The concrete material and mix proportions are listed in Table 1. Crushed stone with maximum aggregate size distributed from 5 to 25 mm was used as the coarse aggregate. River sand with a fineness modulus of 2.71 was chosen as the fine aggregate in the mixes. The water/cement ratio for the mixtures was 0.36.

All test rebars were ribbed with a diameter of 16 mm and yield design strength of 400 MPa. The chemical composition of rebars are listed in Table 2.

A total of 16 categories were considered, including 90 specimens. A construction joint was made in the middle of each specimen, and the specimens were long enough, so that both side parts of specimens that didn’t contain joints could be as control parts, and the test conditions were exactly the same, so this experiment didn’t design separate control specimens. The experimental parameters are described as follows:

1) Two marine environmental conditions (James et al., 2019) were selected, namely a chlorine salt solution immersion condition and a cyclic wet–dry condition. The former was used to simulate a seawater immersion environment, and the latter was used to simulate the tidal environment.

The experimental categories and specimen quantities are listed in Table 3.

The design of specimens referred to the Standard for Testing the Long-Term Performance and Durability of Ordinary Concrete (GB/T 50082-2009, China). Thus, the dimensions 100 × 100 × 800 mm were selected for concrete prism specimens. A construction joint was located in the middle of each specimen as shown in Figure 1. The concrete cover thickness of the steel bar at the bottom was designed as 30 mm. The bottom surface of the specimen was set as the exposed face to experimental conditions, and the side and top surface were sealed with epoxy resin to avoid the intrusion of chloride ions.

The length of each rebar was 900 mm. First, all mild steel and stainless-steel bars were polished to remove the rust spots. For epoxy-coated rebars (Figure 2A), man-made scratches were added to the epoxy coating to simulate irregular damage in the transportation of epoxy rebars and concrete construction. The epoxy coating at scratches was completely damaged and the substrate was exposed. The size of a scratch was about equal to the rib spacing of rebar and the scratches was evenly distributed. The ratio of coating damage was about 5%. Then, all rebars were labeled and the length and weight were accurately measured.

Each specimen was molded in two parts and the fabrication process is shown in Figure 2B. The part of old concrete was first casted. After 3 days, all partitions at joints were removed. In 2 days, the joint interface on old concrete of each specimen with roughened joint was artificially scabbled as shown in Figure 2C, and kept in a saturated surface dry state according to current construction practice in China. Then the part of new concrete was placed as shown in Figure 2D. After three more days, all formworks were removed, and all specimens were placed in a standard maintenance room with a temperature of 20°C ± 2°C and a relative humidity of 95% for 28 days.

The comparative experiments were carried out by the electrochemical acceleration corrosion method. Direct current in a parallel manner was used to accelerate corrosion, as shown in Figure 3.

For Group IC in the immersion condition, all specimens directly and completely immersed in chlorine salt solution (Figures 3A, B). Given that the salinity of seawater is of 3.5% NaCl, a 3.5 wt% NaCl solution was used for corrosion experiments. As the water in the basin evaporated, the solution level in the basin was checked every 3 days and the water was supplemented in time. The steel bar was connected with the positive electrode of the steady voltage DC power supply as the anode of the electrochemical reaction; the graphite rod was directly placed in the solution as a cathode.

For Group WD in the dry-wet cycle condition, all specimens were supported by bolsters above the chloride solution (Figures 3C, D), and the same solution was sprayed periodically onto the specimens twice a day. Graphite rod as cathode, directly flat on the top of the specimen. In order to prevent the solution from evaporating too quickly and length of exposure to the humid environment being too short after spraying, a layer of gauze was spread on specimens, and a thick cotton sack covered the gauze for moisture storage.

Before the experiments were conducted, all specimens were placed in the experimental conditions for 1 week to ensure that the subsequent electrochemical reaction could formed an electrical circuit.

The initial current density was controlled at about 100 μA/cm², which was checked periodically to ensure its stability. In the dry stage of dry-wet cycles, the current density was significantly reduced. In order to reflect the influence of dry-wet cycle conditions, only the current density in the wet stage was stably controlled during the test. The corrosion time was set at 60 days.

After corrosion experiments, the specimens were broken, and the steel bars were taken out. The concrete adhering to the steel bars was scraped off. Then, the steel bars were pickled with 12% hydrochloric acid solution, rinsed, and dried for weighing.

After each full steel bar was accurately weighed, the mild steel bars and two types of stainless-steel bars were cut into multiple sections and three of them were taken. The length and position of the cut sections are shown in Figure 4. The length and weight of each section were accurately measured.

The weight loss rates of full rebars were calculated as follows: L ω = ω 0 − ω ω 0 × 100 % In the equation above L ω is weight loss rate of the steel bar; ω 0 is the weight of the steel bar before corrosion (g); and ω is the mass of the corroded steel bar after pickling (g).

Assuming that the initial mass of steel bar was uniformly distributed along the length, the initial mass of each section was determined proportionally, and then the rust amount of each section was calculated. An influence coefficient of construction joint on local weight loss of a rebar was defined and calculated as follows: k j = 2 m 2 m 1 + m 3 In the equation above m 2 is the rust amount of the rebar in joint section 2 (g); m 1 is the rust amount of the rebar in non-joint section 1 (g); and m 3 is the rust amount of the rebar in non-joint section 3 (g).

