Strange time step selection issue (for information only)

JuPa · February 27, 2017, 06:28

I'm performing a series of transient population balance simulations using the MuSiG model.

With my chosen time step I achieve stable convergence within 3 coefficient loop iterations per time step; the imbalances look sensible as do the monitor points.

When I gradually increase the time step so it is double or triple the original value, I still achieve stable convergence within 3/4 coefficient loop iterations per time step; however the simulations will often randomly crash citing divide by zero errors.

It's strange - all time steps (the small and large ones) give me good convergence behaviour (residuals, imbalances, and monitor points) yet the large ones cause the simulations to diverge, DESPITE showing good residual history.

The population balance models must be sensitive to time step.

To fix the problem I'm using the original small time step.

I guess this thread is for information only

ghorrocks · February 27, 2017, 06:56

When you use highly non-linear and complex models like MuSiG multiphase you can get into situations where small differences make the difference between convergence and non-convergence and the usual signs (like residuals) hardly show a difference.

You might find that other measures are actually a better measure of convergence than the residuals in your case. Imbalances is certainly one to try - with several coupled phase populations to get conservation over them all will be challenging. But there may be other measures of convergence in your case which are better.

February 27, 2017, 06:28	Strange time step selection issue (for information only)	#1
JuPa Senior Member Mr CFD Join Date: Jun 2012 Location: Britain Posts: 361 Rep Power: 15	I'm performing a series of transient population balance simulations using the MuSiG model. With my chosen time step I achieve stable convergence within 3 coefficient loop iterations per time step; the imbalances look sensible as do the monitor points. When I gradually increase the time step so it is double or triple the original value, I still achieve stable convergence within 3/4 coefficient loop iterations per time step; however the simulations will often randomly crash citing divide by zero errors. It's strange - all time steps (the small and large ones) give me good convergence behaviour (residuals, imbalances, and monitor points) yet the large ones cause the simulations to diverge, DESPITE showing good residual history. The population balance models must be sensitive to time step. To fix the problem I'm using the original small time step. I guess this thread is for information only

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Multiple floating objects	CKH	OpenFOAM Running, Solving & CFD	14	February 20, 2019 10:08
Extrusion with OpenFoam problem No. Iterations 0	Lord Kelvin	OpenFOAM Running, Solving & CFD	8	March 28, 2016 12:08
Stuck in a Rut- interDyMFoam!	xoitx	OpenFOAM Running, Solving & CFD	14	March 25, 2016 08:09
plot over time	fferroni	OpenFOAM Post-Processing	7	June 8, 2012 08:56
calculation diverge after continue to run	zhajingjing	OpenFOAM	0	April 28, 2010 05:35

February 27, 2017, 06:56		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	When you use highly non-linear and complex models like MuSiG multiphase you can get into situations where small differences make the difference between convergence and non-convergence and the usual signs (like residuals) hardly show a difference. You might find that other measures are actually a better measure of convergence than the residuals in your case. Imbalances is certainly one to try - with several coupled phase populations to get conservation over them all will be challenging. But there may be other measures of convergence in your case which are better.