Innovation generation for bilinear systems: application to robust fault detection

An algorithm to design an innovation generator for bilinear systems with unknown inputs is presented. Such a generator is a filter which, in the absence of faults, generates a zero-mean white noise sequence, called innovation, from the known inputs and the outputs of the monitored system. The innovation sequence can then be processed by statistical tests to decide whether a fault has occurred or not. The choice of the test depends on the nature of the faults (additive or nonadditive faults)