NDM 7 PDF

Journal of Sound and Vibration 333 (2014) 5356–5377
Contents lists available at ScienceDirect
Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsvi
Analyses of contact forces and vibration response

for a defective rolling element bearing using
an explicit dynamics finite element model
Sarabjeet Singh a,n, Uwe G. Köpke b, Carl Q. Howard a, Dick Petersen a
a
School of Mechanical Engineering, The University of Adelaide, Adelaide, South Australia 5005, Australia
b
Trackside Intelligence Pty Ltd, 17–19 King William Street, Kent Town, South Australia 5067, Australia
a r t i c l e in f o abstract
Article history: This paper provides insights into the physical mechanism by which defect-related impul-
Received 10 January 2014 sive forces, and consequently, vibrations are generated in defective rolling element
Received in revised form bearings. A dynamic nonlinear finite element model of a rolling element bearing with
17 April 2014
an outer raceway defect was numerically solved using the explicit dynamics finite element
Accepted 3 May 2014
software package, LS-DYNA. A hypothesis was developed to explain the numerical
Handling Editor: L.G. Tham
Available online 5 June 2014 noise observed in the predicted vibrations and contact forces, and the noise frequencies
were analytically estimated. In-depth analyses of the numerically estimated dynamic
contact forces between the rolling elements and the raceways of a bearing, which are not
measured in practice, and have not been reported previously, are presented in this paper.
Several events associated with the traverse of the rolling elements through the outer
raceway defect are elaborated, and the impulsive force generating mechanism is
explained. It was found that the re-stressing of the rolling elements that occurs near
the end of a raceway defect generates a burst of multiple short-duration force impulses.
The modelling results also highlight that much higher contact forces and accelerations are
generated on the exit of the rolling elements out of defect compared to when they strike
the defective surface. A bearing with a machined outer raceway defect was tested in a
controlled experiment; the measured acceleration response compared favourably with
the numerically modelled acceleration results, thereby, validating the low- and high-
frequency characteristics of the de-stressing and re-stressing of the rolling elements,
respectively.
& 2014 Elsevier Ltd. All rights reserved.
1. Introduction
Rolling element bearings are widely used in rotating machinery across various industries that include aerospace,
construction, mining, and railways. The damage and failure of bearings contribute to machinery breakdown, consequently
causing significant economic losses and even loss of human lives in certain situations, such as train derailment due to
seizure of a bearing. Undesirable vibrations in rolling element bearings can be caused by faulty installation, localised defects
n
Corresponding author. Tel.: þ61 8 8362 5445; fax: þ61 8 8362 0793.
E-mail addresses: sarabjeet.singh@adelaide.edu.au (S. Singh), uwek@trackiq.com.au (U.G. Köpke), carl.howard@adelaide.edu.au (C.Q. Howard),
dick.petersen@adelaide.edu.au (D. Petersen).
http://dx.doi.org/10.1016/j.jsv.2014.05.011
0022-460X/& 2014 Elsevier Ltd. All rights reserved.
S. Singh et al. / Journal of Sound and Vibration 333 (2014) 5356–5377 5357
[1–26] or distributed defects [27–35]. While the distributed defects are associated with manufacturing imperfections, such
as roughness, waviness, off-sized components, and misaligned races, the localised defects are caused by surface contact
fatigue [36]. In a properly installed and lubricated bearing, the onset of micro-scale subsurface fatigue cracks commences
below the rolling surfaces that continue to progress towards the surface during the operation of a bearing, eventually
causing the material to break loose or flake off, leading to the formation of macro-scale surface spalls or pits [37]. It is well
known that whenever a defective component, a rolling element, outer raceway or inner raceway, within a bearing interacts
with its corresponding mating components, abrupt changes in the contact stresses occur [38]. These changes structurally
excite the bearing resulting in the generation of vibrations, and consequently, acoustic signals which can be monitored to
detect the presence of a defect using appropriate condition-based (vibration and acoustic) diagnostic techniques [37–41]. A
commonly used vibration-based monitoring technique is to measure the acceleration of a bearing at a suitable location
(most commonly its outer shell or a pedestal into which a bearing is mounted), and implement the envelope analysis
technique [41], also referred to as the high-frequency resonance technique [42]. The presence of impulses in measured
acceleration time-traces usually indicates the presence of a defect within a bearing. This paper is concerned with the
vibrations generated due to a localised outer raceway defect, a line spall, within a rolling element bearing.
Understanding the characteristics of the measured vibration spectrum of a rolling element bearing has received
considerable attention in the literature. As a result, numerous researchers have developed analytical models to simulate the
vibration response of defective bearings due to localised defects [1–26]. These models include impulse-train [1–9] and
nonlinear multi-body dynamic models [10–26]. The impulse-train models simulate the generation of forces produced by a
localised point spall as an infinite series of periodic defect-induced force impulses [1–4]. While equi-amplitude impulses
were generated for stationary outer raceway defects [3,4], the amplitude of the impulses for rotating inner raceway and
rolling element defects [1–4] was modulated as per the static load distribution [43]. The effect of different pulse shapes on
the vibration response of a bearing has also been considered [4]. In order to gain close resemblance to actual vibration
signals obtained in practice, the impulse-train models were extended [5–9] to incorporate the slippage of the rolling
elements that causes slight random variations in the spacing between two consecutive defect-related impulses. The insights
into the envelope (demodulated) vibration spectra of defective rolling element bearings, gained through the impulse-train
models [1–9], provided motivation for later researchers to include various components of a bearing in their models that led
to the development of nonlinear multi-body dynamic models for predicting the vibration response of a bearing, bearing-
housing (pedestal) and rotor-bearing systems, due to the localised defects [10–26]. The defects not only included point spalls
[10,15,18,20,21] (as considered for the impulse-train models [1–9]), but also line spalls [11–14,16,17,22,24,26] as a function
of width and depth, circular spalls [19,23], and area spalls (ellipsoids) [25] as a function of Hertzian contact deformation
[44,45]. These models use lumped-parameters, and generally consider the outer and inner rings as lumped masses and the
rolling elements-to-raceways contact interfaces as nonlinear springs. The common feature of all the models [10–26], except
[25,26], is that they neglected the bending deformation of the outer and inner rings and rolling elements; however,
considered the localised nonlinear Hertzian contact deformation at the rolling elements-to-raceways contact interfaces. To
simplify the analysis, the majority of the multi-body models assume the following: (a) the outer and inner rings are rigidly
connected to the housing [10–24] and shaft [10–26], respectively, (b) the rolling elements are excluded or considered
massless [10–15,17–19,21–23,25,26], (c) the inertial and centrifugal effects of the rolling elements are ignored [10–23,25,26],
and (d) the slippage of the rolling elements [7] is ignored [10–16,18–24,26], thus, eventually, resulting in the generation of
periodic defect-induced impulses. In the work presented here, no such assumptions were considered during the explicit
dynamics finite element modelling of a rolling element bearing with an outer raceway defect. Although most of the models
successfully predict the defect-related fundamental and harmonics, some could not reasonably predict their amplitudes
[4,13–16,22,23]; in some cases [4,13–16], the predicted amplitudes have simply been normalised or corrected based on
experimental results, without providing an explanation. The papers [1–26] describe the vibration response of bearings and
can be used to explain the typical characteristics of experimentally obtained vibration signals; however, they do not discuss
the contact forces between mating defective bearing components. A few explicit dynamics finite element models of ball
bearings with raceway defects have also been developed [46–49]; however, like the previous analytical models [1–26], these
models show bearing vibrations, but not the rolling elements-to-defective raceway contact forces. It is the aim of this paper
to present an in-depth analysis of the numerically modelled dynamic contact forces between the rolling elements and the
defective raceway of a bearing.
In the previous finite element models [47–49], the bearing components were modelled as rigid bodies that cause
artificial over-stiffening of the bearing structure resulting in unrealistically high acceleration levels of approximately 4,000g
and 15,000g compared to the corresponding experimental results of 100g and 10g in Refs. [48] and [49], respectively. In
Ref. [47], the numerically predicted acceleration levels of 107g were not compared with experimental results. In the work
presented here, the components of the defective bearing are flexible (meshed with finite elements), which facilitates a more
accurate representation of the stiffness, and hence, the vibration response of the bearing. Some researchers [50,51] only
considered a small section of a bearing raceway, 5 mm wide, containing a defect along with a rigid (meshless) ball, and
conducted an impact analysis without radial or axial loads, whereas the work presented here includes radial load on the
bearing. They [50,51] investigated stress distribution in the vicinity of the defect. Although the ball-raceway model was
described to be a representative of a bearing model, the stresses are less likely to provide realistic estimates. This is because
when the rolling elements interact with a raceway defect, they are generally loaded (compressed) between the raceways
due to radial or axial loading resulting in their movement being constrained within the raceways. Furthermore, the models
5358 S. Singh et al. / Journal of Sound and Vibration 333 (2014) 5356–5377
in Refs. [50,51] do not predict the vibration response, whereas the work presented here describes the vibration response of a
bearing.
In this paper, a dynamic nonlinear finite element model of a rolling element bearing with an outer raceway line spall was
developed. The model was solved numerically using the explicit dynamics finite element software package, LS-DYNA [52].
Unlike the models in the literature, the finite element model in this paper comprises all the components within a bearing
(outer and inner rings, rolling elements, and cage), and models them as flexible bodies so that they can be placed under
compression and vibrate. Although simplified to consider the translations in the plane of the bearing, the explicit
assumptions made by the previous authors were not considered during the current modelling of the defective bearing;
however, input parameters, such as material model, properties, solver time integration method, friction, and damping, were
assumed during the finite element modelling undertaken here. Therefore, the model presented in this paper is
comprehensive in its capacity to predict the impulsive contact forces that generate bearing vibrations.
The physical mechanism that causes the generation of defect-related force impulses in a defective rolling element
bearing is investigated by analysing the rolling elements-to-raceways contact forces in conjunction with the movement of
the rolling elements as they traverse through the outer raceway defect. Contact forces between mating bearing components
are not measured in practice. Therefore, the contact force results from the numerical simulations presented here provide
new insights into how vibration is generated in a bearing as a rolling element passes through a defective region. Multiple
impulses are generally observed in experimentally measured typical vibration signals; however, the reason for their
occurrences has not been clearly understood [53]. The modelling results in this paper show that during the re-stressing of
the rolling elements, a burst of short-duration force impulses, rather than a single force impulse [1–7,15,53,26], occurs that
excites the bearing resonant modes and explains the occurrence of multiple impulses in experimental vibration signals. The
low- and high-frequency characteristics of the de-stressing and re-stressing events, respectively, obtained from both
experimental and numerical results, which compare favourably with each other, are also discussed.
The features of the finite element model of a bearing are presented in Section 2 along with the numerical bearing
acceleration results. The cause of numerical contact noise observed in the predicted results is elaborated in Section 3 along
with the estimation of the noise frequencies and demonstration of the beating phenomenon to justify the proposed
hypothesis. In Section 4, the validation of numerical contact force results is discussed. Section 5 provides a detailed
discussion of the contact force results between the rolling elements and the raceways of the bearing in conjunction with the
movement of the rolling elements as they traverse through the outer raceway defect. The physical mechanism by which
defect-related impulsive forces are generated is discussed in Section 6 along with the correlation of the forces with
the bearing acceleration. The experimental acceleration of a defective bearing is presented in Section 7 along with its
comparison with the numerically modelled acceleration results. Finally, the conclusions are presented in Section 8.
2. Numerical modelling
A commercially available finite element software package, LS-DYNA [52], was used to analyse the dynamics of a rolling
element bearing. It uses an explicit time integration scheme for simulating highly nonlinear transient dynamic events of
very short durations. This type of finite element code is different from the implicit dynamics software, such as ANSYS [54].
2.1. Description of the finite element model
A two-dimensional (2-D) finite element model of a rolling element bearing was built. The model comprises the following
components: an outer ring, an inner ring, a cage retaining a total of 24 rolling elements, and an adapter that distributes the
(radial) load to the outer ring. The dimensions of the modelled components are shown in Table 1. There was a nominal radial
clearance of 0.02 mm between the rolling elements and the inner (running) surface of the outer ring, more commonly
referred to as the outer raceway. The clearance between the rolling elements and their corresponding cage slots was
0.35 mm. These dimensions are representative of a typical railway bearing that was experimentally tested.
Table 1
Dimensions of the components within the finite element model.
Component name Dimensions (mm)
Outer ring Outer diameter ¼220 Thickness¼ 10

