different results on different computers

Replies to This Discussion

Permalink Reply by Elzine Braasch on September 22, 2015 at 4:11pm

Are you able to show us the results, along with the inputs you use to do the daylight simulations?

My guess is that an input within the file is different. I am not sure if different versions of E+ or Daysim would matter, but it does seem odd if you are simply using the same "imaged" computers and the same gh file that the results would be different...

Permalink Reply by Mostapha Sadeghipour Roudsari on September 22, 2015 at 6:30pm

Radiance uses a stochastic sampling for calculation and 5-10% difference is expected. If you want to minimize the difference you can set radPar to be high. Also check this post on Radiance discussion group: http://comments.gmane.org/gmane.comp.graphics.radiance.general/2685

Mostapha

Permalink Reply by Fahimeh Motazedian on September 23, 2015 at 8:59am

Hi Elzine and Mostafa,

unfortunately I didn't take a picture or saved the result, actuly it was conclusion of a series of calculation.

but it had difference was high,

for example I tested some WWR according to UDI in Tehran, and the best one was 40% in the first computer and the same gh file on the other computer showed me 16% which sounds wired!

Now, I am working on the second computer and don't have access to the first one, I guess something is wrong with this computer, I can't trust on the results! Is there a way I get sure that the results are OK?

Permalink Reply by Mostapha Sadeghipour Roudsari on September 23, 2015 at 10:06am

Fahimeh,

I think you're confusing two different topics together. My understanding is that you're getting two different optimum options based on the result of the optimization algorithm. It can have several reasons such as finding a local optima instead of the global optima, or having a weak objective function. This doesn't necessary have to do with Honeybee's daylight analysis. To have a better understanding about what is happening you can re-run the best cases against each other and compare the results. Your problem may have several local optima.