Scarf Adhesive Bonding of 3D-Printed Polymer Structures

Abstract

Additive manufacturing (AM) has swiftly emerged as a substitute for conventional methods such as machining and injection moulding. Its appeal is attributed to accelerated prototyping, improved sustainability, and the capacity to fabricate intricate shapes. Nonetheless, the size constraints of additive manufacturing components require the assembly of smaller 3D-printed elements to create larger structures. This study investigates the tensile properties of scarf joints (SJs) created from several polymers, including ABS, PETG, and PLA, adhered with Araldite^® 2015 and Sikaforce^® 7752 adhesives. The characteristics of the adherends were assessed prior to examining the adhesive efficacy in the SJ configuration. Experimental evaluations quantified failure modes, joint strength, assembly stiffness, and energy at failure, comparing findings with predictions from a cohesive zone model (CZM). The objective was to determine the ideal combination of materials and adhesives for enhanced joint performance. Results indicated that joint performance is greatly affected by the adherend material, adhesive selection, and scarf angle. PLA and Araldite^® 2015 typically exhibited optimal strength and stiffness, but Sikaforce^® 7752 demonstrated enhanced energy absorption for extended bonding lengths.

1. Introduction

The new industrial revolution, the fourth revolution or Industry 4.0, focuses on adapting to new difficulties by creating simpler, more autonomous, and faster processes. Their autonomy involves less human intervention through communication between machines [1]. AM processes, also known as 3D printing, are experiencing growth, and more industries are seeking this technology due to its advantages compared to processes such as machining and injection [2]. AM processes present a huge range of reasons to replace existing processes [3]. AM processes are increasingly sought after by large industries due to the possibility to produce complex geometries and to address economic/environmental issues using sustainable and recyclable materials [4]. The customer will be able to obtain a prototype more quickly, visualise it physically, and outline changes until the final product is obtained [5]. AM can currently be used with titanium and nickel metal alloys, aluminium, resins, polymers such as acrylonitrile butadiene styrene (ABS) and poly lactic acid (PLA), ceramics, and composites [6,7]. These materials can be worked in liquid, solid, and powder forms. However, there is one disadvantage that leaves this technology lagging behind the others: the size of the manufactured parts is quite small [8]. Moreover, the final properties of a printed component are difficult to control because they depend on several factors. In addition, heat transfer can cause distortions and stresses in the component, and surface quality is generally poor and dependent on the process parameters [9,10]. Recently, 3D printing processes were associated with smart materials, the manufacturing process of which is known as 4D printing. This type of material has the characteristic of transforming itself with the introduction of a stimulus, such as heat, light, electricity, or water [11]. Additionally, a hybrid process has been developed, consisting of a combination of AM and machining processes. AM is applied to obtain the complex geometries it is capable of manufacturing, and machining is applied for surface finishing [9].

One of the main challenges for industries is the limited size of AM components [8]. As a result, it is often necessary to join smaller components to build a larger element without losing mechanical and visual characteristics [12]. The joining process must adapt to the components involved without damaging them. To overcome these constraints, it is essential to find effective methods to join AM components. Welding processes are only suitable when the components are made from metal alloys, but these processes can deform the components due to the high temperature concentration [13]. A notch joint involves designing and manufacturing components so that they are directly coupled [14]. Mechanical connections are widely used since the components can be manufactured with internal or external threads. However, it is difficult to achieve the required thread precision. Therefore, it is advisable to use larger threads [15]. Another example of a mechanical connection in AM, especially for easily deformed plastics, is the use of threaded insert nuts [16]. Adhesive bonding is an emerging technique that has addressed many challenges posed by traditional methods. Adhesive connections are easy to make and cost-effective; enable joining of different materials; and offer good fatigue performance, vibration damping, and a more uniform distribution of stresses [17]. However, fabrication can be time-consuming and costly, and the final assembly has poor peel resistance [17]. Adhesive bonding presents opportunities for AM to create larger structures, but the integrity of adhesive bonds needs to be studied first [18]. Although adhesive joining with polymeric 3D printed-adherends currently lack a unified standard, the study of the most common joint, the single-lap joint (SLJ), follows the ASTM D1002 standard, intended for metallic adherends but adapted for the relevant case [19]. The fused filament fabrication (FFF) process is the most requested by engineers studying the combination of adhesives with 3D-printed adherends, but the process parameters require attention, since they affect both the adherend properties and adhesive joint [18]. The adhesive joint parameters, such as adhesive type, adhesive thickness, and surface preparation, must also be optimised, which is often supported by a numerical study [20]. Under this scope, cohesive zone modelling (CZM) is a suitable tool to be used in combination with finite element method (FEM) analysis to promote crack growth in the adhesive and adherends to accurately estimate the complete response of the specimens up to failure, proving a robust design tool for joint optimisation [21]. CZM is based on using cohesive elements with established traction-separation laws that initially have an elastic part, followed by damage initiation and final failure, which allow modelling crack growth governed by strength and energetic principles [22].

