Bond between GFRP bars and concrete. Literature review and evaluation of bond strength proposed by ACI 440.

João Bourbon Ribeiro de Noronha de Alarcão

Abstract The GFRP bar comprises glass fibers which are impregnated in a polymeric matrix composed of resin, filler and additives. These bars reveal as main advantages a high corrosion resistance and mechanical strength. Notwithstanding the aforementioned features, there are some drawbacks as well, GFRP bars cannot be changed on-site, its initial cost is relatively high, they show brittle behavior, and regulations for this type of bars is scarce. The present dissertation aimed to analyze the state of art on bond between steel/GFRP bars and concrete, as well as compare bond strength obtained in experimental tests described in the literature with the theoretical results of the expression proposed by ACI 440 (2006).

The  main issues addressed in the state of the art on the bond between both type of bars and concrete are: bond mechanisms, the failure modes and the bond test specimens. We also have studied the factors affecting bond behavior, analytical models and regulatory requirements. Subsequently it was developed a database comprising several studies on the bond behavior between GFRP bars and concrete, divided by type of experimental  assays:  beam  tests  and 

pullout tests. Finally, it was carried a comparison on the bond strength between the database and theoretical values obtained by equation 1 of ACI 440 (2006). It was concluded that the bond strength is lower in GFRP when compared to steel bars. It was also found that the expression of bond strength by the ACI 440 (2006) should be improved to take into account the type of surface on the GFRP bar and should also provide weightings for different length and diameter of the bar. 

Keywords: bond, concrete, bar, steel, GFRP. 

1   

1. Introduction The GFRP bar is a composite and anisotropic material composed by glass fibers which are impregnated in a polymer matrix that consists in resin, filler and additives, and the production process mostly used is pultrusion (Bakis et al., 2002). This production process ends with a superficial treatment aimed at increasing the bond to the concrete. That treatment may go through mechanic deformations applied in the surface or fiber filaments circled along the bar or in the sand coating, thus increasing the roughness of the bar. Thus, there are four types of GFRP bars: ribbed, twisted, sand coated and spiral wrapped. The values of the mechanical proprieties of the different types of GFRP bars, as the tensile strength longitudinal, the modulus of elasticity and ultimate strain longitudinal are mentioned in table 1. Table 1 – Physical proprieties of the FRP and steel bars (Source: Reis, 2009)

Property 

GFRP  Tensile strength [MPa]  450 ‐ 1600  35 ‐ 60  Modulus of elasticity [GPa]  Ultimate strain [%]  1.2 ‐ 3.7 

Material  CFRP  600 ‐ 3690  100 ‐ 580 

AFRP  1000 ‐ 2540  40 ‐ 125 

Steel  450 ‐ 700  200 

0.5 ‐ 1.7 

1.9 ‐ 4.4 

5 ‐ 20 

Although the use of GFRP bars solves the corrosion problem of the steel bars used in traditional concrete, other inconvenients may appear, such as the bond between GFRP bars and concrete. We started this thesis by collecting an extensive bibliography collection about the state of the art of bond of steel/GFRP bars for concrete. Subsequently we studied the state of the art of bond between steel and GFRP bars and concrete by themes, such as the bar-concrete bond mechanisms, the failure modes and the experimental evaluation of bond. We have also taken in account the factors that may affect the bond between the bar and the concrete, as well as analytic models and regulatory requirements. Lastly we set up a data base about the GFRP bars bond trials with the aim of comparing the bond strength obtained through those trials with the bond strength that would result from the ACI 440 (2006) suggested formula.

2. The state of the art of the bond between concrete and steel and GFRP bars 2.1 – Bond mechanisms The transferences of effort between the bars and concrete occur via three mechanisms: (i) chemical adhesion between the bar and the concrete, (ii) mechanical interlocking arising from the textures on the bar surface and (iii) frictional forces arising from the roughness of the interface between the bar and the surrounding concrete. In the description of the performance of the bond between the bar and concrete, despite the difficulty in isolating these mechanisms,

2   

it is possible to determine which is the predominant mechanism at a specific stage through the bond stress-slip curve (τ-s curve). The (τ-s) curve definition is also useful because it allows the necessary development length determination, ensuring correct effort transference and preventing, or slowing, the bond failure mode. Figure 1 illustrates a typical (τ-s) curve.

