Posted Thu, 30 Aug 2007 13:50:21 GMT by Hydrobert
I try to simulate my tank experiments in FEFLOW 5.015. But i can not simulate the right small dispersion (long. dispersivity 0.01 - 0.0001 and trans. dispersivity 0.001 - 0.00001). I will simulate the hydrodynamically stable case of saltwater (up to 100000 mg/l) underneath de-ionized water (1 mg/l), each half of the tank sized. Always the FEFLOW break down for concentration about 10000 mg/l for heterogenous permeability field. For homogenous permeability field the model works up to highest concentrations and minimal dispersivity.

The model grid was set up to simulate the x-z cross-sectional geometry and the experimental stochastic realization of the permeability field of the tank. The mesh is 9.8 m in x-direction and 1.225 m in z-direction in a vertical 2D-problem. Each of the, in x 0.2 m and in z 0.025 m sized, 2401 sand blocks (made out of 8 different permeability classes ranging from 1,3e-04 to 2,8e-02 m/s) is represented numerically by 8 horizontally and 4 vertically directed elements, resulting in a total 38907 nodes, respective 38416 quadrilateral elements. The models are run in time until steady-state conditions for the transport, likewise to the experiments, are reached (steady flow - unsteady mass, combined).

Therefore i have constant head conditions on both sites of the tank for the flow (10 cm = 4 m/d). For the transport i have constant mass at the inflow. Background mass is 1 mg/l. At the outflow there is no mass condition respectively the default free natural boundary. Viscosity incorporated, Bossinesq approximation applied to density coupling extended and the corresponding density ratio was set by me.

[b]The only warning i get from FEFLOW is the size of the model - only 12.5 square meters.[/b] I try different changes in the solver settings and time step setting of FEFLOW, but nothing helps. How can I stabilize the model at the beginning?

I hope someone have an idea ... Thank you very much.
Posted Sat, 01 Sep 2007 13:37:36 GMT by Boris Lyssenko
The discretization might not be fine enough to deal with your extremely low dispersivity values. Please try to refine the mesh, both horizontally and vertically.
Posted Thu, 06 Sep 2007 09:57:27 GMT by Hydrobert
Thank you very much.

I triangulate my mesh and now i can use the adaptive mesh refinement method. The model is stabilized.

You must be signed in to post in this forum.