Tribology International 122 (2018) 169–178

Contents lists available at ScienceDirect

Tribology International
journal homepage: www.elsevier.com/locate/triboint

A novel method to model effects of natural defect on roller bearing

Yuwei Liu, Yongsheng Zhu *, Ke Yan, Fangzhe Wang, Jun Hong
Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System, Xi'an Jiaotong University, Xi'an 710049, China

A R T I C L E I N F O A B S T R A C T

Keywords: The assumption that the defects extend axially throughout rollers or raceways was widely adopted in previous
Roller bearings studies, however, this is hardly the situation in natural defects. A contact model that considered the geometric
Natural defects feature of natural defect was developed to analyze effects of defect on roller-raceway contact. The contact stiffness
Contact stiffness expression for defected roller-raceway was derived by numerical fitting. The contact stiffness expression was
Contact load distribution introduced into the quasi-static model of roller bearing to model defected roller bearing. The contact load dis-
tribution and stiffness of roller bearing were simulated under different defect sizes and operating conditions.
Results demonstrate that roller bearing was insensitive to natural defects and the prevailing fatigue criteria is not
applicable to roller bearing.

1. Introduction
Defects caused by fatigue spalling will degrade bearing performances
and eventually lead to bearing failure. The modeling for defected bear-
ings has been widely researched, which provided mechanistic models for
simulating and understanding the effects of defect on bearing. Recent
studies attempt to diagnose defects [1,2] and predict useful life [3,4] of
bearings through establishing a correlation model between defect and its
vibration response. These studies have given enough accurate diagnostic
and predictive results for ball bearing, however, they are not suitable for
roller bearings due to the difference between contact type of ball bearings
and that of roller bearings. Experiments have shown that roller bearings
remain useful even when the defect degree is well beyond the failure
criteria [5,6] that the appearance of first spalling or the defect area
growing to 6.5 mm2. The experimental results indicate that the effects of
defect on roller bearings have not been accurately modeled in previous
studies. To provide the foundation for accurate defect diagnosis and life
prediction of roller bearings, the defected roller bearing model consid-
ering the characteristics of contact type and natural defects is required.
Defects of roller bearing is considered as the depression due to ma-
terial spalling. The assumption that defects extend axially throughout the
rollers or raceways is widely adopted in modeling of defected roller
bearings. Rollers passing over such “through” type defect will go into the
defect depression, which will result in additional displacement and cur-
vature variations as shown in Fig. 1. Through the introduction of the
contact load change due to additional displacement and the Hertz contact
stiffness change due to curvature variations [7,8[7,8] into bearing
dynamic model, the vibration response of defected roller bearing can be
modeled. Several researchers analyzed the vibration response of roller
bearing with different width of “through” type defect [9], and proposed
the measuring technique of defect width based on decomposition using
wavelet transform [10,11]. The assumption of “through” type defect is
adopted so that the defect of roller bearing can be modeled in the same
way as that of ball bearing, however, natural defects of roller bearing
rarely extend the entire axial length of roller or raceway [5,12]. Fig. 2
demonstrates a typical natural defect of roller bearing, which shows
clearly that the defect is “non-through” in nature. Rollers passing over
such “non-through” type defect are still supported by non-defect region
as shown in Fig. 3, therefore, the rollers will not go into the defect
depression. In comparison with the prevailing assumption of “through”
type defect, authors believe treating defects as the “non-through” type
depression reflects geometric feature of the natural defects more faith-
fully. Since “non-through” type defect cannot be modeled and analyzed
by the existing methods, it is necessary to propose a new method to
model the “non-through” type defects of roller bearings.
Contact condition of roller-raceway directly affects the motion of
bearings component [13,14,15], therefore, the present paper attempts to
model the effects of “non-through” type defect on roller bearing perfor-
mance by considering the change of roller-raceway contact stiffness. In
general, the roller-raceway contact is simplified to a two-dimensional
plane strain problem and solved using Hertz line contact theory. How-
ever, the geometric feature of roller and raceway, especially the features
of defect, cannot be modeled accurately and completely in this
two-dimensional contact model. Recent contact modeling [9,16,17] have

* Corresponding author.
E-mail address: yszhu@mail.xjtu.edu.cn (Y. Zhu).

