Optimisation of carbon and glass FRP anchor design

Construction and Building Materials 32 (2012) 1–12

Contents lists available at ScienceDirect

Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat

Optimisation of carbon and glass FRP anchor design H.W. Zhang a, S.T. Smith a,⇑, S.J. Kim b,1 a b

Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China Industrial Composite Contractors (ICC) Pty Ltd., 15 Pavers Circle, Malaga, Western Australia, Australia

a r t i c l e

i n f o

Article history: Available online 12 June 2011 Keywords: Anchors Bonding Concrete Fibre reinforced polymers Shear Strength

a b s t r a c t The strengthening and repair of reinforced concrete (RC) civil infrastructure with externally bonded fibrereinforced polymer (FRP) composites is an established technology. One of the limitations of this technology though is the propensity of the FRP to prematurely debond at strains well below its rupture strain. Anchorage of the FRP strengthening is therefore a logical remedy to prevent or delay debonding failure. A promising type of anchor, which is made from bundles of fibre or rolled fibre sheets and can be applied to virtually any shaped structural member, is the FRP anchor. Limited fundamental characterisation of FRP anchors has, however, been undertaken to date. In this paper a series of tests aimed at optimising the FRP anchor design within the confines of the experimental variables are reported. The anchors are tested in an FRP-to-concrete single-lap joint shear test set-up and the main parameters varied are fibre type, fibre content and method of anchor construction. The behaviour as well as strength and failure of the test specimens are discussed and generic load–slip responses are provided which can form the framework for the future development of analytical models. Criterion with which to assess the optimal design of FRP anchors is also provided. Ó 2011 Published by Elsevier Ltd.

1. Introduction Reinforced concrete (RC) members strengthened with externally bonded fibre-reinforced polymer (FRP) composites are susceptible to premature failure by debonding of the FRP. An effective means to delay or halt the propagation of debonding in an FRP strengthened member is by anchorage of the FRP strengthening with FRP anchors [e.g. 1–7]. An FRP anchor (also referred to as an anchor) is essentially a rolled fibre sheet or a collection of bundled fibre strands in which one end of the anchor is inserted into an epoxy filled hole in the concrete member and the other end is threaded through and splayed onto the surface of the FRP strengthening plate (herein plate unless indicated otherwise). Fig. 1 is a schematic representation of a typical installed anchor system, in the context of an FRP shear-strengthened RC beam, in which the individual components of the anchor are clearly identified. Of importance to note in Fig. 1 is the orientation of the anchor fan in the main load carrying direction of the U-jacket. A much more detailed background on the application of FRP anchors to FRP-strengthened RC members is provided in Smith et al. [6]. Research on the fundamental strength and behaviour of FRP anchors in isolation has, however, been much more limited and such research has been predominantly experimental in nature. Studies ⇑ Corresponding author. Tel.: +852 2241 5699; fax: +852 2559 5337. 1

E-mail address: [email protected] (S.T. Smith). Formerly of affiliation (a).

0950-0618/$ - see front matter Ó 2011 Published by Elsevier Ltd. doi:10.1016/j.conbuildmat.2010.11.100

to date have reported the pullout strength and behaviour [8–11] as well as the shear strength and behaviour [12–15] of FRP anchors. It is the result of such limited fundamental understanding and associated design guidance, especially in the case of the shear strength and behaviour of FRP anchors, which is hindering the rational design and wide-scale use of FRP anchors. In a bid to examine the strength and behaviour of FRP anchors, Smith and Kim [15] reported the results of 20 FRP-to-concrete joint tests anchored with FRP anchors. The main variables in these tests were (i) method of plate and anchor installation, (ii) anchor fibre content, and (iii) anchor position. The tests showed the anchored joints to fail in predominantly two different modes, namely (i) complete debonding of the FRP plate followed by near simultaneous failure of the anchor (known as Mode 1 failure), and (ii) complete debonding of the FRP plate followed by failure of the anchor after substantial slippage of the debonded plate (known as Mode 2 failure). Within each of these two major modes the anchors were found to fail differently. More specifically, detailed classifications of all failed specimens were proposed by Smith and Kim [15] as Mode 1A (herein simultaneous plate debonding and anchor rupture), Mode 1B (herein simultaneous plate debonding and anchor fan debonding), Mode 2A (herein plate debonding followed by anchor rupture), Mode 2B (herein plate debonding followed by anchor fan debonding), and Mode 2C (herein plate debonding followed by anchor pullout). The main test variables considered in the experimental study reported herein are aimed at optimising the performance of the

2

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

RC member Anchor dowel FRP plate Anchor fan

(a) U-jacket shear-strengthening

(b) Section of FRP anchor and plate

Fig. 1. Schematic of FRP-strengthened RC member and anchor.

FRP anchor. Such variables consist of (i) anchor fibre type, (ii) anchor fibre content, and (iii) method of anchor construction. In addition, criterion is established with which to optimise the anchor design upon. All test specimens are extensively instrumented with an array of linear variable displacement transducers (LVDTs) and electric strain gauges. The resulting experimental behaviours and failure modes are presented in addition to selected LVDT and strain results. Finally, generic load–slip responses are proposed which form the basis of future load–slip and bond-slip model development.

