Data-Driven and Intelligent Estimation of Aircraft Aerodynamic Parameters via a Transformer-BiLSTM Network
Open Access
- 21 November 2025
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 981-993
Abstract
To address the core challenges of real flight data—including numerous parameters, complex features, strong nonlinearity, and significant noise—this paper proposes an end-to-end aerodynamic parameter estimation method integrating data mining and machine learning. First, the Maximum Information Coefficient (MIC) is employed to uncover nonlinear correlations between flight parameters and aerodynamic coefficients. Combined with Pearson correlation coefficients, redundant features are eliminated to construct a discriminative feature subset. Building upon this foundation, a Transformer-BiLSTM hybrid model is proposed. The bidirectional long short-term memory (BiLSTM) captures local temporal patterns, while the Transformer's self-attention mechanism models global dependencies. The hybrid model is trained and tested using real flight test data. Results demonstrate that the constructed hybrid model achieves outstanding performance in estimating aerodynamic coefficients for civil aircraft. Compared to standalone BiLSTM or Transformer models, as well as state-of-the-art iTransformer sequence models and other mainstream approaches, it exhibits significant advantages in estimation accuracy. This research provides a reliable data-driven solution for flight aerodynamic parameter estimation, fully showcasing the immense application potential and practical value of intelligent technologies in complex industrial scenarios.Keywords
This publication has 25 references indexed in Scilit:
- AFEformer: An adaptive frequency enhancement transformer for time series predictionEngineering Applications of Artificial Intelligence, 2025
- A digital twin-assisted algorithm for diagnosis of permanent magnet synchronous generator interturn short circuit fault and converter open circuit fault in wind power systems using Pearson correlation coefficientEngineering Applications of Artificial Intelligence, 2025
- Short-term offshore wind speed prediction model based on VMD-GDPSO-TCN-BiLSTMOcean Engineering, 2025
- Aircraft parameter estimation using a stacked long short-term memory network and Levenberg-Marquardt methodChinese Journal of Aeronautics, 2023
- System identification method based on interpretable machine learning for unknown aircraft dynamicsAerospace Science and Technology, 2022
- Application of a long short-term memory neural network for modeling transonic buffet aerodynamicsAerospace Science and Technology, 2021
- Estimation of aerodynamic parameters near stall using maximum likelihood and extreme learning machine-based methodsThe Aeronautical Journal, 2020
- Dynamic aerodynamic parameter estimation using a dynamic particle swarm optimization algorithm for rolling airframesJournal of the Brazilian Society of Mechanical Sciences and Engineering, 2020
- Nonlinear aircraft system identification using artificial neural networks enhanced by empirical mode decompositionAerospace Science and Technology, 2018
- Aircraft parameter estimation using Hybrid Neuro Fuzzy and Artificial Bee Colony optimization (HNFABC) algorithmAerospace Science and Technology, 2017