260t转炉终点磷含量预报模型的开发实践  

Development and application of prediction model of end point phosphorus content in 260t converter

作　　者：朱坦华[1] 冯春松[1] 何飞[2] 

机构地区：[1]邯钢集团邯宝钢铁有限公司炼钢厂,河北邯郸056002 [2]北京科技大学冶金与生态工程学院,北京100083

出　　处：《炼钢》2012年第2期10-13,23,共5页Steelmaking

摘　　要：针对邯钢集团邯宝钢铁有限公司西区炼钢厂转炉的冶炼工艺特点和生产数据,建立了基于PCA-GA-BP神经网络的转炉终点磷含量预测模型。通过主成分分析(PCA)将终点磷含量的影响因素降维,并采用遗传算法(GA)对BP神经网络的初始权重进行优化。用Java语言开发了转炉终点磷含量预测模型的软件,在炼钢厂进行了现场使用。结果表明:转炉终点钢水w(P)控制精度在±0.007%时,命中率达到96.67%;控制精度在±0.005%时,命中率达到93.33%;控制精度在±0.004%时命中率达到86.67%。In light of process characteristics of the converter refining and the production data of the Steelmaking plant of hanbao Iron ＆ Steel Co. , Ltd. of Han Steel Group a PCA-GA-BP neural network based end point phosphor content prediction model for the converter is developed. Principal component analysic （PCA） is used to reduce dimension of the factors affecting the end point phosphor content and at the same time the genetic algorithm（GA） is applied for optimization of the initial weights of BP neural network. Then, a software of prediction model of end point phosphor content is developed by Ja va. And the software is applied in the converter mill. The results show that the hit rate to phosphor content predicted by model reaches 96.67 %, 93.33 % and 86.67 % with control precision among ± 0. 007 %, ± 0. 005% and ± 0. 004 % for heats of the experiment respectively.

关 键 词：转炉 主成分分析 BP神经网络 遗传算法 终点磷含量 

分 类 号：TF713[冶金工程—钢铁冶金]