Abstract: Aiming at the problem that the turnout fault category can′t be accurately predicted,an ARMA-based(autoregressive moving average) turnout operation current fault curve prediction algorithm is proposed. Considering the connection between the formation of turnout current curve and turnout fault, the turnout operation current curve data is extracted from the microcomputer monitoring system, and a method for predicting turnout operation current curve based on ARMA model is established. According to actual case, the turnout operation current prediction curve is calculated, then compared and verified based on several indicators. The experimental validation results show that the method can directly predict the curve shape and then determine the fault category, demonstrating well-performed prediction results.