Testing various Counterflow Heat Exchangers designs

[Fradow]

There isn’t much infos on the various designs of counterflow heat exchangers and how they perform, and nowhere could I find a rough size for a “decently efficient” heat exchanger, which is something I wanted to tell on my Steam guide (late-game part, will be published soon).

That’s what started my quest and I decided to test them, to see how they perform under the same conditions (as close as possible as common parameters).

 

The conditions I tested:

- Copper radiant liquid pipes. Gold and Iron are going to have very close performance.

- Strict 95°C Crude Oil input. The mid-point between the 90°C+ Oil Wells and 100°C water boiling point.

- Magma heat source (similar to a Core Tap or Volcano)

- “Decently efficient” heat exchanger: the goal was to test around 500/600kDTU/s, not the extremes.

- Same boiling chamber, without pipe heat exchange in them (403°C temperature sensor, which makes the boiling chamber around 408/410°C in practice).

 

I’ll anticipate a question: “What if you do X / use Y?”: in most cases I don’t know, tests would be needed. Tests take a long time to run, and I probably won’t run any more of them since I got the answers I was looking for. Feel free to run them yourself and report the result!

 

Edit: following several comments, I redid test with only the heat exchanger part, without a boiling chamber. See details below. I've made minor edits to this main post to reflect those results and add some important details.

 

Methodology to test a design:

- Always copy the boiling chamber / magma chamber / crude oil chamber to avoid any variation due to a different boiling chamber design.

- Fill the magma chamber with default Sandbox magma, fill the crude oil chamber with 95°C Crude Oil.

- Initialize it and let it run until it stabilizes (around 20 cycles). Remove any Sour Gas that could form during initialization, fix pipes and remove excess Petroleum as you go.

- Before you start the test, put the pipe thermo sensor to one degree of the actual Crude Oil temperature, and the automated notifier to “Pause”. Use a timer sensor + automated notifier for an accurate 10 cycles pause.

- Fill again the magma chamber with default sandbox magma, so everything is at the same temperature, in order to be able to calculate efficiency from the magma heat loss. Fill the crude oil chamber too so it doesn’t get depleted.

- Run the test for 10 cycles. Each time a thermo sensor pauses the game, put it one degree higher or lower, in order to measure the absolute range during the 10 cycles test.

- At the end of the 10 cycles, take note of both pipe thermo sensors temperature, and the opposite cells of the magma chamber temperature.

 

Here is the designs I tested:

- Layers: the well-known snaking layers

- Staircase: staircase to the left. It’s well-known staircases to the right are bugged.

- Waterfall: introduced by this post: https://forums.kleientertainment.com/forums/topic/119348-waterfall-petroleum-boiler/

- Z-shape: a new design theorized by @Saturnus after the test results of the first 3. Because I did not know about the staircases to the right bug, I tested them mirrored as well, which shows just how much the bug impacts efficiency.

 

“What about design Y?” See above: tests are needed, feel free to do them. @Saturnus have some answers on some details, namely:

- Bead Drop vs Waterfall: doesn’t make any difference This is debated by players more experienced than me. See @nakomaru answer below.

- Staircase incline: it lowers the efficiency of staircases.

 

First, let’s see the actual design tested for the 3 first designs:

 

And here are the results:

 

 

As you can see, each design has some pros and cons. It’s not like there is a perfect design on all metrics. For similar efficiencies:

- Layers have the best stability, by far, but the worst number of pipe segments and number of cells used, even though it’s competitive with Staircases if you don’t use Staircases dead spaces. This experiment also confirms that more layers is better than wider layers for efficiency. On the other hand, each additional layer delete a small amount of heat (see details in stand-alone heat exchanger), contrary to the other 2 designs.

- Staircases have the best efficiency and lowest number of pipe segments (especially as you could have only half the segment be radiant). Their main drawback is their diagonal shape and the fact they can only be built to the left. They also don’t have great stability.

- Waterfalls have the worst stability, and are roughly 8% worse about the same (see standalone heat exchanger experiment below) as staircase in number of pipe segments (and twice as worse if you only count radiant ones). Their advantage is their vertical shape and the fact they don’t need floor/walls, which can be a big advantage for space-constrained scenarios.

