Abstract
The scope of this effort is to evaluate the performance of a link slab transversely connecting press-brake-formed tub girders (PBFTG). Modular PBFTGs were developed by a technical working group within the Steel Market Development Institute’s (a business unit of the American Iron and Steel Institute) Short Span Steel Bridge Alliance, led by the current authors. This working group consists of stakeholders in the steel bridge industry, including mills, fabricators, service centers, industry trade organizations, universities, and bridge owners. A full-scale link slab transversely joining two PBFTGs was fatigue loaded simulating a 75-year fatigue life in a rural environment. Strain and deflection data was recorded and compared throughout the fatigue life to determine the link slab’s effectiveness. Results of this effort show the link slab detail performs adequately throughout its fatigue life.
Overview
This paper presents the experimental assessment of a link slab transversely joining press-brake-formed tub girders (PBFTGs). This technology consists of cold-bending standard mill plate width and thicknesses to form a trapezoidal box girder. Steel shapes are available in either hot-dipped galvanized or weathering steel options, either being an economical solution. Once the box has been formed, shear studs are welded to the top flanges. A reinforced concrete deck is cast on the girder in the fabrication shop and allowed to cure, resulting in a composite modular unit. The composite tub girder module is shipped to the bridge site and joined with an Ultra-High Performance Concrete (UHPC) longitudinal joint. The PBFTGs employed in this effort are of the same dimensions originally optimized in a project conducted by the Short Span Steel Bridge Alliance (SSSBA), as shown in Fig. 1.
Modular composite PBFTG.
PBFTGs have proven economically and structurally competitive in the short span bridge market through multiple laboratory experiments and field demonstrations by Gibbs, Underwood, and Roh [1–3]. The American Association of State Highway and Transportation Officials (AASHTO) Innovation Initiative selected the PBFTG bridge system as a 2021 Focus Technology and will continue to invest time and resources to encourage the adoption of the system throughout the United States. While the system is exceptionally efficient and economical, its applicability has currently only been utilized in simple span configurations.
The goal of this research program is to extend the applicability of PBFTG bridges into continuous span configurations. The issue with the use of PBFTGs in continuous spans is derived from the construction of the tub. As PBFTGs are formed from a single piece of standard plate, the bottom flange is a thin section prone to buckling in negative bending regions. Conventional continuous span design, where the bottom flange is continuous over interior piers, eliminates PBFTGs from consideration. Link slabs are a transverse deck level connection at piers between the decks of two adjacent spans, providing continuity. The deck is made continuous across the pier, but the supporting beams or girders are not connected. In addition to simpler designs consisting of simple spans instead of continuous spans, link slabs allow for prefabricated bridge elements to be implemented, further reducing the total cost of the bridge.
Prefabricated bridge elements and systems
The use of Accelerated Bridge Construction (ABC) in the United States has dramatically expanded over the last decade. ABC methods include bridge construction techniques utilizing innovative planning, design, materials, and construction methods in a safe and cost-effective manner to reduce the on-site construction time when building new bridges or replacing or rehabilitating existing bridges [4]. One of the many methods employed in ABC is the use of Prefabricated Bridge Elements and Systems (PBES). PBES are structural components of a bridge fabricated off-site and shipped to the bridge site to be assembled. Off-site construction of some or all bridge components significantly reduces the time lost to conventional construction practices such as extended concrete curing. In addition to reduced on-site construction time, PBES practices improve construction quality as fabrication occurs in a controlled environment. Efficiency of economy and quality is gained through the standardization of the construction process.
Press-brake-formed tub girders
The body of research regarding PBFTGs at West Virginia University began in late 2011 in conjunction with the SSSBA. Standard mill plate width and thicknesses are cold-bent in a large capacity press-brake to form a trapezoidal box girder. The optimum design of the sections was determined by Michaelson using a computer program to determine section properties of any configuration of PBFTGs [5]. Six different standard plate widths were considered, 1.52, 1.83, 2.13, 2.44, 2.74 and 3.05 (60, 72, 84, 96, 108 and 120 in.), and each plate width was evaluated in three different thicknesses, 11.1, 12.7, and 15.9 mm (7/16, 1/2, and 5/8 in.). Certain geometrical and material properties were held constant in the calculation, including the web slope ratio, the inside bend radii, the top flange width, deck dimensions, concrete compressive strength, and yield strength of the steel. The optimum depth of the noncomposite PBFTG was chosen for each cross-section to maximize the yield moment.
