Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

©Civil-Comp Press, 2013

Proceedings of the Fourteenth International Conference on


Paper 167 Civil, Structural and Environmental Engineering Computing,
B.H.V. Topping and P. Iványi, (Editors),
Civil-Comp Press, Stirlingshire, Scotland

Finite Element Model of the Pull-Out Test for


Concrete Strength Evaluation
D. Meloni, B. De Nicolo and M. Valdes
Department of Civil and Environmental Engineering and Architecture
Faculty of Engineering, Cagliari, Italy

Abstract
In this paper numerical analysis of pull-out tests for concrete are discussed. The
analysis is carried out by means of finite element models based on an experimental
study carried out in the laboratory of the Department of Civil and Environmental
Engineering and Architecture (DICAAR) of Cagliari University on properly made
concrete samples. The experiments have been performed on 24 cubic 300x300x300
mm specimens, made of six different concrete grades with cubic compressive
strength ranging from 15 to 45 MPa. Post-installed expansion anchors have been
forced into the drillings and then extracted from the specimens by means of manual
equipment. Also a specific correlation between the extraction force and the
compressive cubic strength has been provided.
The numerical models have been carried out and calibrated in order to provide the
same results as the experimental tests and at the same time a deeper insight into the
physics of the phenomenon. First a proper idealization of the tests had to be
arranged, discarding all the features considered of minor importance such as
prestressing, friction, geometrical details and taking advantage of the axial
symmetry of the problem. Several material models have been tested within the
framework of plastic and damage theory and the encountered numerical problems
have been discussed, such as mesh sensitivity of the results, convergence difficulties
of classic implicit algorithms and the lack of results objectivity due to the use of
strain softening constitutive laws. Final axisymmetric non linear models have been
proposed and solved by means of an explicit dynamic algorithm capable of
bypassing the above mentioned convergence problems, though involving other
difficulties due to inertial effects, reduced integration finite elements, and mesh
sensitivity. The results obtained have been compared with the experimental ones,
showing a good accordance between the compressive strengths of the material
models and the expected extraction loads. Pros and cons of the chosen approach are
finally discussed and possible improvements and future research lines are proposed.

Keywords: pull-out, finite element, explicit algorithm, size effect.

1
1 Introduction
The Pull-Out test belongs to the family of non destructive tests (NDTs) for structural
elements mechanical properties and homogeneity in-situ evaluation. Together with
tests on cored specimens and the Sonreb Method, it’s part of the standard procedure
for in-situ investigations on new and old concrete structures.
The test commonly consists of the extraction of a metal insert that has been
previously embedded in a concrete element before it was cast (cast-in devices), or
afterwards by means of drillings (post-installed devices). Together with the anchor,
a conic portion of concrete is usually pulled off. The force needed to perform the
extraction can be related to the hardened concrete resistance, commonly in terms of
cubic compressive strength. Proposed for the first time in the ’40s [1,2] in the
subsequent years, especially from the ‘70s, a lot of researches and publications have
been devoted to this topic, both from the theoretical and experimental point of view,
in order to evaluate proper correlation laws between the extraction force and the
material strength, to clarify the nature and the physics of the involved collapse
phenomena and to lay the basis for standard codes. Throughout all these researches
and decades of applications the test proved its effectiveness in reaching its goal with
a good approximation. Nevertheless a lot of features of the problem are still a matter
of discussion, such as: the actual stress state induced in the proximity of the anchor
during extraction, the failure mode of concrete and consequently the material
parameters directly involved.
In Europe the test is performed according to UNI EN 12504-3 [3] and it’s commonly
recognized to ensure a good estimate of concrete compressive strength with an error
not larger than ±20%, that can be regarded as a satisfactory results as compare to
other NDTs. But, accordingly to the mentioned code, to ensure a better level of
reliability, a correlation between the test results and the concrete strengths should be
calibrated case by case by means of a certain number of compression test on cored
samples.

