[pashazanousi]

I want to simulate the case in attachment picture in steady state...I solve the equations for steady state and I want to get desire result in CFX.but CFX didn't get Converge...I don't know why?...
verif_paper.jpg
AAAAA.jpg

[karachun]

Because this problem is transient, not steady state. You can calculate interaction of waves and beach only in transients, there are no possibilities to reduce the problem to steady state.

[pashazanousi]

Right is yours...
But in Analytical solution we obtaine a solution that affected by the reflections...
I have two methods for satisfying this condition :
1-Control the simulation time to get the results affected by the reflections(for example do simulation for 40,50,...,100 and even more and compare results)
2-Steady state simulation(if possible)
Thank you for your attention...!
I work on Wave maker theory and I do my thesis...
I'm Applied Design Student... and I didn't work CFD before... Sorry if I asked a silly Question.. Thank you

[karachun]

I do not understand your answer, maybe you confuse time periodic flow with steady state solution. You can not solve this problem with steady state, only with transient simulation.

I'm Applied Design Student... and I didn't work CFD before... Sorry if I asked a silly Question.. Thank you

This did not matter.

If flow parameters don't change during time then flow may be assumed as steady state. If flow quantities (like velocity, pressure, volume fraction of water and air) vary during time then flow is transient.

Does flow quantities change during time in your problem? -Yes, water "particles" travel along circular or elliptical trajectories. Your case is transient. This is not related to ane theory you use to make analytical calculation.

This is how CFD can perform calculation. Some flow problems can be simplified to steady state, your problem can not be simplified.

[pashazanousi]

I agree with you it's not possible... I should control the simulation time and compare results...I asked a silly question... My solution is function of time... then there is no steady state solution...
Thank you for time you spent for my silly question

[karachun]

Yep, you got it.

Run transient solution, save enough transient results at selected time intervals and somehow post process results to obtain dependency of flow parameters vs. time.

BTW All flow problems ate transient, without exceptions. But sometimes transient fluctuations are too small and we can neglect them without significant loss of accuracy. Sometimes not.

[pashazanousi]

Thank you for everything

[ghorrocks]

BTW All flow problems ate transient, without exceptions. But sometimes transient fluctuations are too small and we can neglect them without significant loss of accuracy. Sometimes not.

Not really. A better way of thinking about it is if the flow is unchanging at all points in the flow for all time the flow is steady state. If the flow changes with time anywhere then it is transient.

The Navier Stokes equations has flows which are completely steady state. There are no fluctuations of any size, even tiny ones. But the Navier Stokes equations are a mathematical model of the actual physics of Molecular Dynamics which has molecules constantly whizzing around, and there is no steady state in Molecular Dynamics. But CFX is a Navier Stokes solver, not a Molecular Dynamics solver, so CFX can have precisely steady state solutions.

[pashazanousi]

Thank you...

[pashazanousi]

hi ghorrocks
I want to simulate standing waves in a flume of piston-type wave maker....My professor said if you increase simulation time,Error due residuals affect on results.
I simulate for 100 seconds with 0.01 time step with RMS 0.0001 with 50 max loop...
But the flow doesn't match with Analytical Solution....
Help me with this problem...Thank you...
Flume.jpg

CFX1.jpg

[karachun]

1. You should make grid independence study, reduce mesh size to ensure that mesh size doesn't affect the solution.


2. Run calculation with smaller residuals, like 1e-5 to ensure that residuals don't affect the result.


3. I recommend to use adaptive timestep size. Set min coef loops to 2 and max coef. loops to 10. Set the initial timestep to some small value like 1e-6 or 1e-5. Set min/max timesteps so the solver never reaches them, like 1e-10 and 1000. Set min/max target coef loops to 3 and 5.


I also have a question, I've read that if I solve flow using one iteration during timestep I can lose some accuracy - some variables are calculated only from the second timestep. Am I right?

[pashazanousi]

Originally Posted by karachun

1. You should make grid independence study, reduce mesh size to ensure that mesh size doesn't affect the solution.


2. Run calculation with smaller residuals, like 1e-5 to ensure that residuals don't affect the result.


3. I recommend to use adaptive timestep size. Set min coef loops to 2 and max coef. loops to 10. Set the initial timestep to some small value like 1e-6 or 1e-5. Set min/max timesteps so the solver never reaches them, like 1e-10 and 1000. Set min/max target coef loops to 3 and 5.


I also have a question, I've read that if I solve flow using one iteration during timestep I can lose some accuracy - some variables are calculated only from the second timestep. Am I right?

Thanks dear karachun
I try as you said...
And about your question if I understand what you ask...in first step some variables have initial values...Initialized variable doesn't need to Calculated......Just simple calculus for other variable base on initialization.....I hope answer your question right....

[pashazanousi]

I mesh the domain....is this good meshing?....How I can make a good mesh?
I use smaller mesh size near water surface....I use 0.001 mm for element size...othes element size is 0.005....Is it good?
I use two method for meshing...I use sweep method for the triangle part...I use all tri meshing for sweep method...help me to make a suitable mesh...thanks

Mesh1.jpg

Mesh2.jpg

Mesh3.jpg

Mesh4.jpg

[ghorrocks]

There is no general answer to those questions. The requirement is different for every application. So do a sensitivity analysis on your case and see what you need.

This means: For mesh size, do a simulation with double the element edge length and half the element edge length and see if the results change.

[pashazanousi]

Hi again...
I have a problem yet....
The problem is about simulation time....
It takes a long time to reach a good result to satisfy our analytical solution...
Do you have idea to reduce the simulation time to get the desire result?
Please note that We solve the equation for standing waves...it means we assume a time periodic solution without progressive waves...
Some results are in this file:
Solution.pdf

[karachun]

- make mesh coarser;
- turn off turbulence model (in free surface problems you can sometimes neglect turbulence).
On each step you should check that the results are still accurate enough.

[pashazanousi]

Originally Posted by karachun

- make mesh coarser;
- turn off turbulence model (in free surface problems you can sometimes neglect turbulence).
On each step you should check that the results are still accurate enough.

hi karachun
if I do as you said, solution affected due coarse mesh...
I ask my Professor to reduce simulation time enter our soulution as initial solution and let CFX Analayse it for 20 seconds and see how its changed and compare the results...But my professor reject my solution....I think my solution is the best solution...what is your idea?

[karachun]

Originally Posted by pashazanousi

I think my solution is the best solution...what is your idea?

Then try it and see if it help.

[pashazanousi]

Thank you karachun...but I expected more help...
I'm involving with this case more than 2 months...
I tried every possible method to validate the solution...
I asked this Question in researchgate too...but I didn't get the desire answer...
I don't know...may there is no way and the only way is coding...

Thanks for your attention

[karachun]

Originally Posted by pashazanousi

may there is no way

It was my first thought but I do not want to write rude comments.
There is no magic if you already have one layer of elements and you do not want to coarse mesh then I don't see other ways to speed up this simulation.
I don't think that you can speed up CFX using some programming.