non-convergence issue

Hi everyone,

I'm running a 2d radial case in petrasim where I'm heating from the top, bottom and one side. I first tried to define an extra cell and connect all those boundary cells to it and then set a heat flux for that extra cell making it a heater. The reason I decided to input heat flux vs heat rate is the fact that the boundary cells have different areas. However, I realized that it only allows to input a heat rate, not a flux meaning that I probably had to define the same number of extra cells as the number of boundary cells. Anyway, then, instead of using an extra cell, I defined very thin layers on the bottom/top and right side of the model and set them as the heater and input the heat flux to inject heat into each cell proportional to its contact area. However, every time I run it, it fails and noticed the temperature goes to something like 280C. Any idea why this is happening?

Thanks,

Alireza

8replies Oldest first
  • Oldest first
  • Newest first
  • Active threads
  • Popular
  • Alireza,

    Specify the total heat rate (in Watts) into the single heater element that is connected to all the surface element (using AddBound). Even if the surface elements have different sizes, the correct "heat flux" will enter each surface element according to the respective cross-sectional areas assigned to all connections.

    Sorry, I have no time to look at your input files (and I cannot look at any .sim files, as I don't have PetraSim). Make sure you have a reasonable volume, density, and heat capacity for the  heater element and a reasonable nodal distance to prevent this element from overheating.

    Stefan

    Like
  • Thanks Stefan. As you suggested, I went back to add a single extra cell and connected all the top/bottom/right-side cells to this extra cell and also define a heat rate (in watts). Keep in mind there is no heating from the left side of this 2d model (i.e. center of the cylinder). However, when I look at the temperature profile, it seems it does not inject the right amount of energy into cells with respect to their contact area in any particular direction. Specifically, the lower-left corner is heated much faster than the lower-right corner (which should be the other way around). The properties of the single extra cell are as follows.

    Any suggestions to fix this, please? 

    Alireza

    Property Value
    Rock Density (kg/m3) 1.225
    Rock Porosity 0.001
    Permeability (m2): i=j=k 0.00E+00
    Specific Heat (J/kg-C) 1.00E+03
    Pore Compressibility (1/Pa) 1.00E-06
    Wet Heat Conductivity (W/m-C) 2.40E-02
    Grid Volume Modifier 1.00E+50
    Like
  • Alireza,

    Just a quick reply (please appreciate the effort it takes to “reverse engineer” a TOUGH2 model, specifically its geometry. Not knowing how you generated the mesh, i.e., radii of interfaces etc., makes it time-consuming to understand how the model looks like).

    Here are a few observations:

    (1) You set the volume of the heater element to 1E50, which typically indicates that you want to have a constant temperature boundary condition. But you also specified heat rates on the order of 1E50, including negative rates. I don’t know how you calculated all these, but it seems you want to specify a time-dependent Dirichlet boundary condition. I trust (but did not check) that this is all correct.

    (2) Are all ATMO elements supposed to represent an air-filled space within your test cell? If so, note that the thermal conductivity you specified is way too high. Should be on the order of 0.024, as you did for the heater. not 3.1

    (3) Set KDATA=3 so you see the heat (and fluid) fluxes in the output file. This will demonstrate that the heat fluxes are indeed proportional to the interface areas, as expected.

    (4) There is something wrong with your interface areas from the heater element to the test cell. For example, the first 20 connections from element “ 12 1” should have surface areas that increase proportionally to the ratios of the element volumes (which I assume have the same vertical thickness in a given layer). This is clearly not the case, i.e., the innermost element “    1” has a proportionally much higher connection area to the heater element than the next element, and so on, thus explaining the temperature increase in the corner element. Please check.

    In short, the overall approach seems okay - just check the input values for geometry and thermal properties.

    Good luck!

    Stefan

    Like
  • Thanks Stefan, really appreciate it. My model is a vertical vessel where I have some sand inside and the rest is just empty and then closed the cap and heating/cooling from all around. I did correct the surface area as suggested and now the amount of heat going in the model is proportional to the surface area of boundary blocks. 

    1) I have also attached the excel file where I have calculated the heat rate based on the measured temperature profile from the lab and the properties f the heater element. But, when I look at the simulated temperature profile of the heater is looks different and has lots of fluctuations. Why TOUGH cannot regenerate the same temperature profile given the heat rate calculated is right?

    2) I had also a very tiny opening in the cap of the vessel to ensure the pressure does not build-up. So, in the simulation, I have defined a producer operating on a deliverable set at atmospheric pressure. However, in case the pressure inside falls below atmospheric I need to define an injector too to operate based on a maximum pressure of equal to atmospheric. The problem is for the injector I need to define a mass rate and not a maximum pressure. Any suggestions on this as well?

    Thanks,

    Alireza

    Like
  • Alireza,

    I'm sorry, but I don't have the time to review your model in detail. I hope somebody from the TOUGH community can jump in. Regarding Point (2), specify a Dirichlet b.c. element at 1 bar, not a well on deliverability.

    Good luck!

    Stefan

    Like
  • Thanks Stefan. I fixed the temperature fluctuations problem with tuning the numerical setting (changed it to the "rigorous step" (mop 12). With regards to point 2, I cannot specify a Dirichlet b.c. element at 1 bar since I'd like the temperature to change but the pressure.

    Thanks,

    Alireza

    Like
      • Stefan Finsterle
      • Finsterle GeoConsulting
      • Stefan_Finsterle
      • 1 yr ago
      • Reported - view

      Hi Alireza,

       

      Yes, I should have alerted you to MOP(12) - but it seems you know your way around TOUGH2!

      I would attach a single Dirichlet element to one model element where the vent is. The sole purpose of that element would be to control the pressure; give it a thermal conductivity of zero. Air flowing out from the test cell won't impact the temperature in the test cell at all; small amounts of air flowing into the cell won't affect the temperature significantly either, as you control the temperature conductively by your other boundary element.

      Stefan

      Like
  • Thanks Stefan, it worked nicely. I much appreciate your help.

    Alireza

    Like
Like Follow
  • 1 yr agoLast active
  • 8Replies
  • 76Views
  • 3 Following