Machine Learning in Additive Manufacturing: A Review of Process Optimization and Strength Prediction

Additive manufacturing (AM) has evolved from a rapid prototyping technique into a key production technology for complex, high-performance components in aerospace, automotive, biomedical, and energy sectors. However, the mechanical reliability of AM parts remains highly sensitive to process parameters, thermal history, and defect formation mechanisms. Traditional empirical and physics-based models struggle to describe the nonlinear and multidimensional interactions that arise in AM processes, limiting their ability to predict strength and structural integrity. Machine learning (ML) has emerged as a powerful alternative, capable of learning complex process–property relationships directly from data and providing accurate predictions for tensile strength, porosity, surface roughness, hardness, and dimensional accuracy. This review synthesizes recent advances in ML applications across polymer-, metal-, and ceramic-based AM technologies, focusing on process parameter analysis, mechanical strength prediction, defect monitoring, and parameter optimization. The discussion highlights commonly used ML algorithms, sensor integration strategies, and hybrid optimization approaches, and identifies key research gaps related to dataset scarcity, model generalization, interpretability, and cross-platform reproducibility. Finally, the review outlines future directions, including digital twins, physics-informed ML, and reinforcement learning, to enable autonomous, industrial-grade intelligent additive manufacturing systems.