Research on the Fatigue Flexural Performance of RC Beams Attacked by Salt Spray

The fatigue flexural performance of RC beams attacked by salt spray was studied. A testing method involving electro osmosis, electrical accelerated corrosion and salt spray was proposed. This corrosion process method effectively simulates real-world salt spray and fatigue loading exerted by RC components on sea bridges. Four RC beams that have different stress amplitudes were tested. It is found that deterioration by corrosion and fatigue loading reduces the fatigue life of the RC and decreases the ability of deformation. The fatigue life and deflection ability could be reduced by increasing the stress amplitude and the corrosion duration time. The test result demonstrates that this experimental method can couple corrosion deterioration and fatigue loading reasonably. This procedure may be applied to evaluate the fatigue life and concrete durability of RC components located in a natural salt spray environment.


Introduction
Owing to the continued economic growth, civil engineers are required to build infrastructure with highly efficient ways and use abundant materials. This is one reason why a large number of coastal reinforced concrete (RC) structures were constructed. Several sea-crossing bridges have built in China, such as the Hangzhou Bay Bridge, the Zhoushan Sea-crossing Bridge and the Qingdao Bay Bridge. All of these bridges should be resistant to the corrosive environment for the sake of the long-term durability of the concrete. However, many upper structures such as the box girder and the bridge deck are located in the salt spray zone. These bridges also suffer from a high vehicle load, which, ultimately, leads to bridge fatigue. Therefore, these upper structures must endure the coupled effect of fatigue load and salt spray.
Currently, there are two categories of corrosion fatigue tests used for reinforced concrete beams. The first category of corrosion fatigue tests involves applying fatigue loads to pre-corroded RC components. The RC components are first corroded by an electrical accelerated corrosion test and then fatigue loads are applied. The fatigue life (Masoud et al., 2001), the stiffness variation (Song and Yu, 2015), the stress-strain relationship (Fernandez et al., 2015), the bonding performance (Al-Hammoud et al., 2010) and the flexural performance of the corroded RC beam (Al-Hammound et al., 2011) are studied. The main disadvantage to this type of experiment is that the steel bars are uniformly corroded which is not typical of real-world scenarios where steel bars are non-uniformly corroded. Generally, the side of the steel bar that is closest to the concrete surface shows a higher degree of corrosion and the other side often shows no corrosion (Oh and Jang, 2003). However, many existing corrosion analysis techniques assume a uniform corrosion pattern around a rebar circumference (Chen and Mahadevan, 2008;Val et al., 2009), which is rarely the case in practice. A non-uniform corrosion layer formed around a rebar perimeter would be qualitative in nature (Muthulingam and Rao, 2014). Also, the effect of non-uniform corrosion induced by various stresses in the RC is more conducive to cover cracking than uniform corrosion (Xia et al., 2012).
A second category to test fatigue caused by corrosion is to apply fatigue loads to the RC components while they are immersed in salt water. Many experiments in the literature (Ahn and Reddy, 2001;Wang et al., 2015) have utilized conditions involving tandem fatigue loading and corrosive medium eroding. For example, Ahn and Reddy (2001) used seawater to wet the RC beams every 12 hours and applied electrical accelerate corrosion tests during the fatigue loading process. However, the upper structures of bridges do not come into direct contact with seawater; instead, they undergo a dry-wet cycle by salt spray and rain erosion. Therefore, the immersion of RC components in corrosive solutions deviates from the actual situation. The main disadvantages of this category of tests include the facts that (1) the corrosive products are dissolved into a corrosive solution which results in a decrease in the corrosive expansion force (Jaffer and Hansson, 2009), and (2) the cracks in the concrete would open and close due to fatigue loading. The open cracks provide a direct pathway for chloride ion to penetrate through the concrete to the steel bars (Coronelli et al., 2013). This would significantly accelerate the corrosion process.
The focus of this work was to investigate the fatigue flexural performance of RC beams attacked by salt spray. An experimental method including the steps of electro osmosis, electrical accelerated corrosion and salt spray (Xia et al., 2011) were carried out using a multifunctional climate simulation laboratory. The equipment to simulate the salt spray environment and the fatigue loads were applied to the top of RC components. Five RC beams were tested under different stress loading ranges. The fatigue life, the failure mode, and the deflection variation for each beam were compared. This work offers an effective experimental method that can be applied in concrete durability research.

