Abstract:

In this paper, the fault detection problem of systems with missing measurements is discussed. The background of the study is the increasing application of networks as data communication medium in complex large-scale systems. First, in order to cope with missing measurement, the structure of standard model based residual generator is modified and dynamic network resource allocation is suggested. The dynamics of the residual generator were characterized by a discrete-time Markovian jump linear system. Then a residual evaluation scheme is developed to reduce false alarm rate caused by missing measurement. Further, a co-design approach of time-variant residual generator and threshold is proposed to improve the dynamics and the sensitivity of the fault detection system to the faults.