Visualisation of airflow network calculation convergence in E+?

asked 2016-02-04 09:54:14 -0500

Hello,

Sometimes my airflow network calculations converge, but most of the time they don't. And it doesn't seem to make sense to me. I must admit that I altered some components in the source code, but I don't think they cause the problem.

Does anyone know a way to visualise the residuals in an energyplus airflow network calculation? ( Or to save the iteration steps to a textfile. )

Thanks!

edit retag flag offensive close merge delete

Comments

Could you be more specific about your modifications?

Jason DeGraw ( 2016-02-05 09:16:20 -0500 )edit

Hi Jason, sure. I created a self-regulating vent component that limits the flow through the vent when a maximum dP value is reached. In the cases that do converge this component seems to be working. I am also in the process of creating an advanced hybrid ventilation manager. However the calculations don't reach the point where this is called, so that can't be the problem. I suppose the Newton-Raphson algorithm is not converging because the initial guess is too far away from the solution. By looking at the residuals I hope to get a better insight.

BBel ( 2016-02-05 09:36:02 -0500 )edit

If that is what you are doing, then I wouldn't be too sure that your modifications are not the problem. It depends on exactly how you've changed things, but it could (effectively) decouple parts of the flow network in a way that the solver does not expect. But good luck!

Jason DeGraw ( 2016-02-05 19:15:58 -0500 )edit

add a comment

Comments

Thanks Aaron. I appreciate it. For me this is a good solution. I was already doing something similar with cout to the screen at the same line for both absolute and relative convergence. However I was hoping someone would have a more general solution that would be useful for other forum members without changing the source code.

BBel ( 2016-02-05 18:44:06 -0500 )edit

Ok, I don't know of a built-in feature like this, but someone else on the forum might still know. You've probably already seen this too, but check out the int called LIST which controls some built-in reporting. It's hard coded to 0, so you'd have to recompile to change this. It might be a leftover feature from AIRNET that has been disabled. If you continue to have issues with convergence (and you have some time and patience), you may want to look into modifying the solver to the Levenberg-Marquardt algorithm. I've had success at modeling hydronic flow networks with this method.