2. Experimental details 2.1. Details of test specimens and test set-up The experimental program consisted of 27 FRP-to-concrete single shear joint tests in which three joints were unanchored (i.e. control joints) and 24 joints were anchored with a single anchor. The resulting test matrix is given in Table 1 in which the three main variables are evident, namely, (i) fibre type (i.e. glass or carbon), (ii) anchor fibre content (fibre sheet widths of 134, 200, and 259 mm), and (iii) method of anchor construction (dry or impregnated anchor). Three tests series were established, namely (i) Control Series, (ii) Glass Series, and (iii) Carbon Series. In addition, three tests were undertaken for each set of variables for repeatability. The Control Series tests on unanchored FRP-to-concrete joints were quite standard and numerous experimental studies have been reported in the open literature on such testing e.g. [16]. No further comment is therefore warranted here. For the Glass Series, two different types of anchors were tested. The anchors were either dry FRP anchors or impregnated FRP anchors. Detailed descriptions of the dry and impregnated FRP anchors are contained in Section 2.2 although brief comments are included as follows. Dry anchors were essentially formed from rolled fibre sheets in which the epoxy was omitted in the manufacturing stage. Impregnated anchors were also formed from rolled fibre sheets, however, the anchor dowel component of the anchor was impregnated with epoxy during the anchor

manufacture stage so as to form a solid ‘dowel’. In the case of the glass anchor series, the same width of fibre (i.e. 200 mm wide glass fibre sheet) was used in order for the method of anchor construction to be directly compared. For the Carbon Series, both dry and impregnated FRP anchors were also tested and the anchors were manufactured in the same manner as the Glass Series anchors. In addition, three different amounts of fibre were used in the construction of the carbon anchors for comparison with the glass anchor counterparts. The three different carbon fibre sheet widths corresponded to equivalent glass anchor properties of (i) equal tensile force capacity (based on flat FRP coupon properties) (i.e. 134 mm wide sheets), (ii) equal fibre sheet width (i.e. 200 mm wide sheets), and (iii) equal fibre content (based on equal cross-sectional area of fibre sheet) (i.e. 259 mm wide sheets). The FRP coupon specimens were made from impregnating the fibre sheet with epoxy of which more detailed information is contained in Section 2.4. Table 2 provides a detailed summary of the tensile properties of the tested glass FRP and carbon FRP coupon specimens though in addition to the nominal fibre sheet thickness and the equivalent fibre sheet widths. Also contained in Table 2 is the amount of carbon fibre to produce an equal axial rigidity as the glass fibre anchor in which the axial rigidity is defined as the elastic modulus of the FRP coupon multiplied by the fibre sheet cross-sectional area. As this value (i.e. 121 mm in Table 2) is quite similar to the carbon fibre sheet width to produce equivalence in FRP coupon force capacity (i.e. 134 mm in Table 2), tests on anchors constructed from such amount of fibre have been omitted from the test program as there is not expected to be any noticeable difference to anchors made from 134 mm carbon fibre sheet width. The test set-up and a typical test specimen are shown schematically in Fig. 2a and b, respectively. All concrete prisms were nominally 200 mm wide, 200 mm deep and 400 mm long. The FRP plates were formed from the same roll of carbon fibre used for the construction of the carbon FRP anchors. The plate, of 50 mm width and 250 mm bonded length, was made from three layers of carbon fibre sheet in a wet lay-up procedure. For such an arrangement, the effective bond length (i.e. the limiting length of plate with which an increase in plate length will not increase the strength of the joint) was 111.7 mm when calculated in accordance with Chen and Teng’s [17] commonly referred to model. An unbonded zone of 40 mm was maintained at the loaded free end of the concrete prism in order to reduce edge effects of the concrete prism and to minimise the formation of a concrete wedge at failure. In all anchored joint tests, the anchor was positioned 75 mm from the bonded loaded end of the plate (i.e. 115 mm from the loaded free end of the concrete

Table 1 Test matrix. Test series and specimen identificationa

Concrete b

Control Series Glass Series

Carbon Series

a b

CN-1–3 GD-200-1–3 GI-200-1–3 CD-134-1–3 CD-200-1–3 CD-259-1–3 CI-134-1–3 CI-200-1–3 CI-259-1–3

FRP plate

FRP anchor

fcu (MPa)

Width (mm)

Bond length (mm)

Width (mm)

Fibre type

Anchor type

Fibre sheet width (mm)

50.3 50.3

200 200

250 250

50 50

– Glass

– Dry Impregnated Dry

50.3

200

250

50

Carbon

– 200 200 134 200 259 134 200 259

CN = control, G = glass fibre, C = carbon fibre, D = dry anchor, I = impregnated anchor, 1–3 = specimens 1, 2, 3. fcu = concrete cube compressive strength (cylinder strength  0.8 fcu).

Impregnated

3

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12 Table 2 Fibre properties and anchor fibre content.

a

Fibre sheet width (mm)

FRP property

Fibre sheet width (mm)

1595

–

3090

–

94.1

–

201.4

–

19,772

–

15,172

–

0.17

–

0.131

–

200

–

121

–

200

–

134

–

200

–

200

–

200

–

259

–

a

15

FRP property

25

40

Carbon fibre anchor

12 14

(a) Dry anchor Anch nchor or fan an

FRP plate

15

50

Anchor bend ben

40

FRP rupture strength (MPa) FRP elastic modulus (GPa) FRP rupture strain (le) Nominal fibre sheet thickness (mm) Equivalence in axial rigidity (EA) Equivalence in force capacity Equivalence in fibre sheet width Equivalence in fibre cross-sectional area

Glass fibre anchor

25

Property

50

– = not applicable to study.

12 14 prism). Upon application of Yuan et al.’s [18] theory for calculation of plate strain at debonding in an unanchored joint based on the experimental parameters considered herein, the strain at 75 mm from the bonded loaded end of the plate is 43% of the peak strain. As a result, an anchor positioned in such a location has been shown by Smith and Kim [15] to significantly enhance the strength of the joint due to the engagement of the anchor in a high strain region. An additional dimension of importance to the strength of the joint is the length between the anchor and the free unloaded end of the plate. In this case the length is equal to 175 mm which is greater than the effective bond length. The embedment depth and dowel hole diameter for all anchored specimens was kept constant at 40 mm and 14 mm, respectively. In addition, the anchor fan was 50 mm in length and was radiated at 60° in a symmetrical manner about the longitudinal axis of the plate. The fan was oriented towards the direction of load as shown in Fig. 2 in order to provide a direct load path from the plate to the anchor dowel. In addition, such a fan angle ensured the fan edges radiated to the edges of the plate as shown in Fig. 2b.

