Sensor validation for turbofan engines using an autoassociative neural network

Abstract

This paper develops a novel sensor validation approach for the F100 turbofan engine by using an autoassociative neural network. The proposed approach utilizes dimensionality reduction to develop the neural network for sensor failure detection and data recovery. In the autoassociative neural network, the redundant sensor information is compressed, mixed and reorganized into a smaller number of network nodes in the first part of the network. The compressed information is then used to regenerate the original redundant data at the output. Due to the information mixture, if some sensors fail, other sensor data still provide enough information to generate the good estimates to replace the faulty measurements. The simulation results demonstrate the applicability of the proposed scheme.