Compound Fault Diagnosis for Motor Drive System Based on Adaptive Filtering Enhanced Physical Polar Diagram and SSA-RICNN

Compound fault diagnosis has long been a research hotspot for rotating machinery, in which it is particularly difficult to classify and isolate the similar fault characteristics in frequency domain. To this end, this paper proposes a compound fault diagnosis for motor drive system based on adaptive filtering enhanced physical polar diagram and sparrow search algorithm optimized rotational invariant convolutional neural network (SSA-RICNN), which can distinguish similar fault features by fusing self-sensing motor drive signals. Firstly, the adaptive filtering method is utilized to remove the irrelevant harmonic components in the q-axis current signal. Then, an improved physical polar diagram (PPD) construction method is designed to fuse dynamic hehavior and fault features. Finally, the rotational invariant feature layer is constructed to improve the generalization ability of the CNN model, while its hyperparameters are optimized by sparrow search algorithm. Comparative analysis and ablation studies were conducted on the lead screw and rotating machinery test bench. The experimental results of the proposed method demonstrate its superior performance on compound fault diagnosis.