基于NARX神经网络系统辨识的振动台迭代学习控制研究
Study on iterative learning control of shaker based on NARX neural network system identification
郭迎庆 1朱文 1刘少帅 1李世东 2景兴建 3徐赵东2
作者信息
- 1. 南京林业大学机械电子工程学院,南京 210016
- 2. 东南大学中国巴基斯坦"一带一路"重大基础设施智能防灾联合实验室,南京 210096
- 3. 香港城市大学机械电子工程学院,香港 999077
- 折叠
摘要
针对传统振动台台面控制效果不佳的问题,提出了一种自适应迭代学习控制算法,该算法在原有的位移三参量控制系统基础上构建外部位移闭环,形成双闭环控制系统.同时为更准确地模拟振动台的动态行为,引入灰狼优化(GWO)算法优化非线性有源自回归(NARX)神经网络对振动台模型辨识.仿真结果表明,利用GWO-NARX神经网络进行振动台模型辨识,取得了较高的辨识效果,精度可达99.8％.在辨识模型的基础上,利用自适应迭代学习控制算法极大地提高了振动台的控制精度,最大误差较原系统下降了 49.6％.与传统的NARX神经网络进行振动台模型辨识相比,GWO-NARX神经网络辨识效果更好,模型更贴近真实系统;与传统的三参量控制系统相比,自适应迭代学习控制算法提高了振动台波形复现精度,并且能够更好地适应系统的复杂性,为实际工程应用提供了可靠的技术支持和解决方案.
Abstract
Aiming at the poor control effect of the traditional shaker,an adaptive iterative learning control algorithm was proposed,which constructs an external displacement closed-loop on the basis of the original displacement three-parameter control system to form a double closed-loop control system.At the same time,in order to more accurately simulate the dynamic behavior of the shaker,the Gray Wolf Optimization(GWO)algorithm was introduced to optimize the nonlinear organic originated regression neu-ral network(Nonlinear Auto-Regressive with exogeneous inputs neural network,NARX)for shaker model identification.The sim-ulation results show that the shaker model identification using GWO-NARX neural network achieves a high identification effect with an accuracy of 99.8％.On the basis of identification model,the control accuracy of the shaker was greatly improved by using the adaptive iterative learning control algorithm,and the maximum error is reduced by 49.6％compared with the original system.Compared with the traditional NARX neural network for shaker model identification,the GWO-NARX neural network has a better identification effect and the model is closer to the real system;compared with the traditional three-parameter control sys-tem,the adaptive iterative learning control algorithm improves the accuracy of shaker waveform reproduction and can better adapt to the complexity of the system.It provides reliable technical support and solutions for practical engineering applications.
关键词
电动式振动台/自适应迭代学习/NARX神经网络/系统辨识Key words
electrodynamic shaker/adaptive iterative learning/NARX neural network/system identification引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据