The efficiency of mechanical anchors in CFRP strengthening of masonry: An experimental analysis

Composites: Part B 64 (2014) 1–15

Contents lists available at ScienceDirect

Composites: Part B journal homepage: www.elsevier.com/locate/compositesb

The efficiency of mechanical anchors in CFRP strengthening of masonry: An experimental analysis Mario Fagone a,⇑, Giovanna Ranocchiai a, Carmelo Caggegi b, Silvia Briccoli Bati a, Massimo Cuomo c a

Department of Civil and Environmental Engineering, University of Florence, Piazza Brunelleschi, 6, 50121 Firenze, Italy Université Paris-Est, Laboratoire Modélisation et Simulation Multi Echelle, MSME UMR 8208 CNRS, 5 bd Descartes, 77454 Marne la Vallée, France c Department of Civil and Environmental Engineering, University of Catania, v.le A. Doria, 6, 95125 Catania, Italy b

a r t i c l e

i n f o

Article history: Received 18 December 2013 Received in revised form 6 March 2014 Accepted 27 March 2014 Available online 18 April 2014 Keywords: A. Carbon fibre B. Debonding B. Mechanical properties D. Mechanical testing Mechanical anchor

a b s t r a c t Fiber reinforced composite materials are widely used for structural rehabilitation and retrofitting of existing buildings. Failure of not anchored CFRP reinforcements, applied to both concrete and masonry, mainly occurs in the substrate or at the bonding surface, for load values lower than the tensile strength of the composite. Mechanical anchors can effectively increase the maximum load of this type of reinforcements. Particular attention should be paid to the design and sizing of mechanical anchors so that these can produce adequate increments of both strength and ductility of the reinforcement. At the moment there are no specific rules or reliable predictive formulas that adequately support designers in the design choices and sizing of anchors. Of course, these can be defined only after collection of an extensive experimental database that highlights the peculiar characteristics of these reinforcements also with reference to the substrate material. In this context, the present work describes the results of an experimental campaign carried out on brick specimens reinforced with CFRP strips, anchored to the substrate by CFRP spike anchors. Almost all research in the literature about this topic refers to the use of this reinforcement technique for concrete structural elements. Studies concerning masonry are limited: this work contributes to bridge part of this gap. Plane CFRP reinforcement strips generally exhibit brittle failure mode. The experimental campaign reported in this paper shows that properly designed mechanical anchors increase both the failure load and the ductility of the reinforcement. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Several consolidation techniques based on bonding composite materials, and particularly CFRP (Carbon Fibers Composite Polymers) strips, are more and more used in the past decades for the rehabilitation and the reinforcement of both concrete and masonry structural elements. The excellent mechanical performance of these reinforcements, combined with lightness and simplicity of application, justify their widespread use in the structural field. Experimental analysis carried out on concrete [1,2] and masonry [3–7] structural elements reinforced by CFRP strips subjected to in-plane loads, have shown that failure generally occurs in the substrate, a few millimetres below the bonding surface. The load is transferred to the substrate mainly through shear stresses that produce cohesive debonding at a failure load lower than the composite tensile strength. Shear stresses are mostly concentrated in a limited portion of the reinforcement whose length is ⇑ Corresponding author. Tel.: +39 055 2756831; fax: +39 055 212083. E-mail address: mario.fagone@unifi.it (M. Fagone). http://dx.doi.org/10.1016/j.compositesb.2014.03.018 1359-8368/Ó 2014 Elsevier Ltd. All rights reserved.

called ‘‘effective bond length’’ [8]. The failure load of a reinforcement strip increases with the bond length until it reaches such a limit length; longer bond lengths do not significantly increase the peak load, but only the reinforcement ‘‘ductility’’: while the load remains constant (at failure), the stress transfer zone moves from the loaded to the unloaded end of the reinforced surface. Various numerical models can describe the mechanical behaviour of plane CFRP reinforcements. Interface cohesive zone models as well as continuous damage models [12] have been proposed in the literature to schematize the mechanical behaviour of such reinforcements applied to concrete [9,10] or masonry [11] structural elements. As previously said, failure of plane CFRP reinforcements, applied to both concrete and masonry, mainly occurs in the substrate or at the bonding surface. Therefore, the maximum load, lower than the composite tensile strength, strongly depends on the mechanical properties of the substrate. Several methods have been proposed in the literature to increase the maximum load of this type of reinforcements, especially with reference to concrete structural elements [13,14]. ‘‘U-wraps’’ anchors have been used, for example, in [15] to increase the bending strength of concrete

2

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

beams strengthened with CFRP reinforcements applied on the lower surface. U-wraps anchors are FRP strips applied perpendicular to the main reinforcement to prevent premature failure. An alternative mechanical anchor is the insertion of CFRP spike anchors to fix the reinforcement to the substrate: this ‘‘fiberanchor’’, made rolling a carbon fiber fabric sheet, is inserted into a hole drilled in the reinforced structural element, and glued upward, in a ‘‘fan’’ shape, to the reinforcement strip [16–18]. The upper edge of the hole should be smoothed [19] in order to avoid stress concentrations that can lead to premature failure of the anchor. In [20] it is reported an experimental program carried out on concrete beams strengthened with CFRP strips anchored with both systems previously described. The beams, notched at the centreline, were subjected to bending tests. The experimental campaign showed that, under the considered conditions, ‘‘fiberanchor’’ is more effective than ‘‘U-WRAP.’’ Moreover, fiber anchors having fan angle less than 90° exhibited a better mechanical performance. Niemitz et al. [16,17] studied the efficiency of carbon fiber anchors by means of single shear tests on reinforced concrete blocks strengthened with CFRP strip. They found that the overall effectiveness of FRP anchors is related to geometric characteristics of the anchor, such as the ratio between the anchor diameter and the fan diameter, the anchor spacing and the anchor depth. A larger fan diameter increases the force applied to the anchor so that, if the amount of fibers in the CFRP spike anchor is insufficient, a shear failure of anchor is possible. The most efficient position of anchors is across the width of the composite strip; spacing anchors longitudinally is effective also to increase ductility of the system. Zhang et al. [21] studied FRP anchors by means of 27 single shear tests on reinforced concrete blocks strengthened by FRP strips. They tested two typologies: dry and impregnated anchors. The second one, more efficient, is realized pre-impregnating the lower part of the anchor (dowel) in the rolling fiber fabric phase. Different design arrangements of impregnated fiber anchors have been studied in bending tests on FRP-strengthened reinforced concrete slabs [22]. Most of research in the literature concerning the efficiency of mechanical anchors refers to concrete structural elements. However, in view of the widespread use of CFRP reinforcements of masonry structural elements, and since the failure load of plain (i.e. not anchored) reinforcements is lower than the tensile strength of the composite, experimental analyses devoted to masonry elements appear appropriate and necessary. Mechanical anchors applied to FRP-to-masonry joints can effectively increase the performance of CFRP reinforcements [23], and can be efficiently used in structural rehabilitation or retrofitting interventions needed, for example, after seismic events, or due to other factors such as environmental degradation, changes in usage resulting in heavier loading conditions. The increase of reinforcement strength produced by mechanical anchors could decrease the portion of building on which such interventions have to be implemented. On the other hand, they require onerous preliminary preparation (surface preparation, drilling of masonry, inclusion of CFRP anchors, etc.) and longer installation time with respect to plane reinforcements. Moreover, such reinforcements are clearly irreversible so that their use in buildings belonging to the historical and cultural heritage must be carefully evaluated. Nevertheless, the pool of existing buildings in which anchored reinforcements can be effectively used is still very large. Their use in technical practice is, however, discouraged by the lack of specific rules that adequately supports designers in the design choices and sizing. In this regard, American Concrete Institute (ACI 440.2R-08) [19] and Italian CNR-DT 200 R1/2013 [8] guidelines stipulate that the use of mechanical anchors must be substantiated by representative experimental testing, not exhaustively specified, that sometimes

