Composite Structures 92 (2010) 401–411

Contents lists available at ScienceDirect

Composite Structures

journal homepage: www.elsevier .com/locate /compstruct

Different CFRP strengthening techniques for prestressed hollow core concreteslabs: Experimental study and analytical investigation

F. Elgabbas a, A.A. El-Ghandour a, A.A. Abdelrahman a, A.S. El-Dieb b,*

a Ain Shams University, Faculty of Engineering, Department of Structural Engineering, Cairo, Egyptb United Arab Emirates University, College of Engineering, Civil and Environmental Engineering Department, P.O. Box 17555, Al Ain, United Arab Emirates

a r t i c l e i n f o

Article history:Available online 11 August 2009

Keywords:Precast prestressed hollow core RC slabsCFRP laminatesNear surface mounted (NSM)Externally bondedTransverse anchorageStrengthening efficiency

0263-8223/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.compstruct.2009.08.015

* Corresponding author. Tel.: +971 3 7133042; fax:E-mail address: amr.eldieb@uaeu.ac.ae (A.S. El-Die

a b s t r a c t

This study presents an experimental–analytical investigation on the structural behavior of precast pre-stressed hollow core RC slabs strengthened in flexure by CFRP laminates. Externally bonded and near sur-face mounted (NSM) laminates were used. The CFRP area and using transverse anchorage were alsoinvestigated. Results demonstrated that NSM technique resulted in optimum strengthening efficiency.The increased bond strength also resulted in full activation of the NSM laminates at failure. However,the NSM flexural strengthening level should be carefully designed to avoid unfavorable shear-tensionfailure mode. Moderate efficiency was associated with the externally bonded technique due to the pre-mature de-bonding. However, this efficiency was optimized by using transverse CFRP laminates asanchorage, which re-directed de-bonding further away from the laminates’ ends and delayed failure,but at much lower deformations than those of the control slab. A rational analytical study was conducted.Comparisons between the experimental and analytical results demonstrated satisfactory agreements.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Nowadays, a considerable number of existing reinforced con-crete (RC) and prestressed concrete (PC) structures need structuralupgrade either to incorporate the codes’ modifications or to allowfor any change of use and the associated changes in superimposedloads. The use of Fiber Reinforced Polymers (FRPs) has emerged asone of the most promising technologies in the field of strengthen-ing, due to their well-established advantages relative to conven-tional steel. However, since bond is the dominant factor affectingthe flexural strengthening efficiency, most of the recent researchhas focused on various ways to enhance this property, among whichusing FRP mechanical ends anchorage to delay/prevent the peelingof the externally bonded laminates [1]. Nonetheless, the efficiencyof the previous ends anchorage systems were mostly limited bythe change in the failure mode to intermediate crack de-bonding(ICD) occurring further away from the laminates’ ends [2]. Enhanc-ing the ICD capacity has been achieved through the provision ofadditional intermediate anchorages within the flexural zone [3].Lately, the near surface mounted (NSM) technique was also intro-duced as another efficient alternative to enhance bond betweenFRP laminates and the concrete substrate for RC [4] and PC elements[5]. Nonetheless, no similar application was investigated, as to theauthors’ knowledge, on prestressed hollow core concrete slabs.

ll rights reserved.

+971 3 7623154.b).

This study focuses on the positive bending CFRP flexuralstrengthening optimization of prestressed hollow core concreteslabs. The strengthening efficiency of the near surface mounted(NSM) technique is tested against the externally bonded CFRPlaminates without and with transverse anchorage. The anchorageconsisted of transverse CFRP laminates at the ends and two inter-mediate locations at the thirds of the slabs’ spans; an arrangementthat demonstrated considerable efficiency in delaying the ICD fail-ure when applied as U-Shaped sheets to RC beams strengthened inflexure by CFRP laminates [3]. Moreover, since the CFRP flexuralstrengthening level – especially when bond is optimized – is a cru-cial factor affecting the strengthening efficiency in terms of chang-ing the mode of failure to unfavorable shear mode, changing thearea of the NSM laminates is included in this study. Finally, arational analytical study is conducted, and validated against testresults.

2. Experimental work

2.1. Specimens’ details and strengthening schemes

Six full-scale precast prestressed hollow core slabs werestrengthened in positive bending with CFRP and tested to failureunder monotonic load. The slabs’ dimensions were 1200 mmwidth, 120 mm depth and 3550 mm total length over a clear spanof 3200 mm, and were internally reinforced by six pre-tensionedstrands of 6 mm diameter each.

(a) Externally bonded slabs (b) Near surface mounted slabs

Fig. 1. Specimen’s details and strengthening schemes.

Fig. 2. Application of the NSM CFRP laminates.

402 F. Elgabbas et al. / Composite Structures 92 (2010) 401–411

The un-strengthened control slab is denoted (S), the other fivespecimens were strengthened in positive bending by CFRP lami-nates, as shown in Fig. 1 and denoted as follows:

� Slab (S-NSM-2C25) was strengthened by two NSM CFRP strips of25 mm width and 1.2 mm thickness (S & P).

� Slab (S-NSM-2C16) was strengthened by two NSM CFRP strips of16 mm width and 2 mm thickness (Aslan 500).

� Slab (S-NSM-2C50) was strengthened by two NSM CFRP strips of50 mm width and 1.2 mm thickness (S & P).

� Slab (S-EB-2C50) was strengthened by two externally bondedCFRP strips of 50 mm width and 1.2 mm thickness (S & P).

� Slab (S-EBA-2C50) was similarly strengthened to slab (S-EB-2C50), but with ends as well as two intermediate lateral CFRPstrips �50 mm width and 1.2 mm thickness (S & P).

The strengthening of the slabs was done according to the man-ufacturer recommendations, and in accordance with ACI 440.2R-08[6]. For the NSM strips, a diamond blade concrete saw was used tocreate grooves in the appropriate locations. The grooves werecleaned with compressed air and manually injected with epoxyresin adhesive (Fig. 2a). The NSM laminates were then cleanedwith a solvent material, inserted in the grooves, and leveled withthe concrete substrate, as shown in Fig. 2b. On the other hand,

the externally bonded strips were traditionally installed by theapplication of the epoxy resin adhesive to the concrete substrateafter grinding and smoothing the concrete surface, followed bymanual laminates’ placement and pressing onto the adhesive witha rubber roller.

