- Email: [email protected]
Interfacial strength study between a concrete substrate and an innovative sprayed coating
Construction and Building Materials 79 (2015) 345–356
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Interfacial strength study between a concrete substrate and an innovative sprayed coating Mathieu Eymard a, Jean-Patrick Plassiard a,⇑, Pascal Perrotin a, Stéphane Le Fay b a b
Laboratoire LOCIE, Campus Scientifique, Savoie Technolac, F-73376 Le Bourget du Lac, France Société PAREXLANKO, 38 rue du Montmurier, 38070 Saint Quentin Fallavier, France
h i g h l i g h t s Light mortar based on a silica aerogel solution for thermal rehabilitation. Mechanical characterisation at the interface with its substrate. Influence of stress concentration on experimental results.
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 20 October 2014 Accepted 17 December 2014 Available online 28 January 2015 Keywords: Interface Strength characterisation Concrete Coating Refurbishment Slant-shear Experimental Numerical
a b s t r a c t This paper investigates the mechanical behaviour, at a local scale, of a solution of a thick thermal insulation pneumatically placed from the outside for refurbishment. In particular, this study shows the strength of the critical area: the interface with its concrete substrate. It is possible to define the failure criterion of this interface using the slant-shear test. However, due to a difference in rigidity between the concrete substrate and the insulating coating, a numerical study is necessary to better understand the experimental results and to assess the possible stress concentrations at this interface, which could distort the experimental results. Ó 2015 Published by Elsevier Ltd.
1. Introduction To meet the thermal refurbishment level required by the current standards in buildings, the addition of an external layer of insulation directly on the structure can be a useful solution in terms of thermal bridges and the thermal inertia of structures [10]. A recently patented insulating coating based on silica aerogels (super)-insulating materials has been developed. The invention is a light mortar (approximately 300 kg/m3) principally composed of a mineral binder and an insulating filler comprising granules of hydrophobic silica aerogel. This coating has a thermal conductivity of 0.027 W/(m K) [11]. If a direct surface contact between this layer and a structural wall is considered, a load transfer from the wall to the insulating ⇑ Corresponding author. E-mail addresses: [email protected] [email protected] (S. Le Fay). http://dx.doi.org/10.1016/j.conbuildmat.2014.12.031 0950-0618/Ó 2015 Published by Elsevier Ltd.
(J.-P.
Plassiard),
layer through the interface can be assumed. This load transfer could first provoke local failures in the coating, which could deteriorate its thermal insulating performance. Secondly, a large surface of coating can experience debonding during seismic solicitations such as during the earthquake in Lorca (Spain) on 11 May 2011 [1]. To better understand this complex behaviour, the mechanical properties of the materials as well as their interfacial strength must be known. During solicitations, this interface can be submitted to multiple stresses and therefore, the failure criterion must be determined to verify whether or not the interface can support them. To assess the failure criterion of the interface, it is necessary to determine a sufficient number of representative points of the behaviour law, which means soliciting the interface with different stress states. In the literature, many local-scale tests have been used to submit an interface to different stress states, in tension, shear, or a combination of normal and tangential stresses (Fig. 1). The patch test developed by [5] is also a very common and well-known bond test.
346
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
According to [16], each of these tests gives partial information on the overall mechanical behaviour of the interface. However, due to differences between the specimen’s geometry and/or the loading path, it can be difficult to directly compare these tests because the bond strength obtained is highly dependent on the tests chosen. For this study, the analysis is limited to two types of tests generally used. The uniaxial tension test, called the pull-off test (Fig. 1(a)), is not presented in this paper because the results could not be used: the core drilling needed prior to this test damaged the coating material and thus disturbed the results, particularly for the thermal insulating coating. The second test is the slant-shear test (Fig. 1(e)), which submits the interface to a combination of normal rn and tangential s stresses. By varying the interface angle, the interface can be submitted to different stress states. By performing each test until failure, the couples obtained (rn ; s) should let us characterise the bond strength of this interface and the shape of its governing criterion in the compression area.
2. Bibliographic study: the slant-shear test 2.1. Background and parameters This test was first presented by Kreigh [14] and used to measure the bond strength between two resinous base materials, [2,6]. However, it is also widely referenced in the literature to assess the bond strength between two cementitious base materials such as in [4,12,13,16,21,23] and in several international standards, such as [7]. It is generally used with a 30° bond angle a (Fig. 2) to test two different concretes or mortars in order to measure the shear strength of a structural repair or reinforcement. By applying a compression stress r0 to the specimen, it is possible to deduce the combination of average normal stress rn;moy (Eq. (1)) and shear stress smoy (Eq. (2)) at the interface. Thus, it is possible, according to [4], to represent the corresponding state of stress in a Mohr landmark. [4] has already used this
Fig. 1. Schematic description of different experimental tests to assess the bond strength between two materials, according to [8].
Fig. 2. Slant-shear test principle and state of stresses at the interface, deducted from the Mohr landmark, from [4].
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
347
Fig. 3. 2D mesh for finite element analysis and stress distributions obtained along the interface for the slant-shear test, according to [13].
test in tension and considered different interface angles: 0, 15, 30, 45 and 60°.
rn;moy ¼ smoy ¼
F max sinðaÞ Sint
F max cosðaÞ Sint
ð1Þ
ð2Þ
F max is the maximum load applied during the test and Sint the interface surface of the slant-shear specimen, calculated according to Eq. (3).
Sint ¼
p D2 4 sinðaÞ
ð3Þ
Here, D is the diameter of the cylindrical specimen. Two main failure modes can occur, adhesive and cohesive failure. Adhesive failure occurs when the failure plane is along the interface surface. Cohesive failure occurs when at least one of the materials reaches its compressive strength (or tensile strength, depending on the test). Many authors, for example [19], state that cohesive failure only gives a lower estimate value of the interfacial bond strength. Therefore, an adhesive failure is necessary to properly assess the mechanical behaviour of an interface. The modified slant-shear test proposed by [19] uses steel reinforcement to enforce adhesive failure. According to [17], the slant-shear test can give an estimation of the shear resistance under no normal stress if at least three different interface angles are considered. He noted, however, that the coefficient of variation of this method was about 23.5%. Several experimental parameters can have a significant influence on the specimen’s bond strength and its failure mode. They have been identified by [12,13,16,21,23]. These papers emphasise that the bond strength of the interface depends mainly on:
Substrate soundness. It is generally accepted that dust and other dirt can significantly deteriorate bond strength. Interface roughness. Many authors have studied the influence of this parameter in terms of strength and failure mode, and it is considered one of the most influential, for instance, by [3,4,9]. The surface can be prepared mechanically. Indeed, increasing roughness leads to a more efficient bond, but the use of heavy methods, such as hammering, can weaken the substrate’s surface and thus deteriorate the sustainability of the interface. It is generally more effective to use abrasive methods, such as hydrodemolition, sandblasting or wire brushing, for example. Increasing roughness can also lead to a change in failure mode, from the adhesive mode to the cohesive mode. According to [22,24], the most effective method is hydrodemolition because: (a) it does not introduce micro-cracking on the substrate; (b) it can be used to simultaneously remove concrete and clean reinforcement bars; (c) large surface areas can be prepared in a shorter time; (d) the prepared surfaces present a uniform roughness; and (e) it usually leads to the highest bond strength values. When water cannot be used, sandblasting and shotblasting are considered the most efficient techniques, according to [12]. It should be noted that [20] shows that wire-brushing is the technique with the lower coefficient of variation (21% at most), considering the correlation between roughness parameters and surface treatments, in comparison with a left as-cast surface (40%) and sandblasting (60%). The results of this study will be used to assess interface thickness for the numerical study presented in Section 4. However, the coefficient of variation of the experimental results shows quite close results between wire-brushing and sandblasting (8.90 and 8.56%, respectively, according to [12]; 19.36 and 18.40%, respectively, according to [21]). The left as-cast specimens seem to result in higher coefficient of variation and lower bond strength.
348
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
the steel–concrete bond connection through-out the pull-out test using this method.
