Optimization of Adhesive Joint Design in Timber–Glass Systems: Enhancing Structural Performance with Primer Treatment

Abstract

The increasing use of large glass surfaces in modern architecture requires robust adhesive solutions that balance aesthetic appeal with structural resilience, particularly in timber–glass applications. This study examines the influence of primer treatments on the shear performance of timber–glass adhesive joints, employing a combination of experimental testing and simulation techniques. Double-lap shear tests with epoxy adhesives assess the impact of various surface treatments on joint stiffness, shear stress distribution, and deformation. Additionally, a finite element model is developed to simulate joint behavior, evaluate failure modes, and analyze displacement patterns. Results indicate that primer applications notably enhance structural integrity by reducing displacement and increasing joint stability, thereby supporting more durable timber–glass assemblies. These findings offer valuable insights for advancing adhesive technologies in architectural components, enabling a closer alignment between structural performance and design innovation in timber–glass systems.

1. Introduction

In contemporary architecture, glass is an increasingly widely used design element thanks to its transparency and lightness. The current construction market asks for ever larger and seamless glass surfaces. This request can be satisfied by combining glass with other materials, such as steel, wood, aluminum, and pultruded components, thanks to adhesive unions. This allows us to create clean and innovative designs, able to guarantee at the same time transparency and low energy consumption, also thanks to the technological advances in the double-glazing sector. This has led to the development of new techniques for glass panel assembling that allow architects to design aesthetically pleasing, structurally robust, and energy-efficient structures. One of the most investigated technologies is the potential of making a joint between glass and other materials. Several studies have been carried out on the feasibility of making joints between different materials [1,2], such as bonding timber and concrete for flooring applications [3] and glass—steel [4] and glass—concrete [5] connections for full-scale beams, or a combination of glass with Glass Fiber-Reinforced Polymers (GFRPs) [6] and plastic [7].

Compared to conventional mechanical joints, adhesive joints offer a number of advantages, including the ability to avoid stress concentration and distribute the external load over a larger area, resulting in light and high-performance structures [8]. In addition, adhesive joints are less prone to corrosion and can be used to join dissimilar materials, making them a versatile option for construction projects [4,9]. On the other hand, some major disadvantages of structural adhesive joints are temperature and moisture sensitivity [10,11,12] (reduced performance, adhesive degradation, or bond failure under high temperatures or moisture condition); bond strength dependence on material compatibility [13]; and challenging disassembly and repair [14], requiring specialized techniques, such as heating or chemical treatments, to break the bond (this can increase maintenance costs and limit the ease of future modifications or renovations). Additional aspects to be considered when choosing and designing an adhesive joint are the curing time (achieving the maximum strength can take hours or even days); the surface preparation requirements (a proper surface preparation is crucial for achieving strong adhesive bonds); and environmental considerations (some structural adhesives may contain hazardous substances (e.g., VOCs), and the disposal of adhesive waste must comply with environmental regulations.

An adhesive joint allows the peculiar properties of two different materials to be combined; in the case of a timber—glass joint, for example, the high stiffness of the glass is coupled with the tensile strength of the wood, resulting in ductile structures with good mechanical strength and durability [15,16,17]. This type of joint is particularly useful in situations where traditional mechanical fastening methods are not suitable or when a seamless appearance is desired. Moreover, the use of adhesive joints can result in cost savings and reduced environmental impact by minimizing the need for additional materials and reducing waste. Vallee et al. [18] used a case study of a pedestrian timber—glass bridge to show how to design and dimension complex bonded structures using traditional engineering methods. They found out that existing timber codes underestimate strength for crucial stress components related to bonding. Blyberg et al. [19] investigated the mechanical behavior, energy performance, and Life Cycle Assessment (LCA) performance of various structural glass—timber composite building components (i.e., beams and shear walls), finding out that it is possible to obtain a non-brittle failure of these elements. In particular, the shear walls can be used as load-bearing structures in 3–4 floors, with no loss of energy performance and a slight increase in LCA performance. Marchione et al. [20] tested the application of adhesive technology to a glazed panel for an innovative curtain wall with an integrated frame. The mechanical performance of the adhesive joint depends on numerous factors, ascribable to both the nature of the bond itself (e.g., type of adhesive [21], geometry of the joint [22], bonded materials [3,23], etc.) and external factors, including environmental exposure (temperature and relative humidity) and service load (duration and type) [24,25,26]. Therefore, it is important to carefully consider all of these factors when selecting an adhesive and designing a bonded joint for a particular application. However, the major factor influencing the performance of the joint is the chemical bond between the polymer chains of the adhesive and the joining substrates to be joined. The strength of this bond, in turn, greatly affects the type of failure (e.g., adhesive, cohesive, stock-break, etc.). For this reason, the surface preparation of the materials being joined plays a crucial role in the mechanical performance of the adhesive joint, meaning that proper surface preparation can improve adhesion and prevent premature failure of the bond [27,28,29,30,31]. Primers and adhesion promoters, for example, are useful surface treatments to improve surface tension at the interface and favor adhesion mechanisms. It is worth mentioning that the choice of surface treatment should be based on the specific materials to be bonded and the intended application.

