[Topologic] Redefining BIM through Spatial Topology, Information, and Grammars

T by topologic on 6 Feb 2021 # +5 votes

@paullee said:

@topologic said:

Hi @paullee Are you looking to try Topologic or Homemaker/Topologise in FreeCAD? If the former: coding Topologic in FreeCAD can now be tested by installing it from http://github.com/wassimj/topologicPy

I am happy to jump on a zoom meeting to walk you through it.

Many thanks !

Not aware what is Homemaker/Topologise :)

I am on Fedora 31, just follow the instruction on github ? (not much idea what each line means :D )

I understand. This effort thus far is not at the point where it is a user-friendly piece of software with a graphical user interface. @brunopostle has it installed on Fedora 32. I have it on Ubuntu 2.10 Groovy Gorilla and Python 3.8. If you are interested, my offer stands for a Zoom meeting to walk you through it. You (or anyone reading this) can book a time using http://jabiw.ycb.me

J by JQL on 6 Feb 2021 # +3 votes

I would love to attend that meeting, or any other around Topologic, even if I'm a windows user waiting for a dumb UI.

P by paullee on 7 Feb 2021 # +2 votes

@topologic said:

I understand. This effort thus far is not at the point where it is a user-friendly piece of software with a graphical user interface. @brunopostle has it installed on Fedora 32. I have it on Ubuntu 2.10 Groovy Gorilla and Python 3.8. If you are interested, my offer stands for a Zoom meeting to walk you through it. You (or anyone reading this) can book a time using http://jabiw.ycb.me

Seems quite some peoples would be interested.

Would it be better if this is a short presentation in the Monthly Meetup ? :D

Probably becoming a series of updates and/or discussion on the usage, direction of development, features etc. ?

T by topologic on 7 Feb 2021 # +7 votes

@paullee I think once we progress with the work a bit more, @brunopostle and I could perhaps give a joint talk and field questions.

B by brunopostle on 11 Feb 2021, edited 11 Feb 2021 # +9 votes

Here's a bit more work in progress, I've ported my Homemaker Topologise add-on code that identifies floors and internal/external walls to use the topologicPy python classes (that use the Topologic C++ library behind the scenes). Then, so there is something to see, I've used Dion's 'dumb wall' to draw the results in blender - This is just temporary code, I intend to reintegrate the old 'Molior' system that generates IFC entities for windows and doors etc..

As you can see, using Topologic to segment some quite roughly drawn geometry into cells is really easy and very fast - with my Perl code in the old add-on I had to ensure that all the vertices in the mesh aligned perfectly, now all that is necessary is that the faces overlap.

L by lukas on 11 Feb 2021 # +1 votes

Woahh.. is this what I think it is? I am wishing an IFC-repairing Tool. We get Revit IFCs from architects that do not respect the state-of-the-art rules. Very often the IFCspaces are not well declared. Then I need to import those IFC to the next tool (DDS-CAD or DIalux Evo) and of course it fails because of wrong IFcSpaces or Spaces-geometry descritpion. So what I see is that this combination of Topologic/Topologise can recognise volumes and declare them back to a clean IFC entity ? This is my repairing Tool ?

Any Building in Blender (IFC or not..) can be scanned and given back as an IFC Model !! Since it is topologised and topologic.ised , it is also open for different geometry descriptions... Cyril may want only internal walls descriptions, Jesusbill only a kind of wireframe.. sounds very cool..

B by brunopostle on 11 Feb 2021 #

@lukas said:

So what I see is that this combination of Topologic/Topologise can recognise volumes and declare them back to a clean IFC entity ? This is my repairing Tool ?

Any Building in Blender (IFC or not..) can be scanned and given back as an IFC Model !! Since it is topologised and topologic.ised , it is also open for different geometry descriptions...

There is still the problem that the raw geometry may have gaps, so two cells can 'leak' into each other. The other issue is complexity, Topologic identifies a cell complex in a mesh with 600 faces in about three seconds (on my ancient laptop), which is incredibly fast; a normal bim model may have hundreds of thousands of faces, plus the resulting cells would have too much detail.

This is why I suggested a two part approach: you convert the raw geometry into voxels to fill the gaps and reduce the resolution; then use this to identify cells, walls and floors; then you use this Topologic model to go back and classify the building elements properly.

J by JQL on 11 Feb 2021 #

What if topologic would ignore cells that are too thin and would then find the average axis between the thinner distance. That would allow us to turn walls into axis instead of cells. From those axis we could then auto remodel spaces. Far fetched?

