Fault Diagnosis of Hydraulic System Based on MIBBPSO-SDAE
For the purpose of monitoring the fault state of the hydraulic system,it is necessary to install a variety of sensors,the col-lected data is huge and complex,a variety of features can be obtained through the multi-feature calculation.In order to make the diagno-sis more accurate,a MIBBPSO-SDAE fault diagnosis method was proposed.Based on the particle swarm optimization based on mutual information(MIBBPSO)feature selection method and the bee colony initialization strategy for label correlation,the correlation between features and class labels were used to accelerate convergence;two local search operators were used to reinforce the utilization perform-ance of the algorithm,and an adaptive flipped variation operator was used to find out the optimal subset of features.The filtered feature subset was used for data fusion and fed into the model of a trained stacked denoising auto encoder(SDAE)for fault diagnosis.The re-sults show that the diagnostic accuracies of the MIBBPSO-SDAE method for the piston pump,cooler,throttle valve and accumulator,are 99.5%,100%,96.52%and 98.1%,respectively,which can identify the fault types more accurately.
万方数据
维普
