基于时域卷积网络的精轧出口厚度预测  

Thickness Prediction for Precision Rolling Exit Based on Time Domain Convolutional Network

作　　者：杨萍萍 马亮[2] YANG Pingping;MA Liang(School of Advanced Engineering,University of Science and Technology Beijing,Beijing 100083,China;School of Automation and Electrical Engineering,University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]北京科技大学高等工程师学院,北京100083 [2]北京科技大学自动化学院,北京100083

出　　处：《矿冶工程》2024年第1期138-142,共5页Mining and Metallurgical Engineering

基　　金：国家自然科学基金(62003030)。

摘　　要：以精轧过程为研究对象,引入时域卷积网络算法,构建了基于时域卷积网络的精轧出口厚度预测模型。利用时域卷积网络模型提取精轧过程时序数据的特征信息,通过优化模型结构和参数,提升精轧出口厚度预测性能。实际钢种数据集仿真实验结果表明,相较于传统方法,本文所提出的时域卷积网络算法在均方根误差、平均绝对百分比误差及决定系数等评价指标方面存在较大优势,可为现场工程师提供重要的决策信息。As for the precision rolling process,a thickness prediction model was constructed for precision rolling exit by introducing a time domain convolutional network algorithm.The feature information of time-series data of the precision rolling process was extracted by using this time-domain convolutional network model,and the prediction performance of the precision rolling exit thickness was improved by optimizing the structure and parameters of the model.The simulation results of the actual steel dataset show that the proposed time-domain convolutional network algorithm,compared to traditional methods,has significant advantages in evaluation indicators,such as root mean square error,average absolute percentage error,and coefficient of determination,which can provide critical information for decision of on-site engineers.

关 键 词：带钢 热轧 厚度预测 时域卷积网络 精轧过程 时序数据 特征提取 均方根误差 

分 类 号：TG335[金属学及工艺—金属压力加工] TP29[自动化与计算机技术—检测技术与自动化装置]