On the evaluation of the viscosity of nanofluid systems: Modeling and data assessment

Viscosity of nanofluids can significantly affect pumping power, pressure drop, workability of the nanofluid as well as its convective heat transfer coefficient. Experimental measurements of this property for different nanoparticles and base fluids at various temperatures is cumbersome and expensive. In this communication, a comprehensive review of the most important modeling works on viscosity of nanofluids including theoretical models, empirical correlations, and computer-aided models is conducted. Next, four multilayer perceptron (MLP) models optimized with Levenberg-Marquardt (LM), Bayesian Regularization (BR), Scaled conjugate gradient (SCG), and Resilient Backpropagation (RB), two radial basis function (RBF) neural network models optimized with Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), and one least square support vector machine (LSSVM) model optimized with coupled simulated annealing (CSA) were developed for the prediction of nanofluid viscosity based on 3144 data points. These data sets include 42 nanofluid systems under a wide range of operating conditions; including temperature from -35 to 80 degrees C, particle volume fraction from 0% to 10%, nanoparticle size from 4.6 to 190 nm, and viscosity of base fluid from 0.24 to 452.6 cP. Then, these seven models were combined in a single model using a committee machine intelligent system (CMIS). The proposed CMIS predicts all of the data with excellent accuracy with an average absolute relative error of less than 4%. Furthermore, the developed model was compared with five theoretical models and four empirical correlations through statistical and graphical error analyses. The results demonstrate that the proposed CMIS model significantly outperforms all of the existing models and correlations in terms of accuracy and range of validity. Finally, the quality of the experimental data was examined both graphically and statistically and the results suggested good reliability of the experimental data.