**1. Introduction**

Glued-in rod (GiR) or bonded-in rod (BiR) connections are increasingly used in construction of timber structures [1], and so far several researchers addressed the mechanical performance of specific solutions of technical use in buildings.

Owning to their versatility, BiR connections are used extensively, and thus the need for proper assessment of their mechanical properties and standardized assembling procedures is ever increasing [2,3]. Research efforts have been spent to offer an accurate detail on BiR connections behaviour, but mainly for limited applications that can be hardly generalized. In this framework, most of the literature involving experimental studies have been focused on the axial pull-out strength of a single BiR connection, and its dependency on geometrical and material parameters. Examples can be found in [4–10] for various configurations, with a focus on the experimental assessment of various failure mechanisms [4], test protocols [5] or monotonic and cyclic loading [8]. Often, Finite Element numerical modelling techniques are applied to bonded joints in timber engineering [11–14]. Various experimental studies have been carried out with the additional goal of proposition and validation, as well as assessment of existing methods, of empirical formulations in support of design [15–17], based on curve-fitting of experimental outcomes. In this regard, the current study further explores the mechanical behaviour and properties of BiR connections for timber applications, but with a special focus on the effects due to different adhesive types and their operational

**Citation:** Barbali´c, J.; Rajˇci´c, V.; Bedon, C.; Budzik, M.K. Short-Term Analysis of Adhesive Types and Bonding Mistakes on Bonded-in-Rod (BiR) Connections for Timber Structures. *Appl. Sci.* **2021**, *11*, 2665. https://doi.org/10.3390/app11062665

Academic Editor: Tore Brinck

Received: 16 February 2021 Accepted: 15 March 2021 Published: 17 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

condition. It is nowadays well recognized that both the environment conditions and the loading configuration severely affect the adhesive properties and thus the mechanical performance of BiR joints [18–20]. Further, the current investigation aims at finding a link between the proposed fracture mechanics failure modes [21] for BiR connections and the impact of adhesive sensitivity to service conditions.

In this paper, original pull-out experiments are carried out on a total of 84 BiR specimens, characterized by different adhesive types (epoxy or polyurethane glue), rod-to-grain arrangements (parallel or perpendicular), and average moisture content in timber (9%, 18% or 27% respectively, see also Table 1). The experimental results are discussed, with a focus on the load-bearing mechanism, fracture mechanisms and BiR performance analysis using simple empirical formulations of literature. Later on, a more refined analytical model is presented, and further assessed against the available experimental data.

**Table 1.** Service classes for timber structures and typical examples.


#### **2. Problem Definition**

Connections and reinforcements with bonded-in rods have been used in building for several decades. For instance, this solution appeared for the first time in 1980, in French historical monuments [22]. Besides, adhesive bonds for timber applications are notoriously sensitive to several aspects, including:


(Point 1) relates to the substrate (type of surface, its treatment, any kind of ageing and chemical modification, etc.), and to the adhesive in use (viscosity, density, chemical affinity with the substrate, etc.). (Point 2) is particularly relevant when the thickness is high (as usual for structural applications in timber buildings). Finally, environment conditions in 3) can include elevated thermal distortions (due to fires or repeated hygrometric variations), such as wood deformations that induce additional coactive stress at the interface between the adhesive and timber.

From a practical point of view, the applied bond strength design values and the explicit strength modification factors are in most cases not retrievable. This is closely related to the fact that the current lack of worldwide standards or commonly accepted specifications exist for assessing and approving adhesives to be used for BiR applications. It is obvious from a chemistry view point that different adhesive types, which may have rather similar shortterm bond strengths, can behave differently under variable climates. The design of BiR joints is implemented in European prestandards and technical documents [23–25], which specify the modification factors for accumulated duration of load in different climates (service classes in Table 1, as described in [26]). The final result takes the form of the well-known *kmod* factor. On the other hand, this factor is irrespective of the adhesive type.

The current investigation presents experimental tests of BiR connections in different humid climates, in order to explore their actual load-bearing capacity (strength and slip modulus, failure mechanism), as a function of two different adhesive types and rod arrangements.

