Smart Furnace - load-balanced, demand-based, and tileable

[XKnight]

XKnight can i ask for blueprint or explanation because i'm staring on it for a while and have no idea what you did.

Sure, why not.

blueprint

Input - left bottom combinator

Code: Select all

H4sIAAAAAAAA/81b7ZKjuA59la78vEvu4g9MqFmeZGorRSfubmoJpIDsTleq332dDvkQYGJZ9OzMv0xjoSMd68hysq2eimqTFU/r9KjLNm9z3aTHY5ntdLrY
VGXTZmW73FS757zM2qpeBPuqMU9VZXr8kS7j4D1dio/APFnqzem/zervC7b4Mz2+1lqbp85m39f5No2DTV5vDnl7+sA+zL/gJS9aXZ9e2eSvZVakx/Z9b179
d163h6xYBGdPzn9cPhcHvTi97VC2KQ+DvNzqHyk7mQqcfVYTPtd6a/eYUz2WPJGJUqtEXTxHOS4Qjq+ojn9mr+d5zJPo6nngZCa8mmDdSoXBzFH8iozh2yfG
3emWt3p38fw5a/LNMi8bXZtnb/4rK/T71UVVvi7fMvPkdsRG3NngkzZesqYdWbzqFovJxc0uq8dWJ91q+Wh1USx1YYJdmyjsq2Jkw0WYFE6xdphC6Zymf6pq
q8vl5k037YBk0znK66q/kDslpmm1LnorBSIrcKV0y4iJaPaqlybIf12XRtdMTC09M7mts7LZV+b1z7q4vV1d9+NDKlosXMgcT1rQP/a1bhqbkQupV2gky7Yy
VcoY2V6NXTieYDGN2GJXeQk98I0ZZDe9eoy12Rd5CzbwhaLMoXgMV19oyqZ5ekEzNHBhK5um6/OhLnU9Un+uynFfPbI6b992ujWA7Xr9/+hUQkzQtnl9riBm
u6LVm0FxECcvvi+4o+qb123z7m03p9PjS1437dpFCXW2efuMxrlepmFQ7XWdfcJZ/LYIqkO7P7iZ6rTZL44cH0fVE9JgWLfZRDg5iLyCwQ1sTQubN+6NPply
WtH1bHfp+R8qPec3IrMjvLMzYKstDwLugPDXiPal0bwL9+9fH+7Qu6gM42aLN3wSUp3JmQIe4qLNYKRxdYf5htmD1St77BLXBCQT8R85HMyTjDkKfAgCvdWb
fGs01XpI8ucyn1DIyKHqY9bb9wmfyFM0rcVdbL62RLHPtO72WX0Kfrr4w6M+GQP79/VnL7R+qavdOi+NgVO7VjR6gc+2x5aarEfscbbl9Ap7ftkvkMPwZ+bQ
bXDG3EY5/e4LPcq5zWBWyOkNG6wcnj3GV/LBKVdg4hOa8IRu4Umok66bq5yNzOXc9mPosR0dtpz/JrvfqlL8N3uO/8w9d55cFdlu39tq7EQm7kYmthpEqovq
eKQmB1JdHyBADFJ3nx1HvTyc32fu4XOCCLOCUpL4I+jN8ggxX6Focu8/H2qas//9kSIBQOLOGRZDAMMGmJgAHwKtUJwH/g9PU9QE+ACIEQxKIIBhhXYH0Bst
Exik3AFwKDE8pmSgN+EmIIgRHILnFR7NngIfDikEANiQc4J22VLgg0AiSATLkCAome2igsClCAEEzta48gdivbQhInElFbxE5QRpfoTEl1yuSKBIC4JIPyCX
DxCBIBfslgRBrCcu4wj04u5YBBRuQRBuhztBAiaBIBrUckHQ8sf58eEad8cioKyLmQvZeH58MDEE56DQC4LQP7469qcc4pgqoPILQpF2ur6mnGQR3IMiKggt
pXOePKiHOJoLWLklob/B5MkDFKaEw3I3MlzCpQl81cCfaZiCsJqtIIx8WYMGwZVZsHOWhJ7ADsGTR44QJGwFJLUsz4QAcXyRsGBJQsEa+9qMP5EQzWVvL0jc
9Ru6fM20UxDHAAkPNJJyoLEnyYNoiA5Twg5GErrmqUR4gEBMjCQsWRGhZFm/ZOvPKMRhv7dHJLlHngkCYnQkoYZHBA1/lAsPUiHGFRJ2wpIwA7PnwgMC4jIh
gps7osyyh9/99qcTYpotYZ8bUed4MyFAXChEUMoj8jB4Jh4h5tkSKnlEnXfNhABxLRiFcyZh9HcM/mRC3K5FvL+j7z/GXwfLa5ByOs9KR2CEDtd21cw8fE4Q
LvdKE0Egxm86ffxfoUIO/Ce0r5abTh8AMQIALEuKoA2WazYfAAoBAPZKitC32q7ZfBBE7ggUrKuKoAyP7nR8kEgEEtgrKersaV4gAgGk96OIGU5082LhCCyw
d1LkcZrLjNMHE0NggpqtCGc851m0BySEdis4B1EEIcTMoj1AYagHxSWmHjzGpiIeCDBEg/2JoswQrYNcDwiIohxDeYkp8mKfsHlgwNRj2GapGW7UZkoEolOJ
oajEBFF5NNXxAIJpWGDLFZNvN2aCgOjb41kPfZaBggcEROcew2YrpjZbMyFAHP7iWac6UyfvERwfQb75fMex+1FBZ2n8d3PG0e4XFsEAvrH1rdatqY1P62+6
3P4LFZpzmrVFAAA=

