Académique Documents
Professionnel Documents
Culture Documents
. 143 150, 1998 1998 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0950 0618 r 98 $19.00 q 0.00
PII:S09500618(97)000160
Department of Civil Engineering, University of Washington, 233 More Hall, FX-10, Seattle, WA 98195, USA Received 20 May 1994; accepted 5 July 1997
Modular expansion joints are commonly used on bridges with expansion and contraction movements in excess of 127 mm. These joints ensure a serviceable bridge surface, and protect the structure and substructure against water damage. However, these joints are dynamically loaded by truck wheels crossing the joint, and severe fatigue damage has been noted. The various types of modular joints are discussed, and fatigue damage on one joint is described. Analytical studies considering the static and dynamic behaviour of the joint are summarized, and the results are correlated to fatigue design methods. Preliminary results from an experimental study of field measurements of joint behavior are provided. The results indicate that the load spectrum is a very important element in the fatigue life of these joints. Acceleration and braking of the truck traffic induces large horizontal forces on the joint, and is a primary contributor to the fatigue damage. 1998 Elsevier Science Ltd. All rights reserved. Keywords: bridges; dynamic response; expansion joints
Introduction
Modular expansion joints are often used on bridges whose expected movements are larger than 127 mm. These systems are relatively expensive, since they are sealed against moisture penetration and are believed to reduce corrosion associated with water penetrating into the structure and substructure. They are separated from other sealed joint systems because they are primarily a structural system designed to accommodate large movements and support large dynamic wheel loads. That is, the design of these joint systems requires consideration of a range of structural design factors including dynamic wheel loads, stress and strain in components, and dynamic response of the joint rather than intuitive rules or guidelines. Modular expansion joints have been used on bridges with movements in excess of one meter, but fatigue cracking has been observed on many of these joints1 3 , and the cracking has signicantly reduced the fatigue life of the expansion joint system. This paper will provide an overview of the fatigue damage noted on one particular bridge, and a summary of the analysis performed to examine the fatigue cracking. Various fatigue design methods for modular expansion joints will be discussed, and a brief description of an experimental program to examine the fatigue loading characteristics will be provided. 143
144
bar is pinned at one end to the bridge superstructure, and on a low friction sliding surface at the other end. In other cases, the support bar is on a low friction sliding surface at both ends, and springs and damping devices are used to control the horizontal position of the support bar and the centerbeam spacing. The bridge movements are accommodated through this support bar sliding action. The support bars are spaced at fairly close intervals commonly 1.5 m.. This spacing limits the spans and bending stress of the centerbeams under
Figure 2 Three types of modular joint: a. mechanical support system; b. multiple support bar system; and c. single support bar system
145
fully established yet. The fatigue problem is quite different for modular joints than its normal steel bridge fatigue design since many, many more cycles of loading will occur, and the fatigue loading depends on the dynamic wheel load rather than the truck weight. The 3rd Lake Washington Bridge in Seattle, Washington, has two 1200-mm modular expansion joints as shown Figure 1 at opposite ends of 1.75 km of oating pontoons. The bridge was opened to trafc in June 1989, and these joints utilize the single support bar swivel joist design. Steel tubes were substituted for the I-shaped centerbeams used in the original design because domestically produced centerbeams were unavailable and FHWAs Buy American steel requirements for federally funded bridge construction would not permit the use of foreign steel. Approximately 6 months after the bridge was opened to trafc, there were numerous complaints of expansion joint noise. Inspection of the joints showed that some elastomeric bearings used to cushion the trafc impact between the centerbeams, stirrups and support bars were loose. Shims were added, but within 1 year cracks in the tubular centerbeams were observed. Most of these cracks started at the toe of the stirrup to centerbeam llet weld and progressed through the centerbeam, as shown in Figure 3. One crack occurred at the end of a ange cover plate. The manufacturer repaired seven of these cracks in April 1991 by rewelding the cracked metal. Additional cracks were noted in the centerbeams after this rst repair, and seven more cracks were repaired in November 1991. Additional cracks were noted after this second repair, and some of the previously repaired cracks reappeared. In the intervening period, more than 20 fatigue cracks have been noted. This research was initiated in response to this fatigue cracking.
