Past and Future Failures

A reprint from American Scientist the magazine of Sigma Xi, The Scientific Research Society This reprint is provided for personal and noncommercial use. For any other use, please send a request to Permissions, American Scientist, P.O. Box 13975, Research Triangle Park, NC, 27709, U.S.A., or by electronic mail to [email protected]. ©Sigma Xi, The Scientific Research Society and other rightsholders © 2004 Sigma Xi, The Scientific Research Society. Reproduction with permission only. Contact [email protected]. ENGINEERING PAST AND FUTURE FAILURES Henry Petroski iterary history would hardly seem to have bridge type whose evolution appeared most much in common with structural engineer- closely to be following the pattern that emerged ing. Still, a recent development in the study from Sibly’s work was the cable-stayed bridge. Lof literature has revealed temporal patterns in Though cable-stayed bridges have become no- the rise and fall of literary genres surprisingly torious for the unwanted motion of their cables, similar to those related to the success and failure no dramatic catastrophic failure has yet occurred of large bridge types. For some years now, the lit- in a completed bridge, so it is reasonable to ask erary scholar Franco Moretti has been applying why not. Furthermore, it is also reasonable to ask quantitative methods to the study of the novel. whether in the past few years any other bridge In the recently published first in a series of three type did suffer an instability significant enough projected articles, he has proposed that the genre to continue the cycle identified by Sibly. First, has had not a single rise but rather that different however, it is be helpful to review the historic forms of it have developed, evolved and disap- bridge failures that establish the pattern and peared in a repeating manner. In fact, according bring thinking about them up to date. to Moretti, different novel types—picaresque, gothic, domestic et cetera—appear to have ex- Four Times Thirty perienced roughly 30-year periods of rising and In the 1840s, at the height of the British expan- falling popularity, indicating that there are forces sion of the railroads known as “railway mania,” at work that transcend any given literary move- the state of the art in iron-bridge building was ment or fashion. Such forces also appear to be at to use cast iron. Since casting beams longer than work in structural engineering, where cycles of a about 30 feet was not common, longer distances similar duration and patterns of success and fail- were spanned by attaching beams end to end ure apply across a wide variety of bridge types. and employing wrought-iron trussing as a sort It is now more than 10 years since I described of belts-and-suspenders safety measure. The lon- in these pages (“Predicting Disaster,” March– gest bridge of this type (spanning almost 100 April 1993) the study of historic bridges carried feet) was constructed across the River Dee on the out by Paul Sibly in his 1977 doctoral thesis, The Chester & Holyhead Railway and was passed Prediction of Structural Failures. Sibly’s research in 1846 by the Inspector General to the Board of was “mainly concerned with the prediction of Trade. According to James Sutherland, “he was one class of structural failure, namely that due to opposed to this form of iron trussing and virtu- the extrapolation of existing design or construc- ally admitted that he had approved the bridge tion procedures to fit new situations.” He consid- because of the number of apparently successful ered four types of large metal bridges that ended ones already built.” in dramatic collapses. In a subsequent article he The Dee Bridge collapsed suddenly in 1847, and his advisor, A. C. Walker, pointed out that just after a load of gravel had been added to the failures followed a temporal pattern: They its roadbed, and this “first serious structural had occurred at roughly 30-year intervals. collapse on the railway network” was the sub- The cycle that Sibly described continued be- ject of an inquest. This in turn prompted the yond the cases that he initially studied, which subsequent appointment of a royal commission led me to speculate that a dramatic bridge fail- charged with looking into the use of iron in rail- ure might be expected to take place sometime way bridges. The bridge had been designed and around the year 2000. In the early 1990s, the constructed under the supervision of the emi- nent engineer Robert Stephenson, who was ex- onerated. Landmark failures like that of the Dee Henry Petroski is the Aleksandar S. Vesic Professor of Civil Engineering and a professor of history at Duke University. His seem never to be fully put to rest, however, and latest book, Pushing the Limits: New Adventures in Engi- the design of the bridge is looked at afresh in a neering (Knopf), is a collection of essays that appeared first in recent biography