Tennessee+River+Bridge+Collapse

By: Chris W. Knipp PSU CE M.Eng. Student Fall 2013, cwk141@psu.edu

**Key Words:** stability; hybrid girder; bracing; bridge; structural; failure; collapse; construction; investigation; Tennessee River Bridge; State Route 69 toc

1.0 Introduction
On the morning of May 16, 1995, a three-span bridge along State Route 69 over the Tennessee River suddenly collapsed. The bridge was under construction and the two completed spans fell into the Tennessee River. There were three men working on the structure at the time of the collapse. One of the men was attached to a collapsed girder via a safety harness and as a result drowned. Two other men working on the bridge were rescued.

The ultimate cause of the failur e was determined to be due to inadequate bracing. This report will look into the cause of the bridge failure and study the codes and standards of the time to determine their efficacy. In addition, the common practice of design of plate girders will be investigated to determine proper bracing during construction. The failure investigation was carried out by Wiss, Janney, Elstner Associates, Inc. (WJE, 1995) and a seperate academic study was conducted by Dr. Daniel A Wojnowski (Wojnowski et al., 2010). The results of both investigations will be discussed to determine ways that this bridge failure could have been avoided. The majority of the information, photos, and diagrams presented in this article are from the final failure report written by WJE. The background and events preceding the collapse are summarized from the findings of WJE's investigation.

2.0 Background
The bridge over the Tennessee River along SR 69 near Clifton, Tennessee was designed by the Tennessee Department of Transportation (TDOT). The bridge was to consist of a 1,690 feet multi-span, precast concrete girder segment up to the river crossing and a 1,205 feet three-span steel portion that crossed the river with a finished roadway width of 48 feet. The construction of the bridge was let to McKinnon Bridge Company of Franklin, Tennessee. Up to this project, McKinnon had constructed approximately 20 major bridges over the Tennessee River. The steel erection for the river crossing was subcontracted by ABC Contractors, Inc. of Lavergne, Tennessee. The steel fabricator on the project was Trinity Industries.



The bridge superstructure consisted of three identical steel hybrid girders constructed in segments. The girders varied in depth from 8 feet at the two ends to 14 feet at the piers and throughout the center span. The flange dimensions of the girders varied throughout the superstructure as well. An elevation view of the three spans over the river is presented in Figure 2 above. The outer spans had lengths of approximately 340 feet while the center span had a length of 525 feet. Cross frames between girders were spaced at 25 feet.

3.0 Events Preceding Collapse
ABC Contractors, Inc. was responsible for the steel erection of the three girder bridge over the river. ABC used a system of barges to transport the girders to the site and cranes to lift the girders into place. While working on the main span of the bridge, ABC would connect cross frames between girder two and three (G2 & G3) on the barge and lift the pair into position where workers would connect the girders to members that were already in place at field splices (FS). The team would construct girder lines 2 and 3 ahead of girder line 1 a certain distance. Then they would erect girder G1 into place in segments and connect the girder at field splices. When the girder was properly positioned, cross frames between girder G1 and G2 would be installed. S till frames from a video shot during construction are presented in Figures 3 and 4 below depicting the erection procedure. As can be seen, girders G2 and G3 are erected with cross frames installed a distance ahead of girder G1.



Prior to the collapse, the work was progressing from pier 14 to pier 15, west to east. Girders G2 and G3 were placed until they were cantilevered past pier 15. ABC then began transporting and setting segment FG of G1 into place up to FS6 and installing intermediate cross frames at FG4 and FG6. The team erected segment H and began work on JK. While the team was setting segment JK of G1 over pier 15, a significant lateral sweep was observed in the east end of segment H of G1. A still frame from construction video is presented below in Figure 5. As can be seen girder G1 is projecting away from the bearing at pier 15. Also, a diagram depicting the locations of cross frames and crane locations is presented in Figure 6. There were two intermediate cross frames installed at the time of the lateral sweep, FG4 and FG6. In order to properly install segment JK of girder G1, ABC needed to correct the sweep at the end of segment H. The erectors did this by attaching a cable hoist to the bottom flanges of girder G1 and G2 and a crane was attached to the top flange of girder G1 to pull the two girders closer. Prior to the girder adjustment, the connections of cross frame FG6 were unbolted to allow for transverse movement. Cross frame FG4 was left fully connected. When the crew moved girder G1 towards G2 via the hoist and crane, a double angle section of cross frame FG4 buckled. A picture taken during the investigation of the buckled cross frame is presented in Figure 7. With the girder in the correct position, ABC erected segment JK of girder G1, cantilevered past pier 15 and aligned with the ends of the other girders. A cross frame at pier 15 between girders G1 and G2 was fully installed. All cranes were released after this section was erected. It should be noted that this is how the bridge was left over the weekend prior to the day of the collapse. This proves that the bridge was in a stable condition with this girder and cross frame arrangement. The following week, ABC began connecting cross frames between girders G1 and G2. The work progressed from the cantilevered ends of the girders towards the main span. Cable hoists between girder G1 and G2 were used to aid installation of the cross frames. During the installation of the cross frames the iron workers noted significant misalignment of bolt holes between many of the cross frame and girders. As a result many of the connections were only partially bolted.