Forgot to mention, using smelting switching + productivity modules it is even possible to convert resources from one to another:
4 iron ore + 1 copper ore = 4 iron plates + 2 copper plates. x2 copper!
20 iron plates + 1 copper ore = 4 steel + 5.2 copper plates. x5 copper!
So, if you don't have copper/iron/stone, but have tons of other resource - you know what to do

*Edit: fixed const array in blueprint

[Neotix]

OMG it's so simple Entire 2^n calculation and binary coding is not necessary. Just signals comparison and sum. ()
It's not improvement, it's total overhaul
And big advantage is that it's easily scalable for more resources.

[Qon]

Forgot to mention, using smelting switching + productivity modules it is even possible to convert resources from one to another:
4 iron ore + 1 copper ore = 4 iron plates + 2 copper plates. x2 copper!
20 iron plates + 1 copper ore = 4 steel + 5.2 copper plates. x5 copper!
So, if you don't have copper/iron/stone, but have tons of other resource - you know what to do

You did it again
That is a nice discovery. You are pretty good at this. I'm impressed.

[Neotix]

Minor issue with that setup. It can't be turned off. If all input signals are 0 then all lamp in bottom row are on. It can be fixed by adding condition to transmit signal from constant combinator in 3rd column only when any input signal is > 0.

[MeduSalem]

The conversion between resources works only between Iron Ore and Copper Ore. It doesn't work for Steel.

For steel the furnace is keeping its own individual progress bar for productivity modules. That's at least the case with 0.12.33 which I am playing.

I guess that the conversion of Copper into Iron Ore (and vice versa) by abusing productivity modules isn't intended either and is rather an oversight, which I expect to be fixed somewhen in the future if someone actually reports it.

[Qon]

The conversion between resources works only between Iron Ore and Copper Ore. It doesn't work for Steel.

For steel the furnace is keeping its own individual progress bar for productivity modules. That's at least the case with 0.12.33 which I am playing.

I guess that the conversion of Copper into Iron Ore (and vice versa) by abusing productivity modules isn't intended either and is rather an oversight, which I expect to be fixed somewhen in the future if someone actually reports it.