of Stephenson. According to this magazine. Address: Duke University, Box 90287, Durham, James Sutherland, “It would be wrong to blame NC 27708-0287. a single individual for the misconception over © 2004 Sigma Xi, The Scientific Research Society. Reproduction 500 American Scientist, Volume 92 with permission only. Contact [email protected]. Paul Raftery/Alamy Figure 1. Since the collapse of the Dee Bridge in 1847, bridge failures appear to recur at approximately 30-year intervals. This suggests that a failure should have been due around the year 2000, 30 years after the Milford Haven Bridge collapsed during construction. Although no bridge fell at the changing of the millennium, several footbridges exhib- ited severe problems, including this one, the Passerelle Solferino in Paris. the behavior of these girders. This was a case porary comments of the president of the court of of group myopia suffered by a large tranche of inquiry, who described the structure as “badly the most distinguished engineers of the day.” designed, badly constructed, and badly main- Such collective nearsightedness is responsible for tained.” Truss bridges did not cease to be built most colossal failures. after the Tay collapse, but they were constructed The Tay Bridge, completed in 1878, was built with more attention to detail and to resist much to carry the North British Railway over the wide stronger winds. The Tay itself was rebuilt as a estuary at Dundee, Scotland. Although it was truss bridge, and its more substantial appearance overall the longest bridge in the world when is testament to the natural response to failure. completed, none of the Tay’s many individual When the Tay Bridge failed, Bouch was in the spans exceeded contemporary experience. In process of building another bridge on the North late December 1879, during a storm, the bridge’s British line, one across the Firth of Forth near highest and longest girders, along with a train Edinburgh, but that commission was understand- with 75 passengers, were blown into the river. ably taken from him. Instead of the flimsy suspen- A court of inquiry found that, “The fall of the sion bridge he had planned, the new engineers bridge was occasioned by the insufficiency of John Fowler and Benjamin Baker designed a mas- the cross bracing and its fastenings to sustain the sive cantilever structure, then a relatively unfamil- force of the gale.” The engineer, Thomas Bouch, iar bridge type. The Forth Bridge was intended had assumed that the maximum wind force to restore confidence in the railroad not only by would be barely one-fifth what it might have its striking difference from anything designed by been that night. Bouch but also in its own right by looking like it The Tay disaster continues to be the focus of could take anything nature could throw at it. attention. In the most recent study, researchers The immediate success of the Forth Bridge scanned archival photos of the accident scene prompted engineers around the world to emulate and used modern digital photography tools to it, and for many of them the cantilever became enhance the photos and inspect the record of the bridge type of choice to carry heavy railroad remains. What they found was evidence of nu- trains across wide rivers, gorges and valleys. One merous broken tie-bar lugs, strongly suggest- example under construction in the early 20th cen- ing that the bridge structure was loosely held tury was the bridge to carry the Canadian trans- together and racked excessively under crossing continental line over the St. Lawrence River near trains, which caused the lugs to fail under the Quebec. With an 1,800-foot main span, the Que- repeated action, a phenomenon known as metal bec Bridge was to be the longest cantilever in the fatigue. The analysis was confirmed by contem- world, bettering the Forth Bridge by almost 100 © 2004 Sigma Xi, The Scientific Research Society. Reproduction www.americanscientist.org with permission only. Contact [email protected]. 2004 November–December 501 feet. The bridge was under construction in 1907 He speculated on how such accidents might be when it collapsed with 86 workers present. Only predicted, while admitting that the amount of 11 survived. data collection and analysis necessary to do so A royal commission found that the bridge’s would involve considerable effort. His conclud- design and construction were carried out in a ing remarks include a brief discussion of the 1970 disorganized manner, and with an inadequate failure of the Milford Haven Bridge, in Wales, a respect for the magnitude of the structure. The steel box-girder structure that collapsed while principal consulting engineer, Theodore Cooper, still under construction. A similar accident that who was de facto also the chief engineer, oversaw same year befell the West Gate Bridge under the project from a distance, not once visiting the construction in Melbourne, Australia.

Load more