appear, however, burdensome, discouraging the use of these devices in current practice. The development of predictive formulas, to be inserted in appropriate regulations, is therefore necessary to bridge this gap. Of course, these can be defined only after collection of an extensive experimental database that highlights the peculiar characteristics of these reinforcements also with reference to the substrate material. In this context, the present work describes the results of an experimental campaign performed on anchored reinforcement strips applied to masonry. Since mortar joints can affect the mechanical performance of the reinforcement [7], in order to evaluate the effects due to anchors only, uniform substrate, made of a single brick, was considered in the experimental program. A total of five anchor systems, differing in geometry and number of anchors, as well as plane reinforcements were tested. The research activity was carried out within the framework of a scientific cooperation agreement among the University of Florence, the University of Catania and the University Paris-Est Marne-LaVallée. Aims of the experimental program described in this paper are both the evaluation of enhancement in strength and dissipative capability of the CFRP reinforcement produced by the mechanical anchor, and the analysis of the influence in the experimental results of different test apparatus implementing the same test scheme. Therefore, in the experimental program, the test set-up and the specimens’ characteristics were the same of those described in [23]. In particular, for each of the six reinforcement systems considered, twelve specimens were made at the ‘‘Laboratorio Ufficiale Prove Materiali e Strutture’’ of the University of Florence, by the same operator, using the same materials and the same procedure [24]. Half of the specimens were tested in the ‘‘Laboratoire de Mode´lisation et Simulation Multi Echelle’’ of University Paris-Est Marne-La-Vallée: the results of the Paris’ experimental session are described in [23,25]. In this paper, the results of the Florence’s experimental session, carried out on the remaining half of the specimens and following the same test procedure considered in Paris’ experimental session, are described. Although the experimental campaigns carried out in Paris and in Florence differ only in the test apparatus, despite similar overall behaviour they show differences in the strength and dissipative capability of the reinforcement, which will be analysed in this work. The experimental program and the specimens’ characteristics are described in the next paragraph; the experimental campaign carried out to determine the mechanical properties of the bricks is described in paragraph 3, while in paragraph 4 the procedure followed to realize the reinforced specimens is described; in paragraph 5 the test apparatus is described and the results of the experimental campaign are reported in paragraph 6. Final remarks and some comparisons with [25] conclude the paper.

2. Experimental program Various experimental setup have been proposed in the literature [1] for analysing the in-plane mechanical behaviour of structural elements reinforced by CFRP strips. Among these, the ‘‘Near-End Supported Single Shear Test’’ (NES-SST) has been selected to evaluate the improvement in the mechanical behaviour of such reinforcements, applied to masonry structural elements, due to CFRP anchors. In order to evaluate the effects of the mechanical anchors, the experimental program presented in this work concerns reinforcements applied on a uniform substrate constituted by a single brick. The experimental setup is illustrated in Fig. 1: the CFRP reinforcement strip is applied on the larger surface of a brick and protrudes in the length direction, so that it can be grasped by the test machine for applying a tensile force. The smaller surface of the brick, near to the protruding strip, is supported:

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

3

3. Mechanical properties of bricks

Fig. 1. Near-End Supported Single Shear Test (NES-SST) – boundary conditions (measures in mm).

the reaction forces, counter-acting the load, produce compression in the brick. The bond length of FRP strip (180 mm) is longer than the effective bond length (103 mm), estimated according to the equation proposed by CNR DT 200/2004 [8]. In this way the crack advancement could be observed during the tests. For the same reason, the anchor is placed at a distance, from the loaded side, close to the effective bond length (see Fig. 2). The width of bonded FRP strip is equal to 100 mm to permit the application of one or more anchors. For each one of the typologies illustrated in Fig. 2, six specimens were prepared in the ‘‘Laboratorio Ufficiale Prove Materiali e Strutture’’ of the University of Florence. Bricks produced by ‘‘SanMarco’’ (Terreal Italia S.r.l.), carbon fibers ‘‘MBRACE-fibre high res.’’, epoxy resin ‘‘MBRACE-adesivo’’ and primer ‘‘MBRACE-primer’’ (BASF chemical company) were employed to prepare the specimens. MBRACE-fiber high res. is a unidirectional carbon fiber fabric having a nominal thickness of 0.165 mm, a characteristic tensile strength of 2500 MPa and an elastic modulus of 230,000 MPa (property declared by BASF). In order to determine tensile strength, compressive strength and Young modulus of the bricks, specific tests, described in the next paragraph, were performed in the ‘‘Laboratorio Ufficiale Prove Materiali e Strutture’’ of the University of Florence.

The bricks used for the specimens devoted to near-end supported single shear tests are SanMarco ‘‘Faccia a vista Classico di colore Rosso’’, that are solid pressed bricks with dimensions 250  120  65 mm. Pressed bricks were preferred to drawn bricks because their material structure (and, consequently, the anisotropy of mechanical properties and the quality of fracture path) resembles the one of traditional soft pressed bricks, that are used in most existing buildings [26]. Twelve bricks were randomly chosen from the brick supply employed for the construction of the specimens. From each of six bricks (labelled from 1 to 6), four cubic specimens 50  50  50 mm and one prismatic specimen 50  50  150 mm were obtained according to the scheme reported in Fig. 3. Three of the four cubic specimens were tested in compression according to the length (y), width (x) and thickness (z) of the brick. These were named with the letter C (Compression) followed by a number from 1 to 6 indicating the brick label, a number from 1 to 4 referring to the position of the specimen in the brick according to Fig. 3, and the letter X, Y or Z that indicates the load direction. The load– displacement diagrams obtained by the (displacement driven) compression tests are reported in Fig. 4. It is apparent that, after an initial non-linear branch, the diagrams show a quasi-linear stroke whose slope is almost constant for all the specimens loaded in the same direction. All the diagrams also show an extended post-peak branch. The tests ended when a load value equal to a half of the peak load was reached. The compressive strength values of the cubic specimens are not very scattered: the Coefficient of Variation (C.V.) is less than 10% for all the load directions (see Table 1). The specimens exhibited the higher compressive strength in the x direction, corresponding to the brick width (Fig. 3), and the weaker compressive strength in the z (thickness) direction. This is due to the production process that leads to a greater compaction of the dough in the directions orthogonal to the thickness. The 50  50  150 mm prismatic specimen (Fig. 3), named with the letter M followed by a number from 1 to 6 indicating the brick,

Fig. 2. Typologies of test specimens considered in the experimental program (measures in mm).

4

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Fig. 3. Cubic and prismatic specimens, obtained from one brick, for compression and Young modulus tests.

Fig. 4. Load–displacement diagrams for cubic specimens tested in compression: in the specimens label, ‘‘C’’ refers to ‘‘Compression’’, the first index to the brick, the second index to the specimen position in the brick (see Fig. 3), the last letter to the load direction.

Table 1 Results of compression tests. Load direction x

Load direction y

Load direction z

Spec.

Comp. strength (MPa)

Mean (MPa)

C.V. (%)

Spec.

Comp. strength (MPa)

Mean (MPa)

C.V. (%)

Spec.

Comp. strength (MPa)

Mean (MPa)

C.V. (%)

C11X C23X C32X C41X C53X C64X

25.39 20.05 23.97 20.93 21.06 21.64

22.17

9.28

C12Y C24Y C33Y C44Y C51Y C63Y

21.04 19.36 19.17 19.05 21.34 19.42

19.90

5.11

C13Z C22Z C31Z C43Z C52Z C61Z

18.39 17.70 18.53 15.86 19.71 19.17

18.22

7.40

were used for the determination of the Young modulus. In each test, three compression cycles were performed, according to the y direction, up to a half of the expected maximum load. For each specimen, three values of the elastic modulus were determined from the slope of the unloading branches, within the linear elastic range of the material, and, successively averaged (see Table 2). Also in this case, the values obtained are not very scattered.

