SUMMARY

This page is intended as a summary of all the other pages combined. Please bear in mind that all other pages were related to Version 1 of VirtualWind, this page is referring to Version 2.1

VirtualWind is a CFD software used for simulating wind behaviors, that has been simplified enough to be specifically used by architects. VW is useful for visualising, using animations (rather than numerical data) what wind is doing in and around buildings. VirtualWind basically works by having a model of the building you want to test constructed using Sketchup, exporting it to VirtualWind, where you can then set your wind parameters (direction, speed), run a simulation and then run an animation to see what is happening. We think this would be useful in the early design phase of a building as you can easily simulate the wind impacts around or through your building and then edit your building influenced by your findings.
 * VirtualWind //-// What is it used for?**

Before beginning please read the tutorials on the VW website http://www.virtualwind.com/learning-center/101-training-videos they provide a very good overview.
 * Tutorials**

Running VW on Mac does not appear to be a great idea. We found that although it performed the calculations in a similar timeframe to a PC, it took a lot longer to output the data; in one instance the PC did it in a few minutes, the Mac 1.5 DAYS! The computer specs were not equal. I was running VW on windows 7, i5 quad core 4gb ram pc. Also the VW being PC based program the parallel program seemed to cause issues
 * Mac Vs. PC**

For animations Virtualwind produces .avi files, which are very large. We found that .mp4 files are much smaller and thus easier to attach to emails, upload online etc. We found QuickTime 10 for Mac to be perfect (just use File - Save for Web). Ironically, the latest free PC version doesn't perform the conversion, you will either need QT Pro, or find one of several free versions online.
 * Producing MP4 files**

This section is meant for use after you have gone through the Virtualwind tutorials. It cuts through to the essence of what we found to be most important when running a study, in succinct dot point form.
 * A QUICK GUIDE TO GETTING A VIABLE SIMULATION RUNNING**

By neat we mean only have what components are necessary and that they are sensibly grouped. Keep in mind that every “group” or “component” you have in your SketchUp will convert into an individual part in Virtualwind, this can make the Canaopy Manager more difficult to navigate than is necessary. Example: Exercise 2 When making this model it was easiest to make all the individual components (columns, beams, purlins) as groups so they could be moved independently. However keeping them as groups then meant a ridiculous amount of unnecessary parts with their own: “Exploding” all these elements in SketchUp and then having only 2 groups (a group for the walls and a group for the framing) makes the model far more manageable:
 * Point 1: Make sure your model is neat and necessary**

We have devised a simple rule of thumb for this - If we are interested in what the wind is doing INSIDE a room select CANOPIES If we are interested in what the wind is doing OUTSIDE and AROUND buildings select BUILDINGS This is as buildings are watertight geometry (e.g bounded boxes) and canopies are not.
 * Point 2: Canopies or Building?**

As soon as you load your model create your 'block actors' You do this by left clicking on the canopies and selecting “create block actor” After you do this select “continuous update” – this will save you the trouble of having to continually update your block actor to see how it looks when you change the wind resolution. The block actor is a very important tool as it allows you to visualise exactly how your model is being perceived by Virtualwind. You can now align the mesh in a way that best suits the model you are testing.
 * Point 3: Block Actors**

An easy step to forget, but if you forget this you will waste time submitting a study and not being able to visualise your results!
 * Point 4: Make sure you select your sampling quantities BEFORE you submit your study.**



This is done through the Vine viewer, left click the icon and select the cores that corresponds with the number of cores your computer has. //Note: make sure it is correct, foe example we assumed our computer had 8 cores when it only had 4 and this made our simulation run VERY slow until we rectified it.//
 * Point 5: Configure Virtualwind to suit your computer**

Sometimes when a study has an extremely high estimated time to run this might be because the time step is too small. If the time step is too large it will crash, if the step is too small it will take too long. Even though 400-800 time steps are recommended for the selected sampling time we have found as a rule this does not always work, and the finer the mesh the smaller your time steps must be, sometimes 1,000 or 10,000 or occasionally hundreds of thousands are required, but if you time step is in the millions or billions then it can be much lower and save you a lot of time! Trial and error is the way to go. Example: This is the default time step given for this model = 4.82208x10^-5 (note the fine mesh required to register the angle of the roof and the columns) This gave an estimates simulation time of roughly 2 days:
 * Point 6: Time Steps**

This is the time step that worked in the end, this took (this is still quite a low timestep, but is required in this case for the fine detail of the model):

This does NOT work in the same way that “save as” does. If you have a VW file open and after working on it you decide you want to save it under a new file name using the ‘copy study to” option it will save the new file at that point, but if you then continue to work on the file you will still be working on the original file. To work on the new saved file you must open the new file from the location you have saved it.
 * Point 7: Copy Study To…**

