Björn Kaiser

Yes, the in-/out-transfer rate is the parameter you may in combination with transfer boundary conditions (Cauchy type). The inflow/outflow at transfer boundary conditions is calculated from the relevant area, the transfer rate, and the difference between reference and groundwater head:

    Q = A*?*(href-h)

where
          Q: inflow or outflow to/from the model
          A: relevant area
          ? : transfer rate
          href: reference water level
          h: current hydraulic head in groundwater

The transfer rate is a conductance term describing the properties of a clogging layer. It is defined as

    ? = K/d

          K: hydraulic conductivity of the clogging layer
          d: thickness of the clogging layer

Yes, you can use the phreatic mode to simulate the unsaturated zone if you intend use the standard (saturated) groundwater flow equation. In a phreatic mode, the model stratigraphy is fixed and, as a consequence, elements may become dry or partially saturated. In contrast, to the unsaturated mode (Richards’ equation), the calculation of the unsaturated zone is simplified. Further detailed are provided by the FEFOW manual, page 51:

http://www.mikepoweredbydhi.com/-/media/shared%20content/mike%20by%20dhi/flyers%20and%20pdf/product-documentation/feflow%206.2%20user%20manual.pdf

If you solve the groundwater flow equations you solve for the process variable hydraulic head on each node including the unsaturated zone. If you want to discriminate the unsaturated zone from the saturated zone in space, I suggest to use the zero isoline for the pressure (=0 KPa). The zero isoline for the pressure represents a visual replica for water table undulations.

As long as you integrated the XY-points of the wells in the supermesh to constrain the mesh generator to locate computational nodes of the FE-mesh at these positions, the following workflow should work:

Assure you use a neighborhood relationship for the regionalization method in the Parameter Link Editor. As a first step please choose a snapping distance larger than 0 and smaller than the distance between well locations and neighboring computational nodes (e.g. 0.1 m). After that, select the shape file containing the wells in the Maps Panel by double-clicking. In the Selection Toolbar please choose Select by Map Points and then Select by all Map Geometries.

If the Parameter Link you created is active you should see the name of the Link in the Input field. Finally, please click on apply.

Another option to simulate the propagating excavation is provided by element deactivation. Elements can be deactivated and reactivated during a simulation run. Of course, the deactivation of elements to simulate the digging of a tunnel requires a fully spatially discretized tunnel geometry in the FE-mesh.

From the simulation point of view I suggest to choose a smooth gradation between highly refined areas within and around the wall to coarser areas slightly away from the wall. Under certain conditions, the GridBuilder may fail, especially if line features are present. A failure may be attributed on how GridBuilder internally treats lines within the supermesh to generate the FE-mesh. More details on the GridBuilder algorithm is provided in the FEFLOW Help.

An invalid mesh topology may be attributed to geometrical characteristics in the supermesh, which may inhibit the generation of a mesh consistent for the FEM.

Obviously, the line in your supermesh causes the problem. Indeed, problems can arise as a results of lines containing sharp kinks or lines which confluence at sharp angles.

Moreover, inconsistencies in geometrical topologies of the input shapefile may also cause the problem as Pete already indicated. Therefore, you may check the shapefile according to possible overlapping lines. You may also check if one large line is represented by several line segments (e.g. several line features). In that case, please assure the last vertex of one line spatially converges with the first vertex of the subsequent line. Small spatial deviations in the magnitude of several orders after the decimal operator between the last vertex of one line segment and the vertex of the subsequent line segment may trigger invalid topologies in the FE-mesh.

Apart from the input data, the mesh generators provided by FEFLOW have advantages and disadvantages with respect to robustness and capabilities of handling complex geometrical structures. Accordingly, it is worth to try another mesh generator.

Instead of using a line in the supermesh you could also split the polygon into two sub-polygons along the trace of the line enabling the application of the Advancing Front mesh generator which originally does not consider lines in the supermesh.

Attached please find also the two files I was mentioning in my previous post:

[i]interpolation_at_current_stages.jpg[/i]
[i]interpolation_of_additional_time_stages.jpg[/i]

Here I give you an example about the difference between the [b]interpolation at current stages[/b] and the [b]interpolation of additional time stages[/b].

Let’s assume two gauges along a river. One gauge is located upstream and the other gauge is located further downstream. The upstream gauge measures a flood wave [b]Curve 1[/b] and the river gauge located downstream measures the flood wave [b]Curve 2[/b]. Please note the following attached file to see both curves: [i]Curve1_Curve2.jpg[/i].

If you interpolate between these two hydrographs spatially only, you get a curve which is characterized by two separated flood waves. Please see the attached file [i]interpolation_at_current_stages.jpg[/i]

In contrast, if you take “transient effects” into account you consider the flood wave dampening effect by retention. Please see the attached file [i]interpolation_of_additional_time_stages.jpg[/i]

Of course, both methods solely interpolate mathematically without taking real physical process into account. However, if you [b]interpolate additional time stages[/b] you mimic retention effects in your interpolation as indicated by a single damped flood wave curve.

Observation points can be located at nodes/slices, or they can be freely placed. In the first case, you have to impose the node number/slice number. In the latter case, you have to specify the field containing the z-coordinates, while the field for the node number needs be set to [i]<no node>[/i] and the field for the slice number needs to be set to [i]<free xyz position>[/i].

The process variable value shown for free xyz positions is then derived by interpolation from the corner nodes of the element the point is placed in.

The settings for the both the Boussinesq approximation as well as for the viscosity looks OK. If you deactivate the checkbox [i]Invoke Boussinesq approximation[/i] you automatically adopt the Extended Boussinesq approximation.

Regarding numerical instabilities I would like to draw your attention to a previous discussion:
http://forum.mikebydhi.com/index.php/topic,2263.msg5242.html#msg5242