Detection and identification of faulty sensors with maximized sensitivity

In this paper we propose a new method for the detection, identification and reconstruction of faulty sensors using a generalized normal process model. The model residual is used to detect sensor faults, and a structured residual approach with maximized sensitivity (SRAMS) is proposed to identify the faulty sensor. An exponentially weighted moving average (EWMA) filter is applied to reducing the effects of noise and dynamic transients. Three different indices are proposed and compared for the identification of faulty sensors. Faulty sensor is reconstructed based on the normal process model and faulty data. The effectiveness of the proposed scheme is tested using the data from an industrial boiler process, where four types of faults are simulated.