According to what has been discussed thus far, the purpose of this study is to evaluate the performance of a wood—glass adhesive joint and the impact pre-treatments such as primers and adhesion boosters have on it. This is performed with the intention of assessing the feasibility of this type of application at an industrial scale, to be used for building components such as windows and curtain walls. For the latter, ductility is a key property, as it ensures that they will withstand various loads and displacements (e.g., due to seismic and wind loads, thermal expansion and contraction, thermal cycling, moisture, Ultraviolet (UV) exposure, etc.) without experiencing failure, excessive deformation, or premature degradation. This can fulfill various structural and environmental requirements, such as performance enhancement, occupant safety, and system longevity.

In this study, an experimental campaign of shear tests on double-lap specimens is performed to evaluate the effects of three adhesion promoters (variously combined with each other) and three paints on the global mechanical performance of an adhesive joint composed of beechwood and glass adherends assembled with one epoxy structural adhesive. Finally, a finite element (FE) model, aimed to simulate the structural response of the adhesive joint, is developed and verified through the numerical—experimental comparison of the results.

2. Materials and Methods

The mechanical performance of double-lap adhesive joints is investigated by assessing the following mechanical parameters: ultimate load, shear strength and strain, displacements, and ductility. Two different adherends are used in this work: transparent float glass panels (Pilkington, Lathom, UK), and beechwood, supplied by Pircher Oberland. The mechanical properties of the materials are listed in

Table 1. Mechanical characteristics of the adherends reported by manufacturers.

Glass [32]			Beechwood [33]
Thermal Coeff. of Expansion	Young Modulus	Tensile Strength	Density	Moisture Content	Young Modulus	Tensile Strength
α (°C−1)	E (GPa)	σR (MPa)	ρ (kg/m3)	%	E (GPa)	σR (MPa)
9 × 10−6	75	30	720	15	1.5	90

.

On the basis of the results of previous experimental studies on the shear performance of double-lap adhesive joints assembled with different commercial structural adhesives [34], the 3M™ Scotch-Weld™ 7240 B/A adhesive [35] is selected as the most suitable for the present experimental campaign. The choice of this adhesive is dictated by its mechanical characteristics as well as its rheological properties. In fact, the ductility and resistance to artificial aging shown by this adhesive [36] make it the most suitable for the investigated purposes (i.e., windows and curtain walls).

Table 2. Mechanical characteristics of 3M™ Scotch-Weld™ 7240 B/A adhesive reported by manufacturer.

Chemical nature	two-part		toughened epoxy base + modified amine accelerator
Viscosity	-		thixotropic
Work life	Wl	(min)	45
Application temperature	At	(°C)	15 ÷ 30
Glass transition temperature	Tg	(°C)	66.87
Service temperature	St	(°C)	−40 ÷ 80
Shear strength	τ	(MPa)	6.2–24.3 *
Young modulus in compression	Et	(MPa)	3500–4000
Use	-		semi-structural

* depending on the substrate.

shows the 3M™ Scotch-Weld™ 7240 mechanical properties listed in the technical datasheet.

