基于HDBSCAN聚类算法的实例推理与规则提取

Case Reasoning and Rule Extraction Based on HDBSCAN Clustering Algorithm

针对复杂装配对象具有结构复杂、开发周期长、装配成本高等特点导致的装配工艺编制较慢、效率低的问题,为实现装配工艺重用,在规则提取过程中,利用Apriori关联规则算法提取出满足约束参数的强关联规则,作为知识检索的条件与结论放入规则库中;在实例推理过程中,提出基于DBSCAN聚类算法快速定位与目标装配对象相似的子实例集,即与目标对象最相似的簇,缩小实例检索的范围以提高匹配的效率。结果表明,该方法使检索范围缩小了50倍,实例匹配速度明显加快。

Due to the complex assembly object’s complex structure, long development cycle and high assembly cost, the assembly process is slow and the efficiency is low. In order to realize assembly process re-use, in the process of rule extraction, the Apriori association rule algorithm is used to extract the strong association rules meeting the constraint parameters and put into the rule base as the condi-tions and conclusions of knowledge retrieval. In the process of case reasoning, the DBSCAN cluster-ing algorithm is proposed to quickly locate the sub-instance set similar to the target assembly ob-ject, that is, the cluster most similar to the target object, and narrow the scope of instance retrieval to improve the matching efficiency. The results show that the retrieval range is reduced by 50 times and the case matching speed is greatly accelerated.