How does a big one-way loop train network work?

[Jupiter]

So regarding train network layout, there are several approaches I know of.

Use RoRo stations everywhere with single headers (locos just in front of the train). This, is in my eyes, the best but it has many problems due to dumb train path finding. Has to do with loops in your stations, creating alternative routes for trains whose paths are blocked.

Use terminal stations with double headers. These stations have less throughput for several reasons but still function. You just need a couple of extra platforms. This layout lessens the problem with train path finding but can still suffer from it if you have some kind of loops in your network (also bigger loops involving multiple intersections). You would need to structure your network in a tree-like fashion to avoid those big loops.

Then there is the idea of one big one-way loop. The promise of this kind of layout is that it has no problems with dumb train path finding while retaining a high throughput. But I'm wondering how this is actually going to work. I never saw a clear example of such a network. So if you need a new outpost far from your base, how is that gonna work with your big loop around your base?? In your loop you would use single headers (it's a one-way loop) so how are you gonna hook up a new outpost without creating a loop for trains to use as a detour in case of blockages?

Or do I have a completely wrong idea about this big loop concept?

[siggboy]

Search for DaveMcW's "One Minute Rocket Defense"

He built a crazy, crazy ass base and all the trains ran on a big loop (actually 4 big loops), no intersections.

[Jupiter]

Search for DaveMcW's "One Minute Rocket Defense"

He built a crazy, crazy ass base and all the trains ran on a big loop (actually 4 big loops), no intersections.

Thanks for the link. I knew about that base but forgot about it. And at the time I was looking at it such a train network seemed just stupid. Now I know better.

[Frightning]

I've been considering an alternative approach to solving the issues of too many trains. Which I will call Hubs. The idea is to effectively spread out the train traffic, while still being able to funnel all of the resources being mined into my main base. I am currently working on a base for which I may eventually implement this idea, but for now, it's not nearly big enough (315MW solar, ~2.5k Iron/min and ~2.5k Copper/min).

So how does a Hub system work?
The idea is to choose (a few) locations around which you have many mines relatively nearby to make Hub stations. The mines feed the Hub much like you would feed a main base, but then the Hub also has a (few) station(s) for LONG trains (think 8-20 wagons, possibly 2 or even 4 locos). It is these long trains that actually feed your main base. This cuts down on the need for an extremely large railyard at your main (and associated train traffic nightmares). A few, long, unloading bays that support long trains can provide the same throughput as many stations that unload short trains. You could even create a 3-tiered system if you REALLY need high throughput (and have real-world sized mega-trains actually feeding your main base; fyi, real freight trains are typically around 75 cars...I would probably have 8-12 locos in factorio for one of those monsters).

Caveats:
This is still theory crafting on my part, still working on the base that may eventually let me really test this idea.
The other detail is that I recommend keeping Oil train system separate from the rest of the raw resources (fussing with barrels in a system that also handles ores could prove to be a logistical nightmare, and there are relatively simple, very high throughput ways of dealing with oil trains on their own; especially with the wait function in 0.13, now you can make the train that serves a depleted oil output actually wait for a full 300-barrel load before returning instead of returning after 200 seconds with scarcely a dozen filled barrels).

[siggboy]

I've been considering an alternative approach to solving the issues of too many trains.

What you need is to use the trains effectively. Put them all on one train line, and send them where they're needed. Sadly it's not possible with Vanilla unless you implement some crude hacks. With SmartTrains you can do it (because it allows you to create logical train lines, and also to tell trains explicitely where to go).

I've created a setup for this a while ago, when I tested it in 0.12 I could make 1 rocket every 3 minutes with only 20 trains in the network -- and that was on a base where everything was distributed into mini factories, only connected by trains (so not one megabase with ore input and rocket output). The trains were LL-CCCC.

This works by having them all go through a central hub (depot), and then from there only go to the pickup station and the dropoff point, then return to the depot. I shipped all the immediate products as well that way, just for giggles and to stress test the concept.

Your hub idea is interesting for me, too. I could try to make a really big base that way, by using a simplified version of the distribution system to the hubs, and the full version to connect the hubs with the consumers.

This even brings us back on topic: the small distribution networks could all be closed-loops, only from the hubs to the depot we use a conventional rail network. That might also be easier to design.