Table 1 Test results on the mechanical properties of the coating materials. Coatings
Reference
rt (kPa)
rc (kPa)
E (MPa)
ISO
I.1
81.54
I.2
64.02
I.3
72.78
I.4
77.25
I.5
68.76
nc 122.38 95.35 123.58 106.62 109.37 124.72 119.78 nc 100.09
nc 3.95 5.30 3.86 3.77 3.77 3.92 4.44 nc 3.94
C.1
1320
C.2
1260
MGF
251.58 256.73 298.15 445.80
Table 2 Mechanical properties of tested materials: mean values.
*
Coatings
E (MPa)
rc (kPa)
rt (kPa)
ISO MGF
4 ± 1.4 300 ± 150
110 ± 30 2900*
70 ± 14 1289 ± 40
Estimated value from
rt and previous tests performed by the manufacturer.
The mechanical properties of the substrate and the added material, particularly the difference in Young’s modulus between the two materials. According to [4,21], this difference can be the cause of stress concentrations at the interface. The case studied in the literature is generally a concrete substrate with usual mechanical properties and an added concrete with higher mechanical properties. No information has been found to assess the behaviour between a concrete substrate and a thermal insulating coating with very low mechanical properties, in terms of Young’s modulus, compression and tension strength. The interface angle also has an influence on the failure mode. For any configuration, there is an angle a from which the failure mode changes from adhesive (Fig. 5) to cohesive. Indeed, the rn =s ratio increases with a, favouring the occurrence of a failure in compression mode in the weakest material, instead of a shear failure at the interface. Shrinkage of the new layer. According to [21], the slant-shear strength increases as the difference in age between the two different concretes increases. Through an experimental and numerical study, the authors stated that compressive loading eliminates the tension stresses at the interface. 2.2. Numerical analysis Finite element analysis on the slant-shear test has already been performed. It is used to evaluate the inherent stress concentrations when two materials with a different Young modulus are used. The studies reported in [4,13,19] used a 2D model (a 3D model was used by [21]) and analysed two different concretes, considered as perfect elastic in order to observe the stress distribution at the interface, thus showing the stress concentrations. Fig. 3 shows the distribution of normal and shear stresses along the interface, for a 30 MPa substrate concrete, tested with added concretes of 30, 50 and 100 MPa. This model shows that the stress concentrations seem to increase with the increase of the ratio of the two Young moduluses. As shown in the literature, adhesion between two different materials can be modelled using the finite element CAST3M software. Moreover, JOINT elements can be used to represent the mechanical behaviour of an interface. For example, [15] analysed
3. Experimental study 3.1. Mechanical properties of the materials Concrete is used as a substrate material here. Two added materials are tested. A topcoat called Monorex GF (MGF), considered as the reference, and the thermal insulating coating (ISO), an aerogelbased material whose detailed composition is not discussed in this paper. To meet their actual placing, these materials were added to the substrate by spraying. Three-point bending tests on 4 4 16 cm3 specimens allows one to assess the coatings’ tensile strengths by flexion rt , using Eq. (4). With compression tests on the remaining semi-specimens, the compressive strength rc (Eq. (5)), as well as the Young modulus E (Eq. (6)) of these materials can be assessed (Table 1). Shrinkage could not be measured due to the fragility of the ISO material.
rt ¼ rc ¼ E¼
1; 5 F t l b
3
Fc S
ðF B F A Þ l0 S ð DB DA Þ
ð4Þ
ð5Þ
ð6Þ
where F t is the maximum bending load, l the length between the two press holds, b the square section of the prism, F c the maximum compressive load and S the applied surface of compressive load. A ready-mixed concrete was used as support material. It consists of hydraulic binder and calibrated aggregates (0–10 mm). According to the manufacturer, its compressive strength is about 28 MPa. Mechanical tests were not performed on concrete, considering that its mechanical properties were sufficiently higher than those of the coatings, which can be considered as mechanically weak, in comparison. Obtaining a stable value for the Young modulus of the MGF coating was quite difficult, yielding only one order of magnitude. A 3 kN load cell was used for these tests; therefore the low mechanical strength values of the ISO coating can explain the significant disparity in the experimental results. Indeed, compressive failure occurred at 6% of the load cell capacity. (See Table 2). 3.2. Slant-shear test Considering the chosen configuration, the concrete substrate is located on the lower part of the specimen, and the added material on the upper part. The height of the specimen was chosen so as to allow for the variation of the interface bond angle (20, 25 and 30° from the vertical, determined to minimise the chances of cohesive
Fig. 4. Reference details of slant-shear specimens. Explanation and possible values.
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
349
Fig. 5. From left to right: 20, 25 and 30° slant-shear specimens during the test and after adhesive failure between the ISO coating and its concrete substrate.
Fig. 6. Force–displacement curves for the ISO coating with 20, 25 and 30° bond angles a.
failure). The specimen is a cylinder 20 cm diameter and 55 cm high. The repeatability was tested with four series of specimens for the ISO coating and two for the MGF. The interface bond angle was obtained with an accuracy of 0.5°. Once unmoulded, the concrete substrate was wire-brushed to create interface roughness. The interface was cleaned before spray-
ing the added materials, which was performed after 28 days of curing the substrate to ensure that most of the shrinkage had occurred. Tests were performed after a 28-day curing of the added materials. The compression force was applied with an electric actuator slaved in displacement (Dv ) and possessing a support ball joint (Fig. 5). The stroke speed of the actuator was kept constant at
350
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 7. Force–displacement curves for the MGF coating with 20, 25 and 30° bond angles a.
A speckel allows the surface to present a random pattern at the surface of each specimen.
Table 3 Experimental results for the slant-shear test. Tested materials
a
Reference
F max (kN)
rn;moy (kPa)
smoy (kPa)
MGF coating
20°
2.20.C.a 2.20.C.b 2.25.C.a 2.25.C.b 2.30.C.a 2.30.C.b
42.38 45.14 63.07 50.00 56.13 52.56
171.23 182.38 389.07 308.44 Cohesive 453.84
470.44 501.08 834.36 661.46 Failure 786.07
3.20.I.a 3.20.I.b 3.20.I.c 3.20.I.d 3.25.I.a 3.25.I.b 3.25.I.c 3.25.I.d 3.30.I.a 3.30.I.b 3.30.I.c 3.30.I.d
1.27 1.23 1.70 1.11 1.58 1.56 1.10 1.86 1.63 1.65 1.85 1.61
5.13 4.97 6.87 4.48 9.75 9.62 6.79 11.47 14.07 14.25 15.97 13.90
14.10 13.65 18.87 12.32 20.90 20.64 14.55 24.61 24.38 24.68 27.67 24.08
25° 30° ISO coating
20°
25°
30°
0.4 mm/min, resulting in the test lasting from 5 to 15 min, depending on the specimen. The specimen references are as described in Fig. 4. Image correlation using 7D software [25], developed by the SYMME laboratory in Annecy-le-Vieux, France, was used to assess strain and displacement values at the surface of the specimen throughout the test. Image correlation appeared in the early 1980s and increasingly used nowadays [18]. During solicitations, displacement fields can be measured at the surface of a specimen by comparing a deformed image to a reference image. The software analyses differences in grey levels at a defined pattern scale. As a result, the grayscale dynamic should be as high as possible.
3.3. Experimental results The tests conducted systematically resulted in adhesive failure (Fig. 5), except for one specimen, referenced 2.30.C.a and corresponding to an MGF-coated specimen. Hence, the choice of 20, 25 and 30° for the interface angles allowed us to use the results to assess the mechanical behaviour of the coating-concrete bond. Average normal and tangential stresses, respectively, rn;moy and smoy were calculated from slant-shear theory (Fig. 2) and Eqs. (1) and (2). Figs. 6 and 7 show the evolution of the compression force F applied to the specimens regarding the actuator displacement Dv for the ISO and the MGF coating, respectively. Good reproducibility of the experimental results can also be seen. It should be noted that, for the ISO material, specimens referenced c and d were also used for the second series of tests. As a consequence, traces of MGF materials remained on the concrete surface after cleaning. This can explain the experimental differences obtained between specimens a and b, which exhibit a force–displacement relationship, and specimens c and d, where more noticeable variations can be observed. Table 3 shows the average normal and tangential stresses at the interface for all tests. As expected, the ISO material shows lower bond strength values compared with the MGF material. Fig. 8 shows the comparison of the experimental criterion obtained that is considered as straight lines on the area studied. The coefficients of determination R2 are 0.93 and 0.92, respectively, for the ISO and MGF coating. A cohesion of 8 kPa was found for the ISO material and 270 kPa for the MGF coating. The friction angle / obtained for the two materials is close to 51°, but no conclusion has been validated yet concerning the matching of these friction angles.
