Sei sulla pagina 1di 14

P. K.

Kankar
e-mail: pavankankar@gmail.com

Fault Diagnosis of High Speed Rolling Element Bearings Due to Localized Defects Using Response Surface Method
In this paper, fault diagnosis of high speed rolling element bearings due to localized defects using response surface method has been done. The localized defects as spalls on outer race, on inner race, and on rolling elements are considered for this study. The mathematical formulation accounted for tangential motions of rolling elements and inner and outer races with the sources of nonlinearity such as Hertzian contact force and internal radial clearance. The nonlinear stiffness is obtained by the application of Hertzian elastic contact deformation theory. The mathematical formulation predicts discrete spectrum having peaks at the characteristic defect frequencies and their harmonics. Experimentation has also been performed to validate the results obtained from the mathematical model and it shows that the model can be successfully used to predict amplitude ratios among various spectral lines with localized surface defects. Combined parametric effects have been analyzed and their inuence has been considered with design of experiments and surface response methodology is used to predict the dynamic response of a rotor bearing system. DOI: 10.1115/1.4003371 Keywords: fault diagnosis, rolling element bearings, radial internal clearance, nonlinear stiffness, design of experiments, response surface method

Satish C. Sharma
e-mail: sshmefme@iitr.ernet.in

S. P. Harsha
e-mail: surajfme@iitr.ernet.in Department of Mechanical and Industrial Engineering, Indian Institute of Technology Roorkee, Roorkee 247 667, India

Introduction

Rotary machines are recognized as crucial equipment in various industries, such as power stations, chemical plants, and automotive industry, that require precise and efcient performance. Rolling element bearings are used in a wide variety of rotating machinery from small hand-held devices to heavy duty industrial systems and are the primary cause of breakdowns in machines. Such breakdowns can lead to expensive shutdowns, drifts in production, and even human casualties. Vibration measurements are widely used for detection of defects in bearings. Rolling bearing defects may be categorized as point or local defects and distributed defects. The vibrations are generated by geometrical imperfections on the individual bearing components and these imperfections are caused by irregularities during the manufacturing process as well as wear and tear. The vibration signals contain information of defective parts and a variety of vibration based techniques have been developed to monitor the condition of bearing. Vibration signals are analyzed using time domain analysis, frequency domain analysis, and time-frequency domain analysis wavelet. Faults are classied using various articial intelligence techniques to predict the meaningful results from the observations. Articial neural networks ANNs 16, support vector machine SVM 711, fuzzy logic classiers 12,13, and other soft computing techniques are widely used tools to classify the faults for further processing. In these works, after a vibration signal is measured, different signal processing techniques are employed to extract the fault sensitive features to serve as the monitoring indices. Then, these techniques are used to diagnose various bearing faults from the vibration signature obtained. Gallina et al. 14 presented an application of coupling between response surface methodology RSM and Monte Carlo simulaContributed by the Dynamic Systems Division of ASME for publication in the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript received September 30, 2009; nal manuscript received October 21, 2010; published online March 24, 2011. Assoc. Editor: Luis Alvarez.

tion in the eld of dynamic analysis of mechanical structures. The error has been introduced by using an approximated metamodel instead of the real model for the Monte Carlo simulation has been analyzed and a solution has been proposed to overcome the problem. Recently, Liang et al. 15 utilized RSM to analyze the effect of design parameters on the sound radiation from a vibrating panel. They showed that the information of RSM can give the direction of design modication such as the weight of material and the thickness of thin panel. Kankar et al. 16 applied RSM to investigate the effects of various defects on the nonlinear vibrations of the rotor bearing system. Ghafari et al. 17 investigated the effect of localized faults on the chaotic vibration of rolling element bearings. The vibration based condition monitoring is mainly emphasized on signal processing techniques and diagnosis of defects based on vibration signature. The present study aims to identify the effect of various bearing component faults on the stability and dynamic behavior of the rotor bearing system. Defects are considered as spall on outer race, inner race, and ball. In this paper, quadratic polynomials to create the response surface have been used. Experiments are carried out for healthy bearings and bearings with faults. Design of experiments DOEs and RSM are used to predict the dynamic response of a rotor bearing system.