Inner ring Outer diameter ¼163.96 Thickness¼ 10
Cage Outer diameter ¼196 Thickness¼ 4
Rolling elements Diameter ¼ 18
Adapter Width¼ 160 Height (central)¼40
The bearing was modelled with a rectangular defect that was located centrally at the top of the outer raceway. The
dimensions of the defect were 10 mm (circumferential length) 0.2 mm (height/depth). These dimensions were selected
based on the inspection of numerous defective bearings from operational use in the railway industry. Such a defect is often
referred to as a line spall.
2-D shell elements were used to model the bearing as a solid structure. A 2-D element is defined by four nodes having
two degrees-of-freedom at each node: translations in the nodal x- and y-directions. The shell elements were modelled as
plane strain elements [52]. The components of the bearing were modelled using the material properties of steel of density
ρ ¼ 7850 kg=m3 , modulus of elasticity E ¼ 200 GPa, and Poisson's ratio ν ¼ 0:3. The isotropic elastic material model was
chosen for the analysis.
2.2. Discretisation of the finite element model
The discretisation of a model into nodes and elements is an important step in a finite element analysis as the accuracy of
the results depends on the quality of the mesh, and size and aspect ratio of the elements. All the components within the
model were meshed using quadrilateral elements, except the rolling elements which due to their geometry could not be
meshed with the quadrilateral elements. They were meshed with a mixture of quadrilateral and triangular elements.
The rolling elements need to maintain a continuous rolling contact with the outer and inner raceways in order to
transmit the load between the raceways. For meshing the rolling elements and raceways, the optimal element size
was determined based on the compliance of the simulation results with the following conditions: (1) the surfaces of the
rolling elements, which are under the influence of load zone, and raceways should be in contact at all times during the
simulation, and (2) the rolling elements should predominantly roll and not slide during the simulation. It should, however,
be noted that in practice, there is a small amount of sliding or slipping associated with the rolling elements [7]. While the
fulfilment of the first condition is necessary to achieve the correct load distribution on the rolling elements as per the
analytical static solution [55, pp. 234–237], the accomplishment of the second condition is necessary to accurately acquire
the rotational speed of the rolling elements (or cage), which would eventually result in the correct prediction of the defect
frequency.
In a typical operation of a bearing in practice, the rolling elements rotate about the axis of the bearing and their own
axes. In order to simulate the real-scenario of a bearing operation, no boundary conditions were applied to the rolling
elements. Therefore, satisfactory achievement of the second condition is significantly important for the reasons mentioned
above. It should, however, be noted that the two conditions are interrelated to each other as the loss of rolling elements-to-
raceways contact will affect the rotational speed of the rolling elements.
The mesh element sizes of 1 mm, 0.75 mm, 0.5 mm and 0.3 mm were tested. For the finite element models meshed using
the element sizes of 1 mm and 0.75 mm, it was found that a continuous contact between the rolling elements and the
raceways could not be maintained during the simulation. The loss of the contact caused incorrect load distribution on
the rolling elements. On the contrary, for the finite element models meshed using the element sizes of 0.5 mm and 0.3 mm,
the two aforementioned conditions were fulfilled, which consequently resulted in the accurate acquisition of the bearing
kinematics. The simulation results, rolling elements-to-raceways contact forces and acceleration, for the models meshed
using 0.5 mm and 0.3 mm element sizes, were very similar to each other. Therefore, the mesh element size of 0.5 mm was
chosen to reduce the CPU run times without compromising the accuracy of the numerical solution.
The manual [52] of the finite element software package used here does not mention a specific criterion for an approp-
riate mesh size in order to accurately predict the dynamic response of a structure. However, it is generally recommended to
use at least 20 elements-per-wavelength (EPW) for a transient structural dynamic analysis for wave propagation studies [56,
Chapter 5]. It is interesting to note that the chosen element mesh size of 0.5 mm corresponds to 97 EPW at 40 kHz, which is
nearly five times the recommended EPW value. The EPW was calculated based on the propagation of bending wave speed in
the outer ring of the bearing. Considering the outer ring as a thin plate, the speed of bending waves, also referred to as
flexural waves, is given by [57, pp. 354–355]:
" #1=4
2
pffiffiffiffi Eh
Cb ¼ ω (1)
12ð1 ν2 Þρ
where E is the modulus of elasticity, ρ is the density, ν is Poisson's ratio, h is the thickness, and ω is the angular frequency; it
corresponds to the highest frequency of interest at which results are sought. Considering that the results are sought at
40 kHz, the bending wave speed at 40 kHz for a 10 mm thick plate equates to 1959.32 m/s, and the corresponding
wavelength λb equals 0.0489 m – this is equivalent to approximately 48 EPW for the mesh element size of 1 mm. Although
one can reasonably predict the structural modal response of a bearing by having at least 20 EPW, ensuring a smooth and
continuous rolling contact requires much more EPW.
Fig. 1(a) shows the meshed finite element bearing model annotated with the names of the components. The geometrical
rectangular defect located centrally at the top of the outer raceway, which cannot be seen in Fig. 1(a), is shown in Fig. 1(b) for
clarity. The centre of the rolling element located immediately to the left-hand side of the defect in Fig. 1(b) is offset by 41
from the y-axis; the rolling elements within the model are 151 apart.
Fig. 1. Images of the meshed 2-D finite element bearing model; (a) the model showing the names of the bearing components, and (b) a zoomed image of
Fig. 1(a) showing the 1-element deep rectangular defect on the running surface of the outer ring highlighted using the ellipse; the centre of the rolling
element to the left-hand side of the defect in Fig. 1(b) is offset by 41 from the y-axis.
2.3. Loads and boundary conditions
The following boundary conditions were applied to the finite element model in order to simulate the real-time operation
of a bearing in a typical railway application. These boundary conditions effectively simulate the experimental setup condi-
tions applicable to a bearing during its testing; the experimental setup is described in Section 7.1.
A radial load of 50 kN on the top edge of the adapter in the negative global y-direction was applied so as to radially load
the bearing; this load represents half of the load carrying capacity of the bearing.
The inner ring was constantly rotated with a uniform angular velocity of 500 revolutions per minute (rpm) in a clockwise
direction.
In a typical three-piece railway bogie, the adapter is located on the top of a bearing, and is sandwiched between the bearing
and one of the two side-frames [58, p. 70]. During the travel of a train (operation of a bearing), the adapter remains at its
position due to the weight of the railway vehicle which is transferred on to the (four) bearings within a bogie through the
side-frame–adapter structural coupling. The coupling between the adapter and the outer ring of a bearing is such that there is
a conformal contact between them, as a result of which the outer ring of a bearing also remains at its position, and does not
rotate. Implementing the coupling between the adapter and the outer ring in the finite element model was challenging as
no direct constraints could be applied to the outer ring because this would result in over constraining its translations
and vibration response, consequently, causing incorrect load distribution on the rolling elements. In order to overcome this
problem, a frictional contact with a high coefficient of friction, 0.1, between the outer ring and the adapter was implemented.
The top edge of the adapter was translationally constrained in the global x-direction, and as a result of the frictional contact,
the outer ring was prevented from being rotated during the simulation.
A frictional contact with a low coefficient of friction, 0.005, was defined for the following contact interfaces within the
model: rolling elements – outer ring, rolling elements – inner ring, and rolling elements – cage. The chosen frictional
coefficient is comparable to that of the friction generally recommended for rolling element bearings in practice. In
addition to modelling the surface-to-surface contact at the aforementioned interfaces, the segment-based contact
formulation [52] was implemented during the numerical simulation. It detects the penetration of one segment into
another and applies a force, proportional to the penetrating depth, to the segment nodes, as opposed to the penetrating
nodes only for the standard node-to-surface penalty-based contacts often used in numerous implicit finite element
solvers including ANSYS [54].
A global (mass-weighted) damping of 2 percent was applied to the finite element model.
The standard Earth's gravity was also applied to the finite element model.
2.4. Analysis and control settings
The termination time of the numerical simulation was set to 30 ms. The results (in the form of binary text files) were
written at the interval of 0.01 ms, which corresponds to a sampling rate of 100 kHz. Despite requesting the results at specific
time intervals, the numerical results are often output at slightly different intervals, which are irregularly spaced, but close
to the requested intervals. This is due to the variation of the time steps that advance the explicit numerical simulation.
Therefore, as part of post-processing the results, the software MATLABs was used to interpolate the results at the specific
time interval.
The accuracy of the finite element analysis was verified by checking the distortion of the elements. In the finite element
software package used here, erroneous or severe distortion of the elements leading to poor element aspect ratio is referred
to as hourglassing or zero-energy modes [52]. It is generally recommended that the hourglass (HG) energy of an element
should be less than 10 percent of its internal energy. For the current simulation, the HG energy was found to be less than
0.2 percent of the internal energy, indicating negligible elemental distortions. One of the reasons for having such a low HG
energy is the uniform discretisation of the bearing model into finite elements using mostly quadrilateral elements.
2.5. Results and observations
A common technique used in vibration-based condition monitoring of rolling element bearings is the measurement of
acceleration levels. Similarly, in these simulations, the nodal acceleration results obtained from the numerical modelling are
presented.
2.5.1. Acceleration time-history