Destructive physical flexural testing was performed on two composite and two noncomposite PBFTG specimens to assess the behavior of the system. Each of the specimens were fabricated from 2.13 m (84 in.) wide×11.1 mm (7/16 in.) thick×10.67 m (420 in.) long plate and loaded in three-point bending. The composite specimens’ method of failure was governed by ductility, where the concrete deck crushed under approximately 1334 kN (300 kip) and 78.7 mm (3.1 in.) of deflection at midspan. The noncomposite specimens’ failure method of was governed by significant lateral torsional buckling under significantly smaller loads.
Long-term behavior of composite PBFTGs was analyzed by performing fatigue testing of two girders under three-point bending, as seen in Fig. 2 [6]. Tennant examined the performance of two separative PBFTG specimens constructed using specimens similar to those used by Michaelson [5] out of ASTM A709 steel, with one being hot dip galvanized and the other being uncoated. Two rows of 22.2 mm (7/8 in.)×102 mm (6 in.) shear studs were welded to each top flange, and a 152 mm (6 in.) concrete deck was cast on top of the specimens. The composite PBFTG modules were fatigue loaded simulating a 75-year design life in a rural environment. At predetermined cycle counts throughout the design life, a Service II load was applied to the modules to observe the performance of specimen throughout its design life. The project concluded the heat of galvanization had no adverse effect on the fatigue performance of PBFTGs.
Experimental fatigue testing of a composite PBFTG [6].
Research was extended by Kozhokin to evaluate the applicability of PBFTGs as modular bridge components and the performance of longitudinal joints between such modules [7]. The longitudinal joint tested consisted of ultra-high performance concrete (UHPC), which is cementitious material containing Portland cement, silica fume, quartz flour, fine silica sand, high range water reducer, water, and steel fibers. To ensure proper bonding between the normal concrete deck and the longitudinal UHPC joint, an exposed aggregate finish on the concrete deck would be required. Four separate slab edge treatment methods were explored, and the best results were produced by the application of a retarder to the shear key formwork and the removal of the concrete paste using a wire brush, as seen in Fig. 3 [7].
Roughened surface for the longitudinal UHPC joint [7].
The proposed longitudinal UHPC joint was evaluated between two PBFTG specimens with identical cross-sections as used by Michaelson and Tennant [5, 6]. The joint was fatigue loaded in three-point bending by applying the Fatigue I load at midspan of one of the composite specimens to simulate infinite fatigue life. At predetermined cycle counts throughout its design life, strains were recorded in both the directly and indirectly loaded modules to determine if the load was being adequately distributed. The research showed the UHPC longitudinal joint satisfactorily transferred load from the directly loaded girder to the indirectly loaded girder.
Conventional continuous span bridges have expansion joints at the end of spans over abutments and piers. The joints allow superstructure movement caused by shrinkage, creep, thermal effects, and rotation due to normal vehicular loading. Bridge owners have historically struggled to maintain these joints as leakage of water carrying corrosive chemicals and debris can lead to extensive and premature deterioration of both the substructure and superstructure elements. Bridge deck joints can be removed at abutments by using integral or semi-integral abutments or deck extensions. Bridge deck joints can be removed at piers by utilizing continuous for live load structures or link slabs.
Link slabs are a transverse deck level connection at piers between the decks of two adjacent spans, providing a jointless bridge without continuity [4]. The deck is made continuous across the pier, but the supporting beams or girders are not connected. In addition to simpler designs consisting of simple span instead of continuous spans, link slabs can allow for prefabricated bridge element to be implemented, further reducing the total cost of the bridge.
