The Effect of Frequency Resolution in Bearing Fault Studies
Arul Muthukumarasamy & Suri GaneriwalaSpectraQuest Inc., 8227 Hermitage Road, Richmond, VA 23228
Tel: (804)261-3300 www.spectraquest.com
Published: Febraury, 01 2010
Abstract
This technical note illustrates the effects and importance of frequency resolution during the diagnosis of inner and outer race bearing faults using spectral analysis. The bearing faults create impulses and results in strong harmonics of the fault frequencies in the spectrum of vibration signals. These fault frequencies can sometimes be smeared by the adjacent frequencies in the spectra due to their little energy. Hence, a very high spectral resolution is often needed to identify these frequencies during FFT analysis. Experiments were performed using the SpectraQuest's Machinery Fault Simulator and deliberately faulted bearings were studied under different rotational speeds using a higher sampling rate. Different cases with and without misalignments were considered to illustrate the significance of the frequency resolution in resolving these faulty components. Desired resolutions can be estimated depending on the data to help in diagnosing the fault frequencies in the spectra even when they are closer to other components. Reducing the sampling rate and increasing the time record helps in achieving a finer resolution.
Keywords: Frequency resolution, Spectral analysis, rolling element bearing faults
Introduction
Rolling element bearings play a prominent role in today’s machinery. These rolling element bearings wear out easily due to the metal-metal contacts and create faults in outer race, inner race or balls. Early detection of these faults in the bearing is necessary to avoid major machine failures and shut-down time. Usually, these faults create impulsive shocks, sometimes along with some modulating phenomenon. The expected bearing fault frequencies can be estimated using the geometry of the bearing, location of defect and shaft speed. Spectrum and envelope analysis are the major methods that are in use to identify these faults in the frequency spectra.
However, detection of these bearing faults in the spectra is often challenging due to various issues. The impulses from the faults usually have very little energy and are easily overwhelmed by noise and other vibrations. Also, poor resolution in the spectra can mask these fault frequency components, especially when they are close to other stronger frequencies. This paper focuses on the importance of the frequency resolution in the spectra during the bearing fault analysis. A very high resolution is often needed to detect these fault frequencies which demands large number of samples and longer time record. On the other hand, this may increase the computation time and amount of memory needed during the FFT processing, which can be demanding.
| Spectral Lines | Resolution (Hz) | Resolution(rpm) |
100 |
400.0000 |
24000 |
200 |
200.0000 |
12000 |
400 |
100.0000 |
6000 |
800 |
50.0000 |
3000 |
1600 |
25.0000 |
1500 |
3200 |
12.5000 |
750 |
6400 |
6.2500 |
375 |
12800 |
3.1250 |
187.5 |
25600 |
1.5625 |
93.75 |
51200 |
0.7813 |
46.875 |
102400 |
0.3906 |
23.438 |
Table 1: Available Frequency Resolutions for maximum frequency setting of 40 kHz
The different cases that are discussed in this section are listed below:
| Case # | Type of Defect | Shaft Speed (rpm) | Misalignment |
| 1 | Outer race | 1510 | None |
| 2 | Outer race | 4507 | None |
| 3 | Inner race | 4507 | None |
| 4 | Inner race | 1510 | None |
| 5 | Inner race | 4507 | Parallel & Angular |
| 6 | Outer race | 1510 | Parallel & Angular |