Fig. 2 shows the (unfiltered) time–history plot of the numerically obtained acceleration results for a node located on the
outer surface of the outer ring. The consecutive defect-related impulses, evident in the plot, are separated by approximately
0.011 s, which corresponds to the outer raceway defect frequency, commonly referred to as a ball pass frequency outer
raceway (BPFO), f bpo , of 90.91 Hz. The analytical estimation of the nominal BPFO, f bpo , is given by [55, p. 994]

N fs Dr
f bpo ¼ b 1 cos ðαÞ (2)
2 Dp
where Nb is the number of rolling elements, fs is the run speed (that is, the rotational speed of shaft or inner ring), Dr is the
rolling element diameter, Dp is the bearing pitch diameter, and α is the contact angle. For the bearing modelled here, the
nominal estimate of the BPFO is 90.07 Hz, which is 0.9 percent different from the numerical simulation results. The slight
difference between the numerical and nominal estimations is because the analytical formula, shown in Eq. (2), does not
account for the slippage of the rolling elements [6,7], which was accounted in the explicit finite element analysis of the
bearing undertaken here.
A close agreement between the numerical and analytical values of the outer raceway defect frequency shows that
the numerical model has satisfactorily simulated the basic bearing kinematics. However, the acceleration signal has a
substantial amount of numerical noise that is explained in the following section.
3. Numerical noise
It can be observed from Fig. 2, while the instantaneous peak impulsive acceleration levels for the three visible defect-
related impacts range from 0 to approximately 7180g, the non-impulsive acceleration levels between the impacts are of
200
150
100
Acceleration [g]
50
50
100
1/f bpo (1/90.91Hz) =0.011s 1/f bpo (1/90.91Hz) =0.011s
150
200
0 5 10 15 20 25 30
Time [ms]
Fig. 2. Numerically modelled unfiltered acceleration results for a node located on the outer surface of the outer ring of the finite element bearing model for
the bearing run speed of 500 rpm.
40
numerical noise frequency 4671 Hz

20
Acceleration [dB re 1 (m/s2 ) 2 /Hz]

0
20
40
60
unfiltered
filtered
80
0 5 10 15 20 25
Frequency [kHz]
Fig. 3. Power spectral density of the numerical acceleration results highlighting the numerical noise frequency in the simulation results.
the order of 750g. The power spectrum of the unfiltered acceleration signal is shown in Fig. 3. A fundamental tone at
4671 Hz, as indicated in the figure, corresponds to the numerical noise. The 5th harmonic at approximately 23 kHz
associated with the fundamental tone can also be seen in the figure. The tonal noise at 4671 Hz was notch filtered from the
numerical acceleration signal, and the corresponding power spectrum is also included in Fig. 3 for comparison. The filter
attenuated the tone at 4671 Hz by 25 dB without affecting the response at other frequencies.
Results from explicit dynamics finite element analysis can contain a significant amount of numerical noise [59, p. 1110].
In the model developed here, the numerical noise could be due to the interaction (contact) of the rolling elements and
raceways. In addition to the dominant numerical noise at 4671 Hz, another frequency component that was intermittently
observed was 4545 Hz. In order to explain the cause of numerical noise along with the estimation of the noise frequencies, a
detailed hypothesis is presented below.
3.1. A hypothesis for explaining the cause of numerical contact noise
Because the circular rolling elements were discretised into a number of finite elements, the edges of the rolling elements
were transformed from circular to multi-point polygons. As they roll during the simulation, the polygonised rolling elements
create small impacts upon their contact with the outer and inner raceways. In this case, the frequencies of these impacts
would be a function of the element size used to mesh the rolling elements and the rotational velocity of the rolling
elements.
In order to estimate the numerical noise frequency components, a basic equation of motion can be used as given by
1 ωr r race
f noise ¼ ¼ (3)
T noise dmesh
where dmesh is the distance between two nodes in the finite element model (mesh element size, 0.5 mm), ωr is the angular
velocity with which the rolling elements roll during the simulation, and r race can either be the radius of the outer raceway
(100 mm) or inner raceway (81.98 mm) which contact the rolling elements. Solving Eq. (3) for the values of r race as 100 mm
and 81.98 mm, the noise frequencies equal 4712 Hz and 3864 Hz, respectively. These frequencies are referred to as the
rolling elements-to-outer raceway f noiseOR , and rolling elements-to-inner raceway f noiseIR noise frequencies, respectively.
The comparison of the analytical estimate of the former noise frequency f noiseIR , 4712 Hz, with the results from the finite
element simulation of 4671 Hz (a 0.8 percent difference) indicates that the presence of the numerical noise at 4671 Hz is
highly likely due to the interaction of the rolling elements with the outer raceway. The difference between the analytical
and numerical estimations is because the rolling elements do not follow a pure rolling movement during the simulation,
indicating a certain degree of slip. It was mentioned earlier that no boundary conditions were applied to the rolling
elements, and they were driven by their interaction with the rotating inner and stationary outer raceways. The slip of the
rolling elements occurs due to the contact between the (finite) elements located at the edges of the rolling elements and
inner raceway as they progress with time.
Another reason for the difference between the analytical and numerical estimations of the noise frequencies is associated
with the interaction between the rolling elements and corresponding cage slots. It was found that at certain instances the
rolling elements were driven (pushed) by the cage slots that consequently result in the slipping of rolling elements. The
interaction between the rolling elements and the cage will be discussed in Section 5.2.
3.1.1. Beating phenomenon

There is a difference of approximately 17 percent between the other numerical noise frequency, 4545 Hz, and the analyti-
cally estimated rolling elements-to-inner raceway contact noise frequency f noiseIR , 3864 Hz. As the difference is significant,
the analytical and numerical noise frequencies cannot be related. Therefore, the concept of beating [57, p. 45] was applied to
explain the occasional presence of 4545 Hz noise frequency component found in the numerical acceleration results.
As mentioned earlier, the polygonised edges of the rolling elements create small impacts with outer and inner raceways
as they roll during the simulation. The interaction of the rolling elements with the raceways would result in the generation
of two sinusoidal waves with a slight difference between their carrier frequencies. The amplitude of the sinusoidal waves
would also slightly differ from each other. For the purpose of verifying the aforementioned hypothesis, and demonstrating
the beating effect, the sum of two interfering sinusoidal waves is
PðtÞ ¼ A1 cos ð2πf noiseIR tÞ þ A2 cos ð2πf noiseIR tÞ (4)
where the amplitudes A1 ¼ A2 ¼ 1, t is the time vector, and f noiseIR and f noiseIR are the analytically estimated noise frequencies
using Eq. (3) (f noiseIR ¼ 4712 Hz and f noiseIR ¼ 3864 Hz).
Fig. 4(a) shows the resultant sinusoidal wave, and the same wave along with its envelope, zoomed from 9 ms to 15 ms for
clarity, is shown in Fig. 4(b). The time separation of the two consecutive peaks, whose data cursors are shown in Fig. 4(b),
corresponds to 4545 Hz. This frequency exactly matches the other noise frequency component, occasionally observed in
the numerically modelled acceleration results. The beating effect can also be clearly observed in Fig. 2 between the first two
defect-related impulses from approximately 5 ms to 16 ms, but not as clear between the second and third impulses from
approximately 17 ms to 28 ms.
The notch filtered numerical acceleration results are shown in Fig. 5. For comparison purposes, the unfiltered accelera-
tion results, shown in Fig. 2, are also plotted along with the filtered results. The performance of the filter is evident in Fig. 5;
the amplitudes of the non-impulsive acceleration signals, which prior to the application of the filter ranged from 0 to
approximately 750g, were reduced to approximately 720g after the filtering. However, there is still some residual noise. It
is very likely that the remaining noise is due to the sliding (slippage) of the rolling elements as a result of their interaction
with the cage slots. The sliding noise frequencies are stochastic, and therefore, could not be estimated. The randomness of
the sliding noise frequencies is one of the reasons for not having the clear beating effect from 17 ms to 28 ms in Fig. 2, and
the intermittent appearance of vibration at 4545 Hz.
A favourable agreement between the numerical and analytical noise frequencies justifies the proposed hypothesis. From
the aforementioned discussion, it can be concluded that the introduction of the noise in the simulation results is an artefact
of the numerical modelling characteristics. Although the time-history acceleration plot, shown in Fig. 5, clearly shows the
defect-related impulses whose frequency also matches well with the nominal defect frequency, further explanations are
required to describe the physical mechanism by which these impulses are generated. This will be presented in Sections 5
and 6 during the analysis of the rolling element-to-raceway contact forces, followed by the validation of the numerically
predicted contact forces in the next section.
2
Amplitude
4
0 5 10 15 20 25 30
Time [ms]
4 X: 11.67 X: 11.89
Y: 1.863 Y: 1.899
2
Amplitude
4
9 10 11 12 13 14 15
Time [ms]
Fig. 4. Demonstration of the beating effect due to the interference of two sinusoidal waves at the two analytically estimated noise frequencies; (a) resultant
sinusoidal wave and (b) the wave along with its envelope zoomed for clarity.
200
unfiltered
150 filtered
100
Acceleration [g]
50
50
100
150
200
0 5 10 15 20 25 30
Time [ms]
Fig. 5. The effect of filtering out the numerical noise on the modelled acceleration results shown in Fig. 2.
10
Contact force Fy [kN]
event # 1 numerical
8
analytical
6
4 event # 2
event # 4 event # 3
2
event # 3
0
0 5 10 15 20 25 30
Time [ms]
10
event # 3
8 event # 1
6
4 event # 2
event # 3
event # 4 numerical
2
analytical
0
0 5 10 15 20 25 30
Time [ms]
10
event # 4
numerical
8 analytical event # 1
6
event # 2
4
event # 3
2
event # 3
0
0 5 10 15 20 25 30
Time [ms]
Fig. 6. Numerically and analytically (quasi-static) modelled results for the vertical (y-direction) contact forces, Fy, between the outer raceway and the three
rolling elements that traversed through the outer raceway defect; (a) outer raceway-to-rolling element 1, (b) outer raceway-to-rolling element 2, and
(c) outer raceway-to-rolling element 3.
4. A discussion on the validation of numerical contact forces
Figs. 6 and 7 show the numerically modelled filtered vertical (y-direction), Fy, and horizontal (x-direction), Fx, contact
force plots between the three rolling elements that traversed through the rectangular defect and the outer raceway, in a
chronological order from top to bottom. The contact force results also contained the numerical noise which was filtered out;
Contact force Fx [kN]

