Re: [DuMuX] BC apparently not respected
Dear Lorenzo, in my opinion it makes no sense to set a Dirichlet and a Neumann condition at the same time. Either you know the flux over your boundary or you know the values of your primary variables on the boundary. If you change your code to setAllNeumann on the left boundary of your domain, do your results change at all (I would assume they do not)? In general, I would use either setAllNeumann or setAllDirichlet on your left boundary. Cheers Beatrix Am 27.03.19 um 3:27 nachm. schrieb Timo Koch: Hi Lorenzo, you still have large capillary pressure gradients at the left boundary at t=600 thus the saturation changes. You also have them in the initial conditions, so they probably never go away. Why this is so, depends on your setup. Does it also occur when your grid is fine from the beginning and you don't have hanging nodes? What are your criterions for Newton convergence? You seem to be in a regime of your p_c-S_w curve where a tiny change in pressure makes a huge difference in saturation. So also numerical error may play a role. As the boundary conditions are only enforced weakly you can get small differences very easily. From a modeling point of view, it's also unclear to me why you need this wavy initial condition? If it's about producing fingers, which should be produced by instability, small numerical pertubations in the boundary conditions should also do the trick. Your initial conditions produces extremely large pressure gradients in y-direction, too. I'm not convinced that this is an issue with Dumux. Best wishes Timo On 27.03.19 14:53, lc wrote: Hello Timo, I enclose 2 solutions, t = 0 and t = 600 days. Indeed, I have a saturation profile as IC as you can see at t = 0.h Best regards, Lorenzo On 27.03.2019 16:41, Timo Koch wrote: Dear Lorenzo, generally for cell-centered discretizations in Dumux, Dirichlet boundary conditions are weakly enforced on the boundary. Your first degree of freedom is half a cell away from the boundary, so it doesn't necessarily have to be the same as the Dirichlet value. In order to see if this is actually a mistake or not, we would need at least a picture of the saturation and pressure fields. If you have some gradients and varying saturation close to the boundary it's perfectly normal that the cell values are different than the boundary values. For example, does you flow field have finger effects close to the boundary? That would explain why the profile in the first cell looks like it does. Best wishes Timo On 27.03.19 14:24, lc wrote: Hello Dumux team, We are, finally, in the post-processing phase, aiming at publishing soon ... but we observed a scaring issue concerning the verification of the BC at the inlet. I'm using Dumux 2.12 with the 2p sequential model for a rectangular domain discretized with a cartesian structured grid with water inflow (q_w = 0.0001446759 kg / m^2 - s). At the inlet I also impose a Dirichlet condition on water saturation, s_w = 0.7856 (smax). The observed issue is visible in the png file (boundary_condition_t=500_kern1.png): the horizontal axis in the figure corresponds to the y axis of the simulation domain (m) and the vertical axis in figure reports the saturation levels, (s_w). We see at t = 500 days from the start of the simulation which runs until 688 days. So, as we can see, the Dirichlet BC at inlet, s_w = 0.7856 is not verified. At t = 0, it is ok, but at every later step not any more, getting a bigger and bigger discrepancy. May you help me to shed some light on this, please? I enclose all necessary files to reproduce such behaviour, in case you have the possiblity to do it. Kind regards, Lorenzo ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux -- ___ Timo Koch phone: +49 711 685 64676 IWS, Universität Stuttgart fax: +49 711 685 60430 Pfaffenwaldring 61email:timo.k...@iws.uni-stuttgart.de D-70569 Stuttgarturl:www.hydrosys.uni-stuttgart.de ___ -- ___ Timo Koch phone: +49 711 685 64676 IWS, Universität Stuttgart fax: +49 711 685 60430 Pfaffenwaldring 61email:timo.k...@iws.uni-stuttgart.de D-70569 Stuttgarturl:www.hydrosys.uni-stuttgart.de ___ ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux -- *** Beatrix Becker Department of Hydromechanics and Modelling of Hydrosystems (LH2) Institute for Modelling
Re: [DuMuX] BC apparently not respected
Hi Lorenzo, you still have large capillary pressure gradients at the left boundary at t=600 thus the saturation changes. You also have them in the initial conditions, so they probably never go away. Why this is so, depends on your setup. Does it also occur when your grid is fine from the beginning and you don't have hanging nodes? What are your criterions for Newton convergence? You seem to be in a regime of your p_c-S_w curve where a tiny change in pressure makes a huge difference in saturation. So also numerical error may play a role. As the boundary conditions are only enforced weakly you can get small differences very easily. From a modeling point of view, it's also unclear to me why you need this wavy initial condition? If it's about producing fingers, which should be produced by instability, small numerical pertubations in the boundary conditions should also do the trick. Your initial conditions produces extremely large pressure gradients in y-direction, too. I'm not convinced that this is an issue with Dumux. Best wishes Timo On 27.03.19 14:53, lc wrote: Hello Timo, I enclose 2 solutions, t = 0 and t = 600 days. Indeed, I have a saturation profile as IC as you can see at t = 0.h Best regards, Lorenzo On 27.03.2019 16:41, Timo Koch wrote: Dear Lorenzo, generally for cell-centered discretizations in Dumux, Dirichlet boundary conditions are weakly enforced on the boundary. Your first degree