一种基于深度神经网络的烧结料层透气性预报模型  

A prediction model of permeability of sintered material layer based on deep neural network

作　　者：赵小青 卜象平 刘颂 刘二浩 王丰 吕庆 ZHAO Xiao-qing;BU Xiang-ping;LIU Song;LIU Er-hao;WANG Feng;LU Qing(College of Mechanical Engineering,North China University of Science and Technology,Tangshan 063009,Hebei,China;Hangzhou Lie Technology Co.,Ltd.,Hangzhou 310000,Zhejiang,China;Department of Computer Science and Technology,Tangshan University,Tangshan 063000,Hebei,China;College of Metallurgy and Energy,North China University of Science and Technology,Tangshan 063009,Hebei,China;Technology Department,Chengde Steel Group Co.,Ltd.,Chengde 067000,Hebei,China)

机构地区：[1]华北理工大学机械工程学院,河北唐山063009 [2]杭州排列科技有限公司,浙江杭州310000 [3]唐山学院计算机科学与技术系,河北唐山063000 [4]华北理工大学冶金与能源学院,河北唐山063009 [5]承德钢铁集团有限公司技术部,河北承德067000

出　　处：《中国冶金》2020年第12期9-17,共9页China Metallurgy

基　　金：国家自然科学基金资助项目(U1360205)。

摘　　要：为了保证整个烧结生产过程的顺利进行,通过对承钢2号烧结厂数据库中的数据进行采集、整合和预处理,运用相关性分析筛选出与料层透气性相关的重要特征变量,基于深度神经网络算法建立料层透气性预报模型,以即时为现场操作人员提供科学的操作指导。通过测试并与随机森林模型、支持向量机模型进行性能对比,结果表明,深度神经网络具有很好的预测效果,可实现对料层透气性的精准预测,对优化烧结现场的操作效果具有很好的指导意义。In order to keep the entire sintering production process smooth,a sintering material layer gas permeability prediction model was established based on the historical data of the actual production of Chenggang No.2 sintering machine to provide immediate scientific operation guidance for field operators.By collecting,integrating and preprocessing the data in the database of Chenggang No.2 sintering plant,the correlation analysis was used to screen out the important characteristic variables related to the gas permeability of the material layer,and the gas permeability prediction of the material layer was established based on the deep neural network algorithm.Through testing and performance comparison with the random forest model and the support vector machine model,the results showed that the deep neural network had a very good prediction effect,achieved an accurate prediction of the permeability of the material layer,and had a good guiding significance for optimizing the operation effect of the sintering site.

关 键 词：烧结料层 透气性 预处理 相关性分析 深度神经网络 

分 类 号：TF046.4[冶金工程—冶金物理化学] TP183[自动化与计算机技术—控制理论与控制工程]