Fatigue Performance of Steel–UHPC Composite Bridge Deck System with Large Longitudinal Rib

,


Introduction
Orthotropic steel decks (OSDs) have been widely used in bridges due to its lightweight, easy to assemble, and high bearing capacity [1,2]. However, the fatigue cracking, which signifcantly reduces the service quality and severely impairs the durability of OSDs, is prominent and frequent [3,4]. Te fatigue cracking of OSDs is mainly caused by combined action of intrinsic and extrinsic factors. Te former includes manufacturing defects, weld residual stress, and weld geometry, the latter is the trafc loads [5,6]. Te fatigue cracks at the rib-to-deck joints, rib-to-diaphragm joints, and the cope holes in the diaphragm are shown in Figure 1.
Scholars have made considerable eforts to promote the fatigue behavior of OSDs [7][8][9][10]. On the one hand, the fatigue resistance of OSD can be enhanced by increasing the deck thickness and improving the welding techniques such as double-side weld [7,8]. On the other hand, the fatigue resistance of OSD can be signifcantly extended using the ultra high performance concrete (UHPC) to form a composite deck, which increases the overall stifness of the deck system and reduces the nominal stress range of fatigueprone details without signifcant increases in the dead weight [9,10]. Recently, a steel-UHPC composite bridge deck ( Figure 2) composed of an OSD with large longitudinal ribs and UHPC layer was proposed for further optimization [11][12][13]. Te large longitudinal rib has opening widths of 400 mm to 450 mm and heights of 300 mm to 330 mm. Compared to the traditional U-rib, it reduces the length of rib-to-deck weld seam and the associated welding cost by 30% to 40%, then the risk of fatigue cracking of rib-to-deck joint is decreased [11,12]. In addition, the large longitudinal rib can also increase the fexural stifness of the deck system in longitudinal direction, thereby alleviating the tensile stress of the UHPC layer in the negative moment region under the vehicle load [13].
To use a steel-UHPC composite bridge deck as a fatigueresistant technology for OSDs, the fatigue performance of OSDs as well as the cracking behavior and fatigue resistance of the UHPC layer must be evaluated. Qin et al. [14] and Abdelbaset et al. [15] studied the efect of UHPC layer on the stress state of fatigue-prone details of OSDs through on-site measurement and full-scale model test, respectively. Te results showed that in steel-UHPC composite bridge deck, the stress range in rib-to-deck joint below the endurance limit, while that of the rib-to-diaphragm and rib splice are still higher than the endurance limit. Feng et al. [16], Lu et al. [17], and Wei et al. [18] investigated the cracking behavior of the UHPC layer under static and cyclic load, and it is conluded that the cracking of the UHPC layer was inevitable and the crack width will increase under the action of the cyclic load. Liu et al. [12] conducted a fatigue test on a fullscale steel-UHPC composite bridge deck model, observing that the UHPC layer cracked under the design fatigue load of approximately 200 000 cycles. Te experience with the use of UHPC composite structures over time indicates that the cracking width and number of cracks is mainly infuenced by the fber content and reinforcement ratio [19,20]. However, studies on the fatigue resistance of composite bridge decks with large longitudinal ribs are limited. In addition, the use of a large longitudinal rib leads to an increase in span of the deck plate. Typically, this increase generates considerable tensile stress in the transverse direction of the steel-UHPC composite bridge deck, rendering the deck cover prone to fatigue failure under external loads. Terefore, a study of the fatigue performance and damage accumulation process of steel-UHPC composite bridge decks with large longitudinal rib is necessary.
Tis research aims to evaluate the fatigue performance of steel-UHPC composite bridge deck with large longitudinal ribs. First, two types of specimens were tested to assess the failure mode and degradation mechanism of the deck system along the longitudinal and transverse directions. Te loaddefection, load-strain, and cracking behaviors of specimens under cyclic loads were studied by fatigue test; then, the fatigue resistance of the composite bridge deck with a large longitudinal rib was evaluated. In addition, the efects of fatigue damage on the UHPC layer and studs with respect to the fatigue performance of the steel-UHPC composite bridge deck were investigated. Figure 2, the mechanical behavior of the OSD (or UHPC layer-stifened OSD) exhibits directional anisotropy due to diferent stifness properties in the orthogonal directions [21]. Generally, three structural component systems must be considered in the design and analysis of OSDs: (1) System I-the deck as part of the main carrying member, (2) System II-the stifened steel deck consisting of longitudinal ribs, transverse diaphragms, and the deck plate, (3) System III-the deck plate supported by the rib wall of longitudinal ribs. After the introduction of the UHPC layer as part of the deck, the stifness of the orthotropic bridge deck in stress systems II and III signifcantly improves. Under vehicle loads, the UHPC above the transverse and longitudinal ribs will be in tension when UHPC participates in the bridge deck system under systems II and III. Terefore, the experiments have  been conducted on two specimens designated as transverse and longitudinal specimens according to the stress characteristics of the composite bridge deck in systems II and III. Te dimensions of the transverse and longitudinal specimens are shown in Figures 3 and 4, respectively. Te specimens have the same transverse section with a width and depth of 2160 and 770 mm, respectively; however, their lengths difer, i.e., 900 and 7000 mm for the transverse and longitudinal specimens, respectively. Te thickness of the deck plates of the specimens is 14 mm. Te diaphragms are 16 mm thick and spaced at 3000 mm. Te height, top width, and thickness of the U-rib are 330, 400, and 8 mm, respectively. Te 70 mm-thick UHPC layer was poured on the deck and connected to the OSD through headed studs with a 200 mm spacing. Te height and diameter of the studs are 50 and 16 mm, respectively. Te diameter of steel rebars spaced at 70 mm is 12 mm. Te fabrication process of the specimens is shown in Figure 5.

