Combining Neural Networks and Genetic Algorithms to Understand Composition–Microstructure–Property Relationships in Additively Manufactured Metals

Abstract

Additive manufacturing (AM) has revolutionized the production of complex metallic components by enabling the direct fabrication of intricate geometries from 3D model data. Despite its advantages in reducing material waste and customization of mechanical properties, AM faces challenges related to microstructural heterogeneity and mechanical property variability. This review highlights the structure–property relationships in additively manufactured metals, emphasizing how heterogeneous microstructure influences yield strength and fracture toughness. Phenomenological equations are provided based on the integration of neural networks and genetic algorithm-based models to predict mechanical properties from composition and microstructural features. We also outline key considerations such as acquiring high-fidelity datasets and understanding mathematical correlations within the data needed to formulate phenomenological equations.