Knowledge- and data-based models for fault diagnosis

This paper is intended to give a survey on the state of the art of knowledge-based model-building for fault diagnosis. Emphasis is placed upon the generation of fault-reflecting signals, the so-called residuals, by using process models based on fuzzy logic or neural networks. The particularities of models needed for fault detection and isolation (FDI) are shown and the differences with respect to the models used in control are pointed out. In contrast to the wide-spread opinion that models for FDI have always to be more complex than those for control, diagnostic models may comprise only of a partial description of the model. Their complexity depends basically on the given situation such as the kind of plant, the kind and number of faults to be detected, the demands for fault isolation and the robustness and the measurements available.