From each of the remaining six bricks (labelled from 7 to 12), a prismatic specimen 40  40  200 mm was obtained, with the main dimension parallel to the brick length (y). Bending tests were performed on these specimens (Fig. 5); they were named with the F letter followed by a number from 07 to 12 indicating the brick label. From one of the two halves of the specimen resulting from the bending test a specimen for a tensile test was obtained: the

Table 2 Results of the tests for the determination of Young modulus. Specimen

Esc1 (MPa)

Esc2 (MPa)

Esc3 (MPa)

E (MPa)

M1 M2 M3 M4 M5 M6

9429 9285 7976 8639 8072 8918

9368 9265 8016 8586 8032 9016

9392 9232 8030 8590 7973 8996

9397 9261 8007 8605 8026 8977

Mean (MPa)

C.V. (%)

8712

6.92

5

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Fig. 5. Three point bending test scheme.

fractured surface was regularized and the new specimen (40  40  90 mm) was named with the letter T followed by a number from 07 to 12 corresponding to the brick, and the letter Y for the load direction. The tensile test was carried out after connecting the smaller surfaces of the specimen to steel plates adequately connected to the loading machine. The results of the bending and tensile tests with their average values are reported in Table 3. The values of the tensile strength, and in particular those obtained by the three point bending tests, are much more scattered than the values of compressive strength reported in Table 2. Furthermore, the bending tensile strength of the bricks is about 35% higher than the direct tensile strength. 4. Specimens for Near-end supported shear tests The geometric characteristics of the specimens used for the Near-End Supported-Single Shear Tests (NES-SST) are schematized in Fig. 2. Since, as far as the authors’ knowledge, there are no specific studies concerning the behaviour of the CFRP spike anchors technique to masonry structural elements, we assumed that the behaviour of anchors in concrete substrate is similar to that of anchors in brick. Therefore, we defined the geometry of CFRP anchors according to the best characteristics defined in the studies referring to concrete [17,18,20,21]. Consequently, the anchors’ typology used in the present experimental session has been the ‘‘impregnated anchor’’ defined by Zhang et al. [21]. The anchor depth was chosen equal to 50 mm to avoid pull out failure [16]. The fan anchor angle has been limited to less than 90° in all series of anchored specimens except one, to avoid stress concentration. A fan opening angle of 75° was chosen because it permits to cover, with two (r = 40 mm) or three (r = 25 mm) anchors placed transversally, side by side, the whole width of CFRP strip. In order to analyse the performance of the anchor as a consequence of the number of anchors, of the fan radius (i.e. the length of fiber used to fasten the anchor to the reinforcement) and of the fan shape (i.e. the geometric distribution of the fan fiber used to fasten the

anchor to the reinforcement), these characteristics were varied in the geometry of the samples. These were labelled with the letter F (samples tested in Florence) or T (samples tested in Paris) followed by a number (from 0 to 3) that refers to the number of anchors, another number that refers to the fan radius (measured in mm) and a letter that refers to the fan shape, distinguishing a fan angle of 360° indicated with ‘‘O’’ and on a fan angle of 75° indicated with ‘‘V’’. The twelve specimens of series F0-T0 represent the reference specimens and were tested in order to determine the behaviour and the strength of the reinforcement system without mechanical anchors (plain reinforcement). The series F1_40_O (T1_40_O) has one anchor with fan angle of 360° and fan radius of 40 mm. The series F1_40_V (T1_40_V) is equal to the previous, except that the fan angle is 75°: the comparison of the F1_40_O (T1_40_O) and F1_40_V (T1_40_V) series permits to evaluate the influence of fan shape. Series F1_25_V (T1_25_V), has a fan radius of 25 mm and a fan angle of 75°, and it has been tested to analyse the influence of the fan radius in comparison with F1_40_V. Series F2_40_V (T2_40_V) and F3_25_V (T3_25_V) both present a fan angle of 75°, but have 2 or 3 anchors placed transversally to the FRP strip. They have been tested for investigating the influence of the number of anchors. The anchor was made with a CFRP spike anchor introduced in a hole drilled in the larger surface of bricks, having a diameter of 14 mm and a depth of 50 mm. Since the anchor depth is close to the bricks thickness, in order to avoid the weakening of the brick in the vicinity of the hole, two bricks were cemented for the construction of specimens with anchors (F1, F2, F3, T1, T2, T3 series) using a high performance epoxy based adhesive for structural bonding (Sikadur 31 – Sika Group). Successively the hole was drilled and the round sharp edge was smoothed on a 13 mm radius to prevent stress concentration as recommended by ACI 440 (2002). The cavity and the surface of the brick were cleaned before inserting the CFRP spike anchor. The two typologies of anchors, having fan lengths of 40 and 25 mm, have been obtained from a 200 mm width sheet according to the following procedure [21] (Fig. 6): the necessary lengths (90 mm and 75 mm for the two different typologies of anchor length) were cut out from the carbon fiber fabric to obtain a strip that was partially impregnated with epoxy resin; then the fabric was rolled in the width direction, so as to form a dowel that was inserted in a pre-formed hole (12 mm of diameter) in a polystyrene mould in order to harden while confined. The anchor was removed from the mould after one day minimum (Fig. 6(c)). After the anchor was hardened, the hole of the brick was filled with epoxy resin and the spike anchor was inserted as indicated in Fig. 7(a). After one day minimum, primer was spread over a portion of 100  180 mm of the surface of the brick; then, the day after, epoxy resin was spread and the carbon fiber fabric was laid on the epoxy resin allowing the passage of the anchor through the fibers of unidirectional fabric (Fig. 7(b)). At this point, epoxy resin was spread on the composite strip and the carbon fiber of the spike anchor, that was free from resin, was splayed over the strip; further epoxy resin was applied over the fan anchor. Note that

Table 3 Tensile strength obtained from three point bending and direct tensile tests. Bending

Tensile test

Spec.

Tensile strength (MPa)

F07 F08 F09 F10 F11 F12

2.16 2.51 4.23 5.03 2.57 3.67

Mean (MPa)

C.V. (%)

Spec.

Tensile strength (MPa)

3.36

33.77

T07Y T08Y T09Y T10Y T11Y T12Y

2.62 2.89 2.59 1.90 2.05 2.89

Mean (MPa)

C.V. (%)

2.49

16.90

6

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

displacement transducers (‘‘T1’’ to ‘‘T4’’ in Fig. 8) measured the vertical displacement of the steel plate as a control of possible translations and rotations of the contrast plate. During the tests, the load has been applied imposing displacement to the upper grip of the testing machine at a constant rate of 0.2 mm/min. The load path was characterized by a succession of displacement steps (Du = 0.2 mm) assigned in a time of 1 min and followed by a 1 min of delay step. In the next sections of the paper, the results of the test campaign are discussed, concerning the analysis of failure modes and the elaboration of the load–displacement diagrams obtained during the tests.

6. Experimental results 6.1. Results overview

Fig. 6. Three phases of the CFRP anchor realization procedure: (a) the CFRP anchor after the realization of the dowel; (b) the anchor with the fan ready to be splayed; and (c) the anchor in the configuration suitable to be inserted inside the brick hole. The fan is grouped to make easy the apposition of the CFRP strip.

the width of the fiber fabric laid on the brick was equal to the width of the reinforcement (100 mm, see Fig. 2), while its length was 660 mm: 180 mm were impregnated with the epoxy resin and bonded to the brick to realize the reinforcement strip, the remainder was used for applying the load to the reinforcement as described in the next paragraph. 5. Test machines and test procedure The test systems was designed to realize the boundary conditions illustrated in Fig. 1. In particular, the test apparatus is composed by two parts: a rigid steel frame (in black in Fig. 8) and an upper grip. The steel frame consists of four columns constraining a rigid plate used to support the upper face of the bricks. Its dimensions are such that specimens of different size can be tested. Following the scheme reported in Fig. 1, a steel cylinder supports the lower end of the brick avoiding its rotation during the test. The specimens were loaded imposing the vertical displacement of the upper grip, consisting of a fork and a steel cylinder around which the carbon fiber fabric was wrapped and glued with the same epoxy resin used for the reinforcement. The instrumentation is composed by a 100 kN load cell and two displacement transducers (‘‘TS’’ and ‘‘TD’’ in Fig. 8) placed at the bottom of the load cell and based on the contrast plate of the brick, so that they could record only the relative displacement between the fork and the upper face of the brick. Moreover, four

The failure phases, the peak loads achieved and the amount of energy supplied and dissipated have been studied by analysing the load–displacement diagrams obtained from the tests. Furthermore the visual analysis of the specimens, during and after the tests, allows the definition and the description of the occurred failure modes. In spite of the differences in the mechanical behaviour better described below, all the load–displacement diagrams referring to the anchored specimens have the shape schematically represented in Fig. 9. Note that, the displacement indicated in abscissa in Fig. 9 (and in the other following diagrams of the same type) is the relative displacement between the rigid plate, constraining the upper specimen surface, and the load grip (measured by the transducers ‘‘TD’’ and ‘‘TS’’ indicated in Fig. 8) minus the elastic deformation of the fiber fabric sheet out of the reinforcement bonding (l0 ), estimated using the elastic modulus and the nominal thickness declared by the producer. In the representation of Fig. 9, the linear stroke O-P1 corresponds to the elastic behaviour of the specimen. A first peak load (FP1) is recorded in P1 (peak 1) when the first crack appears in the loaded extremity of the CFRP to substrate bonded joint: at this point, the load decreases slightly. The stroke PP1-P2 (post-peak 1 – peak 2) corresponds to the crack advancement phase, while in P2 the total debonding of the reinforcement strip occurs. The load value recorded at this point, named FP2, corresponds to the maximum peak load obtained during the test. If no anchors are fastened to the composite strip, the test ends when P2 is achieved by the total detachment of the CFRP reinforcement and the simultaneous fall to zero of the load. The experimental study presented in this paper shows that the presence of CFRP anchors gives a residual not negligible resistance to the reinforced system after the peak load in P2 (stroke PP2-P3). In P3 the ‘‘CFRP strip – CFRP anchor – brick’’ system fails and the load drops again. The increase in dissipative capability of the strengthening systems due to the anchor is analysed through the area C underlying

Fig. 7. Some realization phases of the series fastened by anchors: (a) insertion of the anchor into the specimen; (b) application of the CFRP trough the strip; and (c) bonding of the fan anchor to the CFRP strip.