The Problem Formulation

A schematic diagram of rolling element bearing is shown in Fig. 1. The vertical radial load W is applied due to spindle weight and as the rotor rotates, an unbalance force Fu comes into being. In the mathematical model, the rolling element bearings are considered as a mass-spring system and the balls act as a nonlinear contact spring. The detailed description about mathematical formulation of the problem is taken from Harsha and Kankars previous paper 18. Hence, according to that, the same assumptions are also taken from this paper. The local Hertzian contact force and deection relationship for bearing may be written as MAY 2011, Vol. 133 / 031007-1

Journal of Dynamic Systems, Measurement, and Control Copyright 2011 by ASME


Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

= 2.787 108Q2/3
Hence, the contact force Q is Q = 3.587 107 is Ki = 3.587 107


1/2

1/3

3/2 3/2

The elastic modulus for the contact of a ball with the inner race

1/2

i3/2


N mm3/2

The contact of a ball with the outer race is


Fig. 1 Schematic diagram of a rolling element bearing

Ko = 3.587 107

1/2

3/2 o


N mm3/2

F j = k r i 3/2

Then, the effective elastic modulus K for the bearing system is written as K= 1
/n n 1/Ki1/n + 1/K1 o

Considering the internal radial clearance, the radial displacement at the ith ball is given as r j = x cos i + y sin i Substituting it into Eq. 1, we get
3/2 F j = kx cos i + y sin i +

10

2 3

The effective elastic modulus K for bearing system using geometrical and physical parameters is written as K = 7.481 105


N mm3/2

11

If the expression inside the bracket is greater than zero, then the ball at the angular location i is loaded, giving rise to a restoring force Fi. Thus, the total restoring force components in the x and y directions are Fx =

kx cos + y sin
i i i=1 Nb i i i=1

Nb

3/2 + cos

4a

Fy =

kx cos + y sin

3/2 + sin

4b

2.3 Surface Defects. Spalls, pits, and dents are the major forms of bearing damage. Fatigue cracking of the surface occurs when the repeatedly cycled stress on a surface in rolling contact with another exceeds the endurance strength of the material. This defect propagates and results in a large pit or spall on the surface of bearing components. Corrosion and oxidation pits, true/false brinnelling, and hard particle contamination dents act as locations for incipient fatigue. This can cause bearing endurance to be shorter than that designed and may also lead to rapid failure of the bearing 19. 2.3.1 Outer Race Defect. In this paper, a defect has been considered on the outer race, which is located at an angle from the x-axis. Whenever, a ball passes over the defect location, it has an additional deection s. The outer race is assumed to be stationary. The contact deformation of the ith ball when it coincides with defect angle becomes ri = x cos i + y sin i + s Substituting ri into Eq. 1, we obtain
3/2 F j = kx cos i + y sin i + s+

2.1 Equations of Motion. The system governing equations accounting for inertia, restoring and damping forces, and constant vertical force acting on the inner race are + cx + mx

kx cos + y sin y
i i i=1 3/2 + sin

Nb

3/2 + cos

i
5a

12 13

= W + Fu cost + cx + my

kx cos + y sin y
i i i=1

Nb

i = Fu sint
5b

The equation of motion, Eq. 5, consists of two nonlinear ordinary second order differential equations with a parametric excitation effect, the 1.5 nonlinearity and the summation term. The + sign as subscript in these equations signies that if the expression inside the bracket is greater than zero, then the rolling element at angular location i is loaded, giving rise to a restoring force and if the expression inside the bracket is negative or zero, then the rolling element is not in the load zone, and the restoring force is set to zero. 2.2 Contact Stiffness. The application of the classical theory of elasticity to the problem forms the basis of stress calculation for machine elements as ball and roller bearings. Hertz considered the stress and deformation in such perfectly smooth, ellipsoidal, contacting elastic solids. As per the sign convention followed, negative radius denotes a concave surface. For the contacting bodies made of steel, the relative approach between two contacting and deforming surfaces is given by 031007-2 / Vol. 133, MAY 2011

