Design Complacency: The Dee Railway Bridge Collapse Chester, England - May 24th, 1847 Hector Velez, MS Architectural Engineering, The Pennsylvania State University, Fall 2013
Introduction In the United Kingdom, the 19th century was characterized by rampant industrialization as the civil engineering field matured. England experienced a rapid infrastructural expansion in the 1840's that would eventually be referred to as the "Railway Mania" era. Such projects often relied on trial-by-fire testing (Petroski, 1992): the expected structural performance of a new bridge was relatively unknown until its structure was complete. Bridges that did not immediately deflect or buckle were deemed successes, while those that showed obvious problems were failures. Consequently, it was common for the design details of a single successful bridge to be emulated on other projects. The Dee Railway Bridge, part of the Chester and Holyhead Railway, was constructed in 1846 as part of the country's expanding railway system. Less than a year later, it collapsed under the weight of a passing locomotive, resulting in five fatalities and considerable damages. The Tay Bridge in Scotland, built with cast iron columns and wrought iron cross-braces, collapsed in 1879 and was another victim of the Railway Mania period (Pinsdorf, 1997). Although various sources cite different details as the exact cause of the Dee's collapse, most of them agree on the following: 1. The recent addition of crushed-stone ballast, meant to provide fire-resistance to the bridge's wooden members, added significant weight that overloaded the structure. 2. The combination of cast iron girders with wrought iron tie rods did not perform as intended. 3. The engineer's use of the above-mentioned combination had no precedence, since it had only been used successfully on bridges with much smaller spans 4. An aesthetic detail on the girders created a weak spot. 5. Metal fatigue caused the structural members to degrade over time.
Figure 1: Original newspaper etching of the collapse (used with permission from Wikimedia Commons)
Keywords
Dee Railway Bridge, bridge collapse, Industrial Revolution, 19th century, cast iron, wrought iron, design precedence, fatal, United Kingdom, England, Scotland, Railway Mania, railroad, 1847, Robert Stephenson, deflection, negligence, trial, accident, trussed cast iron girders
Events Preceding the Collapse
Robert Stephenson was at the forefront of England's massive infrastructural expansion in the 1800's (Gagg & Lewis, 2004). He was responsible for the successful design of several railway bridges, specifically using cast iron girders with wrought iron trusses. Cast iron refers to iron that is smelted and cast directly into molds, which produce the desired shape. Iron casting was first successfully done, using coke as a heat source, in the early 18th century in England and was a well-developed process by the mid-1700's (Kemp 1993). Wrought iron, conversely, refers to shapes that are worked and "wrought" out at low temperatures. Stephenson and his contemporaries were well aware of cast iron's poor tensile strength, as well as the cost implications of using large amounts of wrought iron (Beckett, 1984). This, along with other factors, led to the decision to design a composite bridge with cast iron girders and wrought iron tension bars. An early, major design change might have foreshadowed the collapse. Stephenson originally designed the bridge with five arched spans between heavy masonry piers. However, the consensus was that the riverbed could not support such heavy masonry, so the design was changed to a two-pier system supporting three lengthy spans (Gagg & Lewis, 2004). These spans, each one roughly 98 feet in length, necessitated the use of cast iron girders. Such girders were used throughout the Industrial Revolution, both for buildings and bridges, and their span limits were typically dealt with through wrought iron trusses (Gagg & Lewis, 2011). The Dee Bridge was completed in September of 1846, and passed inspection the next month (Beckett, 1984). The inspection coincided with a ceremonious event in which three linked trains successively crossed the bridge in a show of strength. These types of inaugural ceremonies were common ways of proving structural stability, and at times were the only real-world tests conducted. The combined weight of the three trains far exceeded the daily anticipated live loads, assuring the public (and the designer) of the bridge's integrity. It was opened to local traffic in November of 1846. An early warning sign was virtually ignored. Before the bridge was completed, a section of one of the cast iron girders was replaced due to noticeable fractures (Walker & Simmons, 1847). Stephenson assumed that the cracks were due to manufacturing defects during the casting process. He ordered the bridge to be temporarily shored while the section was replaced with a new casting. Thus, the bridge had early indications of failure over half a year before the collapse. Yet another warning sign was evident since the conclusion of construction: the bridge was known to excessively vibrate and deflect under live loads. Painters working underneath the bridge personally measured deflections of 3-5" under passenger trains (Gagg & Lewis, 2004). Apparently, Stephenson and his staff were not aware of the extent of the deflection, but the general vibration issues were common knowledge. The original deflection tests performed at the foundry, which only tested static loads and did not account for oak beams, stone ballast, or eccentric loading, showed that the girders deflected up to 2.5" on center (Gagg & Lewis, 2004).
Bridge Design: Trussed Cast Iron Girders
Figure 2: Railway section created by Hector Velez, adapted from Beckett, 1984 and Walker & Simmons, 1847. The Dee Bridge had 2 railways running parallel: one heading east, and one heading west.
Figure 3: Girder and tie rod detail (image adopted from public domain document: Walker & Simmons, 1847). Each girder was comprised of 3 castings: this image shows part of the middle casting (left side of the image) and the entire right casting (right side of the image).
As previously stated, the designed two-pier system meant that the bridge would have three 98-foot spans. Stephenson chose to use cast iron girders due to his success with previous bridge designs and the material's relatively low cost. However, cast iron was too brittle to cast in sections longer than 35 feet. Thus, each 98-foot span of the bridge would consist of three sections, or "castings," bolted together to form a composite structure (Petroski, 2012). Cast iron's poor tensile strength was well-known at the time; wrought iron tie rods would be used to strengthen the connections between the girder sections, and simultaneously add tensile strength to the bridge. This section of the Chester and Holyhead Railway contained two parallel railways heading east and west. Each railway two girders; each 98-ft girder span consisted of three cast iron sections. The sections were joined by semi-circular wrought iron connectors, which were then connected to each other diagonally and transversely via wrought iron tension bars (Petroski, 2012). The resulting trusses were intended to relieve the tensile forces in the lower halves of the girders, placing the upper halves in compression and taking advantage of the cast iron's compressive strength (Petroski, 2012). Each three-part girder had a 5'-5" bearing on each end, and the completed bridge had a total of 12 cast iron girders: one trussed girder per span, three clear spans per length of the bridge, two bridge lengths for each of the two railway lines (Beckett, 1984). Oak joists, 8x10" in cross section in certain areas and 10x10" in others, were loosely laid perpendicular to the girders and bearing directly on the lower inner flanges (Gagg & Lewis, 2004). 4" planks were laid longitudinally across the oak joists in order to create a flat surface for the rail lines and stone ballast. To reiterate, all components were supported by the lower inner flange of the girders on either side.
