LSTM recurrent neural network assisted aircraft stall prediction for enhanced situational awareness

• Published in: arXiv.org
• Main Authors: Saniat, Tahsin Sejat, Goni, Tahiat, Galib, Shaikat M
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 09.12.2020
• Subjects: Aircraft; Computer architecture; Neural networks; Recurrent neural networks; Situational awareness; Stalling; Warning systems
• ISSN: 2331-8422

Since the dawn of mankind's introduction to powered flights, there have been multiple incidents which can be attributed to aircraft stalls. Most modern-day aircraft are equipped with advanced warning systems to warn the pilots about a potential stall, so that pilots may adopt the necessary recovery measures. But these warnings often have a short window before the aircraft actually enters a stall and require the pilots to act promptly to prevent it. In this paper, we propose a deep learning based approach to predict an Impending stall, well in advance, even before the stall-warning is triggered. We leverage the capabilities of long short-term memory (LSTM) recurrent neural networks (RNN) and propose a novel approach to predict potential stalls from the sequential in-flight sensor data. Three different neural network architectures were explored. The neural network models, trained on 26400 seconds of simulator flight data are able to predict a potential stall with > 95% accuracy, approximately 10 seconds in advance of the stall-warning trigger. This can significantly augment the Pilot's preparedness to handle an unexpected stall and will add an additional layer of safety to the traditional stall warning systems.