Several studies on adhesive bonds with 3D-printed adherends are described next, some of which are reinforced by appropriate numerical analyses. Leicht et al. [23] focused on adhesive bonding of polyamide 12 substrates, produced by laser sintering (LS), using various epoxy adhesives, including DELO-DUOPOX SJ8665 (DELO Industrieklebstoffe GmbH & Co., Ltd. KGaA, Windach, Germany), SikaPower-1277 (Sika Schweiz AG, Zurich, Switzerland), and Easy-Mix HT 180 (WEICON GmbH & Co., Ltd. KG, Muenster, Germany). A key variable was the surface preparation of the substrates: untreated, chemically treated, treated with atmospheric pressure plasma, and a combination of both treatments. The adhesive joints were tested to failure under tensile loading. The authors concluded that surface preparation affects adhesive joint strength, with plasma treatment yielding better results than chemical treatment, while the combination of both provided the best performance. Spaggiari and Denti [24] explored the integration of FFF with polymeric materials and structural adhesives, aiming to improve adhesive performance by customising the adherends’ surfaces and to bond multiple parts without compromising structural integrity. A design of experiment approach was employed to analyse a variety of SLJs using two different adhesives and seven distinct surface textures. Adhesive bonding kept the load-bearing capacity and stiffness of the original structure, and in certain configurations, even enhanced the peak load capacity. Surface texture had a relatively minor effect on joint performance due to peak stresses at the overlap edges. Hiremath et al. [25] examined how print orientation and graphene nanoparticle (GNP) enhancement affected the strength of PLA SJs produced using FFF. Experimental results showed that samples printed at 0° orientation exhibited the highest load-bearing and shear strength compared to 45° and 90° orientations. Incorporating GNP into the epoxy adhesive significantly enhanced performance, with weight percentages of 0.25% to 1.00% leading to strength improvements up to 31.11%. An artificial neural network (ANN) model accurately predicted failure loads, with an overall reliability score of 0.8471. Khosravani et al. [26] analysed the behaviour of PETG SJs to determine the optimal adhesive layer thickness. The research was conducted using a CZM approach for the adhesive behaviour. Simulations allowed the authors to observe distinct stages of adhesive deformation. In the first stage, observed at very low forces, the adhesive acts as a stress distributor across the adhesive layer. Using the stiffness degradation (SDEG) option in Abaqus^®, no deformation was found in the adhesive during this stage. As applied forces increased, the adhesive entered the second stage, in which cohesive deformation began at the joint edges. With further axial load, the third stage was reached, characterised by cohesive failure at the centre of the adhesive layer. The maximum load (P_m) was identified at the onset of the second stage. Morano et al. [27] integrated a FEM/CZM analysis, design exploration, and additive manufacturing to investigate how the architecture of adherends influences crack growth mechanics in adhesively bonded 3D-printed materials. FEM simulations and experimental testing were performed on double cantilever beam (DCB) specimens made from selective laser sintered polyamide (PA) bonded with epoxy. The study considered adherends with either subsurface hollow channels (bulk patterns) of various shapes or sinusoidal interfacial patterns with varying aspect ratios, aiming to promote crack shielding. By tailoring the structure of the bonded layers, these approaches enhanced the toughness and damage tolerance of the joints. Khosravani et al. [28] conducted both experimental and computational analyses to evaluate the effects of various printing parameters on SJs. Factors such as print angle, width, and layer thickness were investigated to determine the optimal adhesive thickness for maximising joint strength. The PLA/LOCTITE^® EA 9466 samples showed multiple fracture modes, with cohesive failure being the most prevalent. Additionally, a numerical simulation was performed to model SJ behaviour, which was then validated against experimental results. Molecular dynamics simulation (MDC) was employed to accurately replicate adhesive performance in the joints. Despite the available studies on adhesive bonding of additive manufactured structures, challenges remain unresolved. The lack of a unified standard for adhesively bonded 3D-printed polymer joints limits direct implementation in industry. The influence of adherend materials and adhesive types on adhesive performance is not yet fully understood by the existence of systematic studies that compare different joining configurations, requiring further study to optimise joint strength and durability. Scarf geometry is also not addressed in the scientific literature, although it brings significant advantages in real applications. Additionally, CZM has demonstrated potential for predicting joint performance, but its accuracy in estimating the failure modes and mechanical behaviour, including the failure energy, remains insufficient so that a tool can be confidently used by designers and industrialists to enable an extended use.

