phoenix and thermal simulation

[super gnu]

Hi.. i have no idea if what im asking is even remotely within the capabilities of phoenix... id like to simulate the heating and movement of air in a space with heat sources. i imagine if it is possible, there wont be much accuracy, but id love to be proved wrong! at the basic end id like to graphically visualise the basic process at whatever accuracy.. ideally id like to use it to test design variants to see which are worth testing in the real world.




effects id like to visualise/simulate (in laymans terms of course, its all ive got!) :

* hot air rising when in contact with warm surfaces, and vice versa, air losing heat to cooler surfaces in a controllable way, i.e specify object thermal properties

* visualisation of airflow graphically, either with volumetric shading or particles

* -the biggy i think- simulate air moving past an opening causing a suction effect

* everything else (think surfaces with sunlight hitting them emitting heat etc...)


any feedback/laughing/ pointers most appreciated.. phoenix seems very powerful and versatile, but can it do this?!

cheers!

[Ivaylo Katev]

no problem if you will use this only for visualization purposes, the air in contact with cold and hot objects will change its temperature , will rise and fall according it, and will cause a suction effect.
however, this will look plausible, but will be not accurate enough for physical predictions. for example the moved air per hour will be not a plausible value, there are no thermal properties of the bodies like thermal capacity and conductivity, the only parameter that you can control is their temperature. the thermal contact between the body and the air is just a side effect, that we are trying to remove long time ago, but in your case it is a feature. you can download the demo and try it. just to say how to setup the bodies: make them sources with very low discharge (0.001), this will make them hot/cold but without to release air. the temperature is set in kelvins, the room temperature is considered 300K.

[super gnu]

excellent!, when i get a mo i shall try this. the results might be accurate enough for my purposes, since even if i dont have accurate values for airflow etc, i might well be able to see if one design is drastically better than another... of course this would be a preliminary to a lot of physical modelling, or possibly some more professional specific software work... in this case its quite likely a 2d simulation would be sufficient.

[super gnu]

just as a personal opinion.. ive done some research, and there is no commercial product i can find that gives a visual simulation of heating/airflow etc for construction. there are building energy simulators definitely, but they are extremely expensive and all seem to deal with tables of numbers.

should it be mathematically possible to include objects with differing thermal properties (i guess by treating them as zones with different properties within the grid) and get something approaching accuracy... you could have a market for such a product. similar to lighting analysis you could calculate solar heat gain, efficiency of shading/heating/cooling systems, ventilation, insulation performance... hell, you could even see the draughts youd get from leaving a window open.. a lot of cool stuff.

seems this would be a seperate product to phoenix in the end, possibly a layer of complexity too far for a (2013)cgi fluid simulator!

[Ivaylo Katev]

this would be a completely different product

[super gnu]

of course... but just in case the idea inspired you, and you had too much free time on your hands

[super gnu]

so anyway i downloaded the demo, and based a -very-quick test on the 2d sample file. a simple room with a vent at the top and at the bottom. i set the roof geometry to be the fuel source with a ver low discharge rate but a high temperature..

sure enough i got heating of the air below the roof, and an updraught through the roof opening and an airflow in through the bottom opening.. very nice.. problem is displaying/rendering.. the viewport display would be fine, but you have very little control over shading.. if i try to use the temperature and adjust the point at which its red (the default colour) i only get a few of the particles shaded.. if i try to use velocity, i can specify a min and a max colour, but with whatever settings i can only ever get a mass of blue particles and only the fastest are red. if i bring the min and max closer together the fastest and slowest particles are simply not shown.

im after a way to visualise the full range of temperatures / velocities i guess with a rainbow or just a gradient which changes linearly from the slowest particles to the fastest.

i considered then rendering it instead.. however the default shading for the fire element is barely visible as i have no fire in my scene, only cool and warm air. i changed this to render the velocity channel, but again only get a dense cluster in a small range of velocities. to be honest ive spent the last 15 mins trying to control the colour gradient in teh rendering options and im close to give up.. how the hell do you use that graph thing? documentation seems to be a bit lacking, or im looking in the wrong place.

[super gnu]

hmm cant seem to edit post... ive tried clearing the gradient in the render settings and placing new points with new colours, but id have imagined they would also appear as points on the graph below for me to drag to the desired velocity or temperature... but i get nothing in the graph. and i cannot see how to add points there... not to mention the scale is a but cryptic and there are even dialogue boxes with small bits of.. chinese is it?

[Svetlin.Nikolov]

Hi,

You've done the right thing moving to rendering - the viewport preview is just meant to aid the artist while working, but you can't show too much there.
If you want to visualize the temperatures, I suggest that you don't use the "Fire/Emission" rollout - the options there are for rendering stuff that shines - the X axis corresponds to the quantity that you have selected in "Based on:" - Temperature, smoke, speed, etc. The Y axis is the Luminance in candelas / m2 ( http://en.wikipedia.org/wiki/Candela_per_square_metre ).

