Exploring real-time fault detection of high-speed train traction motor based on machine learning and wavelet analysis

The implementation of real-time fault detection technology for key components of high-speed train traction electromechanical system is of great significance for improving train motor reliability and reducing guarantee costs. It has become an inevitable trend. Machine learning is a new and powerful means of researching fault detection technology. Based on machine learning, this paper conducts real-time fault detection technology research on the electromechanical actuators of the key components of electromechanical systems. A model of the electromechanical actuator is established, and the mechanism, influence and fault injection method of the three faults of the electromechanical actuator motor shaft jamming, gear broken tooth, excessive ball screw clearance and two faults of internal leakage are analyzed. On this basis, a fault simulation model of the traction motor of a high-speed train was built to obtain simulation fault data. At the same time, based on wavelet packet decomposition and reconstruction, the fault simulation data of electromechanical actuators and hydraulic pumps are analyzed, the wavelet packet energy distribution is calculated, and time-domain statistics are combined to extract energy feature vectors that can reflect component fault characteristics. This paper proposes a fault diagnosis method for electromechanical actuators based on machine learning, designs, and improves the neural network learning algorithm and network parameters, and improves the classification effect of the neural network; proposes a fault diagnosis method based on the GA-SVM algorithm, using real values. The coded genetic algorithm improves the parameter optimization of the support vector machine, and improves the classification speed of the support vector machine. Finally, the effectiveness and superiority of the two fault diagnosis methods designed in this paper are verified on their respective objects.