after dye penetration processing of pre- and post-test damage. Lateral cracks are propagated by lubricant penetration and subsequent brittle-like subsurface failure.
Figure 9 shows the typical fatigue spa11 and contact path, with details of the steel upper ball failure mechanism. Figure 9(a) illustrates the spa11 and contact path while a 90” rotation (Fig. 9(b)) reveals contact edge cracks and subsurface delamination with typical cliff base cracks. Surface micro-cracks are also evident adjacent to the cliff edge (Fig. 9(b)); this further implies subsurface damage.
3.3. Radial crack propagation and sueace delamination Ceramic pre-cracked balls fail from a radial crack propagation and surface delamination fatigue mode when tested with the HiTec 174 lubricant. Two precracked areas (Fig. 8(a)) are connected by fatiguepropagated radial cracks. The subsequent delamination area shows typical fatigue undulations, described previously [ 1I, 121, with adjoining radial propagated cracks. The reversed viewing position (Fig. 8(b)) clearly shows a typical delamination cliff edge and cliff base crack, implying subsurface propagation.
3.4. Radial crack damage and surfitce bulging Radial crack damage and surface bulging failure mode occurs on radial and Iaterai pre-cracked ceramic balls lubricated with kerosene. Figures 10(a)-(d) illustrate the failure mechanisms. An overview of a pre-crack cluster after testing (Fig. 10(a)), shows limited damage from the radial cracks, with no lateral crack damage apparent. Close examination of a radial crack (Fig. 10(b)) shows crack damage, and also wear on the ball surface. Close examination of the bulge (Fig. 10(c)) shows trapped debris within the apex cracks.
et al. / Failure modes
elements under rolling contact
Fig. 7. Lateral crack spall (test E): (a) failure overview, (b) failure overview (UV X280).
Fig. 8. RadiaI crack propagation and delamination radial propagation, (b) surface delamination.
Electron probe microanalysis of the ball surface near the bulge and of the bulge debris (Fig. 10(c)) indicates that the debris is, in fact, ferrous. The debris must originate from the steel upper ball in this case. The steel debris trapped within the ceramic bulges implies the failure mechanism. The lubricant, under high hydrostatic pressure, propagates subsurface cracks from the lateral crack root. The steel debris trapped under the ball surface, suspended in the pressurized lubricant, is forced up through the ceramic surface. A further example of this failure mode is illustrated in Fig. 10(d).
influence on the steel upper ball cycles to failure. The debris size, related to pre-crack type, significantly reduces cycles to failure (Table 1). It can be seen that all ring pre-cracked experiments have shorter lives than the radial and lateral pre-cracks when comparing tests using similar lubricants. The most common type of surface defect, ring cracks, has been examined under different lubrication conditions and extreme loads. The spalling fatigue mode is shown in various stages; subsurface examinations of these stages are described by Hadfield et al. . The lubricant type will affect the overall bearing life if a ceramic element begins to spa11in this mode. The steel upper ball performs satisfactorily when in contact with the damaged ceramic balls, although the high-~scosi~ lubricant prolongs life. Secondary spalling is a cause for concern as remote damage is evident; this occurs during kerosene ring crack tests. Experiments using radial and lateral pre-cracks provide very interesting results. Three failure modes are identified: lateral crack spall; radial crack propagation
4. Discussion All rolling contact fatigue tests of pre-cracked ceramic balls contacting with a steel upper-ball and cup result in non-catastrophic fatigue failure modes. Distortion of ceramic ball surfaces remote from the spalled area only occurs during kerosene lubricant tests. Ceramic debris suspended in the lubricant has a significant
(test D): (a)
et al. I Failure
modes of pre-cracked
elements under rolling c‘ontur~!
(b) Fig. 9. Steel upper ball spa11 and delamination and contact path, (b) spa11 delamination.
