Comparative study of computational frameworks for magnetite and carbon nanotube-based nanofluids in enclosure

Multi-wall carbon nanotubes (MWCNTs) characterize innovative nanoparticles that progress the thermal characteristics of base fluids, compelling them appropriate for utilizing in renewable energy, heat exchanger, and automotive engineering. In this analysis, the buoyancy-driven flow in a superposed spherical enclosure packed with amalgamated porous (Fe₃O₄-MWCNTs/H₂O) hybrid nanofluid layers was explored by employing the procedure of Levenberg–Marquardt with backpropagated artificial neural networks (LMB-ANN) for two temperature models. The exterior wall of enclosure was kept at a constant frigid condition, while the inner surface received partial heating to create a heat flux. The flow situation within the porous cavity was modeled using the Darcy–Boussinesq model. To evaluate the model equations, the control volume-based finite element method (CVFEM) was adopted. The results obtained from numerical method explain the reference data of LMB-ANN for several situations of porous cavity by modifying model variables. By varying the model parameters within the scope of the present numerical approach, a set of proposed data LMB-ANN is generated for cases. The proposed model has equaled for perfection after the numerical findings of various instances have been evaluated using the LMB-ANN train, test, and validating strategy. Several error graphs and statistical visualizations focused on mean square errors, error histogram, and regression assessment are designed to support the proposed methodology (LMB-NN). The proposed approach (LMB-ANN) has been verified based on the correlation of the suggested and benchmark (numerical) outputs, with a validity level ranging from 10^–02 to 10^–09. Also, the principal findings revealed that elevating the Rayleigh and Darcy numbers improves energy transmission inside the enclosure.