Hi All Have anyone use Flow-3D for steady state confined flow before? (even though Flow-3D use unsteady RNS and not a real stead state solver). I found out simulation time is extremely longer ~30hrs to reach steady state. so any experience to share or input on this issue?

Rgds Dan

hello, did you tried by selecting an implicit method? (for example the "implicit advection") it may run faster, reaching the same steady state condition. (however check it)

Stefano i did try all the implicit method available for this turbulent flow + confined flow problem. But, it still took ~20-30hrs to reach steady state. Btw, did you model this kind of problem before? any advice?

Rgds Dan

yes, we did some works on confined flows in the past. If your problem is 3D then maybe 20-30hours are not so strange, or better, your setup (basically the mesh that you draw) don't allow you to go faster. Maybe I can help you a little bit in understanding the speed limit of your simulation, but you'll have to help me in this: open the "solver summary" file, periodically you have a block of lines starting with something like "explicit x-advection stability limits =....". Find the most recurrent block of lines and paste me it here. (if you have the solver summary file coming from a simulation with explicit methods it would be better!)

Hi Stefano sorry i'm out of office for few weeks. anyway, i noticed flowsicene did provided some tips for steady state confined flow problem in recently Fall 2008 news. anyway, it help in reduced iteration for pressure but doesn't help in cycles count as below. so, do you have any better recommendation?

t cycle iter dtstbl/code delt epsi cpu clock

--- ----- ---- ----------- ---- ---- --- ----- 0.00E+00 0 0 1.94E-04/cz 1.00E-05 -1.0E+00 2.4E+01 21:07:59 5.74E-02 364 1 8.06E-05/vs 8.06E-05 1.1E+06 2.7E+03 21:22:05 2.00E-01 1695 1 1.24E-04/vs 1.24E-04 7.2E+05 1.3E+04 22:13:26 4.00E-01 3379 1 1.15E-04/vs 1.15E-04 7.5E+05 2.5E+04 23:18:42 6.00E-01 5163 1 1.09E-04/vs 1.09E-04 7.8E+05 3.8E+04 00:27:55

hello Dan, from the lines showed it seems that your simulation speed is limited by the viscous calculation (/vs code, near the time-step). So you can turn explicit all the physical models except for the viscous one. If you use an implicit method for viscous calculation it should run faster.

If you run the simulation again with this change, you should not see anymore in your solver message the /vs code, but other /.. codes, and the use of a larger time-step.

Let me know if it works,

Hi stefano I don't have a choise of switching to implicit viscous solver as it is not allow for turbulent flow problem. Thus, it become a limiting factor for my steady state confined flow problem. So, do you have any good inputs on this problem?

Dan

On the last FlowScience newsletter there are suggestions on how to reach faster a steady state condition! Following their suggestions in the numerical panel you can speed up your simulation up to 10x. I think you can try it.

And if your fluid flow is turbulent then try to set a TLEN value. As far as I know, sometimes, if you don't set it, the viscosity get excessive values, limiting you excessively the simulation speed with the code /vs.

Hi Stefano I did all the recommended sovler setting (SRO solver and relaxation setting) and also key in TLEN value. But unfortunately it still took 30-40hrs for steady state solution. So, I tried with others CFD codes & suprisingly it took only 3-4hrs to reach steady state for the same model! So guess Flow-3D as generic CFD code may optimize for steady state problem.

yes, this is true, or better: flow-3d solver is optimized for transient solutions. If you wanted to solve a transient simulation probably you may get the benchmark at the contrary: 3-4hours with FLOW-3D, and 40hours with others cfd. But I'm sure FlowScience is working also for a dedicated steady state solver...

I totally agreed that Flow-3D has poor steady state solver but not sure about the tarnsient VOF solver. Pls share with me if you have others benchmarck did before on this. Thanks

hello Dan, I'm glad to write you on that. About the stead-state case, FLOW-3D solver is "time-marching", so you have to reach a steady state condition as in the real world: marching trough the time, starting from an initial condition, untill the flow becomes steady. That's why it may takes more time of a direct steady state solver, but also why I said that for transient solutions is more suitable. (if you simulate 30s for example to reach the steady state solution, the same time almost would have been to make a 30s transient simulation!) About the transient VOF solver: yes, we have done many tests on this. You can have a look at our website (www.xceng.com), in the homepage there's a flash animation of a work that we recently did on a boat that made a World Record. You'll see a very complex simulation with a intensive use of free surface tracking, moving objects, etc... and the time to simulate this was almost like the time you spent to find the steady state solution of your confined flow problem!

Thanks for information sharing but believe Flow-3D may not a suitable option for steady state confined flow problem as it just take too long to reach convergence as compared to others CFD codes. Anyway, the FAVOR option is good for meshing but limited by aspect ratio and multiblock linking error.

Hello Dear All friend I want to use Flow-3D 8.2 for simulation of fluid flow in the submerged nozzle but the solution don't reach to steady state and fluctuate. but if we run this problem with other code same as Fluent we get steady state and correct configuration of fluid flow, where is the problem?

if you want i send you the picture of my simulation... Best Regards

Shahin I believe this is limitation of F3D in simulate steady steady sate confine flow. some time i found out it is not convergence in term of mass flow rate even though the kinetic energy has reach steady trend. So, Fluent should be the right option, no doubt. I had switched to Fluent for my problem and it run 10x faster!

Dan

In pic1( http://shahin1359.tripod.com/pic1.jpg ) you see that after 12.5 second we get good result but after 28.5 sec not only we don't get steady state, we get the wrong result(pic2 http://shahin1359.tripod.com/pic2.jpg)

I use the RNG turbulent model , first order momentum advection, explicit viscose stress omega=1, imp=2, and so on with other variation in the preprocessor else we don't get correct result please help me in this problem!!!!

Shahin i can't open your picture, may be you can try k-e turbulent model. actually i already give up steady state confined flow problem with F3D. you may want to consider to switch to others CFD.

Hi Dear Dan, for open picture copy the address and paste on the browser Address Bar. the k-e model is not correct too I have to do my project with Flow-3D if any one said me this before maybe i did not choose Flow-3D Now do anyone have a suggestion for me?

Hello Dan i think you didn't see pictures please Download this link http://www.mediafire.com/?emyddnzlody

Hi Shahin Sorry i'm not able to help as already give up with Flow-3D and now using Fluent. Structural/rectangular grid/mesh use by Flow-3D some how very easy to mesh geometry however got a lot of limitation as well. Btw, it is not too bad to generate mesh with body fitted mesh (BF) in Fluent. It is more powerful and speed no doubt is at least minimum ~5x faster than Flow-3D in this kind of simulation.