Gas/water relperm test model not running

[EdmundStephens]

[v2025_10] Attached models designed to test the hysteresis scanning curve for gas-water systems. The model has 6 cells initialised with different initial SGAS, a forced water displacement through each cell and reporting from the block summary keywords (BWKR, BGKR, etc.). These are using EHYSTR mode 4.

Test models

• Store2r-F.zip ... CO2STORE model with saturation data in GSF, WSF tables ... this model runs successfully although still shows incorrect scanning curve interpolation, see discussion in Gas/water hysteresis test model run does not time step #6383.
• Swg2r-F.zip ... WATER/GAS model with saturation data in SGWFN tables ... this model does not time step.

So both these models show things are not yet quite working.

Output from the model Store2r-F.data

Output from the model Swg2r-F.data

================ Starting main simulation loop ===============

Report step  0/150 at day 0/100000, date = 01-Jan-2000
Using Newton nonlinear solver.

Starting time step 0, stepsize 1e-06 days, at day 0/0.001, date = 01-Jan-2000
Restart file written for report step   0/150, date = 01-Jan-2000 00:00:00

Problem: [./opm/simulators/wells/StandardWellPrimaryVariables.cpp:755] Infinite primary variable after update from wellState, well: P1

Warning: A zero well potential is returned for output purposes.

Problem: [./opm/simulators/wells/StandardWellPrimaryVariables.cpp:755] Infinite primary variable after update from wellState, well: P1

Warning: Inner well iterations failed for well P1 Treat the well as unconverged.

Problem: [./opm/simulators/wells/StandardWellPrimaryVariables.cpp:755] Infinite primary variable after update from wellState, well: P1

Warning: Inner well iterations failed for well P1 Treat the well as unconverged.

Problem: [./opm/simulators/wells/StandardWellPrimaryVariables.cpp:755] Infinite primary variable after update from wellState, well: P1

Warning: Inner well iterations failed for well P1 Treat the well as unconverged.
flow: ./opm/simulators/flow/FlowProblem.hpp:996: void Opm::FlowProblem<TypeTag>::source(RateVector&, unsigned int, unsigned int) const [with TypeTag = Opm::Properties::TTag::FlowGasWaterProblem; RateVector = Opm::BlackOilRateVector<Opm::Properties::TTag::FlowGasWaterProblem>]: Assertion `isfinite(rate[eqIdx])' failed.
[EdLaptop:00326] *** Process received signal ***
[EdLaptop:00326] Signal: Aborted (6)
[EdLaptop:00326] Signal code:  (-6)
flow: ./opm/simulators/flow/FlowProblem.hpp:996: void Opm::FlowProblem<TypeTag>::source(RateVector&, unsigned int, unsigned int) const [with TypeTag = Opm::Properties::TTag::FlowGasWaterProblem; RateVector = Opm::BlackOilRateVector<Opm::Properties::TTag::FlowGasWaterProblem>]: Assertion `isfinite(rate[eqIdx])' failed.

[totto82]

Thanks for reporting the issue. For me Swg2r-F.data runs but not Store2r-F.data. Is it the PC data that is wrong in the case that runs? You need to enable the PC fix by setting item 14 to 1 in EHYSTR. I hope to change the defaults before the release, but have not done so yet. I will investigate why the failing case fails and report back.

[totto82]

With item 14 = 0 (default)

With item 14 = 1

[totto82]

The Store2r case runs fine without DISGASW. For relperm testing, that may be sufficient. I don't know what's causing the problem with dissolved CO2 yet, but I hope to figure it out.

[EdmundStephens]

Interesting

For me Swg2r-F.data runs but not Store2r-F.data

I get the other way round, including when EHYSTR item 14 is set to 1. There should be no difference in the data as such, coming from the same spreadsheet, but these are loaded differently:

• Swg2r ... data in a SGWFN table
• Store2r ... data WSF and GSF tables

The Swg2r case fails and the Store2r runs (including with the DISGASW keyword).

I tried the same data with SWFN and SGFN tables, where these have the Pc data tabulated on the SWFN and SGFN tables respectively:

• SfnW2r ... data in SWFN and SGFN tables with Pc data in SWFN table
• SfnG2r ... data in SWFN and SGFN tables with Pc data in SGFN table

SfnW2r-F.zip
SfnG2r-F.zip

These fail to run on my laptop, e.g.

Warning: Inner well iterations failed for well P1 Treat the well as unconverged.

Problem: [./opm/simulators/wells/StandardWellPrimaryVariables.cpp:755] Infinite primary variable after update from wellState, well: P1

Warning: Inner well iterations failed for well P1 Treat the well as unconverged.
flow: ./opm/simulators/flow/FlowProblem.hpp:996: void Opm::FlowProblem<TypeTag>::source(RateVector&, unsigned int, unsigned int) const [with TypeTag = Opm::Properties::TTag::FlowGasWaterProblem; RateVector = Opm::BlackOilRateVector<Opm::Properties::TTag::FlowGasWaterProblem>]: Assertion `isfinite(rate[eqIdx])' failed.
flow: ./opm/simulators/flow/FlowProblem.hpp:996: void Opm::FlowProblem<TypeTag>::source(RateVector&, unsigned int, unsigned int) const [with TypeTag = Opm::Properties::TTag::FlowGasWaterProblem; RateVector = Opm::BlackOilRateVector<Opm::Properties::TTag::FlowGasWaterProblem>]: Assertion `isfinite(rate[eqIdx])' failed.

Overall quite confusing, hopefully these additional cases clarify in the debugger where differences might be emerging.

Since you get different run outcomes, possibly there is a compiler library issue ... I have a MS Surface Laptop with the ARM architecture build .. some weird difference there perhaps. On my laptop the Store2r case runs fine, and also when I remove DISGASW and set EHYSTR item 14 = 1:

[totto82]

Thanks for the update. I have dug down and found that the issue is related to a cell that is first filled with gas, then, gradually, the gas disappears, and then the dissolved CO2 disappears as pure water fills the cell from the injectors. In this cell, the Newton iterations stagnate and do not update, so the simulator does not converge. I have not found any fundamental issues that explain why the simulator fails when the amount of dissolved CO2 goes to zero yet, but I will continue to look, as this is not an uncommon situation in CO2 storage. Note that I have not seen this issue in other cases where dissolved CO2 goes to zero.

Also, are the results still wrong for the cases that run when items 13 and 14 are enabled?

[totto82]

As I see it, the key difference between Store2r-F.data and Swg2r-F.data is that the former uses CO2STORE with DISGASW, whereas the latter uses PVTW and PVDG to define the properties. The first will allow CO2 to dissolve in water, while the second will not.

The convergence problems I observe are related to cells where the gas saturation goes below the critical saturation due to dissolution. With master this could lead to negative mobilities. I added a fix (see OPM/opm-common#5073). The PR fixes the issue for the relperm, but I still have convergence issues related to PC hysteresis. I.e. setting item 5 to KR helps. I don't see any obvious issues in the code, but I can understand that the test case is challenging for the simulator when dPc goes towards -10 bars. Note also that I need to set item 13 and item 14 for the simulator to converge on these cases. I have also fixed the output of BGPC and BWPC for gas-water models.