Heat transfer: Current status?

[Gurgel]

In particular I would be grateful for insight into the following:

- Insulated tile <-> liquid: This seems to be zero at the moment. I have an insulated tile in contact with lava and it has been at 46.9C for > 200 cycles.

- Insulated tile <-> insulated tile: This also seems to be zero.

- Insulated tile <-> gas: There seems to be some heat transfer here.
  I remember something about the logarithmic mean of both thermal conductivities as the total conductivity, i.e. (a-b)/(ln(a)-ln(b)). Correct?

This may have been discussed recently, but I seem to be unable to find the discussion.

[crypticorb]

There were some major changes to thermal exchange a few updates ago, but I'll summarize.

All tiles that are not insulated, including raw ore, abyssalite, gasses, liquids, etc all exchange thermal energy with their neighbor by averaging the thermal conductivities of the two materials. Raw abyssalite that is only 1 tile thick with lava on one side and cooler material on the other will eventually leak some heat through. Two tiles of raw abyssalite will perfectly insulate, because they average to near zero between them.

Insulated tiles break from this system, they do not average thermal conductivity with their neighbor. They conduct at 1% of the thermal conductivity of the base material, and the transfer is limited to this (usually) lower number.

The reason you see gas heating up around insulated tiles is because the gas has an incredibly small amount of mass and thermal capacity compared to a solid, so even though the insulated tile is insulated, it'll exchange a small amount of heat with highly conductive gasses, though it's negligible. 

[Gurgel]

Actually, I see no (i.e. plain zero) transfer between magma and an isolated tile. Or between isolated tiles and other solid or liquid tiles. 

I have not actually calculated it, but a change below 0.1C in an indigenous rock isolated tile over 200 cycles when in contact with Magma that is 1400C hotter should be wayyyy below 1%.

As to gas and isolated tiles, I see the isolated tile change temperature as well. This could well be the 1% you say.

 

[Saturnus]

Correct. Currently there seems to be a major bug in heat transfer with insulated tiles. Only gases transfer any heat to or from insulated tiles of any material. There is no heat transfer to or from insulated tiles of any material to other tiles or liquids.

Effectively as long as you're not dealing with gases, insulated tiles of any material is a super insulator. At the moment at least.

[Gurgel]

Thanks for the confirmation.

[FIXBUGFIXBUGFIX]

Note1:

• Surface-Area multipliers for interacting with each of the three states gas/liquid/solid. As a rule of thumb, liquids have a multiplier of 25 for interacting with other liquids, and gas also have a multiplier of 25 for interacting with solids. All other multipliers have a value of 1. Let's note them S for now.

Note2:

• if |ΔQ'| < 1.0E-4, no transfer is done

[Saturnus]

45 minutes ago, R9MX4 said:

Note1:

• Surface-Area multipliers for interacting with each of the three states gas/liquid/solid. As a rule of thumb, liquids have a multiplier of 25 for interacting with other liquids, and gas also have a multiplier of 25 for interacting with solids. All other multipliers have a value of 1. Let's note them S for now.

Note2:

• if |ΔQ'| < 1.0E-4, no transfer is done

It does not seem to obey those rules anymore though as you can quickly see by using insulated tiles of any regular material to encase magma, even overpressurized to the extreme there is absolutely no heat transfer.

I did a test over 10 cycles with 200000kg per tile 2000K magma encapsulated in granite insulated tiles earlier. Here's the result.

Why would anyone want to go to the hassle of getting super insulator materials when an insulated tile of any material will do?

[FIXBUGFIXBUGFIX]

3 hours ago, Saturnus said:

It does not seem to obey those rules anymore though as you can quickly see by using insulated tiles of any regular material to encase magma, even overpressurized to the extreme there is absolutely no heat transfer.

I did a test over 10 cycles with 200000kg per tile 2000K magma encapsulated in granite insulated tiles earlier. Here's the result.

Why would want to go to the hassle of getting super insulator materials when an insulted tile of any material will do?

The rule might change with time, but it can show us Klei's idea about how heat transfer should work.

The heat transfer between solid-gas is 625 times faster than between solid-solid.

Mass doesn't influence heat flux(the amount of transferred heat). see below

[Saturnus]

12 minutes ago, R9MX4 said:

The rule might change with time, but it can show us Klei's idea about how heat transfer should work.