351
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 8. Experimental criterion obtained with the slant-shear test, in red for the ISO coating, in blue for the MGF coating (left). Magnification of the experimental criterion obtained with the ISO coating (right). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 9. Results of the 7D software for specimen 3.20.I.a. From left to right: at the end of the elastic phase; at stress peak; just before failure.
Table 4 Parameters used to model the slant-shear test Materials
Type of parameters
Concrete (Elastic) Coatings (PDP) Interface (Coulomb)
Known
Estimate
Deducted
– Ecoat ; c
Econc ¼ 30 GPa – Eint ¼ Ecoat , eint ¼ 0:5 mm
– – /
rc;coat ; rt;coat
However, due to the Young modulus influence on the stress distribution at the interface, further research is necessary to obtain more precise information on the cohesion values c and friction angles values / at the interface. The 7D software calculates the distribution of shear strain on the surface considered using the principal deformations according to Eq. (7). Fig. 10. Example of force–displacement curve obtained for an unconfined compression test on an ISO coating specimen.
shear ¼
max min 2
ð7Þ
352
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 11. Two different views of the finite element mesh of the half slant-shear specimen, using the CAST3M software.
Table 5 Finite element analysis parameters, with CAST3M. Materials
Models
E (MPa)
rc (kPa)
rt (kPa)
Concrete ISO MGF
Elastic PDP PDP
30,000 5 150
– 110 1900–2500
– 70 1290
with max and min the maximum and minimum principal deformation, respectively. Fig. 9 shows the evolution of the shear strain distribution throughout the test. For the ISO material, it can be seen that during the elastic phase of the test, it is difficult to confirm the presence of stress concentrations at the interface. However, at the stress peak and just before failure, strain concentrations on the upper part of the interface and a progressive interface failure starting from the top can be noted. This suggests that stress concentrations are located in this area. By comparison, the MGF specimens show a more monolithic behaviour during the tests. Given that the MGF coated specimens exhibit a brittle failure, the results confirm that in the elastic phase, stress concentrations are not visible. However, these results indicate that the criterion parameters, considering a homogeneous distribution along the interface, could be not representative of the intrinsic resistance, particularly for the ISO coating.
the CAST3M software [26], in three dimensions and with tetrahedral elements. To consider an elasto-plastic behaviour for the coatings, particularly for the ISO coating (Fig. 10) and considering that its structure is close to concrete or mortar, a Perfect Drucker–Prager (PDP) criterion was used. Due to the level of solicitations, the concrete was considered strictly elastic. The interface was created using JOINT elements governed by a Coulomb criterion; therefore, the post-peak softening behaviour was not taken into account by the model. Only plastification can be observed. As a consequence, simulations were performed until the peak’s strain was reached because the following does not correspond to reality. The objective of this study was to determine the cohesion and friction angle parameters governing the behaviour law of the interface by calibrating the model. Table 4 groups all the necessary parameters of the different laws used in the model. It should be noted that a 0.2 Poisson coefficient was taken for the different materials. Considering the boundary conditions, the lower machine plate was considered by blocking any displacement of the specimen’s lower surface. In addition, to take the ball joint into account, the upper steel plate was created and its mesh was linked to the mesh of the specimen’s upper surface. As a consequence, the friction between the upper steel plate and the specimen was also considered. This configuration seems coherent with the image correlation results. First, a parametric analysis was conducted to assess the most appropriate mesh density, as well as the loading step. Thus, the mesh representing the half slant-shear specimen possesses 87,694 elements (Fig. 11) and the loading step considered is 107 m. This configuration leads to a relative error of 1.5%. This is consistent regarding the error related to the measurement of the load applied to the specimen during the test, which is about 12% for the ISO specimens, and 0.4% for the MGF specimens. The remaining parameters are, firstly, the Young modulus of the interface (Eint ) considered Eint ¼ Ecoat . Indeed, failure occurs at the interface but still on the coating side, and therefore the Young moduluses are considered identical. Secondly, eint ¼ 0:5 mm. In fact, studies conducted by [20] have facilitated the assessment of this interface thickness corresponding to a wire brushed surface preparation. Therefore, it is possible to calculate the interface normal K n and shear K s stiffnesses, from the following equations:
Kn ¼
Eint eint
ð8Þ
Ks ¼
Eint 2 ð1 þ mint Þ eint
ð9Þ
Table 5 displays the values of all the parameters considered. It can be noted that the mechanical properties of the ISO coating are very close to the experimental values. However, for the MGF coating, more significant variations were noted, especially concerning Young’s modulus, with a difference of 60%; however, this was still acceptable regarding the 52% standard deviation of the experimental value. According to Eqs. (10) and (11), with N and T the normal and tangential forces, respectively, the cohesion parameter is not dependent on the level of stress applied.
T¼ 4. Finite element analysis of the slant-shear test 4.1. Principle A numerical study was carried out to provide a better understanding of the stress distribution and its influence on the experimental results. This finite element analysis was performed using
Z
ðrn tanð/Þ þ cÞ ds
T ¼ N tanð/Þ þ c S
ð10Þ ð11Þ
Thus, the cohesion c is not influenced by the possible stress concentrations. As a consequence, the only parameter to be deduced is the friction angle /local of the interface, and the unicity of the solution is guaranteed. Moreover, considering the experimental curves, it seems that the loading peak does not correspond to a generalised
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 12. Comparison of average experimental (a and b) and numerical results for the specimens with a ¼ 25 and 30°, for the ISO coating.
Fig. 13. Comparison of average experimental and numerical results for the specimens with a ¼ 25 and 30°, for the MGF coating.
353
354
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Table 6 Parameters of the interface failure criteria. Experimental values and numerical corrections. Material
Method
c (kPa)
/ (°)
ISO
Experimental Numerical
8 8
51 44
MGF
Experimental Numerical
270 270
51 49–51
ð12Þ
where N is the normal load acting on the interface, /global the friction angle at the sample scale and S the surface of the interface. All of these values can be defined experimentally. It is also possible to express T with:
T¼
Z
s dS
ð13Þ
ð18Þ
Finally, this can be expressed as:
tanð/global Þ ¼ tanð/local Þ
yield of the interface, particularly for the specimen with ISO coating. When the peak is reached, the shear load T acting at the interface can be expressed as:
T ¼ N tanð/global Þ þ cglobal S
N tanð/global Þ ¼ N tan /local DT
DT N
ð19Þ
showing that /local is less than /global . For the MGF material, the model was calibrated using the experimental results of the three bond angles tested. For the ISO coating, the deviation between experimental and numerical results was very significant for the specimen with the 20° bond angle. However, a good match was observed for the 25 and 30° bond angles. These differences could not be explained but could be linked to the geometry of the 20° specimen and/or the differences in the Young modulus between the two materials. Further research would be needed to understand this phenomenon. Moreover, for the ISO coating, numerical analysis only compares the experimental results of the specimens referenced a and b for the reason discussed in Section 3.3. 4.2. Results
S
If the interface was totally plastified, then the corresponding shear load T plast could be given by:
Z
T plast ¼
rn tanð/local Þ þ clocal dS
ð14Þ
S
Since the yield does not occur on the whole interface when the peak is reached, it seems that T < T plast . Let DT be the difference between T and T plast . T can be expressed as:
T ¼ T plast DT
ð15Þ
Then Eq. (13) can be written as:
T¼
Z
rn tanð/local Þ þ clocal dS DT
ð16Þ
S
Integrating this equation gives:
T ¼ ðN tanð/local Þ þ clocal SÞ DT
ð17Þ
Which can be compared with Eq. (12). Remembering the assumption that clocal ¼ cglobal , simplifications exhibit the next relationship:
The results are presented in Fig. 12 for the ISO coating and in Fig. 13 for the MGF coating. According to Table 6, as expected, the cohesion parameter is identical experimentally and numerically. This strengthens the idea of a straight-line criterion, at least in the compression area. Moreover, the numerical friction angle (local) is lower than the experimental angle (global). Using image correlation the experimental and numerical studies were compared by analysing the displacement and strain values on the surface of the specimen throughout the test, to corroborate the numerical parameters deducted. This tool is first used to validate the qualitative behaviour of the model compared with the experimental observations. For example, the behaviour of the ball joint during the loading (Fig. 14). From a quantitative point of view, it is possible to compare the displacement values of the top surface, where r0 is applied. This confirms the Young modulus chosen for the coating considered. By observing the stress distribution at the interface, only a high concentration of normal and tangential stresses on the upper part is noted, indicating that only a small area of the interface is solicited. However, by calculating the fraction r, defined in Eq. (20), it is
Fig. 14. Vertical displacement for the surface of specimen 3.25.I.a, at failure (left). Comparison of the average top surface displacements measured by image correlation and obtained numerically, at failure (right).
