End-to-End Quality Control in Additive Manufacturing: Evaluating Defect Detectability, Dimensional Uncertainty, and Qualification Risk Across the Digital Thread

Abstract

This article presents a reliability-centered quality control framework for AM that treats quality as a decision system spanning in-process monitoring, post-process metrology, and acceptance logic, and that explicitly quantifies uncertainty propagation from process signals to defect detectability and dimensional compliance. A scenario-based quantitative study is developed for powder bed fusion production, comparing quality control strategies that combine melt pool monitoring, layerwise imaging, computed tomography sampling, and coordinate measurement verification. The study uses engineering metrics that translate directly to production decisions, including probability of defect non-detection, probability of tolerance exceedance, time-to-disposition, and cost of quality under false-reject and false-accept trade-offs. Results show that (i) monitoring value is maximized when it is calibrated to the defect and geometry mechanisms that dominate part performance rather than treated as a generic anomaly detector, (ii) decision reliability is governed by how thresholds are engineered under controlled false alarm rates and how measurement systems are validated, and (iii) hybrid inspection policies that allocate high-resolution metrology to the highest-risk builds based on monitored uncertainty can reduce total cost while increasing acceptance confidence. The paper concludes with implementable guidance for designing AM quality control architectures as reliability systems rather than as isolated sensing upgrades.