TMVOC initial conditions Convergence Failure
I am trying to use TMVOC through Petrasim interface to build the attached conceptual model, but when I try to build such model I always have a convergence issues.
P.S I tried to follow the advise that the model build the initial condition from Scratch as in the attached TOUGH2:TMVOC file.
For simplicity I have assumed VOC mole fraction is zero to have only gas and water to initiate a simple one.
I appreciate any help regarding this issue.
Thank you in advance
It is a bit hard to comment about the convergence problem you mention without running the model. Could you share the Petrasim file? If you can, my email is email@example.com.
I can observe that you have initialized both the CLAY and AQFR domains in two-phase G-W conditions, while one would have expected the aquifer in single aqueous conditions. It seems your input file should be used to find the steady-state before the injection of benzene. I guess most of the aquifer elements would switch from two-phase to single aqueous with many phase transitions you may avoid simply assigning aqueous conditions from the beginning.
- CLAY permeability seems too high (0.1 D) and is equal to that of the aquifer domain;
- you have used no capillary pressure for all domains, while I would use capillary pressure at least for the CLAY domain where they should not be negligible. Using PCAP for all domains is any way recommended. The capillary pressure will determine (together with the rainfall recharge) the water saturation profile in the clay layer which will have a considerable role in the migration of benzene when simulating the spill. And you should use 3-phase capillary curves if you want then run the benzene injection
- you have used full mobile phase for the aquifer (IRP=5) and this could be one reason to have convergence problems initializing all the model in two-phase conditions. In addition, you should use 3-phase RELP curves from the beginning for all domain (axcept the ATMO), even if the NAPL is not present, if you want then run the injection of benzene.
you have specified 3-phase conditions for the CLAY zone with SO=0.1, which is a bit strange if you want to simulate first a steady-state before the NAPL spill. From the figure depicting the conceptual model it seems you want to simulate a specific NAPL spill that should eventually start once you have obtained the steady-state.
The ATMO layer at the top has atmospheric P of 1.013E5 Pa. Then you have assigned the initial P distribution in the CLAY and the AQFR domains. The P on the left boundary of AQFR (inlet boundary) is just atmospheric at the aquifer top, but the P is lower than atmospheric at the right boundary top (outlet). This produces tens and tens of gas phase evolutions on the first two aquifer grid layers. Apart for that, it is not physical that a freatic acquifer has a P lower than atmospheric as it is in contact with the ATMO boundary through the vadoze zone (CLAY).
I made a few (cosmetic) changes in the attached .sim file. It runs, but with tiny time steps and lots of phase switching 2-->4. Better if you revise the initial P distribution assigned to the aquifer. And decide about the initial conditions within the CLAY domain (2 or 3 phase?).