正在加载图片...
D0L:10.13374f.issn1001-053x2012.03.012 第34卷第3期 北京科技大学学报 Vol.34 No.3 2012年3月 Journal of University of Science and Technology Beijing Mar.2012 基于案例推理预测精炼开始钢水温度 汪红兵1,2》回 艾立翔》徐安军) 田乃媛》 候志昌) 周正文) 1)北京科技大学计算机与通信工程学院,北京1000832)北京科技大学钢铁流程先进控制教有部重点实验室,北京100083 3)北京科技大学治金与生态工程学院,北京1000834)上海安可科技有限公司,上海200433 ☒通信作者,E-mail:wanghongbing0816@163.com 摘要针对B即神经网络训练时间长的问题,采用基于案例推理的方法预测精炼开始钢水温度.首先,应用层次分析法确定 影响精炼开始钢水温度的各个因素的权值,并使用灰色关联度来计算案例的相似度,克服了传统相似度计算方法在案例信息 不完整的情况下无法获取准确结果的缺点.然后,提出一个包含类选、粗选、精选和择优的四步检索方法,大大缩短了检索时 间.最后,实验比较了人工神经网络和基于案例推理两种方法,结果表明基于案例推理比人工神经网络具有更高的命中率. 关键词炼钢:精炼:温度:预测:基于案例推理 分类号TF703.5:TP391.9 Prediction on the starting temperature of molten steel in second refining by using case-based reasoning WANG Hong-bing,Al Li-xiang,XU An-jun,TIAN Nai-yuan,HOU Zhi-chang,ZHOU Zheng-wen 1)School of Computer and Communication Engineering,University of Science and Technology Beijing,Beijing 100083,China 2)Key Laboratory of Advanced Control of Iron and Steel Process (Ministry of Education of China),Beijing 100083,China 3)School of Metallurgical and Ecological Engineering.University of Science and Technology Beijing,Beijing 100083,China 4)Shanghai Anchor Technology Co.Ltd.,Shanghai 200433,China Corresponding author,E-mail:wanghongbing0816@163.com ABSTRACT Case-based reasoning was used to predict the starting temperature of molten steel in second refining so as to avoid the long training time of a BP(back propagation)neural network.Analytic hierarchy process (AHP)was applied to determine the weights of factors influencing the starting temperature.Grey relational degree was adopted to compute the similarity between cases.Thus the shortcoming of difficulty in obtaining accurate cases with incomplete information is conquered.A four-step search method,including class search,rough search,delicate search,and optimized search,was provided,by which the search time decreases greatly.Experi- mental results using both artificial neural networks and case-based reasoning were compared.It is shown that case-based reasoning has got a higher hit rate and a shorter response time than artificial neural networks. KEY WORDS steelmaking;refining:temperature;prediction:case-based reasoning 二次精炼工艺,如LF、RH和CAS是炼钢连铸 期偏长,钢水温度下降较大和控制精度降低等不利 生产流程中不可缺少的环节.特别对于品种钢治 后果.因此,治金研究人员一直在探索应用在线模 炼,二次精炼是调节钢水温度、合金化、脱氧和去除 型预测炼钢连铸流程中各个关键点的钢水温度,这 夹杂物的重要手段.精炼开始的钢水温度是精炼工 些关键点包括转炉或电炉出钢时刻、精炼开始时刻、 艺的重要初始参数.但是,当前的技术条件下很难 精炼结束时刻和钢水到达连铸中间包时刻等. 实现钢水温度的连续测量.