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
355
to effectively assess the interface shear strength under zero normal stress if at least three different bond angles are considered. Secondly, the global friction angle of the criterion can be influenced by stress concentrations, but only in the case of significant differences between the mechanical properties of the two materials, particularly the Young modulus. Indeed, the ISO material possesses a Young modulus 6000 times lower than concrete, resulting in a 14% gap between the experimental and numerical friction angles. However, the two different approaches give similar results for the MGF coating, even if the Young modulus is 100–200 times lower than concrete. Analysing this on two different concretes to see whether the same tendency occurs would therefore yield valuable information. Further improvements could be made with a constitutive law for the interface which takes into account the loss of cohesion inherent to a local failure and which presents a postpeak softening behaviour. Considering the image correlation, a tetrahedral shape for the slant-shear specimens would be easier to operate. Several elements remain to be determined. Access to experimental data on the tensile behaviour of the interface would help complete the failure criterion. The impact of the substrate roughness on the cohesion and friction factors of the interface would provide additional, useful information. Finally, a higherscale study on walls submitted to in-plane shear stress is ongoing to assess the possible scale effects and to observe the failure mode in a more real solicitation. The interface would be solicited in shear that is within the range of the presented model. Acknowledgements
Fig. 15. The r fraction at the interface for the ISO coating and a ¼ 30 , respectively, for an actuator displacement of 0.5, 1.5, 3.0 (peak) and 3.5 mm corresponding to experimental failure.
The authors would like to thank Parexlanko for providing the necessary materials as well as the technical support of the LOCIE laboratory for the help given during the test campaigns. References
possible to quantify the proportion of the mobilized shear strength in comparison with the mobilizable share.
r¼
s rn tan / þ c
ð20Þ
This fraction varies between 0 (zero shear stress) and 1 (yield stress reached). Fig. 15 shows the evolution of the fraction r at the interface during a test with a 30° bond angle for the ISO coating. It also shows that the upper part of the interface reaches the criterion almost from the beginning of the loading and the rest of the interface reaches this criterion progressively throughout the test. These results confirm the heterogeneous distribution of stress for the slant-shear test. As a consequence, there is a progressive interface failure and not a fragile global failure, as assumed by the calculation of the average experimental stresses with Eqs. (1) and (2). By comparing the friction angles obtained experimentally and numerically, a notable difference can be seen for the ISO coating (44° instead of 51°, corresponding to a gap of almost 14%), whereas for the MGF coating, the gap varies between 0 and 4%. Thus, stress concentrations influence the friction parameters for the ISO configuration. However, for the MGF configuration, the gap can still be within the range of experimental disparities. 5. Conclusions and perspectives The adhesive failure mode obtained for the three different bond angles considered in this study allows us to assess the experimental criterion by slant-shear tests. Firstly, numerical modelling validated that stress concentrations do not influence the experimental cohesion factor of the interface criterion. Therefore, this test seems
[1] AFPS. Association française du génie parasismique – rapport de mission – le séisme de lorca du 11 mai 2011 (in French). 2011. [2] ASTM. Standard test method for bond strength of epoxy-resin systems used with concrete by slant shear. American Society for Testing and Materials, ASTM C882/C882M; 1983. [3] Austin S, Robins P, Pan Y. Tensile bond testing of concrete repairs. Mater Struct 1995;28:249–59. [4] Austin S, Robins P, Pan Y. Shear bond testing of concrete repairs. Cement Concr Res 1999;29:1067–76. [5] Austin S, Robins PJ. Development of patch test to study behaviour of shallow concrete patch repairs. Mag Concr Res 1993;45(164):221–9. [6] BSI. Testing of resin and polymer/cement compositions for use in construction. method for measurement of bond strength (slant-shear method). British Standard Institute, BS 6319-4; 1984. [7] BSI. Products and systems for the protection and repair of concrete structures. Test methods. Determination of slant shear strength. British Standard Institute. BS EN 12615; 1999. [8] Espeche AD, Leon J. Estimation of bond strength envelopes for old-to-new concrete based on a cylinder splitting test. Constr Build Mater 2011;25:1222–35. [9] Garbacz A, Gorka M, Courard L. Effect of concrete surface treatment on adhesion in repair systems. Mag Concr Res 2004;56:1–12. [10] Ibrahim M, Biwole PH, Wurtz E, Achard P. Limiting windows offset thermal bridge losses using a new insulating coating. Appl Energy 2014;123:220–31. [11] Ibrahim M, Biwole PH, Wurtz E, Achard P. A study on the thermal performance of exterior walls covered with a recently patented silica-aerogel-based insulating coating. Build Environ 2014;81:112–22. [12] Júlio ENBS, Branco FAB, Silva VD. Concrete-to-concrete bond strength. Influence of the roughness of the substrate surface. Constr Build Mater 2004;18:675–81. [13] Júlio ENBS, Branco FAB, Silva VD, Lourenco JF. Influence of added concrete compressive strength on adhesion to an existing concrete substrate. Build Environ 2006;41:1934–9. [14] Kreigh JD. Arizona slant shear test: a method to determine epoxy bond strength. J Am Concr Inst 1976;73(3):372–3. [15] Meaud C, Jurkiewiez B, Ferrier E. Steel-concrete bonding connection: an experimental study and non-linear finite element analysis. Int J Adhes Adhes 2014;54:131–42. [16] Momayez A, Ehsani MR, Ramezanianpour AA, Rajaie H. Comparison of methods for evaluating the bond strength between concrete substrate and repair materials. Cement Concr Res 2005;35:748–57.
356
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
[17] Naderi M. Analysis of the slant shear test. J Adhes Sci Tech 2009;23(2): 229–45. [18] Pan B, Qian K, Asundi A. Two-dimensional digital image correlation for inplane displacement and strain measurement: a review. Measure Sci Tech 2009;20. [19] Saldanha R, Júlio E, Dias-da Costa D, Santos P. A modified slant shear test designed to enforce adhesive failure. Constr Build Mater 2013;41:673–80. [20] Santos PMD, Júlio ENBS. Development of a laser roughness analyser to predict in situ the bond strength of concrete-to-concrete interfaces. Mag Concr Res 2008;60(5):329–37. [21] Santos PMD, Júlio ENBS. Factors affecting bond between new and old concrete. ACI Mater J 2011;108:449–56.
[22] Santos PMD, Júlio ENBS. A state-of-the-art review on roughness quantification methods for concrete surfaces. Constr Build Mater 2013;38:912–23. [23] Santos PMD, Júlio ENBS, Silva VD. Correlation between concrete-to-concrete bond strength and the roughness of the substrate surface. Constr Build Mater 2007;21:1688–95. [24] Talbot C, Pigeon M, Beaupré D, Morgan DR. Influence of surface preparation on long-term bonding of shotcrete. ACI Mater J 1994;91(6):560–6. [25] Vacher P, Dumoulin S, Morestin F, Mguil-Touchal S. Bidimensional strain measurement using digital images. Proc Inst Mech Eng 1999;213:811–7. [26] Verpeaux P, Charras T, Millard A. Castem 2000 une approche moderne du calcul des structures. Calcul des structures et intelligence artificielle 1988:261–71.
