An Improved LSSVM Fault Diagnosis Classification Method Based on Cross Genetic Particle Swarm

Abstract

It is difficult to select the parameters of least squares support vector machine (LSSVM) when studying the classification algorithm, A particle swarm optimization algorithm based on crisscross inheritance method is proposed to find the optimal parameters of LSSVM. Further, the wavelet packet is adopted to process the bearing signal and extract time-frequency domain features, which are used as the input of the LSSVM. The classification model is established and applied to identify the fault of bearing. Classification result shows the classification accuracy is improved, and the LSSVM is optimized.