The invention discloses a switched circuit fault diagnosis method based on wavelet transformation and ICA (independent component analysis) feature extraction. The switched circuit fault diagnosis method includes the steps of firstly, performing classifier training and fault dictionary construction, namely, based on circuit simulation, acquiring feature parameters by a method based on wavelet transformation and ICA feature extraction, and constructing a fault dictionary and a training classifier based on the feature parameters; secondly, performing fault diagnosis, namely, acquiring the feature parameters aiming at a switched current circuit to be diagnosed by referring to the fault dictionary and by the method based on wavelet transformation and ICA feature extraction, inputting the feature parameters into a trained classifier, and subjecting the switched current circuit to be diagnosed to fault diagnosis to obtain output signals of the classifier, namely, fault diagnosis results. The switched circuit fault diagnosis method based on wavelet transformation and ICA feature extraction has the advantages of ingenious concept, easiness in implementation, simulation proving and capability of distinguishing various fault types more accurately as compared with an existing method.