2.2. FRP anchor construction and installation The construction of the dry and impregnated anchors used in this study is described in detail in this section. It was the intention to make the method of anchor construction relatively simple and easily carried out in both the laboratory and the field. In this study, both types of anchors were made from 90 mm long fibre sheets of varying width as defined in Tables 1 and 2. As shown in Fig. 3, the 90 mm of length was comprised of a 50 mm length fan with a 40 mm embedded portion and a small allowance for the 90° bend portion. For the dry anchors, the width of fibre sheet prescribed in Table 2 was rolled by hand (Fig. 4a) while being mindful to keep the fibres reasonably firmly compressed together. A 25 mm portion of the anchor dowel component was then defined by

(b) Impregnated anchor Fig. 3. Sectional schematics of installed dry and impregnated anchors.

wrapping with tie wire of 0.5 mm diameter (Fig. 4b). This tie wire was not expected to contribute to the strength of the anchor. The resulting anchor dowel diameter was on average about 12 mm and a typical completed anchor is shown in Fig. 4c. For the impregnated anchors, the 90 mm long fibre sheets of prescribed width were laid flat on a dedicated surface. A 25 mm end portion of the sheet was impregnated with epoxy and then the sheet rolled (Fig. 5a). The rolled dowel end was then inserted into a preformed hole of 12 mm diameter in a polystyrene mould which was pre-filled with epoxy (Fig. 5b). In order to ensure the fibres outside of the epoxy filled hole in the polystyrene remained dry, tie wire was temporarily applied. After the epoxy was allowed to cure for a minimum of 1 day, the anchors were removed from the polystyrene moulds and the tie wire removed. The resulting impregnated anchor is shown in Fig. 5c. The installation of the plate and anchor followed a prescribed procedure. Initially the anchor dowel hole was drilled into the concrete block (Fig. 6a). The region of the block with which the plate was bonded to was prepared using a pneumatic needle scaler (Fig. 6b) and then the whole surface of the block was cleaned by spraying with compressed air. The dry or impregnated anchor dowel was then inserted into the epoxy filled hole in the concrete (Fig. 6c) and such epoxy was allowed to cure for 1 day. Next, the fibre sheet used to form the plates was threaded over the anchor after parting the appropriate region of sheet fibres (Fig. 6d). The wet lay-up system enabled the sheet fibres to be parted in their dry form before application of the epoxy. No drilling of holes in the plate (as would be the case if pultruded plates were used) was therefore required. The fibre sheets

P

Anchor Fan

400

200

400

FRP Anchor FRP Plate Concrete Prism

200

Grip 100

(a) Test set-up

Unbonded Zone 50

FRP Anchor

175

75 40

FRP Plate

Concrete Prism

Anch nchor or dowel do el

(b) Test specimen Fig. 2. Test details.

4

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

(a) Rolling of dry fibres

(a) Rolling of impregnated fibres (foreground) and dry fibres

(b) Tying of anchor dowel fibres (b) Forming of anchor dowel component

(c) Completed anchor

(c) Completed anchor

Fig. 4. Dry anchor construction. Fig. 5. Impregnated anchor construction. used to form the plate were impregnated with epoxy and then the fan fibres of the anchor were splayed (without kinking) and then epoxied onto the surface of the plate (Fig. 6e). The completed joint was then allowed to cure for a period of 7 days in a controlled laboratory environment of approximately 22 degrees Celsius. 2.3. Instrumentation and loading Fig. 7 shows the positions of the 10 mm gauge length strain gauges which were bonded onto the anchored specimens. In addition, four LVDT’s were mounted on the base of the holding frame and strategically positioned to measure in-plane movement in the direction of the load (i.e. LVDTs 1–3) and out-of-plane movement transverse to the direction of load (i.e. LVDT 4). More specifically for the in-plane oriented LVDTs, LVDT 1 was attached to the plate adjacent to the loaded end of the concrete block. In addition, LVDTs 2 and 3 were attached to the ends of a bar connected to the concrete block directly behind the plate. Such a bar set-up was required due to the inability to position an LVDT directly behind the plate. The slip of the plate at the loaded bonded end was then calculated from the difference of LVDT 1 with the average of LVDTs 2 and 3. The determination of slip also entailed considering the elastic extension of the plate in the unbonded region of the block as well as the compressed concrete in the unbonded zone. The same arrangement of instrumentation was utilised for the unanchored control specimens in addition to the placement of an extra strain gauge (SG2a) midway between SG2 and SG3. All block specimens were mounted in a stiff holding frame while the holding frame was positioned in a universal testing machine (MTS 250 kN 810 system) (Fig. 8). In no tests was the out-of-plane movement (LVDT 4) greater than 0.4 mm (on average) thus signifying the effectiveness of the test frame in minimising block rotation. Load was applied by displacing the ram of the universal testing machine at a constant rate of 0.3 mm/min. 2.4. Material properties All 27 concrete prisms were poured in one batch and the concrete material tests were undertaken in the second week out of an 8 week test program. As the concrete was over 6 months old at the time of testing and the fact that all the concrete sam-

ples were kept in a controlled environment, only one set of concrete mechanical property tests was conducted as the properties were not expected to noticeably change over the duration of testing. The mechanical properties of the concrete were determined in accordance with the relevant BS 1881 standards [19–22] and the results (averaged from three test specimens) were cube compressive strength = 50.3 MPa, elastic modulus = 29.08 GPa, splitting strength = 3.44 MPa and modulus of rupture = 5.17 MPa. Tension tests were performed on 30 mm wide flat glass FRP and carbon FRP coupons and the results were calculated with respect to the nominal thickness of the fibre sheet. The coupons were prepared from two layers of fibre sheet and tested in accordance with ACI 440.3R-04 [23]. The results from three glass FRP and three carbon FRP coupons were averaged to produce an elongation at rupture of 19,772 le and 15,172 le (standard deviation, sd. = 1701 le and 126 le), tensile strength of 1595 MPa and 3090 MPa (sd. = 76 MPa and 38 MPa) and elastic modulus of 94.1 MPa and 201.4 GPa (sd. = 2 MPa and 29 GPa), respectively. The nominal thicknesses of the glass and carbon fibre sheets were 0.17 mm and 0.131 mm, respectively.

