【作者】 孙书芳；

【导师】 王建辉；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2006， 硕士

【摘要】 板带钢材是经济建设中的重要物资，它的质量是关系国民经济发展。而板形和板厚是衡量板带钢材质量的主要指标。随着工业的发展，人们对带钢热轧板形板厚控制问题给予了越来越多的重视。热轧的板形板厚控制系统(AFC-AGC)是一个强耦合、非线性的多变量复杂系统。那么提高板形板厚综合系统的控制水平，就成了一个重要的研究难点和研究热点。 粒子群优化(PSO)算法是基于群体智能理论的优化算法。该算法利用生物群体内各个个体的合作与竞争等复杂行为产生群体智能。本文系统的介绍了PSO算法的基本性能，并归纳总结了其发展过程和应用。在此基础上研究了PSO算法在神经网络训练和PID参数整定上的应用，并提出了PSO算法优化神经网络和进行PID参数整定的基本理论和步骤。使用仿真实例证明了方法的有效性，并以板形板厚综合控制为背景研究了这两个方法的使用。 神经网络是一种应用广泛的人工智能技术。它可以利用数学算法模拟人脑的记忆、分析、推理等能力，从而可以达到以任意精度逼近任意非线性函数的目的。本文针对BP算法存在的问题和缺点提出了PSO神经网络的学习算法，并将这种算法应用到板形板厚的解耦控制中。 本文通过分析带材轧制过程，完成了AGC-AFC系统数学公式的推导和数学模型的建立，并建立了AGC-AFC的系统框图；在介绍神经网络技术的基础上提出了一种基于PSO优化神经网络的解耦控制器；并将PSO算法与PID控制方法相结合，设计了一种基于PSO算法的PID控制器；利用上述的基于神经网络解耦器和PID控制器组成了完整的“基于神经网络的AGC-AFC解耦的PID控制系统”。计算机仿真结果表明本文提出控制方案具有良好的解耦性和鲁棒性。 本文提出的基于PSO优化算法的神经网络解耦及基于PSO的PID控制方案不仅结构简单、易于工程实现，且具有较强的解耦性和鲁棒性。为有效解决板形板厚综合控制系统的难题提供了一种新思路、新途径。同时也拓展了神经网络和PSO算法等智能算法在工业控制领域的应用范围，在一定程度上推动了神经网络技术、PSO算法等智能算法的发展。更多还原

【Abstract】 Strip steel is substantial raw materials in national economy and the flatness and gauge are two key indices of performance. As the development of industry, the process of automatic flatness control and automatic gauge control has got more and more attention. In fact the complex system of automatic flatness control and automatic gauge control is a multi-input and multi-output nonlinear system with strong coupling and pure time delay. Therefore how to increase the control level of the flatness and gauge complex system has become hotspot and hard pot of related study.Particle Swarm Optimization (PSO) algorithm is based on swarm intelligence theory. The algorithm can provide efficient solutions for optimization problems through intelligence generated from complex activities such as cooperation and competition among individuals in the biologic colony. Basis PSO algorithm and its potential engineering applications such as neural network training and parameter setting of PID controller is introduced.Firstly, particle swarm optimization is introduced and its developments are reviewed. The basic applications of PSO algorithm and its engineering applications are summarized. The future research directions of PSO algorithm are pointed out and its potential applications are proposed.Secondly, the neural network theory and its structure are reviewed. Because of the great ability of simulation of human memory, analyzing and inference, neural network can approximate arbitrary nonlinear function. In order to avoid the defect in BP training algorithm a training method based on PSO algorithm is put forward. After analysis it is used to decouple in AGC-AFC system.Thirdly, the analysis of roll process is done and the mathematic model is established by the mathematic formula derivation of AGC-AFC system. According to the analysis above a control system of AGC-AFC based on neural network decoupling is designed. In the system a kind of controller is designed which is the combination of PSO algorithm and PID control theory. In the end of this article the whole control system is simulated and the results indicate that the system control scheme has excellent ability of decoupling and robustness.In short, the neural network decoupling scheme based on PSO algorithm has simple structure, good capability of decoupling and robustness, thus it can be realized in engineering easily. This scheme expand the application range of neural network and PSO algorithm in更多还原