The glass components of three out of the thirteen series are treated with paints, in particular one epoxy-based paint provided by Visa Colors (BGS 9200 white pearlescent paint), and two different acrylic-based paints supplied by Racing Colors (“Painting Bilayer RC9000” and “Paint Topcoat 2K satin + Cat”, Racing Colors, Barcelona, Spain), whose properties and auxiliary components are reported in

Table 3. Properties of the paints.

	Epoxy-Based	Acrylic-Based—Type A	Acrylic-Based—Type B
Color	silver	black	black
Type	2-compound pearlescent paint for glass	bilayer is 100% solvent-based, high coverage	High Solids, solvent-based 2-component, satin finish
Composition	1 part paint + 0.6 diluent 1 + 0.45 catalyst 2	1 part paint + 0.5 diluent 3	1 part paint + 0.2 catalyst 4
Auxiliary components	1 DT 810 Epoxy Thinner by Visa Colors 2 AM 85 Glossy Hardener	3 disolvente acrìlico RU RAC medio	4 Catalizador DCP9156, high adherence, by Racing colors (included with the paint)

. Paints are used for purely aesthetic purposes as they guarantee the coverage of the adhesive joint, playing an important role in the design of the glazed component.

Before the bonding phase, all series except for the “untreated” and the “painted” ones are treated with a primer or a combination of primers on the glass side, and a primer or no primer at all on the timber side. The primers used in the study are the 3M™ Silane Glass Primer, 3M™ Adhesion Promoter 111, and 3M™ Primer 94. The first one is a primer for glass, which avoids moisture penetration at the adherend—adhesive interface; the second and third ones are suitable for various surfaces and can be used both on glass and timber substrates. Before the application of the primers, all surfaces of both timber and glass adherends are cleaned with denatured isopropyl alcohol, as recommended by the manufacturer. Physical and chemical properties of the three primers used in the study are retrieved from the product datasheets and listed in

Table 4. Physical and chemical properties of the primers.

		Silane	111	94
Main ingredient	(% by Weight)	isopropyl alcohol (80–95) *	propan-2-ol (98–100)	cyclohexane (30–60) *
Boiling point	(°C)	82.2	82.4	76.7
Flashpoint	(°C)	11.7	11	−17.2
Density	(g/mL)	0.80	0.79	0.82
Vapor Pressure	(mmHg)	43 (at 25 °C)	330 (at 20 °C)	68 (at 20 °C)
Relative Vapor Density		2.07	2.1	no data available

* Trade Secret.

.

Five specimens are assembled for each series. All combinations tested and their nomenclature are reported in

Table 5. Tested combinations of beechwood—glass joints with various adhesion promoters.

CTRL	Untreated
SIL-G	Silane on the glass side
111G	Primer 111 on the glass side
SIL + 111G	Silane + Primer 111 on the glass side
SIL-G_111T	Silane on the glass side and Primer 111 on the timber side
111G-111T	Primer 111 on the glass side and Primer 111 on the timber side
SIL + 111G_111T	Silane + Primer 111 on the glass side and Primer 111 on the timber side
SIL-G_94T	Silane on the glass side and Primer 94 on the timber side
111G_94T	Primer 111 on the glass side and Primer 94 on the timber side
SIL + 111G_94T	Silane + Primer 111 on the glass side and Primer 94 on the timber side
EPX	Glass treated with epoxy paint
A	Glass treated with acrylic paint—type A
B	Glass treated with acrylic paint—type B

.

The test specimens are assembled in accordance with the ASTM D3528 standard [37]. The geometry of the test specimens is shown in Figure 1a. Since wood is an orthotropic material, the timber specimens are shaped in such a way as to have the greater dimension oriented in the direction of the fibers. Similarly, the joints are assembled so that the load is oriented to the fiber direction. The dimensions of the glass panels are 50 mm × 50 mm; the width of the beechwood profiles is 28.0 mm, with a length of 140 mm. Both the adherends have a thickness of 5 mm. The single bonding region has an area of 355.6 mm² (28.0 mm × 12.7 mm). The thickness of the epoxy adhesive used is 0.30 mm, as recommended by the manufacturer.

