基于半监督学习的工况识别方法研究及铜闪速熔炼过程中的应用
|
摘要
铜闪速熔炼过程工况变化频繁,造成经济指标波动大,单位产品能耗高。闪速熔炼过程积累了大量的实际运行数据,包含了反映操作参数与工况之间关系的信息。准确识别当前工况,采取相应的对策,对节能降耗和保证生产的安全稳定具有重要意义。本文利用运行数据中的有标签数据和无标签数据,研究了基于半监督学习的工况识别方法。
针对YATSI算法中无标签样本被误标记,导致分类正确率下降的问题,提出了一种基于数据剪辑技术的DE-YATSI算法。该算法采用基于估计类条件概率的数据剪辑方法,识别并重新标记预标记样本集中的误标记样本,利用有标签样本与经数据剪辑处理的预标记样本所形成的样本集训练带权重的最近邻分类器。实验结果表明,DE-YATSI算法比YATSI算法具有更高的分类正确率。
对影响闪速熔炼过程工况的主要因素进行了分析,分别将半监督算法(DE-YATSI和YATSI)和监督算法(KNN)应用于铜闪速熔炼过程工况识别,实验结果表明,半监督学习算法(DE-YATSI)识别正确率最高。
针对铜闪速熔炼过程积累的工业运行数据有大量的无标签数据的特点,提出一种基于半监督学习工况识别方法,包括数据获取和预处理、离线半监督学习建模和在线工况识别三部分；并对铜闪速熔炼工况识别系统(WSRS)的设计与开发进行了深入研究,给出了WSRS的完整开发方案、架构设计、子系统分解、实现功能和实施步骤,并对系统软件进行了实现。
Working-status of copper flash smelting process frequently changes, which contribute a lot to the poor output and quality of copper. Copper flash smelting process has accumulated lots of industrial operating data. It contains much underlying information between rules of working-status and production parameters. Accurately recognizing working-status and timely taking control are of very grand significance to ensure stable and reliable running of system and promote the economic benefits of enterprise. A method of working-status based semi-supervised learning by using labeled and unlabeled data has been studied in this paper.
YATSI may suffer more from the common problem in semi-supervised learning, i.e. the performance is usually not stable due to the unlabeled examples may often be wrongly labeled. A semi-supervised k-nearest neighbor classifier named DE-YATSI is proposed. A data editing based on estimating the conditional probability of class is used to identify and relabel mislabeled examples of the pre-labeled data set. A k-nearest neighbor classifier with weights is trained by the labeled and the edited "pre-labeled" data set. Experiments on UCI datasets show that DE-YATSI could more effectively and stably utilize the unlabeled examples to improve classification accuracy than YATSI.
Main factors of working-status of copper flash smelting are analysed in this paper. Semi-supervised algorithms (YATSI and DE-YATSI) and supervised algorithm (KNN) are used to recognize working-status of the copper flash smelting. Experimental results show that semi-supervised algorithm can better recognise working-status than supervised algorithm. Considering labeled samples with small number and unlabeled ones with large number of industrial historical data of copper flash smelting process, a framework of working-status recognizes based of semi-supervised are presented. It includes data acquisition and processing, semi-supervised off-line modeling and working-status on-line recognize. The working-status recognition system (WSRS) is designed which includes system theory, architecture design, subsystems decomposing, system function and steps of realization.
针对YATSI算法中无标签样本被误标记,导致分类正确率下降的问题,提出了一种基于数据剪辑技术的DE-YATSI算法。该算法采用基于估计类条件概率的数据剪辑方法,识别并重新标记预标记样本集中的误标记样本,利用有标签样本与经数据剪辑处理的预标记样本所形成的样本集训练带权重的最近邻分类器。实验结果表明,DE-YATSI算法比YATSI算法具有更高的分类正确率。
对影响闪速熔炼过程工况的主要因素进行了分析,分别将半监督算法(DE-YATSI和YATSI)和监督算法(KNN)应用于铜闪速熔炼过程工况识别,实验结果表明,半监督学习算法(DE-YATSI)识别正确率最高。
针对铜闪速熔炼过程积累的工业运行数据有大量的无标签数据的特点,提出一种基于半监督学习工况识别方法,包括数据获取和预处理、离线半监督学习建模和在线工况识别三部分；并对铜闪速熔炼工况识别系统(WSRS)的设计与开发进行了深入研究,给出了WSRS的完整开发方案、架构设计、子系统分解、实现功能和实施步骤,并对系统软件进行了实现。
Working-status of copper flash smelting process frequently changes, which contribute a lot to the poor output and quality of copper. Copper flash smelting process has accumulated lots of industrial operating data. It contains much underlying information between rules of working-status and production parameters. Accurately recognizing working-status and timely taking control are of very grand significance to ensure stable and reliable running of system and promote the economic benefits of enterprise. A method of working-status based semi-supervised learning by using labeled and unlabeled data has been studied in this paper.