Most of the specimens developed rust expansion cracks on the side of the concrete cover exposed to test conditions, as shown in Figure 5. All the cracks appeared along the direction of steel bars, and there were no cracks in the vertical direction of steel bars. No interface cracks developed at the joint, and the joint interface maintained good continuity before and after the test. Because the width of rust expansion cracks is affected by many factors such as the shape of bar and the presence of stirrups (Okazaki et al., 2020), the width of cracks in the experiments was not measured. However, by observing the cracks of all specimens, it was found that the crack widths of WD specimens (Figure 5A) were relatively larger than those of IC specimens (Figure 5B). In addition, the type of steel bar had a significant impact on the widths of the rust expansion cracks. The widths of cracks on mild steel specimens were obviously the largest, while those on stainless steel specimens were relatively small, and there were almost no visible cracks on specimens with epoxy-coated rebars.

The rust morphologies of rebars at different positions are shown in Figure 6.

For mild steel bars (Figure 6A), the most severe corrosion of steel bars occurred at the joints. The corrosion developed from the joints to both sides. Most of mild steel bar surface was relatively smooth, some mild steel bars had longitudinal gullies, which was different from natural corrosion under pitting (Alonso et al., 2019). The corrosion of the non-joint section was more uniform along the surface of mild steel bar, and there was no obvious pitting corrosion. However, the concrete cover side of mild steel bar was more severely corroded than the backside, and the cross section of mild steel bar was elliptical after corrosion.

For S11203 stainless steel bars (Figure 6B), corrosion was mainly concentrated at the joints. The corrosion of the concrete cover side of S11203 bar was more severe, and no obvious corrosion was found on the side away from the cover. Moreover, the corrosion of S11203 bar at the joint was transversely fine and wavy, with less pit corrosion, which was quite different from that of mild steel bar. There were many corrosion pits and some corrosion spots on the surface of S11203 steel bar away from the joint.

The corrosion of S23043 stainless steel bar (Figure 6C) was similar to that of S11203. At the joints, the corrosion surface of S23043 bar was relatively smooth and the pitting corrosion was relatively less. However, far away from the joint, the corrosion pit of S23043 steel bar was larger and less intensive than that of S11203, which was probably because the composition of the two stainless steel bars was different. S23043 was a duplex stainless-steel bar, which was mainly composed of austenite and ferrite (Alonso et al., 2019), while S11203 was mainly composed of ferrite, and the chloride pitting resistance of ferrite was better than austenite (Chen 2020), so S23043 bar corrosion pit was generally larger.

The corrosion of epoxy-coated steel bars (Figure 6D) mainly occurred at the damage locations of epoxy coating (Sohail et al., 2019), not mainly at the joints. There were large corrosion pits at the damage locations, and brown rust adhered to the epoxy coating of the steel bars and the coating tended to peel off. When the epoxy coating at the joint was damaged and the damage was on the side of the concrete cover, the corrosion of the epoxy steel bar was the most severe, but there was no obvious corrosion on the backside.

Compared with mild steel bars, the rust of stainless-steel bars decreased from the joint to both sides more significantly and obviously. This phenomenon was due to the different content and form of iron in the two types of rebars. For mild steel bars with a high iron content, the electrolytic cells formed on the surface of rebars were more uniform, so the rust was more uniform. For stainless steel bars, because the iron was in the form of ferrite, austenite and other complexes, the large potential difference on the surface of rebars led to uneven distribution of the electrolytic cells. Therefore, the rust of the stainless-steel bars at the joints was relatively significant.

In general, except for epoxy-coated steel bars, the most severe corrosion of other steel bars all occurred at the joints, and the corrosion decreased from the joints to both sides, while the corrosion in non-joint sections was relatively uniform and less.

The box chart of the weight loss rate of all groups of full rebars is shown in Figure 7. There are no abnormal outliers in the figure. The average value of the weight loss rate of each group of specimens is shown in Table 4. The loss rate of specimens has the range of 1.18%–15.73% with an average value of 6.22%, which shows that there are big differences between different series.

In Figure 8, the box chart of the weight loss rate of rebars is shown by category. The figure shows that the average relationship of the weight loss rate of full rebars is mild steel bars > S11203 rebars > S23043 bars > epoxy-coated rebars. This means that the corrosion resistance of mild steel bars is the worst, and that of stainless-steel bars is better. Among them, the corrosion resistance of S23043 rebars is slightly better than that of S11203 rebars. The corrosion of epoxy-coated rebars is the smallest, and the fluctuation is also the smallest.