https://doi.org/10.1016/j.triboint.2018.02.028
Received 16 November 2017; Received in revised form 14 February 2018; Accepted 22 February 2018
Available online 27 February 2018
0301-679X/© 2018 Elsevier Ltd. All rights reserved.
Y. Liu et al.
been undertaken to overcome the limitations of the two-dimensional
contact model by using non-Hertz line contact theory in an effort to
model the complex surface features of contact bodies, and got a solution
with enough accuracy needed. However, no studies, to the authors'
knowledge, have analyzed the effects of “non-through” type defect on
roller-raceway contact stiffness.
This paper develops a defected roller-raceway contact model through
the modification of initial clearance function in the non-Hertz line con-
tact theory, and analyzes the effects of different degrees of “non-through”
type defect on roller-raceway contact. Through the roller-raceway con-
tact analysis, the change of contact stiffness due to defect is obtained.
170
Tribology International 122 (2018) 169–178
Fig. 1. The model of "through" type defects
of roller bearing.
Fig. 2. The natural defect of roller bearing: (a) Early
fatigue spalling [5,12]; (b) Severe fatigue spalling.
Fig. 3. The model of "non-through" type
defects of roller bearing.
Fig. 4. The equivalent contact model of
roller-raceway.
Then contact stiffness expressions under different defect degree are
derived by numerical fitting to simplify the subsequent simulation. The
defected roller bearing is modeled by incorporating the change of contact
stiffness into three-degrees of freedom (3-DOF) quasi-static model of
roller bearings. Finally, the effects of defect on roller bearings under
different operation condition are simulated based on the newly devel-
oped model. In contrast to existing model, the proposed model could
faithfully simulate the effects of natural defect, namely “non-through”
type defect, on roller bearings. This is particularly important for accu-
rately diagnosing fault and predicting life of roller bearings.
Y. Liu et al. Tribology International 122 (2018) 169–178

2. The modeling of defected roller-raceway contact

Contact model is concerned with contact pressure distribution and

elastic deformation of two contacting bodies. The contact model for this
study was developed to simulate contact behavior of roller-raceway
under natural defect. In the absence of defects, the roller-raceway con-
tact can be described using an equivalent contact model of a non-classical
cylinder and a plane as shown in Fig. 4. The contact model based on non-
Hertz line contact theory is represented [18] as
8
>
> ∬ pðx0 ; y0 Þdx0 dy0 ¼ Q
> Sc
>
<
> 1 pðx0 ; y0 Þdx0 dy0 (1)
>
>
> 0 ∬ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ δ  Zðx; yÞ
Fig. 5. Schematic drawing of a logarithmically crowned roller. : π E Sc ðx  x0 Þ þ ðy  y0 Þ
2 2

Where Sc is the contact area of roller-raceway; Q is the contact load of

Fig. 6. Calculation flow of contact model.

171
Y. Liu et al. Tribology International 122 (2018) 169–178

Fig. 7. Schematic drawing of defect profile.

8 mn
roller-raceway; E0 is the equivalent elastic modulus of roller-raceway, δ is > X
>
> aj bj pj ¼ Q
the elastic approach of roller-raceway, namely, the approach between >
>
4
>
>
two contact bodies; pðx0 ; y 0 Þ is the contact pressure at pointðx0 ;y 0 Þ; Zðx; yÞ >
>
j¼1
>
>
< 1 X mn
is the initial clearance at pointðx; yÞ. D p ¼ δ  zi ðxi ; yi Þ  Ad  φi ðxi ; yi Þ
0
π E j¼1 ij j (4)
The initial clearance Z(x,y) reflects curvature radius of rollers and >
>
>
>
raceways as well as crowning of rollers. In this paper, the rollers are >
>  
>
> pj  0; xj ; yj 2 Ω
>
>
considered to have logarithmically crowned as shown in Fig. 5. The >
:  
crown drop Zr(y) can be represented as pj < 0; xj ; yj 62 Ω

