Bearing failure could lead to unpredicted productivity loss to the industries or catastrophic failure of mechanical components. Hence, bearing fault detection and diagnosis is an essential part of the predictive maintenance procedure. Nowadays, the usage of micro-electro mechanical system (MEMS) based sensors for machine fault diagnosis have shown increasing trend due to their size, cost, portability, and flexibility. This paper presents the use of cost effective MEMS based accelerometer to acquire vibration signals for three conditions of REB namely, normal (N), defect on inner race (IR), defect on outer race (OR) at variable load and high shaft speed on a customised bearing test rig. The acquired vibration signals are denoised and wavelet packet transform is used to decompose the signal. Statistical features have been extracted from the wavelet packet coefficients. These coefficients are given as inputs to two widely used classifiers namely Artificial Neural Network (ANN) and k-Nearest Neighbor (kNN) to check the reliability of MEMS accelerometer data for REB fault diagnosis. From the results, it is shown that the proposed fault diagnosis scheme using MEMS accelerometers yield 83.3% and 77.8% classification accuracy using ANN and kNN respectively. Accordingly, MEMS based accelerometers are enough to replace conventional accelerometers that are used for REB fault diagnosis.
B et al. (Sun,) studied this question.