On TunaBrains code, i just noticed, rotating and or translating the paddle in the test does not move the Body, and then the solver fails...

[damian-666]

Just wondering if you implemented it and if it is stable enough to use...

I was thinking to apply drag, from thw wikipedia drag equation , then push the fluid away from the normals of the immersed thing. but i'd rather do it that way CBatty did.. i have some of that some... i can put the link.. its on a big page on Variational methods.. it gives away 2d code for gas ( with 3 circles... i could move those.. and it pushes away the liquid.. with a one frame delay, it seems.. but the shape is nice, and it does not miss any of the fluid... I see code that mentions forward Eulear.. so i was hoping you know something about it.. .. I will report to Tunabrain as well...I think the temperature gradients is the focus of this code.....

[g-amador]

Not sure what you mean!?
English appears to have been written to much in a rush...

This code examples use forward explicit Euler.
Implicit Euler commonly know as stable fluids from Jos Stam is something different.
If you mean viscous drag, viscosity is a parameter so I guess you can alter it to simulate viscous drag.
If you mean you want to handle moving solids instead of static ones than the problem that velocities have a different behavior with moving vs static objects:
I couldn't find a more simple article (I know I saw once about 7 years ago) but this thesis fairly overviews the boundary conditions:
https://www.nada.kth.se/utbildning/grukth/exjobb/rapportlistor/2007/rapporter07/bongart_robert_07018.pdf

[damian-666]

Sorry i have been doing CFD research non stop for 5 months. I am porting your code that your ported from Incremental fluiids, to c#. but before i implement it , I was wondering if the Body ( immersed bodies) can rotate and or move... and push the matter out of the way.. and if you had any issues with it. I could help fix the issues maybe.. I see your version of rotate on a body is different than tunabrains version. thast version does not move the paddle in the example at all, and then the solver fails. I tried a simple angular velocity of 0.1 on the body, in sample 6.. or any sample witha body. ill show you your code to refresh your memory, it is different than the c version...

[damian-666]

Well , it looks like you modified the distance normal.. that could push
away stuff. , if I put a angular velocity, i hope that the box pushes the
markers and ink around in a vortex... Since it is Euler, I guess i would
need to limit how fast i move it.
So i think i wlll try it.. unless you tell me it never worked or was too
unstable..

@OverRide
public Vector2D closestSurfacePoint(double x, double y) {
x -= _posX;
y -= _posY;
Vector2D result = rotate(x, y, -_theta);
x = result.getX();
y = result.getY();
double dx = abs(x) - _scaleX * 0.5;
double dy = abs(y) - _scaleY * 0.5;
if (dx > dy) {
x = nsgn(x) * 0.5 * _scaleX;
} else {
y = nsgn(y) * 0.5 * _scaleY;
}
result = rotate(x, y, _theta);
x = result.getX() + _posX;
y = result.getY() + _posY;
return new Vector2D(x, y);
}
@OverRide
public Vector2D distanceNormal(double nx, double ny, double x, double y) {
x -= _posX;
y -= _posY;
Vector2D result = rotate(x, y, -_theta);
x = result.getX();
y = result.getY();
if (abs(x) - _scaleX * 0.5 > abs(y) - _scaleY * 0.5) {
nx = nsgn(x);
ny = 0.0;
} else {
nx = 0.0;
ny = nsgn(y);
}
return rotate(nx, ny, _theta);
}

On Mon, Sep 12, 2016 at 12:49 AM, Gonçalo Amador notifications@github.com
wrote:

Not sure what you mean!?
English appears to have been written to much in a rush...

This code examples use forward explicit Euler.
Implicit Euler commonly know as stable fluids from Jos Stam is something
different.
If you mean viscous drag, viscosity is a parameter so I guess you can
alter it to simulate viscous drag.
If you mean you want to handle moving solids instead of static ones than
the problem that velocities have a different behavior with moving vs static
objects:
I couldn't find a more simple article (I know I saw once about 7 years
ago) but this thesis fairly overviews the boundary conditions:
https://www.nada.kth.se/utbildning/grukth/exjobb/rapportlistor/2007/
rapporter07/bongart_robert_07018.pdf

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_LmWUJSCmAL9EULtWoqd6-QMjXkMECks5qpDEfgaJpZM4J56kt
.

[g-amador]

I just ported the c++ code honestly never attempted to modify it (wanted to port to 3D but haven't got the time since finishing the PhD).

I have prior knowledge with backward Euler in stable fluids from the Masters.

I assume its not just a matter of adding/replacing velocity honestly, it appears it will filter wanted effects.

I would suggest trying to couple two fluids instead, where (the moving solids) density would be tweaked to remain a solid, or if you prefer at each fluid update ignore diffusion of the fluid (solid moving objects).

You could also try to alter locally the velocity in the border cells of the moving object, i.e., the moving obstacle velocity plus the velocity in nearby grid cells, or all together replace the velocity of the cells in the solid border be replaced by the moving obstacle velocity.

