灰铸铁抗拉强度预测的局部加权线性回归建模  

Locally weighted linear regression modeling for tensile strength prediction of gray cast iron

作　　者：任明仑[1] 宋月丽 褚伟[1] Ren Minglun;Song Yueli;Chu Wei(Key Laboratory of Process Optimization and Intelligent Decision Making of Ministry of Education,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学教育部过程优化与智能决策重点实验室,合肥230009

出　　处：《电子测量与仪器学报》2019年第3期65-71,共7页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(71531008)资助项目

摘　　要：为实现在生产现场对灰铸铁抗拉强度进行快速、准确的预测控制,引入非参数化的局部加权线性回归建模方法,过程参数的选择结合铁水热分析仪检测的碳、硅含量及光谱仪检测的锰、磷、硫等主要化学成分值。从安徽一家大型铸造厂的实际生产中采集100多炉铁水化学成分与对应抗拉强度数据,经预处理后,分成训练集和验证集,进一步从新的生产批次采集检测数据组成测试集;并从文献中收集40组数据组成另一数据集,进行两组对比验证实验。将局部加权线性回归模型,与目前灰铸铁强度预测中常用的多元线性回归模型和BP神经网络模型对比,验证了该方法能够达到更高的预测精度,且其非参数化的建模方式能够更好地适应生产现场复杂多变的工况环境。In order to realize fast and accurate predictive control of the tensile strength of gray cast iron in production site,a non-parametric locally weighted linear regression modeling method was introduced.The selection of process parameters combined the content of carbon and silicon detected by hot metal thermal analyzer with other main chemical compositions(such as manganese,phosphorus and sulfur)detected by spectrometer.From the actual production of a large foundry in Anhui Province,the data including chemical compositions and tensile strength values of more than 100 furnaces of hot metal were collected.After being preprocessed,the data were divided into training set and verification set,and the testing set was further composed of the data collected from a new production batch.In addition,40 groups of data were collected from a literature to form another data set,and two groups of comparative validation experiments were carried out.By comparing the locally weighted linear regression(LWLR)modeling method with the multivariate linear regression model and with the BP neural network model commonly used in gray cast iron strength prediction at present,it is proved that the LWLR can achieve higher prediction accuracy,and it can better adapt to the complex and changeable working environment of the production site as a non-parametric modeling method.

关 键 词：铸铁 抗拉强度 局部加权线性回归 性能预测 质量控制 

分 类 号：TG143.2[一般工业技术—材料科学与工程]