Thus, the total restoring force components in the x and y directions given by Eq. 4 and the systems governing Eq. 5 can be modied accordingly. 2.3.2 Inner Race Defect. The inner race is rotating at the shaft speed and the ball center is rotating at the speed of cage, so the contact angle i is

i =

2 i 1 + cage t Nb

14

Defect is on the surface of the inner race, so it rotates at speed in. Additional deection s in contact deformation is obtained when defect angle coincides with the contact angle i of ith rolling element. 2.3.3 Rolling Element Defect. Ball with spall rotates about an axis normal to the plane containing the centers of the inner race and outer race contacts. It is assumed that ball spins about its axis, so it comes in contact with inner race and outer race at regular intervals. Additional deection s in contact deformation can be obtained when sinballt = 0 20. Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 2 Experimental setup with defective and nondefective bearings

Experimental Setup and Data Acquisition

Fig. 4 Inner race with spall

The problem of predicting the degradation of working conditions of bearings and trending of fault propagation before they reach the alarm or failure threshold is extremely important in industries to fully utilize the machine production capacity and to reduce the plant downtime. In the present study, an experimental test rig, as shown in Fig. 2, is used to analyze the effect of various bearing component defects on the stability of the rotor bearing system. Table 1 shows dimensions of the ball bearings taken for this study. The rig is connected to a data acquisition system through proper instrumentation. Accelerometers and proximity pick-ups are used for picking up the vibration signals from various stations on the rig. As a rst step, the machine was run with healthy bearing to establish the base-line data. Then, data are collected for different fault conditions. A variety of faults are simulated on the rig at 1000 rpm and 5000 rpm. The following faults are introduced in the bearing: a b c outer race with spall Fig. 3 inner race with spall Fig. 4 ball with spall Fig. 5

vibration maximum noise conditions, DOE is used with a total of eight trial runs. Table 2 shows parameters used for DOE with their minimum and maximum levels. When the rolling element sets and the cage rotates with a constant angular velocity, a parametrically excited vibration is generated and transmitted through the outer race. The characteristic frequency of this vibration is called the varying compliance frequency VC and is given as VC = Nb cage 15 The rotational speeds of the inner and outer races, the cage, and the rolling elements are different. As a result, defect in the outer race, inner race, and rolling element generate vibrations of distinct frequencies. If cage is the theoretical cage or fundamental train frequency for rolling element bearings, the frequency of rotation of the rolling elements with respect to inner races is given by Nbin 1 +

Vibration responses for all cases are presented as fast Fourier transform FFT. RSM is used to nd the effect of various localized bearing component faults on vibration responses and interactions between faults. In order to perform response surface analysis to determine the combination of defects that gives the chaotic

wp = in cage =

db cos D 2

16

where wp is known as the inner race defect frequency or wave passage frequency.

Table 1 Parameters of bearing Parameter Outer race diameter Inner race diameter Ball diameter Ball No. Contact angle Radial clearance Spall on outer race and ball Spall on inner race Value 28.262 mm 18.738 mm 4.762 mm 8 0 deg 10 m 2 mm 1 mm Fig. 5 Ball with spall Table 2 Parameters for DOE Parameter designation symbols A B C D Fig. 3 Outer race with spall Minimum level 0 0 0 1000 Maximum level + 1 1 1 5000

Parameters Outer race with spall Inner race with spall Ball with spall Speed of rotor

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-3

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Table 3 Characteristic frequencies Rotor speed Hz Characteristic frequency Cage frequency cage Formula 1000 rpm 5000 rpm 33.22

in 1 cage =

db cos D 2

6.65

Varying compliance frequency VC Inner race defect frequency or wave passage frequency wp

Nb cage Nbin 1 +

53.16 80.18

265.79 400.88

wp =

db cos D 2

Outer race defect frequency or ball passage frequency bp

Nbin 1

bp =

db cos D 2

53.16

265.79

Ball spin frequency bsf

bsf =

i D db 1 cos 2 db D

41.14

205.70

Similarly, with Nb rolling element, the frequency of rotation of the rolling elements with respect to outer races considering outer race stationary is given by Nbin 1 +

bp = cage out =

db cos D 2

The method of least-squares is used to estimate the polynomial coefcients in the approximating polynomials such that it minimizes the sum of squares of the model errors. Then, matrix B of polynomial coefcients can be obtained from the formula B = XTX1XTY 21