including impact proposed by Tschemmernegg are: a vertical downward load of q 91.0 kN, a minimum vertical rebound load of y 27.3 kN and a horizontal load of q 18.2 kN. These design loads for fatigue are based on eld measurements from several bridges in Europe 6 8. The stresses then are calculated at critical locations to determine the maximum computed stress range, max . The center beams are treated as continuous beams, and the elastomeric springs and bearings are treated as rigid supports for determination of the moments and the stress level. For normal conditions, each centerbeam carries approx. 50 60% of the wheel load with a 1.8 m wheel spacing since the wheel distributes the load to more than one centerbeam. Note that the primary loading considered in the design method produces compressive stress in the same area as the fatigue cracking on the 3rd Lake Washington Bridge. Curves representing stress-range to number of cycles S N curves. are determined for each critical component or location. The S N curve is constructed with a slope of y 0.33 on the log log S N curve for a stress range of less than 5 million cycles and a slope of y 0.20 for a stress range between 5 million and 100 million cycles. The intercept at 100 million cycles is the theoretical endurance limit, L , proposed by Tschemmernegg 4 6 , but all tests are performed at 2 million cycles or less. The maximum calculated stress range is compared to the theoretical endurance level from the S N curves by max - L 2 1.
where max is the calculated stress range based on the dened range of wheel loads, and L is the limit states fatigue stress range at 100 million cycles. The maximum calculated stress range is divided by two because of the partial safety factors Mf and , where accounts for accumulated fatigue damage. There are reasons for questioning the validity of the Tschemmernegg procedure. Koster 9 believes that the elastic deformation of the system affects the stress distribution and the fatigue potential. He contends that deformability of the joint is desirable because it may spread the load and possibly reduce the critical fatigue stress. However, Tschemmernegg contends that the elastic deformation and stress distribution will not occur because of the very short duration of the wheel loads. Agarwal performed a series of eld measurements on a modular expansion joint on a bridge in Ontario, Canada10 . These eld measurements suggested that the loads and load spectrum recommended by Tschemmernegg may not be universally applicable. Large horizontal forces noted by Tschemmernegg 6 were not detected and the load range and spectrum were different. However, the centerbeam instrumentation that Agarwal used may not have been adequately located or sensitive enough to detect horizontal loads on the joint.
146
147
longer than approximately 30% of the dynamic period of the structure. If the duration is less than 10% of the period, less than 30% of the static load is felt. Comparison of the computed natural periods with the load history and dynamic amplication illustrated in Figure 5 suggests that signicant amplication of horizontal forces should be expected at slower vehicle speeds. High-speed vehicles may cause the expansion joint to experience attenuation of horizontal loading. Amplication of the vertical response will occur over a wide range of vehicle speeds. It should be emphasized that these observations are meaningful for this particular expansion joint because of the transverse exibility of the system. Other modular joint systems, particularly multiple support bar joints, may be stiffer transversely and experience a greater dynamic amplication of horizontal loads.
detail is somewhat analogous to detail 17 in the AASHTO Specications12 where attachments are welded to a longitudinally loaded member with short llet welds. Detail 17 indicates fatigue category D or E. Two million truck passes will cause far more than 2 million cycles of wheel loading. Comparison of the computed stress range with the HS-20 wheel loads produced stress ranges which were applied to the S-N curves for Categories C, D and E. The fatigue life which resulted was not consistent with the number of trucks which have crossed the bridge during the limited life service of the joint. This suggests that either the S N curves are not appropriate for this modular joint or the dynamic loads are quite different from the HS-20 load history postulated in the US design method. The Tschemmernegg fatigue design loads and the S N curve shown in Figure 6, were used to predict a fatigue life of 10 million cycles of total truck wheel loading for the as-built stirrup to centerbeam connection. This estimate is different from the AASHTO12 and AASHTO LRFD13 life estimates in that it includes the total number of truck passings and an estimate of accumulated damage. The accumulated damage estimate is based on a design wheel load spectrum proposed for expansion joints in Europe. Fatigue cracks were noted approximately 18 months after the bridge was opened to trafc, and 10 million cycles would require approximately 18 000 axles for one lane of trafc per day. A trafc count performed in 1990 found that the three westbound lanes of the bridge experienced approximately 6720 axles of bus and truck trafc during the busiest 12-h period of a normal work day. When the trafc was distributed over three lanes and the lighter weekend trafc was considered, the accumulated trafc was less than 20% of the proposed fatigue life estimate 14 . Furthermore, the cracks obtained in the laboratory fatigue test were quite different than those observed on the 3rd Lake Washington Bridge. This observation suggests that the Tschemmernegg procedure may not be applicable for these joints. While the Tschemmernegg method does not duplicate the fatigue cracking noted in the 3rd Lake Washington Bridge, the stress ranges predicted by the test may be fairly realistic. The reason for the approximate
Figure 6
148
accuracy of the S N curve is that the modular joint details are likely to always be close to AASHTO categories D or E because the weld detail is similar to AASHTO fatigue details 9 and 17. The detail may be closer to category D or even category C if the modular joint is less susceptible to fatigue and closer to E if it is more susceptible. However, it is clear that the load history is more important for establishing a fatigue design criteria and it may be quite different for US truck trafc. As a result, an experimental study was started to evaluate the wheel loads experienced by the modular joint under normal trafc conditions.
