Long distance pipes and pumps

[NoriSilverrage]

Just out of curiosity, is the amount of fluid that can move through a pipe related to the pipe's size? For instance, several mods add new pipes that have increased storage. If a pipe could store 20 instead of 10, does this mean more fluid can travel through the pipe?

[BlakeMW]

It does. So will longer underground pipe lenghs.

I don't like to speak of pressure: In Factorio pipe capacity maxes out when the pipe becomes backlogged, that is when it becomes 10.0/10 full at the producers (i.e. pumpjacks) and thus doesn't allow any more liquid to be added at that point. A larger pipe will take more liquid before it backlogs: A pipe at 10.0/20 still has plenty of capacity.

[ribsngibs]

Why do the pumps need to be at right angles? Why not just put them in a line, then connect the inputs and outputs together with pipes?

Water pumps don't have inputs you connect to - only output. Small pumps have the input in the rear and output on the front.

Which did you mean? can you show a picture to demonstrate?

Not the person you replied to, and many months later... but I think what he means (and I have the same question), why is the left superior to the right?

pumpstation.jpg (120.8 KiB) Viewed 19080 times

[BlakeMW]

From my own testing it actually isn't. Both perform identically and as such the design on the right is superior because it is simpler and more compact.

I had to test it because everything I know about factorio fluid dynamics says there should be no difference between the two designs and my experimental setups have failed to expose any difference in favor of either setup.

For what it's worth, here is the same "25 steam engines, 1 pipe" image but using the "naive" (sensible?) pump setup.


As you can see under full load all the steam engines run at 100% performance so the 5 pumps in a line setup is doing a perfectly good job of delivering 150 units of water through the stretches of 4 pipes. (and I set the game speed to 10x and let it run for a while to stabilize first)

I'd be welcome to be proven wrong on this if anyone can come up with a test where the DaveMcW design actually delivers a superior throughput. But my personal testing and theory so far says there is no difference whatsoever.

[DaveMcW]

I'd be welcome to be proven wrong on this if anyone can come up with a test where the DaveMcW design actually delivers a superior throughput. But my personal testing and theory so far says there is no difference whatsoever.

You're right, if you are optimizing for a pump breakpoint, many designs work and you can pick the prettiest one.

I originally tested the pumps past the breaking point, extending the pipes until performance started dropping. Something like this:

fluid-test.png (1.78 MiB) Viewed 18947 times

If you mouse over a radar and run this command, you will see the performance is better for my design.

Code: Select all

/c game.local_player.print(game.local_player.selected.energy)

[BlakeMW]

Okay I did some extensive testing to try and confirm or eliminate a hypothesis, I've observed in the past that that pipe placement order can have at least a fraction of a percent impact on performance, my steam powerplants built close to limits often work 100% with a one pipe placement order, and only about 99% with another pipe placement order so this effect is large enough to be observable without using the console to perform measurements and a test needs to account for it when making very fine measurements.

This is the basic test setup I used (note the small pump flow-limiters after the offshore pumps):


And this is the alternate pumping setup:


I'm going to use the word "prograde" to refer to water flowing in the same direction as the pipes are placed, and "retrograde" to refer to flowing against the direction the pipes are placed in. I ran each test twice, in the first I placed the pipes from south-to-north, in the second from north-to-south. Note the pipe-to-ground is so I can account for a longer or shorter pumping setup while still having 14 pipe segments in series between the pumps and the steam engines. I adjusted the game speed to run the game for long enough for the setup to thoroughly stabilize between measurements and took measurements over a period of time to make sure they were stable (they are, at least to 2 decimal places).

Anyway without further ado, here are the results, the number is a radar energy readout (joules/tick) from

Code: Select all

/c game.local_player.print(game.local_player.selected.energy)

Simple Design:
Prograde: 4903.17
Retrograde: 4901.83

DaveMcW design:
Prograde: 4900.60
Retrograde: 4903.14

The results are inconclusive, except to say that pipe placement order can definitely account for a discrepancy making one design look better than another. The simple design did perform ever so slightly better on average but I wouldn't put too much weight on that because the sample is too small to be statistically significant. I don't have the foggiest clue why the performance is inverted, with the simple design working best prograde and Dave's design working best retrograde, I wouldn't put to much weight on that either because other factors in the experimental setup might make that distinction go away.
Note that while prograde consistently results in much lower latency (liquid will propagate through an empty pipe about twice as fast) it's effect on throughput is inconsistent. It's definitely not correct to say that placing pipes prograde will improve flow, altough at least sometimes it will. Also it's not only the order the pipes are placed in, it's also the order everything else is placed in relative to the pipes, presumably some precise placement order will give optimal results but I don't have a theoretical basis for saying what that order would be.