17

where bp is known as the outer race defect frequency or ball passage frequency. Assuming no slip, the frequency of rotation of rolling elements about their own axes of rotation is given by

bsf =

i D db 1 cos 2 db D

The response surface analysis is done in terms of the tted surface. Once a response surface model is obtained, statistical analysis techniques, such as analysis of variance ANOVA, can be used to check the tness of the model. The process of the whole RSM is shown in Fig. 6. 4.1 Response Surface Model Establishment. This procedure begins with the identication of various bearing component defects that affect the stability of the rotor bearing system. Each parameter affects the nature of the response in mutual interactions with other parameters as well as with separate factors. The following bearing component faults are considered as factors that are inuencing responses that are taken as peaks of vibration response in horizontal and vertical directions. a. b. c. d. outer race with spall inner race with spall ball with spall speed of rotor

18

For the bearing geometry and at rotor speeds 1000 rpm and 5000 rpm, the signicant frequency components are shown in Table 3. Here, in is the rotational frequency of inner race, db is the ball diameter, D is the pitch diameter, Nb is the number of rolling elements, and is the contact angle.

Response Surface Methodology

RSM is a collection of mathematical and statistical techniques that are useful for modeling and analysis in applications where a response interest is inuenced by several variables and the objective is to optimize this response. RSM has found extensive application in a wide variety of industrial settings such as chemical processes, semiconductor and electronics manufacturing, machining, metal cutting, and joining processes 21. RSM constructs polynomial approximations to functional relationships between design variables and performances. The input variables are sometimes called independent variables or factors, and the performance measures or quality characteristics are called responses. For the most responsive surface, the relationship between the response variable of interest y and the factors x1 , x2 , . . . , xk may be described in the following second order equation: y = f x = o +

These four factors are selected with two levels. The upper and lower levels of each factor are dened in Table 2. The reference value for healthy and defective bearings are taken as 0 and 1, respectively, where value 0 indicates absence of defect and value 1 indicates that defect is present in bearing components. DOE considers factors with two levels and eight trial runs, as listed in Table 4.

Results and Discussions

i=1

ix i +

i=1

iixi2 +

i=1, j1

ijxix j +

19

where represents the noise or error observed in the response y , is the polynomial coefcient, and n is the number of factors. Equation 19 can also be expressed in matrix form as Y = XB + 031007-4 / Vol. 133, MAY 2011 20

The equations of motion, Eq. 5, has been solved by modied Newmark- method to obtain the radial displacement and velocity of the rolling elements. Poincar map is a classical technique for analyzing dynamical systems 22. It replaces the ow of an nth order continuous time system with a n 1th order discrete time system called the Poincar map. The usefulness of Poincar maps lies in the reduction of order and the fact that it bridges the gap between continuous and discrete time systems. A periodic nth orTransactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 6 Flow chart for RSM

der nonautonomous system with minimum period T can be transformed into an n + 1th order autonomous system in the cylindrical state space. The Poincar maps of chaotic solutions have fractal structures that repeat as the map is magnied. Periodic response of period-n will appear as n discrete points on the Poincar map and this shows a single curve as it is magnied and a two-period quasiperiodic motion will show up as a close smooth curve. Hence, when the map is magnied, it will show a complete circle while phase plots are plotted for the entire range without magnifying them. For the trials 18 as shown in Table 4, the response plots have been shown in the form of frequency spectra obtained from math-