Before testing, specimens are cured for 30 days under laboratory conditions of 20 ± 1 °C and 50 ± 4% RH. Shear tests are conducted according to ISO 4587 [38] using a Zwick/Roell Z250, ZwickRoell S.r.l., Genoa, Italy, machine with a load cell of 250 kN under displacement control with a crosshead speed of 1.27 mm/min [37]. All tests are performed under the room conditions previously described. The specimen displacement between the wood adherends is measured through a Zwick/Roell extensometer having a gauge length of 70 mm. The experimental setup is shown in Figure 1b.

At the end of the shear tests, all specimens are observed to characterize the failure modes according to the ASTM D 5573-99 standard [39]. For adhesive joints, the main failure modes are

• Adhesive Failure (AF): Failure occurring at the interface between the adhesive and the adherend when the resistance of the interface (adhesion strength) is less than that of the adherend.
• Cohesive Failure (CF): Failure occurring within the adhesive; therefore, the latter is present on both fracture surfaces.
• Thin-Layer Cohesive Failure (TLC): Cohesive failure occurring very close to the adhesive—substrate interface, characterized by a “light dusting” of the adhesive on one substrate surface and a thick layer of the adhesive left on the other.
• Timber-Tear Failure (TT): Failure occurring exclusively within the timber component, characterized by the presence of timber fibers on both ruptured surfaces.
• Light-Timber-Tear Failure (LTT): Timber failure characterized by a thin layer of timber fibers visible on the adhesive.
• Stock-Break Failure (SB): Breaking of the glass substrate outside the adhesively bonded joint region, often occurring near it.
• Mixed Failure (MF): Any combination of two or more of the previous classes of failure modes described.

3. FE Model

The finite element method (FEM) is commonly used in the literature [40,41,42,43] for the prediction of the structural behavior of such structural components as bonded joints. This section presents the modeling approach, based on FEM, for the simulation of the structural behavior of bonded joints. All analyses are performed through Abaqus^® v. 2021 [44] commercial FE code. The FE model, shown in Figure 2a, refers to nominal dimensions shown in Figure 1a. The FE model counts a total of 90,520 linear hexahedral C3D8R elements (from Abaqus^® finite element library), 8 nodes with 3 degrees of freedom, used for the modeling of all adherends, and 1456 linear hexahedral COH3D8 elements (from Abaqus^® finite element library), 0.001 mm thick, 8 nodes with 3 degrees of freedom, used to simulate the adhesive. The cohesive element layer, depicted in gray in Figure 2b, opportunely placed on the adherend area covered by the adhesive, permits us to predict both debonding initiation and evolution without providing to the FE model indications of the zone where the phenomenon starts. Moreover, element nodes of adherends and adhesive layers are merged at the interfaces to guarantee their connectivity, these being characterized by the same in-plane mesh size.

Particular attention has been paid to the material characterization of the adhesive. In detail, the traction—separation law is defined, considering for each opening I mode the bilinear law shown in Figure 3, where σ_i represents stress and δ_i represents the separation displacement. This technique is widely used in FE analyses and provides a simplified representation of the mechanical behavior of adhesive joints. However, it does not capture the scale effect, which is critical in scenarios where the thickness of the adhesive layer influences the stress distribution and mechanical performance. As highlighted by Long et al. [45], the scale effect becomes particularly significant when adhesive layers are reduced to micrometer or nanometer thicknesses (that does not represent the specific case of this study), leading to deviations from classical mechanics predictions. The inability to account for these size-dependent behaviors can result in an underestimation or overestimation of joint strength, especially in applications involving thin adhesive layers.

Regarding the cohesive thickness t, when a traction—separation approach is used to define the cohesive element response, Abaqus assumes by default that the constitutive thickness is set to one. This default setting is based on the premise that the geometric thickness of cohesive elements is typically zero or negligible in applications where a traction—separation-based response is appropriate, such as in the presented model. This choice ensures that nominal strains directly correspond to the relative separation displacements.

