DATA-DRIVEN PROCESS OPTIMIZATION IN AUTOMOTIVE MANUFACTURING A MACHINE LEARNING APPROACH TO WASTE REDUCTION AND QUALITY IMPROVEMENT

Abstract

Data-driven process optimization in automotive manufacturing was examined as a quantitative approach for reducing waste and improving product quality by linking integrated production, quality, maintenance, and batch-traceability data with measurable outcomes. The study was positioned through a structured review of 35 peer-reviewed papers and was executed as a retrospective case analysis using 48,720 eligible production-unit records joined by unit ID, station identity, and synchronized timestamps. Descriptive findings indicated scrap occurrence of 2.60% (n = 1,268), rework routing of 12.55% (n = 6,112), first-pass yield of 87.45% (n = 42,608), and defect occurrence of 6.15% (n = 2,994) at the selected inspection gate; defect cases were distributed across dimensional 36.0% (n = 1,077), weld 24.0% (n = 718), surface/paint 22.0% (n = 659), and functional 18.0% (n = 540) categories. Correlation analysis showed that rework hours per unit were positively associated with torque deviation (r = 0.46) and dimensional deviation (r = 0.41) and negatively associated with a sensor-based stability index (r = −0.38). Measurement reliability supported quantitative interpretation, including inspection label agreement of 0.97, defect-category consistency (κ = 0.78), and high repeatability for key trace-derived indicators (ICC = 0.89). Collinearity screening identified redundancy among correlated telemetry summaries, with maximum VIF reduced from 9.6 to 3.6 after consolidation. Regression results indicated higher defect odds with torque deviation (OR = 1.28, 95% CI [1.21, 1.35]) and dimensional deviation (OR = 1.42, 95% CI [1.31, 1.54]), while stability reduced defect odds (OR = 0.81, 95% CI [0.77, 0.85]). Block wise modeling showed improved explanatory power from context-only (adjusted R² = 0.08) to full models including process, equipment, and batch predictors (adjusted R² = 0.33), supporting an integrated quantitative framework for waste reduction and quality improvement.