Micrstruction Evolution and the Influence on Material Properties in Additive Manufacturing

Metal additive manufacturing (AM) processes face substantial challenges due to their inherent multi-physical nature and material-specific behaviors that vary across different alloy systems. This research establishes a comprehensive set of physics-based closed-form analytical models that fully describe the metal AM process chain, including: (1) precise thermal distribution modeling accounting for moving heat sources and boundary conditions; (2) microstructure evolution models capturing grain size development and crystallographic texture formation; (3) mechanical property prediction models for yield strength, elastic modulus, Poisson's ratio, and residual stress; and (4) defect models. All developed models have been validated against experimental results. 

The study specifically addresses critical industry challenges in processing parameter optimization, where traditional approaches relying on costly trial-and-error methods or computationally intensive numerical simulations often prove impractical for widespread industrial adoption. Our innovative hybrid approach strategically integrates these physics-based analytical models with optimized machine learning (ML) algorithms, creating a synergistic framework that simultaneously reduces computational costs by orders of magnitude, minimizes data acquisition requirements through intelligent sampling strategies, and decreases training expenses via physics-informed neural network architectures. The combined framework enables two-way prediction capabilities: (1) efficient forward prediction of microstructural characteristics and mechanical properties from given processing parameters, and (2) inverse optimization of manufacturing conditions to achieve desired material performance specifications. By successfully establishing quantitative relationships between processing conditions, microstructural evolution, and macroscopic properties, this work not only significantly enhances the industrial applicability of metal AM technologies but also establishes a new paradigm for integrated process-structure-property optimization. The hybrid physics-based/ML methodology offers manufacturers a practical and robust solution for real-world production challenges, while the fundamental insights gained pave the way for future advancements in predictive modeling and optimization strategies for metal AM across various material systems and industrial applications.