Experimental Bond Characterisation and Inverse Numerical Stress-Slip Law Identification for Adhesive Anchor Applications

Abstract

In order to determine the adhesive anchors’ capacity under tensile loading, two test methods (confined and unconfined) are suggested in the guideline. Improvements for one of the two configurations are proposed and tested in this paper. The alternative setup currently being evaluated is a modified version of the one proposed by the RILEM technical recommendation, which is used to determine the bond properties of reinforced concrete elements. Finding a better and more consistent definition of the bonded length, removing the concrete surface contribution from the determined bond resistance, and testing the stability improvements were considered during the proposed setup development. In order to describe the adhesive anchor system’s behaviour, a stress-slip relation (law), which relates the local axial bond stress to the local axial slippage, can be used. Analytical and numerical procedures have been proposed in the literature in order to determine such a law. Here, an inverse calibration method based on Sequentially Linear Analysis (SLA) is applied in order to identify the stress-slip law of pull-out tests.

1. Introduction

Over the last few decades, the industry and researchers have been quite interested in investigating post-installed adhesive anchors for their business and for their safety relevance. Such anchors are used, among others, for rehabilitation and retrofitting purposes, and allow for a more flexible and sustainable design and construction. Steel and concrete are the materials typically involved in mechanical anchor systems [1]. In addition, for adhesive anchors, adhesive mortar is also included in the system (see Figure 1). The adhesive mortar layer transfers the load from the steel element to the concrete along the entire bonded length [2] via adhesion and interlocking [1].

In order to determine the tensile capacity of adhesive anchors, tests were carried out following the recommendation provided by the European Assessment Document (EAD [3]). The code distinguishes two methods: (i) unconfined tests, where the supports allow for an unrestricted formation of the concrete cone failure, and (ii) confined tests, which aim at eliminating the concrete cone failure by moving the test supports close to the anchor and assess the adhesion between the concrete and steel via the adhesive mortar layer. Figure 1a and Figure 1b show a visual representation of the unconfined and the confined test setup, respectively.

A steel plate with a center hole of diameter $1.5d_0<D<2d_0$ is used as a support in the confined configuration, where $d_0$ is the concrete borehole diameter. The supporting area should be designed such that the compressive stress under the support does not exceed $70\%$ of the concrete’s compressive strength. In order to avoid boundary conditions influencing the failure process, the support distance for the unconfined configuration has to be larger than four times the anchor’s embedment depth $h_{ef}$.

In order to avoid concrete cone failure, for high-bond-resistance adhesive anchors, confined configuration is recommended rather than unconfined. Steel failure needs to be avoided by reducing the embedment depth if necessary. If the resistance is obtained with a confined test, the result has to be converted to unconfined with the ${\alpha }_{set-up}$ factor [3].

Confined testing is also preferred to determine the resistance of the bonding layer for numerical simulations [4].

In order to design single-adhesive anchors under a tensile load in uncracked concrete, different models have been proposed in the literature [5,6]. The widely used uniform bond model defines the bond resistance as bond strength and assumes a constant stress along the bonded length [5,6]. Failure occurs when the bond stress reaches the bond strength. Equation (1) provides the adhesive anchor tensile capacity $N_b$, based on its embedment depth $h_{ef}$, on the borehole diameter $d_0$ and on the bond strength ${\tau }_o$. The anchor diameter d, together with the related bond strength $\tau$, can be used alternatively in the equation. According to Cook et al. [6], the uniform bond model is the one that provides the best fit to the large database considered.

$$N_b={\tau }_o{d}_0\pi h_{ef}$$

(1)

The model was shown to be a valid behavioural model, both experimentally and numerically, by Cook et al. [7], if $d_0<1.5d$ and if $4<h_{ef}/d<20$. Cook et al. [6] showed that there is a slight improvement in the assessment when using the anchor diameter as opposed to the borehole diameter, but the results are not conclusive. Within this paper, the anchor diameter d is used for bond strength determination.

The adhesive mortar, during installation, positions itself as a thin layer in the annular gap between concrete and steel. This leads to the development of two interfaces, the “inner” and the “outer” one (see Figure 2a). The so-called outer interface is the physical interface between the concrete and adhesive mortar. This physical interface’s properties are affected by many factors, such as the capability of the adhesive mortar to penetrate into the concrete and into the concrete micro-cracks, the borehole roughness [8,9], the possible chemical interaction between the adhesive mortar and concrete, and the possible presence of residual dust in the borehole. The so-called inner interface is not a physical one; it is, rather, a crack propagation path dictated by the threaded bar geometry. Since the inner interface failure happens through the adhesive mortar material, it could be considered an adhesive mortar Mode II problem only.

