VIBRATION MONITORING OF VERY SLOW SPEED THRUST BALL BEARINGS Vipul M. Patel and Naresh Tandon ITMME Centre, Indian Institute of Technology, Hauz Khas, New Delhi 110016, India e-mail: ntandon@itmmec.iitd.ernet.in Condition monitoring of slow speed bearing is reported to be difficult with conventional vibration monitoring technique because of high the noise to signal ratio. This research work was carried out in order to investigate the applicability of vibration monitoring techniques for different defects seeded on thrust ball bearing race, running under different operating speeds. The measurements were performed on a test rig developed for this purpose. The study reveals that standard accelerometer can be used for vibration monitoring of slow speed bearings. The results show that the defect frequencies measured are different from calculated defect frequencies due to slip.. 1. Introduction Rolling element bearings play vital role in the domestic as well as in industrial equipments. Proper functioning of these equipments is ensured by quiet and smooth running of these bearings. An unexpected failure of bearing in industry can bring production to a standstill. Failure of the bearings can be avoided by early detection of the minor errors and defect present in the bearing. Bearing acts as a source of vibration and noise due to either varying compliance or the presence of defect in them. Whenever a local defect on an element interacts with its mating element, abrupt changes in the contact stresses at the interface occur. This results in generation of very short duration pulse which causes vibration and noise. By monitoring these, we can predict the health of the bearing. Examples of slow speed machinery in industries are wind turbine power plant, rolling machinery of paper mill, machines used in steel pipe and mining industries, mixers air preheater in thermal power plant, rotating biological contractor for processing waste water and many more. Many of these machines have bearings running at less than 30 rpm. Vibration monitoring is a well established technique for monitoring high and medium speed bearings but reported to be not so successful for monitoring slow speed bearings. Difficulties associated with Vibration monitoring of slow speed bearings are; i) the energy released rate from defect reduces as speed reduces. ii) the associated defect repetition frequencies become very low and difficult to detect amongst background noise. iii) very long time records need to be digitized and further processed. iv) conventional vibration system employs accelerometer, though displacement sensors are best suited to measuring low frequency vibration. v) inherent low-frequency instrument noise problems and, low-frequency roll-off filters of typical data analyzers. This requirement has the disadvantage of disguising the low frequencies of interest. vi) similarly, most sensors have roll-off filters that affect the magnitude ICSV20, Bangkok, Thailand, 7-11 July 2013 1
20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 of signals detected within the roll-off frequency range. vii) if frequency analysis (Fourier transformation) is required, then the signal processing itself produces very large mathematical errors at the very low frequency end of the spectral range, which changes the genuine frequency data and alters the overall amplitude values. Slow speed rolling bearings are often very massive and stiff. Vibration monitoring technique has many methods for fault detection such as: overall vibrations, time waveform analysis, high frequency resonance technique (also called, demodulation or enveloping), phase analysis, band trending and cepstrum analysis. In the time domain, typically statistical features of the measured vibration signal such as R.M.S., peak value, kurtosis etc. are trended over the duration of the test and change in the pattern are attributed to the presence of defects. Most published work is related to the applicability of vibration monitoring technique in detecting seeded faults artificially introduced on the bearing. Ray 1 investigated the use of SPM to monitor low-speed bearings (25 rpm) and concluded that the SPM failed to register defects at speeds below 750 rpm. Prashad 2 has reported the effect of cage and roller slip on the measured defect frequency response of rolling element bearings with the help of HFRT as