New assembly concept, maybe without a workbench

[freedman]

I feel like I need to bring this up now before Bodies become a container for multiple solids. I've been thinking about a new assembly concept. I'll try to convey this in a few words, the details can come later.

Assembly now
The user brings in a part and then has to dictate all the placement parameters to the program. All the assembly params are stored and the tree gets really complicated. The assembly concept is almost an after-thought to part creation.
Assembly new (using DatumDock)
The user brings in a part and the program automatically places the part using the DatumDock info. The assembly (docking) data was created and stored in the part file when the part was designed .

Basic concept
Put a datum feature in the file with the body/part. When asked to assemble, the program jams two of the datumdock planes together (opposing) using local coordinates. Properties can be used to add mating discretion.

Usage
Create a block with two tubing connections, after all the design stuff is finished click the new tool icon "DatumDock", this creates (in the tree) a special datum plane that holds docking info, center the two datumdocks on the center of the tubing ports, create a dock for each pipe port. If you now bring in a tube (with a datumdock on the center of the tube) they will be aligned and docked.

Maybe the Assembler is a macro, something like bring in two parts, highlight two parts and click "Dock" in the macro screen, the macro looks in both part files and finds the docking planes.

That's probably enough for now.

[triplus]

Hence adding a plane relation in between datum planes?

Yes, this will likely be supported at some point in the future. That is adding a classical relation in between two datum planes and/or in addition being able to use a coordinate system or a datum plane feature as an interface. Therefore being able to define some relations in between parts, before any geometry exists. We had such discussion regarding Assembly 4 effort and in general relating to assembly structure.

P.S. Currently prototypes are based on utilizing expression engine. Once things mature more, likely some sort of basic assembly solver will need to get coded, or one of the existing one could be used for this purposes.

[galou_breizh]

before Bodies become a container for multiple solids

This surprises me. Where did you read this?

[freedman]

There has been much talk about allowing more than one solid in a body. Apparently it isn't that difficult to do in the code but some of the designers are not sure which way to go. Personally I like it the way it is but other CAD programs are referenced that allow multiple solids.

[HarryGeier]

I was also reading this, but i am not sure if such would add more confusion later than forcing the user to go for single solids .
I think your approach is interesting. I did quite a few multi body designs by relating bodies thru shapebinders and datum planes.
For rather static assemby, that was good enough for me up to now and your idea would add some standards to such approaches.

[wandererfan]

There has been much talk about allowing more than one solid in a body. Apparently it isn't that difficult to do in the code but some of the designers are not sure which way to go. Personally I like it the way it is but other CAD programs are referenced that allow multiple solids.

It isn't so much the creation of multiple solids in a Body that is the problem. The current code will actually produce multiple solids under the appropriate conditions, but there is a result check that prevents the return of multiple results.

The problem comes in subsequent operations that are expecting a single solid and may not handle multiple solid input gracefully.

[freedman]

My vision for the new assembly concept (I'll call it A10) is to make it a passive assembler, it works on rules only. It will scan thru the parts and look at data in the datum docks. I believe that local coordinates can be used to link two objects and then store the global coordinates (and orientation) in the datumdock properties. So every time a file is opened and A10 is started it will scan thru all the objects and put them in their place by coordinates. A copy of the coordinates is used to update locations if an object is moved. This will work something like OCC's approach to building a solid model, hopefully it works better.

The problem comes in subsequent operations that are expecting a single solid and may not handle multiple solid input gracefully.

I want the single solid concept so the rules that are used to manipulate them stay simple.

[DeepSOIC]

My vision for the new assembly concept (I'll call it A10) is to make it a passive assembler, it works on rules only. It will scan thru the parts and look at data in the datum docks.

Sounds familiar, https://forum.freecadweb.org/viewtopic.php?t=32843

[freedman]

DeepSOIC,
Thanks for the link.
I haven't been paying attention to the Assembly section all that much and I should have. I commend those working on Assembly 2,3,4.

I'm trying to visualize a macro based concept that is simple and easy to use with current design tools, something that I can integrated into Part Design or not, something that follows some basic rules, I would expect limitations.

If a few extra properties were added to the DatumPlane I believe there would be plenty of info to do a macro based Assembly tool. I wish I had the time to code it up, maybe next year. I will continue thinking about it.

Thanks