Figure 1- Bond stress-slip curve. (Source: Harajli et al., 2004)

2.2. Failure modes There are two bond failure modes: splitting and pullout. The concrete cracking failure happens when there is crack spread to the concrete’s outer surface, leading to bond loss. The bar pullout occurs when the concrete’s shear stress is reached between the bar’s ribs, enabling slip and extraction of the bar inserted in concrete. It is a failure mode that occurs mainly when bar is well confined and with adequate concrete cover, therefore resisting radial stresses, and/or when there is enough transverse reinforcement to slow or reduce concrete failure spread (Almeida Filho, 2006). 2.3. Bond test specimens The bond strength and the relative slip between the bar and the concrete are determined mostly by pullout and beam tests. The bond strength is calculated by dividing the tensile load applied to the bar until bond failure, by the bar lateral area. The pullout test consists of applying a tensile load on one of the bar’s end which is impregnated in a concrete prism. Then the bar displacements on the concrete prism ends, due to the tensile force applied, are measured. This force is increased until bar or bond failure is verified. The beam test is used when the aim is to determine the bearing capacity of the anchorage and splice zones of the bars that are submitted to tensile, in flexure beams. During the test, two identical and symmetric forces are applied, in relation to the profile steel, on the concrete blocks. As the load increase, we measure the respective displacements that occur in the bars free ends. Figure 2 and 3 illustrate the pullout and beam test specimens, respectively.

3   

Figure 2 – Pullout test specimen

Figure 3 – Beam test specimen

2.4 Factors Most of the factors which affect the bond of the steel bar with the concrete are the same that affect on the case of the GFRP bar. This factors are the type of bar material, the bar diameter, the development length and the type of bar surface, the concrete compressive strength, the bar position in the cast, the presence of transverse reinforcement and the concrete cover and bar spacing. 2.5 Analytical models of bond-slip behavior The bond stress-slip curves (τ-s) are obtained empirically through bond tests. These curves are the foundation for the definition of several analytical models. The analytical models for the behavior of bond between the steel bar and the concrete are: Tassios model (1979), BerteroPopov-Eligehausen model (1983) and the Model Code (2010). On the other hand, the analytical models regarding the bond of GFRP bar to concrete are: Eligehausen, Popov, and Bertero Model (BPE model) (1983), Malvar Model (1994), BPE Modified Model (1996) and CosenzaManfredi-Realfonzo Model (CMR Model) (1995). 2.6 Regulatory requirements The design codes that regulate the behavior of bond between the steel bar and the concrete are: REBAP (1983), CEB-FIP Model Code (1990), fib Model Code (2010), EC2 (1992) and ACI 318 (2008). In the other hand, the design codes witch regulate the behavior of bond between the GFRP bar and the concrete are: JSCE (1997), CSA S806 (2002), CNR (2006), ACI 440 (2006), CSA S6 (2006) e FIB (2007). The parameters that were analyzed are: development length and bond strength.

3. Development of a database of bond of GFRP bars to concrete The database is consists in 298 beam tests and 492 pull-out tests, in splitting mode of failure or pullout mode of failure.

4   

Data collected from beam tests in the literature were the following: paper’s authors, the resin of GFRP bars, presence of steel transverse reinforcement, the type of bar surface, the bar cast position, bar nominal diameter (mm), compressive strength of the concrete (fc, in MPa), modulus of elasticity of GFRP (E, in GPa), tensile strength of GFRP (ft, em MPa), concrete cover (c in mm), embedment length (lemb, em mm), bond strength (τm, in MPa) and failure mode. Regarding the pullout tests, the data collected were those of beam tests, except the presence of steel transverse reinforcement.

4. Evaluation of precision methodology proposed by ACI 440 (2006) for evaluating of the bond strength of GFRP bars to concrete The present chapter intends as a main object the comparison of the bond strength determined in the beam tests and the pullout tests which is the database, with the respective theoretical values obtained by the expression presented in the ACI 440 (2006), given by: 0.33 where: 

0.025

8.3

(1)

– concrete cover;



– bar diameter ;



– development length;



– concrete compressive strength.

