A Neural Network Based Tool for Aircraft SFDIA Modeling and Simulation

This paper presents a Neural Network (NN) based tool for the modeling, simulation and analysis of aircraft Sensor Failure, Detection, Identification and Accommodation (SFDIA) problems. The SFDIA scheme exploits the analytical redundancy of the system to provide validation capability to measurement devices by employing Neural Networks as on-line non-linear approximators. The tool allows evaluating either the open loop or the closed loop performance of the SFDIA scheme. Several kinds of NN approximators and learning algorithms are employed, and a library comprising all these different adaptive neural networks is presented. In particular, Resource Allocating Networks featuring fully tuned Radial Basis Activation Functions are proposed as one of the most effective architectures. Finally, the results of a comparative study of different NN approximators applied to the SFDIA problem on a detailed nonlinear model of a De Havilland DHC-2 “Beaver” aircraft are reported.