CAREER: Highly-Efficient Dynamic Prediction Models for Quality Improvement in Cold Rolling

该学院早期职业发展（CAREER）计划奖支持基础研究，为显着提高冷轧米尔斯生产的铝，钢，铜和黄铜片的质量奠定基础。这些金属板是许多产品的重要原材料，包括飞机，汽车，船舶，冰箱，计算机，电动机和建筑物。几何质量差的轧制金属板需要经历昂贵的附加加工。从这项研究的结果将导致新的轧机控制策略，提高轧制质量和生产率，这些金属板在重要的新兴市场。这项研究将有利于美国的经济竞争力和安全。来自代表性不足的群体（包括妇女和自闭症患者）的工程专业学生直接参与制造业将为教育和社会带来积极的效益。薄铝、钢、铜和黄铜冷轧板的几何缺陷是由于来料带材厚度轮廓与轧机咬入轮廓不匹配而产生的。详细的咬辊行为源于轧机结构偏转、热膨胀、轧辊磨削和轧辊磨损的复杂瞬态效应，并且难以预测。在这项工作中创建的高效的动态模型将集成结构，热，和磨损模式，使用一种新的，混合有限元方法来确定三维，高频非线性动态响应的轧制米尔斯。与单自由度或全尺寸有限元模型相比，该模型将允许对轧机动态进行更有效和更准确的多自由度预测。高效的非线性、多自由度动态模型还将与贝叶斯/马尔可夫链蒙特-卡罗控制方法相结合，以更好地理解磨机行为，尽管存在显著的随机过程变化。所获得的理解可能会导致新的轧机轧辊磨削方法，以纠正几何缺陷，新颖的二维（而不是中心线只有）厚度控制，和概率的方法，以提高轧制金属的几何质量。

This Faculty Early Career Development (CAREER) Program award supports fundamental research to lay the foundation to significantly improve the quality of aluminum, steel, copper, and brass sheets manufactured on cold rolling mills. These metal sheets are important raw materials for many products, including aircraft, automobiles, ships, refrigerators, computers, electric motors, and buildings. Rolled metal sheets of poor geometric quality need to undergo costly additional processing. Results from this research will lead to new mill control strategies that improve rolling quality and productivity for these metal sheets in important emerging markets. The research will benefit the economic competitiveness and security of the United States. The direct involvement of engineering students from the under-represented groups (including women and persons with autism) in manufacturing will provide positive benefits to education and society.Geometry defects in cold rolling of thin aluminum, steel, copper, and brass sheets arise from the mismatch of incoming strip thickness profiles with mill roll-bite profiles. Detailed roll-bite behavior stems from complex transient effects of mill structural deflection, thermal expansion, roll grinding, and roll wear, and is difficult to predict. The highly-efficient dynamic models to be created in this work will integrate structural, thermal, and wear patterns using a novel, mixed finite element approach to determine three-dimensional, high-frequency nonlinear dynamic responses of rolling mills. The models will allow for more efficient and accurate multi degree-of-freedom prediction of mill dynamics than is possible with single degree-of-freedom or full-scale finite element models. The efficient nonlinear, multi degree-of-freedom dynamic models will also be combined with Bayesian/Markov-Chain Monte-Carlo control approaches, to yield a better understanding of mill behavior in spite of significant random process variations. The understanding gained may lead to new on-mill roll grinding methods to correct geometry defects, novel two-dimensional (instead of centerline-only) gauge control, and probabilistic approaches to improve geometric quality of rolled metals.