Figure 2b shows a possible failure mechanism of adhesive anchors tested in a confined configuration, where failure involved both the inner and the outer interfaces. In an unconfined configuration, several failure mechanisms under tensile loading can be obtained [6]. Figure 3a and Figure 3b show the steel failure and the concrete splitting failure, respectively (failure mechanisms also shared by mechanical anchors [10]). If the embedment depth is short, concrete cone failure (or concrete breakout) can be expected (see Figure 3c). The mixed-mode failure begins as a concrete cone failure and continues through the anchor’s outer (see Figure 3d) or inner (see Figure 3e) interface (see [4]). In the former case, the failure can either keep progressing along the outer physical interface until the end of the anchor, or it can cross the adhesive mortar layer and continue along the inner interface until the end of the anchor (see Figure 3f).

In this contribution, a new test setup that can be used to obtain the adhesive anchor bond properties in a confined configuration is presented.

The test setup aims at improving the reproducibility of the bond strength determination, especially for mortars with a very high bond strength. These mortars have to be qualified in a confined configuration to avoid concrete cone failure, with a short embedment depth to avoid steel failure. In this case, the concrete contribution to the anchor failure mechanism and failure load may become non-negligible due to the support used for the confined test and the concrete surface influence.

The test setup also aims at improving the stability of the test, allowing for a consistent post-peak characterisation by controlling the test with the unloaded-end anchor displacement as done for cast-in rebars. This point is important to facilitate the stress-slip law determination with the proposed numerical method.

The proposed setup is an adaptation of the one proposed from RILEM technical recommendations [11] that is used to identify the bond properties of reinforcing steel elements embedded in concrete. The method used for reinforcing steel elements embedded in concrete (cast-in rebars) is suitable for the identification of the bond properties of a system with a unique interface (steel–concrete). Post-installed anchors present two interfaces, and the installation process requires modifications in order to expose the unloaded end of the anchor. In addition, a novel method used for the stress-slip law inverse identification is proposed and applied on a sub-set of the experimental results. The stress-slip law is a relation between anchor axial bond stress and axial slippage that can be used for adhesive anchor modelling purposes.

2. Proposed Test Setup

During the development of a product, it is important to reduce the impact of the below-described sources of uncertainty to be able to observe even small variations in product performance. The support plate used for the confined tests allows for the development of a shallow concrete cone, which may lower the resulting anchor performance. The concrete cone size is also dependent on the inner ring diameter, which can vary within the above-mentioned range, introducing an additional source of uncertainty.

The bond resistance is not always enough to characterise the anchor behaviour, especially if the anchor is modelled numerically. In the case of numerical models using an interface to connect steel and concrete [4], a stress-slip law can be prescribed as interface constitutive law. In this case, the outer and inner interfaces, as well as the mortar annular gap, are merged into a single interface. Also, other systems which operate with adhesion, such as fiber-reinforced polymer strips (FRP) [12], fiber-reinforced cementitious matrix (FRCM) [13], or reinforcement in concrete [14], can be represented by an interface which uses as material constitutive model a stress-slip law. In order to properly identify such a relationship, confined tests should be carried out until full softening occurs. This leads to the need to perform stable pull-out tests since unstable tests typically lead to premature uncontrolled failure [15], which deliver only partial information. The Proportional Integral Derivative (PID) controller should be able to control the tests in the post-peak domain. In order to solve this problem, the displacement measured as closely as possible to the concrete surface on the anchor’s loaded end, or, ideally, the anchor’s unloaded end, should be used as feedback signal. This limits the influence of elastically stored energy which is released during fracture and otherwise causes a snap-back instability.

The bond properties’ identification is addressed differently in the case of concrete reinforcement (rebar). RILEM technical recommendations [11,16] proposed a test configuration which can be used to obtain the bond properties of non-pre-stressed reinforcing steel element. In Section 2.2, an adaptation of the RILEM setup for adhesive anchor testing is described.