The heat transfer between solid-gas is 625 times faster than between solid-solid.

Mass doesn't influence heat flux(the amount of transferred heat).

The problem is that with liquids to insulated tiles, it's not a question of the heat transfer being faster or slower. It literally doesn't exist. There is no heat transfer.

I use high mass to prevent temperatures dropping even a fraction should there be any heat transfer I can then easily calculate the rate.

[crypticorb]

5 minutes ago, Saturnus said:

The problem is that with liquids to insulated tiles, it's not a question of the heat transfer being faster or slower. It literally doesn't exist. There is no heat transfer.

I use high mass to prevent temperatures dropping even a fraction should there be any heat transfer I can then easily calculate the rate.

Have you done any tests with standard tiles made from the superinsulator? From what I can tell, it should be the same as the old abyssalite standard tiles. With zero thermal conductivity from the superinsulator, it should average to zero.

Since the insulated tiles are currently bugged, it may be more worthwhile to do tests with the new materials with standard tiles, rather than insulated.

[Saturnus]

1 minute ago, crypticorb said:

Have you done any tests with standard tiles made from the superinsulator? From what I can tell, it should be the same as the old abyssalite standard tiles. With zero thermal conductivity from the superinsulator, it should average to zero.

Since the insulated tiles are currently bugged, it may be more worthwhile to do tests with the new materials with standard tiles, rather than insulated.

No. I don't do preview branches anymore. I value what remains of my (in)sanity.

[FIXBUGFIXBUGFIX]

4 minutes ago, crypticorb said:

With zero thermal conductivity from the superinsulator, it should average to zero

not 0, it's 0.00001.too small to show, but never be 0.

15 minutes ago, Saturnus said:

The problem is that with liquids to insulated tiles, it's not a question of the heat transfer being faster or slower. It literally doesn't exist. There is no heat transfer

liquid carbon, sand stone tile. just take about 30 seconds.

heat transfer literally exist.

[wachunga]

1000kg 1726.9C magma vs 1000kg 1726.9CK steam. After one cycle on very fast, magma heats insulated granite by 0.6C, steam heats by 14.5C. Liquid to solid is substantially slower than gas to solid. That gas to solid 25x modifier seems to still exist. Behavior is the same in both RU and the preview. I would guess the very high mass in Saturnus' test is causing weirdness.

Spoiler

 

[Saturnus]

@wachunga You are correct. I didn't test with tiles that aren't full tiles of liquid. So although @R9MX4 said mass doesn't affect the heat transfer your test agrees with mine that it does just the opposite direction of what it should be. It seems the less mass you have the more heat is transferred. 

I did a new test now and found something extremely odd. And hence my position remains that heat transfer is profoundly flawed at the moment.

I pasted in two tiles of 1000kg 2000K magma on top of eachother. Naturally, one tile of that would be a full tile of magma, the bottom one, and the top one wouldn't be a full tile. The split is 190kg and 1810kg respectively. Now here's what the status is after about 2 cycles. Look closely. and remember the top tile has about 10 times less magma touching it than the bottom one.

As you can see. The top tile not touching a full tile of magma has heat transferred to it. The bottom that does touch a full tile of magma has no heat at all transferred to it.

[FIXBUGFIXBUGFIX]

9 minutes ago, Saturnus said:

@wachunga You are correct. I didn't test with tiles that aren't full tiles of liquid. So although @R9MX4 said mass doesn't affect the heat transfer your test agrees with mine that it does just the opposite direction of what it should be. It seems the less mass you have the more heat is transferred. 

I did a new test now and found something extremely odd. And hence my position remains that heat transfer is profoundly flawed at the moment.

I pasted in two tiles of 1000kg 2000K on top of eachother. Naturally, one tile would that be full tile of magma, the bottom one, and the top one wouldn't be a full tile. The split is 190kg and 1810kg respectively. Now here's what the status is after about 2 cycles. Look closely. and remember the top tile has about 10 times less magma touching it than the bottom one.

I confirmed your result. You are right about mass. And also looks like it is not the clamping by |Q2-Q1|/8

left,190kg,20.4C; right.1810kg,20C

Pretty strange.

[The Flying Fox]

45 minutes ago, R9MX4 said:

I confirmed your result. You are right about mass. And also looks like it is not the clamping by |Q2-Q1|/8

left,190kg,20.4C; right.1810kg,20C

Pretty strange.

