基于主动学习GASVM分类器的连铸漏钢预报

中国机械工程

基于主动学习GASVM分类器的连铸漏钢预报

方一鸣1,2;胡春洋1;刘;乐1;张兴明3

1.燕山大学工业计算机控制工程河北省重点实验室, 秦皇岛，066004
2.国家冷轧板带装备及工艺工程技术研究中心, 秦皇岛，066004
3. 清华大学天津高端装备研究院，天津，300300

出版日期:2016-06-25 发布日期:2016-06-24
基金资助:
国家自然科学基金委员会与宝钢集团有限公司联合资助项目（U1260203）；国家自然科学基金资助项目（61403332）；河北省自然科学基金-钢铁联合基金资助项目（F201320329）；河北省高等学校创新团队领军人才培育计划资助项目（LJRC013）

Breakout Prediction Classifier for Continuous Casting Based on Active Learning GASVM

Fang Yiming1,2;Hu Chunyang1;Liu Le1;Zhang Xingming3

1.Key Lab of Industrial Computer Control Engineering of Hebei Province,Yanshan University,Qinhuangdao,Hebei,066004
2.National Engineering Research Center for Equipment and Technology of Cold Strip Rolling,Qinhuangdao,Hebei,066004
3.Tianjin Research Institute for Advanced Equipment,Tsinghua University,Tianjin,300300

Online:2016-06-25 Published:2016-06-24
Supported by:

摘要/Abstract

摘要： 针对在小样本数据情况下训练的连铸漏钢预报模型难以获得较高预报准确率的问题，提出了一种基于主动学习遗传算法-支持向量机（GASVM）分类器的漏钢预报算法。该算法首先将采集到的连铸结晶器坯壳温度数据进行预处理，并将有效数据进行标注；然后利用标注后的小样本数据和遗传算法来优化SVM的经验参数，训练并得到支持向量机模型；最后利用某钢厂采集到的连铸结晶器坯壳温度数据进行测试。测试结果表明，在利用小样本数据进行训练的情况下，所提出的基于主动学习GASVM分类器的连铸漏钢预报算法具有较高的漏钢预报率（预报精度）和100%的漏钢报出率，验证了所提漏钢预报算法的有效性。

关键词: 漏钢预报, GASVM, 主动学习, 小样本数据

Abstract: Aiming at the problem that was difficult to obtain a high accurate breakout prediction model of continuous casting in the case of small sample data, a breakout prediction algorithm was proposed based on active learning GASVM classifier. Firstly, the algorithm preprocessed temperature data of continuous casting mold and labels valid data. Secondly, SVM model was obtained after SVM empirical parameters were optimized using labeled small sample data and GA. Finally, the optimized SVM model was tested using the historical data of a steel plant. The results show that in the case of small sample data for training model, the breakout prediction algorithm based on active learning GASVM classifier can obtain higher breakout prediction accuracy and 100% reported ratio. The presented breakout steel prediction algorithm was validated.

Key words: breakout prediction, genetic algorithmsupport vector machine(GASVM), active learning, small sample data

中图分类号:

TF345

方一鸣, 胡春洋, 刘, 乐, 张兴明. 基于主动学习GASVM分类器的连铸漏钢预报[J]. 中国机械工程.

Fang Yiming, Hu Chunyang, Liu Le, Zhang Xingming. Breakout Prediction Classifier for Continuous Casting Based on Active Learning GASVM[J]. China Mechanical Engineering.

参考文献

[1]秦旭, 朱超浦, 尹延荣, 等. 液压伺服振动式板坯连铸机的漏钢预报技术[J]. 钢铁, 2010, 45(11): 97-100.

Qin Xu, Zhu Chaopu, Yin Yanrong, et al. Forecasting of Molten Steel Breakouts for the Slab Continuous Casters with Hydraulic Servo Oscillation Systems[J]. Iron and Steel, 2010, 45(11): 97-100.

[2]Mills K C, Billany T J H. Causes of Sticker Breakout during Continuous Casting[J]. Iron Making and Steelmaking, 1991, 18(4): 253-265

[3]厉英, 王正, 敖志广, 等. BP神经网络漏钢预报系统优化[J]. 控制与决策, 2010, 25(3): 453-456.

Li Ying, Wang Zheng, Ao Zhiguang. Optimization for Breakout Prediction System of BP Neural Network[J]. Control an Decision, 2010, 25(3): 453-456.

[4]张本国, 李强, 王葛, 等. 基于改进BP神经网络的连铸漏钢预报[J]. 中国机械工程, 2012, 23(2): 204-207.

Zhang Benguo, Li Qiang, Wang Ge， et al. Breakout Prediction Based on Improved BP Neural Network in Continuous Casting Process[J]. China Mechanical Engineering, 2012, 23(2): 204-207.

[5]高隽. 人工神经网络原理及仿真实例[M]. 北京:机械工业出版社, 2007.

[6]Vapnik V N. Statistical Learning Theory[M]. New York: J. Wiley, 1998.

[7]张新峰, 赵彦. 基于最小二乘支持向量机的小样本威布尔可靠性分析[J]. 中国机械工程, 2012, 23(16): 1967-1971.

Zhang Xinfeng, Zhao Yan. Weibull Reliability Analysis in Small Samples Based on LSSVM[J]. China Mechanical Engineering, 2012, 23(16): 1967-1971.

[8]陈果. 基于遗传算法的支持向量机分类器模型参数优化[J]. 机器科学与技术, 2007, 26(3): 347-350.

Chen Guo. Optimizing the Parametrs of Support Vector Machine's Classifier Model Based on GeneticAlgorithm[J]. Mechanical Science and Technology, 2007, 26(3): 347-350.

[9]Cheng Jian, Wang Kongqiao. Active Learning for Image Retrieval with Co-SVM[J]. Pattern Recognition, 2006, 40(1): 330-334.

[10]徐海龙, 王晓丹, 廖勇, 等. 一种基于主动学习的SVM增量训练算法[J]. 控制与决策, 2010, 25(2): 282-286.

Xu Hailong, Wang Xiaodan, Liao Yong, et al. Incremental Training Algorithm of SVM Based on Active Learning[J].Control and Decision, 2010, 25(2): 282-286.

[11]曹建, 孙世宇, 段修生, 等. 基于KKT条件的SVM增量学习算法[J]. 火力与指挥控制, 2014, 39(7): 139-143.

Cao Jian, Sun Shiyu, Duan Xiusheng, et al. SVM Incremental Learning Algorithm Based on KKT Conditions[J]. Fire Control&Command Control, 2014, 39(7): 139-143.

[12]Nello C, Hohn S. An Introduction to Suppot Vector Machines and Other Kernal-based Learing Methods[M]. Cambridge: Cambridge University Press, 2000.

[13]Martens D, Baesens B, van Gestel T. Decompositional Rule Extraction from Support Vector Machines by Active Learning[J]. IEEE Transaction on Knowledge and Data Engineering, 2009, 21(2):178-191.