International Journal of Adhesion & Adhesives 21 (2001) 495502

The transfer of stress through a steel to concrete adhesive bond

Richard Andrew Barnesa,*, Georey Charles Maysb
a

Engineering Systems Department, Craneld University, Royal Military College of Science, Shivenham, Swindon, Wilts SN6 8LA, UK b Civil Engineering, Craneld University, Royal Military College of Science, Shivenham, Swindon, Wilts SN6 8LA, UK Accepted 20 March 2001

Abstract This paper describes part of a programme of research aimed at investigating the potential for strengthening reinforced concrete beams in shear by means of externally bonded steel plates. This may be a useful strengthening technique following the assessment of older bridge and building structures designed to outdated codes of practice. In order to produce a design guide for such shear plate bonding, a method for determining the anchorage length needs to be devised. By measuring the strain distribution in a steel plate adhesively bonded to a concrete block, the shear stress distribution within the adhesive and the eective anchorage length can be determined. A series of 15 experimental tests have been conducted to investigate the transfer of stress through a steel-concrete adhesive bond. The experimental programme was supported by theoretical and nite element analysis. The shear stress in a steelconcrete adhesive bond was found to be distributed exponentially, peaking at the loaded end of the specimen. For the specimens used, the stress distribution was distributed over a length of up to 155 mm for serviceability loads. r 2001 Elsevier Science Ltd. All rights reserved.
Keywords: B. Concrete; B. Steel; D. Stress distribution

1. Introduction Existing concrete structures may require strengthening or stiening in order to increase their ultimate exural or shear capacity, or to control deections and cracking. One method for providing this enhanced capacity is to adhesively bond steel plates to the concrete surface. The technique of using externally bonded steel plates has been used worldwide for over 30 years and, since 1975, in the UK. Advantages of external reinforcement over other methods include minimum eect on headroom, low cost, ease of maintenance and the ability to strengthen part of the structure whilst still in use. Since the early 1990s, a notable expansion of interest in the technique by clients, consultants and contractors has occurred, partly as a result of the EC 40 tonne directive for highway bridges. To date most of the plate bonding applications in both buildings and bridges have been to enhance the exural capacity of the beam by bonding plates to the tension face. This is a well documented method and
*Corresponding author. Tel.: +44-1793-785398; fax: +44-01793783192. E-mail address: r.a.barnes@rmcs.craneld.ac.uk (R.A. Barnes).

guidance for such a scheme is available [13]. However, some structures may well be decient in shear reinforcement due to the inadequacy of shear design methods in older design codes of practice, e.g. CP 114 [4]. Some plate bonding strengthening schemes have already included steel plates bonded to the side faces of the beam in order to enhance their shear capacity, an empirical approach being taken for the design of the steel plates [5]. Although some initial work has been conducted on strengthening for shear [6], this is a relatively new technique and there is no guidance available for the design or performance of such schemes. In order to produce a design guide for shear plate bonding a method for determining the anchorage length of the bonded plate needs to be devised. Fig. 1 shows a plated concrete beam under shear loading. The maximum tensile stresses in the bonded plate occur at right angles to the line taken by a typical shear crack. This situation can be reproduced approximately in the experimental specimen shown in Fig. 2. By measuring the strain distribution in the plate along the length of the specimen, the shear stress distribution within the adhesive and the eective anchorage length can be determined.

0143-7496/01/$ - see front matter r 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 3 - 7 4 9 6 ( 0 1 ) 0 0 0 3 1 - 8

496

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502

Fig. 3. Experimental specimen.

Fig. 1. Plated concrete beam under shear loading.

where l =bonded length, t=shear stress, b=width, P=load, x=distance, o=a constant.   G 1 1 2 ; 2 o a Es t Ec h where G=shear modulus for adhesive, a=thickness of adhesive layer, Es =modulus of elasticity for steel, Ec =modulus of elasticity for concrete, t=thickness of steel plate, h=thickness of concrete. More recently, several other authors [911] have used similar linear elastic analysis to demonstrate the exponential decay of deformations and stresses in adhesively bonded plate-to-concrete lap shear specimens. In this paper, nite element analysis is additionally employed to provide comparison with experimental data.

