基于机器学习的砂型铸造涂料悬浮率的预测  

Suspensibility Prediction of the Foundry Coating Based on Machine Learning

作　　者：翟秀云 陈明通[2] ZHAI Xiu-yun;CHEN Ming-tong(School of Intelligent Manufacturing,Panzhihua University,Panzhihua Sichuan 617000,China;Public Experiment Center,Panzhihua University,Panzhihua Sichuan 617000,China)

机构地区：[1]攀枝花学院智能制造学院,四川攀枝花617000 [2]攀枝花学院公共实验中心,四川攀枝花617000

出　　处：《铸造设备与工艺》2021年第3期35-38,共4页Foundry Equipment & Technology

基　　金：四川省钒钛材料工程技术研究中心项目(2020-2FTGCYB-01)。

摘　　要：在铸件浇注过程中,涂料是钢液和型腔壁的重要分隔层,其性能对铸件的外观及内部质量有很大的影响。本文中,基于机器学习方法和文献中的数据集,建立了预测铸造涂料悬浮率的支持向量回归(SVR)和反向传输神经网络(BPNN)模型。交叉验证与外部测试验证结果表明,两个模型都具有很高的预测精度和实用价值。本文通过对已有的铸造涂料数据的训练学习,实现了对未知涂料悬浮率的精确预测。因此,本文的研究方法可以为实现低成本、快速而有效地预测铸造涂料性能提供有价值的参考。During casting,foundry coating is an important separation layer between molten steel and cavity wall because its performance has a great influence on the appearance and internal quality of casting.In this work,the support vector regression(SVR)and back propagation neural network(BPNN)models were established to predict suspensibility of foundry coating based on machine learning methods and data set collected from the literature.According to the results of cross validation and external test,the two models were shown to be promising and practically feasible in accurately predicting suspensibility.The research method in the paper can reveal valuable clues for fast and effective prediction of foundry coating properties.

关 键 词：铸造涂料 悬浮率 支持向量回归 反向传输神经网络 机器学习 

分 类 号：TG221[金属学及工艺—铸造]