[hello-fluenttt]

Thanks Sasquatch for your reply...
the typical order of shock wave is of the order of 100 nm , thats why in order to capture a shock wave as a discontinuity, i thought that the grid cells size should be small enough so as to realize the gradients in flow. for instance for coarse mesh and bigger time domain , the shock wave looks thick(see the attached file) .
my next doubt is,.... does this convergence also depend upon systems i.e, can some system give convergence while others dont, in that case how should we decide ...

[hello-fluenttt]

Thanks Sasquatch for your reply...
the typical order of shock wave is of the order of 100 nm , thats why in order to capture a shock wave as a discontinuity, i thought that the grid cells size should be small enough so as to realize the gradients in flow. for instance for coarse mesh and bigger time domain , the shock wave looks thick(see the attached file) .
my next doubt is,.... does this convergence also depend upon systems i.e, can some system give convergence while others dont, in that case how should we decide ...

[hello-fluenttt]

Thanks Sasquatch for your reply...
the typical order of shock wave is of the order of 100 nm , thats why in order to capture a shock wave as a discontinuity, i thought that the grid cells size should be small enough so as to realize the gradients in flow. for instance for coarse mesh and bigger time domain , the shock wave looks thick.
my next doubt is,.... does this convergence also depend upon systems i.e, can some system give convergence while others dont, in that case how should we decide ...

[ghorrocks]

I discuss this in detail in my PhD thesis: https://opus.lib.uts.edu.au/handle/2100/248

In chapter 5, shock tube modelling.

[hello-fluenttt]

thank u ghorrocks for your reply

i understood from your thesis that
fluent can't capture shock very sharply & accurately...

what commercial software should b used for capturing exact shocks to study complex shock interaction problem

thanks again

[ghorrocks]

The models in my thesis were done in CFX 4.3. But you will get the same results in Fluent, the current version of CFX, CFD-Ace, Star-CD and pretty much all the other codes as well. They will all blur the shock over a few mesh elements and sometimes have little wiggles at the top and bottom of the shock.

Whether this can be called "accurate and sharp" depends on your application. It is plenty good enough for most applications I have seen from race engine manifolds to re-entry vehicles. You must have a very demanding application where it is unacceptable.

Can you describe what you are trying to model?

I can recommend CFX for shock interaction models. The density based solver in fluent could be considered as well. I don't know of any commercial codes which do exact modelling of shock wave flows. I have used the method of characteristics to get exact answers for shock wave simulations with sharply resolved shocks and no wiggles, but that approach gets tricky when you want to include viscosity and other physics (not to mention you have to write this solver yourself ).

[hello-fluenttt]

the problem is a shock tube problem with state 1 and state 2 with diaphragm rupture, generating a moving shock wave. now this shock wave is thick at t>0 like in your thesis where the slope(p vs x) was not very steep. this results in the shock wave not appearing as a discontinuity , this is making a problem in validation with the work of an earlier publication where shock wave is shown as a line clearly visible where they used some GRP scheme. it's important for me to get the shock appear in my work to carry it further.
i am using 2d inviscid flow with density based implicit solver, i have used different combinations of other things like courant number but couldn't get...
i realised that making a very very fine mesh might solve my problem but as u said it won't. the thin the shock line appearing or steep increase in pressure means strong validation and thus more accurate.
what should i do changes to solve my purpose or any other help would be appreciated..
thanks for your time Glenn

[ghorrocks]

If you are careful you can get the shock wave thickness down to about 4 element lengths. So as you refine the mesh the shock wave gets steeper as it just stays at 4 element lengths long. I don't know why you call that "not very steep" as it is it steep enough for almost all shock wave work in history.

You can still compare this to analytical results. Just take the values a few elements after the shock has passed and it has settled. In my experience CFX can get these pressures, temperatures and densities correct to within 1%.

Note that this smearing of the shock is inherent with all mesh based simulation methods I am aware of. The only way I know to model the shock as a sharp discontinuity is using Method of Characteristics and a meshless method. But you will have to code this yourself, I do not know of any available software which uses this technique. My undergraduate thesis used this method for shock wave modelling but I have not published that so cannot share it easily.