 

The stability is generally under-played, but it serves a very important purpose: a high stability allows to run a very efficient heat exchanger without breaking. 

Edit: the instability come from the heat differential from the heat source and boiling chamber and a spiky heat bridge. Instead of working on stability in the heat exchanger, you can also improve your heat bridge and boiling chamber to avoid any heat fluctuation, rendering the heat exchanger stability meaningless.

 

What if we could have the efficiency of staircases/waterfall with the stability of layers?

That’s what the Z-shape is for: combining layers and staircases for both efficiency and stability. Note that the staircase could be replaced by a waterfall for similar results. The original Waterfall article actually does that a bit, and I blame that short top layer for the misleading results there.

 

Here are all the test I did:

 

And here are the results:

 

Conclusions that can be drawn from that:

- Don’t do a left to right staircase. Just don’t. See this bug for more explanations: https://forums.kleientertainment.com/klei-bug-tracker/oni/left-to-right-liquid-flow-heat-transfer-bug-r25140/ The stand-alone heat exchanger results below shows just how much heat is deleted.

- For stability, the top layer is the most important, by far. You can remove the bottom layer without seeing much impact on stability, though it will obviously lower efficiency a bit. Having a single top layer also reduce the amount of heat deleted compared to layers.

 

- A width of about 15 is a good sweet spot in my opinion, with diminishing returns above that, especially above 20. This would need more precise testing to be sure.

- Having several stacked Z-shape isn’t great, especially as it impacts the width of the top layer.

 

I fully expect very similar results by replacing the staircase with a waterfall.

 

As for using the dead space on either side of the staircase, that’s up to you to decide how to do that. With some heat sources (volcano, aquatuner), you can tug them inside the top part. You could for example have all or part of your Petroleum power plant inside the bottom part.

 

Results as a Sheet format: https://docs.google.com/spreadsheets/d/1zpe5I14U4oezAq0V5kUG6v2nh0jVp52w42-Moo4Q-i0/ 

 

Here is the save game with all the designs built. You also need to use the heat gun on the Steam Chambers from time to time, since I don’t use Debug mode.

Note that designs are no longer considered initialized. You have to re-initialize everything again to have meaningful numbers.

Pet boiler sandbox Cycle 145.sav

[nakomaru]

You put in a lot of work for this, great job. I hope you don't mind me giving some feedback.

Methodology:

• I highly recommend you remove all boilers when testing a heat exchanger, especially because they have variable input in your tests. But also because they conduct heat directly into your flow. The boiler design can be studied separately.
  • Instead, pick an input oil temperature and an input petroleum temperature. Send both through your exchanger, either through pipes or by vents. Then you can measure the output temperature of both.
• With an input and an output temperature, you can rate the efficiency of your heat exchanger for dead simple comparison between heat exchangers.
  • If your heat exchanger is 100% efficient, your output oil will be an equal temperature to your input petroleum.
  • If your heat exchanger is 100% efficient, your output petroleum will be equal to the input petroleum's temperature, minus the heat energy difference in crude oil temperature.
  • For real efficiencies, calculate ΔT_Actual/ΔT_Ideal or Q_Actual/Q_Ideal.
  • You can calculate this for the total heat transfer of the system or for each oil and petroleum. All methods should give you the same efficiency, otherwise, your boiler or maybe a bug has influenced your data, or maybe the system is in transient state.
  • For help with the calculations, see this post.

For designs:

7 hours ago, Fradow said:

Bead Drop vs Waterfall: doesn’t make any difference

Waterfalls seem to perform slightly better using equal space for reasonable lengths, but beads have perfect backflow isolation. The most efficient designs will isolate backflow, as with staircases or beads.

7 hours ago, Fradow said:

- A width of about 15 is a good sweet spot in my opinion, with diminishing returns above that, especially above 20. This would need more precise testing to be sure.

I think you might be way over target here. In my experience about 5 is ideal due to the above point, but I was using different materials so more study is needed.

Finally, I think this might have been beyond the scope of your research, but massive improvements (on the order of threefold) can be made by using bridges and tiles in clever ways even with just a waterfall. See links above.

[Fradow]

47 minutes ago, nakomaru said:

You put in a lot of work for this, great job. I hope you don't mind me giving some feedback.

Thanks! Of course I don't mind, though since I achieved my goal and already spent too much time on this, I won't be running more tests.

