Welcome, Guest
Username: Password: Remember me
Forgot your password? Forgot your username? Create an account
Forum
System Modules
TELEMAC-3D
Mass balance: convergence issues with large error
Flat Threaded

TOPIC: Mass balance: convergence issues with large error

Mass balance: convergence issues with large error 7 years 1 month ago #27988

  • SDAC
  • SDAC's Avatar
Hello!

I'm having trouble achieving convergence on my mass balance, and keep getting large errors. For example|:
                MASS BALANCE

  WATER
MASS AT THE PREVIOUS TIME STEP                :     230.8138
MASS AT THE PRESENT TIME STEP                 :     230.8142
MASS LEAVING THE DOMAIN DURING THIS TIME STEP :   -0.4131773E-03
ERROR ON THE MASS DURING THIS TIME STEP       :   -0.8850420E-12
FLUX BOUNDARY    1:   -0.7336823     M3/S  ( >0 : ENTERING  <0 : EXITING )
FLUX BOUNDARY    2:    0.7750000     M3/S  ( >0 : ENTERING  <0 : EXITING )

================================================================================
ITERATION    18990 TIME    0 D  0 H  3 MN   9.9000 S   (        189.9000 S)
================================================================================
 ELEVATION AT THE BOUNDARY READ                       IN THE ASCII FILE
--------------------------------------------------------------------------------
                    ADVECTION STEP
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    1.6412221725393010E-006
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    1.0790969161468380E-006
 MURD3D_LIPS:    2 ITERATIONS
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    2.1705898326252831E-007
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    9.1777896767652854E-008
 MURD3D_LIPS:    2 ITERATIONS
 JACOBI:          18  ITERATIONS, ABSOLUTE PRECISION =    3.1161095241045298E-008
 JACOBI:           4  ITERATIONS, ABSOLUTE PRECISION =    2.9983818474002073E-008
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF WN
 GRACJG (BIEF) :        2 ITERATIONS, RELATIVE PRECISION:   0.8240593E-06
--------------------------------------------------------------------------------
       PROPAGATION AND DIFFUSION WITH WAVE EQUATION
 GMRES (BIEF) :        1 ITERATIONS, ABSOLUTE PRECISION:   0.1880222E-08
 POSITIVE DEPTHS OBTAINED IN    10 ITERATIONS
--------------------------------------------------------------------------------
                DYNAMIC PRESSURE STAGE
 GRACJG (BIEF) :       19 ITERATIONS, ABSOLUTE PRECISION:   0.8984517E-04
--------------------------------------------------------------------------------
                VELOCITY PROJECTION STEP
--------------------------------------------------------------------------------
       ADVECTION-DIFFUSION OF K-EPSILON OR OMEGA STEP
 JACOBI:          20  ITERATIONS, ABSOLUTE PRECISION =    1.9805044769621615E-008
 JACOBI:          21  ITERATIONS, ABSOLUTE PRECISION =    1.9339222359433896E-009
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF AKN
 GRACJG (BIEF) :        2 ITERATIONS, RELATIVE PRECISION:   0.5323668E-06
 JACOBI:          20  ITERATIONS, RELATIVE PRECISION =    2.3347301145346052E-007
 JACOBI:          21  ITERATIONS, RELATIVE PRECISION =    1.6977921593802961E-008
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF EPN
 GRACJG (BIEF) :        2 ITERATIONS, RELATIVE PRECISION:   0.1497274E-07
--------------------------------------------------------------------------------
                MASS BALANCE

  WATER
MASS AT THE PREVIOUS TIME STEP                :     230.8180
MASS AT THE PRESENT TIME STEP                 :     230.8184
MASS LEAVING THE DOMAIN DURING THIS TIME STEP :   -0.4218900E-03
ERROR ON THE MASS DURING THIS TIME STEP       :    0.1840994E-12
FLUX BOUNDARY    1:   -0.7328110     M3/S  ( >0 : ENTERING  <0 : EXITING )
FLUX BOUNDARY    2:    0.7750000     M3/S  ( >0 : ENTERING  <0 : EXITING )

================================================================================
ITERATION    19000 TIME    0 D  0 H  3 MN  10.0000 S   (        190.0000 S)
================================================================================
 ELEVATION AT THE BOUNDARY READ                       IN THE ASCII FILE
--------------------------------------------------------------------------------
                    ADVECTION STEP
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    1.6961916236193058E-006
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    1.1386036348261629E-006
 MURD3D_LIPS:    2 ITERATIONS
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    2.2987037924346075E-007
 JACOBI:           1  ITERATIONS, ABSOLUTE PRECISION =    1.0763102934767386E-007
 MURD3D_LIPS:    2 ITERATIONS
 JACOBI:          13  ITERATIONS, ABSOLUTE PRECISION =    2.8343826207694563E-008
 JACOBI:           5  ITERATIONS, ABSOLUTE PRECISION =    2.9359160572083577E-008
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF WN
 GRACJG (BIEF) :        2 ITERATIONS, RELATIVE PRECISION:   0.5925799E-06
--------------------------------------------------------------------------------
       PROPAGATION AND DIFFUSION WITH WAVE EQUATION
 GMRES (BIEF) :        1 ITERATIONS, ABSOLUTE PRECISION:   0.1593106E-08
 POSITIVE DEPTHS OBTAINED IN    10 ITERATIONS
--------------------------------------------------------------------------------
                DYNAMIC PRESSURE STAGE
 GRACJG (BIEF) :       19 ITERATIONS, ABSOLUTE PRECISION:   0.9151514E-04
--------------------------------------------------------------------------------
                VELOCITY PROJECTION STEP
--------------------------------------------------------------------------------
       ADVECTION-DIFFUSION OF K-EPSILON OR OMEGA STEP
 JACOBI:          21  ITERATIONS, ABSOLUTE PRECISION =    2.1404081125513093E-009
 JACOBI:          28  ITERATIONS, ABSOLUTE PRECISION =    5.1194876800521412E-008
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF AKN
 GRACJG (BIEF) :        2 ITERATIONS, RELATIVE PRECISION:   0.1358405E-07
 JACOBI:          21  ITERATIONS, RELATIVE PRECISION =    1.0815729125060351E-011
 JACOBI:          28  ITERATIONS, RELATIVE PRECISION =    5.1466174711521180E-007
 MURD3D_LIPS:    2 ITERATIONS
 DIFFUSION OF EPN
 GRACJG (BIEF) :        1 ITERATIONS, RELATIVE PRECISION:   0.5516821E-06
--------------------------------------------------------------------------------