2 event # 2
1
event # 3 event # 3
0
event # 4
−1
numerical
−2
event # 1 analytical
−3
0 5 10 15 20 25 30
Time [ms]
3
2
event # 1 event # 2
1
event # 3 event # 3
0
event # 4
−1
numerical
−2 analytical
−3
0 5 10 15 20 25 30
Time [ms]
3
numerical event # 4
2
analytical event # 1 event # 2
1
0
event # 3 event # 3
−1
−2
−3
0 5 10 15 20 25 30
Time [ms]
Fig. 7. Numerically and analytically (quasi-static) modelled results for the horizontal (x-direction) contact forces, Fx, between the outer raceway and the
three rolling elements that traversed through the outer raceway defect; (a) outer raceway-to-rolling element 1, (b) outer raceway-to-rolling element 2, and
(c) outer raceway-to-rolling element 3.
however, some random noise is still present in the results. The quasi-static contact forces on the corresponding rolling
elements, which were analytically estimated using the model presented in Ref. [60], have also been plotted, using dashed
lines, along with the numerically modelled dynamic contact force results for comparison. In order to verify the load
distribution within the finite element bearing model, the static contact forces on the rolling elements can be compared with
the corresponding analytical solution from Ref. [55, pp. 234–237]. The static contact forces from the numerical finite
element simulation were obtained at time t ¼0, which correspond to the mechanically stressed levels prior to the
commencement of the dynamic analysis. As per the modelled parameters, the numerical and analytical static contact forces
on seven rolling elements within the bearing load zone are compared in Fig. 8. The rolling element positions are plotted
relative to the centre of the defect where the radial load was applied; that is, along the y-axis. As mentioned earlier, the
angular separation between the rolling elements is 151, and as shown in Fig. 1, the first rolling element is offset by 41 to the
left-hand side from the y-axis. For the modelled system, the theoretical estimate of the angular extent of the bearing load
zone is approximately 1061 (7531 from the y-axis).
It is necessary for the numerical solution to have the applied radial load (in this case, on the top of the adapter)
appropriately distributed on to the rolling elements as per the static solution. The comparison of the numerical and analytical
results for the outer raceway-to-rolling elements contact forces in Fig. 6-8 shows that both results compare favourably with
each other. In Figs. 6-7, the numerical contact forces are within 3 percent of the corresponding analytical estimates. A potential
reason for the slight mismatch is that during the numerical simulation, the outer ring, the inner ring, and the rolling elements
were modelled as flexible bodies compared to assuming them as rigid bodies during the analytical modelling. Due to the
ability of the finite elements of the bearing model to vibrate during the numerical solution, the interaction of the polygonised
edges of the rolling elements and raceways could result in the excitation of non-physical modes.
For a 2-D finite element analysis, except the contact force, the contact width and area at the interface of contacting
bodies are not available from the calculations. Instead, they can be estimated on the basis of the contact force using Hertz's
theory. The numerical and analytical estimates of the contact width at the loaded rolling element-to-raceway contact
interfaces are compared in Table 2. The numerical estimations are about 2 percent higher than the analytical results. The
contact width at the interfaces is less than 0.5 mm, the element mesh size used to discretise the finite element model. This

2
1
0
1
numerical
2 analytical
3
60 40 20 0 20 40 60
Rolling element position relative to defect centre ( y axis)
10
Contact force F [kN]
numerical
8
analytical
y
0
60 40 20 0 20 40 60
Rolling element position relative to defect centre ( y axis)
Fig. 8. Numerically and analytically modelled load distribution on the rolling elements within the load zone of the bearing; (a) horizontal (x-direction)
contact forces, Fx, (b) vertical (y-direction) contact forces, Fy.
Table 2
Comparison of the analytical and numerical contact width at the rolling element-to-raceway interfaces.
Roller position relative to y-axis (deg) Contact width (mm)
Analytical Numerical
49 0.034 0.035
34 0.142 0.145
19 0.220 0.225
4 0.256 0.262
11 0.245 0.250
26 0.188 0.192
41 0.094 0.096
means that only one node was present at the rolling element-to-raceway contact interfaces during the numerical simula-
tion. Although this would generate inaccurate results for the standard node-to-surface penalty-based contact formulations
used in many other finite element software packages [54], the surface-to-surface segment-based contact formulation
employed in the finite element software package used here does not suffer from this problem. The favourable comparison of
the numerically estimated contact forces with the corresponding analytical results is evident in Figs. 6–8.
5. Contact force analysis
The numerically modelled contact forces between two contact interfaces, (1) outer raceway-to-rolling elements, and
(2) inner raceway-to-rolling elements, were analysed. Figs. 6 and 7 show four events, labelled event #1 to event #4. For the
purpose of relating these events with the movement of the rolling elements, refer to Fig. 9. It shows a schematic of a 2-D
bearing model comprising an outer ring, an inner ring, a few rolling elements, and a rectangular defect. The rolling elements
marked as ‘1’, ‘2’, and ‘3’ are the ones that traversed through the defect in the clockwise direction during the numerical
simulation undertaken for the paper. The starting and ending positions of the defect are also shown in the schematic.
Throughout the discussion to follow in the paper, the ‘top surface of the defect’ is referred to as the ‘defective surface of the
outer raceway’ as shown in the schematic.
As the radial load was applied at the top of the bearing so that the defect is under maximum load, the rolling elements
which traverse through the defect will have higher magnitude of vertical contact forces Fy compared to the corresponding
horizontal forces Fx. Therefore, it is convenient to analyse the variations in the vertical outer raceway-to-rolling elements
contact force results shown in Fig. 6, although similar variations can also be seen in the horizontal force components in Fig. 7
but comparatively less clear.
Starting position Ending position

