Fault diagnosis method of air compressor based on wolf swarm algorithm and improved neural network

Abstract

The faults occurred during the uninterrupted operation of air compressors are characterized by high complexity, high similarity between different fault characteristics and difficult diagnosis. On this premise, to ensure the normal operation of air compressors, the fault diagnosis system is required to ensure high timeliness while accurate diagnosis. The traditional BP network algorithm has some deficiencies in convergence speed and training accuracy, so it is not suitable for direct application in fault diagnosis system. Therefore, in this paper, the Wolf pack algorithm is further fused with the improved LM-BP (Levenberg-Marquardt-Back Propagation) algorithm to improve the speed and accuracy of fault diagnosis. The algorithm was tested on the data set provided by a gas transmission company, and the fault air compressor was quickly identified and classified by analyzing the characterization data. The results show that the fusion algorithm based on Wolf pack algorithm and LM-BP neural network can significantly improve the convergence speed and identification accuracy of the fault diagnosis of air compressor.