Inverse Design of Materials with Lab Automation and AI-driven Experimentation

Recent advances in machine learning models capable of learning process-structure-property (PSP) relationships paired with Bayesian Optimization have enabled inverse design of materials which is rapidly accelerating the design and discovery of fit-for-purpose materials. Inverse design allows properties to be selected first based on design criteria and then the system iteratively identifies optimal experiments to automatically discover or improve the required material. This talk will present foundational knowledge of the machine learning and optimization tools used in inverse design and then discuss the how NASA is leveraging these tools along with a robotic lab platform to accelerate the discovery and optimization of new materials that will enable future missions. The use of automatic image analysis to quantify microstructure to inform better PSP machine learning models will also be discussed.