Bond strength degradation in concrete cracked by expansion agent filled pipes

In this study, the bond behavior of corroded reinforced concrete was investigated by conducting pull‐out tests on concrete specimens with variable induced crack widths and different splitting modes. To simulate cracks due to corrosion and focus on their effect on bond strength degradation, a novel method using expansion agent filled pipes is proposed. With the increase of the crack width over elapsed time from filling the aluminum pipe with an expansion agent, a target crack width was obtained. The results confirm that the type of splitting greatly influences the reduction in bond strength with deterioration of bond strength more severe in “Side‐split” than in “Single‐split.” Furthermore, using the surface crack width as a variable, empirical formulas to predict bond strength degradation due to corrosion of rebar are proposed. These prediction models give a good correlation when compared to the available literature.


| INTRODUCTION
The bond at the steel-concrete interface is one of the most fundamental properties of reinforced concrete (RC). A large volume of publications study the bond behavior of steel reinforcement bars (rebars) in sound concrete. [1][2][3][4][5][6] Also, researchers have attempted to evaluate the impact of damaged concrete on bond properties. The cracking of concrete by loads or corrosion can negatively affect the bond performance between concrete and rebar. [7][8][9] Rebar corrosion is among the most devastating and expensive deterioration phenomena of RC structures. 9 The past thirty years have seen increasingly rapid advances in research into the degradation of bond strength caused by the corrosion of rebars. Some of them have developed predictive models, which are in good agreement with test results. [10][11][12][13][14][15][16] Many studies used accelerated electrical corrosion techniques to assess the degradation of bond with corroded rebars. The relevant literature is reviewed by Lin et al. 17 These authors primarily used the level of corrosion as the main variable to evaluate bond degradation. But these variables are difficult to measure in existing structures under service conditions and unsuitable for practical use. Also, the experimental data are rather controversial. The different typology of specimens or the impressed current density typically caused scattering of the data.
Discussion on this paper must be submitted within two months of the print publication. The discussion will then be published in print, along with the authors' closure, if any, approximately nine months after the print publication.
Alternatively, other authors have focused on the direct relationship between bond degradation and corrosion-induced crack width. [18][19][20][21] They suggest that the crack width is the dominant factor and an excellent parameter to assess bond degradation. Also, fib Model Code 2010 22 provides guidelines to evaluate bond degradation with corrosion crack width. The authors propose a rough association between bond strength decrease and surface crack width. Based on a large number of pull-out tests, the chart gives the variation of bond strength degradation corresponding to a range of crack widths (see Table A1 in Appendix). However, in electrical corrosion, the induced crack width distributes arbitrarily along the specimen because the rebar corrosion spreads nonuniformly. 23 It has also been observed that the impressed current density can influence the quantity of corrosion products, which can lead to a different crack pattern. Moreover, a higher corrosion rate led to bigger cracks and poorer bond strength. 24 The bond degradation is also higher in artificially corroded specimens than naturally corroded specimens. 20 Therefore, models based on the corrosion crack width have not yet been widely adopted because a well-assessed relationship between the crack width and the corrosion level is yet unavailable.
The corrosion rebar changes the surface properties of the steel, the rebar shape, and cracks the surrounding concrete. 17 This phenomenon altogether affects the bond strength of the corroded specimen. However, the combination of these effects leads to difficulties in analyzing the processes at a fundamental level and negates the overall accuracy of the proposed models. Further, studies show that induced crack width plays a more important role in the bond deterioration mechanism than the corroded rebar shape and rust accumulation. 19,25 All of the studies reviewed here support the hypothesis that the surface crack width may be directly linked with the bond deterioration. The alternative approach proposed here will focus on the single effect of crack width on the bond mechanism through interlocking ribs. While ignoring the ambiguity related to corrosion product or rate, the crack width enable quantification of the exclusive effect of corrosion on bond strength. Much uncertainty still exists about the direct relationship between crack width and bond degradation. So, further investigations in this respect are needed.
The study aims to increase the knowledge about the bond behavior in cracked RC structures. Specifically, the link between visual inspection data (crack width and position) and bond degradation is investigated. To focus on the more fundamental effect of the cracking itself, first, we developed a novel method for simulating cracks. Then, we created two series of cracked concrete which are designed to fail either in side-split (leading to delamination) or single-split (leading to spalling) when subjected to a pull-out test. Finally, an empirical model is built from the results and compared with corroded specimen data in the literature.

