Enabling Materials-by-Design using Machine Learning

Kurzfassung

In order to manufacture/create materials with specific target properties it is critical to establish robust and accurate linkages between how the material is made (i.e., the material processing) and its resultant structure, as well as linkages between the structure and its corresponding property. Existing capabilities throttle the development of these linkages given that they require costly experiments and computationally intensive high-fidelity large-scale simulations. Using machine learning, we can address these critical bottlenecks given its unparalleled ability to learn complex nonlinear behavior. In this talk we will showcase how machine learning can extract and reuse knowledge from existing experimental and simulation data to build fast, accurate linkages that reduce months of testing or computing to seconds and enable the unique capability to determine where new data would be most valuable. We will specifically highlight two examples: (i) a set of data-driven models to optimize tribological alloy design by leveraging high-throughput experiments and simulations, and (ii) the innovative Materials Data Driven Design (MAD3) software solution for rapid prediction of the load-dependent behavior of metals that is revolutionizing the forming and stamping industry. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525. SAND No. SAND2026-20766A