Some Observations of the Detection of Rolling Element ...
Some Observations of the Detection of Rolling Element Bearing Outer Race Fault SpectraQuest Inc. 8205 Hermitage Road Richmond, VA 23228 Tel: (804)261-3300 • www.spectraquest.com September, 2006
Abstract: The rolling element bearing is used widely in industry. It is one of the most vulnerable components in a machine because it is most often under high load and high speed running conditions. Prompt diagnostics of rolling element bearing faults is critical not only for the safe operation of machines, but also for the reduction of maintenance cost. Among the three main components in a rolling element bearing, the outer race, the inner race and the ball, this work studies the diagnostics of faults on bearing outer race. There is an argument that a rolling element will excite the natural frequencies of bearing component when it passes the fault on the outer race. Therefore, the bearing outer race natural frequencies have been identified by numerical simulation and hammer test. Vibration data were collected on a Machinery Fault SimulatorTM (MFS) with good as well as faulted bearings. The data were analyzed using the VibraQuestTM software and efforts were made to identify the bearing fault characteristic frequency and its harmonics.
1. Identification of Bearing Outer Race Natural Frequencies The bearing used in this study is illustrated in Fig.1. From the geometries and the number of balls, the outer race fault characteristic frequency can be computed as 3.592.
Figure 1. Rolling Element Bearing Used The natural frequencies and mode shapes of the bearing outer race are simulated. The first three distinctive natural frequencies and mode shapes are illustrated in Fig.2.
(a) Mode 1 (3421 Hz)
(b) Mode 2 (6367 Hz)
(c) Mode 3 (9613 Hz)
Figure 2. Natural Frequencies and Mode Shapes of Bearing Outer Race A tri-axial accelerometer is bolted on a bearing outer race and a hammer test is carried on a bearing outer race. The accelerometer has a frequency range up to 10 kHz. The data were collected using the VibraQuestTM software/hardware system. The results are illustrated in Fig. 3. Three natural frequencies can be identified and they are around 3.5, 5 and 9 kHz respectively. Comparing Figs. 2 and 3, we can see that the hammer test results match the simulation results quite well.
(a) Channel 1
(b) Channel 2
(c) Channel 3
Figure 3. Bearing Outer Race Hammer Test Results However, the above results are for a separate bearing outer ring with free boundary conditions. In operation, the bearing is assembled together and clamped tightly in the bearing housing. The natural frequencies of an assembled and clamped bearing are different than the above results. Another hammer test was carried out on a bearing after it was installed in a MFS. The hammer test results are illustrated in Fig. 4. Figure 4 (a) and (b) are the responses from the horizontal and the vertical accelerometers, respectively. Comparing Fig. 4 and Fig. 3, it can be noticed there are obvious changes of the resonance peaks between them. In Fig. 4, the resonance frequencies concentrate between 2 and 4 kHz. The 5 and 9 kHz peaks in Fig. 3 disappear in Fig. 4. Since the bearing is clamped on the bearing housing of the MFS in the latter hammer test, we can not predict definitely where the resonances illustrated in Fig. 4 come from. It may originate from the bearing components, or it may be created by some other structural parts in the MFS, for instance, the bearing housing.
(a) Horizontal Acceleration
(b) Vertical Acceleration
Figure 4. Bearing Hammer Test Results (bearing clamped in housing)
2. Running Test The running test was carried out on a Machinery Fault SimulatorTM (Fig.5).
Figure 5. Machinery Fault SimulatorTM Four single-channel axial accelerometers were mounted on the inboard and outer board bearing housings in vertical and horizontal directions respectively. The frequency range is up to 10 kHz. 2.1 Faulted Outer Race Bearing Case 1 First, the baseline vibration data for good bearings were collected in different running speed conditions. Next, the inboard good bearing was replaced by a bearing with faulted outer race. The location of the fault on the outer race is in line of the load path. Again, data were collected in different speed conditions. The vibration spectra for these faulted bearing cases are illustrated in Figs. 6 and 7. Figure 6 illustrates the vibration data of inboard bearing in the vertical direction. The frequency
range is from 0 to 2 kHz. The harmonic cursors (parallel dotted lines) indicate the locations of bearing outer race fault characteristic frequency and its harmonics. Please notice that the amplitude is in the dB scale. Some spectrum peaks overlap with the parallel lines. Although the fault characteristic frequency can not be identified, its harmonics indicate an outer race fault. However, these harmonics can not be easily observed in the linear scale.
