Sliding mode fuzzy observers and neural networks applied to solve fault detection and isolation problem

In this paper, non-linear systems (hydraulic tank configurations) were analyzed using a technique of fault detection and isolation based on a Takagi-Sugeno fuzzy model. The system state vector was obtained by means of sliding mode observers, and then the signal-residual is generated by comparing the estimated and measured outputs. The isolation problem was solved using Neural Networks. From the resulting active or inactive signal-residuals, the faulted elements of the system are easily identified. The method proposed represents a hybrid fault diagnosis technique.