A Data-Driven Monitoring Method for Stator Insulation Degradation in Induction Motors Based on ConvNeXt Network

Insulation degradation monitoring is crucial for the reliability of induction motors. Current signal-based methods are non-invasive and easy to acquire. However, the insulation features in current signals are weak, easily affected by interference, and may appear similar under different insulation conditions, leading to potential misdiagnosis and making the task challenging. To address this, this article first establishes a common-mode impedance model of the stator winding to explore the relationship between insulation degradation and current. On this basis, a data-driven new method for the insulation monitoring of induction motors (IM) is proposed. Firstly, high-pass filtering is used to remove low-frequency interference and retain the high-frequency current components containing key insulation information. Then, the proposed dual-spectrum aggregation module (DSAM) integrates the amplitude and phase spectra of the current data, and adopts the squeeze-and-excitation (SE) mechanism to enhance the model's attention to critical features. Subsequently, the enhanced signal is input into the modified ConvNeXt network for insulation condition recognition. Experimental results demonstrate that the proposed method achieves outstanding monitoring accuracy with low computational complexity under various operating conditions, providing an effective health management solution for induction motor systems.