4.0 Collapse
On May 16, 1995 ABC continued work on connecting cross frames between girder G1 and G2. A diagram of the work completed through May 16 is presented in Figure 8. As can be seen, at the beginning of the day on Tuesday, cross frames were at least partially installed up to cross frame H3. Work continued heading west, installing cross frame H2 and H1. Ironworkers again had trouble installing the cross frames because the top flanges of girder G1 and G2 were too close. Therefore, only the bottom flange connections were completed. To correct this issue, ABC decided that a jacking beam needed to be fabricated to spread the top flanges.

While in the process of fabricating the jacking beam, the team erected cross frame FG5 into place between the previously installed FG4 and FG6. Before the cross frame at FG5 could be fully connected to the top flange of girder G1, the workers began removal of the buckled cross frame FG4. To accomplish this the ironworkers either unbolted or torch-cut the bolts of the connection. Shortly after the disconnection of the cross frame the structure collapsed. Three ironworkers were on the superstructure at the time of the collapse. All three fell into the water, two men were rescued with minor injuries and one man died. A photo taken on the day of the collapse is presented in Figure 9.

There are some observations that can be made from this photograph:
 * Cross frame FG4 is still being held by the crane, verifying that this frame was disconnected prior to collapse.
 * All three girders across the main span appear to have rolled in the direction of girder 1, towards the north.
 * Finished span 14 was lifted off of pier 13 and rolled towards the south.

5.0 Cause of Collapse
This portion of the article will focus on the cause of the collapse of the Tennessee River Bridge. A brief discussion of lateral buckling of members in flexure will take place. This will be followed by the two investigations as to the cause of the collapse that were conducted. The Attorney General of the State of Tennessee hired Wiss, Janney, Elstner Associates, Inc. (WJE) to investigate and make a determination of the cause of the collapse. In addition, in 2002 an academic study was conducted by Wojnowski et al., with a focus on design guidelines and analytical methods for stability of a bridge during construction. It should be noted that these studies came to differing conclusions.

5.1 Lateral Buckling of Flexural Members
To preface the failure investigations, a brief introduction to lateral buckling is provided. When a member is subjected to a vertical load it tends to bend about it's cross-sectional horizontal axis. In reference to an I-beam, the girder shape of the Tennessee River Bridge, the bottom flange elongates and is in tension while the top flange shortens and is therefore in compression. At some critical loading, the compression flange will no longer shorten, or compress, but will begin to bend sideways. This movement is resisted to a degree by the tension flange and the web. At some point, if the compression flange is not adequately supported, the girder can become laterally unstable. This usually occurs before yielding of the compression flange. The lateral stability of girders is largely dependent on the the space between lateral supports (i.e. the unbraced length).

Erectors of bridge structures, being "responsible for the means and methods of construction"(Duntemann, p. 375), have historically relied on rules of thumb to determine the adequacy of lateral support during construction (Hastings, p. 322). Using the ratio of the unbraced length to the compression flange width (L/b), contractors could determine where bracing was required. A diagram of values used to calculate this ratio is presented in Figure 10. Values of stability ratios for simply supported girders are listed below:
 * Less than 60, stability is guaranteed
 * Between 60 and 80, stability is doubtful
 * Greater than 80, temporary supports are required

It should be noted that these values have no theoretical basis but are based entirely on experience. A more rational development of adequate L/b ratios will follow the two investigation reports.

5.2 WJE Failure Report
WJE began investigation of the failure on May 17, 1995. The full forensic study included removal and documentation of all recoverable members from the site, interviews with the members of the construction crew on site at the time, and structural analysis of probable in-situ conditions. The final report thoroughly covered all defects discovered upon recovery of the parts of the bridge. However, only the pertinent conclusions will be covered here.