2.2. Material properties

Slabs used in the study were precast prestressed hollow coreslabs manufactured locally. Normal weight concrete of average28-days compressive cylinder strength, f 0c ¼ 32 N=mm2, tensilestrength, fct = 3.14 N/mm2, and modulus of elasticity, Ec = 26,500N/mm2, was produced as per the manufacturer’s data sheet.Internal pre-stressing strands of 1725 N/mm2 tensile strength(fpu), 1380 N/mm2 yield strength (fy), and 200,000 N/mm2 elasticmodulus (Ep), were also provided by the manufacturer. Table 1shows the design guaranteed values for the mechanical propertiesof the CFRP laminates and epoxy adhesive, as per the manufac-turer’s data sheets.

2.3. Instrumentation

Five Linear Variable Displacement Transducers (LVDTs) denotedD1–D5, as shown in Fig. 3 monitored deflections along the slab’sspan. Six electrical strain gauges (of 10 mm length) denoted S1–S6 were used to monitor the strain in the longitudinal CFRP lami-nates whether NSM or externally bonded, while, eight electricalstrain gauges (of length 10 mm) denoted S7–S14 were used to mon-itor the strain in the transverse CFRP laminates, Fig. 3. The top fi-bers’ compressive concrete strain at the mid-span section wasmeasured using one electrical strain gauge of 60-mm length, de-noted (C). The cracks’ propagation was also monitored during thetests until failure.

2.4. Test setup and loading procedure

Simply supported slabs were tested under monotonic load infour-points bending up to failure, as shown in Fig. 4. The loadwas gradually increased by successive 10 kN increments, whichwere progressively reduced near failure. Measurements and obser-vations were recorded simultaneously after each load increment.

Table 1Mechanical properties of used CFRP laminates and adhesive (design guaranteed values).

Material property Epoxy adhesive Sikadur-30 CFRP laminates

Aslan 500 S & P

Dimensions (mm) – 16 � 2 50 � 1.2 25 � 1.2Cross section area (mm2) – 32.0 60.0 30.0Comp. strength (N/mm2) >90 – – –Tensile strength (N/mm2) 25 2068 P2700 –Young’s modulus (N/mm2) 12,800 P131,000 P165,000 –Flexural strength (N/mm2)a 4 – – –Shear strength (N/mm2)a 15 – – –Ultimate strain – 0.017 0.017 –

a Concrete failure.

D2 & D3 & D4

32003550

800 800

S1S3

D3

D2

S4

D4

S6

S1S3 & S4 S2

P/2 P/2

800800

S6

120

Lin

e of

sup

port

Lin

e of

sup

port

1200

175 3200300 983 984 983 300

175

600

600

100

500

500

100

20680 750

800800

750 680 20

800 800

406

388

406

S13 S11 S9 S7

S12 S8S10

S12S14

C(at mid width)

Fig. 3. Locations of deflection and strain measurements.

Fig. 4. Testing setup.

F. Elgabbas et al. / Composite Structures 92 (2010) 401–411 403

3. Experimental results

3.1. Test observations

The crack propagation in all tested slabs followed the similartraditional flexural patterns in simply supported slabs, as shownin Fig. 5. However, strengthened slabs experienced higher numberof tensile cracks with smaller widths and extending further awayfrom the constant moment region towards the supports, hence,

resulting in more uniform distributions when compared to thecontrol slab (S). This is attributed to the stiffening effect of the lam-inates. It was also noticed that this improvements were more pro-nounced in the slabs with NSM CFRP laminates, while graduallybecoming less pronounced in the slab with externally bondedand anchored CFRP, and the improvement was the least for the slabwith externally bonded non-anchored CFRP. This shows that thecrack patterns are primarily dominated by the bond characteristicsbetween the laminates and the concrete substrate, where the NSMtechnique resulted in optimum bond, followed by the anchoredthen the non-anchored externally bonded techniques.

On the other hand, Table 2 shows the slabs’ cracking (Pcr) andfailure (Pu) loads, and the failure modes. For each slab, the crackingload was determined based on the change in stiffness of the load–deflection response. The failure load of each slab was determinedbased on either the abrupt reduction in the slab’s capacity or 20%loss of the maximum applied load. The table first demonstratesthe expected low sensitivity of the slabs’ cracking loads, Pcr, to allstrengthening techniques investigated in this study, due to thenegligible effect of the thin laminates on the gross moment ofinertia of the strengthened slabs’ cross sections. However, the12.5% lower cracking loads of the NSM slabs relative to the controlone may be due to large variability in the concrete tensile strength[7].

Fig. 5. Failure modes of tested slabs.

404 F. Elgabbas et al. / Composite Structures 92 (2010) 401–411

Unlike the pre-cracking stage, Table 2 shows the significantinfluence of the different strengthening techniques on the slabs’capacities and failure modes. In fact, while the control slab (S)

failed in a flexural ductile manner at 50 kN (pre-stressing tendonsyielding followed by compressive concrete crushing), all thestrengthened slabs failed at enhanced capacities, but through var-ious failure modes.