Design of a RC beam
The mixture ratio of concrete is shown in Table 1. The cement used was P·O 42.5 ordinary Portland cement. The coarse aggregate was crushed rock with a 5.0-20.0 mm continuous grading. The fine aggregate used was river sand. The designed compressive strength of the concrete was C30 and the measured compressive strength of the concrete was 34.0 N/mm 2 .
The dimensions of the RC beam were 150 mm×200 mm×1500 mm (width×height×length). The concrete cover was 20.0 mm thick. The beams were designed as under-reinforced beam and ensured a higher shear capacity than a flexural capacity. The reinforcement ratio was 1.22%. Fig. 1 shows details of the specimens.
In order to transfer chloride ions from the concrete surface to the steel bar surface, a stainless steel sheet of the di-mensions 30 mm×1300 mm×0.2 mm (width×length×thickness) was pre-embedded into the center of each beam. Since the corrosive targets were two tensile longitudinal reinforced bars, the stirrup corners placed outside the longitudinal reinforcement were wrapped with the rubber tube. The rubber tube prevents the direct current of the electrical accelerated corrosion test from going through the stirrup and correspondingly prevents the corrosion of the stirrup. The five beams were poured at the same time and cured under the same conditions.

Chloride ion migration by electro osmosis
One way that steel bars corrode is through chloride ions reaching a critical level at the steel. Some researchers who study steel bar corrosion incorporate NaCl powder during the concrete casting. The main disadvantage to this is that NaCl powder can disperse into the components homogeneously and lead to uniform corrosion of the steel bar, which is not coincident with actual corrosion conditions. On the other hand, natural permeation of chloride ions through concrete is a long process that can take several decades and is impractical to reproduce in a lab setting.
Therefore, an electro osmosis method has been proposed to speed up the permeation. In order to ensure that chloride permeates to the surface of the steel bar, controlling parameters were first obtained. According to 'Standard for Test Methods of Long-term Performance and Durability of Ordinary Concrete' (Ministry of Housing and Urban-Rural Construction of the People's Republic of China, 2010), the chloride diffusion coefficient is the main parameter. A rapid chloride migration (RCM) test is performed to determine the chloride diffusion coefficient. (1) α where T is the average temperature of the anodic solution (K), h is the height of the specimens (m), t is the testing time (s), x d is the diffusing depth of chloride ions (m), and presents a non-dimensional parameter. The RCM test was performed on six cylinder specimens with the diameters of 100 mm and heights of 50 mm. To prepare the specimens, they were first dried under vacuum and then immersed into a Ca(OH) 2 solution for 18 h. The voltage was set at 30.0 V. The direct current and temperature of each anodic solution was recorded. After electro mi-  gration, the specimens were separated into two halves. A solution of AgNO 3 was sprayed onto the surface of the con-crete as a color indicator. The diffusion depth of chloride ions was measured and results are shown in Table 2.
By Eq.
(1) and the data in Table 2, the time of electro osmosis for the chloride ions diffusing to the surface of steel bar is calculated using: D RCM where x cov is the depth of concrete cover (m) and is the average chloride diffusion coefficient. In this case, the time of electro osmosis was 87.3 h at a constant voltage of 30 V. The layout for the electro osmosis treatment of beams is shown in Fig. 2.
In the electro osmosis method, ions migrate under the force of an electrical field. The electrical field was constructed by an inner pre-embedded stainless steel sheet and an external wrapped stainless steel net. During treatment, the positive electrode is connected to the inner pre-embedded stainless steel sheet only. No direct current was passed through the steel bars, which ensured that the steel was not corroded by the electro osmosis treatment itself. The target electro osmosis area had a length of 1.0 m and was located at the mid-span of the beam. A stainless steel net and a sponge were wrapped inside the plastic films. Before the electro osmosis treatment, a 5% NaCl solution was sponged onto the concrete to fully moisten the surface. The voltage of the DC power was set to a constant value of 30 V for 87.3 hours.