2 The Pull-Out test


2.1 Methodology
As above mentioned the test requires the introduction of a calibrated metal insert in
the examined volume of concrete, either embedded before the concrete is cast, or
inserted afterwards in hardened concrete by means of drillings. In the latter case a
pachometer survey is needed to avoid any interference with the existing
reinforcements. The extraction is performed by means of an hydraulic jack and a
special equipment comprising a metal ring of inner diameter D, used as a contrast
against the concrete surface (see Figure 1 and 2). The extraction of the insert,
usually causes the removal of a conical portion of concrete, characterized by a
standard geometry due to the dimensions of the extraction equipment and the insert
itself, i.e. its height h and the effective diameter d. In the common practice,

2
depending on the equipment used, the following dimensions are used: D = 55÷70
mm, d = 20÷25 mm, h = 25÷30 mm, 2α = 54°÷70°.
Different kind of expansion inserts can be used to provide the anchorage to the
concrete: expansion anchors, whose edge expansion is forced against the lateral
concrete surfaces of the drilling and undercut anchors, whose expansion is free to
develop inside an undercut at the end of the drilling. Sometimes bonded anchors are
used as well.
h

2
d

Figure 1. Scheme of the test

Figure 2. Equipment for the extraction of the insert

3
Figure 3. Expansion anchors used during the experimentation

2.2 Interpretation of experimental results


In the literature there’s no general agreement on which mechanical property is
directly involved during the extraction of the conical portion of concrete, whether
the tensile strength or the compressive strength, due to the complex state of stress
preceding the rupture, and even on the possibility of correlating the extraction force
to the compressive strength, as it’s usually addressed.
Nevertheless several experimental and theoretical correlations have been presented
between the extraction force or the manometer pressure and the compressive
strength of concrete, as usual dependent on the specific implemented methodology
or equipment.
The form of the experimental correlations are usually linear, bilinear, non linear or
exponential. Among non linear formulas, there’s the one proposed by Bocca [4]:

R = 7.44 ⋅ e 0.013 X (1)

Equation (1) correlates the cubic compressive strength R (MPa) to the pressure X
(Atm) provided by the manometer and is referred to post-installed expansion
anchors methodology. An exponential correlation has been provided by Chabosky
and Briden-Smith [5], but referred to the expansion controlled couple T (Nm):

R = 3.116 ⋅ T 1.69 (2)

Similar formulation has been proposed by Cossu and Pozzo [6] with reference to
the extraction force P (KN):

4
R = 0.0092 ⋅ P 1.647 (3)

Among linear formulations, there’s the general Danish correlation, referred to


expansion anchors methodology:

P = 1.0 + 0.96 ⋅ f 2 ≤ P ≤ 25 (4)


P = 5.0 + 0.80 ⋅ f 25 ≤ P ≤ 65 (5)
P = 2.2 + 0.75 ⋅ R 3 ≤ P ≤ 65 (6)

Expressions (3) and (4) represent a bilinear relationship between extraction force
P (KN) and cylindrical compressive strength (MPa), while the (5) is a unique linear
correlation addressed to cubic compressive resistance (MPa) evaluation.
Analogue expression is proposed by Meneghetti et al. [7] for both pre-installed and
post-installed expansion anchors:

R = 94.1 + 0.092 ⋅ P (7)

These last formulations claim to be general, being elaborated on the basis of a


wide number of experimental results, but the commonly accepted position is that a
proper correlation has to be formulated for specific experimental devices and
concretes. UNI EN 12504-3, the commonly referred standard for the test in Europe
together with ISO/DIS8046, underlines that empirical correlation are needed for the
specific apparatus in use and that such formulations can be adopted for a wide range
of concretes, but are unlikely to have 95% confidence limits better than ± 20%. On
the other hand if a specially prepared correlation for the type of concrete is available,
estimates can be improved to ±10% confidence limits. Coherently the code doesn’t
provide a correlation between compressive strength and extraction force, and ends
up to indicate a generic Pull-Out strength, determined as the ratio between the
extraction force and the area involved in the rupture, that is to say the predetermined
lateral surface area of the conical portion extracted together with the insert.

2.3 Brief review on mathematical models


Several theoretical interpretations of the failure process involved by the test have
been proposed in the last 40 years, accompanying the attempts to provide the
mentioned empirical formulations. Before the spreading of the finite element
method (FEM) a lot of analytical approaches have been pursued, trying to
understand which mechanical property is responsible for the rupture and the
sensitivity of the extraction load to the variation of the different parameters. Some
qualitative remarks widely agreed are: the stress state of the concrete volume
comprised between the anchor head and the bearing ring is complex and non
uniform; circular cracks develop during the first load levels and propagate towards
the bearing ring; after the cracks reach a sufficient length, the conical portion of
concrete is abruptly pulled off.