Concerning the NSM slabs, Table 2 shows that while slab (S-NSM-2C25), with 60 mm2 CFRP area, failed in flexure at 75 kN(50% capacity enhancement) through the rupture of the laminates,slabs (S-NSM-2C16), with 64 mm2 CFRP area, and (S-NSM-2C50),with 120 mm2 CFRP area, failed in shear-tension mode at 70 kN(40% enhancement) and 90 kN (80% enhancement), respectively.For the latter slabs, the shear-tension failure, inclined crack prop-agated from the bottom fibers outside the two line loads till theadjacent line load location, hence penetrating the concrete com-pression zone, as shown in Fig. 5c. The failure of all the NSM slabsalso occurred after undergoing comparable deformations com-pared to the control slab. In addition, all slabs showed no signsof laminates’ de-bonding until failure; i.e., no splitting cracks weremonitored at the slabs’ soffits at the locations of the NSM lami-nates. Despite the highest strengthening efficiency achieved in slab(S-NSM-2C50) in terms of capacity and deformability, its unfavor-able shear-tension failure mode reflects the need for carefully de-sign the flexural strengthening system, since the strengtheninglimit arises as an additional concern when bond is perfectly se-cured. The deformability of the slab was not affected by its failuremode since the shear-tension failure occurred close to the flexuralcapacity at a high level of flexural cracking, hence, revealing thatthe flexural strengthening limit was marginally violated. The unex-pected shear-tension failure of slab (S-NSM-2C16) provided withalmost half the CFRP area of slab (S-NSM-2C50) might be due tothe higher tensile strength of the laminates used in the slab (S-NSM-2C16) compared to their design guaranteed values. More-over, the unexpected lower shear capacity of slab (S-NSM-2C16)compared to the flexural capacity of slab (S-NSM-2C25), withnearly similar CFRP area, can again be attributed to the variabilityin the concrete tensile strength [7], especially when noting thatslab (S-NSM-2C16) failed in shear-tension mode. In general, Table2 shows the high strengthening efficiency of the NSM technique.Strengthened slabs achieved full activation of the laminates (cutoff) at flexural failure, or hit their shear capacities prior to de-bond-ing. In both cases, considerable capacity enhancements from 40%to 80% were achieved, with the latter improvement being associ-ated to doubling the laminates’ area. Nonetheless, extra careshould be taken when designing the NSM systems, in order toavoid the unfavorable shear mode of failure.

On the other hand, Table 2 shows that the externally bondedslabs (S-EB-2C50) and (S-EBA-2C50), without and with transverseCFRP anchorage, respectively, failed in a brittle manner throughlaminates’ de-bonding (Fig. 5d and e) at respective 57.5 kN (15%enhancement) and 85 kN (70% enhancement), and at much smallerdeformations when compared to the control slab. The failure of theslab (S-EB-2C50) was initiated by the end de-bonding of one lam-inate at 57.5 kN and a sudden capacity drop, followed by a partialload recovery till 50 kN, where end de-bonding of the second lam-inate occurred. The large beneficial effect of the transverse anchor-age provided in slab (S-EBA-2C50) was noted through theconsiderable 70% enhanced capacity, where simultaneous ICD ofboth laminates occurred. However, the transverse anchorage didnot fully prevent de-bonding, but only re-directed it further awayfrom the supports and delayed the ICD failure at much higherloads, which is in line with the findings of a previous studyconducted on RC beams [3].

A comparative look at the strengthening efficiencies of the NSMand the externally bonded techniques for similar laminates’ areas(120 mm2), clearly shows the lowest efficiency of the externallybonded system without transverse anchorage in slab (S-EB-2C50), from both capacity enhancement (15%) and deformability

Table 2Slabs’ capacities and modes of failure.

Slab code CFRP laminates Strengthening system Pcr (kN) Pu (kN) Dult (mm) Failure mode Capacity enhancement (%)

No. Area (mm2)

S None None – 40.0 50.0 132.0 Flexural –S-NSM-2C25 2-strips 60 NSM 35.0 75.0 140.0 CFRP Rupture 50.0S-NSM-2C16 2-strips 64 NSM 35.0 70.0 122.0 Shear-tension 40.0S-NSM-2C50 2-strips 120 NSM 35.0 90.0 144.0 Shear-tension 80.0S-EB-2C50 2-strips 120 EB 40.0 57.5 53.5 De-bonding 15.0S-EBA-2C50 2-strips 120 Anchored EB 40.0 85.0 65.0 De-bonding 70.0

F. Elgabbas et al. / Composite Structures 92 (2010) 401–411 405

reduction points of views. This lowest efficiency is significantly im-proved from a capacity enhancement perspective to 70% and 80%,for the externally bonded with transverse anchorage and NSMtechniques of slabs (S-EBA-2C50) and (S-NSM-2C50), respectively.The NSM system provided a further advantage from the ultimatedeformability point. The unfavorable shear-tension failure modeof NSM slab (S-NSM-2C50) should be prevented through carefuldesign of the system in terms of strengthening limits. A similarprecaution is difficult in case of externally bonded systems, sincelaminates’ de-bonding always occurs. Therefore, combining thehighest strengthening efficiency of the NSM technique from bothcapacity enhancement and ductility perspectives to the associatedeconomical savings in the CFRP anchorage material, the use of NSMtechniques from the structural and practical efficiencies becomestriking.

3.2. Deflections

Fig. 6 shows the load versus the mid-span deflection (D3) re-sponses for all tested slabs. First the slabs’ responses are nearlysimilar pre-cracking stiffness and cracking loads, due to the negli-gible effect of the laminates’ dimensions on the gross moment ofinertia of the slab.

Unlike the pre-cracking stage, the different strengthening tech-niques show significant influence on the slabs’ post-cracking re-sponses until the failure. All NSM slabs exhibited much higherpost-cracking stiffness until failure at considerably increasedcapacities and nearly similar deflection when compared the controlslab. In this respect, slabs (S-NSM-2C25) and (S-NSM-2C16) withCFRP area 60 and 64 mm2, respectively, demonstrated similar mod-erately increased post-cracking stiffness till failure at moderatecapacity enhancements of 50% and 40%, respectively, and compara-ble failure deflection to the control slab. However, the lower capac-ity improvement and the reduced deformability demonstrated byslab (S-NSM-2C16) could be attributed to its brittle shear-tensionfailure mode. When the NSM CFRP laminates’ area was increased

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160

Loa

d (k

N)

Deflection, D3 (mm)

S

S-NSM-2C25

S-NSM-2C16

S-NSM-2C50

S-EB-2C50

S-EBA-2C50D3

P/2 P/2

Fig. 6. Load – Mid-span deflection curves (D3) of all tested slabs.

to 120 mm2, slab (S-NSM-2C50), the post-cracking response dem-onstrated the highest stiffness among NSM strengthened slabs untilfailure, and the highest capacity enhancement (80%) and the high-est deflection among all tested slabs; this is despite its unfavorableshear-tension failure mode. This emphasizes that the extremelybeneficial effect of increasing the CFRP area when good bond is pro-vided by the NSM technique should always be guided by the flex-ural strengthening limits considerations preventing eitherunfavorable changes in the mode of failure or more dramatic defor-mability reduction (when the limits are seriously violated).

