MACHINE LEARNING AIDED INTERPRETATION OF FACTORS AFFECTING NI-CATALYZED DRY REFORMING OF METHANE

Abstract

Design principles for active and stable catalysts are well-established for Dry Reforming of Methane reaction &40;DRM&41;&44; however&44; predicting selectivity of DRM over competing side reactions that alter the product H₂&47;CO ratio remains challenging&46; Screening catalysts through experiments is an empirical method&46; It is also time and resource intensive&46; Therefore&44; in this work&44; we devise a Machine learning &40;ML&41; model to predict the catalyst performance while interpreting the model using various techniques&46; We curate a set of 1638 data points from published literature and train tree regressor models to predict the H₂&47;CO ratio as a function of 11 process parameters and catalyst properties&46; The trained CatBoost regressors achieved the best prediction performance with R&178; &61; 0&46;91&46; Our exploratory study highlights ability of data&45;driven ML models to unearth structure&45;property relationships in heterogeneous catalysis by isolating effects of individual design parameters in a manner that would be difficult to achieve experimentally&46;