Data-driven modeling and fault estimation for nonlinear systems: A checksum-redundancy correction based anti-attack strategy

In a network attack environment, this paper investigates the data-driven modeling and fault estimation for a class of Lipschitz nonlinear systems with actuator and sensor faults. Here, the sampled measured output data from the system transmitted over the wireless network, which may pose a risk of being subjected to network attacks, can be captured by the receiving end. A checksum-redundancy correction based anti-attack strategy, which can restore the transmitted data to the original one, is proposed to ensure data reliability. A reliable data-driven differential equation model based observer associated with the original system is constructed, so as to achieve the estimation of system multi faults. LMI-based uniformly ultimate boundedness (UUB) stability conditions are proposed for the error dynamics. Then, an iterative observer related to the original system is designed to further improve the multi-fault estimation accuracy. A convergence theorem is given to guarantee the effectiveness of the designed iterative observer. Finally, a longitudinal dynamics model of the aircraft is exhibited to test the proposed strategy.