Welcome, Guest
Username: Password: Remember me
Forgot your password? Forgot your username? Create an account
Forum
System Modules
TELEMAC-3D
Modelling a floating pier - tips?
Flat Threaded

TOPIC: Modelling a floating pier - tips?

Modelling a floating pier - tips? 10 years 3 months ago #13823

  • jmhervouet
  • jmhervouet's Avatar
Hello,

It is the number of sub-iterations that causes the problem in your case. I try to fix it but you may try with 1 instead of 2 to see if there is nothing else (and maybe 1 sub-iteration is enough, we do it for specific cases like non linear waves).

Regards,

JMH
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13824

  • jmhervouet
  • jmhervouet's Avatar
Hello,

Next try. The problem was that the atmospheric pressure gradient must not be added to the free surface gradient after the first sub-iteration because it has already been added once. So now I have tested your case in static in parallel, with 2 sub-iterations, and with free surface gradient compatibility not equal to 1. Let's see...

Regards,

JMH
Attachments:
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13843

  • pilou1253
  • pilou1253's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 584
  • Thank you received: 106
Hello,

I tried your last version on my case.

I first ran a static case with free surface gradient compatibility = 0,9, and it seemed to work even though I am a bit suprised by the results since the velocities remained constant and equal to 0 through the whole computation. With wave_equation from 6.3 and free surface gradient compatibility = 1,0 I had some (extremely small, 1.E-12) variations, but here, nothing...

Then I tried a dynamic case, with a 10 m/s wind in X direction, in a closed domain. Unfortunately, it seems that something is done wrongly here because the flow pattern is completely different from the one computed with the original wave_equation. I don't observe a flow recirculation inside the pool anymore, now the flow seems to be coming from the atmosphere at the entrance and living back to the atmosphere at the exit, with velocity vectors not parallel to water surface... See the pictures below.

Entrance

enter.png


Exit

exit.png


The images above correspond to a case with free surface gradient compatibility = 1,0. With a value of 0,9 the general pattern is the same, but there is a strong asymetry on the right end of the model (exit side, after the pier).

I also noticed something strange in the listing compared with simulations ran with wave_equation from 6.3. The propagation and diffusion step (GRACJG) is now completed with 0 iterations (!) and indicating SOLUTION X=0 BECAUSE L2-NORM OF B VERY SMALL:
================================================================================
ITERATION    18000 TIME    0 D  0 H 59 MN  60.0000 S   (       3600.0000 S)
================================================================================
--------------------------------------------------------------------------------
       ADVECTION STEP
 DIFFUSION OF WN
 GRACJG (BIEF) :        3 ITERATIONS, ABSOLUTE PRECISION:   0.4591832E-06
--------------------------------------------------------------------------------
       PROPAGATION AND DIFFUSION WITH WAVE EQUATION
 GRACJG (BIEF) :        0 ITERATIONS, ABSOLUTE PRECISION:   0.5586335E-06
 GRACJG (BIEF) :  SOLUTION X=0 BECAUSE L2-NORM OF B VERY SMALL:    0.000000
--------------------------------------------------------------------------------
       ADVECTION-DIFFUSION OF K-EPSILON OR OMEGA STEP
 DIFFUSION OF AKN
 GRACJG (BIEF) :        3 ITERATIONS, ABSOLUTE PRECISION:   0.3774177E-06
 DIFFUSION OF EPN
 GRACJG (BIEF) :        1 ITERATIONS, ABSOLUTE PRECISION:   0.2821719E-06
--------------------------------------------------------------------------------
       ADVECTION STEP
 DIFFUSION OF WN
 GRACJG (BIEF) :        3 ITERATIONS, ABSOLUTE PRECISION:   0.4591832E-06
--------------------------------------------------------------------------------
       PROPAGATION AND DIFFUSION WITH WAVE EQUATION
 GRACJG (BIEF) :        0 ITERATIONS, ABSOLUTE PRECISION:   0.7222320E-06
 GRACJG (BIEF) :  SOLUTION X=0 BECAUSE L2-NORM OF B VERY SMALL:    0.000000
--------------------------------------------------------------------------------


Hopefully this can give you some hints of what can be the possibl cause...

Thanks in advance,

Best regards
PL
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13846

  • jmhervouet
  • jmhervouet's Avatar
Hello,

The 0 iterations are just that the free surface does not move (or not enough accuracy is asked). The 2 sub-iterations are probably not needed and you can revert to 1 (just in case the problem is there), and it will be twice faster.
So I'll do dynamic tests now, but I'll be soon on holidays too, and I provisionally conclude that we are not ready for such computations, it might be a real and long development which has not been scheduled yet.

