[mm]

Transcription

1 . Parameter validation from comparison with tunnel experiment To ensure reasonable definition of parameters used in the numerical simulation, a number of error impact analysis were taken between the results of CFD and of the reference tunnel experiment. Four main groups of parameters are considered, which include domain scale (geometry), mesh, boundary condition, turbulence models and solution method. To evaluate the outcome of different values for the selected principal parameters, two indicators are chosen, respectively, E/U 0 the average percentage of error in portion of the reference velocity from CFD data and that of tunnel, the smaller the better, and P the number of test points with better approching CFD data than a fixed case as reference, the bigger the better. Note that all the velocity data analysd here are in vector X while the inlet is in X direction. Besides, the judgement is more focused on the vertical section performance while for our project the object is for the wind energy, which appears in higher level rather than the pedestrian s Fig. 0 Schematic view of tunnel test model b=m, Rs=:00 for tunnel test Inlet flow data can be simplified as a power law: (U is the inlet velocity in the height of Z) U 0 =7,8m/s (H=00m) Fig. 0 Horizontal section (pedestrian s level Z=,m) and its test points Table. Error Impact analysis of the parameters for CFD Domain Scale Mesh Fig. 0 Vertical section (Y=0m) and its 09 test points E/U 0 P(/) E/U 0 P(/09) B K-e Standard (R=0m) Ba R=0m Bb R=70m Bc R=00m Bd R=00m ,069 7 L=0.8m, N=0, T g =0.m, r g =., T b =0.08m, A r b =., Advanced size function: proximity and curvity, RC (Relevance Centre): Medium, Smoothing: Medium, Q=.8*0 Aa L=0.6m, Q=.6* Ab L=m, Q=.6* E hor E ver

2 Boundary contition Turbulenc e Model and Solution Ac L=.m, Q=.9* Ad L=.m, Q=.076* Aa N=8, Q=.8* Ab N=, Q=.8* Ac N=, Q=.8* Aa T g =0.m, Q=.87* Ab T g =0.m, Q=.8* Aa T b =0.m, Q=.78* Ab T b =0.06m, Q=.9* Aa r b =., Q=.9* Ab r b =., Q=.777* A6a RC: fine, Q=.96* A6b RC: corse, Q=.* A7a Smoothing: high, Q=.86* A7b Smoothing: low, Q=.8* B L=.m, N=0, T g =0.m, r g =., T b =0.08m, r b =., Advanced size function: proximity and curvity, RC (Relevance Centre): Medium, Smoothing: low, Q=.6*0 Ba T g =0.m, r g =., r b =., Q=.9* Bb T g =0.m, r g =., r b =., Q=.6* Bc T g =0.m, r g =., r b =., Q=.9* Bd T g =0.m, r b =., Q=.7* Be T g =0.m, r b =., Q=.6* Bf r b =., Q=.6* Bg r b =., Q=.6* A Ks=m, Cs=0.99, I=0%, L=m Aa Ks=0m Ab Ks=0.m Ac Ks=0.m Ad Ks=0.8m Ae Ks=.m Af Ks=.m Ag Cs= K-e realisable, precision, Scheme: Simple, A Discretization: order (pressure), Quick (momentum, k, e). A Discretization (momentum, k, e) order A k-e RNG A RSM A k-e Standard Note: E hor : horizental section (Z=,m) velocity in vector X ; E ver : vertical section (Y=0m) velocity in vector X; E/U 0 : average error percentage of reference velocity; P: number of test points better approching tunnel data than a fixed reference case; R: radius of the domain. L: the cell size for the faces of the building; N: layers number of inflation for both ground and building; T g : first layer thickness from the ground inflation;

3 r g : ground inflation transient ratio; T b : first layer thickness from the building inflation; r b : building inflation transient ratio; Q: quantity of cells element for the mesh. Ks: roughness height (m) for wall function; Cs: roughness constant for wall function; I: tubulent intensity (%). Fig. 0 Comparison of results CFD with model in different domain scales Fig. 0 Comparison of results CFD with different turbulent models

4 Fig. 06 Comparison of results CFD with meshes of different inflation layer numbers (N) Fig. 07 Comparison of results CFD with meshes of different size precision for the building (N)

5 Fig. 08 Comparison of results CFD with meshes of different inflation ratios for the building (r b ) From the comparison table and plots above we can see that:. For the domain scale a radius of 00m can be a best choice, while increase it to 0m or 00m the error has little change.. For the mesh: a. among different inflation first layer thickness for the building (T b =0.06m-0.m), the smaller the size the better the outcome. However with consideration of mesh cell number (Q) a balanced value like T b =0,08m is chosen; b. there is no best number of inflation layers, while the small average error is more often accompanied by a bad number of better approaching points. Hence we take a balanced value: N=0. c. among different inflation first layer thickness for the ground (T g =0.m-0.m), 0.m has the best d. among different inflation transit ratio for the ground (r g =.-.),. has the best e. among different inflation transit ratio for the building (r b =.-.),. has the best f. for the option of general mesh control, Relevance Centre (RC) in Medium, and low in smoothing get a best performance respectively.. For boundary condtion in wall function, the roughness height Ks and constant Cs in fact both have little influence over the result. However, a bigger Cs seems a better outcome and when Ks>0.m the outcome seems to stop changing. Here we take a Ks=m and Cs=0.99 as the best setting.. Among different turbulent models k-e standard shows a best