Creep behavior of cold-curing epoxy adhesives: analysis

and predictive approach

Karim Benzarti, Nourredine Houhou, Marc Quiertant, Sylvain Chataigner

To cite this version:

Karim Benzarti, Nourredine Houhou, Marc Quiertant, Sylvain Chataigner. Creep behavior of cold-
curing epoxy adhesives: analysis and predictive approach. Composites In Civil Engineering (CICE),
Aug 2014, VANCOUVER, Canada. 7p. �hal-01207683�

HAL Id: hal-01207683

https://hal.science/hal-01207683
Submitted on 1 Oct 2015

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 th
The 7 International Conference on FRP Composites in Civil Engineering

International Institute for FRP in Construction

CREEP BEHAVIOR OF COLD-CURING EPOXY ADHESIVES:

ANALYSIS AND PREDICTIVE APPROACH
Karim BENZARTI
Senior Researcher, PhD, Université Paris-Est, IFSTTAR, Département Matériaux et Structures, France
[email protected]

Noureddine HOUHOU
Civil Engineer, PhD, Chryso, France
[email protected]

Marc QUIERTANT
Senior Researcher, PhD, Université Paris-Est, IFSTTAR, Département Matériaux et Structures, France
[email protected]

Sylvain CHATAIGNER
Researcher, PhD, LUNAM Université, IFSTTAR, Département Matériaux et Structures, France
[email protected]

ABSTRACT: This study investigates the creep behavior of two commercially available cold-curing epoxy
adhesives, intended for the bonding of external composite reinforcements on concrete structures. In a
preliminary stage, the characteristics of the mineral fillers (nature, content and size) contained in the two
systems were determined, and the viscoelastic properties of the unfilled epoxy matrices extracted from
these systems were analyzed. Short-term tensile creep experiments were then carried-out on cured
samples of the two adhesives and their unfilled matrices, in order to evaluate the influence of the fillers
and the polymer network characteristics on the instantaneous and delayed mechanical responses.
Finally, two predictive approaches based on either the Time-Temperature Superposition Principle (TTSP)
or the Time-Stress Superposition Principle (TSSP) were applied to evaluate the long-term creep behavior,
and their suitability in the case of cold-curing adhesives was discussed.

1. Introduction
Cold-curing epoxy adhesives are commonly used to install external fiber reinforced polymer (FRP)
composite systems for repairing or upgrading reinforced concrete (RC) infrastructures. These epoxies are
usually bi-components systems that contain a certain amount of mineral fillers and additives to set the
rheological and mechanical properties. Their formulation is also adjusted according to the type of process
intended for the installation of the FRP materials (wet lay-up, or bonding of pultruded FRP plates, for
instance). For this reason, their mechanical behavior, can significantly vary from an epoxy system to
another, which may affect the durability and performances of the bonded FRP reinforcement, especially
under environmental ageing or sustained loads (Houhou, 2012; 2014)
In this study, it is first proposed to investigate the parameters related to the adhesive formulation (nature
and content of the mineral fillers, characteristics of the polymer network) that govern the viscoelastic
properties in general, and the creep behavior in particular. In this line, physicochemical characterizations
are carried-out on two adhesive systems from the market, and on the unfilled epoxy binders extracted
from these adhesives as well. The viscoelastic and tensile creep behaviors of the bulk adhesives cured at
ambient temperature are then analyzed in the light of the previous results, in order to point out influential
parameters. A second part explores two predictive approaches intended for evaluating the long-term
creep behaviour, based on the Time-Temperature or Time-Stress Superposition Principles. Their
suitability in the case of cold-curing adhesives, which exhibit an incomplete crosslinking, is examined.

2. Description of the selected epoxy adhesives

Two cold-curing and bi-component epoxy adhesives were selected among the various commercial
products available on the French market. Both systems are used for the installation of pultruded carbon

Page 1 of 7
FRP (CFRP) plates on RC concrete structures. The names of the two products are not explicitly cited in
this paper, and the two systems are noted Adhesive 1 and Adhesive 2.
The main physical and mechanical characteristics of these products are gathered in Table 1. The two
systems contain a substantial amount of mineral fillers in their components, but the content and nature of
the fillers are not mentioned in the technical data sheets.