The concept of removing some expansion joints in favor of using continuous bridge decks over simply supported girders began gaining attention in the 1980 s and 1990 s. Gastal began exploring the elimination of structural joints by casting a fully continuous deck over simply supported girders [5]. The researcher developed a finite element model to capture the response of jointless bridge decks. The results of his analytical study found the behavior of jointless bridge decks over non-continuous girders was significantly influenced by the support conditions of the girders. The results also showed the maximum capacity of the specimens was controlled by yielding in the reinforcing steel, but ultimate capacity was significantly higher than that of a conventional non-continuous beam.
Caner and Zia conducted experimental and analytical testing on two link slab specimens, one on a continuous reinforced concrete deck cast on two simple-span steel beams, and the other on a similar deck cast on two simple-span reinforced concrete beams [9]. The experimental testing on the steel specimen was performed on two 6.25 m (246 in.) long W12x26 steel beams with a 50.8 mm (2 in.) gap between the adjacent ends. The 610 mm (24 in.) wide×102 mm (4 in.) thick concrete deck was made composite to the beams with shear connectors welded to the top flange of the beams. Elastic loading was applied at midspan of the specimens up to 40% of the estimated ultimate load capacity to observe the behavior to the specimens under varying support conditions while reaction forces, deflections, and rotations were measured. The results were benchmarked against a finite element analysis model and a design procedure for link slabs was developed.
Due to the unique structural demands on link slabs, Li et al. explored the use of Engineered Cementitious Composites (ECCs), a high-performance cementitious composite with high tensile strain capacity, high tensile strength, and excellent post-crack strain hardening, in the use of link slabs [10]. Specifically, the high ductility provided by ECC allows for small crack widths in the unique loading seen by link slabs. The researchers proposed a modification to the link slab proposed by Caner and Zia [9], which included an additional transition zone outside the standard link slab where the concrete is poured with the link slab, but shear studs are provided. The original link slab detail included termination of the debond zone and additional reinforcement spliced to the existing reinforcement at the deck slab/link slab interface. This imposes a high stress concentration and ultimately makes the interface the weakest part of the link slab detail. The proposed detail, shown in Fig. 4, includes a transition zone of 2.5% of each adjacent span in addition to conventional debonded zone of 5%. The addition of shear connectors in this transition zone allows the development of composite action over a region instead of at the interface, reducing the stress concentration.
Link slab design detail including a transition zone in addition to the debonded zone.
The researchers performed monotonic and cyclic fatigue testing on the improved link slab design detail utilizing ECC. The physical testing of the link slab was conducted on inverted specimens with supports at the inflection points to allow for larger scale testing. The specimens were statically loaded to 40% of the rebar yield strength. Following the static loading, the specimens were fatigue loaded to 100,000 cycles with the static load being the mean and the maximum load corresponding to that which causes a maximum allowable link slab rotation. The researchers concluded ECC was a suitable material in link slabs due to its high ductility and low flexural stiffness.
Setup of experimental testing
In order to determine the effectiveness of link slabs joining two simply supported PBFTGs over a pier, experimental fatigue testing of the system was conducted at the Major Units Laboratory at West Virginia University, as shown in Fig. 5. The cross-section of the noncomposite PBFTGs, before concrete casting, is identical to previous testing performed by Michaelson, Tennant, and Kozhokin [5–7]. With a large capacity press-brake, the PBFTGs used in this study were formed from 2.13 m×11.1 mm (84 in.×7/16 in.) plate resulting in a 584 mm (23 in.) depth. The experimental testing was performed on two separate PBFTG specimens transversely joined by a link slab, as shown in Fig. 6. One specimen consisted of uncoated ASTM A709 steel, while the other consisted of galvanized ASTM A709 steel. As described by Tennant [6], the coating system was found to not have an impact on the fatigue performance of PBFTGs. Specimens with different coating systems were used due to lack of availability of PBFTGs at the time of testing. For the uncoated specimen, the boundary condition furthest away from the link slab simulated a pinned condition and the boundary condition below the link slab simulated a roller condition. For the galvanized specimen, the boundary condition furthest away from the link slab simulated a roller condition and the boundary condition below the link slab simulated a pinned condition. To resist bearing forces, 19 mm (3/4 in.) bearing plates were welded to the girders 76 mm (3 in.) from the edge of the girder at support locations.
Test setup schematic.
Isometric view of the SIP metal formwork and shear studs.
