DEVELOPMENT OF AN INTEGRATED NEURAL SCHEME FOR SENSOR FAILURES DIAGNOSIS AND DISTURBANCE ABSORPTION IN MIMO SYSTEMS

Abstract

In this thesis, (wo integrated artificial neural schemes for diagnosing sensors failures (detection and identification) in MIMO system are developed. One scheme is designed for failure of single sensor at a given instant, while the other one is designed for failure of multiple sensors at a given instant. Fur the first neural scheme, two neural networks are developed for sensor failures detection and identification (SFDI), The first neural network receives sensors output signals and system input signals to serve as on-line estimator for detecting system dist jrbancc or sensor failures, depending upon a predetermined threshold. The second neural network is pretrained to identify tlie failed sensor based on the failure signature it receives from the first neural network. For disturbance absorption (DA), athird neural network is developed