Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
46 CHAPTER 3 INTERPRETATION, APPRAISAL, AND APPLICATION CAPACITY OF CONNECTION DETAILS The results of the experimental work answered several questions about details for positive moment connections. This work examined both bent-strand and bent-bar type con- nections. Connections using embedded girder ends, addi- tional stirrups, and horizontal web bars were also examined. The survey results showed that a bent-strand type of con- nection was used in many states, but the AASHTO LRFD Specifications and Standard Specifications do not provide a means of determining the strand length needed for this con- nection (12, 15). A set of equations from research done for the Missouri DOT was found (8â10, 13) and was used to detail the bent-strand connections in this study. The bent- strand connection, designed using the proposed equations, had adequate capacity to resist positive moment. The con- nections were easy to fabricate and to erect. Using the bent strand connection may also help to reduce congestion at the end of the girder. The AASHTO LRFD Specifications require a check on the amount of longitudinal reinforcement at the end of concrete beams (12). The contri- bution of the prestressing strand is reduced if there is an inad- equate development length. As a result, there is often a need for additional steel in the already congested end region of the girder. The extended bent strand clearly provides anchorage into the diaphragm. For the extended bent strand, the pull-out stress should be the sum of the pull-out stress from the devel- opment length at the end of the girder plus the pull-out stress from the bent strand as calculated by the given equations. The bent-bar connections were designed using the provi- sions of the AASHTO Standard Specifications for hooked bars (15). These provisions are nearly identical to the AASHTO LRFD Specification provisions for hooked bars (12), so the results are applicable to either specification. The connections were capable of developing the nominal moment capacity. In the case of the bent bars, there was some concern that con- gestion in the diaphragm region might reduce the capacity since it is well known that there are interaction effects when bars are in proximity to each other; however, this was not found to be the case. The bent-bar connections were found to be more difficult to construct. In flanged members such as bulb-T or I sections, the tails of the hooked bars extend beyond the top of the bot- tom flange. Pre-bent bars cannot be used because they inter- fere with the formwork. The bars are installed straight and then bent later. This is a labor-intensive operation, and the resulting hooks are usually not uniform, making it difficult to install anchor bars in the corners of the hooks. The uneven hooks may also cause uneven stresses in the bars. One possi- ble solution is to make 180° bends in the bars (see Figure 70). Another concern with bent-bar connections is that the bar pattern must usually be asymmetrical to allow for the mesh- ing of the bars. This results in asymmetrical behavior of the connection, asymmetrical stresses in the bars, and asymmet- rical crack openings. Some states avoid this problem by using a wide diaphragm and not meshing the bars. Assembly of the bent-bar connection was slightly more dif- ficult than for bent-strand connection. The strands are flexible and can be easily moved by hand from side to side during assembly. This is not true of the bars. However, neither type of connection required extraordinary effort to assemble. Embedment of the end of the girders into the diaphragm did seem to reduce the stress in the connection and, in general, the embedded connections had a higher number of cycles to fail- ure. However, this effect may be difficult to quantify. The reduction in stress is due to the weak chemical bond between the girder and diaphragm concrete and/or frictional effects as the girder tries to pull out. The magnitudes of these types of stresses vary widely and may be difficult to assess. It is prob- ably best to ignore any embedment effects in design and to allow the embedment to provide an additional, although vari- able, factor of safety. The placement of additional stirrups in the diaphragm just outside of the girder bottom flange (see Figure 22) does not increase strength. In fact, prior to pull-out of the connection, a vibrating wire strain gage placed in the diaphragm but outside of the area between the girder ends (see Figure 22) showed almost no response. This indicates that there is no stress in the diaphragm outside of the cross section of the girder; therefore, it is not surprising that steel placed in the diaphragm out- side of the area between the girder ends does not affect behav- ior before pull-out. However, after the connection pulls out, the additional stirrups arrest the diagonal cracks that form in the diaphragm and provide additional ductility (see Figure 23). The specimen using the additional stirrups had the girder ends embedded in the diaphragm, and the experimental evidence suggests that embedded ends are needed for the additional stir- rups to provide the ductility. The large diagonal cracks the