#### **3. Experimental Investigation**

#### *3.1. Test Specimens and Materials*

In order to gain a better insight into the behaviour of the joint, an extended series of experimental tests was carried out. In total, 84 specimens were taken into account in the laboratory investigation, with 72 "half-size" and 12 "standard" specimens. Among the half-size specimens, 36 samples were tested parallel and 36 perpendicular to the grains of timber. Furthermore, 12 full-size specimens tested to confirm the observed correlation between half-size and standard specimens (Figure 1).

**Figure 1.** Preparation of bonded-in rod (BiR) connections with M10 steel rod (examples of polyurethane adhesive bonding): (**a**) half-size and (**b**) full-size specimens.

Half-size specimens (with dimensions *B* = 120/*W* = 60/*L* = 60 mm) were drilled in their full height *L* with a concave-notch diameter equal to *d<sup>h</sup>* = 14 mm. This hole was placed at the centre of the cross-section of each timber log, in order to introduce both the rod and the adhesive bond. Standard type specimens were characterized by double size (with dimensions *B* = 120/*W* = 120/*L* = 60 mm) and prepared with a similar approach. Each timber log was drilled in the full height (60 mm), and a *d<sup>h</sup>* = 14 mm hole was drilled at the centre of the wood cross-area.

To get the results for the highest service class, Siberian larch (*Larix sibirica*) wood was used for the timber components [27]. After three weeks in a climate enclosure room with a controlled atmosphere, the moisture content was around 12% (Figure 2). The measured average density was close to 600 kg/m<sup>3</sup> , with a standard deviation of 25 kg/m<sup>3</sup> .

**Figure 2.** Preparation of specimens under controlled atmosphere.

One standard metrically threated steel rod with 8.8 strength, nominal diameter *d* = 10 mm and total length of 200 mm was bonded in each wooden specimen (*L* = 60 mm the bonded length). The bonding effect was investigated with two different adhesive types, being represented by a two-component epoxy (KGK EPOCON '88) and a two-component polyurethane (LOCTITE PUREBOND CR 821). Table 2 summarizes the nominal mechanical properties of materials.

**Table 2.** Nominal mechanical properties for KGK EPOCON '88 (two-component epoxy), LOCTITE PUREBOND CR 821 (two-component polyurethane) and Siberian larch wood (*Larix sibirica*).


Based on the M10 rod in use and the *d<sup>h</sup>* mm hole, an annular bond-line thickness of 2 mm was created for each sample, and the anchorage length was set equal to *H* = 60 mm. The bonding stage was performed under controlled laboratory conditions (9% humidity in wood and a room temperature of 20 ◦C, see Figure 1).

#### *3.2. Test Setup and Instruments*

After the assembly process, all the specimens were subjected to a controlled atmosphere, so as to achieve different degrees of moisture in wood (at the same temperature of 20 ◦C). The experiments reported herein comprised three artificial climates being selected as extreme examples of operational conditions for service class 1, 2 and 3 (Table 1). All test series were in fact performed at a constant temperature of 20 ◦C, with moisture content in wood in the order of 9%, 18% and 27% respectively.

The stiffness and strength characteristics in short-term loading were thus investigated according to EN 15274:2015 recommendations [28]. Further, all tests were carried out based on the EN1382:2016 provisions [29], on a Zwick Roell 50 kN capacity machine, with data recording frequency of 10 Hz. The reference pull-out setup is shown in Figure 3. Each specimen was fixed to the machine with a steel clamping plate. Possible relative displacements of wood logs were restrained by four M8 anchoring bolts. The single rod was hence clamped to the pull machine. To this aim, the nominal cross-head displacement rate was set to ensure a reference value of 0.5–2.0 mm/min (depending on specimen type). The latter was then calibrated (test by test) in order to reproduce a short-term failure mechanism for all the specimens. The axial load *F* applied to each specimen was recorded and compared with the average relative displacement of the rod with respect to the wood log. In support of these experimental investigations, contactless optical measurements of strain (based on digital image correlation (DIC) techniques), were also implemented to provide full-field strain maps of specimens under load, until failure. The experiments of 24 specimens were further supported by a Canon 700D camera with macro lens and photo recording frequency of 0.5 Hz (Figure 3). All specimens had preprepared surface adapted for recording. The whole postprocessing stage of acquired images was carried out by VIC-2D (Correlated Solutions, University Santa Barbara, Santa Barbara, CA, USA). This approach was used to reveal local aspects of load transfer mechanisms from the rod to the adhesive and wood that could be of importance for further analysis using theoretical or even numerical models.

