Abstract: A systematic fault diagnosis method for tolerence circuits based on the combination of wavelet Transform(WT) and principal component analysis(PCA) is presented in the paper. The WT is used as a de-noise and feature extraction tool. The PCA is used to compress the feature data, and results in less feature vector dimension and simplifier classifiers. Practical diagnosis example is described, and results show that diagnosing and locating faults are quick and exact.