ematical model and experimental analysis, Poincar map, and orbit plot, as shown in Figs. 714, respectively. Figure 7 shows the response plots for the minimum level of localized defects of bearing components and rotor speed is 1000 rpm. For this trial, peak amplitude of vibration excitation appears at VC and its multiples in frequency spectra. The multiloop character of Poincar map and orbit plot conrms the multiple peaks of excitation of FFT and shows the periodicity of the system. In presence of spall on outer race with 1000 rpm of rotor speed, the peak amplitude of vibration appears in the spectrum at ball passage frequency bp and its multiples, as shown in Fig. 8. The broad band frequency spectrum that appears around the peak of excitation shows that system losses its periodicity, which can also be conrmed by Poincar map. Figure 9 shows the response with spall on inner race with 1000 rpm of rotor speed. For this trial peak, the peak amplitude of vibration appears in the response at wave passage frequencies wp and 2wp. Poincar map and orbit plot show that a periodic nature of system refers to onset of chaos condition. Next, trial is considered with spall on ball with 1000 rpm of rotor speed and periodic excitation appears at ball spin frequency bsf in frequency spectra, as shown in Fig. 10. The nature of solution under these conditions signies the periodicity of system. For trials 58, the rotor speed has increased by 5000 rpm, i.e., at maximum level, which is considered for the experimentation. When the system is operating with healthy bearings at rotor speed 5000 rpm, the second order subharmonic nature is observed with peak amplitude of vibration and appears at VC and its harmonics in the frequency spectrum, as shown in Fig. 11. The multiloop characteristic of Poincar map and orbit plot conrms the periodicity of the system solution. For trial 6, with spall on outer race, the system shows chaotic nature with dense broad band frequency around peak excitation at ball passage frequency, as shown in Fig. 12. A dominant peak appears in the vibration spectrum at bp. The peak amplitude of vibration appears in the spectrum at wp, wp in, and in when a spall appears on inner race, as shown in Fig. 13. Poincar map and orbit plot clearly show the chaotic nature of response. When a spall appears on rolling element, the system shows the quasi-periodic response, as shown in Fig. 14, by net structure in Poincar map and orbit plot. The frequency spectrum shows the fourth order of subharmonic response mainly due to the amplitude modulation of two excitation frequencies as ball spin frequency and carrier frequency. Dominant peaks in the spectrum appear at ball spin frequency bsf . There appears to be some obvious discrepancies between the numerical and experimental results that are presented in the form of frequency spectra plots. In Fig. 9, for example, the numerical frequency spectrum suggests that the response of the rotor is periodic. The experimental results, on the other hand, shows a broad band frequency content indicating chaos in the response of the rotor. This observation is also true for the results shown in Figs. 7

Table 4 DOE set and results Factor 1 Trial 1 2 3 4 5 6 7 8 A + + Factor 2 B + + Factor 3 C + + Factor 4 D + + + + Vibration response Horizontal amplitude gRMS 0.013418 0.024767 0.015532 0.015004 0.103042 0.157276 0.017408 0.101925 Vertical amplitude gRMS 0.007406 0.011958 0.010312 0.01187 0.159953 0.164743 0.011608 0.114851

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-5

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 7 Response plot for rotor speed 1000 rpm supported on healthy bearings trial 1

Fig. 8 Response plot for rotor speed 1000 rpm supported on bearings with outer race defect trial 2

031007-6 / Vol. 133, MAY 2011

Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 9 Response plot for rotor speed 1000 rpm supported on bearings with inner race defect trial 3

Fig. 10 Response plot for rotor speed 1000 rpm supported on bearings with ball defect trial 4

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-7

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 11 Response plot for rotor speed 5000 rpm supported on healthy bearings trial 5

Fig. 12 Response plot for rotor speed 5000 rpm supported on bearings with outer race defect trial 6

031007-8 / Vol. 133, MAY 2011

Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 13 Response plot for rotor speed 5000 rpm supported on bearings with inner race defect trial 7

Fig. 14 Response plot for rotor speed 5000 rpm supported on bearings with ball defect trial 8

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-9

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Table 5 Comparison of peak excitation of frequency components Excitation frequency at frequency in Hz and amplitude in gRMS Theoretical VC 53.16, 0.015 bp 53.16, 0.035 wp 80.18, 0.021 bsf 41.14, 0.015 VC 265.79, 0.102 bp 265.79, 0.157 wp 400.88, 0.24 bsf 205.7, 0.19 Experimental VC 51.2, 0.014 bp 51.8, 0.015 wp 76.5, 0.014 bsf 44.2, 0.015 VC 266.3, 0.11 bp 254.2, 0.164 wp 410.9, 0.09 bsf 200.8, 0.12

S. No. 1 2 3 4 5 6 7 8

Bearing condition and rotor speed Healthy bearing at rotor speed 1000 rpm Bearing with outer race defect at rotor speed 1000 rpm Bearing with inner race defect at rotor speed 1000 rpm Bearing with ball defect at rotor speed 1000 rpm Healthy bearing at rotor speed 5000 rpm Bearing with outer race defect at rotor speed 5000 rpm Bearing with inner race defect at rotor speed 5000 rpm Bearing with ball defect at rotor speed 5000 rpm

and 8. It can be visualized from Table 5 that the theoretical and experimental results are having same exciting frequencies and their amplitudes are also in close approximation. Table 4 shows that for the trials 18, peak amplitude of vibration appears in the spectrum for horizontal acceleration response and for vertical acceleration response. Through the experiments, the polynomials f x is approximated by the design parameters A , B , C , D. The nal functions of the response surface model are listed as the following. The second-degree polynomial for horizontal acceleration response is R1 = 7.85712 103 5.03125 104 A + 0.022920 B + 2.20681 105 D + 1.10591 105 A D 2.15991 105 B D 22 The second-degree polynomial for vertical acceleration response is