钢水温度的获得一般依 Fernandez等m使用人工神经网络(artificial 赖于事后的点测,这种事后处理方法造成了处理周 neural network,ANN)作为分类器,并设计模糊推理 收稿日期:20110202 基金项目:“十一五”国家科技支撑计划重大项目“新一代可循环钢铁流程工艺技术”(2006BAEO3A07):中央高校基本科研业务费专项(FRF- AS-09-006B)第 34 卷 第 3 期 2012 年 3 月 北京科技大学学报 Journal of University of Science and Technology Beijing Vol. 34 No. 3 Mar. 2012 基于案例推理预测精炼开始钢水温度 汪红兵1,2) 艾立翔3) 徐安军3) 田乃媛3) 候志昌4) 周正文1) 1) 北京科技大学计算机与通信工程学院,北京 100083 2) 北京科技大学钢铁流程先进控制教育部重点实验室,北京 100083 3) 北京科技大学冶金与生态工程学院,北京 100083 4) 上海安可科技有限公司,上海 200433 通信作者,E-mail: wanghongbing0816@ 163. com 摘 要 针对 BP 神经网络训练时间长的问题,采用基于案例推理的方法预测精炼开始钢水温度. 首先,应用层次分析法确定 影响精炼开始钢水温度的各个因素的权值,并使用灰色关联度来计算案例的相似度,克服了传统相似度计算方法在案例信息 不完整的情况下无法获取准确结果的缺点. 然后,提出一个包含类选、粗选、精选和择优的四步检索方法,大大缩短了检索时 间. 最后,实验比较了人工神经网络和基于案例推理两种方法,结果表明基于案例推理比人工神经网络具有更高的命中率. 关键词 炼钢; 精炼; 温度; 预测; 基于案例推理 分类号 TF703. 5; TP391. 9 Prediction on the starting temperature of molten steel in second refining by using case-based reasoning WANG Hong-bing1,2) ,AI Li-xiang3) ,XU An-jun3) ,TIAN Nai-yuan3) ,HOU Zhi-chang4) ,ZHOU Zheng-wen1) 1) School of Computer and Communication Engineering,University of Science and Technology Beijing,Beijing 100083,China 2) Key Laboratory of Advanced Control of Iron and Steel Process ( Ministry of Education of China) ,Beijing 100083,China 3) School of Metallurgical and Ecological Engineering,University of Science and Technology Beijing,Beijing 100083,China 4) Shanghai Anchor Technology Co. Ltd. ,Shanghai 200433,China Corresponding author,E-mail: wanghongbing0816@ 163. com ABSTRACT Case-based reasoning was used to predict the starting temperature of molten steel in second refining so as to avoid the long training time of a BP ( back propagation) neural network. Analytic hierarchy process ( AHP) was applied to determine the weights of factors influencing the starting temperature. Grey relational degree was adopted to compute the similarity between cases. Thus the shortcoming of difficulty in obtaining accurate cases with incomplete information is conquered. A four-step search method,including class search,rough search,delicate search,and optimized search,was provided,by which the search time decreases greatly. Experi￾mental results using both artificial neural networks and case-based reasoning were compared. It is shown that case-based reasoning has got a higher hit rate and a shorter response time than artificial neural networks. KEY WORDS steelmaking; refining; temperature; prediction; case-based reasoning 收稿日期: 2011--02--02 基金项目:“十一五”国家科技支撑计划重大项目“新一代可循环钢铁流程工艺技术”( 2006BAE03A07) ; 中央高校基本科研业务费专项( FRF-- AS--09--006B) 二次精炼工艺,如 LF、RH 和 CAS 是炼钢连铸 生产流程中不可缺少的环节. 特别对于品种钢冶 炼,二次精炼是调节钢水温度、合金化、脱氧和去除 夹杂物的重要手段. 精炼开始的钢水温度是精炼工 艺的重要初始参数. 但是,当前的技术条件下很难 实现钢水温度的连续测量. 钢水温度的获得一般依 赖于事后的点测,这种事后处理方法造成了处理周 期偏长,钢水温度下降较大和控制精度降低等不利 后果. 因此,冶金研究人员一直在探索应用在线模 型预测炼钢连铸流程中各个关键点的钢水温度,这 些关键点包括转炉或电炉出钢时刻、精炼开始时刻、 精炼结束时刻和钢水到达连铸中间包时刻等. Fernández 等[1] 使用人工神经网络 ( artificial neural network,ANN) 作为分类器,并设计模糊推理 DOI:10.13374/j.issn1001-053x.2012.03.012
向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有