神经网络预测控制应用研究

【作者】 王捍兵；

【导师】 曾庆军；

【作者基本信息】 江苏科技大学 ， 控制理论与控制工程， 2011， 硕士

【摘要】 基于神经网络的预测控制是国内外复杂工业过程控制领域中研究的前沿问题之一。本文根据预测控制的三大机理,在分析基本的预测控制算法的基础上,设计了一种新颖的基于BP神经网络模型的广义预测控制算法,并开展了神经网络预测控制在微创手术遥操作机器人系统和三容水箱系统中的应用研究。本文绪论对预测控制和神经网络产生的历史背景、发展历程、研究现状、发展趋势和存在的问题作了简要的综述,并扼要的讲述了神经网络预测控制复合控制方法的应用研究和发展方向。第二章首先针对三种典型预测控制算法讨论了预测控制的基本结构和原理,深入分析了预测控制的模型预测、反馈校正与滚动优化方法。然后通过分析三种常见的神经网络模型对神经网络控制原理进行简单的探讨。第三章在微创手术遥操作机器人系统模型的基础上,针对遥操作机器人微创手术中,由呼吸运动和心跳产生的组织周期性位移干扰设计了一种广义预测控制方案。通过MATLAB仿真实验表明,系统能在遥操作机器人系统模型下较好的消除这种由呼吸运动和心跳产生的周期性扰动。第四章以工业过程中大时延、非线性系统三容水箱为研究对象,分别进行PID算法、广义预测控制算法和神经网络预测控制算法的仿真实验,从而体现了神经网络预测控制复合控制算法在工业过程控制装置上的实用性和先进性。第五章设计了遥操作机器人系统的LabVIEW实验平台,实现实验装置的串口通信、数据采集卡以及网络通讯,并完成了主从手位置跟踪实验。在LabVIEW平台界面上,实验操作更加简便,数据观察更加直观。最后对全文进行总结,并结合自己的研究心得,指出一些可深入研究或有待解决的问题。更多还原

【Abstract】 Predictive control based on neural networks is one of the forward problems on study in the field of complex industrial process control. This paper, according to predictive control theory, designs a novel generalized predictive control algorithm based on BP neural network. Also, in this paper, neural network predictive control algorithm is studied for minimally invasive surgical teleoperation robot system and three-container water tanks system.In Ch.1, the background, the present development, the tendency and problems of the predictive control and the neural networks are summarized briefly. The application research and the development direction of the compound control method the neural network predictive control are discribed briefly.In Ch.2, first, aiming at the typical predictive control algorithm, the basic structure and the theory of the predictive control are discussed. Then, neural network control theory is simply discussed through the analysis of three common neural network models.In Ch.3, In order to eliminate disturbance of periodic deformations of organs due to respiratory movements and heart beating, during minimally invasive surgery with teleoperation robot, this paper designs an algorithm of generalized predictive controller scheme, based on the minimally invasive surgical teleoperation robot system model. The curves of MATLAB show that the system can reject better the cyclic disturbances.In Ch.4, the simulation experiments of PID, GPC and NNPC algorithms are made, taken the three-container water tanks of the long time delay and nonlinear system in industrial process, as the research object. The results of the experiments show that the compound control method of the neural network predictive control algorithm is practical and advanced in industrial process control equipments.In Ch.5, the LabVIEW platform of the teleoperation robot system is established for serial port communication, data acquisition card, network communication and the experiment of master-slave tracking position. In the LabVIEW platform interface, the experimental operation is more convenient and the observation data is more intuitive.In the end, the paper is summed up. Some problems which can be deeply studied are presented according to personal experiences and ideas.更多还原