My intention was to build everything in 0.14, and make minor adaptations for 0.15, adding in rail tankers for the water supply, and nuclear power for energy. But between the recipe changes, the new recipes, more prod modules in chemical plants, and addition of rocket science, I've decided that it's time to start over from scratch. So this is mostly in memorium. I didn't exactly reinvent any factorio wheels, but it's pretty well put together, it works and quite satisfying to watch, so I consider it a success.
Aerial Shot
As seen from wide zoom-out. There's enough belt width that you can track most of the process pretty easily, even from this zoom-out, even without labels on the various zones.
Receiving Yard
Everything arrives via train, with enough waiting bays for every single active train, even if the factory is shut down. 4L/8C is the configuration of choice for all trains, although I want to use something bigger next time. You can see where I clearly intended to use a four lane main intersection, but surprisingly enough, the two lane version handles it all. Edit: Notice that empty section on the copper belts! Accidentally deleted a single exit signal 
This "compact" 8 wagon > 8 blue unloader serves for most of the trains, and can saturate a blue belt out of each wagon. It can also be mirrored and put between two 8 wagon setups, with zero wasted vertical space. Having each train stop singularly responsible for just eight belts means that nothing larger than an 8-8 balancer is ever needed. Belt balancers (Props to Yttrium for these specific arrangements) ensure that each train gets unloaded fairly evenly - I was able to let trains unload right down to less than 100 ore/wagon before any of the wagons began to run empty early.
After repeated attempts and failures to do direct wagon-to-assembler unloading of oil barrels in a balanced manner, I simply gave up. The current configuration unloads barrels, condenses them down to a single blue belt (which still amounts to 1000 crude/second), then unbarrels into each lane as needed, which ensures that the train gets unloaded evenly. (This makes me question whether long strings of rail tankers will be convenient in 0.15). Also visible are the collection of water pumps - my original intention was to eventually bring water in via rail tanker in 0.15, but... stuff happened. Maybe next factory.
This "compact" 8 wagon > 8 blue unloader serves for most of the trains, and can saturate a blue belt out of each wagon. It can also be mirrored and put between two 8 wagon setups, with zero wasted vertical space. Having each train stop singularly responsible for just eight belts means that nothing larger than an 8-8 balancer is ever needed. Belt balancers (Props to Yttrium for these specific arrangements) ensure that each train gets unloaded fairly evenly - I was able to let trains unload right down to less than 100 ore/wagon before any of the wagons began to run empty early.
After repeated attempts and failures to do direct wagon-to-assembler unloading of oil barrels in a balanced manner, I simply gave up. The current configuration unloads barrels, condenses them down to a single blue belt (which still amounts to 1000 crude/second), then unbarrels into each lane as needed, which ensures that the train gets unloaded evenly. (This makes me question whether long strings of rail tankers will be convenient in 0.15). Also visible are the collection of water pumps - my original intention was to eventually bring water in via rail tanker in 0.15, but... stuff happened. Maybe next factory.
Metal Processing
Steel processing, definitely the largest area of all. I strongly suspect this was NOT the most efficient use of speed beacons possible (a recurring pattern, just you wait!), but I liked the look of it. Each row is designed by output - 600 units of steel/minute, which consumes most of a blue belt of iron ore. Feed four of these assemblies into a lane balancer, and you get a blue belt worth of steel output, with even consumption across four iron ore lanes (half a train assembly). The theory held up fairly well in practice, I found.
Copper and Iron were simpler. I used a 2 in/2 out configuration for aesthetics, since I was still dreaming about having a roughly square shape to the entire factory. No real advantage beyond the regular 1 in/1 out design, though. The base design is flexible enough that I could set the copper to input and output from the same end, while the iron inputted and outputted on opposite ends.
Zoomed out, you can see the entire metal assembly, plus the edge of the green circuit production just above. Interestingly, with the exception of just one belt of iron ore that goes to supply satellite/acid production, every other (non-steel) iron plate goes into green circuit production. Definitely going to be thinking harder about some iron ore>green circuit direct feeding in the future.
Copper and Iron were simpler. I used a 2 in/2 out configuration for aesthetics, since I was still dreaming about having a roughly square shape to the entire factory. No real advantage beyond the regular 1 in/1 out design, though. The base design is flexible enough that I could set the copper to input and output from the same end, while the iron inputted and outputted on opposite ends.
Zoomed out, you can see the entire metal assembly, plus the edge of the green circuit production just above. Interestingly, with the exception of just one belt of iron ore that goes to supply satellite/acid production, every other (non-steel) iron plate goes into green circuit production. Definitely going to be thinking harder about some iron ore>green circuit direct feeding in the future.
