Multi-objective optimization of rheological behavior of nanofluids containing CuO nanoparticles by NSGA II, MOPSO, and MOGWO evolutionary algorithms and group method of data handling artificial neural networks

Abstract

In this article, the ability of GMDH artificial neural networks (ANNs) to predict the rheological behavior (RB) of nanofluids (NFs) containing CuO NPs is studied. ANNs are a powerful mathematical tool that can identify the relationship among the parameters without the need to extract the relationship among them. The main purpose of this study is to use the GMDH ANN method to generate and predict the viscosity (mu) parameter using several input variables (IPV) such as solid volume fraction (SVF), nanoparticles (NPs), temperature (Temp), and shear rate (SR). By pairing the GMDH ANN with the evolutionary algorithm, this capability is created so that the values predicted by the ANN are more compatible with the laboratory numbers. The evolutionary algorithms (EAs) used in this study include three algorithms: Non-Dominated Sorting Genetic Algorithm II (NSGA II), Multi-Objective Particle Swarm Optimization (MOPSO), and Multi-Objective Grey Wolf Optimizer (MOGWO). These algorithms are selected for optimization, among which the best performance is related to the coupling of GMDH ANN with the MOGWO algorithm. In the next step, the Genetic Algorithm (GA), the Particle Swarm Optimization (PSO), and the Grey Wolf Optimizer (GWO) algorithms are used. This process is done to minimize the target function (TF) (mu) and evaluate the optimal points. According to the obtained results, among the EAs used in this study, the best performance belongs to the GA algorithm. Finally, in the last part of this study, the most optimal mode for IPV and output variable (OPV) of TF is determined. Numerically, the values of IPV data, such as SVF, T, and SR, are respectively 0.2242%, 50, and 246.7427, and the most optimal value for the OPV of TF (mu) was estimated as 0.96686 cP.