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

7

Fig. 8. Test machines for NES-SST carried out in Florence.

to calculate the area DC* indicated in Fig. 9. However, from the diagrams reported in [6], referred to tests analogous to those considered in this paper but carried out on specimens made with historical bricks, the unstable branch seems to be almost vertical. For this reason, the value of DC* was estimated as the area underlying the parabola interpolating the PP2-P3 branch, evaluated between the displacement corresponding to P2 and PP2. As said, we underline that not all the energy shares described above can be regarded tout-court as energy dissipated during the crack advancement process, but represent a quite good approximation that can be used at least to compare the dissipative capability of the anchors considered in this work. During the tests, the following fracture modes were observed (see Fig. 10):

Fig. 9. Typical load–displacement diagram obtained from the NES-SST performed on anchored specimens. In the figure the characteristic points and the energy shares used in the analyses are indicated.

the load–displacement curve. Specifically, the total area Cend is equal to the sum of the following shares (Fig. 9): – DC1 is the amount of energy stored in the system before the first crack. It is equal to the area underlying the (quasi-)linear elastic stroke O-P1; – DC2 is the amount of energy supplied during the crack advancement phase, until the total debonding of CFRP reinforcement. It is equal to the area underlying the stroke P1-P2; – DC* + DC3 is the amount of energy supplied during the fiber anchor loading phase, after the debonding of CFRP strip until the achievement of peak P3. The area underlying the load–displacement diagram up to P2 is indicated as Cp2 = DC1 + DC2. Note that the tests described in this paper are driven by the relative displacement between the contrast plate and the upper load grip (see Fig. 1). In this way, we are not able to follow an eventual not stable equilibrium path between P2 and PP2 [6]: only the initial and final points of such snap-back branch can be recorded in the tests. Moreover, the deformability of the load system (steel apparatus) contributes to increase the length, in particular the displacement gap, of the unstable branch, as the elastic energy absorbed by the load frame is given back during the collapse events. Given the particular test procedure used, we are not able

– Cohesive Fracture (CF): indicates a fracture occurring in the brick, below the layer impregnated by the primer applied for the realization of the reinforcement [8]. This is the fracture mode that most frequently occurred in the experimental campaign; – Mixed Fracture (MF): indicates a fracture that occurs partly in the brick (as CF) and partly at the interface between the brick and the reinforcement [8]. It is recognized because, after the failure, the lower face of the reinforcement is only partially attached to residues of brick; – Debonding Splay Anchor (DSA): indicates the sliding of the reinforcement strip from the fan anchor. In this case the CFRP spike anchor remains fixed to the brick; – Pull Out Anchor (POA): in this case the reinforcement strip remains glued to the anchor, while the latter slides out from the brick; – Prismatic Failure (PF): this fracture mode occurs with the detachment of a significant portion (wedge-shape) of brick, to which the reinforcement system remains attached [1]; – Fiber Failure (FF): occurs in the carbon fiber fabric, outside the reinforcement strip, which remains, in this case, attached to the brick. In order to directly compare the overall response of all specimens tested in the experimental campaign, the corresponding load–displacement diagrams are reported in Fig. 11. Note that the diagrams of specimens F1_40_V_3 and F2_40_V_1 are not reported in the figure because the tests were affected by positioning mistake. Moreover, in Table 3 the load and the fracture mode corresponding to peaks P2 and P3 (Fig. 9) are reported. The samples

8

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Fig. 10. Fracture modes occurred in the shear tests: (a) ‘Cohesive Fracture’, CF; (b) ‘Mixed Fracture’, MF; (c) ‘Debonding Splay Anchor’, DSA; (d) ‘Pull Out Anchor, POA; (e) ‘Prismatic Failure’, PF; and (f) ‘Fiber Failure’, FF.

Fig. 11. Load–displacement diagrams: (a) sample F0; (b) sample F1_25_V; (c) sample F1_40_O; (d) sample F1_40_V; (e) sample F2_40_V; and (f) sample F3_25_V.

exhibited a substantially different mechanical performance, depending on the strengthening systems. In particular, the tests showed that the anchors increase both the maximum peak load and the dissipative capability of the reinforcement. Moreover, also the failure mode of the specimens with anchored strengthening strip was very different from the failure mode of plain reinforced specimens. In the following, the experimental results are described in detail for all the reinforcement systems taken into account in the test campaign. 6.2. Tests results for specimens without anchor The load–displacement diagrams referring to the samples F0 (see Fig. 11(a)) show an initial quite linear load path, having aver-

age slope K1 approximately equal to 13.0 kN/mm, until reaching a first peak load FP1 approximately equal to 10.9 kN (see Tables 5 and 6). At this point, corresponding to P1 in the scheme reported in Fig. 9, fractures develop in the brick behind the reinforcement surface, mostly near the support plate (loaded side of specimen), according to diagonal lines. After the first fracture occurrence, load decreases of a small amount and the load path becomes more irregular. However, this stroke, corresponding to PP1-P2 in Fig. 9, can be approximated as linear and has a slope equal to 40% of K1. In this phase cracks advance from the loaded to the unloaded extremity of the specimen. In fact, being the bond length longer than the effective bond length, during the tests the stress transfer zone moves from the loaded to the unloaded side of the reinforced surface.

9

M. Fagone et al. / Composites: Part B 64 (2014) 1–15 Table 4 Maximum peak load at P2, peak load at P3 and failure modes occurred during the tests. Specimen

Failure mode

Peak load (N)

Specimen

P2

P3

P2

P3

F0_1 F0_2 F0_3 F0_4 F0_5 F0_6

CF CF CF MF MF CF

– – – – – –

14,852 13,133 12,506 13,337 13,376 14,172

– – – – – –

F1_40_V_1 F1_40_V_2

MF MF

DSA PON

18,642 19,494

8581 11,555

F1_40_V_4 F1_40_V_5 F1_40_V_6

CF MF MF

DSA DSA DSA

22,634 19,793 19,604

9414 11,790 9922

Failure mode

Peak load (N)

Specimen

Failure mode

Peak load (N)

P2

P3

P2

P3

P2

P3

P2

P3

F1_25_V_1 F1_25_V_2 F1_25_V_3 F1_25_V_4 F1_25_V_5 F1_25_V_6

DSA DSA MF CF CF CF

DSA – DSA DSA DSA DSA

18,892 19,418 20,085 19,862 19,425 15,974

13,929 – 13,720 10,124 82,56 63,40

F1_40_O_1 F1_40_O_2 F1_40_O_3 F1_40_O_4 F1_40_O_5 F1_40_O_6

CF MF MF CF MF MF

DSA DSA DSA DSA DSA DSA

17,131 20,780 16,649 16,560 22,493 15,693

7947 5408 6960 9052 10,571 5498

F2_40_V_2 F2_40_V_3 F2_40_V_4 F2_40_V_5 F2_40_V_6

FF MF MF MF CF-DSA

– PF PF PF –

25,095 25,598 23,027 24,290 27,482

– 19,089 17,095 16,946 –

F3_25_V_1 F3_25_V_2 F3_25_V_3 F3_25_V_4 F3_25_V_5 F3_25_V_6

CF–FF CF MF CF CF–FF CF–FF

– FF PF PF – –

28,598 27,734 29,499 30,827 26,867 31,385

– 10,690 21,020 19,001 – –

Table 5 Statistical values for P1 (FP1), P2 (FP2) and P3 (FP3) peak loads. Values corresponding to Fiber Failure have not been taken in account for FP2. Only specimens that exhibited a post peak stroke have been considered for FP3. FP1

F0 F1_25_V F1_40_O F1_40_V F2_40_V F3_25_V

FP2

FP3

n. spec.

