Default Settings of Particle Dynamics

Numerical methods and mathematical models of Elmer
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by kevinarden »

Can you post the working test model sif and mesh?
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

The jump in a single step is nearly the same regardless of the time step (1e-15 to 1e-9) which makes me think it's something about the mesh or spawn conditions.

For example:

Code: Select all

MAIN: 
MAIN: -------------------------------------
MAIN: Time: 1/20000:   1.000E-15
MAIN: -------------------------------------
MAIN: 

ParticleDynamics: -----------------------------------------
ParticleDynamics: Following the path of the particles

 Statistical info on current coordinate values
 Mean:   3.4500000699913172E-002  -6.4860928685100937E-010  -4.8856527129484508E-010
 Abs:    3.4500000699913172E-002   3.9033122218275929E-009   3.7219216500204368E-009
 Var:    4.5625301874860709E-009   4.7380696194091742E-009   4.4414509352605569E-009

 Min:    3.4499991563837581E-002  -8.6450825131470156E-009  -8.6598595886241417E-009
 Max:    3.4500008662648386E-002   8.5945020326808508E-009   8.1028555946225105E-009
 Statistical info on current velocity values
 Mean:   0.0000000000000000        0.0000000000000000        0.0000000000000000     
 Abs:    0.0000000000000000        0.0000000000000000        0.0000000000000000     
 Var:    0.0000000000000000        0.0000000000000000        0.0000000000000000     
 Min:    0.0000000000000000        0.0000000000000000        0.0000000000000000     
 Max:    0.0000000000000000        0.0000000000000000        0.0000000000000000     

ParticleOutputVtu: Saving in VTK XML unstructured format to file: particles.vtu

 Statistical info on current coordinate values
 Mean:   6.2997818918521017E-003  -2.8887769726496839E-003  -7.5848444277543489E-005
 Abs:    1.2591389584284213E-002   6.4008375800768047E-003   9.9370754269542914E-003
 Var:    1.5063945515253769E-002   1.0141367023783468E-002   1.2840787990134817E-002

 Min:   -1.9389048062423943E-002  -2.8829138497424231E-002  -1.9845399150312314E-002
 Max:    3.4500008662647984E-002   2.1189775091820767E-002   2.1009812736142840E-002
 Statistical info on current velocity values
 Mean:  -7.4044194727076865E-002  -4.3445609568104736E-003   2.7900404856991589E-002
 Abs:    8.3251891993981836E-002   8.5805204413712567E-003   4.8344696060490168E-002
 Var:   0.15837987831572073        1.5965178519418165E-002   9.6362636908131433E-002
 Min:  -0.48195943455625362       -5.4163471571248449E-002 -0.14354608453801862     
 Max:    5.9486034454402807E-002   2.6772353932426361E-002  0.29086333977545364     
ParticleDynamics: All done
ParticleDynamics: -----------------------------------------

Last edited by willsprocket on 25 Jun 2022, 01:51, edited 1 time in total.
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

It seems like it's a problem with the mesh being second order. If I switch to 1st order meshing it works as expected. I thought 2nd order would produce much higher quality results, but maybe it isn't stable numerically and a different method should be used to solve it? I would really appreciate hearing thoughts!
Last edited by willsprocket on 25 Jun 2022, 06:05, edited 2 times in total.
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

Also, it would be great if the electric field could be solved steady-state than referenced in a transient simulation and not saved at every step. My simulations are creating a ridiculous number of files of redundant information.
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

False victory. Upon further work, the 1st order mesh doesn't work. The particles still jump in the first timestep.

I've tried the following but there is no change:

Code: Select all

Timestep Distance = Real 1e-4
Max Characteristic Speed = Logical True
Min Timestep Size = Real 1e-15
Max Timestep Size = Real 1e-10 ! Same as simulation setting
Max Timestep Intervals = 25
Will
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

It seems like there are some portions of the mesh that cause the first step to jump. I think the solver thinks the particle isn't in a valid portion of the mesh. The electrostatic field looks smooth and continuous throughout the entire domain.
willsprocket
Posts: 33
Joined: 18 Jun 2022, 05:32
Antispam: Yes

Re: Default Settings of Particle Dynamics

Post by willsprocket »

More testing shows there are regions of the mesh that the particles just don't travel through correctly.

The electric field is as expected everywhere.

Could mesh face orientation be a problem? Could the mesh numbering be all meshed up?

I also get a weird warning when using parallel solve, something about a charged particle boundary condition. The particle motion in some regions is completely non-physical with a partitioned mesh and some particles have the jump on the first step.
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Default Settings of Particle Dynamics

Post by raback »

Hi,

Parallel operation has been tested much less than the serial one. There has not really been that many active users for this and I think most have been content on serial. Perhaps you could test the serial operation 1st.

There are tons of flags for this. It really does not make sense to save the potential field every time step. Just save the "vtu" files once and use the separate particle output routine to save the particles when you need them.

-Peter
Post Reply