3. Experimental results This section contains a detailed description of the behaviour and failure modes of all test specimens. It also contains a discussion of the effect of the three main test variables as defined in Section 2.1. A summary of the key loads and slips are contained in Table 3 in addition to the failure modes. A summary of the strength enhancement of different combinations of anchored joints to that of the control specimen results are contained in Table 4. Overall, the additional of an anchor increased the strength of the joints on average by 53%.

5

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

Fig. 6. FRP plate and anchor installation.

400

400

200

40

SG5 SG4 SG3

50 50

40

LV 4 SG2

90

175

75

40

LV 1 SG1 LV 2,3

100 10 40

50

50

75

200

175

SG4 SG5

SG2 SG3

250

(a) Elevation view

LV 4

60

LV 3 SG1 LV 1 LV 2

(b) Plan view

Fig. 7. Instrumentation (LV = LVDT, SG = strain gauge).

3.1. Behaviour and failure modes All three unanchored control specimens failed by debonding of the FRP plate. Debonding initiated at the loaded bonded end of the plate and propagated to the far unloaded end of the plate. Such debonding, which occurred in the concrete at the FRP-to-concrete interface, has been reported on numerous occasions in the open literature [e.g. 16]. Fig. 9 shows a typical failed specimen in which the failure in the concrete at the FRP-to-concrete interface can be clearly seen. Such a failure mode is desirable as it indicates suitable application of the plate and it also enables the strength of the bonded interface to be related to the strength of the concrete. All 24 anchored specimens failed in one of two modes. In keeping with the terminology established by Smith and Kim [15], the

failure modes with which the anchored specimens failed in this study are defined as:  Mode 2A: Plate debonding followed by anchor rupture.  Mode 2C: Plate debonding followed by anchor pullout. In all anchored specimens, the plate initially completely debonded from the concrete prism. The specimens then experienced a reserve of strength which is primarily believed to be due to shear resistance provided by sliding of the roughened debonded plateto-concrete surfaces and the tensile resistance of the anchor. Such shear resistance was achieved due to the anchor clamping the plate to the concrete. It is hypothesised that the majority of the shear resistance was provided by the region of debonded plate between

6

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

was observed to curl outwards from the concrete face at higher loads and hence it was not considered as effective. In addition to experiencing a reserve of strength after complete plate debonding and before anchor failure, the joints also experienced large plateto-concrete slippage. Upon inspection of the failure modes listed in the last column of Table 3, the majority of the dry anchors failed by pullout (Mode 2C) and the majority of the impregnated anchors failed by rupture of the fibres in the bend region (Mode 2A). More detailed descriptions of the anchor failures are provided as follows. In the case of the dry anchors, the dry anchor dowel fibres pulled out although the dowel fibres which were bonded to the anchor hole face remained attached. Fig. 10ai shows pulled-out carbon fibres and Fig. 10aii shows, for the same test, the pulled-out anchor dowel attached to the plate. The anchor dowel fibres which remained attached to the anchor hole face are evident in Fig. 10ai. Fig. 10b shows glass anchor dowel fibres in the processes of being pulled out from the concrete prism during testing of a different test specimen. In the case of impregnated anchors, the fibres ruptured in the bend region at the junction of the anchor dowel and the anchor fan components. Rupture occurred due to several factors, namely (i) pullout force in the anchor, (iii) shear force in the anchor, (iii) reduction in strength due to the bent fibres, and (iv) concentration of stress in the anchor fibres immediately above the impregnated anchor dowel region. Fig. 11a shows a typical ruptured anchor in which the ruptured failure plane extending across the whole cross-section of the anchor at the face of the concrete prism can be clearly seen. Fig. 11b also shows another typical anchor rupture failure in the largely dry anchor fibres of the bend region. Particularly evident in Fig. 11b are two positions of rupture failure in the anchor, namely (i) rupture failure at the debonded face of the FRP plate as evident by a clean breaking of the fibres, and (ii) tearing

FRP plate

Concrete prism

Fig. 8. Test in progress.

the loaded end of the joint to the anchor. The length of debonded plate between the anchor position to the free unloaded plate end Table 3 Selected test results. Specimen identification.

a

Pdb (kN)

b

ddb (mm)

b

Pmax,1 (kN)

b

dmax,1 (mm)

b

Pmax,2 (kN)

b

dmax,2 (mm)

b

a

Individual Group Individual Group Individual Group Individual Group Individual Group Individual Group result average result average result average result average result average result average

Failure mode d

Control CN-1 Series CN-2 CN-3

16.75 17.59 14.42

16.25

0.18 0.26 0.28

0.24

18.00 18.13 17.73

17.95

0.59 0.67 0.45

0.57

–c

–

–

–

DB DB DB

Glass GD-200-1 Series GD-200-2 GD-200-3 GI-200-1 GI-200-2 GI-200-3

13.50 15.69 12.30 15.97 17.55 13.05

13.83

0.11 0.06 0.11 0.17 0.14 0.10

0.09

22.99 23.24 21.72 27.73 27.22 27.46

22.65

0.91 0.79 0.86 1.13 1.06 1.00

0.85

9.47 3.39 6.09 11.05 8.08 10.77

6.32

6.00  – 3.93 – 2.42

6.00

2A 2C 2C 2A 2A 2A

Carbon CD-134-1 Series CD-134-2 CD-134-3 CD-200-1 CD-200-2 CD-200-3 CD-259-1 CD-259-2 CD-259-3 CI-134-1 CI-134-2 CI-134-3 CI-200-1 CI-200-2 CI-200-3 CI-259-1 CI-259-2 CI-259-3

