The existing particle filter fault prediction methods give the predictive value of the corresponding time by the particle filter algorithm, and then compare the distance between forecasting sequence and observation sequence to predict the fault. However, this fault prediction method can not handle the condition that the length of forecasting sequence is different from that of observation sequence. Dynamic Time Warping is a pattern matching algorithm based on nonlinear, which is suitable for the time sequence of different lengths. This paper is from the new perspective of using the Dynamic Time Warping algorithm to measure the similarity between normal working equipment's time sequence and abnormal sequence caused by potential faults, and design the system normal degree and abnormal degree to distinguish whether the device is operating properly or not, thus predict potential faults. Experimental results demonstrate the feasibility of this method, which can predict the system faults timely and accurately.