buildingSMART Forums

Presenting Blender as a new IFC authoring tool

hmmm. What’s @Moult say? I see your point gester, so, I’d like to see that too. Important though that, I think, Dion’s motivated by the experience of users who ARE familiar with the bim process. tightly controlled coupling breaks down often for several reasons. I think he’s trying to address that. Maybe Dion’s got something for reviewing models, so the right properties sets can be applied to the right element sets, to fill in the gaps created by normal bim usage, where the coupled tools don’t suit every needed case, and where added required properties end up not consistently applied.

and this is the main point here:

  • who’s to decide which information is the proper one for the particular granularity level, and

  • how far this can be applied automatically?

Here I explained what the Live BIM is?
And also, here and here I explained why we need SQLite/SQL-based [1] [2] [3] schemas and approaches

It’s been a little over a week, and I’m very excited to see Blender actively used on some commercial projects! I find that the modeling speed increase is by far the greatest benefit - it allows for rapid experimentation of designs and ideas, similar to sketching on paper, but with the ability to retain semantic BIM data.

As for the discussion about decoupling and coupling, I’m afraid I’m not entirely certain exactly what the problem is. Although I have some questions about the IFC spec, I generally agree with the approach and their level of (de)coupling in the schema. Perhaps you are discussing more about how vendors have chosen to implement IFC authoring and what UI they provide - but this is a separate matter. I have not broken or changed any rules of the IFC schema in my implementation. From a UI perspective, I have adopted a lot of authoring concepts from the CG world - things like mesh reuse and property assignment to geometry are old concepts that have been used for decades.

Some new features include:

  • You can now create nested element relationships: for hosted objects and nested components.
  • Material layer sets can be defined
  • Material constituent sets can be defined
  • Predefined door attributes can now be defined and assigned to door types
  • Predefined window attributes can now be defined and assigned to window types
  • It’s now packaged for more end-user testing! See below:

Blender BIM is also now available in a packaged format for Windows, Mac, and Linux! It now joins FreeCAD as being the next cross-platform BIM authoring application. It is now an installable package available here and has updated installation instructions.

Also, the OpeningDesign architectural studio has been very kind to provide an excellent repository of BIM projects to test on! They work on completely open-source projects where the entire workflow is available to the public under CC-BY-SA 4.0. They are also providing test data to test out data roundtripping between different BIM authoring tools. Big kudos to @Theoryshaw and @yorik!

Here is one of their projects imported into Blender, showing the beautiful real-time rendering capabilities of Blender and clean mesh output which allows for easy geometric manipulation thanks to IfcOpenShell.

1 Like

Dion, you did what I was scared about

You have two choices:

  1. Blender as a main platform for BIM
  2. FreeCAD as a main platform for BIM

And you have chosen the second scenario, which personally I think is wrong
I appreciate efforts on FreeCAD community, but as I said before FreeCAD is not a use friendly software (it’s like Solidworks) and it’s really hard to see it as a software many be comfortable to use it

Also, you have to see things with “national digital (delivery)” and “smart city” and “digital twin(s)” lenses to understand current obstacles related to OpenBIM, and BIM (See this project)

@Moult
i’m having hard time understanding the advantages of blender for architectural activity, and, beside obvious advantages (free modelling of everything, free assignment of data, either relational, or not) i’m missing significant issues, which might be based on my lacking deeper knowledge of blender, though.

what i miss are first of all intelligent systems that understand one another: eg. windows in the walls (including components overwrapping), flat roof drainage, which adjusts particular roof components for the slopes, automatic roof/walls framer, curtain walls with automatic edition of elements, and many, many others.

at first glance the architectural/structural modelling in blender is a cumbersome act.
am i missing something?

rob

At the moment Blender is a raw software for BIM,
But it has some vitally important bases that has the potential to make it one of the most promising BIM software in the Digital Built Environment Industry in the near future

These days all talk about AI, DL/ML, but few know what really they want from them?
If you want the software automatically recognizes the building objects/components and their relations, so some work on it in “national digital (delivery)” projects and in the near future we will see more and more solutions related to this

Blender can be a nD software and can handle all your considerations, because has “Python” consul and you can write appropriate codes for anything you want

It’s something like Grasshopper, it opens a door for many to do what they want

@ReD_CoDE - Blender BIM only contains code for Blender. It has nothing to do with FreeCAD. That said, FreeCAD is a very powerful and capable program developed by talented people. I have nothing but respect for FreeCAD.

