I am running an HB grid based radiation analysis on a fine grid and I started getting these ‘dead patches’ on the results mesh which appear in different sizes and location every time I run the exact same script again.
I have not overridden any RADParameters and have my geometry built as HB surfaces with custom radiance materials (which mostly differ in their rgb reflectance). Note that I had to make some materials with rgb reflectances as 0.0 for this test case, but can this be an issue with Radiance?
It is more helpful to look at your file briefly, but I assume your geometry is not clean. According to your screenshot, you imported your geometry from somewhere else or you did not directly build your 3D model in Rhino. Your script often does not recognize your geometry well if you bring it from somewhere else or it is not clean enough. First, I would recommend you to rerun your analysis after rebuilding a few geometries, such as the middle and the right top buildings which seem to influence your ‘dead patches’ or problematic.
Update: I was able to get a clean result without any glitches with the exact same files after I restarted my computer but the patches started to appear again only after a few runs and all I changed was one material. So strange! Do you think it may be a Radiance caching issue?
If I make the grid coarse, the weird shapes on the mesh are not noticeable so I cannot say whether the problem still persists or not or whether it is related to grid size.
Also, yes the mesh geometry is imported from a cad file. I have done some routine checks for bad objects etc before setting up the analysis and did some test runs. Seemed to work fine until suddenly these patches started to appear. I am checking the geometry again now.
In all my time using LBHB tools, this is the first time I am noticing this kind of error!
I get the same ‘dead patches’ after a second run. I tried to set -aa 0 and -lw 0.000001 in addionalP_, where -lw is found by 1/(ad^2), and then the ‘dead patches’ did not appear - I tried 3 consecutive runs. I’m not sure if these parameters are considered to deliver ‘true’ results for this case and did not compare with the default parameters, but just by a quick comparison with your images it seems to be pretty close.
@mikkel Thank you so much! This solves the issue on my end too.
I will compare these to the previous good runs that I got, because for my purpose, I do need ‘true’ results. But at least I can move forward with this . Still don’t understand what exactly caused it but seems like it may not be something on the model/script side.
What I am doing is subtracting the results of two simulations by only changing one variable (reflectance), so I am hoping that any net effect of these RADparameter values to change the ray weightage, ambient interpolation error etc should cancel out. It will still affect the output but should work for this comparative study.