To investigate the bending performance of steel- UHPC composite bridge decks with epoxy bonded joints, four sets of bending tests were designed. The overall specimen design is illustrated in Figure 2, consisting of a steel plate, epoxy adhesive, carbon fiber cloth, and precast UHPC slab. The overall dimensions are 800 mm in length, 150 mm in width, and 42 mm in height. The steel plate dimensions are 800 mm in length, 150 mm in width, and 10 mm in thickness. The precast UHPC slab dimensions are 800 mm in length, 150 mm in width, and 30 mm in thickness. The epoxy bonded layer dimensions are 800 mm in length, 150 mm in width, and 2 mm in thickness, and the carbon fiber cloth dimensions are 150 mm in length, 20 mm in width, and 2 mm in thickness. To simulate the actual joints between precast UHPC slabs in bridge structures, two sets of specimens with 2 mm thick epoxy bonded joints were designed, and one set was reinforced with a 20 mm width carbon fiber cloth in the joint area. The specimens without joints were subjected to both positive and negative bending moments, while the specimens with joints were subjected to negative bending moments. All precast UHPC slabs had smooth surfaces, and epoxy bonded joints were used to naturally bond the precast UHPC slabs to the steel plates. Specific specimen parameters are detailed in Table 1.

The precast UHPC slabs were naturally bonded to the steel plates using an epoxy bonded layer. The specific procedure was as follows: a) Fabrication of molds and casting of precast UHPC slabs; b) After casting, the UHPC required 48 h of curing at room temperature, followed by demolding and 72 h of steam curing at 95°C; c) Cleaning the surface of the steel plates with acetone to remove all debris and rust; d) Fixing the precast UHPC slabs onto the steel plates with a 2 mm gap maintained between them; e) Preparation of epoxy adhesive; f) Injecting the epoxy adhesive into the 2 mm gap through pre-drilled holes until a small amount of adhesive overflowed from the edges; g) Curing the specimens at room temperature for 7 days until the epoxy bonded layer reached its full strength. The preparation process of the specimens is illustrated in Figure 4.

Figure 5A illustrates the loading method and measurement point arrangement for the specimens. All composite bridge decks are subjected to a four-point loading scheme, enabling a pure bending region of 300 mm at the midspan using a distributing beam. The load is applied using an electronic universal testing machine (MTS Exceed E45.205) at a loading rate of 0.01 mm/s. Displacement sensors D-4 and D-5 are placed 50 mm from the edge of the composite plate to measure deflections at the supports, while displacement sensor D-1 was positioned at the midspan of the specimen to measure midspan deflection. Displacement sensors S-1 to S-2 are arranged on the edge of the side of the specimen to measure the slip at the interface between the UHPC layer and the steel plate. Strain gauges are installed at the top, bottom, and sides of the midspan region to measure strains in the UHPC layer and the steel plate within the pure bending zone.

For specimens subjected to positive bending moment loading, the UHPC is positioned on top with the steel plate underneath. Conversely, for specimens subjected to negative bending moment loading, the arrangement is reversed. The bending tests of the steel-UHPC composite bridge decks follow the loading scheme used by Zou et al. (2021b). As shown in Figure 5B, preloading with an amplitude of 30% of the ultimate load (Pu) was applied three times to check the potential effects during the loading process. Subsequently, continuous loading is applied until specimen failure, with the observation and recording of crack propagation during the loading process.

The failure characteristics of the specimens under positive bending moment without joints (EA-Lm) are as follows: as the load increases, cracks begin to appear in the pure bending region at mid-span. When the ultimate load P _u is reached, the UHPC slab delaminates from the epoxy bonding layer, accompanied by a “bang” sound. The top of UHPC slab in the pure bending section of the specimen collapses under pressure. The failure mode of this type of specimen indicates excellent bonding performance between the epoxy bonding layer and the steel plate, while the bonding strength between the smooth UHPC slab and the epoxy bonding layer is significantly insufficient, resulting in separation of the steel plate from one side of the UHPC slab. The failure mode of the EA-Lm specimen is shown in Figure 6A.

The failure mode of the specimens under negative bending moment without joints (EA-Ln) is shown in Figure 6B. This type of specimen exhibits UHPC rupture and exposed steel fibers on the basis of the failure mode of the EA-Lm specimen. After reaching the peak load P _u , cracks extend upward along the main crack at mid-span, ultimately resulting in rupture of the UHPC slab along the mid-span. At the same time, the bonding strength between the UHPC slab and the epoxy bonding layer is completely lost, leading to separation between the steel plate and the UHPC slab.

