CFD analysis of transonic turbulent cavity flows

This thesis presents the study of transonic cavity flows using CFD. The main focus of the thesis is on the turbulence modelling and simulation of cavity flows. The thesis aims to show the range of applicability of turbulence modelling for cavity flows. Aspects of cavity flow control are also addressed. The cavity models a weapons bay with a length-to-depth (L/D) ratio of 5 and length-to-width (L/W) ratio of 1. The flow is set to transonic speeds (M=0.85) and the Reynolds number based on the cavity length is in the turbulent regime (Re=6.783 million). At these flow conditions, very high noise levels are encountered inside the cavity combined with intense acoustic and turbulent interactions. Unsteady Reynolds-Averaged Navier-Stokes (URANS) was initially applied and the effectiveness of various two-equation turbulence models such as the Wilcox k - ω and the Menter's Baseline k - ω and SST turbulence models was assessed. Computations were first performed with the 2D clean cavity to minimise the computational cost, where the 2D cavity was a reasonable representation of the full 3D cavity with doors-on. Results demonstrated that linear statistical turbulence models generally gave reasonable qualitative predictions of the cavity flow-field but on coarse grids only. The amplitudes of the noise levels and frequencies were however less well predicted and the level of agreement deteriorated with grid refinement for the L/D=5 cavity. Nonetheless, out of the models employed, the SST model proved to be the most robust and provided the best agreement with experimental pressure and PIV measurements. The velocity distributions and the turbulent and acoustic spectra at the cavity floor were also analysed and compared with experiments (where possible) and in doing so the influence of turbulent processes in the cavity highlighted. With the higher acoustic frequencies and the broadband noise less well predicted with linear statistical two-equation turbulence models, attention was diverted towards simulation methods such as Large-Eddy Simulation (LES) and Detached-Eddy Simulation (DES). Numerical results for the 3D L/D=5 cavity with a width-to-depth ratio (W/D) of 1 in both doors-on and doors-off configurations were compared with experiments. Even for coarse grid simulations, better agreement was found between the LES/DES results and experimental pressure and PIV measurements for various grid levels and time-steps than URANS.