Fault detection of affine hybrid systems using finite memory observers

The purpose of this paper is to design a finite memory observer for fault detection of a class of hybrid systems. The hybrid FMO observer is decomposed in two stages: the discrete observer and the continuous observer which estimate the mode and the continuous state vector respectively. Using the past output measurements and the model equations, state and mode estimation is expressed as a minimization problem. In the fault detection approach, we focus about residuals generation using two different approaches. The performances of the proposed method for state estimation and fault detection is illustrated through numerical simulations on a simulated benchmark two-tanks system whose equations are linearized.