of the defect of the defect
1
2
3
Top surface
of the defect
Clockwise
inner ring
rotation
Fig. 9. A schematic of a 2-D bearing model showing an outer ring, an inner ring, a few rolling elements, and a rectangular defect (not to scale).
5.1. Event #1: Entry of the rolling elements into the defect – the de-stressing phase
The application of a radial load on a bearing results in the loading of a certain number of the rolling elements within the
bearing; the extent of the load zone depends on the diametral clearance in addition to the geometry of the bearing. As
a result, the rolling elements, which are under the influence of the load zone, are stressed between the outer and inner
raceways. The maximum stress on the rolling elements at their two respective points of contact with outer and inner
raceways can be analytically estimated using the classical Hertz theory of elasticity [44,45].
The results of the numerically modelled contact forces between the rolling elements and the outer raceway in Figs. 6
and 7 show that as a stressed rolling element starts entering the defect, located on the outer raceway, it gradually starts
losing contact with the outer raceway. From the figures, the time instances at which rolling elements ‘1’, ‘2’, and ‘3’ start
entering the rectangular defect correspond to approximately 1 ms, 12 ms and 23 ms, respectively. Consequently, the contact
forces between the rolling elements and the outer raceway start decreasing gradually to zero. The reduction of the vertical
contact forces Fy from approximately 6 kN to 0 kN implies that the rolling elements have no mechanical stress in contrast to
their location between the raceways where they were stressed and loaded due to the application of the radial load.
The gradual drop in the contact force between the three rolling elements, which traversed through the defect during the
simulation, and the outer raceway is highlighted using the solid elliptical markers in all the subplots of Figs. 6 and 7.
Event #1, which corresponds to the entry of the rolling elements into the defect, can be referred to as ‘the de-stressing of the
rolling elements’. The occurrences of the consecutive de-stressing events are separated by approximately 0.011 s, which
corresponds to (the BPFO of) 90.91 Hz.
The numerically modelled contact force results show that the de-stressing of the rolling elements shows no indication of
an impact, and therefore, cannot excite a broad range of frequencies which generally causes the ringing of a bearing. Sawalhi
et al. [53] discussed the de-stressing of the rolling elements on the basis of their experimental findings. They analytically
modelled this event as a step (vertical) response; however, from the numerical contact force results shown in Figs. 6 and 7, it
is evident that the de-stressing of the rolling elements is not a step response, but a gradual decrease. Because no impulse-
like signals are generated during the entry of the rolling elements into a defect, the de-stressing is a low-frequency event as
mentioned in Refs. [53,26]. The experimental results of the defective bearing with the machined outer raceway line spall
also show the low-frequency characteristics of the de-stressing event; it will be discussed in Section 7.2 of this paper.
5.2. Event #2: Traverse of the rolling elements through the defect – impact of the rolling elements with the defective surface
The zero values of the contact forces between the rolling elements and the defective part of the outer raceway indicate
that the rolling elements unload as they traverse through the defect. The numerically modelled contact force results in
Figs. 6 and 7 show that during the traverse of the rolling elements through the defect, the contact force remains zero for
most of the time except for a few instances. These instances, which are indicated as event #2 in the subplots of Figs. 6 and 7,
correspond to the impact of the rolling elements with the top surface of the defect – the defective surface of the outer
raceway (refer to Fig. 9 to see the top surface of the defect).
During the operation of a bearing, the rolling elements possess centrifugal forces. While the centrifugal forces tend to
project the rolling elements outward from the centre of the bearing, their outward trajectory is restricted between the outer
and inner raceways when the rolling elements are stressed between the raceways. However, when the rolling elements are
de-stressed due to the loss of contact with the outer raceway while they traverse through the defect, they tend to freely
follow their outward trajectory. At some instance, they impact the top surface of the defect. Such impacts are evident in the
subplots of Figs. 6 and 7, and are indicated as event #2. It should, however, be noted that the magnitude of these impacts for
the horizontal contact forces Fx is lower than the corresponding vertical components Fy due to the aforementioned reason,
and therefore, these impacts are not clearly visible in Fig. 7.
It can be seen from the vertical outer raceway-to-rolling elements contact forces Fy that the first and third rolling
elements impact the top surface of the defect twice (Fig. 6(a) and (c)), while the second rolling element impacts the defect
surface three times (Fig. 6(b)). It is also interesting to note that not only the amplitudes of these impacts differ from each
other, but also the timings of the impacts. The reason for the varying period is due to the interaction of the rolling elements
with the adjacent cage slots. The irregular period of the impact timings implies that at some instances the rolling elements
are driven by the cage (slots), and at other instances, the rolling elements drive the cage. The aforementioned interaction is
also another cause for the slippage of the rolling elements observed in practice in addition to the explanation provided in
Ref. [7]. It is one of the reasons for the slight difference (0.8 percent) between the analytical and numerical estimations of
the rolling elements-to-outer raceway contact noise frequency as discussed in Section 3.1.
5.3. Event #3: Re-distribution of load on the rolling elements – the load compensation phase
It has been described during the discussion pertinent to event #2 that the rolling elements lose contact with the
damaged part of the outer raceway as they successively traverse through the defect. The loss of the load (force) on a rolling
element must be compensated in order to have the equilibrium of the system maintained at all times. Therefore, as one
rolling element loses contact (unloads), other rolling elements in the load zone (located at the non-defective sections of the
raceways) take the lost load resulting in the re-distribution of the radial load. Event #3, indicated in the subplots of Figs. 6
and 7, highlights the load compensation.
In the event of a rolling element traversing through a defect, an increase in the magnitude of load on other rolling
elements results in higher contact stresses compared to a non-defective bearing. The higher contact stresses will eventually
result in the onset of localised surface fatigue cracks at approximate roller spacings [61].
5.4. Event #4: Exit of the rolling elements from the defect – the re-stressing phase
When the rolling elements reach the end (position) of the defect (refer to Fig. 9 to see the ending position of the defect),
they again come into contact with the non-defective surface of the outer raceway. At this instance, the contact force on the
rolling elements increases from zero to a certain value that is related to the static load distribution on the rolling element.
The resumption of the contact between the rolling elements and the outer raceway causes the re-stressing of the rolling
elements. Event #4, which corresponds to the re-stressing of the rolling elements between the outer and inner raceways, is
highlighted using the solid rectangular markers in all the subplots of Figs. 6 and 7.
Prior to regaining the full contact with the raceways, the free outward trajectory of the de-stressed rolling elements was
again restricted between raceways. The rolling elements were gradually wedged between the outer and inner raceways, and
during this period, they continuously rattle between the raceways until they are fully loaded (stressed). The rattling of the
rolling elements results in the generation of a few short-duration (contact) force impulses that excite the high-frequency
bearing resonant modes. The subplots of Figs. 6 and 7 were zoomed in the vicinity of the re-stressing of the rolling elements,
and are shown in Fig. 10 in order to highlight the occurrence of the rattling phenomenon. Fig. 10(a)–(c) shows the vertical
outer raceway-to-rolling elements contact forces Fy, and Fig. 10(d)–(f) shows the horizontal contact forces Fx. While the y-
axis for the vertical contact force plots scales from 0 to 10 kN, it scales from 0 to 1 kN for the horizontal contact forces.
Despite being different by an order of magnitude, the timings at which the vertical and horizontal multiple force impulses
occur are comparable.
In contrast to the de-stressing of the rolling elements upon their entrance into the defect where no force impulses were
generated, the contact force results in Fig. 10 show that the exit of the rolling elements from the defect causes a few short-
duration force impulses that can result in the ringing of a bearing. Previous analytical models [1–7,15,53,26] were based on
the assumption that the defect-related impulsive vibration response of a bearing is generated by a single force impulse
which exponentially decays over time. In contrast, the experimentally measured vibration signals shown in Fig. 16 (to be
discussed later in Section 7) and in the literature [17,26,53] clearly show the occurrence of multiple impulses in the time-
domain vibration response of defective rolling element bearings. A tentative explanation about the occurrence of multiple
impulses is provided in Ref. [53] that they could be due to the beating effect related to small differences in the resonance
frequencies of a bearing. However, the numerical modelling results presented here show that a burst of short-duration force
impulses occurs as the rolling elements re-stress between the raceways on their exit out of the defect, which consequently,
generates multiple impulses in the vibration response of a defective bearing. This is an important outcome of the model as
it provides a firm confirmation as to why in typically measured vibration signals: (1) multiple impulses are observed, and
(2) the defect-related impulses do not uniformly decay in a pattern similar to that of a single force impulse.
10 1

8 0.8
y
x
6 0.6
4 0.4
2 0.2
0 0
5.2 5.4 5.6 5.8 6 5.2 5.4 5.6 5.8 6
Time [ms] Time [ms]
10 1

8 0.8
y
6 0.6
4 0.4
2 0.2
0 0
16 16.2 16.4 16.6 16.8 17 16 16.2 16.4 16.6 16.8 17
Time [ms] Time [ms]
10 1
8 0.8
y
6 0.6
4 0.4
2 0.2
0 0
27.2 27.4 27.6 27.8 28 27.2 27.4 27.6 27.8 28
Time [ms] Time [ms]
Fig. 10. Numerically modelled vertical (y-direction) and horizontal (x-direction) outer raceway-to-rolling elements contact forces shown in Figs. 6 and 7,
respectively are zoomed in the vicinity of the rolling elements being re-stressed between the raceways; (a, d) outer raceway-to-rolling element 1,
(b, e) outer raceway-to-rolling element 2, and (c, f) outer raceway-to-rolling element 3. The y-axis in (a–c) scales from 0 to 10 kN, and in (d–f) from
0 to 1 kN.
It is also interesting to note that the results of the numerically modelled vertical outer raceway-to-rolling elements
contact forces Fy (Fig. 6) show that the amplitude of the impacts produced during the re-stressing phase, approximately
5 kN, is higher than those produced when the rolling elements impact the top surface of the defect, which is less than 2 kN
(event #2). This is another important outcome of the modelling as it indicates that higher forces and stresses are generated
during the exit of the rolling elements out of the defect compared to when they impact the defect surface, and hence, could
lead to the gradual expansion or lengthening of the defect. These findings show a firm agreement with the experimental
study conducted by Hoeprich [60]. He investigated the damage progression in rolling element bearings, and found that the
size of a spall progresses in the rolling direction.
6. Correlating contact forces with bearing vibrations
6.1. Physical mechanism that generates defect-related impulsive forces
In order to explain the physical mechanism by which impulsive forces in defective bearings are generated, it is useful to
compare the contact forces on the rolling elements with both raceways. Fig. 11(a)–(c) shows both sets of the numerically
modelled filtered vertical contact force plots for the three rolling elements as they traverse through the defect. The outer
10 10
outer race−roller 1

8 inner race−roller 1
5
outer race−roller 1 6
0
inner race−roller 1
4
−5
2
−10 0
0 5 10 15 20 25 30 5.2 5.4 5.6 5.8 6
Time [ms] Time [ms]
10 10
outer race−roller 2 outer race−roller 2

