Fault detection via optimally robust detection filters

An approach is presented for using optimally robust detection filters to generate analytic redundancy. By introducing an appropriate criterion the design of the filter is formulated as an optimization problem. Its solution shows that the optimally robust detection filter consists of a bandpass filter and a linear system which is obtained by solving a general eigenvalue problem. The algorithm for designing this filter is therefore computationally simple and systematic. Investigations on the physical core of the fault detection procedure are carried out. It is shown that the quality of the fault detection is dependent on the coupling strength of the faults with respect to the system output. It is demonstrated that the optimal detection filter presented is a generalization of the detection filter of R.V. Beard (1971) and H.L. Jones (1973)