Debonding of externally bonded CFRP at low and high temperatures

E.L. Klamer & C.S. Kleinman Eindhoven University of Technology, Eindhoven, the Netherlands

D.A. Hordijk Eindhoven University of Technology, Eindhoven, the Netherlands & Adviesbureau ir. J.G. Hageman BV, Rijswijk, the Netherlands

ABSTRACT: Fiber Reinforced Polymers (FRP) has proven to be an effective strengthening material in the construction industry, due to its low weight, non-corrosiveness and high strength. Extensive research has been carried out into the strengthening of concrete structures with externally bonded FRP. It turned out that debon-ding of the FRP is governing the design of most FRP strengthening applications. One of the parameters, which may affect the bond properties of the FRP-concrete joint and which will induce additional stresses is the ambient temperature. Only very little research into the influence of temperature has been carried out in past studies. This paper will give the results of an exploratory experimental research project in which the in-fluence of temperature on the debonding behavior of CFRP, externally bonded to small scale concrete speci-mens, is investigated.

1 INTRODUCTION

Strengthening of concrete structures with externally bonded Fiber Reinforced Polymers (FRP) has be-come increasingly popular in the construction indus-try. Strengthening with FRP laminates has many ad-vantages over strengthening with steel strips, like the high strength, the non-corrosiveness, the low weight and consequently the easiness of application and maintenance.

Last two decades, extensive research has been car-ried out into the use of FRP as a strengthening mate-rial. In 2001, fib-Bulletin 14 (fib Task Group 9.3. 2001) has been published, in which available infor-mation was gathered and design guidelines for the calculation and application of FRP were given. Gui-delines will most probably be updated in the near fu-ture, as the various related topics are subject of on-going research and development. The main issue governing the design of a FRP strengthening application is the debonding of the FRP. Debonding is initiated before the ultimate ten-sile strength of FRP can be reached. Research into the debonding behavior is rather difficult, due to the explosive and sudden way of failure. Although a lot of research has been carried out into the debonding of externally bonded FRP, still several questions ha-ve to be answered.

2 RESEARCH SIGNIFICANCE

One of the research needs, which has so far received only little attention, is the influence of temperature on the debonding of externally bonded Carbon Fiber Reinforced Polymers (CFRP). Temperature can af-fect the debonding behavior in several ways. Tem-perature changes will induce additional thermal stresses, due to the significant difference in coeffi-cient of thermal expansion between concrete (αc ≈ 10 × 10-6 / °C) and e.g. CFRP (αf ≈ 0 × 10-6 / °C) (longitudinal direction). Furthermore, differential temperature in the strengthened structure causes im-posed deformations. Both influences will induce ad-ditional stresses in the CFRP and the concrete and may result in debonding at a different load level. The material properties of concrete and CFRP and the adhesive layer in between are also affected by changes in temperature. E.g. the strength and stiff-ness of concrete will increase when the temperature drops below zero (Di Tommaso et al. 2001), while the adhesive and CFRP become more brittle due to the properties of the epoxy component in these ma-terials. Increasing the temperature has a negative ef-fect on the adhesive, especially when the glass-transition temperature is reached (Tg ≈ 45°C – 80°C for most epoxies). Reaching this temperature will re-sult in a significant drop in strength and stiffness.

3 REVIEW OF PREVIOUS WORK

3.1 Tadue & Branco Double-lap shear tests were carried out (Fig. 1) on steel laminates externally bonded to concrete speci-mens, while three different concrete grades were ap-plied (Tadeu & Branco 2000). In this case thermal stresses will be small because steel and concrete have almost the same coefficient of thermal expan-sion. The results of the experiments showed a reduc-tion of the failure load with an increase of tempera-ture. Especially for high strength concrete (fcm,cube = 74.1 N/mm2) there was a significant reduction. In-creasing the temperature from 20°C to 30°C resulted in a decrease of the failure load by 32%, while for low strength concrete (fcm,cube = 27.9 N/mm2) a de-crease of only 1% was found.

