• Studying the behavior of penta-hybrid nanofluids in fully 3D MHD flow including the effects of radiation and slip. • Creating a deep learning surrogate model able to imitate complex numerical solutions with very high precision. • Monohybrid, hybrid, tri-, tetra- and penta-hybrid nanofluids were subjected to a unified comparative evaluation at the same physical condition. • Merging of physics and AI technique has been introduced which will provide fast predictions, stability and generalization advantages over purely numerical methods. The research creates a numerical and artificial intelligence framework to study the three-dimensional magnetohydrodynamic (MHD) flow and heat transfer properties of mono-, hybrid-, tri-, tetra-, and penta-hybrid nanofluids (PHNFs) which flow over a bi-directional stretching surface while experiencing thermal radiation and viscous dissipation and velocity slip effects. The study addresses the growing need for efficient predictive tools that can handle highly nonlinear multi-parameter nanofluid systems, which require expensive computational resources when solved through traditional numerical methods. The governing nonlinear boundary-layer equations are transformed into a system of ordinary differential equations through similarity transformations, which are solved numerically with a boundary value problem (BVP) solver. The generated dataset is used to train a deep learning (DL) surrogate model which was implemented in Python to predict velocity and temperature profiles across a wide parametric space. The results show that increasing the magnetic parameter causes a decrease in vertical velocity while horizontal flow patterns experience changes because of complex three-dimensional interactions. The velocity slip parameters determine how momentum gets redistributed between flow directions. PHNFs demonstrate higher temperature distributions and thicker thermal boundary layers because their effective thermal conductivity improves through multiple nanoparticle interactions. The heat transfer performance improves because Nusselt numbers increase when compared to basic nanofluids. The trained DL model demonstrates strong predictive capability with coefficient of determination values exceeding 0.98 and low mean square errors across all scenarios which confirms its effectiveness as a surrogate for complex numerical simulations. The proposed hybrid framework delivers an effective and dependable method for studying advanced nanofluid systems while its built-in system provides solutions for thermal management and energy systems and engineering design optimization.
Shah et al. (Fri,) studied this question.