a surveillance module. Percentage roller slip varies between -5.8 to 7.1. It was found that negative roller slip is predominant at moderate speed under no load and load operation. Positive slip is significant at high speed under no load operation. Murphy 3 suggested that a strong magnetic clamp be used to attach the sensor to the machine, thus avoiding the rocking effects that would be experienced with hand held probe, thereby reducing unwanted noise at very low frequencies. Tandon and Nakra 4 have compared the vibration and acoustic measurement method for defect detection of ball bearing. Measurements have been performed on new bearings and bearings with seeded defects in their elements, for 100 to 1500 rpm range. The authors have measured overall vibration acceleration, envelope detected acceleration, overall sound intensity and sound pressure, shock pulse, and acoustic emission ring down counts and peak amplitude. The results indicate that, in general, the detection of defects by acoustic emission and envelope detected acceleration are better as compared to other measurements. Robinson and Canada 5 described that by using a low noise accelerometer having sensitivity of 500mV/g, and state of the art portable data collector with sufficient dynamic range, a meaningful vibration analysis can be carried out on slow speed machinery. They developed a slow speed technology (SST) system for measuring vibrations on low-speed rotating machinery which was based on separating the high frequency noise of the machine from the low frequency signatures of interest. Furthermore, it is reported that this method could be applied at speeds as low as 10 rpm. Robinson 6 suggested new methodology for vibration monitoring of low speed machinery. This builds on SST method described above. It involves segmenting the signal in to time intervals, depended on the sampling frequency and obtaining peak values for continuing time intervals until the desired number are captured for processing, called peak value techniques. Processing peak values involved spectrum analysis. This method was successfully applied on bearing rotating at 10 rpm with low frequency accelerometer and detected the inner race defect one month prior to failure. Barratt 7 suggested that low frequency piezoceramic sensor can be used for condition monitoring of low frequency. Piezoceramic transducer provides superior performance over the broad frequency and amplitude ranges required in industrial applications. Enveloping method can be used to study signals as enveloping enhances the response signals of small repetitive defect impacts. Mba 8 described main problems associated with vibration analysis in to three sections: i) conventional vibration system employs accelerometer, though displacement sensors are best suited for measuring low frequency vibration. ii) inherent low frequency noise problem and low frequency roll off filters for typical data analyzers, which will have negative effect on low frequency of interest. iii) most sensors have a roll off filters that affects the magnitude of the signals detected within the roll- off frequency range. Shumin 9 mentioned that at low rotating speed, weak impact of rolling element with defect may not be able to excite the resonance of the piezoelectric sensor. Hence, they designed special sensor for monitoring of low speed bearing. Natural frequency of this sensor can be changed ICSV20, Bangkok, Thailand, 7-11 July 2013 2
20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 by varying the length of cantilever plate (integral part of sensor). Selvaraj and Marappan 10 investigated the effect of operating parameters (shaft speed, radial load, viscosity of lubricating oil, number of rollers and bearing temperature) on cage slip in cylindrical roller bearing. They have concluded that the decreased load cause initiation of cage slip whereas the increased shaft speed increase the magnitude of the cage slip. Increased viscosity increases the cage slip. Increased bearing temperature decreases the cage slip because of the reduced viscosity and no significant changes observed in the critical load. The literature indicates that conventional vibration acquisition setup may not be capable for monitoring slow speed bearing. There is a need for emphasis on proper selection of vibration sensor which has high sensitivity and very low cut off frequency and there may be deviation of defect frequency observed because of either slippage or skidding action of balls and cage. Percentage roller slip varies between -5.8 to 7.1, depending on operating condition. Present work was undertaken to study the vibration spectrum of extremely slow speed thrust bearing with seeded d e- fects using standard accelerometer. 