【作者】 单修迎；

【导师】 刘宏民； 贾春玉；

【作者基本信息】 燕山大学 ， 机械设计及理论， 2011， 博士

【摘要】 本文以人工智能理论为基础,选择具有理论和工程实际意义的冷轧带钢板形控制的矩阵模型研究为课题,对平直度模式识别、平直度控制和断面形状控制进行了深入的理论研究,取得了新的研究成果。平直度模式识别是平直度控制系统的重要组成部分。考虑到现代轧机平直度控制手段多样化的实际,为提高平直度模式识别模型的精度,提出了含有3次分量的新型平直度模式识别方法。该模型采用基于最小二乘原理的勒让德多项式回归方法,使用1次、2次、3次和4次勒让德多项式作为平直度基本模式,利用小波消噪技术对平直度原始数据进行预处理,从整体上提高了平直度模式识别的精度,使平直度模式识别结果具有完备性,细化了平直度控制目标,有利于提高平直度控制的精度。为了分析各种平直度控制手段对平直度控制影响矩阵的影响规律,以HC轧机为例,以倾辊、工作辊弯辊、工作辊非对称弯辊和中间辊横移作为平直度控制手段,按照机理模型计算了各控制手段对1次、2次、3次和4次平直度分量的影响系数,系统地揭示了各种控制手段对各次分量的影响规律,为平直度在线控制模型的建立、实现提供了基础。为了提高平直度控制模型的精度,综合运用机理和智能建模方法,从轧制过程的本质和实测数据两个方面出发,建立了机理-智能型动态影响矩阵法平直度控制模型。采用微粒群算法优化的BP网络建立了机理影响矩阵快速计算网络模型,实现了机理影响矩阵的快速计算,避免了由于机理模型庞大复杂、计算时间长而不能满足在线快速计算需要的问题。采用聚类算法优化的RBF网络建立了实测影响矩阵快速计算网络模型,用现场实测影响矩阵数据作为样本,进行在线自学习,使平直度在线控制模型动态跟踪现场实际情况。将机理控制矩阵和实测控制矩阵有机结合,制定了平直度控制量计算方案,提高了平直度在线控制模型的通用性、灵活性和精确性。为了提高带钢断面形状的控制精度,首次提出了带钢断面形状影响矩阵控制模型。该模型采用1次、2次、3次和4次曲线精确描述带钢断面形状,根据影响函数法的原理,给出了断面形状控制影响矩阵的概念,并且建立了断面形状控制方程,为带钢断面形状在线控制提供了新的方法。应用机理模型计算了各种控制手段对断面形状的影响系数,系统分析了各种控制手段对断面形状的影响规律,为制定带钢断面形状控制策略提供了依据。为实现断面形状控制影响矩阵的快速计算,应用BP网络建立了带钢断面形状快速预报模型,使带钢断面形状影响矩阵控制模型能够方便地应用于在线断面形状控制。为了提高连轧机的平直度和断面形状控制能力,综合应用机理-智能型动态影响矩阵法平直度控制模型和断面形状影响矩阵控制模型,制定了连轧机平直度和断面形状综合控制方案,使在线平直度控制和断面形状控制相互配合,实现了各机架的在线并行控制。通过某1550连轧机实测平直度和断面形状数据对该控制方案进行了工业验证,结果表明该方案能具有较高的平直度和断面形状控制精度,控制过程稳定、可靠,提升了连轧机在线控制能力,对提高连轧机产品质量有重要意义。选择具有理论和工程实际意义的冷轧带钢板形控制的矩阵模型研究为课题,对平直度模式识别、平直度控制和断面形状控制进行了深入的理论研究,并进行了相应的仿真分析和工业验证,取得了新的研究成果,对平直度和断面形状控制理论的发展有重要意义。更多还原

【Abstract】 In this paper, author chooses research on matrix model of shape control for cold strip mills as the research subject and done deep theory research on flatness pattern recognition, flatness control and section configuration control, achieving new research findings. Flatness pattern recognition is an important part of flatness control system. Considering the practical situation of modern mills with many different flatness control means, a new flatness pattern recognition method including the cubic flatness is brought forward. Using the Legendre polynomial regression method based on least square theory, the model processes with the original flatness data by wavelet de-noising techniques with linear, quadratic, cubic and biquadratic Legendre orthogonal polynomials as basic patterns. The model increases the precision of flatness pattern recognition, makes the results of flatness pattern recognition self-contained, refines flatness control target and makes for increasing the precision of flatness control.Taking HC mill for example and using tilting roll, bending work roll, asymmetry bending work roll and shifting intermediate roll as flatness control means, the effective coefficients on the linear flatness, the quadratic flatness, the cubic flatness and the biquadratic flatness for different control means are calculated by theory model in order to analyse the effective rules on flatness control effective matrix for different flatness control means, which opens out the effective rules on different order flatness for different control means and provides the foundation for the building and realizing on-line flatness control.In order to increase the precision of flatness control, considering the direction of rolling process essence and the direction of measured data, the theory-intelligent dynamic effective matrix flatness control model is built by using theory and intelligent methods synthetically. The network model for rapid calculating theory effective matrix is built by the BP network optimized by particle swarm algorithm, realizing the rapid calculation of theory effective matrix and avoiding the problem that theory model can not meet the need of rapid on-line calculation. The network model for rapid calculating survey effective matrix is built by the RBF network optimized by cluster algorithm, realizing the rapid calculation of survey effective matrix. The model can track practical situations by on-line self-learning with the measured effective data as samples. The scheme for flatness control quantity calculation is established by combing theory control matrix and survey control matrix, improving the commonality, flexibility and precision of on-line flatness control model. In order to increase the precision of strip section configuration control, the effective matrix control mode for strip section configuration control is brought forward for the first time. Using linear, quadratic, cubic and biquadratic orthogonal polynomials describing strip section configuration precisely, the model provides the conception of the effective matrix for section configuration control and builds the control equation for section configuration control by the theory of effect function method, which provides a new method for strip section configuration on-line control. The effective coefficients on section configuration for different control means are calculated by theory and the effective rules on section for different control means are analysed, which provides basis for establishing strip section configuration control strategy. In order to realize rapid calculating the effective matrix for section configuration control, the rapid prediction model for strip section configuration is built by BP network, making the effective matrix model for strip section configuration can be used in on-line section configuration control conveniently.Using theory-intelligent dynamic effective flatness control model and section configuration effective matrix control model synthetically, flatness and section configuration integration control scheme is built in order to enhance flatness and section configuration control capability of tandem mill, which makes on-line flatness control cooperate with on-line section configuration control and realizes the parallel control of different stands. According to the measured flatness and section configuration data of the 1550 tandem mill, the control scheme is validated and the result indicates that the scheme has high flatness and section configuration control precision with a steady and reliable control process and upgrades the control capability of tandem mill with much significance for increasing product quality.Choosing research on matrix model of shape control for cold strip mills as research subject, author has done deep theory research on flatness pattern recognition, flatness control and section configuration control and done relevant simulation analysis and industry validation. New research findings are acquired, which has much significance for the development of flatness and section configuration control theory.更多还原