A Model of Predicting the Consumption of Fuel for Aircraft in Dynamic Time-Series Based on LSTM-KAN Network
TANG Zhixing
NIU Zhaolun
FAN Yijie
YANG Ruichao
ZHONG Yuming
JIA Meng
TANG Xiaowei
Abstract：In view of the problem that it is very difficult for traditional methods to capture the intricate and nonlinear relationship between flight states and fuel consumption,this paper proposes a method of predic-ting the consumption of fuel for aircraft in dynamic time-based on the LSTM-KAN Network.First,eight key flight state parameters from QAR(quick access recorder)data in the terminal area-including altitude,true airspeed,and wind speed-the model employs KAN layers with B-spline basis functions in combination with a basic output structure are utilized for accurately capturing the nonlinear relationships between flight states and fuel consumption.And then,the KAN network logging on the final time step of the LSTM net-work is to achieve high-precision modeling of the dynamic time-varying patterns in the consumption of fuel for aircraft.The experimental results demonstrate that the model achieves a mean squared error(MSE)of 0.001,with 98.32%of the test set exhibiting a root mean square error(RMSE)below 0.09 1 kg/h.Ad-ditionally,the coefficient of determination(R2)reaches even more 0.989 7,significantly outperforming traditional models such as MLP(Multilayer Perceptron),standalone LSTM,and Transformer.The find-ings can be applied to optimization of airline fuel efficiency and airspace operations,thereby promoting greener practices in civil aviation.
Keywords：aircraft fuel consumption predictionKANLSTMQAR
Publication Date：2025-10-25
Online Publishing Date：2025-10-17(First online date of this platform, not the publication date of the document)
Pages：9( 22-30 )