| CRACK SIMULATION BY EXPANSION AGENT FILLED PIPE
One of the most well-known tools for assessing the service life of RC structures is the monitoring of cracking due to corrosion of rebar. When corrosion occurs, depending on the arrangement of the rebars within the member, the crack location and progression differ. These phenomena considerably affect the bond properties. 26 Figure 12 shows two possible induced crack patterns due to corrosion, which are modeled in this research.
As described previously, the accelerated corrosion techniques have been adopted in several studies on the bond degradation of cracked concrete specimens. In these techniques, it can be challenging to control the width or location of the crack.
Yasojima et al. 27 studied the bond of rebars in precracked specimens. The cracks were formed by inserting thin slits (e.g., propylene sheet). Using this approach, they have been able to simulate internal cracking. In bond tests, while the use of slits is easy to implement, the relationship to real cracks is unclear.
A nonexplosive demolition agent (expansion agent) is mainly used for the destruction of rocks and RC structures. In powder form, it expands when humified. To simulate the expansion of rebars due to corrosion, an aluminum pipe embedded in concrete is filled with an expansion agent (as illustrated in Figure 1). Due to this expansion, cracks are generated in the concrete, as can be seen in Figure 2. The mechanism of crack formation is close to one caused by the expansion of rebars due to corrosion. But a note of caution is due here since the speed to induce crack with expansion agent filled pipes (EAFP) may modify the cracking process.
EAFP allows focusing on the effect of the cracking itself while ignoring the other the effect of corrosion (section loss or rust around the rebar). It is also particularly useful in controlling the cracks (width and position) in a relatively short time. The next subsections highlight this method. Figure 3 shows the specimens used to test the cracking by EAFP. Its dimensions were 150 Â 150 Â 400 mm 3 . An aluminum pipe with 18 mm as outer diameter and 1 mm thickness was embedded in the concrete. The crack width was measured with π-type displacement transducers placed in three positions on the specimen. Table 1 shows the list of specimens. The concrete cover size was set as the main parameter. The six specimens were divided into two groups with different concrete cover (15 and 20 mm). Figure 4 exhibits the crack induced by EAFP. This method cracks the cover side of the specimen along the aluminum pipe. Table 2 shows the average and maximum crack width after 250 h. Figure 6 shows the examples of the relationship between the crack width and elapsed time. In all specimens, the crack width increased over time. The induced crack appeared to be slightly higher in the 20 mm concrete cover. The same tendency was observed by Lin et al. 18 using a high-current density(600 μA/cm 2 ) to accelerate corrosion. However, given that the shallow depths and the small difference, cover depth might not be the main factor governing the surface crack width.

| Confirmation of cracking simulation
To simulate specific cracks width, the elapsed time and crack width relationship shown in Figure 5 could be used. Table 3 presents the specimen list. To better observe the deterioration of the bond, the main parameter is the induced crack width. Also, to investigate the influence of the splitting mode at failure, a total of 27 test specimens were used in series I and II.

| Test series
In series I, the specimen is designed to fail in "Singlesplit." In those specimens, a single crack is simulated in the plane parallel to the longitudinal axis of the bar. The pull-out bar is the ribbed EAFP.
Series II focuses on "Side split," where the induced cracks are located in plane perpendicular to the longitudinal axis of the rebar. Two induced cracks are simulated by two EAFPs located around the D19 rebar.
Here, the effect of concrete strength (18 and 30 MPa) was also investigated. The pull-out bar is the D19 steel rebar. Figure 7 shows the design of the Series I specimen. Its dimensions were 260 Â 260 Â 82 mm 3 , and the aluminum pipe with ribs was embedded at 38 mm from the specimen side. A short bond length of 51.6 mm was chosen to focus on the local bond behavior. To avoid cone failure of concrete, a unbonded part was set by only limiting the ribs in the bond length. A M6 coupler was fixed 100 Â 100 mm 2 at the up-left position to attach a linear variable displacement transducer (LVDT) for measuring the slip at the free end of the pipe. The specimen framework is shown in Figure 6. Figure 7 shows an overview of a machined aluminum pipe with ribs geometry designed according to Japanese is embedded at 47.5 mm from the specimen cover side. To focus on the local bond, a short bond length of four times the diameter of the rebar equal to 76 mm was chosen. Two EAFPs (diameter: 22 mm and thickness: 1 mm) were set to 50 mm from the rebar to simulate the cracking of surrounding concrete. To avoid cone failure of concrete, the bonded length was positioned in the middle of the specimen, and an unbonded length of 47 mm at either side was set. The M6 coupler was fixed 100 Â 50 mm 2 at the upright position to set an LDVT for measuring the slip of the free end of the bar. The reinforcement is a deformed bar of a nominal diameter of 19 mm.

| Crack simulation by EAFP
The ratio of the water to an expansion agent was set to 30%. The specimen was placed to set vertically the longitudinal axis of the pipe. In Figure 9, the expansion agent was filled from the top of the pipe. The crack width increases over elapsed time after filling. Thus, this time is controlled to obtain the target crack width.