Collapse
Figure 4: Original newspaper etching of the collapse (public domain document; used with permission from Wikimedia Commons)
Six separate trains crossed the Dee Bridge on May 24th, 1847 without any issues (Walker & Simmons, 1847). Later that afternoon, Stephenson had five inches of additional stone ballast laid on the railways to provide better fire protection for the wooden railroad ties. This was done in response to recent bridge fires caused by cinders and debris from coal-powered trains. Multiple sources agree that this added roughly 18 tons of dead loads to the weight of the bridge (Walker & Simmons, 1847; Gagg & Lewis, 2004). At approximately 6:15PM, a locomotive pulling a tender (fuel car), four passenger carriages, a luggage van, and 25 passengers crossed the bridge. As it reached the third and final span, the driver noticed that the bridge was sinking underneath him and quickly opened the engine to "full steam" (Walker & Simmons, 1847). The locomotive's momentum drove it past the edge of the bridge and dragged the tender behind it, damaging a stone parapet, and depositing the tender next to the river. All of the remaining cars fell into the Dee River as two of the three castings of the third span gave way below (Walker & Simmons, 1847). The driver in the locomotive, the only uninjured survivor, proceeded to the next train station where he sounded the emergency alarm (Gagg &Lewis, 2004). He then drove back across the Dee Bridge on the opposite track, which was still intact, to prevent more trains from crossing. In total, five passengers were killed and 18 were seriously injured (Beckett, 1984).
Figure 5: Elevations of the girder fractures (image adopted from public domain document: Walker & Simmons, 1847)
Ensuing Events & Investigations
Within days of the collapse, the British government dismissed the inspector who originally approved the bridge's completion (Gagg & Lewis, 2004). Captain John Simmons of the Royal Engineers, an officer with the British Railway Department, was sent to the site two days after the collapse to conduct an initial investigation. His report concluded that the stone piers and girder-to-girder connections were in sound condition, and that overall the bridge did not seem to have construction defects (Walker & Simmons, 1847). He estimated that the train was traveling at a high speed (around 30 mph) since it was behind schedule, but that the train itself had not caused the collapse by abruptly increasing its speed as some had believed. After studying the intact railway, he concluded that the wrought iron tie bars were loose (and thus were not providing tensile reinforcement), passing trains caused significant deflection, and the bridge was subjected to serious vibration on a daily basis. The Commissioners of Railways called on Simmons and James Walker, a civil engineer, to write a formal report detailing the events of the collapse and its primary cause. First, they summarized the characteristics and dimensions of the bridge, and commented on its overall rigidity and workmanship. They also noted that on several occasions, the bridge had been crossed by trains that were much heavier than the one that caused the collapse (Walker & Simmons, 1847). On June 4th, a formal inquest was begun to determine the exact cause of the collapse and fatalities. The trial was also concerned with whether or not Stephenson, the designer and structural engineer, was liable for the five fatalities. Walker and Simmons' report (1847) played a key role in the jury's decision. Stephenson's primary defense was that the tender car had derailed, struck a parapet wall above one of the load-bearing piers, and caused the bridge to collapse. However, eyewitness accounts and Walker's report agreed that the derailment was a product of the collapse, not the cause. Based on witness testimonies and the report, the jury reached several conclusions. First, the deaths were accidental and Stephenson was not negligent or liable for manslaughter (Gagg & Lewis, 2004). Secondly, the collapse was was an accident caused by the insufficient strength of the girders. Lastly, the jury recommended that a Royal Commission be established to determine whether cast iron was a suitable material for bridge construction. The Commission's final report, published two years after the Dee Bridge collapse, stated that cast iron members were generally strong under stationary loads, but tended to fail after less than 900 oscillation cycles under live loads (Gagg & Lewis, 2004).
Causes of the Failure
The primary cause of the failure was the ineffective composite action of the cast iron girders and wrought iron tension bars (Walker & Simmons, 1847). Multiple sources agree that this design flaw caused the failed girder to buckle due to torsional instability (Beckett, 1984; Evan & Manion, 2002). As Walker and Simmons (1847) concluded, when the bridge deflected under live or static loads, the wrought iron tension bars loosened and were thus rendered ineffective. While Stephenson had the right idea of using wrought iron for tensile reinforcement, his design did not allow those members to perform as intended. If the wrought and cast iron members had worked in tandem as designed, the bridge most likely would have been strong enough to avoid collapse (Walker & Simmons, 1847). Even when the tie rods were tensioned under passing trains, this actually increased the compressive forces in the top flange of girder, which was significantly smaller than the bottom flange (Beckett, 1984). The cast iron girders alone had a safety factor of about 1.5; the wrought iron bars were meant to raise that factor to an acceptable level, although Stephenson could not precisely calculate its effect (Beckett, 1984). Conversely, modern analysis of another iron bridge built in 1859 showed that while it had significant rusting, it still had an adequate safety factor of about 4 (Gordon and Knopf, 2005). Robert Stephenson had no real precedence for his design. The equation he used to determine failure loads was based on tests performed on centrally-loaded girders with much shorter spans than the Dee (Beckett, 1984). Although he successfully designed cast and wrought iron bridges for over a decade, none of them were as long or complex as the Dee Bridge. Additionally, the Dee's 98-ft clear spans were nearly 10 feet longer than the longest trussed girder bridges in existence (Beckett, 1984). The application of previously-established norms to bigger and more complex designs, defined by Petroski as "success syndrome," was a common theme of several 19th century failures (Evan & Manion, 2002). Metal fatigue, defined as "when a component fails well below its rated strength owing to crack formation and growth by repeated loading cycles," also played a major role in the collapse. Basically, structural metal tends to crystallize and weaken due to cyclical loading and unloading. 19th century civil engineers were believed to have some understanding of this brittle characteristic (Petroski, 2012). Four years before the Dee Bridge collapsed, a Scottish engineer named William Rankine wrote a paper that essentially defined the metal fatigue phenomenon, and suggested that it might have been the cause of many sudden and unexpected failures (Petroski, 2012). Metal fatigue was especially detrimental in this case because the Dee Bridge was known vibrate and deflect excessively. Still, Petroski (2012) argued that it was difficult to blame Stephenson for this material flaw because, while the engineering community had a basic understanding of metal fatigue, they did not fully understand how it started and progressed.
