Third+exercise+-+urban+scale+investigation

SIEM:

Below is the Urban Block Model thus far.



I'm having some water-tightness issues, I didn't even realise my model making was so dodgy...

I'm going to keep at it a bit longer, but decided I should put this up if anyone feels like giving it a tweak, and posting up a newer version.

It is still missing the heights of the test buildings. They are currently on a hidden layer called "TEST BLDG".

Also there are some interesting shaped buildings on the perimeter of the model - this is where the turn table previously cut through the buildings (compare dwg to dwf). since I used the dwg to make the sketchup model.

Mmmm the size of this model scares me a bit.

STEVE: I might be showing my naivete, but are you having trouble picking up points in space that are 'co-planar' (in the same plane) to make the building surfaces? I would have thought the only way making this model **//quickly//** would make sense, is to get a simplified ground plan for each building as a single plane, and then use the tool to extrude it to the correct height. If there is a variation of the form, then to extrude each building in stages?

ANDY This is a new version of the site model, it still needs some work.

I could not find a way to get the DWF to open in either AutoCAD, ArchiCAD or sketchUp, this is important as there ia a lot of linework missing from the DWG file that would make the extruding easier. Ihave extruded what i can but need the linework for the complicated buildings; while i'm sure we don't need all the detail i felt there was too much variety in the built form to ignore, maybe being too pedantic!!

I'm happy to do the extruding if someone can post the DWF file info in a better format, ideally straight into the skp file.

FYI. i scaled up the DWG by 800% as it said the turntable represented 450m radius

haven't tried to check for watertightness yet as i'm on a different machine, hopefully all will be well

ANDY

OK, one watertight model!

there is still detail missing, all the test buildings are modeled as 1m high (so they are watertight) as a couple of others nearby are too. Still can't get the lines out of the DWF file and don't have any heights for the test buildings.

ANDY I should explain.... in order to create a water tight model i've been using the checking for w/tight functions; this has always encouraged the turning on the hidden geometry. Long story short, the skp file i have is full of hidden geometry which i am unable to add a material or any surface too. Please see where i have gone astray. The main issue when modeling was that everytime i used the check for w/tightness it automatically took away all the surface leaving only hidden geometry, i cannot find a way to get it back.

NEWS FLASH: BY GOING TO EDIT- HIDE-UNHIDE ALL IN SKP IT ALL REAPPEARS!!

Also, i had to take the dwg into archicad, create 'fills' for the buildings and then delete everything else to get rid of all the hidden crap in dwg file.

Over to someone else.......

Sam: Ok we have a model with the 'plaza' modelled with heights and basic form etc. I have attached the sketchup and VW file below. If you have a look and let known your thoughts.

Ok I ran a crude sim for the model above. It worked ok and managed to get a wind pattern that looks like its correct. media type="file" key="urabn 1.mp4" width="300" height="300"media type="file" key="urabn 2.mp4" width="300" height="300"

Model was made as canopy with a wind speed of 20m/s with default setting for direction. The domain was made smaller to focus on the site and to save time etc.

The sim shows the air stream passing through the streets and how that influences the wind path and pattern through the cluster of high rises and open space. as expected, the turbulance and the directions of the velocity vectors on the 'main streets' have a much larger ranges. I will run some more to get familiar with the model and at some point we will have to break down tasks and simulations etc.

I ran this simulation on a PC HP i5 quad core, 1GB graphic, 4MB RAM. Simulation took 15minutes. Bargain:)

(SEAN) : I envy your simulation time. My setup is roughly identical to Andy's computer therefore it will take me the 4 hours to run just as a coarse model. If I run it on even the medium wind resolution I am quite certain it will buff out to a min of 15 hours. I am a bit frustrated by how long it takes on my machine.

Sam can you post a couple of the larger resolution movie files of your simulation so I can get a better understanding of how the model is performing.

Q: Should we set up a structured approach to this urban exercise?

Q: Are we going to apply a considered alteration and then test it?

This simulation was with a new direction and with the analysis grid at different heights to show the wind patterns.

__Can we get an address to apply the appropriate climate conditions.__

media type="file" key="urabn 4.mp4" width="300" height="300"media type="file" key="urabn 12.mp4" width="300" height="300"

//ANDY A technical question. My computer (MacBook with 4GB RAM) actually performs the simulation in around 8 minutes but then takes approx 4 hours to "write frame..." Has VWind every been tested when running on a VirtualPC?

DEREK//

//1) Virtualwind has not been tested by us on a Mac under any conditions; hence, it's difficult for us to diagnosis problems running Virtualwind on Macs.// //Admittedly, we would very much like to test Virtualwind on Mac's running Windows (using whatever means possible). Since Virtualwind is built in a completely cross-platform manner, in theory we could release a Mac version; however, the reality we are currently stretched very thin on the human resource side of things.

