An Integrated Approach Fusing CEEMD Energy Entropy and Sparrow Search Algorithm-Based PNN for Fault Diagnosis of Rolling Bearings.

Abstract

This paper solves the problem of difficulty in achieving satisfactory results with traditional methods of bearing fault diagnosis, which can effectively extract the fault information and improve the fault diagnosis accuracy. This paper proposes a novel artificial intelligence fault diagnosis method by integrating complementary ensemble empirical mode decomposition (CEEMD), energy entropy (EE), and probabilistic neural network (PNN) optimized by a sparrow search algorithm (SSA). The vibration signal of rolling bear was firstly decomposed by CEEMD into a set of intrinsic mode functions (IMFs) at different time scales. Then, the correlation coefficient was used as a selection criterion to determine the effective IMFs, and the signal features were extracted by EE as the input of the diagnosis model to suppress the influence of the redundant information and maximize the retention of the original signal features. Afterwards, SSA was used to optimize the smoothing factor parameter of PNN to reduce the influence of human factors on the neural network and improve the performance of the fault diagnosis model. Finally, the proposed CEEMD-EE-SSA-PNN method was verified and evaluated by experiments. The experimental results indicate that the presented method can accurately identify different fault states of rolling bearings and achieve better classification performance of fault states compared with other methods.