| Loading and measurement method
The cracks induced by EAFP were sketched, and their widths were measured. Then the specimens were subjected to monotonic pull-out loading with a universal testing machine. Figure 10 shows the general setup for the pull-out test. The specimen was placed on the Teflon sheet to limit the friction. Also, not to restrict the lateral deformation of concrete, the loading plate included a hole with a diameter corresponding to the concrete cover. The tested bar was subjected to monotonic pullout loading at a speed of 0.5 mm/min as recommended by ASTM234-91A. The measurement items are pull-out load and slippage of the pipe or D19 rebar at the free end.

| RESULTS OF EXPERIMENTS
All experimental results are summarized in Table 4. Specimens S.II.18.NC and S.II.30.NC were subjected to pull-out loading without filling the expansion agent, so there was no induced crack by EAFP in these specimens. Figure 11 shows examples of crack patterns after filling the expansion agent. The maximum crack width before pull-out loading in every specimen can be seen in Table 4.

| Failure mode
All specimens experienced failure due to splitting of the concrete. A group failed by newly generated crack and others by the widening of crack induced by EAFP, as shown in Figure 12.
In the case of series II, some specimens presented newly side crack (named N1 with one new crack and N2 with two new cracks). The specimens without induced crack failed by single splitting with crack on the cover side. The induced crack can switch the failure mode from single to side-splitting. Moreover, in 30 MPa specimens, a small crack width (0.15 and 0.25 mm) even leads to a single splitting. This may be explained by the fact that concrete does not lose all strength at once on cracking, and some ability to transmit stresses across cracks is retained. Figure 13 shows the pull-out and slippage curves. In series I, the specimens with crack width from 0.1 to 1 mm, indicated almost linear relationships between load and slip up to the maximum load (see upper-left Figure 13). After that, the load rapidly drops while the slip keeps increasing. The sudden drop in the load-slip curve at the maximum load is likely to be related to the brittle splitting of the concrete cover. In addition, no significant correlation was found between the slip at maximum load and the induced crack width. Further, when the crack F I G U R E 8 Series II specimen detail width is higher than 2 mm, the stiffness is negatively affected by the induced crack width and significant slip occurred at maximum load (see the upper-right figure). And the maximum load decreases until 4.11-3.77 kN for 1.5 and 2 mm crack width, respectively. These observations were also reported by Cabrera et al. 28 These big cracks severely affect the confinement provided by the concrete cover. Consequently, the bearing action between the rebar rib and concrete is deteriorated and the deformed bar become similar to a round one.

| Pull-out load and slippage
In series II, the maximum load decreases up to 90% when the crack width increases, whereas Similarly, the corresponding slip at maximum load shows a decreasing tendency (see in lower Figures 13). At the same applied load, less slip was obtained for the 30 MPa specimens than for the 18 MPa specimens. However, as soon as the maximum is reached, a larger slip is seen for 18 MPa specimens, resulting in a more gradual descending of the curve. A possible explanation for this might be that the 30 MPa concrete has a stiffer response and shows brittle behavior ( Figure 13). Figure 14 summarizes the relationships between the maximum pull-out load and induced crack width. There is a clear trend of decreasing the maximum load when the crack width increases. As expected, a significant correlation between residual bond strength and induced crack width can be seen from this result.

| Bond degradation and surface crack width relationship
In this section, residual bond strength ratio, which is defined as the ratio of the bond strength of a cracked specimen to that of the uncracked specimen, is evaluated in order to better understand the process of bond degradation. In series I, for uncracked concrete, the specimen failed due to the rupture of the aluminum pipe before pulling out. Consequently, the bond strength of these specimens could not be obtained. Therefore, the following equation, as reported in the previous study, 29 is used to calculate the pull-out splitting strength of uncracked specimen: F I G U R E 9 Filling of the expansion agent F I G U R E 1 0 Loading and measurement setup where, τ b,max : bond splitting strength, σ t : splitting tensile strength of concrete, r u :C + d b /2, d b : diameter of the pipe (19 mm), C: the thickness of cover concrete, α: the angle between the longitudinal axis and splitting force (=34 degrees). For series II specimens, the bond strengths of uncracked concrete are 48.44 and 52.35 kN for 18 and 30 MPa specimens, respectively. Figure 15 displays the relationships between residual bond strength ratio and crack width. The following formulas were obtained by the regression analysis, where the coefficient of correlation (R 2 ) was 0.91 and 0.82 for series I and series II, respectively. Where P (Wcr) : bond strength in cracked concrete; P 0 : bond strength of specimen without crack; W cr : crack width, in mm. Figure 16 compares the two prediction formulas. In the range of 0.15-1 mm crack width, the formula gives a residual bond strength ratio from 0.91 to 0.54 for singlesplitting specimens. However, with in the same crack width range, for side-splitting specimens, the residual bond strength ratio deteriorates to between 0.73 and 0.12. Moreover, when the crack width reaches 2 mm, the sidesplitting specimens have a residual bond strength ratio close to zero. This is in contrast to single-splitting specimens, for which it remains at around 0.30. It can be seen that the deterioration of the bond strength is more severe in side-splitting specimens (series II) than in singlesplitting specimens (series I). This result confirms the fundamental influence of the type of splitting in reducing the bond strength. This may be explained by the fact that the wedging action of ribs causes circumferential tensile stresses in the concrete surrounding the bar and tends to split the concrete along the weakest plane. The position of cracks in series II specimens seems to create a weaker side-splitting plane when compared to the single splitting one.

