A novel approach for analog circuit incipient fault diagnosis by using KECA as a preprocessor

In order to diagnose incipient fault of analog circuits effectively, an analog circuit incipient fault approach by using kernel entropy component analysis (KECA) as a preprocessor is proposed in the paper. Time responses are acquired by sampling outputs of the circuits under test. Raw features with high dimension are generated by wavelet transform. Furthermore, lower dimensional features are produced through KECA as samples which are used to construct a classification model based on least squares support vector machine. Bandpass filter and leapfrog filter incipient fault diagnosis simulations demonstrate the diagnose procedure of the proposed approach, and also validate proposed approach by using KECA as a preprocessor can produce higher diagnosis accuracy than the commonly used methods. Key-Words: Analog circuits, Incipient fault diagnosis, Wavelet transform, KECA, Least squares support vector machine