Aiming at the problem of classification difficulty caused by the high dimensionality of fault dataset, a fault dataset dimension reduction algorithm local centroid mean minimum-distance discriminant projection (LCMMDP) was proposed. The algorithm can maintain the local geometric structure information of the sample while considering the cohesion and separation of the sample, reflecting the close relationship between the sample and the local centroid mean. The hybrid characteristics of rotor vibration signals were extracted from multiple angles, the original high-dimensional feature sets were constructed, and low-dimensional sensitive feature subsets were extracted by LCMMDP. The improved K-nearest centroid neighbor classification based on local mean and class mean was used (KNCNCM) for fault pattern recognition. The proposed method integrated the advantages of LCMMDP in dimension reduction and KNCNCM in pattern recognition and obtained higher fault identification accuracy. The validity of the proposed method is verified by the instance of the fault diagnosis of a double span rotor system data-set and simulation data-set. This research has important significance for the rapid processing and effective utilization of big data in the rotor system.