Can you explain why smearing the shock over 4 elements is unacceptable? You have to remember that all CFD methods are numerical approximations and no method will be 100% perfect. If you are looking for perfection then just draw what you want to see using photoshop.

[hello-fluenttt]

Thanks glenn i got my problem solved..
but i have another issue
now i got my ansys uninstalled and re installed (student version)
but now when i run fluent , it shows error as
Error: writing compressed file "FFF.1-1-00000.cas.gz".

i am new to ansys ,

could you please help...
thanks

[ghorrocks]

This is the CFX forum. Try the fluent forum.

[sidaero]

Hi this siddharth. My problem is regrading moist air flow condensation on airfoil in transonic flow. Velocity=258 m/s (M=0.8) Temperature=259 and Pressure=65600 Pa
I am working Komega model with a Y+=2
The continuity residual does not converge when i perform the analysis in pressure based solver- coupled solver.. I am getting divergence and sometimes the continuity residual gets stuck near some place and does not converge.. My mesh aspect ratio is 740 and the transition is also smooth. But still the solution is not converging..
I even tried by putting first order scheme for few iterations and second order later and it also did not work..
can anyone give some suggestions about to how to proceed with this???
Thanks in advance..

[ghorrocks]

This sounds like a fluent question. This is the CFX forum. Try the fluent forum.

[Gerhard]

Originally Posted by TB
;72209

Hi,

Thank you for your comments. I've already extrapolated the coarse mesh solution onto the fine mesh definition file before I start the solver. The max residual will drop rapidly in the first 100 iterations but stay at about same level (with small oscilation, sort of like a noise) thereafter (I run totally 3000 iterations and there's no sign of convergence and residuals just stay almost the same).

This situation just happen when I start refining the mesh close to the wall (O mesh is used). I found out that I can only reduce the first node distance to a certain value. If I define a smaller value than this, I will get the convergence problem as stated in the first post regardless of whether I'm using omega- or epsilon-based model. Surprisingly, standard ke model doesn't seem to have problems reaching the target max residual (5e-6).

I believe this is not a new problem for simulation. Please kindly give me some more feedback if possible.

Do you think increasing turbulent dissipation artificially will help the convergence here? If so, what method will you use to do this?

Hi
I had a similar problem when trying to solve a steady state, axisymmetric simulation of the ERCOFTAC conical diffuser.
In my case, the turbulence residuals dropped rapidly and then remained roughly constant for the rest of the simulation. The rest of the equations then do not converge.



That is because after the residuals of one of the equations drop so rapidly, that equation is not solved any longer for the rest of the simulation (track that variable in paraFoam and see if that is your problem). That is only if "tolerance" in "fvSolution" is too high (for me it was 1e-6). I changed "tolerance" to 1e-16 for all my equations and my problem was solved.


It sounds simple, but it worked for me.

[ghorrocks]

Sorry, Gerhard; your comment is not correct. It might be correct in OpenFOAM but this is the CFX forum and it is not correct in CFX.

CFX does not stop solving an equation when it reaches a criteria, but continue solving the others. CFX continues solving all equations until the criteria for convergence is reached then the entire run is declared converged.

[Gerhard]

Dear Glenn
Thanks for the correction.
I was not sure how to know what kind of forum it is, but I figured it out quite quickly after your reply
Perhaps an OpenFOAM user stumbles over my post and finds it helpful. Who knows.
Kind regards.

[ghorrocks]

Yes, if this page came up in a search it can be difficult. The way to tell is the forum tree on the top of the page. This page is "Home > Forums > Software User Forums > ANSYS > CFX"

[sam_cfdd]

Hi,

I am granulating particles with initial size range of 20-500 micron by using PBM (Descrite). I am looking for a proper mesh size to run simulation and successful convergence?

[ghorrocks]

Please start a new thread for a new question.

You need to determine the mesh size required for your application through a mesh size sensitivity study.