Fig. 2. Section through AA plated beam.

This is a somewhat simplistic approach as stresses derived from exural behaviour will also be present in the beam but not represented in the steel-concrete adhesively bonded specimen. Also, there may well be more than one shear crack present in the plated beam. However, the experimental specimen should be sucient to provide a representation of the anchorage required and typical strain distribution.

3. Experimental procedure The manufacture and testing of 15 steel-concrete adhesively bonded specimens was conducted to produce a series of stress transfer curves. The main objectives were to:
*

2. Previous work
*

Prior to using externally bonded steel plates for strengthening of concrete members, Bresson [7] studied the distribution of shear stress in a steel-adhesiveconcrete bond. Two pieces of steel were bonded either side of a concrete block (280 mm 70 mm 70 mm) and pulled in tension with the concrete block restrained, (similar to the specimens used by the author and shown in Fig. 3), thus generating a shear force in the adhesive layer. The failure was observed to occur in the concrete. In a theoretical study of the shear stress distribution the following relationship was derived, based upon the work by Volkersen [8]: tx o Pl coshox ; b sinhol 1

derive a simple means by which the shear stress transfer through a bonded steel plate over a shear crack in a reinforced concrete beam can be studied, investigate the eect of plate and adhesive thickness on the shear stress distribution, examine the distribution of strain and shear stress with respect to both magnitude and anchorage length.

Bonding plates either side of a concrete block produces a symmetrical experimental specimen that is easier to load and provides two steel plates for the measurement of strain. From the review of the literature an anchorage length between 100 and 200 mm was expected. Therefore, a bonded length of 280 mm was chosen. A 70 mm 70 mm concrete block had been successfully used in earlier work by Bresson [7] and so this cross section was adopted. Steel-concrete specimens with the dimensions shown in Fig. 3 were manufactured with electrical resistance

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502

497

also being used, their physical properties are broadly similar to that of Sikadur 31 PBA. The modulus of such adhesives needs to be such that ecient stress transfer is achieved through the adhesive from the steel plate to the concrete substrate. In order to characterise the adhesive for analytical and design purposes, the mechanical properties of the adhesive were determined by testing. To determine the exural modulus of elasticity, three adhesive prisms 200 mm long by 12 mm deep by 25 mm wide were tested in four point bending. Each sample was loaded at the third points at a rate of 1 mm/min and the central deection recorded. From the load deection curve the secant modulus at 0.2% strain was calculated. The mean value of exural modulus of elasticity determined in this manner was 7470 N/mm2. The shear modulus was calculated from the compressive elastic modulus and the Poissons ratio of the adhesive using the equation: G E =21 n; 3

Fig. 4. Diagram of specimen in loading jig.

strain gauges applied to the bonded face and the external face of the steel. Pressure was applied using mechanical clamps and the adhesive allowed curing for a minimum period of 7 days at 201C prior to test. The specimens were loaded at a rate of 1 mm/min using a tensile testing machine incorporating a purpose built loading jig in order to ensure correct alignment (Fig. 4). The loading pad was carefully positioned on the concrete to avoid direct loading through the adhesive. Strain readings were taken at 5 kN intervals.

where G=shear modulus (N/mm2), E =elastic modulus (N/mm2), n=Poissons ratio. Three rectangular prisms of dimensions 40 40 160 mm were tested in compression to BS 6319 : Part 6 [14] with compressive strain being recorded on two faces. The other two faces were used to record lateral expansion, in order to determine the Poissons ratio. The mean values of compressive modulus, Poissons ratio and shear modulus determined in this manner were 7270 N/mm2, 0.268, and 2870 N/mm2. The shear stress vs. normal stress failure envelope for the adhesive was determined according to BS EN 12188 [15] and is shown in Fig. 5. The mean tensile pull-o strength determined in this process was 29.0 N/mm2.

