Machine Learning for Material Designs

Traditional material testing methods have been established and standardized for decades to study material capabilities and durability. However, material testing efforts are often laborious as each specimen requires tens to hundreds of repeated tests in order to generate sufficient statistical data. While standard material test methods provide a reproducible basis for characterizing material behaviors, the resulting data are insufficient to inform novel material designs. This limitation impedes progress in advanced material research – an integral area for maximizing the efficiency and functionality of engineered systems. The increasing interest and success in machine learning have motivated a new paradigm for material testing methods and designs, such that material properties can be extracted more efficiently and accurately. In this presentation, we will first review the development and application of deep learning methods in material science. Subsequently, we will present our initial work using physics-informed neural works for material identification. Lastly, we will conclude with thoughts on opportunities to further research in material testing and discovery.