MASS BALANCE

  WATER
MASS AT THE PREVIOUS TIME STEP                :     230.8222
MASS AT THE PRESENT TIME STEP                 :     230.8227
MASS LEAVING THE DOMAIN DURING THIS TIME STEP :   -0.4307286E-03
ERROR ON THE MASS DURING THIS TIME STEP       :    0.1745646E-11
FLUX BOUNDARY    1:   -0.7319271     M3/S  ( >0 : ENTERING  <0 : EXITING )
FLUX BOUNDARY    2:    0.7750000     M3/S  ( >0 : ENTERING  <0 : EXITING )

--------------------------------------------------------------------------------
                FINAL MASS BALANCE
T =         190.0000

--- WATER ---
INITIAL MASS                        :     281.4202
FINAL MASS                          :     230.8227
MASS LEAVING THE DOMAIN (OR SOURCE) :     50.59751
MASS LOSS                           :    0.4618528E-12
 CALL OF P_EXIT IN ITS VOID VERSION

 CORRECT END OF RUN

There's a large error on the mass for each time step, and the difference in mass between the initial and final mass is also large. The differences between current and previous time-steps come close to E-03 but at most it stays like that for one-three time-steps.

One thought has been accuracy: if mass balance is the order of magnitude of the accuracy asked for propagation, then by using TREATMENT OF NEGATIVE DEPTHS = 2 (accuracy of the machine) and changing ACCURACY FOR PROPAGATION to 1.E-12 I hoped to find better balance. This is not the case.

I imagine it's likely due to the mesh. With a low resolution for my mesh I seem to get better results.

I think it's due to the mesh quality. There are no holes in the mesh bank from what I can see, but there are some elongated triangles at various points along my mesh where there are sudden changes in height between channel and bank. These larger triangles with poor ratios may be increasing the error. To change this I'm using Blue Kenue to alter the density of the banks using hard-lines but the results aren't great. From the image you can see the mesh line doesn't conform to the outline.

Why might this be happening? I followed the instructions in the manual for density:

Elongated bank:


poormesh3.jpg



Bank density:


poormesh2.jpg


T3 Mesh generator outcome:



poormesh4.jpg



Are these elongated triangles likely to be the main source of error, or am I looking in the wrong direction?
Attachments:
The administrator has disabled public write access.

Mass balance: convergence issues with large error 7 years 1 month ago #27991

  • cyamin
  • cyamin's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 997
  • Thank you received: 234
Hello SDAC,

Addressing to your mesh generation issue:
  • If you are using a density mesh to control mesh density, make sure the density mesh is a bit larger than the mesh you are applying it to, i.e. they do not share boundary definition!
  • If your mesh density is constant, why not use a polygon instead?
  • Why don't you use softlines instead of hardlines and let BK resample the lines according to the density you have specified.
Regards,
Costas
The administrator has disabled public write access.
The following user(s) said Thank You: SDAC

Mass balance: convergence issues with large error 7 years 1 month ago #28010

  • SDAC
  • SDAC's Avatar
Thankyou Costas.

Unfortunately even with high densities the gradients are still too sharp and the triangles stretched. Going too refined on the mesh density would severely lengthen the computational time.

Splitting the triangles seems to help, but there are many stretched triangles whcih probably need to be split multiple times. Does Blue Kenue have an option to split multiple triangles, or is there any software that can?

Many thanks!
The administrator has disabled public write access.

Mass balance: convergence issues with large error 7 years 1 month ago #28013

  • cyamin
  • cyamin's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 997
  • Thank you received: 234
Hello SDAC,

Make sure you follow the first bullet in my previous post! I used to have the exact problem at the boundaries when applying a density mesh. Offset the boundary for the density mesh a few meters outwards and apply it again. However, since the density is constant, it is much easier to draw a polygon around the area and set the density for it.
You can post your BK project files if you need more help.

Regards,
Costas
The administrator has disabled public write access.
The following user(s) said Thank You: SDAC

Mass balance: convergence issues with large error 7 years 1 month ago #28016

  • riadh
  • riadh's Avatar
Hello

I think that it is simply the matter of resampling boundary outline. you need just to resample the line with a dx close to the density you are using.

kind regards
Riadh
The administrator has disabled public write access.
Flat Threaded
Moderators: pham
Forum
System Modules
TELEMAC-3D
Mass balance: convergence issues with large error

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