With best regards,

Jean-Michel Hervouet
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13849

  • pilou1253
  • pilou1253's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 584
  • Thank you received: 106
Hello,

I forgot to mention that I also made tests with 1 and 2 subiterations, but that it does not have any significant effect on the results.

Regarding the 0 iterations, it seemed strange to me because this step was solved with something like 5 iterations when using the original wave_equation.

I agree with you that these tests show that further and maybe deeper investigations are needed to get this functionality fully working (for free surface gradient compatibility not equal to 1).
But since the tests made with the original wave_equation and free surface gradient compatibility = 1 seemed to give satisfaying results, I will continue with this methodology for now on, unless if it is really not advised to use a fully compatible free surface gradient (?) or if it brings too many oscillations, ...

A last remark: this methodology has been previously applied with Telemac-3D for a few projects. One is presented in "Numerical modelling of the Vaugris reservoir on the Rhône with Telemac 3D" (Sogreah and CNR, proceedings TUC 2011), the other one was a 3D model simulating the filling of a lock pool with the presence of a ship (cannot find the reference?).
There are not lots of details given in the article from 2011 apart from that simulations were performed in hydrostatic mode. Maybe those who did the modelling can also give hints and feedback if this functionnality is to be implemented in a future release...

Once more, thanks again for your time and precious help on this topic!

Best regards,
PL
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13865

  • jmhervouet
  • jmhervouet's Avatar
Hello,

I investigated the last problem of 0 iterations, and actually it was a big mistake: an unchanged line had strangely disappeared from the last version of wave_equation I sent.

Anyway I wanted to do a safer implementation, by systematically adding the atmospheric pressure gradient to the free surface gradient anywhere it is computed, it is now done in FSGRAD, and WAVE_EQUATION is then less impacted. For this I had to move the call METEO in telemac3d.f because one free surface gradient was done before knowing the atmospheric pressure.

This is all in the enclosed file. I hope it is better now, I tried static and also with wind and it does not show the curious pattern you had yesterday, it really looks like wind.

I also noticed that the way you add the pressure does not work in parallel, but you probably know.

With best regards,

Jean-Michel Hervouet
Attachments:
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13888

  • pilou1253
  • pilou1253's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 584
  • Thank you received: 106
Hello,

Thank you for this correction.
I made a test and yes, the result looks like what is expected with wind, but I still observe some significant differences with the same computation performed with the original wave_equation.f and free surface copatibility gradient = 1,0.

The velocities in the recirculation stream near the bottom are significantly lower with the new correction, see images below.

Original version

6.3.png



With new correction and free surface gradient compatibility = 0,9

newcorrection.png



It seems that this new implementation needs to be validated properly.

Regarding the way I added pressure, you are right, I only made my tests in scalar. I have just tried to adapt it for parallel, for fun, but I face some problem. I intented to use function GLOBAL_TO_LOCAL_POINT but I think it is not a good idea since this function seems to be defined for 2D meshes only. Am I right?

Regards
PL
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13890

  • pilou1253
  • pilou1253's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 584
  • Thank you received: 106
Hello,

I made a test with only wind, without pressure. Results with the new correction are in perfect accordance with the original wave_equation.f, as one can expect.

The difference observed and pictured in the previous post might be due to the different treatment of water surface gradient in the zone of high air pressure. This difference actually causes a very different flow pattern in my model, with a higher part of the plunging stream being recirculated without flowing towards the lee of the model. Results for this specific case are therefore method-dependent...

Regards,
PL
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13891

  • jmhervouet
  • jmhervouet's Avatar
Hello,

You are right on the validation, though the first problem would be to know against what data we could do it (and turbulence is important too). However we can study the influence of the free surface gradient compatibility on non linear waves, which is a very sensitive case.

In your case I noticed that we are far from non linear waves, the free surface does not move a lot, maybe you can come back to one sub-iteration and remove the DYNAMIC PRESSURE IN WAVE EQUATION, at least it would be faster.

For the function GLOBAL_TO_LOCAL_POINT limited to 2D you might be right, because the first plane has same numbers as in 2D, and then for upper planes you just add NPOIN2 at every new plane for node numbers (but actually we could use this trick to write a more general function...).

Regards,

JMH
The administrator has disabled public write access.

Modelling a floating pier - tips? 10 years 3 months ago #13917

  • jmhervouet
  • jmhervouet's Avatar
Hello,

I checked and actually GLOBAL_TO_LOCAL_POINT works in 2D and 3D as well, for all points.

JMH
The administrator has disabled public write access.
Flat Threaded
Moderators: pham
Forum
System Modules
TELEMAC-3D
Modelling a floating pier - tips?

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.