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Interfacial strength study between a concrete substrate and an innovative sprayed coating Mathieu Eymard a, Jean-Patrick Plassiard a,⇑, Pascal Perrotin a, Stéphane Le Fay b a b
Laboratoire LOCIE, Campus Scientifique, Savoie Technolac, F-73376 Le Bourget du Lac, France Société PAREXLANKO, 38 rue du Montmurier, 38070 Saint Quentin Fallavier, France
h i g h l i g h t s Light mortar based on a silica aerogel solution for thermal rehabilitation. Mechanical characterisation at the interface with its substrate. Influence of stress concentration on experimental results.
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 20 October 2014 Accepted 17 December 2014 Available online 28 January 2015 Keywords: Interface Strength characterisation Concrete Coating Refurbishment Slant-shear Experimental Numerical
a b s t r a c t This paper investigates the mechanical behaviour, at a local scale, of a solution of a thick thermal insulation pneumatically placed from the outside for refurbishment. In particular, this study shows the strength of the critical area: the interface with its concrete substrate. It is possible to define the failure criterion of this interface using the slant-shear test. However, due to a difference in rigidity between the concrete substrate and the insulating coating, a numerical study is necessary to better understand the experimental results and to assess the possible stress concentrations at this interface, which could distort the experimental results. Ó 2015 Published by Elsevier Ltd.
1. Introduction To meet the thermal refurbishment level required by the current standards in buildings, the addition of an external layer of insulation directly on the structure can be a useful solution in terms of thermal bridges and the thermal inertia of structures [10]. A recently patented insulating coating based on silica aerogels (super)-insulating materials has been developed. The invention is a light mortar (approximately 300 kg/m3) principally composed of a mineral binder and an insulating filler comprising granules of hydrophobic silica aerogel. This coating has a thermal conductivity of 0.027 W/(m K) [11]. If a direct surface contact between this layer and a structural wall is considered, a load transfer from the wall to the insulating ⇑ Corresponding author. E-mail addresses: [email protected] [email protected] (S. Le Fay). http://dx.doi.org/10.1016/j.conbuildmat.2014.12.031 0950-0618/Ó 2015 Published by Elsevier Ltd.
(J.-P.
Plassiard),
layer through the interface can be assumed. This load transfer could first provoke local failures in the coating, which could deteriorate its thermal insulating performance. Secondly, a large surface of coating can experience debonding during seismic solicitations such as during the earthquake in Lorca (Spain) on 11 May 2011 [1]. To better understand this complex behaviour, the mechanical properties of the materials as well as their interfacial strength must be known. During solicitations, this interface can be submitted to multiple stresses and therefore, the failure criterion must be determined to verify whether or not the interface can support them. To assess the failure criterion of the interface, it is necessary to determine a sufficient number of representative points of the behaviour law, which means soliciting the interface with different stress states. In the literature, many local-scale tests have been used to submit an interface to different stress states, in tension, shear, or a combination of normal and tangential stresses (Fig. 1). The patch test developed by [5] is also a very common and well-known bond test.
346
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
According to [16], each of these tests gives partial information on the overall mechanical behaviour of the interface. However, due to differences between the specimen’s geometry and/or the loading path, it can be difficult to directly compare these tests because the bond strength obtained is highly dependent on the tests chosen. For this study, the analysis is limited to two types of tests generally used. The uniaxial tension test, called the pull-off test (Fig. 1(a)), is not presented in this paper because the results could not be used: the core drilling needed prior to this test damaged the coating material and thus disturbed the results, particularly for the thermal insulating coating. The second test is the slant-shear test (Fig. 1(e)), which submits the interface to a combination of normal rn and tangential s stresses. By varying the interface angle, the interface can be submitted to different stress states. By performing each test until failure, the couples obtained (rn ; s) should let us characterise the bond strength of this interface and the shape of its governing criterion in the compression area.
2. Bibliographic study: the slant-shear test 2.1. Background and parameters This test was first presented by Kreigh [14] and used to measure the bond strength between two resinous base materials, [2,6]. However, it is also widely referenced in the literature to assess the bond strength between two cementitious base materials such as in [4,12,13,16,21,23] and in several international standards, such as [7]. It is generally used with a 30° bond angle a (Fig. 2) to test two different concretes or mortars in order to measure the shear strength of a structural repair or reinforcement. By applying a compression stress r0 to the specimen, it is possible to deduce the combination of average normal stress rn;moy (Eq. (1)) and shear stress smoy (Eq. (2)) at the interface. Thus, it is possible, according to [4], to represent the corresponding state of stress in a Mohr landmark. [4] has already used this
Fig. 1. Schematic description of different experimental tests to assess the bond strength between two materials, according to [8].
Fig. 2. Slant-shear test principle and state of stresses at the interface, deducted from the Mohr landmark, from [4].
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
347
Fig. 3. 2D mesh for finite element analysis and stress distributions obtained along the interface for the slant-shear test, according to [13].
test in tension and considered different interface angles: 0, 15, 30, 45 and 60°.
rn;moy ¼ smoy ¼
F max sinðaÞ Sint
F max cosðaÞ Sint
ð1Þ
ð2Þ
F max is the maximum load applied during the test and Sint the interface surface of the slant-shear specimen, calculated according to Eq. (3).
Sint ¼
p D2 4 sinðaÞ
ð3Þ
Here, D is the diameter of the cylindrical specimen. Two main failure modes can occur, adhesive and cohesive failure. Adhesive failure occurs when the failure plane is along the interface surface. Cohesive failure occurs when at least one of the materials reaches its compressive strength (or tensile strength, depending on the test). Many authors, for example [19], state that cohesive failure only gives a lower estimate value of the interfacial bond strength. Therefore, an adhesive failure is necessary to properly assess the mechanical behaviour of an interface. The modified slant-shear test proposed by [19] uses steel reinforcement to enforce adhesive failure. According to [17], the slant-shear test can give an estimation of the shear resistance under no normal stress if at least three different interface angles are considered. He noted, however, that the coefficient of variation of this method was about 23.5%. Several experimental parameters can have a significant influence on the specimen’s bond strength and its failure mode. They have been identified by [12,13,16,21,23]. These papers emphasise that the bond strength of the interface depends mainly on:
Substrate soundness. It is generally accepted that dust and other dirt can significantly deteriorate bond strength. Interface roughness. Many authors have studied the influence of this parameter in terms of strength and failure mode, and it is considered one of the most influential, for instance, by [3,4,9]. The surface can be prepared mechanically. Indeed, increasing roughness leads to a more efficient bond, but the use of heavy methods, such as hammering, can weaken the substrate’s surface and thus deteriorate the sustainability of the interface. It is generally more effective to use abrasive methods, such as hydrodemolition, sandblasting or wire brushing, for example. Increasing roughness can also lead to a change in failure mode, from the adhesive mode to the cohesive mode. According to [22,24], the most effective method is hydrodemolition because: (a) it does not introduce micro-cracking on the substrate; (b) it can be used to simultaneously remove concrete and clean reinforcement bars; (c) large surface areas can be prepared in a shorter time; (d) the prepared surfaces present a uniform roughness; and (e) it usually leads to the highest bond strength values. When water cannot be used, sandblasting and shotblasting are considered the most efficient techniques, according to [12]. It should be noted that [20] shows that wire-brushing is the technique with the lower coefficient of variation (21% at most), considering the correlation between roughness parameters and surface treatments, in comparison with a left as-cast surface (40%) and sandblasting (60%). The results of this study will be used to assess interface thickness for the numerical study presented in Section 4. However, the coefficient of variation of the experimental results shows quite close results between wire-brushing and sandblasting (8.90 and 8.56%, respectively, according to [12]; 19.36 and 18.40%, respectively, according to [21]). The left as-cast specimens seem to result in higher coefficient of variation and lower bond strength.
348
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
the steel–concrete bond connection through-out the pull-out test using this method.