2.1. RILEM Standard for Bond Properties Identification

In order to obtain bond properties of reinforcing elements used in reinforced concrete, RILEM recommends a test setup which is shown in Figure 1c. The specimen used is a concrete cube with a minimum edge length of 200 mm. A reinforcing bar is placed in the mould before the casting of the concrete, which, therefore, hardens around it. The load is applied on the longer reinforcing element side, while the displacement is measured on its unloaded end. A certain unbonded length at the loaded end of the reinforcing bar is required by the standard in order to avoid damage close to the concrete surface. Moving away from the concrete surface, the beginning of the failure avoids the development of the shallow concrete cone, which may influence the experimental results. Even though concrete reinforcement and adhesive anchors share similarities in geometry and working principles, there are two fundamental differences between them: the number of interfaces and the installation procedure. These two differences limit the direct applicability of the RILEM test setup for adhesive anchors; thus, a modified setup is required.

2.2. Alternative Test Setup for the Identification of Bond Properties

For the present experimental campaign, unreinforced concrete slabs of sizes $300\times 300\times 150$ mm were cast. The threaded bars used for the system were M12 class 12.9, which required a borehole diameter of 14 mm. In order to be able to control the test with the adhesive anchor unloaded end, a hole needs to be drilled through the concrete specimen. In general, during the hammer-drilling procedure, the concrete toward the other end fails with a concrete breakout of unknown geometry.

Such breakout needs to be avoided or controlled in order to ensure the repeatability of the tests. One of the methods which can be used to control the concrete breakout is to force it consistently to localise on a notch. In order to create such a notch, a wooden cylinder of dimensions $35\times 40$ mm was placed in each one of the moulds before the concrete casting, creating a cylindrical void after demoulding. The concrete breakout localises at this void during the hammer drilling (see Figure 4a). Once the hole is drilled and properly cleaned, the lower end of the borehole needs to be sealed in order to avoid leakage of adhesive mortar during the anchor installation (see Figure 4b). Additionally, plastilin is placed in the notch to facilitate the removal of the excessive adhesive mortar that would otherwise fill the notch. After the hole is sealed, the anchor is installed and left to cure for the required time according to the manufacturer product installation instruction. Once the adhesive mortar hardens, the lower part of the slab is cleaned from the excessive adhesive mortar (see Figure 4c) in order to expose the threaded bar’s unloaded end. A specially designed LVDT holder is coupled with the lower end of the threaded bar, which was previously milled to obtain a precise sliding coupling with the holder and to minimise the bond with the mortar. This procedure was chosen to measure the lower end displacement along the same axes of the threaded bar. The holder is then fastened in the correct position with a rapid-hardening adhesive mortar. A thin layer of plastilin is introduced to isolate the two adhesive mortars from each other (see Figure 4d). The applied testing speed was 0.001 mm/s for all tests.

Figure 5 shows the experimental setup used for the testing. The specimen is positioned on a wooden platform that allows access to the unloaded end of the anchor, in order to setup the displacement transducer. The anchor displacement was measured at the loaded end and at the machine stroke, and measured and controlled at the anchor’s unloaded end. At the unloaded end, a single LVDT was used, while at the loaded end, three LVDTs were placed in a 120° configuration to record the pull-out displacement at 60 mm from the concrete surface. A cardan joint is used to connect the specimen to the machine in order remove moments. The two beams that hold the supporting plate are connected with spherical couplers to threaded bars, which are themselves connected to the base of the machine.

Three configurations of debonding methods and support geometry were tested in this experimental campaign, as shown in Figure 6. Additionally, two configurations were used as controls. In the first one (see Figure 6a), in order to unbond the threaded bar from the concrete, a 30 mm portion of threaded bar close to the concrete surface was wrapped with paper tape and Teflon tape. This configuration is referred to as Teflon. The paper tape was used to create a smooth surface over the ribs of the threaded bar, while the Teflon tape was used for its poor adhesion with the adhesive mortar. This method, however, forces the failure to start and progress along the inner interface. This may influence the obtained result in cases where the failure would otherwise initially involve the outer interface as well. In order to identify the influence of such unbonding layer on the measured capacity, a control test was performed on an anchor entirely wrapped with paper tape and Teflon tape (see Figure 6d). In the second configuration (see Figure 6b), the forcing of the failure to start at the inner interface is solved by unbonding the threaded bar from the concrete with a 30 mm deep notch cored across the outer interface with a coring bit. This configuration is referred to as Cored. The third configuration (see Figure 6c) was not unbonded and was considered as the Reference case and used to compare the results of the proposed configurations with the results of the guideline configuration.

