AI-assisted PDFP-based PINNs for heat regulation of nanofluid in industrial cooling enclosures around circular barriers

Abstract

Significance:

This study explores the vital aspects of applied side of heat transfer in different areas like in heat exchangers, electronic device cooling, and energy storage units. This article is significant because it investigates the flow around circular obstacles and makes vital contribution for modern industrial systems.

Cause of the article:

This article implements the Physics Informed Neural Networks (PINNs) on established study related to heat transfer enhancement in diamond based nanofluid inside a square cavity having a fixed circular obstacle. The inclusion of circular obstacle induces localized temperature gradient sand recirculating vortices. Basically, this article tells how geometric barriers influence convective heat transfer. Furthermore, it computes the involved mathematical model using data-efficient PINNs optimization

Scheme:

The mathematical model based on square cavities configuration around circular obstacle is formulated and system of Partial Differential Equations (PDEs). After making them dimensionless, the PINNs framework has been utilized to get the predicted results. Preconditioned Davidon–Fletcher–Powell (PDFP) optimizer is employed to enhance convergence rate, prediction accuracy and training stability of the PINN framework.

Key outcomes:

Conclusion of this attempt shows that the configuration and data used to train the PINN model found well and discretization of this model separates boundary points from interior points. The velocity components maintain near-zero values at all walls, and this confirms the precise accurate suppression of artificial penetration velocities. Thermal surfaces exhibit well-defined gradients that compress toward the hot wall as Pr increases.