Fault Diagnosis of Rolling Bearing Based on Variable Mode Decomposition and Multi-Class Relevance Vector Machine

Under variable load conditions, the bearing vibration signal is non-stationary, which renders ineffective the techniques used for bearing fault diagnosis under constant running conditions. A fault diagnosis model of the variational mode decomposition (VMD) and multi-classification correlation vector machine (MRVM) based on chaotic quantum particle swarm optimization (CQPSO) is proposed. First, the number of intrinsic mode function (IMF) and penalty parameter of VMD is optimized by QPSO algorithm to search the optimal combination value of two parameters. Then, the optimal combination of parameter values corresponding to the parameters of VMD algorithm are set, and decompose the known fault signal. The two dimensional marginal spectral entropy of the IMF component is used as the input eigenvector of the multi-classification RVM. Finally, the experimental data under variable load conditions are used to verify the method. The experimental results show that the proposed method can accurately diagnose the type and degree of the bearing fault in the variable load condition with high diagnostic accuracy and strong robustness.