This study investigates the tensile behaviour of scarf joints (SJs) formed with additively manufactured adherends of different polymers: acrylonitrile butadiene styrene (ABS), polyethylene terephthalate glycol-modified (PETG), and polylactic acid (PLA). The joints were bonded with the Araldite^® 2015 and Sikaforce^® 7752 adhesives. The properties (elastic, plastic, and fracture) of the adherends were experimentally evaluated prior to analysing the adhesive performance in the SJ configuration. Experimental assessments were conducted to measure failure modes, joint strength, assembly stiffness, and energy at failure, and these results were compared with CZM predictions. The goal is to identify the optimal combination of materials and adhesives for enhanced joint performance.

2. Materials and Methods

2.1. Adherends’ Characterisation

The selection of 3D printing material for the adherends was conducted taking into account a previously published paper by the authors [21]. The materials were PLA, ABS, and PETG. The chosen 50% infill presented a good ratio between mechanical properties and fabrication time. Plus, if a 100% infill were to use used, plastic injection would be a better fabrication process. By using 50% infill, an advantage of the AM (FFF) process can be used and analysed. For the sake of completion, the printing parameters are summarised and included in

Table 1. PLA, PETG, and ABS printing parameters (adapted from [21]).

Material	PLA	PETG	ABS
Printer model and make	Ender 3 Max	Ender 3 Max	Ender 6
Adhesion table type	Vidro Carborundum	Vidro Carborundum	Vidro Carborundum
Adhesion table temperature [°C]	60	80	70
Extruder nozzle temperature [°C]	215	230	230
Material deposition speed [mm/s]	75	75	120
Layer height [mm]	0.2	0.2	0.2
Extruded line thickness [mm]	0.4	0.4	0.4
Outer wall thickness [mm]	1.2 (3 lines)	1.2 (3 lines)	1.2 (3 lines)
Bottom and top shell thickness [mm]	0.8	0.8	0.8
Extruded filament thickness [mm]	0.4	0.4	0.4
Infill pattern	Lines
Infill density [%]	50	50	50
Printing direction
Build plate adhesion type	Skirt	Skirt	Skirt

.

For the determination of the elastic/plastic properties, the authors conducted the tests [21] according to the applicable standard [29]. For illustration purposes, Figure 1 shows the tensile test sample dimensions.

Table 2. Relevant average mechanical properties of PLA, PETG, and ABS (adapted from [21]).

