【作者】 陈湘涛；

【导师】 刘业翔； 李劼；

【作者基本信息】 中南大学 ， 有色金属冶金， 2004， 博士

【摘要】 本文针对传统的铝电解控制系统存在着网络处理能力较差，信息孤岛现象严重，数据分析能力欠缺等问题将数据仓库与数据挖掘技术引入到铝电解控制系统，并结合计算机网络技术，进行了深入的理论和实验研究，主要内容及创新如下： 1) 将数据仓库和数据挖掘技术引入到铝电解控制系统的过程中，结合铝电解专家的经验和知识，充分利用控制系统产生的各种数据，设计并开发了铝电解数据挖掘器； 2) 在数据仓库的预处理过程中，针对重复记录的处理，提出了属性重复量与记录重复量的概念，设计了重复记录辩识和噪音识别的相关算法：对“空穴”数据，则根据新老信息对其的影响，构造了加权算子实现空穴填充，形成等时距序列，最终实现数据挖掘； 3) 将灰关联分析引入数据挖掘，提出了灰关联规则的挖掘框架；在框架内对控制系统产生的数据进行挖掘，获得的灰关联规则表明在给定的时间属性区间内，非主属性对主属性影响程度的大小，为调整电解槽的控制参数提供决策支持；对灰关联挖掘模型，设计了挖掘算法Gray_CTL； 4) 提出了基于连通分支的聚类分析模型，通过该模型，实现对单槽和系列槽的槽况分类；不属于任何连通分支的的孤立点，表明存在异常槽况；通过对时间窗口的滑动，实现槽况的趋势分析；针对聚类分析模型，设计了挖掘算法CC_CTL，算法通过空间中点的邻域确定连通分支，可以挖掘任意形状的聚类； 5) 在铝电解控制系统中构建了偏移系数模型，从管理的角度对生产数据进行挖掘;在模型中，通过对电解槽的历史数据进行分析，获得属性集的最佳考核指标范围，即满足最小支持度的属性划分;在此基础上，实现对单维和多维属性的偏移系数挖掘，获得电解槽当前各考核目标的指标偏移状况，及时调整和修改偏差较大的控制参数，提高管理效率; 6)基于网络与数据仓库技术，设计并开发了新型铝电解网络监控系统;通过网络，用户根据权限可以在任意一个节点实现历史数据与报表的远程访问、动态信息的远程监控，以及各种控制参数的远程访问或修改。实际的使用结果表明，这种新型的铝电解网络监控系统加强了铝电解槽的管理，提高了电解系列的监控水平。更多还原

【Abstract】 In the traditional aluminum electrolysis control systems, there were many problems, such as bad abilities of network processing, phenomena of information alone islands, shortcomings of data analysis. To solve the problems, this dissertation introduced data warehouse and data mining concepts into the aluminum electrolysis control systems. Combined with network technology, a thorough study was made in theory and experimentation. The main contents and innovations of this dissertation were summarized as follows:1. In process of introducing the technology of data warehouse and data mining into the aluminum electrolysis control systems, the aluminum electrolysis data miner was designed and developed, which combined with experience and knowledge of aluminum electrolysis experts, and made full use of all kinds of data produced by the control systems.2. The concepts of the measure of attributes repeating and the measure of records repeating were brought forward to process the repeating records in the data preprocessing of data warehouse, and the algorithms correlated with identifying of repeating records and noise were designed. Based on the influence of the old and new information on the cavity data, arithmetic operators with weight were built to fill in the cavity for the formation of the equal interval sequence and the completion of the data mining in the end.3. The gray association analysis was introduced into data mining, and the framework of the gray association rules was put forward. The gray association rules obtained from the result in mining the data produced from the control systems in such framework showed the degree to which the non-main attributes affected the main attribute in a certain area of time attribute, and provided the decision support for adjusting the control parameters of the aluminum electrolysis cells. The algorithm for mining the gray association rules named GrayCTL wasdesigned.4. A model of cluster analysis based on connected components was proposed. The cell conditions of a single one or a series of cells was grouped into classes by this model. The outliers not belonging to any connected component showed the abnormity cell condition. Moving the time-window could complete a tendency analysis of the cell conditions. The clustering algorithm called CCCTL was designed, and it could discover clusters with arbitrary shape since the algorithm regarded the regions of points in the data space as connected component.5. A model of offset coefficient was built in the aluminum electrolysis control system, and it was used to mine the production data from the point of view of management. The best target range of attribute sets for examining could be obtained through analyzing the history data of aluminum electrolysis cells, namely attribute partition satisfied minimum support. Based on the attribute partition, the offset coefficient of single dimension attribute and the one of multidimensional attribute were mined, which gained the status excursion of current targets of the aluminum electrolysis cells, and the control parameters with obvious warp could be adjusted and modified in time to improve the efficiency of management.6. A new type of aluminum electrolysis control system was designed and developed based on the technology of network and data warehouse. The target such as remote accessing of history data and reports, remote controlling of dynamic information and remote accessing and modifying of all kinds of control parameters could be realized that a user lay in any node of the network based on his rights. The result of real application showed that thus new type of aluminum electrolysis control system strengthened the management of cells and enhanced the control level of aluminum electrolysis series.更多还原