Mean (N)

C.V. (%)

Incr. w.r.t. F0 (%)

n. spec.

Mean (N)

C.V. (%)

Incr. w.r.t. F0 (%)

n. spec.

Mean (N)

C.V. (%)

6 6 6 5 6 6

10,860 11,095 12,168 9966 12,537 10,957

6.66 8.87 9.09 6.87 12.61 14.83

– 2.2 12.0 8.2 15.4 0.9

6 6 6 5 5 3

13,562 18,943 18,218 20,033 24,134 29,353

6.10 7.98 15.06 7.58 11.26 5.29

– 39.7 34.3 47.7 78.0 116.4

– 5 6 5 4 3

– 10,474 7573 10,253 16,906 16,904

– 31.88 26.86 13.51 11.13 32.39

Table 6 Statistical values for the slope of O-P1 (K1) and PP1-P2 (K2) strokes. K1

F0 F1_25_V F1_40_O F1_40_V F2_40_V F3_25_V

K2

n. spec.

Mean (N/mm)

C.V. (%)

Incr. w.r.t. F0 (%)

n. spec.

Mean (N/mm)

C.V. (%)

Incr. w.r.t. F0 (%)

6 6 6 5 6 6

12,980 12,814 13,298 12,604 13,544 11,968

18.23 15.49 17.61 28.07 31.00 21.42

– 1.3 2.4 2.9 4.3 7.8

6 6 6 5 6 6

5167 8961 8761 9023 9917 10,810

37.32 20.40 24.71 13.02 16.18 13.40

– 73.4 69.6 74.6 91.9 109.2

When the maximum peak load FP2 is reached (point P2 in Fig. 9), a sudden failure is recorded. The average value of the maximum peak load is 13.6 kN (see Table 5), approximately 25% higher than FP1. As it is apparent in Table 5, the peak load values FP1 and FP2 recorded in the tests are not very scattered being the coefficient of variation about 6% (see also the graphical representation of Fig. 12). The slopes of the two branches, especially of PP1-P2, are more irregular. CNR-DT 200/R1 [8] proposes the following equation for evaluating the axial stress design value of plain CFRP reinforcements applied to masonry:

ffdd

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2Ef CFd ¼ cf ;d tf 1

Ef ¼ ð1Þ

Using the previous relation, design strength of plane reinforcements can be determined as follows:

F fdd

bf qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2tf Ef CFd ¼ ffdd bf t f ¼

cf ;d

In order to evaluate the strength of not anchored reinforcements according to CNR-DT 200/R1, regardless to limit state design, we set cf,d = 1. For the CFRP reinforcement strips considered in this paper bf = 100 mm and tf = 1 mm. Moreover, as previously reported, the values of nominal thickness and elastic modulus of the unidirectional carbon fiber fabric used in the reinforcement strips are respectively tc = 0.165 mm and Ec = 230,000 MPa. The ‘‘MBRACE-primer’’ has a Young modulus Em = 3300 MPa. Therefore, the elastic modulus of the composite Ef, in the longitudinal direction, can be evaluated as follows:

ð2Þ

being cf,d is the partial factor for limit state design; Ef is elastic modulus of composite; bf,tf is resp. width and thickness of the reinforcement strip and CFd is the specific fracture energy of brick.

Ec t c þ Em ðtf  t c Þ ¼ 40706 MPa tf

ð3Þ

The specific fracture energy can be estimated according to the following relation [8]:

CFd ¼

kb kG pffiffiffiffiffiffiffiffiffiffiffiffiffiffi fbm fbtm FC

ð4Þ

qffiffiffiffiffiffiffiffiffiffiffi 3b =b being FC is the confidence factor; kb is 1þbf =b (geometric correction f factor); b is width of the reinforced structural element; for the specimens considered in this paper b = 120 mm; kG is correction coefficient depending on the material: kG = 0.031 mm for brick masonry

10

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Fig. 12. Box charts of statistical values related to (a) K1, (b) FP1, (c) K2, and (d) FP2. Filled diamonds indicate the values obtained from the tests; unfilled squares indicate the average values; box half height corresponds to the standard deviation. Values refering to Fiber Failure have not been taken in account.

and fbm, fbtm is the respectively compressive and tensile strength of brick. The mechanical properties of bricks have been determined trough an exhaustive experimental program, so that we can set FC = 1. Using for fbm the mean value of compressive strengths reported in Table 1 (fbtm = 20.10 MPa) and for fbtm the average tensile strength obtained from direct tensile tests (fbtm = 2.49 MPa, see Table 3), Eq. (2) provides the following value for plain reinforcement strength:

F fdd ¼ 13:9 kN This value compare well with the average maximum peak load (13.6 kN) obtained in the experimental campaign for F0 series. This confirms the reliability of the procedure proposed by the CNR-DT 200/R1 for the evaluation of the strength of this type of reinforcements. As indicated in Table 4, the fracture mode of specimens F0 is mainly cohesive (CF in Fig. 10): it is characterized by a fracture path developing in the brick, some millimetres behind the strengthening system. Only in two cases a mixed fracture mode occurred (MF in Fig. 10), that is the strengthening system detached from the brick partially in correspondence of the bonding surface. All the specimens in the samples F0 exhibited, substantially, a homogeneous mechanical behaviour. The average value of energy CP2 = DC1 + DC2 in series F0 is equal to 13.51 J as reported in Table 7. This value will be used as a reference for assessing the increase in dissipation capability supplied from anchorage to the reinforcement.

6.3. Tests results for specimens with one anchor Almost all the tests carried out on specimens fastened by only one anchor (series F1), showed a reserve of resistance after the load fall in P2 and a prominent post peak branch in the load–displacement path (see Fig. 11(b)–(d)). This indicates a higher dissipative capacity of the anchored reinforcement than the plain one (F0 series). In order to facilitate the global comparison among the load–displacement diagrams relating to different series, these are represented through an ‘‘average broken line’’ having the initial endpoint coinciding with the origin of the load–displacement axes and the other vertex in a position that averages the points P1, PP1, P2 PP2 and P3 of all the diagrams of the same series (see Fig. 13). In particular, the abscissa (ordinate) of the first vertex is defined as the average of the abscissa (ordinate) of peak P1 of the diagrams of the same series; similarly for the others characteristic points. Note that the last vertex and the end point, representative of PP2 and P3, have been determined considering only the specimens that showed a post-peak branch. This representation allows an immediate qualitative comparison among the characteristic points and the average slopes of the linear strokes of the diagrams referring to different series. Moreover, in the diagram of Fig. 13 the envelope of the load–displacement diagrams of the same series is also reported. Specimens provided by one anchor showed an almost linear load–displacement path up to point P1 (see Fig. 11). As it is apparent in Table 6, the average slope K1 of F1 series is similar to the corresponding value of series F0. The first peak load FP1 of F1 series

11

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Table 7 Statistical values for CP2 and Cend. Only to the population of specimens that exhibited a post peak branch has been considered in order to properly calculate the values reported in the last column. Sample

F0 F1_25_V F1_40_O F1_40_V F2_40_V F3_25_V

n. specimens

CP2 Average (J)

Coeff. of variation (%)

Increment w.r.t. F0 (%)

Average (J)

Coeff. of variation (%)

Increment w.r.t. F0 (%)

(Cend  CP2)/Cend (%)

6 5 6 5 4 3

13.51 18.38 18.56 23.37 30.01 44.03

16.12 6.14 13.15 20.53 26.14 6.84

– 36.04 37.38 72.98 122.13 225.91

13.51 30.59 35.86 47.61 60.77 60.15

16.12 16.08 24.97 15.65 11.85 7.50

– 126.42 165.43 252.40 349.81 345.23

– 39.92 48.24 50.91 50.62 26.80

Cend

Fig. 13. Average line and envelope of the load–displacement diagrams of series F0 and F1.