(test D): (a) spa11
and surface delamination; and radial crack damage and surface bulging. All are related to lubricant type. The lateral crack spa11 is localized to the point defect area and hence adverse effects on the contact zone are limited. The radial crack propagation and delamination failure mode does produce relatively large areas of damaged ceramic element surfaces. The spa11 and delamination of the steel upper ball illustrate the fact that this failure mode is not restricted to ceramic elements in this contact configuration. Previously measured residual stresses of delamination areas  suggest high shear stresses, which are probably related to the lubricant’s mechanical properties. The radial crack damage and surface bulging failure mode is restricted to the pre-damaged area. Surface distortion of the ceramic element and steel upper ball wear occurs when a kerosene lubricant is used. The steel debris found in the ceramic bulges confirms a subsurface mechanism. The bulges also illustrate the subsurface fatigue mechanisms which may occur when
Fig. 10. Radial crack propagation and delaminal .ion (tes F): (a) pre-crack damage, (b) radial crack damage, (c) aldjalcent XIlging. (d) additional bulging example.
high hydrostatic material.
et al. I Failure modes of pre-cracked
stresses are applied to the ceramic
Non-catastrophic failure modes of ceramic elements have been identified from pre-cracked rolling contact fatigue tests. A spalling fatigue failure mode is identified in the case of ring pre-cracks. Spa11 ring crack stages are described which illustrate the fracture mechanism. Three failure modes of lateral and radial pre-cracks are identified. These fatigue failure modes are strongly related to lubricant properties.
References G. Hamburg, P. Cowley and R. Valori, Operation of an allceramic mainshaft roller bearing in a J-402 gas-turbine engine, J. ASLE, Lubr. Erg., 37 (1980) 407-415. H. Aramaki, Y. Shoda, Y. Morishita and T. Sawamoto, The performance of ball bearings with silicon-nitride ceramic balls in high speed spindles for machine tools, J. Tiibol., 110 (1988) 693-698.
elements under rolling contact
J.I. 3 F.R. Morrison, load-life relationship
McCool and T.M. Yonushonis, for M50 steel bearings with
75 The silicon
nitride ceramic balls, J. ASLE, Lubr. Eng., 40 (1984) 153-159. into the 21st century, 4 R.T. Cundill, Rolling element bearings IMechE Seminar, London, November 1990, June 1993. Zur theorie de Beruhrung fester elastische 5 M.T. Hiiber, Korper, Ann. Phys. Lpz., 14 (1904) 153. Y. Li and D.A. Hills, The Hertzian cone crack, Trans. ASME, 58 (1991) 120-127. K. Ueda, Contact-stress deformation and fracture in ceramics, Jpn. J. Tribal., 34 (1989) 123-131. R. Tourret and E.P. Wright, Rolling contact fatigue: performance testing of lubricants, Proc. Znt. Symp., Inst. Petroleum, London, October 1976, Heyden & Son, London, 1977. Hot pressed silicon nitride as a 9 D. Scott and J. Blackwell, rolling bearing material - a preliminary assessment, Wear, 24 (1973) 61-67. and P.J. McCullagh, Silicon nitride as 10 D. Scott, J. Blackwell a rolling bearing material - a preliminary assessment, Wear, 17 (1971) 73-82. T.A. Stolarski and R.T. Cundill, Delamination 11 M. Hadfield, of ceramic balls in rolling contact, Ceram. Int., 19 (3) (1993). T.A. Stolarski and S. Tobe, Analysis of silicon 12 M. Hadfield, nitride ball failure due to rolling contact fatigue, ICR 92, Birmingham, Sept. 7-10, 1992, MCE, Vol. 1, pp. 297-302. T.A. Stolarski, R.T. Cundill and S. Horton, 13 M. Hadfield, Failure modes of ceramic elements with ring-crack defects, Tribol. Int., in press. 14 M. Hadfield, T.A. Stolarski, G. Fujinawa and S. Tobe, Residual stresses in failed ceramic rolling contact balls, Ceram. Int., 19 (5) (1993).