According to Figure 3, the traction—separation law consists of a first linear A–B path, characterized by a prefixed slope (penalty stiffness K_p), where σ_ic and δ_0i represent the critical stress value and the displacement at the debonding initiation. The following B–C path represents the debonding evolution phase. The complete decohesion, and then the cohesive element deletion, occurs when δ_max,i is reached. For inclusion within the simulation, the effects of the combination of multiple opening modes on debonding the quads’ damage criterion and the power law [44] are implemented as initiation and evolution criteria, respectively. Concerning the former, debonding is assumed to initiate when a quadratic interaction function involving the nominal stress ratios reaches a value of one, Equation (1).

$${\left\{{\displaystyle \frac{{\sigma }_n}{{\sigma }_n^C}}\right\}}^2+{\left\{{\displaystyle \frac{{\sigma }_s}{{\sigma }_s^C}}\right\}}^2+{\left\{{\displaystyle \frac{{\sigma }_t}{{\sigma }_t^C}}\right\}}^2=1$$

(1)

where ${\sigma }_n^C$, ${\sigma }_s^C$, and ${\sigma }_t^C$ represent the critical values of the nominal stress ${\sigma }_n$, ${\sigma }_s$, and ${\sigma }_t$ for induction in the adhesive under the normal, first, or second shear modes. The latter, instead, permits us to include within the simulation the effects of the interaction of the energies spent for the adherends debonding under mixed-mode conditions, Equation (2).

$${\left\{{\displaystyle \frac{G_n}{G_n^C}}\right\}}^{\alpha }+{\left\{{\displaystyle \frac{G_s}{G_s^C}}\right\}}^{\alpha }+{\left\{{\displaystyle \frac{G_t}{G_t^C}}\right\}}^{\alpha }=1$$

(2)

where $G_n^C$, $G_s^C$, and $G_t^C$ are the critical fracture energies required to cause the debonding under normal, the first, and the second shear opening modes, respectively. Specifically, the adhesive mechanical properties implemented within the FE model are summarized in

Table 6. Numerical characterization of adhesive: material properties of adhesive layer.

Longitudinal Young’s modulus	En	[MPa]	4000
Transversal Young’s moduli	Es = Et	[MPa]	1350
Nominal stress I mode	${\sigma }_n$	[MPa]	42
Nominal stress II mode	${\sigma }_s$	[MPa]	15
Nominal stress III mode	${\sigma }_t$	[MPa]	15
Critical fracture energy, mode I	$G_n^C$	Jm−2	380
Critical fracture energy, mode II	$G_s^C$	Jm−2	190
Critical fracture energy, mode III	$G_t^C$	Jm−2	190

. The critical fracture energy values represent, for each opening mode, the area subtended to the cohesive traction—separation law. For example, if the traction—separation law is represented by a perfect isosceles triangle (Figure 3), $G_i^C$ can be computed as in Equation (3).

$$G_i^C={\sigma }_i^C{\delta }_{0i}$$

(3)

when the thickness of the adhesive is t = 1 mm, δ_0i can be computed with Equation (4).

$${\delta }_{0i}={\displaystyle \frac{{\sigma }_i^C}{E_i}}$$

(4)

In the presented model, the traction—separation law, for each mode, is expressed by a triangle with the B–C side longer than the A–B side, which typically occurs in reality.

For material properties of all adherends introduced in the FE model, refer to Table 1. Boundary conditions are defined to reproduce the real ones, Figure 4. According to Figure 4, ux, uy, and uz represent the three translational degrees of freedom with respect to the reference system depicted in the same picture. All degrees of freedom of the orange highlighted nodes are fully constrained except for the ux ones of the right end, where a displacement is imposed to simulate the tensile test. In detail, the FE analysis has been carried out according to the implicit and large displacement schemes.

4. Results and Discussion

In this section, experimental and numerical results are illustrated and compared.

4.1. Experimental Results

The experimental results are discussed in terms of ultimate strength, displacement, shear stress (τ, the ratio between load and total adhesive area), shear strain (γ, the ratio between displacement and thickness of the adhesive layer), and global stiffness (k, the ratio between load and displacement) of the joint. Failure modes are also analyzed. For the calculation of the stiffness, only the linear portion of each curve is used since the beginning of nonlinearity indicates the occurrence of damage in the adhesive joint. Figure 5 reports the load—displacement curves for all specimen series up to the failure point, namely until the last load increment is recorded.