5
Some researches [8] have correlated the rupture to the concrete mortar tensile
strength. Some others have underlined that crushing of concrete localized near the
anchor head precedes the propagation of the fracture due to tensile cracking. Ottosen
[9] proposed one of the first F.E. analysis in the non linear field for pre-installed
devices, questionably assuming a perfect connection between the insert and the
surrounding concrete and describing the resulting crack evolution. The final
extraction load was detected in correspondence to the failure of analysis
convergence after numerous iterations and was ascribed to concrete crushing.
Similar approach was pursued by Peier [10], who, via axisymmetric F.E. models,
examined several failure modes for different equipment settings. The chosen
constitutive model was a three-dimensional non linear model, accounting for
cracking via a deformation controlled criterion and crushing with a strain softening
post-failure behaviour. Some of the detected failures were due to the crushing of the
concrete surrounding the anchor head. Yener [11] in his researches underlines the
complexity of the three-dimensional stress state involved by the test, examined by
means of a modified Von Mises elastoplastic constitutive model, and points out the
crushing of concrete under the bearing ring. Also Etse [12] ascribes an important
role to the compressive strength of concrete in driving the collapse mode and
analyzes the problem by means of a non linear finite element model based on an
elastoplastic constitutive model with isotropic hardening, associated flow rule and a
fracture energy driven softening behaviour. Other researchers, on the other hand,
have proposed a F.E. modelling based on non linear fracture mechanics (NLFM) and
the Fictitious Crack Model [13], disregarding the contribution of compressive failure
to the general phenomenon. Bortolotti [14], starting from the slabs punching load
theory, has invoked a tensile hardening to explain some mismatches between
experimentations and theoretical analytical models when low diameter bearing rings
are used. The author used a modified Mohr-Coulomb constitutive model and came
to the following expression of the extraction load (for the symbols refer to Figure 1):

P = f *t ⋅π(D + d) ⋅ h/ 2 cos α (8)

The expression depends uniquely on the tensile strength, properly modified as


a function of a variable internal friction angle, as the result of a fictitious hardening
induced by the constrained geometry of fracture surface. Finally the correlation to
the cylindrical compressive strength is provided and dependence on the curing time
is also taken into account. As a last reference, Pivonka, Lackner and Mang [15]
proposed different F.E. models based on single- and multi-surface plasticity
constitutive models. It’s interesting that in the this work the mathematical model
takes into account the different stages of the test, including the deformation of the
metal jacket during the expansion and the contact to the concrete surface, before
applying the extraction force.
From this brief review it’s clear that different interpretations of the complex collapse
phenomenon still coexists, in terms of the main parameters driving the collapse, the
role of compressions up the anchor head and under the bearing ring, and ductility of
damage propagation before the extraction, for example due to lateral confinement.

6
3 Experimentation
3.1 Overview
In the laboratories of the Department of Civil and Environmental Engineering and
Architecture (DICAAR) of Cagliari a wide experimental campaign has been recently
carried out on different kind of NDTs applied to different grade concrete samples, in
order to investigate proper correlations for single or combined methodologies aimed
to determine the concrete compressive strength. For these purpose 24 cubic
specimens have been prepared with different concrete grades from 15 to 45 MPa
cubic compressive resistance, that have been subjected to Ultrasound method test
and Pull-Out test. Subsequently, from the above specimens cored samples have been
extracted and subjected to destructive compressive tests in order to evaluate the
reference compressive resistance.

3.2 Description of the experimentation


The 24 specimens were distinguished in 6 groups of 4, respectively characterized by
the following expected characteristic cubic compressive strength Rck: 15, 20, 25, 30,
35, 45 MPa. For each specimen, after the detection of ultrasonic propagation
velocity, five extractions have been performed on different faces, discarding the
confection face.

