Large eddy simulation of turbulent flows with variable property heat transfer using a compressible finite volume formulation

A compressible finite volume formulation has been developed to solve the Favre filtered Navier-Stokes equations to perform the large eddy simulation (LES) of turbulent flows with variable property heat transfer. The efficient finite volume formulation was developed using a dual time stepping approach with time derivative preconditioning and multigrid acceleration. Time marching was done with either an explicit Runge-Kutta scheme or an implicit lower-upper symmetric-Gauss-Seidel (LU-SGS) scheme. The code was developed in a multiblock framework so it could be applied to complex geometries and to provide a means for parallelization.;The second-order accurate finite volume LES formulation was validated with simulations of turbulent incompressible benchmark flows. The results were compared to experimental data and incompressible direct numerical simulation (DNS) results. The LES formulation was subsequently applied to a plane channel flow with constant wall heating rates of magnitudes large enough to cause significant property variations across the channel. The effects of high heating versus high cooling on the turbulence quantities and turbulent structure were studied. Finally, the LES formulation was evaluated for a complex geometry by attempting to simulate the turbulent flow and heat transfer for a plane channel with transverse square ribs on one wall.;Multigrid acceleration and time derivative preconditioning were very effective for steady and unsteady laminar flows. However, multigrid provided no benefit for LES, most likely due to insufficient numerical damping to drive the multigrid since artificial dissipation was not used.;The LES formulation provided excellent agreement with DNS and experimental results for simple turbulent flows (i.e. homogeneous, isotropic, decaying turbulence and smooth wall channel flows). For the constant heat flux channel flows, high heating tended to reduce the velocity fluctuations, while high cooling tended to promote the fluctuations. The mean and fluctuation velocity profiles collapsed towards the incompressible results when normalized by local properties, as opposed to wall values. The temperature fluctuation profiles were largely independent of the heating rate when normalized by the wall-to-bulk temperature difference. Problems related to significant odd-even decoupling, grid distributions, and SGS modeling were identified for the rib-roughened channel.