Figure 6. Harmonics of Outer Race Fault Characteristic Frequency (1000RPM) Fig. 7 illustrates the spectra of acceleration signals obtained on the MFS with the same faulted bearing case but at a different running speed. Fig. 7 (a) is the spectrum of the data collected on the inboard bearing housing (remember the faulted bearing is always mounted inboard). Fig. 7 (b) is the spectrum of the data collected on the out board bearing housing. Comparing Figs. 7 (a) and (b), it can be noticed that the data collected on the out board bearing housing reveal the harmonics of the bearing fault characteristic frequency. The inboard data (nearer to the faulted bearing) do not show these harmonics.
(a) spectrum of inboard data
(b) spectrum of outboard data
Figure 7. Vibration Spectra for a Bearing with Faulted Outer Race (1403RPM)
2.2 Faulted Outer Race Bearing Case 2 A bearing, whose outer race has a hole, was installed on the inboard bearing housing and the vibration data were collected in different speed conditions. During the running tests, the load on the bearings was kept constant. Fig. 8 illustrates the spectrum of the data for a bearing outer race with a hole on it. Fig. 8 (a) shows the harmonics of the fault frequency. Fig. 8 (b) displays the fault characteristic frequency.
(a) Harmonics of Fault Frequency
(b) Fault Frequency
Figure 8. Vibration Spectrum for a Bearing Outer Race with a Hole (1529RPM) Fig. 9 illustrates the spectrum of the data obtained on the MFS with the same bearing outer race with a hole but at a different running speed. However, Fig.9 (a) does not reveal harmonics of fault frequency distinctively. Alike Fig. 8 (b), Fig. 9 (b) displays the fault characteristic frequency.
(a) Harmonics of Fault Frequency (b) Fault Frequency Figure 9. Vibration Spectrum for a Bearing Outer Race with a Hole (1840RPM)
2.3 Faulted Outer Race Bearing Case 3 During another running test with a seriously faulted outer race bearing, data were collected with 40 kHz frequency bandwidth and high enough frequency resolutions. Fig. 10 illustrates the acceleration signal in the time domain. The impact pulses caused by ball passing of outer race fault in the load zone can be identified clearly.
Figure 10. Impact Pulses Caused by Ball Passing Outer Race Fault (1800 RPM) Fig.11 presents the spectrum of the inboard bearing acceleration in the vertical direction. The resonance around 10 kHz is most probably from the accelerometer itself. The resonance around 3 kHz should be identifies as a bearing component resonance.
Figure 11. Vibration Spectrum with Structural Resonances (1800 RPM) Fig.12 (a) and (b) illustrate the vibration spectra of the horizontal and vertical accelerations, respectively. A lot of BPFO harmonics arise. Compared to Fig. 12 (a), in the bearing component resonance area of Fig. 12 (b), the BPFO harmonic components have been amplified significantly.
(a) Horizontal
(b) Vertical
Figure 12. Vibration Spectrum for a Bearing Outer Race with Serious Fault (1800RPM) Using SpectraQuest’s VibraQuest system, the envelope analysis (amplitude demodulation algorithm) was applied to the vertical vibration signal in the resonance area and the modulation frequency, i.e. the BPFO, can be distinguished in the envelope spectrum as illustrated in Fig. 13.
Figure 13. Identification of BPFO using the Envelope Analysis If we take a closer look at the accelerometer resonance area in the spectrum, no clear BPFO harmonic component is present, as shown in Fig. 14.
Figure 14. No BPFO harmonics in Accelerometer Resonance Area.
3. Summary and Remarks In this work, the natural frequencies of an outer race of one rolling element bearing have been simulated as well as measured. The simulation and test results match well. Running tests were carried out with different faulted outer race bearings. The data collected during the running tests were analyzed and several observations can be reported. 1.
2. 3. 4. 5. 6.
The natural frequencies of a rolling element bearing with the free boundary conditions are over 3 kHz. In order to use the bearing component resonance bandwidth to detect the bearing fault at an initial stage, high frequency range accelerometer should be adopted and long period of data need to be acquired. The fault characteristic frequency can only be identified when the fault extent is severe, such as a hole in the outer race. The harmonics of fault frequency are a more sensitive indicator of the bearing outer race fault. The dB amplitude scale or log amplitude scale, in contrast to a linear scale, will reveal the fault frequency harmonics at an earlier stage of a fault for some cases. For the seriously faulted bearing case 3, waveform, spectrum and envelope techniques all reveal the fault. The high frequency demodulation used in the envelope analysis to detect bearing fault characteristic frequencies has to be carried out carefully as some resonances do not contain the fault frequency components.
