Encyclopaedia Index

 

MICA Project

ENVIRONMENTAL EXTERNAL FLOWS

Authors: Yanna Nikou & Christina Christidou

Organisation: NTUA

Report date: 30 April 1998

 

1 Introduction to the industrial problem

The role of NTUA within the MICA project was to specify the VR Interface and validate MICA for flow and concentration field around buildings and in street canyons. In particular, air pollution monitoring consultants, urban planners and traffic engineers as well as safety engineers for the assessment of hazards arising from technological accidents were to be considered.

According to the Council Directive "on ambient air quality assessment and management (96/C59/02)", the more severe deviation of the air quality limit values in modern cities could be happened in street canyons. Following this statement, the monitoring of street canyons is specified, in various street types, by using in a combined manner measuring stations and dispersion models.

Plume dispersion in the vicinity of buildings and more specifically in street canyons is an extremely complicated process since in the near field the plume dispersion does not follow a simple Gaussian behavior and some knowledge of the velocity and turbulence fields is required as these determine the diffusion characteristics of the flow. The dispersing plume interacts with various features of the wind flow around the building, such as the recirculation region, and the aerodynamic disturbances from the building generally encourage downwash of the plume in the building wake.

The widely used Gaussian models in the prediction of pollutants dispersion, emitted by tall stacks, are not appropriate for the simulation in the urban canopy, with complex terrain, buoyant flows, chemical reactions among pollutants, 3-D flow fields etc. For the specific case of street canyons semi-empirical models are usually used which are very simplistic concerning their mathematics and necessitate measurements campaign for every road category. CFD flow and dispersion models are capabl e of simulating valuably the complex flow field in a small urban region with multiple street canyons and variable meteorological conditions.

However, despite the potential benefits of CFD, the urban engineers and consultants are not exploiting the technology to its full potential. The main reason is the non user friendliness associated with traditional CFD programs as well as the high cost of the hardware and software required.

An excellent solution to this problem would be to make the CFD models simpler to use and to provide easy access to high performance computing facilities. MICA and the associated Internet Network of MICANET have the potential to satisfy both these requirements.

 

2 Description of the simulated cases

NTUA has studied three scenarios in order to simulate the flow field and predict the pollutant concentrations in street canyons. The first one consists of the development of a two-dimensional flow model applicable to street canyons under ambient winds perpendicular to the street. Different cases of geometrical configuration were examined according to the aspect ratio W/H.

The second scenario concerns the development of three-dimensional street canyons. Different cases of various ratios W/H and L/H were examined.

The third scenario concerns the prediction of flow and concentration field around a block of buildings.

Scenarios 1 and 2 were described in previous reports whereas the third one is new and represents the remaining part of NTUA contribution to the validation exercise.

  1. Two - Dimensional Street Canyon

The flow is considered to be two-dimensional. This is happening in the middle of a road which is not affected from the adjacent streets.

In order to predict the flow and concentration field within the street canyon for various geometrical configurations, four cases were examined with respect to the ratio W/H.

In all cases the building height (H) is remaining constant and equal to 20m whereas the street canyon width (W) is depending on the ratio W/H. The height of the computational domain is equal to 160m whereas its width varies and is equal to (60+W+100)m.

The wind has an atmospheric boundary layer profile using the logarithmic law formulation over a surface with a roughness length of 0.5m and it is perpendicular to the canyon axis. The wind speed is set to 5m/s at the top boundary. The atmospheric boundary layer is extended 120m above the ground that is 100m above the street canyon roof.

Figure 1 shows a schematic diagram of the computational domain and building’s configuration for Case 1 (W/H=1.0).

 

Figure 1 : Geometry used for the simulation of Case_1

The ground has been set up as BLOCKAGE thermally adiabatic. The material selected to represent the ground was asphalt/felt/bitumen layers (material No 129). The buildings are represented as solid blockages with smooth wall friction (material No 198).

In order to produce a fine grid inside the street canyon and around the buildings that is in regions with steep gradients, the number of cells was increased using NULL objects.

The pollutant source is represented in the form of a BLOCKAGE object consisting of air (and the ideal gas law).

Figure 2 illustrates the geometry used for Case_2 (W/H=0.5) as it can be seen in the VR-editor (v.3.1).

Figure 2: Geometry used seen in the VR-Editor for Case_2

The number of iterations needed for convergence was 500-1000.

All cases that were examined during the project have also been set-up with the final version.