inner race−roller 2 8 inner race−roller 2
5
6
0
4
−5
2
−10 0
0 5 10 15 20 25 30 16 16.2 16.4 16.6 16.8 17
Time [ms] Time [ms]
10 10
outer race−roller 3
8 inner race−roller 3
5
outer race−roller 3 6
0
inner race−roller 3
4
−5
2
−10 0
0 5 10 15 20 25 30 27.2 27.4 27.6 27.8 28
Time [ms] Time [ms]
Fig. 11. Numerically modelled results for the vertical contact forces, Fy, between two contact interfaces: (1) outer raceway and the three rolling elements that
traversed through the defect, and (2) inner raceway and the rolling elements; (a) raceways-to-rolling element 1, (b) raceways-to-rolling element 2, (c) raceways-
to-rolling element 3, and (d–f) contact forces shown in (a–c), respectively, are zoomed in the vicinity of the rolling elements being re-stressed between the
raceways. The inner raceway-to-rolling elements contact forces represented by broken lines in (a–c) are inversed and changed to solid thick lines in (d–f).
raceway-to-rolling elements and inner raceway-to-rolling elements contact forces closely follow each other; however, in
opposite directions. Although not annotated, the four events highlighted in Fig. 6 (for the outer raceway-to-rolling elements
contact forces) can also be seen in Fig. 11(a)–(c) for the inner raceway-to-rolling elements contact forces.
The contact forces in Fig. 11(a)–(c) show that the de-stressing and re-stressing of the rolling elements, upon their entry
into and exit out of the defect, respectively, occur at the same instances of time. In contrast to the de-stressing, the analysis
of the re-stressing phase is significantly important to gain an understanding of the physical mechanism by which impulsive
forces are generated in defective rolling element bearings. As discussed in Section 5.4, a burst of short-duration impulses is
generated at the re-stressing of the rolling elements in the vicinity of the end of the defect. Therefore, the raceways-to-
rolling elements contact forces plotted in Fig. 11(a)–(c) are zoomed in the vicinity of the re-stressing phase, and the
corresponding plots are shown in Fig. 11(d)–(f). For the comparison of the time instances related to the interaction of
the rolling elements with outer and inner raceways, the inner raceway-to-rolling elements contact forces (thick lines) are
inversed. It can be seen from the zoomed contact force results in Fig. 11(d)–(f), as the rolling elements re-stress between the
raceways, they alternate between striking the outer and inner raceways. Compared to the outer raceway-to-rolling elements
contact forces (thin lines), the amplitude of the inner raceway-to-rolling elements contact forces is higher during the re-
stressing of the rolling elements. The amplitude of the forces is also higher at the beginning of the re-stressing phase, and it
reduces gradually as the rolling elements fully stress between the raceways.
In summary, it can be concluded that defect-related force impulses in defective rolling element bearings are produced
during the re-stressing phase, which occurs in the vicinity of the end of the defect, where the rolling elements alternatively
impact the outer and inner raceways.
As the contact forces between the rolling elements and the raceways are not measured in practice, they can be correlated
with the acceleration results from simulations to correlate the timings of the re-stressing events with the timings of
impulses observed in the acceleration results. The correlation is discussed in the following section.
6.2. Cause of impulsive signals in acceleration results
Fig. 12 comprises four subplots; subplot (a) shows the numerically modelled filtered nodal acceleration results (shown
in Fig. 5), and subplots (b)–(d) show the numerically modelled vertical outer raceway-to-rolling elements contact forces
(shown in Fig. 6). Out of the four events discussed in Section 5, the time instances corresponding to the occurrences of
event #2 and event #4 were correlated between the acceleration and contact force plots. Event #2 and event #4
correspond to the rolling elements impacting the defective surface of the outer raceway and the re-stressing of the rolling
elements, respectively. These events are highlighted using elliptical and rectangular markers, respectively, in Fig. 12. For
clarification, the acceleration results were zoomed in the vicinity of the second and third impulses, and are shown in
Fig. 13(b) and (c), respectively; Fig. 13(a) is the same as shown in Fig. 12(a). It can be seen in Figs. 12 and 13 that the
dominant impulsive signals, visible in the acceleration plots, correspond to the re-stressing of the rolling elements. The
numerical acceleration results also highlights the occurrence of multiple impulses rather than a single impulse with a
uniform decay.
During the discussion of the contact force analysis, it was also observed that the amplitude of the contact forces related
to the impacts of the rolling elements with the defective surface of the outer raceway (event #2) is lower than those related
to the re-stressing of the rolling elements between the raceways (event #4). This is also reflected in the acceleration results.
The amplitude of the acceleration during the re-stressing of the rolling elements is approximately twice the amplitude
of the acceleration during the period where the rolling elements impact the top surface of the defect.
event # 4
200
Acceleration [g]
event # 4 event # 2 event # 4 event # 2