Figure 6: Polariscope image of girder with cavetto molding detail (reproduced with the author's permission)
The fourth major cause of the collapse is related to an aesthetic detail. Cavetto moldings, popular in contemporary carpentry designs, were added to the corners of the cast iron girders where the web connected to the flanges. Gagg & Lewis (2004) created a scaled girder section out of a polymer material, bent one of the lower flanges to replicate the bridge's asymmetric loading, and studied the section under a polariscope. Their experiment showed that the loads would have been concentrated on the sharp corners of the cavetto moldings, which acted as stress raisers. Since metal fatigue can aggravate small cracks and imperfections in the metal, it is believed that the girder originally fractured near the corners of the moldings (Gagg & Lewis, 2004). Wide flange beams and other modern structural members purposely avoid such stress-raising corners. The addition of stone ballast just hours before the collapse was the last straw, metaphorically speaking. This action was preceded by the crossing of six trains, none of which caused the bridge to collapse. Gagg and Lewis (2004) argued that by the date of the collapse, the bridge was already significantly weakened. The trains that successfully crossed, along with the 18 tons of additional ballast, must have caused an existing crack to "grow to a critical length, and it grew catastrophically when the 6:15 pm train entered the span" (Gagg & Lewis, 2004). While the previous four factors were significantly involved and would have led to an inevitable collapse, the added weight of the stone expedited the process.
Prevention
Not every facet of this collapse would have been easily preventable during the mid-19th century. In terms of materials, cast and wrought iron were both well-developed and extensively used by the time the Dee Bridge was constructed. The minimal use of wrought iron compared to cast iron was a matter of cost effectiveness, which is still an important component of design and engineering. Due to the state of technology, Stephenson had no means of testing the design's reaction to compressive and tensile forces, which is why the inaugural ceremonies were so crucial for 19th century designers. The overextension of previous designs to new applications was definitely preventable. Stephenson's design was unprecedented on several levels; it was inexcusable to assume, without any empirical data, that the trussed girder design could be expanded or modified without consequence. The metal fatigue problem represents another nebulous issue. On one hand, the engineering community was somewhat aware of metal's progressive deterioration. At the same time, a renowned engineer like Stephenson should have questioned the bridge's noticeable vibration and deflection. The cavetto detail was popular in 19th-century wooden beam carpentry (Gagg & Lewis, 2004). Considering the limited nature of empirical testing at the time, it would have been difficult to condemn such a seemingly innocuous detail. However, the cast iron girders were originally tested at the foundry with the flanges loaded symmetrically. If the same load tests had been performed on one of the lower flanges, in accordance with the girders' intended use, the weakness of the cavetto moldings might have been more apparent. Stephenson's addition of such a large amount of track ballast (5" deep and weighing 18 tons) suggests that he did not consider the bridge to be in any immediate danger. Thus, the perpetual warning signs were deemed inconsequential at most. His reaction to the recent railway fires was understandable, but to simply add more weight to a life-threatening structure without some form of testing or reassurance was regrettable. Modern professional engineers are bound to protect the safety of the public before all else. Periodic inspections were not mentioned in the literature, either.
Lessons Learned
Perhaps the biggest lesson to learn from this collapse is the importance of thorough testing. Evan and Manion (2002), in discussing the Dee Bridge, Hyatt Regency walkway, and similar collapses, argued that there will always be limits to the tests that can be performed. Petroski (1994) previously made a similar statement: "virtually all design is conducted in a state of relative ignorance of the full behavior of the system being designed." This was especially true during the Industrial Revolution. Due to a lack of sophisticated testing equipment and measures, the common practice was to ensure a structure's integrity by purposely overloading it in a parade-like event. The Dee Bridge's inauguration involved roughly 90 tons, far exceeding the expected daily live loads (Walker & Simmons, 1847). Although the structure survived this event, and was thus deemed structurally sound, the need for more rigorous testing presented itself half a year later. While a lack of testing is a major concern, the suitability of the tests that are performed is equally important. Deflection tests performed on the girders at the foundry involved evenly-distributed loads that were much less than the expected daily live loads, which Stephenson acknowledged (Walker & Simmons, 1847). If performed tests do not reflect the anticipated loading conditions of the completed structure, then they are virtually useless. Finally, the relevance of prevailing code requirements is paramount. William Fairbairn, a Scottish engineer and close associate of Robert Stephenson, was one of several noted engineers who testified that Stephenson was not negligent. In 1864, several years after the collapse of the Dee Bridge and other notable failures, Fairbairn published a paper that sought to definitively describe wrought iron's performance under continued vibrations and load changes (Fairbairn, 1864). His findings reiterated the issue of metal fatigue, and concluded that one of the strength requirements established by the Board of Trade for iron bridges (a maximum of 5 tons per square inch) was vague and did not distinguish between tension and compression. He further argued that such requirements were unenforceable until the means of construction, the methods of measuring strength, or both were standardized. Lastly, Fairbairn asserted that the current standards written by the Lords Commissioners for Trade were based on limited data and heavily relied on "good material and workmanship" for them to be applicable (Fairbairn, 1864).
Industry Effects
Figure 7: Stephenson's High Level Bridge (public domain image; used with permission from Wikimedia Commons)
Figure 8: Stephenson's Britannia Bridge (public domain image, used with permission from Wikimedia Commons)
The inquest of the Dee Bridge collapse led to the establishment of a Royal Commission that studied iron's feasibility as a bridge material (Beckett, 1984). The jury's decision, along with the findings of the Royal Commission's report two years later, was that the Dee Bridge and others built in a similar manner were unsafe. Stephenson consequently changed his focus from trussed iron girders to tubular shapes. The High Level Bridge, completed over the River Tyne in 1849, was a wrought iron bowstring girder bridge, meaning that its arches were tied between each masonry pier. Stephenson purposely avoided extending the trussed girder design, as he had done with the Dee Bridge, and his new design properly used cast iron for compressive loads and wrought iron for tensile roads (Beckett, 1984). Similarly, the Britannia Bridge was completed in 1850 and used a tubular shape to carry railroad traffic. This bridge was under construction while the Dee Bridge investigation was still underway (Petroski, 1993). I.K. Brunel, an English civil engineer and close associate of Robert Stephenson, adapted the wrought iron tubular design for Royal Albert Bridge in 1859 (Beckett, 1984). During the Dee Bridge investigation, Brunel testified that cast iron bridges were inherently faulty, but said that the Royal Commission was impeding the industry's progress by enacting new regulations. Overall, Stephenson and Brunel's work after the Dee Bridge collapse represented a transition point for the civil engineering community: larger bridges would continue to be designed out of necessity, but engineers were increasingly experimental and tried to avoid failed design principles.
