Closed MFraters closed 8 years ago
ian-r-rose
commented
8 years ago Hi Menno,
This does indeed seem like a bug. I am having a bit of a hard time reproducing the error here -- can you confirm that it only happens when there are free-slip boundary conditions? Also, any chance for a prm file that I can run on my workstation (preferably in 2D)?
Ian
MFraters
commented
8 years ago Hey Ian, Thanks for your quick reply. The 2d case is a lot faster (cookbooks/shell_simple_2d.prm), and also shows the bug. Just change the same things as above (but you need left and right in stead of east,west and south). I forgot to say that in the 3d case you also need to add set Opening angle = 90. In the 2d case I can say that the problem only occurs when the inner boundary is set to tangential, left and right can be either zero or tangential without a problem.
ian-r-rose
commented
8 years ago Hmm, I just tried shell_simple_2d.prm, and it seems to work okay on my machine. Care to attach your prm with the appropriate modifications so we can be 100% sure that we are doing the same thing?
I am on the ASPECT master branch, using deal.ii 8.3.
MFraters
commented
8 years ago I have noticed a difference in behaviour between my laptop (with an oder version of deal.ii) and the cluster with a recent version (june 2015 from github). My laptop does generate the error with the 2d version, but the cluster doesn't. But the 3d version crashed in both cases. The problem is dependent on the resolution which is used, so this is the smallest 3d case I found (of I take a lower adaptive refinement the problem disappears):
# A simple setup for convection in a 3d shell. See the
# manual for more information.
set Dimension = 3
set Use years in output instead of seconds = true
set End time = 1.5e9
set Output directory = output
set Pressure normalization = no
set CFL number = 0.1
subsection Material model
set Model name = simple
subsection Simple model
set Thermal expansion coefficient = 4e-5
set Viscosity = 1e22
end
end
subsection Geometry model
set Model name = spherical shell
subsection Spherical shell
set Opening angle = 90
set Inner radius = 3481000
set Outer radius = 6336000
end
end
subsection Model settings
set Zero velocity boundary indicators =
set Tangential velocity boundary indicators = inner,east,west,south
set Prescribed velocity boundary indicators =
set Fixed temperature boundary indicators = inner, outer
set Free surface boundary indicators = outer
set Include shear heating = false
end
subsection Boundary temperature model
set Model name = spherical constant
subsection Spherical constant
set Inner temperature = 1973
set Outer temperature = 973
end
end
subsection Initial conditions
set Model name = function
subsection Function
set Function expression = 1.473e3
end
end
subsection Gravity model
set Model name = radial earth-like
end
subsection Mesh refinement
set Initial global refinement = 2
set Initial adaptive refinement = 2
set Strategy = temperature
set Time steps between mesh refinement = 15
end
subsection Postprocess
set List of postprocessors = visualization, velocity statistics, temperature statistics, heat flux statistics #, depth average
subsection Visualization
set Output format = vtu
set Time between graphical output = 1e6
set Number of grouped files = 1
end
subsection Depth average
set Time between graphical output = 1.5e6
set Output format = vtu
end
end
subsection Checkpointing
set Steps between checkpoint = 50
end
I am testing the ellipsoidal geometry model, and for this I am testing how it works together with different parts of ASPECT, and especially the free surface. During the testing I stumbled upon a curious bug which seems to be in the constraints part of the free surface code. I have been able to get some constrains on the bug, in the sense that it occurs in combination with at least the spherical shell, chunk and ellipsoidal chunk geometries when using tangential velocity boundary conditions (all the geometries which use manifolds).
The simplest way to reproduce this error is by taking the shell_simple_3d.prm cookbook file and add
set Pressure normalization = no set CFL number = 0.1
and set in the model section:
set Zero velocity boundary indicators = set Tangential velocity boundary indicators = inner,east,south,west set Prescribed velocity boundary indicators = set Free surface boundary indicators = outer
I also had to remove depth average from the list of postprocessors, because it gave some strange error.
The bug shows itself at the beginning of timestep 2:
and in the error output the following:
I am not exactly causes the problem. Could it for example be caused by the decoupling of the manifold by the free surface code? If needed I can also attach input files where the problem occurs for both the chunk and ellipsoidal chunk geometry.
Many thanks!
Menno