Research On Non-fragile State Estimation For Non-linear System

In this paper,a corresponding non-fragile state estimator is designed for the discrete-time nonlinear neural network system to ensure the validity and accuracy of the state estimation.The main work is as follows:(1)At present,most of the state estimation research results are obtained based on the known and accurate parameters of the estimator.However,in complex real environments,the observer is susceptible to various environmental influences and can not maintain itself in an ideal state.Based on this kind of situation,a non-fragile state estimation method is studied for a typical class of nonlinear discrete-time neural networks with sensor saturation.Using Lyapunov-Krasovskii function and matrix analysis techniques,the sufficient conditions for delay-dependent correlation are obtained.By solving linear matrix inequalities(LMIs),the gain matrix of the estimator can be obtained.Finally,numerical simulations show that the proposed method has better state estimation accuracy for discrete-time neural networks with sensor saturation.(2)In classical control theory,the system is considered to be asymptotically stable if the system is Lyapunov stable for a finite period of time.However,in real industrial processes,it is more valuable to study the stability under time constraints in the real state estimation problem.Therefore,for nonlinear neural networks,it is necessary to consider the impact of finite time on the system.In this paper,a finite time non-fragile state estimation method is studied for discrete neural networks with sensor failure and random non-linear sensor disturbances.Considering the influence of time-varying delay,finite time and sensor constraints,a finite time non-fragile state estimator is designed by using Lyapunov-Krasovskii function and augmented matrix.Finally,the numerical simulation results show that the proposed non-fragile state estimator has a good observational effect.(3)In practical networked control system,in order to save limited network bandwidth resources,an event triggering mechanism is introduced to study the non-fragile state estimation problem of a class of Markov discrete-time neural networks with communication delay and nonlinear disturbance.Among them,when the system sampling data to meet certain event trigger criteria,the sensor data is transmitted.In this paper,we propose a non-fragile state estimator based on event-triggered mechanism that can react to stochastic parameters such as Markov jump,communication delay,nonlinear interference and sensor failure.Finally,numerical simulations show that the proposed non-fragile state estimator can achieve desired state estimation.