Technology
Review: Machine Learning Defect Detection in Powder-Based Metal Additive Manufacturing Concentrates in PBF, Evidence for DED/BJ Remains Sparse
2026-10-31
Materials Today Communications published an online review on October 31, systematically examining the application of machine learning (ML) in defect detection and classification for powder-based metal additive manufacturing, covering processes such as powder bed fusion (PBF-LB/M, PBF-EB/M), directed energy deposition (DED), and binder jetting (BJ). Methods include supervised, semi-supervised, and unsupervised learning, as well as CNNs, autoencoders, random forests, and GANs. Trend data: research papers in this field grew from 2 in 2015 to 234 in 2025; the review explicitly notes that ML evidence is mainly concentrated in PBF, while DED and BJ remain relatively sparse. Implications for powder AM industrialization: online quality closed-loop control is a key bottleneck for scaling production, and the review defines maturity differences in ML across processes, providing a roadmap for technology selection in monitoring, detection, and classification.
📰 News Source
This article was translated from a Chinese source.
Source URL not available
Show Chinese original
综述：机器学习缺陷检测在粉末基金属增材制造中集中成熟于PBF，DED/BJ证据仍稀疏
《Materials Today Communications》10月31日在线刊发综述，系统梳理机器学习（ML）在粉末基金属增材制造缺陷检测与分类中的应用，覆盖粉末床熔融（PBF-LB/M、PBF-EB/M）、定向能量沉积（DED）、粘结剂喷射（BJ）等工艺，方法囊括监督/半监督/无监督学习及CNN、自编码器、随机森林、GAN等。趋势数据：该领域研究论文由2015年2篇增至2025年234篇；综述明确指出ML证据主要集中于PBF，DED与BJ仍较稀疏。对粉末AM量产化的含义：在线质量闭环是扩产关键瓶颈，综述界定了各工艺ML成熟度差异，为「监测—检测—分类」技术选型提供路线图。