2. Experimental set up and tests Test rig used for the present work is shown in Fig. 1. In this rig, a shaft can be driven at very low rpm (2 to 18) by a motor using a variable speed drive which controls the frequency of incoming power and helps in maintaining desired speed. Test bearing used is SKF 51204 thrust ball bearing. The test bearing is at the top end of the shaft. Vibration transducer was mounted on the top of the hollow cup with the help of magnetic clamp. The defects of 500 µm, 1 mm and 1.5 mm diameters with 300 µm depth were created in the test bearing. To measure the low frequency vibrations, frequency range of FFT analyzer is important. The transducer used for measuring vibration signal was piezoelectric type transducer Bruel & Kjaer type 4368 accelerometer along with Bruel & Kjaer type 2635 charge amplifier. B & K Pulse analyzer 3560C was used for this work because it has frequency range starting from direct DC, hence suitable to measure low frequency vibration as well as it is compatible with vibration transducer. Transducer was mounted just above the test bearing to ensure that it was kept just close to loading zone. Motor Variable speed drive Test Bearing Weight holder Figure 1. The test rig assembly. ICSV20, Bangkok, Thailand, 7-11 July 2013 3
Vibration amplitude (µm/s 2 ) 20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 The test rig was operated under constant load of 10 kg and different operating speeds of 12 rpm, 14 rpm and 16 rpm. Three different sizes of defects were seeded with spark erosion technique on the upper race of the bearing. Frequency span was set up to be 6.25 Hz. Number of lines selected were 800 and signal enhancing mode was used. Bearing defect frequency is given by: Nb Bd cos 1 RPM 2 Pd Where N b is number of balls in the bearing, B d is ball diameter, P d is pitch circle diameter of the bearing and α is the contact angle. Putting the dimensions for SKF 51204 bearing in the above equation, the calculated bearing defect frequency is 6 times the RPM (α = 90 ). 3. Results and discussion Vibration spectra of new test bearing at speeds of 12, 14 and 16 rpm were observed and did not show peak at bearing defect frequencies. The spectra of bearing with 500 µm size defect are shown in Figs. 2 to 4. At 10 kg load and for the operating speeds of 12, 14 and 16 rpm, peaks were observed at 1.146 Hz, 1.336 Hz and 1.484 Hz instead of bearing defect frequencies i.e. 1.2 Hz, 1.4 Hz and 1.6 Hz respectively. This appears to be because skidding action of balls and cages. Under skidding action, balls rotate at single position rather than travelling with pure rotary motion. Hence less number of balls pass through defect and defect frequency lesser than calculated one is obtained. Percentage deviation of bearing defect frequencies from calculated one is shown in Table 1. 950 850 750 650 550 450 350 250 150 50 Figure 2. Vibration spectrum of bearing with 500 µm defect at 12 rpm. Speed (rpm) Table 1. Deviation of bearing defect frequencies for defect size of 500 µm. Calculated bearing defect frequency (Hz) Observed bearing defect frequency (Hz) % Deviation 12 1.2 1.148 4.33 14 1.4 1.336 4.57 16 1.6 1.461 8.68 ICSV20, Bangkok, Thailand, 7-11 July 2013 4
Vibration amplitude (µm/s 2 ) Vibration amplitude (µm/s 2 ) 20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 950 850 750 650 550 450 350 250 150 50 Figure 3. Vibration spectrum of bearing with 500 µm defect at 14 rpm. 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 Figure 4. Vibration spectrum of bearing with 500 µm defect at 16 rpm. The defect size was raised to 1 mm with spark erosion machine. The FFT was taken at the same speeds (12, 14, 16 rpm). Figure 5 shows frequency spectrum of defective bearing at 14 rpm. Here also peaks were observed at frequencies which are deviated from calculated defect frequencies are observed because of skidding action of balls. Percentage deviations of frequencies are shown in Table 2. ICSV20, Bangkok, Thailand, 7-11 July 2013 5
