Convergence Failures

All about WUFI Pro
Post Reply
Toby Cambray
WUFI User
WUFI User
Posts: 16
Joined: Mon Jan 11, 2016 1:28 am -1100
Contact:

Convergence Failures

Post by Toby Cambray » Tue Apr 11, 2017 1:03 am -1100

I am interested in how the number fo convergence failures is defined.
If I have a 1 year simulaiton with 8760 hours, and I get say 8 failures, does that mean eight of the hours has failed, or about 0.1%? Or does it refer to the calculations at every node in the model, e.g. if I had a 100 node model, my 8 failures is compared to 876000 calculations, about 0.00001%?

Toby

Christian Bludau
WUFI SupportTeam IBP
WUFI SupportTeam IBP
Posts: 1129
Joined: Tue Jul 04, 2006 10:08 pm -1100
Location: IBP Holzkirchen, the home of WUFI
Contact:

Re: Convergence Failures

Post by Christian Bludau » Tue Apr 11, 2017 1:11 am -1100

Dear Toby,
yes, 8 hours where the iteration does not reach the convergence criterion. The convergence failures are counted per calculated time step.

Convergence failures
To limit the computation time the solver is stopped after a maximum number of iterations, even if the convergence criterion is not yet reached, and a convergence failure is registered. This may lead to only small differences in the moisture balances but is usually acceptable.

Diverging iterations also lead to convergence failures - but usually with significant differences in the moisture balances - such results are not reliable

Please note: the convergence failures should always be evaluated together with the balances!

Balances:
Balance 1: Change in the total water content during the calculation period (negative: drying, positive: moisture accumulation) – this amount of moisture is displayed in kg/m² over the calculation period
Balance 2: Sum of moisture fluxes through the surfaces and released by sources Differences in balance: Numerical errors or inaccuracies can cause water to „appear“ or to „disappear“ in the assembly (longer periods cause higher differences in balances) - Differences in balances should be zero or small
If the differences in balances are small relative to the total amount of water taken up or dried out, or relative to the total water content, this may be harmless (depends also on the type of construction).
- High relative and absolute differences in the balances indicate unreliable results!!

Hope that helps,
Christian

Toby Cambray
WUFI User
WUFI User
Posts: 16
Joined: Mon Jan 11, 2016 1:28 am -1100
Contact:

Re: Convergence Failures

Post by Toby Cambray » Tue Apr 11, 2017 1:27 am -1100

Many Thanks Christian

Kostas
WUFI User
WUFI User
Posts: 5
Joined: Wed Nov 17, 2021 3:59 am -1100

Re: Convergence Failures

Post by Kostas » Thu Dec 02, 2021 3:12 am -1100

Hello there,

In a previous answer in this thread, it is stated that differences between balance 1 & 2 should be zero or small.
How can we define a 'small difference? Can we assume that it should be close to 10% for example?

I have also noticed that if the values of balance 1 and 2 are very small (below 1 kg/m2) then the relative error can be large.
In this case I understand that we can we use the absolute difference between balance 1 & 2 over the total water content. It is just not clear to me which value to use (i.e. the total water content in the end of the simulation for example?)

Regards,
Kostas

saricioglu pelin
WUFI User
WUFI User
Posts: 20
Joined: Fri Jun 03, 2022 12:07 am -1100

Re: Convergence Failures

Post by saricioglu pelin » Sun Oct 23, 2022 12:40 am -1100

Hello everyone,
When i finished my calculation and can not see the graph results because of this warning;
Adsız.png
Adsız.png (171.05 KiB) Viewed 2937 times
What is the problem ? i don't understand please reply me ?
Thank you.

Christian Bludau
WUFI SupportTeam IBP
WUFI SupportTeam IBP
Posts: 1129
Joined: Tue Jul 04, 2006 10:08 pm -1100
Location: IBP Holzkirchen, the home of WUFI
Contact:

Re: Convergence Failures

Post by Christian Bludau » Wed Oct 26, 2022 11:59 pm -1100

Hi,
please check if there are values in the table "Status of Last Calculation". If all fields are filled with "NaN", something went wrong with your calcualtion and there are no results to show.
Christian

saricioglu pelin
WUFI User
WUFI User
Posts: 20
Joined: Fri Jun 03, 2022 12:07 am -1100

Re: Convergence Failures

Post by saricioglu pelin » Sat Dec 03, 2022 7:12 am -1100

Thank you Christian. It is solved.

Post Reply