0

Problems with ECO2N V. 2.0

Hi All,

I am experiencing some problems with the latest version of ECO2N. I have a series of simulations that run fine using ECO2N V1.0. However in version 2.0, they fail when repeated convergence within one Newton-Raphson iteration was achieved.

Is there a set of SELEC flags to replicate the behaviour of the original ECO2N V1.0, i.e. is setting SELEC parameter 16 to 1.0  equivalent to running the old version of ECO2N?

Also, does anyone have any idea why this would happen when injecting CO2, it is not an equilibration run. I would have expected it to fail cycling small time steps, rather than the system achieving steady state!

I realise that the new version of TOUGH uses a different set of solvers, not the ma28.f package in ECO2N V1.0. I clearly need to change something as I am struggling to get anything to run!

Any ideas would be greatly appreciated!

Thanks for your help,

Gareth

3 replies

null
    • Finsterle GeoConsulting
    • Stefan_Finsterle
    • 7 yrs ago
    • Reported - view

    Gareth,

     

    I don't have an answer, just a few questions:

    (1) SELEC parameter IE(16) can be set to 1 (an integer, not a real!) to revert to the ECO2N V1.0 solubility model, but it does not affect anything else.

    (2) Be aware that FE(1) and FE(2) of ECO2N V1.0 is FE(9) and FE(10) in ECO2N V2.0. Did you change that?

    (3) I'm not aware that ECO2N was ever distributed along with MA28. In fact, MA28 was never distributed by LBNL (only by ESTSC). Can you please confirm?

    (4) Do all your simulations fail in the same way?

    (5) Are your injection rates non-zero (please check this in the output file).

    (6) Are you willing to share your input files with the TOUGH community?

    In case of persistent problems, please contact Lehua Pan.

    Thanks,

    Stefan

    • Gareth_Williams
    • 7 yrs ago
    • Reported - view

    Hi Stefan,

    Thanks for your suggestions. I have turned off the permeability dependence on salt (IE11 = 0).

    With regard to ma28, this is slightly embarrassing. I have inherited all this from a colleague who retired, leaving very little in the way of any documentation for me to work with (a sore point with me at the moment!). I am starting to find many modifications (uncommented) to the original TOUGH code. He may well have got it from ESTC as far as I know, it certainly uses ma28.f!

    I  believe I have narrowed down the problem by a process of elimination, and the simulations appear to fail when they try to take a large time-step 'across' a step boundary in my GENER (injection) table. I am now running a constant injection rate example to test this theory and it has progressed further than the previous simulations (still running).

    If that does not solve it I am happy to upload the files. 

    Once again,

    Thanks for your suggestions,

    Gareth

    • Gareth_Williams
    • 7 yrs ago
    • Reported - view

    Using a smoothed function for the injection rates allowed the simulations to complete.

    Thanks,

     

    Gareth

Login to reply

Content aside

  • 7 yrs agoLast active
  • 3Replies
  • 825Views
  • 2 Following