Test programs
Two types of tests were performed. First, two series of controlled tests were performed. During 17 19 August 1993, the right lane of trafc was shut down for several hours each day, and loads were applied by a moderately heavy, three-axle dump truck. The dimensions and the static wheel loads of the truck were measured before testing. The truck passed over the joint at known speeds and locations. The truck was sometimes at constant speed, braking, accelerating, or at rest. A second series of controlled tests was performed during 1 2 February 1994. This second series of controlled tests were performed when the bridge was closed to other trafc. Forty-two load passes were made with the same truck used in the earlier tests and a nearly identical loading. Nearly all of the truck passes were made at various points within the outside southern most. lane, so that the maximum useful information with regard to truck location could be obtained. Most of the tests were performed with eastbound truck trafc, but a few passes were performed with a westbound truck. These tests may be useful in interpreting future reversible lane measurements. Two tests were performed with the truck passing in the center lane so that the effect of such a truck passing on the measured results could be determined. The results of these tests will be used to establish basic elements of joint behavior such as the effect of truck position, truck braking or acceleration, and distribution of load between centerbeams. The results will also be used as an aid in interpreting the second type of the test results. The second type of tests consists of measurements of the uncontrolled truck and bus trafc and its affect on the joint. The recorders were set to record data when a large truck crossed the joint in the outside lane. The signal was set such that a truck in the middle lane would not trigger the device nor would a heavy car or light truck in the right lane. Data was recorded continuously in blocks of 10 trucks. A few trucks may be missed during the short period required to transfer the block of data from the wave form recorder to the computer, but the percentage is small. Furthermore, a few lighter trucks are missed because they are not heavy enough to trigger the instruments, but they do not contribute much to fatigue. To illustrate the number of trucks recorded, it can be noted that during the 24-h period on Wednesday, 1 September, 990 vehicles were measured in this right hand lane of the eastbound I-90. This is not inconsistent with the truck trafc count of 13 November 1990 where a total of 6720 axles were counted in all of the westbound lanes during a 12-h period. The 990 trucks should result in nearly
149
Figure 7 Measured bending moment in centerbeam due to vertical wheel loads at 90 kmrh and static conditions
Figure 9 Measured bending moment in centerbeam due to horizontal wheel loads at 90 kmrh with moderate braking
5000 axles in a single lane for a 24-h period. These measurements will be used to establish a fatigue load spectrum for vertical and horizontal wheel loads on the joint. It should be emphasized that these loads are not static wheel loads. They are dynamic loads including the effect of impact and dynamic amplication or attenuation. The variation and distribution of stress is quite interesting since it affects fatigue behavior. However, the stress and bending moment at the strain gage locations depends upon the placement of the wheels on the centerbeam, and this placement is not known for the uncontrolled trafc measurements. The rst four channels of measurements allow determination of the vertical plane bending moment at four locations on the centerbeam under one lane of trafc. These four moments are enough to estimate both the magnitude of the dynamic wheel load on the centerbeam and position of the wheel on the joint through inuence lines developed for each gage location. It should be noted that the inuence lines are quite complicated and depend upon dynamic effects as well as static loading. More measurements are made than are actually needed to determine load and position. This provides redundancy which can be used to verify and improve the
accuracy of the estimates. The position of the vehicle wheels combined with the two horizontal plane bending moments allow estimation of the horizontal wheel load on the centerbeam.
