[Nek5000-users] What is the accuracy of the particle tracking?

[nek5000-users at lists.mcs.anl.gov]

Dear Alexis,

If I am not mistaken (or my info is outdated), there are two examples 
available with Lagrangian particle integration; one is the hemi case in 
the examples directory, and then there is a description on the homepage 
at https://nek5000.mcs.anl.gov/index.php/Data_processing_example.

Both of these examples work with fluid particles, that is dxp/dt = 
vf(xp). The temporal scheme is AB3 for hemi, and Ab4 on the webpage, 
with the same time step as for the flow.

Azad Noorani at KTH has implemented and validated inertial particles 
(with nonlinear Stokes drag) in comparably complex geometries, see
kth.diva-portal.org/smash/get/diva2:698903/FULLTEXT01.pdf. The 
corresponding code is fairly general, and we could certainly collaborate 
further on that.

We have however not implemented other forces (maybe somebody else did), 
but there are no larger difficulties in doing so. The Basset force might 
be a bit tricky though.

Hope this helps,
Philipp


On 2015-08-07 06:21, nek5000-users at lists.mcs.anl.gov wrote:
> Hello,
>
> I need to track particles in flow field solutions, and I saw that
> NEK5000 has particle tracking capabilities.
> I checked the basic online documentation and searched in the mailing
> list and I was not able to find a full description of the particle
> tracking algorithm.
>
> So my questions are the following?
> 1) What is the time-accuracy of the particle tracking solver? I mean, do
> you use Runge-Kutta 4th order or equivalent?
> I guess that your interpolation in space is very accurate, but
> time-accuracy matters a lot for estimating particle trajectories right.
>
> 2) Do you solve for particle dynamics equations or just for perfect
> tracers?
> I mean, is the code prepared to solve dVp/dt=f(Xp,t) rather than Vp=Up?
> (where Xp is the particle position, Vp is the particle velocity, Up is
> the fluid velocity at the particle position.)
> A simple example is the typical aerosol particle equation:
> dVp/dt=-1/St(Vp-Up),
> where St is the Stokes number.
>
> 3) In my case, I need to solve a bit more complicated equation (See Loth
> and Dorgan (2009) Environ Fluid Mech, 9, pp187-206),
> where "additional" terms such as lift, added mass, fluid stress and
> particuarly 'history' terms are considered.
> Are any of these "additional" terms considered in your current
> implementation?
> If not, is the code flexible enough to implement these additional terms
> using the current code structure?
>
> Many thanks.
> Alexis Espinosa-Gayosso
> University of Western Australia
> PS. It is my first time using this list, and I did not know how to set a
> particular thread or even If this is the right email address to write.
> PS2. I did not know either how to post to an existing thread. Can you
> tell me how?
>