To begin the analysis, WJE investigated the state of the cross-frame connections at the time of the collapse. Using the accounts of the ABC ironworkers, the recovered material, and video from a recreation boat, WJE was able to reasonably determine the number of bolts in the cross frame connections between girders G1 and G2. A diagram depicting their finding is presented in Figure 11. As can be seen there were no top flange connections complete between cross frames E5 and FG6, a length of approximately 150 feet. Directly adjacent there was an unbraced length between cross frame FG6 and H3 of 75 feet. Using this configuration, structural analyses were performed for all three girders. The capacity of girders G2 and G3 were found to be adequate in this state. The stability of girder G1 was investigated using several different procedures. Using the direct AASHTO procedure and the most stable section between cross frames E5 and FG6, an ultimate flexural capacity of 21,000 ft-kips was found. This is much less than the expected applied moment of 33,000 ft-kips. WJE noted that the AASHTO procedure applied directly did not account for torsional stiffness provided by the adjacent girder segments, and therefore may be a low approximation of the girders actual strength. Taking the torsional stiffness into account by a modified effective length method, WJE found the girder had a critical flexural capacity of 36,400 ft-kips, which is relatively close to the estimated applied moment at the time of the collapse.

Finally, a finite element model was constructed to more accurately analyze the interaction of the entire superstructure. Using this model and assuming initial imperfections, such as lateral sway, WJE found the that the top flange would have yielded near cross-frame H3. This represents an inelastic instability; as the top flange of girder G1 between E5 and FG6 displaces to the north, the section between FG6 and H3 displaces laterally to the south and a hinge forms near H3. Because the top flange at H3 has yielded, it is not possible to stabilize the girder. The top flange will continue to displace laterally and more yielding will occur at H3. Then, WJE concluded, cross frame FG6 failed and girder G1 continued to displace laterally. Eventually, the connection at cross frame H3 failed and a plastic hinge formed near cross frame H4, at which point the bridge failed catastrophically. Photos depicting the hinge near cross frame H3 is presented below in Figures 12. Also, a series of diagrams depicting the failure sequence is presented in Figure 13.

5.3 Wojnowski et al. Study
In 2002, Dr. Daniel A. Wojnowski published a finite element analysis of the Tennessee River Bridge (Wojnowski et al., 2002). The paper makes no reference to the WJE official failure report and therefore, it can be assumed, that the authors were not privy to some of the information presented within this article. However, the goal of the paper was not solely to find the cause of the failure but also cover "design guidelines and analytical methods that can be used by engineers to determine stress states and stability of a bridge structure during erection." (Wojnowski et al., p. 87)

The authors could not conclude exactly what the status of the installation of cross frames at the time of the collapse. Therefore, five different probable cross frame configurations were analyzed. A list of the configurations is presented below. None of the analyses included the H3 cross frame which is a direct contradiction to the findings by WJE. Also, it was the authors' belief that the cross frame that was being held by the crane after the collapse, presented in Figure 9 above, was actually H4, not FG4 as WJE concluded. They concluded that retro-fitting of H4 was necessary due to fabrication errors. By running the analyses of the five configurations, Wojnowski found that an unbraced length of 45.7 m (150 ft) or more would result in a catastrophic failure, while an unbraced length less than would not fail. The authors' concluded that the likely failure of the bridge occured when the cross frame at H4 became unattached which increased the unbraced length from 30.5 m (100 ft, between H4 and JK3) to 61 m (200 ft, between FG6 and JK3). Again, this is a deviation from the findings of WJE which found evidence of connections of cross frames H3, H4, H5, JK1 and JK2.
 * A1-E5, JK3
 * A1-E5, FG6, JK3
 * A1-E5, FG6, H4, JK3
 * A1-FG6, H4, JK3
 * A1-JK3

5.4 Rule of Thumb
As was stated previously, erectors are responsible for maintaining stability of a structure during construction. Without advanced software available in the field, rules of thumb have historically been used, listed in section 5.1. Using the smallest compression flange width between cross frames E5 and FG6, 30 inches, and an unbraced length of 150 feet, a ratio of 60 is found. Following the rule of thumb of the industry at the time, this would guarantee a stable structure. It should be noted, that there is no evidence that this rule of thumb was used by the erectors during construction of the Tennessee River Bridge. However, it does reveal the industry standard at the time.

In 2010, John Hastings, a civil engineer at the Tennessee DOT, undertook a study to determine the efficacy of rules of thumb used by contractors in the field for girder stability (Hastings et al., 2010). As was stated, the rules of thumb have no rational basis. Therefore, Hasting's performed analyses for a wide range of girder sizes using AASHTO specifications. Unfortunately, the study did not include dimensions similar to the Tennessee River Bridge, but some conclusions can still be made. The original L/b limits of 60 and 80 for a simply supported girder are grossly unconservative. Also, Hastings found that the stability of the girder depends largely on the web height, a parameter not directly taken into account in the rule of thumb calculation. Hasting's found that new limits of 45 and 60 should be adopted. These new limits apply to a wide range of girder dimensions. Using these new values, instability would be expected with the unbraced length produced by the removal of cross frame FG4 (150 feet).