Precise bonded length was ensured by milling the thread of the excessive section of the bar. As previously mentioned, milling the threaded bar surface also allows for the coupling of the anchor to the LVDT holder. The bond between the milled section of the bar and the adhesive mortar is avoided by a sleeve made of paper and duct tape that surrounds the milled section. In order to identify the influence of the sleeve on the measured capacity, another control test was performed on an anchor entirely milled and covered by such a sleeve (see Figure 6e). The supporting ring used for the Cored and Teflon configurations pull-out had an inner diameter of 36 mm, while for the other configurations, the support inner diameter was 24 mm, which is within the range prescribed by the guideline. Bonded lengths of 30 and 40 mm were tested for the three configurations. Note that both bonded lengths are not in accordance with the guideline since they are below 40 mm and four times the anchor diameter. However, in this comparative study, due to the capacity of the anchor testing machine limitations, these two dimensions were chosen. Nevertheless, 30–40 mm is much larger than the characteristic length of the studied problem, i.e., thread spacing, annular gap, or size of the filler inside the adhesive. Note that the concrete coarse aggregates and cone break-out are not contributing to the studied interface failure. Four repetitions were performed to estimate mean value and standard deviation of the obtained bond strength and to determine the variability of the load displacement curves for each one of the three configurations.

3. Materials Description

In this chapter, the materials involved in the anchoring system evaluated in this study are described.

3.1. Epoxy Based Mortar

The adhesive mortar used in this experimental campaign is a compound composed of a polymer matrix and fillers. Additional elements are also included to enhance the product’s properties. The adhesive mortar is based on a bisphenol-A/F epoxy resin with an amine hardener. The matrix is filled with a relatively high amount of inorganic broken shaped fillers. Since, for adhesive anchor applications, the mortar is injected and cured in situ, different adhesive mortar curing degrees can be expected. Therefore, the properties of the bonding material, and thus of the bonding interface, may vary. The time-dependent change in the mechanical properties of this material has been investigated by Singer et al. (see [17]). An additional study has been performed on the long-term creep behaviour of adhesive mortar specimens. Tensile and shear master creep curves were determined (see [18] and [19], respectively).

3.2. Concrete Properties

Since concrete is the base material for the bonded anchor system used in this study, a comprehensive characterisation of its mechanical properties was performed at 28 and 70 days.

Table 1. Mixed design of the analysed concrete.

Water content	173.2 kg/m3	Aggregate type	Quartz
Cement content	275.0 kg/m3	Aggregate shape	Round
Aggregates content	1976.1 kg/m3	Max. aggregate size	16 mm

shows the concrete mix design.

The obtained properties at these ages are Brazilian splitting indirect tensile strength $f_{t,B}$, cube compressive strength $f_{cu}$, cylinder compressive strength $f_{cy}$, elastic modulus E, and fracture energy $G_f$. In addition, at 28 days, cores were taken from spare specimens in order to obtain the cored cylinder compressive strength $f_{cy}^{\ast }$. Anchor tests and additional compression tests on cored cylinders were performed at the concrete age of 400 days. Eurocode or modelcode ageing functions [14,20] were used to estimate the aforementioned concrete properties at 400 days of concrete age. The ageing function parameter s is a coefficient that depends on the cement class. According to the Eurocode, $s=0.2$ for the cement type used in the current investigation. The s parameter was also fitted ($s=0.32$) to the cube compressive strength results at 28 and 70 days and used for the estimation of the concrete properties at 400 days of concrete age.

Table 2. Summary of concrete properties for different specimens and ages. Predictions made according to Eurocode ageing functions using cement-related s parameter (ECa) and using s parameter calibrated ($s=0.32$) on cubes’ results (ECb).

Prop.	Unit	28 Days	70 Days	400 Days	ECa	ECb
$f_{cu}$	MPa	$34.3\pm 6.2\%$	$38.7\pm 7.6\%$	-	$39.7$	$43.4$
$f_{cy}$	MPa	$32.3\pm 7.7\%$	$36.5\pm 2.9\%$	-	$37.4$	$40.9$
E	GPa	$30.7\pm 8.1\%$	$31.2\pm 7.5\%$	-	$32.1$	$33.0$
$f_{t,B}$	MPa	$3.3\pm 9.8\%$	$3.4\pm 8.6\%$	-	$3.64$	$3.86$
$G_f$	N/m	$105.8\pm 14.7\%$	$117.8\pm 7.3\%$	-	-	-
$f_{cy}^{\ast }$	MPa	$26.1\pm 8.0\%$	-	$34.0\pm 6.5\%$	$30.2$	$33.0$

summarises the concrete properties obtained and estimated for the considered concrete ages. The cored cylinder compressive strength development was used to choose which of the two s parameters is more suitable for the properties’ estimation. The fitted coefficient provides a 400-day $f_{cy}^{\ast }$, which is closer to the experimental result than the one calculated using the s parameter obtained from the Eurocode.

