You are on page 1of 6

Fourth Asia-Pacific Conference on FRP in Structures (APFIS 2013)

11-13 December 2013, Melbourne, Australia


© 2013 International Institute for FRP in Construction

CHARACTERIZATION OF COHESIVE BOND INTERFACES BETWEEN FRP AND


UHPC UNDER MODE II LOADING USING FINITE ELEMENT METHODS

D. Chen 1 and R. El-Hacha 1,2


1
Department of Civil Engineering, University of Calgary, Calgary, Alberta, Canada.
2
Email: relhacha@ucalgary.ca

ABSTRACT

The behaviour of bond interfaces in composite structures is critical in the overall performance of the system.
The ability to characterize and effectively approximate this behaviour without a considerable experimental
sample size can be achieved through the use of finite element method software packages. This investigation will
focus on the performance of double lap shear specimens, with Fibre Reinforced Polymer (FRP) plates bonded to
Ultra-High Performance Concrete (UHPC) prism with an adhesively bonded aggregate interlock system. A
finite element model will be developed to model the global behaviour of the specimens, which will then be
validated using experimental test results. After model validation, the development of bond stresses leading up to
damage initiation as well as during damage evolution will be evaluated and compared with the trends observed
during experimental testing as well as those predicted using numerical methods. The influence on the overall
structural behaviour due to areas that have reached maximum bond stress will be examined. The findings from
this investigation will provide more clarity into the bond behaviour throughout the entire interface rather than at
distinct and discontinuous locations, allowing for improved predictions for pre- and post-damage behaviour at
material interfaces.

KEYWORDS

FRP, UHPC, bond, finite element, interfacial shear stress.

INTRODUCTION

Continued research into the development of optimized composite structural elements has led to many promising
and feasible options that can perform as well as, if not better than, conventional construction materials, namely
reinforced concrete and structural steel. The design of such composite structural element does involve an
element of additional challenge, where a possible region of weakness can arise if the bond interface between
adjacent materials has insufficient strength to ensure that the connected components perform as one entity.

The body of research into the performance of cohesive bonds is diverse, ranging from the study of interlaminar
interactions between neighbouring layers of Fibre Reinforced Polymer (FRP) sheets within a laminate sheet
(Allix and Ladavèze, 1992; Hashemi et al., 1990), its potential as a connection method at attachment points
between structural elements in modular construction (Nelson and Fam, 2013), and its application in hybrid
structural elements composed of more than one material within its cross-section (Bakay et al., 2009). Note that
the list above is by no means comprehensive and that numerous other branches exist in this field of research.
Due to its importance, researchers have attempted to develop and characterize the behaviour of different types of
bonding agents, through experimentation, modeling and numerical analysis (Keller and Gürtler, 2006; Sheppard
et al., 1998; Reeder and Crews, 1990; Tvergaard and Hutchinson, 1996; Zuo, 1990; Cho et al., 2006; De Moura
et al., 2008). Finite element modeling techniques have proven to be a useful tool in the design and analysis of
structural elements that incorporate bond interfaces. Once validated with experimental results, the developed
models can be applied towards the analysis of similar structural elements, without the need for additional
experimentation.

The goal of this paper is to determine the properties of the cohesive layer, a bonded silica sand layer, under pure
Mode II loading. Due to the composite nature of the bond layer, experimental test results were compared with
the finite element model analysis for validation of the results. Different approaches were used to evaluate the
suitability of the model. Parameters that were investigated include: a) the usage of linear or quadratic elements,
b) the accuracy of results obtained through a two-dimensional or three-dimensional model and c) the idealization
of the bond layer using cohesive elements compared with using an interaction with cohesive properties. In
addition, a closer inspection of the results obtained from the finite element analysis was taken to observe the
progression of crack propagation and the distribution of stresses along the bond interface. At the end of the
study, a recommended set of parameters are offered that will provide the optimum combination of accuracy and
required computational energy and time.

BACKGROUND INFORMATION
The experimental program was conducted on double shear bond specimens. A thin layer of silica sand was
bonded to the coarsened surface of a pultruded Glass FRP (GFRP) plate. The epoxy adhesive used for bonding
of the silica sand was a moisture insensitive epoxy adhesive originally intended for bonding between hardened
concrete and steel. After allowing for a seven day curing time, two of the prepared GFRP plates were bonded
directly to a cast-in-place Ultra-High Performance Concrete (UHPC) prism, to two of its opposing sides. The
configuration of the test specimen is shown in Figure 1.

