The invention relates to a fuzzy-preference-relation-based multi-vibration-signal fusion method. The method comprises: (a), obtaining a filtering signal; (b), calculating a fuzzy preference relation; (c), calculating an integrated fuzzy preference relation; (d), calculating a weight value; (e), calculating a fusion signal value; (f), calculating a fusion signal value within an overall collection period; and (g), obtaining an ideal signal. According to the method, in order to improve accuracy of a fault diagnosis by using information of multiple sensors comprehensively, a weight fusion algorithm is studied and a fuzzy-preference-relation-based multi-vibration-signal fusion algorithm is put forward; and irrelevant information is deleted and effective fusion of data is realized. Therefore, accurate information is provided for feature extraction and identification of a fault diagnosis.