Research On Event-triggered Fault Detection Of Discrete System Based On Neural Network

Nowadays,the scale of control systems has become increasingly large and complex,the probability of system faults has also increased significantly.Therefore,the reliability of the system has become a research hotspot.In order to improve the reliability of the system,it is necessary to detect faults in time.In the framework of model-based fault detection,this thesis proposes a neural network-based event-triggered fault detection strategy for discrete-time systems with unknown nonlinear functions,unmodeled dynamics and disturbances.The main research contents are as follows:Firstly,a neural network-based fault detection observer is constructed by using the ability of neural network to approximate unknown nonlinear functions.An event-triggered mechanism is designed by considering the communication burden and computational complexity.The current output measurement transmission and the execution of the neural network weight update are determined by the event-triggered condition.The current transmission instruction is executed and the weight of the neural network is updated,when the ratio of the event-triggered error to the system output exceeds the preset threshold.Secondly,the weight update law of aperiodic neural network is designed.A novel Lyapunov function is constructed and the weight update law of the aperiodic neural network is designed.The stability of the state estimation error dynamic system and the convergence of the neural network weight estimation error are proved in the event-triggered time and trigger time interval under two different situations.It is guaranteed that the state estimation error,residual error and neural network weight estimation error are ultimately bounded in fault-free case.Again,on the basis of fault detection and analysis,the fault detection decision logic is proposed and the fault sensitivity is characterized by analyzing the set of detectable faults.The proposed scheme reduces the communication burden and computational complexity while ensuring the performance of fault detection.Finally,a model of the double-link flexible manipulator system is established by considering the actuator's offset fault.The effectiveness of the proposed method is verified through simulation experiments.