Distributed Fault Estimation through Neural Network Unknown Input Observer

This article proposes an integrated robust neural network and unknown input observer to address fault estimation in continuous interconnected systems affected by unknown input disturbances. Initially, a neural network-based joint observer was developed by enhancing the unknown input observer and leveraging a neural network to model the system’s interconnection terms. Subsequently, it was shown that the combined stability of the enhanced unknown input observer and the neural network modeling system leads to improved overall system stability. Finally, the accuracy of fault estimation using the neural network-based joint observer was validated through a model of a ground-air interconnection system.