I do not believe in a “main platform for BIM”. The industry is too diverse to be delivered by any one software. I believe in using the best tool for the task, and using open data standards like IFC to integrate together. Objectively, with Revit’s 5 modeling tools, we can conclude that Blender, a dedicated modeling program, is preferred to the task of rapid and complex architectural modeling. Equally objectively, with Blender’s lack of structural analysis capabilities, we may use a superior tool such as Tekla for producing steel work. I would encourge diversity in the industry where possible.

@gester - thank you for raising this point. Although it may differ in other disciplines, I find that in the architectural discipline the use of parametric behaviour allows for speedy modeling with one caveat: that your building is very simple. The ability to place a wall and a window within that wall in 5 clicks is alluring, but quickly reaches its limits on the “design complexity vs modeling speed” graph.

That said, once the design progresses further, or if the designer is more ambitious early on, these parametric modeling often hinders rather than helps, I find. Evidence for this is seen when architects often use other software to prototype their ideas, and then painfully redraw later in their documentation tool. Often, they still use renders from other software to communicate the designs. Another example would be stairs - where a parametric stair can be set up with a single click, but is often not quite correct, and often generates a 3D form which is simply insufficient for a certain level of detail. A third example would be the difficulty in using a parametric modeller to provide LOD400 BIM files - it would be “mostly” correct, but “mostly” is not good enough when we are using for fabrication. The simple issue of wall joins, despite the intelligence created by parametric priorities of IfcWallStandardCase, simply creates more frustrations than it solves, leading many architects to simply model layer by layer. You may also see evidence in designers using Rhino + Grasshopper to do custom parametric design (these features also available in Blender, though to a lesser extent).

That said, if you do need these functions, they exist in the Blender ArchiPack. They are also rather trivial to build and extend - creating and splitting walls is practically instant, and automatic voids for inserted elements are not far off. If you are missing drag and drop content libraries, there is a huge ecosystem out there from the Blender Cloud, and trivial to create. Reusable content libraries and parametric geometry is not unique to programs like Revit and ArchiCAD. They are plentiful and easy to create.

You would also be surprised at the speed of modeling, if you have not yet seen a fully trained CG modeler employed to model a building - a building for them is a very simple exercise compared to the complex Hollywood-style scenes that they aspire to. A project I am on has demonstrated that we were capable of modeling a 5 storey building down to the curtain panels, screw flutes, baffles in the precast panels, backing rods, sealant and packers, in a meagre few weeks, whilst simultaneously designing the building (and doing things like changing the grid - a very parametric modification). This model can then be brought back into a documentation tool like Revit to have sections / plans cut out and placed on sheets. The model has been animated, rendered (in real-time), experimented with lighting and texture, and had prototype sample designs in minutes while they were being discussed in meetings. If you have not seen some speed modelers, you may be very surprised to see what you might discover :slight_smile:

Modeling features aside, I am building as close to native IFC support as you would find anywhere - leading to a large amount of control over the BIM data. At the moment, I believe I have implemented more OpenBIM features than Revit, although I would love to be proven wrong.

Blender has a huge ecosystem: animation, building physics, amazing lighting simulation capabilities, and, of course, a very customisable setup with an interactive Python shell.

Features aside, the industry needs a bit of open-source :slight_smile: We’ve had monopolies for a while now, and I’m not sure how much benefit they have brought.

