[TRNSYS-users] Error creating transfer function coefficients

Michaël Kummert michael.kummert at gmail.com
Tue Nov 20 06:28:58 PST 2007


There are a few known workarounds when Type56 is not able to calculate
transfer functions. I believe they have been discussed in the past but here
are the ones I know:

- Sometimes the problem occurs for highly insulated walls that are not
particularly thick or heavy. In that case replacing some insulation layers
with mass-less layers can do the trick. Usually the thermal mass of the
insulation layer can be neglected so this is my preferred approach.

- Split the wall in two (or more) parts. You should split the wall as you
describe (e.g. half of the thickness in each wall for a uniform wall) but
then add a fictitious zone between the two parts of the wall. You basically
use that fictitious zone as an infinitely small and infinitely conductive
air layer. You should set a very small volume and infinite convection
coefficients in the fictitious zone. I think it is best to handle windows
separately from the wall if you adopt that approach (windows do not need to
be part of a wall in Type56). 

- Increase the time base. This is obviously not a great idea if you also
have much lighter walls in the building, but if you are simulating a castle
with 3-m thick walls all over I guess that would be acceptable. 

In any case I would setup a test model with standard walls that are OK,
apply the workaround (e.g. add the fictitious zone) and check that the
results are the same. Then you can replace the walls with the one that
causes the problem.

I hope this helps,

Michaël Kummert
ESRU, University of Strathclyde

> -----Original Message-----
> From: Kim Goethals [mailto:kim.goethals at ugent.be] 
> Sent: Monday, 19 November, 2007 17:01
> To: trnsys-users at engr.wisc.edu
> Subject: [TRNSYS-users] Error creating transfer function coefficients
> Hello,
> I found the following message. My model suffers from the same 
> problem. Isn't there an easier way to simulate a very thick 
> insulated wall? I tried the following: I splitted the wall 
> with for example 5 layers in two walls, respectively with 2 
> and 3 layers. I define the same surface, orientation and the 
> same CHTC's. I only incorporate the window in one of the two 
> walls. From comparing the simulations of both test cases, I 
> conclude that this option doesn't work. Please some feedback 
> and hopefully a solution.
> Thanks in advance,
> best regards.

More information about the TRNSYS-users mailing list