Similar Failures
Henry Petroski, a Civil Engineering and History professor at Duke University, classified the Dee Bridge failure as the first of a chain of collapses (Pinsdorf, 1997; Subramanian, 2008). Petroski asserted that, starting with the Dee in 1847, there have been major bridge collapses roughly every 30 years. Thus far, they have included the Tay Bridge, St. Lawrence Bridge near Quebec City, Tacoma Narrows Bridge, "two box girder bridges in Wales and Austria," and finally the I-35W Bridge in Minneapolis (Pinsdorf, 1997). Of these failures, the Tay Bridge collapse in England, 1879, is the most similar to the Dee Bridge collapse. It was built towards the end of England's Railway Mania, and collapsed due to hurricane-like winds that exceeded the design loads (Pinsdorf, 1997). While the collapse of the Tay Bridge caused 76 fatalities, Pinsdorf (1997) argued that it was a critical turning point for Victorian-era overconfidence and was part of the transition from iron to steel.
Another notable failure occurred in England roughly one year before the Dee Bridge's collapse. The Grays Mill building in Manchester utilized a similar combination of cast iron girders and wrought iron tensions bars. As trusses, the wrought iron members did not reinforce the girders at all; a center girder fractured and caused the entire building to collapse (Gagg & Lewis, 2011).
The Dee Bridge collapsed due to several problems that had cumulative effects. In a broad sense, it was a victim of an era that emphasized progress and expansion before research and development. More specifically, the engineer did not fully understand the interplay of tension and compression within his design, and grossly overestimated its strength as a result. Although Robert Stephenson was not found negligent nor responsible for the five fatalities, his fate would probably have been much worse in today's courts. Engineering negligence is typically determined by using a "reasonable man standard" in which the engineer has a duty to perform his tasks and make judgments based on the accepted standards and practices of his time. By this definition, Stephenson was not negligent because the industry as a whole was relatively ignorant. On the other hand, the blatant disregard of the various warning signs has to account for something. Additionally, the determination that Stephenson's design and execution were flawed did not arise decades after the accident, but within weeks. This alone implies that his judgment might not have been so aligned with the state of the industry, after all.
Annotated Bibliography
Beckett, D. (1984). Stephenson's Britain. David & Charles, North Pomfret, VT.
Derrick Beckett chronicles the professional careers of father-and-son engineers George and Robert Stephenson. Robert's design for the Dee Bridge are described in detail, along with the underlying circumstances of the collapse.
Evan, W.M., and Manion, M. (2002). Minding the Machines: Preventing Technological Disasters. Prentice Hall, Upper Saddle River, NJ.
This comprehensive study of technological disasters discusses the Dee Bridge collapse as part of a group of failures related to improper testing.
Fairbairn, W. (1864). "Experiments to Determine the Effect of Impact, Vibratory Action, and Long-Continued Changes of Load on Wrought-Iron Girders." Philosophical Transactions of the Royal Society of London, 154, 311-325.
Written less than two decades after the fall of the Dee Bridge, this paper exemplifies the engineering industry's struggle to understand materials performance during different loading conditions.
Gagg, C.R., and Lewis, P.R. (2004). "Aesthetics versus function: the fall of the Dee bridge, 1847." Interdisciplinary Science Reviews, 29, 177-191.
This article summarizes the circumstances of the bridge collapse, and concludes that the bridge most likely failed due a strictly aesthetic feature of the cast-iron girders.
Gagg, C.R., and Lewis, P.R. (2011). "The rise and fall of cast iron in Victorian structures- A case study review." Engineering Failure Analysis, 18, 1963-1980.
The collapse of the Tay and Dee bridges are presented, among other case studies, to show the weaknesses of cast iron that led to its declining use for large structures.
Gordon, R., and Knopf, R. (2005). "Evaluation of Wrought Iron for Continued Service in Historic Bridges. Journal of Materials in Civil Engineering, 17(4), 393-399.
The authors of this paper used modern analysis to show that the Aldrich Bridge, an iron bridge also built in the mid-1800's, was still in decent condition.
Kemp, E.L. (1993). "The Introduction of Cast and Wrought Iron in Bridge Building." The Journal of the Society for Industrial Archeology [sic], 19, 5-16.
Emory L. Kemp provides an insightful view into the development of cast iron and wrought iron, which play a pivotal role in the Dee Bridge collapse. Topics range from the 18th-century development of these materials to more advanced uses in the 19th century.
Petroski, H. (1992). "Making Sure." American Scientist, 80, 121-124.
This paper describes the gradual paradigm shift from heavy masonry bridges supporting lighter loads (carriages and pedestrians) to lightweight bridges supporting heavier loads (trains, vehicular traffic). Civil engineering in the 18th and 19th centuries heavily relied on trial-by-fire testing, which ultimately led to the Dee Bridge's demise.
Petroski, H. (1993). "Predicting Disaster." American Scientist, 81, 110-113.
This article discusses the engineering features and advancements that the Metrodome had employed at the time of its construction. Included are parallels of the Metrodome and previous air-supported fabric roofs.
Petroski, H. (1994). "Success Syndrome: The Collapse of the Dee Bridge." Civil Engineering, 64, 52-55.
Petroski argues that most failure studies are so specific that they only appeal to the few experts and specialists that have experienced such collapses. The Dee Bridge collapse is an example of a historic, well-covered case study that has not been over-analyzed or politicized, as is the case with many modern failures.
Petroski, H. (2004). "Past and Future Failures." American Scientist, 92, 500-504.
In describing the cyclical nature of major bridge collapses, Petroski shows the tendency of the industry to move from one bridge type to another as major failures decreased the use of certain designs and led to the development of others.
