Turbulent Incompressible Wake Flow in a Rectangular Channel: High speed PIV experiment and LES simulation

Abstract

Turbulence is a ubiquitous phenomenon of highly complex flow. With increasing computer capacity and steadily improving numerical methods Computational Fluid Dynamics (CFD) is becoming a beneficial tool in the study of turbulence. In this work the performance of a Large Eddy Simulation (LES) of turbulent flow is evaluated by comparing simulated data to experimental values. The turbulent incompressible wake flow of the Particle Image Velocimetry (PIV) work of Feng was simulated using LES. A digital filter method was employed for the inlet condition. In a series of presimulations the effects of several central parameters were tested by varying one parameter at a time. With support in the presimulations the final simulation setup was determined. The final simulation was run for a simulation time of 80 seconds collecting 10000 instantaneous velocity fields. The PIV data gave two dimensional velocity fields that were used to evaluate the LES. The evaluation was carried out in traditional ways comparing one-point statistics. In addition, the nature of the PIV data gave the opportunity to compare spatial values such as two-point correlations and vortex statistics. The spatial information was used to study the coherent structures of the flow and to evaluate to what extend LES was able to capture them. The simulations showed in general good congruence with the experimental data and the chosen inlet method proved to be successful. It was found that an inlet transition region forming in the simulations caused some deviation of the simulation data from the experiment data especially regarding turbulence intensities and diffusion. The presimulations testing inlet condition, timestep, grid resolution and subgrid models showed that the mean velocity values were not sensitive to the different parameters in the range tried. From the tested parameters, the inlet condition proved to be the most influential on the fluctuating components of velocity. Changing the other parameters led to only minimal changes in the results. Furthermore, the study of coherent structures showed that the main coherent motion was swirling motion induced by shear in the wake and wall regions. Two-point spatial correlations indicated good consonance between the simulation and the experiment. However, experimental results were over all slightly noisier. Conformity was also found between vortex statistics from the PIV data and from the simulation data when the different grid resolutions were accounted for. This work showed that two dimensional PIV data provides a mean for thorough evaluation of LES especially with regards to turbulent coherent structures. Findings indicate that LES can be successfully employed for prediction of turbulence even in more complex flows such as the one used in this study.