This study aims to propose a reliable numerical framework for investigating the momentum and heat transfer behavior of a kerosene oil-based hybrid nanofluid consisting of silica and alumina nanoparticles, using similarity-based numerical modelling and physics-informed deep learning approaches.
The governing mathematical model, which describes combined flow and thermal fields by nonlinear partial differential equations, is converted into geometrical similarity transformations and finally reduced to coupled ordinary differential equations. Further, shooting technique combined with the Runge–Kutta scheme is used to solve the resulting boundary value problem and study the effects of main physical parameters on velocity distribution, temperature profile, skin friction coefficients and Nusselt number. The numerical finding is compared with benchmark solutions available in the literature for validation purposes. Moreover, a Physics-Informed Neural Network (PINN) is used to learn the solution of the reformulated equations by means of encoding the governing physics as part of the loss function.
This research results show that the performance of kerosene-based silica–alumina hybrid nanofluid in enhanced thermal transport is better than conventional nanofluids. It is confirmed that the model was correct, as the numerical solution based on a similarity transform fits well with published data. The PINN approach reliably predicts flow and heat transfer characteristics over the domain of parameters, verifying its robustness as a surrogate modeling method for solving nonlinear transport problems.
The authors proposed a similarity-transformed computational scheme using a PINN to investigate fluid transport dynamics in an oil-based hybrid nanofluid, which is unified numerical scheme that integrates classical numerical discretization approaches with deep-learning-driven, physics-constrained prediction.
