Enhancing thermal field prediction in a jet in cross flow through turbulent heat flux model optimization by using large eddy simulation

High fidelity big data obtained from highly accurate large eddy simulation (LES) was utilized to enhance current models for turbulent heat transfer, boosting the precision of the realizable Reynolds averaged Navier Stokes (RANS) k-epsilon model in forecasting film cooling thermal field. The LES analysis was conducted on a flat plate with a jet in crossflow, at a primary flow Reynolds number (Re) of around 17382. Post-validation of the LES outcomes, a data analysis, and a numerical optimization (gradient descent algorithm) technique were employed to fine-tune three chosen models for turbulent Prandtl number using data near the coolant wall. In order to demonstrate the effectiveness of these refined models for RANS simulations of various film cooling setups with distinct shapes and flow situations, a series of simulations were executed, encompassing two curved surfaces to simulate flow around leading-edge and over the blade suction side. The optimized models exhibited marked enhancement in forecasting film cooling effectiveness in both averaged lateral and longitudinal directions (achieving reductions in numerical errors of up to 68.0 % and 43.67 % respectively). These fine-tuned models were proven suitable for use across different geometries and flow conditions according to their performance in three analyzed problems.