Petroski, H. (2012). To Forgive Design: Understanding Failure. The Belknap Press of Harvard University Press, Cambridge, MA.
This is Petroski's second book about the fundamental design principles that often lead to failures. He briefly discusses the Dee Bridge collapse, along with the investigation that was conducted involving the bridge's design engineer, Robert Stephenson.
Pinsdorf, M.K. (1997). "Engineering Dreams into Disaster: History of the Tay Bridge." Business and Economic History, 26, 491-504.
This paper describes the events surrounding the Tay Bridge, which collapsed in England roughly 30 years after the Dee Bridge and was part of the broader movement towards lighter truss bridges.
Subramanian, N. (2008). "I-35W Mississippi river bridge failure- Is it a wake up call?" The Indian Concrete Journal, February 2008, 29-38.
A civil engineer's opinions regarding the industry-wide effects of this major bridge collapse in Minneapolis. Also discusses the cyclical nature of major bridge failures including the Dee, Tacoma Narrows, and I-35W bridges.
Walker, J. and Simmons, J.L.A. (1847). "Report to the Commissioners of Railways, by Mr. Walker Captain Simmons, R.E., on the Fatal Accident on the 24th day of May 1847, by the Falling of the Bridge over the River Dee, on the Chester and Holyhead Railway; together with any Minutes of the Commissioners thereupon."
This technical report, written by a civil engineer and a railway inspector in response to the Dee Bridge collapse, summarizes the most likely causes of the bridge's failure. The authors' findings eventually led to a thorough investigation of the contemporary uses of cast iron in bridges and railways.
Wood, J.G.M. (2004). "Failures from Hazards, a Short Review." International Association for Bridge and Structural Engineering (IABSE), July 2004, 1-4.
A cautionary review of the construction industry's reaction to past failures. This paper argues that factors such as robustness and deterioration have to be carefully considered, even in extreme failure cases such as terrorist attacks.
Chester, England - May 24th, 1847
Hector Velez, MS Architectural Engineering, The Pennsylvania State University, Fall 2013
Table of Contents
IntroductionIn the United Kingdom, the 19th century was characterized by rampant industrialization as the civil engineering field matured. England experienced a rapid infrastructural expansion in the 1840's that would eventually be referred to as the "Railway Mania" era. Such projects often relied on trial-by-fire testing (Petroski, 1992): the expected structural performance of a new bridge was relatively unknown until its structure was complete. Bridges that did not immediately deflect or buckle were deemed successes, while those that showed obvious problems were failures. Consequently, it was common for the design details of a single successful bridge to be emulated on other projects.
The Dee Railway Bridge, part of the Chester and Holyhead Railway, was constructed in 1846 as part of the country's expanding railway system. Less than a year later, it collapsed under the weight of a passing locomotive, resulting in five fatalities and considerable damages. The Tay Bridge in Scotland, built with cast iron columns and wrought iron cross-braces, collapsed in 1879 and was another victim of the Railway Mania period (Pinsdorf, 1997). Although various sources cite different details as the exact cause of the Dee's collapse, most of them agree on the following:
1. The recent addition of crushed-stone ballast, meant to provide fire-resistance to the bridge's wooden members, added significant weight that overloaded the structure.
2. The combination of cast iron girders with wrought iron tie rods did not perform as intended.
3. The engineer's use of the above-mentioned combination had no precedence, since it had only been used successfully on bridges with much smaller spans
4. An aesthetic detail on the girders created a weak spot.
5. Metal fatigue caused the structural members to degrade over time.
Keywords
Events Preceding the Collapse
Robert Stephenson was at the forefront of England's massive infrastructural expansion in the 1800's (Gagg & Lewis, 2004). He was responsible for the successful design of several railway bridges, specifically using cast iron girders with wrought iron trusses. Cast iron refers to iron that is smelted and cast directly into molds, which produce the desired shape. Iron casting was first successfully done, using coke as a heat source, in the early 18th century in England and was a well-developed process by the mid-1700's (Kemp 1993). Wrought iron, conversely, refers to shapes that are worked and "wrought" out at low temperatures. Stephenson and his contemporaries were well aware of cast iron's poor tensile strength, as well as the cost implications of using large amounts of wrought iron (Beckett, 1984). This, along with other factors, led to the decision to design a composite bridge with cast iron girders and wrought iron tension bars.
An early, major design change might have foreshadowed the collapse. Stephenson originally designed the bridge with five arched spans between heavy masonry piers. However, the consensus was that the riverbed could not support such heavy masonry, so the design was changed to a two-pier system supporting three lengthy spans (Gagg & Lewis, 2004). These spans, each one roughly 98 feet in length, necessitated the use of cast iron girders. Such girders were used throughout the Industrial Revolution, both for buildings and bridges, and their span limits were typically dealt with through wrought iron trusses (Gagg & Lewis, 2011).
The Dee Bridge was completed in September of 1846, and passed inspection the next month (Beckett, 1984). The inspection coincided with a ceremonious event in which three linked trains successively crossed the bridge in a show of strength. These types of inaugural ceremonies were common ways of proving structural stability, and at times were the only real-world tests conducted. The combined weight of the three trains far exceeded the daily anticipated live loads, assuring the public (and the designer) of the bridge's integrity. It was opened to local traffic in November of 1846.
An early warning sign was virtually ignored. Before the bridge was completed, a section of one of the cast iron girders was replaced due to noticeable fractures (Walker & Simmons, 1847). Stephenson assumed that the cracks were due to manufacturing defects during the casting process. He ordered the bridge to be temporarily shored while the section was replaced with a new casting. Thus, the bridge had early indications of failure over half a year before the collapse.
Yet another warning sign was evident since the conclusion of construction: the bridge was known to excessively vibrate and deflect under live loads. Painters working underneath the bridge personally measured deflections of 3-5" under passenger trains (Gagg & Lewis, 2004). Apparently, Stephenson and his staff were not aware of the extent of the deflection, but the general vibration issues were common knowledge. The original deflection tests performed at the foundry, which only tested static loads and did not account for oak beams, stone ballast, or eccentric loading, showed that the girders deflected up to 2.5" on center (Gagg & Lewis, 2004).