A detailed description of the failure modes, specimen by specimen, is provided in

Table 7. Failure modes. The (G) and (T) mean that the failures happen at the glass and timber interfaces, respectively.

Series	Specimen 1	Specimen 2	Specimen 3	Specimen 4	Specimen 5
CTRL	SB + TLC (G)	SB	SB + AF (G) + CF + LTT	SB	SB + CF + LTT
SIL-G	SB	SB + TT + TLC (T)	SB + LTT + TLC (T)	SB + TT	TT + LTT + TLC (T/G)
111G	SB + TT + TLC (G)	SB + TLC (T/G) + LTT	SB + TT + TLC (G)	SB	SB + TT
SIL + 111G	SB + LTT + TLC (T)	SB	SB	SB	SB + LTT + TLC (T)
SIL-G_111T	SB	SB + TLC (T) + LTT+ AF (G)	CF + LTT + TLC (T/G)	SB	SB + TT + LTT + TLC (T)
111G_111T	SB + TT	AF (G) + LTT + TLC (T)	SB + LTT + TLC (T)	SB	SB + TLC (T) + LTT
SIL + 111G_111T	TT + LTT	SB + TT	SB + TT	LTT	SB + TLC (T) + LTT
SIL-G_94T	SB	TLC (T) + LTT	SB + TT + TLC (T)	SB	SB + TLC (T) + LTT
111G_94T	TLC (G) + LTT	SB + LTT + TLC (T)	SB + LTT + TLC (T)	SB + LTT + TLC (T)	SB + TT
SIL + 111G_94T	SB + TLC (T)	SB + TT + LTT + TLC (T)	SB + LTT + TT + TLC (T)	SB + TLC (T) + LTT	SB + LTT + TLC (T)
EPX	SB	SB	SB + CF + LTT + TLC (T)	SB + LTT + TLC (T)	SB + LTT + TLC (T)
A	AF (G) + TLC (G)	AF (G) + TLC (G)	AF (G)	AF (G)	AF (G)
B	TLC (G) + AF (G)	TLC (G) + SB + AF (G)	TLC (G) + AF (G)	TLC (G) + AF (G)	TLC (G) + AF (G)

, while Figure 6 depicts all the observed failure modes, using specimens CTRL_1, CTRL_3, and SIL-G_5 as an example.

The results obtained are compared in terms of average values and summarized in

Table 8. Mechanical properties and failure mechanisms of beechwood—float glass double-lap adhesive joints.

Series	Load max	Displ max	τmax	γmax	k	Failure Modes [39]
	(kN)	(mm)	(MPa)	(-)	(kN/mm)
CTRL	8.90 ± 0.64	0.31 ± 0.05	7.26 ± 0.50	1.09 ± 0.16	30.39 ± 3.66	3 MF + 2 SB
SIL-G	8.61 ± 1.06	0.20 ± 0.05	6.67 ± 0.83	0.66 ± 0.16	44.62 ± 5.97	4 MF + 1 SB
111G	9.01 ± 0.73	0.18 ± 0.03	6.98 ± 0.57	0.62 ± 0.08	50.73 ± 5.97	4 MF + 1 SB
SIL + 111G	9.13 ± 1.47	0.18 ± 0.03	7.07 ± 1.14	0.63 ± 0.04	46.21 ± 10.89	3 SB + 2 MF
SIL-G_111T	9.18 ± 1.04	0.19 ± 0.03	7.11 ± 0.80	0.58 ± 0.09	49.88 ± 4.48	3 MF + 2 SB
111G_111T	9.20 ± 1.41	0.23 ± 0.07	7.13 ± 1.09	0.66 ± 0.21	43.19 ± 11.77	4 MF + 1 SB
SIL + 111G_111T	9.06 ± 0.82	0.17 ± 0.02	7.02 ± 0.64	0.53 ± 0.07	50.27 ± 3.02	4 MF + 1 LTT
SIL-G_94T	8.91 ± 0.87	0.18 ± 0.01	6.90 ± 0.67	0.60 ± 0.04	50.18 ± 6.09	3 MF + 2 SB
111G_94T	9.44 ± 0.65	0.20 ± 0.03	7.32 ± 0.50	0.68 ± 0.11	48.50 ± 4.85	5 MF
SIL + 111G_94T	8.59 ± 0.81	0.18 ± 0.03	6.66 ± 0.63	0.63 ± 0.10	48.46 ± 6.14	5 MF
EPX	10.10 ± 0.47	0.29 ± 0.04	7.82 ± 0.36	1.17 ± 0.30	34.87 ± 3.20	3 MF + 2 SB
A	2.72 ± 1.12	0.09 ± 0.03	2.11 ± 0.87	0.30 ± 0.12	30.06 ± 6.21	3 AF (G) + 2 MF
B	6.46 ± 0.72	0.21 ± 0.05	5.01 ± 0.55	0.70 ± 0.16	31.45 ± 4.75	5 MF

