Artificial neural network (ANN) analysis of non-similar solution of MHD nanofluid flow past a curved stretching surface

Abstract

The objective of this study is to predict the non-similar solution of Magnetohydrodynamics (MHD) mixed convection transports of viscous nanofluids over a curved stretching surface by using numerical simulation with an artificial neural network (ANN) algorithm, namely the back-propagation Levenberg–Marquardt scheme. In this article, aluminum oxide and iron oxide are used as nanoparticles to enhance the thermal conductivity of the base fluid. Additionally, aluminum oxide and iron oxide nanoparticles are used in certain biomedical applications, including composite materials, photocatalysis, wastewater treatment, and nanofiltration membranes. The radiation, Joule heating, and viscous dissipation effects are studied along with the convective boundary condition. The flow problem is attained in the form of a system of Partial differential equations (PDEs) and converted into dimensionless PDEs via non-similarity transformation. These dimensionless systems of PDEs are converted into a system of Ordinary differential equations (ODEs) via local, non-similar techniques up to second-order truncations. The Bvp4c technique is used to simulate the numerical results. The data sets are collected using the Bv-p4c technique. These data sets are utilized for the validation, training, and testing phases of the developed Back-propagated Levenberg-Marquardt Scheme- Artificial Neural Network (BLMS-ANN) model to determine the solution of the problem for a variety of physical scenarios. The ANN model is utilized to choose data, construct and train a network, and evaluate the performance of the network through the use of mean square error and regression analysis. The ANN models are used to predict the performance of the Nusselt number for both nanofluids and the results demonstrated that ANN is capable of making accurate predictions of the local Nusselt number for both nanofluids. In the entire study, the ANN model predicts the result of the local Nusselt number of nanofluids with higher accuracy as compared to nanofluids.