Abstract: The rolling element bearing is used widely in industry. It is one of the most vulnerable components in a machine because it is most often under high load and high speed running conditions. Prompt diagnostics of rolling element bearing faults is critical not only for the safe operation of machines, but also for the reduction of maintenance cost. Among the three main components in a rolling element bearing, the outer race, the inner race and the ball, this work studies the diagnostics of faults on bearing outer race. There is an argument that a rolling element will excite the natural frequencies of bearing component when it passes the fault on the outer race. Therefore, the bearing outer race natural frequencies have been identified by numerical simulation and hammer test. Vibration data were collected on a Machinery Fault SimulatorTM (MFS) with good as well as faulted bearings. The data were analyzed using the VibraQuestTM software and efforts were made to identify the bearing fault characteristic frequency and its harmonics.
1. Identification of Bearing Outer Race Natural Frequencies The bearing used in this study is illustrated in Fig.1. From the geometries and the number of balls, the outer race fault characteristic frequency can be computed as 3.592.
Figure 1. Rolling Element Bearing Used The natural frequencies and mode shapes of the bearing outer race are simulated. The first three distinctive natural frequencies and mode shapes are illustrated in Fig.2.
(a) Mode 1 (3421 Hz)
(b) Mode 2 (6367 Hz)
(c) Mode 3 (9613 Hz)
Figure 2. Natural Frequencies and Mode Shapes of Bearing Outer Race A tri-axial accelerometer is bolted on a bearing outer race and a hammer test is carried on a bearing outer race. The accelerometer has a frequency range up to 10 kHz. The data were collected using the VibraQuestTM software/hardware system. The results are illustrated in Fig. 3. Three natural frequencies can be identified and they are around 3.5, 5 and 9 kHz respectively. Comparing Figs. 2 and 3, we can see that the hammer test results match the simulation results quite well.
(a) Channel 1
(b) Channel 2
(c) Channel 3
Figure 3. Bearing Outer Race Hammer Test Results However, the above results are for a separate bearing outer ring with free boundary conditions. In operation, the bearing is assembled together and clamped tightly in the bearing housing. The natural frequencies of an assembled and clamped bearing are different than the above results. Another hammer test was carried out on a bearing after it was installed in a MFS. The hammer test results are illustrated in Fig. 4. Figure 4 (a) and (b) are the responses from the horizontal and the vertical accelerometers, respectively. Comparing Fig. 4 and Fig. 3, it can be noticed there are obvious changes of the resonance peaks between them. In Fig. 4, the resonance frequencies concentrate between 2 and 4 kHz. The 5 and 9 kHz peaks in Fig. 3 disappear in Fig. 4. Since the bearing is clamped on the bearing housing of the MFS in the latter hammer test, we can not predict definitely where the resonances illustrated in Fig. 4 come from. It may originate from the bearing components, or it may be created by some other structural parts in the MFS, for instance, the bearing housing.
(a) Horizontal Acceleration
(b) Vertical Acceleration
Figure 4. Bearing Hammer Test Results (bearing clamped in housing)
2. Running Test The running test was carried out on a Machinery Fault SimulatorTM (Fig.5).
Figure 5. Machinery Fault SimulatorTM Four single-channel axial accelerometers were mounted on the inboard and outer board bearing housings in vertical and horizontal directions respectively. The frequency range is up to 10 kHz. 2.1 Faulted Outer Race Bearing Case 1 First, the baseline vibration data for good bearings were collected in different running speed conditions. Next, the inboard good bearing was replaced by a bearing with faulted outer race. The location of the fault on the outer race is in line of the load path. Again, data were collected in different speed conditions. The vibration spectra for these faulted bearing cases are illustrated in Figs. 6 and 7. Figure 6 illustrates the vibration data of inboard bearing in the vertical direction. The frequency
range is from 0 to 2 kHz. The harmonic cursors (parallel dotted lines) indicate the locations of bearing outer race fault characteristic frequency and its harmonics. Please notice that the amplitude is in the dB scale. Some spectrum peaks overlap with the parallel lines. Although the fault characteristic frequency can not be identified, its harmonics indicate an outer race fault. However, these harmonics can not be easily observed in the linear scale.