Bridge Design: Trussed Cast Iron Girders
As previously stated, the designed two-pier system meant that the bridge would have three 98-foot spans. Stephenson chose to use cast iron girders due to his success with previous bridge designs and the material's relatively low cost. However, cast iron was too brittle to cast in sections longer than 35 feet. Thus, each 98-foot span of the bridge would consist of three sections, or "castings," bolted together to form a composite structure (Petroski, 2012). Cast iron's poor tensile strength was well-known at the time; wrought iron tie rods would be used to strengthen the connections between the girder sections, and simultaneously add tensile strength to the bridge.
This section of the Chester and Holyhead Railway contained two parallel railways heading east and west. Each railway two girders; each 98-ft girder span consisted of three cast iron sections. The sections were joined by semi-circular wrought iron connectors, which were then connected to each other diagonally and transversely via wrought iron tension bars (Petroski, 2012). The resulting trusses were intended to relieve the tensile forces in the lower halves of the girders, placing the upper halves in compression and taking advantage of the cast iron's compressive strength (Petroski, 2012). Each three-part girder had a 5'-5" bearing on each end, and the completed bridge had a total of 12 cast iron girders: one trussed girder per span, three clear spans per length of the bridge, two bridge lengths for each of the two railway lines (Beckett, 1984).
Oak joists, 8x10" in cross section in certain areas and 10x10" in others, were loosely laid perpendicular to the girders and bearing directly on the lower inner flanges (Gagg & Lewis, 2004). 4" planks were laid longitudinally across the oak joists in order to create a flat surface for the rail lines and stone ballast. To reiterate, all components were supported by the lower inner flange of the girders on either side.
Collapse
Six separate trains crossed the Dee Bridge on May 24th, 1847 without any issues (Walker & Simmons, 1847). Later that afternoon, Stephenson had five inches of additional stone ballast laid on the railways to provide better fire protection for the wooden railroad ties. This was done in response to recent bridge fires caused by cinders and debris from coal-powered trains. Multiple sources agree that this added roughly 18 tons of dead loads to the weight of the bridge (Walker & Simmons, 1847; Gagg & Lewis, 2004). At approximately 6:15PM, a locomotive pulling a tender (fuel car), four passenger carriages, a luggage van, and 25 passengers crossed the bridge. As it reached the third and final span, the driver noticed that the bridge was sinking underneath him and quickly opened the engine to "full steam" (Walker & Simmons, 1847). The locomotive's momentum drove it past the edge of the bridge and dragged the tender behind it, damaging a stone parapet, and depositing the tender next to the river. All of the remaining cars fell into the Dee River as two of the three castings of the third span gave way below (Walker & Simmons, 1847).
The driver in the locomotive, the only uninjured survivor, proceeded to the next train station where he sounded the emergency alarm (Gagg &Lewis, 2004). He then drove back across the Dee Bridge on the opposite track, which was still intact, to prevent more trains from crossing. In total, five passengers were killed and 18 were seriously injured (Beckett, 1984).
Ensuing Events & Investigations
Within days of the collapse, the British government dismissed the inspector who originally approved the bridge's completion (Gagg & Lewis, 2004). Captain John Simmons of the Royal Engineers, an officer with the British Railway Department, was sent to the site two days after the collapse to conduct an initial investigation. His report concluded that the stone piers and girder-to-girder connections were in sound condition, and that overall the bridge did not seem to have construction defects (Walker & Simmons, 1847). He estimated that the train was traveling at a high speed (around 30 mph) since it was behind schedule, but that the train itself had not caused the collapse by abruptly increasing its speed as some had believed. After studying the intact railway, he concluded that the wrought iron tie bars were loose (and thus were not providing tensile reinforcement), passing trains caused significant deflection, and the bridge was subjected to serious vibration on a daily basis.
The Commissioners of Railways called on Simmons and James Walker, a civil engineer, to write a formal report detailing the events of the collapse and its primary cause. First, they summarized the characteristics and dimensions of the bridge, and commented on its overall rigidity and workmanship. They also noted that on several occasions, the bridge had been crossed by trains that were much heavier than the one that caused the collapse (Walker & Simmons, 1847).
On June 4th, a formal inquest was begun to determine the exact cause of the collapse and fatalities. The trial was also concerned with whether or not Stephenson, the designer and structural engineer, was liable for the five fatalities. Walker and Simmons' report (1847) played a key role in the jury's decision. Stephenson's primary defense was that the tender car had derailed, struck a parapet wall above one of the load-bearing piers, and caused the bridge to collapse. However, eyewitness accounts and Walker's report agreed that the derailment was a product of the collapse, not the cause. Based on witness testimonies and the report, the jury reached several conclusions. First, the deaths were accidental and Stephenson was not negligent or liable for manslaughter (Gagg & Lewis, 2004). Secondly, the collapse was was an accident caused by the insufficient strength of the girders. Lastly, the jury recommended that a Royal Commission be established to determine whether cast iron was a suitable material for bridge construction. The Commission's final report, published two years after the Dee Bridge collapse, stated that cast iron members were generally strong under stationary loads, but tended to fail after less than 900 oscillation cycles under live loads (Gagg & Lewis, 2004).
Causes of the Failure
The primary cause of the failure was the ineffective composite action of the cast iron girders and wrought iron tension bars (Walker & Simmons, 1847). Multiple sources agree that this design flaw caused the failed girder to buckle due to torsional instability (Beckett, 1984; Evan & Manion, 2002). As Walker and Simmons (1847) concluded, when the bridge deflected under live or static loads, the wrought iron tension bars loosened and were thus rendered ineffective. While Stephenson had the right idea of using wrought iron for tensile reinforcement, his design did not allow those members to perform as intended. If the wrought and cast iron members had worked in tandem as designed, the bridge most likely would have been strong enough to avoid collapse (Walker & Simmons, 1847). Even when the tie rods were tensioned under passing trains, this actually increased the compressive forces in the top flange of girder, which was significantly smaller than the bottom flange (Beckett, 1984). The cast iron girders alone had a safety factor of about 1.5; the wrought iron bars were meant to raise that factor to an acceptable level, although Stephenson could not precisely calculate its effect (Beckett, 1984). Conversely, modern analysis of another iron bridge built in 1859 showed that while it had significant rusting, it still had an adequate safety factor of about 4 (Gordon and Knopf, 2005).
