Hi, Just an update on this matter.
I implemented a thermal gradient, where before I had a constant temperature throughout the domain. This pretty much solved the problem, although I don’t fully understand why and the velocities are a still a bit different from what I expect (and intend) them to be. Presumably, there is still some difference between the definition of density/viscosity etc at the boundary compared to how its calculated in the model. Anyways, for now this will do. I guess I will have to visit Stuttgart again to work out the last details :). Cheers, Gilian From: Schout, G. (Gilian) Sent: 15 August 2018 11:32 To: 'Timo Koch' <[email protected]>; DuMuX User Forum <[email protected]> Subject: RE: [DuMuX] problem inducing strictly lateral groundwater flow in 2p2c model Hey Timo, Thanks for helping out. I have tried your suggestions but unfortunately it doesn’t seem to change the strange velocity fields I get. In the examples below I have Dirichlet conditions on the lateral boundaries only between 10 < y < 50 (domain is 60 m high in total) to avoid the corner vertices. Also, I removed the outflow boundary and changed all the other boundaries to a solDependentNeumann (zero flux). That gives the following result: [cid:[email protected]] Which is still relatively ok, but if I for example increase the overall pressure in the domain, conceptually moving the model to a depth of 240 m while keeping hydrostatic pressure, I get the following: [cid:[email protected]] The fact that increasing pressures makes the problem worse makes me believe that the problem is related to the implementation of the density/pressure on the boundaries, but looking at the results in Paraview the pressure and density distribution don’t appear to me as they should be producing these velocities. I also get weird kind of convection cells when I completely remove the lateral head gradient, but keep the dirichlet conditions. In earlier communications with Dennis about the same problem is sent him the results of 2 simulations through wetransfer (https://we.tl/OxogWezIhV ), maybe they can give you some clue as to the solution. With regards to your question, I would like to have the groundwater table / atmosphere at the top of my domain in most scenarios. So, ideally, I would have an unsaturated zone at the top of the model as well. I noticed one test problem (3p3c/infiltrationproblem.hh) which has the same, but haven’t been able to implement that so far. That’s another subject for another day though ;) Kind regards, Gilian From: Timo Koch [mailto:[email protected]] Sent: 14 August 2018 20:46 To: DuMuX User Forum <[email protected]<mailto:[email protected]>> Cc: Schout, G. (Gilian) <[email protected]<mailto:[email protected]>> Subject: Re: [DuMuX] problem inducing strictly lateral groundwater flow in 2p2c model Hi Gilian, I’m just guessing what the problem could be.. Have you tried not to set Dirichlet boundary conditions for the corner vertices? This can be tricky with the box method and its dual grid as you can’t control the flow over the sub control volume face adjacent to the corner vertex on the bottom and the top boundary if the corner is set to Dirichlet. This may cause the water to escape this weirdly through that single scvf. Also, the outflow boundary condition only works, if you’re flow direction is always strictly outwards. So in the beginning it might not work out that well. You can use solDependentNeumann to weakly enforce any boundary condition, like a Robin/Cauchy-type. What’s on top of your domain conceptually? Best wishes Timo On 13. Aug 2018, at 17:27, Schout, G. (Gilian) <[email protected]<mailto:[email protected]>> wrote: Dear dumux-team, I hope you can help me solve a problem in my models that I’ve been trying to figure out for some time now without success. Conceptually, I have been modelling leakage of natural gas into a fully saturated unconsolidated sedimentary aquifer. In the most basic version I use a 2D, 2p2c model where methane enters the from the base of a fully saturated aquifer. The domain is 120 by 60 m, and the injection occurs by way of a Neumann boundary between x=30 and x=31 m. Besides this I have no flow boundaries at the top and bottom, with an outflow boundary at the top for methane so that it can leave the domain once it reaches the top (due to buoyancy). Then, to induce lateral flow, I use a hydrostatic pressure distribution with a minor increase in pressure towards the left of the domain, such that water flows to the right. This distribution I have assigned as both the initial pressure and at sides as dirichlet boundaries. The code can be seen on my GitLab of this version of the model here<https://git.iws.uni-stuttgart.de/gilianschout/CH4Leak/blob/master/2p2c/leak2p2c.hh>. Now the problem is that however I try to implement the lateral boundaries, I always get unexpected velocity vectors around the corners of the domain as you can see in this figure: <image002.jpg> Now this is not even that bad - in the middle of the domain the velocity profile is as I would like it to be. However, when I increase the absolute pressure in the domain, it becomes much worse. For example, here is the same image for a depth of 240 m. <image005.jpg> This obviously influences the results of the simulation significantly. I have been trying different solutions the last couple of days but haven’t been able to put my finger on it. Presumably, there is some discrepancy between the model calculated pressures and the pressures/densities set at the boundaries. My hope is that someone in the team has perhaps run into similar issues and has a solution. The goal is obviously to have strictly lateral groundwater flow, without a vertical component (unless induced by the gas injection at the bottom). Thanks in advance, kind regards, Gilian Schout | PhD Candidate | Faculty of Geosciences | Utrecht University | +31 6 16523607 | [email protected]<mailto:[email protected]> _______________________________________________ Dumux mailing list [email protected]<mailto:[email protected]> https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
_______________________________________________ Dumux mailing list [email protected] https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
