How to identify the cause of an incomplete simulation run


I have a problem to identify the cause of the incomplete simulation run for the attached input file. The simulation ends before the specified total simulation time without error messages! The simulation is for CO2 injection in a saline aquifer 2D radial model. How can I identify the reason for the simulation termination?

8replies Oldest first
  • Oldest first
  • Newest first
  • Active threads
  • Popular
  • I don't see any obvious problems with your input file.  Could you post your output file?

    • Christine Doughty Sure. The output file is attached below.

    • Refaat G Hashish 

      You asked for the code to take 9999 time steps (PARAM block, first line, columns 5-8) and the code took 9999 time steps and stopped.  It is doing just what you asked.  If you want more than 9999 steps, see the manual for a description of the variable MCYC.  If it is positive, it is the number of steps.  If it is negative, the number of steps is 2 to the power (-MCYC).  

      But the real question here is why is the code taking such small time steps (~2 hours), so that it requires 9999 steps in the first place.  I am not sure, but I see that you are getting a lot of salt precipitation near the well, and in my experience this can cause the code to slow down.  I see that you are not having permeability decrease with salt precipitation - that may be unrealistic when most of the pore space is filled with salt.

      I am also not too happy with your choice of capillary pressure function (ICP=7).  When there is a lot of salt precipitation, liquid saturation goes to zero, and this puts you in a bad part of the capillary pressure curve.  Have you downloaded the capillary pressure spreadsheet from Thunderhead Engineering?  If not, please do so (https://www.thunderheadeng.com/petrasim/petrasim-resources/ and scroll down to Capillary Pressure Spreadsheet) and put in your capillary pressure parameters and see what your Pap curve looks like.  Near the well you are in the flat part at Pc=Pcmax.  It is also bad that there is a kink in the Pcap curve when you go from the curvy part to the flat part.  It is not in the User's Guide, but if you put in a value less than 1 for Pcmax, the code will not create the flat part of the Pcap curve and will have a linear extension instead.  This may help.  You can try different values of Pcmax<1 and test it (unfortunately, I don't think this feature is included in the Thunderhead spreadsheet).

      Also, I recommend putting IE(16)=0 in the SELEC block.  I do not know what IE(16)=2 does.

      As a quick check, you can decrease the amount of dissolved salt in the problem and rerun, and see if you get bigger time steps with less salt precipitation.

    • Refaat G Hashish You have reached the maximum number of cycles (9999). Rename the output SAVE file as INCON, remove the INCON block in your input file, and restart the simulation (it will start where it left).

    • Christine Doughty You mean by Pcmax less than 1 is that Pcmax less than 1 or 1E7.

  • In the ICP=7 formulation for capilliary pressure, Pcmax < 1.0 acts a flag that changes how the Pc versus Sl curve is calculated for small values of Sl.  I have not used this feature very much, but it did seem to work with a value of Pcmax=CP(4)=1E-3.  

    • Christine Doughty Thank you.

  • Refaat G Hashish, regarding the flat part of your curve at early saturations, consider also to change the irreducible water content that you set in the CP(2). As TOUGH manual indicates, this has to be l chosen smaller than the corresponding parameter in the relative permeability function. 

Like Follow
  • 1 yr agoLast active
  • 8Replies
  • 91Views
  • 4 Following