Abstract:

This paper deals with problems related to designing fault detection systems for linear uncertain systems, where model uncertainties are characterised by stochastic changes in the model parameters. First, statistical features of the residual signals generated by means of an observer are studied. Based on the achieved results, a residual evaluation strategy is then developed, which combines the norm-based and statistic testing schemes. The major mathematical tool used in our study is the well-established LMI-technique.