非稳态轧制过程的有限元分析、板厚建模与智能控制

【作者】 段建辉；

【导师】 李涵雄； 邓华；

【作者基本信息】 中南大学 ， 机械制造及自动化， 2008， 硕士

【摘要】 板带轧制咬入过程是一个非稳态过程,轧制工艺参数的不稳定导致轧件头部厚度高于中间厚度,一般情况下该差值要大于100μm,尖锋值可达200—300μm,故而在考核质量时一般都除去头部若干米,只对本体部分进行考核。为了提高成材率,有必要研究非稳态轧制过程中轧件的变形特点,分析工艺参数对非稳态轧制过程的影响规律,建立非稳态轧制过程的数学模型,并研究一种新型的控制系统来缩短板带头部厚度不考核的长度。本文的主要研究工作如下:(1)在研究热力耦合刚塑性有限元基本理论的基础上,利用MSC.MARC非线性有限元软件,建立铝板带非稳态轧制过程的三维有限元模型。该模型不仅考虑了轧辊的弹性变形,而且还考虑支撑辊对轧件变形的影响,此外还考虑了轧件与乳液对流换热的边界条件;(2)利用三维大变形热力耦合刚塑性有限元法对非稳态轧制过程进行模拟,并将仿真结果与实际生产数据进行比较,结果表明本文所采用的方法对非稳态轧制过程进行仿真是可行并且可靠的。在此基础上研究了非稳态阶段的轧制力分布特点和轧件的变形特点,分析了轧件表面温度对轧件“头部厚跃”现象的影响规律;(3)在前述研究基础上,分析了轧制温度、前后张力等工艺参数对非稳态轧制过程中厚度的影响规律,提出了缩短板带头部厚度不考核长度的措施;(4)利用有限元分析所得结果,选择合理的工艺参数作为神经网络的输入输出参数,建立了非稳态轧制阶段轧件出口厚度的BP神经网络预测模型。模型计算的相对误差控制在1.200%以内,平均相对误差为0.507%,说明利用BP神经网络建立的模型对非稳态轧制阶段的出口厚度进行预测具有很高的精度。(5)研究了神经网络PID控制器对非稳态轧制过程板厚控制的控制性能,仿真结果表明该控制系统收敛快,稳定性强,超调量小,具有良好的控制性能,将非稳态轧制过程的时间从原来的6s缩短为3s,具有较高的工程应用价值。更多还原

【Abstract】 Biting stages is unsteady process in the plate rolling and the unsteady process parameters result in that the head and tail is thicker than the middle.Generally speaking,the difference is more than 100μm and the maximum value may be up to 200-300μm.Therefore,only main part of the rolled piece is examined for quality.In order to increase the yield and reduce the unsteady process,it is essential to study the deformation characteristic of the unsteady rolling,analyze the effect of process parameters on the unsteady rolling process,set up mathematical model of the unsteady process,and propose a new control system to improve control precision,The main research work in this thesis includes:(1)Based on the theory of rigid plastic finite element,the 3-D finite element model of unsteady process was built with nonlinear finite element analysis software MSC.MARC.The elastic deformation of the roll and the influence of emulsion were considered in simulation.(2)The unsteady process of the plate rolling was simulated by using three-dimension large deformation thermo-mechanical coupled rigid-plastic finite element methods.The simulated result was compared with the real rolling data,which showed that the method used to simulate the unsteady rolling process was feasible and reliable.The rolling force distribution and the rolled piece deformation in the bite stage were studied and the rolled piece temperature influence on strip-head bending was analyzed.(3)Based on the proposed method,the effect of process parameters such as rolling temperature and tension on the unsteady rolling process was analyzed.Moreover,the way to shorten the unqualified length of the plate head was presented.(4)Based on the simulation results of FEM,the main process parameters were selected as the inputs and outputs of a neural network, and a BP neural network prediction model of.unsteady state rolling was built.The maximum calculated error of the model was 1.200%and the average relative error was 0.507%,and showing that the accuracy of the prediction model was high. (5)A neural network PID controller was built to control the gauge of plate head.Simulation results showed that this control system has the characteristics of fast convergence,strong stability and small overshoot. The unsteady state rolling was reduced from 6s to 3s by using the controller that has good performance for engineering application.更多还原