Figure 8 Measured bending moment in centerbeam due to horizontal wheel loads at 90 kmrh and static conditions
150
Acknowledgements
Figure 10 Measured bending moment in centerbeam due to horizontal wheel loads at 90 kmrh with emergency braking
This research is being funded by the Washington State Transportation Center TRAC. and the Washington State Department of Transportation. The assistance of John Van Lund, Myint M. Lwin and Alan H. Walley is gratefully acknowledged. The opinions and conclusions expressed or implied in the paper are those of the author and are not necessarily those of funding agency.
References
1 Roeder, C. W., Fatigue cracking in modular joints. Report No. WA-RD 306.1, Washington State Transportation Center TRAC., March 1993 2 Van Lund, J. A., Improving the quality and durability of bridge modular expansion joints. Transportation Research Record 1393, TRB, National Research Council, Washington, D.C., 1993, pp. 9 16 3 Mayrbaurl, R. M., Analysis of the Manhattan Bridge modular expansion joints. 1994 TRB Annual Meeting, Washington, D.C 4 Pattis, A. and Tschemmernegg, F., Fatigue testing and design of modular expansion joints. Report published by University of Innsbruck and The D.S. Brown Co., Innsbruck, Austria, March 1992 5 Fatigue design and testing for expansion joints. Technology Bulletin for Bridge Bearings, Expansion Joints and Components. No. 1, The D.S. Brown Co., North Baltimore, Ohio, October 1991, pp. 2 6 Tschemmernegg, F., The design of modular expansion joints. Preprint, Volume 1 Joints and Sealants, Third World Congress on Joint Sealing and Bearing Systems for Concrete Structures, Toronto, Canada, American Concrete Institute, October 1991, pp. 67 86 7 Ostermann, M., Stresses in elastically supported modular expansion joints under wheel impact load. Bauingenieur, 1991, 66, 381 389 in German. 8 Braun, C., Strain on expansion joints for road bridges under trafc loads. Bauingenieur, 1992, 67, 229 237 in German. 9 Koster, W., The principle of elasticity for expansion joints. In Joint Sealing and Bearing Systems for Concrete Structures, Vol. 2, SP-94. American Concrete Institute, Detroit, MI, 1986, pp. 675 711 10 Agarwal, A. C., Static and dynamic testing of a modular expansion joint in the Burlington Skyway. In Proceedings of the Third World Congress on Joint Sealing and Bearing Systems for Concrete Structures, Toronto, Canada, October 1991 11 Stanton, J. F., Roeder, C. W., Elastomeric bearings design, construction and materials. NCHRP report 248, TRB, National Research Council, Washington, D.C., 1982 12 Standard Specications for Highway Bridges, 15th ed. AASHTO, Washington, D.C., 1992 13 Proposed LRFD Standard Specications for Highway Bridges. 17 April 1992 Draft, NCHRP 12 33, Washington D.C., 1992 14 Pattis, A., Fatigue testing and design of modular expansion joints the BrownrMaurer swivel joist system repair procedures for 3rd Lake Washington Bridge. Short version of an unpublished report, August 1992
Figure 11 Measured horizontal displacement of centerbeam due to horizontal wheel loads at 90 kmrh with emergency braking
Figure 11 shows typical centerbeam movement if the truck is braking to an emergency stop. It can be seen that a large horizontal deection occurs under this severe braking condition. The maximum movement is approximately 10 mm and there is a permanent set of approximately 3 mm. The largest centerbeam movements appeared to occur at slower speeds because of the dynamic characteristics of the joint. This is consistent with the observations made in analysis that greater dynamic amplication of horizontal loads occurred at slower speeds because the duration of loading more closely matches the longer periods noted for horizontal displacement. The major portion of the deection is due to deformation and sliding of the elastomeric springs. That is, the centerbeam approximately moves as a rigid body. The elastomeric deection is elastic and is recovered after the load is removed, while sliding results in permanent set and is not recovered. This again is quite different than assumed in existing fatigue evaluations procedures. Additional interpretation of this data is needed. Ultimately, it is expected that a statistical load spectrum will be developed. Examination of the results to date clearly indicates that the fatigue evaluation procedures