4. Material characteristics The concrete was a type CM (0.45) mix as specied in EN 104-801-1 Reference concretes for testing [12] allowed curing for 28 days and then gritblasted. Mild steel plates of three dierent thicknesses (3, 5 & 6 mm) were gritblasted to Swedish Standard Sa 2 1 2 [13] immediately prior to bonding. A two component, room temperature cure epoxy resin adhesive, Sikadur 31 PBA (Plate Bonding Adhesive) was used to bond the steel plate to the concrete. Historically this has been the adhesive predominantly used for plate bonding works throughout the UK. Although other manufacturers adhesives are now

Fig. 5. Failure envelope for the adhesive.

498

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502

5. Results 5.1. Failure loads Table 1 shows the failure loads of the steel-concrete shear specimens. Most of the specimens which failed did so through a layer of concrete near to the adhesive, demonstrating that the adhesive is stronger in shear than the concrete. This would suggest that the ultimate shear strength of the concrete is the upper limit of the joint strength rather than the shear strength of the adhesive. These failure loads show a consistent strength increase with increasing plate thickness. The thicker adhesive layers between 3 and 5 mm produced a stronger specimen than the 1 mm adhesive layer, although there is little dierence between the specimens with 3 and 5 mm thick layers. In both cases the variance at the 5% level of signicance between the samples (i.e. adhesive thickness) is greater than that within the samples. 5.2. Distribution of shear stress The strain gauges measured tensile strain in the steel plate, which was transmitted as a shear stress through the adhesive into the concrete substrate. To calculate the adhesive shear stresses from the measured strain readings a linear method was used. tx Es ts De=Dx; 4

The strains from the two plates were averaged to give an average strain distribution for each sample and from this the shear stress distribution was derived. It should be noted that although the rst strain gauge was at 10 mm the rst point at which stress levels could be calculated was between the rst and second gauge (at 25 mm), thus peak stress levels would be much higher than those shown. In all cases the shear stress increases with increasing load and in most of the specimens the stress is distributed exponentially, increasing towards the loaded end of the specimen. 5.3. Theoretical analysis Early work by Bresson [7], reviewed above, used a technique developed by Volkersen [8] to predict the shear stress distribution throughout the joint (Eq. (1)). This same analytical method was applied for the samples tested. The following assumptions are made:
*

* *

the materials are homogenous, isotropic and linear elastic. the adhesive is only exposed to shear forces. the thickness of the materials is constant throughout the bondline. the width of the steel plate is constant along the bondline.

where tx=shear stress at point x; ts =steel thickness, Es =modulus of steel, De=change in strain, Dx=distance between gauges.
Table 1 Failure loads No. Plate thickness Adhesive thickness Failure load Failure modea (mm) (mm) (kN) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
a

Due to the material assumptions made, the analysis is only applicable at loads below that which cause concrete cracking. 5.4. Finite element analysis Finite element analysis of the steel-concrete bonded specimens was undertaken, the main objectives were to
*

3 3 3 5 5 5 5 5 5 5 5 5 6 6 6

1 1 1 1 1 1 3 3 3 5 5 5 1 1 1

65 80 80 80 80 80 90 100 100 85 100b 100 100b 100b 95

C1 C1 C 1.5 C1 C2 C2 C 1.5 C2 C2 A1 F A1 F F C2

validate the use of nite element analysis in such models, conrm the trends derived from the experimental test programme, gain a greater understanding of the stress transfer process in the concrete-adhesive-steel bonded joint.

C=failed in concrete. A=failed at adhesive interface. 1=failed in one side. 1.5=main failure in one plate, partial failure in other. 2=failed in both plates. b Reached capacity of testing apparatus without failing.

The ABAQUS Standard [16] nite element analysis code was used to model the steel-concrete specimen. ABAQUS oers the facility for non-linear modelling of both plain and reinforced concrete. A two-dimensional model was used, two-dimensional plane stress elements with four nodes being used to dene the concrete, adhesive and steel plate. The aspect ratio of the elements was not greater than four. Although the symmetry of the experimental specimen could be utilised to produce a smaller model, greater stability was achieved by modelling the entire 2D aspect of the specimen as shown in Fig. 6.

