Sensor failure detection with a bank of Kalman filters

Abstract

This investigation presents a new approach for detecting sensor failures which affect only subsets of system measurements. In addition to a main Kalman filter, which processes all the measurements to give the optimal state estimate, a bank of auxiliary Kalman filters is also used, which process subsets of the measurements to provide the state estimates which serve as failure detection references. After the statistical property of the difference between the state estimate of the main Kalman filter and those of the auxiliaries is derived with an application of the orthogonal projection theory, failure detection is undertaken by checking the consistency between the state estimate of the main Kalman filter and those of the auxiliaries by means of the chi-square statistical hypothesis test.