YATSI may suffer more from the common problem in semi-supervised learning, i.e. the performance is usually not stable due to the unlabeled examples may often be wrongly labeled. A semi-supervised k-nearest neighbor classifier named DE-YATSI is proposed. A data editing based on estimating the conditional probability of class is used to identify and relabel mislabeled examples of the pre-labeled data set. A k-nearest neighbor classifier with weights is trained by the labeled and the edited "pre-labeled" data set. Experiments on UCI datasets show that DE-YATSI could more effectively and stably utilize the unlabeled examples to improve classification accuracy than YATSI.
Main factors of working-status of copper flash smelting are analysed in this paper. Semi-supervised algorithms (YATSI and DE-YATSI) and supervised algorithm (KNN) are used to recognize working-status of the copper flash smelting. Experimental results show that semi-supervised algorithm can better recognise working-status than supervised algorithm. Considering labeled samples with small number and unlabeled ones with large number of industrial historical data of copper flash smelting process, a framework of working-status recognizes based of semi-supervised are presented. It includes data acquisition and processing, semi-supervised off-line modeling and working-status on-line recognize. The working-status recognition system (WSRS) is designed which includes system theory, architecture design, subsystems decomposing, system function and steps of realization.
引文
[1]谢锴.现代铜闪速熔炼炉发展中若干理论与操控优化问题：[博士学位论文].长沙：中南大学,2005
[2]Crowther R H, Pitrak J E, Ply E N. Computer control at american oil. Chemical Engineering Process,1961,57(6):39～50
[3]胡志坤.复杂有色金属熔炼过程操作模式智能优化方法研究：[博士学位论文].长沙：中南大学,2005
[4]彭晓波.铜闪速熔炼过程智能优化方法及应用：[博士学位论文].长沙：中南大学,中国,2008
[5]颜青君.铜闪速熔炼操作参数优化的研究与应用：[硕士学位论文].长沙：中南大学,2007
[6]桂卫华,阳春华,李勇刚,等.基于数据驱动的铜闪速熔炼过程操作模式优化及应用.自动化学报,2009,6(51)：717-724
[7]Morari M, Stephanopoulos G, Arkun Y. Studies in the synthesis of control structures for chemical processes:Part I. AIChE Journal,1980,26(2):220～232
[8]Narraway L T, Perkins J D. Selection of process structure based on linear dynamic economics. Industial & Engineering Chemistry Research,1993,32(11): 2681～2692
[9]Zheng A, Mahajanam R V, Douglas J M. Hierchical procedure for plantwide control system synthesis. Aiche Journal,1999,45(6):1255～1265
[10]Kassidas A, Patry J, Marlin T. Integrating process and controller models for the design of self-optimizing control. Computers and Chemical Engineering,2000, 24(12):2589～2602
[11]Skogestad S. Control structure design for complete chemical plants. Computers and Chemical Engineering,2004,28(12):219～234
[12]Young R E. Petroleum refining process control and real-time optimization. IEEE Control System Magazine,2006,26(6):73～83
[13]Singh A, Forbes J F, Vermeer P J. Model-based real-time optimization of automotive gasoline blending operations. Journal of Process Control,2000,10(1): 43～58
[14]Nath R, Alzein Z. On-line dynamic optimization of olefins plants. Computers & Chemical Engineering,2000,24(2):533～538
[15]Zanin A C, Tvrzska M, Odloak D. Integrating realtime optimization into the model predictive controller of the FCC system. Control Engineering Practice, 2002,10 (8):819～831
[16]Zanin A C, Tvrzska D, Gouvea M, et al. Industrial implementation of a real-time optimization strategy for maximizing production of LPG in a FCC unit. Computers & Chemical Engineering,2000,24 (2):525～531
