深色名贵硬木家具零件分类成组研究  

Research on the classification of dark rare hardwood furniture parts

作　　者：任毅 王雨凡 杨越淳 张仲凤[1,2] REN Yi;WANG Yufan;YANG Yuechun;ZHANG Zhongfeng(Green Furniture Engineering Technology Research Center of National Forestry and Grassland Administration,Central South University of Forestry&Technology,Changsha 410004,Hunan,China;Green Home Engineering Technology Research Center in Hunan,Central South University of Forestry&Technology,Changsha 410004,Hunan,China)

机构地区：[1]中南林业科技大学国家林业和草原局绿色家具工程技术研究中心,湖南长沙410004 [2]中南林业科技大学湖南省绿色家居工程技术研究中心,湖南长沙410004

出　　处：《中南林业科技大学学报》2022年第8期166-175,共10页Journal of Central South University of Forestry & Technology

基　　金：湖南省科技人才托举工程中青年培养计划项目(2020TJ-Q18);湖南省林业科技创新计划项目(XLK201946);中南林业科技大学研究生创新基金项目(CX202102052)。

摘　　要：【目的】成组技术可以有效应对由于定制生产导致的效率低下问题,本研究基于零件的加工工艺,使用模糊聚类分析和BP神经网络技术提出了一种适合深色名贵硬木家具的零件分类成组方法,旨在实现零件合理分类成组,进而帮助企业提升定制生产能力。【方法】首先收集所有零件的工艺信息,将其汇总为零件-工艺表,在此基础上通过模糊聚类分析将零件聚类成组,形成零件族。再使用F统计量对成组结果做有效性评价,F统计量的数值越大代表聚类效果越好,企业可根据生产条件选择合适的成组结果。最后使用BP神经网络技术通过对旧有零件工艺信息和成组结果进行分析迭代,构建新型零件归组神经网络,实现其高效分类,准确归入现有的零件族。【结果】案例分析部分对MT公司柜类家具共118种工艺流程不同的零件进行分类成组。F统计量表明,当零件被分为5或16个零件族时成组效果较好,结合企业实际生产条件,5个零件族的成组方案为最优选择。将所有零件的工艺信息和5个零件族的成组信息加载至BP神经网络,该网络经8次迭代后误差降到1×10^(-6)以下,所测试的零件准确归组,达到实际使用标要求。【结论】本研究基于成组技术使用模糊聚类分析和BP神经网络技术,提出了一种深色名贵硬木家具零件分类成组的方法,该方法下零件可以按照加工工艺相似度分类成组,并实现新型零件高效归入现有零件族,帮助企业应用成组技术提升定制生产能力。【Objective】Inefficiencies in customized production can be effectively solved by group technology.Based on the processing technology of parts,this paper proposed a part classification grouping method suitable for dark rare hardwood furniture by using fuzzy cluster analysis and BP neural network.The method aimed to achieve the reasonable classification of parts into groups,thereby helping enterprises to improve customized production capacity.【Method】Firstly,the process information of all parts was collected and summarized to generate a part-process table.Based on that,the parts were clustered into groups through fuzzy clustering analysis to form a part family.Then the validity of the grouped results was evaluated by the F statistic,and the larger the F value statistic,the better the clustering effect.The enterprise can choose the appropriate grouping results according to the production conditions.Finally,in order to construct a new type of parts grouping neural network,the old part process information and grouping results were analyzed and iterated through the BP neural network,so as to achieve the efficient classification of parts and accurately classify them into the existing parts families.【Result】The 118 technological processes of the cabinet furniture parts of MT company were classified into groups.The F statistic showed that the grouping effect was better when the parts were divided into 5 or 16 families.Combined with the actual production conditions of the enterprise,it was shown that the grouping scheme of 5 parts families was the best choice.The process information of all parts and the group information were loaded into the BP neural network.After 8 iterations,the error of the network was reduced to less than 1×10^(-6).The tested parts were accurately grouped,which met the actual use standard requirements.【Conclusion】Based on grouping technology,through clustering fuzzy analysis and BP neural network,this paper has proposed a method for classifying dark rare hardwood furniture parts into g

关 键 词：深色名贵硬木家具 榫卯结构 成组技术 模糊聚类分析 BP神经网络 

分 类 号：S784[农业科学—林学]