Time step iteration criteria exceeded

[ele...@lbl.gov]

Good morning,

Sorry for the next stupid question,
it appears that my model is not computing till the end of the time I'm asking,
"  Stopping: Time step cut criteria exceeded!
     icut = 17, max_time_step_cuts= 16
     dt   =  4.2385525E-10, dt_min=  0.0000000E+00
"
Which criteria do I have to increase deacrease ?

Here are parts of my input file

SIMULATION
  SIMULATION_TYPE SUBSURFACE
  PROCESS_MODELS
    SUBSURFACE_FLOW flow
      MODE TH
    /
    SUBSURFACE_TRANSPORT transport
      GLOBAL_IMPLICIT
    /
  /
END

#===================== Discretization =========================================
GRID
  TYPE structured
  ORIGIN 0.d0 0.d0 0.d0
  NXYZ 100 100 1
  BOUNDS
    0.d0 0.d0 0.d0
    1.d0 1.d0 1.d0
  /
END


#===================== Solver options =========================================
NEWTON_SOLVER FLOW
  ATOL 1D-12
  RTOL 1D-8
  STOL 1D-15
  ITOL 1D-8
  MAXIT 100
  MAXF 1000
END

#========================= output options =====================================
OUTPUT
  TIMES min 10. 20. 30. 40. 50.
#  SCREEN PERIODIC 1
#  PERIODIC_OBSERVATION TIMESTEP 1
  PRINT_COLUMN_IDS
  FORMAT TECPLOT POINT
#  FORMAT HDF5
  VELOCITY_AT_CENTER
END

#======================== times ===============================================
TIME
  FINAL_TIME 1. h
  INITIAL_TIMESTEP_SIZE 0.1d0 s
  MAXIMUM_TIMESTEP_SIZE 100.d0 s
END


Thanks for the incoming help,

Have a good one

e

[Nathan Collier]

You can change the number of cuts in the timestepper card, see here:

https://bitbucket.org/pflotran/pflotran-dev/wiki/Documentation/QuickGuide/Timestepper

16 is the default number, but increasing this may not help you. This tends to happen when your residual cannot be reduced to zero, might be an error.

Nate

[Peter Lichtner]

Why don’t you just send the entire input file. Is this happening in flow or transport? If in chemistry the tolerances might be too tight—you appear to be using default values which I have found are not that good.
Cheers,
…Peter

[Emmanuel LEGER]

Good morning Peter,
Thanks for your answer,

Here is the file.

[Emmanuel LEGER]

Hi Nate,
An error in my input file you mean ?

e

[Peter Lichtner]

For me it crashes in flow on the first time step. You inlet and outlet regions seem to overlap—is this what you want to do?

…Peter

[Emmanuel LEGER]

Oups,
no, but same issue
...e

[Peter Lichtner]

Yes—seems necessary to reduce the injection/production rates to 1.e-5 kg/s. Might be a bug in TH! 

…Peter

<th_1d_el.in>

[Emmanuel LEGER]

Yes, but even with it, does not want to go through all the time step..
e

[Peter Lichtner]

Works for me but I added:

NEWTON_SOLVER TRANSPORT

  ATOL 1D-12

  RTOL 1D-8

  STOL 1D-30

  ITOL 1D-8

  MAXIT 100

  MAXF 1000

END

Also changed the chemistry so you may want to do a diff. of the input files.

…Peter

[Nathan Collier]

Emmanuel,

I should just never write mail on my phone, it is inevitably too brief. I meant that often when you see a lot of time step cuts in one time step, it means that you have reached a point where the nonlinear solver cannot find a solution. This can be real and the nonlinearity is just hard to resolve, but in my experience it is more often about some unphsyical state that I have forced in the problem (incompatible boundary conditions, etc). Peter's specific advice gets to just that.

Nate

[Emmanuel LEGER]

Ok, yes, Thanks for the precision.

The problem is mine then :).

Thanks Nate

e

[Peter Lichtner]

I agree with Nate: usually when something that catastrophic happens there is something wrong with the problem formulation—or a bug!

