I was slotted on friday morning. I'm almost always on the last day; in fact it was the first day where I looked once I got the schedule :-) So this set of notes also includes my own presentation.
My server went down today for the first time in half a year. Exactly when I'm 1500 km away. That's why this file contains the entire day instead of having the talks splitted out.
Some news: the 2006 ecppm conference will be held in Valencia, Spain.
See the overview page for the links to the rest of my notes.
Note: The site was down since friday and there may be some more problems in the next days as I'm doing a speed-move of the site to my home machine. After three years the university suddenly found it necessary to disallow the hosting of vanrees.org on my university machine. Despite 75% work-related content... "The domainname contains your personal name".And with just a few hours warning while I was in Turkey... Berthold Firmenich (Bauhaus university Weimar) - A novel modelling approach for the exchange of CAD information in civil engineering
The objective is CAD data exchange. Normally, when you transform data from one model to another, you lose information. When a second user edits the data and transforms it back again, you lose even more data. In this traditional workflow scenario, you lose a lot of data. The problems:
Two possible approaches are the version modelling approach and the change-oriented approach. With versioning you store the various versions, with change-orientation you just store the differences.
There are existing software version control system, but they are textfile-based and cannot handle the semantics of CAD models. (Note: subversion is the best open source software versioning system. Recommended by Reinout.)
He makes a distinction between an evaluated model and an unevaluated model. If I understand it correctly, an evaluated model is a CAD model as intended by the one using it. Unevaluated means that it is a CAD model where some changes were dumped in it that still have to be handled.
The change-based system he proposes only communicates the changes. It assumes both parties (from the original example) start out with a similar model in both their different CAD systems. User A changes something and this is communicated (using the exchange format?). User B changes his model with this. Changes back are handled the same way. As the basic model on both sides is OK, the sending back and forth of changes in this way doesn't wreak havoc. Or at least not in the cumulative normal way.
Conclusion: this has advantages. The main research to be done is the design of a programming language or API for the change handling.
(Note: This sounds like a good idea. The only thing is that it sounds like the changes are communicated using some API instead of using a standard exchange format. But that's just my opinion. The idea itself is OK.)
Can we develop a common schema for core construction metadata? Three dimensions are proposed: entity, supporting concepts, categorizing concepts. Supporting concepts supports the description of the entity.
(No, I can't write down 50-item boxes-and-arrows diagrams, so you'll have to read the paper for those :-)
Entities are "encoded" using something resembling genes. (A nice way of illustrating the use of metadata, imho).
The taxonomy data behind this system can be used as a lattice, which allows a wealth of mathematics to be used to handle it all. You can also combine those lattices.
You can use those combined lattices and the genes to combine data from two companies, allowing collaborating.
Comment during the questions: this might be one of the things that is needed to push building and construction IT over the tipping point.
Another comment: lattices is a well-understood mathematic concept, so the algorithms are in place and are quite speedy.
Connet Turkey is the Turkish node in a European network. It is a subscription based bundle of internet enabled information and b2b trading services.
They are planning a whole new set of services like news server, materials market, web hosting, education, etc.
The whole technical infrastructure is open source and java-based. Java, servlets, mysql, linux, tomcat, etc.
They typed over a lot of information from paper catalogues and websites in a database. The goal is to build a building material information system.
A problem is that a lot of the suppliers are only working regionally and aren't available elsewhere in Turkey. But, most of the companies find it attractive to supply or to introduce building material information on the web.
(Note: don't literally read the sheets from the screen...)
Sometimes the building and construction industry is like herding cats. If only you could get them to cooperate. But if they cooperate they can do great things.
The semantic web
The semantic web is one step further than the normal web. On the web, you can access lots of documents with a URL and link from document to another document anywhere. On the semantic web, you can point from object to another object anywhere in some file. The semantic web extends the URL mechanism to a GUID mechanism. Build into the semantic web is the idea of ontologies. There is also a special format for ontologies: OWL.
Before we go to the semantic web practical experience, first a little bit of background on building specifications. My own research is on building specifications and the semantic web. You'll see the link with this presentation later on.
Specifications are quite central and they have many information links to other "documents" which are owned by a lot of different parties. This makes it an ideal target for the semantic web. Links exist to:
Apart from Huns, Goths and Venetians, small contractors also can bring the house down. The scenario used for this research was that of a private house owner that wanted to have an extension made to his livingroom. Normally there is no architect involved, just the owner and the contractor. The client has three basic problems:
As a certain professor fell victim to a contractor, he wanted a client support web application. Semantic-web based.
Semantic-web based means ontologies, multiple ontologies. But just simple ones dealing with what was needed ad-hoc. Objects and properties were needed, which ruled out a simple reuse of an 12006-2 classification. IFC has properties, but wasn't available in a nicely reusable XML or OWL form. So a few ad-hoc ontologies were made in a UML-like interface to protege, which helped to keep the ontologies small as the screen became crowded real soon.
The application worked. You have a house? What kind of house (subclass relation). What components? All from the ontologies, external to the application, which is good. It was T-ford like, though. Not a lot of choice.
A test was done to generate a simple building specification out of the resulting data. A ready-made Dutch specification was converted to an RDF file with the plain specification items and one with the chapter/classification structure. The specification items were linked with the data from the house example, copying over specific values etc. Coupling it with the classification structure produced a specification.
The Dutch SfB was, as a second test, also converted to the RDF structure used by the original classification and, after being coupled, also resulted in a specification (but with a different chapter structure now). Good.