, where the prevalent failure modes are also listed.

The highest failure load is achieved by the series treated with the epoxy paint, while the two acrylic-painted series register the worst behavior; in particular, for paint A (Figure 5n), a drastic reduction in ultimate load is observed (70% lower than the CTRL series, Figure 5a), probably due to the inhibition of the cohesive process between the adhesives caused by the interposition of the paint layer. All other series have a slightly higher failure load than the untreated series, except for the SIL-G_94T (Figure 5h) series (similar maximum load) and the SIL-G (Figure 5b) and SIL + 111G_94T (Figure 5l) series (slightly lower maximum load). The untreated series exhibits the highest ultimate displacement, comparable only to that of the EPX series (Figure 5m); all other series (A excluded) have a maximum displacement equal to approximately two thirds of that of the untreated series. The EPX series also has the highest shear stress and strain; the average value of shear strain of the EPX series is comparable only to that of the untreated series (Figure 5a), while the shear strain of all other series with primers is halved (0.53 ≤ γ ≤ 0.68). The highest values of the global stiffness (around 50 kN/mm) are reached by four series, for which at least one primer on the glass side is used (111G—Figure 5c, SIL-G_111T—Figure 5e, SIL + 111G_111T—Figure 5g, and SIL-G_94T—Figure 5h). This results from the improved connection of the adhesive to the glass surface, made possible using adhesion promoters. As a matter of fact, due to the environmental exposure, the bonding surface on the glass side may be subject to moisture penetration at the interface between the glass adherend and the adhesive. This causes a weakening of the bonding and may lead to the detachment of the adhesive from the glass. The use of a Silane primer effectively counteracts this phenomenon. Moreover, the stiffness of the untreated series (Figure 5a) is comparable to that of the painted series, particularly the acrylic types (Figure 5n,o). The series treated with acrylic paint type A (Figure 5n) shows the worst results for all the parameters considered, followed by type B (Figure 5o) series.

Once the shear tests were completed, the failure modes of all specimens were analyzed. Only 18 out of the total 65 specimens show a single failure mode, while in all the others, the failure occurs as a mix of two to four different mechanisms. The most recurring one is the stock-break (SB) failure mode, which occurs in three to five out of five cases for all series, except acrylic-treated specimens. The latter all show an adhesive-type failure mode, AF, alone (only for the A type) or in combination with a thin-layer cohesive fracture; both those collapse mechanisms occur at the interface between the adhesive and the glass, which is the most undesirable form, since it indicates low bonding forces between adherends. The TLC mode takes place close to both interfaces, but at the timber interface, it is two times more frequent. The AF never happens at the timber interface, a phenomenon attributable to the fibrous and porous structure of the wood. This is particularly evident in beechwood, which, despite its thin and delicate bark that lends itself well to carving, is still classified as a hardwood due to its high density. In fact, such a structure makes possible the development of greater cohesion forces between the adhesive and the timber adherends. The presence of an interface characterized by a limited roughness (i.e., glass Ra = 0.0012–0.0017 mm outer adherends) determines the presence of a preferential sliding plane for the specimen according to an adhesive failure, AF.

