ArcGIS Schematics - Preventing overlapping links and consistent layout?

Question

I'm trying to recreate some network diagrams on a consistent basis based on our geometric network data.

The layout algorithms I'm selecting are getting me close but I can't seem to fine tune the layout task to generate a consistent diagram with the desired layout.

This is the approximate style I'm looking to recreate. The important part is the roughly even distribution of the links / branches.

This is roughly what I have going (some of the time) when using the Schematic Editor.

1A / 1B: These are links that overlap that should not.

2: This is a link that need not extend to the right at all. It would great if it was just a straight link downwards.

3: This branch is an example of where better placement would be on the other side of the main line.

I used a Composite algorithm of the following tasks:

1 - Orthogonal

2 - Relative - Main Line (using a line attribute)

3 - Geo - Rotate Nodes Along Links

4 - Separate Overlapping Links

5 - Square Links

6 - Bypass Nodes

Anyone have a clue as how these tweaks can be achieved?

RickAnderson · Answer

Actually, instead of running the Separate Overlapping Links algorithm, you really should be running the Geo-Partial Overlapping Links algorithm.  Separate only works on lines that are completely overlapping, i.e. they have the same end points.  The partial overlapping algo just looks for any overlapping segment.

POLA algorithm

RickAnderson · Answer

Looks like you need to play with the order of the tasks a bit.  Separate overlapping links would solve your 1a/b issues.  However, immediately after you are running that algo, you then run square links which then perfectly squares things again causing the overlap.  So just flipping 4 and 5 should clarify the overlaps.

For items 2 and 3, you can manually adjust afterwards.  The algorithms will normally only get you 80 to 90 percent of the way, then manual white space management is needed for the final 10 to 20 percent.  A combination of algorithms that works perfectly for 1 sample area of data may not work well for the next sample area.  White space management is tricky.  The only way you would get 100% is to write a custom algorithm that does everything for you and rechecks several times to make sure it doesn't violate the desired results.