B by brunopostle on 11 Feb 2021 #

@JQL I'm not sure, what about walls that are drawn as studs and panels? or openings that are modelled rather than cut with ifcopening* elements?

T by topologic on 11 Feb 2021 # +4 votes

@JQL said:

What if topologic would ignore cells that are too thin and would then find the average axis between the thinner distance. That would allow us to turn walls into axis instead of cells. From those axis we could then auto remodel spaces. Far fetched?

Ever since we published Topologic, people ask us if they can use it to go from existing BIM models to conceptual models to add information, or fix them or analyse them. Our original vision was always to do the exact opposite and we need to change people's minds about this: Start with a simple conceptual model (e.g. a CellComplex), stay with it for as long as possible to analyze it, optimize it etc and when ready, use shape-grammar rules to 'thicken' it and create very well-formed BIM/IFC models. If I understand the video, this is what @brunopostle is doing. He is creating a Cell, slicing it with Faces and when done, creating a CellComplex and then 'thickening' it into a wall,floor, ceiling type of model.

Going from Complexity to Simplicity is always going to be fraught with errors. Having said that, we are doing some interesting things with grouping rooms in a Revit model into apartments by creating the graph and checking for islands in the graph. It took a 5-day job and reduced to mere minutes.

L by lukas on 11 Feb 2021 # +1 votes

Once everybody is opensourced, sure the simple to complex way is the right choice.. but for now, we have to deal with BIM-unaware/incomplete/mismanaged models and try to use them.. so the correction/repair tool is a kind of go in between as long as the world is not openBIMed. I do not get any chance to experiment myself.. think I need 3 weeks holiday to install and learn. I pay a xxxx.-€ to Osarch for a tutorial and workflow that would bring me the one revit-ifc here to a usable DDS-CAD model that would also go into Dialux-Evo. Probably not enough.. need to deal with import-capabilities of DDS-CAD and Dialux-Evo. But in fact I should ask these companies to pay. That is why I xx.ed the amount. Should go for xxxxx . This IFC-compatibility tool would be a runner to my opinion. (just walls, doors, windows, stairs and 2nd-ceilings)

J by JQL on 11 Feb 2021 #

@topologic said:

@JQL said:

What if topologic would ignore cells that are too thin and would then find the average axis between the thinner distance. That would allow us to turn walls into axis instead of cells. From those axis we could then auto remodel spaces. Far fetched?

Ever since we published Topologic, people ask us if they can use it to go from existing BIM models to conceptual models to add information, or fix them or analyse them. Our original vision was always to do the exact opposite and we need to change people's minds about this: Start with a simple conceptual model (e.g. a CellComplex), stay with it for as long as possible to analyze it, optimize it etc and when ready, use shape-grammar rules to 'thicken' it and create very well-formed BIM/IFC models. If I understand the video, this is what @brunopostle is doing. He is creating a Cell, slicing it with Faces and when done, creating a CellComplex and then 'thickening' it into a wall,floor, ceiling type of model.

Going from Complexity to Simplicity is always going to be fraught with errors. Having said that, we are doing some interesting things with grouping rooms in a Revit model into apartments by creating the graph and checking for islands in the graph. It took a 5-day job and reduced to mere minutes.

And that's exactly why I think topologic is the way to go. However the opposite would be an interesting possibility nonetheless.

@brunopostle said:

@JQL I'm not sure, what about walls that are drawn as studs and panels? or openings that are modelled rather than cut with ifcopening* elements?

Maybe the threshold for how thin would ignore the thickness of those altogether?

T by topologic on 11 Feb 2021 #

@JQL said:

Maybe the threshold for how thin would ignore the thickness of those altogether?

The problem is if the model is like the picture below, the wall lines wouldn’t close and different rooms will blend into each other as one cell.

P by paullee on 11 Feb 2021, edited 11 Feb 2021 # +4 votes

One of the approach I envisioned / experimented in FreeCAD to build the Villa Savoye model is simply use of the Sketch object there to 'outline' the layout of the floor plan / CellComplex.

Then, currently it can be used to build with ArchWall to form the model.

And I have initial success to use this Sketch to extrude ArchSpace from it, that can be CellComplex. This ArchSpace was early @topologic demonstrate to 'import' into Topologic to turn into CellComplex I understand

1st August 20:00 UTC Monthly Meetup

J by JQL on 11 Feb 2021 #

And Sketch based architectural workflow with FreeCAD is really a cool concept. If topologic comes into play with it, FreeCAD will truly become an amazing architectural modeler.

