Fault Detection in DC Electro Motors Using the Continuous Wavelet Transform

Two time–frequency methods were used to detect typical faults in DC electro motors: the windowed Fourier transform and the continuous wavelet transform. Four groups containing three electro motors each were manufactured with typical faults and examined. These faults included a bearing fault, an increased unbalance, a fragmented brush and a fragmented collector. The velocity of the vibrations at selected points on the electro motors was measured with a laser probe. The parameters of both transforms were selected in order to make both methods comparable. Because of the poor frequency variance, the windowed Fourier transform was, in this case, proven to be inferior to the continuous wavelet transform. Therefore, the continuous wavelet transform was chosen as the primary tool for fault detection.Three criteria were found that successfully discriminated between the typical faults. These were the highest magnitude level, the frequency of the first and second harmonics and the time period between the magnitude pulses in the third (highest) frequency region. If the maximum magnitude levels versus the period of the pulses in the third frequency range are plotted, four distinct regions corresponding to four different faults are obtained. Since the regions do not overlap, linear classifiers can be used with the presented criteria.