Table 1 Test results on the mechanical properties of the coating materials. Coatings
Reference
rt (kPa)
rc (kPa)
E (MPa)
ISO
I.1
81.54
I.2
64.02
I.3
72.78
I.4
77.25
I.5
68.76
nc 122.38 95.35 123.58 106.62 109.37 124.72 119.78 nc 100.09
nc 3.95 5.30 3.86 3.77 3.77 3.92 4.44 nc 3.94
C.1
1320
C.2
1260
MGF
251.58 256.73 298.15 445.80
Table 2 Mechanical properties of tested materials: mean values.
*
Coatings
E (MPa)
rc (kPa)
rt (kPa)
ISO MGF
4 ± 1.4 300 ± 150
110 ± 30 2900*
70 ± 14 1289 ± 40
Estimated value from
rt and previous tests performed by the manufacturer.
The mechanical properties of the substrate and the added material, particularly the difference in Young’s modulus between the two materials. According to [4,21], this difference can be the cause of stress concentrations at the interface. The case studied in the literature is generally a concrete substrate with usual mechanical properties and an added concrete with higher mechanical properties. No information has been found to assess the behaviour between a concrete substrate and a thermal insulating coating with very low mechanical properties, in terms of Young’s modulus, compression and tension strength. The interface angle also has an influence on the failure mode. For any configuration, there is an angle a from which the failure mode changes from adhesive (Fig. 5) to cohesive. Indeed, the rn =s ratio increases with a, favouring the occurrence of a failure in compression mode in the weakest material, instead of a shear failure at the interface. Shrinkage of the new layer. According to [21], the slant-shear strength increases as the difference in age between the two different concretes increases. Through an experimental and numerical study, the authors stated that compressive loading eliminates the tension stresses at the interface. 2.2. Numerical analysis Finite element analysis on the slant-shear test has already been performed. It is used to evaluate the inherent stress concentrations when two materials with a different Young modulus are used. The studies reported in [4,13,19] used a 2D model (a 3D model was used by [21]) and analysed two different concretes, considered as perfect elastic in order to observe the stress distribution at the interface, thus showing the stress concentrations. Fig. 3 shows the distribution of normal and shear stresses along the interface, for a 30 MPa substrate concrete, tested with added concretes of 30, 50 and 100 MPa. This model shows that the stress concentrations seem to increase with the increase of the ratio of the two Young moduluses. As shown in the literature, adhesion between two different materials can be modelled using the finite element CAST3M software. Moreover, JOINT elements can be used to represent the mechanical behaviour of an interface. For example, [15] analysed
3. Experimental study 3.1. Mechanical properties of the materials Concrete is used as a substrate material here. Two added materials are tested. A topcoat called Monorex GF (MGF), considered as the reference, and the thermal insulating coating (ISO), an aerogelbased material whose detailed composition is not discussed in this paper. To meet their actual placing, these materials were added to the substrate by spraying. Three-point bending tests on 4 4 16 cm3 specimens allows one to assess the coatings’ tensile strengths by flexion rt , using Eq. (4). With compression tests on the remaining semi-specimens, the compressive strength rc (Eq. (5)), as well as the Young modulus E (Eq. (6)) of these materials can be assessed (Table 1). Shrinkage could not be measured due to the fragility of the ISO material.
rt ¼ rc ¼ E¼
1; 5 F t l b
3
Fc S
ðF B F A Þ l0 S ð DB DA Þ
ð4Þ
ð5Þ
ð6Þ
where F t is the maximum bending load, l the length between the two press holds, b the square section of the prism, F c the maximum compressive load and S the applied surface of compressive load. A ready-mixed concrete was used as support material. It consists of hydraulic binder and calibrated aggregates (0–10 mm). According to the manufacturer, its compressive strength is about 28 MPa. Mechanical tests were not performed on concrete, considering that its mechanical properties were sufficiently higher than those of the coatings, which can be considered as mechanically weak, in comparison. Obtaining a stable value for the Young modulus of the MGF coating was quite difficult, yielding only one order of magnitude. A 3 kN load cell was used for these tests; therefore the low mechanical strength values of the ISO coating can explain the significant disparity in the experimental results. Indeed, compressive failure occurred at 6% of the load cell capacity. (See Table 2). 3.2. Slant-shear test Considering the chosen configuration, the concrete substrate is located on the lower part of the specimen, and the added material on the upper part. The height of the specimen was chosen so as to allow for the variation of the interface bond angle (20, 25 and 30° from the vertical, determined to minimise the chances of cohesive
Fig. 4. Reference details of slant-shear specimens. Explanation and possible values.
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
349
Fig. 5. From left to right: 20, 25 and 30° slant-shear specimens during the test and after adhesive failure between the ISO coating and its concrete substrate.
Fig. 6. Force–displacement curves for the ISO coating with 20, 25 and 30° bond angles a.
failure). The specimen is a cylinder 20 cm diameter and 55 cm high. The repeatability was tested with four series of specimens for the ISO coating and two for the MGF. The interface bond angle was obtained with an accuracy of 0.5°. Once unmoulded, the concrete substrate was wire-brushed to create interface roughness. The interface was cleaned before spray-
ing the added materials, which was performed after 28 days of curing the substrate to ensure that most of the shrinkage had occurred. Tests were performed after a 28-day curing of the added materials. The compression force was applied with an electric actuator slaved in displacement (Dv ) and possessing a support ball joint (Fig. 5). The stroke speed of the actuator was kept constant at
350
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 7. Force–displacement curves for the MGF coating with 20, 25 and 30° bond angles a.
A speckel allows the surface to present a random pattern at the surface of each specimen.
Table 3 Experimental results for the slant-shear test. Tested materials
a
Reference
F max (kN)
rn;moy (kPa)
smoy (kPa)
MGF coating
20°
2.20.C.a 2.20.C.b 2.25.C.a 2.25.C.b 2.30.C.a 2.30.C.b
42.38 45.14 63.07 50.00 56.13 52.56
171.23 182.38 389.07 308.44 Cohesive 453.84
470.44 501.08 834.36 661.46 Failure 786.07
3.20.I.a 3.20.I.b 3.20.I.c 3.20.I.d 3.25.I.a 3.25.I.b 3.25.I.c 3.25.I.d 3.30.I.a 3.30.I.b 3.30.I.c 3.30.I.d
1.27 1.23 1.70 1.11 1.58 1.56 1.10 1.86 1.63 1.65 1.85 1.61
5.13 4.97 6.87 4.48 9.75 9.62 6.79 11.47 14.07 14.25 15.97 13.90
14.10 13.65 18.87 12.32 20.90 20.64 14.55 24.61 24.38 24.68 27.67 24.08
25° 30° ISO coating
20°
25°
30°
0.4 mm/min, resulting in the test lasting from 5 to 15 min, depending on the specimen. The specimen references are as described in Fig. 4. Image correlation using 7D software [25], developed by the SYMME laboratory in Annecy-le-Vieux, France, was used to assess strain and displacement values at the surface of the specimen throughout the test. Image correlation appeared in the early 1980s and increasingly used nowadays [18]. During solicitations, displacement fields can be measured at the surface of a specimen by comparing a deformed image to a reference image. The software analyses differences in grey levels at a defined pattern scale. As a result, the grayscale dynamic should be as high as possible.
3.3. Experimental results The tests conducted systematically resulted in adhesive failure (Fig. 5), except for one specimen, referenced 2.30.C.a and corresponding to an MGF-coated specimen. Hence, the choice of 20, 25 and 30° for the interface angles allowed us to use the results to assess the mechanical behaviour of the coating-concrete bond. Average normal and tangential stresses, respectively, rn;moy and smoy were calculated from slant-shear theory (Fig. 2) and Eqs. (1) and (2). Figs. 6 and 7 show the evolution of the compression force F applied to the specimens regarding the actuator displacement Dv for the ISO and the MGF coating, respectively. Good reproducibility of the experimental results can also be seen. It should be noted that, for the ISO material, specimens referenced c and d were also used for the second series of tests. As a consequence, traces of MGF materials remained on the concrete surface after cleaning. This can explain the experimental differences obtained between specimens a and b, which exhibit a force–displacement relationship, and specimens c and d, where more noticeable variations can be observed. Table 3 shows the average normal and tangential stresses at the interface for all tests. As expected, the ISO material shows lower bond strength values compared with the MGF material. Fig. 8 shows the comparison of the experimental criterion obtained that is considered as straight lines on the area studied. The coefficients of determination R2 are 0.93 and 0.92, respectively, for the ISO and MGF coating. A cohesion of 8 kPa was found for the ISO material and 270 kPa for the MGF coating. The friction angle / obtained for the two materials is close to 51°, but no conclusion has been validated yet concerning the matching of these friction angles.
