Fault diagnosis for drivetrain gearboxes using PSO-optimized multiclass SVM classifier

A novel method consisting of an adaptive feature extraction scheme and a particle swarm optimization (PSO)-optimized multiclass support vector machine (SVM) classifier is proposed for condition monitoring and fault diagnosis of drivetrain gearboxes in variable-speed operational conditions. The adaptive feature extraction scheme consists of an adaptive signal resampling algorithm, a frequency tracker, and a feature generation algorithm for effective extraction of the features of gearbox faults from the stator current signal of the AC electric machine connected to the gearbox. The multiclass SVM classifier is designed to identify different faults in the gearbox according to the fault features extracted. The PSO algorithm is utilized to optimize the parameter setting of the SVM classifier to obtain the best classification accuracy. The proposed method is testified on a drivetrain gearbox connected with a permanent-magnet synchronous machine with three different faults. Experimental results show that the faults can be effectively classified by the proposed method.