On the other hand, the externally bonded slabs (S-EB-2C50) and(S-EBA-2C50) with 120 mm2 CFRP area, without and with trans-verse anchorage, respectively, demonstrated the highest post-cracking stiffness improvements of all tested slabs, as shown inFig. 6. This might be due to the external positioning of the lami-nates bonded horizontally to the concrete surface, thus increasingthe cracked inertia of the section and the lever arm of the internalforces in the cross section when compared to that of the NSM lam-inates. Nonetheless, this improvement was achieved at a compro-mised lower capacity enhancement and higher deformabilityreduction (failure deflection) when compared to those of theNSM slab (S-NSM-2C50) provided with similar CFRP area, due tothe premature de-bonding of the external laminates. The lowercapacity enhancement (15% compared to control slab) and higherdeformability reduction (60% reduction compared to control slab)were more pronounced in slab (S-EB-2C50) without transverseanchorage. The deflection response of the slab (S-EB-2C50) alsoconfirms the previous observation of its successive laminates’ de-bonding at failure. Finally, Fig. 6 shows that the provision of thetransverse anchorage in slab (S-EBA-2C50) recovered most of thelost capacity with enhancement of 70%, which is nearly compara-ble to the 80% achieved by the NSM slab (S-NSM-2C50). However,transverse anchorage showed much lower efficiency in recoveringits dramatically reduced failure deformation. This provides clearconfirmation that the anchorage only re-directed de-bonding fur-ther away from the supports and retarded the ICD until higher loadlevels.

The previous deflection results reconfirm the optimumstrengthening efficiency of the NSM technique compared to theexternally bonded one, in terms of all structural aspects includinghighest capacity enhancement at comparable deformability to thecontrol specimen. This makes the NSM technique attractive for usein both strength governed and seismic applications. In parallel,optimizing the externally bonded technique through the provisionof transverse anchorage only resulted in comparable capacity in-crease, but at the expense of dramatically reduced deformabilityand increased material and installation cost. Nevertheless, thehigher improved stiffness of the externally bonded technique com-pared to the NSM technique makes it more attractive for use in ser-viceability-governed applications.

3.3. Tensile strains at the CFRP laminates’ mid-spans

Fig. 7 shows the load versus the mid-span tensile strain curvesof the CFRP laminates at positions (S3) and (S4), for strengthened

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

0.0000 0.0020 0.0040 0.0060 0.0080 0.0100 0.0120 0.0140 0.0000 0.0020 0.0040 0.0060 0.0080 0.0100 0.0120

Loa

d (k

N)

Tensile strain, S3 Tensile strain, S3

S-NSM-2C50

S-EB-2C50

S-EBA-2C50

Loa

d (k

N)

S-NSM-2C50

S-EB-2C50

S-EBA-2C50P/2

S3

P/2 P/2

S4

P/2

710.0=niartSPRFCerutpuR710.0=niartSPRFCerutpuR

Fig. 7. Load – Mid-span CFRP tensile strain curves at locations (S3) and (S4) of strengthened slabs with equal CFRP area (120 mm2).

0

10

20

30

40

50

60

70

80

90

100

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012

Loa

d (k

N)

Tensile strain, S1

S-NSM-2C50

S-EB-2C50

S-EBA-2C50

S1

P/2 P/2

Rupture CFRP Strain = 0.017

Fig. 8. Load – tensile strain curves at the laminates end (S1) of strengthened slabswith equal CFRP area (120 mm2).

406 F. Elgabbas et al. / Composite Structures 92 (2010) 401–411

slabs with equal CFRP area of 120 mm2. Fig. 7 shows the nearlysimilar pre-cracking responses of the slabs, as well as the signifi-cant post-cracking increases in the CFRP tensile strains until fail-ure, as noted by the reduced post-cracking stiffness of all slabs’responses. Moreover, the observed responses further highlightthe key findings of this study, including the optimum efficiencyof the NSM technique which is demonstrated by the highest strainvalues of S3 = 0.0127 and S4 = 0.0113 developed in the laminates ofslab (S-NSM-2C50) at the highest enhanced capacity. This showsthat the NSM technique’s success in taking advantage of the con-fined bonding of the CFRP laminates inside the concrete as failurewas delayed until 75% of the CFRP rupture strain, where shear-ten-sion failure occurred. It is believed that the NSM CFRP area of120 mm2 violated the flexural strengthening limit. Also wasresponsible for the unfavorable failure mode, as well as for pre-venting the laminates from being fully activated at failure, since50% reduction in the CFRP area, slab (S-NSm-2C25), resulted in fullactivation and rupture of the laminates at flexural failure of slab (S-NSM-2C25), which confirms the existence of full bonding betweenCFRP laminates and concrete.

On the other hand, the significant efficiency reduction associ-ated with the externally bonded technique becomes obvious inthe response of slab (S-EB-2C50), where end de-bonding of the firstlaminate occurred at only 15% capacity enhancement and at a verylow CFRP tensile strain level of 0.004 at location (S3), only 24% ofthe CFRP rupture strain. However, this efficiency reduction waspartially compensated by the provision of the transverse anchor-age, as can be noted by the response of slab (S-EBA-2C50), wherethe laminates’ ICD was delayed till 44% of the CFRP rupture strain(S3 = 0.0074 and S4 = 0.0071) at 70% enhanced capacity.

Moreover, the response of slab (S-EB-2C50) clearly shows itslaminates’ successive de-bonding, where the first laminate’s de-bonding caused the sudden drop in its second laminate’s response(location S4) from 57.5 kN to 45 kN. Afterwards, the figure showsthat the load recovered to 50 kN, where the second laminate’sde-bonding occurred at a strain value of S4 = 0.0042 (nearly 25%of the CFRP rupture strain).

