Advanced Neural Network Approach to Analyzing Heat Transfer and Chemical Reactions in Radiative Triple Diffusion Nanofluids Flow

Abstract

This study introduces a mesh-free framework for steady boundary-layer flow of a tangent-hyperbolic nanofluid over a stretching sheet, incorporating magnetohydrodynamic forcing, mixed convection, thermal radiation, and coupled heat, mass, and nanoparticle transport. Similarity transformations reduce the governing equations to nonlinear ordinary differential equations solved by a Gaussian neural network optimized via a hybrid cuckoo search algorithm; the network enforces momentum, energy, and species equations through residual minimization at Gauss collocation points while automatically satisfying boundary conditions. Validation against high-resolution numerical benchmarks yields mean squared errors below 10⁻⁸ and Nash–Sutcliffe efficiencies near unity, with velocity, temperature, solute concentration, and nanoparticle volume-fraction profiles matching reference solutions to within 10⁻⁶ pointwise error. Parametric studies show that increasing buoyancy accelerates near-wall flow and thins the hydrodynamic layer, whereas higher Hartmann and Prandtl numbers suppress convection and sharpen thermal decay; solute and nanoparticle distributions respond predictably to Schmidt, Soret, Brownian motion, and thermophoretic effects. This combined approach offers a computationally efficient tool for designing complex nanofluid systems with applications in thermal management, chemical processing, and environmental remediation.

Date of Award	2025
Original language	American English
Supervisor	Abdallah Berrouk (Supervisor)