This might also be a bug in TH so you might have two simultaneous problems! I set the injection pH to 3 but not seeing that in the output even with the reaction shut off (calcite rate const = 0). Also the temperature goes to 10 C which I don’t understand.

…Peter

[Emmanuel LEGER]

Ok, so please Peter, let me know if you find something unphysically possible in my codes, I'd be glad to understand what I'm doing wrong. I just modified a bit the th_1d.in which was available in the regression folders. Maybe doing it in 2d is asking too much computing ? Well, I juste have 100*100 nodes

Thanks to both of you

e

[Hammond, Glenn E]

I will add that the code usually dumps the most recent solution (as long as I/O is turned on) when it cuts to failure and that may provide additional information.

 

Glenn

[Karra, Satish]

Emmanuel,

Can you please pull the latest and see it works for you with the mass_rate source/sink?

—Satish

======================================

Satish Karra, PhD

Staff Scientist

EES-16: Computational Earth Science Group

Los Alamos National Laboratory

Los Alamos, NM 87545

E-mail: satk...@lanl.gov

Phone: 505-606-1894

Webpage: http://ees.lanl.gov/staff/satkarra

======================================

[Emmanuel LEGER]

Good evening Satish,
Thanks for your concern, I thought you forgot about me, well it doesn't change that much...
it seems, that on short scale (time and distance) the code has trouble on solving Thermal flux.
Could you have a look on my input file ? I don't think I ask unphysical state... Am I ? I switched to volumetric rate


Cheers

[Peter Lichtner]

Emmanuel: your input file still seems to have problems: you have

REGION all

  COORDINATES

    0.d0 0.d0 0.d0

    100.d0 100.d0 1.d0

  /

END

which should be:

REGION all

  COORDINATES

    0.d0 0.d0 0.d0

    1.d0 1.d0 1.d0

  /

END

Try the attached input file (you will need to change back the database path to your setup).

…Peter

[Emmanuel LEGER]

Well,
As usual I was stupid...
Thanks Peter...
Sorry for the time you (wasted) spent on that

e

    0.d0 0.d0 0.d0
    1.d0 1.d0 1.d0
  /
END

<th_1d_el.in>

[Peter Lichtner]

Note I also reduced the rates to get convergence. 

...Peter <iPhone>

[Emmanuel LEGER]

Good afternoon,

Yes I have seen, unfortunately, because of that the modeling doesn't fit what I'm looking for.

I'd like an input and output volumetric flux at around 50-100 ml/min, but cannot, it does not converge when I use these higher fluxes.
I'm trying to make a modeling, very small scale, in agreement with what I do during lab experiment.


Thanks for your help,,
best

e

[Peter Lichtner]

There might still be a problem with the Jacobian that limits the rate. 

...Peter <iPhone>

<th_1d_el.in>

[Emmanuel LEGER]

You mean from pflotran side or from my side ( wrong parametrization from me for instance)

e

[Peter Lichtner]

I'm thinking PFLOTRAN bug. 

...Peter <iPhone>

[Hammond, Glenn E]

The issue here is that you have a closed system with strong thermal gradients and only source/sinks.  Densities/pressures are all over the place. 

One quick question: Is the prescribed injection rate the total injection rate or should it be prescribed to each of the three inlet cells?  If it is the total injection rate, use TYPE SCALED_VOLUMETRIC_RATE VOLUME for the inlet.

Otherwise, you need to set a Dirichlet pressure/temperature condition at the outlet (you can just use the initial condition).  This will allow the outlet boundary to lose mass without over/under pressurizing the domain.  It is too difficult with just a source and a sink.  Or perhaps I am missing something.

See the attached input deck.

​Glenn

---

From: pflotra...@googlegroups.com <pflotra...@googlegroups.com> on behalf of Peter Lichtner <peter.l...@gmail.com>
Sent: Saturday, November 14, 2015 2:26 PM
To: pflotra...@googlegroups.com
Subject: [EXTERNAL] Re: [pflotran-users: 2978] Time step iteration criteria exceeded

 

[Hammond, Glenn E]

​Should use a NEUMANN boundary condition with a specified FLUX at REGION 'inlet' with the REGION being defined at the WEST FACE.