Stay-in-place (SIP) metal formwork was utilized between the top flanges of each specimen. 51 mm (2 in.) deep pans were utilized on the uncoated specimen and 76 mm (3 in.) deep pans were utilized on the galvanized specimen. The difference in SIP metal formwork flute depth for uncoated and galvanized specimens was due to fabrication error. The difference between the composite moment of inertia of the galvanized specimen and the uncoated specimen was slightly over 1%, which was deemed acceptable for the purpose of this test. The SIP metal formwork ran from the exterior support of each girder longitudinally until approximately 229 mm (9 in.) from the interior support, allowing for a purely unbonded region to develop between the concrete deck and the steel PBFTGs at the interior support region. To develop composite action, two rows of 22.2 mm×152 mm (7/8 in.×6 in.) shear studs, longitudinally spaced at 305 mm (12 in.) were welded through the SIP metal formwork to each top flange, as seen in Fig. 6. The final four bottom flutes near the interior support did not have shear studs and were instead plug welded to connect the formwork to the girder. This was performed to aid in the development of a transition zone between the normal concrete deck and the link slab.
Wooden concrete formwork was placed around the two longitudinally placed 7.62 m (25 ft.) PBFTGs. Contrary to previous testing by Michaelson, Tennant, and Kozhokin [5–7], the concrete deck in this research was made to simulate the thickness of full scale 203 mm (8 in.) bridge deck. During deck placement for the concrete deck, four 152 mm (6 in.) diameter cylinders were poured for future compressive strength testing. The concrete was vibrated after placement to minimize air pockets and honey combing. Wet burlap was placed over the curing concrete for 14 days and allowed to dry cure for an additional 14 days.
The American Association of State Highway and Transportation Officials’ Load Resistance and Factor Design Bridge Design Specifications’ (AASHTO LRFD BDS) Article 9.7.2 was used to design the reinforcement pattern with the empirical deck method [11]. The bottom mat of reinforcing consisted of #5 rebar with a bottom clear cover 25.4 mm (1 in.) between the mat and the 76 mm (3 in.) SIP metal formwork. The top mat of reinforcement consisted of #4 rebar with a top clear cover of 60.3 mm (2 3/8 in.) between the top of the concrete and the top of the transverse bars. The longitudinal rebar of both layers was spaced at 305 mm (12 in.) with an edge spacing of 51 mm (2 in.). The transverse rebar of both layers was spaced at 305 mm (12 in.) with an edge spacing of 51 mm (2 in.), which coincided with placement of the shear studs.
A combination of the methodologies provided by Caner and Zia and Li et al produced the size and layout of the link slab reinforcement [9, 10]. The bottom mat of reinforcement located 57 mm (2¼ in.) above the 51 mm (2 in.) deep SIP formwork consisted of 4 #5 bars spaced 305 mm (12 in.) on center and the top mat of reinforcement, located 127 mm (5 in.) above the 51 mm (2 in.) deep SIP formwork consisted of 4 sets of 2 #7 bars tied together spaced 305 mm (12 in.) on center. As the use of the link slab does not behave similarly to two simple spans or continuous spans, the calculation of the moment at the link slab is not easily calculated. The rotation in the link slab was calculated to be 0.0029 radians (0.16 degrees) from the applied load by assuming the link slab was simply supported at the deck/link slab interface and under three-point bending (load applied by the supports in the center of the link slab). The moment carried in the link slab was then calculated from the rotation. From this moment, the required area of steel was calculated to be 2307 mm2/m (1.09 in2/ft) and rebar size was determined to be 2 #7 = 2540 mm2/m (1.2 in2/ft). Another key difference in the construction of the link slab was the need to debond the concrete deck from the underlying steel girders. This was achieved by placing plywood over the open ends of the tubs, not covered by the SIP metal formwork, and covering the entire bottom face of the link slab with standard roofing paper. The roofing paper allows movement of the link slab independently from the underlying girders. The corrugation of the SIP formwork with the roofing paper allowed for the transition zone to develop between the flat link slab and the main deck, as seen in Fig. 7.
Isometric view of the link slab transition zone.