Conclusions and further work
Conclusion: it basically worked.
Important: it will be an all-open-source operation. We think that is essential in this area.
The focus will stay on the small contractors and small clients.
Perhaps of interest: http://www.funsiec.org/
(Note: I'm still not really comfortable giving a presentation... They're taping the talks on cassette, I'll ask them for an mp3 of my talk to see how it went. Right now I'm not happy with how it went. This is something that doesn't come naturally to me, I'll only be able to improve it with experience).
Current CAD systems lack object specialisation and incremental class extensibility.
Incremental design is the process of incrementally adding
Even though terms like
width are well-known, the
computer needs lots of help to understand them correctly. A system
that does that is called a semantics.
Views can be seen as so-called abstract interpretations of the same conceptual model.
His software architecture has property- and language-orientation instead of object- and class-orientation. Properties are partly handled by a template system that can create a new view out of a model.
A design tool prototype has been implemented (in the Moscow ML).
His thesis can be gotten on the net: Construction informatics - issues in engineering, computer science and ontology note 2005-04-12: link corrected). The paper relies on work in this doctoral thesis.
The research was on the information flow in collaborative design meetings. In those meetings you often refer to the meeting minutes. These are structured in chunks of natural language. Relations between decisions are vaguely defined, which is a problem to later know what and why. It is difficult to reach common understanding. There is a lot of tacit knowledge, not contained in the design meeting minutes.
What is the problem in storing them electronically in project webs? It is just a transfer from paper to electronic storage, the content is the same. Often there is poor use of metadata, no references to a standard or an ontology.
They made a prototype that made use of the semantic web possibilities. The concept behind the prototype:
All the metadata was defined in RDFschema and stored in RDF. (Note: good!) The RDF is stored in a mysql database with a tomcat webserver to serve the content. Protege was used as an editor for the information.
Also a part of the prototype was a decision context editor (again protege), where you could graphically edit decisions ("this was done because of that", with "this" and "that" refering to information chunks in the database).
The prototype needs more improvements, for instance improving the web functionality. Also the practicallity of the system needs to be examined.
e-sharing allows businesses to pool resources to improve their usage. This way usage peaks and lows can be smoothed. You can select resources from a virtual resource pool that includes the idle resources of the other users.
They used bcxml to store their information. (Purrrr Purrrr, I made parts of bcxml :-) Nice to see it being used, even two years after the project ended. It was a good solution, I'm also still using ideas from that project.)
Resource matching: the user can search for an exact resource or can search for similar resources:
Included are various bidding possibilities. The prototype is already available to the consortium members.
In the 70's you had little global competition, IT was a dream and there was much public funding. 80's? IT started, east asia was on the rise. 90's: global competition, new contract types with lot of private development instead of public funding. And IT has come of age.
Also in the 90s you saw a big consolidation.
The drivers of change
There are some global drivers of change. Like aging. The population gets older and older. That has an effect on hospital buildings etc. The birth rate is dropping almost everywhere. The dependency ratio is getten bigger too: the percentage of the population that is dependent on the portion of the population that works.
Urbanisation continues. 1900: 12 cities over 1 milion. 1950: 83 2004: 410. So: congestion. Infrastructure. IT can have influence here. An airplane talks to the runway, but a car doesn't do a thing. Volvo's Gothenburg lab seems to be looking into this. Safety improvements, intelligent lighting, road lights that change color when you approach a bend to fast or when there's ice on the road. This is capitalising on IT.
Video he showed: a plastic (but very strong) bridge in the south of England. Two companies invested 1 milion in it. Now how to move on, how to capitalise on this development? That's what he wants to press on us: stay with your research, see it through; don't just chase the next new research.
Doing things differently
He introduced the project to build a new car racing track in Bahrein. 1.5 year to build it and they'd lose half the money if it wasn't ready on time. A 2000 man workforce. In that project, IT wasn't a luxury. It was an absolute necessity.
New ways of financing projects. You see more and more new contract and investment types. For big projects, the risk is big. It drives you mad financially. IT is essential here.
For really big projects, the entry costs are high. You need specialised software. You need to buy that software. You need the guys who can do it. Example was some MIT building with wildly horrendous hair-raising difficult form and materials. They needed Dassault's "CATIA" system (used to design airbusses and boeings etc.). Ouch, that's a big bag of money.
The answer? Use technology to leverage competetive advantage. Technology is a huge lever. Exploit change as an opportunity: innovate!
This was mostly the same talk he gave in the w78 workshop in Toronto, read the summary there.
Discussion about research and software. Do we build our own tools or do we do research in cooperation with the software vendors? Sometimes the software vendors are years ahead of the research community. They have the money to develop great software.
The building industry is concentrating on cheaper and faster. Quality is a victim, fallen by the wayside. We need to build quality buildings again, and (in the opinion of the speaker) IT can be a driver there.
Robin Drogemueller: They didn't want everybody going out and defining their own interface. And therefore they mandated IFC for the research in his group.
As universities, we've steered away from outputting knowledge and knowledgeable people; instead we have tried to make profits with products.
(Note: I didn't get everything).
My name is Reinout van Rees and I work a lot with Python (programming language) and Django (website framework). I live in The Netherlands and I'm happily married to Annie van Rees-Kooiman.
Most of my website content is in my weblog. You can keep up to date by subscribing to the automatic feeds (for instance with Google reader):