Problem with zoned earth dams

I'm working with FEFLOW 5.4 for my masterthesis and I want to calculate zoned earth dams with a constant Head on the upside of the dam and a free seepage face on the downside (unsaturated condition, steady, flow only). I don't have any problems to simulate homogenious dams.

When I want to simulate a zoned earth dam, I'm not able to reach a convergence. The error is always bigger than something like 6.7E-2.
When I then start a Re-Run, the results are not the same. Anytime I do a Re-Run the program calculates a different result. I've already tried to enrich the mesh but there's no change of the program-behavior.
Then I found the SAMG solver package in the manuals, which should reach a convergence for problems where the PCG-solver has difficulties. But when I do the simulation, FEFLOW calculates just nonsense. There's a constant hydraulic head on the whole dam and absolutely no water flux.

Then I found the example in the FEFLOW folders "dam6.fem". But when I start the calculations there are the same problems:
a) The program isn't able to reach a convergence for a error tolerance 1E-3 (which is the default value)
b) Every Re-Run calculates different results.

I don't think that this is normal. What am I doing wrong? Or is there just something that I don't understand?

I've just found the option "AMR". Is it possible that this options solves my problem?

AND: I don't realy now what the "Storativity" and the "Storage copressibility" are supposted to be. Could you explain what the parameters describe and which values I should use?

I recommend to use the search function of the forum. You will - for example find the following task: http://feflow.info/forum/index.php/topic,656.0.html.

Yes, with the search function I can answer the question about Storytivity and Storage compressibility. But I can't find any answer for the first question. Why can't I/FEFLOW produce a convergence?

1. The results are not the same because FEFLOW always uses the previous result as the new initial conditions for the subsequent run if the model is not re-loaded in between.
2. Steady-state convergence in unsaturated models highly depends on the parametric relationships and their respective parameters. Steep curves may cause difficulties in convergence. Mesh refinement may help, but does not necessarily have to. I'd recommend to run in transient for a long time to get an indication where the model is not stable.
3. AMR probably will not help.

Thank you for your answer. I will try to run a transient simulation!

I've just run a transient simulation on a very simple example dam with a less permeable core than the rest of the dam. The model is not stable on the landside-border of the core and the "sustain body". But what can I do?

Another question: I still don't understand why the example dam6.fem isn't converging? I have not changed anything from the original file.