3.4. Tensile strains near the CFRP laminates’ ends

Fig. 8 shows the load versus tensile strain curves at location (S1)near the laminates’ ends, for strengthened slabs with CFRP area120 mm2. The figure shows that all the pre-cracking responsesdemonstrated nearly zero strains. These strains relatively in-creased gradually at post-cracking, due to the concrete crackingand the associated tensile strains’ transfer to the laminates. How-

ever, the strain values remained negligible relative to the lami-nates’ ultimate tensile strain.

Near failure, Fig. 8 shows the different trends followed by theNSM and externally bonded slabs’ responses. It shows that theNSM slab exhibited no sudden changes in its strain response untilfailure, at which the end strain remained negligible. This confirmsthat no de-bonding of the CFRP laminates occurred at failure,hence, reconfirming the NSM bonding efficiency.

On the other hand, Fig. 8 shows the different trend of the exter-nally bonded slab (S-EB-2C50) response near failure, where the endstrains jumped at 95% of the failure load until reaching 0.0011 atfailure, hence, revealing the premature laminate’s end de-bondingas a major disadvantage of the externally bonded technique. How-ever, the absence of a similar jump in slab’s (S-EBA-2C50) responsenear failure does not contradict its de-bonding mode, since thetransverse anchorage re-directed bond failure further away fromthe laminate’s ends (Fig. 5e), where delayed ICD failure occurred.

3.5. Longitudinal tensile strains’ profiles in the CFRP laminates

The longitudinal tensile strain profiles at failure along the lam-inates of the strengthened slabs with equal CFRP area of 120 mm2

is shown in Fig. 9. These profiles are plotted using the strainreadings at locations S1–S6, and they clearly follow the bendingmoment requirements of simply supported condition, hence,reflecting effective composite action between the laminates and

0.0000

0.0020

0.0040

0.0060

0.0080

0.0100

0.0120

0.0140

0.00 0.25 0.50 0.75 1.00

Tens

ile s

trai

n

Position / Span

S-NSM-2C50

S-EB-2C50

S-EBA-2C50

S6

S3 S2

S4S5

S1

Rupture CFRP Strain = 0.017

Fig. 9. CFRP tensile strains’ profiles at failure of strengthened slabs with equal CFRParea (120 mm2).

0

10

20

30

40

50

60

70

80

90

0.0000 0.0002 0.0004 0.0006 0.0008 0.0010

Loa

d (k

N)

Tensile strain

S7

S8

S9

S10

S8

S7

S10

S9

Bottom surface

Rupture CFRP Strain = 0.017

Fig. 10. Load - CFRP tensile strain curves at transverse anchorage of slab (S-EBA-2C50).

0

10

20

30

40

50

60

70

80

90

100

0.0000 0.0010 0.0020 0.0030 0.0040 0.0050

Loa

d (k

N)

Compressive strain, C

S

S-NSM-2C25

S-NSM-2C16

S-NSM-2C50

S-EB-2C50

S-EBA-2C50

C

Top surface

Fig. 11. Load – Concrete compressive strain at location (C) of all tested slabs.

F. Elgabbas et al. / Composite Structures 92 (2010) 401–411 407

the concrete substrate. However, the de-bonding of the laminatesin the externally bonded slab (S-EB-2C50) without transverseanchorage could be noticed by the very low strain values. Themid-span drops in the profiles are also due to the different strainmeasurements of the two laminates (S3 and S4) at mid-span.

The NSM technique’s efficiency in preventing de-bonding isshown in Fig. 9 by slab’s (S-NSM-2C50) CFRP strain profile at fail-ure, in which mid-span strain values are 75% of the CFRP rupturestrain indicating considerable laminates’ activation, which reflectseffective composite action at failure. While full activation wasachieved due to the shear-tension failure mode of the slab. Onthe other hand, the very low strain levels (24% of the CFRP rupturestrain) together with the sudden end strain’s jump of slab’s (S-EB-2C50), clearly reveal the severe degradation in the externallybonded technique’s efficiency, and the associated partial laminates’activation at premature end de-bonding. Furthermore, the effect ofusing transverse anchorage to the externally bonded technique canbe noted by the relatively higher strain values developed at failure(44% of CFRP rupture strain) of slab (S-EBA-2C50). The suddenstrain jump at the quarter span at location (S2) of slab (S-EBA-2C50) provides further evidence that the anchorage did not pre-vent de-bonding, but only re-directed it further away from thelaminates’ ends, and delayed the ICD occurrence.

3.6. Tensile strains in the CFRP transverse anchorage of slab S-EBA-2C50

Fig. 10 shows the load versus tensile strain curves at positionsS7–S10 of the end and third-span transverse laminates for slab (S-EBA-2C50). These curves were selected where de-bondingoccurred.

All transverse laminates were nearly inactivated and developedsimilar negligible strains during the pre-cracking stage. At post-cracking, the strains increased gradually, due to concrete crackingresulting in higher strains developed in the longitudinal laminatesand the associated partial activation of the transverse ones. How-ever, the strain values remained negligible relative to the CFRP rup-ture strain, due to the bond between the longitudinal laminatesand the concrete. It should be noted that the third-span positions(S9) and (S10) consistently showed higher strain values than thoseat the end positions (S7) and (S8), due to the more intense cracking

within the mid-span region. Strain values increased gradually withincreasing loads until failure, where sudden strain jumps occurredat positions (S9) and (S10), which indicates the ICD of the longitudi-nal laminates within the third -pan’s region where the transverselaminate is located. This gives further evidence that the transverseanchorage system re-directed bond failure further away from thesupports’, and delayed the ICD failure. Finally, Fig. 10 shows that,for each transverse laminate, the middle strain (S9 or S7) was al-ways higher than the edge one (S10 or S8), due to the longitudinalde-bonded laminates’ proximity to the transverse laminates’ mid-dle region, resulting in de-bonding of transverse laminates at fail-ure at the middle region while their ends remained bonded,(Fig. 5e).

