Physics-Informed Deep Learning Optimisation for Heat Transfer Analysis of Non-Similar Flow over Porous Moving Surface

Abstract

The investigation of fluid flow and thermal behaviour near continuous moving surfaces is an active and important area of research with direct applications across numerous engineering processes. Physics-informed neural networks gain a significant role in fluid dynamics and engineering applications due to their ability to integrate physical laws directly into the learning process of neural networks. This work expands the use of physics-informed neural networks to address thermal and momentum boundary layer issues of non-Newtonian fluid. The primary concern is to investigate momentum and thermal boundary layer under tangent hyperbolic non-Newtonian fluid flow circumstances. Due to the shear thinning and shear thickening behaviour of the non-Newtonian fluid, it has wide application in various domains. The physical constraints of the governing equations are solved using physics-informed neural networks. The models are constructed and trained using TensorFlow, and the predictions produced by this method are contrasted with those derived using Richardson extrapolation and adaptive moment estimation optimisation. The adaptive moment estimation optimiser is a well-known gradient-based optimisation technique that is frequently used in fluid dynamics, particularly when governing equations are solved using machine learning, deep learning, or physics-informed neural networks. The impact of various parameters has been analysed and shown graphically. The Lorentz force opposes fluid motion, producing a retarding effect that decreases velocity. The Darcy and Forchheimer drag forces increase with decreasing porosity, resulting in substantial momentum suppression and less convective heat evacuation, which raises the local temperature.