基于信息融合的钢包精炼炉元素收得率预报  

Element yield prediction in ladle furnace based on information fusion

作　　者：徐喆[1] 毛志忠[2] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110819 [2]东北大学流程工业综合自动化国家重点实验室,沈阳110819

出　　处：《仪器仪表学报》2012年第4期795-800,共6页Chinese Journal of Scientific Instrument

基　　金：国家高技术研究发展计划(2007AA041401;2007AA04Z194)资助项目

摘　　要：为提高元素收得率的预报精度,针对元素收得率主要影响因素无法及时获得的问题,提出一种基于信息融合的预报方法。该方法对原始数据层与特征层分别进行融合操作。在原始数据层,首先将相关性较为明显的可测变量按照机理进行结合,然后使用基于改进遗传算法的特征构建方法对由可测变量与结合变量组成的变量集合进行融合,使融合后的特征作为元素收得率主要影响因素的评价指标。在特征层,使用自适应神经-模糊推理系统对构建后的特征进行融合,从而得到元素收得率预报模型。将此方法应用于Q345B钢种Si、Mn元素收得率的预报,实验结果表明所提出的方法不仅可以满足生产要求,而且比其他预报方法有更高的预报精度。To improve the element yield prediction accuracy,a prediction method based on information fusion is proposed aiming at the problem that the main affecting factors of element yield cannot be obtained in time.This method carries out fusion operation for raw data level and feature level respectively.At raw data level,the measurable variables that have obvious correlations are combined according to the mechanism;then,improved genetic algorithm is used to fuse the variable set that consists of measurable variables and combined variables,so that the constructed features can be used to indicate the main affecting factors of element yield.At feature level,adaptive neuro-fuzzy inference system is used to fuse the constructed features;therefore,the element yield prediction model can be obtained.The proposed method is applied to establish yield prediction models for Si,Mn elements in Q345B steel.Experiment results show that the proposed method can meet the production requirements,and is more accurate than other prediction methods.

关 键 词：信息融合 钢包精炼炉 元素收得率 遗传算法 自适应神经-模糊推理系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]