Now, if I understand you correctly, you may want to achieve the following effect:



* In the "Smoke color" rollout (this is the diffuse color) you base your display on Temperature and lay out your colors in the Color gradient box.

* Don't forget that as Ivaylo mentioned above, the temperature is in absolute units and 0 Celsius is 300 in the simulator, that's why in the screenshot above we start from ~360 until ~1000 for the mushroom cloud.

* You can control the opacity depending on temperature in the rollout below - "Smoke density". You choose Based on Temperature and lay out the Opacity Diagram again according to the Temperature on X, and the opacity is on Y. Again, don't forget to start from above 300, otherwise the entire box of the simulator will fill up with color


Also, if you literally see dialog boxes with chinese, I would like a screenshot with it, because it sounds like a bug that needs fixing

Cheers!

[super gnu]

ok getting somewhre now with visualising the results, i just had to persevere with the fire shading as it defaults to much higher temperature differentials than i am trying to show.

however next question is, how to get the most accurate simulation possible.. i understand that the geometry will not behave correctly as i can only assign a fixed temperature to a given object.. but what about the simulation parameters..? i was trying with "forward transfer" since if i understand correctly it conserves the volume of gas better than semi lagrangian.. however i get vastly different results between the two methods.. semi lagrangian gives a plausible looking flood of hot air rising from the external part of the roof, and the air flows generally quite convincingly.. but with forward transfer, its like the hot roof object has barely any effect on the surrounding air temp.. it takes 500 frames just to see any movement, and it generally doesnt look plausible.. (especially considering for testing purposes the roof is at 5000k.. should be a virtual tornado of hot air around it..)

-and ive not even got into wether i should have "cooling" or "vorticity" enabled, or what their values should be.

in the end im hoping to simulate a room at 20 odd degrees with a roof at 60 or so...



anyway what i think im after in the end, is advice on how to get phoenix to do the most physically correct hot air convection possible (within its limitations..)

i appreciate this is possibly a big ask for a demo user, but if anyone fancies helping, i may find a reason to buy it at last

[super gnu]

oh and btw, i can do a screengrab if you need it, but the chinese (possibly) text appears when i right click on a point in either the gradient or luminance graphs, and click "edit" o get a simple floater with a few numeric values and title, thats definitely not in english ..or bulgarian.

[Ivaylo Katev]

FT gives less effect of the hot bodies over the air, this is true. as i said, we consider this thermal diffusion undesired and are trying to avoid it. the diffusion at zero discharge is uncontrollable and depends on the advection method, you can make more precise and controllable simulation changing the entire source setup. make two instances of the body, and apply a shell modifier over the second instance. put only the second instance in the source and put it in the exclude list of the simulator. this will make it to work in brush mode, i.e. it will affect directly the air, but does not interact mechanically with it. the first instance will interact mechanically but will not interact thermally. because the second instance is bigger (the shell mod) it will form a thin layer of air that is heated in controllable way, by the source discharge. be sure that the size difference between both the instances is at least two cells wide, otherwise the heat production will be not equal in the different surface points. this should produce result that is relatively equal with both the advection methods. i think that really FT is the better method for physically accurate predictions, you can see in the animation explaining the conservation parameter how the smoke disappears. with FT it will keep its quantity, the low conservation in this case affects only the compressibility of the air.

[wxyz]

Hi,

You've done the right thing moving to rendering - the viewport preview is just meant to aid the artist while working, but you can't show too much there.
If you want to visualize the temperatures, I suggest that you don't use the "Fire/Emission" rollout - the options there are for rendering stuff that shines - the X axis corresponds to the quantity that you have selected in "Based on:" - Temperature, smoke, speed, etc. The Y axis is the Luminance in candelas / m2 ( http://en.wikipedia.org/wiki/Candela_per_square_metre ).



Now, if I understand you correctly, you may want to achieve the following effect:

[ATTACH=CONFIG]15335[/ATTACH]

* In the "Smoke color" rollout (this is the diffuse color) you base your display on Temperature and lay out your colors in the Color gradient box.

* Don't forget that as Ivaylo mentioned above, the temperature is in absolute units and 0 Celsius is 300 in the simulator, that's why in the screenshot above we start from ~360 until ~1000 for the mushroom cloud.

* You can control the opacity depending on temperature in the rollout below - "Smoke density". You choose Based on Temperature and lay out the Opacity Diagram again according to the Temperature on X, and the opacity is on Y. Again, don't forget to start from above 300, otherwise the entire box of the simulator will fill up with color


Also, if you literally see dialog boxes with chinese, I would like a screenshot with it, because it sounds like a bug that needs fixing

Cheers!

How is the blue palette achieved? Is that the preview mode?

[Svetlin.Nikolov]

You just need to disable the emission/fire and set rainbow colors for the diffuse/smoke color as in the screenshot, and then render.