基于案例推理的软测量方法及在磨矿过程中的应用  

Soft-sensor Approach with Case-based Reasoning and Its Application in Grinding Process

作　　者：周平[1] 岳恒[1] 赵大勇[1] 柴天佑[1] 

机构地区：[1]东北大学教育部暨辽宁省流程工业综合自动化重点实验室,沈阳110004

出　　处：《控制与决策》2006年第6期646-650,655,共6页Control and Decision

基　　金：国家973重大基础研究计划项目(2002CB312201);国家自然科学基金重点项目(60534041)

摘　　要：针对复杂工业过程中一些关键工艺参数难以用仪表进行在线检测的问题,提出了基于案例推理的软测量方法.案例表示由案例产生时间、工况描述、解及相似度组成;案例检索采用具有多相似度阈值计算的最近相邻策略;案例重用采用基于静态相似度阈值和基于动态相似度阈值两种算法,并给出了新的案例修正和存储策略.用该方法建立的磨矿粒度软测量模型已成功应用在某选矿厂磨矿过程中,应用结果表明提出的方法效果显著,具有推广应用前景.A soft-sensor approach with case-based reasoning（CBR） is developed to deal with the problem that some key technical parameters cannot be measured directly online in complex industrial processes. The structure of case representation is composed of case time, case descriptors, case solution and case similarity. The improved k-nearest neighbor（k-NN） technique with multiple case similarity threshold computation is used in case retrieval. The methods based on static and dynamic similarity threshold are employed in case reuse. New strategies of case revision and case retaining are presented. A soft-sensor model for grinding particle size based on the proposed CBR is established for a grinding process in a mineral processing plant, and the successful industrial application shows the effectiveness and future prospect of the proposed method.

关 键 词：案例推理 软测量 磨矿粒度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]