The previous expression presented was determined resorting the beam tests with the bars placed in the bottom of the concrete element, based on a splitting failure. This comparison has in consideration the following elements: the test type and the failure mode, the surface type of the GFRP bars as well as the bar position in the cast. 4.1. Beam tests 4.1.1. Beam test with splitting failure On figure 4, we can see that the majority points (τEXP,τACI) are close to the unity gradient line. This shows that, in the majority of the tests, the experimental bond strength is the same of the theoretical value. The average of the ratio (τACI/τEXP) is 1.13. This fact goes against the security, since the theoretical results suggest that the failure occur for a superior load than the experimental value. The best adjustment occurs on tests with SW bars, while the tests with N bars have a worst adjustment. These are also the tests having a higher value of the average (τACI/τEXP), 1.19. Even though there are different ways of adjustment to the unity gradient line, it can be observed that the type bar surface doesn’t make a significant difference for this type of test and with this type of failure.

5   

In some tests, in Ehsani et al. (1996, 1997), occurs high ratio (τACI/τEXP). This can occur because for a certain value of normalized embedment length (lemb/db), the theoretical bond strength value doesn’t change, since the equation (1) depends on (db/lemb), for any bar diameter. But, in laboratory, keeping constant the relationship (db/lemb), it can be shown that increasing the diameter leads to a decrease of the experimental bond strength. So, if is only increased the diameter, the average of (τACI/τEXP) will also increase Analysing the totality of beam tests represented by the points (τEXP,τACI) on figure 4, it was proposed to substitute the coefficients from equation 1 (0.33, 0.025 and 8.3) for the next ones: 0.02, 0.144 and 8.63 respectively. Such modification leads to a ratio (τACI,rev/τEXP) of 1.05±0.24. One more time, the adjustment between theoretical results and experimental results is good. Only analyzing the points (τEXP,τACI) of the SW bars in beam tests illustrated on figure 4, it was proposed to change the coefficients of the equation 1 (0.33, 0.025 e 8.3) for: 0.34, 0.000 and 7.73 respectively. There is proposed only to change these coefficients for theses surface because of the higher approximation between the theoretical and experimental values. Such modification leads to a ratio (τACI,rev/τEXP)  of  1.04±0.13.  Despite the average of the ratio (τACI,rev/τEXP) done only for this type of bar surface are similar than the average of the ratio (τACI,rev/τEXP) considering the totality of tests, it can be seen a considerable decrease of the standard deviation. It is despised the influence of the (c/db) in the theoretical bond strength value, due the associate coefficient is 0. This is a normal conclusion, based on the fact that concrete splitting occurs preferably in tests with a low (c/db). The modification factor is calculated dividing the mean of the ratio (τACI/τEXP) of bars placed in the top by the mean (τACI/τEXP) of bars placed in the bottom of the concrete element. Considering the totality of the data, the modification factor obtained is 1.43, very closed with the suggested by the ACI 440 (2006), 1.5. Taken in account only the tests with SW bars that value is 1.22 and in the tests with bars N, the value is 1.82. 16,0 14,0

y = 0,8002x + 1,1973 R² = 0,7922

12,0 10,0 τACI [MPa]

8,0 6,0 4,0 2,0 0,0 0,0

2,0

4,0

6,0

8,0 τEXP [MPa]

10,0

12,0

14,0

16,0

 

Figure 4 - (τEXP,τACI) in beam tests with splitting failure, for type of bar surface: sand coated bars (SC), spiral wrapped bars (SW), ribbed bars (N) and spiral wrapped and sand coated bars (SW+SC).

6   

4.1.2. Beam test with pullout failure The figure 5 make a relationship with τACI and τEXP, obtained from beam tests with pullout failure and bars located on the bottom of the concrete element. Analyzing the figures 4 and 5, it is clear that, in beam tests, the failure by pullout of the bar occur, on average, for a bond strength markedly superior than the splitting failure, as mentioned on the chapter of the state of art. This difference occurs because of the high embedment length value of the bars on beam tests with splitting failure compared on the tests with pullout failure. In these, the embedment length values represents, on average, 36% of the bars tested on splitting failure. In figure 5, nearly up to τEXP ≈ 8 MPa, the most points (τEXP,τACI), are in the unitary gradient line, or above it, mainly on the points with lower τEXP. Typically, these points corresponds the tests with bars SW and SW+SC. From τEXP ≈ 8 MPa, there is a zone characterized by experimental results greater than the theoretical ones, independently of the bar surface. At the end there are a last group of points (τEXP,τACI) that corresponds to tests with SC bars, characterized by the higher dispersion and also by τACI > τEXP. These points are located above the unitary gradient line. However the big dispersion results also in an average of (τACI/τEXP) of 1.15, so the theoretical bond strength is 15% superior than the experimental bond strength. It is a similar value of the one verified for the beam tests with splitting failure, with the bars allocated on the bottom. Although the equal values of the averages of (τACI/τEXP), it can be seen a larger gap in the case that separate the tests by the bar surface type. Analyzing the totality of the beam tests in figure 5, the coefficients on the equation 1 (0.33, 0.025 and 8.3) will be substituted to: 0.00, 0.455 and 2.90 respectively. Such modification leads to a ratio (τACI,rev/τEXP) of 1.23±1.01, so it can be concluded that this modification conducts to a poor adjustment to the experimental results. It must be notice that the first coefficient change only depends on

