Fault Detection And Control For Networked Systems With Incomplete Measurements

With the rapid development of network technologies, networked systems, due to their many advantages, such as low cost, reduced weight and power requirement, simple installation and maintenance, and high reliability, have been finding application in a board range of areas. However, the network inserted will lead to unavoidable incomplete measurements, such as packet dropouts, time delays, data disorder, access constraints, and signal quantization which will degrade the performance of the systems significantly. Therefore, study on networked systems with the above incomplete measurements is very important in both theory and applications.Compared with the rich results on state estimation, stability analysis and controller design, there are few results on fault detection of networked systems. Moreover, very limited papers simultaneously take some network-induced incomplete measurements into account. Therefore, based on a few newly proposed models, the dissertation investigates fault detection and control for networked systems with multiple incomplete measurements. The main work and contributions are as follows:We give a review of the previous work on networked systems, including the research topics, the modeling approaches, and the research directions. Meanwhile, some problems which need further study are pointed out.A new model is proposed which simultaneously describes access constraints and packet dropouts. In the existing literature investigating the similar problem, periodic communication sequence (PCS) and Bernoulli stochastic variable are adopted to describe access constraint and packet dropout, respectively. Considering that PCS law may not be available and the current communication status may have some relationship with the prior-known information, we use two independent Markov chains to describe the access constraint and packet dropout, respectively. In addition, different from most of the existing literature, when utilizing this new model to investigate fault detection, by introducing two performance indexes, the sensitivity of residual to fault and the robustness of residual against disturbance are simultaneously taken into consideration. A more general and in some cases more useful model is proposed to describe time delays and packet dropouts. On the one hand, this new model can be regarded as an extension of some existing models. On the other hand, it is not easy to get the completely shown probability distribution of delay taking each element in a finite set in practice, especially when the delay is quite long. In this case, it is much easier to obtain the probability distribution of delay within certain intervals.A new model describing packet dropouts, time delays, and signal quantization is proposed. Compared with some existing models, signal quantization and packet dropout compensation are more properly considered in this new model.A new model describing packet dropouts, time delays, signal quantization, and access constraint is proposed for the following reasons. Firstly, in the existing literature, access constraints and time delays are not studied simultaneously. Secondly, considering that PCS law may not be available during transmission, a prescribed switching law is adopted to describe access constraint. Thirdly, different form most existing models, at each instant, more than one channel are available for transmission, and different transmission channels have different packet dropout and time delay characteristics.Based on the above new models, by Lyapunov stability theory, some LMI-based conditions for the existence of the desired fault detection filter or controller are established. When these LMIs are feasible, the explicit expression of the corresponding parameters can also be given. Some numerical simulations which demonstrate the effectiveness of our results are given at the end of each subsection.