R2 = 0.03082 + 4.671 103 A + 0.040808 B + 0.016945 C + 3.81665 105 D 3.78425 105 B D 1.24213 105 C D 23 The performance prediction of vibration amplitude in horizontal and vertical acceleration responses has been shown in Fig. 15. The actual and predicted values of response in both directions are very close and verify the tness of polynomial for the response. In order to verify whether the obtained polynomials are valuable or not, we perform variance analysis and F-ratio test on them. Table 6 shows the analysis of variance table for horizontal displacement. The model F-value of 4713.87 implies that the model is signicant. Values of prob F less than 0.0500 indicate that model terms are signicant. In this case, A, B, D, AD, and BD are signicant model terms. Values greater than 0.1000 indicate that the model terms are not signicant. Analysis of variance for ver-

Fig. 15 The performance prediction: a horizontal and b vertical acceleration response

031007-10 / Vol. 133, MAY 2011

Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Table 6 Analysis of variance table for horizontal acceleration partial sum of squares: type III Source Model A: outer race defect B: inner race defect D: speed of rotor AD BD Residual Cor total Sum of squares 0.022 1.42 103 2.33 103 4.51 103 6.52 104 2.48 103 1.82 106 0.022 Df 5 1 1 1 1 1 2 7 Mean square 4.43 10 1.42 103 2.33 103 4.51 103 6.52 104 2.48 103 9.4 107
3

F-value 4713.87 1513.09 2485.48 4799.08 693.36 2644.76

p-prob F 0.0002 0.0007 0.0004 0.0002 0.0014 0.0004

tical displacement is shown in Table 7. F-value of 4.2 105 implies that the model is signicant. In this case, A, B, C, D, BD, and CD are signicant model terms. The ANOVA tests of both polynomials show that they are valuable for the corresponding problems. They can be used to analyze the relation between factors and their corresponding responses. After testing the polynomials using ANOVA, RSM of horizontal and vertical acceleration responses are developed. The response surface in Fig. 16a shows the interaction effect of outer race defect and inner race defect on horizontal response. The presence of spall at inner race and outer race increases amplitude of vibration. The response surfaces in Figs. 16b and 16c show the interaction effect of ball defect with outer race defect and inner race defect on horizontal response, respectively. Amplitude of vibration increases with outer race defect and decreases with inner race defect in the presence of ball defect. Ball defects have very less effect on horizontal amplitude of vibration. Figures 16d and 16e show the interaction of rotor speed with outer race defect and inner race defect, respectively. Rotor speed has considerable effect on the horizontal amplitude of vibration. In presence of outer race defect in the bearing, horizontal amplitude of vibration increases more at rotor speed 5000 rpm compared when the rotor is running at speed 1000 rpm, as shown in Fig. 16d. With inner race defect on bearing and at rotor speed 5000 rpm, the catastrophic stage of bearing started and due to self-peening of the bearing defects, amplitude of vibration decreased, as shown in Fig. 16e. Figure 17 shows the interaction effect of various bearing component defects on vertical response. Figure 17a shows that in the presence of outer race defect and inner race defect, vertical amplitude is mainly affected by inner race defect. Figures 17b and 17c show the interaction of ball defect with outer race defect and inner race defect, respectively. Ball defect has considerable effect on vertical amplitude, although it is least affecting horizontal amplitude of vibration. The response surfaces in Figs. 17d and 17e show the interaction effect of rotor speed with inner race defect and ball defect on vertical response, respectively. Vertical amplitude of vibration increases signicantly with rotor speed changes from 1000 rpm to 5000 rpm, as shown in Figs. 17d and 17e.