Figure 1. Configuration and dimensions of the double shear bond specimens (Chen and El-Hacha, 2012)

The specifics details relating to the experimental program and its results were described in detail in a separate
publication and will not be repeated here in this paper (Chen and El-Hacha, 2012; Chen and El-Hacha, 2011).
Only key results required in order to construct the finite element (FE) model will be presented.

FINITE ELEMENT ANALYSIS

Model development

The FE model was developed using ABAQUS 6.9. The material properties for the pultruded GFRP plate are
provided in Table 1. The associated parameters for cohesive elements obtained from a related preliminary study
regarding the behaviour of full length hybrid FRP-UHPC beams (Chen and El-Hacha, 2011), based on the
method introduced by Diehl (2008), are provided in Table 2, along with the material properties for the UHPC.
The test matrix showing the different developed models and their designation IDs is shown in Table 3.

Table 1. GFRP material properties. (Chen and El-Hacha, 2011)


Parameters Value
Longitudinal modulus 17200 MPa
Transverse modulus 6890 MPa
Longitudinal tensile strength 207 MPa
Longitudinal compressive strength 207 MPa
Transverse tensile strength 48.3 MPa
Transverse compressive strength 103 MPa
Longitudinal shear strength 31 MPa
Transverse shear strength 31 MPa
Poisson’s ratio 0.33
Table 2. Cohesive parameters and UHPC material properties. (Chen and El-Hacha, 2011)
Parameters Value
UHPC Young’s modulus 50000 MPa
UHPC Poisson’s ratio 0.2
Cohesive shear stress 6 MPa
Cohesive shear fracture energy 1.2 J/m2
Cohesive penalty normal stiffness 100000 MPa
Cohesive shear stiffness 300 MPa

Table 3. Model designation IDs.


Model description Designation ID
2D model with linear elements (cohesive elements) 2D-L-CE
2D model with quadratic elements (cohesive elements) 2D-Q-CE
3D model with linear elements (cohesive elements) 3D-L-CE
3D model with quadratic elements (cohesive elements) 3D-Q-CE
2D model with linear elements (cohesive interaction) 2D-L-CI
2D model with quadratic elements (cohesive interaction) 2D-Q-CI

In the developed models, either plane stress bilinear or bi-quadratic elements were used to represent the GFRP
and UHPC materials, along with four-noded cohesive elements or a cohesive interaction used to idealize the
bond layer. The GFRP plate was modeled using 3 layers of elements, the UHPC prism modeled with 19 element
layers and the cohesive layer modeled as a single layer of cohesive elements. The approximate size of the mesh
along the length of the specimen is 4 mm. Due to symmetry of the specimen dimensions as well as the applied
load, only one-half of the specimen was modeled, with the appropriate boundary conditions used at the lines of
symmetry to restrict the degrees of freedom. A diagram of the FE model is provided in Figure 2.

Figure 2. Diagram of the developed finite element model

Comparison of model accuracies

As a general validation method, the results obtained from the four FE models performed using the cohesive
elements were first compared with the experimental load-deflection curves of the three bond specimens tested.
The graphical comparison is shown in Figure 3.

200

180

160
Applied Shear Load (kN)

140

120
Specimen 1
100 Specimen 2
Specimen 3
80
2D-L-CE
60 2D-Q-CE
40 3D-L-CE
3D-Q-CE
20

0
0 0.1 0.2 0.3 0.4 0.5
Slippage (mm)

Figure 3. Comparison of FE models with cohesive elements with experimental data.


From the results, it is evident that the use of more complex and computationally intensive models do not provide
any additional increase in the accuracy, in terms of the analysis results for global response. All of the load-
deflection curves derived from the FE analysis essentially overlap one another and as a group, shows a lower
stiffness and maximum load reached than the tested specimens. The accuracy of the FE analysis in relation to
the actual experimental results is quite acceptable, with a percent error for the maximum load of approximately
15%. Judging from the similarity of the FE analysis regardless of the dimension and element type used, for
further analysis, the simpler two-dimensional FE model using linear elements will be used.

Comparison between cohesive elements and cohesive interactions


Using the penalty-based cohesive zone finite element approach proposed by Diehl (2008), it imposes a large
number of simplifying assumptions and limitations on the parameters used in order to model the cohesive
behaviour solely based on the fracture energy. By modeling the interface as a layer of cohesive elements with a
quantifiable thickness, one of the required parameters is the specification of the cohesive layer stiffness. This
parameter is not explicitly necessary when the cohesive layer is defined as a cohesive interaction, which
idealizes the bond layer as a zero-thickness area, and allows for the substitution of element stiffness values with
a contact enforcement criteria.

