Test stimulus generation is a very effective tool in analog circuit troubleshooting. With the extensive usage of machine learning methods in analog circuit fault diagnosis, their good interpretability and diagnosis effectiveness have been verified. This paper proposes a test stimulus generation method for analog circuits based on the integration of four machine learning methods. It achieves frequency selection by classifying the frequency domain signals, fusion of feature importance and one-to-one correspondence between features and frequencies. To verify the effectiveness of the proposed method, it is validated using Sallen-Key bandpass filter circuit and four-op-amp biquadratic high-pass circuit. The experimental results show that the test stimulus obtained by the proposed method can effectively excite the fault features and improve the diagnosis accuracy.

Analog circuits,test stimulus generation,machine learning,fault diagnosis