Quantifying Defect Escape Risk in Additive Manufacturing: The Impact of Sensor Uncertainty and Control Latency

Abstract

This article presents an engineering-oriented framework that models AM quality control as an end-to-end decision system, quantifying how uncertainty propagates from monitoring signals through feature extraction, thresholds, verification inspection, and corrective interventions into distributional outcomes that matter in production: probability of defect escape, probability of false alarm, scrap risk, rework burden, time-to-detection relative to layer deposition, and expected quality cost per part. A scenario-based quantitative study is developed for laser powder bed fusion (LPBF) manufacturing of safety-relevant components, comparing four quality control architectures: baseline post-build inspection, enhanced in-situ monitoring without governance, model-based anomaly scoring with limited drift handling, and a governance-optimized two-tier architecture that constrains nuisance alarms, uses drift-aware verification triggers, and applies staged corrective actions aligned with evidence strength and layer timing. Results show that (i) defect escape risk is dominated by decision latency and sampling limitations rather than by average signal quality, (ii) adding sensors without governance can increase scrap and operator fatigue due to nuisance triggers, and (iii) a governed two-tier approach reduces defect escape while stabilizing operational workload by using quantile-based alarm governance and early-stage corrective action windows. Up to three copy-ready tables and figure prompts are provided for Techne submission.