  1. Three - Dimensional Street Canyon

The flow is considered to be three-dimensional.

In order to predict the flow and concentration field within the street canyon and around the buildings for various geometrical configurations, two cases were examined with respect to the ratio W/H and L/H (L is the building’s length):

  • Case_1: W/H=4.0, L/H=1.0
  • Case_2: W/H=4.0, L/H=2.0

In both cases the building height (H) is remaining constant and equal to 20m whereas the street canyon width (W) is depending on the ratio W/H.

The inlet velocity profile follows the logarithmic law (atmospheric boundary layer type) over a surface with a roughness length of 0.5m. The wind speed is set to 5m/s at the top boundary. Inlet boundary is extended 200m above the ground and 100m upstream of the first building. Outlet boundary is set 200m downstream of the second building.

The height of the computational domain is equal to 200m whereas its width varies and is equal to (100+L+W+L+200)m. Half of the real domain was discretized due to the symmetry of the problem.

The geometry of the computational domain is shown below in Figure 3.


Figure 3 : Geometry used for the simulation of Case_1

In order to produce a fine grid inside the street canyon and around the buildings and to obtain correct solutions, NULL objects were introduced to the computational domain. The number of cells was about 90.000 -120.000.

The ground is set up as BLOCKAGE thermally adiabatic. The material selected to represent the ground was asphalt/felt/bitumen layers (material No 129). The buildings are represented as solid blockages with smooth wall friction (material No 198).

In Case_2 it is assumed that there is a linear pollutant source along the street located on the ground. It occupies 4.5m of road and it is represented in the form of a BLOCKAGE object consisting of air (and the ideal gas law). Emission coefficient which deducted by the composition of the car fleet is 29.86 mg per vehicle. This leads to a rate of emission of 8.4mg/sec/m.

The number of iterations needed to achieve convergence was 5.000-10.000. All cases that were examined during the project have also been set-up with the final version.

Figure 4 illustrates the geometry used for Case_2 (W/H=4, L/H=2.0) as it can be seen in the VR-editor (v.3.1).

Figure 4: Geometry used seen in the VR-Editor for Case_2

  1. Block of buildings

Two cases were examined. The first one (Case_1) represents a four building area which has symmetries and for that reason it is divided. The second one (Case_2) represents a part of Patission Street which is at the center of Athens and presents high traffic load.

  1. Case_1

Due to the symmetry of the problem two buildings are introduced to the computational domain. The results need to be reflected in y-dimension in order to give the whole 4-building area.

The two buildings have the same height (H) which is equal to 25m. Building’s length is also remaining constant and it is equal to 40m whereas their width is twice their length (80m).

Inlet boundary is located 125m (5H) upwind of the first building. Outlet boundary is set 10 building height downstream of the second building. Canyon width is equal to 25m. The height of the computational domain is equal to 205m.

The inlet velocity profile follows the power law according to the expression Uz=Uh(z/d)a.The velocity at reference height (which is set to 205m) is equal to 5.4m/s. Alpha (a) is equal to 0.143.

The atmospheric boundary layer is extended 200m above the ground with a roughness length of 0.05m.

Symmetry conditions are assumed for the frontages of the plane which are perpendicular and parallel to x-dimension

Free boundary condition is assumed for the frontage of the plane which is perpendicular and parallel to y-dimension

The geometry of the computational domain is shown below in Figure 5.

Again a fine grid of about 57,000 cells was generated by the method of NULL objects.

The ground is set up as BLOCKAGE thermally adiabatic. The material selected to represent the ground was asphalt/felt/bitumen layers (material No 129). The buildings are represented as solid blockages with smooth wall friction (material No 198).

There are two linear sources of the pollutant. The one linear source is situated along the x-axis and it occupies the space beside the buildings, neither upwind nor downwind them. It occupies 5m of road, one meter away from the walls of buildings. It is assumed that it comprises of one lane of cars. The second linear source is situated along the y-axis and occupies 12 m at the center of the road. It comprises of two lanes of cars.

The two sources are represented in the form of BLOCKAGE ("cube4") objects consisting of air using the ideal gas law.

It is assumed that every lane of cars is comprised by 24300 vehicles/day. Emission coefficient which is deducted by the composition of the car fleet is 29.86mg/vehicle/m. This results in a rate of emission for one lane equal to 8.4mg/sec/m.

Figure 6 illustrates the geometry used for Case_1 as it can be seen in the VR-editor (v.3.1).