17.09 14.06 15.67 14.39 16.63 16.28 13.18 14.09 14.15 13.96 19.97 17.31 13.94 18.64 17.40 15.39 16.06 10.92

15.61

0.14 0.12 0.15 0.15 0.38 0.22 0.17 0.09 0.11 0.12 0.14 0.19 0.03 0.09 0.10 0.17 0.07 0.15

0.14

26.60 25.10 22.71 29.04 25.99 28.05 24.94 26.95 27.28 27.58 29.20 30.29 33.17 29.41 30.81 30.91 28.85 30.64

24.80

1.07 0.92 0.57 0.88 1.09 1.26 0.91 0.97 1.12 0.98 0.86 1.06 0.97 1.02 0.87 1.18 0.93 1.53

0.85

8.57 9.20 7.71 8.98 10.33 3.98 12.05 12.35 8.58 11.44 6.88 11.33 13.24 13.70 14.93 15.06 18.56 17.01

8.49

9.27 6.68 6.80 6.87 – – 9.81 7.08 4.91 4.90 3.96 4.10 2.94 3.69 5.02 4.35 3.72 3.54

7.58

15.52

15.77

13.81

17.08

16.66

14.12

0.14

0.25

0.12

0.15

0.07

0.13

27.47

27.69

26.39

29.02

31.13

30.13

1.06

1.08

1.00

0.97

0.95

1.21

9.97

7.76

10.99

9.88

13.96

16.88

3.18

6.87

7.27

4.32

3.88

3.87

2A 2C 2C 2C 2C 2C 2C 2C 2C 2A 2A 2A 2A 2A 2A 2A 2C 2A

See Table 1 for definition of specimen identification. Pdb and ddb = load and slip at initiation of plate debonding (Fig. 15). Pmax,1 and dmax,1 = load and slip at first peak (Fig. 15); Pmax,2 and dmax,2 = load and slip at second peak (Fig. 15). c Result not available. d DB = plate debonding; Mode 2A = plate debonding followed by anchor rupture; Mode 2C = plate debonding followed by anchor pullout. b

7

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12 Table 4 Strength enhancement of anchored specimens to unanchored control specimens. Grouping

a b c

No. of specimens

First load peak, Pmax,1 (kN) (% increase b)

Second load peak, Pmax,2 (kN) (% increase c)

Anchor type

Anchor fibre sheet width (mm)

Control All anchors

– All

–

3 24

17.95 {Pcon} 27.41 (53%)

– 10.53 (41%)

Glass anchor

All Dry Imp.

– 200 200

6 3 3

25.06 (40%) 22.65 (26%) 27.47 (53%)

8.14 (55%) 6.32 (65%) 9.97 (44%)

Carbon anchor

All Dry Dry Dry Imp. Imp. Imp.

– 134 200 259 134 200 259

18 3 3 3 3 3 3

28.20 24.80 27.69 26.39 29.02 31.13 30.13

(57%) (38%) (54%) (47%) (62%) (73%) (68%)

11.33 (37%) 8.49 (53%) 7.76 (57%) 10.99 (39%) 9.88 (45%) 13.96 (22%) 16.88 (6%)

Dry anchors Imp. anchors Mode 2A Mode 2C

– – – –

– – – –

12 12 13 11

25.38 29.44 28.77 25.81

(41%) (64%) (60%) (44%)

8.39 (53%) 12.67 (29%) 11.66 (35%) 9.20 (49%)

a

Imp. = Impregnated. (Pmax,1  Pcon)/Pcon  100%. (Pmax,2  Pcon)/Pcon  100%.

Interfacial failure in concrete

Debonded plate surface

Fig. 9. Typical failure mode of an unanchored control joint (specimen CN-1) glass fibres.

within the anchor dowel due to the incomplete impregnation of epoxy into the anchor fibres in the bend region.

3.2. Load–slip response: test results and generic relationships The load–slip responses for all three control joints are shown in Fig. 12. All three responses are quite similar and differences (especially in the initial linear region) are due to the inherent variability involved in testing concrete. The slips in this initial elastic region are very small and in the order of 0.2–0.3 mm. Debonding cracks, as evidenced by a change in slope of the load–slip responses, generally occurred near or at the maximum load. While it was very difficult to observe the exact extent of the debonding cracks in the experiment, the cracks initially appeared to propagate from the loaded end of the joint to a length approximately equal to the effective bond length (i.e. about 110 mm). The gradual propagation of debonding cracks in the remaining bonded portion of the plate caused the peak load plateau effect evident in Fig. 12. The peak load plateau was able to be captured due to the displacement controlled nature of the testing. The load–slip responses of all joints anchored with glass anchors are presented in Fig. 13 and all joints anchored with carbon anchors are presented in Fig. 14. For ease of comparison, all responses in Figs. 13 and 14 are plotted on vertical and horizontal axes of the same ranges. In some cases, the bracket mounted to

