Identification of nonlinear processes and model based fault isolation using local linear models

Deals with identification of nonlinear processes and model-based fault detection/isolation (FDI). The applicability of the proposed methods is illustrated on a three-tank laboratory setup. The process identification is based on the local linear model tree (LOLIMOT) algorithm and leads to local linear models. The parameters of the local models are used for generation of structured residual equations, similar to the well-known parity space approach. This enables detection and isolation of five different sensor faults of the three-tank process, continously over all ranges of operation