How Realistic of a Tornado?

Alpha Omega · Post by **Alpha Omega** » Sun Jan 23, 2011 2:34 am

I am building the Irrlicht Dedicated Physics engine and generated some video to share. My question is could you please tell me how real looking this tornado simulation is?

Video is here

http://www.youtube.com/watch?v=_Cdx2GFDFsg

I am using the planetary laws of motion with different constants and had to invent a new constant to put in the equation but it came out alright.

Radikalizm · Post by **Radikalizm** » Sun Jan 23, 2011 3:13 am

I have to say it looked pretty good, reminded me a little of this demo http://www.youtube.com/watch?v=IJ0HNHO5Uik
skip to about 2:15 to see what I'm talking about

I'm pretty amazed at how fast your physics project seems to be going, I hope development keeps going at this pace and I hope to see some awesome results

ent1ty · Post by **ent1ty** » Sun Jan 23, 2011 11:36 am

Nice!

Virion · Post by **Virion** » Sun Jan 23, 2011 5:43 pm

that's cool man. yout wrote the physics engine from scratch or using an existing one (bullet, physx etc)?

Alpha Omega · Post by **Alpha Omega** » Sun Jan 23, 2011 6:30 pm

Virion wrote:that's cool man. yout wrote the physics engine from scratch or using an existing one (bullet, physx etc)?

Totally from scratch. Its something I can learn from and get better in C++ and irrlicht while stilling doing stuff I know. I am a total noob in graphics lol.

fmx · Post by **fmx** » Sun Jan 23, 2011 6:47 pm

It looks really cool, very promising progress!

In terms of realism though (or as "real" as simulated tornados get) its still a long way off.
Looks more like a donut vortex, you probably need to increase the forces vertically so the cubes vary their heights as they get flung around and around.

Most likely random upward forces should get applied occasionally flinging the cubes up, with gravity and attraction automatically pulling them down again

I could mention tapering at the bottom, but I dont think you're interested in really simulating a tornado

Alpha Omega · Post by **Alpha Omega** » Sun Jan 23, 2011 6:58 pm

fmx wrote:It looks really cool, very promising progress!

In terms of realism though (or as "real" as simulated tornados get) its still a long way off.
Looks more like a donut vortex, you probably need to increase the forces vertically so the cubes vary their heights as they get flung around and around.

Yea I could make it look different with different constants in the equations. I am trying to structure the engine on first principles and not just randomly throwing in forces for different effects. Thus the only real equations I could work with are Keplar's Planetary Motion laws. That was simulated with gravity off but I agree more vertical rotation would make it nicer.

ent1ty · Post by **ent1ty** » Sun Jan 23, 2011 7:43 pm

It would be nice if you implemented some sort of force fields. Everything that is inside a box, would get some force applied on it, depending on the setting user would pass. Could be good for simulating a river with flowing garbage in it, or a fan.

fmx · Post by **fmx** » Tue Jan 25, 2011 10:32 am

You keep talking about Keplar's Planetary Motion laws but you forget that tornados are wind powered and derive their physics from airflow dynamics in differential pressure environments.
So offcourse your simulation will not work correctly without introducing lots of hacky constants and additional forces

A better application of your work would be to simulate objects in outer space, maybe a galaxy with planets, asteroids and lots of random spaceships or something

IIRC there was an impressive space demo made by someone on the forums a few years back, now I can't remember if it used a physics engine or not

Radikalizm · Post by **Radikalizm** » Tue Jan 25, 2011 12:02 pm

It's true that Kepler's laws do not give a realistic description of how a real tornado would function, but it does give a nice approximation without having to keep track of too many factors, since I assume that you are simulating the tornado by creating motion through the gravitational pull of a central upwards facing axis, instead of a central point which would be used for actual planetary simulation (correct me if I'm wrong here)

Actual 'particle-based' (as I can't really find a better term to describe this) wind simulation of a tornado using differential pressure fields as you proposed would not be viable in a real-time environment, unless you use some extreme optimizations, but even then you'd be better off by using an approximation

Alpha Omega · Post by **Alpha Omega** » Tue Jan 25, 2011 9:38 pm

The method used is a gravity transform onto another object and that is what it was meant to do. I merely found out that by manipulating parameters different effects could be produced.