Digital Twin of Fixed-Wing Aircraft Powertrain Performance

A digital twin model for a fixed-wing unmanned aerial vehicle (UAV) equipped with an internal combustion engine and a variable pitch propeller is developed in this paper. The modeling framework begins with the analysis of key flight performance parameters, derived from first-principles of flight mechanics and calibrated using laboratory and flight test data. Subsystem-level models capture the behavior of individual components, while system-level integration represents the coupled airframe–propulsion dynamics under steady-state assumptions. Machine learning models are employed to represent multidimensional mappings across data spaces, including engine fuel efficiency, propeller aerodynamic coefficients, and thrust-power performance maps. These models establish relationships between flight state variables and operational outputs such as fuel burn rate, brake-specific fuel consumption, engine settings, and propeller pitch. The selection and optimization of machine learning architectures, including neural networks and symbolic regression techniques are detailed in the paper. The structured generation of synthetic training data to ensure representative and uniform coverage of the flight envelope is also discussed. The resulting digital twin models are computationally efficient, generalizable, and scalable. These models support a broad range of applications, including energy-aware trajectory planning, low-level feedback control, and system-level diagnostics. The framework is designed for extensibility across a wide class of fixed-wing platforms and propulsion types, contributing toward the integration of data-driven methods and flight physics for energy-aware autonomy in ultra-long endurance UAV missions.