the plate upon which LVDT 1 measured against detached due to the large release of energy when the plate debonded. For these tests, the portion of the load–slip responses after the peak load could therefore not be plotted. Fig. 15 contains generic responses of the anchored joints failing in Modes 2A and 2C as extracted from the load–slip responses presented in Figs. 13 and 14. The generic response of the unanchored joints, as extracted from Fig 12, is also contained in Fig. 15. Finally, Fig. 15 also shows a simplified load–slip response for anchored joints which is described by lines A–B–C–D1–E–F1. Such a response forms the basis of the development of a load–slip model, however, such development is left for future research. A detailed description of the behaviour and failure sequence of the anchored specimens is now provided based upon the non-simplified generic load–slip responses presented in Fig. 15. A detailed summary of the key loads and slips are in turn contained in Table 3. 3.2.1. Region A–B The first linear portion extending from points A to B in Fig. 15 was of a similar slope to the unanchored joints. In most cases the anchored joint response was slightly stiffer on account of two main factors, namely (i) the presence of the anchor, and (ii) the plate being considerably stiffened due to the bonded fan. The slopes of both the unanchored and anchored joints are therefore drawn as one on Fig. 15 for simplicity. At point B, the change in slope signified debonding of the plate from the concrete substrate. Debonding initiated at the loaded bonded end of the plate and generally propagated to the position of the anchor for the anchored joints and slightly beyond quite rapidly. Such debonding initiated either slightly below or at approximately the same load as the unanchored control joints. The similarity of this initial debonding load, Pdb, to those of the control joints is evident in Table 3. The corresponding slips (ddb) for the anchored joints were, however, less than the control joints possibly on account of the stiffening effect of the anchor. These differences are also evident in Table 3 although they are not shown in Fig. 15. 3.2.2. Region B–C From points B to C n Fig. 15, the contribution of the anchor is most evident as the anchor was activated upon the debonding crack propagating past the anchor. In this region, joint resistance

8

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

(ii) FRP plate (i) Concrete block (a) Pulled-out carbon fibres of Specimen CD-200-3

Glass fibres

(b) Pulled-out glass fibres of Specimen GD-200-2 Fig. 10. Typical anchor pullout failures of dry fibre anchors.

was offered via three means, namely (i) remaining portion of plate bonded to the concrete, (ii) strength and stiffness of anchor, and (iii) roughened interfaces of debonded plate and concrete substrate sliding against each other. Point C corresponds to complete debonding of the FRP plate and at this point the anchor has still remained intact. The peak load Pmax,1 (at C) is consistently higher than that of the unanchored control joints as observed in Table 3. The corresponding slip, dmax,1, at Pmax,1 is also higher than the slip at maximum load of the control joints as also observed in Table 3. For the control joints, the load remained approximately constant once debonding had commenced due to the long bond length and this is represented by the line of zero slope extending from B–B0 in Fig. 15. 3.2.3. Region C–D Slippage of the plate is represented in Fig. 15 from points C to D. Physically, slippage was due to (i) rigid body movement of the plate due to complete debonding, and (ii) elastic shortening of the plate due to the sudden release of load upon complete debonding. Plate slippage was eventually halted by bearing of the anchor fibres onto the face of the anchor dowel hole nearer to the loaded end of the joint. The fact there was little to no epoxy impregnated into the fibres in the bend region meant the fibres in the bend region were able to compress.

tance of the anchor. The clamping effect provided by the anchor ensured the plate remained in contact with the concrete. As commented upon in Section 3.1, the region of debonded plate between the loaded end of the plate to the anchor was believed to contribute to the increase in joint strength as the plate region past the anchor was observed to curl away from the concrete prism. The peak load after plate debonding, Pmax,2, was a function of the amount of fibre in the anchor and the method of construction as evident from the results in Table 3. The corresponding slip, dmax,2, at Pmax,2 was higher for the dry anchors due to the ability of the dry anchor fibres to compress and accommodate plate slippage. 3.2.5. Region E–F The final portion of the load–slip response in Fig. 15 was dependent upon the failure mode of the anchor. The load reduced rapidly for anchor rupture failure (Mode 2A), however, a more gradual descent was observed for anchors which were pulled out (Mode 2C). In addition, in the case of Mode 2A failure, Figs. 13 and 14 generally show some post-rupture reserve of strength often less than 1 kN. In reality the effectiveness of the anchor can be considered to have been lost once Pmax,2 (and dmax,2) has been reached. The contribution of the anchorage to the joint can for all intents and purposes is therefore assumed to cease at position E. 3.3. Effect of anchor fibre type

3.2.4. Region D–E Region D–E in Fig. 15 corresponds to an increase in joint strength offered by the sliding of the roughened planes of the debonded plate and the concrete substrate as well as the tensile resis-

Inspection of the average peak loads (Pmax,1) in Table 4 reveals the dry and impregnated carbon anchor loads to be higher than their glass counterparts. Inspection of the average post-peak load

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

9

(i) Concrete block (ii) FRP plate (a) Ruptured carbon fibers of Specimen CI-134-3 FRP plate

Concrete prism

Ruptured anchor fibres

(b) Ruptured carbon fibres of Specimen CI-200-2 Fig. 11. Typical anchor rupture failures of impregnated anchors.

20

CN-1

CN-2

Load (kN)

15

CN-3 10 5 0 0.0

0.5

1.0

1.5

2.0

Slip (mm) Fig. 12. Load–slip responses of unanchored control joints.

(Pmax,2) also reveals the carbon fibre anchors to be higher than the glass fibre anchors. The higher rupture strength and higher elastic modulus of carbon fibres to glass fibres are the reasons why carbon fibres are superior. 3.4. Effect of anchor fibre content Of the three different widths of fibre used in the carbon anchors, 200 mm wide sheets gave the greatest Pmax,1 as observed in Tables 3 and 4. This occurred for both the dry and impregnated carbon anchors. Contrary to expectation, the anchors made from the greatest

amount of fibre were not the most effective. This was possibly on account of the difficulty of impregnating epoxy into the more densely packed fibres. The carbon anchors made from the greatest amount of fibre were found to be the most effective for both the dry and impregnated anchors in terms of Pmax,2 though. This trend was due to the anchors being more stressed in the post-peak region due to the large slips. In terms of the slip dmax,1, both dry and impregnated anchors containing fibres of all widths were approximately equal as evident in Table 4. Inspection of the slips dmax,2 in Table 3 reveal the dry carbon anchors to be consistently higher than the impregnated carbon anchors. This trend was also evident with the glass anchors. Such a phenomenon is due to the unimpregnated dry anchors being allowed to deform more easily than the stiffened impregnated anchors especially in the anchor bend region as described in Section 3.1.