. Comparatively with the coefficient values of (τACI,rev/τEXP) obtained by the

beam tests with splitting failure, it can be observed an increase of the value indexed to (c/db) (0.00 to 0.455) and a significantly decrease of the values indexed to (db/lemb) (8.63 to 2.9). The average and the standard deviation of the ratio (τACI,rev/τEXP) are superior comparatively than the ratio (τACI/τEXP), showing that the modification of the coefficients leads on a poor approximation of the tests results. Only the SW bars, allocated on the superior face of the concrete element, had a pullout failure in beam tests. The averages of (τACI/τEXP) are 1.07 and 1.18, for bars allocated in the inferior face and the superior face respectively. The modification factor is 1.10. This value is lower than the one obtained in the beam test with splitting failure (1.22). This means that in the beam tests with pullout failure of SW bars, the influence of the position of the bar on the concrete element is less significant than the beam tests with splitting failure.

7   

45 40 y = 0,9756x + 0,6635 R² = 0,4531

35 30 τACI 25 [MPa] 20 15 10 5 0 0

5

10

15

20

25

30

35

40

τEXP [MPa]

 

Figure 5 - (τEXP,τACI) in beam tests with pullout failure, for type of bar surface: sand coated bars (SC), spiral wrapped bars (SW), ribbed bars (N) and spiral wrapped and sand coated bars (SW+SC). 4.2. Pullout tests 4.2.1. Pullout tests with splitting failure In the Figure 6 below it is presented the relation between τACI and τEXP, as a result of pullout tests from the main database, with splitting failure, being the bars centered in the concrete element. It is noted in the majority of the tests, τACI > τEXP to a value of τEXP ≈ 10 MPa. Considering higher values, the opposite is observed, it is also worth mentioning that the expression proposed by ACI 440 (2006) underestimates the values in comparison to those experimentally obtained. The average and standard deviation of (τACI/τEXP) are 1.20 and 0.87 respectively, showing once more a considerable dispersion of results. As a consequence it becomes convenient to assess figure 6 by type of surface. Studying the total of beam tests relative to (τEXP,τACI) of figure 6, we suggest to replace the coefficients from equation 1 (0.33, 0.025 and 8.3), with the following values: 0.48, 0.061 and 7.24, respectively. As a result the ratio (τACI,rev/τEXP) becomes (1.32±0.87). In spite of the standard deviation’s value being preserved, the average of the new ratio is higher. Once again, the revision of equation 1 coefficients translates into a worse adjustment between theoretical and experimental results. The splitting failure only occurred in tests with SC bars, located in the top surface of the concrete element, tested in Esfahani et al. (2005). In the tests where the SC bars are located in the inferior surface, the average ratio (τACI/τEXP) is 1.06 resulting in a modification factor of 1.33 (1.41/1.06). This value is inferior to the ACI 440 (2006) proposal, 1.5. Therefore, a lower increment of development length in the bars located in the top surface, is needed, comparatively to ACI 440 (2006).

8   

20 18 16 14 12 τACI 10 [MPa] 8 6 4 2 0

y = 0,423x + 5,804 R² = 0,5438

0

2

4

6

8

10 12 τEXP [MPa]

