A learning approach to the SFDIA problem using radial basis function networks

This paper presents an online learning approach for the problem of sensor failure detection, identification, and accommodation (SFDIA) using neural networks. The SFDIA scheme exploits the analytical redundancy of the system to provide accommodation for a set of sensors without physical redundancy. A modified version of Gaussian radial basis function network (GRBF) is used to approximate the unknown nonlinearities of the dynamic system. The properties of RBF networks provide a learning law with guarantee of stability. A modified form of GRBFN reduces the computational burden typical of the RBFN, while preserving the stability of the learning. The scheme is then applied to the SFDIA problem within the longitudinal flight control system of an F-16 aircraft