复杂设备诊断与维修中基于改进最近相邻法和线索的案例检索

基于案例推理系统的强大功能，来源于它能够从案例库中迅速地检索到与当前新问题“相似”的案例，而这种相似性的比较主要依赖于如何建立问题的特征索引。结合目前三种主要的案例索引技术：最近相邻法（Nearest－neighbor）、归纳法和知识索引法，提出了改进最近相邻法的案例索引技术；并以设备故障诊断与维修为背景，给出了一种基于改进最近相邻法的线索推理算法。

CASE RETRIEVING ALGORITHM BASED ON MODIFIED NEAREST-NEIGHBOR FUNCTION AND THREAD REASONING ON COMPLEX DEVICEDIAGNOSIS AND MAINTENANCE

The strong functionality of case - based reasoning (CBR) depends on the ability of retrieving previous cases which are similar to the current case. Central to retrieval methodologies is the assessment of the similarity of a new case with previous cases. The key of the compare of the similarity is to build the character index of relevant previous cases. Integrating the current three kinds of main case index technology (nearest - neighbor index, inductive index, knowledge index), we present a modified form of the nearest - neighbor function which has capabilities in addition to those of inductive and knowledge index functions. Our empirical evaluation of a CBR application to a kind of equipment diagnosis and repair task shows that the modified nearest - neighbor function has a superior retrieval performance when compared to that of single matching function. At last, we address in detail the implementation of the thread reasoning algorithm based on modified nearest - neighbor function.