The present invention discloses an analog circuit fault mode classification method comprising the following implementation steps: (1) acquiring M groups of voltage signal sample vectors as shown in the specification of each fault mode as shown in the specification of an analog circuit by using a data acquisition board; (2) sequentially extracting fault feature vectors as shown in the specification of the voltage signal sample vectors as shown in the specification by using subspace projection; (3) performing standardized treatment on the extracted fault feature vectors as shown in the specification to obtain standardized fault feature vectors as shown in the specification; (4) constructing a supporting vector machine based fault mode classifier, inputting the standardized fault feature vectors as shown in the specification, performing learning and training on the classifier, and determining structure parameters of the classifier; and (5) finishing determination of fault modes according to fault mode determination rules. The fault mode classifier is simple in learning and training and reliable in mode classification accuracy.