DNN-Based Surrogate Modelling-Based Aircraft Performance: Take Off and Landing Distance

Abstract

In the early stages of aircraft design, available information is often limited to theoretical models supported by sparse experimental or flight test data for calibration. While these preliminary models provide a useful foundation, they can be unreliable due to simplifying assumptions inherent in theoretical formulations. In such cases, data-driven methods such as neural network–based surrogate models can serve as valuable alternatives, effectively leveraging limited experimental data to enhance the accuracy and reliability of predictions. In this work a real-world dataset comprising take-off and landing distances from Boeing 737 aircraft was utilized to develop a data-driven model. This model relies on a limited set of basic input features to predict take-off and landing distances, without requiring knowledge of the aircraft's aerodynamic properties or solving physics equations.