基于k-means聚类算法的电弧炉终点碳预报  

Study of end-point carbon prediction in EAF based on k-means clustering algorithm

作　　者：杨凌志[1] 朱荣[1] 秦云广 宋水根[3] 马国宏[1] 魏光升 

机构地区：[1]北京科技大学冶金与生态工程学院,北京100083 [2]杭州钢铁集团公司 [3]新余钢铁集团有限公司

出　　处：《冶金自动化》2014年第5期22-26,共5页Metallurgical Industry Automation

摘　　要：为了实现对电弧炉冶炼过程碳质量分数的预测,根据炼钢过程的碳氧反应机理建立电弧炉脱碳模型。在此基础上采用数据挖掘技术中的k-means聚类算法对电弧炉炼钢历史数据进行分析,选取8个影响终点碳质量分数的因素,得出不同冶炼情况下的聚类结果。通过计算当前炉次与聚类结果加权欧氏距离,将相似度高的聚类结果炉次作为当前炉次的预测参考炉次,最终实现对电弧炉终点碳质量分数的预测。仿真结果表明钢水碳质量分数预报的命中率在75%以上,模型具有较高的预报精度。In order to forecast the carbon content in arc furnace smelting process, decarburization model of EAF is established based on the C-O reaction mechanism in the steelmaking process. The steelmaking history data of EAF are analyzed by using the k-means clustering algorithm,which is be-longing to the data mining technology. Clustering results in different situations are found out by selec-ting eight factors affecting the end-point carbon content. By calculating the weighted Euclidean dis-tance between the current furnace and clustering results, the furnace which clustering results are of high similarity is used as the reference for the current furnace prediction,and the end-point carbon content prediction for EAF is finally realized. Simulation results show that the hit rate of the carbon content prediction for molten steel is above 75%,which means the model has high precision of fore-cast.

关 键 词：聚类算法 终点碳 电弧炉 脱碳模型 

分 类 号：TF748.41[冶金工程—钢铁冶金]