6.0 Relevant Codes and Standards
The Tennessee River Bridge along SR 69 was designed according to the governing Standard Road and Bridge Specifications of the Tennessee Department of Transportation (1981 Edition) and the AASHTO 1992 Standard Specifications for Highway Bridges. This section will contain an investigation into the standards of the time for stability of plate girders during construction. It should be noted that exact specifications used for the design were not able to be obtained. Therefore, the AASHTO specifications of the 1994 edition will be used (AASHTO, 1994). Also, this section will contain a comparison between the standards of the time, including the AISC LRFD Speications from 1994 (AISC, 1994), and more modern standards, specifically the 2010 edition of AASHTO (AASHTO, 2010).

For the lateral torsional buckling limit state, the specifications treat the compression flange as a column. The buckling of a column will almost always occur before it reaches its full yield stress. The critical buckling stress is dependent on a number of factors including the applied moment shape, the unbraced length, and effective radius of gyration. By comparing the standards, they are all essentially the same but in slightly different forms. The only major difference between both editions of AASHTO and AISC is the resistance factor for flexure. AASHTO uses a value of 1.0 while AISC uses 0.9. This is due to the service limit states for bridges being based on experience and engineering judgment rather than statistically based (Barker & Puckett, p. 710).

7.0 What We Learned
Most bridge failures during construction occur due to inadequate bracing (Feld, p. 150). The failure of the Tennessee River Bridge was an unfortunate reminder of the importance of stability during construction. As has been noted, the stability of the girders during the construction phase is often left up to the means and methods of the contractor. As a result the successful construction of a bridge relies heavily on the contractors experience and the rules of thumb that the contractor chooses to employ. But even the most experienced contractor makes mistakes. This section focuses on what was learned in the wake of this tragic bridge failure.

7.1 Inadequacy of Construction Standards
The contractor is responsible for the means and methods of construction. Because of this, "the engineer of record for the bridge relinquishes some control of the project, which, in turn, increases the probability of construction complications or failures." (Duntemann, p. 375) Studies of the efficacy of the rules of thumb historically used by contractors during bridge erection were found to be grossly unconservative. The failure of the Tennessee River Bridge prompted the industry to make adjustments, albeit 16 years later, to provide more accurate and conservative L/b ratios (Hastings et al., 2010). Use of the new ratio limits could have prevented the catastrophic collapse of the Tennessee River Bridge. However, even with better L/b ratios, it is clear through this case study that unforeseen circumstances can arise and improved standards need to be developed.

7.2 Precaution Before Production
Errors can happen and when they do the contractor needs to have the knowledge to provide flexibility in the construction process. As was seen in this case study, the misaligned bolt holes and misaligned girders, preventing cross frame connections, caused an unforeseen cross frame configuration. Also, the additive effect of the buckling and removal of cross frame FG4 was an unexpected event. In both cases, precaution should have been taken by setting temporary bracing or installing cross frames between E5 and FG4 to reduce the unbraced length and lessen the chance of buckling.

8.0 Similar Bridge Collapses
Unfortunately, bridge failures during construction are not uncommon and stability is only one issue that needs to be considered. Below are a list of other bridge collapses that occurred during construction:
 * Quebec Bridge, Canada, 1907 and 1916 - One of the only bridges to be known for two failure incidents during construction. The first collapse occurred in 1907 due to inadequate design. The bridge was to have the world's longest span and the subsequent investigation revealed stresses "being higher than any established by past practice." The incident claimed 74 lives. The second failure occurred in 1916, when a reassembled span was floated to the site and was being lifted into place by jacks. When the structure was lifted only 12 feet a sudden collapse occurred. Eleven workers were killed in the incident. (Feld, p. 152)
 * The Westgate Bridge, Melbourne, Australia, 1970 - Steel box girders were preassembled in two halves and transported to the site where they would be fitted together with field connections. The box girders were to act compositely with the deck when finished. However, during construction the compression flanges of the box girders were unstable and suffered local buckling. In an attempt to correct the local buckling, transverse connections between the two halves of the boxes were released. This transferred a large load to an adjacent span ultimately causing a complete collapse of the structure. The collapse killed 35 people. (Ghosh, p. 28)

9.0 Conclusion
The collapse of the Tennessee River Bridge along state route 69 was yet another fatal disaster of a bridge during construction that could have been avoided. Through multiple investigations, although with slightly differing conclusions, it was determined that inadequate lateral bracing, in the form of cross frames or temporary bracing, was ultimately the cause of the collapse. The industry has responded by studying the criterion used by the construction industry prior to the collapse. The study found that the common L/b ratios used by contractors were grossly inadequate and therefore were revised in accordance with modern standards. However, it is clear that standards of practice for the construction industry need to be improved to avoid future catastrophic collapses of bridges during construction.