47 stirrups arrest form in the embedded specimens, but are less pronounced or nonexistent in the nonembedded specimens. This type of detail may be useful in providing additional duc- tility in seismic zones. The use of horizontal bars through the web enhanced the performance of the connection. This type of connection was stiffer and more resistant to fatigue. However, when the con- nection failed, there was considerable cracking in the beam. This may not be desirable. The decision to test a connection with horizontal web bars was made after the girder sections were cast. Holes for bars were drilled into the webs between the shear stirrups. There was no additional reinforcement in the web around the holes. Had the holes been cast into webs and additional reinforcing added, the cracking in the girders might have been less severe. The test results showed that all of the connection details performed adequately, and each had advantages and dis- advantages. There was no detail that performed markedly better or worse than any other. Thus, the selection of a spe- cific detail should be left to the preference of the engineer, state DOT, or both. BRIDGE BEHAVIOR Temperature Effects The most striking result in the full-size tests was the influ- ence of temperature in the system. During a single 24-h period, the end reactions often changed as much as ±5 kips, 20% of the average value of 25 kips. This change generated diaphragm moments of ±250 k-ft, approximately 60% of the positive cracking moment for the section or 2.5 times the live- load positive moment. The AASHTO LRFD Specifications require that temperature effects be considered (12), but these effects are often ignored in design. The experimental results show that temperature effects can be significant. Alabama attributes cracking in bridges to temperature effects (20, 21). Partial Diaphragm and Slab Casting Effects In the first full-size specimen, a partial diaphragm was used. Initially, the bottom one-third of the height of the diaphragm was cast. After 28 days, the remaining part of the diaphragm was cast along with the slab. The idea was that the weight of the slab would cause the ends of the girder to rotate into the partial pour, compressing the diaphragm and the girder ends. This would then prevent positive moment cracking due to creep shrinkage. It is not clear how well this actually worked. During the process of casting the slab, about 50 microstrain of compression was seen at the diaphragm or at the girder ends. This would translate to a stress of 200 psi. However, the system was already in tension from positive moments, which had occurred since the partial diaphragm was cast. Even with the compressive stresses generated when the slab was cast, the system was still in a net state of tension. After a few hours, the slab concrete began to heat up from hydration. The top of the specimen expanded, the system cambered up, the center reaction decreased, and the end reactions increased. This caused a positive moment on the connection, which relieved all compression caused by cast- ing the slab. Later, the slab cooled and contracted. This caused the system to deflect downward, increasing the cen- ter reactions and decreasing the end reactions. It appears that a net negative moment on the order of 500 k-ft was created and that the girder ends were compressed into the diaphragm. A net negative strain of 100 microstrain was observed in the girder ends and in the diaphragm. This would correspond to a stress of approximately 400 psi. Thus, it appears that the partial diaphragm worked, but not by the mechanism assumed. A tension tie at the top of the girders should improve per- formance because it seems that much higher compressive stresses were generated at the bottom of the diaphragm once continuity was established. This tension tie either can be a mechanical tie between the girder tops or it can be made by pouring the entire diaphragm and a portion of the slab in the negative moment region. Effects of Creep and Shrinkage The first full-size specimen was constructed and then mon- itored for a period of 4 months after the deck was cast. Accord- ing to the model of the system, the girders should camber upward due to creep and shrinkage. Before the deck was cast, a partial diaphragm was cast. A total of 28 days elapsed between the time the partial diaphragm was cast and the time the deck was added. During this time, the girders cambered upward. This increased the end reactions, causing positive moment at the connection. Tensile strains were seen in the girder end and in the diaphragm. When the deck is added, the models predict that the deck will shrink and the differential shrinkage between the slab and Figure 70. Positive moment connection with 180o hooks.
