The existence of dual solutions in magnetohydrodynamic (MHD) convective flow is investigated, focusing on single and multi-walls carbon nanotubes (SMWCNTs). The examination assesses the impact of suction and injection on flow characteristics and heat transfer. The primary purpose of this study is to improve predictive accuracy for these phenomena through the application of a novel artificial neural network (ANN)-based methodology. This approach is intended to examine a more comprehensive understanding of complex fluid dynamic behavior.
This study examines a two-dimensional MHD flow with SMWCNTs in kerosene and water. The main equations are converted to ordinary differential equations via similarity alterations and solved numerically with the BVP4C algorithm to provide reference data. An ANN is trained with the backpropagation Levenberg–Marquardt algorithm to predict the local skin friction and local Nusselt number (LNN). The network uses 80% of the data for training and 10% each for validation and testing. Stability analysis reveals stable and unstable dual solutions due to suction effects.
The findings confirm the existence of a dual solution for suction, and the stability analysis is required to find out the physically possible solution. ANN-BLMA model has high level of accuracy, and the error of prediction is between 10–4 and 10–6. The regression plots, error histograms, function and autocorrelated analyses confirm that the predicted values of SFC and LNN are close to the numerical values. It has been established that comparative analysis of the ANN-based and numerical solutions has a high level of agreement with respect to various flow and thermal parameters. The mean error rate of the ANN model in predicting the skin friction coefficient is −0.21%, whereas the mean error rates of Sherwood and Nusselt number predictions are 0.01% and 0.05%, respectively. The values of MSE are 0.13e–10, 1.49e–11 and 0.22e–12, the gradient values are 3.351e–09, 2.031e–09 and 0.67e–09 and the mu values are 1e–09, 1e–09 and 1e–09, respectively. The numerical and ANN model is also compared to various data sets.
The study provides a new application of ANN-BLMA in the modeling and prediction of MHD nanofluid flow with carbon nanotubes a technique that has not been used before. The study uses a combination of both numerical and intelligent modeling methods to provide a very precise and computationally efficient method of solving complex heat/fluid flow problems in carbon nanotube-based systems. The results and methodology offer good insights in the future in terms of thermal management, energy systems and modeling of nanofluids.