Corrosion by a dry-wet cycle
In realistic RC components, harmful chloride ions mainly come from sea spray, rain and deicing salt. One notable feature of real-world situations is that the salt spray cycles through a dry and wet cycle. To achieve the type of dry-wet cycle, a multifunctional climate simulation laboratory was designed. The structure of the multifunctional climate simulation lab is shown in Fig. 3. The multifunctional climate simulation laboratory is made of stainless steel plates. It can simulate salt spray, salt rain, ultraviolet irradiation and a corrosive gas environment. The temperature can be varied from -20°C to 60°C and the humidity can be varied from 30% RH to 98% RH. There are salt spray sprinkler heads installed on the side plates and the spray particle size can be as small as 5.0 μm. Salt rain sprinkler heads are also installed along the water pipe located at the top surface.
The dry-wet cycle treatment was applied as 8.0 h of wetting and 8.0 h of drying. During the wetting period, a 5% NaCl solution was sprayed. The drying period consisted of an environment at 35°C and with a humidity of 80% RH. Meanwhile, the fluorescent light and centrifuge were turned on during the drying period to simulate illumination and nature wind, respectively.

Electrical-accelerated corrosion
The electro osmosis method and the dry-wet cycle experiment have each been shown to simulate the deterioration of the RC components. They are available under a dead-loading or a no-loading condition. In the test, the loading frequency of the fatigue experiment was 2.0 Hz and the fatigue failure of the RC components occurred in a short  time. At this frequency, the corrosion process accomplished by the electro osmosis method and the dry-wet cycle is a process relatively slower than that of the fatigue experiment. In order to have the corrosion process occurred at the same time as fatigue, the electrical-accelerated corrosion experiment is applied during the dry-wet cycle test. The layout for the electrical-accelerated corrosion experiment is shown in Fig. 4. Fig. 4 show the differences between our set up and traditional electrical-accelerated corrosion experiments as there is no sponge wrapped around the RC components. In this experiment, the harmful chloride ions are introduced via electro osmosis and the dry-wet cycle. The electrical-accelerated corrosion experiment was used to speed up the electrochemical process.

Fig. 2 and
The electrical-accelerated corrosion, dry-wet cycling and fatigue experiments were carrying out at the same time in the multifunctional climate simulating laboratory. The current density of natural corrosion is normally arranged at 50 to 100 μA/cm 2 (Andrade et al., 2002). The current density in an electrical-accelerated corrosion experiment has normally been set at 0.20 mA/cm 2 to 1.0 mA/cm 2 (Ballim et al., 2001;Higgins and Farrow, 2006;Minh et al., 2007). In order to simulate the salt spray, the current density of 0.20 mA/cm 2 was selected for the wet cycle and 0.10 mA/cm 2 was selected for the dry cycle.
The corrosion quantity of the steel bars in the RC beams could be estimated by Faraday's law as follows: ∆w where is the mass loss caused by the corrosion of the re-inforced bar, M is the atomic weight of iron (M = 56), I is the corrosion current (A), i is the current density (mA/cm 2 ), A is the initial surface area of the reinforced bar (cm 2 ), t is the corrosion time (s), Z is the valence electron count of iron (Z=2), and F is the Faraday constant (F=96490).
2.3 Design of fatigue loading 2.3.1 Experiment group Five RC beams were tested in this paper and the loading details are shown in Table 3.
The B-Ref beam acts as a reference beam. None of the corrosion process was conducted on the B-Ref beam. Only the static loading experiment was performed to obtain the ultimate bearing capacity, which was further used to determine the maximum and minimum stress of the fatigue loading experiment. The FB-0.6 beam and FB-0.5 beam were subjected to the fatigue loading experiment and no corrosion processes were applied. The maximum stress ratios for these two beams were 0.6 and 0.5, respectively, and the minimum stress ratio for both was 0.1. The CFB-0.6 beam and CFB-0.5 beam were each subjected to the fatigue loading and corrosion processes.
The normal fatigue loading frequency on the RC beams was set to be 1.0-5.0 Hz. The loading frequency for the FB-0.6 and FB-0.5 beams was set to be 4.0 Hz. Many researches show that the fatigue loading frequency (when lower than a certain value) has no significant influence on the fatigue life and failure mode of reinforced concrete beams (ACI Committee 215, 1974). However, while taking the corrosion ratio into consideration, the loading frequency for Beam CFB-0.6 and Beam CFB-0.5 was set to be 2.0 Hz. The relatively lower loading frequency allows the RC beams to obtain higher corrosion ratios.