Crude Processing
After the barrels are broken down and used to fill 8 lanes of crude pipes, they're sent to the initial processing. This horrific display of speed beacon carpet-bombing can process roughly 180 units of crude/second, converting everything into light oil and petroleum. Four of these supply ingredients for the rest of the stages. Later on, I properly Excel-ed the numbers, and found out that 3 of them would have been sufficient - such is life. Notice how many pumps are in parallel? Loving the new higher power pumps in 0.15!
The ratio of light oil to petro means that I need SOME of the former converted, but not too much. Balance is achieved via actively forcing all Light Oil towards the Solid fuel production, where it also fills a single tank. If the tank gets above a certain level, the array of light-to-petro plants kicks in. Once the light oil drops, the conversion stops. Simple, but pretty much idiot-proof, I've found. One of the very few uses of the circuit network anywhere in this entire factory.
That handled, everything either pushed up to the solid fuel production, or over to plastic/sulphur production, where generic chemical plant arrangements output blue belts full of their designed product.
Despite this section taking longer than any other to design, I'm probably the least happy with it. Factorio's liquid handling meant that, even after carefully re-laying a lot of pipes in flow order, there were still random low and high spots, with some plants inexplicably starving. Future designs will avoid mass condensing of liquid products, instead matching smaller numbers of refineries to plastic/sulphur plants independently.
The ratio of light oil to petro means that I need SOME of the former converted, but not too much. Balance is achieved via actively forcing all Light Oil towards the Solid fuel production, where it also fills a single tank. If the tank gets above a certain level, the array of light-to-petro plants kicks in. Once the light oil drops, the conversion stops. Simple, but pretty much idiot-proof, I've found. One of the very few uses of the circuit network anywhere in this entire factory.
That handled, everything either pushed up to the solid fuel production, or over to plastic/sulphur production, where generic chemical plant arrangements output blue belts full of their designed product.
Despite this section taking longer than any other to design, I'm probably the least happy with it. Factorio's liquid handling meant that, even after carefully re-laying a lot of pipes in flow order, there were still random low and high spots, with some plants inexplicably starving. Future designs will avoid mass condensing of liquid products, instead matching smaller numbers of refineries to plastic/sulphur plants independently.
Electronics Assembly
After running the numbers on a "4 iron/4 copper in, 4 green circuits out" setup, I realized that there was almost exactly enough unused iron and copper for a fifth lane of green circuits out. And by coincidence, the number of belts full of green circuits that I wanted was 20, which was divisible by 5, so...
(image rotated sideways for easier viewing)
Red circuits. Nothing special here, pretty sure I've seen similar elsewhere on the forum. Easily stacked, good beacon use, uses a full blue belt of plastic and green circuits, while outputting a full red belt of red circuits.
Blue circuits. Nothing special here either. The splitter blob in the bottom left corner is what you do when you need to condense 6 red belt volumes into 5 blue belt outputs, EACH with more than a red belt volume on them. Sometimes there is no elegant solution. I spent way too long googling "5-to-6 balancer" before realizing that.
(image rotated sideways for easier viewing)
Red circuits. Nothing special here, pretty sure I've seen similar elsewhere on the forum. Easily stacked, good beacon use, uses a full blue belt of plastic and green circuits, while outputting a full red belt of red circuits.
Blue circuits. Nothing special here either. The splitter blob in the bottom left corner is what you do when you need to condense 6 red belt volumes into 5 blue belt outputs, EACH with more than a red belt volume on them. Sometimes there is no elegant solution. I spent way too long googling "5-to-6 balancer" before realizing that.
Rocket Assembly
Kind of boring, to be honest. At 2 rockets/minute, each of the three rocket components can be supplied on a single blue belt, which made things convenient. A little bit of the components diverted away for satellite purposes, but otherwise just a brute force loading into all four rocket silos. Launches are cycled via rationing the supply of satellites, and sending them through a 4-way splitter, one path to each silo in turn. Not scientific, but worked well.
By The Numbers
2000 launches later, with a launch every 27 seconds or so, things are good. Learned lots, good and bad. Still love standing in the middle and staring at the acres of belts in tandem, thinking hard about applying the same concept to an infinite research factory in 0.15.
Not shown in 0.14, but my excel charts suggest that I'm using roughly 28000 crude/minute, which is getting turned into about 10000 light oil and 24000 petroleum.
Not shown in 0.14, but my excel charts suggest that I'm using roughly 28000 crude/minute, which is getting turned into about 10000 light oil and 24000 petroleum.
