Any explanations for this observed slow cooling effect?

[allmhuran]

I was playing around with some part files to investigate the stock thermal system, and observed something curious.

I adjusted an engine to the following parameters:

mass = 0.1

maxTemp = 2250

thermalMassModifier = 125000

heatConductivity = 0.05

emissiveConstant = 2 // yes, double what is physically possible. I also tried 1, no appreciable difference.

radiatorHeadroom = 0.2 // I still don't think there's any official explanation as to the effect of this one

...

heatProduction = 10 // this is a very low value

I connected this to another part with a mass of 3, an equally high thermal mass modifier, and a conductivity of 0.1 (low, but higher than the engine).

When I turned on the engine and put it at 100% throttle, sure enough, its temperature rose, albeit very slowly. I turned on physics warp and let it climb to 500 Kelvin. Not super hot, but "hot enough". The int flux value in the debug menu was 2.4 million. It took about 5 - 10 minutes to reach this temperature.

Then I throttled down the engine to zero. The int flux went to zero (not negative?). I turned on 4x physics warp and left it for an hour. Cond flux was negative 740-odd during this phase. When I came back the part had only dropped 0.3 degrees. Now, that seems odd to me. That's essentially four hours passing in "real time", and yet the engine has barely cooled down at all. Just from every day experience I think people would agree that doesn't seem right, think about your stovetop, your car engine, etc. Your car engine might get to ~400K and it's trapped in an enclosed space, but turn off your car (which will obviously also stop any active cooling from the radiator fan and pump) and it will probably drop 20 to 30 degrees in half an hour.

Edited September 2, 2015 by allmhuran

[sal_vager]

Well, if you change things without knowing what's going on it's going to do some odd things

It's all Dutch to me, but I do know this isn't suited to Support, moved to development discussion.

[allmhuran]

Not sure what you mean by "without knowing what's going on". In the sense that I don't have the source code? True, no idea. In the sense that I understand that something 200 degrees above ambient should be radiating heat pretty damn quickly into the environment, and I've done tertiary physics? That would be a different answer.

I bumped the conductivity of both the engine and its parent up to 0.5. Now the cond flux of the engine went to about -35000 at ~500K. But the rate at which the temperature of the part increased did not noticeably change from the previous test, nor did the rate of heating change as the part temperature rose (rate of heating should have slowed as the temperature rose relative to its surroundings). Also curious is that conv flux was +3 and rad flux was only -1. That' can't be right... the atmosphere should be convecting heat away rather a lot more than 1/10000th (and with the wrong sign?) as quickly as the connected part is conducting it away.

No doubt the game is not modelling convection from air currents, such as wind (there isn't any!) or the movement of air generated by the hot part itself. But putting in some kind of fudge factor to approximate that might be a good idea.

Edited September 2, 2015 by allmhuran