Re: [sparta-users] Surface and flow mesh resolutions

[Steve P. <sj...@gm...>]

Hi - glad you figured things out.

All the values you have mentioned: Ntouch, Nscoll, Nscheck, etc are
described
on the stats_style doc page.

Ntouch is the number of grid cells "touched" by all the particles during a
timestep.
So if a particle moves from cell A to B to C in a single timestep, it would
"touch" 3 cells.
If Ntouch is large (normalized by number of particles), that means your
timestep is big
and particles are moving a long ways relative to grid cell size.  That
could also be
by Nscheck is large, b/c each time a particle "touches" a cell it has to
check all the
surfs in that cell for a possible collision.  Shrinking the timestep should
also reduce Ntouch.

Steve

On Fri, Mar 24, 2023 at 12:55 AM Zuheyr Alsalihi <zuh...@vk...>
wrote:

> Dear Steve,
>
> Thank you for your kind reply. What I showed was for single time steps.
> Just to help fellow colleagues, I have solved the problem also taking a
> smaller time step, please see the attached screenshot.
> Apparently one has a compromise between Ntouch (I do not know the
> significance) and Nscheck.
>
> Kindest regards,
>
> Zuheyr
>
>
> Dr. Zuheyr Alsalihi
> Senior Research Eng.
> von Karman Institute
> 72, Chee. de Waterloo
> 1640 Rhode-St-Genese, Belgium.
> +32 359 98 65
> +32 473 69 32 27
> ------------------------------
> *From:* Steve Plimpton <sj...@gm...>
> *Sent:* Friday, March 24, 2023 12:28 AM
> *To:* Zuheyr Alsalihi <zuh...@vk...>
> *Cc:* spa...@li... <
> spa...@li...>
> *Subject:* Re: [sparta-users] Surface and flow mesh resolutions
>
> Warning: This is an email sent from outside the von Karman Institute. Do
> not open links or attachments unless you were expecting them and have
> verified they come from a safe source. Contact IT in case of doubts.
> Hi - not sure what your question is.  Are the outputs you are showing for
> a single timestep,
> or cumulative Nscoll Nscheck for a long run?  You could also post your
> input files.
> How many procs are you running on?  When you say "slow", is that a Q or
> comment or what?
>
> Nscheck of 44M with only Nscoll of 400K means you probably have grid cells
> with lots of surface
> elements in them.  That can indicate you need a finer resolution grid.
>
> Steve
>
> On Thu, Mar 16, 2023 at 6:56 AM Zuheyr Alsalihi <zuh...@vk...>
> wrote:
>
> Dear Colleagues,
>
> I am not new to Sparta but I am mostly moving in the dark due to lack of
> time, despite the very valuable community help I get here.
>
> I have a small model of 5mm x 4mm x 0.5mm at high Kn flow but still with
> collisions, density about 1.4e+16 /m^3.
>
> I normally choose the flow mesh resolution from the principles of at least
> around 15 particles in a cell, and particle travel time in the cell,
> usually time step is 1/3rd of the particle travel time. The surface mesh
> usually coarsest mesh with well surface definition.
>
> *My question is the meaning of the  Nscoll Nscheck variables and their
> reasonable orders.*
>
> With 100 x 50 x 30 flow mesh and  31000 element surface mesh:
> Nscoll=401,414, Nscheck=289,249,273, Natt=0, Ncoll=0, 11k particles.
>
> Same flow mesh with a surface mesh of 3500 elements, and same number of
> particles:
> Nscoll=398264  and Nscheck=44803563.
>
> Slow... Can you please share your ideas? This is a very basic question.
>
> Many thanks for reading.
>
> Zuheyr
>
>
>
>
>
>
> Dr. Zuheyr Alsalihi
> Senior Research Eng.
> von Karman Institute
> 72, Chee. de Waterloo
> 1640 Rhode-St-Genese, Belgium.
> +32 359 98 65
> +32 473 69 32 27
> _______________________________________________
> sparta-users mailing list
> spa...@li...
> https://lists.sourceforge.net/lists/listinfo/sparta-users
>
>