HI Asltan and Kera (et.al)
I am running some simulations on four models (Daylight Autonomy). All are replicas of the first one, but I am just varying the rotation. (studying the rotation and WWR). I was ready with all my calculations but today I repeated a simulation of one of the models , to be sure of the results, big surprise, the results vary, and a lot. I started re-reunning some of the calculations and results tend to vary, sometimes by 30% to 40%!. What could be producing this data oscillation. Am I doing somenthing wrong , or is it part of the software?
The grid-map does not vary, "The predicted annual electric lighting energy use" is what is changing. (I have 2 nodes selected)
There could be a few explanations for this. Could you please send us your results and a snapshot of your model?
I'm not sure exactly what problem you are having yet, but if you have fine geometry like venetian blinds, if the on center spacing of your sensor grid nodes is far apart, and if your settings are low, a swing like this is quite possible. DIVA runs Radiance with stochastic randomness turned on, so the rays that are sent out during each simulation can vary from run to run. This effect is amplified the lower your settings are. If you can imagine: a ray might land between your nodes on one simulation, but hits a node on the next one, causing your results to vary greatly.
You can turn off the stochastic randomness by editing your Radiance Parameters adding a "-u" in the field, such as before "-ab" (a space should separate the -u from the next variable). I would also recommend increasing your settings.
Let us know about your results so we can figure out what is going on in your case,
HI Kera, am using the normal ,0.8 height and 0.425, spacing grid. I was wondering if it could be about having more than one model in the Workspace, but they are really simple boxes. (imported from DesignBuilder as DXF, if it says anything).
I have all the models in Dropbox, just as a plus data, in case running the simulations from files in a special place, says anything,.....again. :)
It must be somenthing about my model, I don't think the stochastic randomness is producing such variation.
It shouldn't have anything to do with having more than one model in the space.
I'm having trouble downloading your files. Can you send a link to the files in Dropbox?
Here it goes, for all.
Thanks for uploading these, but I still need some more information. What Radiance settings are you using? I want to repeat the simulations, but I don't know what parameters you are using. It sounds like you are running the advanced shading and lighting modules. What parameters are you using for those? Do you have the Daysim Reports that you refer to in your emails above? Can you send those along as well?
Hi Kera, thanks for your answer.
I upgraded the folder.
the radiance settings are: -ab 2 -ad 1000 -as 20 -ar 300 -aa 0.1
and the lightning control is dimming w. Occupancy Off Sensor (2 points I guess should be able to see them.).
Tell me if you can run them.
Well, after looking over your model, I wasn't able to reproduce the discrepancy. I have attached 3 DIVA Daysim Reports. The first two are exact same runs of the first model that you send to me "Rhino 30WWW.3dm" (which I renamed "Rhino30WWW-KL.3dm"). You can see that I get the exact same results in both runs. These are also very similar to your "A" results. The two nodes are transposed, but the results are nearly the same.
One thing I did notice is that when I opened your model, the ground plane was locked. Any geometry that is hidden, locked or not assigned a material is not factored into the calculations. I ran the model again correctly with the ground plane unlocked, and the DA went up to 84%, although the electric lighting usage didn't change much (see third attached file).
The last thing I can think of is that DIVA doesn't like spaces, in model names (or View names). If you have two models named: "Rhino 30WWW-A.3dm" and Rhino 30WWW-B.3dm", to DIVA these both look like "Rhino", because it stops reading after the first space. I recommend, if you are running variations, to save each file as a separate name, without spaces, to minimize confusion in the results. Perhaps this is another explanation for the discrepancies?
Having looked over your models, I will note the following in general
*Unlike energy simulations, with daylighting simulations, you need to model wall thickness
*You should reduce the size of your ground plane. While the ground plane is critical for the calculations, if it is too big with relation to your model, it reduces overall accuracy.
*Your need to increase your settings for accurate results. I don't know the goals of your design, but the settings are very low.
*Run a low-quality visualization to check that your geometry is exporting properly.