Material Properties.
Te steel used for the test specimens were fabricated from Q345qD steel with a nominal yield strength of 345 MPa [22]; HRB400 grade steel rebars were used in the reinforcing mesh. Te yield and ultimate strengths of the steel rebars were 400 and 570 MPa, respectively [23]. Steel fbers with diameter, length, and volume ratio of 0.2 mm, 13 mm, and 2%, respectively, were employed in UHPC. After 24 h of natural curing, the castings were maintained at 80°C with 90% humidity for 3 d. Table 1 lists the UHPC ingredients and mix proportion. Te mechanical properties of UHPC are summarized in Table 2. Te test methods are consistent with techniques employed in previous research [24,25].

Test Setup, Instrumentation, and Loading Protocol.
Te feld test setup of specimens is shown in Figure 6. Te test specimens were connected with supporting beams fxed on the ground. A servo-hydraulic loading system with a load capacity of 1000 kN was used to apply the cyclic load. Te loading position of the transverse specimens is shown in Figure 3. Te two loading positions are symmetric with respect to the center of the specimen. Te middle spacing is 760 mm, and the single loading area is 270 mm (longitudinal direction) × 320 mm (transverse direction) [26]. Te loading position of longitudinal specimens is shown in Figure 4. Te two loading positions are symmetric with respect to the middle diaphragm. Te middle distance is 1200 mm, and the single loading area is 200 mm (longitudinal direction) × 600 mm (transverse direction) [27].

