3D modeling with tomography data

7 years 1 month ago #46 by Montoya

At this Moment we are working in the modeling of migration experiments of redox sensitive species in fracture media with iCp.
Our COMSOL model is based on a 3D mesh (surface of the fracture and an interpolation function to map the apertures onto this surface/mesh) without using any geometry, then we cannot select or define domains.
In all the ICP examples we have there is always a domain which has to be selected and referenced in the JSON file. This is not possible for us and we do not have a solution to this.

Thus, we wanted to ask, if we can at all use our COMSOL model/approach with the ICP?

Thank you!
Best regards

Please Log in or Create an account to join the conversation.

7 years 1 month ago #47 by JLP
Hi Montoya,
I heard (but have not tested it) that version 5.0 of Comsol should have the ability to convert imported meshes into a geometry. This would imply, that once iCP is updated for its use with Comsol 5.0 you should be fine with it. (This might happen eventually by end of this year?).

Cheers, Johanna Lippmann-Pipke

Please Log in or Create an account to join the conversation.

7 years 1 month ago #48 by Jorge.Molinero
Hi Vanessa and Johanna,
I was about replying the same thing than Johanna. Check the geometry and mesh options at: www.comsol.com/release/5.0 (there is one video explaining).
If your problem is homogeneous in chemistry (all the fracture is one chemical domain) that should work. If you need to incorporate full heterogeneous chemistry (a different chemical system for each element in the model), then the current version of iCP is not prepared for that (Tis is one of the main developments we have for the next version).

Please Log in or Create an account to join the conversation.

7 years 1 month ago - 7 years 1 month ago #49 by Flo
Hi all,
Thanks for the answer!

yes, our domain is homogeneous. Our model is based on the idea of using an aperture field mapped directly on a 3D surface (=mesh; midplane of our fracture) and then solving the cubic law and solute transport (similar (but without geometry) to the rock fracture flow example in the ICP workshop). That is, although our model is in 3D, it is just a surface (without any thickness/volume, ...). Thus, we can not select a domain, only boundaries and edges. We will upload the model later if you want to have a look at it?!

all the best,
Last edit: 7 years 1 month ago by Flo.

Please Log in or Create an account to join the conversation.

7 years 1 month ago #51 by osilva
Hi Florian,

Effectively, as you said, if you select to build a 3D model, the molal solute transport physics needs to be defined in 3D. According to our understanding, to solve your problem there are two options:

1. To define the problem as 2D both for flow and transport and work with porosity and transmissivity based on the fracture aperture distribution.

2. To define the problem as 3D by including a small portion of "ficticious" matrix ( the matrix being unreactive with almost zero porosity and permeabilitty). Then you have a 3D domain where you can define your chemistry.

Orlando et al.

Please Log in or Create an account to join the conversation.

6 years 11 months ago #59 by Flo
Dear Orlando,

We decided to go for the 2D model at the moment and were successful in setting up the COMSOL model. The model solves the Navier-Stokes flow and uses the Molal Solute transport module needed for ICP.
(The option of the cubic law flow field is still pending due to some problems with the transformation of the fracture geometry and aperture distribution from 3D to 2D, but we are working on this)
The actual model so far has ~45000 number of elements and solves the conservative tracer transport in around 20 min (in COMSOL only; 8 core machine and ~2 Gb of memory needed). Our next step was to setup the ICP and try to solve the conservative tracer again
with ICP and compare to the COMSOL result. We took the ICP example of the 1D Cl- transport as a basis for our 2D model and changed things accordingly. We observed that the simulation really takes a VERY VERY long time to solve
when using ICP instead of using only COMSOL. We are at the moment not sure if this is due to the ICP input file or if everthing is alright and it just takes this long (which would be really a pitty!). It may be likely the case that the parameters
stated in the json input file are not the best ones and can be improved to speed up the whole simulation. For this, we attached our json ICP input file. The COMSOL file will be made available to you as download (link will be send via email to you)!

If you have any questions, please contact use anytime!
Thanks a lot in advance!

Happy Christmas! :P
flo and vanessa

Please Log in or Create an account to join the conversation.

Time to create page: 0.238 seconds
Powered by Kunena Forum