48 girders will cause the formation of negative moment at the connection. This was not observed. In fact, over the monitor- ing period, the end reactions increased and the center reaction decreased, indicating the formation of positive moment. Similar results were seen in three projects using high- performance concrete (26). The girders in these projects ranged from 60 to 300 days old. The data did not show the expected downward camber or formation of negative moment. The data presented by Ramirez et al. (14), show that the effect of differential shrinkage is overestimated by the models. This is also consistent with anecdotal evidence as there have been no reports of severe negative moment dis- tress occurring in continuous-for-live-load bridges. The problem does not appear to be the structural model. As noted in the previous section, there was a clear differen- tial contraction of the deck due to temperature after casting, and the behavior was consistent with the model predictions for a contraction of the deck. The girders deflected down- ward, the end reactions increased, the bottom of the connec- tion showed compressive strain, and there were tensile strains at the top of the connection. The problem appears to be that values used for deck shrinkage are not correct. In most mod- els, the values for deck shrinkage are based on unrestrained shrinkage values. These values either are obtained experi- mentally or are from empirical equations based on the results of experimental studies. These values of shrinkage do not seem to duplicate the actual field conditions. Many models do not correctly account for the restraining effects of the rein- forcing bar or the girder, actual field relative humidity, and rewetting of the deck by rain or snow. This difficulty in assessing the effects of differential shrink- age has an impact on the assumed stresses in the connection. The models generally show that negative moment caused by differential shrinkage will mitigate positive moments cause by creep and shrinkage of the girders. However, if these negative moments do not form, the actual positive moments on the con- nection will be much greater. This issue needs further study. Some states attempt to solve the creep and shrinkage issues by specifying maximum ages, minimum ages, or both for the girders at the time continuity is established. The max- imum age limit is to prevent the formation of the large neg- ative moments caused by differential shrinkage of the deck slab. Since the data suggest the models overestimate this effect, there does not appear to be a reason to limit the age of the girder at the time continuity is created. However, a min- imum age does seem advisable. If continuity is formed when the girders are young, creep and shrinkage in the girders cause large positive moments to form. Since the data suggest that these positive moments are not mitigated by the deck slab shrinkage as much as the models predict, the actual pos- itive moments that develop may be worse than predicted. If the girders are allowed to age, much of the creep and shrink- age will occur before continuity is established. This will lower the magnitude of the positive moments that form. Continuity In the first full-size specimen, cracks formed at interface of the girders and the diaphragm. These cracks varied in width (due to the asymmetrical bar pattern), but some were as wide as 0.02 in. The crack did not extend into the slab. Continuity was assessed by loading the specimen such that the moment at the diaphragm was either the positive or neg- ative live-load moment and observing changes in reactions and strains in the cross section. This specimen showed con- tinuity under all loads. This is consistent with observation from Alabama (20, 21). The second full-size specimen had cracks between the diaphragm and the girders of 0.07 in. These large cracks were induced by increasing the positive moment by raising the end supports. This specimen maintained continuity until the con- nection appeared near failure. This determination that the con- nection was near failure was based on the observations that the interface crack had propagated into the slab, the bottom of the diaphragm was cracked, and a diagonal crack had formed on the face of the diaphragm. The tests of the stub specimens indicated that these cracks were signs of the impending fail- ure of the connection. At the point were the connection seemed near failure, the system still maintained approximately 70% continuity. It therefore appears the cracking at the diaphragm does not affect continuity unless the cracking is severe. One reason for this is probably the condition of the center support. In most models, the system is modeled as having a single center sup- port. In reality, the center support consists of two supports, one under each girder end. The girder ends and the diaphragm then form a âmemberâ between the two center supports. The presence of a crack on either side of the diaphragm does not appear to alter the stiffness of this intermediate member enough to cause a complete loss of continuity. However, as the diaphragm cracks and the cracks propagate into the slab, this center section softens enough to cause some loss of con- tinuity. From the results of the stub specimen tests, it appears that continuity will only be completely lost when the con- nection fails and a hinge forms. The models predict some loss of continuity as soon as the section cracks; however, this is a limitation of the models. In theory, as soon as the section cracks, the crack propagates into the slab (see Figure 12). In reality, the crack starts at the bottom of the joint and propagates upward to the deck slab. Since the joint between the girder and the diaphragm is a cold (or construction) joint, the crack can follow this joint easily, but seems to be arrested when it hits the deck slab. Only when the connection is about to fail does the crack propagate into the slab. At this point, the cracked section at the joint matches that predicted by the model, and only then does the predicted loss of continuity occur. This experimental program tested a single system using I girders. This certainly does not cover the entire range of bridge types, so only limited recommendations can be made.
