| 摘要: |
| 气动参数辨识对于大气层内飞行器来说至关重要，通过在线气动参数辨识可规划更准确的飞行轨迹，并对控制参数进行自适应调整。传统辨识方法的模型较为复杂，运算量大，无法满足飞行器在线辨识的要求。而基于神经网络的智能参数辨识方法，不仅可以离线对网络模型进行训练，并利用历史飞行数据进行模型修正，也可在线时直接利用训练好的网络对参数进行快速调整，在保证参数估计精度的同时，保障参数估计的快速性。提出了一种基于支撑向量机（SVM） 的样本扩充和神经网络参数在线快速修正方法。通过仿真和统计，证明了基于SVM的神经网络方法对飞行器气动参数进行在线快速智能辨识的可行性。 |
| 关键词: 气动参数辨识 智能辨识 在线快速辨识 神经网络 |
| DOI： |
|
| 基金项目:国家自然科学基金 (61403100) |
|
| Research on Intelligent Online Identification Technology for Aerodynamic Parameters of Aircraft |
| PU Jialun,HAN Yepeng,ZHANG Liang |
| (School of Astronautics , Harbin Institude of Technology) |
| Abstract: |
| Aerodynamic parameter identification is essential for aircraft in the atmosphere. By utilizing the online aerodynamic parameter identification method, more accurate flight trajectory can be planned, and the control parameters can be adaptively adjusted. Traditional identification method needs to establish the complex mathematical model and the calculation amount is large, which cannot meet the requirements of the online identification of the aircraft. Based on the intelligent parameter identification method of neural network, the network model can be fully trained in offline phase, and the historical model data can be used to modify the network model. The network model can adjust the parameter rapidly in the online phase to guarantee the precision and rapidity of parameter estimation. In this paper, a sample expansion combined with Support Vector Machine(SVM) and the online fast correction method for neural network parameters is proposed. Simulations results are carried out to prove the feasibility of the proposed method. |
| Key words: Aerodynamic parameter identification Intelligent identification Online quick identification Neural networks |