100
0
−100
−200
0 5 10 15 20 25 30
Time [ms]
10
event # 4: re−stressing of a
roller produces impulsive
acceleration signal
5
event # 4
0
0 5 10 15 20 25 30
Time [ms]
10
5
event # 4
0
0 5 10 15 20 25 30
Time [ms] event # 4
10
5
event # 2: a roller’s impact with the defect surface produces
lower impulsive signals compared to the re−stressing phase
0
0 5 10 15 20 25 30
Time [ms]
Fig. 12. Correlation between the numerically modelled acceleration results shown in Fig. 5 and vertical contact forces, Fy, between the outer raceway
and the first three rolling elements that traversed through the defect shown in Fig. 6; (a) nodal acceleration, (b) contact force: outer raceway-to-rolling
element 1, (c) contact force: outer raceway-to-rolling element 2, and (d) contact force: outer raceway-to-rolling element 3.
event # 4
200
event # 2 event # 4 event # 2
Acceleration [g]
100
100
200
0 5 10 15 20 25 30
event # 2: a roller’s impact with event # 4: re stressing of a
Time [ms]
the defect surface produces roller produces high
low acceleration amplitude acceleration amplitude
200
Acceleration [g]
event # 2 event # 4
100
100
200
15 15.5 16 16.5 17 17.5 18
Time [ms]
200
event # 2
Acceleration [g]
event # 4
100
100
200
26 26.5 27 27.5 28 28.5 29
Time [ms]
Fig. 13. Numerically modelled acceleration results; (a) full time-history showing the three impulses that occurred during the numerical simulation,
(b) partial time-history zoomed in the vicinity of the second impulse, and (c) partial time-history zoomed in the vicinity of the third impulse.
From the detailed discussion of the contact forces in conjunction with the acceleration results, it can be concluded
that although a rolling element can impact the surface of the defect and generate a low amplitude acceleration signal,
a much higher acceleration signal is generated when the rolling elements are re-stressed between the raceways. These
higher acceleration signals are the ones which are generally observed in practice.
7. Experimental verification
7.1. Description of the bearing defect and test rig setup
In order to verify the numerical acceleration results, experiments were conducted on a defective bearing. The bearing
considered for the testing has the number of rolling elements Nb ¼ 24, nominal pitch diameter Dp ¼ 180:2 mm, mean roller
diameter Dr ¼ 17:9 mm, and contact angle α ¼ 9:11. A line spall of dimensions, circumferential length 10 mm and depth
0.2 mm, was manufactured on the outer raceway of the bearing using an electric spark erosion. The dimensions of the
bearing as well as the defect are similar to those of the finite element bearing model. A photo of the machined line spall is
shown in Fig. 14.
The bearing test rig available at the premises of Trackside Intelligence Pty. Ltd. (Track IQ), the Industrial Partner of this
project, was used to conduct the testing of the defective bearing. Fig. 15 shows a photo of the test rig. It comprises a 15 kW
three-phase induction motor, a steel structure to support the wheelset axle, a hydraulic piston-cylinder arrangement for
applying radial load, and a variable frequency drive unit (not shown in the figure) for controlling the speed of the motor. The
defective bearing, referred to as the test bearing in Fig. 15, and a non-defective bearing on the drive-end are press-fitted onto
the axle. The motor is coupled to the axle using V-belts and pulleys. The piston-cylinder arrangement applies radial loads of
up to 100 kN to the wheel seat area on the axle via two metallic rollers approximately spaced 1201 apart. The test and drive-
end bearings are seated on standard adapters which are supported on wooden blocks. The movement of the axle and
bearings in the horizontal direction is constrained by sandwiching rubber-lined wooden blocks between the steel structure
and the bearings. The test rig setup simulates the operation of the bearings in a typical railway application.
An accelerometer was stud mounted on a base that was glued to the test bearing using a super glue. It was mounted on
the top of the bearing, diametrically opposite to the location of the outer raceway defect. A tachometer was mounted near
the axle in order to measure its rotational speed. The experimental vibration data were acquired using the data acquisition
system that comprises a National Instruments (NI) CompactDAQ along with a 3-channel NI 9232 I/O module for the
accelerometer and a NI 9234 I/O module for the tachometer signal. The signals were sampled at the rate of 102.4 kHz.
Fig. 14. A photo of the manufactured line spall of circumferential length 10 mm and depth 0.2 mm machined on the outer raceway of the bearing using an
electric spark erosion.
Motor
Axle
Piston-cylinder arrangement
for radially loading the bearing
Test bearing
Fig. 15. A photo of the experimental test rig.
7.2. Comparison of experimental and numerical acceleration results
Fig. 16 shows the measured acceleration time-trace of the defective bearing for the axle rotational speed of 500 rpm
under the radial load of 50 kN; the speed and load are similar to those applied to the finite element bearing model. The
elliptical and rectangular markers indicate the de-stressing and re-stressing of the rolling elements as they enter into and
exit from the line spall, respectively. While the peak instantaneous level of the de-stressing event is about 720g, it is
approximately 7150g for the impulses due to the re-stressing of the rolling elements. The time separation between the
consecutive events, either de-stressing or re-stressing, corresponds to the BPFO, f bpo , of 90.91 Hz. It agrees favourably with
the numerically estimated f bpo of 90.91 Hz from the numerical acceleration results in Fig. 2.
In contrast to the de-stressing event, it can be seen from the experimental acceleration results that the re-stressing of the
rolling elements is a high-frequency impulsive event that causes the generation of multiple short-duration impulses. The
low- and high-frequency characteristics of the de-stressing and re-stressing events, respectively, were also observed in the
numerically estimated raceways-to-rolling elements contact forces; these events were discussed in Sections 5.2 and 5.4.
From the numerical acceleration results in Figs. 2 and 5, the occurrences of the high-frequency defect-related impulses
due to the re-stressing of the rolling elements are evident; however, the occurrences of the de-stressing events are
not visible despite filtering. This is because even after the filtering, the numerical acceleration results are still noisy.
As mentioned in Section 3.1.1, the high-frequency noise is stochastic, and therefore, could not be estimated.
Although the modelled and measured instantaneous peak acceleration levels of the defect-related impulses are
comparable, approximately 150g, the levels between the impulses could not be compared due to the presence of numerical
noise in the modelled acceleration results. Based on the low-frequency characteristics of the de-stressing event, the
numerical acceleration signal was low-pass filtered using a third-order Butterworth filter with a cut-off frequency of
200
150
de stressing re stressing
event event
100
Acceleration [g]
50
50
100
150
200
0 5 10 15 20 25 30
Time [ms]
Fig. 16. Experimentally measured acceleration of the defective bearing for the axle rotational speed of 500 rpm under the radial load of 50 kN. The low-
frequency de-stressing and high-frequency re-stressing events are highlighted using elliptical and rectangular markers, respectively.
50
40
de stressing re stressing
event event
30
20
Acceleration [g]
10
10
20
30
40
50
0 5 10 15 20 25 30
Time [ms]
Fig. 17. A low-pass filtered numerically modelled acceleration results in Fig. 5 highlighting the low-frequency de-stressing and high-frequency re-stressing
events using elliptical and rectangular markers, respectively.
2500 Hz, and the corresponding results are plotted in Fig. 17. On the one hand, the occurrences of the de-stressing event,
indicated using the elliptical markers, which could not be previously seen in the acceleration signal, can be clearly
distinguished with the elimination of the high-frequency content. On the other hand, the application of the low-pass filter
resulted in the loss of high-frequency defect-related impulsive characteristics, but the impulsive signals related to the re-
stressing event are still visible; the amplitude of the impulses is significantly reduced from approximately 180g to 20g.
The absolute amplitudes of the (unfiltered) experimental and low-pass filtered numerical acceleration signals cannot be
compared; however, the low- and high-frequency characteristics of the de-stressing and re-stressing events, respectively,
predicted by the numerical finite element model compare favourably well with the experimental results. This provides
further confidence to the numerical modelling results presented in the paper.
The modelled and measured acceleration signals were band pass filtered from 16 to 20 kHz, and the well-known
envelope analysis technique [41] was implemented. The corresponding envelope spectra are shown in Fig. 18. The peaks
correspond to the fundamental and harmonics of the outer raceway defect frequency f bpo as highlighted in the figure.
Although the frequencies compare favourably, the magnitude of the modelled results are higher than the experimental
results. This is due to the difference in the characteristics of the defect-related impulses: the numerically modelled impulses
are sharp and narrow compared to the experimentally measured impulses. The favourable comparison of the numerical and
experimental envelope spectra in Fig. 18 indicates that the numerical noise did not affect the bearing diagnosis, which is
generally conducted using the envelope analysis by demodulating the vibration signals in high-frequency regions.
70
fbpo experimental
numerical
60 2 x fbpo
3 x fbpo
Acceleration [(m/s2)2/Hz]
50
4 x fbpo
40
5 x fbpo
30
20
10
0
0 200 400 600 800 1000
Frequency [Hz]
Fig. 18. Envelope power spectrum of the numerically modelled and experimentally measured acceleration results.
From the comparisons presented in this paper, it can be concluded that the developed finite element model of a defective
rolling element bearing has accurately modelled the complicated dynamics of the system.
7.3. Spall size estimation
The time separation between the de-stressing and re-stressing events for the simple case of a line spall can be used to
determine the average size (length) of the defect as
ωr Do Δt event
lspall ¼ (5)
2
where lspall is the average length of the defect, ωr is the angular velocity of the rolling elements, Do is the outer race diameter,
and Δt event is the time difference between the consecutive de-stressing and re-stressing events.
Depending on the signal-to-noise ratio of measured vibration signals, which can be quite low in noisy field environments, the
above-described signal processing technique may be implemented in order to distinguish between the de-stressing and the
re-stressing of the rolling elements. Otherwise, relatively sophisticated techniques and algorithms may be employed on practical
data for estimating the spall size of a bearing defect [26,53].
8. Conclusion
This paper presented results from the explicit finite element analysis of a rolling element bearing with an outer raceway
defect. A detailed interpretation of the numerically modelled contact forces between the rolling elements and the raceways
of a bearing, which are not measured in practice, was presented. The entry and exit of the rolling elements into and from the
defect, respectively, were specifically highlighted by discussing the characteristics of the rolling elements-to-raceways
contact forces.
It was found that the re-stressing of the rolling elements (on their exit from the defect) between the raceways produces a
burst of short-duration (high-frequency) impulses which are observed in the acceleration signals of a defective rolling
element bearing. The generation of multiple force impulses during the re-stressing phase provides pragmatic insights that
explain the occurrence of multiple impulses in measured vibration signals. The correlation of the contact forces with
acceleration results showed that the impulsive forces and corresponding vibrations generated during the re-stressing of the
rolling elements are much greater than when the rolling elements strike the surface of the defect.
A rolling element bearing with a machined outer raceway defect was experimentally tested. The experimental and
numerical acceleration results were compared for model verification purposes. While the de-stressing and re-stressing
of the rolling elements were clearly observed in the experimental results, the re-stressing was also observed in the
numerical acceleration results but not the de-stressing event as it was masked by the high-frequency noise. It was shown
that the de-stressing of the rolling elements was enhanced by filtering out the high-frequency noise signals. The low- and
high-frequency characteristics of the de-stressing and re-stressing of the rolling elements, respectively, obtained from the
numerical and experimental results agreed favourably with each other.
The phenomenon that caused the generation of numerical contact noise in the simulation results was demonstrated
using the beating effect. A favourable match between the analytical and numerical estimates of the rolling elements-to-
raceways noise frequencies justified the proposed hypothesis for explaining the presence of numerical contact noise.
The numerical modelling results also showed that as a rolling element unloads in the defective region, load on other rolling
elements, located at the non-defective sections of the raceways, increases in order to maintain the equilibrium of the
system. The analytical solution of force distribution on the rolling elements of a defective bearing agreed favourably with the
numerical solution. The results indicate that while high magnitudes of the impulsive forces generated near the end of a
defect would lead to its lengthening in the rolling direction, the re-distribution of the forces would eventually lead to
surface crack initiation on the raceways at approximate roller spacing.
The finite element model presented in the paper can be used to investigate the vibration characteristics of bearings with
more complex defect geometries, which may, subsequently, lead to better understanding of force and vibration generation
mechanisms and to designing better signal processing techniques to distinguish between defects of various sizes.
Acknowledgements
This work was conducted as a part of ARC Linkage funded project LP110100529 with Industry Partner, Track IQ, who
provided extensive project facilities and funding. The principal author of the paper would like to thank Dr. David Rennison of
Trackside Intelligence Pty. Ltd., Dr Michael Kidner of Vipac Engineers and Scientists Ltd., and Emeritus Professor Colin
Hansen of the University of Adelaide for many useful discussions during the course of the research.
References
[1] P.D. McFadden, J.D. Smith, Model for the vibration produced by a single point defect in a rolling element bearing, Journal of Sound and Vibration 96 (1)
(1984) 69–82.
[2] P.D. McFadden, J.D. Smith, The vibration produced by multiple point defects in a rolling element bearing, Journal of Sound and Vibration 98 (2) (1985)
263–273.
[3] Y.T. Su, S.J. Lin, On initial fault detection of a tapered roller bearing: frequency domain analysis, Journal of Sound and Vibration 155 (1) (1992) 75–84.
[4] N. Tandon, A. Choudhury, An analytical model for the prediction of the vibration response of rolling element bearings due to a localized defect, Journal
of Sound and Vibration 205 (3) (1997) 275–292.
[5] D. Brie, Modelling of the spalled rolling element bearing vibration signal: an overview and some new results, Mechanical Systems and Signal Processing
14 (3) (2000) 353–369.
[6] D. Ho, R.B. Randall, Optimisation of bearing diagnostics techniques using simulated and actual bearing fault signals, Mechanical Systems and Signal
Processing 14 (5) (2000) 763–788.
[7] R.B. Randall, J. Antoni, S. Chobsaard, The relationship between spectral correlation and envelope analysis in the diagnostics of bearing faults and other
cyclostationary machine signals, Mechanical Systems and Signal Processing 15 (5) (2001) 945–962.
[8] J. Antoni, R.B. Randall, Differential diagnosis of gear and bearing faults, Journal of Vibration and Acoustics 124 (2) (2002) 165–171.
[9] J. Antoni, R.B. Randall, A stochastic model for simulation and diagnostics of rolling element bearings with localized faults, Journal of Vibration and
Acoustics 125 (3) (2003) 282–289.
[10] N.S. Feng, E.J. Hahn, R.B. Randall, Using transient analysis software to simulate vibration signals due to rolling element bearing defects, Proceedings of
the Third Australian Congress on Applied Mechanics, Sydney, 2002, pp. 689–694.
[11] J. Sopanen, A. Mikkola, Dynamic model of a deep-groove ball bearing including localized and distributed defects. Part 1: theory, Proceedings of the
Institution of Mechanical Engineers, Part K: Journal of Multi-body Dynamics 217 (3) (2003) 201–211.
[12] J. Sopanen, A. Mikkola, Dynamic model of a deep-groove ball bearing including localized and distributed defects. Part 2: implementation and results,
Proceedings of the Institution of Mechanical Engineers, Part K: Journal of Multi-body Dynamics 217 (3) (2003) 213–233.
[13] Z. Kiral, H. Karagülle, Simulation and analysis of vibration signals generated by rolling element bearing with defects, Tribology International 36 (9)
(2003) 667–678.
[14] Z. Kiral, H. Karagülle, Vibration analysis of rolling element bearings with various defects under the action of an unbalanced force, Mechanical Systems
and Signal Processing 20 (8) (2006) 1967–1991.
[15] A. Choudhury, N. Tandon, Vibration response of rolling element bearings in a rotor bearing system to a local defect under radial load, Journal of
Tribology 128 (2) (2006) 252–261.
[16] S. Sassi, B. Badri, M. Thomas, A numerical model to predict damaged bearing vibrations, Journal of Vibration and Control 13 (11) (2007) 1603–1628.
[17] N. Sawalhi, R.B. Randall, Simulating gear and bearing interactions in the presence of faults. Part I: the combined gear bearing dynamic model and the
simulation of localised bearing faults, Mechanical Systems and Signal Processing 22 (8) (2008) 1924–1951.
[18] M. Cao, J. Xiao, A comprehensive dynamic model of double-row spherical roller bearing-model development and case studies on surface defects,
preloads, and radial clearance, Mechanical Systems and Signal Processing 22 (2) (2008) 467–489.
[19] A. Ashtekar, F. Sadeghi, L.-E. Stacke, A new approach to modeling surface defects in bearing dynamics simulations, Journal of Tribology 130 (4) (2008)
041103.
[20] H. Arslan, N. Aktürk, An investigation of rolling element vibrations caused by local defects, Journal of Tribology 130 (4) (2008) 041101.
[21] A. Rafsanjani, S. Abbasion, A. Farshidianfar, H. Moeenfard, Nonlinear dynamic modeling of surface defects in rolling element bearing systems, Journal
of Sound and Vibration 319 (3–5) (2009) 1150–1174.
[22] V.N. Patel, N. Tandon, R.K. Pandey, A dynamic model for vibration studies of deep groove ball bearings considering single and multiple defects in races,
Journal of Tribology 132 (4) (2010) 041101.
[23] M.S. Patil, J. Mathew, P.K. Rajendrakumar, S. Desai, A theoretical model to predict the effect of localized defect on vibrations associated with ball
bearing, International Journal of Mechanical Sciences 52 (9) (2010) 1193–1201.
[24] M. Nakhaeinejad, M.D. Bryant, Dynamic modeling of rolling element bearings with surface contact defects using bond graphs, Journal of Tribology
133 (1) (2011) 011102.
[25] M. Tadina, M. Boltežar, Improved model of a ball bearing for the simulation of vibration signals due to faults during run-up, Journal of Sound and
Vibration 330 (17) (2011) 4287–4301.
[26] S. Zhao, L. Liang, G. Xu, J. Wang, W. Zhang, Quantitative diagnosis of a spall-like fault of a rolling element bearing by empirical mode decomposition
and the approximate entropy method, Mechanical Systems and Signal Processing 40 (1) (2013) 154–177.
[27] T.E. Tallian, O.G. Gustafsson, Progress in rolling bearing vibration research and control, Transactions of the American Society of Lubrication Engineers
8 (3) (1965) 195–207.
[28] C.S. Sunnersjö , Rolling bearing vibrations—the effects of geometrical imperfections and wear, Journal of Sound and Vibration 98 (4) (1985) 455–474.
[29] F.P. Wardle, Vibration forces produced by waviness of the rolling surfaces of thrust loaded ball bearings, Part 1: Theory, Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 202 (5) (1988) 305–312.
[30] E. Yhland, A linear theory of vibrations caused by ball bearings with form errors operating at moderate speed, Journal of Tribology 114 (2) (1992)
348–359.
[31] Y.-T. Su, M.-H. Lin, M.-S. Lee, The effects of surface irregularities on roller bearing vibrations, Journal of Sound and Vibration 165 (3) (1993) 455–466.
[32] S.P. Harsha, K. Sandeep, R. Prakash, Non-linear dynamic behaviors of rolling element bearings due to surface waviness, Journal of Sound and Vibration
272 (3–5) (2004) 557–580.
[33] G. Jang, S.W. Jeong, Vibration analysis of a rotating system due to the effect of ball bearing waviness, Journal of Sound and Vibration 269 (3–5) (2004)
709–726.
[34] B. Changqing, X. Qingyu, Dynamic model of ball bearings with internal clearance and waviness, Journal of Sound and Vibration 294 (1–2) (2006) 23–48.
[35] C.K. Babu, N. Tandon, R.K. Pandey, Vibration modeling of a rigid rotor supported on the lubricated angular contact ball bearings considering six
degrees of freedom and waviness on balls and races, Journal of Vibration and Acoustics 134 (1) (2012) 011006.
[36] W.E. Littmann, The mechanism of contact fatigue, in: P.M. Ku (Ed.), Interdisciplinary Approach to the Lubrication of Concentrated Contacts, no. NASA SP-237,
Washington, DC, National Aeronautics and Space Administration, 1969, pp. 309–378.
[37] I. Howard, A review of rolling element bearing vibration: “detection, diagnosis and prognosis”, Tech. Rep. DSTO-RR-0013, Defence Science and
Technology Organisation, Aeronautical and Maritime Research Laboratory, Australia, 1994.
[38] N. Tandon, A. Choudhury, A review of vibration and acoustic measurement methods for the detection of defects in rolling element bearings, Tribology
International 32 (8) (1999) 469–480.
[39] D.R. Houser, M.J. Drosjack, Vibration signal analysis techniques, Tech. Rep., U.S. Army Air Mobility Research and Development Laboratory, 1973.
[40] D. Mba, B.K.N. Rao, Development of acoustic emission technology for condition monitoring and diagnosis of rotating machines; bearings, pumps,
gearboxes, engines and rotating structures, Shock and Vibration Digest 38 (1) (2006) 3–16.
[41] R.B. Randall, J. Antoni, Rolling element bearing diagnostics—a tutorial, Mechanical Systems and Signal Processing 25 (2) (2011) 485–520.
[42] P.D. McFadden, J.D. Smith, The vibration monitoring of rolling element bearings by the high-frequency resonance technique—a review, Tribology
International 17 (1) (1984) 3–10.
[43] R. Stribeck, Ball bearings for various loads, Transactions of the American Society of Mechanical Engineers 29 (1907) 420–463.
[44] H. Hertz, Über die Berührung fester elastischer Kö rper (on the contact of elastic solids), Journal of Reine and Angewandte Mathematik 92 (1882) 156–171
(reprinted in English, in: Jones and Schott (Eds.) Miscellaneous Papers by H. Hertz, Macmillan, London, 1896 [90, 243]).
[45] H. Hertz, Über die Berührung fester elastischer Kö rper and über die Harte(on the contact of rigid elastic solids and on hardness), Verhandlungen des
Vereins zur Befö rderung des Gewerbefleisses (reprinted in English, in: Jones and Schott (Eds.), Miscellaneous Papers by H. Hertz, Macmillan, London,
1896 [90, 156]).
[46] Y. Shao, W.Tu, F. Gu, A simulation study of defects in a rolling element bearing using FEA, International Conference on Control, Automation and Systems,
Gyeonggi-do, Korea, 2010, pp. 596–599.
[47] L. Guochao, P. Wei, L. Yongcai, G. Lixin, J. Zhang, Simulation and dynamic analysis of outer ring fault on rolling bearings using explicit finite element
method, China Mechanical Engineering 23 (23) (2012) 2825–2829.
[48] J. Liu, Y. Shao, M.J. Zuo, The effects of the shape of localized defect in ball bearings on the vibration waveform, Proceedings of the Institution of
Mechanical Engineers, Part K: Journal of Multi-body Dynamics 227 (3) (2013) 261–274.
[49] A. Utpat, Vibration signature analysis of defective deep groove ball bearings by numerical and experimental approach, International Journal of Scientific
and Engineering Research 4 (6) (2013) 592–598.
[50] N.K. Arakere, N.A. Branch, G. Levesque, V. Svendsen, N.H. Forster, Rolling contact fatigue life and spall propagation of AISI M50, M50NiL, and AISI
52100, Part II: stress modeling, Tribology Transactions 53 (1) (2009) 42–51.
[51] N.A. Branch, N.K. Akakere, V. Svendsen, N.H. Forster, Stress field evolution in a ball bearing raceway fatigue spall, Journal of ASTM International 7 (2)
(2010) 1–18.
[52] J.O. Hallquist, LS-DYNA Theory Manual, Livermore Software Technology Corporation, March 2006.
[53] N. Sawalhi, R.B. Randall, Vibration response of spalled rolling element bearings: observations, simulations and signal processing techniques to track the
spall size, Mechanical Systems and Signal Processing 25 (3) (2011) 846–870.
[54] ANSYS, Release 14.5, SAS IP, Inc.
[55] T.A. Harris, Rolling Bearing Analysis, fourth ed. John Wiley & Sons, 2001.
[56] ANSYS, Release 14.5, Help System, Structural Analysis Guide, SAS IP, Inc.
[57] D. Bies, C. Hansen, Engineering Noise Control: Theory and Practice, fourth ed. Taylor & Francis, 2009.
[58] S. Iwnicki, Handbook of Rail Vehicle Dynamics, CRC Press, Taylor & Francis Group, 2006.
[59] Livermore Software Technology Corporation, Livermore, California 94551-0712, LS-DYNA Keyword User's Manual, Vol. I, February 2012.
[60] D. Petersen, C. Howard, N. Sawalhi, A.M. Ahmadi, S. Singh, Analysis of bearing stiffness variations, contact forces and vibrations in radially loaded
double row rolling element bearings with raceway defects, Mechanical Systems and Signal Processing 50-51 (2015) 139–160, http://dx.doi.org/10.1016/j.
ymssp.2014.04.014.
[61] M.R. Hoeprich, Rolling element bearing fatigue damage progression, Journal of Tribology 114 (2) (1992) 328–333.

NDM 7 PDF

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

NDM 7 PDF

Încărcat de

Drepturi de autor:

Formate disponibile

Journal of Sound and Vibration 333 (2014) 5356–5377

Contents lists available at ScienceDirect

Journal of Sound and Vibration

Analyses of contact forces and vibration response

2.1. Description of the finite element model

Component name Dimensions (mm)

Outer ring Outer diameter ¼220 Thickness¼ 10

2.2. Discretisation of the finite element model

2.3. Loads and boundary conditions

2.4. Analysis and control settings

2.5. Results and observations

2.5.1. Acceleration time-history

numerical noise frequency 4671 Hz

Acceleration [dB re 1 (m/s2 ) 2 /Hz]

3.1. A hypothesis for explaining the cause of numerical contact noise

3.1.1. Beating phenomenon

PðtÞ ¼ A1 cos ð2πf noiseIR tÞ þ A2 cos ð2πf noiseIR tÞ (4)

4. A discussion on the validation of numerical contact forces

Contact force Fx [kN]

Contact force Fx [kN]

Roller position relative to y-axis (deg) Contact width (mm)

5. Contact force analysis

Starting position Ending position

Contact force F [kN]

Contact force F [kN]

Contact force Fx [kN]

Contact force F [kN]

6. Correlating contact forces with bearing vibrations

6.1. Physical mechanism that generates defect-related impulsive forces

Contact force Fy [kN]

Contact force Fy [kN]

Contact force Fy [kN]

Contact force Fy [kN]

6.2. Cause of impulsive signals in acceleration results

event # 4 event # 2 event # 4 event # 2

7.1. Description of the bearing defect and test rig setup

Fig. 15. A photo of the experimental test rig.

7.2. Comparison of experimental and numerical acceleration results

7.3. Spall size estimation

S-ar putea să vă placă și