Optimal Selection of Parameters for Production of Multiwall Carbon Nanotubes (MWCNTs) by Electrolysis in Molten Salts using Machine Learning

Abstract

The production and use of carbon nanotubes (CNTs) have become extremely wide within the last decade. Hence, the high interest in producing non-expensive and quality CNTs has motivated many research projects. This research considers the design and development of new technology for producing MWCNTs by electrolysis in molten salts using non-stationary and stationary current regimes. The electrolysis is simple, ecological, economical, and flexible, and it offers possibilities for accurate control of various parameters, such as applied voltage, current density, or temperature. We infer the underlying relationship between the parameters and the quality of the experimentally produced MWCNTs by using explainable tree-based Machine Learning (ML) models. We train several models in a supervised manner, whereas in model covariates, we use the parameters of the MCWNTs, and as a target variable, the quality of the produced MWCNT. Domain experts label all the experimental examples in our data set. Controlling these parameters enables high-yield production and, particularly important, obtaining MWCNTs, which are up to ten times cheaper than other existing technologies.