Figure 6: Geometry used seen in the VR-Editor for Case_1

 

  1. Case_2

Case_2 represents a block at the center of Athens that consists of eleven buildings. The buildings are surrounded of different streets which present high traffic load. A ground plan of the area is given below (Figure 7). The buildings have the same height (25m).

 

Figure 7: Geometry of the block area. Top view

Figure 8 illustrates the geometry used as it can be seen in the VR-editor.

Figure 8: Geometry used seen in the VR-Editor for Case_2

3 Preprocessing of the case

During the MICA-project several versions of PHOENICS-VR have been released and developed according to the comments reported from the users. At the end of the project we can claim that the present 3.1 version for AQUA SPP is significantly improved so that it can meet all the important aspects of MICANET and the requirements of the industrial end-users for the cases of environmental external flows.

3.1 Availability of all the essential objects. Limitations

All the essential objects needed for the cases and geometries studied are now available. Some items that were reported previously and need to be improved are:

  • Possibility to create a fine grid using a "grid menu" option in spite of using the method of NULL objects.
  • It would be convenient for the user to have the possibility to re-view an object after is hidden.
  • More explanation is required for the FINE_GRID_VOL and SUB_GRID_SOLID objects.
  • Explanation is required of how the different shapes of clip-art objects affect the solution.

3.2 Availability of all the essential mathematical models. Limitations

It would be useful to introduce additional turbulence models that perform better in cases of separated flows and flows with recirculation regions. Such models may be the Chen-Kim and RNG models which are modifications of the widely used KE-EP model.

3.3 Speed of the VR editor

The speed of the VR editor need to be improved. In three-dimensional street canyon simulations it was impossible to run the case in our PC (with Pentium processor). All 3D cases (Scenarios 2&3) were submitted to MICANET in order to obtain solution and retrieve the results. Due to the fine grid used around the buildings it takes a long time to change the view of the geometry, to move objects or to zoom in and out.

3.4 Ease of use of the VR editor

We think that VR editor became quite user-friendly during the different stages of the software development. A few points to note:

  • It would be useful to preview the different object geometries supplied in the clip-art menu.
  • Very small objects are not selected by the mouse.
  • It would be useful to use the mouse in order to move the created objects.
  • Rotation of objects is not easy to handle

3.5 Visual appearance of the VR world

  • In large cases the visualisation is slow.
  • As previously reported it is very important in our geometries to have the possibility to zoom-in to an off-centered location such is the street canyon region.

3.6 Need to supplement VR settings with PIL

As mentioned in previous reports it would be very useful for the user to define the inlet profile directly from the VR-editor and not to have to worry about PIL fragments. In the latest VR version this is actually happening. We now think that all the essential settings and commands are correctly inserted from the VR-editor menu and no need to supply additional PIL commands is required.

3.7 Errors encountered

The latest version of PHOENICS-VR seems to have minimised errors that were previously appeared.

4 Submission of the case, remote calculation and data retrieval

We have the VR software installed at two 586 PC’s. The first one that is most frequently used has a 166MHz processor and 80Mbytes of RAM and it is used for job submission to MICANET. The second one has a 100MHz processor and 64 MBytes RAM. Both PC’s are connected to the NTUA network and to an internal (at CFDU) NOVEL Network. The access to Internet is quite good, including both e-mail, web-browser and ftp. The local PHOENICS 2.2.1 unix versio n which is served for comparing the results with the VR data, is installed at 5 workstations connected to a UNIX network with no communication to the outside world.

The communications software was successfully installed and no problems have been encountered. The MICANET management system was improved since the first release of the software and the job is easily submitted since the contact server is most of the times present.

As mentioned earlier in this report, NTUA submitted all the cases that scenarios 2 and 3 include. This was due to the large computational time needed to perform 3D simulations even in a 586 PC processor. The most available centers were CHAM and INRIA. A summary of the remote calculations is given below:

Description of the case

number of submissions

Scenario 1. Case_1: W/H=1.0

2

Scenario 2. Case_1: W/H=4.0, L/H=1.0

4

Scenario 2. Case_2: W/H=4.0, L/H=2.0

2

Scenario 3. Case_1: Block of 4 buildings

5

 

The speed of data transmission is satisfactory even in the large cases. The data are correctly translated and submitted to the remote computing centers. The results are usually retrieved (by e-mail and zipped) after 1-3 days --which is adequate-- and they are easily handled.

