Oscillating residuals in steady flow problem

Mohit Singh · February 22, 2016, 07:08

Hello,
I am recently trying to run a transonic flow simulation in cfd++ and have come across this phenomenon of oscillating residuals, even when the problem is specified as steady flow. I'm not able to interpret that and also not able to extract the data as it changes with changing residual. For your reference, htte snapshots of two of the cases have been attached herewith. Does anyone has faced such a problem or do anyone knows how to deal with such a case? Any kind of help would be appreciable.

Thanks in advance,
Mohit

fluid23 · February 23, 2016, 13:38

Have you considered the possibility that your flow isn't steady state? If you analyze an unsteady flow assuming it's steady your convergence can often behave as you have shown.

If you are using an implicit solver, an iterative approach is used to advance the solution from a starting state (initialization) to a final (converged) state. This is true if the solution is one step in a transient problem or a final steady-state result. You can think of the iterations as a pseudo temporal progression that will still allow non-steady state behavior to develop.

Mohit Singh · February 23, 2016, 15:30

Quote:

Originally Posted by MBdonCFD

Have you considered the possibility that your flow isn't steady state? If you analyze an unsteady flow assuming it's steady your convergence can often behave as you have shown.

If you are using an implicit solver, an iterative approach is used to advance the solution from a starting state (initialization) to a final (converged) state. This is true if the solution is one step in a transient problem or a final steady-state result. You can think of the iterations as a pseudo temporal progression that will still allow non-steady state behavior to develop.

Hello Matt Sir,
First of all, thanks for your quick reply. As you mentioned flow may be an unsteady state since I'm simulating flow over payload fairing in transonic regime of M=0.9. But I would also like to mention that I have specified the simulation to be steady state in cfd++ i.e. the flow is supposed to be in steady state.
I would be grateful to you if you could please elaborate the meaning of solution being one step in a transient problem and the term pseudo temporal regression. What would you recommend that could be done in such a case?
Can you describe the significance of Courant-Freidrich-Lewy (CFL) number in a steady flow simulation?

fluid23 · February 23, 2016, 15:39

Just because you tell the solver it's steady doesn't mean it's actually steady. You may need an unsteady solution to truly solve the problem. However, residuals can be a little misleading. You should look at other figures of merit (forces are usually a good indicator) to determine if there are any unsteady effects influencing your residuals or if it is a purely numeric issue.

What I was trying to do with the above statement was illustrate the fact that an implicit steady solution actually uses a time step to advance the solution from it's initialized state (whatever that may be) to it's final converged state. Look up and learn how the solver works and you should understand what that means but suffice it to say that even a steady solution uses a time step to get from beginning to end. Each iteration of your solution can be thought of as a time step even though it's steady state. If your flow is truly steady then once the solution converges, the relative change from one iteration to the next is negligible.

Here is a decent discussion on CFL number for steady flow. Hopefully this helps.

http://www.cfd-online.com/Forums/mai...mulations.html

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Help! Oscillating velocity residual for steady flow	ripperjack	Main CFD Forum	0	December 4, 2012 14:23
Flow Across Tube Banks - Transient vs Steady State	HeatTransferFan	CFX	11	September 28, 2012 15:21
Back flow problem in gas cyclone	lakhi	FLUENT	0	August 31, 2012 05:27
convergence problem in 3D steady state, laminar flow in a bath	vajiheh	FLUENT	0	July 10, 2009 13:18
abt. algorithm for a steady flow problem	Sandy	Main CFD Forum	1	February 21, 2008 08:59

February 23, 2016, 13:38		#2
fluid23 Senior Member Matt Join Date: Aug 2014 Posts: 947 Rep Power: 18	Have you considered the possibility that your flow isn't steady state? If you analyze an unsteady flow assuming it's steady your convergence can often behave as you have shown. If you are using an implicit solver, an iterative approach is used to advance the solution from a starting state (initialization) to a final (converged) state. This is true if the solution is one step in a transient problem or a final steady-state result. You can think of the iterations as a pseudo temporal progression that will still allow non-steady state behavior to develop.

February 23, 2016, 15:39		#4
fluid23 Senior Member Matt Join Date: Aug 2014 Posts: 947 Rep Power: 18	Just because you tell the solver it's steady doesn't mean it's actually steady. You may need an unsteady solution to truly solve the problem. However, residuals can be a little misleading. You should look at other figures of merit (forces are usually a good indicator) to determine if there are any unsteady effects influencing your residuals or if it is a purely numeric issue. What I was trying to do with the above statement was illustrate the fact that an implicit steady solution actually uses a time step to advance the solution from it's initialized state (whatever that may be) to it's final converged state. Look up and learn how the solver works and you should understand what that means but suffice it to say that even a steady solution uses a time step to get from beginning to end. Each iteration of your solution can be thought of as a time step even though it's steady state. If your flow is truly steady then once the solution converges, the relative change from one iteration to the next is negligible. Here is a decent discussion on CFL number for steady flow. Hopefully this helps. http://www.cfd-online.com/Forums/mai...mulations.html