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502 Table 2 Model variables Model S1 S2 S3 S4 S5 Plate thickness (mm) 3 5 6 5 5

499

Adhesive thickness (mm) 1 1 1 3 5

Fig. 6. Two-dimensional model used for analysis of bonded steelconcrete specimen.

Load was applied as a uniform pressure on the steel loading block at the end of the concrete with the steel plate ends being restrained in the x and y direction. No other restraints were imposed upon the model. The ABAQUS software allows for non-linear representation of concrete and the stress strain relationship used was that recommended in BS8110 Pt 2 [17]. The adhesive was modelled as a linear elastic material using the values determined experimentally for modulus of elasticity, shear modulus, Poissons ratio and tensile strength. Further details of the FE model are reported elsewhere [18]. In order to apply load to the specimen and allow concrete crushing to develop, an incremental static loading option was chosen. This automatically applied increasing load to the model and was used to load the model up to a load below 55 kN. The loading process was then halted and a single load increment up to 55 kN applied to the cracked model. This allowed for comparison between models at the same load level. A number of models were analysed, as summarised in Table 2. This allowed the eect of changing the variables plate thickness (t) and adhesive thickness (a), to be assessed. Behaviour of the model, in terms of displacement was considered realistic, the deformations corresponding to expected behaviour. Within this model increases in plate and adhesive thickness led to decreases in the tensile stress within the concrete. As failure occurs by tensile shear cracking within the concrete increasing these variables would lead to an increased capacity (within the limits of plate and adhesive thicknesses studied). No attempt was made to predict the ultimate load capacity of any of the specimens. The modelling of this aspect of the specimens behaviour is dependent upon the concrete cracking behaviour of the material model. Examination of the cracking behaviour of the concrete suggests that the material model for concrete is not ideally suited for use in shear. The material model for concrete that is used by ABAQUS Standard provides an elasticplastic damage theory, including tension cracking, compression crushing, concrete rebar interaction and postcrack response using damaged elasticity concepts. The more advanced FE program, ABAQUS Explicit, which is tailored for highly non-linear transient dynamic phenomena, uses a dierent concrete material model. The main ingredients

of its elastic-cracking model for concrete are: a strain rate decomposition into elastic (concrete) and cracking strain rates; elasticity; a set of cracking conditions; and a cracking relation (the evolution law for cracking behaviour). This brittle cracking model is deemed suitable for applications in which the concrete behaviour is dominated by tensile cracking and this may be better suited to studying the ultimate limit state behaviour of the specimens failing in shear. 5.5. Comparison with nite element and Volkersen [8] analysis Both the nite element analysis and Volkersen methods were compared simultaneously with the experimental results for shear stress, comparisons being made at a load level of 55 kN. With the exception of the experimental specimen no. 1 (failure load=65 kN) this load is less than 70% of the ultimate capacity of the steel-concrete adhesively bonded specimens. Therefore, comparisons were made prior to widespread cracking occurring in the concrete. This is signicant, because of the material assumptions made the Volkersen analysis is only applicable at loads below that which cause concrete cracking. At the loaded end of the specimens some cracking may have occurred in the FE analysis, therefore good correlation would not necessarily be expected in this region. Referring to Fig. 7, the Volkersen analysis predicts a shorter anchorage length (70 mm) than either the FE model (105 mm) or the experimental specimens 1 & 2 (155 mm) (t 3 mm, a 1 mm). Here, the anchorage length has been dened as the distance from the loaded ends to the point where the shear stress drops to less than 0.1 N/mm2. The shear stress distribution for the FE analysis corresponds closely to that measured experimentally beyond 60 mm from the loaded end. Closer to the loaded end, however, the correlation is less good. This could well be due to concrete cracking at the loaded ends of the experimental specimens. Similar trends were observed for specimens 46 (t 5 mm, a=1 mm) and 1315 (t 6 mm, a=1 mm). With the adhesive thickness increased to 3 mm (t 5 mm, a 3 mm) both the Volkersen and the FE methods predict a similar anchorage length (120 and 130 mm, respectively) to specimens 7 & 9 (115 mm) as

500

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502

Fig. 7. Shear stress distribution in specimens 1&2 at 55 kN load compared with FE and theoretical analysis.