47 minutes ago, nakomaru said:

Methodology:

• I highly recommend you remove all boilers when testing a heat exchanger, especially because they have variable input in your tests. But also because they conduct heat directly into your flow. The boiler design can be studied separately.
  • Instead, pick an input oil temperature and an input petroleum temperature. Send both through your exchanger, either through pipes or by vents. Then you can measure the output temperature of both.
• With an input and an output temperature, you can rate the efficiency of your heat exchanger for dead simple comparison between heat exchangers.
  • If your heat exchanger is 100% efficient, your output oil will be an equal temperature to your input petroleum.
  • If your heat exchanger is 100% efficient, your output petroleum will be equal to the input petroleum's temperature, minus the heat energy difference in crude oil temperature.
  • For real efficiencies, calculate ΔT_Actual/ΔT_Ideal or Q_Actual/Q_Ideal.
  • You can calculate this for the total heat transfer of the system or for each oil and petroleum. All methods should give you the same efficiency, otherwise, your boiler or maybe a bug has influenced your data, or maybe the system is in transient state.
  • For help with the calculations, see this post.

Duly noted. My goal was to test things under a realistic scenario and pit designs against each other rather than have a perfect heat exchanger or perfect data. I doubt results would change significantly by using your methodology, though I don't dispute it's better.

I took special care about letting every boiler initialize for long enough to be stable and to have the exact same heat source, boiling chamber design and temperature, meaning any impact on results will be minimal, and won't change the conclusions.

47 minutes ago, nakomaru said:

Waterfalls seem to perform slightly better using equal space for reasonable lengths, but beads have perfect backflow isolation. The most efficient designs will isolate backflow.

Since Saturnus and you have conflicting results on that, I don't know who is right. I'll edit the original post to say it's debated.

47 minutes ago, nakomaru said:

I think you might be way over target here. In my experience about 5 is ideal due to the above point, but I was using different materials so more study is needed. This is a matter of opinion, but segments which are not conducting (going up and down in a drip design) can be made of almost any material, so probably shouldn't be counted.

Aluminium is indeed going to make a huge difference (which is what I assume was used in the post you linked, since it's spelled out several times). 5 Aluminium segments is about 15 Copper/Iron/Gold segments when going by the TC values, so we actually got about the same target! I used Copper because my guide is for Terra, and Aluminium isn't available there.

47 minutes ago, nakomaru said:

Finally, I think this might have been beyond the scope of your research, but massive improvements (on the order of threefold) can be made by using bridges and tiles in clever ways even with just a waterfall. See links above.

Indeed, that's beyond the scope of what I tested, since I wanted to keep to simple designs. Even using a waterfall was barely in scope.

[wachunga]

I second @nakomaru's comments regarding testing methodology. Try something like this and see what you get:

The flats in the Z design are deleting heat by the way. Doing the SHC math will show this, it's only a couple degrees worth but FYI. The stair and waterfall are bang spot on regarding conservation of heat energy.

Then once you have your "spherical cow" numbers, play around with boiling chamber designs to see how close you can get to that ideal. Frankly, the drip into a petroleum bucket is pretty bad. It has rather a burpy outflow which is biasing your results. It's not particularly difficult to design a chamber that outputs petroleum that is smoothly consistent in both flow and temperature. If one wanted to be crazy, they could flake it like @mathmanicandid here. That doesn't seem to want to link properly, look on page 3 of that discussion if interested. He probably has a better design by now anyhow.

[Craigjw]

Your boiler design is sub-optimal, instead submerge the liquid vent instead of dripping it in from above, submerging stops a bunch of bad headaches happening, caused by dripping it in from above.

Most people don't take my advice and doubt the efficacy of submerging it, then after countless headaches will then try it and realise, it's just better that way.

Obviously you've had some oil overspill into the heat exchange channels, otherwise they wouldn't be 2 tiles high in order to allow dupes access to mop up oil.  Well, that just doesn't happen if you submerge it.

[mathmanican]

1 hour ago, wachunga said:

they could flake it

The nice part about flaking it is you avoid just about all of the complications of burping boiler.  

I also spent a bit of time today building some boilers.  95C crude, 401C petro. 

