Fault Detection for Uncertain Systems Using Adaptive Sliding-mode Disturbance Observer

When a plant has faults, the plant will be corrupted by unknown fault signals and this paper proposes a design method to detect the fault signals in the case of plants with uncertainties. This paper estimates the unknown fault signals by using a disturbance observer. The disturbance observer estimates state variables of a system augmented by a dynamic model of the fault signals. Hence the fault signals are a part of state variables of the augmented system, when the state variables of the augmented system is estimated, then the signal also is estimated. The plant considered in this paper is supposed to have two kinds of uncertainty; unstructured one and structured one. The unstructured uncertainty is unknown disturbance with known upper bound. The structured uncertainty is unknown disturbance of a sum of unknown parameters multiplied by known functions. To make the observer robust to the unstructured uncertainty, this paper includes sliding mode scheme in the observer, and to the structured uncertainty, the observer in this paper has an adaptive law to estimate the unknown parameters. The observer in this paper treats the two kind of unknown signals, the fault signals and the disturbances, in the different way, it estimates the fault signals exactly and reject the effect of the disturbances completely.