基于PSO算法的RBF神经网络在板形板厚综合控制中的应用
发布时间：2018-10-12 09:48
【摘要】：钢铁是发展国民经济的重要物资基础,板带材是广泛应用于国民经济各部门的重要原材料。其中板形质量和板厚精度是衡量带钢质量的两个重要指标。板形板厚控制是一个复杂多变的控制系统。影响板形板厚的各个参数之间有着很强的耦合性。因此,实现板形板厚控制(AFC-AGC)的研究已经成为一个前沿的热点问题。 随着智能技术的研究和发展,很多专家学者将智能技术应用到AFC-AGC综合控制中。由于AFC-AGC是一个非线性、强耦合、大时滞的多变量实时控制系统。对于这样的非常规的复杂系统,常规的方法很难实现理想的控制。因此,采用现代控制方法和智能方法相结合的控制手段成为必然的趋势。 本文针对AFC-AGC综合控制的特点,主要进行了以下工作： 1.通过分析板带材轧制过程,完成AGC-AFC系统数学公式的推导和数学模型的建立,并建立AGC-AFC的系统框图。 2.对粒子群优化算法(Particle Swarm Optimizition.PSO)进行了深入的分析和研究,针对PSO算法在使用过程中存在易于陷入局部最优,收敛精度不高等缺陷,提出适合本文的改进PSO优化算法,并通过Matlab仿真验证改进的PSO算法具有很好的精度。 3.介绍了神经网络的基本内容,比较了RBF神经网络和BP神经网络的优缺点,选择适合本文的RBF神经网络。同时用改进的PSO算法对RBF神经网络的结构和参数进行优化,并比较分析了各种优化效果。 4.设计出一种基于PSO算法的RBF神经网络的解耦控制器,并将其用于对板形板厚的综合控制系统中,完成系统的解耦实现分别控制。采用Matlab仿真结果证明了本文提出的方案具有很好的解解耦性,满足了所需要的精度。 本文提出的控制方案,结构简单从而方便工程实现,与此同时还有较好的解祸效果和鲁棒性。为板形板厚综合控制系统提供了一种新思路、新途径。
[Abstract]:Iron and steel is an important material base for the development of national economy, and sheet and strip is an important raw material widely used in various sectors of national economy. The shape quality and plate thickness accuracy are two important indexes to measure strip quality. Plate thickness control is a complex and variable control system. There is a strong coupling between the parameters that affect the thickness of the plate. Therefore, the research of shape and plate thickness control (AFC-AGC) has become a hot topic. With the research and development of intelligent technology, many experts and scholars apply intelligent technology to AFC-AGC integrated control. Because AFC-AGC is a nonlinear, strongly coupled, large time delay multivariable real-time control system. For such unconventional complex systems, conventional methods are difficult to achieve ideal control. Therefore, the combination of modern control methods and intelligent methods has become an inevitable trend. The main work of this paper is as follows: 1. By analyzing the rolling process of sheet and strip, the mathematical formula and mathematical model of AGC-AFC system are deduced, and the system block diagram of AGC-AFC is established. 2. The particle swarm optimization (Particle Swarm Optimizition.PSO) algorithm is deeply analyzed and studied. In view of the disadvantages of PSO algorithm which is easy to fall into local optimum and low convergence precision, an improved PSO optimization algorithm suitable for this paper is proposed. Matlab simulation shows that the improved PSO algorithm has good accuracy. 3. 3. The basic content of neural network is introduced, the advantages and disadvantages of RBF neural network and BP neural network are compared, and the RBF neural network suitable for this paper is selected. At the same time, the structure and parameters of RBF neural network are optimized by the improved PSO algorithm. A decoupling controller of RBF neural network based on PSO algorithm is designed and applied to the integrated control system of shape and plate thickness. The Matlab simulation results show that the proposed scheme has good decoupling and meets the required precision. The control scheme presented in this paper is simple in structure and convenient for engineering implementation. At the same time, it also has good effect and robustness. It provides a new idea and new way for the integrated control system of shape and plate thickness.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F426.3;TP18
本文编号：2265675
[Abstract]:Iron and steel is an important material base for the development of national economy, and sheet and strip is an important raw material widely used in various sectors of national economy. The shape quality and plate thickness accuracy are two important indexes to measure strip quality. Plate thickness control is a complex and variable control system. There is a strong coupling between the parameters that affect the thickness of the plate. Therefore, the research of shape and plate thickness control (AFC-AGC) has become a hot topic. With the research and development of intelligent technology, many experts and scholars apply intelligent technology to AFC-AGC integrated control. Because AFC-AGC is a nonlinear, strongly coupled, large time delay multivariable real-time control system. For such unconventional complex systems, conventional methods are difficult to achieve ideal control. Therefore, the combination of modern control methods and intelligent methods has become an inevitable trend. The main work of this paper is as follows: 1. By analyzing the rolling process of sheet and strip, the mathematical formula and mathematical model of AGC-AFC system are deduced, and the system block diagram of AGC-AFC is established. 2. The particle swarm optimization (Particle Swarm Optimizition.PSO) algorithm is deeply analyzed and studied. In view of the disadvantages of PSO algorithm which is easy to fall into local optimum and low convergence precision, an improved PSO optimization algorithm suitable for this paper is proposed. Matlab simulation shows that the improved PSO algorithm has good accuracy. 3. 3. The basic content of neural network is introduced, the advantages and disadvantages of RBF neural network and BP neural network are compared, and the RBF neural network suitable for this paper is selected. At the same time, the structure and parameters of RBF neural network are optimized by the improved PSO algorithm. A decoupling controller of RBF neural network based on PSO algorithm is designed and applied to the integrated control system of shape and plate thickness. The Matlab simulation results show that the proposed scheme has good decoupling and meets the required precision. The control scheme presented in this paper is simple in structure and convenient for engineering implementation. At the same time, it also has good effect and robustness. It provides a new idea and new way for the integrated control system of shape and plate thickness.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F426.3;TP18
【参考文献】
相关期刊论文 前7条
1 侯志荣,吕振肃;IIR数字滤波器设计的粒子群优化算法[J];电路与系统学报;2003年04期
2 舒迪前;奉川东;尹怡欣;;多变量系统神经网络解耦广义预测控制及其应用[J];电气传动;1996年04期
3 舒怀林;PID神经元网络对强耦合带时延多变量系统的解耦控制[J];控制理论与应用;1998年06期
4 郑毅 ,吴斌;由鸟群和蚂蚁想到的——基于主体的仿真与群集智能的研究[J];微电脑世界;2001年16期
5 邱东强,涂亚庆;神经网络控制的现状与展望[J];自动化与仪器仪表;2001年05期
6 余欣梅,李妍,熊信艮,吴耀武;基于PSO考虑谐波影响的补偿电容器优化配置[J];中国电机工程学报;2003年02期
7 周旭东,李连诗,王先进,傅作宝;人工神经网络板形板厚综合控制[J];轧钢;1995年02期
相关硕士学位论文 前1条
1 宋保强;前向神经网络学习理论[D];大连理工大学;2002年
，本文编号：2265675
本文链接：https://www.wllwen.com/guanlilunwen/shengchanguanlilunwen/2265675.html