Advances in Materials Science and Engineering
For the rib-to-deck joints on the transverse specimen, strain gauges are fxed to the bottom side of the deck plate and arranged along the weld root to measure the transverse strain distribution, as illustrated in Figure 7(a). Te strain gauges are placed close to the weld toe of the rib-to-diaphragm joints of the longitudinal specimen, as shown in Figure 8 Te steel rebar mesh is mainly used to test stress and its distribution along the longitudinal direction of the bridge. Tis stress indirectly refects the deterioration characteristics of the concrete layer. A pair of strain gauges attached to the opposite sides of the headed studs was used to measure the bending strain, and six strain gauges were utilized to measure the steel rebar strain on the transverse specimen ( Figure 7(c)). A total of 40 resistance strain gauges are arranged in 8 rows and 5 columns on the longitudinal rebars of the steel mesh; they are symmetric across the diaphragm section on the longitudinal specimen, as shown in Figure 8(c).
Tree cyclic loading stages, labeled as Stages I, II, and III, were executed sequentially with a constant amplitude load range. Te minimum load was set at 10 kN and remained constant during the entire experimental process. Table 3 summarizes the applied cyclic loads to each stage.
Te fatigue load amplitude of the tests in Stage I is calculated according to the standard wheel weight of 60 kN and impact coefcient of 1.3. Wheel type C was selected for transverse specimens, as suggested in [26], and standard fatigue model III was selected for longitudinal specimens based on literature [27]. Te positions of the applied loads A � πr 2 are shown in Figures 3 and 4. Te fatigue load applied to the destructive test in Stage II is based on the frst stage load considering an overload factor of 1.5. Te upper limit of the frst stage test load is increased. If the number of cycles continues to increase up to 4 × 10 6 , the specimen does not undergo fatigue failure, and the accelerated failure test is implemented. Tat is, considering an overload factor of 2.0 based on the load in Stage II, the upper limit of the test load in this stage is increased until the number of cycles reaches 6 × 10 6 . Te list in Table 3 indicates that the objective of Stage I is to verify the durability of the composite bridge deck under fatigue. Te other stages are implemented to evaluate the fatigue life and failure mode of the specimens.
A static loading test was conducted to acquire the initial mechanical properties of each specimen before they are subjected to fatigue loading. Te test was periodically conducted after the completion of a certain number of cycles to record the deterioration process of the test specimens.      Figure 9. When the number of loading cycles reached 2 000 000, visible cracks were not observed on the UHPC layer in Stage I. Te strain in the UHPC layer of the test specimens was stable, and the strain was lower than the elastic limit, which was in the elastic state of the UHPC material. When the loading cycles reached 2 050 000, the initial crack (with a maximum width of 0.01 mm) developed on the UHPC layer surface above the joints of the rib, deck,  [19,20]. Due to the low stress range in the details of the OSD that are not vulnerable to fatigue under the frst two loading stages, the fatigue test in Stage III was implemented using a fatigue load that was 2.0 times that in Stage II to accelerate the fatigue damage to specimens. After the cracks appeared, distinct cracks propagated along the longitudinal direction  Figure 10. Te variation in rebar stress under fatigue loading is shown in Figure 11. Te stress in transversely loaded rebars in the concrete layer remains stable in Stage I. In Stage II, with the appearance of cracks, the stress in the rebars increased to a certain extent; however, it swiftly became constant. With the crack propagation in the UHPC, stress redistribution occurred on the UHPC layer, and the surface stress was continuously transferred to the rebars. Te rebar stress increased with the number of fatigue cycles under Stage III.
By monitoring the strain distribution around the welded joint, monitoring the fatigue crack initiation and propagation process also became possible [28]. When the number of fatigue cycles reached 4 250 000, strain gauge U2-DS-LC indicated a sharp drop in strain (Figure 12), indicating that a fatigue crack initiated at the strain measurement point. Te variation in strain distribution along the weld root during the fatigue test is shown in Figure 12. Te strain redistribution process shows that the fatigue crack propagation is symmetric relative to the central section of the diaphragm. Te fatigue crack formation on the welded joint as reported in literature is shown in Figure 13 [12]. Te crack initiates at the root of the rib-to-deck and diaphragm joint and extends to the deck thickness.
Te variation laws of bending strains in the studs are shown in Figure 14. Te strains were stable at the beginning of the fatigue test, indicating that the mechanical properties of the studs had no distinct degradation. Te stress redistribution during crack propagation increased the bending strain in the studs as strains rapidly increased in Stage III. Te cyclic load was terminated at 6 000 000 cycles. Te failure process of the transverse specimen can be U1 U2 Stage I Stage II Stage III Figure 9: Crack distribution in UHPC layer. increase. Te displacements versus the number of cycles are described in Figure 15. Te increase in displacements indicated that damage occurred in the headed studs and Advances in Materials Science and Engineering 9 UHPC layer, leading to a reduction in transverse stifness. Te variation law of local stifness at the loading positions is shown in Figure 16. Te fatigue endurance of the OSD must surpass 2 000 000 cycles under specifc load models in accordance with current design codes [29,30]. Te transverse specimen underwent 2 000 000 cycles without evident fatigue damage and performance degradation. Due to the excellent crack width control capability of UHPC, the test specimen also satisfed the crack width limitation. Based on the foregoing, the transverse specimen satisfes the design requirements in terms of fatigue and durability. However, the infuence of fatigue damage on the mechanical behavior of test specimens is more evident; this damage is caused by UHPC cracking and stud fracture. Te addition of loading cycles (Stages II and III) decreases the transverse rigidity of the transverse specimen and increases the peak strain in welded details.