4.2. Numerical Results and Experimental Data Comparison

As aforementioned, material characterization in the FE model does not consider the effects of both adhesive promoters and paints, but it refers only to material properties reported in Table 1 and Table 2. Result postprocessing is performed to identify the level of accuracy of the developed FE model in simulating the structural behavior of all specimen series. For this purpose, a representative load vs. displacement curve for each specimen series was selected and is reported in Figure 7, where the curve provided by FE simulation is also reported. According to Figure 7, it can be noticed that all series where adhesive promoters are used are quite overlapped. CTRL, A, B, and EPX specimen curves, instead, appear as dissociated from the previous ones, being characterized by a reduced stiffness. The predicted curve (solid line) is quite overlapped with the curves’ group linked to the specimens where adhesive promoters are used. It must be highlighted that the FE model, in this case, permitted the simulation of an ideal specimen, characterized by a uniform layer of glue, by the lack of adherends misalignment and all those imperfections usually associated with specimen manual manufacturing. Due to the good and acceptable level of accuracy provided by the FE simulation with respect to all series including adhesive promoters, it can be concluded that the usage of such additives permits mitigating imperfections in specimens. In terms of the prevalent failure mode, FE simulation appears to be coherent with those experimentally detected. The main reason for collapse of the joint, in fact, lies in the glass break.

Figure 8a shows the von Mises stress distribution over the whole joint with focus on the glass adherends. According to Figure 8a, it can be observed that the maximum stress amplitude is reached. In Figure 8b, showing the deformation distribution at the glass adherends, it can be observed that the axial (with respect to the joint) strain component reaches the typical value of elongation at break, equal to 5%. Following the glass break, due to the load redistribution, the adhesive layers start to fail, as depicted in the contour plot shown in Figure 8c extracted some increments later. According to Figure 8c, blue zones represent the undamaged areas of the adhesive. Fully collapsed finite elements are deleted during the simulation.

As demonstrated by Figure 9, the adhesive failure is mainly induced by mixed opening modes with a prevalence of the first shear mode (mode II), followed by mode I. All stress components are measured before and after the finite element deletion along the path illustrated in Figure 9.

5. Conclusions

In this study, the influence of different pre-treatments and coatings on the mechanical behavior of an adhesive joint between glass and timber is investigated in order to identify an appropriate technique for including adhesive junctions in the industrial production of building components such as windows, glazing panels for curtain walls, etc. In addition, an FE model was developed and verified through an experimental—numerical comparison of the results.

The results highlight that

(1)

The pre-treatments on the glass and/or timber side of the bonded joint reduce the maximum displacement, shear stress, and shear strain. They may also contribute to increase the maximum failure load of the joint, but this behavior is not observed for all the tested series; in particular, adhesion promoters (111 and 94) seem to have a better effect than a moisture inhibitor (Silane) in this respect. As a consequence of the decrease in the ultimate displacement, all pre-treated series have higher stiffness than the untreated ones. These treatments, in fact, improve surface adhesion at the adherend—adhesive interface, which in turn enhances load transmission between the adherents up to the maximum permissible stress of the glass, before the adhesive failure (AF) occurs. This aspect of stiffness is relevant according to the different fields of application (i.e., depending on the displacements that the joint can bear, a combination is more or less suitable for a specific purpose).

(2)

The behavior of the painted joint strongly depends on the type of paint used on the glass: a paint of the same nature of the adhesive (i.e., epoxy paint) can increase the ultimate load but has no significant effect on the remaining considered parameters (as summarized in Table 8), nor on the type of failure mechanisms. Conversely, the use of an acrylic paint causes a worsening of the overall mechanical behavior and shifts the collapse mode from glass failure to interface failure.

(3)

According to the results of the numerical modeling, it can be concluded that the simulation appears to be more representative of all those series where adhesive promoters are used. This aspect can be attributed to the fact that the usage of adhesive promoters reduces the effects on the joint structural behavior of the uncertainties affecting all specimens because of the manufacturing process, which cannot be easily included in the modeling.