Fault Detection and Classification of Machinery Bearing Under Variable Operating Conditions Based on Wavelet Transform and CNN

Abstract

The rolling bearing, one of the most critical components of wind turbines, is subject to variable operating conditions because of the unsteadiness of environmental aspects. The development of an efficient technique for predicting and classifying rolling bearing faults is a critical task. Condition-based maintenance (CBM) techniques are used to plan the maintenance procedure depending on the actual state of the asset. The convolutional neural network (CNN) structure was built using a neural architecture search (NAS) approach based on reinforcement learning. The time-series data of vibration signals are preprocessed using the continuous wavelet transform (CWT) before delivering to the CNN. The results confirmed that the proposed approach would automatically learn and discover distinct features from vibration signals, as well as identify various rolling bearing health conditions.