Research On Distributed Fault Estimation Problems Over Sensor Networks

In recent years, theoretical and practical analysis on large-scale networked systems has become a research hotspot gradually. including the disciplines such as control engineering, signal processing, computer science and mathematics. As a core part of the research area, the distributed fault detection and estimation problems over sensor networks have attracted more and more researchers’ attention. The sensor network is a kind of computer network, which is composed of a lot of automatic devices distributed in space, and these devices use sensors to monitor physical or environmental conditions in different locations cooperatively. In practical network environments, a series of network induced phenomena may occur easily in the control systems, including time delays, channel fadings, nonlinearities, randomly occurring uncertainties, and so on. Thus, it has far-reaching theoretical and practical significance of researching on distributed fault estimation problems with taking full account of these networked phenomena, which also has many challenges. In this paper, we discuss the design problems of distributed fault estimators for several classes of systems over sensor network, and the work done in this paper is as follows:Firstly, in view of the rarely existing research results of randomly occurring uncertainty, the distributed fault estimator is designed for the discrete time systems with the phenomenon of randomly occurring uncertainties. By the Lyapunov stability theorem, sufficient conditions that guarantee the existence of the distributed fault estimator are obtained, which make the distributed fault estimation dynamic system be exponentially mean square stable and satisfy the given average H? performance constraints, and the expression of the estimator is obtained as well. Then making simulation experiment via the Matlab software, and whether the designed fault estimator satisfies the design requirements is verified.Secondly, taking into account the fact that the sensor networks are often deployed in the complex environment, the model of randomly occurring nonlinearity is utilized to depict the network induced complexity, the distributed fault estimator for a class of time-varying systems with randomly occurring nonlinearity is designed. By the 2L gain theory, sufficient conditions that let the fault estimation dynamic system satisfy the given average H? performance constraints are obtained, and a recursive distributed fault estimator design algorithm is proposed. Through the simulation experiments, the effectiveness of distributed fault estimator design algorithm is verified.Thirdly, for the problem that the fault estimator parameters cannot be accurately executed in the actual situation, the non-fragile distributed fault estimator design method is researched for time-varying systems with randomly occurring nonlinearity and randomly occurring gain variations. Among the existing fault estimation research methods, most of them assume that the estimator parameters can be executed accurately, and the drifting phenomenon of estimator parameter matrices and the resulted problems of stability decline or even instability of fault estimator have been neglected.In this paper, random variables and uncertainty in the norm-bounded multiplicative form are utilized to describe the random occurrence of the estimator gain variations, sufficient conditions that make fault estimator dynamic system satisfy the average H? performance index are acquired and a recursive non-fragile distributed fault estimator design algorithm is put forward. Then make the numerical simulation to indicate the accuracy of the proposed non-fragile distributed fault estimator design algorithm.