摘要
通过脱磷热力学并结合灰色关联理论系统研究了转炉终点磷含量的影响因素,筛选出终点磷含量预测模型的最佳输入维度,建立BP神经网络模型和粒子群算法(Particle Swarm Optimization)优化BP神经网络模型(PSO-BP)。基于国内某钢厂生产的70号钢种200组实际生产数据的训练和预测,获得结论如下:PSO-BP模型的误差范围更小,命中率更高,转炉终点磷质量分数预测平均相对误差为3.55%,绝对误差在0.0002%以内的命中率为30%,误差在0.0004%以内的命中率为60%,误差在0.0008%以内的命中率达到80%。对钢铁企业实际生产中转炉终点磷含量的控制及预测具有较高的指导意义。
In this paper,the influencing factors of the phosphorus content at the end point of converter were systematically studied by dephosphorization thermodynamics combined with the gray relational theory.The optimal input dimension of the prediction model for the end point phosphorus content was screened out,and the BP neural network model and particle swarm optimization(PSO)were established to optimize the BP neural network model(PSO-BP).Based on the training and prediction of 200 sets of actual production data of No.70 steel produced by a domestic steel plant,the conclusions were as follows.The PSO-BP model had a smaller error range and a higher hit rate,and the average relative error of the phosphorus content prediction at the end of converter was 3.55%.The hit rate was 30%if the absolute error was within 0.0002%,60%if the error was within 0.0004%,and 80%if the error was within 0.0008%.It had high guiding significance for the control and prediction of the phosphorus content at the end point of converter in the actual production of iron and steel enterprises.
作者
杨城
包燕平
顾超
王达志
YANG Cheng;BAO Yanping;GU Chao;WANG Dazhi(State Key Laboratory of New Technology of Iron and Steel Metallurgy,University of Science and Technology Beijing,Beijing 100083,China)
出处
《炼钢》
CAS
北大核心
2023年第5期27-32,共6页
Steelmaking
关键词
磷含量
粒子群算法
灰色关联分析
BP神经网络
phosphorus content
particle swarm algorithm
grey relational analysis
BP neural network
