Inter-Turn Short-Circuit Fault Diagnosis and Severity Estimation for Five-Phase PMSM

Abstract

In this article, an inter-turn short-circuit (ITSC) fault diagnosis and severity estimation method based on extended state observer (ESO) and convolutional neural network (CNN) is proposed for five-phase permanent magnet synchronous motor (PMSM) drives. The relationship between fault parameters and motor parameters is analyzed and the equivalent model of ITSC faults in the natural reference frame is accordingly derived. To achieve fault detection and location, the short-circuit turn ratio and short-circuit current are integrated as the fault diagnosis index. According to the model of the shortcircuit current, an ESO is designed for the estimation of the fault diagnosis index. Further, the sensitivity analysis among fault parameters is conducted to evaluate the short-circuit turn ratio and the short-circuit resistance. Subsequently, the postfault current, back electromotive force, electrical angular velocity, q1-axis current reference and the fault diagnosis index are selected as the input signals of CNN to estimate the short-circuit turn ratio. This approach not only resolves parameter coupling challenges but also provides a quantitative assessment of fault severity. Finally, simulations and experiments under different operating points validate the effectiveness of the proposed method.