351
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 8. Experimental criterion obtained with the slant-shear test, in red for the ISO coating, in blue for the MGF coating (left). Magnification of the experimental criterion obtained with the ISO coating (right). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 9. Results of the 7D software for specimen 3.20.I.a. From left to right: at the end of the elastic phase; at stress peak; just before failure.
Table 4 Parameters used to model the slant-shear test Materials
Type of parameters
Concrete (Elastic) Coatings (PDP) Interface (Coulomb)
Known
Estimate
Deducted
– Ecoat ; c
Econc ¼ 30 GPa – Eint ¼ Ecoat , eint ¼ 0:5 mm
– – /
rc;coat ; rt;coat
However, due to the Young modulus influence on the stress distribution at the interface, further research is necessary to obtain more precise information on the cohesion values c and friction angles values / at the interface. The 7D software calculates the distribution of shear strain on the surface considered using the principal deformations according to Eq. (7). Fig. 10. Example of force–displacement curve obtained for an unconfined compression test on an ISO coating specimen.
shear ¼
max min 2
ð7Þ
352
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 11. Two different views of the finite element mesh of the half slant-shear specimen, using the CAST3M software.
Table 5 Finite element analysis parameters, with CAST3M. Materials
Models
E (MPa)
rc (kPa)
rt (kPa)
Concrete ISO MGF
Elastic PDP PDP
30,000 5 150
– 110 1900–2500
– 70 1290
with max and min the maximum and minimum principal deformation, respectively. Fig. 9 shows the evolution of the shear strain distribution throughout the test. For the ISO material, it can be seen that during the elastic phase of the test, it is difficult to confirm the presence of stress concentrations at the interface. However, at the stress peak and just before failure, strain concentrations on the upper part of the interface and a progressive interface failure starting from the top can be noted. This suggests that stress concentrations are located in this area. By comparison, the MGF specimens show a more monolithic behaviour during the tests. Given that the MGF coated specimens exhibit a brittle failure, the results confirm that in the elastic phase, stress concentrations are not visible. However, these results indicate that the criterion parameters, considering a homogeneous distribution along the interface, could be not representative of the intrinsic resistance, particularly for the ISO coating.
the CAST3M software [26], in three dimensions and with tetrahedral elements. To consider an elasto-plastic behaviour for the coatings, particularly for the ISO coating (Fig. 10) and considering that its structure is close to concrete or mortar, a Perfect Drucker–Prager (PDP) criterion was used. Due to the level of solicitations, the concrete was considered strictly elastic. The interface was created using JOINT elements governed by a Coulomb criterion; therefore, the post-peak softening behaviour was not taken into account by the model. Only plastification can be observed. As a consequence, simulations were performed until the peak’s strain was reached because the following does not correspond to reality. The objective of this study was to determine the cohesion and friction angle parameters governing the behaviour law of the interface by calibrating the model. Table 4 groups all the necessary parameters of the different laws used in the model. It should be noted that a 0.2 Poisson coefficient was taken for the different materials. Considering the boundary conditions, the lower machine plate was considered by blocking any displacement of the specimen’s lower surface. In addition, to take the ball joint into account, the upper steel plate was created and its mesh was linked to the mesh of the specimen’s upper surface. As a consequence, the friction between the upper steel plate and the specimen was also considered. This configuration seems coherent with the image correlation results. First, a parametric analysis was conducted to assess the most appropriate mesh density, as well as the loading step. Thus, the mesh representing the half slant-shear specimen possesses 87,694 elements (Fig. 11) and the loading step considered is 107 m. This configuration leads to a relative error of 1.5%. This is consistent regarding the error related to the measurement of the load applied to the specimen during the test, which is about 12% for the ISO specimens, and 0.4% for the MGF specimens. The remaining parameters are, firstly, the Young modulus of the interface (Eint ) considered Eint ¼ Ecoat . Indeed, failure occurs at the interface but still on the coating side, and therefore the Young moduluses are considered identical. Secondly, eint ¼ 0:5 mm. In fact, studies conducted by [20] have facilitated the assessment of this interface thickness corresponding to a wire brushed surface preparation. Therefore, it is possible to calculate the interface normal K n and shear K s stiffnesses, from the following equations:
Kn ¼
Eint eint
ð8Þ
Ks ¼
Eint 2 ð1 þ mint Þ eint
ð9Þ
Table 5 displays the values of all the parameters considered. It can be noted that the mechanical properties of the ISO coating are very close to the experimental values. However, for the MGF coating, more significant variations were noted, especially concerning Young’s modulus, with a difference of 60%; however, this was still acceptable regarding the 52% standard deviation of the experimental value. According to Eqs. (10) and (11), with N and T the normal and tangential forces, respectively, the cohesion parameter is not dependent on the level of stress applied.
T¼ 4. Finite element analysis of the slant-shear test 4.1. Principle A numerical study was carried out to provide a better understanding of the stress distribution and its influence on the experimental results. This finite element analysis was performed using
Z
ðrn tanð/Þ þ cÞ ds
T ¼ N tanð/Þ þ c S
ð10Þ ð11Þ
Thus, the cohesion c is not influenced by the possible stress concentrations. As a consequence, the only parameter to be deduced is the friction angle /local of the interface, and the unicity of the solution is guaranteed. Moreover, considering the experimental curves, it seems that the loading peak does not correspond to a generalised
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Fig. 12. Comparison of average experimental (a and b) and numerical results for the specimens with a ¼ 25 and 30°, for the ISO coating.
Fig. 13. Comparison of average experimental and numerical results for the specimens with a ¼ 25 and 30°, for the MGF coating.
353
354
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
Table 6 Parameters of the interface failure criteria. Experimental values and numerical corrections. Material
Method
c (kPa)
/ (°)
ISO
Experimental Numerical
8 8
51 44
MGF
Experimental Numerical
270 270
51 49–51
ð12Þ
where N is the normal load acting on the interface, /global the friction angle at the sample scale and S the surface of the interface. All of these values can be defined experimentally. It is also possible to express T with:
T¼
Z
s dS
ð13Þ
ð18Þ
Finally, this can be expressed as:
tanð/global Þ ¼ tanð/local Þ
yield of the interface, particularly for the specimen with ISO coating. When the peak is reached, the shear load T acting at the interface can be expressed as:
T ¼ N tanð/global Þ þ cglobal S
N tanð/global Þ ¼ N tan /local DT
DT N
ð19Þ
showing that /local is less than /global . For the MGF material, the model was calibrated using the experimental results of the three bond angles tested. For the ISO coating, the deviation between experimental and numerical results was very significant for the specimen with the 20° bond angle. However, a good match was observed for the 25 and 30° bond angles. These differences could not be explained but could be linked to the geometry of the 20° specimen and/or the differences in the Young modulus between the two materials. Further research would be needed to understand this phenomenon. Moreover, for the ISO coating, numerical analysis only compares the experimental results of the specimens referenced a and b for the reason discussed in Section 3.3. 4.2. Results
S
If the interface was totally plastified, then the corresponding shear load T plast could be given by:
Z
T plast ¼
rn tanð/local Þ þ clocal dS
ð14Þ
S
Since the yield does not occur on the whole interface when the peak is reached, it seems that T < T plast . Let DT be the difference between T and T plast . T can be expressed as:
T ¼ T plast DT
ð15Þ
Then Eq. (13) can be written as:
T¼
Z
rn tanð/local Þ þ clocal dS DT
ð16Þ
S
Integrating this equation gives:
T ¼ ðN tanð/local Þ þ clocal SÞ DT
ð17Þ
Which can be compared with Eq. (12). Remembering the assumption that clocal ¼ cglobal , simplifications exhibit the next relationship:
The results are presented in Fig. 12 for the ISO coating and in Fig. 13 for the MGF coating. According to Table 6, as expected, the cohesion parameter is identical experimentally and numerically. This strengthens the idea of a straight-line criterion, at least in the compression area. Moreover, the numerical friction angle (local) is lower than the experimental angle (global). Using image correlation the experimental and numerical studies were compared by analysing the displacement and strain values on the surface of the specimen throughout the test, to corroborate the numerical parameters deducted. This tool is first used to validate the qualitative behaviour of the model compared with the experimental observations. For example, the behaviour of the ball joint during the loading (Fig. 14). From a quantitative point of view, it is possible to compare the displacement values of the top surface, where r0 is applied. This confirms the Young modulus chosen for the coating considered. By observing the stress distribution at the interface, only a high concentration of normal and tangential stresses on the upper part is noted, indicating that only a small area of the interface is solicited. However, by calculating the fraction r, defined in Eq. (20), it is
Fig. 14. Vertical displacement for the surface of specimen 3.25.I.a, at failure (left). Comparison of the average top surface displacements measured by image correlation and obtained numerically, at failure (right).
