Clear Air Turbulence Detection with Neural Networks

Neural Networks are applied for the combination of 8 experimental hazard indices, in the search for a more reliable and more robust hazard model of Clear Air Turbulence. The experimental data consists of actual flight data from 5 turbulence encounters, as provided by Japan Aerospace Exploration Agency (JAXA). Measurement Integrated Simulation has been conducted with this data, and the hazard indices were computed from the simulation results of the calculated flow field. Results suggest that neural networks might be useful to address the robustness problem of the hazard indices, since the consol idated index derived by the network consistently presented a performance that was close to or slightly superior to the best indices for each flight case. Overall results showed a probability of detection (pd1) = 0.47 for a false alarm ratio of (1- pd0) = 0.1.