The Importance of Failure

Article excerpt

Using what goes wrong in engineering to teach how to do things right

Do engineering students think about failure? Not about failing a test or a course, but about failure as an engineering concept? If they don't, they should. Failure can teach students that yesterday's mistakes can lead to today's solutions and tomorrow's innovations; it can offer them new career options; and it can introduce them to ethics and professional responsibility.


Author, engineer, and educator Henry Petroski is a historian of engineering's negative (and positive) consequences and outcomes, and a believer in the axiom that those who forget the mistakes of the past are doomed to repeat them. He urges engineers and engineering students to look to what's gone wrong before as a way to anticipate what can happen again if they don't take proper precautions.

"Failure is a unifying theme of engineering," Petroski says, yet engineering curricula often focus on successful designs and neglect unsuccessful ones. Ironically, this reliance upon past successes can lead to future failures.

"One of the paradoxes of engineering is that successes don't teach you very much. A successful bridge teaches you that that bridge works," Petroski says. It does not teach you that the same bridge, built at a different location or made longer or taller, will also be successful. "It's all theory until it's completed," he explains.

A case in point is the Tacoma Narrows Bridge, which shook apart in the wind just a few months after opening in 1940. Leon Moisseiff based the bridge design on the designs of several successful bridges, yet at the same time ignored the wind-related problems that had damaged other bridges.

A civil engineer investigating the Tacoma Narrows collapse wrote that the bridge's failure "came as such a shock to the engineering profession that it was surprising to most to learn that failure under wind action was not without precedent," Petroski writes in Design Paradigms: Case Histories of Error and Judgment in Engineering. In fact, "between 1818 and 1889,10 suspension bridges were severely damaged or destroyed by the wind."

Moisseiff's reliance on engineering successes and exclusion of engineering failures has a modern-day counterpart: computer simulations. "There is clearly no guarantee of success in designing new things on the basis of past successes alone, and this is why artificial intelligence, expert systems, and other computer-based design aids whose logic follows examples of success can only have limited application," Petroski writes.

The initial failure of the Hubble Space Telescope is an example of problems caused by relying on computer simulations. In 1990, when the orbiting telescope sent its first photographs back to Earth, the images were unexpectedly fuzzy and out of focus. NASA determined that the problem was the result of a human error made years before the launch: the telescope's mirror had been ground into the wrong shape. The mirror, tested prior to launch like the telescope's other separate components, functioned properly on its own. However, the manufacturers did not actually test the mirror in conjunction with the other components. The manufacturers relied on computer simulations to determine that the separate components would work together. The simulation didn't take into account the possibility of a misshapen mirror.

Because of the Hubble problems, NASA learned "a great lesson" about "the merits of actually testing a system rather than depending upon theory and simulation," explains Doran Baker, founder and vice-president of Utah State University's Space Dynamics Laboratory.

Relying on computers to predict and counteract failure can cause other problems as well.

Smaller Failures, Bigger Problems

On May 19,1998, the Galaxy 4 Satellite spun out of position and disrupted radio, television, pager, bank machine, and other satellite-linked services across North America. …