[17]Qin S J, Badgewell T. A survey of industrial model predictive control technology. Control Engineering. Practice,2003,11(7):733～764
[18]王雅琳.智能集成建模理论及其在有色冶炼过程优化控制中的应用研究：[博士学位论文].长沙：中南大学,2001
[19]Goh A T C. Back-propagation neural networks for modeling complex systems. Artificial intelligence in Engineering,1995,9(3):143～151
[20]Sequeira E, Graells M, Puigjaner L. Real-time evolution of online optimization of continuous processes. Industrial & Engineering Chemistry Research,2002, 41(7):1815～1825
[21]Basak K, Abhilash K S, Ganguly S, et al. On-line optimization of a crude distillation unit with constraints on product properties. Industrial & Engineering Chemistry Research,2002,41(6):1557～1568
[22]Theodora K, John E M. Process analysis, monitor and diagnosis, using multivariate projection methods. Chemometrics and Intelligent Laboratory Systems,1995,28(1):3～21
[23]Boggs P T, Tolle J W. Sequential quadratic programming for large-scale nonlinear optimization. Journal of Computational and Applied Mathematics, 2000,124(1):123～137
[24]Chang K Y, Jong M L, Peter A, et al. On-line monitoring of batch processes using multiway independent component analysis. Chemometrics and Intelligent Laboratory Systems,2004,71(2):151～163
[25]何宁,谢磊,郭明,等.基于独立成分的动态多变量过程的故障检测与诊断方法.化工学报,2005,56(4)：646-652
[26]桂卫华,王雅琳,阳春华.基于改进模拟退火算法的锌电解过程分时供电优化控制,控制理论与应用,2001,18(1)：127-130
[27]Guo M, Xie L, Wang S Q. Research on an integrated ICA-SVM based framework for fault diagnosis. IEEE International Conference on Systems, Man & Cybernetics, Washington, USA,2003,2710～2715
[28]Chu Y, Qin S J, Han C. Fault detection and operation mode identification based on pattern classification with variable selection. Industrial & Engineering Chemical Research,2004,43(7):1701～1710
[29]Chapelle 0, Scholkopf B, Zien A. Semi-Supervised Learning. Cambridge, MA: The MIT Press, London, England,2006
[30]Scudder H J. Probability of error of some adaptive pattern-recognition machines. IEEE Transactions on Information Theory,2005,11:363～371
[31]Fralick S C. Learning to recognize patterns without a teacher. IEEE Transactions on Information Theory,1997,13:57～64
[32]Agrawala A K. Learning with a probabilistic teacher. IEEE Transactions on Information Theory,2000,16:373～379
[33]Zhou Z H. Co-training Paradigm in Semi-supervised Learning. In Proceedings of the Chinese Workshop on Machine Learning and Applications, Nanjing, China, 2007
[34]Seung H, Opper M, Sompolinsky H. Query by committee. In Proceedings of the 5th ACM Workshop on Computational Learning Theory,1992,287-294
[35]Merz C J, Clair D C, Bond W E. Semi-supervised adaptive resonance theory. In Proceedings of International Joint Conference on Neural Networks,1992, 851～856
[36]Vladimir V. The Nature of Statistical Learning Theory. Springer,1995
[37]Vladimir V. Statistical Learning Theory. Wiley-Interscience,1998
[38]Yarowsky D. Unsupervised word sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics,1995
[39]Maeireizo B, Litman D, Hwa R. Co-training for predicting emotions with spoken dialogue data. In Proceedings of the 42nd Annual Meeting of the Association for Computational Linguistics,2004
[40]Rosenberg C, Hebert M, Schneiderman H. Semi-supervised self-training of object detection models. In Seventh IEEE Workshop on Applications of Computer Vision,2005
[41]Blum A, Mitchell T. Combining labeled and unlabeled data with co-training. In Proceedings of the Workshop on Computational Learning Theory,1998
