ITSC fault diagnosis for PMSM by using adaptive filtering and tree-structured parzen estimator optimized-automated random forest

Inter-turn short circuit (ITSC) fault is one of the most common faults in permanent magnet synchronous motors. Early identification and diagnosis of ITSC faults are especially critical to prevent possible secondary and cascading faults. In this paper, an intelligent fault diagnosis method based on adaptive filtering and tree-structured parzen estimator optimized-automated random forest is proposed for the strong noise interfering current signals of ITSC faults. Firstly, the three-phase current signals are fused into a new modal signal by data-level fusion, and adaptive filtering using ensemble empirical mode decomposition and Pearson’s correlation coefficient is performed to effectively filter out outliers and high-frequency noise. Secondly, the tree-structured parzen estimator is used to automatically optimize the hyperparameters of the random forest model. The experimental results show that the fault diagnosis accuracy of the filtered signal is increased to 95.83% compared to the raw signal. Under the training condition of small samples, the diagnosis accuracy of the filtered signal reaches 93.61%, which proves the effectiveness and practicality of the method.