废品
炼钢
产量(工程)
铸铁
随机森林
人工神经网络
过程(计算)
冶金
碱性氧气炼钢
均方误差
原材料
工艺工程
相关系数
工程类
材料科学
统计
数学
计算机科学
人工智能
化学
有机化学
操作系统
作者
Chaojie Zhang,Yi Nian,Liqiang Zhang,Jinjun Cheng,Zhen Zhang
标识
DOI:10.1002/srin.202400713
摘要
Steel scrap is a primary raw material in basic oxygen steelmaking. However, its yield is influenced by numerous factors, making accurate prediction challenging. This study explores and predicts the steel scrap yield in the basic oxygen steelmaking process using machine learning techniques. First, the interquartile range method is applied to clean the collected steelmaking process data. By analyzing the blow loss of molten iron and the amount of steel obtained from the scrap, a deviation coefficient of scrap yield is defined and calculated. Next, a correlation analysis and a feature importance analysis using the random forest algorithm identify the factors influencing the deviation coefficient of scrap yield. Finally, a multilayer neural network regression model is constructed to predict the deviation coefficient of scrap yield. The model achieves a mean squared error of 0.00051 on the test set, with an accuracy rate of 96.89% for absolute errors within ±0.05. This method not only effectively predicts scrap yield but also provides a reference for calculating steel materials and controlling costs in the steelmaking process.
科研通智能强力驱动
Strongly Powered by AbleSci AI