Application of nano-quantitative structure-property relationship paradigm to develop predictive models for thermal conductivity of metal oxide-based ethylene glycol nanofluids

In this work, the relatively thermal conductivity of metal oxide-based ethylene glycol nanofluids is being predicted by using quantitative structure-property relationship methodology. The structural features of studied nanoparticles are represented by quasi-SMILES which is a coded linear structure. The gathered dataset includes ten types of nanoparticles (including Al2O3, MgO, TiO2, ZnO, Co3O4, CeO2, CuO, Fe2O3, Fe3O4, and SnO2) suspended in the same base fluid, ethylene glycol. The calculated optimal descriptors acquired by applying the Monte Carlo method in the free software available on the Web (named CORAL) and four random splits into the training, invisible, calibration, and validation sets were appraised. The statistical characteristics confirmed the predictive power and reliability of the developed models; all splits had R-m(2) more than 0.5 and Delta R-m(2) less than 0.2, and also the validation set showed the correlation coefficient (R-2) in ranges 0.8611-0.6816 and cross-validated correlation coefficient (Q(2)) in ranges 0.8518-0.6668. The presented models accurately predicted the thermal conductivity of all considered nanofluids, and the technique is expected to provide a novel way for future theoretical projects.