A Wavelet Entropy-Based Approach to Select Structure Element of Morphological Filter for Bearing Fault Detection

Abstract

Vibration signal acquired from the bearing often contains components from other elements. Background noise is also involved in the measurement of vibration signals, which creates a big challenge in extracting the fault features. A novel method with a modified morphological filter is proposed to extract the fault features from faulty rolling element bearing vibration signal by reducing the noises as well as other unwanted components. Since outcome of a morphological filter depends on the proper selection of structure elements (SE), a novel technique to select the optimum length of SE based on wavelet entropy is proposed. The modified morphological filter is verified by simulating impulse signals—as well as experimental vibration signals of faulty bearing. Both simulation and experimental results of the proposed method show that noises are reduced effectively and the impulse components (i.e., fault frequencies and rotational frequencies) are extracted efficiently, which implies that the proposed method results in superior performance, particularly for the bearing incipient faults.