
[Sponsors] 
September 28, 2005, 05:37 
convergence in 3d

#1 
Guest
Posts: n/a

in phoenics i am able to solve 2d problems with ease but for 3d problems i am not getting convergence. even laminar flows too ! whether any special care is required to get convergence in 3d problems in setting up parameters like endit, liter, dtfals etc ? Thanks in advance !


September 28, 2005, 11:41 
Re: convergence in 3d

#2 
Guest
Posts: n/a

Generally of course 3d convergence is alot more work then 2d.
Experimentation with relaxation settings can help and experience helps, it can be a bit of a black art, but thats numerics for you. With v3.6 you might try the automatic convergence which will generally be on by default. (MenuNumericsRelaxationAutomatic convergence control). If this does not help switch it off and try sensible dtfals values for the velocities, I usually gauge mine by (smallest cell dist in main direction of flow)/(fastest velocity) and then loosen them off some, i.e. make them a bit bigger often my first guess is 1 order of magnitude higher. If PARSOL is on and the geometry is not so simple, I might put some linear relaxation on P1 (0.75 to 0.5), but as a general rule you should never go < 0.3. Tem1 I usually use falsdt and experiment with the value, 99 times out of 100 for me its always orders of magnitude greater (looser) then the value I have used for velocity. However linear relaxation with TEM1 can also work well and if the flow is airidealgaslaw then I may relax den1 as well. Also if using Immersol I go for linrlx and use the same value for TEM1 as T3, I generally start with values of (0.5). Hope this helps, but experience and experiment is the best aid. But also convergence is a relative thing I look for spot values to flatten and residuals to drop at least a couple orders of magnitude, mass balance from the result file, and normalised residuals for variables to be small. For NORMALISED RESIDUALS see : http://www.simuserve.com/support/converg/cona3.htm or pasted below: Question:  How do I convert the residuals obtained with SELREF=T to residuals normalised by reference inflow quantities? Answer: This question is best answered by means of an example, which in this case is taken to be 2d singlephase flow in the yz plane with solution for mass continuity (P1), momentum (V1 and W1), enthalpy (H1) and mixture fraction (MIXF). Note that in this example the units of 'resref' and 'res sum' for P1 are kg/s rather than m^3/s because DENPCO=T in the Q1 input file. The conversion procedure is as follows: Near the bottom of the RESULT file you will find the following normalised residuals printout for the last sweep: Wholefield residuals before solution with resref values determined by EARTH & resfac= 1.000E03 variable resref (res sum)/resref P1 3.718E09 1.562E+00 V1 7.078E11 1.819E+02 W1 1.880E08 6.684E01 H1 8.555E03 4.698E01 MIXF 1.317E10 5.908E01 In addition, the following nett source printout is given for all boundary conditions and source terms, including the reference inflow rates of each dependent variable: Nett source of W1 at patch named: INLETF = 1.202E06 Nett source of W1 at patch named: INLETA = 2.238E04 Nett source of W1 at patch named: OUTLET =5.453E04 Nett source of W1 at patch named: WFNN =2.016E04 nett sum=5.219E04 pos. sum= 2.250E04 neg. sum=7.470E04 Nett source of R1 at patch named: INLETF = 2.003E06 Nett source of R1 at patch named: INLETA = 3.799E05 Nett source of R1 at patch named: OUTLET =3.999E05 nett sum=3.638E12 pos. sum= 3.999E05 neg. sum=3.999E05 Nett source of H1 at patch named: INLETF = 9.902E+01 Nett source of H1 at patch named: INLETA = 4.615E+01 Nett source of H1 at patch named: OUTLET =1.452E+02 nett sum= 2.136E04 pos. sum= 1.452E+02 neg. sum=1.452E+02 Nett source of MIXF at patch named: INLETF = 2.003E06 Nett source of MIXF at patch named: OUTLET =2.003E06 nett sum=6.821E13 pos. sum= 2.003E06 neg. sum=2.003E06 The actual residuals can be recovered by multiplying the (resref) column by the (res sum)/resref column, as follows: variable resref (res sum)/resref res sum P1 3.718E09 1.562E+00 5.8075E09 kg/s V1 7.078E11 1.819E+02 1.2875E08 N W1 1.880E08 6.684E01 1.2566E08 N H1 8.555E03 4.698E01 4.0191e03 W MIXF 1.317E10 5.908E01 7.7808E11 kg/s These 'res sum' values can now be normalised by dividing them by the appropriate reference inflow rates (taken from the Nett source printout), as follows: variable ref value res sum %error P1 3.999E05 5.8075E09 kg/s 0.0145 V1 2.250E04 1.2875E08 N 0.0057 W1 2.250E04 1.2566E08 N 0.0056 H1 1.452E+02 4.0191e03 W 0.0028 MIXF 2.003E06 7.7808E11 kg/s 0.0039 

October 5, 2005, 04:01 
Re: convergence in 3d

#3 
Guest
Posts: n/a

hi mike regarding convergence in phoenics !
do u feel that it is a good idea to keep the conwiz ON and put some further restrictions through MAXINC ? if at all DTFALS approach is so good, why CONWIZ uses linear relaxation for velocities ? 

October 12, 2005, 09:22 
Re: convergence in 3d

#4 
Guest
Posts: n/a

>>do u feel that it is a good idea to keep the conwiz ON >>and put some further restrictions through MAXINC
Yes this is a very good idea try this with restricting MAXINC I am not sure why CONWIZ uses linear relaxation for velocities as I am presently not sure what it doe's deep inside the code, this may be a roose. 

Thread Tools  
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
URF and convergence in natural convergence  MarieAnne  Main CFD Forum  11  September 11, 2009 10:07 
convergence  vijay  FLUENT  6  February 1, 2006 04:04 
convergence  vijay  Main CFD Forum  1  January 30, 2006 14:13 
too bad convergence  Davoche  Main CFD Forum  2  November 20, 2005 06:08 
About convergence  LQ  CFX  3  June 2, 2005 23:43 