Current-Sensor and Switch-Open Fault Diagnosis Based on Discriminative Machine Learning Model for PMSM Driving System

This paper presents a diagnosis method for current-sensor and switch-open faults using an Artificial Neural Network(ANN) in PMSM driving systems. The proposed uses a discriminative machine learning model. It utilizes the operating condition for machine learning. The ANN model is designed to do diagnosis both in steady and dynamic conditions over a wide driving area. In this research, data are acquired within experiments, and a machine learning model is trained using Keras with TensorFlow backend. The performance of the proposed model is validated in a statistical way. With the proposed model, diagnosis accuracy is improved from the existing ones. Also, experimental validation is performed using a TI28379S control board and 200W Permanent Magnet Synchronous Motor(PMSM).