49 The 2nd edition of the AASHTO LRFD Specifications (12) states: 5.14.1.2.7c: If the calculated stress at the bottom of the joint for the combination of superimposed permanent loads, set- tlement, creep, shrinkage, 50% live load and temperature gradient, if applicable, is compressive, the joint may be con- sidered fully effective. Given the experimental evidence, this specification seems both reasonable and justified. In cases in which the calculated stress at the bottom of the joint for the combination of super- imposed permanent loads, settlement, creep, shrinkage, 50% live load, and temperature gradient is tensile, it might be jus- tifiable to treat the joint as partially effectiveâthat is, some portion of the live load would be carried as a continuous load and some portion as a simple-span load. However, the data from the experiments reported herein are insufficient to make a strong recommendation concerning degree of continuity as only a single structural system is considered. EFFECT OF DIFFERENT CONFIGURATIONS ON THE CONNECTION This experimental program tested six types of connec- tions, but the number of possible variations on these types of connections is large, as was seen in the survey. From the experimental results, it is possible to comment on some of these other configurations. Connecting Different Depth Girders One of the concerns with the positive moment connections is congestion in the diaphragm area. The bars or strands used to create the positive moment connections tend to be meshed, leaving little or no clearance between adjacent bars. The inter- action between the bars might limit the connection strength, but this was not found to be the case. The tests showed that the connections had adequate strength even though the dia- phragm area was congested. Anything that relieves the con- gestion will improve the connection. When girders of differ- ent depths are connected into opposite sides of the diaphragm, the positive moment connection steel from each girder will not lie in the same plane, leaving more clearance between the bars and reducing interaction effects. However, when the girders are the same depth, the forces in the positive moment connections are at the same level and will act in opposition to each other. When the girders are of different depths, the forces in the positive moment connec- tions are now offset by the difference in the depths of the gird- ers. This may lead to additional cracking in the diaphragm. The forces in the diaphragm area should be analyzed using an appropriate model, such as the strut-and-tie model, and addi- tional reinforcing should be provided as needed. Skewed Connections Skewed connections present a particular difficulty. The skewed connection, by nature, has asymmetries, and the exper- imental evidence suggests that asymmetry affects connec- tion performance. Unfortunately, no skewed connections were tested in this program, so it is difficult to make recommenda- tions. More research is required on skewed connections. DISCUSSION ON IMPLICATIONS OF SEISMIC EVENTS ON CONTINUITY CONNECTIONS As outlined in the research problem statement of this proj- ect, the ability of these types of connections to maintain continuity during and after seismic events has been ques- tioned and, therefore, needs to be evaluated. The authors were charged to prepare a discussion on the subject and to identify future research needs in this area. Connection failures are the most common type of damage in bridge structures. Seismic events tend to overstress the entire bridge and, particularly, the connections. For the sake of this discussion on the seismic implications, three types of continuity connections are identified: ⢠Type I: Continuity diaphragms with positive moment connections made of bent bars or bent strands and negative moment reinforcement in the composite deck or even in the continuity diaphragm. The gird- ers and/or continuity diaphragms are placed on bearing pads, which act as pin or roller supports. Therefore, the superstructure is basically isolated from the substructure in that no horizontal force is transferred from ground motions in the longitudinal or transverse direction. Vari- ations of this type of connection has been used in many states, particularly in Tennessee and Missouri. ⢠Type II: Integral continuity connections, in which the continuity diaphragm, the composite deck, and the columns are cast monolithically. This type of con- nection provides fully rigid continuity and, therefore, the superstructure must be designed for a portion of the plastic hinge moment of the columns. This type of con- nection is primarily used in California. ⢠Type III: Similar to Type I, but the diaphragm is connected to the pier cap and, therefore, limited shear and moment would be transferred during ground motion. This type of connection is primarily used in the state of Washington, where the bottom 2 ft of the dia- phragm is cast before the composite deck and the rest of the diaphragm. Previous studies on the subject are very few and not directly relevant. Ductility of continuity diaphragm-deck connections is an important issue. Failure at the interior supports could be brittle and catastrophic because of the smaller compression area at the bottom of the girder-diaphragm. NCHRP Report
