i'm trying to drop some bolts onto a (growing) pile of bolts.
(each bolt is in a group{} )

http://jwstolk.afreehost.net/mechsim_problem.png

but:
collision { 0,0,2 ...} does not seem to work
(the connection-connection collisions do not happen.)
collision { 0,0,1 ...} also does not work.
collision { 1,1,1 ...} still does not do connection-connection collisions.
i'm using 80000 steps, witch very low speeds, so i don't think it will be
possible that the connections move past each other within a single step.
(face-connection collisions and face-face collisions also do not work,
but i think they are not supposed to work anyway)

is adding lot's of masses and then doing mass-face collisions the
only solution?

any suggestions?
Jaap.

Post a reply to this message

Jaap wrote:
> i'm trying to drop some bolts onto a (growing) pile of bolts.
> (each bolt is in a group{} )
> 
> http://jwstolk.afreehost.net/mechsim_problem.png
> 
> but:
> collision { 0,0,2 ...} does not seem to work
> (the connection-connection collisions do not happen.)
> collision { 0,0,1 ...} also does not work.
> collision { 1,1,1 ...} still does not do connection-connection collisions.
> i'm using 80000 steps, witch very low speeds, so i don't think it will be
> possible that the connections move past each other within a single step.

I am pretty sure collisions are calculated in all those cases although i 
can't completely rule out an internal problem at the moment.  To 
actually tell you why you think they don't occur i would need to see the 
complete scene.

Note however that your topology is insufficient to accurately simulate 
the movement of a bolt.  The approximation is simply too sparse. 
Remember: the simulation works by calculating the forces from the 
deformations of the bodies it simulates.  If the topology does not 
actually deform in the way it would in reality this won't lead to 
realistic results.

> (face-connection collisions and face-face collisions also do not work,
> but i think they are not supposed to work anyway)

Where do the docs mention anything about face-face collisions or 
face-connection collisions?  The don't exist and are not necessary either.

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 27 Feb. 2005 _____./\/^>_*_<^\/\.______

Post a reply to this message

Christoph Hormann <chr### [at] gmxde> wrote:
> I am pretty sure collisions are calculated in all those cases although i
> can't completely rule out an internal problem at the moment.  To
> actually tell you why you think they don't occur i would need to see the
> complete scene.

http://jwstolk.afreehost.net/bolts_mechsim.pov.gz

> Where do the docs mention anything about face-face collisions or
> face-connection collisions?  The don't exist and are not necessary either.
it does not.

I removed group{}, changed bounding to 0 (off) and method to 0,
but that dit not help.
(interaction of bolts with the enviroment (floor) looks very nice!)

jaap

Post a reply to this message

Jaap wrote:
> Christoph Hormann <chr### [at] gmxde> wrote:
> 
>>I am pretty sure collisions are calculated in all those cases although i
>>can't completely rule out an internal problem at the moment.  To
>>actually tell you why you think they don't occur i would need to see the
>>complete scene.
> 
> 
> http://jwstolk.afreehost.net/bolts_mechsim.pov.gz
> 

Sorry for the late reply.

You code does not even run with the current version (immediate 
instability, did not check with 1.1 though).  Your models are simply way 
too stiff (extremely small masses combined with high stiffness 
connections).  You can make it run by changing this but will never get 
reasonable movements with this topology.

So the way to go is create reasonable bolt shapes (with a reasonable 
mass distribution) and make them soft enough (start with very soft and 
increase stiffness until the look good enough, you will have to decrease 
time stepping at the same time).  Start with one bolt and see if it 
collides well with the environment.  Then add a second one and turn on 
mass-mass and mass-face collisions, you probably do not even need the 
con-con collisions.

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 27 Feb. 2005 _____./\/^>_*_<^\/\.______

Post a reply to this message

Christoph Hormann <chr### [at] gmxde> wrote:
> Sorry for the late reply.
i've been playing with lightSys+hcrEdit, which does kill time very well. :-)

-it does runs in 1.1, but yes, i had some stability problems.
-i will try with very soft connections.
-i know that the simulation will not be very realistic, for a
 start the mass is not distributed evenly threwout the bolt.
 I was actually only interested in the final result: a pile
 of (non-moving) bolts.
-i will add more masses along the thread of the bolt and add
 faces, and then do mass-face collisions. i guess the increase
 in number of masses will be compensated by a more stable
 simulation, which allows for larger time steps.
-i'm still wondering why the connection-connection collisions
 dit not work, even with near-instability the system should at
 least do something. is it possible that the connections end
 up with a zero radius, even if they are drawn with a large
 radius when rendered?
-i picked 3 (perpendicular) masses from the simulation and
 converted there position to a matrix, which positions the
 iso-surface bolts: http://jwstolk.afreehost.net/bolts_test.jpg
 (the bolt-head in the simulation was bigger, and the radius of
 the masses was not counted for, which leads to the "hovering")
jaap.

Post a reply to this message

Jaap wrote:
> -i'm still wondering why the connection-connection collisions
>  dit not work, even with near-instability the system should at
>  least do something. is it possible that the connections end
>  up with a zero radius, even if they are drawn with a large
>  radius when rendered?

No, the radius used in the simulation is taken from the masses the 
connection connects.

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 27 Feb. 2005 _____./\/^>_*_<^\/\.______

Post a reply to this message

I'm simulating two M16x75 bolts (0.075m long),
each about 120 grams. 75 masses per bolt, about
1.5 grams per mass. All connections are generated
with the default length. I set stiffness to 750 and
dampening to 375, to keep deformations acceptable.

Everything works fine, but slow, because I need to set
time_step to 1/1400000 to prevent the instability warning.
From the source I understand that the warning is
generated when acceleration is more than 1e7. F=m*a so
there is a 10000 N force acting on (one of) my masses.
All of this is happening while by bolts are still
free-floating in mid-air. I never get an instability
warning during the rest of the simulation when the actual
collisions happen.

Can someone tell me more about what might course this
large force? Is it just that some masses are moved by
gravity, and are pulling in the not-yet-moved masses?

Can there be small rounding errors in the initial position
of the masses or the length of the connections?
The .DAT file seems to have sufficient deciaml places for
the mass length position and speed in my situation.

jaap

Post a reply to this message

Jaap schrieb:
> 
> Can someone tell me more about what might course this
> large force? Is it just that some masses are moved by
> gravity, and are pulling in the not-yet-moved masses?

You have to understand what instability in a numerical simulation
means.  This is nothing about rounding errors or inaccuracy of math
operations.  

If you start with very small time steps this won't lead you anywhere
since you will have to decrease them later anyway when you optimize the
simulation to look good.  So if your simulation is instable in the first
run don't use smaller time steps but modify the simulated system.

-- Christoph

Post a reply to this message