[42]Zhou Z H, Li M. Tri-training:Exploiting unlabeled data using three classifiers. IEEE Transactions on Knowledge and Data Engineering,2005,17(11): 1529～1541
[43]王娇.多视图的半监督学习研究：[博士学位论文].北京：北京交通大学,2010
[44]李士进,陶剑,万定生.多分类器实例协同训练遥感图像检索.遥感学报,2010,14[3]：0-7
[45]郝建柏.基于图的半监督学习模型研究与分类器设计：[硕士学位论文].合肥：中国科技大学,2009
[46]Bennett K, Demiriz A. Semi-supervised support vector machines. Advances in Neural Information Processing Systems,1999
[47]Chapelle 0, Zien A. Semi-supervised classification by low density separation. In Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics,2005
[48]王添强,余养强,秦小麟.结构复杂数据的半监督聚类.控制与决策,2010,25(1)：14-17
[49]龙军,殷建平,祝恩.主动学习研究综述.计算机研究与发展,2008,45：300-304
[50]Zhu X. Semi-supervised learning with graphs. Ph.D.thesis. Carnegie Mellon University,2005
[51]Miller D J, Uyar H S. A mixture of experts classifier with learning based on both labelled and unlabelled data. In Proceedings of Neural Information Processing Systems,1996,571～577
[52]Kurt D, Peter R, Bernhard P, et al. Using Weighted Nearest Neighbor to Benefit from Unlabeled Data. In Proceedings of the Asia-Pacific Conference on Knowledge Discovery in Databases(PAKDD 2006), Singapore,2006,918:60～69
[53]王平.基于半监督学习的障碍物检测：[硕士学位论文].南京：南京理工大学,2009
[54]Cover T M, Hart P E. Nearest neighbor pattern classification. IEEE Transaction on Information Theory,1997,13(1):1～27
[55]Ripley B D. Pattern Recognition and Neural Networks. Cambridge:Cambridge University Press,1996.198～201
[56]Angluin D, Laird P. Learning from nosiy examples. In machine learning,1988
[57]Wilson D L. Asymptotic properties of nearest neighbor rules using edited data sets. IEEE Trans on Systems, Man and Cybernetics,2002,2(3):408～421
[58]Barandela R, Gasca E. Decontamination of training data for supervised pattern recognition methods. Advances in Pattern Recognition, Lecture Notes in Computer Science,2000,1876:612～620
[59]Ferri F J, Albert J V, Vidal E. Considerations about sample-size sensitivity of a family of edited nearest-neighbor rules. IEEE Trans on Systems, Man and Cybernetics, Part B-Cybernetics,1999,29(5):667～672
[60]Guan D H, Yuan W W, Lee Y K, et al. Nearest Neighbor Editing Aided by Unlabeled Data. Information Sciences,2009,179(13):2273～2282
[61]UCI repository of machine learning databases [EB/OL]. http: //www.its.uci.edu/mleam/mlrepository.html
[62]李欣峰.炼铜闪速熔炼过程的数值分析和优化：[博士学位论文].长沙：中南大学,2001
[63]张朝阳.闪速熔炼中制氧的节能.有色冶金节能.1999,63(3)：9-11
[64]郭遴,郑明.冶金工艺导论.长沙：中南工业大学出版社,1991
[65]尹文新.中国铜工业发展初探：[博士学位论文].长沙：中南大学,2003
[2]Crowther R H, Pitrak J E, Ply E N. Computer control at american oil. Chemical Engineering Process,1961,57(6):39～50
[3]胡志坤.复杂有色金属熔炼过程操作模式智能优化方法研究：[博士学位论文].长沙：中南大学,2005
[4]彭晓波.铜闪速熔炼过程智能优化方法及应用：[博士学位论文].长沙：中南大学,中国,2008
[5]颜青君.铜闪速熔炼操作参数优化的研究与应用：[硕士学位论文].长沙：中南大学,2007
[6]桂卫华,阳春华,李勇刚,等.基于数据驱动的铜闪速熔炼过程操作模式优化及应用.自动化学报,2009,6(51)：717-724
[7]Morari M, Stephanopoulos G, Arkun Y. Studies in the synthesis of control structures for chemical processes:Part I. AIChE Journal,1980,26(2):220～232
[8]Narraway L T, Perkins J D. Selection of process structure based on linear dynamic economics. Industial & Engineering Chemistry Research,1993,32(11): 2681～2692
[9]Zheng A, Mahajanam R V, Douglas J M. Hierchical procedure for plantwide control system synthesis. Aiche Journal,1999,45(6):1255～1265
[10]Kassidas A, Patry J, Marlin T. Integrating process and controller models for the design of self-optimizing control. Computers and Chemical Engineering,2000, 24(12):2589～2602
