Predicting the thermal distribution in a convective wavy fin using a novel training physics-informed neural network...

Area of Science:

• Heat Transfer Engineering
• Computational Fluid Dynamics
• Artificial Intelligence in Engineering

Background:

• Fins are crucial components in heat exchangers due to their cost-effectiveness, light weight, and compact design.
• Understanding heat transfer in fin structures, especially with internal heat generation and convective effects, is vital for optimizing thermal performance.

Purpose of the Study:

• To investigate the thermal distribution in a wavy fin under convective conditions with internal heat generation.
• To explore the application of physics-informed neural networks (PINNs) for analyzing complex heat transfer phenomena in fin structures.
• To examine the impact of varying hyperparameters on the accuracy of PINN models for nonlinear heat transfer equations.

Main Methods:

• A one-dimensional steady-state heat transfer model for a wavy fin was developed.
• The governing non-linear ordinary differential equation (ODE) was reduced to a dimensionless form.
• Physics-informed neural networks (PINNs) were implemented as a machine learning strategy to solve the ODE.
• The Runge-Kutta Fehlberg's fourth-fifth order (RKF-45) method was used for numerical validation.
• A PINN model was trained using a mean squared error-based loss function, bypassing traditional data-driven approaches.

Main Results:

• The study demonstrated that increased thermal conductivity enhances the overall thermal distribution within the wavy fin.
• An augmentation in the convective-conductive variable was found to decrease the temperature profile.
• PINNs effectively predicted the heat transfer properties without relying on conventional data-driven methods.

Conclusions:

• PINNs offer a powerful and novel approach for solving complex heat transfer problems in engineering applications like wavy fins.
• The findings provide valuable insights into optimizing fin design by understanding the interplay between thermal conductivity and convective effects.
• This research highlights the potential of machine learning strategies in advancing the analysis of thermal systems.