4 Likes

I shared the answer for all your considerations, but just few people can recognize what the solution is:

BIM HAS BASE ISSUES

BIM has some base issues that even repeated in ISO 19650 that if we don’t solve them we won’t see a real Digital Twin(s)

“BIM IS A STATIC APPROACH BUT WE NEED DYNAMIC ONE”
“BIM HAS DEVELOPED FOR DESKTOP AND SIMPLE SYSTEMS, IT IS NOT EFFICIENT FOR COMPLEX SYSTEMS LIKE DIGITAL TWIN(S)”

I included the answer/solution inside the content:
Keys: Dynamic + Complex
So, who can recognize the answer? Those who are in Building Performance and Building Energy area :wink:

BIM+Simulation is the answer. “The industry needs a new schema/approach”

I talk about a live model, a real representation of reality inside the software

And to achieve this level we need to combine IFC with at least two other things (which both are from simulation territory)

So, I shared the solution with buildingSMART friends, especially to @jwouellette, @berlotti, and a little bit with @jonm
Plus @Moult which knows clearly what am I talk about

However, what am I thinking needs time, and I ensure that we will see it soon, from buildingSMART or maybe other organizations

The industry is competitive and those who plan better and act better in the long-run are the winners

@Moult
‘leading many architects to simply model layer by layer’

no, neither archicad nor vectorworks have this necessity for the walls, it may only concern revit, which is actually a software for engineers, not designers.

@ReD_CoDE
i must say your ideas are very convincing. i have to start processing those thoughts, but i think i may be advocating the direction.

how about the uniclass 2015 schema?
rob

Uniclass is a classification which personally like its taxonomy

that’s what i thought :slight_smile:

There is now a short example video of BlenderBIM in use, for creating a detailed wall.

You can check it out here:

which video format it is? i can’t open it on my mac…

I also can’t see it on my windows pc. Only black screen and a hint about a non public ip-address.

I have mirrored it here:

it’s great, of course, to have such control over all elements (reminds me somehow of times way back in the middle of 90s in vienna as i’ve modelled the whole existing attic for 3 flats in microstation, without any building element and no free modelling).

otoh, imagine a project with 10’000+ msq usable area…

Of course, you can choose fully how much detail you want to go to. If you want to model at a lower geometry fidelity, you can choose to do so.

I have two projects currently testing this, one with ~750 elements (not counting repetitive instances) which is a relatively small building, and the other with 31,000 individual elements. This is similar to what the film industry deals with and it most certainly is manageable. The process shown in the video is manual, but there are bulk assign/editing tools available. This is certainly being tested on big projects, because we really do work on big projects :slight_smile:

Update time: apart from the COBie developments which are already covered in another thread, BlenderBIM now has an interface for people to use (including friendly drop downs and filters for classifications) which allow people to do bulk classifications and assignments of types.

It also has a QA module which allow for QA, which is really exciting! Because very often we get IFC models which are very badly misclassified.

image

The QA module includes a feature to allow applying a temporary colourscheme to the model based of a property. For now, the interface allows you to colour it by IfcClass, but you could theoretically colour by anything (fire ratings, acoustic ratings, any property …). Here is an example, showing walls, doors, columns, slabs, and some floating building element proxies:

You can bulk approve or reject classifications and keep track of what elements you’ve audited for various properties. Any IFC property auditing will get tracked in plaintext format in Gherkin syntax, so it reads like English:

Even though the sentences read like English, they function like unit tests in software, so they are run automatically on model publish to a CDE. If any regressions are made in the BIM model, we can get notified instantly. Standard reporting like no. of issues, changes over time are all there because it publishes to JUnit XML, so pretty much anything can read it including any CI system like Jenkins. Here’s an example of the tests being run on the model - red ones are QA failures.

We now have workflows that allow us to rapidly audit and continuously monitor the OpenBIM quality of projects.

6 Likes