Application of association rules in woven wire mesh defects analysis

In this study, association rule algorithm, one of data mining techniques, was applied to analyze the relationship between manufacturing defect types. Association rules technique is a technique to uncover relationships between data variables which have been applied to various problems both in manufacturing industry and services. In this paper, FP-Growth (frequent-pattern growth) algorithm has been applied to a large manufacturer of stainless steel wire mesh in Thailand. The company is experiencing a defects problem that the cause remains unknown. Defect data are available but there has not been analyzed properly. There are in total 19 types of defects. 2,281 records of defect data on mesh 0.8mm grade 316 products were collected and used for association rule mining. The result showed some relationship between three main types of defects which are HP (hard warp) defect, OM (open mesh) defect and OF (open mesh full) defect with difference confidence level. Moreover, if HP defect occur it is possible that OM defect will also occur at 96.3 percent confidence. On the other hand if OM defect occur it is possible that HP defect will also at 99.9 percent confidence. These results can be used to develop the production processes to monitoring of defects, solving a root cause of defects, defects inspection method and judgment criterion in future.