Intelligent Analog Circuit Soft Fault Diagnosis Based on Multi-level SWT and EM-PCA

Analog circuit soft fault diagnosis is an important method to ensure the continuous and stable operation of electronic systems, which directly affects the maintenance cost of the system. To improve the accuracy of soft fault diagnosis in analog circuits, this paper proposes a new diagnostic method. The method utilizes stationary wavelet transform to decompose the output response of analog circuits, obtaining their approximate coefficients and detail coefficients. Expectation maximization principal component analysis is used to reduce the dimensionality of detail coefficients, extracting essential information from high-frequency details, and simplifying data complexity. Then, the approximation coefficients are decomposed, and the aforementioned process is repeated. Finally, the deepest approximation coefficient is combined with all high-frequency information and inputted into the light gradient boosting machine for feature extraction and fault classification. This paper uses this method to test and analyze the soft fault diagnosis of the Sallen-Key band-pass filter circuit, Four-op-amp biquad high-pass filter circuit, and Leap-frog low-pass filter circuit. The experimental results show that the method proposed in this paper has high diagnostic accuracy and good noise interference tolerance. When the deviation between the component value and the nominal value is 30%, both the Sallen-Key band-pass filter circuit and the Four-op-amp biquad high-pass filter circuit can achieve a 100% diagnostic rate. Moreover, the diagnostic accuracy of the Leap-frog low-pass filtering circuit exceeds 98.89%.