3.5. Effect of anchor construction method Comparison of Pmax,1 in Table 4 for the dry and impregnated fibre anchors reveals that the impregnated fibre anchors performed better than the dry anchors. The same conclusion can also be reached upon inspection of Pmax,2. The presence of the epoxy in the impregnated anchors stiffened the anchor and thus enabled them to be more effectively activated. The slip dmax,1 is independent of the method of anchor construction. The slips dmax,2 for the impregnated anchors were less than the dry anchors due to

10

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

3.6. Load–strain response

35 GD-200-2

Load (kN)

30

LVDT reading terminated

25 20

Fig. 16a and b shows typical load–strain responses up to Pmax,1 for an unanchored and an anchored joint, respectively. The load– strain response along the length of the unanchored control joint was consistent with that observed on numerous occasions and as reported in the open literature [e.g. 16]. More specifically as observed in Fig. 16a, upon debonding the strain reading of the second strain gauge (SG2) became larger than that of SG1. In addition, at Pmax,1 a longer length of the plate had become activated (albeit for the portion from SG4 to SG5) and at the peak slip (i.e. 17 kN) the plate completely detached from the concrete block. Specimen CD-134-2 has been selected to be representative of the anchored joints in Fig. 16b although the load–strain response for all anchored joints was reasonably consistent. In this figure the plate strains were reasonably low prior to initiation of plate debonding (Pdb). Upon Pdb being reached, there was a marked increase in strain between the anchor fan and the loaded end of the joint. At 25 kN, there was a marked increase in strain between the anchor and the free unloaded end of the plate. After Pmax,1 was reached the remaining portion of the plate debonded rapidly. The level of strain in the plate region between the anchor and the free end of the plate in Fig. 16b was lower than the strains in the same region of the unanchored joint. This was on account of the contribution of the anchor.

GD-200-3

15 10

GD-200-1

5 0

0

3

6

9

12

15

Slip (mm)

(a) Dry glass FRP anchor (anchor fibre sheet width = 200 mm) 35

GI-200-2

Load (kN)

30 LVDT reading terminated

25 20

GI-200-1

15 10 5

GI-200-3

0

0

3

6

9

12

15

Slip (mm)

(b) Impregnated glass FRP anchor (anchor fibre sheet width = 200 mm) Fig. 13. Load–slip responses of joints anchored with glass FRP anchors.

3.7. Optimal anchor design The impregnated carbon anchors made from 200 mm wide fibre sheets are considered as the most optimally designed anchor in

35

35

30

30

25

Load (kN)

Load (kN)

the epoxy in the former inhibiting the anchor from deforming considerably.

CD-134-3

20

CD-134-1

15

CD-134-2

10

5

0

0

3

6

9

12

CI-134-3

10

5

0

CI-134-1

25 20 15

15

CI-134-2

0

3

6

Slip (mm)

(a) Dry carbon FRP anchor (anchor fibre sheet width = 134 mm) CD-200-3

25 CD-200-2

20 15

CD-200-1

10

20 10 5 0

6

9

12

CI-200-3

15

0

3

CI-200-1

25

5

0

CI-200-2

0

15

3

6

35

35

30

30

25

25

CD-259-2

CD-259-1

15 10 5

3

15

CI-259-3

20 15

CI-259-1

10 CI-259-2

5 CD-259-3

0

12

(e) Impregnated carbon FRP anchor (anchor fibre sheet width = 200 mm)

Load (kN)

Load (kN)

(b) Dry carbon FRP anchor (anchor fibre sheet width = 200 mm)

20

9

Slip (mm)

Slip (mm)

0

15

30 LVDT reading terminated

Load (kN)

Load (kN)

30

12

(d) Impregnated carbon FRP anchor (anchor fibre sheet width = 134 mm) 35

35

9

Slip (mm)

0

6

9

12

15

Slip (mm)

(c) Dry carbon FRP anchor (anchor fibre sheet width = 259 mm)

0

3

6

9

12

15

Slip (mm)

(f) Impregnated carbon FRP anchor (anchor fibre sheet width = 259 mm)

Fig. 14. Load–slip responses of joints anchored with carbon FRP anchors.

11

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

Plate Pre plate debonding debonding

Load

Post plate debonding

Anchor failure

Unanchored joint (A-B-B’)

C

Pmax,1

Anchored joint (A-B-C-D-E-F)

Pdb

B

B’

Anchored joint (Simplified) (A-B-C-D1-E-F1)

E

Pmax,2

D1

D F1 δdb

δmax,1

Slip

Fig. 15. Generic load–slip responses for unanchored and anchored joints.

P

SG1

SG2

SG2a

SG3

SG4

SG5

Unbonded region 10000

5kN 10kN

Plate Strain (µε)

8000

15kN N (Pdb) 16kN N (Pdb)

6000

(Pmax,1) 18kN N (Pmax,1)

17kN N (Failure)

4000

2000

0 0

50

100

150

200

250

Distance from bonded loaded end of plate (mm)

(a) Unanchored control specimen (CN-1) P

SG1

SG2

SG3

SG4

SG5

10000

Anchor fan region

8000

5kN 10kN

14kN Pdb (Pdb) 6000

15kN 20kN

4000

25kN

25.1kN Pmax,1 (P max,1 ) 2000

0 0

50

F

F

δmax,2

Plate Strain (µε)

A

Mode 2C

Mode 2A

100

150

200

Distance from bonded loaded end of plate (mm)

(b) Anchored specimen (CD-134-2) Fig. 16. Typical strain distribution.