3.7. Compressive concrete strains

Load versus mid-span concrete compressive strain (at location(C) at the top surface of all tested slabs) is shown in Fig. 11. It couldbe observed the nearly similar pre-cracking responses, followed bythe significant influence of the different strengthening techniqueson the post-cracking responses. The efficiency of NSM techniquein enhancing the capacity and deformability was demonstrated byconcrete compressive strain values at failure (0.0033, 0.0027 and0.0039 for slabs (S-NSM-2C25), (S-NSM-2C16) and (S-NSM-2C50),

408 F. Elgabbas et al. / Composite Structures 92 (2010) 401–411

respectively), which were either comparable or higher than that ofthe control slab (0.0034). Thus, reflecting the advantage of NSMCFRP technique due to the confined bonding of the laminates insidethe concrete. The high compressive concrete strains at failure alsojustify the considerable deformability of slabs (S-NSM-2C16) and(S-NSM-2C50) despite their shear-tension mode of failure.

For the externally bonded slab (S-EB-2C50) without transverseanchorage, end de-bonding resulted in minor capacity enhance-ment and reduced deformability at failure. This was reflected inthe very low concrete compressive strain at failure with a valueof 0.0018 compared to that of the control slab (0.0034). Fig. 11 alsoconfirms that although the transverse anchorage in slab (S-EBA-2C50) resulted in considerable capacity enhancement (70%), itdid not prevent de-bonding, but only delayed the ICD; hence,resulting in a low concrete compressive strain at failure of0.0019 which is nearly similar to that of the unanchored slab.

4. Analytical study

4.1. Cracking and ultimate capacity prediction

The cracking moment (Mcr) was calculated based on the grosscross section properties using Eq. (1), where the maximum tensileconcrete stress was assumed to be equal to the tensile strength (fct)of the concrete, and the effect of the thin CFRP laminates wasneglected.

MCr ¼ fct � Zb þPe:Zb

Acþ Pe � e ð1Þ

where Pe is the pre-stressing force after losses, e the eccentricity ofpre-stressing reinforcement and Zb is the section modulus of ten-sion side.

ActualCross Section

EquivalentCross Section

dt d s C

FRP

1

d t d s CFR

P 1

dC

FRP

2

(a) NSM tec

d t d

s

CFR

P d t d

s CFR

P

ActualCross Section

EquivalentCross Section

(b) EB tech

Fig. 12. Stress and strain distributions for FR

The flexural failure moments (Mfc or Mft), corresponding toeither compressive concrete crushing or tensile CFRP rupture,respectively, were calculated as per Fig. 12. The figure is basedon the ACI 440.2R-08 [6] stress and strain distributions for FRPstrengthened sections under flexure, after setting the strengthreduction factors equal to unity and incorporating the ACI440.2R-08 [6] modifications for prestressed members. While anelastic-perfectly plastic stress–strain relationship was associatedto the pre-stressing steel and the CFRP laminates were given a lin-ear elastic response until failure, using the manufacturer’s datasheet values. Calculations were then conducted at the mid-spansection according to the following steps:

(a) Assume maximum compressive concrete strain, ec = ecu =0.0035.

(b) Assume a neutral axis depth, c.(c) Determine the internal forces in the compression and the

tension zones based on the tensile strains at the level ofthe prestressed strands (eps) and the CFRP laminates (ef).Note that the total tensile strain at the level of pre-stressingstrands (eps) is:

hniqu

nique

P stren

eps ¼ epe þ es þ ece ð2Þ

where epe is the effective strain in pre-stressing reinforce-ment after losses, es the strain in pre-stressing steel resultingfrom strain compatibility at ultimate moment es ¼ ec

ds�cc

� �and ece is the concrete strain at level of the pre-stressing rein-forcement due to the pre-stressing force.

Also note that the tensile strain at the level of the CFRPlaminates (ef) is the one resulting from the strain compatibil-

ity at ultimate moment ef ¼ ecdf�c

c

� �.

(d) Check the equilibrium of the section according toEq. (3):

c

f 1

s

c c

f

1

1 c

C

Tf 1

Ts

Strain DiagramActual

Concrete StressEquivalent

Concrete Stress Internal Forces

CFRP 2 Tf 2

pe

e

c

f

s

c c

f

1

1 c

C

Tf

Ts

Strain DiagramActual

Concrete StressEquivalent

Concrete Stress Internal Forc

pe

gthened sections under flexure.

F. Elgabbas et al. / Composite Structures 92 (2010) 401–411 409

Z Z Z

Table 3Compar

Spec

SS-NSS-NSS-NSS-EBS-EB

Ac

fcdAc þAps

fpsdAps þAf

ff dAf ¼ 0 ð3Þ

where fc is the concrete stress in compression, fps the stress ofthe bottom prestressed reinforcement, ff the stress of theCFRP laminates, Ac the area of concrete, Aps the area ofthe bottom prestressed reinforcement and Af is the area ofthe CFRP laminates.

(e) Change the assumption of the neutral axis depth until equi-librium is satisfied.

(f) Calculate the failure moment (Mfc) of the section accordingto Eq. (4):

Z Z Z

Ac

fcydAc þAps

fpsydAps þAf

ff ydAf ¼ Mfc: ð4Þ

where y is the distance measured from the neutral axis.

(g) Assume rupture CFRP strain, ef = efu = 0.017 and repeat

steps (b)–(f), where the failure moment (Mft) of thesection can be calculated from Eq. (4). Note that thecorresponding concrete and prestressed strands strains arecalculated in step (c) based on the 0.017 CFRP rupture strainvalue.

The de-bonding moments, Md, of the externally bonded slabswere calculated using the sectional analysis of Fig. 12, but the fail-ure was defined by the tensile strains in the CFRP laminates reach-ing their de-bonding value, (efd), as per the ACI 440.2R-08 [6]equation:

efd ¼ 0:41

ffiffiffiffiffiffiffiffiffiffiffif 0c

nEf tf

s6 0:9efu: ð5Þ

where f 0c is the cylinder concrete compressive strength, efd the effec-tive strain level in FRP reinforcement, efu the design rupture strainof FRP reinforcement, n the number of plies of FRP reinforcement,Ef the tensile modulus of elasticity of FRP and tf is the nominal thick-ness of one ply of FRP rein-forcement.

Bond capacity predictions were not performed for theNSM slabs, since no de-bonding was monitored during theirtesting.