2) In general, the following information and checks are useful in diagnosising or reporting performance of Virtualwind://

//info//

//a) The resolution, or specifically, the size of computational grid (e.g. 32x32x23). b) The number of cores the solver was run on (e.g. run on a single core). c) The characteristics of the machine (e.g. a single CPU quad core machine with a processor speed of 3.2GHz, 3 GB RAM and 500GB Hard Drive). d) The simulation time (e.g. It took 15 minutes to run). e) The number of animation frames written and amount of data produced (e.g. 20GB of simulation data written to file for 50 animation frames). f) The max simulated real time (e.g ran the simulation out to 200 sec). g) The amount of RAM used in running the solver (vwArchitectMPISolver.exe: see below).

checks//

//a) While running the Virtualwind solver, one can check the amount of RAM it is using up, and the % of CPU processing power spent on Virtualwind. To check this, look in the "Task Manager" (or whatever tool lists the processes running) and look for "vwArchitectMPISolver.exe". This process is the Virtualwind solver, and looking up these runtime properties will give you a sense of how much memory the solver is using up, **for the particular simulation running**, and what % of the CPU's effort is at the moment spent on running the simulation.

b) Checking that one has not run out of disk space on the maching running Virtualwind. Since it is reasonably easy to set-up a simulation that can generate 10-100 GB of data rather quickly, when things appear to be going wrong, this is another good reality check.//

//c) One final comment. Depending on the configuration of your machine, sometimes when a process is running and requires more memory than available, it can go into a "hard disk swapping" state, where it trys to copy "stuff" (things stored in memory) back and forth from RAM to hard disk as they are needed by the process. The Virtualwind solver will grind to a haul if this situation arises, and the % of CPU processing power spent running the solver will be less than 1%. In this event, the simulation is too large for the machine.//

Sam: Urban wind study with appropriate wind direction. Wind Rose for Kuala Lumpur:

Wind Study for January with a NW wind direction: media type="file" key="NW 00.mp4" width="300" height="300"media type="file" key="NW 03.mp4" width="300" height="300"media type="file" key="NW 06.mp4" width="300" height="300"media type="file" key="NW 09.mp4" width="300" height="300"media type="file" key="NW 12.mp4" width="300" height="300"media type="file" key="NW SECTION 01.mp4" width="300" height="300"media type="file" key="NW SECTION 02.mp4" width="300" height="300"

Wind Study for May with a WSW wind direction: media type="file" key="WSW 00.mp4" width="300" height="300"media type="file" key="WSW 03.mp4" width="300" height="300"media type="file" key="WSW 06.mp4" width="300" height="300"media type="file" key="WSW 09.mp4" width="300" height="300"media type="file" key="WSW 12.mp4" width="300" height="300"media type="file" key="WSW SECTION 01.mp4" width="300" height="300"media type="file" key="WSW SECTION 02.mp4" width="300" height="300"

Wind Study for September with a SW wind direction: media type="file" key="SW 00.mp4" width="300" height="300"media type="file" key="SW 03.mp4" width="300" height="300"media type="file" key="SW 06.mp4" width="300" height="300"media type="file" key="SW 09.mp4" width="300" height="300"media type="file" key="SW 12.mp4" width="300" height="300"media type="file" key="SW SECTION 01.mp4" width="300" height="300"

As a group we did some research to get the wind direction and run the simulation with the three (3) identified wind directions - see wind rose's above.

It would be good if we can all have a look and pass on thoughts and comments of the analysis.

The wind path on all three studies do show high velocity vectors where 'canyoning' occurs and also how the buildings influence the flow and wind path. Where clusters and same height building occur there is low vectors and wind shadows. The sections show this well.

I will write a more detailed analysis and try and simulate a 3D study.

Technical comment: Simulations run on PC i5 quad core, 1GB graphic, 4GB RAM. Simulation took 37minutes. Model: The cage was set at 512/512/64 with a 20m/s wind flow with the directions mentioned above. No major issues occurred while running these simulations, only a human error which I sussed out myself.(remember to update cages) :)

ANDY These are great! Maybe as a final run we could look at the velocity vector animations and zoom right in on the areas around the base of the towers, then adjust the colour limts map to see if we can pick up if there are any areas of turbulence in what appear to be dead zones. I'm thinking this is where the colour limits function will be most useful to highlight smaller differences.

Sam: Sounds good. We could look at areas that we think will be highly used in the plaza. Should we do this for all three(3) widn directions or should we concentrate on one direction?!

The colour contrast will be easy enough to change to highlight the differing vectors.

Sean: I agree, the colour map will come in handy when start focusing in closer. Note: (I think it would be intuitive if you could in the VW 3d window nudge the cages into the correct position using the arrow keys. It would save time fiddling about trying to remember which is the x and y directions.)

Sam: The error message that reoccurred all day!!