Robust Fault Detection For Lipschitz Nonlinear Networked Control Systems

Networked control systems (NCSs) have many advantages, for example,flexible architectures, lower cost of installation, easy expansibility and maintenance, etc.Network-based analysis and design have many industrial applications in the areas such asindustrial automation, robots, and aircraft. However, the complexity of a network resultedfrom delays and packet dropouts that inevitably degrade the performance and could be asource of instability. A large amout of research achievements have been made on controlof NCSs, but there are few on the study of fault detection, and of nonlinear systems. Thisthesis fouses on robust fault detection for Lipschitz nonlinear networked control systems.Firstly, this paper considers the fault detection problem of Lipschitz nonlinearnetworked control systems with random packet loss induced by network congestion,actuator fault and external interference. Two kinds of random delays in networkcommunication channels are simultaneously considered: the channel from sensor tocontroller and the channel from controller to actuator. The random packet loss is modeledas some linear functions of stochastic variables satisfying Bernoulli random binarydistribution. In terms of linear matrix inequality (LMI), a sufficient condition is derived,which makes the error systems be exponentially mean-square stable and achieveprescribed H∞disturbance-rejection-attenuation level. In addition, the scheme of faultdetection is obtained by using the mean squares of the residual.Secondly, this paper is concerned with the problem of fault detection for Lipschitznonlinear networked control systems with random communication delays induced byexternal interference. An observer-based fault detection filter is then designed to detectfault. The control input of the observer and that of the plant are different because of thecommunication delay from controller to actuator in the network link. The observer-basedfault detection filter can effectively detect fault, and make the system be mean-squareexponentialiy stable and satisfy the robust H∞performance index.Finally, the problem of robust fault detection for Lipschitz nonlinear networkedcontrol systems with multiple sensors and different packet-loss rates is studied. It issupposed that in the communication channels from the multiple sensors to the controller have individual random data missing probability. The random packets loss is assumed toobey the Bernoulli random binary distribution, and then the model is established. In termsof linear matrix inequality (LMI), a sufficient condition of the existence of fault detectionfilter is given. The fault detection filter is not only sensitive to fault, but also ensures goodrobustness under external disturbance. A simulation example illustrates the feasibility ofthe proposed approach.