Chemical and Biomolecular Engineering
Faculty Advisor: Yamil J Colón
Automated Detection of Defects in Porous Materials with Machine Learning
It is understood that the presence of defects can be a determining factor for a multitude of energy-related applications including catalysis, gas storage, separations, etc. Determining the presence and quantity of defects in a material can be a challenging endeavor requiring advanced characterization techniques. Our project seeks to identify and quantify defects in porous materials with standard characterization techniques like a simple adsorption isotherm using machine learning algorithms. We train anomaly detection algorithms to determine if a material has a defect and how many are present. We are currently testing algorithms on data sets that contain known numbers of anomalies and are moving on to characterizing material properties. By the end of this project we will have tested our algorithms on data from molecular simulations and experimental systems. We expect our results will be of great value to the porous materials and energy communities.