Design for loading
The loading equipment used was a type of pulsating fatigue testing machine that could achieve constant amplitude loading with a sine waveform. The oil pump and hydraulic station was located beside the multifunctional climate simulate laboratory. The fatigue actuator was fixed onto the reaction frame that stretched over the multifunctional climate simulate laboratory. The fatigue actuator was connected to the pump by oil pipe. A galvanized steel pipe was used as a link between the fatigue actuator and the RC beam. The layout of this equipment is shown in Fig. 5.  The RC beams were subjected to a three-point loading. The loads were then transferred to the test beam through a distributing beam using a hinged bearing. The clear span of concrete was 1200 mm and the pure bending section in the mid-span was 400 mm. The fatigue loading spectrum is shown in Fig. 6.
During this process, two static load-unload processes were applied before the fatigue loading in order to record initial information for each beam. Then, a normal fatigue loading experiment was applied. Once the fatigue cycle number reached a predetermined value (1, 5, 10, 20, 30, 50, 100 million cycles), the fatigue loading was paused to statically load the upper limit of the fatigue loading and to measure the deflection and crack width. Because of the corrosive environment, the strain gauges of the steel bar and the concrete could not survive. Therefore, only the deflection was recorded by LVDT displacement sensors that were distributed at the mid-span and supports. Finally, the residual deflection for each beam was measured after unloading to the beam.

Fatigue life of the RC beam
The ultimate bearing capacity of the RC beams by static loading is 116.0 kN. Thus, the maximum stress ratios of 0.6 and of 0.5 are 69.6 kN and 58.0 kN, respectively. The fatigue life for each beam is listed in Table 4.
The fatigue lives of Beams CFB-0.6 and CFB-0.5 are reduced by 3.3% and 12.9%, respectively, compared with their no-corrosion beam counterparts FB-0.6 and FB-0.5. This indicates that the corrosion process shortens the fatigue lives of the beams. The corrosion time of Beam CFB-0.5 is four times longer (242 h vs. 62 h) than that of CFB-0.6 and the fatigue life reduction ratio is four times bigger (12.9% vs. 3.3%). It confirms the explanation that the less corrosion time gives rise to the less reduction of the fatigue life in CFB-0.6 than CFB-0.5. From another point of view, the reduction in the fatigue life is related to an increase in the stress amplitude. When increasing stress amplitude from 0.5 to 0.6 (FB-0.5 vs. FB-0.6), the fatigue life decreases by 77.3%. For CFB-0.5 and CFB-0.6, the value increases by 74.9%. This result shows that the stress amplitude is a critical parameter that must be considered. The steel bar corrosion should also be considered for its contribution to the reduction of the fatigue life.

