Artificial Intelligence in Additive Manufacturing: A Critical Review of Methods, Applications and Industrial Challenges

Additive Manufacturing (AM) has grown from a rapid-prototyping tool into a legitimate production technology, yet its broader industrial adoption continues to be constrained by process variability, stochastic defect formation, and the near-impossibility of mapping complex Process Structure Property (PSP) relationships through conventional experimental methods. Artificial Intelligence (AI) and Machine Learning (ML) have emerged as the primary technical response to these constraints, offering data-driven pathways to optimize parameters, detect anomalies in real time, and reduce development lead times substantially. This review examines AI integration across the complete AM workflow: generative design and topology optimization in the pre-processing stage; surrogate modeling, Bayesian optimization, and reinforcement learning (RL) during the build stage; convolutional neural network (CNN)-based in-situ monitoring and sensor fusion for quality assurance; and AI-enhanced Digital Twins (DTs) for post-process certification. Representative quantitative outcomes from the literature are synthesized into comparative tables, and persistent barriers-particularly data scarcity, model interpretability, and the absence of industry-wide standardization-are critically analyzed. The review concludes that federated learning, edge deployed inference, and explainable AI (XAI) represent the most promising near-term research directions for achieving autonomous, self-correcting AM systems. This is a critical-synthesis review. Targeted searches of Scopus, Web of Science, and Google Scholar using 'additive manufacturing', 'machine learning', 'artificial intelligence', 'digital twin', 'in-situ monitoring', and process-specific terms (e.g., 'laser powder bed fusion', 'directed energy deposition') identified publications from 2015-2026, with an emphasis on 2023-2026 work Relevance, methodological rigor, and current literature influenced source selection.