Incipient Fault Diagnosis of Cylindrical Roller Bearings Using Dynamic Response Analysis
Abstract
Detecting faults early on is important in order to maintain the health of rotating machinery. Incipient faults in rolling-element bearings lead to the generation of micro-defects which create weak transient impact signals. However, these are often buried within the background vibration of the machine and structural transmission effects. Indicators based on traditional time-domain methods possess sensitivity to impulsive damage, but when the signal exhibits non-stationarity or a noisy trend. The current study compares time-domain statistical parameters and FFT spectral analysis for progressive inner race damage in an NJ307 cylindrical roller bearing. A wire electrical-discharge-machining technique produces artificial inner-race defects with sizes of 0.25 mm, 0.50 mm, 0.75 mm and 1.00 mm, while a healthy bearing serves as the reference condition. Vibration is measured with a single-axis accelerometer connected to an eight-channel Dewesoft data- acquisition system with a sampling frequency of 20 kHz; the shaft speed is 800 rpm by use of a variable-frequency drive. The RMS, standard deviation, peak value, crest factor, kurtosis, and form factor are taken into account during the analysis. FFT spectra are observed at the calculated inner-race fault frequency and its harmonics. The findings reveal that at the earliest damage stage, there is only a gradual change in both RMS and standard deviation. In contrast, changes in peak value, crest factor, kurtosis, and amplitudes of the BPFI-related FFT are more robust indicators of progressive damage. The FFT provides statistically interpretable frequency information, which scalar time-domain features cannot. As suggested by the findings, the proposition is for a combined diagnostic procedure in which the time-domain features provide a rapid screening while the FFT confirms the bearing-fault frequency. Keywords: cylindrical roller bearing; incipient fault; dynamic response; condition monitoring