Fatigue failure mode of the RC beam
The fatigue failure modes of the RC beams are shown in Fig. 7.
The no-corrosion beams FB-0.6 and FB-0.5 are shown in Figs. 7a and 7b. In Beam FB-0.6, normal and diagonal cracks rapidly develop at the beginning of the fatigue loading and the crack width increases rapidly. Beam FB-0.6 is finally broken by a fatigue fracture of the steel bar in the pure bending section. The fatigue failure mode is shown in Fig. 7a. In Beam FB-0.5, because the stress amplitude is smaller than that for Beam FB-0.6, cracks are only observed at the pure bending section during early fatigue load- ing. By the time approximately 5000 fatigue cycles are performed, diagonal cracks emerge and rapidly propagate. The width growth of the diagonal cracks and the transverse cracks grow slowly. After two million fatigue cycles, there are no apparent signs of damage. The final failure mode is achieved after applying the static loading experiment (Fig.  7b).
The fatigue failure modes of the RC beams subjected to the corrosive treatment, CFB-0.6 and CFB-0.5, are shown in Figs. 7c and 7d. Similar cracking patterns are observed for both the corroded beams and the non-corroded beams. However, since the fatigue cracks are opened and closed during fatigue loading periodically, partial rust products of the steel bar are forced out. Consequently, light yellow rust products are observed inside the fatigue cracks. It is also found that the color of the rust products deepens to a tawny color gradually over the erosion time. After approximately 900000 fatigue cycles for CFB-0.5, longitudinal corrosive cracks are observed at the bottom and the side of the RC beam. The corrosive cracks propagate and finally become connected. Meanwhile, there are no corrosive cracks for the CFB-0.6 beam over the whole fatigue loading time. The CFB-0.6 and CFB-0.5 beams are both subjected to damage caused by the fatigue fracture of the steel bars. The fatigue life of reinforced concrete beams applied to larger loading (FB-0.6 and CFB-0.6) is smaller than that of the beams applied to smaller loading. It indicates that the fatigue loading amplitude is the main factor affecting the life of reinforced concrete beams. It is worth noting that Beam CFB-0.6 experiencing corrosion effect failed by shear, which is different from other beams. The corrosion degree of steels besides diagonal cracks is higher under heavy loading and cyclic loading, which is the main reason causing the fatigue fracture of steels and shear failure.

Corrosion characteristics of steel bars in concrete beams
After the fatigue test, the reinforced bars were excavated from Beams CFB-0.6 and CFB-0.5 and the surfaces of the reinforced bars are shown in Fig. 8. The steel bar's rusts were removed by being immersed into hydrochloric acid. The mass loss represents the corrosion rate of tensile longitude steel bars. The corrosion rate of steels in CFB-0.5 is shown in Table 5. Owing to the low corrosion degree, the accurate corrosion rate of steels in CFB-0.6 could not be measured.
The corrosion time for Beam CFB-0.6 is very short at 62 h (four dry-wet cycles). There are no longitudinal corrosion cracks observed at the bottom surface of CFB-0.6. The only rust spots are around the diagonal cracks. This means that the open cracks provide a direct channel for chloride ions to contact the steel bar. Very low corrosion of the steel bar occurs. The ribs around the steel bar remain intact and the fatigue life of Beam CFB-0.6 is similar to that of Beam FB-0.6.
The corrosion time for CFB-0.5 is relevantly long at 242 h (fifteen dry-wet cycles). One of the main differences between CFB-0.6 and CFB-0.5 is that there are longitudinal corrosion cracks at the bottom surface of Beam CFB-0.5. Besides, the surface of the steel bars has corrosion pits distributed along the surface and the rib volume decreased. This means that the steel bars suffer serious corrosion. The corrosion cracks and the diagonal cracks have separated the concrete completely into several pieces. These results in the fatigue life of Beam CFB-0.5 are far less than that of Beam FB-0.5.

