Intelligent predictive networks for MHD nanofluid with carbon nanotubes and thermal conductivity along a porous medium

Abstract

Recurrent neural networks have been able to capture the interest of the academia as they are able to compute very complex models which are non linear in nature. It is in this light that recurrent neural networks are well suitable for use in complex areas including fluid dynamics, biological computing, and biotechnology since they are capable of learning patterns. In this study we examine the ability of Levenberg-Marquardt algorithm with recurrent neural networks (LMA-RNNs) in simulating the MHD heat transfer properties of nanofluid consisting of carbon nanotubes model via porous media in terms of thermal conductivity. In the contemporary era SWCNTs and MWCNTs based nanofluids have many applications in drug delivery, cancer treatment, tissues regeneration, mechanical engineering, optical devices, electrically powered devices, and industrial production, particularly in solar thermal systems where heat transfer performance can boost energy capture and storage efficiency. The data used in this study is collected using Adams numerical technique and is then fine-tuned using the LMA-RNNs. The LMA-RNNs is performed with 80% of data to train the model and another 10% is used to test the model and the left over 10% is used for validation. In terms of plots for SWCNTs and MWCNTs further information is provided regarding the impacts of critical physical aspects on the field of velocity and temperature, and from the results presented in this article it is observed that an increase in suction parameter leads to a decrease in velocity and temperature profiles. It is found that increase in the form of higher Eckert number, it can come to an abstraction that the temperature profile increases as well. Furthermore, it is noticed that higher Pr values lead to lower values of thermal diffusivity, which mean that the thermal layer is thinner, or that there is less diffusivity of heat. For the assessment of the performance of the applied LMA-RNNs, the fitness of mean squared error, regression plots and error distribution in histograms is presented. The reduced MSE reveals that predictions of model are less likely to deviate from the true values and are more accurate this validates the proposed approach.