Robert Stephenson had no real precedence for his design. The equation he used to determine failure loads was based on tests performed on centrally-loaded girders with much shorter spans than the Dee (Beckett, 1984). Although he successfully designed cast and wrought iron bridges for over a decade, none of them were as long or complex as the Dee Bridge. Additionally, the Dee's 98-ft clear spans were nearly 10 feet longer than the longest trussed girder bridges in existence (Beckett, 1984). The application of previously-established norms to bigger and more complex designs, defined by Petroski as "success syndrome," was a common theme of several 19th century failures (Evan & Manion, 2002).
Metal fatigue, defined as "when a component fails well below its rated strength owing to crack formation and growth by repeated loading cycles," also played a major role in the collapse. Basically, structural metal tends to crystallize and weaken due to cyclical loading and unloading. 19th century civil engineers were believed to have some understanding of this brittle characteristic (Petroski, 2012). Four years before the Dee Bridge collapsed, a Scottish engineer named William Rankine wrote a paper that essentially defined the metal fatigue phenomenon, and suggested that it might have been the cause of many sudden and unexpected failures (Petroski, 2012). Metal fatigue was especially detrimental in this case because the Dee Bridge was known vibrate and deflect excessively. Still, Petroski (2012) argued that it was difficult to blame Stephenson for this material flaw because, while the engineering community had a basic understanding of metal fatigue, they did not fully understand how it started and progressed.
The fourth major cause of the collapse is related to an aesthetic detail. Cavetto moldings, popular in contemporary carpentry designs, were added to the corners of the cast iron girders where the web connected to the flanges. Gagg & Lewis (2004) created a scaled girder section out of a polymer material, bent one of the lower flanges to replicate the bridge's asymmetric loading, and studied the section under a polariscope. Their experiment showed that the loads would have been concentrated on the sharp corners of the cavetto moldings, which acted as stress raisers. Since metal fatigue can aggravate small cracks and imperfections in the metal, it is believed that the girder originally fractured near the corners of the moldings (Gagg & Lewis, 2004). Wide flange beams and other modern structural members purposely avoid such stress-raising corners.
The addition of stone ballast just hours before the collapse was the last straw, metaphorically speaking. This action was preceded by the crossing of six trains, none of which caused the bridge to collapse. Gagg and Lewis (2004) argued that by the date of the collapse, the bridge was already significantly weakened. The trains that successfully crossed, along with the 18 tons of additional ballast, must have caused an existing crack to "grow to a critical length, and it grew catastrophically when the 6:15 pm train entered the span" (Gagg & Lewis, 2004). While the previous four factors were significantly involved and would have led to an inevitable collapse, the added weight of the stone expedited the process.
Prevention
Not every facet of this collapse would have been easily preventable during the mid-19th century. In terms of materials, cast and wrought iron were both well-developed and extensively used by the time the Dee Bridge was constructed. The minimal use of wrought iron compared to cast iron was a matter of cost effectiveness, which is still an important component of design and engineering. Due to the state of technology, Stephenson had no means of testing the design's reaction to compressive and tensile forces, which is why the inaugural ceremonies were so crucial for 19th century designers.
The overextension of previous designs to new applications was definitely preventable. Stephenson's design was unprecedented on several levels; it was inexcusable to assume, without any empirical data, that the trussed girder design could be expanded or modified without consequence. The metal fatigue problem represents another nebulous issue. On one hand, the engineering community was somewhat aware of metal's progressive deterioration. At the same time, a renowned engineer like Stephenson should have questioned the bridge's noticeable vibration and deflection.
The cavetto detail was popular in 19th-century wooden beam carpentry (Gagg & Lewis, 2004). Considering the limited nature of empirical testing at the time, it would have been difficult to condemn such a seemingly innocuous detail. However, the cast iron girders were originally tested at the foundry with the flanges loaded symmetrically. If the same load tests had been performed on one of the lower flanges, in accordance with the girders' intended use, the weakness of the cavetto moldings might have been more apparent.
Stephenson's addition of such a large amount of track ballast (5" deep and weighing 18 tons) suggests that he did not consider the bridge to be in any immediate danger. Thus, the perpetual warning signs were deemed inconsequential at most. His reaction to the recent railway fires was understandable, but to simply add more weight to a life-threatening structure without some form of testing or reassurance was regrettable. Modern professional engineers are bound to protect the safety of the public before all else. Periodic inspections were not mentioned in the literature, either.
Lessons Learned
Perhaps the biggest lesson to learn from this collapse is the importance of thorough testing. Evan and Manion (2002), in discussing the Dee Bridge, Hyatt Regency walkway, and similar collapses, argued that there will always be limits to the tests that can be performed. Petroski (1994) previously made a similar statement: "virtually all design is conducted in a state of relative ignorance of the full behavior of the system being designed." This was especially true during the Industrial Revolution. Due to a lack of sophisticated testing equipment and measures, the common practice was to ensure a structure's integrity by purposely overloading it in a parade-like event. The Dee Bridge's inauguration involved roughly 90 tons, far exceeding the expected daily live loads (Walker & Simmons, 1847). Although the structure survived this event, and was thus deemed structurally sound, the need for more rigorous testing presented itself half a year later.
While a lack of testing is a major concern, the suitability of the tests that are performed is equally important. Deflection tests performed on the girders at the foundry involved evenly-distributed loads that were much less than the expected daily live loads, which Stephenson acknowledged (Walker & Simmons, 1847). If performed tests do not reflect the anticipated loading conditions of the completed structure, then they are virtually useless.
Finally, the relevance of prevailing code requirements is paramount. William Fairbairn, a Scottish engineer and close associate of Robert Stephenson, was one of several noted engineers who testified that Stephenson was not negligent. In 1864, several years after the collapse of the Dee Bridge and other notable failures, Fairbairn published a paper that sought to definitively describe wrought iron's performance under continued vibrations and load changes (Fairbairn, 1864). His findings reiterated the issue of metal fatigue, and concluded that one of the strength requirements established by the Board of Trade for iron bridges (a maximum of 5 tons per square inch) was vague and did not distinguish between tension and compression. He further argued that such requirements were unenforceable until the means of construction, the methods of measuring strength, or both were standardized. Lastly, Fairbairn asserted that the current standards written by the Lords Commissioners for Trade were based on limited data and heavily relied on "good material and workmanship" for them to be applicable (Fairbairn, 1864).
