Original research

A SEQUENTIAL ANALYTICAL FRAMEWORK FOR DEFECT DIAGNOSIS IN DIRECT INJECTION PROCESS MANUFACTURING: APPLICATION IN THE FOOTWEAR INDUSTRY OF BANGLADESH

Pages 437-446

10.61552/JIBI.2026.04.007

Syed Tahmid Jamil, Md. Sarfuddin Anowar Khan, Md Tasfiq Alam, Md Jahid Hasan, Mohammed Jubaed

• View Article
• PDF 807 K

Abstract: In the growing footwear industry of Bangladesh, defect management in Direct Injection Process (DIP) production is still a challenge, especially in manufacturing production areas where data availability is limited and defect logs are the only means to get quality information. This study suggests a sequential analytical framework for systematic diagnosis of defect sources based on log-only production data, which includes Pareto analysis, machine-wise defect attribution, chi-square test of independence to determine the significant influences, Cramér's V effect size estimation to measure the influence size, and linear regression-based trend analysis. Applied across five injection machines over a 47-week observation period, with 39 weeks containing usable defect records. The four priority defect types determined were Short Injection, Spew, Cavity, and Upper Cut, which covered 78.3% of all recorded failures. The highest contribution of defects was from machine MC_3 with a statistically significant but negligible machine-defect association (Cramér's V = 0.077), suggesting that the machine identity is not a significant contributor to the defect variation, and the variation is due to the process-level variables. The slope of defect volumes as a function of time was determined and was found to be positive but not significant; this indicates that defect volumes are being driven up and should continue to be monitored. As defect counts were analyzed without production volume denominators, the framework is scoped to distribution diagnosis rather than rate estimation. The proposed framework offers a reproducible, data-driven diagnostic tool for quality improvement in manufacturing environments where comprehensive process data are unavailable.

Keywords: Trend analysis, Chi-square, Cramér's V, Machine attribution, Pareto analysis.

Received: 11.06.2026. Revised: 29.07.2026. Accepted: 31.08.2026.