wong wai lun-005008

8/2/2019 Wong Wai Lun-005008

1/12

Numerical Studies on Airflow in Urban Street Canyons

WONG Wai Lun, UNIMKL-005008

Computational Fluid Dynamics (MM4CFD)

Department of Mechanical, Materials & Manufacturing Engineering

Supervisor: Prof. Andy Chan

Abstract

The objective of this study is to investigate numerically the effect of the ratio of street width,

W to building height, H on wind flow and predicting the pollutant dispersion in a street

canyon within an urban environment. Three-dimensional numerical models based on

Reynolds-Averaged Navier-Stokes - RNG model and Large Eddy Simulation (LES) werecreated to analyze the air flow development within an urban canyon. The dispersion of thepollutant thus can be predicted based on the air flow simulation using Fluent code. Its learnt

that the geometry and configuration of the building play important roles in determining the

complex flow pattern and then the pollutant concentration within urban canyon streets.

The model generated has strong agreement with the literature data. It is observed that

Large Eddy Simulations LES was able to capture the unsteady and intermittent fluctuations

of the flow. However, it did not give much differences compared to the - RNG model forthe cost of higher computational time and cost. In the light of this, - RNG model is morepreferred in this case. Further improvement in ventilation can be done with alterations of

building height and roof shape.

8/2/2019 Wong Wai Lun-005008

2/12

Introduction

In the light of continuing urbanization and the increase of vehicles in urban area, the

dispersion of the pollutants that mainly emitted by vehicles exhaust has become subject of

interest to environmental analysts, building occupants and so forth. The air quality is

depending upon the air ventilation within an urban area, which is influenced by the traffic

flow, building geometry, ambient wind speed and direction, street configurations such as

roof shape, building height, width of the street and so on. The air quality has direct impact

on the health of the pedestrians, cyclists as well as people living in the urban area. Without

doubts, poor ventilation will lead to high concentration of pollutant. Several studies

targeting on the ventilation issues had been conducted with field experiments (eg. DePaul

and Sheih, 1986; Nakamura and Oke, 1988; Rotach, 1995; Croxfrod, 1988) and numerical

modeling (eg. Lee and Park,1994; Sini et al., 1996; Hassan and Crowther, 1998) in order to

understand the flow patterns and pollutant removing in urban street canyon. Increasingly

pollutants density has been the momentum for researchers such as Baik and Kim (2003), Xieet al. (2005) to develop numerical studies on pollutant dispersion using Computational Fluid

Dynamics (CFD) models to enrich understanding of pollutant transport and hence the

development of pollutant removal mechanism. Xie et al.(2006) studied the flow field and

pollutant dispersion characteristics in the street canyons with different configurations to

identify the influence of the street geometry on the wind flow and the dispersion of

pollutants in the street canyon. However, most numerical studies have been modeled based

on canyons formed by flat-roof building. The main goal of the present work is to provide the

numerical simulations of airflow patterns within urban street canyon of difference height of

building and width of street combinations to understand the effect of these configurations

on pollutants distribution/characteristic within the canyon. Besides that, the effect of

slanted-shaped roof buildings will also be analyzed in this study.

In this study, the geometries with different width of urban streets are employed with

Reynolds-Averaged Navier-Stokes - RNG and Large Eddy Simulation (LES) model and theperformance of both models is compared. The LES model explicitly solve for the large eddies

and implicitly account for the small eddies, which enable the study of the unsteadiness of a

flow, and provides the detail information on the flow structure including turbulence

statistics. The minus point is that it requires a lot of computational effort, time and cost. LES

has to be run for a sufficiently long flow time to obtain stable statistics for research purpose.

Consequentially, the computation cost involved is higher than RANS model in terms of

memory (RAM) and CPU time.

Further improvement which possibly enhance the air ventilation in the densely build urban

area can be done by finding the critical roof shape and width of the street relative to the

height of the buildings.

