The Intermittent Fault Diagnosis of Analog Circuits Based on EMD-CrossFormer

In order to improve the stability of analog circuits, intermittent faults caused by suboptimal component soldering and performance degradation, which are prevalent in analog circuits, need to be detected in a timely manner. This paper proposes an intermittent fault diagnosis method for analog circuits based on empirical modal decomposition (EMD) and CrossFormer. The method addresses the intermittent fault characteristics of analog circuits, information entropy is initially employed as a fault-triggered detection mechanism to analyze only voltage data segments that contain intermittent faults, thereby reducing the amount of stored data. Proposed method then decomposes the corresponding intermittent fault signal interval into multiple intrinsic mode functions (IMFs) using EMD, and inputs the resulting multiple IMFs as multivariate time series into CrossFormer for fault diagnosis. This approach is able to capture the correlation of different variable sequences in the multivariate time series well. The simulation was validated by the four-opamp biquad high-pass filter circuit, and the results demonstrated that the method achieved a superior diagnostic effect with high diagnostic accuracy.