Nonlinear Fault Diagnosis based on RBF with Sliding Window Error Feedback

Nonlinear fault diagnosis is one of the difficulties in fault diagnosis field. The paper presents the nonlinear fault estimator based on RBF with sliding window error feedback for a class of nonlinear system. The input of estimator is input and output of the system, and the output is the fault estimate. The neural network weight adjusting algorithm adopts sliding window error feedback, which enforces the amount of fault information and speed up the convergence. The paper analyses the robustness of algorithm and the window length influence upon fault estimate, gives the variable window length strategy, and qualitatively presents a method of choosing window length. The simulation results prove that the method improves greatly the response speed and accuracy in fault diagnosis under the circumstances of choosing the proper window length.