50 322 (11) carried out a parametric study of the negative moment strength of composite sections and found the typical sections to have adequate rotational ductility to allow formation of a failure mechanism. The static analysis indicated that an upper limit on deck reinforcement equal to 50% of its balanced rein- forcement ratio would ensure sufficient ductility (of about 3.4) to develop the full-failure mechanism as well as to pro- vide enough deformation to give adequate warning of failure. Most recently, Holombo et al. (30) studied the continuity of precast/prestressed spliced-girder bridges under seismic loads for a typical construction in California. Since the gird- ers were spliced, continuity was present for both self-weight and live load. Two 40% scale-model continuous bridge struc- tures, one with bulb-T and the other with bath-tub girders, were tested under longitudinal seismic loads with up to eight and six times the first-yield displacement in the bulb-T and bathtub girder systems, respectively. Continuity was estab- lished by post-tensioning the girders together with strand. This is a common detail in spliced girders and a possible, but not common, detail for continuous-for-live-load systems. Although this system is not necessarily representative of con- tinuous-for-live-load systems, it is presented here to give some indication of the performance of a continuous system under seismic load. In the tests, significant ductility with only minor strength degradation was observed. An essentially elastic performance of the bent cap and superstructure was observed during the simulated seismic displacement cycles. Holombo et al. sug- gested approximating the superstructure seismic moments by giving two-thirds of the column plastic hinge moment to the girders adjacent to the column and the remaining one-third to the nonadjacent girders (30). Since seismic events predominantly develop horizontal in- plane loads, the continuity connection may be subjected to longitudinal or transverse movements. If the superstructure is isolated from the substructure through pin or roller sup- ports, the only implication of seismic event may be the unseat- ing of the girders in the longitudinal or transverse direction. Piers may experience a significant level of damage or be âlostâ after a major earthquake. At this stage, the connection is also expected to have cracked and damaged. From a life safety performance point of view, it would be important that the girders still be able to resist their self weight plus other superimposed gravity loads. Continuity connections do pro- vide enhanced structural integrity and redundancy for the entire system. However, the positive moment connection is not designed for carrying, nor should it be expected to carry, such moments for the two adjacent spans. Previous experiences with continuity connections have demonstrated the benefits of such redundancy. For example, in the late 1970s, in a bridge on State Route 1 over the Wolf River in Memphis, Tennessee, a bent supporting two spans failed completely due to scour, but the simple spansâwidened with continuous-for-live-and- composite-dead-load box beamsâonly sagged. Failure was averted by the alternate load path created by the continuous widened portion and the New Jersey parapet. To avoid catastrophic failures due to unseating of the gird- ers during seismic events, appropriate detailing must be devel- oped based on the expected longitudinal and transverse movements of the support columns and the pier cap. In Type II connections, the superstructure is actually designed for the proper transfer of moments. In Type III connections, as dis- cussed earlier, the dowel bars transfer some horizontal forces to the superstructure. However, rigidity of the girders in the longitudinal direction and that of the continuity diaphragm in the transverse direction should be more than adequate to resist such loads. Another important implication of seismic events is that vertical ground accelerations may induce a significant level of inertia force for near-fault bridges. However, such forces have to first overcome the gravity accelerations before devel- oping an uplift force at the continuity connections. Based on the above discussion, future research on the sub- ject should be focused on the analysis and testing of Type III connections, where some limited force transfer occurs, but perhaps is not designed for. Proper detailing must be devel- oped to avoid unseating of the girders. Use of shear keys or dowel bars may be considered. Effect of vertical ground motion on the vibration of the superstructure and the conti- nuity diaphragm also need to be studied. Finally, in this study, no experimental work was done that was specifically related to seismic behavior, but the experi- mental work that was done does provide some relevant infor- mation that may be applicable to seismic design. The results of the connection capacity tests showed that the bent-strand connections tend to slip under cyclic loads. This type of behav- ior might not be desirable in a seismic area. Tests on the fifth stub specimen demonstrated that if the girder ends are embed- ded in the diaphragm and the additional stirrups are placed in the diaphragm just outside of the girders, the connection will be more ductile after the pull-out failure occurs. This addi- tional ductility would be useful in seismic designs.
