Transient signal period detection for bearing fault diagnosis using amplitude demodulation and signal segmentation and labeling

When a localized fault occurs in a bearing, the transient signals will appear in the acquired signal. The bearing fault mode can then be determined by analysis of the period of the transient signals. Many advanced signal processing algorithms have been proposed to realize period detection, but majority of them are complicated and require adequate computational resource. This paper proposed a new method characterized by low computational cost and low memory occupation, to detect the period of transient signals for bearing fault diagnosis. First, the fault-induced impulse signals are demodulated using the squaring and low-pass filtering envelope algorithm. Subsequently, the period of the envelope signal is estimated using the signal segmentation and labeling algorithm. Note that the proposed method avoids the complex convolution operation and hence it is suitable for implementing in an embedded system. Finally, experimental verification is conducted on a micro control unit to diagnose the bearing faults.