Data-Driven Identification of Product Quality Defect Drivers in Manufacturing Systems

Md Foysal Hossain; S M Habibullah; Md Jahedul Islam Arif; Maksud Miah; Ariful Islam; Fahim Rahman; Al Akhir

doi:10.5935/jetia.v12i59.3603

Authors

Md Foysal Hossain Industrial and Systems Engineering Lamar University, Beaumont, Texas, USA. https://orcid.org/0009-0009-9906-5376 (unauthenticated)
S M Habibullah Industrial and Systems Engineering Lamar University, Beaumont, Texas, USA. https://orcid.org/0009-0008-8094-332X (unauthenticated)
Md Jahedul Islam Arif Industrial and Systems Engineering Lamar University, Beaumont, Texas, USA. https://orcid.org/0009-0009-0104-6565 (unauthenticated)
Maksud Miah Business Analytics and Systems University of Bridgeport, Bridgeport, Connecticut 06604, USA https://orcid.org/0009-0003-0130-038X (unauthenticated)
Ariful Islam Business Analytics and Systems University of Bridgeport, Bridgeport, Connecticut 06604, USA https://orcid.org/0000-0001-6609-3324 (unauthenticated)
Fahim Rahman University of Findlay, Findlay, Ohio 4584, USA. https://orcid.org/0009-0002-5007-2187 (unauthenticated)
Al Akhir Business Analytics and Systems University of Bridgeport, Bridgeport, Connecticut 06604, USA https://orcid.org/0009-0001-3721-8439 (unauthenticated)

DOI:

https://doi.org/10.5935/jetia.v12i59.3603

Abstract

The Product quality defects lead to substantial operational inefficiencies, financial losses, reputational costs, and customer dissatisfaction in contemporary manufacturing systems. Although Industry 4.0 environments generate extensive operational data, there remains limited empirical clarity regarding which factors most strongly drive defect occurrence and how such insights can support manufacturing decision-making. This study presents an empirical, data-driven assessment to evaluate operational, quality, and supply chain factors that influence product defects. The analysis is based on 3,240 manufacturing observations and 17 operational variables spanning production, maintenance, quality, efficiency, and supply-chain dimensions. Descriptive statistics and correlation analysis were conducted, followed by a Random Forest classifier to rank influential predictors of defect occurrence. Analysis of variance (ANOVA) was used to examine the effect of delivery delays. Model performance was evaluated using precision, recall, and F1-score. Maintenance hours and defect rate jointly accounted for approximately 44% of predictive contribution, with the model achieving precision and recall above 95% for defective items. Delivery delay showed no significant effect. The findings support the use of machine-learning (ML)-based analytics for data-driven quality improvement and process optimisation in manufacturing systems.