After becoming familiar with the software we spent 2 days using the university lab computers to assess the viability of VW being run by students in a lab environment. Our ideal aim was to have the urban study running on a powerful lab computer in a time of ~30 minutes. We were confident we could get this result, and if we did it would mean that students could run simulation at university in a sensible timeframe.
 * Version 2 Conclusions**
 * Lab Computer:**

Unfortunately we ended up spending two full days at university with numerous computers at our disposal without running any viable simulations, with the following problems:
 * Simulations that on out personal computers were having estimated times of up to 16 days (unreasonable)
 * Numerous Blue screen errors
 * Simulations crashing after several seconds of running
 * Simulations simply disappearing from the vine viewer 2 seconds from completion
 * Program freezing

To resolve these problems we tried adjusting the parameters of our files, e.g. number of time steps, domain size, mesh size.

We do not know why this is the case.

We have some questions that we have tried to find the answers to but are running out of time to investigate.

Q: Are there differences in how the computers are dealing with old //vs// new versions of VirtualWind?

Q: Could it possibly be because the lab computers are on a network?

Q: Could it have something to do with the high-level anti-virus programs required for these computers (as they are public computers)?

The lab computers have the following specifications:

Operating System: Windows 7 Enterprise 64-bit (6.1, Build 7600) (7600.win7_gdr.100226-1909) Language: English (Regional Setting: English) System Manufacturer: Hewlett-Packard System Model: HP Z400 Workstation BIOS: Default System BIOS Processor: Intel(R) Xeon(R) CPU W3565 @ 3.20GHz (4 CPUs), ~3.2GHz Memory: 6144MB RAM Available OS Memory: 6128MB RAM Page File: 2806MB used, 9445MB available Windows Dir: C:\Windows DirectX Version: DirectX 11 DX Setup Parameters: Not found User DPI Setting: Using System DPI System DPI Setting: 96 DPI (100 percent) DWM DPI Scaling: Disabled DxDiag Version: 6.01.7600.16385 32bit Unicode

We also tried a more powerful computer in the computing unit office which had the below specifications:

Operating System: Windows 7 Enterprise 64-bit (6.1, Build 7600) (7600.win7_gdr.100226-1909) Language: English (Regional Setting: English) System Manufacturer: System manufacturer System Model: System Product Name BIOS: BIOS Date: 11/03/09 13:28:23 Ver: 08.00.15 Processor: Intel(R) Core(TM) i7 CPU 975 @ 3.33GHz (8 CPUs), ~3.3GHz Memory: 12288MB RAM Available OS Memory: 12280MB RAM Page File: 1775MB used, 22780MB available Windows Dir: C:\Windows DirectX Version: DirectX 11 DX Setup Parameters: Not found User DPI Setting: Using System DPI System DPI Setting: 96 DPI (100 percent) DWM DPI Scaling: Disabled DxDiag Version: 6.01.7600.16385 32bit Unicode


 * SIMPLE BUILDING EXERCISE

This exercise is a development on Exercise 2.

The Exercise 2 sketch-up model was made before becoming completely familiar with VirtualWind and properly understanding how complicated a model really had to be. Initially I had even modeled all the purlins as I assumed even the most minor texture change to the roof might affect the wind behavior upon it. I now realise there are numerous unnecessary elements in that model and it woul dbe more beneficial to take the lessons learned nd apply them to a new and simpler modeled building of a similar size and intention.

This is a reworked version of Exercise 2. It still had a skillion roof, and many opening facing the predominant wind, and only a small opening to the rear. There are also some masses inside to indicate rooms, and all these rooms are pushed to the rear of the structure. One room has the door open and a high level window. **



Initially my computer refused to run the simulation, it would disappear from the the vine viewer after less than a minute with no error message given.

I tried to solve this issue by making further simplified versions of the building, trying lower windspeeds, larger time steps, but nothing would work.

I then decided to actually look in VirtualWind Help in the trouble shooting section and the following entry helped me:

//The solver doesn't start// //Two reasons may cause the solver not to start, i.e., in the ViNE viewer, you will see the preprocessor finishes but then the solver does not start or it quits without any error massage. If you check the simulation's "joboutput.txt" file (located in the "Raw" subfolder of the simulation folder), the last line is something like "Preprocessor completed normally".// //The first reason is that the wind resolution is set too high and it needs too much memory to run the solver. If this happens, what you can do is to reduce the wind resolution. Or, if you ran the simulation on a single CPU, you can try to re-run it on multiple CPUs – by distributing the job to multiple CPUs you might be able to use more memories if they are physically available on your computer (You need to have a license to support the parallel version of Virtualwind though).// //The second reason is in the situation where you try to run a simulation on multiple CPUs. The cause of this is likely that MPI, a service that the Virtualwind solver uses when running a simulation in parallel mode (supported by the Trail and Pro versions of Virtualwind), failed to install. This typically only happens on a non-XP operating system. (It might happen on Windows XP if you don't have sufficient privileges.)// //If this happens, you can try to install and start MPI manually. Here is what you need to do:// //The Virtualwind solver should now be able to correctly run in parallel mode.// //You should only have to do this once for each time that you install Virtualwind – Once MPI is installed, Windows knows that it should start MPI automatically.
 * //Open an command shell (or "DOS prompt"), with full administrative privileges. (Some OSs like Server 2008 have multiple command shells available. If you use one with lesser privileges, it won't work.)//
 * //In the command shell, navigate to the [...]/bin/ subfolder in the installation directory. If you used the default installation path, this is probably something like C:\Program Files\Virtualwind\Virtualwind 2.0\bin\. (On 64-bit operating systems, it might be in C:\Program Files (x86)\[...].)//
 * //In the command shell, type: smpd.exe -install//

