基于神经网络的矫直机弯辊量自适应控制系统

【作者】 杨红光；

【导师】 黄庆学； 胡鹰；

【作者基本信息】 太原科技大学 ， 控制理论与控制工程， 2012， 硕士

【摘要】 由于板材在加热、轧制和冷却过程中受到多种因素的影响，所以轧制出的板材存在中浪、边浪和瓢曲等板型缺陷，欲消除此类板型缺陷，矫直机需配备弯辊装置。中厚板辊式矫直机的弯辊系统是决定板材质量、提高矫直效率的关键环节。由于矫直过程是多变量、非线性、慢时变、强耦合的一个过程，而采用传统技术构建的弯辊量模型并未涉及时变因素和耦合因素的影响。所以本文将现有的弯辊量模型和神经网络技术相结合构造弯辊量自适应模型，使其具有自适应性和自学习的能力。本文的主要研究内容如下：（1）弯辊量样本值管理系统的设计与实现。该部分的创建是以临钢十一辊全液压强力矫直机的二级监控系统为背景，该二级监控系统由自动化站、主操作站、主监控站、工程师站和服务器集群共同构成的分布式的计算机控制系统。本文以此为基础创建样本值管理系统，此系统具备管理应用样本表中现有样本值、通讯数据及实时数据的功能，也具有根据一定条件查询样本表中的样本值和添加、删除样本值的功能。（2）弯辊量自适应模型的设计和实现。该部分创建了由解析数学模型和神经网络自学习系统构成的具有自适应功能的智能弯辊模型。通过神经网络模型将历史数据及标准样本对系统的影响导入模型，提高了模型的精确性和适应能力。通过试验对神经网络自学习系统的传递函数、学习函数、训练函数和隐层神经元个数进行了设计，使弯辊量自适应模型具有学习速度快、精度高和泛化能力强的特性。（3）弯辊量自适应模型的应用。利用数据库技术、OPC技术、工业现场总线技术等，将弯辊量自适应模型、矫直机二级控制系统及基础自动化系统有机地结合起来，从而构建了完整的矫直机智能工艺设定系统。系统通过OPC，指挥矫直机监控系统控制PLC，再由PLC控制矫直机弯辊装置中的液压缸，进行对应的弯辊操作，以改善板型。整个系统在某厂的大型第三代全液压中厚板矫直机控制系统中投入应用。更多还原

【Abstract】 Because the plates are affected by many factors during the heating, rollingand cooling process，the rolled plates have shape defects, such as wave in thepart of plates、waves in the edge of plates、buckling. In order to eliminate defects,the roller must be equipment the bending parts. The bending system of the plateroller leveler is a key part which decides sheet quality and improves thestraightening efficiency. Because the straightening process is a multivariable,nonlinear, slowly time-varying, and strong coupling process. The model usedconventional techniques does not involve time-varying and coupling factors.Therefore, the paper combines the Existing model of bending value with neuralnetwork to construct the bending amount of models and neural networktechnology combining construct the adaptive model of bending value, whichmakes that have adaptive capacity and the ability of learning by itself. Thispaper designs a neural network as a bending roll control model. This articlecontains the following sections:(1) The design and implementation of bending value sample valuemanagement system. The creation’s background of this part is the monitoringsystems of11-high hydraulic powerful leveler of a factory. The two stepmonitoring system is consisted of the automation station, the main operatorstation, the main monitoring station, the engineer station and server clusterswhich together constitute the distributed computer control system. The papercreates the management system of sample value depend on the system. Thisprocedure interface owns the ability of managing sample value, communicationdata and real-time data, querying sample value from the sample table accordingto certain conditions and adding、deleting sample value in sample table.(2) The design and implementation of the adaptive model of Bending value.The part creates a intelligent roll bending model with adaptive function which ismade up of analytical mathematical model and neural network. The neuralnetwork model makes the impact of historical data and standard samples to thesystem imported into the model to improve the accuracy of the model and ability to adapt. After many experience, it designs the learning function, the transferfunction, the training function, and the number of neurons in the hidden layer ofneural network to make it own fast learning speed, high accuracy andgeneralization.(3) The application of the adaptive model of bending value. Using thedatabase technology, OPC technology, the industrial field-bus technologycombines the adaptive model of bending value with the monitoring system ofstraightening machine and the basic automation system, so that build acomplete process system of straightening machine. The system commands theleveler monitoring system to control the PLC by OPC, and then the PLCcontrols the roller and operate the bending rollers to improve the plate. Thesystem is applied in the control system of the large third-generation fullhydraulic plate straightening machine.更多还原