8/2/2019 Wong Wai Lun-005008

3/12

2. Methodology

2.1 Physical model configuration

Three dimensional urban street canyon with different widths of street relative to the

building height were used in this simulation, as shown in Figure 1.

Figure 1: The schematic diagram of the 3-D urban street canyon and the structure of the slanted roof.

The model street canyon consisted of 6 rows of shop lots and 5 streets. All the urban street

buildings have a square cross section of length H=20m with a triangle slanted roof profile as

shown in the Figure 1. The air ventilation in the urban area is depending upon the width of

the street relative to the height of the building. The width of the street, W solely is not able

to justify how well the ventilation would be. Hence, the street width to building height ratio

is employed, where

The width of the street is altered where W=0.5H, H, 2H which corresponding to R=0.5, 1.0

and 2.0. The wind direction is orthogonal to the direction of the street.

The coordinate system has its origin at the bottom, middle of the domain inlet, with x

measured as positive in the downwind direction and y is positive for upward.

5H

7H

5H

HH

W

H

45

Z

3H

20H

8/2/2019 Wong Wai Lun-005008

4/12

2.2 Mathematical model/ Numerical Method

In this study, the transient incompressible Reynolds-Averaged Navier-Stokes (RANS)

equation is employed for the mean flow field in street canyon using the Re-Normalisation

Group (RNG) k-epsilon model. The commercial code FLUENT is adopted in the CFD

simulations. The governing equations with the RNG model are expressed as follow:

Mass conservation (Continuity Equation)

= 0

Momentum Equation

+

) = -

+

kand transport equation (rupanya for convection- diffusion) +

=

+

-

+

=

+

-

Where,

= ith mean velocity componentP = the deviation of pressure from its reference value

= air density

= inverse effective Prandtl number for k = inverse effective Prandtl number for = effective turbulent viscosity = scalar measure for the deformation tensor

RNG k- constants: = 1.42 = 4.38 = 0.012

Large Eddy Simulation (LES)

The original Smagorinsky-Lilly model is used due to its algorithmic simplicity and numerical

stability. LES is suitable in complex flow simulation owing to its less approximation but direct

resolving is achieved as opposed to RANS. This would, however, require substantially finer

meshes and needs to be run for sufficiently longer flow-through time in order to obtain

stable statistics of the flow. Therefore, higher memory (RAM) and CPU time are required.

The governing equations are discretized by the finite volume method and the SIMPLE

algorithm is used to handle the pressure-velocity coupling. The second-order unwind

scheme is adopted for the approximation of the convection terms, and the second-order

central difference for the diffusion terms. The scaled residual criteria for all the flow

properties were set at 1x10-5

.

8/2/2019 Wong Wai Lun-005008

5/12

2.3 Model domain

A three dimensional computational domain is created, which is 100m x 140m x 400m. Biased

mesh is applied due to the large domain in simulating the urban area and the atmosphere.

Mesh cells are biased towards to the buildings. The grid is finer close to the building and

ground in which the grid intervals near the wall of the building is 2m. The number of grid

cells is about 342,640. Structured hexahedron cells are believed to give more even mesh

distributions and accurate result since false diffusion can be prevented in the case where

second order upwind scheme is adopted for the approximation of the convection terms, the

alignment of the flow direction with the grid is rather important.

For both transient models, the number of time steps employed was 5000 with time step size

of 0.07 and maximum 20 iterations per time step.

2.4 Boundary Conditions

Velocity inlet with boundary layer profile is used in the main inlet wind flow. The initial free

stream wind speed is 5m/s, and this inlet velocity profile is developed with turbulence of 1/7

in the power law in the user defined function (UDF) which was then implemented in the CFD

code.

