Intelligent fault diagnosis of rolling bearings based on refined composite multi-scale dispersion q-complexity and adaptive whale algorithm-extreme learning machine

Abstract

In order to extract the non-linear fault characteristics of rolling bearings more accurately, a novel nonlinear dynamical analysis method, referred to as the refined composite multi-scale dispersion q-complexity (RCMSDQC), is proposed for fault feature extraction of rolling bearings. To improve further the overall performance of the extreme learning machine (ELM) algorithm, the adaptive whale optimization algorithm (AWOA) is used to determine the input weights and hidden layer biases of the ELM. The RCMSDQC has the advantages of strong feature extraction ability and stability compared to the composite multi-scale weighted permutation entropy (CMSWPE) and composite multi-scale permutation entropy (CMSPE) methods. Furthermore, compared to the whale optimization algorithm, particle swarm optimization, and genetic algorithm, the AWOA shows a superior performance in the benchmark function comparison experiment. Based on the experimental rolling bearing data from the Paderborn University, the performance of the proposed method is further evaluated. The experimental results indicate that the proposed fault diagnosis method can identify the type and severity of rolling bearing faults with an accuracy of 99.1%.