The failure modes of the specimens under negative bending moment with joints (EA-J-Ln) and the specimens with joints reinforced with carbon fiber cloth under negative bending moment (EA-JC-Ln) are shown in Figures 6C, D, respectively. The failure characteristics of the two types of specimens are consistent: once the joint cracks, the cracks continue to extend upward, eventually leading to delamination of the prefabricated UHPC slab from the epoxy bonding layer near the joint, while the interface far from the joint remains intact.

Figure 7A presents the load-deflection curves for the EA-Lm specimen, which can be divided into three stages: a) Elastic stage: Before reaching the ultimate load P _u , the interface of the specimen remains undamaged, and it is in the elastic stage. During this stage, the load continues to increase, and the deflection increases slowly. b) After reaching the ultimate load P _u , there is a sudden failure at the interface, resulting in a sharp drop in load, indicating irreversible damage to the interface. c) Continued loading: The upper load is borne by the steel plate, and the deflection of the specimen continues to increase, but the load increases slowly. Local fragmentation occurs in the UHPC layer near the loading point, resulting in bending failure of the composite bridge deck. The load-deflection curves for these three types of specimens exhibit a similar trend, and their ultimate load P _u is approximately 27.67 kN.

The load-deflection curves for the negative bending moment tests are shown in Figure 7B, including three types of specimens: EA-Ln, EA-J-Ln, and EA-JC-Ln. It can be observed that the specimen EA-Ln without joint is capable of resisting external loads under negative bending moment loading, with an ultimate load P _u of approximately 16.58 kN. Before reaching the ultimate load P _u , the composite bridge deck is able to jointly bear the external load, and a primary crack appears at the mid-span of the UHPC board and continues to propagate upwards during this process. After reaching the ultimate load P _u , the UHPC board detaches from the epoxy bonding layer in the mid-span bending section. As the load continues to increase, the UHPC board is pulled apart along the primary crack at the mid-span, eventually the composite function was lost, and only the steel plate beared the external load.

The load-deflection curves for EA-J-Ln and EA-JC-Ln specimens exhibit similarities and can be analyzed in three stages: a) Elastic Stage: During this stage, the load continues to increase gradually, and the deflection increases slowly, resulting in a straight upward-sloping curve. The composite bridge deck are jointly stressed and the interface is intact in this stage. b) Ultimate Load P _u Stage: When reaching the ultimate load P _u , cracks propagate upward along the seam near the mid-span. The load experiences a sudden drop, and the deflection increases rapidly, leading to a steep decline in the curve. This phase continues until the UHPC board detaches from the epoxy bonding layer near the seam. c) After Reaching Ultimate Load: In this phase, the UHPC board and epoxy bonding layer at the seam have detached. The composite structure loses its capacity, and all external loads are borne by the steel plate. Comparing the test results of all specimens, it is evident that the failure of EA-J-Ln and EA-JC-Ln specimens occurs at the seam. This indicates that the composite bridge deck has poor resistance under negative bending moment loading, whereas the seam-less specimen EA-Ln can effectively resist external loads in negative bending moment mode. Additionally, the epoxy bonding layer-bonded composite bridge deck exhibits effective resistance to external loads in positive bending moment mode. Therefore, in the design of orthotropic steel bridge decks using prefabricated UHPC deck combined with epoxy bonding layers, it is recommended to place the seam at the location where positive bending moments are generated in the bridge deck. From Figure 7B, it can be observed that the ultimate load of the EA-JC-Ln specimen is approximately twice as high as that of the EA-J-Ln specimen. This indicates that reinforcing the joint area with carbon fiber cloth can enhance its bending performance.

Figure 8 shows the load-slip curves of the specimens. As shown in Figure 8A, the load-slip curve of the EA-Lm specimen indicates that, prior to reaching the ultimate load, the epoxy adhesive layer ensures an effective connection between the UHPC panel and the steel plate, with negligible slip occurring at this stage. Subsequently, upon reaching the ultimate load, the slip at both ends of the specimen escalates and continues to increase. Larger slip occurs on the S-1 side, with a maximum slip value of S-1 = 13.9 mm. Hence, a smooth UHPC panel can achieve effective bonding through the epoxy adhesive layer. The load-slip curves of the EA-Ln, EA-J-Ln, and EA-JC-Ln specimens, as shown in Figures 8B, C, are characterized by the fact that the failure of all three types of specimens occurs at the mid-span joint, and debonding occurs between the UHPC panel and the epoxy layer at the joint, resulting in minimal and almost negligible slip at the ends. The maximum slip value for specimens with only epoxy joints is S-1 = 0.08 mm. For the EA-JC-Ln specimen, which was reinforced at the joint with carbon fiber cloth, the slip at the ends is significantly larger, indicating that the reinforcement of the joint with carbon fiber cloth enhances the bending performance of the epoxy joint, with a maximum slip value of only S-1 = 0.31 mm.