Some points to note:

  • An item that need to be introduced in the MICANET service is the possibility to perform a restart simulation at the remote centre. This is very important in our cases which need 5000 or 10000 iterations in order to achieve convergence.
  • The "Abort" message which appears when trying contacting server is not functioning when the remote centre is not present.

5 Post-processing of the case

Although improvements have been made since the first version release, the post-processing tool is not as user-friendly as we expect to be. As previously reported selective zooming, which is very important in our cases, is still missing. We still use PHOTON and AUTOPLOT to produce results and graphics. Focusing on the street canyon region which is the region of interest with the use of VR-viewer, is a complex procedure and not always successiv e especially in 2D simulations.

5.1 Availability of all the essential tools for dataset interrogation. Limitations

We think that all the essential tools are available. But they need to be improved with regard to the following items:

  • possibility to draw the velocity field in a particular region: in our cases the region of concern (the street canyon) represents only a small piece of the overall domain. Drawing vectors outside this region is useless. Besides, above the canyon height the velocity vectors are too big comparing with these inside the canyon which make the flow field graphical representation ugly! An example is given below, for a 3D case:

 

 

Figure 9: The flow field for Scenario 2, Case_1 as it can be seen from the VR-Viewer

  • The same observation applies also to contours plotting.
  • Streamlines are difficult produced as you have always to move the probe position in the point of interest. In PHOTON this is done much more easier.
  • The different operation buttons are not informative.
  • The change of the range for contours plotting doesn’t work very good.

5.2 Speed of the VR viewer

The speed of the VR Viewer is a very serious problem even for 2D geometries. The response of the commands is most of the times restrictive to do anything.

5.3 Ease of use of the VR viewer

The comments made above are applied also here. It is still not very easy to work with the VR-viewer.

5.4 Visual appearance of the results in VR viewer

As in 5.2, 5.3 and 3.5.

5.5 Errors encountered

No errors were encountered.

6 The CFD simulation

6.1 Analysis of results

As written in Chapter 2, NTUA has developed 8 cases which apply to three scenarios. The results of the simulations are now described and commented for each scenario.

  1. Two - Dimensional Street Canyon

Four cases were developed according to the ratio W/H.

Case_1: W/H=1.0

Case_2: W/H=0.5

Case_3: W/H=0.7

Case_4: W/H=0.33

The simulations were performed locally on our PC except for Case_1 which was also submitted to MICANET remote centre.

The results are almost similar for the four cases. Some typical figures are given below to show the flow and concentration field predicted inside the street canyon region (all figures are not submitted to avoid creating an enormous file).

The flow pattern seems correct. The canyon is filled with one main vortex. The velocity values inside the canyon are reached a reduction of about two orders of magnitude from the free-stream velocity. In Case 1 (W/H=1.0) a minor secondary vortex is generated at the leeward lower corner which extends over only a small fraction of the street width (Figure 8). A similar secondary vortex is also predicted for the geometry W/H=0.7 (Case_3) but it is appeared at both the canyon corners (Figure 9).

Concerning the concentration field (Figures 10-11), in all cases the higher values of the pollutant are observed in the leeward wall of the street canyon which is what we expected.

fig 10

Figure 11: Streamlines inside the street canyon region for W/H=0.7

Figure 12: The concentration field inside the street canyon region for W/H=0.33

fig 13

  1. Three - Dimensional Street Canyon

Two cases were developed according to the ratio L/H:

Case_1: W/H=4.0, L/H=1.0

Case_2: W/H=4.0, L/H=2.0

Both cases were submitted to MICANET. Figure 12 shows the flow field with the recirculation regions formed upstream of the first building, downstream of the second building and inside the street canyon region, for Case_1. The length of the vortex formed downwind of the second building is increased with the ratio L/H.

The concentration field is predicted in Case_2. The pollutants are entrapped inside the canyon. The higher concentration values are observed in the leeward wall of the street canyon and extend over the half canyon width (Figure 13).

 

Figure 14: The flow field around the buildings for W/H=4.0, L/H=1.0

 

Figure 15: The concentration field inside the canyon for W/H=4.0, L/H=2.0

  1. Block of buildings

6.2 Validation data

The results of all the studied cases are compared against our local PHOENICS (v.2.2.1) predictions. The second case of Scenario3, that is the block of eleven buildings, is also validated using concentration measurements of the air quality monitoring network of Athens from the station of Patission street.

6.3 Validation of results

6.4 Parametric analysis

7 MICANET in the context of an industrial environment

 

8 Final recommendations