Hi,

I am trying to simulate melting of ice. I wrote a c++ code to generate frames
for the animation. The cube is a set of particles initially in a solid state and
later melt and become liquid and fall off.

To render this, I used super epsilloid to represent each solid particle as a
small cube. The 'merge'd set of cubes make the whole cube. Liquid particles
become small spheres and form blobs with other nearby spheres. This is the video
link. (http://www.youtube.com/watch?v=fMul9JbGBiw) There are various problems I
am trying to fix.

1) The still solid part of the cube has sharp ridges. I am not able to figure
what to do to avoid that. Maybe Isosurface. If someone could suggest how to make
that look better it would be great.

2) The particles I think are too heavy for surface tension effects. But
disturbing the particle mass changes the rate of temp change drastically. So if
you could suggest what to use there it would be great as well.

Any inputs would be helpful to me.

Thanks,
pH

Post a reply to this message

I suggest using the 'blob' object.

Post a reply to this message

> I suggest using the 'blob' object.
>
>

I tend to agree. With a blob, it will also be faster than using many 
superellipsoids in a merge.

The superellipsoid is a relatively slow primitive. When using a merge, 
once you hit the bounding box of the merge, you need to test and 
evaluate each and every components.

With a blob, there is a great deal of optimisations done, starting with 
an internal bounding hierarchy. You only evaluate the closest components.
A blob tends to produce rounder shapes.

As for the particles been to heavy, have you tryed making them lighter 
but increase the thermal inertia?
You can also increase the surface tention to compensate.
You can also decrease the heat imput proportionately to the mass.



Alain

Post a reply to this message

Alain <aze### [at] qwertyorg> wrote:

> > I suggest using the 'blob' object.
> >
> >
>
> I tend to agree. With a blob, it will also be faster than using many
> superellipsoids in a merge.
>
> The superellipsoid is a relatively slow primitive. When using a merge,
> once you hit the bounding box of the merge, you need to test and
> evaluate each and every components.
>
> With a blob, there is a great deal of optimisations done, starting with
> an internal bounding hierarchy. You only evaluate the closest components.
> A blob tends to produce rounder shapes.
>
> As for the particles been to heavy, have you tryed making them lighter
> but increase the thermal inertia?
> You can also increase the surface tention to compensate.
> You can also decrease the heat imput proportionately to the mass.
>
>
>
> Alain

Thanks guys.. I'll tinker with the values more to get a more realistic effect.
The problem i had with blobs was that I had to keep changing the threshold and
radii to make sure the cube is large enough and the molten water droplets are
not too big.

-pH

Post a reply to this message

> Alain<aze### [at] qwertyorg>  wrote:

>>> I suggest using the 'blob' object.
>>>
>>>
>>
>> I tend to agree. With a blob, it will also be faster than using many
>> superellipsoids in a merge.
>>
>> The superellipsoid is a relatively slow primitive. When using a merge,
>> once you hit the bounding box of the merge, you need to test and
>> evaluate each and every components.
>>
>> With a blob, there is a great deal of optimisations done, starting with
>> an internal bounding hierarchy. You only evaluate the closest components.
>> A blob tends to produce rounder shapes.
>>
>> As for the particles been to heavy, have you tryed making them lighter
>> but increase the thermal inertia?
>> You can also increase the surface tention to compensate.
>> You can also decrease the heat imput proportionately to the mass.
>>
>>
>>
>> Alain
>
> Thanks guys.. I'll tinker with the values more to get a more realistic effect.
> The problem i had with blobs was that I had to keep changing the threshold and
> radii to make sure the cube is large enough and the molten water droplets are
> not too big.
>
> -pH
>
>
>
>

Don't play with the threshold, work with the strenght and radius instead.

Post a reply to this message

pH wrote:

> The problem i had with blobs was that I had to keep changing the threshold and
> radii to make sure the cube is large enough and the molten water droplets are
> not too big.

Of course, if you basically have a 3d grid of particle positions
you might simply be able to use differently rounded shapes instead
of boxes depending on which neighboring positions are occupied.

Post a reply to this message

Thanks all. I used blobs played with the settings and came up with this
(http://www.youtube.com/watch?v=J_GUjGDknvI). The scene has a heat source (The
Orange sphere near the left top of the cube)  The only problem I see now is the
size of the molten droplets.

What I thought of doing is, flattening the droplet once it melts by scaling. I
think scaling it in the direction perpendicular to the normal at the point the
droplet is on the remaining solid would do the trick. But I am not able to find
a way to calculate the normal of a blob at a given point on its surface. Can
someone please suggest how to do that? Or is there any other way apart from what
I thought?

Post a reply to this message

"pH" wrote:
> Thanks all. I used blobs played with the settings and came up with this
> (http://www.youtube.com/watch?v=J_GUjGDknvI). The scene has a heat source (The
> Orange sphere near the left top of the cube)  The only problem I see now is the
> size of the molten droplets.
>
> What I thought of doing is, flattening the droplet once it melts by scaling. I
> think scaling it in the direction perpendicular to the normal at the point the
> droplet is on the remaining solid would do the trick. But I am not able to find
> a way to calculate the normal of a blob at a given point on its surface. Can
> someone please suggest how to do that? Or is there any other way apart from what
> I thought?

I believe you may be looking for the four-parameter trace() function.

Post a reply to this message