Performance analysis of backpropagated networks for entropy optimized mixed convection nanofluid with second-order slip over a stretching surface

Abstract

Due to their excellent generalization performance, artificial neural networks (ANN) are becoming more and more popular for solving difficult nonlinear real-world scientific and engineering challenges. The article in hand addressed the comparative study of artificial intelligence-based backpropagated neural networks (ABPNN) with Levenberg Marquardt scheme (LMS) and Bayesian Regulation scheme (BRS) for mixed convection-based nanofluid flow model (MCNM) optimized with entropy and second-order slip past a stretching sheet. The system is converted into an ordinary differential system and then solved with the Homotopy analysis method (HAM) to generate datasets by varying different parameters. The datasets are divided into training, testing, and validation to find the solution through LMS and BRS and further compared with reference technique i.e Homotopy analysis. The proficiency of the proposed ABPNN is proved by error histogram, mean squared error (MSE), and regression fitness analysis of LMS and BRS. Furthermore, flow field, concentration, thermal gradient, heat transfer rate, Sherwood number, Bejan number, skin friction, and entropy generation are found graphically and numerically. The results indicate that for higher values of slip parameters the velocity decays. Ec and Nt behave as a growing function of temperature. The fluid’s concentration is increased by an increase in E, whereas K shows an opposite trend. As the slip parameters rise, the system’s overall entropy reduces. Bejan number indicates a rising trend against slip parameters. For higher Nt and E, heat and mass transfer rates decrease. The solution is accurate up to 3–7 decimal places for all scenarios which are considered to fulfill the consistency standards.