250

12

H.W. Zhang et al. / Construction and Building Materials 32 (2012) 1–12

this study. This conclusion has been reached for several reasons, namely (i) greatest average peak load Pmax,1, (ii) slip at Pmax,1 greater than control specimens, (iii) substantial post-peak load and slip capacity, (iv) carbon fibres not too densely packed, and (v) the failure mode was constant. The impregnated carbon anchors constructed from 259 mm wide sheets were better performing in Pmax,2, however, at the expense of more fibre. The main drawback with the anchors made from 134 mm wide sheets was with Pmax,2 being considerably lower than that for the anchors made with more fibre. Regardless, anchors made from 134 mm wide fibre sheets did perform very well. According to Fig. 15, the mode of anchor failure really only appeared to be significant once Pmax,2 and dmax,2 had been reached. The mode of anchor failure is therefore not to be of concern in design provided the anchors are not expected to contribute after dmax,2 has been reached. 4. Conclusions A description of the test set-up and the results of three unanchored FRP-to-concrete control joints and 24 FRP-to-concrete joints anchored with FRP anchors have been presented in this paper. It was found that the addition of a single anchor can on average increase the strength of the joint by 53% above the unanchored control joint capacity. In addition, an anchor can considerably increase the slip capacity of the joint by more than ten times the slip capacity of control joints in some cases. Such enhanced joint behaviours are most important for the anchorage of externally bonded strengthening systems in which higher bond strengths can ensure more efficient use of the FRP material and large slip capacity can induce deformability into the strengthened system. Generic and simplified load–slip responses have also been described in this paper which can form the basis of the future development of analytical models for FRP-to-concrete joints anchored with FRP anchors. An optimal anchor design has been established based on the criteria of joint strength increase, post-peak reserve of strength and slip capacity, consistency of failure mode and amount of fibre material. Impregnated carbon anchors constructed from 200 mm wide fibre sheets were found to be optimal based on the range of variables considered in this experimental program. Depending on the design requirements though, anchors formed from other amounts of fibre sheet may also deemed to be optimal. As a result of this study, it is evident that one now has the power to manipulate the load and slip characteristics of an anchored joint by varying the amount of fibre in the anchor and the method of anchor construction. Future research will need to establish the effectiveness of this optimal anchor design in the anchorage of externally bonded FRP for the strengthening and repair of structural members such as beams, slabs and walls.

Acknowledgement Funding provided by the Hong Kong Research Grants Council in the form of General Research Fund Grant HKU715907E is gratefully acknowledged. References [1] Teng JG, Lam L, Chan W, Wang J. Retrofitting of deficient RC cantilever slabs using GFRP strips. J Compos Constr, ASCE 2000;4(2):75–84. [2] Lam L, Teng JG. Strength of RC cantilever slabs bonded with GFRP strips. J Compos Constr, ASCE 2001;5(4):221–7. [3] Oh H-S, Sim J. Interface debonding failure in beams strengthened with externally bonded GFRP. Compos Interfaces 2004;11(1):25–42. [4] Eshwar N, Ibell TJ, Nanni A. Effectiveness of CFRP strengthening on curved soffit RC beams. Adv Struct Eng 2005;8(1):55–68. [5] Orton SL, Jirsa JO, Bayrak O. Design considerations of carbon fibre anchors. J Compos Constr, ASCE 2008;12(6):608–16. [6] Smith ST, Hu S, Kim SJ, Seracino R. FRP-strengthened RC slabs anchored with FRP anchors. Eng Struct 2011;33(4):1075–87. [7] Micelli F, Rizzo A, Galati D. Anchorage of composite laminates in RC flexural beams. Struct Concr, fib 2010;11(3):117–26. [8] Özdemir G. Mechanical properties of CFRP anchorage, Master of Science Dissertation, Middle East Technical University, Turkey; 2005. [9] Kim SJ, Smith ST. Behaviour of handmade FRP anchors under tensile load in uncracked concrete. Adv Struct Eng 2009;12(6):845–65 [special issue on bond behaviour of externally bonded FRP reinforcement]. [10] Kim SJ, Smith ST. Pullout strength models for FRP anchors in uncracked concrete. J Compos Constr, ASCE 2010;14(4):406–14. [11] Ozbakkaloglu T, Saatcioglu M. Tensile behaviour of FRP anchors in concrete. J Compos Constr, ASCE 2009;13(2):82–92. [12] Bizindavyi L, Neale KW, Erki M-A. Behaviour of bonded FRP/concrete joints with glass fibre anchors. In: Proceedings, third international conference on advanced composite materials in bridges and structures, ACMBS III, Ottawa, Canada, 15–18 August, 2000. p. 719–26. [13] Aiello MA, De Lorenzis L, Galati N, La Tegola A. Bond between FRP laminates and curved concrete substrates with anchoring composite spikes. In: Proceedings, innovative materials and technologies for construction and restoration, IMTCR 2004, Lecce, Italy, 6–9 June, 2004. [14] Smith ST. FRP anchors: recent advances in research and understanding. In: Proceedings, second Asia-Pacific conference on FRP in structures, APFIS 2009, Seoul, Korea, 9–11 December, 2009. p. 35–44. [15] Smith ST, Kim SJ. Strength and behaviour of impregnated carbon FRP anchors in FRP-to-concrete joint assemblies. Adv Struct Eng, in press. [16] Yao J, Teng JG, Chen JF. Experimental study on FRP-to-concrete bonded joints. Composites Part B 2005:36; 99–110. [17] Chen JF, Teng JG. Anchorage strength models for FRP and steel plates bonded to concrete. J Struct Eng, ASCE 2001;127(7):784–91. [18] Yuan H, Teng JG, Seracino R, Wu ZS, Yao J. Full-range behaviour of FRP-toconcrete bonded joints. Eng Struct 2004;26(5):553–65. [19] BS 1881-116. Method for determination of compressive strength of concrete cubes. British Standards (UK); 1983. [20] BS 1881-117. Method for determination of tensile splitting strength. British Standards (UK); 1983. [21] BS 1881-118. Method for determination of flexural strength. British Standards (UK); 1983. [22] BS 1881-121. Method for determination of static modulus of elasticity in compression. British Standards (UK); 1983. [23] ACI 440.3R-04. Guide test methods for fiber-reinforced polymers (FRP) for reinforcing or strengthening concrete structures. Michigan (USA): American Concrete Institute (ACI); 2004.