I followed the steps, and found that my problem was only fixed to an extent. After doing the above if I then selected to run the simulation using only 1 core (even though I had assigned 2) // the simulation would run.

Every time I selected to use both cores the pre-processing would work, and then the solver would crash immediately after. I figured 1 core was enough, as if I adjusted my time steps to be suitable I would still get acceptable simulation times (under 5 hours).

I also found my computer was unwilling to cope with high wind speeds, or similarly if I assigned an acceptable wind speed to be measured from a low height (e.g. 3m/s at a wind speed height of 10m would cause problems whereas 3m/s with a wind speed height of 250m is fine.) I decided low wind speeds to be the compromise required to have any results.

Just for the record my computers specification are :



This is above the minimum requirements for VirtualwWnd but below the recommended, which are:

//Minimum Requirements// //Recommended System Parameters//
 * //2+ GHz processor//
 * //2 GB RAM//
 * //200 GB hard drive space//
 * //3D video card with 128+ MB memory//
 * //2+ GHz processor for Standard version and multi-cores for Pro version//
 * //4 GB RAM//
 * //500 GB hard drive space//
 * //3D video card with 512+ MB memor//

// **Simulation 1:

Unfortunately after running the simulation for 5 hours it crashed just before completion as my C: drive was full.

I tried running the simulations anyway but nothing had saved, and then the program crashed.

I cleaned up my computer, rebooted, and ran the simulation again, with a wind speed of 3m/s, a wind speed height of 250m and a timestep interval of 0.03 this estimated 5 hours. However when processing the final time step the study disappeared from the vine viewer with no error message.

I "copied study to" a new file name and re-submitted the study.

This time it completed successfully in just under 4 hours.

This is the animation:

media type="file" key="simplehouse2 plan view just above floor.mp4" width="300" height="300"

This animation is very interesting. Initially it looks like the wind is moving quite actively through the building, but then as the wind has to exit the building the small size of the outlet VERY significantly slows down the airflow, so much so that initially I thought the animation had froze. Th wind then accelerates rapidly again immediately after exiting the building.

I used these observation to change the building so that it had a small inlet, and multiple larger outlets to see the effect on the wind movement inside. I also moved the inner volumes to the side.



After several attempts to get submit the study, the animation below was the result:

media type="file" key="simple house edited plan near floor,.mp4" width="300" height="300"

This time the airstream goes in a straight line quite rapidly and then turns along the back wall towards the outlet opening. There is some movement in front of the outlet as some wind circulates the room before exiting. This is not exactly the result I was expecting, I assumed the air flow would not be so linear.

This is why I wanted to run a third simulation, I again edited the building so the rear outlet was twice as large as the inlet, with a small window on the southern end windward face of the building. However after trying to run this simulation a total of 4 times and having the study disappear when the final time step was computing I ran out of time. I am very disappointed by this, and this time could not explain the reason for the disappearance. ** //


 * Past Third Exercise

Andy and Siems Exercise**

On my personal computer I have not been able to manage a sensible period of time for this simulation (this means under 2 days on a fine mesh). Other times my computer seemed to give up in the pre-processing stage (which on average took slightly less than half an hour to simply process the model).


 * SUMMARY OF THE SUMMARY

Unfortunately whilst running the simulations using the latest version of VirtualWind more time was spent trying to resolve problems then actually running simulations. When the simulations did run there was valuable insight given that at times proved assumptions and in other instances completely disproved them. For this VirtualWind is a very useful tool! If the time spent ironing out problems was instead apllied to running more simulations I could really see the benefits of using VirtualWind in early stage sketch design. However at this stage the program may be too slow to keep up with the rapid pace of early sketch designing. If VirtualWind could focus on running simpler simulations faster then it would be an ideal program for Architecture students who want simple conclusions and do not have a great understanding of wind behaviors. Personally I know I used VirtualWind for it's visual data animations rather than for numerical data, and I suspect this is how most students would be using the program. **