[g-amador]

You might also require to add artificial vorticity for this around the moving obstacle.

[damian-666]

i guess by question is... does this code handle moving or rotating boxes.
? it looks different there...more promising..

On Mon, Sep 12, 2016 at 2:36 AM, Damian Hallbauer <
damian.hallbauer@gmail.com> wrote:

Well , it looks like you modified the distance normal.. that could push
away stuff. , if I put a angular velocity, i hope that the box pushes the
markers and ink around in a vortex... Since it is Euler, I guess i would
need to limit how fast i move it.
So i think i wlll try it.. unless you tell me it never worked or was too
unstable..

@OverRide
public Vector2D closestSurfacePoint(double x, double y) {
x -= _posX;
y -= _posY;
Vector2D result = rotate(x, y, -_theta);
x = result.getX();
y = result.getY();
double dx = abs(x) - _scaleX * 0.5;
double dy = abs(y) - _scaleY * 0.5;
if (dx > dy) {
x = nsgn(x) * 0.5 * _scaleX;
} else {
y = nsgn(y) * 0.5 * _scaleY;
}
result = rotate(x, y, _theta);
x = result.getX() + _posX;
y = result.getY() + _posY;
return new Vector2D(x, y);
}
@OverRide
public Vector2D distanceNormal(double nx, double ny, double x, double y) {
x -= _posX;
y -= _posY;
Vector2D result = rotate(x, y, -_theta);
x = result.getX();
y = result.getY();
if (abs(x) - _scaleX * 0.5 > abs(y) - _scaleY * 0.5) {
nx = nsgn(x);
ny = 0.0;
} else {
nx = 0.0;
ny = nsgn(y);
}
return rotate(nx, ny, _theta);
}

On Mon, Sep 12, 2016 at 12:49 AM, Gonçalo Amador <notifications@github.com

wrote:

Not sure what you mean!?
English appears to have been written to much in a rush...

This code examples use forward explicit Euler.
Implicit Euler commonly know as stable fluids from Jos Stam is something
different.
If you mean viscous drag, viscosity is a parameter so I guess you can
alter it to simulate viscous drag.
If you mean you want to handle moving solids instead of static ones than
the problem that velocities have a different behavior with moving vs static
objects:
I couldn't find a more simple article (I know I saw once about 7 years
ago) but this thesis fairly overviews the boundary conditions:
https://www.nada.kth.se/utbildning/grukth/exjobb/rapportlist
or/2007/rapporter07/bongart_robert_07018.pdf

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_LmWUJSCmAL9EULtWoqd6-QMjXkMECks5qpDEfgaJpZM4J56kt
.

[damian-666]

thanks.. that is my plan... i tried it with C Batty variational liquid 2d
code ( cant find it now)...andi could push the particles..i will have to
interpolate i guess... and see how C Batty code worked.. unless i use
flip , then just push the particle...

On Mon, Sep 12, 2016 at 3:08 AM, Gonçalo Amador notifications@github.com
wrote:

I just ported the c++ code honestly never attempted to modify it (wanted
to port to 3D but haven't got the time since finishing the Ph.D..

I have prior knowledge with backward Euler in stable fluids from the
Masters.

I assume its not just a matter of adding velocity honestly.
As far as I know I would suggest trying to couple two fluids instead,
where (the moving solids) density would be tweaked to remain a solid, or if
you prefer at each fluid update ignore diffusion of the fluid (solid moving
objects).

You could also try to alter locally the velocity in the border cells of
the moving object, i.e., the moving obstacle velocity plus the velocity in
nearby grid cells, or all together replace the velocity of the cells in the
solid border be replaced by the moving obstacle velocity.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_LmTo346xHNZbjmroakApgOr3E30wsks5qpFHBgaJpZM4J56kt
.

[damian-666]

but what trobules me is how the c++ code didnt solve if i put a 0.001
velocity on the box... are you sure you didnt fix it? i woulld not
know how to fix this if it happen.. i guess i should test it... i dont even
have java working...

On Mon, Sep 12, 2016 at 3:15 AM, Damian Hallbauer <
damian.hallbauer@gmail.com> wrote:

thanks.. that is my plan... i tried it with C Batty variational liquid 2d
code ( cant find it now)...andi could push the particles..i will have to
interpolate i guess... and see how C Batty code worked.. unless i use
flip , then just push the particle...

On Mon, Sep 12, 2016 at 3:08 AM, Gonçalo Amador notifications@github.com
wrote:

I just ported the c++ code honestly never attempted to modify it (wanted
to port to 3D but haven't got the time since finishing the Ph.D..

I have prior knowledge with backward Euler in stable fluids from the
Masters.

I assume its not just a matter of adding velocity honestly.
As far as I know I would suggest trying to couple two fluids instead,
where (the moving solids) density would be tweaked to remain a solid, or if
you prefer at each fluid update ignore diffusion of the fluid (solid moving
objects).

