Research Information System
International Journal of Numerical Methods for Heat and Fluid Flow, pp.1-35, 2026 (SCI-Expanded, Scopus)
Purpose – The purpose of this study is to investigate the influence of different magnetic source locations on the flow behavior, heat and mass transfer and oxytactic bacterial distribution in a square cavity filled with Cu-water nanofluid, and to develop a neural network (NN) modeling of key transport characteristics. Design/methodology/approach – Four distinct magnetic source configurations, left-located, bottom-located, right-located and top-located are examined for different physical parameters by using polyharmonic spline radial basis function (RBF) method. A comprehensive data set of 42, 862 simulation cases is generated from numerical solutions and used for NN modeling of average Nusselt number, Sherwood number, and bacterial density. Model performance is evaluated using mean squared error (MSE) and feature importance analysis. Findings – The results reveal that the bottom-positioned source produces the weakest damping, reducing flow velocity by just 18.30%, whereas the top-positioned source generates the most significant suppression, decreasing velocity by 95.77%. As Hartmann number increases, heat transfer degradation varies across configurations, with left-located source showing the largest reduction at 38.32%, while bottom-located source experiences the smallest decline at 10.16%. NN models demonstrate excellent predictive capability, achieving MSE values on the order of 10−5 even when trained with only 50% of the data set. The feature importance analysis reveals that Rayleigh number is the most critical parameter for predicting heat and mass transfer, while bacterial density predictions are governed by the Peclet number. Originality/value – This study presents a novel combined numerical and machine learning framework for bioconvective nanofluid systems with varying magnetic source locations.