Data-Driven Prediction of Aircraft Aerodynamic Coefficients Using GA-LM-BP Neural Network

Accurately predicting aerodynamic coefficients like lift and drag during aircraft landing is crucial for aircraft performance modeling and simulation. This paper proposes a new approach using QAR flight data and GA-LM-BP neural network to predict lift and drag coefficients online during landing. QAR data from Airbus A320 flights, including airspeed, altitude, flap settings, etc., were used to train the LM-BP network optimized with a GA. The model was evaluated on test data, achieving a mean absolute error of 0.024% for lift coefficient prediction and 0.054% for drag coefficient compared to other machine learning simulation data. The model demonstrates a promising capability to predict aerodynamic coefficients online during landing using only typical QAR flight parameters.