Qd 1 Where Ω is the solution domain, which is divided into m  n units; m and

Zr ðyÞ ¼ ln (2)
π E 0 le 1  ð2y=le Þ2 n is the number of grid nodes in the solution region width and length
direction, respectively; aj and bj is half length and half width of j unit,
Where le is the length of roller; Qd is the design load of roller; respectively; Dij is the flexibility matrix [18]; zi ðxi ; yi Þ and φi ðxi ; yi Þ is
Natural defect is of surface depression of roller and raceway, which initial clearance and defect topography function of i unit, respectively.
will change the initial clearance. As shown in Fig. 4, the defect is The details of the solution process of discretized contact model are
described using topography function φðxd ; yd Þ in the defect-fixed coor- illustrated in Fig. 6. It is noted that the initial solution domain needs to be
dinate system of Od-XdYdZd, while the origin of the coordinate system of larger than the actual contact area. In this paper, the length and width of
Od-XdYdZd is described by position vector Ad in the O-XYZ coordinate solution domain is the roller length and the 1.4 times Hertz line contact
systems of contact area. The position vector Ad describes relative position width, respectively.
between defects and contact area during kinematic condition. By intro- In this study defects are considered to be a rectangular depression
ducing the effects of defect on initial clearance, the present paper de- with uniform depth dd as shown in Fig. 7, the length of defect along roller
velops the contact model of defected roller-raceway as axial and radial direction is denoted as defect length ld and defect width
8 bd , respectively. The defect profile is chosen so as to simulate the effects
>
< of “non-through” type defect on roller-raceway contact while reducing
1 pðx0 ; y0 Þdx0 dy0
∬ pðx0 ; y0 Þdx0 dy0 ¼ Q ∬ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the model complexity due to irregular natural defect profiles. However,
>
: Sc π E Sc ðx  x0 Þ2 þ ðy  y0 Þ2
0
the developed contact model can be use in future to simulate more
¼ δ  Zðx; yÞ  Ad  φðxd ; yd Þ (3) complex defects.

Since there is no analytical solution for equation (3), the equation was 3. The modeling of defected roller bearing
discrete and solved numerically. The discretization equations of the
contact model are presented as The model of defected roller bearing was performed to demonstrate
the effects of defect on roller bearing. Taking tapered roller bearings

Fig. 8. The TRB with a outer-raceway defect.

172
Y. Liu et al. Tribology International 122 (2018) 169–178

Fig. 9. Solution flow of defected roller bearing

model.

(TRB) as an example, the three-degrees of freedom (3-DOF) quasi-static Q ¼ Kδn (6)

model, namely the equilibrium equations of bearing ring, is given in
simultaneous equation (5). The relationship between contact load and WhereK; n is the stiffness coefficient and stiffness index between roller
elastic approach of roller-raceway can be represented by the contact and raceway, respectively; δ is the elastic approach between roller and
stiffness expression as shown in equation (6). Therefore, the present raceway; Q is the contact load between roller and raceway.
paper attempt to model defect of roller bearing by changing the contact The defect region is determined by angular position and angular
stiffness, namely the stiffness coefficient K and index n. The values of distance of defects as shown in Fig. 8. The defect width can be considered
stiffness coefficient and index under different degree defects can be ob- as the arc length of angular distance of defect. As the roller enters defect
tained by the following steps:1) The defected roller-raceway contact region, the roller-raceway contact stiffness changes. Taking the defect on
analyses are performed to obtain the elastic approaches over a range of outer-raceway for example, the contact stiffness of j th roller is repre-
contact loads. 2) The stiffness coefficient and index coefficient are sented as
derived by numerical fitting based on the relationship between contact 
  ðKd ; nd Þ jθj  ϕd j < Δϕd =2
loads and elastic approaches. Kj ; nj ¼ (7)
ðKn ; nn Þ non  defect area
8X Z
>
> Qj cos αcos φj þ Fc cos φj þ Fx ¼ 0 Where θj is the angular position of jth roller; ϕd is the angular position of
>
>
>
> j¼1 defect; Δφd is the angular distance of defect; Kd ; nd are the stiffness co-
>
<X Z
Qj cos αsin φj þ Fc sin φj þ Fy ¼ 0 (5) efficient and index under defect condition; Kn ; nn are the stiffness coef-
>
> ficient and index under non-defect condition.
>
>
j¼1
>
> XZ
>
: The quasi-static model of defected roller bearing is simultaneous
Qj sin α þ Fz ¼ 0
nonlinear equations, which can be solved by the Newton-Raphson iter-
j¼1
ation algorithm. The detailed solution process of defected roller bearing
Where Qj is the contact load betweenjth roller and raceway; Fc is the model is illustrated in Fig. 9.
centrifugal force of roller; α is the contact angle between roller and
raceway; φj is the angular position of jth roller; Fx ; Fy ; Fz are the external 4. Simulation results
loads of outer ring in different direction.
The modeling of defected roller bearing is carried out on TRB 32008J

173
Y. Liu et al. Tribology International 122 (2018) 169–178

Table 1 position of defect (Fig. 8) is 0 rad.

