Intelligent Architectural Modeling

By Martyn Day

September 2002

AEC intelligent Objects and the Single Building Model have all been with us for many years, but few practices have taken the leap to modeling their designs in their entirety. Martyn Day finds out why.

The AEC software developers have been marketing and trying to up-sell the arrival of their current generation of intelligent modelers for quite a while now, however the adoption and usage has left a lot to be desired. While companies like Autodesk and Bentley have chosen to develop this functionality using their existing 2D systems as a foundation, developers like Graphisoft, Nemetschek and Revit approached the problem with ground-up 3D architectural systems.

The software developers are trying to produce design software that will model an entire building in 3D, using not just lines, circles and arcs but geometric and intelligent compound objects, such as 'virtual' doors, walls, windows etc. Each object has a graphical display component (depending on the view required - 2D/3D, British Standard symbols etc.), attribute data (as to what its parameters are, this could be cost or finish) and 'behavior'. The concept here is that these 'objects' will all interact within the developer's CAD package, with the architect or engineer creating a virtual model of the building/structure - e.g. walls will recognize windows once inserted and trim a correctly sized hole - and will heal, should a window get moved or deleted.

As the Architectural, Engineering and Construction process is heavily reliant on 2D drawings, the idea is that from this one single model, the CAD software will create all the necessary sections and elevations, which will automatically update, should the core Single Building Model change. As one could expect, the model would also contain quantities and costs which could also assist the QA part of the job. The model would also prove useful for structural analysis, clash detection, services and structural member placement and provide the core geometry for any rendering or fly-through animations. Essentially, the ideal is that the entire design process, from concept, to fabrication, through maintenance to decommission, becomes 'model-centric', with all data being in the single model. Well, at least that's the concept.

The challenge

All this sounds simple enough but in practice developing an architectural modeling tool that's this comprehensive, whilst being infinitely flexible to accommodate the designer's geometric whim, together with a high degree of ease of use, is a very long-term computer science project.

To deliver on the Single Building Model (SBM; sometimes called the Virtual Building Model), software developers had to 'define' standard architectural objects, walls, doors and windows - analyzing what parameters architects would want to alter, working out what additional information should be included in each 'object' and then establish the degree of interoperability between the SBM objects in the environment i.e. what happens when a wall meets a ceiling? How should a staircase and wall interact? etc. Now multiply that by the number of individual components (objects) in a typical building - tens of thousands, each component not only has a geometric representation to model but also a behavior to capture.

Software doesn't stay still, it's always under development, so it's also important to think about the revisioning of objects from release to release. As 'Objects' are programs within programs it's important to maintain backward compatibility with each and every 'object', otherwise the model intelligence could be compromised. So if you're modeling a Nuclear Power Station you might want to be able to load up your Single Building Model in subsequent releases of the software and still get an integral, working model e.g. you might have modeled your Nuclear Plant in SBM CAD release 1 and used the intelligent object called 'Important Nuclear Valve' which is at 'revision 1' but three years later you now want to now open the model in SBM CAD version 23, in which the 'Important Nuclear Valve' object is at revision 23.8. It's essential that version 23.8 can understand all the behavior, characteristics and graphical representation of version 1. While you might have had problems in the past because you had trouble getting the right version of Word to open a document, 'objects' will offer this issue with every instance within a model, so it's a potential problem at a much more granular level.

Then of course there's the issue of non-standard geometry, fixtures, fittings and detail that may not have any intelligence - all of which the developer couldn't possibly anticipate or accommodate in the big plan. And I forgot to mention that all the developers created their own SBM systems with not only geometric differences (the old data translation problems that we know and love) but also new potential minefields as they all have completely different ideas on how building objects (walls, doors, windows) should be defined, creating a potentially massive data sharing problem should the market move en masse to these new systems.

The one common theme from developers is that an intelligent Single Building Model would utilize a single database file for the project - not a million 2D abstract drawing files to define a building, but one database which generates all sections and elevations from the 'live' model, on demand. While this sounds fine and dandy, it ignores the reality that the construction process is highly federated. The architect is responsible for geometric configuration of the building while the engineer is responsible for a structure's performance and the contractor is responsible for fabrication (on average there's 27 companies involved in a construction project). Each has their own representation standards and symbology and as we all know in this blame-culture ridden industry, each profession guards its role and data. To adopt a single database CAD system would mean getting highly integrated, sharing knowledge and accountability i.e. changing the very way the construction industry has come to work.

So, there you have it, it's not just a question of moving from 2D to 3D or from dumb graphics to intelligent objects, it's also an issue of transitioning from working on layered and referenced 'files' to a information rich centralized database. So while there may be benefits to moving to intelligent modeling, the implications of adopting a Single Building Model system are incredibly far reaching and influence not only your firm's technology infrastructure, process, training requirements and standards but have the potential to impact the whole industry's culture and methodology.

New software old problem

In the early 70's CAD systems became commercialized for larger mechanical engineering firms and these systems were surprisingly advanced, offering features such as 3D and parametric design (resolved relationship-based geometry). Software developers appear to have been pretty ambitious back then, considering the processing power, lack of tools, poor system storage and low levels of system memory available.

Software like the Building Design System (BDS) and General Drafting System from Applied Research of Cambridge (ARC) and another Cambridge-based development, RUCAPS pioneered CAD within the architecture market and again, here, the developers were not looking to mimic the drafting process but were thinking of a different three dimensional way to design. Many of the concepts within RUCAPS were the same as those now being presented by products like Revit and TriForma, a central 3D parametric database offering multi-user access to a single building model, with drawings treated as 'graphical reports'.

