Pentagon Shield: Experiments with a nonosillatory forwardin-time CFD code (EuLag) to simulate flow around the Pentagon

Transcription

1 Pentagon Shield: Experiments with a nonosillatory forwardin-time CFD code (EuLag) to simulate flow around the Pentagon 8 th GMU Conf. on Transport and Dispersion Modeling 15 July 2004 Piotr Smolarkiewicz NCAR/MMM Bob Sharman NCAR/RAP

2 Pentagon CFD modeling goals Evaluate ability of CFD models to accurately represent flow around the Pentagon To better understand the nature of flow around the Pentagon under a larger variety of conditions than can be investigated with limited field campaigns Provide baseline against which faster CFD models can be evaluated for operational use

3 EuLag Description Nonhydrostatic, anelastic ( ρ o v=0), Navier-Stokes eqns. Eularian/semi-Lagrangian options Multidimensional Positive Definite Advection Transport Algorithm (MPDATA) Extended terrain-following mesh (upper, lower, lateral) Structured, time-dependent grids Preconditioned Generalized Conjugate Residual (GCR) nonsymmetric solution of elliptic p eqn. (Eisenstat, 1983) LES type turbulence closure (1 ½ order, prognostic tke) Successfully used for a wide variety of engineering and geophysical applications Mountain-valley flows, convective clouds DNS of turbulence in a box Aircraft wake vortices Solar convection, oceanic flows, etc. Relative of HIGRAD

4 Some EuLag References Smolarkiewicz and Margolin, 1998, "MPDATA: A finite-difference solver for geophysical flows, J. Comput. Phys., 140, Smolarkiewicz and Margolin, 1997, "On forward-in-time differencing for fluids: an Eulerian/Semi-Lagrangian non-hydrostatic model for stratified flows," Atmosphere-Oceans, 35, Wedi and Smolarkiewicz, 2003: "Extending Gal-Chen and Somerville terrain-following coordinate transformation on timedependent curvilinear boundaries." J. Comput. Phys., 193, Prusa and Smolarkiewicz, 2003: " An all-scale anelastic model of geophysical flows: dynamic grid deformation," J. Comput. Phys., 190, Margolin, Smolarkiewicz, and Sorbjan, 1999: "Large-eddy simulations of convective boundary layers using nonoscillatory differencing," Physica D, 133, Margolin, Reisner and Smolarkiewicz, 1997: Application of the volume-of-fluid method to the advection-condensation problem, Mon. Wea. Rev., 125, Grabowski and Smolarkiewicz, 2002: A multiscale anelastic model for meteorological research. Mon. Wea. Rev., 130,

5 EuLag Validations Analytic solutions Basic pde solution (simple advection, diffusion, etc.) Linear and nonlinear mountain waves (2d and 3d) Hydraulic jumps Tank experiments Wake vortex breakdown Boundary current separation Stratified flow over surface mounted obstacles CFD intercomparisons Convective BL Boulder downslope windstorm intercomparisons Comparisons to observations Flow over Alps Flow over island of Hawaii Comparison to nonlinear analytic solution (- -)

6 EuLag (MPDATA) validations Initial condition upwind leapfrog MPDATA MPDATA-FCT

7 EuLag Validations (contd.) LES of unstable BL Comparisons to laboratory experiments ( ) Comparisons to aircraft measurements (o) Comparisons to other LES (Schmidt and Schumann, JFM, 2000) <w w >/w* 2 <θ θ >/T* 2

8 Pentagon Model 1Jul 2002 Derived from 1m resolution lidar dataset (April 2001) Stated accuracy: 0.5 m H, 0.3 m V Data noisy had to be simplified Untouched 1m data simplified model

9 Pentagon setup Boundary-fitted representation z = H(z-h(x,y))/(H-h(x,y)) x= y= z=2m 7200 time t=0.05 s Rigid upper boundary Eulerian option 2 nd order in space and time sgs 1 ½ order closure Specified CD on building and sfc. Neutral with prescribed velocity profile from previous LES simulation (Moeng and Sullivan) NCAR supercomputer ( older IBM MPI using 200 processors): 10 ½ hrs wallclock time for 7400 time steps (with 3 tracers)

10 z =0 Some results: w after 7200 time steps (6 min) z=10 m y=600 m z=25 m

11 Some results (cont): w (z=10m) after 7200 time steps (6 min)

12 z =0 z=10 m Some results (cont.): sgs tke after 7200 time steps (6 min) y=600 m z=25 m

13 Future work Validate against wind tunnel results (neutral) Perform more sensitivity studies (e.g., to resolution, SGS parameterization, etc.) Investigate other stability regimes (stable, unstable) Incorporate actual terrain around Pentagon Simulate one or more field campaign cases Couple Lagrangian particle model (Jeff Weil) to wind output from CFD calculations Assess mesh refinement techniques (with Joe Prusa) Wind tunnel model

14 Pentagon model simplifications Remove terrain Neglect peripheral obstructions roof structures, antennas, etc. Corridor bridges

15 Some results (cont.): p after 7200 time steps (6 min) z=10 m z=25 m