Hi, everyone

I am using steady coupled dpm to calculate my case.

During each dpm iteration, the residual changes from 10**-4 to 10**-6, and during about each 20-30 dpm iterations, the largest residual changes from 10**-4 to 10**-5. So the plot is changing like a wave.

And when I plot the particle trajectory at each lowest residual point, the trajectory is different. So my question is that since the residual is less then 10**-3, why the field is still changing and when I can get the final result?

I really need your help. Thank you in advance!

Best regards

winnie

Hi,

The same is with my case and what i think , since the residue is less than 10**-3 or -4, and no longer with decreasing trend, the solution can be said converged. But the trajectories are different for each DPM iteration and so the DPM results,u wanna present,should be average of few DPM iterations. I am doing the same way and getting good agreement with experimental ones.

Regards,

Umesh

Hi, my friend

Thank you very much. I will try it.

Thanks a lot!

Best regards

winnie

hi, In my case the convergence of 1e-3 does not mean the flow has converged and i believe in driving the solution to 1e-6. Plus there are other options which winni has not mentioned like how many continous phase iterations/DPM you are using? How many stochastic tries are you demanding? how much is length scale factor? All these factors directly affect the convergence of your solution and by keeping first two factor high(~200) and last factor minimum ,i do get convergence of 1e-8 in 10000 iterations (24 hours).Of course,I am paying for my accuracy with time!

Then as winni said I start getting a wave , but my solution does not change much(what i mean is from 1e-8 to 1e-5).

regards kulwinder

Hi Winnie

Long time without talking to you! I was thinking so deeply in your case, and I guess you have two potential cases:

1) temporal accumulation of particles that can generate periodicity of your flow. In experiments, we have been able to see some sort of particles acummulation that can form a clog. This acculation of particles follows a cycle of formation and destruction. This potentially could explain the behavior of your residuals

2) You are working with gas/liquid phase! this type of simmulations can generate an imbalance of the equations that FLUENT uses to solve the flow field.

I sugguest that you get in contact with professor Kumar at the U of A. In my opinion he can give you the most accurate answer to your case.

Other explanation that people has posted on this forum are just especulation.

If you average your data you are ignoring the ramdom component of the DPM model!!! like always think twice before carrying with some suggestions.

Your trully friend

Alex Munoz

Hi, Alex

So glad to hear the voice from you! You can always give me some valuable suggestions. All that I can say is thank you.

These days, I have tried different methods to make my cases converge. I have tried using different force coefficients¡¢different particle injection types, and as a result, the cases sometimes converge with continuous equation and two momentum equations' residuals level out and another momentum equation's residual changing from 10**-6 to 10**-7, I think it is enough and the results also stop changing. But I still don't know why some converge and why others don't. I need to think it deeply to find some useful experience.

Thank you, Alex. By the way, can you tell me the full name of 'U of A'. Sorry for my ignorance.

Best regards

winnie

Hi, kulwinder

Thank you for your reply!

I agree with you since in my case when the residual fall to between 10**-5 and 10**-7, the field is still changing a little and after fall to between 10**-6 and 10**-7, the field stop changing.

I choose 20 continous phase iterations/DPM, and I think more iterations per dpm iteration has little effects on the convergence since in my experience, when I use 30-40 iterations per dpm iteration, the only effect is that during each dpm, continuous phase can get better convergence and the lowest residual becomes lower, but the highest residual has little change.

As for length scale or length scale factor, in my experience, a smaller value for the length scale or a larger value for the length scale factor decease the discrete phase integration time step which improve the accurate of the result but make the calculation less converge.

My case is a laminar flow so I don't use the stochastic try.

It seems that my experience is a little different from yours. Maybe I should be more patient to get more steps.

I will try. Thank you!

Best regards

winnie

Hi Winnie

Could you please send a e-mail explaining your case to Dr Nandakumar, this is his e-mail address.

kumar.nandakumar@ualberta.ca

He has an amazing knowledge in fluid mechanics and the implementation of turbulence models in CFD codes.

I hope he will be able to give you a satisfactory unswer.

The following paragraphs are a summary of Dr Nandakumar works.

Krishnaswamy Nandakumar, Ph.D. (Princeton), P. Eng. Professor 504C Chemical and Materials Engineering Building (780) 492-5810 Fax: (780) 492-2881 kumar.Nandakumar@Ualberta.ca

Appointments for current month

Research Areas: Fluid Mechanics; multiphase transport processes

My research interests center around computational aspects of fluid mechanics and transport phenomena. The series of structured transitions exhibited in a class of hydrodynamic stability problems are being investigated using recent algorithms from bifurcation theory and are also being verified in experiments using Laser Doppler anemometer.

In collaboration with Prof. Minev, we are developing next generation algorithms for direct simulation of multiphase flows with many particles using fictitious domain methods and deformable bubbles and droplets.

In another area, we are learning about issues related to modeling multiphase flows in industrially important flows. These include modeling the tray hydraulics of a distillation column, flow in packed towers, design of structured packings, erosion/corrosion in pipelines, SOFC fuel cells, sedimentation, bubble column dynamics, polymeric flows in twin-screw extruders etc.

Sample Publications

F. H. Yin, M. Song, A. Afacan, K. Nandakumar and K. T. Chuang, CFD Modeling of Mass Transfer in Randomly Packed Distillation Columns, Ind Eng Chem Res. 395) (2000) pp1369-1380

Nandakumar, K., Y. Shu and K .T. Chuang Predicting geometrical properties of randomly packed beds from computer simulation, AIChEJ, 45 (1999) pp2286-2297.

Selmi, M. and K. Nandakumar, Bifurcation study of flow through rotating curved ducts, Phys. Fluids, 11 {1999} pp2030-2043.

Minev, P. D., U. Lange and K. Nandakumar, A comparative study of two-fluid models relevant to bubble column dynamics, J. Fluid Mech., 394 {1999} pp.73-96.

Lange, U., K. Nandakumar and H. Raszillier, Symbolic computation as a tool for high-order long-wave stability analysis of thin film flows with coupled transport processes, J. Comp. Phys. 150 {1999} pp.1-16.

Lakshminarayanan, S., Sirish L. Shah and K. Nandakumar, Modeling and control of multivariable processes: dynamic PLS approach, AIChEJ. 43 (1997) pp2307-2322.

Alleborn, N., K. Nandakumar, H. Raszillier and F. Durst, Further contributions on the two-dimensional flow in a sudden expansion, Journal of Fluid Mechanics 330 (1997) pp169–188.

Basu, S., K. Nandakumar and J. H. Masliyah, A study of oil displacement on model surfaces, J. Colloid and Interface Science 182 (1996) pp82-94.

Best regards

Alex Munoz

Hi, Alex

Thank you very much for your kind help.

Best regards

winnie