基于二进制粒子群优化算法的零件特征与机床加工关联知识挖掘

Binary Particle Swarm Optimization-based Association Rule Mining for Machine Capabilities and Part Features

为了充分利用企业历史数据,提出一种基于二进制粒子群优化(BPSO)的关联规则挖掘方法,从数据中提取有用的工艺知识反映产品设计与制造的映射关系;对知识挖掘问题进行描述,建立用于制造过程综合的关联规则挖掘方法框架;对BPSO本身进行改进,提出考虑多个评价指标的适应度函数,并加入相似度指标以消除较差的规则,提高方法在实际问题中的适用性;将所提方法应用于汽车零部件机床加工数据的关联规则挖掘。结果表明,该方法与现有方法相比,平衡了可靠性、相关性及理解性等多个指标,能有效地进行规则挖掘。

To fully utilize the available historical data,a binary particle swarm optimization(BPSO)-based association rule mining(BPSO-ARM)method was proposed.The main idea was to extract knowledge that identified the mapping relationships between product design and manufacturing.Firstly,the research problem was described and the association rule mining framework was established.Improvements were made to enhance the search capability of BPSO.A fitness function considering multiple measures was designed.Moreover,an overlapping measure indication was further proposed to eliminate redundant rules to improve the applicability of the method.Finally,BPSO-ARM was applied to knowledge extraction for the automotive part manufacturing.The results show that BPSO-ARM outperforms other regular methods and is capable of mining association rules with a balance among various measures,including reliability,relevance and comprehensibility.