Sustainable Aviation Technologies and AI-Based Fuel Optimization

Abstract

A method was developed and evaluated that estimates and reduces the fuel consumption, and the corresponding carbon-dioxide (CO₂) emissions, of a general-aviation training aircraft (a Tecnam P2006T) from real recorded avionics data and the aircraft's certified performance model. Because the avionics logs record the parameters that govern fuel flow but not fuel flow itself, per-second fuel burn was reconstructed from the Pilot's Operating Handbook (POH) cruise tables and integrated over fifteen real flights totalling 28.1 flight hours. The reconstructed burn was 771 litres (1.79 t CO₂). A flat planning figure, evaluated at the setting the operator actually plans with, was shown to over-estimate the profile-implied burn by 12 %, systematically and in the same direction for every flight. A cruise optimiser that holds the useful work fixed found savings of 16.6–18.8 % of cruise fuel, mostly from the engine power setting rather than altitude, corresponding to about 2.6–3.0 t CO₂ per aircraft per year at typical training utilisation. A small, cross-validated machine-learning model was also trained to convert how a flight plan deviates from reality into a recommended fuel safety margin.