of freedom is half a cell away from the boundary, so it doesn't necessarily have to be the same as the Dirichlet value. In order to see if this is actually a mistake or not, we would need at least a picture of the saturation and pressure fields. If you have some gradients and varying saturation close to the boundary it's perfectly normal that the cell values are different than the boundary values. For example, does you flow field have finger effects close to the boundary? That would explain why the profile in the first cell looks like it does. Best wishes Timo On 27.03.19 14:24, lc wrote: Hello Dumux team, We are, finally, in the post-processing phase, aiming at publishing soon ... but we observed a scaring issue concerning the verification of the BC at the inlet. I'm using Dumux 2.12 with the 2p sequential model for a rectangular domain discretized with a cartesian structured grid with water inflow (q_w = 0.0001446759 kg / m^2 - s). At the inlet I also impose a Dirichlet condition on water saturation, s_w = 0.7856 (smax). The observed issue is visible in the png file (boundary_condition_t=500_kern1.png): the horizontal axis in the figure corresponds to the y axis of the simulation domain (m) and the vertical axis in figure reports the saturation levels, (s_w). We see at t = 500 days from the start of the simulation which runs until 688 days. So, as we can see, the Dirichlet BC at inlet, s_w = 0.7856 is not verified. At t = 0, it is ok, but at every later step not any more, getting a bigger and bigger discrepancy. May you help me to shed some light on this, please? I enclose all necessary files to reproduce such behaviour, in case you have the possiblity to do it. Kind regards, Lorenzo ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux -- ___ Timo Koch phone: +49 711 685 64676 IWS, Universität Stuttgart fax: +49 711 685 60430 Pfaffenwaldring 61email:timo.k...@iws.uni-stuttgart.de D-70569 Stuttgarturl:www.hydrosys.uni-stuttgart.de ___ -- ___ Timo Koch phone: +49 711 685 64676 IWS, Universität Stuttgart fax: +49 711 685 60430 Pfaffenwaldring 61email: timo.k...@iws.uni-stuttgart.de D-70569 Stuttgarturl: www.hydrosys.uni-stuttgart.de ___ ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
Re: [DuMuX] BC apparently not respected
Dear Lorenzo, generally for cell-centered discretizations in Dumux, Dirichlet boundary conditions are weakly enforced on the boundary. Your first degree of freedom is half a cell away from the boundary, so it doesn't necessarily have to be the same as the Dirichlet value. In order to see if this is actually a mistake or not, we would need at least a picture of the saturation and pressure fields. If you have some gradients and varying saturation close to the boundary it's perfectly normal that the cell values are different than the boundary values. For example, does you flow field have finger effects close to the boundary? That would explain why the profile in the first cell looks like it does. Best wishes Timo On 27.03.19 14:24, lc wrote: Hello Dumux team, We are, finally, in the post-processing phase, aiming at publishing soon ... but we observed a scaring issue concerning the verification of the BC at the inlet. I'm using Dumux 2.12 with the 2p sequential model for a rectangular domain discretized with a cartesian structured grid with water inflow (q_w = 0.0001446759 kg / m^2 - s). At the inlet I also impose a Dirichlet condition on water saturation, s_w = 0.7856 (smax). The observed issue is visible in the png file (boundary_condition_t=500_kern1.png): the horizontal axis in the figure corresponds to the y axis of the simulation domain (m) and the vertical axis in figure reports the saturation levels, (s_w). We see at t = 500 days from the start of the simulation which runs until 688 days. So, as we can see, the Dirichlet BC at inlet, s_w = 0.7856 is not verified. At t = 0, it is ok, but at every later step not any more, getting a bigger and bigger discrepancy. May you help me to shed some light on this, please? I enclose all necessary files to reproduce such behaviour, in case you have the possiblity to do it. Kind regards, Lorenzo ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux -- ___ Timo Koch phone: +49 711 685 64676 IWS, Universität Stuttgart fax: +49 711 685 60430 Pfaffenwaldring 61email: timo.k...@iws.uni-stuttgart.de D-70569 Stuttgarturl: www.hydrosys.uni-stuttgart.de ___ ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
[DuMuX] BC apparently not respected
Hello Dumux team, We are, finally, in the post-processing phase, aiming at publishing soon ... but we observed a scaring issue concerning the verification of the BC at the inlet. I'm using Dumux 2.12 with the 2p sequential model for a rectangular domain discretized with a cartesian structured grid with water inflow (q_w = 0.0001446759 kg / m^2 - s). At the inlet I also impose a Dirichlet condition on water saturation, s_w = 0.7856 (smax). The observed issue is visible in the png file (boundary_condition_t=500_kern1.png): the horizontal axis in the figure corresponds to the y axis of the simulation domain (m) and the vertical axis in figure reports the saturation levels, (s_w). We see at t = 500 days from the start of the simulation which runs until 688 days. So, as we can see, the Dirichlet BC at inlet, s_w = 0.7856 is not verified. At t = 0, it is ok, but at every later step not any more, getting a bigger and bigger discrepancy. May you help me to shed some light on this, please? I enclose all necessary files to reproduce such behaviour, in case you have the possiblity to do it. Kind regards, Lorenzo BUG.tar.gz Description: application/gzip ___ Dumux mailing list Dumux@listserv.uni-stuttgart.de https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