You could also try to alter locally the velocity in the border cells of
the moving object, i.e., the moving obstacle velocity plus the velocity in
nearby grid cells, or all together replace the velocity of the cells in the
solid border be replaced by the moving obstacle velocity.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_LmTo346xHNZbjmroakApgOr3E30wsks5qpFHBgaJpZM4J56kt
.

[damian-666]

with the tunabrain code.. i at a angular vel of 0.01 and it blows up ..
never converges..

On Mon, Sep 12, 2016 at 3:16 AM, Damian Hallbauer <
damian.hallbauer@gmail.com> wrote:

but what trobules me is how the c++ code didnt solve if i put a 0.001
velocity on the box... are you sure you didnt fix it? i woulld not
know how to fix this if it happen.. i guess i should test it... i dont even
have java working...

On Mon, Sep 12, 2016 at 3:15 AM, Damian Hallbauer <
damian.hallbauer@gmail.com> wrote:

thanks.. that is my plan... i tried it with C Batty variational liquid
2d code ( cant find it now)...andi could push the particles..i will have to
interpolate i guess... and see how C Batty code worked.. unless i use
flip , then just push the particle...

On Mon, Sep 12, 2016 at 3:08 AM, Gonçalo Amador <notifications@github.com

wrote:

I just ported the c++ code honestly never attempted to modify it (wanted
to port to 3D but haven't got the time since finishing the Ph.D..

I have prior knowledge with backward Euler in stable fluids from the
Masters.

I assume its not just a matter of adding velocity honestly.
As far as I know I would suggest trying to couple two fluids instead,
where (the moving solids) density would be tweaked to remain a solid, or if
you prefer at each fluid update ignore diffusion of the fluid (solid moving
objects).

You could also try to alter locally the velocity in the border cells of
the moving object, i.e., the moving obstacle velocity plus the velocity in
nearby grid cells, or all together replace the velocity of the cells in the
solid border be replaced by the moving obstacle velocity.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_LmTo346xHNZbjmroakApgOr3E30wsks5qpFHBgaJpZM4J56kt
.

[g-amador]

For forward Euler you can use this code from the
"Fluid Simulation for Computer Graphics"

For forward with obstacles you will need Stable Fluids modified to use obstacles
http://www.intpowertechcorp.com/GDC03.pdf

Another old alternative would be using
http://www.gamasutra.com/view/feature/1549/practical_fluid_dynamics_part_1.php?print=1

[g-amador]

The only difference I handled in the code was that I had to make sure the right values appeared in some cases because c++ handles out of bound (representable numbers) differently from Java so I slightly altered that in the source code.

I literally tested visually and numerical the C++ vs Java code to address this.

[g-amador]

The C++ code slightly cheats using C++ behavior common in C++ programming but in Java or C# might be a problem with ports sometimes.

[g-amador]

If that code considers rotating or moving obstacles: No!

Moving and rotating obstacles velocity requires you to:
1-Identify the cells in which their borders are within the grid (rough aproximation), e.g.:

---

|__ ||__ ||__ || S ||__ ||__ ||__ |
|__ ||__ || S || S || S ||__ ||__ |
|__ || S || S || S || S || S ||__ |
|__ ||__ || S || S || S ||__ ||__ |
|__ ||__ ||__ || S ||__ ||__ ||__ |
|__ ||__ ||__ ||__ ||__ ||__ ||__ |

[g-amador]

The object must be within the cells but not all cells are entirely engulfed by the objects in this case accuracy would state use the percentage of the cell occupied (more processing time).
Instead use the mean between the cell old velocity and the moving object velocity at the cell (might work).

[damian-666]

thanks... i just went over your thesis.. its in 3d.. very ambitious.. im
staying with 2d.. i hope that heat will stir up stuff.. i also have some
IB peskin like code... i have seens the stam.. and now i might do the FFT
way. after 5 months the FFT way is as good as any multigrid... i
use panel an a estimated drag .. and i guess i push the stuff.. .. i
may use FLIP scheme after all...i just thikn this one is too slow... but
the rest is great... up to mixed density... i hope for stratified
dust.... in the air from a volcano... using this solver, with4 types of
particles .. ash sand dust... and something.. and with the heat...

On Mon, Sep 12, 2016 at 3:22 AM, Gonçalo Amador notifications@github.com
wrote:

For forward Euler you can use this code from the
"Fluid Simulation for Computer Graphics"

For forward with obstacles you will need Stable Fluids modified to use
obstacles
http://www.intpowertechcorp.com/GDC03.pdf

Another old alternative would be using
http://www.gamasutra.com/view/feature/1549/practical_fluid_
dynamics_part_1.php?print=1

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#2 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AP_Lmc-s_bP8ldjLA7-N4nKwgQWd8ziPks5qpFTegaJpZM4J56kt
.