Peter Schätzl Grundwassermodellierer

Most probably at one or several locations the vertices of the lines are very close to each other or close to a polygon vertex (or point vertex). Triangle then tries to nearly infinitely refine the mesh to meet the vertices, potentially leading to an overflow (or using all RAM you have and then crashing). Make sure to snap them together where vertices are very close.

Jimmy,
if the two wells are exactly above each other, you should make sure that in between you have at least one or two extra slices. Otherwise one well will always immediately influence the other one, leading to non-convergence.
Peter

If I interpret the help file of OpenLoop version 1.6 correctly, this is possible:

"2. A nodal reference distribution with the name 'OpenLoop' is created to identify participating nodes for a temperature differential. For single-species mass transport, the distribution is called 'OpenLoop Mass', for multi-species transport 'OpenLoop Mass 1' for species 1, 'OpenLoop Mass 2' for species 2, etc..In these reference distributions, all nodes belonging to a specific group (injection and extraction nodes) are assigned a value equal to the number of the time-varying power function containing the corresponding temperature or concentration differential. All other nodes need to be given values less than or equal to 0 (default in FEFLOW is -99999 = nodata). Be careful with mesh refinement: In case of refining the mesh after assigning the power function number, the data interpolation will lead to a number of nodes with non-zero values around the wells, too. "

For heat transport, this would then be OpenLoop Heat 1 etc. The citation is from the last version that we have very recently received from DHI WASY.

****************  EDIT *******************

I apologize, I was completely wrong here, misreading the help file - now I checked again. The OpenLoop Mass 1 etc. are for multiple species, not for multiple groups of wells. Groups of wells with separate temperature differentials are simply assigned different values in the User Data distribution, corresponding to the time series id to be used for the temperature differential. So if the temperature differential of the first pair of wells is in time series 50, the two (or more) well nodes get the value 50 in the User Data Distribution OpenLoop. Assuming the second pair has time series 60, those nodes need to get 60 as their value.


****************  EDIT *******************

There is an automatic mesh smoothing option in FEFLOW that may help to lower the number of Delauney criterion violations. When generating the mesh, Triangle has an option to produce Delauney-complient mehes. However, triangles violating the criterion may point towards a non-ideal mesh geometry, but they will not inhibit FEFLOW from running well. So I would typically not do a manual correction.

You can find a tutorial for using IFM in the FEFLOW help: http://www.feflow.info/html/help74/feflow/13_Programming/tutorial_ifm.html
The use of the Python interface is also described in the help system. Note that FEFLOW 7.0 did not have a python interpreter in the FEFLOW GUI, but you could use Python via the command-line interface or by loading FEFLOW as a module in your own python code (I'd recommend the latter). The help also contains some examples.

Yes, this can be done. You will, however, by setting source/sink define what the discharge from each of the pipes is. By setting a well in the center the discharge is distributed by the model according to acutal pipe lenghtes, flow field, etc.

Yes, indeed, but even internally in FEFLOW for the simulation run a multi-layer well is a combination of a well bc (on the lowest node) plus a discrete feature representing the well pipe along the screen (only the input/data storage is different). Thus you can mimick a multi-layer well by setting this combination yourself rather than using a multi-layer well. For the Ranney well, you can use a well bc at the bottom of the center shaft, and discrete features for the collector pipes. I'd typically use the Hagen-Poiseuille law (laminar flow) for the collector pipes, and from the FEFLOW book you can figure out what their properties would have to be for a circular pipe with a certain diameter.

When having tetrahedral parts in the model domain, you will need to do the selection as a basis for the multi-layer well in a 3D view. I'd probably use discrete features that I set myself plus a well bc rather than multi-layer wells in such a case.

You're welcome :)

I don't have much experience with the DNC, especially when it comes to particle tracking - so I do not know whether they follow the DNCs and the DFs. If this has not been explicitly added to the particle tracking scheme, I would expect that the particles in this case do not see the DF/DNC velocities.