ENGINE FAULT DIAGNOSIS BASED ON DEEP BELIEF NETWORK IMPROVED BY ADAPTIVE CROW SEARCH ALGORITHM (MT)

Abstract

Based on the crow search algorithm（CSA）, by the adaptive crow search algorithm（ACSA） the adaptive value strategy of sensing probability and flight distance, which effectively enhanced the performance of the algorithm is designed. In view of the fact that the selection of deep belief network（DBN） model parameters has great influence on the engine fault diagnosis results, ACSA is used to optimize the selection of its model parameters, and an engine fault diagnosis method based on DBN improved by ACSA is proposed. The Engine fault diagnosis example results show that the ACSA algorithm can obtain better DBN model parameters, and obtain higher engine fault diagnosis accuracy in less time than other methods.

Keywords

• Deep belief network;Crow search algorithm;Adaptive parameter;Engine;Fault diagnosis