The ground and building surfaces are defined as walls with no-slip boundary condition. In

FLUENTTM

, the surface roughness is expressed in terms of sand grain roughness, Ks in order

to circumvent problem with coarse grid resolution near the ground due to large Ks value. The

sand grain roughness, Ks is set to same as aerodynamic roughness length, z0 which was

found to be z0=0.0033m in wind tunnel experiments. They agreed that setting K s equal to z0

was not correct in a strong sense, but justify the choice from the result obtained, where only

minor difference in terms of velocity profile and turbulence intensity. The top plane and

both sides of the domain are applied with symmetry boundary condition. Zero gradient

boundary condition is set at the outflow.

The time step size

=

=

=0.07

8/2/2019 Wong Wai Lun-005008

6/12

3. Result

W/H = 0.5

Figure 2: Air velocity vector for W/H=0.5 modelled by k- RNG (left) and LES (right).

Figure 3: The graph of distance from the ground against horizontal velocity in the W/H=0.5 canyon.

Figure 4: The velocity vector in the W/H=0.5 urban area by k- RNG model.

0

20

40

60

80

100

120

140

160

-1 0 1 2 3 4 5 6

Distancefromt

heground(m)

Velocity in x direction (m/s)

X-velocity in the W/H = 0.5 canyon

k-e RNG

LES

8/2/2019 Wong Wai Lun-005008

7/12

Figure 5 The air velocity vector in the W/H=0.5 urban area by LES model.

W/H = 1.0



0

20

40

60

80

100

120140

160

-1 0 1 2 3 4 5 6Distancefromt

heground(

m)


X-velocity in the W/H=1.0 canyon

k-e RNG

LES

8/2/2019 Wong Wai Lun-005008

8/12


Figure 9: The velocity vector in the W/H=1.0 urban area by LES model.

WH = 2.0


8/2/2019 Wong Wai Lun-005008

9/12



Figure 13: The velocity vector in the W/H=2.0 urban area by LES model.

0

20

40

60

80

100

120

140

160

-4 -2 0 2 4 6 8

Distancefromt

heground(m)


X-velocity in the W/H=2.0 canyon

k-e RNG

LES

8/2/2019 Wong Wai Lun-005008

10/12

4. Discussion

From the numerical result obtained from the CFD code, it is obvious that the airflow pattern

inside the canyon is strongly influenced by the ratio of the width of the canyon, W to the

height of the surrounding buildings, H. The airflow pattern can be directly related to the

distribution of the pollutants in the canyon.

4.1 W/H=0.5

For the first case where the ratio W/H=0.5, it can be seen that two vortices appeared in the

canyon as shown in the Figure 2. The upper one is in clockwise direction, which located in

between the two slanted roof. This vortex is generated due to the ambient wind. The lower

one is driven by the circulation above, therefore it is in counter clockwise direction. As

shown in the Figure 3, the horizontal velocity from ground to approximate 10m above ispositive (counter clockwise circulation) and the velocity slowly reduced as the distance from

the ground increases. It has finally become negative, which indicates a clockwise vortex.

Owing to this lower counter clockwise circulation, pollutants tend to accumulate on the

windward side of the canyon and can hardly escape from the canyon. Comparing the k-RNG and LES model, the LES model present a more realistic result, where the airflow pattern

is not steady and there could have more than two vortices in the canyon. This is illustrated in

Figure 3, where the direction of the horizontal velocity component is fluctuating at the base

of the canyon. Nonetheless, the k- RNG model tends to give higher horizontal velocitycompared to LES model.

4.2 W/H = 1.0

As shown in Figure 6, the upper clockwise vortex has enlarged, and the centre of this vortex

moves downwards. This is because the main stream flow has more space to create a large

circulation in the canyon and generate greater effect on the lower vortex. As a result of it,

the lower counter-clockwise vortex is pressed downwards and smaller in size. Both

numerical models give similar horizontal velocity in the canyon as shown in the Figure 7.