differs of about ±10% from the analogous value in the F0 series (see Table 5). This shows that the anchor does not substantially modify the overall response of the reinforcement up to the first peak P1. After P1, the load decreases of a small amount and a second, almost linear, branch is exhibited; its slope is almost the same for all the F1 series. Also the FP2 peak load is quite constant for all the F1 series. In particular, the second linear branch slope and the maximum peak load of F1 series are respectively about 63% and 41% higher than the corresponding values of F0 series: the anchor transfers the load to the inner core of the brick, increasing both the stiffness and the strength of the reinforcement. The global behaviour of all the F1 series before P2 is quite homogeneous. This

shows that inserting a CRP anchor, the stiffness and strength of the reinforcement improve regardless of the fan shape. As well as for F0, the failure mode corresponding to point P2 is mainly cohesive and mixed (see Table 4). Beyond P2, part of the load can still be carried by the anchor and by the strengthening strip still adhered to the substrate. As we previously said, the test procedure allows to record just the points P2 and PP2 and not the entire, eventually unstable, equilibrium path. The post-peak load (PP2) is quite scattered (see Fig. 13) as it depends on the particular crack pattern occurred in the specimen after the peak P2. The trend of the post peak branch (PP2-P3) is quite scattered too for the F1_25_V and F1_40_O series, while it is more regular for F1_40_V (see Fig. 13). The ultimate failure of the specimens is almost always due to the sliding (debonding) of the reinforcement strip from the fan anchor (DSA). Pull out anchor (POA) occurred only for one specimen of the F1_40_V series (see Table 4): this failure mode is induced by stresses in the anchor, parallel to the spike anchor axis. The average values of FP3 achieved at the end of the further loading phase in the series F1_25_V, F1_40_O and F1_40_V are respectively equal to the 77%, 56% and 76% of the average maximum peak load for F0 series. Therefore, the post peak load FP3 is strongly influenced by the anchor shape: the ‘‘V’’ shaped anchors (regardless of the length of the reinforcement) had better performance respect to the ‘‘O’’ shaped. The energy CP2 (Fig. 9) and Cend are reported in Table 7 and schematized in Fig. 14. It is apparent that the series F1_25_V and F1_40_O show similar values for CP2, about 40% higher than the value determined for not anchored strengthening. The series F1_40_V shows larger values of CP2: about 73% higher than the not anchored strengthening. As Cend is concerned, the highest and most regular values refer again to the series F1_40_V, which shows increments of approximately 250% with respect to plain

Fig. 14. Statistical distribution of (a) CP2 and (b) Cend. Filled diamonds indicate the values obtained from the tests; unfilled squares indicate the average values; box half height is equal to the standard deviation. Only to the population of specimens that exhibited a post peak branch has been considered.

12

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

strengthening. As mentioned above, beyond the peak load the strength of the reinforcement is mostly related to the anchor. Therefore, the increasing of dissipation capability attributed to the reinforcement by the anchor was evaluated trough the ratio between (DCend  DCP2) and Cend (last column of Table 7). The highest increase in dissipative capability conferred by the anchor refers to the series F1_40_V, which also showed a low coefficient of variation. As far as described, among the reinforcements with one anchor the series F1_40_V showed the greatest increase, with respect to plain reinforcements, in both the maximum peak load and the dissipative capability. Moreover, the post-peak branches and dissipative properties of this series are the most regular. Therefore, among the types of reinforcement considered in this section, they showed the more satisfactory mechanical performance. On the other hand, since the fan anchor is mainly oriented in a specific direction, these reinforcements are suitable to loads having constant direction and sign. Otherwise, anchors having circular fan could be more appropriate. However, as far as the authors’ knowledge, for this kind of reinforcements, tests with cyclic loads have not been performed. These actions may damage the fan anchor because of instability phenomena induced in the compressed portion of the fan, which may lead to a premature failure.

6.4. Tests results for specimens with two or three anchors The load–displacement diagrams obtained from the tests carried out on specimens with two or three anchors (see Fig. 11(e) and (f)) exhibit an initial almost linear branch until the peak P1 (Fig. 9), that is not always well recognizable, especially for F3_25_V series. In fact, even if fractures occurred at P1 in the brick below the reinforcement, the anchors were able to distribute the load to the inner core of the brick, without a significant decrease of the specimen stiffness (27% for F2_40_V and 10% for F3_25_V series). It is apparent from Table 6 that, for the reinforcements considered in this paragraph; the slope K1 of the first linear branch is similar to the corresponding value obtained for F0 series while K2 is 92% and 109% higher than the corresponding value referred to F0 respectively for F2_40_V and F3_25_V series. The first peak load FP1 occurred for F2_40_V and F3_25_V series is similar to the value obtained for F0 series (Table 5), while the maximum peak load is 78% and 116% higher than FP2 of F0 series respectively for F2_40_V and F3_25_V. This indicates that the anchor is effective for loads higher than FP1. Note that the mechanical behaviour of the reinforcement considered in this paragraph is quite regular: the coefficient of variation for the maximum peak load ranges, in fact, from 5% to 11%. The comparisons between F1_40_V and F2_40_V series and between F1_25_V and F3_25_V series highlight the influence of the number of anchors irrespective of the fan radius and shape; it is apparent that the number of anchors increases mainly the values of FP2 and, slightly, the values of K2. The stiffness K2 of the specimens with more than one anchor is, in fact, about 20% higher than the corresponding value of the specimens with only one anchor: this occurs because both reinforcement systems (F2_40_V and F3_25_V) transfer the load to a similar portion of the substrate. Otherwise, the maximum peak load FP2 increases of about 30% using two anchors rather than one, and of about 55% using three anchors rather than one: the latter reinforcement system generates a more uniform stress distribution in the brick that leads to a higher maximum peak load. Also in this case, in order to facilitate the global comparison among the load–displacement diagrams relating to different series, these are compared through the average line and the envelope of the diagrams obtained by the tests of the same series as described in the previous paragraph (see Fig. 15).

Fig. 15. Average line and envelope of the load–displacement diagrams of series F0, F2_40_V and F3_25_V.

Neither F2_40_V nor F3_25_V series showed a significant post peak branch. Actually, five up to eleven specimens exhibited ‘‘fiber failure’’ (the reinforcement was stronger than the fabric) or ‘‘debonding splay anchor’’ at P2 without a post peak stroke. Furthermore, the specimens that showed a reserve of post peak strength, collapsed because of ‘‘prismatic failure’’ (five up to six) or ‘‘fiber failure’’ (one up to six). In the F2_40_V and F3_25_V series each CFRP spike anchor is less stressed with respect to reinforcements having only one anchor, therefore failure occurs in the brick as Prismatic Failure (PF). By considering only the population of tests characterized by a loading phase after the achievement of P2, the average energy DCP3 = DC* + DC3 recorded during the tests is equal to about 30 J in the F2 series and 16 J in the F3 series. The increase in dissipation capability conferred by the anchors to the reinforcement (last column in Table 7) is about 50% for F2 and 27% for F3 series. Obviously, these values decrease substantially if we refer to the entire specimens in the series, considering also those that do not showed a post peak branch. Although the fracture mode at P2 involves effects that we referred to as dissipative (see Table 4), depending on fractures occurring in the brick below the reinforcement, this is often associated to ‘‘brittle’’ phenomena. For this motivation, the ductility and the dissipative capability of this type of reinforcement can be neglected. 7. Comparison with CFRP reinforcements of concrete As previously mentioned, the experimental campaign described in this paper draws on Zhang et al. [21], concerning mechanical anchored CFRP reinforcements of concrete prisms. In order to highlight how the substrate material can affect the mechanical performance of CFRP reinforcement, the key similarities and differences between some of the results of the two experimental campaigns will be described in this paragraph. Both experimental campaigns show that in such reinforcements failure occurs in the substrate, below the reinforcement: it is expected, therefore, that the characteristics of the material to be strengthened substantially affect the mechanical performance of the strengthening system. The experimental campaign described in Zhang et al. [21] was carried out on concrete prisms having elastic modulus and compressive strength respectively equal to 29.08 GPa e 50.3 MPa. The ratio between the values of these parameters reported by Zhang et al. for concrete and the values determined on bricks in this campaign (see Tables 1 and 2) is equal to Sstr = 2.5 for the compressive strength (Substrate strength ratio) and Smod. = 3.3 for the elastic modulus (Substrate modulus ratio). Analogously to brick (see