It does work, it just becomes 5 times smaller when you go from the 3.5 second recipe to the 17.5 second recipe. The reason you get 5 copper/iron plates from one ore after making 4 steel plates is because it goes to something like 400% bonus progress when it become 5 times bigger in the other diretcion. So the conversion works even if it doesn't look like it.

[Qon]

Standard process:
0.4 iron + 0.8 iron =[*from 2 eq]=> 0.4 iron + 0.192 steel

Result: new process gives (0.2 - 0.192)/0.192 = 4.16% more steel than standard! and takes (2 - 1.8)/1.8 = 11.1% more time.

0.8 irons gives 0.8*1.2*1.2/5 steel in the standard process. Which is 0.2304. Too bad, no extra productivity :/

[XKnight]

Standard process:
0.4 iron + 0.8 iron =[*from 2 eq]=> 0.4 iron + 0.192 steel

Result: new process gives (0.2 - 0.192)/0.192 = 4.16% more steel than standard! and takes (2 - 1.8)/1.8 = 11.1% more time.

0.8 irons gives 0.8*1.2*1.2/5 steel in the standard process. Which is 0.2304. Too bad, no extra productivity :/

0.8 iron gives 0.8*1.2/5 steel = 0.192
0.8 ORE gives 0.8*1.2*1.2/5 steel

in my original post "iron = iron plate", "ore = iron ore".

[kann_]

I tried your circuit XKnight. It is...simple
Just needed a similar fix to turn it off as already suggested.

One steel, iron, copper tile is 3x10. So not much smaller than the combinator version, but I like the simplicity and I can use belts with it

furnace.png (132.31 KiB) Viewed 12489 times

[XKnight]

And the main holy-crap part:

When you switch back and forth between steel and iron smelting you lose your productivity bonus progress some since they have different smelting times (prod3 modules in everything).

I decided to test this because it looks very very important:

The attachment Animation.gif is no longer available

Math

Some math for productivity 20%(1 time unit = iron smelting time, 5 time unit = steel smelting time):
obvious process: ore => iron plate => steel
1 ore + 1 time => 1 iron plate + 0.2 iron plate from productivity (iron) = 1.2 iron plate (1*)
5 iron + 5 time => 1 steel + 0.2 steel from productivity (steel) = 1.2 steel
Divide by 5:
1 iron + 1 time => 0.24 steel (2*)

Using smelting process from gif:
5 iron + 5 time + 5 ore + 5 time => 1 steel + 1 iron from productivity (steel) + 5 iron + 1 iron from productivity (iron) = 1 steel + 7 iron
Divide by 5:
1 iron + 1 ore + 2 time => 0.2 steel + 1.4 iron
Subtract 1 iron from left and right part:
1 ore + 2 time => 0.2 steel + 0.4 iron

Standard process:
1 ore =[*from 1 eq]=> 1.2 iron = 0.4 iron + 0.8 iron =[*from 2 eq]=> 0.4 iron + 0.192 steel
1 ore + 1.8 time => 0.192 steel + 0.4 iron

Result: new process gives (0.2 - 0.192)/0.192 = 4.16% more steel than standard! and takes (2 - 1.8)/1.8 = 11.1% more time.

In-game proof:

Builds

Simple process:

1.png (297.82 KiB) Viewed 12479 times

Advanced process:

2.png (669.48 KiB) Viewed 12479 times

Test: 600 iron ore in input, only steel in output.

Results in the end:
simple process - produced 172 steel
advanced process - produced 180 steel (4.6% more steel)

Results at the same moment in time:
simple process - 172 items
advanced process - 158 items (8.1% slower)

Animation

Animation.gif (12.3 MiB) Viewed 12479 times

[MeduSalem]

It does work, it just becomes 5 times smaller when you go from the 3.5 second recipe to the 17.5 second recipe. The reason you get 5 copper/iron plates from one ore after making 4 steel plates is because it goes to something like 400% bonus progress when it become 5 times bigger in the other diretcion. So the conversion works even if it doesn't look like it.