| Comparison of the test results with fib model
The fib Model Code 2010 22 introduces the reduction in bond strength depending on the surface crack width (hereafter, called fib model). For a certain range of surface crack width, the variation of the residual bond strength is suggested (see Table A1 in Appendix). To illustrate the upper and lower limit of the bond F I G U R E 1 1 Crack after filling the expansion agent degradation provided by fib model, Figure 17 plots the degradation range against the median values of the crack width range.
In series I, the prediction with the fib model is in good agreement with the test results for specimens with crack width smaller than 0.5 mm. However, the fib model gives underestimations for the specimens with crack widths between 0.5 and 1.5 mm.
A comparison of the series II results reveals that the prediction with fib model overestimates for 18 MPa specimens. The overall deterioration of our specimen was found to be 15% (crack width smaller 0.4 mm) to 40% (crack width bigger than 0.75 mm), lower than that of fib model code. However, for 30 MPa specimens, the fib model is in relatively good agreement with test results.
A comparison of the test results with the prediction given by the fib model shows that the decrease in bond strength remained largely on the safe side. It should be noted that the authors of the fib model have assumed that the residual strength of concrete structures is also affected by the cross-section loss of both steel and concrete. However, our study did not include the effect of the rebar section change or rust accumulation.

| Comparison of the proposed formulas with experimental results in the literature
To validate the effectiveness of the proposed model, a comparison with relevant experimental data available in the literature is carried out (Table A2 in Appendix). To induce cracks, specimens in these studies were electrically corroded to various degrees.
In Figure 18(a), the models are compared with the results of Lin et al. 18 (hereafter, called Lin-SA0 and Lin-SC0). The test was performed on eccentric pull-out specimens with a different concrete cover. Longitudinal rebar of 20 mm in diameter with 100 mm as the bond length was used. The concrete cover was 35 mm (1.5D) and 60 mm (3D) for Lin-SA0 and Lin-SC0, respectively. It can be seen that our model series II is in good agreement with those obtained in Lin-SA0 and Lin-SC0. An example of corrosion-induced crack patterns from Lin et al. can be seen in Figure 19(a). Focusing on induced crack location, this study suggests that side-splitting is more dominant when longitudinal and transversal induced cracks are both present. In the test by Zhao et al. 30 in Figure 18(b), pull-out was carried out with the reinforcing bars (18 mm in diameter) centrally embedded in the 150 mm cubic specimens (hereafter, called Zhao). The bond length was 100 mm. Our model from Series I is in good agreement with Zhao, and, thus, can be considered suitable to evaluate the bond deterioration when the induced crack is only along the rebar. This confirms the importance of crack location in evaluating the bond degradation. Figure 18(c) shows the comparison with test results of Law et al. 22 Beam end specimens which had a dimension F I G U R E 1 3 Pull-out load versus free end slippage curve F I G U R E 1 4 Maximum pull-out load versus crack width of 200 Â 200 Â 300 mm 3 , were used in pull-out test. Two types of rebars with a diameter of 12 and 16 mm were employed, and the concrete cover (3D) was 36 and 48 mm, respectively. An example of corrosion-induced crack patterns from Law et al. can be seen in Figure 19(b). For specimens with induced crack width smaller than 0.5 mm, the model predictions from series II are in good agreement with the test results. However, when the crack width is bigger than 0.5 mm, the model tends to slightly underestimate the bond strength. This can be attributed to section loss or change of interfacial layer due to corrosion.
Wang et al. 31 propose an interesting formula to evaluate the bond degradation in corroded RC. Based on beam end test, the proposed formula integrates two parameters (crack width and the ratio between concrete cover and bar diameter) as reported in the review by Lin et al. 17 It is expressed in Equation (4).
Where τ u w ð Þ : bond strength in cracked concrete; τ u : bond strength of specimen without crack; w: crack width, c: concrete cover and d: rebar diameter, in mm.
In Figure 18(d), our proposed formulas are compared with the model proposed by Wang et al. It can be seen that their model shows a very good correlation with our prediction from series II.
These experimental results highlight the significant influence of the single effect of induced crack width. Also, the form of cracks inside the concrete members has a great influence on the residual bond strength and should be taken into account in models of degradation. This study illustrates how corrosion-induced crack widths and locations can be used as a damage indicator, which may be of assistance in estimating the bond degradation.