Table 1- Properties of the two epoxy adhesives under study, as provided by the manufacturers

Adhesive 1 Adhesive 2
Organic components Part A: diglycidylether Bisphenol A Part A: diglycidylether Bisphenol A
Part B : aliphatic amines hardener Part B : modified polyamines and
cyclo-aliphatic amines
Density at +23°C 1.65 ± 0.1 kg/l (mix A+B) Part A : 1.43 kg/l Part B : 1.52 kg/l
Pot life in mn 120 (+8°C), 90 (+20°C), 20 (+35°C) 120 (+10°C), 60 (+23°C), 30 (35°C)
Glass transition Temp. 62°C (after 7 days at +45°C) 54 ± 2°C
(Tg) (ISO 11357-2)
Tensile strength (ISO 527) 24-27 MPa (after 7 days at +15°C) 29,5 ± 1 MPa
26-31 MPa (after 7 days at +35°C)
Tensile modulus (ISO 527) 11200 MPa (at +23°C) 4940 ± 170 MPa

3. Preliminary characterizations of the adhesives

In a first step, basic characterizations were carried-out in order to:
- assess precisely the nature and content of the mineral fillers in the 2 epoxies under study,
- determine the viscoelastic properties of the cured adhesives, and that of the organic binders (i.e,
the epoxy matrices whose mineral fillers have been removed prior to cure). At this stage, the
objective is to evaluate the influence of the fillers on the viscoelastic properties of the two
commercial systems. In a further step, this will help us to understand possible differences in the
experimental creep behaviors of the two products.
Observations by optical and scanning electron microscopy (SEM) techniques, coupled with energy
dispersive X-ray (EDS) elemental analysis were first conducted to assess the size, the volume distribution
and the nature of the mineral fillers. Fig. 1 shows polished sections of the two cured adhesives. In
Adhesive 1 (Fig. 1.a), there is a uniform distribution of large smooth particles in the polymer matrix and
the largest particle sizes are around 250µm. The elemental analysis reveals that these fillers are mainly
composed of silica sand. As regards Adhesive 2, a large density of voids can be observed (air bubbles in
black on fig 1.b), and filler particles appear to be much smaller (size < 50 µm) and exhibit an elongated
angular shape. In this case, EDS analyses show that these fillers consist of a mix of calcium carbonate,
silica, alumina and titanium dioxide.

(a) (b)
Fig. 1 - Micrographic observations of polished sections for Adhesive 1 (a) and Adhesive 2 (b)

Page 2 of 7
 The filler contents in the two pasty components (part A and B) of each adhesive were also assessed, and
the global content in the mix as well (part A+ part B, according to the mix ratio recommended by the
manufacturer). A centrifugation in a solution of methyl ethyl ketone (MEK) followed by a filtration made it
possible to separate the organic binders from the fillers, hence to quantify their relative weight contents.
Corresponding results are reported in Table 2. It was found that the filler content of Adhesive 1 (mix A+B)
is close to 80 wt.%, and much higher than that of Adhesive 2 which is around 60 wt. %. This result is
consistent with properties reported on the technical datasheet (Tab. 1), which indicate a higher density
and a much higher tensile modulus for Adhesive 1 compared to the other system.

Table 2 - Relative weight contents of organic binders and fillers in the 2 epoxy adhesives
System Component Organic binder (wt. %) Filler content (wt. %)
Adhesive 1 A 18,4 ± 0,8 79,7 ± 0,4
B 15,4 ± 0,8 85,2 ± 0,9
Mix A+B (ratio 3/1) 17,7 81,1
Adhesive 2 A 48 ± 2 52 ± 2
B 31 ± 2 68 ± 2
Mix A+B (ratio 1/1) 39 60

