Data-driven Adaptive Filtering for Fault Diagnosis of A Class of Nonlinear Discrete-time Systems

Data-driven adaptive filtering technique has been widely concerned due to its extensively used in industry control system. In order to handle actuator fault diagnosis problem with a class of nonlinear discrete-time systems, this paper investigates a novel data driven adaptive filtering based on dynamic linearization technique in the framework of Kalman Filter and Recursive Least Square(RLS) algorithm. Compared with the conventional fault diagnosis algorithm, by modeling actuator faults changes, the proposed data-driven adaptive filtering for fault diagnosis (DDAF-FD) method is performed through joint state-parameter estimation and designed by only using measurement I/O data without precise mathematical model or linearization approximation. Numerical examples are presented to verify the effectiveness of the proposed algorithm.