Capturing Upstream Turbulence Information by Modeling the Turbulent Generation Region and Comparison of 0 and 30 Flow and Dispersion CFD Modeling and Macdonald's MUST Experiment

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014: 9:30 AM
Room C206 (The Georgia World Congress Center )
John P. Keady, George Mason University, Fairfax, VA; and P. Huq, W. G. K. Bradbury, P. E. Price, and F. E. Camelli

Handout (5.8 MB)

Modeling of laboratory dispersion experiments often begins with setting the modeled upstream velocity profile to the experimentally reported velocity profile, with the upstream turbulent profile modeled separately. One objective of this study is to model the pre-roughness section of a laboratory experiment to obtain turbulent information that can then be associated with the reported upstream velocity profile then compare modeled flow values with experimental values. The computationally derived values of a Computational Fluid Dynamics (CFD) model were compared with experimentally measured values of flow through a MUST array in a water tunnel. The approach flow to the experimental test section was a fully turbulent simulation of a neutral stability atmospheric boundary layer (ABL). The ABL was generated in the roughness section using a combination of a short barrier wall, turbulence spires, and an extended fetch of surface roughness. The surface roughness has two fetches. The first consists of 2.4-m of 19-mm high aluminum plates with packing density λf=0.063. This is followed by a 2.4-m fetch of 10-mm high x l6-mm wide Lego@ blocks on 48-mm centers with a packing density λf=0.069. The roughness section was modeled so that the computationally derived entry velocity profile of the test section matched the experimentally measured profile, retaining turbulent information in the simulated profile. The experimental array in the test section consisted of 50-mm (H) high obstacles with a 4:1 (width-to-height) aspect ratio and a packing density of 13.3%. A plume was experimentally obtained by injected fluid at a temperature of 65C with respect to the temperature of 20C of the ambient fluid before injection. Temperature was measured using Resistance Temperature Detectors (RTD) The Computational Fluid Dynamic model used 39 million tetrahedral elements solving the incompressible Navier-Stokes equations using Large Eddy Simulation and Smagorinsky WALE Turbulence Model. Computationally derived values of mean velocity, turbulence statistics and concentration were compared to the experimental values which were measured in the obstacle arrays behind selected rows from 1 to 12. Computationally derived values of the upstream (of the experimental array) average velocity were within one standard deviation of the reported experimental values, while the upstream simulated turbulence intensities generally coincided with the peak locations reported within a vertical location of about +/-0.2 z/H. The simulated upstream normalized turbulent-u peak intensities overpredicted the reported peak value by about 40%, while the lateral (turbulent-v) and vertical (turbulent-w) simulated peak values underpredicted the reported values by about 15% and 10% respectively. In the experimental array the simulated normalized turbulent kinetic energy at row 1 underpredicted the peak value reported by about 28%, while the simulated values at row 3 and 6 overpredicted the reported values by about 6% and 4% respectively by row. The underpredicted peak value at row 1 is thought to occur during a region where the flow is not yet fully developed, while by rows 3 and 6 the flow is thought to be more fully developed. The simulated normalized temperature profile lateral peak location at row 3 for the four reported vertical locations z/H =0.1, 0.6, 1.0, and 2.4 varied from the reported lateral peak location by about -1y/H, 0y/H, -2y/H, and -1.0y/H respectively by vertical location. In summary the overall simulated values were similar to the reported values for both the 0 and 30 degree orientation, with the turbulent kinetic energy maintaining a close match between the simulated and reported values for later rows (3 and beyond). Thus, simulating the flow through the roughness section provided more accurate upstream turbulence information which then propagated through the flow solutions through the test section.