基于BP神经网络的烧结矿性能预报模型  

Prediction model of sinter properties based on BP neural network

作　　者：赵路朋 吴铿[1] 朱利[1,2] 陈小敏[1] 秦喧柯 

机构地区：[1]北京科技大学钢铁冶金新技术国家重点实验室,北京100083 [2]首秦金属材料有限公司炼铁部,河北秦皇岛066000

出　　处：《钢铁》2017年第9期11-15,共5页Iron and Steel

基　　金：国家自然科学基金资助项目(51474022)

摘　　要：为解决烧结矿预报模型中未考虑铁矿粉高温基础特性的情况,在预报模型中添加了反应铁矿粉高温性能的同化反应特征数,即流动性特征数。采用BP神经网络建立烧结矿性能预报模型。选择影响高炉生产的烧结矿指标作为输出,分析影响这些指标的烧结操作制度,铁矿粉的高温、物化特性作为输入;通过BP神经网络建立预测模型,并对BP神经网络的算法进行优化。预报模型采用8-17-4的BP神经网络结构,经过训练后,预测精度达到85%以上,具有很好的准确性和自适应性。To solve the problem of neglevting the high temperature characteristics of iron ore powder in sinter forecast model,the assimilation reaction characteristic number and liquidity characteristic number which reflect the high temperature performance of iron power are added into the model. BP neural network is used to establish prediction model of sinter performance. The sinter indexes that affects the production of blast furnace were chose as the output. The high temperature and the physical and chemical properties of the iron ore powder were analyzed as input. Thus,the prediction model was established by BP neural network,and the algorithm of BP neural network was optimized. The structure of BP neural network for prediction model was 8-17-4. After training the neural network,the prediction accuracy of the predicted characters was more than 85%,which meant that the neural network had good accuracy and adaptability.

关 键 词：铁矿粉 烧结矿性能 高温性能特征数 BP神经网络 预报模型 

分 类 号：TF046.4[冶金工程—冶金物理化学]