Modeling time series based fault prediction for model-unknown nonlinear system

An optimal modeling approach is proposed and applied in unknown nonlinear system, which is based on the identification of linear time series. Firstly, using the CARMA model, the time series is translated into a linear time-varying state-space model at the sampling time. According to this control object, an optimal tracking control law is designed to compensate the errors of modeling and prediction. Furthermore, it is proved that the prediction error is bounded and robust. Apply this method to predict the fault of nonlinear system; the fault is predicted earlier than use the prediction error to do so. At same tine, the false alarm rate is reduced also. A simulation example about the fighter F-16 is also included to illustrate the method efficiency.