航空学报 > 2015, Vol. 36 Issue (10): 3359-3369   doi: 10.7527/S1000-6893.2014.0333

一种非线性自适应切换控制混合方法及其在倾转旋翼机上的应用

王奇, 吴文海   

  1. 海军航空工程学院 青岛校区, 青岛 266041
  • 收稿日期:2014-09-28 修回日期:2014-12-02 出版日期:2015-10-15 发布日期:2014-12-09
  • 通讯作者: 王奇, Tel.: 0532-58833632 E-mail: wq_navy@163.com E-mail:wq_navy@163.com
  • 作者简介:王奇 男, 博士研究生。主要研究方向: 飞行力学与飞行控制。 Tel: 0532-58833632 E-mail: wq_navy@163.com;吴文海 男, 博士, 教授。主要研究方向: 飞行控制。 Tel: 0532-58833632 E-mail: austin@qingdaonews.com
  • 基金资助:

    航空科学基金 (20100785001)

A nonlinear adaptive switching control blending method and its application to tiltrotor

WANG Qi, WU Wenhai   

  1. Qingdao Branch, Naval Aeronautical Engineering Institute, Qingdao 266041, China
  • Received:2014-09-28 Revised:2014-12-02 Online:2015-10-15 Published:2014-12-09
  • Supported by:

    Aeronautical Science Foundation of China (20100785001)

摘要:

为避免子控制器切换时控制量的跳变,提出了一种非线性自适应切换控制混合方法。针对输入输出反馈线性化子控制器在使用中存在的逆误差及模型不确定性,采用多层神经网络进行在线补偿,为实现此类非线性自适应子控制器的平滑切换,实际控制律采用各子控制律的凸组合,各组合系数值由切换参数确定。通过合适的设计参数选取与神经网络权值更新律设置,寻找到了闭环切换系统的公共Lyapunov函数,保证了此类系统在切换控制混合下的稳定性。在倾转旋翼机轨迹跟踪控制的应用中,设计了直升机模式、过渡模式与飞机模式的非线性子控制器,应用神经网络在线补偿与随短舱角的控制混合,仿真结果表明该方法具有对系统不确定性的鲁棒性及平滑切换的特性。

关键词: 非线性切换系统, 平滑切换, 神经网络, 倾转旋翼机, 飞行控制

Abstract:

To avoid control signals of sub-controllers bumping at switching time, a nonlinear adaptive switching control blending method is presented. Multilayer neural-net is used to compensate the inversion error of feedback linearization controllers and unmodeled uncertainties. The convex combination of each sub-controller output is utilized as the online control signal in which the weight value of each sub-controller is determined by a switching parameter, and then the continuous and smooth switching can be achieved. A common Lyapunov function of the closed-loop switching control system is found by choosing design parameters and weights update laws of neural-net appropriately, then the stability of this system is guaranteed. With its application to tiltrotor trajectory tracking control, three sub-controllers are designed based on input-output feedback linearization for helicopter mode, transition mode and airplane mode, which are combined with neural-net online compensation and switching control blending by nacelle angle for all controllers. Simulation results show that this method is robust for plant uncertainties and has a smooth switching characteristic in mode conversions.

Key words: switched nonlinear system, smooth switching, neural networks, tiltrotor, flight control

中图分类号: