The bolt at the bottom of multiple unit train plays a key role in the overall braking system of the train. The bolt missing will bring serious challenges for the train safety braking and safety running. By the example of fault inspection of bolt missing, an online inspection and recognition algorithm is proposed for the fault of components missing in a train, which provides a guidance for the targeted inspection on the key parts of the train. Due to the characteristics of the bolts geometry, a complete local binary patterns feature extraction algorithm based on the image Sobel gradient edge is proposed, and the training and learning of a binary classifier is combined to complete the automatic fault inspection of bolt missing. The results show that the proposed algorithm has strong robustness to inspect the fault in complex scenes with high inspection efficiency and precision, which can meet the demand of the site application.