0

Non-convergence when upper boundary saturated in CO2

Hi everyone,

I am using the module ECO2N_V2.0 to model the hydrothermal fluid and heat circulation into high-topographic volcanoes.  I want to create a vadose zone in the upper part of my mesh. For this, I saturated a part of my upper boundary with CO2 setting a Sg value of 10.9 (in the INCON) and I ran the simulation for 1000 years without injecting anything (my mesh was already initiated, so with normal geothermal gradient and hydrostatic pressure). But I have a convergence failure at step 13. I tried to reduce the time step, but it did not work.

Can someone help me? I really appreciate it.

Thanks in advance,

Nathalie.

12 replies

null
    • Mikey_Hannon
    • 5 yrs ago
    • Reported - view

    Hi Natalie,

    I'm afraid without more details, it'll be hard to parse out the problem exactly.  I suspect it has something to do with an unintended instability that you may have set up with the initial conditions.  But to start with something easier (and I doubt this is your issue), what do you have set for your value of RE1 in the third line (record) of the PARAM block (PARAM.3)?  It is usually set to 1.0E-5, but often it'll go down to 1.0E-7. 

    If you wouldn't mind, it may be easiest to understand what's happening if I can get a look at your input file.

      • Nathalie_Collard
      • 5 yrs ago
      • Reported - view

      Mikey Hannon 

      Hi Mikey,

      First, thanks for your reply.

      I set a value of 1.0E-5 for my convergence criterion. I usually let the default value. 

      You can find attached my input file and my INCON file, too. 

      If you need something else, please let me know.

      I really appreciate your help.

      Cheers,

      Nathalie.

      flow.inp INCON
      • Nathalie_Collard
      • 5 yrs ago
      • Reported - view

      Mikey Hannon 

      Good news!

      With a convergence criterion of 1.0E-7 it ran properly =).

      Thank you very much.

      • Mikey_Hannon
      • 5 yrs ago
      • Reported - view

      Nathalie Collard 

      Interesting. Congrats!

      For what it's worth, I was able to run the file you sent me as-is (with the 1.0E-5 relative convergence threshold).  I've attached the output here.

      flow.out
      • Nathalie_Collard
      • 5 yrs ago
      • Reported - view

      Mikey Hannon 

      One question,

      Did you modify something in the input file to be able to run it without changing the convergence criterion? 

      I am still using TOUGH2. I don't if you use TOUGH3 (it could be the reason why it did not run?).

      Thanks!

      • Mikey_Hannon
      • 5 yrs ago
      • Reported - view

      Nathalie Collard 

      I personally run iTOUGH2 even for (forward) TOUGH simulations.   It's based on TOUGH2, so I'd imagine it would run similarly.  I do have a TOUGH2-ECO2N available to me, though, so I can give that a shot as well and let you know how it goes.  I haven't tried it with TOUGH3.

      • Nathalie_Collard
      • 5 yrs ago
      • Reported - view

      Mikey Hannon 

      Mmm very strange.

      That would be nice of you to run it with TOUGH2-ECO2N.

      • Mikey_Hannon
      • 5 yrs ago
      • Reported - view

      Nathalie Collard 

      Unfortunately, I only have a tutorial version from the TOUGH2 short course, which doesn't allow me to run a problem this big.  Perhaps someone like  Curt Oldenburg or Christine Doughty can give it a shot?

      • Nathalie_Collard
      • 5 yrs ago
      • Reported - view

      Mikey Hannon 

      Thank you. 

      It is suprising that it runs with iTOUGH2 and not with TOUGH2. 

      That would be nice if Curt Oldenburg or  Christine Doughty could run the input file in order to understand what is happening. 

      In any case, many thanks!

      P.-S. I attach mi output file, just in case and mi input and INCON files, too.

      flow.inp flow.out INCON
    • Staff Scientist
    • Christine_Doughty
    • 5 yrs ago
    • Reported - view

    Hi Nathalie and Mikey,

    I tried running the problem with TOUGH2/ECO2N V2 on my Windows 7 32-bit laptop.  It failed after 12-15 steps when RE1 was 1.e-5 or 1.e-6.  With RE1=1.e-7 it ran about 24 years, with lots of time step reductions and phase appearances and disappearances , before failing for non-convergence.  Finally, with RE1=1e-8 it ran fine.  I don't really know why it ran better for Mikey, but I suspect this is a difficult problem for TOUGH, maybe because CO2 is so compressible.  The machine you are running on might be as important as the version of TOUGH you are using.  I don't think I have ever run a gravity-equilibration problem with a lot of CO2 present.  Usually I start with water or brine, and do a gravity equilibration with single-phase liquid, then inject CO2 as a second step.  Nathalie, you said you want a vadose zone -  to me that means a near-surface region with water and air present.  ECO2N will not create this, since it does not include air.  If you do want a vadose zone, with CO2 also in the problem, you could try using EOS7CA (available with TOUGH3).  But if you really want a region with lots of CO2 as an upper boundary, then I guess ECO2N makes sense, but you need to realize that it is a difficult problem and use a small value of RE1.

    Christine

      • Mikey_Hannon
      • 5 yrs ago
      • Reported - view

      Thanks for your help, Chris!

    • Nathalie_Collard
    • 5 yrs ago
    • Reported - view

    Thanks for your reply, Christine.

    First, I ran a simulation of brine and CO2 injection (in two steps as you explained) into a fully water-saturated volcanic edifice, but my recharge was too high. It’s the reason why I want a no fully saturated region whether it is with CO2 or air. My aim is to control the recharge. I am going to follow your advice and use a small value of RE1.

    Once again, I thank you for your time.

    Nathalie.

Login to reply

Content aside

  • 5 yrs agoLast active
  • 12Replies
  • 149Views
  • 3 Following