The last pairs of FE models conducted were performed by replacing the layer of cohesive elements with an
equivalent cohesive interaction between the upper FRP plate and the lower UHPC prism. The global load-
deflection curves of models 2D-L-CE, 2D-L-CI, and 2D-Q-CI are compared in Figure 4.

200

180

160
Applied Shear Load (kN)

140

120

100
2D-L-CE
80 2D-L-CI
2D-Q-CI
60

40

20

0
0 0.1 0.2 0.3 0.4
Slippage (mm)

Figure 4. Comparison of behaviour using cohesive elements and cohesive interaction.

From the comparison shown in Figure 4, it is evident that the substitution of cohesive elements with a cohesive
interaction results in noticeably different behaviour in the load-deflection of the double shear specimens prior to
the initiation of crack growth, where the stiffness of the specimen appears to be nearly infinite prior to crack
initiation. This is to be expected since the contact enforcement criteria would assume complete contact and zero
relative movement above and below the interface prior to reaching the limiting shear stress. It appears that the
use of contact interactions should not be applied when linear elements are used for modeling the material
adjacent to the interface, as it leads to irregular digressions in the load-deflection behaviour. Nevertheless, it is
worth noting that as the load approaches the critical failure load, both methods converge to the same point.
Thus, if the aim of the model is to determine the ultimate behaviour of the system, without granting heavy
importance on the behaviour of the system during the initial loading prior to cracking, then a model using a
cohesive interaction rather than a layer of cohesive elements would provide similar, if not identical, solutions.

INVESTIGATION OF CRACK GROWTH AND INTERFACE SHEAR STRESS

To track the progress of crack growth along the interface, snapshots of the shear stress within the area of interest
at regular load intervals are provided for model 2D-L-CE and model 2D-Q-CI in Figures 5 a) and b),
respectively. The color scaling provided in both figures are consistent, where the blue regions indicate the
highest stress concentration, which is approximately equal to 6 MPa, the ultimate shear stress of the interface.
For reference to approximate dimensions, a scale is provided at the bottom of both figures.
Comparisons between the crack progression shown in Figures 4 indicate that, in general, the location of the peak
stress, and thus the location of the crack locus, is consistent between the two models. It can be noted that, in the
case of model 2D-Q-CI where the cohesive interaction was used to model the interface, uneven stress fields are
evident in the surrounding materials, particularly where high stress gradients are present. For model 2D-L-CE,
the cohesive element layer provides consistent cohesive layer nodes locations that line up with the nodes of the
adjoining materials, leading to the smooth stress fields seen in Figure 5a.

(a) (b)
Figure 5. Shear stress concentration plot of model (a) 2D-L-CE and (b) 2D-Q-CI at loads (from top to bottom) of
80 kN, 120 kN, 140 kN and 160 kN.

The detailed investigation into the stress and crack growth during loading shows that crack initiation would
commence at a load level of 80 kN, reaching an ultimate load of approximately 160 kN. At the ultimate
condition immediately prior to failure, the crack would extend to a length of around 180 mm along the interface.
It should be noted that, since the two-dimensional models used for this part of the investigation was performed
using plane stress elements, the location of the crack front obtained would be correlated to the crack location at
the outer edges of the cohesive layer in the equivalent three-dimensional model. This is due to the fact that, in
three-dimensional space, the crack front along the entire width of the specimen would be parabolic in shape,
where the crack location in the interior of the interface would lag behind the crack location at the edges. This
behaviour is shown in Figure 6.

.
Figure 6. Shape of crack front in a three-dimensional model.

CONCLUSIONS

This investigation into the FE modeling of a double shear bond specimen loaded under pure Mode II conditions
can provide the following conclusions:
• There is no discernible advantage in the use of higher order elements in the FE analysis when cohesive
elements are used;
• Simplification of the analysis into a two-dimensional model will provide near similar results to those
obtained from a three-dimensional model;
• The penalty-based cohesive zone finite element approach proposed by Diehl [18] performs well and
provides accurate results when compared with experimental data;
• The use of a cohesive interaction to model the interface layer requires higher order elements to be used
in the surrounding material in order to prevent irregularities in the obtained behaviour;
• Modeling with cohesive elements rather than a cohesive interaction will allow for smooth stress fields
in the adjacent bonded materials; and,
• The locus of the crack front determined from models using both cohesive elements and cohesive
interaction are equivalent.

