Heat nonconservation on Conduction Panel

Can confirm, something thermally illegal here, even on 1x speed:

with 4 of panels, steam at 320C

with 5 of them at 380C

That looks far beyond theoretically possible 42 kdtu(which never achived), and i dont get from the fly what dependency is exactly

 

I third this. Also I found that an aquatuner can self-cool in vacuum using the panel (with super coolant going through the whole pipe loop). Eventually the aquatuner, the panel and the super coolant will be all at -271C, unless other objects involved.

Changed Status to Fixed

While the worst issues with the Conduction Panel have been fixed, the math still does not work out in a few cases. Specifically:

• If one Conduction Panel overlaps more than one building (you can put gas pipes, wires, foreground buildings, tempshift plates and so forth under the active tile), the heat transferred is incorrect
• If multiple Conduction Panels overlap the same building, the heat transferred is incorrect

In both cases, the problem is that the Temperature of a building (either the Conduction Panel itself, or one/more of its target buildings) is set twice in the same Sim tick. This posts a temperature update request to the Sim, but reading the Temperature returns the same value until the next Sim tick, so any changes from previous updates are not taken into account. Which building updates first depends on build order, memory layout and so forth, and can change across save/reload.

For example, this Hydrogen Generator overlaps two Conduction Panels built out of the same material which do not contain any liquids. One is at 11 C, the other at 26 C. At 17.7 C, one would expect the Hydrogen Generator to warm up, as it is farther away from the hot Panel's temperature than the cold one. However, the temperature (and Heat Energy from the enhanced popup) of the Hydrogen Generator is steadily decreasing, as the temperature update from the right side Panel is lost when the left Panel updates the Hydrogen Generator's temperature.

I discovered this same bug in the mod Thermal Interface Plate, and my fix (not yet published) was to batch the updates of all Thermal Interface Plates into a single class that could store all intermediate temperatures of each building and plate between Sim updates to ensure that calculation results are not lost. While it would be a huge rewrite to the Conduction Panel, it may be the only way to correctly transfer heat in all scenarios.

Fueling the fire here to highlight the issue that Peter Han pointed out. The issue with multiple conduction panels can be easily observed (and exploited) with self-cooled steam turbines (SCST) using input steam at absurd temperatures (above 300 C).

SCST Background. Recall that SCSTs self-cool by using their output water at 95 C to keep the turbine temp under 100 C. Theoretically, this should work with input steam up to 140 C; in practice, most optimized designs are reliable with input steam at 135 C.

Experiment (on build U44-537329). In the following two experiments, we have a steam turbine (aluminum) over steam at 300 C. The steam turbine is cooled by a sequence of 3 conduction panels (aluminum) followed by a valve to initialize the water storage (to avoid the bug with stored contents only staying for 1/5 ticks). According to both theory and practice, an SCST should not be possible with input steam at 300 C and the turbine should overheat.

To start the system:

1. set the valve to 0 g/s;
2. allow the turbine to pump water to the valve until the lines back up, i.e. the conduction panels are full of water;
3. set the valve to 2000 g/s to match the turbine outflow;

Test at 1x game speed. This result depends on game speed like many other game behaviors, so please be cautious with tab on game speed.

Results. The following two gifs show the results with the game speed at 1x speed and 3x speed.

1x speed.

Spoiler

At 1x game speed, observe three things:

• input steam at 300 C;
• steam turbine at 95.7 C (converges to 95 C);
• conduction panel at 95 C (appears stuck, likely due to bug pointed out by Peter Han);

This should not be possible.

3x speed.

Spoiler

At 3x game speed, observe three things:

• input steam at 300 C;
• steam turbine at 100 C (overheats);
• conduction panel at 100 C (quickly);

This result is included to show the dependence on game speed. In other experiments we have conducted, game speed has an inversion effect on the heating/cooling. For example, if a system using conduction panels is slowly heating up at 1x speed, then increasing the game speed to 3x speed yields slow cooling instead. Hence, we use 1x speed as the normative behavior and consider 3x speed as an additional effect.

Note. These results primarily apply to conduction panels in series (or in sequence). For conduction panels used in parallel, the results have more dimensions that I will omit from this bug report to keep it simple.

Edited January 18, 2023 by hydrotoast
Note series vs parallel conduction panels