The parameters of TRB 32008J.
Geometrical characteristics Value
4.1. Effects of defect on roller-raceway contact
Small diameter of taper roller (mm) 6.131
Large diameter of taper roller (mm) 6.846
Length of roller (mm) 13.66 The processing of roller passing over a defect is simulated to
Number of rollers 23 demonstrate the effects of defect on contact pressure and contact stiff-
Outer raceway angle (rad) 0.2473 ness. As shown in Fig. 10, the relative displacement between defect and
Inner raceway angle (rad) 0.1949
contact area Dx increases from negative to positive, as the roller passes
Roller angle (rad) 0.0262
Flange angle (rad) 1.5621 over defect. When the sum of half contact width bc/2 and half defect
Elasticity modulus (N/mm2) 2.1  105 width bd/2 is larger than jDxj, the roller will enter the defect area. Ac-
Poisson's ratio 0.278 cording to the effect of defect on roller-raceway contact area, the contact
state can be divided into the following stages: In state 1or 5, the roller is
not located in defect region, and the contact area remains constant; In
state 2 or 4, the roller is in contact with both axial and radial edge of
defect, and the contact area gradually changes with the changing of Dx; In
state 3, the roller is only in contact with axial edge of defect, and the
contact area remains constant.
Under 3 kN contact load, when roller passes over a defect (defects
length ld, width bd and depth dd is 2 mm, 1 mm and 0.2 mm, respectively)
from left to right as shown in Fig. 10, the dimensionless contact pressure
distribution of roller-raceway are shown in Fig. 11. In order to show the
pressure concentration at defect edge obviously, the number of grid
nodes m and n is 40 and 400, respectively. In this figure, X and Y axis is
radial direction and axial direction of roller, respectively. The dimen-
sionless pressure is the ratio of actual contact pressure to maximum
contact pressure obtained from non-defect condition. In Fig. 11 (a) and
(i), the contact pressure distribution is the same as that under non-defect
state, which corresponds to the State 1 and State 5 in Fig. 10, respec-
tively. In Fig.11(b) ~(d) (correspond to the State 2 in Fig. 10). and (f)
Fig. 10. The process of roller passing over defect.
~(h) (correspond to the State 4 in Fig. 10), the pressure concentration
appears simultaneously at the axial and radial edge of defect, which
with the presented defect modeling method. The parameters of TRB demonstrates the roller contacts with both axial and radial edge of defect
32008J are shown in Table 1. The defect is positioned to have its axis region. The pressure concentration gradually intensifies with the
(Fig. 7) on the center of roller-raceway contact area, and the angular decreasing of contact area caused by defect. Since the contact width is

Fig. 11. The dimensionless contact pressure

distribution of roller-raceway: (a)
Dx ¼ 0.6 mm; (b) Dx ¼ 0.54 mm; (c)
Dx ¼ 0.5 mm; (d) Dx ¼ 0.46 mm; (e)
Dx ¼ 0.36~-0.36 mm; (f) Dx ¼ 0.46 mm;
(g) Dx ¼ 0.5 mm; (h) Dx ¼ 0.54 mm; (i)
Dx ¼ 0.6 mm.

174
Y. Liu et al. Tribology International 122 (2018) 169–178

Fig. 12. The contact pressure profile on axial symmetry section of contact area.

Fig. 14. The relationship between elastic approximation and contact load.