The failure mode was also affected by the tem-perature. For 20°C and 30°C failure of the concrete occurred, while for higher temperatures, failure of the adhesive occurred. For specimens at 60°C, the failure load was reduced to 45 - 51% of the initial failure load, while for 90°C only 24 - 29% was left. At 120°C, hardly any capacity was left (7 - 9%).

steel strip

concreteload

bond length

loadsupport

steel stripadhesive

Figure 1. Double-lap shear test (Tadeu & Branco 2000)

3.2 Blontrock

Double-lap shear tests (Fig. 2) were carried out on concrete specimens, strengthened with externally bonded CFRP laminates (Blontrock 2003). Anchor-age sheets have been applied to make sure debond-ing occurs at one side, which reduces the number of needed strain gauges. The experimental results were different from the results of Tadeu & Branco for steel plates. Increasing the temperature from 20°C to 40°C resulted in a significant increase of the failure load (41%) in stead of a decrease (fcm,cube = 40 N/mm2).

FRP laminate

concrete load

bond length

load

anchorage sheet

precrack adhesive

Figure 2. Double-lap shear test (Blontrock 2003)

Blontrock mentioned two possible causes for the dif-ferent results compared to Tadeu & Branco:

1. The dimensions of the specimens: A significant difference in bonded surface resulted in a different failure mode. Failure occurred in the concrete near the bonded surface (depth approximately 0-1 mm), whereas Tadeu & Branco found failure of the con-crete at a depth of 30 mm from the bonded surface.

2. The difference in coefficients of thermal expan-sion. For steel and concrete, the coefficient of ther-mal expansion is about the same, whereas it differs significantly in case of CFRP laminates. This will induce additional interfacial stresses between CFRP and concrete. These stresses obviously had a posi-tive effect on the capacity.

Further increasing the temperature to, respec-tively, 55°C and 70°C resulted in a decrease of the failure load, although at 55°C the failure load was still higher than the initial failure load at 20°C. The decrease of the failure load was caused by the sof-tening of the adhesive.

3.3 Di Tommaso et al.

Three-point bending tests on small scale concrete specimens (Fig. 3) without internal reinforcement were performed (Di Tommaso et al. 2001). The specimens were strengthened at room temperature with two different types of CFRP (high and normal E-modulus laminates) and tested at four different temperatures, -100°C, -30°C, 20°C and 40°C.

The results showed that increasing the tempera-ture to 40°C decreased the load capacity. This was explained by the additional thermal stresses and sof-tening of the adhesive. This result is opposite to the results obtained by Blontrock.

Load

Failure surfaceAdhesiveCFRP

Load

Failure surfaceAdhesiveCFRP

Load

Failure surfaceAdhesiveCFRP

T = 40°C

T = 20°CT = -50°C (Low CFRP E-modulus)

T = -50°C (High CFRP E-modulus)T = -100°C

(c) CFRP delamination

(b) Concrete shear failure

(a) Cohesive failure

Figure 3. Failure modes (Di Tommaso et al. 2001)

As a result of a decrease of the temperature to re-spectively -30°C and -100°C, the maximum load in the linear part of the load-deflection relation in-creased. However, the ultimate failure load was in most tests lower than the failure load at 20°C, which means that the ductility decreased.

In the experiments, three different types of failure were observed, mainly governed by the applied tem-perature (see Fig. 3). For specimens at 40°C cohe-sive failure in the adhesive layer was found, due to the softening of the adhesive. For moderate tempera-tures, concrete shear failure was found and for lower temperatures delamination of the CFRP was found.

4 EXPERIMENTS The available experimental results in literature have shown that the influence of temperature can be sig-nificant. However, it looks as if the temperature in-fluence depends on the experimental test setup that is used. Further research is needed to gain a better insight in the influence of temperature on the bond-ing capacity of CFRP to concrete. In this paper, the results are given of a series of experiments in which the influence of temperature is investigated in a double-lap shear test setup. Basically, the test setup was similar to that used by Blontrock. For the CFRP laminates different dimensions have been applied and only one threaded rod was used to make the connection of the specimen to the loading device. Besides the effect of temperature increase also the effect of freezing (-10°C) is investigated.

4.1 Preparation of the test specimens

The specimens were produced in three series of four concrete specimens each (150 × 150 × 800 mm3). To be able to apply the specimen in the tensile testing machine, one steel threaded rod (M24) of about 1 m was put into the center of the specimen (see Fig. 4). The specimens, including the rod, were cut in half after curing of the concrete (at 20°C at 60% relative humidity). This has the advantage that the two parts have the exact same width and height. After about 28 days, the concrete surfaces were sandblasted and two CFRP laminates (50 × 1.2 × 650 mm3) were bonded to the two side faces of the concrete speci-mens. The adhesive had a thickness of about 1.5 mm. 50 mm of cardboard has been placed in be-tween the CFRP and concrete at the location of the saw cut. In this way, 50 mm in the middle of the CFRP (25 mm at each side of the saw cut) remains unbonded (see Fig. 4), which was done in order to avoid local stress concentrations at the saw cut.