[11]Skogestad S. Control structure design for complete chemical plants. Computers and Chemical Engineering,2004,28(12):219～234
[12]Young R E. Petroleum refining process control and real-time optimization. IEEE Control System Magazine,2006,26(6):73～83
[13]Singh A, Forbes J F, Vermeer P J. Model-based real-time optimization of automotive gasoline blending operations. Journal of Process Control,2000,10(1): 43～58
[14]Nath R, Alzein Z. On-line dynamic optimization of olefins plants. Computers & Chemical Engineering,2000,24(2):533～538
[15]Zanin A C, Tvrzska M, Odloak D. Integrating realtime optimization into the model predictive controller of the FCC system. Control Engineering Practice, 2002,10 (8):819～831
[16]Zanin A C, Tvrzska D, Gouvea M, et al. Industrial implementation of a real-time optimization strategy for maximizing production of LPG in a FCC unit. Computers & Chemical Engineering,2000,24 (2):525～531
[17]Qin S J, Badgewell T. A survey of industrial model predictive control technology. Control Engineering. Practice,2003,11(7):733～764
[18]王雅琳.智能集成建模理论及其在有色冶炼过程优化控制中的应用研究：[博士学位论文].长沙：中南大学,2001
[19]Goh A T C. Back-propagation neural networks for modeling complex systems. Artificial intelligence in Engineering,1995,9(3):143～151
[20]Sequeira E, Graells M, Puigjaner L. Real-time evolution of online optimization of continuous processes. Industrial & Engineering Chemistry Research,2002, 41(7):1815～1825
[21]Basak K, Abhilash K S, Ganguly S, et al. On-line optimization of a crude distillation unit with constraints on product properties. Industrial & Engineering Chemistry Research,2002,41(6):1557～1568
[22]Theodora K, John E M. Process analysis, monitor and diagnosis, using multivariate projection methods. Chemometrics and Intelligent Laboratory Systems,1995,28(1):3～21
[23]Boggs P T, Tolle J W. Sequential quadratic programming for large-scale nonlinear optimization. Journal of Computational and Applied Mathematics, 2000,124(1):123～137
[24]Chang K Y, Jong M L, Peter A, et al. On-line monitoring of batch processes using multiway independent component analysis. Chemometrics and Intelligent Laboratory Systems,2004,71(2):151～163
[25]何宁,谢磊,郭明,等.基于独立成分的动态多变量过程的故障检测与诊断方法.化工学报,2005,56(4)：646-652
[26]桂卫华,王雅琳,阳春华.基于改进模拟退火算法的锌电解过程分时供电优化控制,控制理论与应用,2001,18(1)：127-130
[27]Guo M, Xie L, Wang S Q. Research on an integrated ICA-SVM based framework for fault diagnosis. IEEE International Conference on Systems, Man & Cybernetics, Washington, USA,2003,2710～2715
[28]Chu Y, Qin S J, Han C. Fault detection and operation mode identification based on pattern classification with variable selection. Industrial & Engineering Chemical Research,2004,43(7):1701～1710
[29]Chapelle 0, Scholkopf B, Zien A. Semi-Supervised Learning. Cambridge, MA: The MIT Press, London, England,2006
[30]Scudder H J. Probability of error of some adaptive pattern-recognition machines. IEEE Transactions on Information Theory,2005,11:363～371
[31]Fralick S C. Learning to recognize patterns without a teacher. IEEE Transactions on Information Theory,1997,13:57～64
[32]Agrawala A K. Learning with a probabilistic teacher. IEEE Transactions on Information Theory,2000,16:373～379
[33]Zhou Z H. Co-training Paradigm in Semi-supervised Learning. In Proceedings of the Chinese Workshop on Machine Learning and Applications, Nanjing, China, 2007
[34]Seung H, Opper M, Sompolinsky H. Query by committee. In Proceedings of the 5th ACM Workshop on Computational Learning Theory,1992,287-294
