基于案例推理的氧化铝生料浆配料优化系统

【作者】 舒军；

【导师】 王雅琳；

【作者基本信息】 中南大学 ， 控制科学与工程， 2008， 硕士

【摘要】 配料作为氧化铝生产的第一道工序,配制生料浆指标的好坏直接关系到熟料质量的高低,“配料是基础,烧结是关键”的思想已是广大科技人员的共识。基于规则的专家系统相对于传统手工计算配比方法,提高了计算速度和配比精度。但专家系统的知识“瓶颈”问题使得系统扩展性差,不能适应生产工序变化。专家系统是基于规则的系统,规则的修改、更新和学习非常困难,从而导致调整生产工序和增添新设备后,原有专家系统不能正常使用,需要进行二次开发,增加了系统的成本,影响了企业的正常生产。本文基于氧化铝生料浆配料长期积累的配料知识,引入了基于案例推理技术,研究了适用于生料浆配料生产特点的基于案例推理的生料浆配料优化系统。论文首先详细讨论了案例表示的内容与原则,以给定质量指标和原料成分为条件属性,下料配比为决策属性,设计了基于框架的配料知识表示,研究了配料案例的组织方法及其案例库的维护算法,设计了配料系统案例库。其次针对案例的表示特点和组织原则,构造了合成相似性度量函数,分析了现有特征属性权值确定方法客观性差、算法复杂、实现困难等问题,提出了一种基于覆盖度的案例特征属性权值确定方法,根据不同属性对案例平均覆盖度的影响,自动为案例的特征属性确定权值,提高了系统案例检索精度。然后,以适应生料浆配料工艺变化为目的,设计了调整参数与步长自由选取的案例调整模式,将机理模型预测质量指标与给定质量指标差值作为案例调整输入,提出了基于模型的案例调整策略,将机理模型设计为CBR系统的实时案例评价模型,并与专家系统进行仿真比较,结果表明该系统既能适应生产工序的变化,又能保证生料浆的质量。最后,介绍了基于案例推理的配料系统功能模块及关键技术的实现,并将CBR系统试运行结果与专家系统进行比较,证明了CBR系统在保证生料浆质量的同时,稳定了生产。更多还原

【Abstract】 Blending is the first work procedure of the alumina production by sintering method, and the quality of the pulp directly relates to the sintered mixture. Most technique men agree with that blending is the basic and the sintering is the key. Comparing with the classical manual method of compute ratio, rule-based expert system improved the compute speed and the precise. However, the problem of bottle neck in expert systems results to bad expansibility, and inadaptable with the changes of the produce process.Expert system is based rule base, but it is difficult to modify, update or learn the rules; so it can not use normally when the produce work procedure has adjusted or new equipments have added. In this case, the expert system needs to develop again, then the maintenance expanse is added, and the normal production is effected. Based on the long-period knowledge of produce experience and the technology of case-based reasoning (CBR), a blending CBR system that adapts productive features is studied in this paper.Above of all, in order to realize the blending CBR system in production of alumina, according to the content and the principles of the case expression, the blending case expression based frame is discussed in detail, the given quality index and material component are designed to the condition features, and the mixture ratio is designed to the decision features. The organizatition methods and the maintenance algorithms of the case base are researched. Then the case base of the blending system is designed. Secondly, aiming at solving the problems of the existing algorithms such as poor objectivity and high complexity, a method based on coverage for determining the case feature weights is proposed, which calculates the case feature weights by the affecting of each attribute to the case average coverage, it improved the precise of the system retrieve. Thirdly, in view of the characteristics the technologic changes in blending system, a case adaptation mode with its adjust parameters and steps could be chosen was put forward, the error between the quality index predicted by the mechanism model and the given quality index was designed as an input of the case adaptation, then the strategy of the case adaptation based on the mechanism model was proposed, the mechanism model was designed to the case review. The simulation results which were gotten by the CBR system and the blending expert system are compared, which verifies effectiveness of the method and the ability of the adaptation in the changes of the work procedure. Lastly, the realization of the system function and the key technologies is introduced. The running results which were gotten by the CBR system and the blending expert system are compared, which proved that the CBR system can ensure the quality of pulp and stabilize industrial production.更多还原