Akademik Veri Yönetim Sistemi
Results in Engineering, cilt.28, 2025 (ESCI, Scopus)
This investigation presents the application of an artificial neural network (ANN) model to predict the flow and heat transfer parameters for Fe3O4-water nanofluid flow over a confined cylinder, subjected to a non-uniform magnetic source. In this ANN model, Reynolds numbers, 25 ≤ Re ≤ 150, Hartmann numbers, 0 ≤ Ha ≤ 22, volume concentrations of nanoparticle 0 % ≤ Φ ≤ 4 % and cylinder surface coordinates, 0˚ ≤ φ ≤ 360˚ were used as input variables, while the local Nusselt number, NuL drag coefficient, CD and separation angle, φs were selected as the output variables. The training and validation data for the developed ANN model were acquired from computational fluid dynamics (CFD) simulations, where the governing equations were analyzed utilizing the finite volume technique. The ANN showed perfect predictive accuracy, yielding a mean absolute error (MAE) of 0.894, root mean square error (RMSE) of 1.247 and correlation coefficient (R) of 0.985 during the prediction of NuL in the testing phase. Nevertheless, the developed model reached an MAE of 0.232, RMSE of 0.3307 and R of 0.996 during the prediction of CD in the testing phase. These outcomes demonstrated the high reliability of the ANN model’s estimations. One of the most important benefits of the recommended approach is its computational performance, decreasing the time needed for test case evaluations from 12 h (via CFD) to only 17 min using ANN. This situation is especially valuable in simulating complex, nonlinear flows where conventional CFD simulations often face convergence issues and stability problems. This work supplies a novel contribution by being among the first to employ an ANN approach for estimating hydrothermal flow parameters over the confined cylinder subjected to non-uniform magnetic fields and nanofluid flows.