Aircraft range fuel prediction study based on WPD with IAPO optimized BiLSTM-KAN model.

Area of Science:

• Aerospace Engineering
• Computational Intelligence
• Sustainable Development

Background:

• Accurate aircraft fuel consumption prediction is crucial for sustainable development in civil aviation.
• Existing models have limitations in predicting aircraft range fuel expenditure.
• There is a need for advanced methods to improve the precision of fuel consumption forecasts.

Purpose of the Study:

• To propose a novel fuel consumption prediction model integrating Wavelet Packet Decomposition (WPD) and an Improved Arctic Puffin Optimization (IAPO) algorithm to optimize a Bidirectional Long-Short-Term Memory network-Kolmogorov-Arnold network (BiLSTM-KAN).
• To enhance the accuracy and reliability of aircraft fuel consumption predictions for sustainable aviation practices.
• To address the shortcomings of current aircraft range fuel prediction methodologies.

Main Methods:

• Feature selection using Pearson's correlation coefficient.
• Data decomposition into frequency subsequences using Wavelet Packet Decomposition (WPD).
• Integration of Bidirectional Long-Short-Term Memory (BiLSTM) for temporal dependencies and Kolmogorov-Arnold Network (KAN) for nonlinear relationships.
• Optimization of BiLSTM-KAN hyperparameters using the Improved Arctic Puffin Optimization (IAPO) algorithm, incorporating SPM chaotic mapping and adaptive strategies.

Main Results:

• The proposed WPD-IAPO-BiLSTM-KAN model demonstrated superior prediction accuracy across B737, A320, and B747 aircraft models.
• Achieved low error metrics: e.g., for B737, MSE=34.57, NRMSE=0.0061, MAPE=3.81, R²=0.9974.
• Outperformed comparative models in predicting aircraft fuel consumption, indicating high precision and low prediction error.

Conclusions:

• The WPD-IAPO-BiLSTM-KAN model offers a novel and effective approach for predicting airline fuel consumption.
• The model provides valuable insights for reducing aircraft fuel consumption and supporting sustainable aviation.
• The integration of WPD, IAPO, BiLSTM, and KAN significantly enhances prediction performance.