Figure 4. Execution of the test on one specimen

7
Figure 5. Conical portion extracted together with the anchor

The equipment used during the experimentation consisted of a manual pump with
60 KN maximum load manometer, an hydraulic jack, a bearing ring with 69.30 mm
internal diameter and 109.30 mm external diameter. The used insert are Fischer M10
14x40 mm expansion anchors.
After the drilling, the anchors have been inserted and, by pushing with a hammer its
internal rod, the final 10 mm expanding part of the anchor external jacket is forced
to expand against the drilling lateral surface, providing the needed interlocking with
the concrete. The procedure described by UNI EN 12504-3 has been strictly
followed with care of the positioning and centring of the drilling and load
application speed, though the code is addressed mainly to pre-installed and undercut
expansion devices. The Guidelines [16] for NDT evaluation of concrete properties
issued by Italian Consiglio Superiore dei Lavori Pubblici (CSLLPP) have been
regarded as well. For each specimen face one extraction has been performed and the
results have been averaged for the subsequent comparison with the strength
evaluated by means of the compressive tests on cored specimens. The integrity of
the extracted conical portions has been regarded too, resulting in the observation that
for low resistance concretes it’s hardly possible to pull off an integer portion.

3.2 Results and elaborations


Though the above mentioned Italian Guidelines suggest a linear correlation between
the extraction force P (KN) and the cylindrical strength of concrete fc (MPa), the one
proposed on the basis of the experimentation results has an exponential form, that
has proved to ensure better estimates (see Figure 6):

f c = 0.056 ⋅ P 1.874 (8)

8
Figure 6. Correlation between the force P and the experimental strength fc

Table 1. Comparison of tested compressive strength fc and estimated strength fext

9
With the above formulation a good agreement between predicted compressive
strengths of a wide range of concrete grades and the ones determined with
compressive tests, can be reached. Table 1 shows the comparison between the tested
strength fc, and the strength fext evaluated by formulation (8). The corresponding
extraction forces and difference in the estimation are also presented. The Table
clearly shows that a good agreement is reached in most of the cases. In fact, the
differences range from 1 to 20%, but in the 71% of cases a confidence limit better
than 10% is reached.
The experimentation has gone further examining the possibility of combining the
Pull-Out and the Ultrasonic methodologies to ensure better estimates, but this is
beyond the goal of this paper.

4 Finite Element Model


4.1 General description
As mentioned above, the Pull-Out test consists of different stages before performing
the extraction. In the present modelization only the last one corresponding to the
extraction has been taken into account, disregarding the insert-concrete interlocking
stage, that in the case of expansion anchors is obtained by forcing the expanding
metal jacket, next to the insert head, against the surrounding concrete, causing a
local compressive damage that is likely to influence the initiation of cracking.

Figure 7. Images of the F.E. model

No frictional contact with the drilling internal surface has been taken into account
along the insert rod. According to accurate observations of the drilling and the insert

10
surfaces, only the upper part of the anchor head (about 10 mm) has been assumed to
be responsible of the mechanical interlocking and thus in contact with the
surrounding concrete (see Figure 8). The contact hasn’t been strictly modelled, being
the F.E. mesh of the two parts simply connected, under the hypothesis that only
compressive stresses and no friction are acting on the above mentioned surfaces. On
this basis only the anchor head has been modelled.

Figure 8. Detail of the F.E. model

The problem is globally axialsymmetric, both geometrically and in terms of loads


and constraints, therefore it has been reduced to a 2D problem, with a considerable
resources saving. Thus, only a half diametral section has been actually modelled.
The deformations of the metal insert have been judged negligible, so the anchor
head has been modelled as a rigid body. Almost the entire specimen has been
modelled and translational constraints have been applied in correspondence of the
bearing ring contact surfaces and in the axis of symmetry. In correspondence of the
upper edge of the insert head a vertical displacement has been applied to perform the
extraction. The final model, implemented in Abaqus v. 6.5 [17], has been meshed
with 1273 nodes and 1229 four nodes, axialsymmetric, reduced integration elements
(CAX4R), with a denser mesh in the area were fracture is likely to propagate.