For me, both the waterfall and stair design collect crude at the top at 371.8C.  The beaded design is actually worse (temp only reaches 370.8).  My guess is that the beaded version is worse because my radiant pipes go all the way to top of the beading, and if I dropped things one tile I would see better. Oh well. 

What I got from this is that the right-to-left stair is essentially the same (per pipe segment) as the waterfall).

If the real goal is to use as little magma as possible to boil stuff, and you're willing to design something to pump >400C petroleum (using pump dynamics in a clever way, or have a space material pump), then THE BEST WAY to boil stuff is a left-to-right stair design (with everything flipped from normal where the heating occurs at the bottom, not top). 

The BUG is not a heat deletion bug, rather it amplifies heat difference because of an average value issue.  As such, you can generate extra heat, rather than delete it, if you flow the crude in the open against upflowing petro. 

FYI, the crude here that collects at the bottom is right around 397.9 (with 401C petro). That hands down beats anything else (though clearly uses a bug to do this). The outflowing petro is much warmer, so if you care about the temp of the outflowing petro, then this design is garbage.  If you want to maximize magma usage, then this design is your friend.  By the time you have space materials to pump 400C magma, this would drastically reduce aquatuner power consumption, but power is not an issue by then. 

[mathmanican]

If anyone wants an examples of a flaking boiler design, here's a basic prototype.  Just heat the steam (however you want). I didn't optimize it here, rather just wanted to slap 3 down to make some quick comparisons. 

 

[wachunga]

1 hour ago, mathmanican said:

burping boiler

Alas, apparently you need 1000 posts to set a custom title. If any Klei folk are watching, I hereby request "Burping Boiler" for myself.

 

I quite like the benefits of flaking. I'll have to try it. I'm thinking a small ledge maybe 4 or 5 tiles wide catching the beads and easing the transition from the per second vent regime to the per tick free liquid regime. Ought to improve exchanger efficiency a slight bit as the vent can be thought of as a burping boiler needing some smoothing out.

[nakomaru]

I tried to make a flaking boiler in January and didn't know about that part of the thread. I basically made the same design, though I might have used a ceramic tile. I also buffered the input to let me run 10020g/s for 99.8% of the time.

The problem I had was, most of the time the 10g of oil would get deleted, but small bits would occasionally accumulate. Did you ever solve this or does it just not happen for you?

[mathmanican]

1 hour ago, nakomaru said:

the 10g of oil would get deleted,

This always occurs because upon flaking this excess above 5kg is thrown in other directions (split) so each blob is under the 10g limit. If you have two beads close then there is a chance the thrown beads of crude might survive. As such you have to plan for possible crude in your petro collection zone.

It's a 0.02% loss in mass. Worth it for the savings in other ways.

[nakomaru]

I'm happy with the deletion, it's the undeleted oil that falls down which bothers me. I guess you'll have to add storage for it at least.

[mathmanican]

46 minutes ago, nakomaru said:

it's the undeleted oil that falls down which bothers me

I've ran this for several cycles (element sensor set to crude to guarantee only 10kg packets pass the shutoff). No crude has ever appeared. 

However, we could design a crude catching region just in case. This would require making sure the steam temp for flaking never gets hot enough to heat the petro enough to phase change the crude catch region. Definitely doable. 

[wachunga]

I couldn't find the thread discussing the fixed flaking behavior. I know the math works out correctly but I forget how the to flake or not to flake cutoff points are determined. In the case of crude flaking to petro, I'm seeing the cutoff point being whatever causes a 10C drop in temp of the donor. Is that correct or am I doing something weird? For example, a 200kg igneous tile dropping 10C gives up enough heat to flake 5kg of ~166C crude. Actual temperature is a bit less since conduction brings it up to the magic 166C point. Colder crude does not flake regardless of the igneous temperature. Is that what everyone else is seeing or is my version busted?

With that in mind and looking at bootstrapping the boiler, we really want to be able to flake cold crude. Which means more mass/SHC in the donor, since donor temperature seems irrelevant to the cutoff point. Which means an insulated tile, but those are slow to dump heat into. I put as many bridges and plates contacting the insulated donor as I could and ended up with this:

So far it's working great at a constant 5kg/s flow of 388.6C crude. I got the intermediary steam chamber down to about 500C so far while still dumping enough heat into the insulated tile.  I've done better with conventional boiling chambers, but that's still pretty decent. Startup involved bumping the steam temperature higher and only sending every 8th, then 4th, then 2nd crude packet while the exchanger warmed up.

