基于DNN的飞机俯仰运动响应预测研究  

Aircraft pitch motion response prediction research based on DNN

作　　者：张晓敏[1,2] ZHANG Xiaomin(Chinese Flight Test Establishment,Xi'an 710089,China;AVIC Aeronautical Science and Technology Key Laboratory of Flight Simulation,Xi’an 710089,China)

机构地区：[1]中国飞行试验研究院,陕西西安710089 [2]航空工业飞行仿真航空科技重点实验室,陕西西安710089

出　　处：《飞行力学》2022年第2期53-60,共8页Flight Dynamics

基　　金：航空科学基金资助(2018ZA51003)。

摘　　要：针对人工智能的辨识方法在飞行器模型辨识应用中存在间接依赖数学模型以及泛化能力较低的局限性问题,基于深度学习思想,提出了一种新的数据处理方式,完成飞行器的系统模型辨识。首先,针对飞行器动态模型的特点,提出一种基于时序性的飞行数据处理方式;其次,采用交叉熵损失函数进一步优化深度神经网络;最后,针对飞行器纵向非线性模型进行仿真计算。仿真结果表明,训练好的模型成功提取了飞行器输入与输出之间的非线性映射关系,使得基于深度神经网络的飞行器模型能够对未知输入进行状态预测,克服了目前基于神经网络辨识算法的局限性。Although the identification method based on artificial intelligence has been more and more applied in aircraft model identification, it has the limitations of indirect dependence on mathematical models and low generalization ability. To solve this problem, based on the idea of deep learning, a new data processing method is proposed to complete the system model identification of aircraft. Firstly, according to the characteristics of aircraft dynamic model, a flight data processing method based on timing is proposed. Secondly, the cross entropy loss function is used to further optimize the deep neural network. Finally, the longitudinal nonlinear model of aircraft is simulated.The simulation results show that the trained model successfully extracts the nonlinear mapping relationship between the input and output of the aircraft, so that the aircraft model based on deep neural network can predict the state of the unknown input, which overcomes the limitations of the current neural network identification algorithm.

关 键 词：系统辨识 深度神经网络 飞行器建模 交叉熵损失函数 

分 类 号：V211[航空宇航科学与技术—航空宇航推进理论与工程] TP183[自动化与计算机技术—控制理论与控制工程]