Well my fault...

I have to admit... I really thought it doesn't work... or at least it didn't really work out the way I expected it when I first tried abusing the behaviour past year. Back then I discarded the idea when it first seemed like the productivity progress isn't carried over from Iron/Copper to Steel due to the "marginal" effect... 4.6% like Xknight says.

But now that I tried the other way around with XKnight's conversion of 20 Iron Plates + 1 Copper Ore into 4 Steel Bars + 5 Copper Plates I realized that it somehow... works... because the effect is much, much more noticeable... I mean like 4 copper plates out of thin air in one go, which can't be overlooked.

From pure logic I can tell that this "black magic" is an absurd, but funny "exploit"... sort of. Not that it matters much since resources are basically infinite anyways.

I still have to get into the math of XKnight though... Sometimes I get the most complex stuff right away, other times I can look at simple relations like the above and I don't get why the hell it works like that.

Though for practicability I think I would never really use the exploit because of how it requires combinators on a per furnace basis, which is just not worth the space for a 4.6% increase in Steel Output (and increased smelting time coming with it). The other way around of turning excess Iron Ore into Copper Plates by abusing Steel Plates for the insane productivity multiplier might actually be worthwhile doing in the special case of running out of Copper Ore.


But it makes me think about how ridiculous this would turn out to be in Assemblers with 40% Productivity if the Devs ever decide to make Assembler recipes switchable by circuit network signal or something like that.

[Qon]

In-game proof:
Test: 600 iron ore in input, only steel in output.

Results in the end:
simple process - produced 172 steel
advanced process - produced 180 steel (4.6% more steel)

Results at the same moment in time:
simple process - 172 items
advanced process - 158 items (8.1% slower)

Awesome! Can I have your blueprint string?

I still have to get into the math of XKnight though... Sometimes I get the most complex stuff right away, other times I can look at simple relations like the above and I don't get why the hell it works like that.

Though for practicability I think I would never really use the exploit because of how it requires combinators on a per furnace basis, which is just not worth the space for a 4.6% increase in Steel Output (and increased smelting time coming with it). The other way around of turning excess Iron Ore into Copper Plates by abusing Steel Plates for the insane productivity multiplier might actually be worthwhile doing in the special case of running out of Copper Ore.

But it makes me think about how ridiculous this would turn out to be in Assemblers with 40% Productivity if the Devs ever decide to make Assembler recipes switchable by circuit network signal or something like that.

Yeah I have a headache atm so it's physically painful to think. :\

Maybe it can be compressed further. I did manage to get it alternating between 5 iron ore/5 iron plate in the same space I woul have used anyway. But that is not the correct ratio because it doesn't use up all the plates. Maybe it's worth doing.

Yeah I was thinking the same about switchable assemblers. If you don't lose bonus with circuit switching that would be awesome!

[MeduSalem]

Yeah I have a headache atm so it's physically painful to think. :\

And I get one too trying to understand why it happens. The Factorio gods are merciless today. :S

Maybe it can be compressed further. I did manage to get it alternating between 5 iron ore/5 iron plate in the same space I woul have used anyway. But that is not the correct ratio because it doesn't use up all the plates. Maybe it's worth doing.

Depending on playstyle yeah, it might be worthwhile for some.

If you are using Beacons with Speed Modules though, which I happen to do, the rule is basically to cover as many furnaces as possible. At least I get a bad feeling in my guts when using Beacons and not being able to get a good coverage. It strongly hurts my sense for perfectionism. So wasting space with excessive combinator stuff is absolutely out of question at least in combination with Beacons.

A global load balancer (like the original posts in this thread initially tried to do) where a bunch of combinators is doing the job for any amount of furnaces attached to the logic is then the more practical solution.