Conclusions

In the present investigation, an analytical model of a rotor bearing system has been developed to obtain the nonlinear vibration response due to localized defects. The results have been validated under various bearing localized defects with experimentation performed in the laboratory. The present work uses DOE and RSM procedures to conduct several trials for investigating simultaneous effects of localized defects with two levels of rotor speed and to check the combined effect of the various localized parameters. All trials are designed to be able to consider two-way interactions between various factors as well as main effects of individual factors. From the obtained responses, the following conclusions are drawn. 1. Nonlinear dynamic responses are found to be associated with high rotor speed and due to localized defects on inner and outer races Fig. 12 trial 6 and Fig. 13 trial 7. 2. The system shows periodic nature when ball defect is at its maximum level with rotor speed 1000 rpm Fig. 10 trial 4. 3. The system shows quasi-periodic nature with strange attractor when rotor speed 5000 rpm and ball defect are at their maximum level Fig. 14 trial 8. 4. When the outer race defect is present on the bearing components, the peak amplitude of vibration appears at the ball passage frequency Fig. 12 trial 6. 5. When the inner race defect is present on the bearing components, the peak amplitude of vibration appear at the wave passage frequency and at its interaction with inner race frequency Fig. 13 trial 7. 6. When the ball defect is present on the bearing components, the peak amplitude of vibration appear at the ball spin frequency Fig. 14 trial 8.

Acknowledgment
Authors are truly thankful to all the reviewers for their critical comments, which improved the quality of this paper.

Nomenclature
Fi local Hertzian contact force, N Fu force due to unbalance rotor, N

Table 7 Analysis of variance table for vertical acceleration partial sum of squares: type III Source Model A: outer race defect B: inner race defect C: ball defect D: speed of rotor BD CD Residual Cor total Sum of squares 0.036 2.18 105 5.28 103 4.12 104 2.71 103 7.63 103 8.22 104 1.41 108 0.036 Df 6 1 1 1 1 1 1 1 7 Mean square 6.02 103 2.18 105 5.28 103 4.12 104 2.71 103 7.63 103 8.22 104 1.41 108 F-value 4.2 105 1540.73 3.73 105 29,154.85 1.92 105 5.39 105 58,107.93 p-prob F 0.0012 0.0162 0.0010 0.0037 0.0015 0.0009 0.0026

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-11

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 16 Response surfaces showing interaction of a parameters A and B, b parameters A and C, c parameters B and C, d parameters A and D, and e parameters B and D

031007-12 / Vol. 133, MAY 2011

Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Fig. 17 Response surfaces showing interaction of a parameters A and B, b parameters A and C, c parameters B and C, d parameters B and D, and e parameters C and D

Journal of Dynamic Systems, Measurement, and Control

MAY 2011, Vol. 133 / 031007-13

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

i angular location of ith rolling element, rad/s k constant for Hertzian contact elastic deformation, N / m3/2 internal radial clearance, mm Nb number of balls m mass of the rotor, kg c equivalent viscous damping factor, N s / m W radial load, N t time, s ri displacement at the ith ball, mm deformation, mm dimensionless contact deformation Q normal force between rolling element and raceway, N curvature sum, mm1 s additional displacement at the ith ball due to localized defect, mm cage angular speed of the cage, rad/s in angular speed of the inner race, rad/s out angular speed of the outer race, rad/s ball angular speed of the ball, rad/s bp ball passage frequency, Hz VC varying compliance frequency, Hz wp wave passage frequency, Hz bsf ball spin frequency, Hz Do outer race diameter, mm Di inner race diameter, mm D pitch diameter, mm db ball diameter, mm n number of factors noise or error observed in the response polynomial coefcient R1 horizontal acceleration response R2 vertical acceleration response

References
1 Chow, M., Mangum, P. M., and Yee, S. O., 1991, A Neural Network Approach to Real-Time Condition Monitoring of Induction Motors, IEEE Trans. Ind. Electron., 38, pp. 448453. 2 Alguindigue, I. E., Loskiewicz-Buczak, A., and Uhrig, R. E., 1993, Monitoring and Diagnosis of Rolling Element Bearings Using Articial Neural Networks, IEEE Trans. Ind. Electron., 40, pp. 209217. 3 McCormick, A. C., and Nandi, A. K., 1997, Classication of the Rotating Machine Condition Using Articial Neural Networks, Proc. Inst. Mech. Eng.,

