First Roof calc has lots of convergence errors

[Ned Lyon]

I ran a roof simulation with default long-wave radiation parameters. I used some membrane roof material properties. The model runs error free without radiation, but has several hundred convergence failures with radiation. I'll try to isolate the inputs that cause the problem, but watching the calc time while it runs seems to indicate a rather random time point that is usually early to mid evening. The results do not seem to be greatly affected. Perhaps the radiation models only need a little more calculation time to finish converging.

Night temperature depression seems reasonable. Decrease in daytime high temperature puts roof surface more in line with other calculation estimates I have made for solar heated roofing temperatures.

Not of great importance - I like the ability to get the moisture transfer data, but it doesn't graph well because the numbers are so small. The graph numerical format fills up with lead 0's.

[Manfred Kehrer]

I ran a roof simulation with default long-wave radiation parameters. I used some membrane roof material properties. The model runs error free without radiation, but has several hundred convergence failures with radiation. I'll try to isolate the inputs that cause the problem, but watching the calc time while it runs seems to indicate a rather random time point that is usually early to mid evening. The results do not seem to be greatly affected. Perhaps the radiation models only need a little more calculation time to finish converging.

Night temperature depression seems reasonable. Decrease in daytime high temperature puts roof surface more in line with other calculation estimates I have made for solar heated roofing temperatures.

How did you design the roof. If it has an air gap mybe the explicit long-wave radiation is not a good choice. I will think about and answer again.

Can you send me all the input data. I´m interrested in the convergence errors

Not of great importance - I like the ability to get the moisture transfer data, but it doesn't graph well because the numbers are so small. The graph numerical format fills up with lead 0's.

Yes, I know, i still haven´t found a good solution for this. But i Think it´s OK for the BETA version.