【作者】 薛涛；

【导师】 杜凤山；

【作者基本信息】 燕山大学 ， 机械设计及理论， 2014， 博士

【摘要】 冷轧板带轧制过程的控制水平对冷轧板带产品的质量起着决定性的作用。在研究冷轧板带轧制过程时，通常采用近似线性化的数学模型进行描述。但是在实际生产过程中，冷轧板带过程是多因素、强耦合、非线性的。本文采用非线性有限元法结合神经网络的非线性映射对冷轧板带轧制过程进行了分析研究，为冷轧板带轧制过程的精确、快速控制提供了基础。本文为了对冷轧板带轧制力进行精确、快速地预报，采用非线性有限元法建立了板带和轧辊耦合变形的二维动态弹塑性有限元模型。利用该模型分析了入口厚度、压下率、前张应力、后张应力、摩擦系数、变形抗力和工作辊直径等因素对轧制力的影响。采用神经网络建立了冷轧板带轧制力预报的网络模型，利用有限元模型的模拟结果对神经网络进行了训练，得到了能够精确、快速预报冷轧板带轧制力的网络模型。在冷轧板带轧制过程中，板形控制和板厚控制是耦合在一起的。对辊缝进行调整不仅会影响板厚也会影响板形，对弯辊力进行调整不仅会影响板形也会影响板厚。为了对板形板厚进行准确的设定控制，必须综合考虑板形板厚之间的影响关系，研究板形板凸度综合预设定策略。本文分析了六辊UCM轧机板形板凸度综合预设定所需的方程和模型，包括轧机刚度方程、板厚方程、板凸度方程、板形板厚耦合方程、中间辊横移位置设定模型、弯辊力设定模型和空载辊缝设定模型等，给出了板形板凸度综合预设定策略。采用非线性有限元法建立了板带和轧辊耦合变形的六辊UCM轧机三维有限元模型。利用该模型计算了轧制力的横纵向刚度、工作辊弯辊力的横纵向刚度、中间辊弯辊力的横纵向刚度，分析了中间辊横移对轧制力横纵向刚度的影响。采用有限元模型验证了板形板凸度综合预设定后的板形、板厚均满足目标要求。本文系统分析了入口厚度、压下率、前张应力、后张应力、摩擦系数、变形抗力、板带宽度、入口板凸度、工作辊弯辊力、中间辊弯辊力、中间辊横移量、工作辊直径、中间辊直径、支承辊直径等因素对板凸度的影响。为了对冷轧板带轧制过程中的板凸度进行快速、准确地预报，采用神经网络建立了冷轧板带板凸度预报的网络模型。为了对板带横向厚差进行控制，本文分析了板形调控功效的基本概念，给出了板带横向厚差调控功效的表达形式。推导了板形设定过程的板形目标曲线，即板带横向厚差目标曲线。利用有限元模型分析了工作辊弯辊、中间辊弯辊、中间辊横移对板带横向厚差、横向位移、轧辊弯曲和轧辊压扁的影响，建立了板带出口横向厚差的神经网络预报模型。利用有限元模型计算了工作辊弯辊、中间辊弯辊、中间辊横移的横向厚差调控功效，建立了横向厚差调控功效的神经网络预报模型。分析了板形设定的计算过程，采用等比例凸度法和调控功效法对板带板形进行了设定，比较了两种方法的计算结果和特点。分析了工作辊弯辊、中间辊弯辊、中间辊横移对边部减薄的影响，研究了边部减薄与板形调节量计算值的关系，对基于调控功效法的板形设定模型进行了改进，提出了基于中心板凸度和边部减薄量的板形设定模型。综合以上研究成果，在四机架六辊900UCM冷连轧机组上进行了工业实验并验证了本文所建立的轧制力预报模型、板凸度预报模型和有限元模型的正确性和准确性。更多还原

【Abstract】 In the cold-rolled strip rolling process, control level of product quality plays adecisive role. In general, the cold-rolled strip rolling process is researched by theapproximate linear mathematical model. But in the actual production process, thecold-rolled strip process includes multiple factors, strong coupling, nonlinear. In this paper,the nonlinear finite element method and the neural network method are combined toresearch cold-rolled strip rolling process. The combined method provides the foundationfor cold-rolled strip rolling process control accurately and quickly.In this paper, in order to forecast the rolling force accurately and quickly, atwo-dimensional dynamic model is established by nonlinear elastic-plastic finite elementmethod. The effects of entry thickness, reduction rate, forward tension, backward tension,friction coefficient, deformation resistance and work roll diameter on the rolling force areanalyzed by the model. The neural network of rolling force prediction is established. Theneural network model is trained by the simulation results of finite element model. Thetrained network model can forecast rolling force accurately and quickly.In the cold-rolled strip rolling process, shape control and gauge control are coupledtogether. Adjusting roll gap affects not only gauge but also shape. Adjusting roll bendingforce affects not only shape but also gauge. In order to set shape and gauge accurately, therelationship of shape set and gauge set should be researched. This paper systematicallyanalyzes the equations and models of UCM mill, including mill stiffness equations, gaugeequation, crown equation, shape and gauge coupling equation, intermediate roll shift setmodel, roll bending force set models and roll gap set model. The set strategy of shape andcrown integrated is presented. The three-dimensional dynamic UCM mill model isestablished by nonlinear elastic-plastic finite element method. The horizontal and verticalstiffness of the rolling force, the horizontal and vertical stiffness of the work roll bendingforce, the horizontal and vertical stiffness of the work roll bending force are calculated bythe model. The effects of the intermediate roll shift on the horizontal and vertical stiffnessof rolling force are calculated by the model. The set strategy of shape and crown integrated is verified by the finite element model. The gauge and shape meet therequirements after set.This paper systematically analyzes the effects of entry thickness, reduction rate,forward tension, backward tension, friction coefficient, deformation resistance, strip width,entry crown, work roll bending force, intermediate roll bending force, intermediate rollshift, work roll diameter, intermediate roll diameter and backup roll diameter on stripcrown. In order to forecast strip crown quickly and accurately, neural network is used toestablish a network model of cod-rolled strip crown forecast.In order to control the lateral thickness difference, this paper analyzes the basicconcept of actuator effectiveness and presents the lateral thickness difference actuatoreffectiveness expression. The lateral thickness target curve of shape set process is derived.The effects of work roll bending, intermediate roll bending, intermediate shift on striplateral thickness difference, lateral displacement, roll bending and roll flattening areanalyzed by finite element model. The neural network model of strip lateral thicknessdifference forecast is established. The lateral thickness difference actuator effectiveness ofwork roll bending, intermediate roll bending, intermediate shift are calculated by finiteelement model. The neural network model of strip lateral thickness difference actuatoreffectiveness forecast is established. The shape set process is analyzed. Shape is set byproportional crown method and actuator effectiveness method. The results andcharacteristics of the two methods are compared. The effects of work roll bending,intermediate roll bending, and intermediate shift on edge drop are analyzed. Therelationship of edge drop and adjustment amount is researched. The shape set methodbased on actuator effectiveness is improved. The shape set model based edge drop andcentral crown is presented.Based on above research findings, industrial experiments are conducted on six-high900UCM cold tandem mills. Rolling force prediction model, strip crown predictionmodel and FEM model are verified by the experiments.更多还原