Search

In order to avoid the costs of experimental evaluations, soft computing methods like artificial neural network (ANN) and genetic algorithm have remarkably grabbed the attentions of investigators for predicting the hydrothermal characteristics of different types of nanofluids. In this paper, the implementation of ANN and genetic algorithm for modeling and multi-criteria optimizing the hydrothermal behavior of SiO 2 /water nanofluid has been investigated. Using the data obtained from the experimental analysis, an ANN model is developed to estimate the pressure drop and Nusselt number as a function of volume concentration, Reynolds number, and inlet temperature. Different network structures were assessed and it has been achieved that a network with 2 hidden layers and 6 neurons in every layer provides the most accurate prediction. It is revealed that the developed network is satisfactorily accurate to determine the Nusselt number and pressure drop of SiO 2 /water nanofluid compared to the empirical correlations. To optimize the hydrothermal behavior of the nanofluid (i.e. to find the optimal cases with highest Nusselt number and the relatively least pressure drop), the genetic algorithm coupled with compromise programming approach has been implemented considering decision maker's attitude. [ABSTRACT FROM AUTHOR]

The costly and time-consuming determination of thermophysical properties of nanofluids through the experimental analysis leads the current investigations to use the soft computing methods like correlating, artificial neural network (ANN) and genetic algorithm. In this study, the application of ANN, empirical correlations and genetic algorithm for modeling and multi-criteria optimization of the thermophysical properties of clove-treated MWCNTs nanofluid which has been synthesized through a facile and eco-friendly procedure has been investigated. In this contribution, totally 6 structures are assessed: networks including one and two hidden layers with 2, 4, and 6 neurons. From assessment of the ANN, it is found that the network including two hidden layers with 4 neurons in every layer results in the least difference between the network outputs and the experimental data, providing the best performance. It is concluded that the optimal ANN model is a more precise and accurate way to predict the thermal conductivity and viscosity of environmentally friendly C-MWCNT/water nanofluid compared to empirical correlations obtained from non-linear regression method. Moreover, the evolutionary algorithm has been implemented for achieving the optimal conditions to maximize the thermal conductivity and to minimize the viscosity of nanofluid. In this regard, based on real-world engineering experience, the final optimal solutions opted from several distinguished procedures of decision-making including the Bellman-Zadeh, TOPSIS and LINMAP approaches has been investigated in parallel. The results of this study revealed that the obtained outputs using TOPSIS and LINMAP procedures are the closest to ideal solution. [ABSTRACT FROM AUTHOR]