Fig. 9. Shear stress distribution in specimens 1012 at 55 kN load compared with FE and theoretical analysis.

Fig. 8. Shear stress distribution in specimens 79 at 55 kN load compared with FE and theoretical analysis.

Fig. 10. The eect of adhesive thickness on shear stress (t 5; load=55 kN, a 1; 3 & 5 mm, line thickness increasing with adhesive thickness).

shown in Fig. 8. The shear stress distribution for the Volkersen and FE methods correspond closely but the predicted levels are below those measured experimentally. At an adhesive thickness of 5 mm (t 5 mm, a 5 mm) the theoretical and FE predictions correspond closely with the experimental results, both in terms of anchorage length (145, 155 and 150 mm, respectively) and shear stress levels (see Fig. 9). Examining the eect of adhesive thickness on shear stress distribution in Fig. 10 reveals a reduction in peak shear stress with increasing adhesive thickness for each of the experimental results, the FE analysis and the Volkersen analysis. Both analytical techniques predict an increase in anchorage length with increasing adhesive thickness. However, it was more dicult to detect such a

trend with experimental specimens due to some variabilities in the results. When the eect of plate thickness was examined, a reduction in peak shear stress was also found to occur with increasing plate thickness for all three methods. However, there appears to be no discernable eect on anchorage length. Some of the graphs show peak shear stresses deduced from the experimental specimens well in excess of 10 N/ mm2 which is considerably higher than typically quoted shear strength values for concrete (up to 5 N/mm2). However, these quoted values are mean shear strengths and it can be seen from the graphs that the mean shear strength of the concrete in the experimental specimens was considerably lower than the peak values.

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502

501

epoxy adhesive. By measuring the strain distribution in the steel plates when subjected to tensile load, the shear stress distributed within the adhesive and the eective anchorage length could be determined. The following conclusions were drawn from the results:
*

The shear stress in a steel-to-concrete adhesive joint is distributed exponentially, peaking at the loaded end of the specimen. For the specimen congurations used, the strain was distributed over a 130 mm anchorage length. Increases in either plate or adhesive thicknesses led to a general reduction in peak stress levels and an increase in total bond capacity.

Fig. 11. Comparison of strain distribution.

5.6. Comparison with strain data obtained from reinforced concrete beams To test the validity of the design approach, the strain distributions obtained from the experimental programme described above were compared with those obtained from externally plated reinforced concrete beams. The testing of 30 such beams is described elsewhere [18,19] but Fig. 11 shows the strain distribution perpendicular to an assumed shear crack in one beam immediately prior to failure. Superimposed are the strain distributions for the steel-concrete adhesively bonded specimens 1 & 2 (t 3 mm, a 1 mm) at a load of approximately 75% of ultimate. Although, a direct comparison of absolute strains is not possible, it can be seen that an anchorage length of approximately 150 mm occurs in both cases. The divergence in the slope of the strain distributions is believed to be due to two possible factors. Firstly, there was more than one shear crack present in the beam and secondly, the strain distribution in the beam will be aected by exural behaviour. Taking into account these factors, the steel-concrete adhesively bonded specimens nevertheless provide a simple means for investigating the shear stress transfer mechanism for design purposes.