Should the CAD software model
the construction process - or should the construction process work like the software?

RUCAPS, in particular, established the idea of a 3D model composed of an arrangement of 3D components (or objects), from which 2D views could be extracted and those views arranged on a drawing sheet, with a relational database loaded from the components within the models, and non-graphic reports run on the database. Sounds familiar?

This type of system made a clear distinction between the definitive 'model' data and reports (drawings). The benefit here is that one would always know that the set of drawings produced from the model were always up to date. Users of the early systems found that they had to flesh out their designs more at the conceptual stage, so they may produce a drawing. Once the concept was modeled, it was easier for other project members to work on details. However the systems were too rare and too short lived for there to be any major re-working of the construction process to accommodate these innovative first generation 3D systems.

RUCAPS used the concept of 'box geometry', where pre-prepared 2D views of a 3D object were pasted onto rectangular prism and viewed. These components could be modeled, created by user-defined code or generated by basic parameter lists. In modern terms, its developers will admit that RUCAPS was extremely crude but it's essentially the same concept that Bentley, Autodesk, Revit, Nemetschek and Graphisoft are all selling within their respective AEC products. RUCAPS’ main problem was inflexibility, as the geometry was highly limited (no curves, surfaces or solids!!!) and the hardware it ran on was expensive and slow (early CAD adopters would have paid approx £150,000 for a seat including hardware).

RUCAPS transformed into Sonata, developed by T2 Solutions (T Squared Solutions), which was eventually bought by Alias|Wavefront. While it improved over time, at £30,000 a seat in the early 80s (plus Silicon Graphics or Apollo Workstation), it was soon mothballed by the combination of IBM's low-cost Personal Computer and 2D drafting products like AutoCAD from Autodesk.

These 2D drafting applications allowed traditional draftsmen to use their existing skills and just replace the drawing boards with a computer - everything continues as was before with the same workflow, same management structure and strategy, although with the added benefit of being able to quickly edit and plot. And there the industry has remained.

According to Robert Aish, Director of Research at Bentley Systems and one of the key figures behind RUCAPS, "CAD took a different direction in electronics, automotive and aerospace (compared to AEC) because the geometry models had to be unambiguously defined and there were real advantages to directly link these geometric representations to digital analysis, simulation and the fabrication processes. In AEC, the fractured workflow, where communication is via graphical shorthand (drawings) which are an incomplete, partial representation, still required human interpretation."

So the early developers and adopters in AEC CAD were keen on 3D and introducing a new way of working. They had high expectations on their substantial return on investment but unfortunately the inflexibility and performance of the systems let them down. That with the introduction of low-cost software put the final nail in the coffin.

A new hope?

So is this a case of history repeating itself or is there more to today's intelligent architectural systems? The most obvious difference is that if you buy one and it's not right for you, you won't have made a £150,000 bad decision! Many have modeling tools available either as part of their subscription or for a minimal incremental price.

From looking at the array of systems on offer, there is a great deal of work still to be done to deliver on the original concept. The biggest problem seems to be ease of use, especially with SBM products based on existing drafting tools, with products ending up slightly lost in what was a 2D system. That said, all the products I have seen need work to complete the data model (define more objects like lintels as an example). Some have a high bar of entry, as they require you to define your own suite of components (standard parts) prior to even starting a project. Interoperability of SBM models currently isn't a big issue as there are so few users, however, it's a ticking time bomb. The International Alliance for Interoperability (IAI) has been developing an independent standard, called Industry Foundation Classes (IFCs) but this appears to have been put on hold for now.

The good news is that the processing power is cheap, the 3D geometry capability is pretty advanced and computer software has advanced to the point where new capabilities may be possible for model exploration and development, such as the automation of glazing schedule for complex designs or maintaining real-world relationships like keeping the number of seats in a sports Stadium constant, as the pitch of the stands are altered. This is where CAD software can really increase productivity, by allowing a vast number of iterative design studies to be produced in a fraction of the time.

However, the big question is about process. Should the CAD software model the construction process or should the construction process work like the software? Unfortunately, I think there's a novel in the answer to that question! According to Aish, users should look to take a "systems" approach to building design and engineering and view a building as a set of 'interacting engineering systems', structure, services etc.. By looking at the existing 2D process, it's obvious that design data could be more effectively created and used to do 'new' things like directly drive a fabrication system - cut steel beams, timber frames, and provide further savings.

Perhaps unsurprisingly, the area where the Single Building Model is getting the most attention is within Design/Build firms that own and run the design and construction process for a particular project. These new intelligent CAD systems map more easily to their internal infrastructure and processes and offer improved productivity. So, perhaps the construction industry is already in the midst of changing its process. The rise of the design/build player isn't being driven by the advent of new CAD technology, the process is one of straight business competition - design build firms guarantee cost and delivery, providing accountability. A system which offers reduced design time, with reduced errors, is attractive to companies that are looking beyond providing a single task of 'just creating the geometry' within the federated construction industry.

There's still work to be done by all the software vendors, of that there's no doubt, and their varied attempts to oversell the first generation SBM products certainly burned many early adopters and muddied the technology waters. Despite the false starts and all the potential transition problems, the software is steadily improving and with the construction industry consolidating, there's a more chance for the Single Building Model to succeed this time around.

www.autodesk.com, www.bentley.com, www.graphisoft.com,  www.nemetschek.com, www.revit.com, www.iai.org.uk

Next month

I will investigate how today's generation of Single Building Modeling products could be deployed within a 2D dominated design process to improve drafting productivity, design exploration and communication.

Related Articles