Yeah I was thinking the same about switchable assemblers. If you don't lose bonus with circuit switching that would be awesome!

Yeah, but I guess it wouldn't take long for the devs to notice the exploit themselves and eventually fix it by introducing a per recipe productivity progress, which is something they are probably going to do anyways after reading into this thread eventually.

But they would also have to fix the other severe problem of residual source items which would block the recipe switch, which is at smart furnaces already a pain in the butt with Stone/Bricks, even more so in Assemblers with a multitude of recipes. Resource Pools or Recipe Queues might help with that to some degree but apart from a few minor discussions no one really came up with a serious solution for that major problem.

After about a year of thought on the matter I would actually recommend Recipe Queues within Assemblers to deal with the problem, so when an assembler receives a circuit signal with another recipe it basically temporarily suspends the current recipe in favor of the new recipe. Each recipe would have its own productivity progress associated as well. Resources already in the assembler would be kept in the background until the corresponding recipe gets activated again. If no signal is active the assembler could switch back to the default recipe if the player has one selected, otherwise it sits there waiting for a signal. Inserters that were already active/holding stuff during the recipe switch would put the resources into the buffer of the recipe under which they were activated to avoid blocked inserters, which means they could temporarily go over the input limit for a recipe to free themselves of resources in flight.

It would be the safe method, but I admit that it would be quite messy if you tear down an assembler that went through 20 different recipes and which has source items for every single one of them. xD

Another method would be to use the value send with the circuit signal as the actual "amount" of how many times the assembler should produce that particular item before retiring the recipe. That would mean that with the last item the inserters would have to know not to put more source items into the buffer so it can't get blocked when the actual switch to the next recipe in the queue happens. Once the order has been completed the recipe gets retired and the next recipe in the queue starts. But that means that in case of a resource shortage the assembler is locked in place waiting for the last resources to arrive to terminate the recipe. It couldn't switch to something else until the order is completed = waste of precious time. That approach would require a lot of tuning on the player's end to avoid shortages.

Maybe even some sort of combination of both worlds. Every time a new recipe signal comes it switches immediately and the value specifies an amount after which a recipe gets terminated. After that it resumes with the previous recipe in the queue until the amount is reached and so on.

But that's already far offtopic.

[Qon]

Depending on playstyle yeah, it might be worthwhile for some.

If you are using Beacons with Speed Modules though, which I happen to do, the rule is basically to cover as many furnaces as possible. At least I get a bad feeling in my guts when using Beacons and not being able to get a good coverage. It strongly hurts my sense for perfectionism. So wasting space with excessive combinator stuff is absolutely out of question at least in combination with Beacons.

A global load balancer (like the original posts in this thread initially tried to do) where a bunch of combinators is doing the job for any amount of furnaces attached to the logic is then the more practical solution.

Works with speed beacons. No combinators. Might not be completely speed optimal, need to test some more. Got 181 steel from 600 ore. The extra 1 was most likely from earlier testing, I didn't reset the bonus before doing the 600 run.

Blueprint string:

Code: Select all

H4sIAAAAAAAA/8VU7WrjMBB8laDfNthN0hZS3YuUYnTSOl1QJN9KKQ3G7971xzVx7LQJFPorsVc7OzM7svEL67Wyi0LW4CJGhCDr2qkdSGH9FkNEnepXCDFV
OuIbpBX5NzRAIql84AbvZP0u0zw5yHTZNMl8M8G/Pf+etw1dyVAuSrR8pqWAzkBb79GQvEs9gUi037so8yxrjrPCTlFM0QWgyYSe2F2TGCTQ3dtVcmkMEwSb
VlZFECf4YLmTWEu5J6c0jCdkPCC7jsx/Lto715NhDs8iFy+yJjDMp1vCoUAj7xlySwBu9DZvdXcABod+jaT3GGVdIrGFAbdOWVnHQ9UaF2EnTk0cxLUIISq2
cs1/d5UiFT1J8SQmvnYLnDGV1/ZXEWu6Rs2aQY9PD6Onx6a5ZZnX+/ezRnHmJlZdTNIpxHXast7SY0zvf0fmaiTyz+0iSxW+yv/4Kn6ftW+jNr0k66ZPFOru
3BnryW1mDv2Ju2R2TZ/15ZxxQ201/wnZEEQesyg24MwH7wFZ2m4FAAA=

I don't want the devs to remove this
This is depth.

