Detection and isolation of sensor faults on nonlinear processes based on local linear models

The development of a reliable fault detection and isolation (FDI) scheme for nonlinear processes is often time consuming and difficult to achieve due to the complexity of the system. Neural networks and fuzzy models, able to approximate nonlinear dynamic functions offer a powerful tool to cope with this problem. In this paper, a new approach for FDI of sensor faults on nonlinear processes is introduced, based on local linear models of the process. The parameters of this model are used for generation of structured residuals, similar to the parity space approach. The practical applicability is illustrated on an industrial scale thermal plant. Here, four different sensor faults can be detected and isolated continuously over all ranges of operation.