Figure 6. Harmonics of Outer Race Fault Characteristic Frequency (1000RPM) Fig. 7 illustrates the spectra of acceleration signals obtained on the MFS with the same faulted bearing case but at a different running speed. Fig. 7 (a) is the spectrum of the data collected on the inboard bearing housing (remember the faulted bearing is always mounted inboard). Fig. 7 (b) is the spectrum of the data collected on the out board bearing housing. Comparing Figs. 7 (a) and (b), it can be noticed that the data collected on the out board bearing housing reveal the harmonics of the bearing fault characteristic frequency. The inboard data (nearer to the faulted bearing) do not show these harmonics.
(a) spectrum of inboard data
(b) spectrum of outboard data
Figure 7. Vibration Spectra for a Bearing with Faulted Outer Race (1403RPM)
2.2 Faulted Outer Race Bearing Case 2 A bearing, whose outer race has a hole, was installed on the inboard bearing housing and the vibration data were collected in different speed conditions. During the running tests, the load on the bearings was kept constant. Fig. 8 illustrates the spectrum of the data for a bearing outer race with a hole on it. Fig. 8 (a) shows the harmonics of the fault frequency. Fig. 8 (b) displays the fault characteristic frequency.
(a) Harmonics of Fault Frequency
(b) Fault Frequency
Figure 8. Vibration Spectrum for a Bearing Outer Race with a Hole (1529RPM) Fig. 9 illustrates the spectrum of the data obtained on the MFS with the same bearing outer race with a hole but at a different running speed. However, Fig.9 (a) does not reveal harmonics of fault frequency distinctively. Alike Fig. 8 (b), Fig. 9 (b) displays the fault characteristic frequency.
(a) Harmonics of Fault Frequency (b) Fault Frequency Figure 9. Vibration Spectrum for a Bearing Outer Race with a Hole (1840RPM)
2.3 Faulted Outer Race Bearing Case 3 During another running test with a seriously faulted outer race bearing, data were collected with 40 kHz frequency bandwidth and high enough frequency resolutions. Fig. 10 illustrates the acceleration signal in the time domain. The impact pulses caused by ball passing of outer race fault in the load zone can be identified clearly.
Figure 10. Impact Pulses Caused by Ball Passing Outer Race Fault (1800 RPM) Fig.11 presents the spectrum of the inboard bearing acceleration in the vertical direction. The resonance around 10 kHz is most probably from the accelerometer itself. The resonance around 3 kHz should be identifies as a bearing component resonance.
Figure 11. Vibration Spectrum with Structural Resonances (1800 RPM) Fig.12 (a) and (b) illustrate the vibration spectra of the horizontal and vertical accelerations, respectively. A lot of BPFO harmonics arise. Compared to Fig. 12 (a), in the bearing component resonance area of Fig. 12 (b), the BPFO harmonic components have been amplified significantly.
(a) Horizontal
(b) Vertical
Figure 12. Vibration Spectrum for a Bearing Outer Race with Serious Fault (1800RPM) Using SpectraQuest’s VibraQuest system, the envelope analysis (amplitude demodulation algorithm) was applied to the vertical vibration signal in the resonance area and the modulation frequency, i.e. the BPFO, can be distinguished in the envelope spectrum as illustrated in Fig. 13.
Figure 13. Identification of BPFO using the Envelope Analysis If we take a closer look at the accelerometer resonance area in the spectrum, no clear BPFO harmonic component is present, as shown in Fig. 14.
Figure 14. No BPFO harmonics in Accelerometer Resonance Area.
3. Summary and Remarks In this work, the natural frequencies of an outer race of one rolling element bearing have been simulated as well as measured. The simulation and test results match well. Running tests were carried out with different faulted outer race bearings. The data collected during the running tests were analyzed and several observations can be reported. 1.
2. 3. 4. 5. 6.
The natural frequencies of a rolling element bearing with the free boundary conditions are over 3 kHz. In order to use the bearing component resonance bandwidth to detect the bearing fault at an initial stage, high frequency range accelerometer should be adopted and long period of data need to be acquired. The fault characteristic frequency can only be identified when the fault extent is severe, such as a hole in the outer race. The harmonics of fault frequency are a more sensitive indicator of the bearing outer race fault. The dB amplitude scale or log amplitude scale, in contrast to a linear scale, will reveal the fault frequency harmonics at an earlier stage of a fault for some cases. For the seriously faulted bearing case 3, waveform, spectrum and envelope techniques all reveal the fault. The high frequency demodulation used in the envelope analysis to detect bearing fault characteristic frequencies has to be carried out carefully as some resonances do not contain the fault frequency components.