[MeduSalem]

Works with speed beacons. No combinators. Might not be completely speed optimal, need to test some more. Got 181 steel from 600 ore. The extra 1 was most likely from earlier testing, I didn't reset the bonus before doing the 600 run.

Ah, well then.

I don't want the devs to remove this
This is depth.

Arguably it is an exploit, not depth. I might be wrong but unless a dev comments on the matter I am pretty sure this behaviour was never intended and just happens because the devs currently have no code in place to keep track of productivity progress for individual recipes after recipe switching.

As I have written in other module balancing threads in the past Productivity Modules themself are already a questionable concept in a game with infinite resources in my opinion, but as some might say it can be explained as further "refining a recipe" by increasing the "yield per resource". So I can understand why it is there because it is something that happens in reality too with better optimized manufacturing processes.

BUT creating Steel Bars out of thin air by abusing an entirely different resource other than Iron Plates is an exploit. In reality one can't convert one resource type into a entirely different one either, especially when looking at converting Copper into Steel and vice versa. The way it works currently is more like performing alchemy than anything else. So it eventually should be fixed.

That said I would be open to improving the entire smelting process anyways to give it more depth with more advanced, multistage refining processes. In real life there are various methods for steel production and for other metals that are quite interesting. It could be a mid/late game technology to improve resource consumption, just like advanced oil refining is.

Make it an actual feature, not a weird exploit due to oversight.

[Qon]

Exploit doesn't mean it's not depth or not good gameplay :>
I'm enjoying it. 4% bonus isn't really game breaking either.
It having no basis in reality doesn't disqualify it either as a valid game mechanic.
And the way my blueprint manages to follow the valid sequence with no combinators and just simple smart inserters and game mechanics is depth as well. But that is gonna break with v0.13 though if there's no way to make rapids only grab 5 at once. The second 5-batch of ore is actually inserted when the iron plate box is empty, which means it starts inserting it after the plates are in the fast inserter on their way into the furnace. But it still manges to get into the furnace first because of residual iron plates in the furnace output slot, I think. Quite fascinating to see it actually work q:

But I'm not saying that they shouldn't improve the game with intentianal depth. I just don't want them to break this on purpose with no replacement mechanics that are equally interesting.

[MeduSalem]

Exploit doesn't mean it's not depth or not good gameplay :>
I'm enjoying it. 4% bonus isn't really game breaking either.
It having no basis in reality doesn't disqualify it either as a valid game mechanic..

I see we have different philosphies about that and I respect yours.

But that is gonna break with v0.13 though if there's no way to make rapids only grab 5 at once.

Well, if I remember right I commented on the FFF where they announced the decision of the damn rapid inserters and I stated that I absolutely dislike the concept because of how it breaks so many things at once just for the sake of breaking it. Ironically Rapid Inserters, or the Loaders concept before that, weren't even necessary in the first place and were uncalled for except by a minority of elitist players with their train-based mega bases getting sick of building train stations crying for a better way to load/unload train wagons with compressed belts and whatnot. Belt-balancing maniacs with their ugly splitter contraptions which are not necessary either.

I am quite unhappy about the entire development because it turned from providing an "optional" or "alternative" way of doing things into the upcoming "mandatory" way of doing things and on the path breaking a lot of things just to improve one little part of the game (train stations) without caring about the rest.