From the summarized results, it is recommended for future FE model developments of this kind to use a layer of
linear cohesive elements to represent the bond interface, since it provides superior and consistent results on both
the local and global scales. The use of the cohesive interaction method, which assumes an initial contact
enforcement criteria, would only be recommended for use in preliminary studies or in the case where the global
behaviour at the ultimate load condition is desired.

ACKNOWLEDGEMENTS

The authors would like to thank Lafarge Canada and Sika Canada Inc. for their generous donations of materials
used. We would also like to acknowledge the University of Calgary and the Natural Sciences and Engineering
Research Council of Canada (NSERC) for their financial support towards this research project.

REFERENCES
Allix, O. and Ladavèze, P. (1992). “Interlaminar interface modeling for the prediction of delamination”,
Composite Structures, 22, 235-242.
Bakay, R., Sayed-Ahmed, E.Y. and Shrive, N.G. (2009). “Interfacial debonding failure for reinforced concrete
beams strengthened with carbon-fibre-reinforced polymer strips”, Canadian J. of Civil Eng., 36(1), 103-121.
Chen, D. and El-Hacha, R. (2012). “Bond strength between cast-in-place ultra-high-performance-concrete and
glass fibre reinforced polymer plate using epoxy bonded coarse silica sand”, Journal of ASTM International,
9(3), 17p.
Chen, D. and El-Hacha, R. (2011) “Investigation of contact behaviour in hybrid FRP-UHPC beams using finite
element methods”, Proceedings of the ACI Fall 2012 Convention.
Cho, K., Cho, J-R., Chin, W-J. and Kim, B-S. “Bond-slip model for coarse sand coated interface between FRP
and concrete from optimization technique”, Computers & Structures, 84(7), 439-449.
Correia, J.R., Branco, F.A. and Ferreira, J.G. (2007). “Flexural behaviour of GFRP-concrete hybrid beams with
interconnection slip, Composite Structures, 77(1), 66-78.
De Moura, M.F.S.F., Gonçalves, J.P.M., Chousal, J.A.G. and Campilho, R.D.S.G. (2008). “Cohesive and
continuum mixed-mode damage models applied to the simulation of the mechanical behaviour of bonded
joints”, International Journal of Adhesion and Adhesives, 28(8), 419-426.
Diehl, T. (2008). “On using a penalty-based cohesive-zone finite element approach, Part I: Elastic solution
benchmarks”, International Journal of Adhesion and Adhesives, 28(4-5), 237-255.
Hashemi, S., Kinloch, A.J. and Williams, J.G. (1992). “The analysis of interlaminar fracture in uniaxial fibre-
polymer composites”, Proceedings of the Royal Society of London, Series A, Mathematical and Physical
Sciences, 427(1872), 173-199.
Hassan, T. and Rizkalla, S. (2007). “Investigation of bond in concrete structures strengthened with near surface
mounted carbon fiber reinforced polymer strips”, Journal of Composites for Construction, 7(3), 248-257.
Keller, T. and Gürtler, H. (2006). “Design of hybrid bridge girders with adhesively bonded and compositely
acting FRP deck”, Composite Structures, 74(2), 202-212.
Lu, X.Z., Teng, J.G., Ye, L.P. and Jiang, J.J. (2005). “Bond-slip model for FRP sheets/plates bonded to
concrete”, Engineering Structures, 26(6), 920-937.
Nelson, M. and Fam, A. (2013). “Structural GFRP permanent forms with T-shape ribs for bridge decks
supported by precast concrete girders”, Journal of Bridge Engineering, Vol. 18, Issue 9, Sept. 813-826.
Mi, Y., Crisfield, M.A., Davies, G.A.O and Hellweg, H.B. (1998). “Progressive delamination using interface
elements”, Journal of Composite Materials, 32(14), 1246-1272.
Reeder, J.R. and Crews, J.R. Jr. (1990). “Mixed-mode bending method for delamination testing”, American
Institute of Aeronautics and Astronautics Journal, 28(7), 1270-1276.
Sheppard, A., Kelly, D. and Tong, L. (1998). “A damage zone model for the failure analysis of adhesively
bonded joints”, International Journal of Adhesion and Adhesives, 18(6), 385-400.
Suo, Z. (1990). “Failure of brittle adhesive joints”, Applied Mechanics Reviews, 43(5), S276.
Tvergaard, V. and Hutchinson, J.W. (1996). “On the toughness of ductile adhesive joints”, Journal of the
Mechanics and Physics of Solids, 44(5), 789-800.

You might also like