Mass-based nanofluid transport in decelerating separated-stagnation point flow over EMHD Riga plate: shape-factor effects and LMB-ANN prediction
Abstract
This research examines the unsteady, decelerating flow and thermal transfer of a mass-based nanofluid across an electromagnetic Riga plate, directly contrasting spherical and cylindrical nanoparticles. The base-fluid mass is set at 𝑤 𝑓 = 100 g, and the nanoparticle loading is defined by mass ( 𝑤 𝑠 = 2.5–10 g). Thermal radiation, viscous dissipation, wall mass suction, and electromagnetic forcing are integrated. The converted boundary-value problem is solved using MATLAB's bvp4c, and the obtained velocity and temperature data are learned using a Levenberg-Marquardt backpropagation artificial neural network (LMB-ANN). Elevating 𝑤 𝑠 from 2.5 to 10 g enhances the decreased skin-friction metric by 10.41% for spherical particles and 95.85% for cylindrical particles; concurrently, the reduced Nusselt number varies by +0.97% and −16.23%, respectively. Increasing the suction from S = 0.5 to 2.0 raises the Nusselt number by 158.79% and 162.41%, respectively, and increases skin friction by 58.24% (sphere) and 48.28% (cylinder). Electromagnetic forcing from Z=0.5 to 2.0 more than doubles the skin friction (116.49% and 116.56%) but lowers the heat transmission by 70%. Radiation elevates the overall radiative–conductive Nusselt number by 27.15% for spherical particles and by 25.70% for cylindrical particles. At large Eckert number, the Nusselt number for the spherical particle changes sign suggesting a dissipation-induced thermal overshoot and local heat-flux reversal. The LMB–ANN generates mean-square errors for training, validation, and testing ranging from 2.71×10⁻¹⁰ to 6.44×10⁻⁹, indicating a precise replication of the bvp4c solutions. These results demonstrate the potential of combining particle mass, shape, and electromagnetic control to tailor the wall drag and heat transmission. Elevating α from 0.5 to 2.0 amplifies skin friction by 87.46% and 122.72%, as well as the Nusselt number by 168.88% and 152.51% for spherical and cylindrical particles, respectively.