I expect people will be taking the bitter pill and accept it because of how awesome the additional circuit network features are going to be. No, I am not sarcastic, I really think they are going to be awesome. But on the other hand some things won't work anymore the way they did and there is not going to be a replacement, like for example stack size bonus for Long-Handed Inserters. There will be mods with long-handed rapid inserters, yes, but I honestly don't give a damn about mods for the most part for several reasons.

And I am also concerned about not being able to control the rapid inserter in how many items it grabs from chests. Yeah, it may be possible to control it when it grabs from belts when there are some ticks in between grabing each item, but I bet the rapid inserter won't give the same control about that if it grabs from a chest/assembler/furnace. If it does then it would be truly awesome, especially for smart furnace setups. Otherwise robot-based smart furnaces will be basically dead since one would lose entire control over item flow.

But I'm not saying that they shouldn't improve the game with intentianal depth. I just don't want them to break this on purpose with no replacement mechanics that are equally interesting.

Maybe it is better that I don't have a thing to say on the game development because I would have attached a "purity"-factor to each piece of resource, which would have determined how much items you get out of it during processing it, which could have been increased with multi-pass processing like smelting it multiple times, merging several low-quality resources of the same type together. It wouldn't be necessary to do that, but if you don't the output product would have less quality, affecting certain other aspects of the end products. Like for example a tank that has less armor because you used low quality iron/steel and things like it.

I know that there are certain people who like that kind of stuff, myself included, while others absolutely hate crap like that because of how a game turns into spreadsheet madness that way.

[kann_]

Just to understand what happen:
One steel crafting process accumulates enough bonus production to turn the next 5 iron ore into 7 iron plates?

By using the full production bonus of the steel production on iron we produce one extra iron per steel.
So for X iron we get X/5 steel and X/5 iron and this continues recursively.
So the total steel from X iron = X/5 + X/(5*5) + X/(5*5*5)...

Now I used some buried knowledge from ages ago to figure this one out.

This is actually the "geometric series".

Let me put this into a picture to display the formulas.

steel.png (10.98 KiB) Viewed 6061 times

Indeed the experimental results are correct ;D
This is the same result as XKnights calculation, but maybe it helps someone else to understand it.

Geometric series in wiki: https://en.wikipedia.org/wiki/Geometric_series

[Neotix]

Back to the topic
My Furnaces tile. It's able to produce Iron, Copper, Steep plates and Stone bricks.
After some tests so far it's safe from jam (no iron plates or stone left in any furnace).

furnace

For this setup I used some mods like bob logistic, color lamps and expand signals. Of course after some changes it can be done in vanilla but I like mods so i'm using them
Building furnaces array is really painful. Each tile have to be setup manually because each use different signal. Good that in every tile i have to change signals in 4 inserters and copy/paste it (Thanks Devs for that feature )
Each tile have combinator that send signal if some resources are low (more tile have low resource then higher value of signals). I will use it latter to modify proportions.

Furnace Array

Input values
0 - Iron plates
1 - Copper plates
2 - Steel plates
3 - Stone Bricks

Furnaces

White - Stone Bricks
Yellow - Steel plates
Red - Copper Plates
Blue - Iron Plates

Expanded controller (test setup) that can handle 60 tiles (120 furnaces).

Controller

Tis will requite a lot of bots.

[kann_]

Alright, so I tried my robot less version. As far as I can tell it works as expected.

Basically I use two furnaces, the left furnaces (yellow circle) are for steel but if it happens to be empty it will be used for iron/copper as well.
When crafting steel the smart inserter (red circle) is competing against the fast inserter (red circle) to pick up the iron. Usually the smart inserter wins.
I tested it a little and it seems that the "last activated" wins. The only time the fast inserter gets activated is when it gets placed.

Furnace Array

furnace2.png (810.44 KiB) Viewed 6041 times

I though about using passive splitter sorting, but I have the feeling it is not really worth the effort with 3 items and high throughput. Is there a good example?