[mathmanican]

29 minutes ago, wachunga said:

the thread discussing the fixed flaking

 

[Fradow]

Thanks a lot for all the comments, it's appreciated.

There are a lot of good points made that I didn't know previously, and it's certainly going to be useful to whoever lands there in the future.

On the other hand, while the experiments were not perfect and should be a few % off, I didn't see anything in replies that would indicate a critical flaw (unless the blurping causes more than a few %, but the linked post doesn't indicate numbers, so for now I guess that's a no).

 

Discussions on boiling chamber design, bead drops, flaking are not things that I want to experiment with at the moment, but feel free to continue the dicussion without me I'll be reading with interest.

[wachunga]

@mathmanicanThank you, I see your discussion of the 10C rule now.

[Saturnus]

39 minutes ago, Fradow said:

Discussions on boiling chamber design, bead drops, flaking are not things that I want to experiment with at the moment, but feel free to continue the dicussion without me I'll be reading with interest.

As I was part of the discussion on Discord, I know that the goal of the experimentation was to test designs that are easy to implement for the regular player in a survival game. Not a theoretical discussion on the most efficient designs.

The goal was to serve as an alternative and counterpoint to highly popular but sadly wildly inefficient and/or overly complex designs popularized by several youtubers. Something that any regular player could easily build and get a good result. I think that point somehow got lost on several people in this thread but it's probably because they did not know the actual design criteria and goals.

[nakomaru]

5 hours ago, mathmanican said:

(element sensor set to crude to guarantee only 10kg packets pass the shutoff)

This was also true for me, but unfortunately I didn't notice that when 10kg stops coming through.. I was just sending out the remainder. I was sure the flaking physics were unreliable so I gave up.

Fixing that, I also observe long term reliability. Still, it's nice to have a backup. This setup also works well, and the 10g of naphtha in the middle partially insulates the oil from boiling.

Spoiler

Drip (I prefer this)

Bead:

[mathmanican]

10 hours ago, Saturnus said:

I think that point somehow got lost

Data was presented about counterflow heat exchangers. However the boiler itself causes problems. As such we shared data independent of what was presented which contradicts the presented data.

We began a discussion because the provided data was not reproducible. The boiler makes a difference.

I'll move beaded flaking to a new thread.

[Fradow]

11 hours ago, mathmanican said:

Data was presented about counterflow heat exchangers. However the boiler itself causes problems. As such we shared data independent of what was presented which contradicts the presented data.

We began a discussion because the provided data was not reproducible. The boiler makes a difference.

I just published my guide yesterday, and since you insist on that, I'll take the time to run the tests again with a better boiler design. I'm determined to test it everything with a boiler because I find it closer to survival conditions.

I'll try to test it with only Petroleum/Crude Oil afterwards; as you suggested.

If I read the linked thread correctly, using Angpaur design (since I need a Thermo Senso with magma), does that boiler modification look like it would provide accurate data? I'd prefer keeping the lip, since it's required for waterfalls and I'd rather keep the same design before radiant pipes. The left-to-right heat bug should only impact data minimally with a single step.

I'll await a confirmation before beginning, since tests are lengthy to run.

[nakomaru]

2 hours ago, Fradow said:

I'll take the time to run the tests again with a better boiler design

This statement alone should make you realize why you should decouple the boiler from the test of a heat exchanger.

If you can't be persuaded, at least raise the output by 1 tile so that it drips into your exchanger. You can add any extra stuff you need to create a waterfall after that.

[Saturnus]

6 hours ago, Fradow said:

I just published my guide yesterday, and since you insist on that, I'll take the time to run the tests again with a better boiler design. I'm determined to test it everything with a boiler because I find it closer to survival conditions.

I'll try to test it with only Petroleum/Crude Oil afterwards; as you suggested.

If I read the linked thread correctly, using Angpaur design (since I need a Thermo Senso with magma), does that boiler modification look like it would provide accurate data? I'd prefer keeping the lip, since it's required for waterfalls and I'd rather keep the same design before radiant pipes. The left-to-right heat bug should only impact data minimally with a single step.