Table 2
The fitting results of stiffness coefficient and index.
Defect length Stiffness coefficientK Stiffness index Maximum error
l(mm) n

0 9.386  105 1.201 2.21%

2 9.052  105 1.215 2.37%
4 8.270  105 1.231 2.24%
6 7.469  105 1.244 2.50%
8 6.162  105 1.262 2.45%

Corresponding to the contact state shown in Fig. 10, the change of

elastic approach can be divided into the following stages: a) the elastic
Fig. 13. The elastic approach of roller-raceway under defect condition.
approach remains constant at the minimum; b) the elastic approach
gradually increases; c) the elastic approach remains constant at the
very small, the contact states showed in Fig.11(b)~(d) and (f)~(h) last
maximum; d) the elastic approach gradually decreases; e) the elastic
for a short time compared with others contact state. In the process of
approach remains constant at the minimum. It is clear that the elastic
roller passing over defect, the most likely case is the State 3 shown in
approach is negatively correlated with contact area of roller-raceway.
Fig. 10, in which the pressure concentration appears only at the axial
Since the increasing and decreasing process of elastic approach is very
edge as shown in Fig. 11(e). The pressure concentration at defect edge
rapid, the change of elastic approach can be approximated with a rect-
lead to acceleration of material fatigue and plastic deformation of roller
angular function in the numerical fitting of contact stiffness.
and raceway, which will in turn accelerate bearing fatigue failure.
Fig. 14 plots the elastic approach of roller-raceway under different
The pressure profile on axial symmetry section of contact area (sec-
defect lengths and contact loads. As shown Fig. 14, the elastic approach
tion y ¼ 6.83) is plotted to demonstrate the contact condition between
increases with the increase of defect length, however, the relationship
roller and defect bottom when the roll passes through the defect. As
between elastic approach and contact load is still near linear. Since the
shown in Fig. 12, the pressure is zero where the defect is located. This
elastic approach of roller-inner raceway is approximately equaled with
phenomenon indicates that the roller does not contact with defect bottom
that of roller-outer raceway, the difference between roller-inner raceway
when roll passes over a sufficiently deep defect. With developments of
and roller-outer raceway was ignored in the numerical fitting of contact
materials and manufacturing technology, the rolling contact fatigue
stiffness.
(RCF) resulting from subsurface crack has become the main cause of
From the above analysis, it can be found that the defect length is the
bearing defects. The depth of defects caused by the RCF is approximately
main factor affecting value of elastic approach and contact stiffness.
equal to a quarter of the contact width [6,19], which is sufficiently large
Based on equation (6), the contact stiffness expression of roller-raceway
so that there is no contact between rollers and defect bottom. Therefore,
under different defect length was derived by numerical fitting. Table 2
it is reasonable to ignore the effect of defect depth on roller-raceway
summarizes the fitting results of stiffness coefficient and index. The
contact stiffness in following analysis.
maximum error between fitting results and simulation results is less than
Fig. 13 plots the change of dimensionless elastic approach (δ/δ0)
2.5%, which validate the usability and accuracy of fitted contact stiffness
between roller and raceway in the processing of roller passing over the
expressions.
defect. The δ0 is the elastic approach of roller-raceway under non-defect
condition.

175
Y. Liu et al. Tribology International 122 (2018) 169–178

Fig. 15. The effects of rotating speed and load on

contact load distribution: (a) At low load; (b) At high
load.

4.2. Effects of defect on contact load distribution and stiffness of roller

bearing

The quasi-static analysis of TRB 32008J with a defected outer-

raceway is performed to demonstrate the effects of defect on contact
load distribution and bearing stiffness. The contact load distribution and
stiffness of roller bearing were simulating over a range of defect length,
defect width, load and rotating speed.

4.2.1. Effects of rotating speed

Fig. 15 plots the contact load distribution between roller and outer-
raceway under different rotating speeds and loads, in which the defect
length and width is 4 mm and 7 mm, respectively. At low load (Fx ¼ 0 N,
Fy ¼ 120 N, Fz ¼ 60 N), the increase of contact loads due to centrifugal
action is obvious (Fig. 15(a)), while at the high load (Fx ¼ 0 N,
Fy ¼ 9000 N, Fz ¼ 4500 N), the contact load increment is too small to be
clearly observed (Fig. 15(b)). The contact load of roller in defect region
suddenly decreases, however, the contact load does not reduce to 0 under Fig. 16. The effects of defect length on contact load distribution.
any rotating speeds and loads. As a result, it is believed that the previous
assumption of “through” type defect, which usually result in the contact