14

16

18

20

22

Figure 6 - (τEXP,τACI) in pullout tests with splitting failure, for type of bar surface: sand coated bars (SC), spiral wrapped bars (SW), ribbed bars (N) and spiral wrapped and sand coated bars (SW+SC). 4.2.2. Pullout tests with pullout failure The figure 7 presents the results from pullout tests with centered bars, in which the failure occurred by pullout. Comparing figures 6 and 7, it is observed that the failure by pullout is more expressive in these types of tests. This is the more likely mode of failure when the standardized cover (c/db) is high, and a low normalized embedment length to the bar diameter (lemb/db). Evaluating figure 7, we can observe that the theoretical bond strength is lower than the correspondent experimental value, in the majority of assays with τEXP >17 MPa. For that reason, the expression proposed by ACI 440 (2006), does not provide a good adjustment for experimental results of a higher value, independently of the bar surface. In average the theoretical bond strength is higher in 14%. As a result of the analysis we suggest a replacement in equation 1 (0.33, 0.025 e 8.3) by the following values 0.67; 0.068 and 5.84. The result is a (τACI,rev/τEXP) de 1.25±1.05. Once again the average of the revised mean manifests a higher value. The means of (τACI/τEXP) to SW and SW+SC bars in pullout tests, with the bars being in the top surface of the concrete element and with pullout failure are 1.07 and 1.13 respectively. Hence, the modification factors to apply in tests with SW and SW+SC bars are 1.13 and 1.28 respectively. Again, the modification factor calculated in assays with these types of surfaces is inferior to 1.5, value proposed by ACI 440 (2006).

9   

30 y = 0,4168x + 7,0993 R² = 0,3578

25 20 τACI 15 [MPa] 10 5 0 0

5

10

15 τEXP [MPa]

20

25

30

Figure 7 - (τEXP,τACI) in pullout tests with pullout failure, for type of bar surface: sand coated bars (SC), spiral wrapped bars (SW), ribbed bars (N) and spiral wrapped and sand coated bars (SW+SC).

5. Conclusion According to the analysis given to the state of the art, it is concluded that the bond strength of steel bars to concrete is superior to GFRP bars. The bond strength recommended by ACI 440 (2006) has a better adjustment in beam tests with splitting failure. As a matter of fact, only in figure 4 results the new coefficients of equation (1) translate into a better approximation between theoretical and experimental results. This consideration was expected, since the theoretical expression was based in these types of tests. It is also concluded that the theoretical expression must encompass the surface type of the GFRP bar. The failure by concrete splitting tends to occur when the bond strength is inferior in comparison to the failure by pullout. This consideration was verified in both types of tests. Through the analysis of Figures 4-7, it is concluded that in the case, reaches a high value, τEXP>τACI, being the prediction in the safety side. This behavior was not so clear in rehearsals with sand coated bars. The expression of ACI 440 (2006) should not estimate a coefficient to (db/lemb). The reason for this is the following, if the bar length is derived in function of its diameter, if the latter increases the value of the theoretical bond strength remains while the experimental value decreases. In relation to the correction factors, it was monitorized that only in beam tests with ribbed bars, and where the failure occurred by concrete splitting, the experimental value was higher than the ACI 440 (2006). Nevertheless, more tests should be developed in order to reassure this tendency. The lower modification factor was determined in beam and pullout tests with SW bars, with values of 1.10 and 1.13 respectively.

10   

Reference Almeida Filho, F. M. (2006), “Contribution to study of the bond between steel bars and selfcompacting concrete”, PhD Thesis, Universidade de São Paulo, Brazil (in Portuguese). ACI Committee 440 (2006). “Guide for the design and construction of structural concrete reinforced with FRP bars (ACI 440.1R-06).” American Concrete Institute, Farmington Hills, MI, 44 pp. Ehsani, M. R., Saadatmanesh, H. e Tao, S, “Design recommendations for bond of GFRP rebars to concrete”, Journal of Structural Engineering-ASCE, Vol. 122, No. 3, pp. 247-254, 1996. Ehsani, M. R., Saadatmanesh, H. e Tao, S, “Bond behavior of deformed GFRP rebars.” Journal of Composites Materials, Vol. 31, No 14, pp. 1413-1430, 1997. Esfahani, M. R., Kianoush, M. R. and Lachemi, M., “Bond strength of glass fiber reinforced polymer reinforcing bars in normal and self-consolidating concrete”, Canadian Journal of Civil Engineering, Vol. 32, pp. 553-560, 2005. Focacci, F., Nanni, A., Fellow, ASCE e Bakis, C. E., “Local bond-slip relationship for FRP reinforcement in concrete”, Journal of Composite for Construction, Vol. 4, No. 1, pp. 24-31, 2000. Reis, V. L. F. (2009), “Construction of concrete with FRP bars”, Thesis for the degree Master in Civil Engineering, FEUP, Porto.

 

11