摘要
针对转炉的动态吹炼过程,建立了基于BP神经网络的转炉炼钢总吹氧量预测模型和二次吹氧量预测模型。通过相关性分析确定影响转炉总吹氧量和二次吹氧量预测的主要因素;利用五数概括法筛选数据;采用LM优化算法改进BP神经网络。并利用历史生产数据对不同拓扑结构的神经网络模型进行了训练和比较,确定了最优网络结构模型,对模型的性能进行了评价,总吹氧量预测模型预测误差小于800 m3的命中率达到87.88%,二次吹氧量预测模型预测误差小于400 m3的命中率为91.99%。
In light of the dynamic blowing process of converter a BP neural network based prediction model is proposed for the purpose of determination of total blow oxygen and end blow oxygen in converter steel-making. Critical influence factors for total blow oxygen and end blow oxygen prediction can be determined by correlation analysis. Fivenumber summary method is used to discriminate the data. BP neural network is improved by LM algorithm. Then, several neural network models with different topology structure are trained and compared with the history data,and the best network topology is selected. The performance of the best neural network model is evaluated. The results show that the hit rate to total blow oxygen predicted by model is 87. 88 % with control precision among 800 m3, and the hit rate to end blow oxygen predicted by model is 91.99 % with control precision among 400m3 respectively.
出处
《炼钢》
CAS
北大核心
2013年第2期34-37,41,共5页
Steelmaking
关键词
转炉
BP神经网络
吹氧量
预测
converter
BP neural network
blow oxygen
prediction