The strain distribution curves were drawn after averaging the values for specimens in the same group. However, the strain distribution curve for the EA-JC-Ln specimen could not be plotted due to damage to the strain gauges during initial loading. Figure 9A shows the strain distribution curve along the height direction at mid-span for the EA-Lm specimen. Before the applied load reaches the ultimate load P _u = 27.3 kN, the strain at mid-span exhibits a linear distribution. As the load continues to increase, the strain at mid-span shows an increasingly evident nonlinear distribution. However, the strain distribution across the mid-span section always remains linear. The distribution is discontinuous at the steel-UHPC interface, where both the thin UHPC panel and the steel plate exhibit a linear strain distribution. Once the ultimate load P _u = 27.3 kN is reached and the epoxy adhesive layer interface is damaged, the assumption of plane sections being plane no longer holds.

Figure 9B depicts the strain distribution curve along the height direction at mid-span for the EA-Ln specimen. Throughout the loading process, the strain distribution of the specimen consistently adheres to the plane section assumption, with the neutral axis maintaining a constant position as theoretically expected, as shown in Figure 13B. This is consistent with the low level of interfacial slip for the EA-Ln specimen shown in Figure 8B. In Figure 9C, for the EA-J-Ln specimen, when the load is below 2.5 kN, the strain distribution at mid-span is approximately linear. Before reaching the ultimate load, the epoxy adhesive interface can achieve reliable connection between the thin UHPC panel and the steel plate. After the load exceeds 2.5 kN, sudden extensive damage occurs at the interface around the joint, leading to a sharp increase in strain, which indicats that the joint has fractured and damaged.

In this study, a finite element model of the steel-UHPC composite bridge deck was established using the static analysis method in ABAQUS. Based on the symmetry of the steel-UHPC composite bridge deck, a half-model was developed to improve computational efficiency. The symmetrical plane is the x-plane. The steel plate, ultra-thin UHPC, and epoxy adhesive layer were simulated using C3D8R (three-dimensional, 8-node linear solid integrated elements). According to the experimental results, the bonding behavior between the UHPC panel and the epoxy adhesive layer was simulated using cohesive contact properties. The specific finite element model is shown in Figure 10.

Figure 14 displays the load-deflection curves for both the EA-Lm and EA-J-Ln specimens (experimental and modeled). Both types of specimens exhibit good consistency in their curves. Compared to the experiment, the finite element model for the EA-Lm specimen exhibits debonding between the UHPC panel and the epoxy adhesive layer, while for the EA-J-Ln specimen, debonding occurs at the joint between the UHPC panel and the epoxy adhesive layer, with the area of failure being consistent. This indicates that the bending behavior of the composite bridge deck can be simulated by Cohesive Zone Model (CZM) to represent the bonding between the UHPC panel and the epoxy adhesive layer. The deviation between the experimental ultimate load and the model calculated value is within 5%, demonstrating that the finite element analysis can accurately simulate the bending behavior of the EA-Lm and EA-J-Ln specimens.

Figure 15 presents the numerical simulation results of the EA-Lm and EA-J-Ln specimens. CSDMG is an indicator used to determine the degree of interface damage. When CSDMG reaches 1, it indicates delamination between the interface layers, and when its value ranges between 0 and 1, it signifies interface damage. DAMAGEC and DAMAGET represent the indices for evaluating compression and tensile damage in UHPC, respectively. Their values also range between 0 and 1, indicating damage to the UHPC matrix when they reach 1. According to Figure 15A, under the ultimate load P _u = 24.87 kN, there is extensive debonding between the epoxy adhesive layer and the UHPC panel, rendering them ineffective in working together. Additionally, significant tensile damage and minor compressive damage are observed near the loading point on the UHPC panel, but no destruction occurs. From Figure 15B, it can be seen that under the ultimate load P _u = 2.45 kN, extensive debonding occurs between the UHPC panel and the epoxy adhesive layer near the joint, while the UHPC matrix exhibits only minor tensile damage and no compressive damage.