Abstract: Aiming at the problem of large prediction errors caused by the selection of background values in the traditional gray prediction model, a fault prediction model combined with the improved gray wolf algorithm is proposed. An improved gray wolf algorithm is designed to improve the algorithm parameters of the basic gray wolf algorithm and used to optimize the background value in the gray prediction model to obtain the optimal prediction model. Taking 8 fault data of CNC lathe spindle as an example, the prediction accuracy of the proposed method is compared with other gray models to verify the fitness and stability of the method.