导航

Acta Aeronautica et Astronautica Sinica ›› 2025, Vol. 46 ›› Issue (S1): 732159.doi: 10.7527/S1000-6893.2025.32159

• Excellent Papers of the 2nd Aerospace Frontiers Conference/the 27th Annual Meeting of the China Association for Science and Technology •    

Prediction of whirl flutter boundary for tiltrotor aircraft based on BPNN with adaptive data

Lixiong ZHENG1,2, Zhe CHEN1,2, Xin WANG1,2, Qijun ZHAO1,2()   

  1. 1.National Key Laboratory of Helicopter Aeromechanics,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China
    2.Helicopter Research Institute,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China
  • Received:2025-02-25 Revised:2025-03-27 Accepted:2025-04-27 Online:2025-05-09 Published:2025-05-06
  • Contact: Qijun ZHAO E-mail:zhaoqijun@nuaa.edu.cn
  • Supported by:
    National Natural Science Foundation of China(12032012);Postgraduate Research & Practice Innovation Program of Jiangsu Province(KYCX25_0563);the Priority Academic Program Development of Jiangsu Higher Education Institutions

Abstract:

To address the aeroelastic instability issues of tiltrotor aircraft, a prediction method for whirl flutter boundary of tiltrotor aircraft based on Back-Propagation Neural Network (BPNN) was proposed. Firstly, a multi-modal coupled aeroelastic stability analysis model for tiltrotor aircraft was established based on Hamilton’s principle and multi-body dynamics methods. Secondly, minimal modal damping ratio data under strongly correlated parameters characterizing system stability were generated, and an artificial neural network prediction model was constructed. Finally, an adaptive data refinement method was proposed to enhance the prediction accuracy of the neural network model. The results show that within the training range, the maximum error between the calculated and predicted values is 3.24%, with an average relative error of 0.031%; outside the training range, the maximum error is 6.51%, and the average relative error is 0.089%. The trained BPNN model exhibits good generalization and high fitting accuracy, enabling efficient and high-precision predictions with fewer sample data, both within and outside the training range. The refinement method effectively improves prediction accuracy, particularly excelling in the prediction of the critical point of whirl flutter, significantly mitigating the peak-to-peak fluctuation effects. Moreover, BPNN provides new tools and methods for handling large-scale complex data, offering new perspectives for the research and application of aeroelastic dynamics in tiltrotor aircraft.

Key words: BP neural network, tiltrotor aircraft, whirl flutter, adaptive data, boundary prediction

CLC Number: