Number of Navier-Stokes Iterations

Numerical methods and mathematical models of Elmer
Post Reply
MT_Keg
Posts: 9
Joined: 02 Jun 2013, 19:36
Antispam: Yes
Location: Tennessee

Number of Navier-Stokes Iterations

Post by MT_Keg »

I am running into an issue where the IterSolve indicates

Code: Select all

NUMERICAL WARNING:: IterSolve: Too many iterations was needed.

at

Code: Select all

FlowSolve:  NAVIER-STOKES ITERATION          33

Looking at the solver settings I am not sure what is limiting the NS Iterations to 33. I am running this simulation in Parallel using 4 processes.

Question 1: What setting needs to change to get more Navier-Stokes Iterations (i am not interested in changing the convergence tolerance, preconditions, etc.)?

Question 2: Looking at the solver log it appears that there are 4 different iterations happening (we will look at the last instance). I believe the steady state iteration is 1 because we are not solving a coupled system but I don't necessarily understand what is going on in items 2, 3, and 4.
  • 1. A Steady State Iteration = 1
    2. An overall Navier-Stokes Iteration = 33
    3. A compute change: NS (ITER 65)
    4. A compute change: NS (ITER 66)

Code: Select all

FlowSolve: 
FlowSolve: 
FlowSolve: -------------------------------------
FlowSolve:  NAVIER-STOKES ITERATION          33
FlowSolve: -------------------------------------
FlowSolve: 
FlowSolve: Starting Assembly...
FlowSolve: Assembly:
: ..........Assembly done
DefUtils::DefaultDirichletBCs: Setting Dirichlet boundary conditions
DefUtils::DefaultDirichletBCs: Dirichlet boundary conditions set
FlowSolve: Dirichlet conditions done
CRS_IncompleteLU: ILU(3) (Real), Starting Factorization:
CRS_IncompleteLU: ILU(3) (Real), NOF nonzeros:  67795903
CRS_IncompleteLU: ILU(3) (Real), filling (%) :       281
CRS_IncompleteLU: ILU(3) (Real), Factorization ready at (s):     1.69
IterSolver: Calling real valued iterative solver
       1 0.5071E-07
       2 0.2545E-07
	    .
	    .
	    .
     500 0.4331E-09
     501 0.4331E-09
NUMERICAL WARNING:: IterSolve: Too many iterations was needed.
ComputeChange: NS (ITER=65) (NRM,RELC): ( 0.13897368     0.17096309E-06 ) :: navier-stokes
FlowSolve: iter:   33 Assembly: (s)   10.73  340.00
FlowSolve: iter:   33 Solve:    (s)  241.31 8074.57
ComputeChange: NS (ITER=66) (NRM,RELC): ( 0.13897367     0.85481509E-07 ) :: navier-stokes
FlowSolve:  Result Norm     :   0.13897367194452842
FlowSolve:  Relative Change :    8.5481508828119667E-008
ComputeChange: SS (ITER=1) (NRM,RELC): ( 0.13897367      2.0000000     ) :: navier-stokes
Any help will be greatly appreciated. I have attached the *.sif and solver log files.

Thanks,

MT
Attachments
case.sif
(2.08 KiB) Downloaded 305 times
Solver_Log.txt
(364.17 KiB) Downloaded 286 times
mika
Posts: 230
Joined: 15 Sep 2009, 07:44

Re: Number of Navier-Stokes Iterations

Post by mika »

Hi,

Unfortunately it seems that if relaxation is applied (the value of the keyword Nonlinear System Relaxation Factor differs from the unity), the value of the nonlinear iteration counter is hassled and not intuitive. I found that one would get more sensible values for the iteration counters if relaxation were not used. This should be fixed in the code. Anyhow, it seems that the nonlinear stopping condition is met at NAVIER-STOKES ITERATION 33 which stops the solution procedure.

That the error in the linear system does not meet the stopping criterion at the time when the maximum number of linear system iterations is reached causes the warnings (there are many of them) "NUMERICAL WARNING:: IterSolve: Too many iterations was needed." To get rid of these one could set a larger value for the keyword Linear System Max Iterations or try to use a different iteration method (try for example with Linear System Iterative Method = BiCGStabl and note that the last letter is now "l").

Best regards,
Mika
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Number of Navier-Stokes Iterations

Post by raback »

Hi,

Yes, there seems to have been some missteps in the development that resulted to performing some of the steps twice. This is now fixed in "devel" branch. Thanx for reporting!

-Peter
MT_Keg
Posts: 9
Joined: 02 Jun 2013, 19:36
Antispam: Yes
Location: Tennessee

Re: Number of Navier-Stokes Iterations

Post by MT_Keg »

No problem. Thanks for looking into it and providing a fix! Keep up the good work!

MT
Post Reply