176
Y. Liu et al.
Fig. 18. The effects of defect length on contact load distribution.
load of roller in defect region is 0, will incorrectly estimate the contact
load distribution.
4.2.2. Effects of defect length
The defect length directly affects the contact stiffness of roller-
raceway, therefore, the contact load distribution is also dependent on
defect length. Fig. 16 plots the contact load distribution between roller
and outer raceway, in which the defect length ranges from 2 to 8 mm and
177
Tribology International 122 (2018) 169–178
Fig. 17. The relationship between bearing stiffness
and defect length: (a) Radial stiffness (b) Axial
stiffness.
the defect width is 7 mm. In the defect region, the contact load decreases
significantly with increasing of defect length. On the other hand, the
contact loads of roller-raceway, near to the defect region, increase with
increasing of defect length, and the increment of contact loads positively
correlate with the relative distance between the roller and the defect
(Fig. 16 View A). It is clear that the defect results in the concentrate of
contact load distribution from non-defect region to defect region.
The changes in TRB stiffness due to increase of defect length are
shown in Fig. 17. The stiffness of TRB, both radial and axial, decrease
gradually with increasing of defect length and the stiffness change trend
is unaffected by the load. The maximum loss of stiffness along radial and
axial direction is approximately 8.7% and 3.5%, respectively. The pre-
vailing fatigue criteria of bearing is the first spalling formation [19] or
the area of defect upping to 6.5 mm2 [5], which does not consider the
difference between line contact (roller bearings) and point contact (ball
bearings). Assuming there is a square defect on outer-raceway, when
defect area goes up to 9 mm2 that is greater than the prevailing fatigue
criteria, the stiffness losses of TRB along both radial and axial direction
are less than 3%. This result demonstrates that the roller bearing stiffness
is less affected by natural defect, therefore, it is suggested that the fatigue
criteria could be relaxed to accommodate some acceptable larger spall
size for roller bearing running at medium or low speed.
4.2.3. Effects of defect width
As the defect grow further, especially the increases of defect width,
the number of rollers in the defect region increases and the contact load
distribution changes accordingly. Fig. 18 shows the effects of defect
Fig. 19. The relationship between bearing stiffness
and defect width: (a) Radial stiffness (b) Axial
stiffness.
Y. Liu et al.
width on contact load distribution, in which the defect width ranges from
7 to 49 mm (the corresponding angular distance of defect ranges from
0.24 to 1.71 rad) and the defect length is 4 mm. With increasing of defect
width, the contact loads of roller-raceway in defect region increase, and
the contact load distribution is close to that under non-defect condition.
In comparison with the previous defected roller bearing models which
ignores the effects of “non-through" type defect, the proposed model
simulates the contact load distribution at different defect degree more
accurately.
The effect of defect width on bearing stiffness is dependent on the
number of rollers located in defect region, therefore, the change of TRB
stiffness caused by defect is not continuous. As shown in Fig. 19, the
stiffness of TRB, both radial and axial, decrease obviously with increasing
of the number of rollers in defect region increases. When defect width
goes up to 49 mm, there are seven rollers in the contact region, and the
stiffness losses of TRB along both radial and axial direction is approxi-
mately 22% and 11%, respectively. This result demonstrates that the
“non-through" type defect covering multiple rollers will lead to a sig-
nificant decrease in the roller bearing stiffness.
5. Conclusions
A roller-raceway contact model based on non-Hertz line contact
theory was developed to investigate the effects of defect on roller-
raceway contact. The geometric features of natural defect were consid-
ered in the contact modeling. The roller-raceway contact was analyzed
under different load and defect size. The results showed that the adverse
effects of defect on roller-raceway contact were mainly the great pressure
concentration and contact stiffness decrease. Then the contact stiffness
expressions under different defect size were derived by numerical fitting.