Fatigue Failure Mode.
After the longitudinal specimen underwent 750 000 cycles, the initial crack in the UHPC layer occurred at the region above the intermediate diaphragm (Stage I in Figure 17). When the number of loading cycles was increased, multiple cracks were observed between the two loading positions and extended horizontally along the specimen. During the early stage of each load increase, many small cracks were observed. When the load increase reached Stage II, cracks mainly propagated near the middle area. Cracks also appeared at the edge of the specimen until the load increase reached Stage III. Te variation laws of the maximum crack width under the fatigue load are shown in Figure 18. Te maximum crack width remained virtually constant at each stage; this width was controlled within the 0.05-mm maximum limit throughout the stage I and II, but not in the stage III.
Te response shown by the number of cycles-strain curve ( Figure 19) indicates that crack initiation does not occur in the rib-to-diaphragm joints, and no signifcant mechanical degradation occurred in the longitudinal specimen. When the loading cycles reached 6 000 000, the equivalent stress range of the rib-to-diaphragm joints exceeded the 71 MPa fatigue strength at 2 000 000 cycles specifed in Eurocode 3 [30]. Because the equivalent stress range of the welded joint exceeds the fatigue strength, fatigue cracking was expected. Te longitudinal fatigue performance of the steel-UHPC composite deck seems to mainly depend on the fatigue strength of the rib-to-diaphragm joints.

Mechanical Response.
Te displacements versus the number of cycles are shown in Figure 20. Te local stifness of the loading point varies with the number of actions, as shown in Figure 21. Te local stifness of the composite bridge deck below the loading point is defned as the ratio of applied loads causing the vertical displacement. In the frst two stages, the local stifness of the specimen remained virtually constant with the increase in the number of actions and degenerated during the loading process of Stage III. Te reduction in longitudinal stifness can be attributed to the UHPC cracking because it was the only damage observed during the fatigue test ( Figure 21).
Distinct fatigue damage and mechanical degradation were not observed in the longitudinal specimen after 6 000 000 fatigue cycles under 1.5-3.0 times the designated vehicle loads. Terefore, the specimen has sufcient antifatigue capacity during its life expectancy and can withstand repeated trafc loads.
In the composite bridge deck, the stress concentration efect at the end of the weld at the rib-to-diaphragm joints is prominent; furthermore, the weld quality at the end of weld is difcult to ensure. Accordingly, this structural details with severe fatigue damage in the system are analyzed. Te test results shown in the experiment indicate that the local strain of structural details is constant, and no fatigue crack is observed during the test. However, considering the essential attribute of large discreteness of fatigue strength of structural details, focus must remain on the fatigue performance of these details in an actual bridge structure.

Fatigue Strength.
Based on the test results of longitudinal and transverse full-scale models, the composite deck structure system with large longitudinal ribs has satisfactory fatigue resistance under longitudinal and transverse fatigue loads. After 200 000 cycles of fatigue load as specifed in the standard, 1.5 times the standard fatigue load was implemented; the structural mechanical properties remained stable. Although the UHPC layer cracked to some extent, the crack width was smaller than the nominal initial crack index (0.05 mm). In the accelerated failure test stage (Stage III), the longitudinal fatigue failure mode of the composite bridge deck structure with a large rib is manifested by UHPC layer cracking; the crack width exceeds the limit. In the transverse direction, the fatigue failure mode is indicated by UHPC layer cracking in which the crack width also exceeds the limit and weld root fatigue cracking at the details of the weld joint between the deck and longitudinal rib. In the full-scale longitudinal specimen test, although no fatigue cracking occurs at the rib-to-diaphragm joint, the stress level exceeds the fatigue limit in the accelerated failure test stage.
Te nominal stress approach was applied to determine the fatigue strength of welded joints according to the measured fatigue strain range. Based on existing research, the fatigue failure criterion of rib-to-deck joints was defned as 25% of the drop in strain [31]. Te points determined by the fatigue test results are located above the S-N curves [30], indicating that the fatigue performance of the welded joints fundamentally satisfes the design requirement ( Figure 22).