**Figure 3.** Pull-out test set-up: schematic view and details (nominal dimension in mm).

#### *3.3. Test Results*

The analysis of experimental results was carried out at different levels, including (a) qualitative analysis of observed failure mechanisms, (b) measured load-displacement laws, (c) DIC measurement of displacements in the bonded region.

For sake of clarity, the specimens are labelled to detect the type of glue ("E" or "P" for epoxy and polyurethane), the moisture content (9%, 18% or 27%), the rod-to-grain orientation ("0◦" or "90◦" for parallel or perpendicular arrangement) and the sample number *n* for each group. The full set of pull-out test results is presented in Appendix A.

Three regimes can be distinguished from the collected load-displacement curves, as shown in the examples of Figures 4 and 5 (specimens under 9% moisture and load parallel or perpendicular to the grain, respectively).

Firstly, a linear elastic stage can be noticed in the load-bearing response of all the specimens, from which the initial stiffness *Kser* = *Fax/d* can be estimated from a linear regression procedure.

After the yield point, a progressive decrease of stiffness occurs followed by a sudden failure of the connection. Worth to be noted, in this regard, that the failure path of all tested connections was located in the wood substrate in the vicinity of the wood–adhesive interface, as also emphasized in Figure 6. Thus, the nonlinearity observed in the collected force–displacement responses can be rationally justified in the quasibrittle damage of wood.

**Figure 4.** Experimental force–displacement results for specimens with (**a**) epoxy or (**b**) polyurethane glue, under 9% moisture and parallel rod-to-grain arrangement.

**Figure 5.** Experimental force–displacement results for specimens with (**a**) epoxy or (**b**) polyurethane glue, under 9% moisture and perpendicular rod-to-grain arrangement.

Finally, the ultimate load measurement for the experimental curves as in Figures 4 and 5 allows estimating the overall shear strength of the examined connections. Following Figure 6 and that the failure mechanism of BiRs is dependent on the mechanical properties of solid wood (strength and stiffness), the typical collapse of BiRs can be assumed as quasi-brittle for general applications.

More in detail, for rods bonded parallel to the grains, wood failure was observed to start in the area around the adhesive matrix, while for rods bonded perpendicularly to the grains, failure typically originated in line between adhesive matrix and wood (Figure 6). The qualitative observations at failure, for a selection of specimens, were further explored by DIC system as in Figure 7.

**Figure 6.** Example of failure configuration for the tested samples: (**a**) epoxy and (**b**) polyurethane bonded rods for half-size specimens (18% moisture, parallel rod-to-grain arrangement) and (**c**) full-size specimens.

**Figure 7.** Example of the typical shear strain distribution along the rod, as obtained by digital image correlation (DIC).

Average values of maximum force (*Fmax*) and displacement (*dmax*), as well as of slip modulus (shear stiffness *Kser*) and their corresponding coefficient of variation (CoV.) and standard deviation (St.Dev.) are listed in Tables 3 and 4. Major scatter of grouped predictions is found for the "9%" set in Table 4, which was found characterized by 36% CoV. in terms of slip modulus. On the other side, such an experimental outcome was severely affected by few specimens (like specimen #2 in Figure 5a).


**Table 3.** Maximum axial force, displacement and slip modulus for specimens with parallel rod-to-grain arrangement (mean experimental values).

**Table 4.** Maximum axial force, displacement and slip modulus for specimens with perpendicular rod-to-grain arrangement (mean experimental values).


Considering the specimens grouped by adhesive type, from Tables 3 and 4 it is possible to notice a less pronounced sensitivity and scatter of polyurethane bonded rods, compared to the epoxy bonded samples. This effect is even more pronounced for service classes 2 and 3, with higher moisture content.

