The mechanical fault diagnosis algorithm of DC circuit breaker is the core part of on-line monitoring system of mechanical status of DC circuit breaker. In this paper

the mechanical fault simulation experiment of DC circuit breaker is performed. The coil current and vibration signals under different faults are collected. After extraction of its feature

the current feature

vibration short-time energy feature and wavelet packet frequency band energy feature are permuted and combined and the fault diagnosis model is constructed by using the support vector machined. In this paper

the principal component analysis (PCA) method and Relief-F algorithm are used for dimensionality reduction of different feature combinations to analyze further the diagnosis effect after dimensionality reduction of feature combinations. Moreover

the K-Fold cross validation algorithm is used to assesses the diagnosis model which is output by the single feature and feature combination training output so to select the diagnosis model with the optimal classification performance.