From the perspective of joint type, the average rust amount of roughened wet joint (RJ) specimens is slightly greater than that of direct wet joint (DJ) specimens, and the fluctuations of the two categories are not much different. The ratio of the weight loss rate of RJ specimens to DJ specimens has an average value of 1.09. It is generally believed that the roughening is beneficial to the joint interface, which can make it better connect into a whole and conducive to load transfer (Santos and Julio 2011; Santos et al., 2012). However, experiments have shown that the interfacial scabbling deteriorates the compactness of the interface to a certain extent. The main reason may be that scabbling by a mechanism or high-pressure water flow causes certain damage to the concrete surface. This damage leads to micro-cracks on the surface of the concrete and makes it easier for chloride ions to pass through the concrete cover to the surface of the steel bars.

From the perspective of condition type, the rust amounts of rebars in the immersion environment are greater than that in the wet–dry cycle environment. The average ratio of weight loss rate is 1.71, but this does not mean that the immersion environment is more unfavorable than the wet–dry cycle environment. From the point of view of the test conditions, the energization time under the immersion environment was significantly longer, and the current in the dry environment stage under the wet–dry cycle environment was relatively small, which led to a slower electrochemical corrosion reaction.

Considering that the corrosion resistance of epoxy steel bars was mainly determined by the quality of epoxy coating during the test time, the effect of concrete cover was less evident. Therefore, when the steel bars were cut off to compare the corrosion of joints and non-joints, the epoxy coated steel bars were excluded, and only the mild steel bars and two types of stainless steel bars were cut and weighed and analyzed.

The box diagram of the influence coefficient of construction joint on local weight loss ( k j ) in 48 specimens is shown in Figure 9. The average rust amounts of rebars in the interception section and the average influence coefficient of construction joint on local weight loss are shown in Table 4. It can be seen that the rust amounts of steel sections at the joints of all concrete specimens are significantly greater than those of the non-joint sections. The influence coefficient of construction joint is in the range of 1.15–3.33, with an average value of 1.76. According to the experimental phenomena, the corrosion of each steel bar at the non-joint sections was relatively uniform, so the corrosion rate could be deduced by local weight loss in non-joint sections. But the corrosion of steel bars at the joint sections are significantly higher than those of non-joint sections, the corrosion rate of steel bars may not effectively reflect this uneven phenomenon. So the influence coefficient of construction joint on local weight loss ( k j ) can negatively reflected the influence of joints on the rust uniformity of rebars in concrete specimens with joints to some extent, and the construction joints are indeed a weakness of concrete durability.

In Figure 10, the box chart of the influence coefficient of construction joint on local weight loss is shown by category. The average influence coefficients of construction joint of mild steel bars, S11203 stainless steel bars, and S23043 stainless steel bars are 1.38, 1.92, and 1.97, respectively. The influence coefficients of construction joint of mild steel bars have relatively small fluctuation. The amount of rust in mild steel sections at construction joints is higher than that of stainless-steel sections as shown in Table 4, which is consistent with the aforementioned rust distribution of mild steel and stainless-steel bars. However, it can be concluded that the rust uniformity of mild steel bars affected by construction joints is relatively better, and the two types of stainless-steel bars have worse uniformity. Therefore, in coastal concrete structures using stainless steel reinforcement, it is recommended to consider the concentrated corrosion of steel bars that may occur at the joints.

The average influence coefficients of RJ specimens and DJ specimens are 1.72 and 1.80, respectively, and DJ specimens have more fluctuation as shown in Figure 10. From the perspective of the relationship of joint section and non-joint section, the interfacial roughening has no significant effect on the corrosion uniformity of rebars. However, the rust amounts of joint sections in RJ specimens are significantly greater than in DJ specimens as shown in Table 4. The ratio of the two ranges from 1.07 to 1.24, with an average value of 1.13, which is slightly larger than the ratio of the overall weight loss rate (1.09). This shows that roughening does cause certain defects to the faces on old concrete and affects the density of the joints. That is, roughening is to a certain extent unfavorable to the resistance to the penetration of chloride ions at construction joints.

The average influence coefficients of IC specimens and WD specimens are 1.44 and 2.07, respectively, and WD specimens have more fluctuation as shown in Figure 10. This indicates that the rust uniformity of rebars in the immersion environment is significantly lower than in the wet–dry cycle environment. The ratio of the two ranges from 0.54 to 0.97, with an average value of 0.72. This shows that the corrosion of rebars at the joints is more severe in the wet–dry cycle environment; that is, to say, the rust amounts of rebars in the immersion environment may be higher than in the wet–dry cycle conditions, but the wet–dry cycle conditions are relatively more unfavorable for the joints.

The test results showed that the construction joint had become the weakness of durability in concrete. According to the analysis of joint details and concrete characteristics, there were three main reasons. First, because of the difference of concrete pouring sequence, the construction joint is different from the general continuous casting concrete, and there is no coarse aggregate crossing distribution on the construction joint interface. Without the block of coarse aggregate, the transmission capacity of harmful substances in cement mortar is stronger. Secondly, in the early stage of hydration heat, the water in the newly casted concrete will migrate to the concrete substrate, resulting in an increase in the water-cement ratio of the fresh concrete near the interface, thereby reducing the durability of the interface. Finally, there are shrinkage differences in different casting ages of concrete. The restrained shrinkage will produce tensile strain in fresh concrete, which further deteriorates the durability of the interface.