In view of the fault complex and diagnosis difficult problem of present numerical control (NC) machine tool, the fault diagnosis method based on artificial neural network was put forward. In the study of traditional fault diagnosis model based on BP neural network, the improved BP algorithmLM algorithm was introduced. The LMBP neural network fault diagnosis model of spindle system of machine tool was established, to carry out its fault analysis and diagnosis. Then the Matlab simulation was compared with the traditional BP neural network. The simulation results show that the traditional BP neural network is existed of more difficult to achieve rapid and accurate fault location problem, while LM algorithm based on BP neural network as the core of the fault diagnosis algorithm has fast convergence speed and recognition accuracy. The scheme design is reasonable and feasible, has good application prospect, and the application example is given.