Question-and-Answer Resource for the Building Energy Modeling Community
Get started with the Help page
Ask Your Question

E+ Fatal error: Convergence error in SolveForWindowTemperatures

asked 2016-01-14 01:12:51 -0600

jperick1's avatar

updated 2016-01-14 10:19:37 -0600

Having adjusted the Relative Airflow and Absolute Airflow convergence tolerances, as well as the Maximum Number of Iterations under AirflowNetwork:SimulationControl, which solved an earlier fatal error (Error: Too many iterations in Airflow Network simulation), I have encountered the fatal error below and have run out of ideas for resolving it. A google search turned up a similar problem with E+ v7 over on the Design Builder Forum but without a solution.

The Airflow Network I am attempting to simulate includes several large window openings which I understand can cause errors due to the high airflow transfer between zones. Reducing the aperture sizes is an obvious solution, however, high airflow/transfer rates is a desired quality of this test simulation so I would prefer to retain the large window openings.

(Using e+ v8.3.0 x64) Thank you in advance

From the *.err file:

** Severe  ** Convergence error in SolveForWindowTemperatures for window ZONE_1_WINDOW_005MM
**   ~~~   ** During Warmup, Environment=RUN PERIOD 1, at Simulation time=09/29 08:30 - 08:32
**   ~~~   ** Glazing face index = 1 ; new temperature =NANC  ; previous temperature = 9.4479C    
**   ~~~   ** Glazing face index = 2 ; new temperature =NANC  ; previous temperature = 13.9687C  
**  Fatal  ** Program halted because of convergence error in SolveForWindowTemperatures for window ZONE_1_WINDOW_005MM
...Summary of Errors that led to program termination:
..... Reference severe error count=1
..... Last severe error=Convergence error in SolveForWindowTemperatures for window ZONE 1 WINDOW 005MM
edit retag flag offensive close merge delete


The error is related to the calculation of conduction through the window panes. The problem could be related to the calculated zone temperature with the airflow network. Does the problem go away if you disable the airflow network? (This will help diagnose where the problem is, we can turn it back on after we've identified the root issue.)

Neal Kruis's avatar Neal Kruis  ( 2016-01-14 10:23:22 -0600 )edit

2 Answers

Sort by » oldest newest most voted

answered 2020-02-06 12:10:02 -0600

updated 2020-02-06 14:02:23 -0600

Did someone find a solution to this?

I´ve tried @aaron comments, but the error Convergence error in SolveForWindowTemperatures persists.

I´m running the same model with two different weather files and only the second one (same location but Future weather 2050), throws this error.

UPDATE: For future reference, I´ve increased the minimum supply air temperature from the default 13ºC to 15ºC and the model runs without problems

edit flag offensive delete link more

answered 2016-01-14 13:31:40 -0600

There are a couple things you can try for a quick fix.

  1. Set the Initialization Type = LinearInitializationMethod. This will calculate the first iteration of flows and pressures assuming a linear relationship. All subsequent iterations will be nonlinear. A better initial guess will improve you're chances of convergence.
  2. Adjust the Convergence Acceleration Limit field. The solver uses a "Steffenson like" algorithm to reduce oscillations when steps in pressure are too large. One of the downfalls of Newtonian solvers is that they can fall outside of their radius of convergence when step sizes are large, so the E+ solver uses a kind of damper on the Newton step in an attempt to prevent this. However, it only applies the damper when the ratio of pressure corrections for the current iteration to the previous iteration is less than a critical value. This critical value is the Convergence Acceleration Limit. This field also controls the amount of damping in the Newton direction. A value between -1 and 0 will reduce the step size , and a value between 0 and 1 will increase the step size. Since you are trying to improve convergence, a value of Convergence Acceleration Limit = 0 will be the most conservative choice. You'll also see an increase in computation time. To my knowledge, no one has actually proven that this method improves the chances for convergence, so this may not work for you, but it's worth a shot.
edit flag offensive delete link more


@aaron sorry i cannot understand what have i do ??'s avatar  ( 2016-10-18 05:29:36 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer


Question Tools

1 follower


Asked: 2016-01-14 01:04:07 -0600

Seen: 1,199 times

Last updated: Feb 06 '20