Hmmm, I repeated mine using liquid tungten starting around 5700C.  The left at only around 100KG, the right at 3907KG and it seemed to work as 'expected'.  No strange heating behavior with any of the insulated tiles.  All four on the right side are the same temperature and the four on the left side are as well.  Perhaps the strangeness that's being seen is that magma, with it's reduced conductivity compared to liquid tungsten, is so close to the clamping that rounding errors are creeping in?  Perhaps simulation speed has something to do with it?

In either case, it is pretty strange that liquids are effected less by liquids then gases.  Or course, I can somewhat understand the reasoning behind the increased transfer rate from a solid to a gas.  A solid will thermally dominate a gas due to their extreme differences in mass, but also a gas can convect heat around readily.  Along with a gas being only partially transparent to the radiant heat just pouring out of said solid.  Imagine a hot round sphere inside the center of a 1m x 1m x 1m enclosed box.  The radiant heat pouring out of that sphere effectively has a massive 'surface area' with the gas inside that tiny room.  The heat can really only go one way out of the sphere.. into the gas.

 

On the other hand, increasing the temperature transfer from a gas to a solid, the other way around, doesn't make as much sense.  A hot gas's thermal energy is chaotic and goes in all directions, mostly missing a cool sphere sitting inside the box.  In this case, the surface area really is only the actual surface of the sphere.

 

Of course, I'd imagine this is one of those areas where the simplification of ONI's heat stimulation is biting us.  Anyway, I filled out a bug-report on this same thing a few days ago since the behavior is certainly weird in how magma can be held with almost no temperature change, but gases slowly heat up insulated tiles.

 

[wachunga]

Here's what I think is happening. First the game calculates how much heat transfer there is supposed to be. It then tries to apply that transfer to each material. If the change in temperature is below a certain threshold (if the material is sufficiently massive and/or the heat to be transferred is sufficiently small) then the game scraps the whole transfer. The threshold could simply be a result of rounding errors if not intentional.

For magma this magic threshold occurs when the mass of magma is above 1166.3kg (for 1726.9C magma interacting with a 20C insulated granite tile). To test this, I tried it with molten glass. Molten glass has the same conductivity but 1/5 the SHC of magma. You would need 5x the mass of molten glass to reach this magic threshold. Sure enough, molten glass transfers at 5831kg but not 5832kg.

Reading the decrypting heat transfer thread seems to confirm this. There is an initial heat transfer calculated. That transfer is used to calculate new temperatures. These new temperatures are used to calculate the final heat transfer (presumably to prevent temps overshooting each other and seesawing back and forth). If the new temperature is the same as the old temperature (due to a rounding error), then the final heat transfer is zero. I think, my brain hurts.

Regardless heat transfer really needs to be reworked and cleaned up.

[Gurgel]

Thanks for the experiments! In particular the influence of the weight on the tile is a nice find. 

As ONI seems to be using IEEE754 single (only 23 significant bits) for calculations, resolution is only about 1/8'000'000. Quite possibly this problem is caused because the difference added to the absolute temperature does not change the absolute temperature. 

I don't see how they can fix that except by going to IEEE754 double. That would probably slow things down significantly. 

 

[Yothiel]

"Decrypting heat transfer" guy here! I haven't really dived in those mechanics since then but for information, when experimenting ingame, I was using a temperature of "0" for one of the two bodies (i.e. 0°C or 0°F depending on your display mode, I don't remember if 0K was a valid value for non-vacuum cells).

This allows for a far better accuracy in the readout, since normal temperatures will only get displayed up to 1 digit, while near-zero will get displayed with something like 4 or 5 digits.

Also, low masses where possible are usually better to check heat transfer on a very low time scale.

 

In any case, I think IEEE757 track is the correct one here. In Saturnus' experiment. We have a very low heat flow (because of the low conductivity from isoled tiles, and a relatively low temperature delta of a couple thousand K) and a huge heat capacity (because of the amount of magma), so the temperature delta for magma is probably too small (i.e. the computation T'1 = T1 + ΔQ' / C1 will give exactly the same float values for T1 & T'1)

[wachunga]

@YothielThat thread has been a huge help to me btw. Still reference it nearly a year later. Some things have changed but most of it still seems accurate. So thanks for it.