Following the centrifugation stage, the organic binders that were extracted from components A and B of
the two systems, were recombined into a mix (binder A + binder B), which was used to mold
parallelepipedic samples (25mm x 4mm x 1 mm ) of unfilled epoxy matrices. It is to note that the mix ratio
(binder A/binder B) was recalculated, so that the stoïchiometry is respected for the two systems, and the
resulting samples were cured at ambient temperature for at least 1 month before further characterization.
Similar specimens were produced with the filled (as received) adhesives, to serve as control samples.
Finally, the viscoelastic properties were determined by Dynamical Mechanical Analyses (DMA) for the
cured samples of filled adhesives and unfilled extracted matrices. Tests were carried-out under
traction/compression mode with an imposed dynamic displacement of amplitude 10 µm, at a frequency of
1 Hz. The test specimen was simultaneously subjected to a ramp of temperature between 20 and 140°C
at a heating rate of 2°C/min. This technique made it possible to record the evolution of the storage
modulus (E’) and the loss factor (tanδ) over the temperature interval. Results are plotted in Fig. 2.
1.00E+11 1
Adhesive 1 Adhesive 1
Adhesive 2 0.9 Adhesive 2
Matrix extracted from Adhesive 2 0.8 Matrix extracted from Adhesive 2
1.00E+10
Matrix extracted from Adhesive 1 Matrix extracted from Adhesive 1
0.7

1.00E+09 0.6
E' (Pa)

tanδ

0.5
1.00E+08 0.4

0.3
1.00E+07 0.2

0.1
1.00E+06 0
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Temperature (°C) Temperature (°C)
(a) (b)
Fig. 2 - DMA experiments at 1 Hz: Evolutions of the storage Modulus E’ (a) and the loss factor tanδ
(b) versus temperature, for Adhesive 1 and Adhesive 2 and their unfilled extracted matrices

As regards the filled adhesives, a large drop of the storage modulus is observed through the glass
transition (Fig 2.a.) for both systems. However, the moduli of Adhesive 1 in the vitreous and rubbery
domains remain much higher than that of Adhesive 2, due the higher filler content. On Fig 2.b., the
temperature at the maximum of the tanδ peak, called Tα and closely connected to Tg, is found to be
higher for Adhesive 2 compared to Adhesive 1 (deviation of 12°C). This feature is contradictory with the

Page 3 of 7
values of Tg mentioned in the technical datasheet; which may be due to the unrepresentative cure
conditions chosen by the manufacturer of Adhesive 1 (7 days at 45°C, as reported in Table 1).
In the case of the unfilled extracted matrices, a drop of E’ is also observed through the glass transition
region; however, differently from the filled epoxies, the unfilled matrices extracted from Adhesive 1 and
Adhesive 2 exhibit very similar values of the vitreous and rubbery storage moduli. This confirms the major
role played by fillers on the mechanical elastic properties of commercial adhesives. As regards the loss
factor peak, the matrix extracted from Adhesive 1 exhibits lowers values of both Tα and the peak
amplitude, as compared to the matrix extracted from Adhesive 2. This feature suggests a lower crosslink
density, hence a higher molecular mobility of the epoxy network in Adhesive 1.
Finally, DMA characterizations have shown that the elastic characteristics (storage moduli in the vitreous
and rubbery domains) of the two commercial adhesives are mainly dependent on the filler content,
whereas the dispersive characteristics (Tα and amplitude of the loss factor peak) are rather related to the
structure of the epoxy network (nature of the reactive components, crosslink-density).

4. Analysis of the short-term creep behavior

Short-term tensile creep experiments (25°C, applied stress of 5 MPa) were carried-out on samples of the
two commercial adhesives and their unfilled extracted matrices. Creep curves are displayed in Fig.3.
0.006 0.006 Adhesive 2
Adhesive 1
Matrix extracted from Adhesive 1 Matrix extracted from Adhesive 2
0.005 Burger's Model 0.005 Burger's Model

