Software for the identification of nonlinear dynamic systems in noisy environment using multiple model approach and its application to fault diagnosis

This software tool implements the researches into the development of methods for rapid detection and diagnosis of faults in input-output control sensors of industrial processes. The tool relies on the use of analytical redundancy methods based on information implicit in functional or analytical relationships which always exist between a number of measurements taken from a process (e.g. industrial plant). The software has exploited methods, using robust state observers, which have provided a capability for reliable detection of faults in the presence of typical plant parameter variations. These methods enable faults to be isolated, in addition to being detected. Furthermore, the reliable detection and isolation of multiple faults has also been experimented. Applications studies range from industrial gas turbines and power plants (e.g. Pont sur le Sambre). The research was driven by industrial application studies, although a particular interest has been the development of a unifying theory for model-based diagnosis algorithms. I am trying to apply the experimented methods within the context of robustness to uncertainty.