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Tables 2 and 3), the tensile strength of concrete is about one tenth of the compressive strength. The reinforcement strips used by Zhang et al. were made from three layers of carbon fiber sheet in a wet lay-up procedure, having elastic modulus and nominal thickness respectively equal to 201.4 GPa and 0.131 mm. No information is provided on either the resin used for the realization of the reinforcements or on the composite thickness. Therefore, assuming as a first approximation that stiffness of the reinforcement is almost equal to that of the carbon fabric, the axial stiffness per unit width of the reinforcement trips used by Zhang et al. can be estimated as ka(Z) = (Young modulus)  (nominal thickness)  (n. layers) = 79.2 kN/mm. Similarly, the axial stiffness per unit width of the reinforcement strips used in this experimental campaign (Ef = 230 GPa, tf = 0.165 mm) can be evaluated as ka(F) = 38.0 kN/ mm. The reinforcement strip stiffness ratio is equal to Rstiff. = 2.1. Since Smod and Rstiff are quite similar, a direct comparison between the two experimental campaigns appears reasonable. Specimens of ‘‘control series’’ by Zhang et al. were reinforced with plane reinforcement strips having width bf = 50 mm. The reinforcement length is greater than the effective length as well as for plane reinforcement of F0 series, having width bf = 100 mm (see Fig. 2). Similarly to F0 series, load–displacement diagrams of ‘‘control series’’ show an initial quasi-linear stroke until initiation of plate debonding (average FP1 = 16.25 kN). Then, load quickly decreases of a small amount and a second almost linear stroke, having slope slightly smaller than the first branch, is recorded until reaching the maximum load (average FP2 = 17.95 kN), little higher than FP1. Failure occurs with a very long plateau, corresponding to crack advancement due to debonding below the reinforcement strip, unlike what recorded for series F0 which exhibited instantaneous failure at P2. The average ratio FP2/FP1 is equal to 1.10 for Zhang et al. ‘‘control series’’ and 1.25 for F0 series. In specimens of F0 series debonding occurs with a crack pattern that propagates also within the brick, while in concrete specimens debonding occurs a few tenths of millimetres below the bonding surface [21]. Failure in F0 series specimens involves ‘‘damage’’ in a greater substrate volume, so that greater decrease in stiffness and a higher difference between FP1 and FP2 than ‘‘control series’’ occurs. Maximum load per unit width (FP2/bf) is equal to 0.36 kN/mm for ‘‘control series’’ and 0.14 kN/mm for F0 series. The ratio between these values, equal to 2.6, is very close to the Substrate strength ratio Sstr = 2.5. This indicates that, despite the differences in stiffness of the reinforcements and of the substrates, plane reinforcement strength is mainly related to the substrate strength. Note that this result cannot be immediately generalized as the strength of plane reinforcement of masonry elements does not vary linearly with bf due to scale effects experimentally observed especially for small values of bf [4]. In Zhang et al. [21], also reinforcements with both dry and impregnated anchors have been analysed. At least for the considered typologies, impregnated anchors shown better mechanical performance, which is independent on the fiber content. However note that that the amount of fiber in the anchor should be at least more than the amount of material contained in the main FRP strip [20]. Mechanical anchor used in F1_40_V and F2_40_V series are very similar to those used by Zhang et al. in CI-200 series, where an impregnated anchor has been applied to a reinforcement strip having width bf = 50 mm. Since the reinforcement strip is a fibrous material and therefore it is expected to have poor capability to ‘‘spread’’ the stress state, it seems natural to compare CI-200 series with F2_40_V series: in both cases there is an impregnated anchor every 50 mm of width of the reinforcement, albeit in CI-200 series it is placed at a distance from the loaded side of the reinforcement smaller than the effective length, while in F2_40_V series this distance is close to that effective length. Load–displacement diagrams of CI-200 and F2_40_V series show analogous characteristics

13

(schematized in Fig. 9), especially before the maximum load, but the failure modes and the dissipative capability are very different. The two series have similar values of FP2/FP1 ratio (1.86 for CI-200 series and 1.93 for F2_40_V series) as well as similar increments of maximum load FP2 with respect to plane reinforcement (+73% for CI-200 series, +78% for F2_40_V series). The reinforcement strength per unit length FP2/bf is 0.62 kN/mm for CI-200 series and 0.24 kN/ mm for F2_40_V series. The ratio between such specific strengths, equal to 2.58, is, also in this case, very close to the Substrate strength ratio Sstr = 2.5. This confirms that the reinforcement strength strongly depends on the substrate strength. Against similar behaviour up to P2, the two series differ a lot in the post peak branch: load–displacement diagram of all the specimens of CI-200 series shows very wide post-peak branch, while only three samples (out of five) of F2_40_V series show (very short) post peak branch. The difference between FP3 and maximum load of plane reinforcement (FP3  FP2(‘‘plain’’))/FP2(‘‘plain’’)) is higher for F2_40_V series (+25%) with respect to CI-200 series (22%). Furthermore, specimens of the two series show very different failure mode. In CI-200 plate debonding is followed by anchor rupture occurring in the bend region at the junction of the anchor dowel and the anchor fan components. In specimens of F2_40_V series that showed post peak branch, ‘‘prismatic failure’’ occurred because of substrate failure. Ultimately, as revealed by the previous analysis, at least for the examined reinforcement typologies, the substrate material affects the maximum load per unit length, producing variations proportional to Substrate strength ratio Sstr. However, anchored CFRP reinforcements applied to bricks showed much more brittle behaviour with respect to concrete. This may be because in F2_40_V series anchors are very close to one another and their distance is of the same order of magnitude as the dimensions of the cross section of the reinforced specimens: this produces almost uniform stress state in the inner core of the brick that leads to instantaneous collapse of the specimen. Further analysis would be required to check what might be the optimal spacing for anchors applied to the masonry and the possible influence of scale effects. Further analyses are required to check the anchor optimal spacing and possible scale effects that can affect the mechanical behaviour of such reinforcements. However, it must be observed that, unlike concrete structures, the brick is the unit of material that has to be taken into account whatever is the dimension of the brickwork structure. Anchor systems with higher spacing in brickwork may involve adjoining bricks.

8. Discussion The experimental analysis reported in this work shows that, although the reinforced specimens exhibit similar mechanical behaviour and similar characteristics in the load–displacement diagram (see Fig. 9), the stiffness of the reinforcement, the maximum peak load, the post-failure behaviour and the failure modes are strongly affected by the type of reinforcement system. In particular, as described in the previous paragraphs, a typical load–displacement diagram shows an initial linear elastic branch, which slope K1 is almost independent on the presence, and thus on the type, of mechanical anchor (see Fig. 16(a)), as well as the first cracking load FP1 (Fig. 16(b)). This demonstrates that the anchor, placed at a distance from the loaded side of the reinforcement close to the effective bond length [8], does not affect this phase of the mechanical response. After the first fracture occurrence, load decreases of a small amount and the load path, although more irregular, can still be approximated as linear. The slope K2, as well as the maximum peak load FP2, strongly depends on the number of mechanical anchor, regardless of the fan shape

14

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

Fig. 16. Average values of (a) stiffness and (b) peak loads for each series.

Fig. 17. Average values of energy shares obtained for each series.

and the fan length. In particular, K2 increases, with respect to the plain reinforcements of the series F0, of about 70–75%, 92% and 109% respectively for the specimens reinforced with one, two or three anchors; the increments in FP2 are respectively about 34– 48%, 78% and 116%. Therefore, for the reinforcement considered in this paper, the phase that precedes the maximum peak load FP2 depends on the number of anchors, but is almost independent on the fan length and fan shape. These latter characterize the postpeak P2 behaviour: specimens of the series F1_25_V, having a small fan length, show short post-peak branch and very scattered value of FP3; specimens having higher fan length, especially the F1_40_O series, exhibit a very long post-failure branch. Anyway, among the specimens provided with one anchor, the series

