CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   SOLUTION DOES NOT CONVERGE (http://www.cfd-online.com/Forums/fluent/48742-solution-does-not-converge.html)

Jibran Haider July 11, 2008 03:25

SOLUTION DOES NOT CONVERGE
 
can anyone tell me what should i do if my solution doesnot converge or it diverges.

joao July 11, 2008 03:28

Re: SOLUTION DOES NOT CONVERGE
 
Hi Jibran,

It is difficult to say what to do without any information about your particular simulation.

Jibran Haider July 11, 2008 03:37

Re: SOLUTION DOES NOT CONVERGE
 
I am trying to run a simulation of steady, laminar flow of air around a circular cylinder in cross flow placed in a duct. For this purpose I have created a 2 dimensional mesh in GAMBIT 2.2.30 and solved it in FLUENT 6.3.26. when i iterate for lower velocities such as less than 0.1 m/s the solution converges but it doesnot converge for higher velocities.

joao July 11, 2008 04:45

Re: SOLUTION DOES NOT CONVERGE
 
What is the behaviour of the residuals? Is there a divergence or is it that they simply won,t come down to the level you've prescribed?

umesh javiya July 11, 2008 12:06

Re: SOLUTION DOES NOT CONVERGE
 
I dont know...but....may be ur flow becomes turbulence at higher velocity.... so switch on turbulence model.....

Jibran Haider July 12, 2008 00:47

Re: SOLUTION DOES NOT CONVERGE
 
occasionally they diverge but most of the time they donot come down to the required criteria of 10^-3.

Jibran Haider July 12, 2008 00:52

Re: SOLUTION DOES NOT CONVERGE
 
Umesh the Reynold's Number i am checking for is around 200 so i dont think i need to turn on the turbulence model as the critical reynold's number for crossflow over a cylinder is 5*10^5.

Osman Turan July 13, 2008 03:03

Re: SOLUTION DOES NOT CONVERGE
 
I think the critical reynold's number for crossflow over a cylinder is not 5*10^5. This value is critical for surface such as flat plat in external flow. For flow around an obstacle, the critical Re > = 20000.


umesh javiya July 13, 2008 04:58

Re: SOLUTION DOES NOT CONVERGE
 
hi jibran....

Yes Osman is right...re=1e5 is criteria for flate plate....and for flow around obstacles have transition reynolds number is very low....I really dont know but you could check in literatures....it may happen even below 200

Micael July 13, 2008 12:12

Re: SOLUTION DOES NOT CONVERGE
 
I agree, transition may be well below Re=200 for external flow.

Hossam July 14, 2008 02:50

Re: SOLUTION DOES NOT CONVERGE
 
i believe for flow over cylinder at Re=200 the flow is still laminar, transition to turbulence on the cylinder wall doesn't occur untill Re=1000, at Re =200 the well known Von Karman vortex street will occur, and the wake will propagate untill far downstream the cylinder (see aerodynamics for enginnering students chapter 7 for example)this flow is an unsteady laminar flow. you may enhance convergence by extending the flow domain downstream the cylinder,try the unsteady solver or if you want to keep the steady olver try the coupled solver instead of the segragated, use velocity inlet and outflow boundary conditions for the inlet and outlet respectively. also as this flow is dominated by vortices so you have to ensure that your mesh is fine enough for example i did a similar computation and a grid independent solution was reached with 250000 volume cells for 2d computations

Isaac July 14, 2008 12:20

Re: SOLUTION DOES NOT CONVERGE
 
Hossam is right. The flow inherently becomes unsteady due to vortex shedding. If your residuals do not diverge but show oscillating behavior, then even if they go below your set critereon, you cannot consider the solution to be converged. Try unsteady and if you are using outflow condition, make sure that the normal flux across that boundary is zero as is required by the BC. You might have to extend your flow domain beyond what you currently have.

David July 16, 2008 00:48

Re: SOLUTION DOES NOT CONVERGE
 
I just met a similar problem and I have solved it by switching my laminar model to the SKE turbulence model.

Jibran Haider July 19, 2008 03:52

Re: SOLUTION DOES NOT CONVERGE
 
Hey guys! Take a look at this paragraph taken from Heat Transfer by Cengel (2nd edition - chapter 7) "The characteristic length for a circular cylinder or sphere is taken to be the external diameter D. Thus, the Reynolds number is defined as Re = VD/υ where V is the uniform velocity of the fluid as it approaches the cylinder or sphere. The critical Reynolds number for flow across a circular cylinder or sphere is about Recr ≈ 2x10^5. That is, the boundary layer remains laminar for about Re ≤ 2x10^5 and becomes turbulent for Re ≥ 2x10^5."


Masood Kamal July 28, 2008 13:20

Re: SOLUTION DOES NOT CONVERGE
 
Hi Jibran

U scaled ur model according to the scale u amde the model in. FLUENT by default take meters as units for length.

Dale July 30, 2008 11:23

Re: SOLUTION DOES NOT CONVERGE
 
Flow is genrally turbulent for Re over 1e3 for a cylinder. What Cengal is discussing is the boundary layer, which will remain laminar until the Re number increaces over 2e5. Boundary layer and general flow regime are different concepts. If Re is less than 150 then you should expect laminar vortex street, over 300 and turbulent. In between is transistional. Over 2e5 becomes a turbulent boundary layer with a characaristic low drag coeficient.


All times are GMT -4. The time now is 18:11.