Al2O3-CO2 Nanofluid Transport Properties: A Molecular Dynamics Study and Machine Learning Predictive Modelling

Abstract

Nanofluids are recognized to have significant difference in thermal/transport properties in contrast to the corresponding heat transfer fluids. The viscosity and thermal conductivity of carbon dioxide, which are the transport properties, play the vital role in swiftly growing applications of enhanced oil recovery process and industrial refrigeration. The current study presents different machine learning models to predict the transport properties of alumina-carbon dioxide nanofluid along with the molecular simulation approach. Several machine learning methods with linear regression, K-Nearest Neighbors, and Decision Tree are used to see the accuracy in determining these transport properties. The input variables taken to predict these transport properties are temperature, nanoparticle volume fraction and size. Molecular dynamics simulations using Large-scale Atomic/Molecular Massively Parallel Simulator are executed to determine the properties. Pearson correlation was established between the independent and dependent variables to check the dependency of the input variables on thermal conductivity and µ. Finally, we performed the statistical coefficients of determination to resolute the accuracy of the results obtained. It is concluded from the study that, the decision tree model with an accuracy of 99% is the best suited model for the prediction of transport properties of current nanofluid over the temperature range, volume fractions, and varied nanoparticle sizes.