[35]Merz C J, Clair D C, Bond W E. Semi-supervised adaptive resonance theory. In Proceedings of International Joint Conference on Neural Networks,1992, 851～856
[36]Vladimir V. The Nature of Statistical Learning Theory. Springer,1995
[37]Vladimir V. Statistical Learning Theory. Wiley-Interscience,1998
[38]Yarowsky D. Unsupervised word sense disambiguation rivaling supervised methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics,1995
[39]Maeireizo B, Litman D, Hwa R. Co-training for predicting emotions with spoken dialogue data. In Proceedings of the 42nd Annual Meeting of the Association for Computational Linguistics,2004
[40]Rosenberg C, Hebert M, Schneiderman H. Semi-supervised self-training of object detection models. In Seventh IEEE Workshop on Applications of Computer Vision,2005
[41]Blum A, Mitchell T. Combining labeled and unlabeled data with co-training. In Proceedings of the Workshop on Computational Learning Theory,1998
[42]Zhou Z H, Li M. Tri-training:Exploiting unlabeled data using three classifiers. IEEE Transactions on Knowledge and Data Engineering,2005,17(11): 1529～1541
[43]王娇.多视图的半监督学习研究：[博士学位论文].北京：北京交通大学,2010
[44]李士进,陶剑,万定生.多分类器实例协同训练遥感图像检索.遥感学报,2010,14[3]：0-7
[45]郝建柏.基于图的半监督学习模型研究与分类器设计：[硕士学位论文].合肥：中国科技大学,2009
[46]Bennett K, Demiriz A. Semi-supervised support vector machines. Advances in Neural Information Processing Systems,1999
[47]Chapelle 0, Zien A. Semi-supervised classification by low density separation. In Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics,2005
[48]王添强,余养强,秦小麟.结构复杂数据的半监督聚类.控制与决策,2010,25(1)：14-17
[49]龙军,殷建平,祝恩.主动学习研究综述.计算机研究与发展,2008,45：300-304
[50]Zhu X. Semi-supervised learning with graphs. Ph.D.thesis. Carnegie Mellon University,2005
[51]Miller D J, Uyar H S. A mixture of experts classifier with learning based on both labelled and unlabelled data. In Proceedings of Neural Information Processing Systems,1996,571～577
[52]Kurt D, Peter R, Bernhard P, et al. Using Weighted Nearest Neighbor to Benefit from Unlabeled Data. In Proceedings of the Asia-Pacific Conference on Knowledge Discovery in Databases(PAKDD 2006), Singapore,2006,918:60～69
[53]王平.基于半监督学习的障碍物检测：[硕士学位论文].南京：南京理工大学,2009
[54]Cover T M, Hart P E. Nearest neighbor pattern classification. IEEE Transaction on Information Theory,1997,13(1):1～27
[55]Ripley B D. Pattern Recognition and Neural Networks. Cambridge:Cambridge University Press,1996.198～201
[56]Angluin D, Laird P. Learning from nosiy examples. In machine learning,1988
[57]Wilson D L. Asymptotic properties of nearest neighbor rules using edited data sets. IEEE Trans on Systems, Man and Cybernetics,2002,2(3):408～421
[58]Barandela R, Gasca E. Decontamination of training data for supervised pattern recognition methods. Advances in Pattern Recognition, Lecture Notes in Computer Science,2000,1876:612～620
[59]Ferri F J, Albert J V, Vidal E. Considerations about sample-size sensitivity of a family of edited nearest-neighbor rules. IEEE Trans on Systems, Man and Cybernetics, Part B-Cybernetics,1999,29(5):667～672
[60]Guan D H, Yuan W W, Lee Y K, et al. Nearest Neighbor Editing Aided by Unlabeled Data. Information Sciences,2009,179(13):2273～2282
[61]UCI repository of machine learning databases [EB/OL]. http: //www.its.uci.edu/mleam/mlrepository.html
[62]李欣峰.炼铜闪速熔炼过程的数值分析和优化：[博士学位论文].长沙：中南大学,2001
[63]张朝阳.闪速熔炼中制氧的节能.有色冶金节能.1999,63(3)：9-11
[64]郭遴,郑明.冶金工艺导论.长沙：中南工业大学出版社,1991
[65]尹文新.中国铜工业发展初探：[博士学位论文].长沙：中南大学,2003