From Figure 7, it can be seen that within the canyon, the changes of horizontal velocity with

respect of distance from ground is similar to first case. The difference is that the positive

horizontal velocity is up to approximately 5m from the ground only, which means the lower

counter clockwise vortex is lowered, due to the enlarged upper clockwise vortex. In the light

of this, some pollutants are carried towards the leeward face by the upper vortex and some

to the windward face by the lower vortex. However, when comparing the k- RNG and LESmodel, k- RNG did not show the effect of the roof shape on the free stream. LES allowsbetter predictions of the transient flow as circulation above the building roofs are observed

in Figure 9. Besides, k- RNG model has higher magnitude of velocity compared to LES modelalthough both model give similar trend in horizontal velocity as shown in the Figure 7.

8/2/2019 Wong Wai Lun-005008

11/12

4.3 W/H=2.0

When the width of the street is twice the height of the buildings, one strong clockwise

vortex is generated without a driven counter clockwise vortex like previous cases. The wider

street allows better ventilation in the canyon. The main stream flow is able to ventilate the

canyon better without the lid-driven-cavity effect. The graph of horizontal velocity in the

canyon also different compared to the previous cases, where there is only negative (reverse

direction) flow in the base of the canyon which represent that the strong clockwise vortex is

taking place in the canyon. Pollutants tend to accumulate on leeward side of the street due

to the clockwise strong vortex as shown in the Figure 10. Again, the k- RNG model hashigher horizontal velocity of wind in the urban canyon.

Conclusion

In this study, the effects of the street width relative to the buildings height have been

numerically investigated. It can be concluded that different street width will have different

distribution of the pollutants as presented in the Table 1:

Street width to buildings

height ratio, W/H

Number of vortices

generated

Expected pollutant

distribution in the canyon

0.5 2 main vortices, about equal

in size

Windward side of the canyon

1.0 2 main vortices, lower

counter clockwise vortex is

smaller than upper clockwise

vortex.

Leeward side of the canyon

by upper vortex and

windward side by lower

vortex.

2.0 1 main clockwise vortex

driven by the free stream

flow.

Leeward side of the canyon

by the clockwise vortex.

Table 1: Summarized results.

The narrow (W/H=0.5) canyon has poorer air ventilation properties compared to wider

(W/H=2.0) canyon.

Both CFD models (k- RNG and LES model) give good agreements with the literaturenumerical result in Reference [2]. However, the numerical results of k- RNG model aregenerally higher than LES model. LES model is believed to be a model with higher accuracy

compared to k- RNG model, in which the simulated result of transient flow is more realistic.It has also shown the effects of the roof shape to the main stream flow and the circulation

formed. In terms of resources demand and computational cost, the k- RNG took about4hrs while the LES model consumed about 10days for the same geometry. Since the flow

pattern and flow velocity simulated by both models give similar numerical result, the k- RNG model is more suitable to simulate an urban canyon case where computational time

and cost are the constraints.

8/2/2019 Wong Wai Lun-005008

12/12

References

1. Mohamed F.Yassin (2011). Impact of height and shape of building roof on air qualityin urban street canyons. Atmospheric Environment 45 (2011) 5220-5229.

2. HUANG Yuan-dong, JIN Ming-xia, SUN Ya-nan (2006). Numerical studies on airflowand pollutant dispersion in urban street canyons formed by slanted roof buildings.

Journal of Hydrodynamics (2007) 100-106.

3. HUANG Yuandong, Xiaonan Hu, Ningbin Zeng. Impact of wedge-shaped roofs onairflow and pollutant dispersion inside urban street canyons. Building and

Environment 44 (2009) 2335-2347.

4. Salim Mohemed Salim, Riccardo Buccolieri, Andrew Chan, Silvana Di Sabatino.Numerical simulation of atmospheric pollutant dispersion in an urban street canyon:

Comparison between RANS and LES. Journal of Wind Engineering and Industrial

Aerodynamics (2011)103-113.

wong wai lun-005008

Documents