800 mm650 mm

150 mm

150 mm

50 mm 50 mm (unbonded)

CFRP saw cutadhesive

Figure 4. Dimensions of the test specimen

4.2 Material properties

The concrete compressive strength and tensile split-ting strength were determined using 150 × 150 × 150 mm3 cubes. The bond strength was determined according to CUR Recommendation 20 (CUR 1990) by bonding steel cylinders (∅ 50 mm) on the sand-blasted concrete surface and pulling them off by a hydraulic jack. The average concrete properties are given in Table 1. The concrete for series C had a much lower strength than intended. Probably in the mixing there was a mistake. Nevertheless, the specimens have been used in the exploratory inves-tigation.

The CFRP (SIKA CarboDur S512) has a tensile strength of 2800 N/mm2, an elastic modulus of 165,000 N/mm2 and a thermal expansion coefficient of 0.3 × 10-6 / °C in the longitudinal direction and is temperature resistant till at least 150°C (all accord-ing to the manufacturer). The adhesive (SikaDur-30) has a bond strength of about 4 N/mm2 (due to con-crete failure) and a bond strength of 15 N/mm2, when tested on sandblasted steel. It has an elastic modulus is 12,800 N/mm2

, a coefficient of thermal expansion of 90 × 10-6 / °C and a glass transition temperature (Tg) of 62°C. Table 1. Concrete material properties Material properties Series A Series B Series C N/mm2 N/mm2 N/mm2

Compressive str. (fcm,cube) 37.7 31.5 18.6 Tensile splitting st. (fctm,sp) 2.72 2.40 1.69 Bond strength (fcbm) 3.68 3.10 2.64

4.3 Test setup

The specimens are tested in a 250 kN tensile testing machine (see Fig. 5a). Steel clamps have been used at the bottom side of the specimen, to make sure that debonding of the CFRP is initiated at the upper side. Series A and B have been used to investigate the in-fluence of elevated temperatures. The specimens we-re heated in an oven for at least 12 hours before test-ing. Series C has been used to investigate the influence of low temperature (-10°C). These have been frozen for at least 24 hours.

All heated and frozen specimens were packed with isolation during the tests (see Fig. 5b). Di Tommaso et al. have shown that only 2.5 hours are needed to bring the core of a specimen (100 × 100 × 700 mm3) to the desired temperature. Though not measured, it is expected that for the slightly bigger specimens in this investigation, 12 hours of heating is enough to reach a uniform temperature and entirely warm up the specimen. The surface temperature is measured during the experiment and changed with a maximum of 2°C.

Figure 5a. Test setup 5b. Isolated part of test setup

4.4 Measurements

Two Linear Variable Differential Transformers (LVDT’s) placed diametrically to the center of the specimen were used to measure the displacement over the saw cut (see Fig. 6).

Figure 6. Test setup including the measurement devices In order to determine the strain development over the length of the CFRP laminate, five strain gauges have been applied per laminate (only one specimen per investigated temperature). The strain gauges (10 mm) have been applied at 10, 80, 150, 220 and 290

5 THEORETICAL ANCHORAGE FORCE

mm from the end of the laminate. The measured strains were temperature corrected.

The maximum CFRP anchorage force at 20°C can be calculated with the model of Holzenkämpfer,

ximum CFRP anchorage force (N ) is related to the frac-modified by Neubauer & Rostásy. The ma

fa,maxture energy (GF) of concrete, assumes a bilinear bond-slip relation (Holzenkämpfer 1994) and can be calculated with equation (1).

fa ,max c f F f f

c b f F f f cbm

N k b 2 G E t

k k b 2 c E t f

= α ⋅ ⋅ ⋅ ⋅ ⋅ ⋅

= α ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ (1)

where α = reduction factor to account for the influ-ence of inclined cracks on the bkc = factor to account the state of compaction of the concrete; k = geometry factor; b = width of FRP;

imited du

ond strength;

b fEf = elastic modulus of FRP; tf = thickness of FRP; fcbm = mean value of concrete bond strength.