Two-dimensional non-linear nite element analysis of these bonded steel-to-concrete specimens produced a realistic model upon which to base design, both in terms of stress levels and anchorage lengths. With an adhesive thickness of 1 mm the shear stress distribution derived from the nite element analysis more closely resembled the experimental results than a theoretical analysis based upon the method of Volkersen [8]. However, with the thicker adhesive layers (3 & 5 mm) both the Volkersen [8] and the nite element analysis provided similar shear stress distributions. Both analysis methods are only applicable at lower load levels, where signicant concrete cracking is not present. The steel-concrete adhesively bonded specimens provide a simple means for investigating shear stress transfer and their results are transferable to strengthened reinforced concrete beams in shear for design purposes.

References
[1] Concrete Society, Design guidance for strengthening concrete structures using bre composite materials (Technical Report 55). Crowthorne, UK: Concrete Society, 2000. [2] Mays GC. The use of bonded external reinforcement in bridge strengthening: structural requirements of the adhesive. Bridge management 2. London: Thomas Telford, 1993. p. 67280. [3] Department of Transport. BA 30/94 Design manual for roads and bridges, volume 3 Highway structures: Inspection and maintenance, Section 3, Repair and strengthening, part 1, Strengthening of concrete structures using externally bonded plates. London: HMSO, 1994. [4] British Standards Institute (BSI). The structural use of concrete in buildings, CP 114: 1957 (amended 1965). London: BSI, 1957. [5] Mays GC, Hutchinson AR. Adhesives in civil engineering. London: Cambridge University Press, 1992. [6] Mays GC, Peh LC. Strengthening of reinforced concrete beams in shear. Proceedings of the Conference on: Structural Faults and Repair. London, 1989. p. 139148. [7] Bresson J. Nouvelle recherches et applications concernant ! ton plaque ! . 1971. 1utilisation des collages dans les structures. Be Annales de 1Institut Technique due Batiment et des Travaux Publics. Suppl. No. 278, Paris, 1971. p. 2354.

6. Conclusions A study of the transfer of stress through a steel-toconcrete adhesive bond was undertaken in specimens which comprised a concrete block with steel plates bonded to two opposite faces using a two part structural

502

R.A. Barnes, G.C. Mays / International Journal of Adhesion & Adhesives 21 (2001) 495502 determination of modulus of elasticity in compression. London: BSI, 1984. British Standards Institute (BSI) BS EN 12188. Determination of adhesion steel-to-steel for characterisation of structural bonding agents. London: BSI, 1999. ABAQUS. ABAQUS Users manuals, Version 5.6. New York: Hibbittt, Karlsson and Sorensen, 1993. British Standards Institute (BSI). Structural use of concrete, part 2: code of practice for design and construction. BS 8110, part 2. London: BSI, 1985. Barnes RA. Strengthening of reinforced concrete beams in shear by the use of externally bonded steel plates. Ph.D. thesis, 2000. Craneld University (Royal Military College of Science). Barnes RA, Mays GC. The shear strengthening of rectangular reinforced concrete beams using bonded external reinforcement. Proceedings of the Symposium on Performance of Concrete Structures in the Arabian Gulf Environment. Dhahran, 1998.

[8] Volkersen O. Die Nietkraftverteilung in zugbean spruchten nietverbindungen mit konstanten laschenquerscnitten. Luftfahrforschung 1938;15:417. [9] Taljsten B. Dening anchor lengths of steel and CFRP plates bonded to concrete. Int J Adhesion Adhesives 1997;17:31927. [10] Bizindavyi L, Neale KW. Transfer lengths and bond strengths for composites bonded to concrete. ASCE J Compos Constr 1999;3(4):15360. [11] Lee YJ, Boothby TE, Bakis CE, Nanni A. Slip modulus of FRP sheets bonded to concrete. ASCE J Compos Constr 1999;3(4):1617. [12] British Standards Institute (BSI). EN 104-801-1 Reference concretes for testing. London: BSI, 1996. [13] Swedish Standards Institution. Pictorial surface preparation standards for painting steel surfaces. Stockholm: SIS (SIS 055900), 1967. [14] British Standards Institute (BSI) BS6319. Testing of resin compositions for use in construction part 6: method for

[15]

[16] [17]

[18]

[19]