The model for fault diagnosis of hydraulic systems based on the PSO SES BPNN algorithm

DOI:	10.11809/bqzbgcxb2023.04.024
Keywords:	hydraulic system; fault diagnosis; BPNN; SES; PSO; hyper parameter optimization
Abstract:	In order to improve fault diagnosis accuracy of a hydraulic system, this paper firstly analyzes data characteristics of the system, and clarifies its data processing method and fault type. Secondly, based on the back propagation neural network (BPNN) algorithm, the gradient optimization process of the BPNN algorithm is improved by using the single exponential smoothing (SES) method. In the third step, particle swarm optimization (PSO) is used to optimize the main hyper parameters of the BPNN algorithm and the smoothing coefficients of the SES, and the optimal hyper parameter model is obtained. Lastly, the hydraulic system fault diagnosis model based on the PSO SES BPNN algorithm is experimentally verified by UCI public data. The experimental results show that the model has a fault diagnosis accuracy of 98.06% for hydraulic cooling and filtering systems with 144 fault states, and the diagnostic performance is much higher than that of the similar literature studies, which is helpful in improving fault diagnosis efficiency of hydraulic systems during operation.
Issue:	Vol. 44 No. 4 (2023)
Published:	2023-04-28
PDF HTML