The corresponding maximum anchorage length (ℓb,max) can be determined with equation (2). A hig-her anchorage length does not result in a higher an-chorage force because the anchorage force is l

e to the fracture energy of concrete (Neubauer & Rostásy 1999).

F f fb,max 2

max

2 G E t2

F f f

cbm

c E t1.57f

⋅ ⋅ ⋅= ⋅α ⋅

τ

⋅ ⋅= ⋅α ⋅

The maximum anchorag(Nfa,max) and corresponding anchorage length (ℓb,max) for the applied concrete mixes are given in Table 2. The available anchorage length (ℓ ) is 300 mm,

ℓb

(2)

e force (per laminate)

bwhich is longer than the maximum anchorage length. The analytical failure load (Fmax) for the spe-cimen with two laminates can therefore be taken twice the maximum anchorage force. Table 2. Anchorage force, anchorage length and failure load Series Nfa,max ℓb,max Fmax kN mm kN

LVDT

strain gauges

saw cut

clamps

thermo couple

10 mm

290 mm

mm A 31.5 148 63 300 B 28.9 160 58 300 C 26.6 174 53 300

6 TEST ULT

Elevated temperatures

The measured failure loads as function of the ap-plied temperature are plotted in Fig. 7. The two specimens from series A, tested at 20°C, failed ex-plosively by debonding of the concrete at the joint

RES S

6.1

with the adhesive. The average failure load of well with the theoretical

in failure load when increasing th

m of th

61.4 kN corresponds very failure load (63 kN).

For series A and B the same tendency can be ob-served. When increasing the temperature, first the failure load slightly increases, while for tempera-tures 65°C and 75°C the failure load was signifi-cantly lower than at 20°C. This is similar to what Blontrock 2003 had found, though he reported a much larger increase

e temperature from 20°C to 40°C. The reduction in strength for the higher temperatures can be under-stood when it is realized that the glass transition temperature of the applied adhesive was 62°C.

The measured load-displacement curves for the series A are plotted in Figure 8. As can be seen, the stiffness decreases with increasing temperature. In Figure 9 the strain distributions in the CFRP for an arbitrarily chosen load level of 35 kN are compared. It appears that where for 20°C there is a concentra-tion of strains in the first approximately 70 m

e strip adjacent to the saw cut, it is spread over a longer length for the strip at 50°C. At 75°C there is an almost linear strain distribution from the begin-ning to the end of the strip.

67.363.958.9

48.950.7

40

50

60

70

80

load

(kN

)

39.6

35.9

46.7

0

10

20

30

0 20 40 60 80

Temperature (oC)

Failu

re Series A

Series ASeries BSeries B

Figure 7. Failure loads of the specimens (series A and B)

Figure 8. Load displacement curves of series A

eries A and B

oncrete attached to the CFRP strip. For both failure

)Figure 9. Strain development in CFRP (35 kN, s

0

500

1000

1500

2000

0 100 200 300

Distance from CFRP end (mm)

Stra

in ( m

m/m

)

50°C

20°C

40°C

75°C

Strain in unbonded laminate

The strain in the unbonded part of the FRP laminate (300 mm from CFRP end) can be calculated with F / (Af × Ef). At a load of 35 kN, the strain in the FRP is 17,500 / (50 × 1.2 × 165.000) = 1768 μm/m, whichcorresponds very well with Figure 9.

Two different types of failure were found in the experiments (see Fig. 10). At moderate temperatures (20°C and 40°C) failure occurred in the concrete, leaving about 1-3 mm of concrete attached to theunbonded CFRP strip. At higher temperatures (50°C, 65°C and 75°C) debonding occurred in beween the adhesive and concrete, leaving hardly any

-tcmodes debonding occurred in a rather explosive way.

Based on the presented results it is expected that the behavior at elevated temperatures can be ex-plained as follows. Due to a higher temperature the stiffness of the adhesive reduces. As a result the bond stresses along the length of the laminate be-come more equally distributed. For temperatures up to at least 50°C most probably the bond strength will not significantly be reduced so that the more uni-form bond stress distribution yields the higher fail-ure load.

