Hi, thanks for your answers. So your boundary particles do not take part in the pressure gradient term? But they do in the viscous acceleration term and in the mass conservation equation... I have ran the same case and the dynamics are correct too, but the pressure field not very pretty at all lol (I have posted another similar case, but with fluid column size L=2H).
Hi yonarw, very nice animation! what are the boundary conditions at the solid boundary? are you using dummy particles? I think there is something wrong on your left wall... particles should not leave the boundary (at 0.02 sec there is a gap between the fluid column and the wall). Aside from that... fantastic!
Do you have pressure contours for this simulation?
@maettino hi and thanks for your comment. as boundary i used a simple position and velocity "reset" (setting positions to boundary and inverting velocity vectors )
it might look wrong but i did not add any static attraction forces between particles and walls so nothing sticks particles to walls.
i tried to draw pressure values of the particles but i never got any good looking image ;)
Good simulation!. But I want to ask about the strange initial behavior, it's like a atraction force bretween particles in the begining of the simulation.Interestingly, I got a similar behavior due to a miscalculation of the initial density using the Tait equation for pressure
@jahkr thanks. i think it's only looking strange but physics are correct. the problem is that particles are initialized in an grid which then brings up these gaps because they are not initialized close enough to each other ...
How do you find a neighbour particles for pressure forces and color functions? Do you use a hash table indexing?
When i use hash tables (i mean i give a structured 3D grid in R^3 and then use it to store particle numbers that are near) on CPU i get 0.01 fps for 118k particles =)
I;m just interested cos i had my doctorate in CFD (traditional mesh methods) a long time ago and now i'm investigating some mesh free methods for free surfaces. SPH seems to be the best.Thanks!
@NoctumDeVir hmm i'm using a grid in this simulation too. it helped a bit to store neighbors for the next frame and check whether their distance is larger than the interactionradius in this next frame after the move and then only check particles for neighbors when they haven't kept their "neighbor link" from the last frame.further it helped to not use any dynamic datatypes. hope this will help somehow. thanks for your interest
@NoctumDeVir well its not wrong .. its doing as expected with given values ... but i agree it "looks" wrong ;) should have set the surface tension a bit higher
Hi, thanks for your answers. So your boundary particles do not take part in the pressure gradient term? But they do in the viscous acceleration term and in the mass conservation equation... I have ran the same case and the dynamics are correct too, but the pressure field not very pretty at all lol (I have posted another similar case, but with fluid column size L=2H).
maettino 1 year ago
Hi yonarw, very nice animation! what are the boundary conditions at the solid boundary? are you using dummy particles? I think there is something wrong on your left wall... particles should not leave the boundary (at 0.02 sec there is a gap between the fluid column and the wall). Aside from that... fantastic!
Do you have pressure contours for this simulation?
maettino 1 year ago
@maettino hi and thanks for your comment. as boundary i used a simple position and velocity "reset" (setting positions to boundary and inverting velocity vectors )
it might look wrong but i did not add any static attraction forces between particles and walls so nothing sticks particles to walls.
i tried to draw pressure values of the particles but i never got any good looking image ;)
yonarw 1 year ago
Good simulation!. But I want to ask about the strange initial behavior, it's like a atraction force bretween particles in the begining of the simulation.Interestingly, I got a similar behavior due to a miscalculation of the initial density using the Tait equation for pressure
jahkr 1 year ago
@jahkr thanks. i think it's only looking strange but physics are correct. the problem is that particles are initialized in an grid which then brings up these gaps because they are not initialized close enough to each other ...
yonarw 1 year ago
I see. Well, that's nice.
How do you find a neighbour particles for pressure forces and color functions? Do you use a hash table indexing?
When i use hash tables (i mean i give a structured 3D grid in R^3 and then use it to store particle numbers that are near) on CPU i get 0.01 fps for 118k particles =)
I;m just interested cos i had my doctorate in CFD (traditional mesh methods) a long time ago and now i'm investigating some mesh free methods for free surfaces. SPH seems to be the best.Thanks!
NoctumDeVir 1 year ago
@NoctumDeVir hmm i'm using a grid in this simulation too. it helped a bit to store neighbors for the next frame and check whether their distance is larger than the interactionradius in this next frame after the move and then only check particles for neighbors when they haven't kept their "neighbor link" from the last frame.further it helped to not use any dynamic datatypes. hope this will help somehow. thanks for your interest
yonarw 1 year ago
HI!
That's a nice demo.
Is the breach of particles in the beginning ( 0:02 ) due to the wrong free surface tension?
Thanx!
NoctumDeVir 1 year ago
@NoctumDeVir well its not wrong .. its doing as expected with given values ... but i agree it "looks" wrong ;) should have set the surface tension a bit higher
thx for watching!
yonarw 1 year ago