4.2 Constitutive Models


4.2.1 Introduction

The code in use provides several constitutive models to threat concrete elements,
that mix plasticity and damage theories differently in the compression and in the

11
tensile fields. Some of them are available both for the implicit and explicit algorithm
for equilibrium equations resolution. They all implement the concept of smeared
cracking, i.e. that the damage associated to cracks onset and propagation is not
discretely modelled, but smeared in each finite element volume in the form of
stiffness matrix weakening.
This models generally introduce a pathology in the F.E. analysis, consisting in a
mesh dependency of the results. In other words, the fracture propagation and finally
the overall result may depend on the elements dimensions. This inconvenient can be
solved, or regularized, as will be discussed in the following, with the definition of a
further property of the material such as the characteristic length or alternatively the
specific fracture energy G, that are forced to be mesh independent.
Strain softening at least in tension is available, according to the cohesive zone model
or fictitious crack model [13], usually expressed as a function of crack opening, so
that the damage can develop gradually and fracture energy can be evaluated.
Preliminary a certain number of experiments have been attempted to evaluate the
suitability of these constitutive models in the specific case. Apart from the
convergence difficulties encountered, as it was expected due to the high non
linearity of the problem, the tested plasticity models failed in reproducing the
conical crack surface, due to a premature compressive crushing near the anchor
head, which is acceptable for low strength concretes. Given the highly triaxial
compressive state of stress present in that area, the models seemed inadequate in
treating the compressive behaviour of concrete in the case of high lateral
confinement.
Finally, as a simplification, the rupture leading to the extraction has been ascribed
entirely to the brittle tensile cracks development and no attention has been paid to
the crushing of concrete, nor in the form of damage, or plasticity.

4.2.1 Brittle model for concrete

The chosen material model for the concrete is a brittle constitutive model, whose
behaviour is merely elastic in the compression field and also in the tensile field, but
up to a limit represented by a crack detention surface; then damage starts developing
till a total loss of stiffness, which corresponds to the complete crack opening and the
stress orthogonal to the crack plane falls to zero.
The crack detection criterion is shown in Figure 9, where a Rankine-like failure
surface for the general three-dimensional case is represented in the deviatoric plane.
Therefore crack onset is determined when maximum principal tensile stress is
reached with reference to only Mode I fracture.
The crack opening is treated according to Hilleborg’s theory of fictitious crack
model [13] with a strain softening driven by the crack opening displacement (Figure
10), accounting for the Mode I fracture energy Gf. Cracks can also close, but are
unrecoverable with respect to their directions. This is not of concern in this specific
case where a monotonic load history is dealt with.
After first crack develops, Mode II fracture is also taken into account, in the
previously detected crack plane, with a Shear Retention Model, according to which
shear stiffness depends on crack opening as well, in order to account for the loss of

12
Figure 9. Fracture detection criterion (taken from [17])

Figure 10. Strain softening (taken from [17])

aggregates interlocking. The influence of this shear model on crack propagation has
been preliminary investigated and a linear evolution of the Shear Retention Factor,
affecting the shear modulus on crack surfaces, has been discussed.

4.2.2 Tensile softening and mesh dependency

As previously mentioned, the main feature of the model in use, is a deformation


driven damage with strain softening. The latter can be defined in terms of stress-
strains curves, but also as a stress vs crack opening curve, or, equivalently, directly
defining the Fracture Energy corresponding to the area under this curve provided a
fix shaped softening is assumed. To correlate the displacement and particularly the

13
ultimate displacement corresponding to zero stress across the crack to the strain
field, a geometrical characteristic length needs to be invoked, equivalent to a crack
band width, in which fracture energy is dissipated (Figure 11). Apart being in
accordance to experimental evidences, this feature provides at the same time a mean
to overcome the pathologic dependency of these models on mesh density due to
strain or damage localization. In fact the fracture energy or the displacement at zero
stress can be imposed as a material characteristic and the strain-stress constitutive
curve can be adapted to the mesh density keeping the final result independent.
Unfortunately in general cases and meshes it’s not easy to automate the
determination of this characteristic length, depending on the shape of the elements
and the local direction of principal stresses. That’s why some kind of dependency
can still persist.

Figure 11. Stress Vs Strain or Displacement softening curves

Particularly, since in Abaqus the characteristic length for a quadrilateral planar


element is defined according to equation (8), regardless the local stresses directions
and thus the fracture plane direction,

Lch = Lex ⋅ Ley (8)

we found more suitable to avoid too stretched elements and to force the mesh to
be as much regular as possible in the part of it where fracture is expected to occur
(see Figure 8). This has been very important in order to avoid spurious propagation
paths of the fracture.

