On-Line Switch-Open Fault Detection of PMSM Using Artificial Neural Network

Abstract

This paper proposes an artificial neural network which can detect a switch-open fault in PMSM. The proposed method does not require any post-processing of the data for fault diagnosis. Also, using the electrical property of the fault, the required number of neuron in the input layer is only 12, so the DSP may perform on-line analysis of the system. The training dataset is obtained through wide operation area considering errors in parameter values. The proposed artificial neural network could classify the status of the machine correctly unless there is zero-speed crossing moment. A fault diagnosis governor concept is introduced, which is to utilize the ANN based on the speed of the machine, and the ANN became able to give right answers all the time. The proposed method was also verified with an experiment, and a switch open fault occurred during the operation could be detected.