A double-threshold-testing robust method for fault detection and isolation in dynamic systems

"Robustness" is one of the most important issues for model-based fault detection and isolation (FDI) in dynamic systems. This paper proposes a double threshold statistical testing method for general residuals that are created by a model-based residual generator. After all available knowledge and information has been applied to the residuals, they are still nonzero while there is no fault. The double threshold statistical testing represents the nonzeros, model-errors or residues, by a standard Bernoulli and binomial random model so that, by appropriate design, the false-alarm probability can be controlled and the detection power can be maximized. Consequently, the FDI scheme becomes more "robust" while maintaining a higher level of fault detection sensitivity. The theoretical properties of this robust method are presented and the testing criterion is examined.