The associated cracking (Pcr) and different failure loads (Pfc, Pft,Pd) were then calculated using the four-point bending test setupand the associated bending moment diagram.

The slabs’ shear capacities, Ps, were also calculated according tothe CPCI design manual [8], using the following equations forconcrete shear resistance (Vc) calculation at a specific section aftersetting the strength reduction factor equal to unity:

dv ¼ the greater of0:9d or 0:72t ð6Þ

Sze ¼35Sz

15þ agP 8:5Sz ð7Þ

ex ¼Mf =dv þ Vf � Vp � Apfpe

2ðEsAs þ EpsApsÞð8Þ

ison between predicted and test results.

imen code Analytical prediction

Pcr (kN) Pfc (kN) Pft (kN) Pd (kN) Ps (kN) Pu (kN)

34.4 43.5 – – 73.0 43.5M-2C25 34.4 – 66.1 – 73.0 66.1M-2C16 34.4 – 63.7 – 73.0 63.7M-2C50 34.4 73.8 – – 73.0 73.0-2C50 34.4 90.3 – 56.7 73.0 56.7A-2C50 34.4 90.3 – 56.7 73.0 56.7

b ¼ 0:401þ 1500ex

13001000þ sze

� �ð9Þ

Vc ¼ /ckbffiffiffiffif 0c

qbwdv ð10Þ

where t is the slabs’ thickness, dv the effective shear depth, taken asflexural lever arm but need not be taken less than 0.9 d, bw the sum-mation of webs’ thickness of hollow core slab, Sze the equivalent Sz

allowing for aggregate size, Sz the crack spacing parameter, depen-dent on the longitudinal reinforcement; i.e., Sz = dv for one bottomsteel layer, and Sz = bars’ spacing for multiple layers, ex the longitu-dinal strain of flexural tension chord of the member, Mf the momentdue to applied load, Vf the applied shear force, Vc the concrete shearresistance, Vp the effective pre-stressing force component in thedirection of the applied shear (Vp = zero for straight tendons), f 0c cyl-inder concrete compressive strength, fpe the effective stress in pre-stressed reinforcement after allowance for all prestress losses, Aps

the area of bottom prestressed reinforcement, Eps the modulus ofelasticity of pre-stressing reinforcement, As the area of bottomnon-prestressed reinforcement (if any), Es the modulus of elasticityof non-pre-stressing reinforcement, k the factor accounting for lowdensity concrete (k = 1.0 for normal weight concrete), uc theStrength reduction factor (uc = 1.0) and ag is the aggregate’s size.

The Vc calculation is repeated at different sections along theslab’s span and the corresponding diagram is superimposed onseveral shear force (Vf) diagrams associated to the used testing set-up and drawn at different load levels. The shear capacity (Ps) isthen determined as the load level which results in the first tangen-tial interference between the corresponding shear force (Vf) dia-gram and the concrete shear resistance (Vc) one.

Table 3 shows the comparison between the slabs’ predicted andtest cracking load (Pcr), failure load (Pu) and failure modes. For eachslab, the predicted failure mode is the one resulting in the leastanalytical capacity (Pu). The table shows very good correlations be-tween the experimental and predicted results. However, for slab(S-NSM-2C16) failure mode prediction (rupture of CFRP) is differ-ent than the experimental (shear-tension failure) might be due tothe higher actual tensile strength of the CFRP laminates comparedto the manufacturer’s design guaranteed value used in the predic-tions. Moreover, for slab (S-EBA-2C50) the higher experimentalcapacity compared to the predicted value (49.9% increase) can bedue to the beneficial transverse anchorage effect provided experi-mentally and were not accounted for analytically. This increase ishigher than the 30% value of the ACI 440.2R-08 [6]. This also ex-plains why the predicted behavior of slab is similar to that of theexternally bonded slab without transverse anchorage, as shownin Table 3.

4.2. Deflection prediction

The analytical deflection was calculated using numerical inte-gration of curvature at many sections along the span [9] accordingto Eq. (11) and Fig. 13.

Experimental result PcrðExp:ÞPcrðPre:Þ

PuðExp:ÞPuðPre:Þ

Failure mode Pcr (kN) Pu (kN) Mode of failure

Flexural failure 40.0 50.0 Flexural 1.163 1.150CFRP rupture 35.0 75.0 CFRP rupture 1.017 1.135CFRP rupture 35.0 70.0 Shear-tension 1.017 1.099Shear failure 35.0 90.0 Shear-tension 1.017 1.233De-bonding 40.0 57.5 De-bonding 1.163 1.014De-bonding 40.0 85.0 De-bonding 1.163 1.499

Fig. 13. Numerical integration of curvature for deflection calculation.

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140

Loa

d (k

N)

Deflection, D3 (mm)

(S)Exp.(S)Pred.(S-NSM-2C25)Exp.(S-NSM-2C25)Pred.(S-NSM-2C16)Exp.(S-NSM-2C16)Pred.

(a) Slabs S & S-NSM-2C25 and S-NSM-2C16

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100 120 140 160

Loa

d (k

N)

Deflection, D3 (mm)

(S-NSM-2C50)Exp.

(S-NSM-2C50)Pred.

(S-EB-2C50)Exp.

(S-EB-2C50)Pred.

(S-EBA-2C50)Exp.

(S-EBA-2C50)Pred.

(b) Slabs S-NSM-2C50, S-EB-2C50 and S-EBA-2C50

D3

P/2 P/2

D3

P/2 P/2

Fig. 14. Comparison between the experimental and analytical deflection responses(D3).

410 F. Elgabbas et al. / Composite Structures 92 (2010) 401–411

D ¼X/ixi þ /iþ1xiþ1

2Dxi ð11Þ

where D is the mid-span deflection, /i the curvature correspondingto a distance xi from the support, /i+1 the curvature correspondingto a distance xi+1 from the support and Dxi is the xi+1 � xi.

