Convergence in Nat conv problem - simple geometry

gradstudent2430 · April 11, 2012, 18:23

I am simulating natural (Rayleigh Benard convection) in a rectangular 2D cavity, heated at the bottom, colder at the top (~5K delta T), with a Rayleigh number of ~10e6. I cannot get convergence with the steady state solver for this Rayleigh number. I know that flow is laminar at this Ra from several books.

As suggested in earlier discussions I did a stepwise solve, by increasing g from 0.098 to 0.98 to 9.8 (to modify Rayleigh number) and using each solution as the initialization for the next. Things converge upto Ra~10e5. After that the residuals diverge/oscillate - the velocity and contunuity specially, with the net heat flux actually being lower for same delta T at higher rayleigh number (less convection)!!

Relaxation factors:
Momentum - 0.15
Energy - 0.5

Boussinesq model, Presto for pressure, SIMPLE scheme. I entered in starting density and thermal exp coeff. in the material model

MESH:
Regular mesh with the rectangular area (25mm W, 15mm H) split into 40x100 cells.

What else can I do? I tried starting with a transient solution with a time step of 0.0006 s and 1000 iterations and going on to steady stater from there but same divergence issue. It looks like this is a very simple problem for fluent, why is it tricky?

Thanks

foamcfd · April 19, 2012, 06:23

Hi Indu,

Generally this type of problems are very tough to converge with steady state solver. You said that you also tried with transient solver. How did you calculate the time step?

What I can suggest you to use linear variation of density instead of boussinisque approximation. This is something strange as later is very simple to converge but give a try.

Thanks

April 11, 2012, 18:23	Convergence in Nat conv problem - simple geometry	#1
gradstudent2430 New Member Indu Join Date: Apr 2012 Posts: 1 Rep Power: 0	I am simulating natural (Rayleigh Benard convection) in a rectangular 2D cavity, heated at the bottom, colder at the top (~5K delta T), with a Rayleigh number of ~10e6. I cannot get convergence with the steady state solver for this Rayleigh number. I know that flow is laminar at this Ra from several books. As suggested in earlier discussions I did a stepwise solve, by increasing g from 0.098 to 0.98 to 9.8 (to modify Rayleigh number) and using each solution as the initialization for the next. Things converge upto Ra~10e5. After that the residuals diverge/oscillate - the velocity and contunuity specially, with the net heat flux actually being lower for same delta T at higher rayleigh number (less convection)!! Relaxation factors: Momentum - 0.15 Energy - 0.5 Boussinesq model, Presto for pressure, SIMPLE scheme. I entered in starting density and thermal exp coeff. in the material model MESH: Regular mesh with the rectangular area (25mm W, 15mm H) split into 40x100 cells. What else can I do? I tried starting with a transient solution with a time step of 0.0006 s and 1000 iterations and going on to steady stater from there but same divergence issue. It looks like this is a very simple problem for fluent, why is it tricky? Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
convergence problem in the turbomachinery problem.	u k jha	CFX	1	September 7, 2010 19:41
Problem in Initializing Blocking for 2d geometry	nitin Dewangan	CFX	6	October 24, 2008 02:03
CONVERGENCE PROBLEM - oil boiler	MM	FLUENT	1	February 15, 2007 06:24
Cooling simple 3D geometry problem	Luk	FLUENT	4	May 12, 2006 10:47
convergence problem with SIMPLER	NURAY KAYAKOL	Main CFD Forum	1	February 24, 1999 14:43

April 19, 2012, 06:23		#2
foamcfd New Member Yogesh Patil Join Date: Nov 2009 Posts: 20 Rep Power: 17	Hi Indu, Generally this type of problems are very tough to converge with steady state solver. You said that you also tried with transient solver. How did you calculate the time step? What I can suggest you to use linear variation of density instead of boussinisque approximation. This is something strange as later is very simple to converge but give a try. Thanks