基于贝叶斯优化LightGBM的热轧中厚板终冷温度预测  

Prediction of final cooling temperature for hot rolled plate based on Bayesian optimized LightGBM

作　　者：王义铭 杜岩 张田 杜平 田勇[1] 王丙兴[1] WANG Yiming;DU Yan;ZHANG Tian;DU Ping;TIAN Yong;WANG Bingxing(State Key Laboratory of Rolling and Automation,Northeastern University,Shengyang 110819,China;School of Mechanical Engineering,Shenyang Jianzhu University,Shenyang 110168,China;Institute of Research of Iron and Steel,Shagang,Jiangsu Province,Zhangjiagang 215625,China)

机构地区：[1]东北大学轧制技术及连轧自动化国家重点实验室,辽宁沈阳110819 [2]沈阳建筑大学机械工程学院,辽宁沈阳110168 [3]江苏省(沙钢)钢铁研究院,江苏张家港215625

出　　处：《冶金自动化》2024年第1期54-64,共11页Metallurgical Industry Automation

摘　　要：中厚板热轧生产是典型的流程工业,依次要经历连铸、加热、除鳞、轧制、冷却、卷取等工艺过程,因冷却过程钢板温度变化幅度大、速率快的特点,使得冷却过程对钢板组织性能的影响最大,其中终冷温度是冷却过程的一项关键控制参数。为了提高终冷温度预测的精度,基于LightGBM(light gradient boosting machine)模型对终冷温度进行回归预测。以坯料尺寸、化学成分和上下游工艺参数作为模型的输入,终冷温度作为模型的输出,使用贝叶斯优化方法完成模型超参数调优;此外,使用Shapley加性解释(Shapley additive explanation, SHAP)方法检验输入参数对预测参数的影响程度。结果表明,贝叶斯优化LightGBM(BO-LightGBM)模型在训练集和测试集上均实现了较低的误差,95%的预测数据绝对误差控制在±10℃以内,相较其他集成学习模型,耗时最多减少了97%,同时提高了对中厚板热轧工艺流程温度的预测精度和预测效率。Plate hot rolling is a typical process industry,which goes through continuous casting,heating,descaling,rolling,cooling,coiling and other technological processes in turn.Because of the large range and fast speed of temperature change in the cooling process,the cooling process has the greatest influence on the microstructure and properties of steel plate,and the final cooling temperature is a key control parameter in the cooling process.In order to improve the accuracy of the final cooling temperature prediction,the LightGBM(light gradient boosting machine)model was used for regression prediction of the final cooling temperature.The size of plate,chemical composition and upstream and downstream process parameters are used as inputs of the model,and the final cooling temperature is used as output of the model.Bayesian optimization method is used to complete the super-parameter optimization of the model.In addition,Shapley additive explanation(SHAP)method is used to test the influence of input parameters on the predicted parameters.The results show that Bayesian optimized LightGBM(BO-LightGBM)model achieves lower error in both training set and test set,95%of the absolute error of predicted data is controlled at±10℃,and the time consumption of the model is reduced by 97%compared with other ensemble learning models,the prediction accuracy and prediction efficiency of hot rolling process temperature of plate are improved.

关 键 词：中厚板 终冷温度 机器学习 贝叶斯优化 LightGBM 

分 类 号：TG335[金属学及工艺—金属压力加工]