T by topologic on 11 Feb 2021 # +1 votes

@JQL said:

And Sketch based architectural workflow with FreeCAD is really a cool concept. If topologic comes into play with it, FreeCAD will truly become an amazing architectural modeler.

I agree. How far fetched would it be to find something like sverchok (or sverchok itself) for FreeCAD? That would be amazing.

P by paullee on 12 Feb 2021, edited 12 Feb 2021 # +1 votes

The @brunopostle workflow and mine seems different in quite some aspect.

Mine : -

2D Sketch --> (may have Height dimension built-in) --> each Edge 'thickened+extruded' to become ArchWall

..................... --> ............................................................................. --> areas defined by edges become ArchSpace

Then, conceiving to use Togologic to do analysis / generate bubble diagram on Spaces

Brunopostle :-

3D 'Boxes' --> Using Topologic at this point ? --> each surface 'thickened' to become Walls / Slabs ?

Seeking Algorithm

BTW, as I am not trained as programme, I do not have a good algorithm to deduce the 'cells' from the 2D Sketch.

Any idea ? :D

B by brunopostle on 12 Feb 2021, edited 12 Feb 2021 # +2 votes

@paullee said:

The @brunopostle workflow and mine seems different in quite some aspect.

Brunopostle :-

3D 'Boxes' --> Using Topologic at this point ? --> each surface 'thickened' to become Walls / Slabs ?

I'm drawing boxes, or just polygons, these don't have to align perfectly, as long as they enclose space Topologise will identify the cells and create a cellcomplex. You don't actually see the cellcomplex, I iterate through it to extract 2d floorplans for each level - these are equivalent to your 2d sketch, except I can link every element back to a position in the cellcomplex. Then I'm doing the same thing, extruding wall centrelines etc..

Seeking Algorithm

BTW, as I am not trained as programme, I do not have a good algorithm to deduce the 'cells' from the 2D Sketch.

Any idea ? :D

This isn't too hard a problem, I solved it in 3d for my previous Topologise tool. Basically iterate through all your lines, allocate a 'room' to the left and to the right of the line, then propogate this information to adjacent lines until there are no lines left. Alternatively, you could convert to 3d and let Topologic [edited] sort it out for you.

T by topologic on 12 Feb 2021 # +3 votes

@brunopostle my recommendation for assigning 'rooms' to spaces is the following:

Get the Cells from the CellComplex (or draw initial volumes and convert them to Cells)
For each Cell, create a corresponding Dictionary with any key/value pair that you want (e.g. "room_name"/"bedroom") (Dictionary.ByKeysValues)
Assign the Dictionary to the Cell (Topology.SetDictionary)
Build a new CellComplex from the Cells (CellComplex.ByCells -- now all the Cells in the CellComplex have a dictionary)
You can now get the interior and exterior walls, bottom slabs, intermediate slabs, and ceiling s (Faces) of the CellComplex, ask them to return the Cells that they are separating and then ask each Cell what room type they are and set the thickness and other attributes of that Face accordingly.

I know your process is more about lines and extruding them, but I cannot help but think a fully 3D approach would give you more flexibility (e.g. inclined walls) which Topologic can handle just fine. I am also not 100% sure the above works well in 2D. We have not tested that extensively. But you can try a 2D shell of faces separated by Edges and you can ask an Edge to return the two faces it separates (or one Face if it is an external boundary edge). It might works exactly the same as in 3D.

Regardless, I am looking forward to seeing the amazing IFC models out of Topologic/Topologise/HomeMaker! :)

A by arquitextonica on 12 Feb 2021 # +2 votes

Amazing work @paullee. I´m currently working on my thesis and my research approach is almost identical to yours, though I got some advantages from the tools. I´m working with Rhino and GH and there the 2D bounds recognition algorithms are already strong and stable, and above all, @topologic gives a lot of strength.

So as I saw you already built the spaces from the edges, I´ll throw my two cents.

I´m currently working by making closed space contours with 2d polylines. I then convert it to a topologic object with properties dictionaries and operate within them. Then, the resulting objects have properties dictionaries and allow me to perform topological operations among them. The result for me is very interesting because through the overlap of the properties, new properties can airse and we get "emergence" for design.

B by brunopostle on 12 Feb 2021 # +2 votes

@topologic said:

I know your process is more about lines and extruding them, but I cannot help but think a fully 3D approach would give you more flexibility (e.g. inclined walls) which Topologic can handle just fine. I am also not 100% sure the above works well in 2D. We have not tested that extensively. But you can try a 2D shell of faces separated by Edges and you can ask an Edge to return the two faces it separates (or one Face if it is an external boundary edge). It might works exactly the same as in 3D.