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
355
to effectively assess the interface shear strength under zero normal stress if at least three different bond angles are considered. Secondly, the global friction angle of the criterion can be influenced by stress concentrations, but only in the case of significant differences between the mechanical properties of the two materials, particularly the Young modulus. Indeed, the ISO material possesses a Young modulus 6000 times lower than concrete, resulting in a 14% gap between the experimental and numerical friction angles. However, the two different approaches give similar results for the MGF coating, even if the Young modulus is 100–200 times lower than concrete. Analysing this on two different concretes to see whether the same tendency occurs would therefore yield valuable information. Further improvements could be made with a constitutive law for the interface which takes into account the loss of cohesion inherent to a local failure and which presents a postpeak softening behaviour. Considering the image correlation, a tetrahedral shape for the slant-shear specimens would be easier to operate. Several elements remain to be determined. Access to experimental data on the tensile behaviour of the interface would help complete the failure criterion. The impact of the substrate roughness on the cohesion and friction factors of the interface would provide additional, useful information. Finally, a higherscale study on walls submitted to in-plane shear stress is ongoing to assess the possible scale effects and to observe the failure mode in a more real solicitation. The interface would be solicited in shear that is within the range of the presented model. Acknowledgements
Fig. 15. The r fraction at the interface for the ISO coating and a ¼ 30 , respectively, for an actuator displacement of 0.5, 1.5, 3.0 (peak) and 3.5 mm corresponding to experimental failure.
The authors would like to thank Parexlanko for providing the necessary materials as well as the technical support of the LOCIE laboratory for the help given during the test campaigns. References
possible to quantify the proportion of the mobilized shear strength in comparison with the mobilizable share.
r¼
s rn tan / þ c
ð20Þ
This fraction varies between 0 (zero shear stress) and 1 (yield stress reached). Fig. 15 shows the evolution of the fraction r at the interface during a test with a 30° bond angle for the ISO coating. It also shows that the upper part of the interface reaches the criterion almost from the beginning of the loading and the rest of the interface reaches this criterion progressively throughout the test. These results confirm the heterogeneous distribution of stress for the slant-shear test. As a consequence, there is a progressive interface failure and not a fragile global failure, as assumed by the calculation of the average experimental stresses with Eqs. (1) and (2). By comparing the friction angles obtained experimentally and numerically, a notable difference can be seen for the ISO coating (44° instead of 51°, corresponding to a gap of almost 14%), whereas for the MGF coating, the gap varies between 0 and 4%. Thus, stress concentrations influence the friction parameters for the ISO configuration. However, for the MGF configuration, the gap can still be within the range of experimental disparities. 5. Conclusions and perspectives The adhesive failure mode obtained for the three different bond angles considered in this study allows us to assess the experimental criterion by slant-shear tests. Firstly, numerical modelling validated that stress concentrations do not influence the experimental cohesion factor of the interface criterion. Therefore, this test seems
[1] AFPS. Association française du génie parasismique – rapport de mission – le séisme de lorca du 11 mai 2011 (in French). 2011. [2] ASTM. Standard test method for bond strength of epoxy-resin systems used with concrete by slant shear. American Society for Testing and Materials, ASTM C882/C882M; 1983. [3] Austin S, Robins P, Pan Y. Tensile bond testing of concrete repairs. Mater Struct 1995;28:249–59. [4] Austin S, Robins P, Pan Y. Shear bond testing of concrete repairs. Cement Concr Res 1999;29:1067–76. [5] Austin S, Robins PJ. Development of patch test to study behaviour of shallow concrete patch repairs. Mag Concr Res 1993;45(164):221–9. [6] BSI. Testing of resin and polymer/cement compositions for use in construction. method for measurement of bond strength (slant-shear method). British Standard Institute, BS 6319-4; 1984. [7] BSI. Products and systems for the protection and repair of concrete structures. Test methods. Determination of slant shear strength. British Standard Institute. BS EN 12615; 1999. [8] Espeche AD, Leon J. Estimation of bond strength envelopes for old-to-new concrete based on a cylinder splitting test. Constr Build Mater 2011;25:1222–35. [9] Garbacz A, Gorka M, Courard L. Effect of concrete surface treatment on adhesion in repair systems. Mag Concr Res 2004;56:1–12. [10] Ibrahim M, Biwole PH, Wurtz E, Achard P. Limiting windows offset thermal bridge losses using a new insulating coating. Appl Energy 2014;123:220–31. [11] Ibrahim M, Biwole PH, Wurtz E, Achard P. A study on the thermal performance of exterior walls covered with a recently patented silica-aerogel-based insulating coating. Build Environ 2014;81:112–22. [12] Júlio ENBS, Branco FAB, Silva VD. Concrete-to-concrete bond strength. Influence of the roughness of the substrate surface. Constr Build Mater 2004;18:675–81. [13] Júlio ENBS, Branco FAB, Silva VD, Lourenco JF. Influence of added concrete compressive strength on adhesion to an existing concrete substrate. Build Environ 2006;41:1934–9. [14] Kreigh JD. Arizona slant shear test: a method to determine epoxy bond strength. J Am Concr Inst 1976;73(3):372–3. [15] Meaud C, Jurkiewiez B, Ferrier E. Steel-concrete bonding connection: an experimental study and non-linear finite element analysis. Int J Adhes Adhes 2014;54:131–42. [16] Momayez A, Ehsani MR, Ramezanianpour AA, Rajaie H. Comparison of methods for evaluating the bond strength between concrete substrate and repair materials. Cement Concr Res 2005;35:748–57.
356
M. Eymard et al. / Construction and Building Materials 79 (2015) 345–356
[17] Naderi M. Analysis of the slant shear test. J Adhes Sci Tech 2009;23(2): 229–45. [18] Pan B, Qian K, Asundi A. Two-dimensional digital image correlation for inplane displacement and strain measurement: a review. Measure Sci Tech 2009;20. [19] Saldanha R, Júlio E, Dias-da Costa D, Santos P. A modified slant shear test designed to enforce adhesive failure. Constr Build Mater 2013;41:673–80. [20] Santos PMD, Júlio ENBS. Development of a laser roughness analyser to predict in situ the bond strength of concrete-to-concrete interfaces. Mag Concr Res 2008;60(5):329–37. [21] Santos PMD, Júlio ENBS. Factors affecting bond between new and old concrete. ACI Mater J 2011;108:449–56.
[22] Santos PMD, Júlio ENBS. A state-of-the-art review on roughness quantification methods for concrete surfaces. Constr Build Mater 2013;38:912–23. [23] Santos PMD, Júlio ENBS, Silva VD. Correlation between concrete-to-concrete bond strength and the roughness of the substrate surface. Constr Build Mater 2007;21:1688–95. [24] Talbot C, Pigeon M, Beaupré D, Morgan DR. Influence of surface preparation on long-term bonding of shotcrete. ACI Mater J 1994;91(6):560–6. [25] Vacher P, Dumoulin S, Morestin F, Mguil-Touchal S. Bidimensional strain measurement using digital images. Proc Inst Mech Eng 1999;213:811–7. [26] Verpeaux P, Charras T, Millard A. Castem 2000 une approche moderne du calcul des structures. Calcul des structures et intelligence artificielle 1988:261–71.