基于互信息-贝叶斯-案例推理的智能制造零部件特征识别与自动匹配模型
Feature recognition and automatic matching model of intelligent manufacturing components based on mutual-information-Bayesian Network-case-based reasoning
沈江 1胡月 1李晋1
作者信息
- 1. 天津大学 管理与经济学部,天津 300072
- 折叠
摘要
智能制造过程被认为是多系统联合决策的过程,尤其是在装备制造过程中,需要在众多零部件供应商中做出选择,迅速完成采购任务.针对这一问题,文中在智能采购决策单元中,引入基于案例的推理机制(CBR),而CBR数据集通常存在特征冗余,使得推理效率和准确度较低.传统的基于贝叶斯网络案例推理特征的选择模型(BN-CBR)对先验知识利用效率不高,且不能选择出消除冗余的特征子集.为此,文中提出了采用互信息的贝叶斯-案例推理(MI-BNCBR)特征选择模型,采用特征冗余度和互信息计算出案例特征的综合权重,以改善BN-CBR模型对先验知识的利用效率不高的问题,并且该模型采用的互信息方法可消除数据集中的冗余特征并得到最优特征子集.最后,采用基于远端最近距离计算的K-D树方法进一步提高基于互信息的贝叶斯-案例推理的效率,并利用基准数据集进行试验.结果表明:所引入的方法提高了案例推理的准确度和推理效率.
Abstract
The intelligent manufacturing process is considered as a multi-system collaborative decision-making process.Es-pecially in the equipment manufacturing process,it is necessary to make choices among numerous component suppliers and quick-ly complete the procurement task.In this article,in order to solve this problem,the case-based reasoning(CBR)mechanism is introduced in the intelligent procurement decision-making unit,but the CBR dataset often suffers feature redundancy,thus resul-ting in a low standard of reasoning efficiency and accuracy.However,the traditional Bayesian Network case-based reasoning fea-ture selection model(BN-CBR)has low efficiency in utilizing prior knowledge and fails to select the feature subsets that eliminate redundancy.To this end,the mutual-information Bayesian Network case-based reasoning(MI-BNCBR)feature selection model is set up.Both feature redundancy and mutual information are used to calculate the case features'comprehensive weight;as a re-sult,the BN-CBR model's efficiency in utilizing prior knowledge will improve.Thanks to the mutual-information method,the re-dundant features in the dataset will be eliminated and the optimal feature subset will be worked out.Finally,the K-D tree method based on remote nearest distance calculation is adopted to improve the efficiency of the mutual-information Bayesian Network case-based reasoning,and a series of experiment are conducted with the help of the benchmark dataset.The results show that this method ensures case-based reasoning a higher standard of accuracy and efficiency.
关键词
智能制造/案例推理/贝叶斯网络/零部件/互信息Key words
intelligent manufacturing/case-based reasoning/Bayesian Network/component/mutual-information引用本文复制引用
出版年
2024
NETL
NSTL
万方数据