MHD convective thermal performance in a nanofluid-filled wavy enclosure: An artificial neural network model

The present study intends to observe the effects of natural convection on fluid flow and thermal performance in an undulating enclosure subjected to a nanofluid, addressing the critical need for enhanced heat transfer mechanisms in engineering applications. This study might optimize thermal management systems, especially in heat-dissipating sectors. The Galerkin type's numerical scheme finite element technique has resolved the non-dimensional governing non-linear equations. The neural network approach is applied to investigate steady two-dimensional free convection flows inside a wavy enclosure. The model's performance is assessed using statistical metrics, the determination coefficient (R²), and mean squared error (MSE). The simulation results will be shown graphically as streamlines, isotherms, undulation effectiveness, and mean Nusselt numbers. The effect of substantial parameters such as the Rayleigh number (10 3 ≤ Ra ≤ 10 6 10 3 ≤ Ra ≤ 10 6 ), Hartmann number (0 ≤ Ha ≤ 100), undulation number (0 ≤ λ ≤ 3) and the nanoparticle volume fraction (0.02 ≤ φ ≤ 0.05)on heat transmission and fluid flow will be examined at constant Prandtl number 6.83. The results reveal that increasing the Rayleigh number enhances heat transfer rates significantly, with the mean Nusselt number showing an increase of up to 35 % as Ra rises from 10 3 to 10 6 . Additionally, the study finds that higher nanoparticle volume fractions lead to improved thermal performance, with a notable increase in heat transfer efficiency. Conversely, the Hartmann number's increase reduces heat transfer rates, highlighting the complex interplay between these parameters. This research indicates a correlation between the actual and predicted Nusselt values, indicating the accuracy and solidity of the model under various circumstances. This innovative Fe 3 O 4 nanoparticles and MHD study of undulating geometries provide vital insights for improving thermal systems in complicated schemes. The results of the current investigation and the previously published work show an excellent level of consistency.