AN INPUT ESTIMATION METHOD FOR FDI USING MULTIPLE ASYNCHRONOUS SENSORS

Abstract This paper addresses the problem of fault detection in dynamic linear objects. A fault is modeled as an unknown input to the object. The presented detection algorithm is based on the input estimation method, where the fault is computed from the Kalman filter innovations. To perform fault detection with minimal delay, multiple sources of data are considered. The data sources (sensors) generate measurements asynchronously at arbitrary time moments. An appropriate use of continuous-time object model allows effective analysis of these measurements. The presented method is able to estimate the moment of the fault and the corresponding fault value.