Fluids, Pipes and Pumps

[Kamikatze]

Hi everyone,

I just tested pumping water from one tank to another and count the ticks it takes with combinators and stuff.

Blueprint String

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

I believe the results might be off by 1 or 2 ticks because it doesn't stop counting exactly on the tick when the tank gets full.

If I use one pump connecting both tanks directly, the combinator reads 181 ticks,
so I'd say the pump can theoretically transfer 25k of fluid in ~3 seconds which means 138.12/tick or 8287.29/s (The tooltip says 200 pumping speed, which apparently means per tick, but I guess the in-/outputs can't keep up)


Then I used one pump with a pipe at the input...


...and then 2 pumps:

I tried this with 0, 1, 2 and 4 pipes (pipe-to-ground counts still as 2 normal pipes[tested]) and later with more.

Summary:

The results for one pump factor the tank in, which seems to count as something between 2-3 pipes.
When i used 15 pipes and one tank it took 1246 ticks (1203.85 fluid/s), so there it was ~3 pipes, but i haven't tested with different fill levels.

Since the pumps aren't the bottleneck anymore, it makes no sense to use pumps parallel now:

This setup takes 618 ticks to fill the tank, but if i remove the upper and lower pump, it takes 642 ticks, so it's really not worth it.

It looks like the sweet spot is 18 pipes between pumps for 1200 fluid/s (which the offshore pump can produce), so 18 pipes-to-ground would get you 99 tiles far.

[MeduSalem]

Oh well good to know.

I already thought that the pipes are now the bottleneck and not the pumps. I figured as much while messing around with Nuclear Power setups, but I didn't make any notes or hard numbers.

So basically one wants to avoid having too many pipes... but at least a single pump every once a while is now enough to keep up the pressure.

[Mehve]

Very interesting. I botched my earlier testing when I never accounted for the pumps exceeding the offshore output, but this looks to be done much better. I'm hoping to put together some setups that can use multiple offshore pumps, I'll definitely be keeping these numbers in mind.

[BlakeMW]

It looks like the sweet spot is 18 pipes between pumps for 1200 fluid/s (which the offshore pump can produce), so 18 pipes-to-ground would get you 99 tiles far.

Confiremd in sandbox. I made a 40 steam engine (1200 water/s) setup, 36 pipe-to-grounds from the pump.
Placing a pump at the 18th and 36th pipes allowed full pressure.
Placing a pump at the 17th pipe also allowed full pressure

Interestingly though, the degradation is not that bad, I tried some different lengths of pipes (more-or-less at random) and checked how much power was produced (full power = 36MW):

• 36 pipes: 32.7MW : 1090 water/s
• 187 pipes: 30.5MW : 1020 water/s
• 281 pipes: 22.4MW : 746 water/s
• 563 pipes: 11.8 MW : 370 water/s

So as in previous versions, if you're willing to settle for 1000 water/s rather than 1200water/s you can go 10x further, in general it would be expected that parallel pipes will be a more effective long distance volume solution than pumps every 18 pipes - unless space is *extremely* constrained (which seems inconsistent with long distance...). And these throughput limits are really no constraint at all on piping oil unless the pumpjack output or distances are absurd.

Note: As always with pipes, placement order, flow direction and chunk boundary crossings do effect the real flow rate. Therefore actual figures may be +/- a few percent.

[Kamikatze]

187 pipes-to-ground would reach over 2000 tiles, someone should calculate how many fluid wagons a train(and which fuel of course^^) would need to transport stuff over 2000 tiles at 1000fluid/sec

[MeduSalem]

187 pipes-to-ground would reach over 2000 tiles, someone should calculate how many fluid wagons a train(and which fuel of course^^) would need to transport stuff over 2000 tiles at 1000fluid/sec

Ugly to lay that many underground pipes.

[cerberusti]

I had a bunch of similar test setups, where it counted the time to fill 20,000 capacity of the tank.

