Using artificial neural networks and self-organizing maps for detection ofairframe icing

A method of using artificial neural networks (ANNs) and Kohonen self-organi zing maps (SOMs) to detect airframe ice is proposed and investigated. It is hypothesized that ANN systems trained on the aircraft dynamics in real tim e would converge to different connection weights for iced and clean aircraf t. Kohonen SOMs are proposed for detecting these differences automatically and, therefore, recognizing airframe ice accretion, This approach is shown to be capable of acting in an advisory role for the flight crew. The fideli ty of the approach is shown to depend on the level of atmospheric turbulenc e, as well as on the magnitude of the elevator input.