Part C: J. Mech. Eng. Sci., 211, pp. 439450. 4 McCormick, A. C., and Nandi, A. K., 1997, Real-Time Classication of Rotating Shaft Loading Conditions Using Articial Neural Networks, IEEE Trans. Neural Netw., 8, pp. 748757. 5 Tse, P. W., and Atherton, D. P., 1999, Prediction of Machine Deterioration Using Vibration Based Fault Trends and Recurrent Neural Networks, ASME J. Vibr. Acoust., 121, pp. 355362. 6 Samanta, B., and Al-Balushi, K. R., 2003, Articial Neural Network Based Fault Diagnostics of Rolling Element Bearings Using Time-Domain Features, Mech. Syst. Signal Process., 172, pp. 317328. 7 Samanta, B., 2004, Gear Fault Detection Using Articial Neural Networks and Support Vector Machines With Genetic Algorithms, Mech. Syst. Signal Process., 18, pp. 625644. 8 Jack, L. B., and Nandi, A. K., 2002, Fault Detection Using Support Vector Machines and Articial Neural Network, Augmented by Genetic Algorithms, Mech. Syst. Signal Process., 16, pp. 373390. 9 Samanta, B., Al-Balushi, K. R., and Al-Araimi, S. A., 2003, Articial Neural Network and Support Vector Machine With Genetic Algorithm for Bearing Fault Detection, Eng. Applic. Artif. Intell., 16, pp. 657665. 10 Sugumaran, V., Muralidharan, V., and Ramachandran, K. I., 2007, Feature Selection Using Decision Tree and Classication Proximal Support Vector Machine for Fault Diagnostic of Roller Bearing, Mech. Syst. Signal Process., 212, pp. 930942. 11 Hu, Q., He, Z., Zhang, Z., and Zi, Y., 2007, Fault Diagnosis of Rotating Machinery Based on Improved Wavelet Package Transform and SVM Ensemble, Mech. Syst. Signal Process., 212, pp. 688705. 12 Wang, W. Q., Golnaraghi, M. F., and Ismail, F., 2004, Prognosis of Machine Health Condition Using Neuro-Fuzzy Systems, Mech. Syst. Signal Process., 18, pp. 813831. 13 Wang, W., Ismail, F., and Golnaraghi, F., 2004, A Neuro-Fuzzy Approach to Gear System Monitoring, IEEE Trans. Fuzzy Syst., 12, pp. 710723. 14 Gallina, A., Martowicz, A., and Uhl, T., 2006, An Application of Response Surface Methodology in the Field of Dynamic Analysis of Mechanical Structures Considering Uncertain Parameters, ISMA 2006 Conference, Leuven, Belgium. 15 Liang, X., Lin, Z., and Zhu, P., 2007, Acoustic Analysis of Damping Structure With Response Surface Method, Appl. Acoust., 68, pp. 10361053. 16 Kankar, P. K., Harsha, S. P., Kumar, P., and Sharma, S. C., 2009, Fault Diagnosis of a Rotor Bearing System Using Response Surface Method, Eur. J. Mech. A/Solids, 28, pp. 841857. 17 Ghafari, S. H., Golnaraghi, F., and Ismail, F., 2008, Effect of Localized Faults on Chaotic Vibration of Rolling Element Bearings, Nonlinear Dyn., 534, pp. 287301. 18 Harsha, S. P., and Kankar, P. K., 2005, Nonlinear Dynamic Analysis of a Complex Rotor Bearing System, Int. J. Acoust. Vib., 101, pp. 3339. 19 Harris, T. A., and Kotzalas, M. N., 2007, Advanced Concepts of Bearing Technology, CRC, New York. 20 Arslan, H., and Aktrk, N., 2008, An Investigation of Rolling Element Vibrations Caused by Local Defects, ASME J. Tribol., 1304, p. 041101. 21 Myers, R. H., and Montgomery, D. C., 2002, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, 2nd ed., Wiley, New York. 22 Nayfeh, A. H., and Balachandran, B., 1995, Applied Nonlinear Dynamics: Analytical, Computational and Experimental Methods, Wiley, New York.

031007-14 / Vol. 133, MAY 2011

Transactions of the ASME

Downloaded From: http://www.asmedl.org/ on 04/28/2013 Terms of Use: http://asme.org/terms

Potrebbero piacerti anche