divergence of scaled residuals (steady simulation)

hanka · November 23, 2010, 07:55

Hi everybody,

do you have any suggestions why scaled residuals (mainly continuity) in steady simulation drop at first, but started to grow after ca. 20th iteration? Smaller mesh cells and second order upwind discretisation quite helped, but the trend is the same.

Are these results even relevant?

Thanks for any suggestions or ideas

siri · November 24, 2010, 02:06

what flow are you solving for ? Your problem might be due to

- improper BCs
- complicated physics or
- inefficient mesh field

other causes might exist, but could say the above in general

hanka · November 30, 2010, 04:09

Hi Siri, thanks.

It is a non-spreading fire in an empty cuboidal room. The fire is defined only by energy source (W/m3), which is generated in a cylinder area. There is a door opening (interior) to the next empty cuboidal room, where the walls are pressure outlet. I am interested especially in temperatures in the room and a flow through the door.
The mesh is hexahedral with the same size in the whole domain. If I made the cell size smaller, the residuals got better - increasing of continuity and energy was only very slight (other residuals were nicely decreasing).

I am only a beginner in Fluent (I have more experiences only with CFD programmes specialised to compartment fires), so I am sorry, if my questions are quite stupid or irrelevant...But is it possible to use these results or does this behaviour mean that there is a serious problem? Or what do you think about it in general?

hanka · November 30, 2010, 05:16

I have probably detected the main reason: the energy source of the fire is too high. The residuals do not increase if this value is much lower. The problem is, that this simulation should correspond to a real fire test, so the released heat and the fire size can not be changed.

siri · November 30, 2010, 06:07

Quote:

Originally Posted by hanka

The residuals do not increase if this value is much lower. The problem is, that this simulation should correspond to a real fire test, so the released heat and the fire size can not be changed.

I don't think that would be a right approach to draw conclusions to your observations. Mass and energy should both get conserved regardless of the magnitudes involved.

Whatever your real test data be, it does not matter. As long as you successfully fed the computer with the right mathematics & physics governing your problem, your residuals MUST converge in all circumstances.

If not, erroreous results will be conveyed. Every CFD user must appreciate the implicit pitfalls such as these while using codes.

siri · November 30, 2010, 06:30

Hana,

I was writing this reply when I read your other post. I submitted a long response, but it didn't get posted unfortunately.
Let me summarize instead brief pointers. I am sure you will appreciate this if you have worked on other CFD codes. As I still suspect the problem lies either in material BCs or equations being solved.

Can you check if all fluids, solids/walls are exactly what they should have been?
What are you using to solve Continuity equation? How you compute the pressure/density - does it comply to the actual physics - it is very important
You may check for the net mass flux thru inlet to outlet. If this is positive, it means mass conservation is violated. Then results mean null even if energy got conserved !
Increasing mass residuals can mean many things.. even justifiable in some cases... does your model include species/reactions/combustion kind of stuff ?

Lastly what did you mean by "There is a door opening (interior) to the next empty cuboidal room, where the walls are pressure outlet"

I wonder, why should you bother to model the adjecent room walls at all ? when you are only "interested especially in temperatures in the room and a flow through the door"

Why should you not just define the outlet for the flow at the location of the door opening?

hanka · December 16, 2010, 04:45

siri, thank you very much for your detailed reply. First, I was thinking and studying the influence of the adjacent room, which you mentioned in your reply. This is also the reason why I did not reply sooner. It sounds maybe weird, but it has an influence on temperatures and mainly on the flow through the door. If I consider no adjacent room, the flow through the door is not correct.
Now, I will finally check your suggestions...

November 23, 2010, 07:55	divergence of scaled residuals (steady simulation)	#1
hanka New Member Hana Matheislova Join Date: Oct 2010 Posts: 7 Rep Power: 15	Hi everybody, do you have any suggestions why scaled residuals (mainly continuity) in steady simulation drop at first, but started to grow after ca. 20th iteration? Smaller mesh cells and second order upwind discretisation quite helped, but the trend is the same. Are these results even relevant? Thanks for any suggestions or ideas

November 30, 2010, 06:30		#6
siri Member Join Date: Oct 2010 Posts: 45 Rep Power: 15	Hana, I was writing this reply when I read your other post. I submitted a long response, but it didn't get posted unfortunately. Let me summarize instead brief pointers. I am sure you will appreciate this if you have worked on other CFD codes. As I still suspect the problem lies either in material BCs or equations being solved. Can you check if all fluids, solids/walls are exactly what they should have been? What are you using to solve Continuity equation? How you compute the pressure/density - does it comply to the actual physics - it is very important You may check for the net mass flux thru inlet to outlet. If this is positive, it means mass conservation is violated. Then results mean null even if energy got conserved ! Increasing mass residuals can mean many things.. even justifiable in some cases... does your model include species/reactions/combustion kind of stuff ? Lastly what did you mean by "There is a door opening (interior) to the next empty cuboidal room, where the walls are pressure outlet" I wonder, why should you bother to model the adjecent room walls at all ? when you are only "interested especially in temperatures in the room and a flow through the door" Why should you not just define the outlet for the flow at the location of the door opening?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Quarter Burner mesh with periosic condition	SamCanuck	FLUENT	2	August 31, 2011 11:34
Steady state simulation of a turbine?	milos	OpenFOAM Running, Solving & CFD	0	November 17, 2009 07:52
divergence in energy residuals	swarley	FLUENT	2	May 22, 2008 08:09
About the difference between steady and unsteady problems	Lisa	Main CFD Forum	11	July 5, 2000 14:37
3-D Contaminant Dispersal Simulation	Apple L S Chan	Main CFD Forum	1	December 23, 1998 10:06

November 24, 2010, 02:06		#2
siri Member Join Date: Oct 2010 Posts: 45 Rep Power: 15	what flow are you solving for ? Your problem might be due to - improper BCs - complicated physics or - inefficient mesh field other causes might exist, but could say the above in general

November 30, 2010, 04:09		#3
hanka New Member Hana Matheislova Join Date: Oct 2010 Posts: 7 Rep Power: 15	Hi Siri, thanks. It is a non-spreading fire in an empty cuboidal room. The fire is defined only by energy source (W/m3), which is generated in a cylinder area. There is a door opening (interior) to the next empty cuboidal room, where the walls are pressure outlet. I am interested especially in temperatures in the room and a flow through the door. The mesh is hexahedral with the same size in the whole domain. If I made the cell size smaller, the residuals got better - increasing of continuity and energy was only very slight (other residuals were nicely decreasing). I am only a beginner in Fluent (I have more experiences only with CFD programmes specialised to compartment fires), so I am sorry, if my questions are quite stupid or irrelevant...But is it possible to use these results or does this behaviour mean that there is a serious problem? Or what do you think about it in general?

November 30, 2010, 05:16		#4
hanka New Member Hana Matheislova Join Date: Oct 2010 Posts: 7 Rep Power: 15	I have probably detected the main reason: the energy source of the fire is too high. The residuals do not increase if this value is much lower. The problem is, that this simulation should correspond to a real fire test, so the released heat and the fire size can not be changed.

December 16, 2010, 04:45		#7
hanka New Member Hana Matheislova Join Date: Oct 2010 Posts: 7 Rep Power: 15	siri, thank you very much for your detailed reply. First, I was thinking and studying the influence of the adjacent room, which you mentioned in your reply. This is also the reason why I did not reply sooner. It sounds maybe weird, but it has an influence on temperatures and mainly on the flow through the door. If I consider no adjacent room, the flow through the door is not correct. Now, I will finally check your suggestions...