Fault Diagnosis of Non-Linear Dynamic Systems

The majority of model-based fault diagnosis methods are based on linear system models. For non-linear systems, the fault diagnosis problem has been traditionally approached in two steps. Firstly, the model is linearized at an operating point, and then robust techniques are applied to generate residual signals which are insensitive to model parameter variations within a small neighborhood of the operating point. The robustness issue is tackled using techniques developed for linear system models. The strategy only works well when the linearization does not cause a large mismatch between linear model and non-linear behavior, the residual has been designed to be robust enough to tolerate small model perturbations around the operating point, and the system closely operates around the operating point specified. However, for systems with high nonlinearity and a wide dynamic operating range, the linearized approach fails to give satisfactory results. A linearized model is an approximate description of the non-linear system dynamics around the operating point. However, when the system operating range becomes wider, the linearized model is no longer able to represent the system dynamics. One solution is to use a large number of linearized models corresponding to a range of operating points. However, this would involve a large number of FDI systems corresponding to all operating points. This is not very practical for real-time application.