A META-ANALYSIS OF MACHINE LEARNING-ENHANCED LEAN QUALITY CONTROL PRACTICES IN MANUFACTURING: OPTIMIZING DEFECT DETECTION AND PROCESS EFFICIENCY

Abstract

The convergence of machine learning (ML) and lean quality control (LQC) represents a transformative shift in modern manufacturing, offering the potential to significantly enhance defect detection accuracy, reduce process waste, and improve overall operational efficiency. While individual studies have reported promising results from the application of ML in specific industrial contexts, a systematic synthesis of these outcomes has been lacking. This meta-analysis bridges that gap by evaluating 112 empirical studies published between 2010 and 2025, spanning multiple manufacturing sectors including automotive, electronics, textiles, pharmaceuticals, and aerospace. Defect detection accuracy improved by 18% to 45%, rework and scrap were reduced by up to 40%, and unplanned downtime declined by 25% to 50% following ML integration. Moreover, FPY and OEE showed measurable gains of 15% to 30% and 10% to 20%, respectively, while inspection time was reduced by up to 60%, enabling more agile and synchronized production cycles.  However, notable gaps were identified, including inconsistent methodology, limited cross-sector validation, and disparities in adoption between large enterprises and SMEs. Furthermore, concerns surrounding data governance, model explainability, and workforce integration remain underexplored, posing potential barriers to widespread adoption. The meta-analysis offers critical insights for researchers, engineers, and policy-makers seeking to operationalize artificial intelligence within lean production systems and sets the groundwork for future research on scalable, ethical, and inclusive ML applications in industrial quality control.