[aspect-devel] Problem with active tracers & compositional fields
John Naliboff
jbnaliboff at ucdavis.edu
Fri May 18 15:12:41 PDT 2018
Hi all,
Apologies, the last few emails I accidentally sent directly to Felipe
rather than the whole list. See the messages below for a series of replies.
Felipe - yes, if you are using active tracers then they need to be
everywhere in the domain. If you are modeling a dense layer in the lower
mantle, then the particle values will determine where that interface is.
Cheers,
John
On 05/18/2018 03:04 PM, FELIPE ORELLANA ROVIROSA wrote:
> John,
>
> No ! My tracers are purposely located on a subset of the
> domain, on a layer at the bottom in this particular case.
>
> is this a problem?
>
> Felipe
>
> On Fri, May 18, 2018 at 3:01 PM, John Naliboff <jbnaliboff at ucdavis.edu
> <mailto:jbnaliboff at ucdavis.edu>> wrote:
>
> Hi Felipe,
>
> I had filled a lower layer of the domain with 10k, 20k..200k, and
> there were plenty of tracers..
> But there are still particles throughout the mesh, yes? Not just
> in the lower layer of the domain?
>
> So, maybe other ways to generate the tracer particles are more
> secure with regards to the mesh.
> Yes, if the ascii file is not working I would try one of the other
> particle generator options ...
>
> Cheers,
> John
> On 05/18/2018 02:58 PM, FELIPE ORELLANA ROVIROSA wrote:
>> Hi John,
>>
>> Thanks a lot for your help.
>>
>> I added the lines
>>
>> set Load balancing strategy = remove and add particles
>> set Minimum tracers per cell = 10
>> set Maximum tracers per cell = 20
>>
>> and tried having up to 200k tracers; but it fails by
>> segmentation fault, again with the same message:
>> #######################
>> The name and call sequence of the exception was:
>> ExcMessage("A cell and all of its neighbors do not contain
>> any particles. " "The 'cell average' interpolation scheme does
>> not support this case.")
>> Additional Information:
>> A cell and all of its neighbors do not contain any particles. The
>> 'cell average' interpolation scheme does not support this case.
>> --------------------------------------------------------
>>
>> Aborting!
>> ######################
>> I am pretty sure this is geometrically not tue, as I have
>> seen the mesh.
>> For example, when the mesh starts, there are like 10^3 active
>> cells, thus, ~10 nodes per side. I had filled a lower layer of
>> the domain with 10k, 20k..200k, and there were plenty of tracers..
>>
>> So, maybe other ways to generate the tracer particles are
>> more secure with regards to the mesh.. Certainly, I tried to
>> implement the uniform box, but I could not make the syntax..I
>> found no examples. In the end generating my own tracers and using
>> a file worked perfect for me (I did this for passive tracers).
>> The thing is, I need my tracers to be located in well-defined
>> regions of my model box.
>>
>> cheers,
>>
>> Felipe
>>
>>
>>
>> On Fri, May 18, 2018 at 2:26 PM, John Naliboff
>> <jbnaliboff at ucdavis.edu <mailto:jbnaliboff at ucdavis.edu>> wrote:
>>
>> Hi Felipe,
>>
>> You are generating the initial particle positions from an
>> ASCII file, so hard to know if in fact there is a particle in
>> each cell. ASPECT seems to think not.
>>
>> There are particle generator schemes that can ensure you
>> always have a particle in each cell, but another option is to
>> enforce a minimum (and maximum) number of particles per cell.
>> These parameters are defined in particle 'world':
>> https://github.com/geodynamics/aspect/blob/master/source/particle/world.cc#L748
>> <https://github.com/geodynamics/aspect/blob/master/source/particle/world.cc#L748>
>>
>> In the parameter file, enforcing a minimum (ex: 10) and
>> maximum (ex: 20)number of particles per cell in the particle
>> subsection would look like:
>> set Load balancing strategy = remove and add particles
>> set Minimum number of particles per cell = 10
>> set Maximum number of particles per cell = 20
>>
>> I think you may need to do at least one refinement during the
>> initial time step to have this enforced?