4.3 Solution algorithm


As previously discussed these types of analysis imply severe numerical difficulties
due to the high non linearity of the fracture process leading to failure and to the
localization of damage and deformation. Though several attempts have been made, it
was generally impossible to get to the conclusion of the analysis without

14
convergence the classical non linear algorithms, such as Newton Raphson, or the
arc-length method, even under displacement control.
That’s why we moved to the explicit algorithm provided by Abaqus. This algorithm
uses an application of the central difference method to solve the dynamic equations
of equilibrium for the system degrees of freedom. For every i-th increment, velocity
is evaluated according to equation (9), without any iteration or balance check, then
integrated to evaluate the solution displacements.

∆t (i +1) + ∆t i (i )
 1  1
 i+   i− 
u  2
= u  2
+ ⋅ u (9)
2

Originally formulated to perform step by step solution of dynamic analysis, this


algorithm has been successfully used also to solve high non linearity problems in the
quasi static field, such as impact analysis, crushing materials, delamination process.
This algorithm ensures robustness, for it doesn’t imply any convergence iteration,
provided a sufficiently small time integration increment is chosen, preferably shorter
than a conservative estimation of a the stability limit showed by the equation (10),
where Le is the minimum dimension of a generic finite element, cd is the sound wave
speed in the material of the element and the minimum is established with respect to
the entire analyzed elastic domain.

L 
∆t cr = min e  (10)
 cd 

At certain conditions, this stability limit can cause strong computational efforts.
For example in our analysis the need to ensure quasi static conditions of analysis,
making possible to neglect inertial effects and the kinetic energy in the overall
energy balance, suggested a low rate of load application, about half the experimental
prescribed speed (0.2 mm in 15 seconds). This caused the elaboration of millions of
intermediate steps.

4.4 Results

4.4.1 The analysis

In the following the analysis results are presented. Four grades of concrete have
been taken into account from the experimentation performed in the labs of
DICAAR. Table 2 shows the nominal cubic strength Rck of the concretes, the
average cylindrical strength fcm as evaluated during the experimentation, the
correspondent theoretical tensile strength fctm according to Eurocode formulae and
the two different values for the Mode I fracture energy Gf, evaluated according to
Italian Guidelines for Structural Concrete (IG) [16] and CEB-MODEL CODE 2010
(MC), that for low grade concretes look quite different. Accordingly in Table 3 the
correspondent values of the displacements at zero stress u0 in the end of a linear
branch of strain softening are shown.

15
Given the uncertainties of fctm and Gf, that haven’t been experimentally evaluated,
and the sensitivity of the model to these material parameters, their values, especially
the latter, have been adopted as a mean for analysis calibration with respect to final
extraction force load.

Table 2: Main properties of the tested concretes.

Table 3: Ultimate displacement for the tested concretes according to different codes.

Figure 12. Principal compressive stress field before collapse

16
Figure 13. Typical Force vs Displacement curve till extraction of the anchor

Figure 14. Max principal strain at different load level till complete extraction

17
4.4.2 Discussion

The results of the analysis have been summarized in the above figures, showing the
stress field far before collapse in terms of principal minimum stress (Figure 12), a
typical Force vs Control Point Displacement provided by the analysis (Figure 13)
and a sequence of images showing the crack propagation at different load stages
until complete extraction of the anchor (Figure 14), in term of max in plane strain.
The last Figure points that element deletion option was also used, that is a complete
removal of the element at a certain crack opening level beyond u0.
The curve of Figure 13 reveals that nonlinearities due to crack onset and propagation
arise since the first load steps, but the extraction comes abruptly, starting at about
85% of final load, with a complete extraction of the anchor together with a conical
portion of concrete, as it was expected, as soon as the peak load is reached. The
model has no regard to local ductile modes of failure linked to compressive damage,
nevertheless it seems quite well capable to represent the extraction phenomenon
with a surprisingly good prediction of final load (see Table 4). As mentioned above
the model revealed to be highly sensitive to Tensile Strength and Fracture Energy
parameters. While the first has been assumed according to theoretical formulae the
latter has been used as a mean to apply some correction to the final results. For the
lowest grade concrete (concrete 1) the value of Gf given by Italian Guidelines gave
the best results, but moving to higher grade concretes (concrete 2 and 3) we
gradually switch to an intermediate value, than to the one proposed by the Model
Code 2010. For the highest grade concrete (concrete 4) we needed to adopt an
overestimated value of this parameter (about 60%) with respect to the formulae
prediction to gain the best fit to the experimental results.