0.004 0.004
Strain

0.003 0.003
Strain

0.002 0.002

0.001 0.001

0 0
0 500 1000 1500 2000
0 500 1000 1500 2000
(a) Time (s) (b)
Time(s)
Fig. 3 - Short-term creep tests for Adhesive 1 (a) and Adhesive 2 (b) and their unfilled matrices, at
25°C for an applied stress of 5 MPa. Fitted curves of Burger’s model are also displayed.
The two extracted matrices exhibit a similar level of elastic strain (instantaneous strain) following the initial
loading stage, which is consistent with their similar storage moduli in the vitreous state. Nevertheless,
matrix extracted from Adhesive 1 presents a higher creep rate (slope of the asymptote), which can be
explained by a higher molecular mobility of the epoxy network, as shown previously by DMA analyses.
Filled adhesives show globally lower levels of deformation than their extracted matrices, due to the
presence of mineral fillers. When comparing the 2 filled systems together, large differences are observed
in terms of initial elastic strain (it is much lower for Adhesive 1 which has the highest filler content and
vitreous elastic modulus) and in terms of creep rate (which appears higher for Adhesive 2). This latter
result contradicts the trend observed for the extracted matrices, as the matrix of Adhesive 1 exhibits the
highest creep rate. It is thus suggested that the creep rate of filled adhesives in the vitreous state is not
only dependent on the molecular mobility of the polymer matrix, but also on other factors such as
interactions between filler particles. It is likely that friction phenomena between sand particles hamper the
creep process in Adhesive 1, due to the high filler content and the large size of the sand particles.
Finally, the well-known rheological Burger’s model (Fig. 4) was used to simulate the creep behaviors of
both the filled adhesives and the extracted matrices. Parameters of the model were identified by fitting the
experimental data, according to the method described in (Houhou et al, 2012; 2014). Values of these
parameters are not reported in this paper, but it is obvious from Fig.3 that a fair agreement was obtained
between experimental and modelled curves.

Page 4 of 7
  

Fig. 4 - Schematic description of Burger’s rheological model

5. Predictive approaches
This last part aims at comparing two predictive methods for evaluating the long-term creep response of
cold curing adhesives, based on the Time Temperature Superposition Principle (TTSP) and the Time
Stress Superposition Principle (TSSP). Both approaches are applied to the case of Adhesive 1.
The TTSP method consists in conducting short-term creep experiments at various temperatures and build
a creep compliance master curve by shifting the experimental compliance curves horizontally along the
log-time scale axis, so that they fit the curve at a chosen reference temperature. Fig. 5.a and 5.b display
respectively the experimental compliance curves recorded at several temperatures between 25 and 57°C
(for an applied stress level of 5 MPa), and the compliance mater-curve at the reference temperature of
25°C for Adhesive 1. From this master curve, it is then easy to determine the long-term evolution of the
creep strain for the material under study, and to fit this curve with the Burger’s model (Fig. 5.c.).
In the case of adhesive systems that exhibit a highly non-linear creep behavior, Houhou et al. (2014)
have demonstrated that the previous TTSP procedure can be reproduced for various levels of applied
stress, and one can finally identify a non-linear Burger’s model. However, a drawback of the TTSP
method is that the polymer adhesive is heated during the multiple isothermal tests, and this may affect its
microstructure and provide unrepresentative results. This is especially true for cold-curing epoxies, which
present usually an incomplete crosslinking after curing at ambient temperature.
  25°C 27°C 29°C 31°C   25°C - Reference 27°C - Shifted
Creep Compliance (1/MPa)

0.007 33°C 35°C 37°C 39°C

Creep Compliance (1/MPa)

0.006 29°C - Shifted 31°C - Shifted

41°C 43°C 45°C 47°C 33°C - Shifted 35°C - Shifted
0.006 49°C 51°C 53°C 55°C 37°C - Shifted 39°C - Shifted
57°C 0.005 41°C - Shifted 43°C - Shifted
45°C - Shifted 47°C - Shifted
0.005 49°C - Shifted 51°C - Shifted
0.004 53°C - Shifted 55°C - Shifted
0.004 57°C - Shifted
0.003
0.003
0.002
0.002

0.001 0.001

0 0
(a) 1.5 2 2.5 3 3.5 (b) 1 3 5 7
log t (s)
9 11 13
log t (s)  
1500
Creep strain (x10 - 6)

1000

TTSP
500
Burger's model

0
(c) 0.0E+0 5.0E+5 1.0E+6 1.5E+6 2.0E+6 2.5E+6 3.0E+6
Time (s)
Fig. 5 - Creep compliance vs. time curves obtained from short-term creep tests under isothermal
conditions for a stress of 5 MPa (a). Master curve for the reference temperature of 25 °C (b).
Comparison of creep strain vs time curves obtained from TTSP approach and Burger’s model (c)