Vibration amplitude (µm/s 2 ) Vibration amplitude (µm/s 2 ) 20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 950 850 750 650 550 450 350 250 150 50 Figure 5. Vibration spectrum of bearing with 1 mm defect at 14 rpm. Table 2. Deviation of bearing defect frequencies for defect size of 1 mm. Speed (rpm) Calculated bearing Observed bearing defect frequency (Hz) defect frequency (Hz) % Deviation 12 1.2 1.164 3 14 1.4 1.352 3.42 16 1.6 1.484 7.25 The defect size was then raised to 1.5 mm and the FFT was obtained taken at 12, 14 and 16 rpm. The frequency spectrum at 14 rpm is shown in Fig. 6. Here again there is a deviation in defect frequencies as compared to calculated defect frequencies of the thrust bearing used, as given in Table 3. 950 850 750 650 550 450 350 250 150 50 Figure 6. Vibration spectrum of bearing with 1.5 mm defect at 14 rpm. ICSV20, Bangkok, Thailand, 7-11 July 2013 6
20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 Table 3. Deviation of bearing defect frequencies for defect size of 1.5 mm. Speed (rpm) Calculated bearing Observed bearing defect frequency (Hz) defect frequency (Hz) % Deviation 12 1.2 1.172 2.8 14 1.4 1.359 2.92 16 1.6 1.484 7.25 Figures 2 to 6 show a general decrease of the vibration amplitudes at defect frequencies, with increase in defect size. This is opposite to that of usual expected trend and could be due to skidding. From Fig. 7, it is observed that % deviation decreases with increasing defect size and hence % of skidding and thus impact energy associated with it is also decreasing. It is also observed that at higher speeds the skidding is more (Fig. 11). Figure. 7 Variation of % deviation of defect frequency from calculated defect frequency with defect size. 4. Conclusions Defects of extremely slow speed rolling bearings can be detected even by using a standard piezoelectric type transducer. The instrument settings are very important for capturing the signal. Because of skidding of balls, bearing defect frequency may deviate from theoretically calculated bearing defect frequency. This deviation increases with increase in bearing speed. The percentage deviation in this study varied from 2.8 % to 8.68 % for the operating conditions used. REFERENCES 1 2 Ray AG, Monitoring rolling contact bearings under adverse conditions, Proceedings of 2 nd International Conference on Vibration in Rotating Machinery, Cambridge, 187 94 (1980). Prashad H, The effect of cage and roller slip on the Measured Defect Frequency response of Rolling Element Bearings, Tribology Transactions, 30(3), 360-367, (1987). ICSV20, Bangkok, Thailand, 7-11 July 2013 7
20th International Congress on Sound and Vibration (ICSV20), Bangkok, Thailand, 7-11 July 2013 3 4 5 6 7 8 9 10 Murphy TJ, The development of a data collector for low speed machinery, Proceedings 4 th International Conference on Profitable Condition Monitoring,, Stratford-upon-Avon, UK, 8 10 December, 258, (1992). Tandon N and Nakra BC, Comparison of vibration and acoustic measurement techniques for the condition monitoring of rolling element bearings, Tribology International, 25(3), 205-212, (1992). Robinson JC, Canada RG. Vibration Measurements on Slow Speed Machinery, Proceedings Predictive maintenance technology national conference, 33-37, (1995). Robinson JC, Vibration monitoring on slow speed machinery: new methodologies covering machinery from 0.5 to 600 rpm, Proceedings 5 th International Conference on Profitable Condition Monitoring Fluids and Machinery Performance Monitoring, BHR Group Publication 22, Cranfield, 169 182, (1996). Barratt Mel, Low speed bearing monitoring-a case study of low speed bearing monitoring in a paperboard plant. An application paper from SKF reliability system, www.skf.com/portal/skf/home/aptitudexchange?contentid=0.237932.237933.237934.23795 9.238383; (2002) Mba D, Applicability of Acoustic emission to monitoring the mechanical integrity of bolted structures in low speed rotating machinery, NDT &E International, 35, 293-300, (2002). Shumin Hou, A New Low Frequency Resonance Sensor for low speed Roller bearing monitoring, Journal of Vibration and Acoustics, 132(1), 014502, (2010). Selvaraj A and Marappan R, Experimental analysis of factors influencing the cage slip in cylindrical roller bearing, International journal of advance manufacturing technology, 53, 635-644, (2010). ICSV20, Bangkok, Thailand, 7-11 July 2013 8