Study on Mechanical Concurrent Fault Diagnosis Method Based on PSO-MRLSSVM

Abstract

In view of those characteristics such as nonlinearity and high dimensional features for mechanical concurrent faults and any conventional classifier being not able to fit multi-output needs, a concurrent fault classification method based on PSO-MRLSSVM was put forward, where the simple and fast searching ability of PSO may be utilized to optimize the penalty and Kernel parameters for the MRLSSVM algorithm; thus, the aimlessness (manually specifying parameters) may be prevented and the prediction precision of MRLSSVM may be improved. Experimental results indicate that our mechanical concurrent fault diagnosis model based on PSO-MRLSSVMs works well for effective identification of concurrent fault types and diagnosis effects are good.