Event-triggered State Estimation And Fault Diagnosis For Stochastic Systems Innetwork Environment

Dynamic systems in network environment have been brought into being because of the growth of the sensor technology,embedded computing and wireless communication,which applied to the remote monitoring,operation and diagnosis in practical industry.However,the resource-constrained network environment has become the bottleneck re-stricting its development.At the same time,the event-triggered transmission mechanism has been paid widespread attention to,especially to save the limited communication bandwidth and the battery capacity of node.Regarding above as the motivation,this thesis studies the state estimation and fault diagnosis in no-ideal network environment.Based on many problems to be solved in that environment and the close attention to the event-triggered transmission mechanism,the thesis delves deeper on event-triggered state estimation and fault diagnosis for a class of discrete-time stochastic systems in no ideal network environment although the study of the event-triggered transmission mechanism gains some success.The thesis studies the following issues in particular.(1)The problem of state estimation is studied for discrete-time stochastic systems under the event-triggered transmission framework.First of all,an event-triggered state estimator is designed for stochastic systems with unknown inputs using unbiased and minimum variance indices.The corresponding scheme threshold is computed via an ap-proximate quadratic performance function.Subsequently,the stability of event-triggered steady Kalman filtering is analyzed for discrete-time stochastic systems subject to inter-mittent observations.In addition,an experimental platform of two coupled water tanks with a wireless sensor node is established to evaluate and verify the proposed transmission scheme.(2)We focus on how to achieve the balance between estimation performance and bat-tery energy consumption by the designed event-triggered data-forwarding scheme.Specif-ically,our study considers the measurements cannot be sent directly from sensor node to remote estimator side.A series of relay nodes are used to forward data packets to remote estimator by multi-hop links.An output estimation error-based data-forwarding scheme and an error covariance-based data-forwarding scheme are presented based on remote estimation of a linear stochastic process.Two types of forwarding schemes are consid-ered such that each relay node can perform:forwarding the measurements to the remote estimator;forwarding the estimated values to the remote estimator.The correspond-ing scheme thresholds are derived via minimizing upper bound of approximate quadratic performance function.Finally,the effectiveness of the proposed schemes is extensively evaluated using two coupled water tanks.(3)The problem of event-triggered fault detection and fault estimation is investigat-ed for discrete-time stochastic systems.Firstly,an event-triggered fault detector for a class of stochastic systems subject to sensor nonlinearities,deception attacks and exter-nal disturbances.A random variable obeying the Bernoulli distribution is employed to characterize the phenomena of the randomly occurring sensor nonlinearities.H_-/H_?per-formance indices are introduced to derive estimator gains and residual weighting matrix so that the fault-detection residual has the best sensitivity to faults and the best robust-ness to unwanted factors including additive disturbances and false information injected by attacker.Secondly,a coordinate transformation approach is exploited to achieve the purpose that the fault-detection residual is only sensitive to system faults while robust to additive disturbances.This approach can transform the considered system into two subsystems,and the disturbances are removed from one of the subsystems.The optimal gain for each filter is derived,which applies to the optimization criteria of unbiasedness and minimum mean-square estimation error.An event-triggered sensor data transmission scheme is provided via deriving an upper bound of the approximate quadratic performance function.In particular,a new fault-alarming strategy is presented so as to eliminate the impact of the event-triggered transmission scheme for achieving the sensitivity and real-time capacity on fault alarming.Thirdly,a synthesized event-triggered design of fault detector and fault estimator in discrete-time stochastic systems is proposed against the phenomena of both external disturbances and randomly occurring deception attacks.The upper bound of estimation error covariance is obtained in form of Riccati-like difference equations by utilizing the mathematical induction approach.The explicit form of estima-tor gain is obtained to minimize this upper bound through a recursive algorithm.Finally,the considered event-triggered fault estimator is extended to the case of multi-hop relay networks.The problem of fault estimator is studied under the error covariance-based data-forwarding scheme for stochastic systems subject to randomly occurring packet dropouts.An application example of one-dimensional target tracking is presented and the benefits of the obtained theoretical results are demonstrated by comparative experiments.(4)The problem of event-triggered fault tolerant control is explored for state-saturated stochastic systems with deception attacks.We first study the problem of passive fault tolerant control:a dynamic output feedback controller is designed for state-saturated s-tochastic systems subject to deception attacks and randomly occurring actuator faults.A sufficient condition is established for the existence of controller using Lyapunov func-tional,where the corresponding gain is calculated via singular value decomposition and linear matrix inequalities techniques.Then,the problem of active fault tolerant control is considered:a fault compensator is designed using the estimated states and faults for par-tial state-saturated stochastic systems subject to randomly occurring deception attacks.The estimator and controller gains are independently derived based on the principle of separation property and system decomposition.Finally,an application example of DC motor is given and the benefits of the obtained theoretical results are demonstrated by comparative experiments.(5)An event-triggered estimator-based lithium-ion battery string health management system is designed according to the proposed theoretical results of event-triggered fault detection and fault estimator.In particular,implementation issues of the theoretical results are discussed.A new data-forwarding communication protocol that could be applied to our addressed topology is designed;this involves hardware design and the corresponding procedure implementation.The introduction of this system is divided into the following aspects.On one hand,the equivalent battery circuit model is presented using Kirchhoff laws.On the other hand,the state space model about three ANR26650 batteries is constructed via the considered equivalent battery circuit model.The benefits of the designed health management system are evaluated including the estimation of terminal voltages,fault detection and reconstruction of various types of faults.Furthermore,our health management system is extended to the situation of multi-hop networks.The effectiveness of the estimation of terminal voltages and reconstruction of various types of faults is evaluated in the case of packet dropouts.