|
 |
Here's a different method of simulating fluid flow in rivers, inspired by
Tim Nikias' videos. The algorithm I'm using here is the one described in my
reply to his most recent post.
The algorithm is a simulation of fluid movements that go through the whole
depth of the fluid, as opposed to surface ripples. Perhaps for this reason,
the waves in this simulation are quite large, and my experiments with
putting ripples from rain here didn't work - the ripples didn't propagate
properly.
I suspect the best results may be achieved by combining the large-scale
waves around objects of this simulation with the results of another
simulation of small ripples, caused by rain/splashing etc. I haven't yet
tried this, and I'm not sure how well the two simulations could be layered:
ideally the ripples could be distorted by the velocity field of the
large-scale simulation, which should avoid the wave stacking problem that
Tim had.
As the large-scale flow quickly settles down into a fairly stable state, it
might be possible to render one large-scale height-field computation and
only animate the small scale details. Currently, the fluid dynamics is
implemented in C++: if only one of the complex fluid dynamics calculations
is needed for a whole animation, it may be possible to implement this in
Pov-Ray to calculate once before the start of the whole animation render and
save the height-field for use in every frame.
Please excuse the fact that the camera obviously rotates around the wrong
point (: Also, for the eagle-eyed, there is a slight discontinuity at the
joins between the copies of the water height-field - I think this is a bug
in the boundary conditions of my simulation which should be quite easy to
fix.
Post a reply to this message
Attachments:
Download 'water.avi.dat' (243 KB)
|
|
|
|
Although one can observe a little bit of how
it looks, the rotation is far too much (and as you
noted about the wrong point), and perhaps even
the resolution of the image is a little too small.
Perhaps one or two larger still images in
binaries.images would be sufficient.
Another thing of interest is the resolution of the
water-mesh and the parsing time.
Aside of that, it does look promising, but you
could showcase it better.
As for saving the mesh to disk: I don't know
how you're saving the data, but you could easily
use my I/O Scripts to save an array to disk
and load it for every frame.
Additionally, if the nodes of mesh are structured
in a 2D-Array, you may even use my Mesh-Modifying-
Macros to save the mesh to disk and load that
for every frame...
Regards,
Tim
--
Tim Nikias v2.0
Homepage: http://www.digitaltwilight.de/no_lights
Email: Tim### [at] gmx de
> Here's a different method of simulating fluid flow in rivers, inspired by
> Tim Nikias' videos. The algorithm I'm using here is the one described in
my
> reply to his most recent post.
>
> The algorithm is a simulation of fluid movements that go through the whole
> depth of the fluid, as opposed to surface ripples. Perhaps for this
reason,
> the waves in this simulation are quite large, and my experiments with
> putting ripples from rain here didn't work - the ripples didn't propagate
> properly.
>
> I suspect the best results may be achieved by combining the large-scale
> waves around objects of this simulation with the results of another
> simulation of small ripples, caused by rain/splashing etc. I haven't yet
> tried this, and I'm not sure how well the two simulations could be
layered:
> ideally the ripples could be distorted by the velocity field of the
> large-scale simulation, which should avoid the wave stacking problem that
> Tim had.
>
> As the large-scale flow quickly settles down into a fairly stable state,
it
> might be possible to render one large-scale height-field computation and
> only animate the small scale details. Currently, the fluid dynamics is
> implemented in C++: if only one of the complex fluid dynamics calculations
> is needed for a whole animation, it may be possible to implement this in
> Pov-Ray to calculate once before the start of the whole animation render
and
> save the height-field for use in every frame.
>
> Please excuse the fact that the camera obviously rotates around the wrong
> point (: Also, for the eagle-eyed, there is a slight discontinuity at the
> joins between the copies of the water height-field - I think this is a bug
> in the boundary conditions of my simulation which should be quite easy to
> fix.
>
>
>
>
Post a reply to this message
|
|
