Maik Posted May 11, 2009 Report Share Posted May 11, 2009 Hello I am using the FEMWATER modul. In my model there are layers which have in some part a vertical extent of up to 10 m. But in the rest of the model the layer has only a minimum vertical extent of 0.001 m. Running the model the water table reacts to variations of the parameters strongly in the part where the layer has a real vertical extent. In the rest of the model the water table reacts very weakly. Rebuilding the model with an expanded minimum vertical extent of 2 m, the model reacts strongly in the whole model area. Lowering the vertical extent to for example 0.5 m the model reacts less except for the part where the layer has a real vertical extent. My problem is that I need a very low vertical extent of such layers. And the model shall react to variations of parameters in the whole model region. Has someone made similar experiences with GMS FEMWATER? Has somebody an idea how to overcome this problem? Thank you, Maik Quote Link to comment Share on other sites More sharing options...
woodward Posted May 11, 2009 Report Share Posted May 11, 2009 Hi Maik Great to hear from you. We are doing some work with GMS:FEMWATER too on a small site with strong stratification (although conductivities don't vary so much) and a variable thin layer. My colleague told me that he got some issues with convergance with thin layers or pinchouts. He has also been doing some inverse modelling with the same model. I passed your query on to him. The question that occurred to me was: but does it *make sense* that the water table should be more responsive in the thin layers? Would it be better to let the vertical extent go to zero? (Not sure how FEMWATER would like this tho). Simon Quote Link to comment Share on other sites More sharing options...
cnr3 Posted May 12, 2009 Report Share Posted May 12, 2009 Hi Maik I’m not sure I correctly understood your question. I also find it is harder to explain this without a graphical aid, but I’ll try. Try having a reasonable thickness for the whole layer – I guess you can have different thickness at different locations. Then I assume you would have a few computational layers within this layer. After you define the material properties during the 3D mesh construction, go to the 3D mesh and select the elements from the area where thickness is actually less, and change it to the material property as same as the upper/lower layer as appropriate. As you’ve a number of computational layers, you can change a few. This is tedious work and wouldn’t give you very smooth transition in material properties, but one way to overcome the numerical instabilities. Hope this would help Quote Link to comment Share on other sites More sharing options...
Maik Posted May 13, 2009 Author Report Share Posted May 13, 2009 Hi Simon and cnr3, @ Simon: Thank you for passing my query. @cnr3: Thank you for your response. This time I send a 2D mesh and a cross section of my model. The locations with the large vertical extents are tailings. I tried to do as you told me. I changed the material outside the tailings to the material 'Mutterboden' for all layers. This can be done quite comfortable because by draging several elements the elements in all layers are marked and the material can be changed for them. As victim for this I loose my lowest layer. Leaving the vertical extent of the layers which pinch out at 0.001 m the model still doesn't react outside the tailings. Changing the material inside the tailings is practically not possible since you have to select every single element with one mouseclick, for every sublayer! So I decided now to increase the vertical extent of the pinched out layers to 0.5 m although it is not correct. So the model reacts everywhere, but still it reacts most in the tailings. I also send a cross section of the new model. If you or someone else has a better idea, let me know. Thanks Maik 1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.