4. Experimental Results

Tests were performed to identify the influence of the unbonding methods (sleeve, paper tape, and Teflon tape) on the measured bond properties. The results show that their influence on bond strength can be considered negligible, being at most around 0.3% of the bond strength obtained from the Cored type. The failure mechanisms resulting from the pull-out tests performed on the Reference, the Teflon, and the Cored specimens are shown in Figure 7a, Figure 7b, and Figure 7c, respectively, for both tested bonded lengths.

Figure 7a shows the failure mechanisms of the Reference anchors having 30 mm and the 40 mm of unbonded length. For the Reference configuration, since the embedment depth is small, the concrete cone developed during the pull-out tests cannot be neglected since it covers a relevant fraction of the bonded length. This is reflected in a strong reduction of the calculated bond strength. Thus, the method suggested in the guideline is conservative and may hide a reserve of capacity which is bigger when smaller embedment depths are considered.

Figure 7b shows the Teflon configuration failure. Since the unbonding layer behaves as a crack located on the threads of the threaded bar, the failure starts and propagates along the inner interface until the anchor’s end. For the Cored configuration specimens, the failure does not start at the inner interface but rather at the outer one. The failure immediately crosses the adhesive mortar layer since the aged concrete is too strong, and the outer physical interface is stronger than the inner one. This is confirmed by the Reference configuration, where the failure starts at the concrete surface as the concrete cone and progresses through the mortar layer to the inner interface without the outer interface’s involvement. In this case, little difference can be expected between the Cored and Teflon cases since, for both of them, the failure mostly involves just the inner interface. Figure 8 shows the experimental results for the three configurations. For each one of the three configurations, the results of the tests performed on both bonded lengths are plotted in the same diagram. The results are also presented in

Table 3. Bond strength results for the three configurations obtained using the uniform bond model.

	Cored [MPa]	Teflon [MPa]	Reference [MPa]
30 mm	40.60 ± 2.2%	39.89 ± 2.7%	28.03 ± 7.0%
40 mm	37.87 ± 1.9%	38.28 ± 6.9%	31.12 ± 9.3%

in terms of the bond strength obtained by assuming the uniform bond model.

As can be seen in Figure 8a,b, both the Cored and Teflon configurations lead to the same peak load and, consequently, to the same bond strength. The Cored configuration, however, shows very consistent results, having a smaller scatter both in terms of peak load and stiffness. Additionally, the post-peak behaviour of the tests performed on the Cored configuration is very consistent compared with the remaining test results. This can be explained by the fact that the cored section excludes contact and, therefore, excludes possible friction within the system. The pull-out tests on the Reference specimens, since they were affected by the mixed-mode failure, exhibit a kink in the loading branch due to the concrete cone that is more pronounced for the 30 mm anchors in comparison with the 40 mm anchors. The cone development additionally leads to a strong bond strength reduction in comparison with the Teflon and Cored cases.

After the verification of the improvement on the bond properties’ determination, the stability improvement is also verified. The test stability and test controllability are dependent on the system stiffness and on the machine PID control settings. The PID controller calculates the difference between the prescribed process variable value and the measured one. This is defined as an error and is continuously calculated. The aim of the controller is to minimise the error over time. More information about the PID and its numerical implementation in an explicit solver can be found in [21].

Figure 9 shows one of the pull-out tests performed in the current experimental campaign showing the recorded displacement from three different points along the threaded bar. Three curves are plotted to explain the improvement of stability and controllability brought on by the proposed setup. The dashed blue line represents the unloaded end displacement, the red dash-dotted line stands for the loaded end displacement, and the solid green line is the stroke displacement. As previously mentioned, all the tests were controlled with the unloaded end LVDT displacement. The loaded and unloaded end displacement curves are stable and controllable because they exhibit no snap-back. The remaining curve, the stroke displacement curve, is a classical example of a curve affected by snap-back, which makes the test unstable. In stroke displacement control, such a situation would lead to a failure at the snap-back point. However, it may still be possible to control this test if the proper PID setting is chosen. In this case, the test would be unstable but controllable. Such a situation leads to a dynamic vertical force drop due to the motion of the structure, except for at the loading point [22]. This phenomenon may lead to a load-displacement curve that is locally influenced by the machine’s attempt to keep the test running and does not reflect the proper specimen behaviour.