Anyway my conclusion for now is that the simple design is generally the best one to use since it definitely did win in this controlled test. Altough perhaps if a pipeline is placed retrograde (like running a pipe from the refinery to the oil field rather than vice-verca) the alternative design would work better.

[Karelipoeg]

Is the throughput of 30 fluid/sec minimal, or does it go even slower beyond pipe length of 759?

[kinnom]

Is the throughput of 30 fluid/sec minimal, or does it go even slower beyond pipe length of 759?

slower

[mrvn]

fluid-test.png

What are the numbers for the boilers? Do they behave like a pipe in throughput? How much do they heat up water at what throughput? Has anyone tested this?

[Acarin]

It has been tested somewhere, I am fairly sure, because I seem to remember that as being the source of the oft-quoted 1:14:10 ratio (offshore pump:boilers:steam engines). However, you would need to do a search for the post - I am otherwise occupied at the moment and don't have time to trawl...

[mrvn]

The 1:14:10 ratio comes from the fact that you can power 10 steam engines from one offshore pump and you need 13.x boilers to make the water 100° hot. I think boilers add energy to the water that passes through.

But I was asking about maximum throughput. Can I connect 2 pumps to one boiler row? 3? 4?

[Duegin]

A quick experiment shows you can certainly connect more than 1 pump to a boiler in order to get additional fluid flow to your steam engines. However it appears that like pipes, boilers and/or steam engines don't have infinite flow capacity. It would take more targeted experiments to gather evidence that they are identical to pipes, but my experiment showed I could get 6 pumps in parallel running through 1 boiler to supply 26 steam engines in series with > 0 water each at steady state. If steam engines consume 6 units of water/sec each then the boiler in this configuration appears to be sustaining 26*6 = 156 units/sec. And of course this is a lousy setup for power generation as the temperature is too low. In order to keep it at 100 I would hypothesize you will need to preserve the 13.x:10 ratio of boilers:steam engines as you scale up the water flow.

Water -> Pump -> Pipe
Water -> Pump -> Pipe
Water -> Pump -> Pipe
Water -> Pump -> Pipe
Water -> Pump -> Pipe
Water -> Pump -> Boiler -> Engine -> Engine -> ...

[mrvn]

You have a pretty unsymetric layout. Water flow is somewhat tricky and symetric setups seem better working. Merging output one by one into a pipe seems to be the worst idea from my experience. Have you thought about feeding the boiler from 3 sides with 2 pumps each?

Code: Select all

  P  P
  +++
P+ |
 +-B--SSSSS
P+ |
  +++
  P P

[DRY411S]

Apologies for resurrecting an old topi.

Is the loss of pressure linear between the steps, of on some kind of parabolic curve?

[mrvn]

Seems to be linear. The amount flowing depends on the difference in pressure between the 2 pipes.

On the other hand there seems to be some inertia or precedence. If you split a pipe then the straight seems to get more flow or so.

[obstructor]

So I hate to necro this thread, but it has been of such use to me.

With the increase of fluid in pipes in 0.15 has anyone tested if these numbers (x10) are still valid?

[Jürgen Erhard]

So I hate to necro this thread, but it has been of such use to me.

With the increase of fluid in pipes in 0.15 has anyone tested if these numbers (x10) are still valid?

The "official wiki" says 70/s. Which is a far cry from the "expected" 600 or 300 (600 because the offshore pump went from 60 to 1200, doubling the amount after the x10 fluid adjustment). If 70's right, it's way too low. Can't really see this, but haven't tested anything.

[obstructor]

I did some quick testing and here is my data:

2 pipes: 2400/s
4 pipes: 1906/s
9 pipes: 1384/s
25 pipes: 1134/s
147 pipes: 1017/s
271 pipes: 773/s
519 pipes: 430/s

Underground pipes still behave as single pipes.

It looks like 1000/s is the amount you can assume is flowing in all practical situations. For long distance its probably best to put a pump every 150 pipes or so.

Also just in case anyone is wondering for bobs mods, smaller pipes are objectively better in every way:

Copper Pipes
2 pipes: 2400/s
4 pipes: 2360/s
9 pipes: 2000/s
25 pipes: 1713/s
147 pipes: 1540/s
271 pipes: 1523/s
519 pipes: 1517/s

So with size 5 pipes you can assume 1500/s is flowing through your pipes pretty much regardless of distance.

As for pumps:
I don't know if this is intended but it reads 24000/s and hooking it up between to 2 tanks it seems like it can move 8000/s, so I fully believe that its trying to move 24000/s.

I am guessing that ever since they removed the decimal the flow doesn't lock to certain speeds like it did before, now it follows a general flow falloff. I haven't tested boosting the pressure with pumps throughout the pipe though. I would assume that as long as you divide the pipe to pump segments into equal lengths you would keep the flow at the rate of that length of pipe.