▎ 摘　　要

Regarding the importance of accurate predictions in industrial applications, this research aims to investigate the ability of artificial neural networks (ANNs) to carry out modeling and multi-criteria optimization of the rheological and thermophysical properties of an environmentally-friendly covalently functionalized nanofluid containing graphene nanoplatelets (CGNPs). In this contribution, different ANN structures are assessed and the NNs with 2-7-1, 2-4-1, 2-7-1 and 2-5-1 structures having a linear transfer function (purelin) and a hyperbolic tangent sigmoid (tansig) transfer function in the output and hidden layer are found to give the least difference between the network outputs and the experimental data for the thermal conductivity, viscosity, specific heat capacity, and density, respectively. Moreover, new correlations for thermal conductivity and viscosity of the nanofluid are proposed and the LASSO (Least Absolute Shrinkage and Selection Operator) and SVM (Support Vector Machine) methods are also presented for comparative purposes. It is observed that all models performed in a very comparable fashion, giving an idea of the easy nature of the problem. So it is recommended to train simple linear models like LASSO or SVM for such problems and perhaps the complex NNs are not necessary in some cases of practical prediction applications such as cooling or heating systems containing nanofluids. Furthermore, finding the optimal conditions is another crucial aspect of engineering problems. In this regard, a multi-criteria optimization of the hydrothermal characteristics of the nanofluid (i.e., to find the optimal cases with highest thermal conductivity and the relatively least viscosity) is conducted using the genetic algorithm coupled with a compromise programming approach. (C) 2017 Elsevier Ltd. All rights reserved.