11

1) Is there any proof that simulations that mix Rigid Body phisix and fluids (say SPH) can provide modeling for real world?
2) How does a frame of such mix work?

Say we have a woden swing inside a box with 2 difrent liquids (oil and water) and a soft body - ball. Start condition looks like: enter image description here

How could we simulate such situation forward? With what tools?

How I currently see it in my had (please tall me if I am wrong)

  1. we have all bodies and fluids at state (*)
  2. Modern engines such as Bullet support soft bodies and rigid bodies - so we can load (*) and make a frame in RB\SB phisix engine.
  3. Take all vertices and compare there positions with previous frame. getting out some form of a vector force field.
  4. Represent our rb/sb at tate (*) in fluid engine as particles with impulses from our force field
  5. load (*) with RB and SB as particles with density corresponding to material they are made and impulses corresponding to previosly calculated vector feild
  6. make a frame in fluid engine
  7. update rb/sb positions with respect to particles movment - get new state (**)

And this is at least in theory may be computable in near real time with current fluids 3 + bullet 3 on gpu for simple cases... Yet I woner how much this makes system not real?

Doron Yaacoby
  • 9,412
  • 8
  • 48
  • 59
DuckQueen
  • 772
  • 10
  • 62
  • 134
  • the problem starts with interactions like collisions ... because you need to interact with all colliding particles at once which is sometimes really hard especially with cyclic interactions (when particle 1 interacts with 2,3,4,5... and some of them interacts back to 1 in not direct-counterforce manner) the precision (realism) drops down very quickly ... most physics engines are meant for games so they replace accuracy with computation speed... For 'precise' simulations are FEM and or field approaches better. at least in my opinion – Spektre Feb 23 '14 at 16:36
  • What do you need this for? Games? Scientific sim? Particle interactions in fluids are incredibly computationally expensive to be realistic. – webuster Feb 23 '14 at 19:05
  • Scientific simulation – DuckQueen Feb 25 '14 at 22:08

2 Answers2

0

It looks as if this question has not received enough attention, so I'm just throwing in my 2 cents here.
A quick Google search reveals this presentation from the University of Pennsylvania and this paper from the University of Tokyo. They may have some of the information you're looking for.

McLovin
  • 3,554
  • 1
  • 14
  • 15
0

There's no need for particle level computation for this to be realistic. Viscous CFD and Fluid Structures Interaction are going to give you the behavior of the system to a very reasonable degree of accuracy. A software package like COMSOL or Ansys could get you the answer in a few hours. If you need to animate this in real time though, you're gonna need to simplify the problem a bit. You said in the comments that you're doing scientific simulation, so I assume you're looking for accurate behavior over frames per seconds.

gankoji
  • 853
  • 1
  • 10
  • 23