Abstract:Abstract: Prediction to the noisy running sound of internal combustion engine is of great help to decrease the interference of the running sound, thus can greatly improve the accuracy of engine fault diagnosis based on engine’s fault-featured running sound. By exploiting Volterra adaptive filter, two engine running sound prediction schemes are proposed. The proposed methods take the advantage of the nonlinear function mapping power of Volterra series effectively, and fully take the priori knowledge of engine running sound into accounts, such as inherent cyclic correlation, periodicity and phase information. The taken experiments show that, the proposed methods are robust to a range of fluctuation of engine rotating speed; the prediction accuracy can meet engineering application requirements.