In order to effectively reduce the probability of machine tool fault, based on graph theory, a method of machine tool fault diagnosis based on optimized grey correlation degree was proposed. The joint fault propagation digraph was used to defined the representation form of machine tool fault, and the grey correlation degree algorithm was introduced to determine the correlation between faults, so as to realize the accurate location of fault points. Starting from the characteristics of machine tool fault data, a complete integrated diagnosis mechanism was established, and the machine tool fault location and detection were realized. Taking FB260 NC lock machine ram feeding system as an example, the effectiveness of the proposed fault diagnosis method was verified, and the fault diagnosis and positioning results were compared with the traditional FMEA method. The results show that the consistency of node depth control of the proposed method is better.