>>
>> Cheers,
>> John
>> On 05/18/2018 02:09 PM, FELIPE ORELLANA ROVIROSA wrote:
>>>
>>> Hi all,
>>>
>>> I am implementing a 3D model with one layer of
>>> compositional fields, I want the convecting system to advect
>>> active tracers (they have a density differential).
>>>
>>> The model seems to have correct syntax, as it starts
>>> to run. But it fails, it fails by segmentation fault: The
>>> last liens of the log file state that there some cells and
>>> all of their neighbors do not contain any particles, and
>>> that makes the interpolation to crash.
>>>
>>> But, I have been increasing the number of tracer
>>> particles, all the way up to 10^5. I have plotted them, and
>>> they would be waaay more than enough to fill all the cells
>>> of the model, at least in its early stages.
>>>
>>> what is going on here?
>>>
>>> I am pasting the pertinent lines of my prm file:
>>>
>>> subsection Postprocess
>>> set List of postprocessors = visualization, temperature
>>> statistics, velocity statistics, composition statistics,
>>> tracers #,particle count statistics, dynamic topography
>>>
>>> # subsection Dynamic Topography
>>> # set Density above = 1000
>>> # end
>>>
>>> subsection Visualization
>>> # set List of output variables = dynamic topography
>>> set List of output variables = density
>>> set Time between graphical output = 3.1558e14 # 10 Ma
>>> set Output format = vtu
>>> set Number of grouped files = 1
>>>
>>> # subsection Dynamic Topography
>>> # set Density above = 1000
>>> # end
>>>
>>> end #end Visualization
>>>
>>> subsection Tracers
>>> set Number of tracers = 10000
>>> set List of tracer properties = initial position,
>>> initial composition, velocity, pT path
>>> set Interpolation scheme = cell average
>>> set Update ghost particles = true
>>> set Time between data output = 3.1558e14
>>> set Data output format = vtu
>>> set Particle generator name = ascii file
>>> # set Number of grouped files = 1
>>> #
>>> subsection Generator
>>> subsection Ascii file
>>> set Data directory = ./
>>> set Data file name = MyTracers_100k_lower.txt
>>> end
>>> end
>>>
>>> end
>>>
>>> end #end Postprocess
>>>
>>> subsection Compositional fields
>>> set Number of fields = 1
>>> set Names of fields = lower #, upper
>>> set Compositional field methods = particles #, particles
>>> set Mapped particle properties = lower:initial lower #,
>>> upper:initial upper
>>> end
>>>
>>> This is a fraction of the output log:
>>>
>>> Number of active cells: 4,096 (on 5 levels)
>>> Number of degrees of freedom: 184,598
>>> (107,811+4,913+35,937+35,937)
>>>
>>> *** Timestep 0: t=0 seconds
>>> Solving temperature system... 0 iterations.
>>> [compute-0-3:29728] *** Process received signal ***
>>> [compute-0-3:29728] Signal: Segmentation fault (11)
>>> [compute-0-3:29728] Signal code: Address not mapped (1)
>>> [compute-0-3:29728] Failing at address: 0xffffffff000186a0
>>> [compute-0-4:28813] *** Process received signal ***
>>> [compute-0-25:06630] *** Process received signal ***
>>> [compute-0-25:06630] Signal: Segmentation fault (11)
>>> ...
>>>
>>> ...
>>> The violated condition was:
>>> non_empty_neighbors != 0
>>> The name and call sequence of the exception was:
>>> ExcMessage("A cell and all of its neighbors do not
>>> contain any particles. " "The 'cell average' interpolation
>>> scheme does not support this case.")
>>> Additional Information:
>>> A cell and all of its neighbors do not contain any
>>> particles. The 'cell average' interpolation scheme does not
>>> support this case.
>>> --------------------------------------------------------
>>>
>>> Aborting!
>>> ##################
>>> cheers,
>>> Felipe
>>>
>>>
>>> _______________________________________________
>>> Aspect-devel mailing list
>>> Aspect-devel at geodynamics.org
>>> <mailto:Aspect-devel at geodynamics.org>
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>> <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20180518/3941a8d9/attachment-0001.html>
More information about the Aspect-devel
mailing list