This comes down to a question about 'what architecture is'; saying that a wall can be any-thing introduces a huge amount of complexity that in itself constrains action later-on. Whereas accepting that floors are horizontal, and that walls are vertical and follow a 2d path brings real flexibility - I can put a door, window, shelves wherever I need to.

When we think of BIM, we generally think of wonky upside-down buildings, but the IFC specification actually has all this great stuff for describing 'ordinary' buildings: IfcStorey, IfcWall, IfcDoor, IfcOpeningElement - if we use these things then it will be possible to have the software check the design completely, and to automatically generate clear drawings that can be used to construct it. A frustration is that useful ideas that would make this easier, like the *StandardCase building elements, are being dropped from the IFC standard.

I do plan to support inclined faces, but these will be 'roof' - I'm beginning to think that Topologic may be the perfect tool for designing pitched roofs; dealing with hips, gables, combinations of hip and gable, dormers etc..

J by JQL on 12 Feb 2021 # +3 votes

@arquitextonica

What you might not have realized about Freecad is that it's Sketches are parametric geometries that are the foundation for a Freecad model. Change the parametric relationships between a sketch and all the model changes accordingly. It's way more powerful than a set of 2D polylines. It's an incredible methodology. Having sketches have a meaning and having topologic build everything else beyond sketches, based on that meaning, would be amazing.

@brunopostle

I think it's very interesting to think on a tool to develop standard architecture in a zip if that's what you're aiming for. However, topologic allows for much more and what's brilliant about the future is how we can free ourselves from constraints. When all software is capable of creating standard architecture with 2 clicks of the mouse, what will matter is creativity and the freedom that software will allow the human mind. I'm interested in that and I'm afraid of making the standard way of doing things too easy, as it might result in the production of the average. Your work is amazing but I would love to see it with a wider scope. If Topologic allows it, why not?

P by paullee on 12 Feb 2021 #

@brunopostle said:

This isn't too hard a problem, I solved it in 3d for my previous Topologise tool. Basically iterate through all your lines, allocate a 'room' to the left and to the right of the line, then propogate this information to adjacent lines until there are no lines left. Alternatively, you could convert to 3d and let Topologic [edited] sort it out for you.

I try to imagine, not sure I understand how to move forward :)

Say, below layout, 6 edges ( 3 'vertical', 3 'horizontal'). It defines 4 areas.

I can't figure out the algorithm to find out the 4 areas automatically. Currently, I need to do that manually - i.e. pointing out which 4 edges define an area and need to do that 4 times :(

T by topologic on 12 Feb 2021 # +1 votes

@paullee said:

I try to imagine, not sure I understand how to move forward :)

Say, below layout, 6 edges ( 3 'vertical', 3 'horizontal'). It defines 4 areas.

I can't figure out the algorithm to find out the 4 areas automatically. Currently, I need to do that manually - i.e. pointing out which 4 edges define an area and need to do that 4 times :(

This is a very easy workflow for Topologic. Feed the edges into a Cluster, self-merge it, and feed it to wire-cycles with maximum 4 vertices (this is a custom node written in pure python available on TopologicDynamo. Planning to port it to Blender/sverchok. It will return 4 closed wires.

P by paullee on 12 Feb 2021, edited 12 Feb 2021 #

@arquitextonica said:

Amazing work @paullee. I´m currently working on my thesis and my research approach is almost identical to yours, though I got some advantages from the tools. I´m working with Rhino and GH and there the 2D bounds recognition algorithms are already strong and stable, and above all, @topologic gives a lot of strength.

So as I saw you already built the spaces from the edges, I´ll throw my two cents.

I´m currently working by making closed space contours with 2d polylines. I then convert it to a topologic object with properties dictionaries and operate within them. Then, the resulting objects have properties dictionaries and allow me to perform topological operations among them. The result for me is very interesting because through the overlap of the properties, new properties can airse and we get "emergence" for design.

I do not know Revit, Rhino, GH etc. except AutoCAD probably upto 2008, and recently FreeCAD (on Fedora linux) :D

I am trying to understand the capabilities of Topologic, and see if and how the worflow in FreeCAD / daily Architect design development could fit in. Maybe you can show more what you are doing to demonstrate your idea ? Thanks in advance.

BTW, there is an software called TAD which seem have similar concept - Room/ Space as fundamental building block of design development.