As fault prediction has the characteristics of uncertainty, we design a method for fault prediction, which combines fuzzy mathematics membership function with particle filter algorithm to predict fault. The new method uses particle filter algorithm to calculate the future state of the device operation, and then designs the normal membership function and the abnormal membership function of the device operation state, calculates and compares the value of the normal and abnormal membership function by using the calculated results and based on the comparison result to predict potential failure. The feasibility of the proposed method is verified by experiments, which can predict the failure of the system in time.