Figure 10. (a

0

10

20

50

60

70

0 0,2 0,4 0,6 0,8 1

Displacement (mm)

F (k

N) 20°C

50°C

30e lo

40adai

lur

75°C

) Failure in the concrete and (b) in the adhesive

Or putted in other words, the influence of the tem-perature on the stiffness and resulting in a more uni-form bond stress distribution governs over a possible negative influence on the bond strength. For tem-peratures 65°C and 75°C (> Tg of the adhesive: 62°C) the strain distribution along the laminate is even more linear, resulting in an almost uniform bond stress distribution. However, at these tempera-tures, a reduction of the strength of the adhesive and bond strength is governing, resulting in a significant lower failure load and o type of failure. ther

6.2 Low temperatures

Because of the rather poor concrete quality of the concrete used in series C, it was not combined with the tests for the other series and used separately to investigate the effect of low temperatures. Two ref-erence specimens are tested at 20°C and failed at 56.7 kN and 60.1 kN respectively (see Fig. 11), which is slightly higher than the analytical calcu-lated failure load (see Table 2).

39.8

0

10

20

30

40

50

-20 -10 0 10 20 30

Temperature (oC)

Failu

re lo

ad (k

56.060

N) 56.7

60.170

Series C

igure 11. Failure load of the specimens (series C)

Decreasing the temperature to -10°C resulted for one specimen in a significantly lower failure load (39.8 kN), while for the other the failure load did not differ much (56.0 kN) from the reference tests. For the specimen at -10°C that resulted in an almost equal failure load as for the specimens tested at 20°C the measured load-displacement curve was also similar. Based on these two experiments no conclusions can be drawn on the influence of low temperatures on the bond behavior.

CONCLUSIONS AND RECOMMENDATIONS

pared to 20°C a slightly increased anchorage capacity was found, which is explained by a more uniform bond

esive at these temperatures. For temperatures above

e CFRP and adhesive.

ts strengthened with externally bonded FRP’s. Ghent: Ghent University (in Dutch).

0. Determination of the crete. Gouda: Stichting

CUR (in Dutch).

onded FRP reinforcement for RC structures. Lausanne: fédération internationale du béton.

1994. Ingenieursmodelle des Verbundes hrung für Beton-bauteile. Braunschweig:

Tad

F

7 For the influence of temperature on the bond behav-ior of externally bonded CFRP contradictory results are reported in literature. In order to increase insight in the significance of the parameter temperature, a research program is started at Eindhoven University

of Technology. In the preliminary experiments with double lap shear tests it was found that the capacity is not much influenced as long as the glass transition temperature is not reached. At 50°C com

stress distribution due to a lower stiffness of the ad-hthe glass transition temperature of the adhesive a significant decrease in anchorage capacity is found, while the failure mode was changed from failure in the concrete to failure in the contact plane between concrete and adhesive.

Further research on the influence of temperature is recommended in which also the effect on other debonding failure mechanisms, which can be distin-guished for FRP strengthened concrete structures, should be investigated.

8 ACKNOWLEDGEMENTS

The authors are indebted to the Dutch Ministry of Transport, Public Works and Water Management that supported this investigation and SIKA Neder-land BV for supplying th

REFERENCES

Blontrock, H. 2003. Analysis and modeling of the fire resis-tance of concrete elemen

CUR B35 1990. CUR Aanbeveling 2bond strength of mortars on con

Di Tommaso, A., Neubauer, U., Pantuso, A., & Rostásy, F.S. 2001. Behavior of adhesively bonded concrete-CFRP joints at low and high temperatures. Mechanics of Composite Ma-terials 37(4): 327-338.

fib Task Group 9.3. 2001. fib Bulletin 14. Externally b

Holzenkämpfer, P. geklebter BeweTechnische Universität Braunschweig (in German).

Neubauer, U. & Rostásy, F.S. 1999. Bond Failure of Concrete Fiber Reinforced Polymer Plates at Inclined Cracks – Ex-periments and Fracture Mechanics Model. In Dolan C.W., Rizkalla, S.H., & Nanni, A. (ed.), Proceedings of the Fourth International Symposium, Non-Metallic (FRP) Re-inforcement for Concrete Structures, Baltimore, 1999. 369-382. Baltimore: ACI International. eu, A.J.B. & Branco, F.J.F.G. 2000. Shear tests of steel plates epoxy-bonded to concrete under temperature. Jour-nal of Materials in Civil Engineering 12(1): 74-80.