一种飞行器弱模型依赖自适应控制方法
A Weak Model Dependent Adaptive Control Method for Flight Vehicle
金泽宇 1安帅斌 1周大鹏 2郑智 3刘凯1
作者信息
- 1. 大连理工大学力学与航空航天学院,大连 116081
- 2. 沈阳飞机设计研究所,沈阳 110035
- 3. 沈阳飞机工业(集团)有限公司工程技术中心试飞试验室,沈阳 110850
- 折叠
摘要
为了降低飞行控制设计对模型准确度的依赖程度,提升复杂环境干扰情况下飞行控制系统鲁棒性,放宽稳定性设计边界,研究了飞行器弱模型依赖自适应控制方法.该方法以经典动态逆控制为基础,离线阶段通过神经网络训练建立模型不确定性与动态逆最优控制增益之间的映射关系,在线阶段采用非线性正交递归最小二乘方法实时辨识模型不确定性参数,依据神经网络输出在线调节最优控制增益,并结合状态观测器捕获辨识误差对控制性能影响,进行进一步补偿,实现动态逆控制的自适应优化.通过数学仿真与飞行试验,验证了弱模型自适应控制方法的鲁棒性与工程适用性;通过与经典工程控制方法对比验证了提出方法的优势.
Abstract
To mitigate the reliance of flight control design on model accuracy,enhance the robustness of flight control systems under complex environmental disturbances,and broaden the stability design boundaries,an adaptive control method with weak model dependence for aircraft is investigated.This approach is grounded in classical dynamic inverse control principles.In the offline phase,a mapping relationship between model uncertainty and dynamic inverse optimal control gain is established through neural network training.During the online phase,a nonlinear orthogonal recursive least squares method is employed to identify model uncertainty parameters in real time,allowing for online adjustments of the optimal control gain based on neural network outputs.Additionally,a state observer captures identification errors and compensates for their impact on control performance,thereby facilitating adaptive optimization of dynamic inverse control.The robustness and engineering applicability of this weak model adaptive control method are validated through mathematical simulations and flight tests;furthermore,its advantages are demonstrated by comparisons with traditional engineering control methods.
关键词
飞行器控制/弱模型依赖控制/动力学在线辨识/离线神经网络Key words
Control of flight vehicle/Weak model dependency control/Dynamic online identification/Offline neural network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据