F1_40_V showed the best mechanical performance; the post failure branches are quite regular and FP3 is not negligible, almost coincident with the first crack load FP1. Moreover, they exhibited better dissipative capability (see Fig. 17). Inserting a greater number of anchors, the stiffness K2 and the maximum peak load FP2 increase, but the global behaviour becomes more brittle: two and three up to six specimens respectively for F2_40_V and F3_25_V series collapsed at P2, while the other specimens of the same series exhibited very short and scattered post failure branches and not dissipative collapse mode. Anyway, it should be noted that, considering only the population of the specimens of series F2_40_V and F3_25_V that exhibited a post failure branch, the average value of FP3 is higher than the maximum peak load FP2 of the series F0. Therefore, among the reinforcements examined in this work, the series F1_40_V showed the best mechanical performance: it exhibited, with respect to the series F0, significant increase in the maximum peak load and substantial improvement in the dissipative characteristics. The ultimate load FP3 is also not negligible. The results and conclusions of the experimental campaign carried out at the ‘‘Laboratorio Ufficiale Prove Materiali e Strutture’’ of the University of Florence, described in this paper, are broadly in agreement with those obtained at the ‘‘Laboratoire de Modélisation et Simulation Multi Echelle’’ of the University Paris-Est Marne-La-Vallée, reported in [23]. In spite of similar overall behaviour and similar failure modes, there are some differences, not always negligible, pertaining to load values higher than the first crack peak FP1. In particular, the maximum peak load FP2 and the energy shares obtained from tests performed in Paris are lower than that obtained from tests carried out in Florence. In particular the differences in the maximum load range from 10% to 30% (see Table 8). Since all the specimens tested in both experimental sessions were made at the ‘‘Laboratorio Ufficiale Prove Materiali e Strutture’’ of the University of Florence, by the same operator, using the same materials and the same procedure, such discordance does

Table 8 Comparison among the peak loads FP1 and FP2 obtained from the test sessions performed in Florence and Paris. Sample

F(T)0 F(T)1_25_V F(T)1_40_O F(T)1_40_V F(T)2_40_V F(T)3_25_V

FP1 (N)

FP2 (N)

‘‘F’’ series (N)

‘‘T’’ series (N)

(‘‘F’’-‘‘T’’)/‘‘T’’ (%)

‘‘F’’ series

‘‘T’’ series

(‘‘F’’-‘‘T’’)/‘‘T’’ (%)

10,860 11,095 12,168 9966 12,537 10,957

10,956 10,365 10,728 9961 11,380 9668

0.88 7.04 13.42 0.05 10.17 13.33

13,562 18,943 18,218 20,033 24,134 29,353

12,298 15,490 15,590 16,163 20,444 23,005

10.28 22.29 16.86 23.94 18.05 27.59

M. Fagone et al. / Composites: Part B 64 (2014) 1–15

not seem attributable to differences in the specimen preparation, including surface preparation, mixing and application, which rather are responsible for C.V. reported in Table 5. In the authors’ opinion, the differences in mean peak loads FP1 and FP2 shown in Table 8 can be attributed to the deformability of the test setup. This last is, in fact, the main difference between the two experimental sessions. Even if they implement the same test scheme (Fig. 1), the device used in Florence is slightly stiffer than the one used in Paris: the rigid plate used in the Florence test session to support the upper face of the bricks (Fig. 8) is constrained by four steel columns and, during the tests, four displacement transducers (T1 to T4) check that it remains horizontal. The coplanarity of the load and the reinforcement strip is checked at the beginning of the test and a steel cylinder avoids the rotation of the specimen with respect to the rigid plate. The supporting device of the test machine used in Paris is almost a spherical hinge. During the test, it has been observed that this constraint undergoes deformations that, especially for high values of the load, allow small rotations of the specimen inducing out of plane actions to the reinforcements. In this case failure is favoured by a small peeling that combines to debonding. These considerations are consistent with results reported in Table 8 as the maximum peak loads obtained from tests performed in Paris are systematically lower than that obtained from tests carried out in Florence. Moreover, higher differences occur for high load values, for which the deformability of the test machine is more influential. Acknowledgements The ReLUIS program of the Italian Civil Protection Agency supported this research. This support is gratefully acknowledged. We thank arch. Giuliana Di Iacovo for making the specimens and assisting the tests. The bricks were provided free of charge by ‘‘SanMarco’’, Castigion Fiorentino factory. We acknowledge its support. References [1] Yao J, Teng JG, Chen JF. Experimental study on FRP-to-concrete bonded joints.. Compos B Eng 2005;36:99–113. [2] Chen JF, Teng JG. Anchorage strength models for FRP and steel plates bonded to concrete. J Struct Eng 2001;127:784–91. [3] Briccoli Bati S, Fagone M. Caratterizzazione della modalità di rottura di elementi in laterizio fibrorinforzati: test sperimentali e simulazioni numeriche. Conference ‘‘Materiali e Metodi innovativi nell’Ingegneria Strutturale’’. Catania 4–6 July 2007. Atti del convegno; 2007. p. 363–72. [4] Briccoli Bati S, Fagone M. An analysis of CFRP-brick bonded joints: XVIII Conference of the Italian Group of Computational Mechanics (GIMC); Siracusa 22–24 September 2010.

15

[5] Capozzucca R. Experimental FRP/SRP-historic masonry delamination. Compos Struct 2010;92:891–903. [6] Carrara P, Ferretti D, Freddi F. Debonding behavior of ancient masonry elements strengthened with CFRP sheets. Composites Part B 2013;45:800–10. [7] Carloni C, Subramaniam KV. FRP-masonry debonding: numerical and experimental study of the role of mortar joints. J Compos Constr 2012;16(5):581–9. [8] CNR-DT 200 R1/2013, Istruzioni per la Progettazione, l’Esecuzione ed il Controllo di Interventi di Consolidamento Statico mediante l’utilizzo di Compositi Fibrorinforzati – Materiali, strutture in c.a. e in c.a.p., strutture murarie Roma; October 2013. [9] Lu XZ, Teng JG, Ye LP, Jiang JJ. Bond-slip models for FRP sheets/plates bonded to concrete. Eng Struct 2005;27:920–37. [10] Cottone A, Giambanco G. Minimum bond length and size effects in FRPsubstrate bonded joints. Eng Fract Mechanism 2009;76:1957–76. [11] Toti J, Marfia S, Sacco E. Coupled body-interface nonlocal damage model for FRP detachment. Comput Methods Appl Mech Eng 2013;260:1–23. [12] Fedele R, Milani G. A numerical insight into the response of masonry reinforced by FRP strips. The case of perfect adhesion. Compos Struct 2010;92:2345–57. [13] Kalfat R, Al-Mahaidi R, Smith S. Anchorage devices used to improve the performance of reinforced concrete beams retrofitted with FRP composites: astate-of- the-art-review. J Compos Constr, ASCE 2011;17:14–33. [14] Grelle SV, Sneed LH. Review of anchorage systems for externally-bonded FRP laminates. Int J Concr Struct Mater 2013;7(1):17–33. [15] Bramblett RM. Flexural Strengthening of reinforced concrete beam using carbon fiber reinforced composites. Master of sciences thesis. The University of Texas at Austin; 2000. [16] Niemitz CW. Anchorage of carbon fiber reinforced polymers to reinforced concrete in shear application. Thesis of master of Science, University of Massachusetts; February 2008. ~ a S. Experimental behavior of carbon fiber[17] Niemitz CW, James R, Bren reinforced polymer (CFRP) sheets attached to concrete surfaces using CFRP anchors. J Compos Constr 2010;14(2):185–94. [18] Le Tuan Pham. Development of a quality control test for Carbon Fiber Reinforced Polymer Anchor. Thesis of master of sciences in engineering. University of Texas at Austin; May 2009. [19] American Concrete Institute (ACI). Guide for design and construction of externally bonded FRP systems for strengthening concrete structures, ACI 440.2R-08, Farmington Hills, Mich; 2008. [20] Orton SL, Jirsa OJ, Bayrak O. Design considerations of carbon fiber anchors. J Compos Constr 2008;12:608–16. [21] Zhang HW, Smith ST, Kim SJ. Optimisation of carbon and glass FRP anchor design. Constr Build Mater 2012;32:1–12. [22] Smith ST, Hu S, Kim SJ, Seracino R. FRP-strengthened RC slabs anchored with FRP anchors. Eng Struct 2011;33:1075–87. [23] Caggegi C, Pansée V, Fagone M, Cuomo M, Chevalier L. Experimental global analysis about the efficiency of carbon fiber anchors applied over CFRP straightened bricks. Constr Build Mater 2014;53:203–12. http://dx.doi.org/ 10.1016/j.conbuildmat.2013.11.086. [24] Di Iacovo G. Rinforzi in CFRP di strutture in muratura: analisi dell’efficienza di sistemi di ancoraggio. Master Thesis, University of Florence; March 2013. [25] Caggegi C. Experimental analysis of the efficiency of carbon fiber anchors applied over CFRP to brick bonded joints. Phd thesis, University of Catania (Italy)–University Paris Est (France); March 2013. [26] Briccoli Bati S, Ranocchiai G. A critical review of experimental techniques for brick material. 10th International brick/block masonry conference. Calgary; 5– 7 July 1994. p. 1247–55.