Interval Model-based Fault Detection for Nonlinear Dynamic System with Parameters Uncertainty

In view of the problems of traditional fault detection methods used for parameter uncertainties of non-linear dynamic systems, in which is incapable of obtaining adaptive threshold and improving the generalization performance of model structure, an adaptive fault detection method based on the optimal interval model is proposed for nonlinear dynamic system with parameters uncertainty in this paper. The proposed method combines the idea of infinite norm of approximation error, structural risk minimization theory and linear programming to ensure the optimality of the interval model used for fault detection. When the actual output caused by parameters uncertainty of nonlinear dynamic system exceeds the boundary of optimal interval model in absence of fault, It is considered that fault of the detected plant is judged. Finally, the experimental results from parameter uncertainties of linear and nonlinear dynamic control systems show that the optimality of the interval model and the availability of fault detection.