Chloride concentration
After the corrosion and fatigue experiment, the beams were dried in the air. The chloride content in concrete cover was determined by the rapid chloride testing method (RCT). For the fatigue beams, symmetrical sampling was applied. Drilling and powder extracting were carried out at the place every 150 mm from the midspan section. And the position of the pure bending, shear bending and cracks should be taken into consideration to determine sampling position. For Beam CB-Ref, only unilateral samples were taken. Sampling position was shown in Fig. 9. CB-Ref is a reference beam in corrosive environment, but it did not apply loading. The corrosion time of CB-Ref was the same as Beam CFB-0.5. The red line, the yellow line and the blue lines represent cracks at the bottom of the beam. Fig. 10 shows the chloride ion distribution in the concrete cover at different sections. Minus sign means that measuring point is located at the left of the midspan.
Learned from Fig. 10, the chloride ion content at different cross sections shows a decreasing trend with the increasing of the concrete cover depth. For Beam CFB-0.6, it is obvious that the chloride ion content besides the crack (150 mm away from the midspan) is higher than that of the section without cracks. It indicates that crack provides a path for chloride ion transferring into the concrete cover. The chloride ion content in certain concrete cover depth increases suddenly at some fatigue cracks in CFB-0.5. It is because that the periodic opening and closing of fatigue cracks make the NaCl solution absorbing into concrete, resulting in the accumulation of chloride ions at a certain depth.

Corrosion potential of steel
The corrosion degree of the steel bar was determined by the half cell potential testing method during the fatigue experiment. While the fatigue time reached the predetermined number (0, 50, 100, 200, 300, 500 thousand), the fatigue loading paused and the load reduced to 1.0 kN. At the same time, the half potential of steel was measured with the distance of 100 mm. The potential of steel bars in the reference beam was also measured. Measuring points are shown in Fig. 11. Fig. 12 shows the variation of the corrosion potential at different fatigue cycle number. Horizontal ordinate presents the distance of measuring points from the midspan. Minus sign means that measuring points are located at the left of the midspan. Corrosion potential of Beams CB-Ref and CFB-0.5 were measured at the same time.
Learned from Fig. 12, the initial value of the corrosion potential has almost reached the value of -350 mV. It means that the corrosion of steel bars is imminent or has already started. The reason is that those chloride ions have migrated to the steel surface and destroyed the passive film before the fatigue test during the period of beam arrangement and machine debugging. At the initial corrosion stage, the corrosion potential decreased significantly. After then, the corrosion potential continued decreasing at a lower rate. The corrosion potential of the steel bar distribution along the beam is not uniform. It reflects the non-uniformity of steel bars corrosion increasing the randomness of the failure location. The corrosion potential of steel bars at the shear bending areas of Beam CFB-0.6 and pure bending areas of Beam CFB-0.5 are relatively low. Meanwhile, the localized corrosions of the mentioned areas where the final fracture position happened are more serious.
The effects of different stress on corrosion can be analyzed by comparing the reducing rate of the corrosion po-   Fig. 11. Measuring points of the half cell potential test (unit: mm). MAO Jiang-hong et al. China Ocean Eng., 2018, Vol. 32, No. 2, P. 179-188 tential. Fig. 13 shows the variation of corrosion potential at the midspan and regression curves at different stress situation. The slope of fitted lines of Beams CB-Ref, CFB-0.6 and CFB-0.5 are respectively -1.094, -2.695 and -1.300. It indicates that the corrosion rate of the beams subjected to the fatigue loading is bigger than that without loading. And on the other hand, the corrosion rate of the beams subjected to larger fatigue stress amplitude is bigger.

