Applications of Machine Learning in Additive Manufacturing

Abstract

The research aims to solve common challenges in additive manufacturing processes, including strong multi-parameter coupling, difficulty in suppressing defects, and unstable performance, while overcoming the limitations of traditional empirical trial-and-error and physics-based modeling approaches and advancing additive manufacturing toward intelligent, automated, and closed-loop upgrades. The methodology centers on machine learning’s data-driven paradigm, integrating various algorithms such as neural networks, random forests, and support vector machines. From the full life cycle of additive manufacturing, the study systematically explores technical pathways and application modes in process monitoring, defect control, parameter optimization, performance prediction, structural design, and material development. Results show that machine learning can effectively construct the “process–structure–performance–defect” mapping relationship in additive manufacturing. In online monitoring, defect identification accuracy exceeds 95%; process development cycles can be shortened from weeks to days, and new material R&D cycles from months to weeks. High-precision prediction of component mechanical properties is also achieved, with some models attaining R² values above 0.9. The conclusion is that machine learning provides a full-process data-driven optimization solution for additive manufacturing and effectively addresses many pain points of traditional methods. Current applications in this field still face challenges such as insufficient data quality and quantity and poor model interpretability. Future efforts should focus on physics-informed machine learning, digital-twin closed-loop control, and other directions, while advancing standardization, to promote the engineering implementation and scaled production of intelligent additive manufacturing and provide support for the development of high-end equipment, biomedicine, and other fields.