The defected roller bearing was modeled by incorporating the change
of contact stiffness into quasi-static model of roller bearings. The effects
of defect on contact load distribution and stiffness of roller bearings were
analyzed over a range of load, rotating speed and defect size. The analysis
demonstrates that the stiffness of roller bearing is less affected by early
natural defect. When defect area grows to the prevailing fatigue criteria,
the stiffness losses along both radial and axial direction are less than 1%.
As the defect grows further, the number of rollers in the defect region
increases and the roller bearing stiffness decreases significantly. There-
fore, the fatigue criteria should be relaxed to accommodate some
acceptable larger defect size for roller bearing running at medium or low
rotating speed.
Adopting the proposed modeling method, the effects of natural defect
on the motion of roller bearing components (i.e., inner and outer race-
ways, cage, and rollers) will be analyzed more accurately, which can lay
the foundation for more accurate defect diagnosis and life prediction of
roller bearings.
178
Tribology International 122 (2018) 169–178
Acknowledgment
The research work is financially supported by the National Natural
Science Foundation of China (Grant No. 51675410).
References
[1] Lei Y, Lin J, He Z, et al. Application of an improved kurtogram method for fault
diagnosis of rolling element bearings. Mech Syst Signal Process 2011;25(5):
1738–49.
[2] Jia F, Lei Y, Lin J, et al. Deep neural networks: a promising tool for fault
characteristic mining and intelligent diagnosis of rotating machinery with massive
data. Mech Syst Signal Process 2016;72:303–15.
[3] Zhu X, Zhang Y, Zhu Y. Bearing performance degradation assessment based on the
rough support vector data description. Mech Syst Signal Process 2013;34(1):
203–17.
[4] Ali JB, Chebel-Morello B, Saidi L, et al. Accurate bearing remaining useful life
prediction based on Weibull distribution and artificial neural network. Mech Syst
Signal Process 2015;56:150–72.
[5] Hoeprich MR. Rolling element bearing fatigue defect propagation. J Tribol 1992;
114(2):328–33.
[6] Weinzapfel N, Sadeghi F. Numerical modeling of sub-surface initiated spalling in
rolling contacts. Tribol Int 2013;59:210–21.
[7] Niu L, Cao H, He Z. Dynamic modeling of rolling ball bearing with localized surface
defects considering three dimensional motions and relative slippage. J Mech Eng
2015;19:53–9.
[8] Niu L, Cao H, He Z, et al. Dynamic modeling and vibration response simulation for
high speed rolling ball bearings with localized surface defects in raceways. J Manuf
Sci Eng 2014;136(4):041015.
[9] LIU J, SHAO YM, QIN XM, et al. Dynamic modeling on localized defect of
cylindrical roller bearing based on non-hertz line contact characteristics. J Mech
Eng 2014;01:91–7.
[10] Kumar R, Singh M. Outer race defect width measurement in taper roller bearing
using discrete wavelet transform of vibration signal. Measurement 2013;46(1):
537–45.
[11] Jena DP, Panigrahi SN. Precise measurement of defect width in tapered roller
bearing using vibration signal. Measurement 2014;55:39–50.
[12] Renaudin L, Bonnardot F, Musy O, et al. Natural roller bearing fault detection by
angular measurement of true instantaneous angular speed. Mech Syst Signal Process
2010;24(7):1998–2011.
[13] Yadav HK, Pandya DH. Non linear dynamic analysis of cylindrical roller bearing.
Procedia Engineering 2017;173:1878–85.
[14] Yan K, Wang Y, Zhu Y, Hong J, Zhai Q. Investigation on heat dissipation
characteristic of ball bearing cage and inside cavity at ultra high rotation speed.
Tribol Int 2016;93:470–81.
[15] Ashtekar A, Sadeghi F, Stacke LE. A new approach to modeling surface defects in
bearing dynamics simulations. J Tribol 2008;130(4):041103.
[16] Yu HL, Yi JH, Hu X, et al. Study on teeth profile modification of cycloid reducer
based on non-Hertz elastic contact analysis. Mech Res Commun 2013;48(1):87–92.
[17] Shao Y, Liu J, Ye J. A new method to model a localized surface defect in a
cylindrical roller-bearing dynamic simulation. ARCHIVE Proceedings of the
Institution of Mechanical Engineers Part J Journal of Engineering Tribology 1994-
1996 (vols 208-210) 2013;228(2):140–59.
[18] Luo JW, LUO TY. Rolling bearing analysis calculation and application [m]. China
Machine Press; 2009.
[19] Slack T, Sadeghi F. Explicit finite element modeling of subsurface initiated spalling
in rolling contacts. Tribol Int 2010;43(9):1693–702.