Predicting Physical Properties of Nanofluids by Computational Modeling Natalia Sizochenko Michael Syzochenko Agnieszka Gajewicz Jerzy Leszczynski Tomasz Puzyn 10.1021/acs.jpcc.6b08850.s003 https://acs.figshare.com/articles/dataset/Predicting_Physical_Properties_of_Nanofluids_by_Computational_Modeling/4558621 The focal point of the current contribution was to develop global quantitative structure–property relationship (QSPR) models for nanofluids. Two target properties, thermal conductivity and viscosity of nanofluids, were thoroughly investigated. Under this investigation, a new database of thermal conductivity and viscosity of nanofluids (more than 150 data points) was created. A hierarchical system of molecular representation reflecting features of nanoparticle’s structure at the different levels of organization was introduced. Also, size-dependent, volume-dependent, and intensive parameters were calculated. The model for thermal conductivity is characterized by determination coefficient <i>R</i><sup>2</sup> = 0.81 and root-mean-squared error RMSE = 0.055; the model for viscosity is characterized by <i>R</i><sup>2</sup> = 0.79 and RMSE = 0.234. Developed models are in agreement with modern theories of nanofluids behavior. Size- and concentration-related behavior of target properties were discussed. Findings suggest that the increase in surface area ratio and interfacial layer thickness and decrease in nanoparticles size lead to thermal conductivity and viscosity increase. Thermal conductivity and viscosity increase with an increase in weighted fraction-dependent parameters. Up-to-date, reliable theoretical models were created only for a single type of nanoparticles. In this article, developed models can simultaneously predict the thermal conductivity and viscosity in an effective way using both size and volume concentration of nanofluid. 2016-12-27 00:00:00 QSPR root-mean-squared error RMSE nanoparticle target properties conductivity surface area ratio viscosity increase nanofluid 150 data points determination coefficient R 2 model