Identifying Quality Factors for Surface Defects on Small Steel Bars Using a Two-step Method of Statistical Difference Testing and k-means Clustering

Small steel bars (SSBs) are common steel products, and their surface defects pose a significant quality issue. Despite numerous studies conducted to detect and predict such defects, identifying the underlying causes, or 'quality factors,' remains a challenge. Most of the existing methods focused on validation of known factors based on the theories of physical properties and material composition; thus, they lack exploring new quality factors. This study proposes a two-step method for identifying these quality factors, considering two common issues that often arise when analyzing operational data in SSB manufacturing processes: 'merged measurement' and 'variation due to operation dates.' These issues are resulted because the rolling process is lack of measurement capability and exposed to external environment. They make it difficult to use various analytical methods such as statistical analysis and data mining methods. The proposed method sequentially employs statistical difference testing and k-means clustering, thereby providing a practical heuristic for overcoming the aforementioned difficulties. The effectiveness of the proposed method was demonstrated through a case study that highlighted its ability to effectively identify the quality factors critical to surface defects on SSBs.