Fault Diagnosis For Systems Over Sensor Networks With Communication Protocol

A sensor network is composed of a large number of sensor nodes with limited computing and communication capabilities,which are distributed over a spatial region to monitor different kinds of information.In practical applications,sensor networks are generally deployed in harsh environments,which are prone to failure.It reduces the system performance and may causes serious consequences such as casualties,economic losses.Fault diagnosis technology can detect,isolate and identify faults to improve reliability of systems over sensor networks.In addition,due to the limitation of network bandwidth,it is easy to cause data conflict when the nodes exchange information with their neighbors.The emergence of communication protocol can solve the problem of data conflict.Summarizing the above discussions,the fault diagnosis problems are studied for several kinds of systems over sensor networks in this thesis.The main content of this thesis can be divided into the following three parts:In the first part,the problem of sensor node fault diagnosis is studied for systems over sensor networks with uncertainty and nonlinearity.In order to reduce the risk of data conflict at the same time,stochastic protocol is introduced to schedule the transmission order of nodes.A group of distributed fault estimators are designed by Kalman filter algorithm,then the parameters of the estimators are obtained by solving two recursive equations.On this basis,the fault diagnosis logic is designed to judge whether the fault occurs,the location of the fault and the size of the fault.Finally,a numerical example is given to illustrate the effectiveness of the proposed design scheme.In the second part,the problem of fault estimation is studied for linear time-varying systems over sensor networks under Round-Robin protocol.Different from the previous part,we consider energy bounded noise,multiplicative noise and missing measurements in this system.Missing measurements are characterized by a set of random variables.Firstly,a set of distributed fault estimators based on observer are designed.Then,stochastic analysis and linear matrix inequality(LMI)techniques are used to obtain the sufficient conditions which satisfy the constraints.Finally,an example is provided to verify the feasibility and effectiveness of the proposed scheme.In the third part,the problem of fault estimation is studied for a class of time-varying nonlinear systems over sensor networks under stochastic protocol.In order to avoid data conflict and reduce the transmission pressure,stochastic protocol is used to control the signal transmission order of sensor nodes.Taylor series expansion method is used to deal with nonlinearity.The high order term of linearization error is compensated which is discarded in traditional extended Kalman filter.Different from the former two parts,the estimator designed for each sensor node does not need to be augmented.The gain of the estimator can be calculated separately,which reduces the amount of calculation.In addition,a sufficient condition for the estimation error to be bounded is derived on stochastic analysis theory.Finally,an example is given to illustrate the feasibility and effectiveness of the proposed fault estimation scheme.