Mean experimental data can be helpful for the analysis of climate and operational conditions on bonded rods for timber applications. However, an in-depth discussion can be carried out in terms of characteristic mechanical properties that can be obtained from the test observations.

Based on [28], the characteristic axial force at failure for each series of specimens was calculated as:

$$F\_{ax, char} = \exp\left(\overline{y} - k\_s s\_y\right) \tag{1}$$

with:

$$\overline{y}^\* = \frac{1}{n} \sum\_{i=1}^n \ln F\_{\text{ax},i} \tag{2}$$

$$s\_y = \sqrt{\frac{1}{n-1} \sum\_{i=1}^{n} \left( \ln F\_{\text{ax}} - \overline{y} \right)^2} \tag{3}$$

where *n* denotes the number of test repetitions for each series, *k<sup>s</sup>* is a coefficient adopted from [28].

Furthermore, the maximum axial force at failure (both in terms of mean and characteristic values) was correlated with the resisting surface of the bond-line of each specimen, *Abond*, given that:

$$
\sigma\_{\text{max}} = \frac{F\_{\text{ax}}}{A\_{\text{bond}}} = \frac{F\_{\text{ax}}}{0.5 \,\pi d\_h L} \tag{4}
$$

where:

$$A\_{bond} = 0.5 \,\,\pi d\_{\text{lt}} L \tag{5}$$

for the half-size specimens.

The estimated results are shown in Figure 8, in terms of stress peak at failure for each set of specimens (under the assumption of uniform stress distribution for the bond-line as a whole). Mean and characteristic values of ultimate stress are grouped by adhesive type and rod arrangement, as a function of the service class/moisture content.

**Figure 8.** Stress peak at failure, as observed for (**a**) epoxy or (**b**) polyurethane bonded rods in various arrangements. Comparison of mean and characteristic experimental results.

As expected from Tables 3 and 4, a major scatter of mean and characteristic values was observed especially for the epoxy-bonded rods, rather than polyurethane samples. Besides, the global decrease of stress peak can be observed, both in mean and characteristic parameters, as far as the moisture level increases. This is a further confirmation of sensitivity of different adhesive types to operational conditions.

In this regard, it is worth mentioning that the imposed displacement rate (in the range of 0.5–2.0 mm/min), as previously discussed, was adapted test by test. The postprocessing stage of experimental measurements was quantified in average rate values that are summarized in Figure 9, as obtained for each series of specimens. Worth noting are the lower rate values for epoxy or polyurethane specimens under high moisture (27%) and bonding parallel to the grain. This was required by the pronounced viscous response of adhesives used.

**Figure 9.** Average experimental displacement rate for the investigated series of specimens.

#### **4. Discussion of Experimental Observations**

#### *4.1. Service Class and Adhesive Behaviour*

Undoubtedly, the experimental investigations revealed significant differences of the mechanical behaviour of bonded-in rod connections with different adhesives when exposed to wet climate. The test set-up and the support of the DIC system helped in obtaining empirical results of practical use. Rheological behaviour indicates that, in terms of reliability, special attention should be paid to the joints exposed to the extreme climatic conditions. Additional requirements in standards should be included or certification from the adhesive manufacturer should be sought to ensure the safe use of this type of joints.

Tests indicated a significant effect of moisture content on the adhesive stickiness. While only small changes were observed for service class 2, for service class 3 the adhesive stickiness began to recede dramatically. For epoxy specimens, the entire adhesive matrix started to slip smoothly, what is especially characteristic for rods bonded perpendicular to the grains, while on the rods bonded parallel, any pieces of wood grains on the adhesive matrix is not visible. This is indicating that bearing capacity of the joint is defined by shear strength of the interface on exact line between the adhesive matrix and the wood. Polyurethane specimens showed enhanced behaviour, especially for rods bonded perpendicular to the grains. The reason for such behaviour can be justified in higher chemical properties of the new generation of this type of adhesives, which is recommended for use in moist environments. Indicatively, it can be stated that the use of epoxy adhesives is not recommended for service class 3, while polyurethane adhesives can be still used, but with careful consideration for the technical characteristics given by the manufacturer.
