Toward grid-independent modeling of natural convection with fire dynamics simulator

This study investigates methods to enhance the accuracy and grid independence of convective heat transfer predictions in engineering-scale fire simulations using Fire Dynamics Simulator (FDS). Direct numerical simulations (DNS), validated against experiments involving a vertical, heated plate, served as benchmarks for large eddy simulations (LES) performed with grid sizes ranging from 1 mm to 38.4 mm. Three approaches were evaluated to improve LES predictions. The first approach modified how FDS applies natural convection correlations by considering the boundary layer thickness and selecting the free stream temperature from beyond the boundary layer, as opposed to the current method, which uses the gas temperature in the first cell adjacent to the wall. The second approach introduced a new Nusselt number correlation based on dimensionless parameters (Grashof and Reynolds numbers), normalized energy flux, and surface-free stream temperature difference. The third approach applied a logarithmic law of the wall using normalized temperature profiles. Both the first and second approaches improved accuracy and reduced grid dependency compared to the existing FDS model, with the first approach demonstrating greater effectiveness. The third approach still exhibited grid-dependent behavior, indicating the need for further refinement. The findings suggest that integrating one of the first two approaches into FDS would enhance its prediction accuracy, particularly for convective heat transfer. This study also establishes a foundation for applying these improvements to more complex fire scenarios, providing a pathway for advancing the modeling of real-world fire scenarios. • Grid-dependent CFD predictions of natural convective heat flux was improved. • DNS data matched experimental data for local temperature and velocity profiles. • A theoretical model was developed to predicted boundary layer thickness. • The boundary layer model led to accurate natural convective heat flux modeling. • A new Nusselt number correlation led to grid-independent heat flux predictions.