Artificial Neural Network–Based Parameter Estimation for Convective Heat Transfer in a Magnetized Porous Trapezoidal Cavity Under the Channel

This study examines heat transport in porous media containing an enclosed trapezoidal cavity under a horizontal channel when affected by an inclined magnetic field. The system fluid motion together with thermal behavior responds to the mutual effects of buoyancy forces and magnetic fields and viscosity conditions. The non‐dimensionalized governing equations require solution through the FEM Galerkin‐weighted residual method. This study performs a complete parametric assessment of four key factors which include magnetic parameter alongside Reynolds' number, Rayleigh number, Darcy number, and magnetic field inclination angle. It was found that heat transfer behaved an increasing function of Rayleigh number, decreasing function of Magnetic parameter and Inverse Darcy numbers. Heat transfer behavior forecasting consumes less computational time through the application of artificial neural network (ANN) models. The training process of these models utilizes simulation results obtained from FEM simulations. The performance of heat transport property capturing by neural networks is examined through a comparison of FEM and two ANN approaches which use either parameters as input points (ANN) or Gauss‐Lobatto transformation (ANN‐GLT). It is worth mentioning that Gauss‐Lobatto scaling yields better prediction results across a wide range of domain data. Therefore, the ANN‐GLT model provides more accurate predictions than a standard ANN. The prediction time of FEM simulations which takes several hours exceeds the significantly shorter processing time of ANN models. The results prove ANN models can serve as reliable alternatives to heat transfer analysis to generate fast and precise engineering‐focused predictions including thermal management system and energy storage and geophysical processes applications.