Construction, models, and pre-fab houses

Modeling is the word for the new millennium. I don’t think people realize how powerful it is to have an accurate computer model of whatever it is you want to build. It frees you to simulate, iterate, and optimize your design in entirely new ways. Back ten or so years, aerospace geeks (that’s me) were excited about the fact that the Boeing 777 was being “built” entirely inside a CAD (computer-aided design) package. People are used to seeing blueprints, schematics, and design plans, but this was something else again. Not only was the aluminum skin being modeled, but also the wiring, the plumbing, the seats, the carpet, all of the thousands and thousands of parts large and small. This let the Boeing engineers make sure that everything would actually fit before it was assembled. The project was a great success, and every plane since then has been assembled in a computer long before any metal gets cut.

A process that works with airplanes ought to work with buildings too, and so it does. The big difference is that the construction industry moves much more slowly than the aerospace industry. There’s less pressure to go high tech. But once contractors get used to working with CAD systems, the payoff will be huge. Here’s an article from Computerworld about this phenomenon: GM builds on 3-D model. The author follows the story of a factory that General Motors built, and it’s very much like the Boeing story above. Instead of printing out thousands of 2-D blueprints, they worked straight from the computer model. The computer tells you when two pipes are colliding. As a result, they were able to eliminate the costly delays that are endemic to the culture of construction.

Because collisions in 2-D projects are unavoidable, tradespeople try to get their work done first, Lemley says. When a collision occurs, everything stops while the drawings are reviewed. “You go through hundreds of drawings, and you call the architect, and they have to come down and bring a mechanical [drawing] down,” he says. That puts everyone else behind and results in expensive change orders. Building to the model eliminated the problem.

The GM project came in 5% under budget and 25% ahead of schedule. That adds up to real money on a $1.5 billion factory.

A process that works on big buildings ought to work on small ones too, and so it does. In the latest issue of Metropolis, I came across this article on bolt-together pre-fab housing: Bursting Out. Pre-fab housing conjures up images of shoddy workmanship, cheap materials, and bad taste. But in the future it will mean customized pre-cut panels delivered in an Ikea-like flat pack and quickly assembled on site. From the article:

The process borrowed heavily from industrial-design mass manufacturing. After hollowing out the solid model and developing a structural diagram based on the ribs, the architects ran commands to unfold the computer model, break up the surfaces into production-size triangles, label each piece and rib, and then organize them onto sheets for the laser cutter. This information was then run through String IT, a program used in furniture design, which “nests” it—calculating an optimum layout of the various shapes on the given dimensions of the plywood sheets to minimize waste—reducing the amount of plywood required by about 20 percent. At the laser cutter this file was run to produce 1,100 nonidentical plywood pieces, each cut, drilled, and etched to determine its location in the house. In January 2005 these arrived flat-packed in North Haven, where a team of 12 students from the architecture program at nearby Newcastle University was prepped for a fast-build process that the architects likened to a barn raising.

This technique is already proving useful in places, like post-Katrina New Orleans, where old-school house construction is too expensive and slow, too medieval to serve the needs of the community.

The first fruits of modeling are in narrow and specialized domains, but the real value comes when you start to integrate the efforts of multiple teams across multiple domains. It takes a long time to get everybody in the game, but the results can be stunning.