Industry Effects
The inquest of the Dee Bridge collapse led to the establishment of a Royal Commission that studied iron's feasibility as a bridge material (Beckett, 1984). The jury's decision, along with the findings of the Royal Commission's report two years later, was that the Dee Bridge and others built in a similar manner were unsafe. Stephenson consequently changed his focus from trussed iron girders to tubular shapes. The High Level Bridge, completed over the River Tyne in 1849, was a wrought iron bowstring girder bridge, meaning that its arches were tied between each masonry pier. Stephenson purposely avoided extending the trussed girder design, as he had done with the Dee Bridge, and his new design properly used cast iron for compressive loads and wrought iron for tensile roads (Beckett, 1984). Similarly, the Britannia Bridge was completed in 1850 and used a tubular shape to carry railroad traffic. This bridge was under construction while the Dee Bridge investigation was still underway (Petroski, 1993).
I.K. Brunel, an English civil engineer and close associate of Robert Stephenson, adapted the wrought iron tubular design for Royal Albert Bridge in 1859 (Beckett, 1984). During the Dee Bridge investigation, Brunel testified that cast iron bridges were inherently faulty, but said that the Royal Commission was impeding the industry's progress by enacting new regulations. Overall, Stephenson and Brunel's work after the Dee Bridge collapse represented a transition point for the civil engineering community: larger bridges would continue to be designed out of necessity, but engineers were increasingly experimental and tried to avoid failed design principles.
Similar Failures
Henry Petroski, a Civil Engineering and History professor at Duke University, classified the Dee Bridge failure as the first of a chain of collapses (Pinsdorf, 1997; Subramanian, 2008). Petroski asserted that, starting with the Dee in 1847, there have been major bridge collapses roughly every 30 years. Thus far, they have included the Tay Bridge, St. Lawrence Bridge near Quebec City, Tacoma Narrows Bridge, "two box girder bridges in Wales and Austria," and finally the I-35W Bridge in Minneapolis (Pinsdorf, 1997).
Of these failures, the Tay Bridge collapse in England, 1879, is the most similar to the Dee Bridge collapse. It was built towards the end of England's Railway Mania, and collapsed due to hurricane-like winds that exceeded the design loads (Pinsdorf, 1997). While the collapse of the Tay Bridge caused 76 fatalities, Pinsdorf (1997) argued that it was a critical turning point for Victorian-era overconfidence and was part of the transition from iron to steel.
Another notable failure occurred in England roughly one year before the Dee Bridge's collapse. The Grays Mill building in Manchester utilized a similar combination of cast iron girders and wrought iron tensions bars. As trusses, the wrought iron members did not reinforce the girders at all; a center girder fractured and caused the entire building to collapse (Gagg & Lewis, 2011).
Links to Tacoma Narrows and I-35W Wiki pages: http://failures.wikispaces.com/Tacoma+Narrows+Collapse; http://failures.wikispaces.com/I+35+W+Minneapolis
Conclusion
The Dee Bridge collapsed due to several problems that had cumulative effects. In a broad sense, it was a victim of an era that emphasized progress and expansion before research and development. More specifically, the engineer did not fully understand the interplay of tension and compression within his design, and grossly overestimated its strength as a result.
Although Robert Stephenson was not found negligent nor responsible for the five fatalities, his fate would probably have been much worse in today's courts. Engineering negligence is typically determined by using a "reasonable man standard" in which the engineer has a duty to perform his tasks and make judgments based on the accepted standards and practices of his time. By this definition, Stephenson was not negligent because the industry as a whole was relatively ignorant. On the other hand, the blatant disregard of the various warning signs has to account for something. Additionally, the determination that Stephenson's design and execution were flawed did not arise decades after the accident, but within weeks. This alone implies that his judgment might not have been so aligned with the state of the industry, after all.
Annotated Bibliography
Beckett, D. (1984). Stephenson's Britain. David & Charles, North Pomfret, VT.
Evan, W.M., and Manion, M. (2002). Minding the Machines: Preventing Technological Disasters. Prentice Hall, Upper Saddle River, NJ.
Fairbairn, W. (1864). "Experiments to Determine the Effect of Impact, Vibratory Action, and Long-Continued Changes of Load on Wrought-Iron Girders." Philosophical Transactions of the Royal Society of London, 154, 311-325.
Gagg, C.R., and Lewis, P.R. (2004). "Aesthetics versus function: the fall of the Dee bridge, 1847." Interdisciplinary Science Reviews, 29, 177-191.
Gagg, C.R., and Lewis, P.R. (2011). "The rise and fall of cast iron in Victorian structures- A case study review." Engineering Failure Analysis, 18, 1963-1980.
Gordon, R., and Knopf, R. (2005). "Evaluation of Wrought Iron for Continued Service in Historic Bridges. Journal of Materials in Civil Engineering, 17(4), 393-399.
Kemp, E.L. (1993). "The Introduction of Cast and Wrought Iron in Bridge Building." The Journal of the Society for Industrial Archeology [sic], 19, 5-16.
Petroski, H. (1992). "Making Sure." American Scientist, 80, 121-124.
Petroski, H. (1993). "Predicting Disaster." American Scientist, 81, 110-113.
Petroski, H. (1994). "Success Syndrome: The Collapse of the Dee Bridge." Civil Engineering, 64, 52-55.
Petroski, H. (2004). "Past and Future Failures." American Scientist, 92, 500-504.
Petroski, H. (2012). To Forgive Design: Understanding Failure. The Belknap Press of Harvard University Press, Cambridge, MA.
Pinsdorf, M.K. (1997). "Engineering Dreams into Disaster: History of the Tay Bridge." Business and Economic History, 26, 491-504.
Subramanian, N. (2008). "I-35W Mississippi river bridge failure- Is it a wake up call?" The Indian Concrete Journal, February 2008, 29-38.
Walker, J. and Simmons, J.L.A. (1847). "Report to the Commissioners of Railways, by Mr. Walker Captain Simmons, R.E., on the Fatal Accident on the 24th day of May 1847, by the Falling of the Bridge over the River Dee, on the Chester and Holyhead Railway; together with any Minutes of the Commissioners thereupon."
Wood, J.G.M. (2004). "Failures from Hazards, a Short Review." International Association for Bridge and Structural Engineering (IABSE), July 2004, 1-4.