Mechanical Fault Diagnosis Method Based on Machine Learning

This paper proposes a novel mechanical fault diagnosis method using a hybrid QPSO and SVM model. Mechanical fault diagnosis refers to the recognition and diagnosis of fault mechanism, fault causes, and the fault positions. Particularly, five types of mechanical faults are considered in this paper, which are 1) quality not balancing, 2) Rotor thermal bending, 3) Shaft crack, 4) Bearing fault and 5) Permanent bending. The main innovations of this paper lie in that we introduce the SVM classifier to solve the mechanical fault diagnosis problem, and then Quantum behaved particle swarm optimization is utilized to optimized the parameters of SVM. Experimental results demonstrate that, using the proposed algorithm, the accuracy of mechanical fault diagnosis is greatly enhanced than SVM and PSO-SVM model.