Integrating In Situ Monitoring, Physics Simulation, and AI Analytics for Quality Control in Additive Friction Stir Deposition

As a new solid-state additive process, additive friction stir deposition (AFSD) circumvents the challenges of fusion-based metal additive manufacturing through severe plastic deformation at elevated temperatures, resulting in fully-dense deposition with refined microstructures and forging-standard mechanical properties. However, maintaining consistent part quality for large builds requires a comprehensive understanding of the intricate thermomechanical interactions during deposition. Here, I present a framework that synergizes real-time monitoring, physics-based modeling, and AI-driven analytics to enable real-time quality assurance and process optimization in AFSD. In situ monitoring techniques, including infrared imaging and force/torque sensing, provide critical data on temperature distribution and material flow, which refines physics-based process modeling and improves prediction capability. AI-driven approaches, particularly physics-informed Bayesian learning, can further enhance prediction accuracy and speed by integrating prior knowledge with real-time data, while calibrating unknown physical parameters and quantifying prediction uncertainties. Beyond conventional wisdom, this explainable AI framework facilitates in-depth understanding of the underlying physics by identifying deviations between model predictions and observed behavior. By connecting both physics-based and data-driven modeling insights, the proposed framework will transform AFSD through robust thermal control, improved part consistency, and inverse processing design, which will accelerate the adoption of this technology in aerospace, defense, and other high-performance applications.