Page 5 of 7
An alternative predictive method based on the Time-Stress Superposition Principle (TSSP) might be more
adapted for cold curing adhesive systems, as it doesn’t require to heat the samples. It involves short term
creep experiments at various loading levels, and the master-curve is again constructed by shifting the
experimental compliance curves along the log-time scale axis. Fig 6. shows an example of application for
Adhesive 1. Successive creep/recovery cycles (15 min creep/15 min recovery) were carried-out at several
levels of applied stress in the range 3 - 25 MPa. Unfortunately, for the highest stress levels (above 20
MPa, which is lower than the expected tensile strength), a rapid failure of the samples was observed, due
to the brittleness of Adhesive 1. For this reason, a master curve was built by shifting only the available
5
creep compliance curves, providing a prediction of the creep behavior limited to 5 decades (10 s).
7.5
Nevertheless, several authors have reported successful predictions up to 10 s (one year) for other
polymer materials (Jazouli et al, 2007), using this TSSP approach which remains of high interest.
0.0002 0.0002
3 MPa 5MPa 5 MPa 6.3 Mpa shifted
6.3 MPa 7.8 MPa

Creep Compliance (1/MPa)

9.4 MPa 11 MPa 7.8 MPa shifted 9.4 MPa shifted
0.00018 12.6 MPa 14.3 MPa 0.00018 11 MPa shifted 12.6 MPa shifted
16 MPa 17.6 MPa
Creep Compliance (1/MPa)

19.2 MPa 14.3 MPa shifted 16 MPa shifted

17.6 MPa shifted 19.2 MPa shifted
0.00016 0.00016

0.00014 0.00014

0.00012 0.00012

0.0001 0.0001

0.00008 0.00008
1.5 2 2.5 3 1.5 2 2.5 3 3.5 4 4.5 5
(a) Log t (s) (b) Log t (s)
  from short-term creep tests at various load
  Fig. 6 - Creep compliance vs. time curves obtained
levels and at a temperature of 25°C (a). Master curve for a reference stress level of 5 MPa (b).

6. Conclusion
The present study has investigated the creep behavior of two selected cold curing epoxy adhesives,
which are commercially available and commonly used in construction for the bonding of CFRP plates.
Preliminary physicochemical characterizations revealed significant differences in the composition and
properties of the two systems, in terms of nature, content and size of the mineral fillers. Viscoelastic
analyses performed on the unfilled matrices extracted from these two systems and cured at room
temperature also demonstrated differences in terms of crosslink density of the epoxy networks.
Short-term tensile creep tests were then carried-out on cured samples of the two adhesives and their
unfilled extracted matrices, showing that the level of the instantaneous elastic strain associated to the
loading stage strongly depends on the filler content of the adhesive, while the creep rate depends upon
both the molecular mobility of the epoxy network and the possible interactions between filler particles.
Finally, two predictive approaches were applied to evaluate the long-term creep behavior, based either on
the Time-Temperature or the Time-Stress Superposition Principles. The first method appears to be very
efficient but may provide unrepresentative results, as it requires heating the samples, which can alter their
microstructure. The second method doesn’t involve heating, but requires creep tests at high stress levels,
which may be difficult, hence limiting the predictions to a few decades.

7. References
HOUHOU, N, QUIERTANT, M., BENZARTI, K., CHATAIGNER, S., RABASSE, M., FLETY., A.
Innovative experimental device for the creep characterization of the concrete/FRP adhesive bond,
6th Int. Conf. on FRP Composites in Civil Engineering –CICE 2012, Rome, Italy, 13-15 June, 8 p.
on CD.
HOUHOU, N., BENZARTI, K., QUIERTANT, M., CHATAIGNER, S., FLETY, A. & MARTY, C., Analysis of
the nonlinear creep behavior of concrete/FRP-bonded assemblies. Journal of Adhesion Science
and Technology, vol. 14-15, 2014, 1345-1366.
JAZOULI, S., LUO, W., BREMAND, F., VU-KHAN, T, Application of time-stress equivalence to non-linear

Page 6 of 7
creep of polycarbonate, Polymer Testing, vol. 24, 2005, 463_467

Page 7 of 7