Glenn

---

From: pflotra...@googlegroups.com <pflotra...@googlegroups.com> on behalf of Hammond, Glenn E <geh...@sandia.gov>
Sent: Saturday, November 14, 2015 4:09 PM
To: pflotra...@googlegroups.com
Subject: Re: [EXTERNAL] Re: [pflotran-users: 2979] Time step iteration criteria exceeded

 

[Peter Lichtner]

The system is not closed---injection is one one side of the domain and production on the other side. 

...Peter <iPad>

<th_1d_el.in>

[Hammond, Glenn E]

Call it whatever you want.  It is a system without boundary conditions and just prescribed source/sinks.  If you inject warm water, how can you expect it not to pressurize?  Especially with volumetric source/sinks.  But even with mass rate, there are issues.  Perhaps the small domain.

Glenn

[Emmanuel LEGER]

Ok Glenn,
Thanks,
"Is the prescribed injection rate the total injection rate or should it be prescribed to each of the three inlet cells?"
It is the total injection rate, divided into 3 cells, so I kept what I had.
I went trough your advices and scripts.
I understand now, where the problem is. I can run my code with Dirichlet condition for the outlet.
However since I model what's happening in the lab, I now that I inject and pump at the same rate (hundred ml/min). But the top boundary condition is open to atmospheric pressure, so that might solve my issue actually, since I won't have over pressure. I haven't said anything about this and I think it will solve the issue. Is that a seepage to set ? well no, not really, what is the water table opened into atm boundary condition ?

Thanks

e

On 14/11/2015 16:09, Hammond, Glenn E wrote:

The issue here is that you have a closed system with strong thermal gradients and only source/sinks.  Densities/pressures are all over the place. 

One quick question: Is the prescribed injection rate the total injection rate or should it be     RATE VOLUMETRIC_RATE
int heprescribed to each of the three inlet cells?  If it is the total injection rate, use TYPE SCALED_VOLUMETRIC_RATE VOLUME for the inlet.

[Emmanuel LEGER]

Well, based on what I've found on your bucket,
seems to be smthg like:
FLOW_CONDITION Piezometric_Surface
  TYPE
    PRESSURE HYDROSTATIC
    TEMPERATURE dirichlet
  /
  PRESSURE X.d0
TEMPERATURE X.D0
END

e

[Hammond, Glenn E]

The problem is that your heat boundary condition will be too dominant.

Glenn

[Emmanuel LEGER]

Yes, but still the pressure can release at the top BC.
Well, it doesn't work, so you must be right

e

[Hammond, Glenn E]

Make the pressure at the outlet slightly above atmospheric and you should be fine.

Glenn

[Emmanuel LEGER]

Yes sure, it compile, and works.
However this is not what I'd like to model.
What the difference between Neumann  and source/sink ?
 I'd like to have my output only with a rate, not a temperature, neither a pressure, since the only info I have about my output (sink) is the mass flow output.

Cheers

e

[Hammond, Glenn E]

Since it is an outflow (zero gradient) boundary, the temperature should have no effect, unless the TH developers implemented the boundary conditions incorrectly.  The pressure will have no effect of you leave it at the initial pressure.  

If you are worried, play around with the initial pressure and see the impact on outflow by turning on the MASS_BALANCE card in OUTPUT, checking the -mas.dat output file.

Glenn

<th_1d_el.in>

[Peter Lichtner]

Emmanuel: I tried increasing the initial pressure to 1.5e6 Pa and that runs using your injection rate. The input file is attached.

Not sure what the problem is with the lower pressure you used.

…Peter

[Emmanuel LEGER]

thanks Peter,
I'm sure I'll have question during the following days, keep you posted

e

[Karra, Satish]

The temperature will have no affect at the outlet — it will be zero gradient.

Satish

[Karra, Satish]

As Glenn suggested you can also specify FLUX boundary conditions with 50cm2/min at inlet and outlet instead of source/sink. You’ll get the same result as the source sink with 50cm3/min. See attached input deck.

[Emmanuel LEGER]

ok great,
thanks!

e

[Hammond, Glenn E]

But with volumetric fluxes leave the outlet as a pressure (DIRICHLET) BC or mass balance will be problem and cause the code to cut.

 

Glenn