数据挖掘技术在轿车白车身装配偏差溯源中的应用

Application of Data Mining Technology in Body-in-white Assembly Error Affiliation of Passenger Car

提出一种将数据挖掘技术与装配工艺知识相融合的白车身装配尺寸偏差源快速诊断方法。该方法采用最大树法对测量得到的大量分散孤立的多维尺寸数据进行相似性聚类,通过多元统计分析进行偏差模式识别,以偏差主模式为基础提取相关产品、工艺、工装等深层知识,通过构造诊断决策树实现对白车身装配过程尺寸偏差源的快速诊断。实际应用表明,该方法有利于缩小问题空间,提供有效的诊断决策支持;能够克服单纯数据分析导致的偏差源误判与漏判,可快速准确地找到偏差源。

A rapid diagnosis method synthesized data mining technology and assembly process knowledge for body-in-white assembly dimension error sources is presented.It is to do similarity clustering to the massive dispersive and isolated multi-dimension data from measurements with the maximal tree method,to do error pattern recognition by multiple statistical analyses,to extract deep knowledge with regard to products,processes and tools based on the main error pattern,to realize rapid diagnosis of dimension error sources during body-in-white assembly process by constructing diagnosis decision tree.It is shown by practice that the method is of great advantage to narrow problem range,able to offer effective decision support,to overcome erroneous judgments and missing of error sources resulting from simple data analysis,and to find out error sources rapidly and correctly.