Network-Based Fault Detection Of Stochastic Systems

The controlled system or process in actual application is often influenced by random phenomenon,which increases the complexity of the systems and brings a great challenge to the system analysis and synthesis simultaneously.On the other hand,with the rapid development of computer technology and network communication technology,networked control systems are widely applied in practical industrial systems.Networked control systems mainly use a shared communication network to connect the controlled objects,sensors,controllers and actuators.It has many advantages such as low cost,high stability and easy installation and maintenance.However,the communication resources of actual network systems are limited,which increases the difficulty of system analysis and design.For instance,the network-induced delay and network packet loss problems.Therefore,it is of great significance to develop fault detection methods for stochastic systems under network environment.This thesis takes several different network stochastic systems as the research objects and studies the fault detection problem with network-induced delay,network packet loss and communication constraints.The main research contents are summarized as follows:The fault detection filter design problem for a class of linear stochastic systems with communication constraints is investigated.The event-triggered scheme is applied to reduce data transmissions between the communication networks so that the limited network communications resources can be saved.In order to detect faults,the event-triggered fault detection filter is designed to build the fault detection model.For the continuous-time network stochastic system,the influence of fault detection model on external disturbances and faults is minimized to achieve the fault detection objective by means of H? optimal performance methods.For the discrete-time network stochastic system,the fault detection model is obtained by H-/H? optimal performance methods,which is both sensitive to the faults and robust to the external disturbances.Taking both the stochastic model and the event-triggered data transmission scheme into account,the sufficient conditions guaranteeing the stability of fault detection system are derived according to the stochastic theory and the Lyapunov stability theory.Then the parameters of the event-triggered fault detection filter are obtained by solving a set of linear matrix inequalities.This thesis discusses the fault detection filter design for a class of uncertain network stochastic systems with network packet loss and communication constraints.The eventtriggered scheme is applied to save the limited network communication resources.The white noise sequence satisfying Bernoulli distribution is used to describe the phenomenon of random packet loss occurring in the network channel;the uncertainties of norm boundedness are used to describe the parameter uncertainty of the stochastic system model.Take the above mentions into consideration,an event-triggered fault detection filter is designed to construct the fault detection model.Moreover,the sufficient conditions for the fault detection system satisfying the mean-square robust exponential stability and the desired fault detection objective are given according to the stochastic theory,Lyapunov stability theory and robust control theory,which further derives the event-triggered fault detection filter parameters.This thesis presents the simultaneous fault detection and controller design for a class of uncertain discrete-time network stochastic systems with communication constraints.An integrated module including a filter and output feedback controller is designed to construct the fault detection and control model,which guarantees the stochastic stability of the model,the desired control performance and the fault detection objective.Two mutually independent event triggers are designed to reduce the transmission of data from the sensor to the filter/controller module and the filter/controller module to the original system channel,respectively.Taking both the uncertain stochastic model and the event-triggered data transmission scheme into account,an optimized algorithm is proposed to obtain the parameters of the designed fault detection filter,controller and two event triggers.This thesis presents a fault detection filter design problem for a class of nonlinear network stochastic systems with network packet loss,data transmission delay and communication constraints.The white noise sequence satisfying Bernoulli distribution is used to describe the phenomenon of random packet loss occurring on the network channel.Aiming at the phenomenon that the traditional event-triggered scheme with fixed threshold can not fully save the limited network resources,an adaptive event-triggered scheme is proposed,which can adjust the event-triggered threshold dynamically.The first type is a nonlinear network stochastic system with random nonlinear terms.A quadratic constraint is used to describe the characteristics of the nonlinear term.The white noise sequence satisfying Bernoulli distribution is used to describe the occurrence probability of a nonlinear term.The second type is a more general nonlinear network stochastic system using the Takagi-Sugeno(T-S)fuzzy modeling method.In order to detect faults,the adaptive event-triggered filters are designed to construct the fault detection models.Then,the performance analysis and filter parameter design are performed for two types of fault detection models respectively.Sufficient conditions guaranteeing for the fault detection systems satisfying the mean square asymptotic stability and the desired fault detection objective are given,which further derives the adaptive event-triggered fault detection filter parameters.