The following editorial reflects the opinion of the author and not necessarily that of Progressive Engineer.

It's Dangerous to Over-Rely on Computer-Created Designs

By Stephen M. Benz, P.E.

As I write this, the year 2001 draws to a close. I'm reminded of the film 2001- A Space Odyssey, in which Arthur C. Clarke gave us of his vision of the past year -- the Moon would be colonized, man would voyage to Saturn, and our fates would be determined by a computer named HAL. The real-life version of 2001 didn't quite work out that way, but you've still got to give the guy credit for his futurist vision. In the 1940s, Clarke predicted our society would communicate between continents using geostationary satellites. I think of this when I call friends in other countries or sit down to watch one of the several hundred television channels available on my satellite dish.

Using my own crystal ball to look ahead to the future of civil engineering, I see exciting potential on the horizon, but the changes could cause problems if not managed properly. These apply to other engineering disciplines as well. Rapid advances in computer technology have accelerated change in the design industry, beginning with the advent of CADD. And today, we communicate across the World Wide Web to share design files and project information. However, this is just the beginning.

Automated design systems, construction robotics, and intelligent engineering systems are coming. Recent advances in the use of construction robotics underscore the imminent nature of this technology. For the past decade, the Japanese have experimented with automated systems that construct entire buildings using computer-driven robots. One world-renowned architect is designing free-form titanium structures using 3-dimensioning CADD and turning the electronic files over to the metal fabricator for computer-controlled fabrication of the titanium skin panels.

The availability of powerful and affordable technology is pushing the design industry into uncharted territory. How long will it be until someone develops a black box that allows the designer to input parameters defining a project and get a complete design in a few minutes? Does this sound far-fetched? It shouldn't! Black boxes may develop as a continuation of our current computer design practices.

Herein lies the problem. Engineers use a variety of computer-based tools in their practice, ranging in complexity from simple spreadsheets to sophisticated design and analysis programs. As the complexity of a program rises, the calculations and algorithms are more deeply embedded. Many of us old-timers are accustomed to performing designs using step-by-step procedures. After one calculation is validated, subsequent calculation steps follow, eventually leading to a complete design. In designing this way, we developed a feel for the validity of our overall design. If something didn't look right at the end, we could revisit the steps, check each calculation, and revise the design until it came out the way "it should have" based on our experience with similar work.

With modern computer design methodology, however, the calculation steps performed by the computer are hidden from the designer, meaning they must trust the results of the computer without seeing the design process unfold. In many cases, designers may not even know which algorithms are used by the computer. If experienced enough, they may be able to get a feel for the results produced by the computer, based on their individual experience. For example, a structural design program may tell a designer to use a certain beam section for a building, but it just doesn't "feel right" based on his or her experience. He or she may then spot-check to confirm the output of the computer design program. In this role, designers become validators of the computer program results, drawing from their experience to check the output .

But what about junior designers, who haven't developed a similar level of experience using manual calculation techniques? When faced with computer results, they can't apply the same level of reality check as their more seasoned associates. Computer program logic can contain flaws, but more likely, input data has errors that may not be caught if the operator can't validate the results.

Modern design programs continue to evolve and take different forms. The current state of the art is interactive design, where the designer provides input through a CADD-based interface. The computer input is a series of graphic commands upon which the design system bases subsequent calculations and provides detailed output. While this process seems to make the designer's job easier, there is often no clear record of the input used to create the end result. For example, a designer could sketch a roadway centerline using a flawed technique such as non-tangent curves. The CADD-based system would use that input to create an entire flawed roadway layout. If the results aren't carefully checked against the input, the problem may never surface until the contractor lays out the design in the field.

As computer design systems mature, they will become more powerful (read-complicated) and shield their algorithms from the designer even more. It may be possible soon to specify input criteria as simply as "25,000 GSF footprint, seismic zone X, Building Code Y, three floors" and get a building designed in a matter of minutes. A list of input will drive the program to spit out a series of CADD plans and supporting documents. This black box technology will bring the role of the designer to the forefront. What responsibilities will the designer have? Will design responsibility be shared with the author of the black box program? Should these systems even be allowed to develop in the future?

One thing is clear: the design industry is changing. CADD was merely the beginning. In many respects, our early forays into CADD were an emulation of the manual drafting process. In essence, we substituted our pens and mylar for the mouse and plotter. As the power of computer-based design continues to evolve, our roles as designers will change also. Let's make sure we're ready for it.

Steve Benz is Chief Engineer at Boston-based Judith Nitsch Engineering, Inc., a civil engineering and surveying firm. He has developed several automated design systems, published dozens of articles on design automation topics, and authored The Project Manager's CADD Survival Guide (ASCE Press, 1997).

Want to write an Editorial? We welcome them from members of the engineering community. Phone 570-568-8444 or e-mail progress@jdweb.com for details.

Progressive Engineer
Editor: Tom Gibson
2049 Crossroads Drive, Lewisburg, PA 17837
570-568-8444 * progress@jdweb.com
©2004 Progressive Engineer