Foil-resistor uniaxial strain gauges (Micro Measurements CEA-06-250UN-350) measured strain throughout the testing of the specimens. They recorded strain at four different cross-sections on the main span specimens and on the rebar in the link slab. Nine strain gauges were used at each cross-section, three placed along the bottom flange and three placed along each web to capture longitudinal bending strain. A typical cross-section of the strain gauge layout can be seen in Fig. 8. The gauges were placed at following longitudinal locations: (1) quarter span toward the exterior support of the uncoated specimen; (2) offset 1.17 m (46 in.) from midspan toward the interior support of the uncoated specimen; (3) quarter span toward the interior support of the galvanized specimen; (4) offset 1.17 m (46 in.) from midspan toward the exterior support of the galvanized specimen. The strain gauges connected to a Micro-Measurements Model 5100 Scanner, which in turn utilized StrainSmart software to record the strain data [12].
Strain Gauge Layout.
The load was applied at midspan of the uncoated specimen by an MTS 1468 kN (330 kip) servo-hydraulic actuator, and the load was applied to the galvanized specimen by an MTS 489 kN (110 kip) servo-hydraulic actuator. Large spreader beams were placed under the head of each actuator to reduce localized concrete crushing due to concentrated load effects. Displacement measurements were also recorded using the MTS servo-hydraulic actuators. Displacement readings were recorded sporadically throughout the fatigue loading.
The loads required to induce the Fatigue I moments into the system were computed prior to the testing of the specimen. The Fatigue I load combination reflects load levels found to be representative of the maximum stress range of the truck population for infinite life design and is determined by using AASHTO LRFD BDS Tables 3.4.1-1 and 3.4.1-2 [11].
The Fatigue I moment is applied to the system through a load of 387 kN (87 kip). Four assumptions were made in the determination of the number of fatigue cycles for the system. First, the average daily traffic (ADT) was assumed to be 800 vehicles, representing low to moderate volume roads. Second the bridge assumed to be located on a rural non-interstate highway. Third, the design life is assumed to 75 years. Fourth, the bridge is assumed to have two lanes available to traffic. AASHTO LRFD BDS Table C3.6.1.4.2-1 states the fraction of trucks in traffic on a rural non-interstate highway is to be 15% of the ADT [11]. With two lanes available to traffic, AASHTO LRFD BDS Table 3.6.1.4.2-1 states the fraction of truck traffic in a single lane is equal to 85% of the average daily truck traffic, or p = 0.85 [11]. Therefore, the number of fatigue cycles representing a 75 year design life was determined as follows:
Number of Cycles = 800 (ADT)×0.15 (representative of 15% truck traffic)×0.85 (p)×365 (days/year)×75 (years)=2,792,250 cycles
A static base test was conducted before fatigue loading began. The load was increased in ten 38.7 kN (8.7 kip) intervals and strain data was recorded at each interval. The static load was allowed to sit on the specimen for five minutes to allow for load effects to settle before data was recorded and the load was increased. The Fatigue I moment was reached twice during each static test and the readings from the strain gauges were averaged. Periodically throughout the fatigue testing, the cyclic load was halted, and a static test was performed. The purpose of these static tests was to assess the performance of the system throughout its projected service life.
After approximately 300,000 fatigue cycles, composite action was lost between the concrete deck and the galvanized PBFTG. The lateral movement between the concrete deck and the top flange of the girder suggested a failure in the welds between the girder and the shear studs.
A procedure provided by Kreitman et al. was adopted to retrofit the galvanized specimen with post-installed shear connectors composed of adhesive anchors [13]. Four threaded rods per cross-section, two in each top flange, were placed in the weak location of the SIP formwork as the strong location had the failed studs. The following procedure was followed to install each of the connectors.
