Degradation prediction method by use of autoregressive algorithm

Forecasting for nonlinear time series is important in many practical applications. This paper proposes a degradation prediction method by use of revised autoregressive (AR) model. The autoregressive (AR) model in this paper is applied to the analysis of gas enginepsilas condition based on temperature signals of engine. In the application of revised AR model, the data preconditioning is important and directly influences the prediction results. Besides, Final Prediction Error criterion is used to estimate the order of AR model. For the simulation, this AR model takes into account Yule-Walker equations, least-squares algorithm and some other algorithms to estimate the AR coefficients. Finally, the prediction mainly uses two algorithms, one is the prediction based on revised residuals by previous simulation, and the other one named dynamic simulation mainly consider the AR coefficients revising. Both of the results of two prediction algorithms are effective. Also, the model can supply some indications besides the simulation and prediction. Anyway, simulation results demonstrate the effectiveness of the AR model for the diagnosis of the condition of gas engine, and prediction results are mainly used for diagnosing gas engines and can be supplied as a reference for the maintenance.