5. Inverse Stress-Slip Law Identification

The method utilised in this paper to inversely identify the stress-slip law of adhesive anchors is based on the Sequentially Linear Analysis (SLA) [23,24], which was specifically developed to overcome the difficulty of snap-back behaviour [25,26]. This method aims at identifying the stress-slip law; no assumptions are needed in terms of a tensile strength and a functional form since the procedure directly outputs the determined stress-slip relation. Figure 10a shows the mesh of the model used for the stress-slip law identification. Four-node quadrilateral axisymmetric finite elements and interface elements with a linear approximation of the displacement jump were utilised. The damage is assumed to be localised in zero-thickness cohesive interface elements [27]. The material outside the potential crack paths is assumed to be linear-elastic.

In order to use this method, the experimentally measured load-displacement curve needs to be known. In each step, a reference load L, which needs to be larger than the maximum load observed experimentally, is applied in the model. Since a sequence of linear (secant) steps is assumed, the response is a straight line. The critical load multiplier $\lambda$ is then obtained as the minimum of (i) the intersection of the calculated and experimental load-displacement curves, the point A in Figure 10b, or (ii) the value by which the normal stress, induced in any integration point by the reference load L, would have to be scaled to become equal to its current critical strength, given by the already determined TS diagram, point B in Figure 10c. Once the critical multiplier is determined, the load L, stress, and strain fields are scaled. In case (i), the extension of stress-slip relation is recorded. Then, the stiffness in the critical element (element in which the minimum $\lambda$ was found) is decreased, and the process continues to identify the next points in the stress-slip relation. Note that the initial stiffness of the computational model has to be the same as the experimentally measured stiffness. For more details regarding the inverse analysis, the interested reader is referred to [23,24].

The experimental data obtained for Cored specimens of bonded lengths 30 and 40 mm are further employed since no damage in concrete is assumed during the inverse analysis. Figure 11a shows the results of the inverse analysis compared with the one (a cored configuration specimen with 30 mm of embedment depth) of the experimentally obtained load-displacement curves. Furthermore, the smoothed interface stress-slip relations and their means are depicted in Figure 11b. The ideal stress-slip law would be the same for all concrete and geometrical parameters. The stress-slip laws obtained for the cored configuration, with an embedment depth of 30 mm (C-30 in Figure 11b) and 40 mm (C-40 in Figure 11b), are very similar to each other, especially in the post-peak part. Using the Cored configuration and controlling the test with the displacement at the unloaded end made it possible to obtain a precise softening behaviour.

6. Conclusions

The proposed approach allows for the accurate determination of the adhesive anchor bond properties by improving the experimental procedure and numerical determination of the stress-slip law of the bond connection. The guideline testing approach used for qualification is revealed to be on the conservative side. However, for development purposes, when small product improvements need to be identified, the proposed approach can be used for its capability of reducing the experimental scatter and allowing for stable testing until full softening occurs.

This method enables access to the unloaded end displacement of post-installed anchors, which offers a more complete view of the anchor behaviour during the pull-out. Additionally, controlling the tests with the anchor’s unloaded end displacement substantially reduces the chance of unstable tests, ensuring the accurate system response acquisition necessary to determine the complete stress-slip law.

Pull-out tests on reference specimens exhibit mixed-mode failure with a pronounced kink in the loading branch. The concrete cone development reduces bond strength in Reference cases compared with configurations where the top part of the anchor is debonded. The proposed so-called “Cored” configuration consistently shows smaller scatter in terms of the peak load and stiffness. Post-peak behaviour in Cored tests is also highly consistent compared with other configurations.

As a debonding method, it removes the influence of the concrete surface, not forcing the failure at the inner interface as the “Teflon” configuration does. Moving the failure away from the concrete surface precludes the development of a shallow concrete cone that can influence the acquired bond properties. Two configurations were proposed: one forces the inner interface to fail (Teflon), while the other leads the weaker interface to fail (Cored). In addition, the inner diameter of the supporting ring has limited influence on the failure. A larger inner diameter can be chosen in order to limit the compressive stress parallel to the interface, which is higher for smaller ring sizes.

The proposed method for the inverse identification of the stress-slip relation from the experimental results provides consistent estimates for different bonded lengths.