摘要
高炉铁水的硅含量是描述铁水质量的一个重要指标。为了在出铁之前了解铁水中硅含量的高低,建立预测模型是必要的。结合遗传算法(GA)和BP神经网络,建立了优化的GA-BP预测分析模型,从某高炉选取生产数据进行学习和预测。运行结果表明,模型具有较高的预测精度,当要求绝对误差为±0.05时,命中率可达70%;绝对误差为±0.08时,命中率可达92.3%。同时,应用该模型分析回归了高炉风量、热风压力、富氧量与铁间料批数等参数与铁水硅含量之间的相关关系,其结果与高炉冶炼理论基本吻合,可为高炉生产提供一定的指导。
Silicon content is an important index to describe hot metal quality. Building a model is necessary to predict silicon content. Combined genetic algorithms （GA） and back-propagation neural network （BP）, an optimized GA-BP model was established to predict silicon content. Some data were chosen to train the network model. The results showed that the model had higher accuracy, when required absolute error was within ±0.05, the accuracy of model can reach 70% and when absolute error was within ±0.08, the accuracy can reach 92.3%. At the same time, the relation between some operating parameters and silicon content had been analyzed, such as blast volume, blast pressure, charging batch, etc, the results were consist with ironmaking theory, and these can provide theoretical basis for ironmaking production.
出处
《钢铁研究学报》
CAS
CSCD
北大核心
2007年第11期60-62,F0003,共4页
Journal of Iron and Steel Research
基金
国家自然科学基金资助项目(59974006)
关键词
遗传算法
BP神经网络
硅含量
预测
genetic algorithm
back-propagation neural network
silicon content
prediction