Table 4. Final comparison between F.E. model and experimentation results

5 Conclusion
In the present paper the finite element modelling of the pull-out test, commonly
adopted as a mean to evaluate in situ concrete compressive strength, has been
presented. The analysis stemmed from an experimental campaign carried out in the
laboratories of DICAAR on properly made samples made of different grade
concretes. The need to get to a better understanding of the failure mechanism at the
basis of the test and its representativeness led to the discussed computational
models. Among different approaches for continuum models of Pull-Out test widely

18
discussed in the literature, we decided to discard any reference to compressive
damage or plasticity, emphasizing the role of crack onset and propagation due to
mode I NLFM in a continuum field, by means of a damage model capable of
detecting crack initiation and represents its propagation with a strain softening
behaviour with final complete loss of stiffness. Element deletion has also been
implemented for high values of crack opening. The model proved to be in good
agreement with experimental results: the complete extraction of the anchor occurs
with the removal of a conical concrete portion of expected geometry; the failure
develops abruptly; the F.E. model extraction loads are quite close to experimental
ones.
The proposed model is deeply sensitive to the fracture energy Gf parameter, which is
difficult to evaluate experimentally, while theoretical formulae generally don’t
agree, especially for low grade concretes. This parameter has been used to perform a
slight calibration of the model with respect to experimentation results, but further
research is advisable on the role of this parameter.

References
[1] B. G. Skramtajew, “Determining Concrete Strength for control of Concrete in
Structures”, ACI J., Vol. 34, 285-303, 1938.
[2] B. Tremper, “The measurement of concrete strength by embedded Pull-out
bars”, Proc. ASTM, 880-887, 1944.
[3] UNI EN 12504-3: 2005 - Testing concrete in structures - Part 3: Determination
of pull-out force.
[4] P. Bocca, “Extraction test: non destructive test of in situ evaluation of concrete
strenght”, Proc. of Istituto di Scienza delle costruzioni, Politecnico of Torino,
n. 375, 1977.
[5] A.J. Chaboski, D.W. Briden-Smith, “Assessing the strength of in situ Portland
cement concrete by internal fracture test”, Magazine of concrete Research, vol.
32, n. 112, 1980.
[6] P. Cossu, E. Pozzo, “Correlation between extraction load and compressive
strength of concrete in the Pull-Out method at variation of curing grade”,
CESMACOS, University of Cagliari, 1991.
[7] F. Meneghetti, T. Meneghetti, “Non destructive methods for compressive
mechanical concrete strenght evaluation, with particular reference to Pull-Out
method”, Proc. of CTE congress, Industrializzazione Edilizia, Firenze, 1984.
[8] W.C. Stone, N.J. Carino, C.P. Reeve, “Statistical Methods for In-place
Strength predictions by Pull-out test”, ACI J., Proc., 745-756, 1986.
[9] N.S. Ottosen, “A failure criterion for concrete”, J. Engrg. Mech., ASCE, No.
103, 1977.
[10] H.W. Peier, “Model for Pull-Out Strength of Anchors in Concrete”, Journal of
Structural Engineering, Vol. 109, No. 5, 1983.
[11] M. Yener, G.C. Li, “Progressive Finite Element Fracture Analysis of Pullout
Concrete”, Journal of Structural Engineering, Vol. 117, No. 8, 1991.
[12] G. Etse, “Finite element analysis of failure response behavior of anchor bolts
in concrete”, Nuclear Engineering and Design 179, 1998.

19
[13] A. Hilleborg, M. Modéer, and P.E Peterson, “Analysis of crack formation and
crack growth in concrete by means of fracture mechanics and finite elements”,
Cement and Concrete Research, 6:773-782, 1976.
[14] L. Bortolotti, “Strenght of Concrete Subjected to Pullout Load”, Journal of
materials in civil engineering, 2003.
[15] P. Pivonka, R. Lackner, H.A. Mang, “Concrete Subjected to Triaxial stress
States: Application to Pull-Out Analyses”, Journal of engineering mechanics,
2004.
[16] Guidelines for concrete structures and for evaluation of hardened concrete
mechanical parameters by mean of non destructive test, Consiglio Superiore
Dei Lavori Pubblici, February 2008.
[17] Abaqus 6.5, theory manual.

20

You might also like