Discussion
Te test results show that the damage to the UHPC layer and head studs degrades the fatigue performance of the transverse specimen. Te infuence of damage on this performance can be determined by gaining understanding of the baseline mechanical properties of the transverse specimen when the UHPC layer and studs are not damaged through fnite element (FE) analysis.  To verify the accuracy of using the FE method for determining the stress feld in the welded joints of the test specimens, numerical simulation of the transverse specimen was conducted using ANSYS software. Due to the symmetry of load and geometry, a semimodel is established to impose symmetric boundary conditions on the symmetric surface, as shown in Figure 23(a). Te bottom of the FE model is fxed, which is consistent with the test specimen. Te material constitutive laws of UHPC were identical to those in [32] and the adopted material parameters were consistent with the material properties in Section 2.2. Te steel plate and rebars were simulated with elastic modulus of 200 GPa. Te UHPC and steel plate were simulated using solid element (Solid45), and the steel rebars were simulated by truss element (Link8). Te nodes of rebars were coupled with the UHPC layer. As shown in Figure 23(b), surface-to-surface contact element (Conta173 and Targe170) was used to model the normal and tangential interaction behavior between the UHPC and deck plate. Te normal behavior was set as hard contact, and the tangential behavior was defned by a friction coefcient of 0.4. At the position of headed studs, two spring elements (Combin39) were used to simulate the transverse and longitudinal load slip behavior. Te properties adopted for the headed studs were acquired from literature [33]. In vertical direction, the vertical displacement of the coincident nodes at the location of studs was coupled. A minimum refnement grid size of 0.5 mm is used to resolve the stress concentration problem near the welded joint. Figure 24. Te simulation results are consistent with experimental data, and the error between the two is within 10%, indicating the efectiveness and accuracy of the FE method.

Evolution of Fatigue Damage. Te comparison between simulation and experimental results is shown in
Te fatigue failure criterion of the rib-to-deck joint was defned as 25% of the drop in strain, which can be calculated using equation (1): where m represents the S-N curve parameters; ∆σ i is the loading stress; ∆σ eq is the equivalent fatigue strength; N eq � 2 000 000 cycles; and n i is the number of load cycles.
Te accumulative damage with the number of cycles at the rib-to-deck joint can be calculated by equation (2): where D is the damage index, and N i is the number of cycles in the S-N curves shown in Figure 24 for ∆σ i .
To determine the infuence of the UHPC and studs of the composite bridge deck, another damage index can be calculated using equation (3): where D US is the damage index caused by the UHPC layer and headed stud; D a is the index obtained from the actual test; and D is the damage calculated using equation (2) by the FE model. Te failure index of the UHPC layer and studs at the cracked rib-to-deck joint on the transverse specimen is shown in Figure 25. At the beginning of the fatigue test, there were almost no additional damage, then, tiny additional damage occurred after the initiation of cracks in the UHPC layer. At the loading stage II, the cumulative rate of additional damage increased to a certain  extent due to the several cracking of the UHPC layer. At the stage III, as the fatigue cracking of the headed stud, the additional damage increased rapidly. After 6 × 10 6 cycles of load, the additional damage value is approximately 0.53, which has a considerable infuence on the fatigue performance of the composite bridge deck.

Conclusions
Tis study experimentally investigated the fatigue behavior of a steel-UHPC composite deck with large longitudinal ribs. Te following conclusions are drawn.
(i) After 2 000 000 cycles of standard fatigue load and 1.5 times the standard fatigue load, no evident damage and mechanical degradation were observed in the transverse and longitudinal specimens. In addition, due to the excellent crack width control ability of UHPC, the test specimens satisfed the crack width limitation. Based on the foregoing, the proposed composite bridge deck satisfes the design requirements in terms of fatigue strength and durability. (ii) Te stifness of the specimen distinctly degrades during the failure loading stage (3.0 times the standard fatigue load). Te degradation mechanism of the transverse specimen involves the appearance of fatigue cracks at the longitudinal rib and weld of the steel bridge deck during the failure loading stage. (iii) Te efect of UHPC crack and headed stud fatigue failure on the damage accumulation of welded joints was quantifed by FE simulation and experimental test; this infuence accounted for 53% of the total damage.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest. Advances in Materials Science and Engineering 13