Variation on deflection of RC beams
The maximum deflections (D max ) and the residual deflections (D res ) of the beams under different fatigue loading cycles are listed in Tables 6-9.
The fatigue loading cycles in Tables 6-9 are not the same. Because of the long-term process of the fatigue loading experiment, it is difficult to stop the machine at a constant time over different experiments. Although the materials of the four beams are the same, the maximum deflec-tions at the first static loading are different. Therefore, the deflections after the first static loading are taken as the baseline to analyze the data, and then subtract to define 'relative maximum deflection' and 'relative residual deflection'. The results are plotted in Figs. 14 and 15.    By comparing Beam CFB-0.6 with Beam FB-0.6, Fig.  14 shows that the relative maximum deflection of CFB-0.6 and FB-0.6 are very similar. The amount of the deflection grows quickly in the early loading (10000-50000 cycles), and then grows more slowly at later loading cycles. The relative maximum deflection of Beam CFB-0.6 was almost coincident with that of the maximum deflection of Beam FB-0.6 under the same fatigue cycles. Therefore, it can be considered that there is no significant difference between the two beams. By comparing Beam CFB-0.5 with Beam FB-0.5, Fig. 14 shows that there exist periods of the rapid growth and steady growth of the maximum deflections. However, the values of the relative maximum deflection between CFB-0.5 and FB-0.5 are evident, and the difference grows larger as the fatigue loading cycles continue. The maximum deflection of the RC beams is related to the rigidity of the cross-section. As discussed above, the corrosion cracks that existed in Beam CFB-0.5 could significantly reduce the rigidity of the cross-section, which could lead to this increase in the maximum deflection.
Residual deflections allow the beam to recover from the deformation during the fatigue experiment. Fig. 15 shows that the residual deflections of Beams CFB-0.5 and CFB-0.6 are larger than those of Beams FB-0.5 and FB-0.6 under similar fatigue loading cycles. The relative residual deflections between 50000 to 400000 fatigue loading cycles (which is considered as the second stage of the fatigue damage process) are linearly fit and the growth rate of the relat-ive residual deflections of Beams CFB-0.6 and FB-0.6 are 0.0019 and 0.0015 mm/10 4 cycle, respectively. This shows that the short-term corrosion has little impact on the deflection growth of the beams. However, for Beams CFB-0.5 and FB-0.5, the growth rates of the relative residual deflection are 9.327×10 -4 and 4.773×10 -4 mm/10 4 cycle, respectively. This illustrates that long-term corrosion can speed up the deflection development of the reinforced concrete beams and can degrade the flexural performance of the beam. The residual deflections of Beams FB-0.6 and CFB-0.6 are much larger than those of Beams FB-0.5 and CFB-0.5, respectively. This suggests that high loading amplitude can degrade the flexural performance of the beam and the corrosion of the reinforced steel can be more severe when coupling the long-term fatigue and corrosion.

Conclusions
The focus of the study is to test a fatigue flexural performance experiment of RC beams attacked by salt spray. An experimental method including electro osmosis, electrical-accelerated corrosion and salt spray was designed. First, a corrosion process for the fatigue loading beams was proposed. The chloride ions were migrated to the surface of the steel bar via electro osmosis. Then, a dry-wet cycle experiment was applied to simulate the actual salt spray environment. Finally, the electrical-accelerate corrosion was applied to synchronize the fatigue loading and concrete corrosion. This simulated corrosion process effectively reproduces the effects that the RC components feel from salt spray and the fatigue loading in the real world.
Four RC beams subjected to different stress amplitudes were tested. The different stress amplitudes led to different fatigue loading duration times and caused various degrees of corrosion. Beam CFB-0.5 subjected to 242 h of the corrosion duration time corroded more seriously than Beam CFB-0.6. There were corrosion cracks observed at the bottom surface of Beam CFB-0.5 and many corrosion pits were distributed around the steel bar extracted from Beam CFB-0.5. These phenomenon do not appear on Beam CFB-0.6 which was subjected to a shorter corrosion duration time. The experimental results show that the coupling of corrosion deteriorating and fatigue loading not only reduces fatigue life but also decreases the ability of deformation. The reduction in the fatigue life and deflection ability correlates with the stress amplitude and corrosion duration time.
The results revealed that a reasonable experimental design achieves the coupling of the corrosive deterioration and fatigue loading. They also provide sufficient data to evaluate the fatigue life and the concrete durability of the RC components located in salt spray environment. However, some improvements could be made to enhance the applicability. The fatigue loading spectrum for an indoor experiment could be optimized by using data from an actual bridge. In that way, a targeted evaluation could be  MAO Jiang-hong et al. China Ocean Eng., 2018, Vol. 32, No. 2, P. 179-188 187 achieved. A theoretical method or numerical simulation could also be useful to develop a strategy to guide the experiment. Finally, advanced structural health monitoring technology such as optical fiber sensors could be applied in the experiment in order to obtain life-cycle data.