Condensation Potential seems very "snappy" - even random - are there rounding errors?

[Georg Reichard]

I've explored THERM (7.4.3) for assessment of condensation risks with my students and came across some strange behavior with a very simple problem.

We built a corner of just 8" Concrete (medium weight) and applied boundary conditions of 10°F on the outside and 70°F, 60% RH (fc 1.46 BTHUh/sf.F), which brings the interior surface temperatures into a range of 56.4°F to 47.4°F in the corner - all good so far.

Now when calculating the dew point, I'd come to a temperature of 55.48°F - which would mean that some surface points are above, and others closer to the corner are below the DP - indicating a condensation potential.

THERM however shows the entire interior surface temperature at risk.

Is this now a rounding error?

I tried to set the RH to 55%, which would result in a DP of 53.09°F - still the same result, the entire interior surface is show with a condensation potential.

I then set the RH to 50%, (DP 50.51) - and suddenly there is no potential indicated any more - even though my corner reads 47.4°F. (I went into mesh view and checked the temperatures at the individual nodes)

Funny thing is that the entire interior surface "switches sides" between 52%RH and 53%, which does not correspond to the surface temperatures vs DP relationship along these boundaries.

At this point I have no explanation, other than there could be some issue as I only use one material, one domain, etc. - but I doubt that rounding is the issue here.

I can share the THM files if anyone wants to look into this.

Any pointers highly appreciated ...

Best,

Georg

[Christian Kohler/LBNL]

Georg,

This certainly seems like a bug. The condensation potential has only been lightly tested, so we appreciate your feedback. 

As a test, could you insert a few points on the interior surface of the concrete, and rerun the boundary conditions? This should break the boundary into segments. I'm curious if all of the segments switch at the same time, or if it is by boundary condition segment. It should color it based on the mesh, but that does not seem to happen based on your report.

Thanks

Christian

[Georg Reichard]

Christian - sorry for the delay, for some reason the notification on your reply didn't come through (maybe I just overlooked it) - thanks for the follow up and pointer!

Your suspicion was correct - I've added a point along the interior boundary and that breaks the condition. Seems like the condition is checked at vertices only and then "painted" along the entire edge.

Here's the comparison - the right/upper leg of the corner (shown horizontally) has an additional vertex in the middle of the segment - and that's exactly where the condition switches. Once again, this is not exactly where the surface risk really changes, but at least we are on the correct track resolving the issue.