0

# representing an oven using extra cell

Hi everyone,

I’m trying to use TOUGH2 to history match a column test where some soil was put inside a glass column and sealed and then put the column itself inside an oven set at some elevated temperatures. The oven and column's temperatures were recorded. Now, I have defined a mesh to represent the column and the soil inside. However, I wonder what the best way is to represent the oven and it heating rate? I thought I would define a very thin layer around entire the mesh set at oven's temperature to represent it, but this would lead to a very large number of gridblocks and potentially long simulation run time. Then, I said it might be best to represent the oven using an "extra cell" or "dummy cell" connected to all the actual side gridblocks in the domain as in the petrasim's manual. However,  the problem is how to define the extra cell connections given it is supposed to be connected to all the boundary gridblock in the domain. Any suggestions would be greatly appreciated.

Thanks,

9replies Oldest first
• Oldest first
• Popular
• Alireza,

I hope you have good reasons to use a full 3D model to simulate and match your experimental data. If your test is axial, a 1D model would suffice; if your test is radial (e.g., heating from the outside), a 2D, radially symmetric model (easily generated by TOUGH2's MESHM option RZ2D) would suffice. Only if your data and processes are truly 3D would you need the model you show in your figure.

Good luck,

Stefan

Like
• Alireza
• Alireza
• 3 yrs ago
• Reported - view

Hi Stefan,

I just need clarification with regards to 2D RZ axisymmetric model. Let say I have an injection rate of a fluid into a well in a reservoir. Now I would like to model that injection well using a 2D R-Z mesh. As the 2D radially symmetric mesh is only a 2D slice, my question is: Do I need to scale down the injection rate in a 2D R-Z mesh or should I inject the full rate the same as in the actual well? Are the results are reported for a full cylinder (doughnut-shaped cells around the center of the model) or just for that 2D slice?

Thanks again,

Alireza

Like
• Alireza It's visually and computationally just a 2D slice (i.e. there is not point to run a 3D calculation or visualize the result in 3D since it is symmetric along the theta coordinate).

When you create a radial mesh, element volumes and interface area of connections are scaled to simulate a 3D symmetric vertical cylinder. You do not need to scale down the injection rate.

Like
• Alireza
• Alireza
• 3 yrs ago
• Reported - view

Keurfon Luu

Sure, thanks a lot.

Like
• Thanks Stefan, much appreciated. To physically represent the actual experiment given I had multiple sensors inserted into the column, I decided to use a 3D domain. However, I could still use a quarter of the domain instead of the entire volume assuming a perfect homogeneity in the material inside the column.

Thanks again,

Alireza

Like
• Stefan Finsterle
• Finsterle GeoConsulting
• Stefan_Finsterle
• 3 yrs ago
• Reported - view

If the material is perfectly homogeneous, and the boundary conditions for heat and fluid flow are axial and/or radial (as they seem to be, given that you consider reducing the problem to a quarter), I think you could get away with a radial RZ mesh, because the sensor response does not depend on the "azimuth", just on the location along the core and the radial distance from the center of the column.

Stefan

Like
• Thanks stefan. I just need clarification with regards to 2D RZ axisymmetric model. Let say I have an injection rate of a fluid into a well in a reservoir. Now I would like to model that injection well using a 2D R-Z mesh. As the 2D radially symmetric mesh is only a 2D slice, my question is: Do I need to scale down the injection rate in a 2D R-Z mesh or should I inject the full rate the same as in the actual well? Are the results are reported for a full cylinder (doughnut-shaped cells around the center of the model) or just for that 2D slice?

Thanks again,

Alireza

Like
• Stefan Finsterle
• Finsterle GeoConsulting
• Stefan_Finsterle
• 3 yrs ago
• Reported - view

Alireza

The model is 2D radial, i.e., it is not a 2D "slice". Element volumes increase with the radius squared. In radial direction, the mesh consists of a series of doughnuts with increasing size, with the hole of the bigger one having the radius of the previous, smaller doughnut; and the doughnuts are stacked in Z direction. This means you inject the full amount into the center element (which is not a doughnut, but a cylinder; see MESHM block and start with a RADII of zero)  - no down-scaling needed!

Think of it like solving an analytical well solution in an RZ rather than XYZ coordinate system; in TOUGH2, this is accomplished not by reformulating the governing equations, but simply by adjusting the geometry of the elements.

Hope this is clear,

Stefan

Like
• Alireza
• Alireza
• 3 yrs ago
• Reported - view

Stefan Finsterle

Thanks Stefan, your explanation is absolutely clear, much appreciated.

Alireza

Like
Like Follow
• 3 yrs agoLast active
• 9Replies
• 85Views
• 3 Following