In the analysis herein, the integration was performed at foursections along half of the slab’s span. In general, the sectionsshould be selected at any expected abrupt curvature change. In thiscase, the selected sections were at the support, location where ap-plied moment equals cracking moment of cross section, location ofthe concentrated load and at mid-span. Needless to mention thatincreasing the number of sections would inevitably increase theaccuracy of the calculations. For each slab, the mid-span deflection(D) of Eq. (11) was calculated at increasing load levels until theanalytically predicted slab capacity. At each load level, the selectedsections curvatures in Eq. (11) were calculated using the straincompatibility approach and iterative procedure previously de-scribed in Section 4.1 and Fig. 12. For each section, the momentcorresponding to the chosen load level is first determined(Fig. 13) and the iterative procedure is accomplished backwardsto determine the top fibers compressive concrete strain (ec) - neu-tral axis depth (c) combination resulting in the equilibrium of thesectional tensile and compressive forces. The sectional curvature(/) at the chosen load level is then calculated using the followingequation:

/ ¼ ec

cð12Þ

Fig. 14 shows the comparisons between the test and analyticalload–deflection responses of all slabs. Although Fig. 14 shows verygood correlation between the experimental and analytical re-sponses, the slightly higher analytical pre-cracking stiffness couldbe due to either overestimated concrete elastic modulus in theanalysis or to the reduction of the experimental pre-cracking stiff-ness associated with micro-cracks induced during slabs’ handlingor the limited number of sections selected for the numerical inte-gration process.

5. Conclusions

Although the stiffening effect of the CFRP laminates resulted inimproved and more uniform crack patterns relative to the controlslab, the pre-cracking response was nearly unaffected by strength-ening, since it is governed by the gross concrete section’s proper-ties, rather than the thin laminates. Hence, the slight reducedcracking loads of the NSM slabs (12.5% reduction) may be due tothe variability in the concrete tensile strength.

The NSM technique and the associated prevention of the CFRPde-bonding failure resulted in optimum strengthening efficiencyin terms of all structural aspects including most uniform crack dis-tribution, as well as the highest capacity increase (80%) at compa-rable deformability and failure strains to the control slab.

The good bond of the NSM technique initiated the crucial issueof flexural strengthening limits, since doubling the CFRP areachanged the full activation and rupture of the laminates at 50%

F. Elgabbas et al. / Composite Structures 92 (2010) 401–411 411

flexural capacity enhancement to the unfavorable shear-tensionfailure mode at 80% capacity enhancement. Nonetheless, no fur-ther negative influences on deformability was monitored, sincethe shear-tension failure occurred at a high level of flexuraldamage.

The unanchored externally bonded technique and the associ-ated CFRP de-bonding failure resulted in the lowest strengtheningefficiency in terms of all structural aspects, including least uniformcrack distribution, as well as the lowest capacity enhancement(15%) at a seriously reduced deformability and failure strains.The 24% activation of the laminates at failure was a clear conse-quence of the technique’s disadvantages.

Although the provision of end and two-thirds span transverselaminates to the externally bonded technique did not totally pre-vent bond failure, it successfully re-directed de-bonding furtheraway from the laminates’ ends. The intermediate transverse lami-nates also delayed the ICD failure, hence resulting in partiallyrecovered strengthening efficiency relative to the none-anchoredsystem in terms of all structural aspects. This included uniformcrack distribution and high capacity enhancement (70%) at 44%partial activation of the laminates. Nonetheless, the system didnot efficiently recover the reduced deformability and failurestrains.

Although, the externally bonded slabs demonstrated the high-est post-cracking stiffness improvements which might be due tothe external positioning of the laminates bonded horizontally tothe concrete surface, thus utilizing the whole cross section of thelaminate at once and increasing the lever arm when compared tothat of the NSM laminates, which are utilized progressively asthe load increases. Nonetheless, premature de-bonding of thelaminates compromised the enhancement in the capacity andsignificantly reduced deformability. Accordingly, while the NSMtechnique appears attractive in strength- and seismic-governed

applications, the externally bonded one is more suitable for ser-viceability-governed applications.

A rational analytical study is conducted, where the strain com-patibility approach, the ACI 440.2R-08 and the CPCI design manualwere used for flexural capacity, CFRP de-bonding and shear capac-ity predictions, respectively. Analytical deflections were alsopredicted using curvature integration along the span, and incorpo-rating the CFRP tensile and bond characteristics. The validity of theprevious predictions is established through comparisons againsttest results.

References

[1] Hosny A, Sayed-Ahmed E, Abdelrahman A, Elhlaby A. Strengthening precast –prestressed hollow core slabs to resist negative moments using carbon fiberreinforced polymer strips: An experimental investigation and a critical reviewof Canadian Standards Association S806-02. Can J Civ Eng (CJCE)2006;33:955–67.

[2] El-Ghandour A, Abdelrahman A, Nasr E, Aly H. Moment redistribution in RCcontinuous beams strengthened in flexure by CFRP laminates. In: Proceedings ofthe second international conference on FRP composites in civil engineering(CICE), Adelaide, Australia 2004, 8-10 December. p. 227–35.

[3] Abdelrahman A, El-Ghandour A. Effect of U-shaped FRP warps on the bondbehavior of CFRP strips. In: Proceedings of the eighth international symposiumon fiber reinforced polymer reinforcement for concrete structures (FRPRCS-8),Patras, Greece, 2007, 16-18 July. p. 96–7 [volume for extended abstracts].

[4] El-Hasha R, Rizkalla S. Near surface mounted FRP bars reinforcement for flexuralstrengthening of concrete structures. ACI Struct J 2004;101(5):717–26.

[5] Hassan T, Rizkalla S. Flexural strengthening of prestressed bridge slabs with FRPsystems. PCI J 2002;47(1):76–93.

[6] ACI 440.2R-08.Guide for the design and construction of externally bonded FRPsystems for strengthening concrete structures. Emerging technology series, ACICommittee 440, Farmington Hills, MI 48331, USA, July 2008. p. 76.

[7] Leonhardt F. Cracks and crack control in concrete structures. PCI J1988;33(4):124–45.

[8] Canadian prestressed concrete institute CPCI. Design manual of precast andprestressed concrete. 3rd ed. Ottawa, Canada, 1996. p. 445.

[9] Collins M, Mitchell D. Prestressed concrete structures. Englewood Cliffs (NJ,USA): Prentice Hall; 1991. p. 766.