基于图形处理器加速的连铸坯实时凝固传热模拟  

GPU-accelerated real-time simulation of heat transfer and solidification process in continuous casting billets

作　　者：刘晓玉 刘文红[1] LIU Xiaoyu;LIU Wenhong(College of Information Science and Engineering,Northeastern University,Shenyang 110819,China)

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《冶金自动化》2022年第3期57-64,共8页Metallurgical Industry Automation

基　　金：国家自然科学基金青年科学基金项目(61703084)。

摘　　要：实时模拟连铸坯的传热过程对于保证二冷动态控制的精确性和压下效果十分重要。实际生产过程中,为了适应工艺参数的不断变化,连铸的动态控制系统需要调整二冷水量以保证铸坯表面温度稳定。对于基于模型的动态系统,其核心是准确的在线传热模型。然而,在线模型对求解时间的严苛要求往往使得精确的模型在实际应用中受限。针对这个问题,在保障模型计算精度的基础上,应用图形处理器(graphics processing unit,GPU)技术来加速连铸凝固传热过程模拟,并针对连铸凝固传热过程和GPU硬件特征,对求解算法进行设计和优化。与基于28核中央处理器(central processing unit,CPU)并行的模型相比,GPU加速的传热模型实现了78倍的加速比。此外,模型的准确性通过表面温度测量得到了验证,最大相对误差为0.728%。The real-time simulation of the heat transfer and solidification process in continuous casting billets plays a central role to ensure the accuracy of dynamic control of secondary cooling and the effect of compression.In the actual continuous casting process,the dynamic control system needs to adjust the secondary cooling water quantity to adapt to the change of process parameters and thus ensure a stable surface temperature of the strand,which depends on the accurate online heat transfer model.However,due to the strict requirements on solving time of online models,the accuracy of models is often limited in practical applications.Hence,GPU acceleration technology into the process simulation of continuous casting process was introduced.The accelerated model was implemented based on an algorithm,which was designed and optimized according to the characteristics of casting process and GPU architecture.Compared with the 28-core CPU-paralleled model,the proposed GPU accelerated heat transfer model achieves a 78 speed-up.In addition,the accuracy of the model is verified by surface temperature measurements,where the max relative error is 0.728%.

关 键 词：凝固传热 GPU 加速模拟 连铸 实时模拟 

分 类 号：TF777[冶金工程—钢铁冶金]