I'll await a confirmation before beginning, since tests are lengthy to run.

You can always test a few boiler design to see the output temperature, and if it's burping or not without running a full test with the heat exchangers.

One thing I've found makes a difference is to ditch the temperature sensor, and use a hydrosensor instead and infer optimal temperature from that. If it's above 740kg/tile (I use 750kg setting) then the tile must be crude oil, otherwise it's petroleum. I find that much more accurately controls the temperature to just enough to boil the crude oil without wasting heat as using a temperature sensor does. It will burp in the beginning though.

[Fradow]

Thanks for the comments. I finally realized I'm way over my head with boiling chamber design and it's best to leave that part out, so I've changed my mind and will hear your advice @nakomaru and test only the heat exchanger part, using the methodology you described in your first answer.

Results coming soon.

I'll be reading with interest if someone decide to put all those boiling chamber design to the test.

 

 

[Fradow]

And I’m back with the results!
Long story short, there is no significant difference in performance compared to the previous experiment with coupled parts.

For this experiment, I kept 95°C Crude Oil and used 408°C Petroleum (about the same temperature as boiling chambers in the previous experiment). The experiment was only runned for 5 cycles after stabilizing heat exchangers, which was more than enough to get stable values.

Here are all the designs (not numbered, they have the same position as in the first post):

Sorry about the lack of readability, I had to fit crude and petroleum tanks wherever I could without altering heat exchangers.

Here is how they performed:

See below for efficiency numbers calculated by @nakomaru according to Crude Oil output temperature.

 

I hope I didn’t make a mistake when calculating KDTU/s: I’m not sure how much heat is used during state change. I saw a 2.5 x crude SHC in @nakomaru spreadsheet and decided to use that, but couldn’t find any source for that.

Screenshots updated to use the proper number of 1.5 x crude SHC as pointed out by @nakomaru

I didn’t put any stability number: that’s because all heat exchangers were mostly stable when using a fixed Petroleum temperature rather than a boiler with a fluctuating temperature. On the other hand, while it was too low to measure, I could still see the same trend as in the first post: the first decimal number was mostly fixed for layers, but varied widely for waterfalls, with staircases in between. In my opinion, that means that some designs' poor stability can be counteracted by having a very stable boiling chamber (using a more precise heat bridge and/or a heat source with less heat differential).

Another very interesting fact, although that was to be expected, is how visible the heat deletion bug is, and how it affects each design: while more layers are better than wider layers, they also delete a bit more heat, though I consider that negligible.That’s also something Z-shape performs well on as it only has 2 layers, and even better if you only have a top layer.

When going into the details, we can see that waterfalls were negatively impacted by the burping boiler, by about 17%, while layers were the least impacted.

In particular, this means my previous statement about waterfall being 8% worse than staircase is invalidated as both performed exactly the same this time. For now, I’ll think of them as “roughly equal”, with waterfall performing worse in some specific cases (and perhaps better in others) as heat exchangers.

 

If anyone wants to re-run the tests, here is the save file when the experiment ended:

heat exchanger only end exp.sav

I used the same spreadsheet, here is the link again if someone wants to delve into the numbers and formulas: https://docs.google.com/spreadsheets/d/1zpe5I14U4oezAq0V5kUG6v2nh0jVp52w42-Moo4Q-i0/edit

 

[nakomaru]

Nice work again @Fradow. Sorry for misleading you - it's actually 1.5K*SHC that is lost after phase changing upwards. (The new material spawns in 1.5K less than the temperature of the previous material.) But this is the same for everyone except maybe for flaking boilers (?). Apparently it doesn't happen when you melt debris, but that's not applicable here.

Personally, I don't think the stability is usually too big of a concern. At high efficiencies you recover almost all of the extra work you accidentally do in your boiler, and your output will always stabilize in your tank. Still a nice thing to be aware of when going for the best design possible.

For 95C and 408C inputs, a 100% efficient exchanger would have 408C and 107.4C outputs. I think you can tell how bugged the staircase is once again by looking at those 363.7C & 99.5C outputs.

Just considering oil output (it's more important for determining how much work your boiler has to do, usually it matches petrol but is bugged when the game doesn't conserve energy), here's the efficiency of those designs: