Convergence problem

lalula2 · July 7, 2009, 22:36

I am having problem with my convergence for simulate wind flow over urban street canyon with different angle as shown in the figure below.

After 1000 iteration, the convergence became fluctuate up and down and did not converge as shown in figure below.

Anyone can tell me why this happened?
Any solution to solve for this. Thanks.
Btw i am using standard k-epsilon turbulent model with UDF on velocity inlet.

-mAx- · July 8, 2009, 00:58

are you sure that your solution may be a steady one?

lalula2 · July 9, 2009, 10:53

I am not sure wheter is steady or unsteady.
Any guideline to determine?
I nvr simulate unsteady flow before.
Can give me some advice on it?
Thanks.

lalula2 · July 9, 2009, 12:16

Sorry for ask such noob question,
can i know how to determine the time step size and number of time step required for my simulation?
Can someone mind to explain to me?
Thanks.

-mAx- · July 9, 2009, 12:46

if you don't have any unsteady BC (function with t), then you compute with steady solver and you monitor a variable such as pressure on inlet, or force on body etc...
either you get an steady monitor (constant with iterations) or you get an oscillating one.
If it oscillates, it could be due to an unsteady phenomenon.
Switch to unsteady solver and adapt the time step dt following.
dt # 1/2 . dx/(maxU) with dy is the minimal cell size in your domain and maxU the max velocity expected.
There is no right way to choose an dt, but this one is a good approximation ( for me )
You may also adapt the dt, if you see that the convergence is too slow (then reduce the dt) of too fast (then increase it)
Once again this how I handle such problems, but I am sure other people uses a different way

lalula2 · July 29, 2009, 11:16

why the convergence cant be too slow or too fast?
I hv set my time step size to 1 sec...
the convergence is very fast..but if i wana simulate for 10 minutes...it took me few days to complete it...
should i set my time step size larger?

-mAx- · July 30, 2009, 01:50

you can enable the adapt time stepping.
Solver will automatically adjust the time step.

July 7, 2009, 22:36	Convergence problem	#1
lalula2 New Member VLKOH Join Date: Mar 2009 Location: Malaysia Posts: 20 Rep Power: 17	I am having problem with my convergence for simulate wind flow over urban street canyon with different angle as shown in the figure below. After 1000 iteration, the convergence became fluctuate up and down and did not converge as shown in figure below. Anyone can tell me why this happened? Any solution to solve for this. Thanks. Btw i am using standard k-epsilon turbulent model with UDF on velocity inlet.

July 8, 2009, 00:58		#2
-mAx- Super Moderator Maxime Perelli Join Date: Mar 2009 Location: Switzerland Posts: 3,297 Rep Power: 41	are you sure that your solution may be a steady one? __________________ In memory of my friend Hervé: CFD engineer & freerider

July 9, 2009, 12:46		#5
-mAx- Super Moderator Maxime Perelli Join Date: Mar 2009 Location: Switzerland Posts: 3,297 Rep Power: 41	if you don't have any unsteady BC (function with t), then you compute with steady solver and you monitor a variable such as pressure on inlet, or force on body etc... either you get an steady monitor (constant with iterations) or you get an oscillating one. If it oscillates, it could be due to an unsteady phenomenon. Switch to unsteady solver and adapt the time step dt following. dt # 1/2 . dx/(maxU) with dy is the minimal cell size in your domain and maxU the max velocity expected. There is no right way to choose an dt, but this one is a good approximation ( for me ) You may also adapt the dt, if you see that the convergence is too slow (then reduce the dt) of too fast (then increase it) Once again this how I handle such problems, but I am sure other people uses a different way __________________ In memory of my friend Hervé: CFD engineer & freerider

July 30, 2009, 01:50		#7
-mAx- Super Moderator Maxime Perelli Join Date: Mar 2009 Location: Switzerland Posts: 3,297 Rep Power: 41	you can enable the adapt time stepping. Solver will automatically adjust the time step. __________________ In memory of my friend Hervé: CFD engineer & freerider

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 01:17
convergence problem	alexandre	FLUENT	2	June 14, 2007 14:31
convergence problem	ammi reddy	FLUENT	1	March 28, 2007 17:01
3D Fluid Flow Convergence problem	Emily	FLUENT	2	March 21, 2007 22:18
Non Convergence of 3D Heat transfer cfd problem	Balraj	Main CFD Forum	3	December 9, 2004 00:24

July 9, 2009, 10:53		#3
lalula2 New Member VLKOH Join Date: Mar 2009 Location: Malaysia Posts: 20 Rep Power: 17	I am not sure wheter is steady or unsteady. Any guideline to determine? I nvr simulate unsteady flow before. Can give me some advice on it? Thanks.

July 9, 2009, 12:16		#4
lalula2 New Member VLKOH Join Date: Mar 2009 Location: Malaysia Posts: 20 Rep Power: 17	Sorry for ask such noob question, can i know how to determine the time step size and number of time step required for my simulation? Can someone mind to explain to me? Thanks.

July 29, 2009, 11:16		#6
lalula2 New Member VLKOH Join Date: Mar 2009 Location: Malaysia Posts: 20 Rep Power: 17	why the convergence cant be too slow or too fast? I hv set my time step size to 1 sec... the convergence is very fast..but if i wana simulate for 10 minutes...it took me few days to complete it... should i set my time step size larger?