My tank to tank transfer through one pump and nothing else consistently came in at 135 ticks. Two pumps at opposite sides can fill 20,000 in about half that (but on the same side still take as long as a single pump). The time to fill from a single pump to 10,000 supports a pump throughput of 200 a tick with a falloff with pressure differences (12,000 per second). The tank also seems to have a similar limit for each input side. It maintains that 135 ticks to fill even through five connected pumps.

With one pipe in the way 227 ticks is the best I could manage to fill 20,000, with a setup of tank pump pipe pump tank (5286 fluid per second).

The next best was tank pump pipe tank at 416 ticks (2885 fluid per second).

Both an offshore pump as well as an offshore pump with a pump attached fill at a rate of 1,200 per second. It looks like the pump can move 10x the fluid of the offshore pump, but the only item with enough capacity to do that is the tank.

[orzelek]

This would also explain why when pumping to rail tanker it works best when you attach pump directly to tank. It can pump it all into wagon very quickly - under 5 seconds for sure. Good research here
(Not that I would want to use 200 underground pipes in a row )

[Engimage]

This confirms what we were saying in old times - you should never try to "pull" water with a pump from a pipe if that one is not fed by another pump. Natural water flow (which is the case when a pipe is connected to a tank) is the worst possible bottleneck. So the best practice is to connect pump directly to a tank.

[Jarin]

So I'm having a bit of trouble following here. What's the best bet now for moving fluid over distance? In my case, I'm running short trying to feed these new heat exchangers, if not right next to the water.

[terror_gnom]

What's the best bet now for moving fluid over distance?

In my opinion: Barrels!

1 blue belt can feed roughly 1 GW of power and bots are extremly efficient transporting them over short to medium distances inside the power plant

[iceman_1212]

Nice analysis - good to know for 0.15.

187 pipes-to-ground would reach over 2000 tiles, someone should calculate how many fluid wagons a train(and which fuel of course^^) would need to transport stuff over 2000 tiles at 1000fluid/sec

1000 fluid per second throughput is achieved if a train with 2 fluid wagons (LCC or LCCL) passes a point every two and a half minutes, or a train with 4 fluid wagons passes a point every five minutes. One fluid wagon holds 75k fluid. Two such wagons passing a point every two and a half minutes = 150k fluid / 150 seconds = 1000 fluid per second. Neither of these are unreasonable.

[BlakeMW]

187 pipes-to-ground would reach over 2000 tiles, someone should calculate how many fluid wagons a train(and which fuel of course^^) would need to transport stuff over 2000 tiles at 1000fluid/sec

187 pipes actually stretches about 1000 tiles, because each pipe-pair is 11 tiles, or 5.5 per pipe-to-ground.

I actually tried it in game to see how fast a 1-1-1 train could do a round trip over 1000 tiles between pumping stations (max tech, solid fuel, 2 pumps at each station): it came to 92s per round trip which would be 815 fluid/s. A 1-2-1 would have nearly double the fluid/s.

So I'm having a bit of trouble following here. What's the best bet now for moving fluid over distance? In my case, I'm running short trying to feed these new heat exchangers, if not right next to the water.

For feeding power stations, parallel underground pipelines will usually be the most logical solution: generally one offshore pump to one pipeline, and assume the pipeline will carry ~1000 water/s for ratios (increasing that to 1200 water/s isn't really worth it, unless the distance is very short). If for some inexplicable reason you need to transport the water more than ~1000 tiles, add repeater pumps.

It would not technically be impossible to transport water by train, but you'd need a vast amount of infrastructure (i.e. large arrays of storage tanks to buffer the trains) and I can't see what the benefit would be over the low-effort and highly efficient pipelines.

[Kamikatze]

187 pipes actually stretches about 1000 tiles, because each pipe-pair is 11 tiles, or 5.5 per pipe-to-ground.

ohh right, forgot to cut that in half. *whoops*