First, a 25.4 mm (1 in.) diameter hole was drilled through the top flange of the PBFTG at the connector location using a portable magnetic drill press. Second, through the hole in the top flange, a 23.8 mm (15/16 in.) diameter hole was drilled 203 mm (8 in.) into the concrete deck. This smaller diameter hole in the concrete deck was to allow the head of the tungsten carbide drill bit to easily fit into the hole made in the steel. Third, the hole was cleaned by alternating abrasion with a wire brush and compressed air until no dust was removed by the compressed air. Fourth, Hilti HIT-HY 200-A structural adhesive was injected into the hole. Care was taken to ensure the hole was filled from the top down, so no air bubbles were present. Fifth, a 305 mm (12 in.) long threaded rod, with the nut and washers already located on the rod, was inserted into the hole using a twisting motion to ensure the adhesive filled the threads. Sixth, the rods were held in place for five minutes and then allowed to cure for one hour. Finally, the nuts were further tightened to not allow movement between the threaded rod and the steel girder.
This retrofit performed exceptionally well to recover composite action. Fatigue testing continued for another 900,000 cycles, totaling to 1,200,000 cycles when composite action was lost in the uncoated specimen. Instead of performing a full retrofit on the uncoated specimen, the load applied to both specimens was reduced to 311 kN (70 kips) to achieve approximately the same deflection in the girder and the same rotation in the link slab.
Experimental results
The concrete compressive strength determined from the concrete cylinders from the regular deck and link slab are summarized in Table 1. The concrete in the deck and the link slab were purposefully made weaker than concrete normally found in bridge decks to ensure testing of the specimen could be completed given the capacity of the actuators available. The first transverse cracks in the link slab appeared immediately after loading began during the first static test. Additional link slab cracking was recorded up to 200,000 cycles after which no additional cracking in the link slab occurred. The maximum recorded crack size in the link slab was 0.64 mm (0.025 in.).
Concrete compressive strength results
Concrete compressive strength results
The strains recorded in the system were converted to stresses utilizing a method adopted by Helwig and Fan [14]. The longitudinal stresses induced by axial forces and bending moments are assumed linearly distributed across the cross-section of the members. Using this method, strain readings from only three non-collinear points are needed to determine the distribution of stresses in the cross-section.
To further reduce the collected data, three-dimensional linear regression techniques based on least square regression were employed. The moments induced into the system were calculated using a procedure developed from Michael, Imhoff, McDaniel and Frederick [15]. These values were used to back-calculate the applied load using the equation for moment induced by a point load at midspan. It should be noted the amount of support rigidity provided by the link slab is small but noticeable.
Figures 9 and 10 provide a summary of the comparisons between the loads applied to the girders and the loads computed using the collected strain data and three-dimensional linear regression. This approach was taken to show consistency of link slab performance over the course of fatigue testing. By comparing the applied load to the back calculated load derived by strains, it can be shown the system’s behavior did not substantially change through the course of the fatigue testing. As shown, the girders’ and the link slab’s behavior remained constant throughout the 75-year life. Select static tests have been removed due to recording errors. A slight discrepancy can be seen in Fig. 9 where the back-calculated vertical load is higher than the applied load. This is attributed to the rigidity retained in the galvanized specimen when composite action was lost in the uncoated specimen, and the loads were adjusted accordingly. Additionally, as shown in Fig. 11, the deflection of both specimens remained consistent throughout the fatigue life of the experiment, further showing the link slab continued to act as a stable structural element, even after composite action was lost in both PBFTGs at different times.
Correlation of the applied load to the back calculated load of the galvanized specimen.
Correlation of the applied load to the back calculated load of the uncoated specimen.
Midspan deflection of each specimen throughout testing.
The scope of this effort is to evaluate the performance of a link slab transversely connecting PBFTGs through full-scale laboratory testing. The link slab and the PBFTGs were fatigue loaded simulating a 75-year design life in a mid-volume rural traffic environment. Deflections were recorded throughout the fatigue loading, and static testing was performed by inducing the Fatigue I moment into the link slab intermittently to record strains through the range of the fatigue loading.
The results can be split between pre- and post-loss of composite action in the uncoated specimen. The data obtained from the strain gauges were linear and consistent for each phase of testing. Transverse cracks appeared on the link slab during the first phases of testing, however, as these cracks arrested at the top mat of reinforcement and no additional cracking occurred after the first 200,000 cycles, the cracking in the link slab was deemed acceptable. Overall, the link slab performed satisfactorily as a transverse connection between PBFTGs. With further long-term performance monitoring and analytical modeling, link slabs between PBFTGs can be an effective bridge type in short continuous span applications.