Adherend	Stress at 0.2% [MPa]	Peak Stress [MPa]	Peak Strain [%]	Young’s Modulus [MPa]
PLA	32.07	35.55	2.06	2353.74
ABS	20.11	25.28	3.77	1362.01
PETG	17.15	22.30	5.27	1559.02

summarises the required mechanical properties for the numerical models [21].

2.2. Adhesives Characterisation

Two adhesives with a different ductility degree and behaviour were selected, namely Huntsman Araldite^® 2015 and Sikaforce^® 7752 FRW L60. Actually, Araldite^® 2015 has smaller ductility and higher strength, while Sikaforce^® 7752 is more ductile and flexible, even though less strong, as depicted in the σ-ε curves of Figure 2. The mechanical and fracture properties of the adhesives presented in

Table 3. Mechanical and fracture properties of the adhesives [30,31].

Properties	Araldite® 2015	Sikaforce® 7752
Young’s modulus, E [GPa]	1.85 ± 0.21	0.49 ± 0.09
Poisson’s ratio, ν *	0.33	0.30
Yield tensile strength, σy [MPa]	12.63 ± 0.61	3.24 ± 0.48
Ultimate tensile strength, σf [MPa]	21.63 ± 1.61	11.48 ± 0.25
Tensile strain at failure, εf [%]	4.77 ± 0.15	19.18 ± 1.40
Transverse elastic modulus, G [GPa]	0.56 ± 0.21	0.19 ± 0.01
Shear yield strength, τy [MPa]	14.60 ± 1.30	5.16 ± 1.14
Shear ultimate strength, τf [MPa]	17.90 ± 1.80	10.17 ± 0.61
Shear strain at failure, γf [%]	43.90 ± 3.40	54.82 ± 6.38
Tensile toughness, GIc [N/mm]	0.43 ± 0.02	2.36 ± 0.17
Shear toughness, GIIc [N/mm]	4.70 ± 0.34	5.71 ± 0.47

* Supplier data.

were determined in previous published works [30,31].

2.3. Joint Geometry, Fabrication, and Testing

This research considers the study of adhesively bonded scarf joints (SJs) manufactured using the FFF process. The scarf angle varies (α = 90, 28.9, and 13.9°) to consider bonding lengths (L_O) = 5, 10, and 20 mm, respectively, aiming to access the adherend and adhesive type effect on the mechanical behaviour of the joints. The selection of L_O = 5, 10, and 20 mm was based on a balance between practical manufacturability and joint loading/efficiency. A smaller L_O (5 mm), which in this case is actually a butt joint, represents a scenario with high peel stress concentration and lower load-bearing capacity, while an extended L_O (20 mm) allows better stress distribution, with predominant load transfer by shear. However, it may introduce excessive adhesive usage and process complexity. Higher L_O values were not evaluated due to adherend weakening due to the pronounced taper, which will result in premature adherend failures. Figure 3 presents the SJ geometry and remaining dimensions (in mm): total joint length between grips L_T = 180, adhesive thickness t_A = 0.2, adherend thickness t_P = 5, and joint width B = 25 mm, all kept constant.

Both adherends of each joint were printed in approximately 1 h 45 min. The specimens were gripped directly by the joint edges, which assured proper joint alignment in the testing machine without shims being required. When printing SJs, the inclined faces present small steps and can lead to adhesion failures in the joint (Figure 4a), which is why all the faces of the adherends were previously sanded with 80-grit sandpaper (Figure 4b) and then cleaned with acetone in the case of PLA and PETG and isopropyl alcohol in the case of ABS. Applying such procedures on the bonding surfaces of the specimens promotes an increase in the wettability, improving contact with the adhesive and spread over the substrate.

To ensure a constant t_A along the bonding line, two copper wires with a diameter of 0.2 mm were used. To avoid adherend misalignment and guarantee t_A = 0.2 mm, clamps were used to grip the specimens during the room temperature curing process. The excess of adhesive that commonly appeared at the joints overlapping edges is removed with a clipper and a drilling machine equipped with a griding wheel. Thus, a smooth surface was created, and the theoretical joint geometry is emulated to promote a proper comparison between experimental and numerical data. Four specimens per each joint configuration were tested accordance with ASTM D638 [32] using a Shimadzu Autograph AG-X 100 universal testing machine (Kyoto, Japan) equipped with a 100 kN load cell at a speed of 1 mm/min. Figure 5 shows an ongoing tensile test on an ABS SJ bonded with Araldite^® 2015. All tested joints were loaded until failure, and the relevant data were collected for analysis.

2.4. Model Pre-Processing

The implicit finite element analysis was conducted with Abaqus^® 2021 (Dassault Systèmes, Vélizy-Villacoublay, France) using the embedded CZM framework, incorporating geometric non-linearities due to significant deformation and rotation particularly at the joining region. The pre-processing involved various stages, including drawing the geometries, inserting material properties, defining the meshing techniques, and applying the boundary conditions. For all cases, two-dimensional (2D) models were utilised, reflecting the specimens’ constant width design. Adherends and adhesive layers were initially represented as a unified structure and subsequently segmented into partitions to allow assignment of specific material properties to the adherend and adhesive regions. The adhesive layer was defined with cohesive elements with a bilinear traction-separation curve, as detailed in reference [33], incorporating both longitudinal and shear moduli (E and G) to provide the elastic stiffness. Damage initiation in mixed mode was described using the quadratic stress (QUADS) criterion based on tensile and shear cohesive strengths (t_n⁰ and t_s⁰ equal to σ_f and τ_f, respectively), with mixed-mode failure evaluated with a linear energy criterion using G_IC and G_IIC values, as outlined in Table 3. Adherend properties included both elastic and plastic characteristics, with isotropic elasticity defined by E and Poisson’s ratio, and plasticity represented by a stress–plastic strain relationship, assuming perfectly plastic behaviour beyond the yield point to simplify the analysis. The mechanical properties of adherends are listed in Table 2. The analysis was set up as a static general step, with increment control to capture crack progression in the adhesive. Boundary conditions simulated the experimental setup, with clamping on the left cross-section and a 5 mm horizontal displacement applied to the right cross-section, while constraining vertical displacement. While this arrangement aimed to replicate pure tensile loading, perfect alignment can be challenging to achieve in physical tests. Mesh details are given in Figure 6. The adherend regions were mostly meshed with plane-strain quadrilateral elements (CPE4) using a structured mesh and plane-strain triangular elements (CPE3) for the sloped scarf portion using a free mesh. The adhesive was meshed with quadrilateral elements (COH2D4) after applying sweep mesh controls between the adherend surfaces to be joined. For the adhesive layer, only one cohesive element was used across its thickness, aiming to fully represent its tensile and shear stiffness. By varying α, the mesh density was kept constant, leading to a different element and node count. Key model outputs included visualisations of failure progression and load–displacement (P-δ) curves, which form the basis for the subsequent analysis and discussion.

3. Results and Discussion

3.1. Failure Modes

Two different failures were obtained: cohesive in the adhesive layer and adherend failure. Between these two types of failure, the cohesive failure indicates that the adhesive strength was the limiting factor of the joint. This type of failure occurred multiples times with PLA, always when combined with Sikaforce^® 7752 adhesive (Figure 7a). For ABS and PETG, the failure always occurred in the adherend (Figure 7b).

The CZM/FEA method allows the visualisation of adhesive degradation throughout the loading process, from the initial load to failure, using the SDEG (stiffness degradation) parameter. SDEG ranges from 0 (no damage) to 1 (full damage). Additionally, adherend plasticisation is captured through Abaqus^® internal variable PEEQ (equivalent plastic strain). For all tested geometry conditions, failures in joints were observed as cohesive failure in the adhesive or as adherend plasticisation leading to tensile failure. Figure 8 shows the gradual cohesive failure of the adhesive layer for the SJs between PLA adherends with L_O = 20 mm with Araldite^® 2015. In the initial stages of loading, it is possible to visualise a small deformation at the overlap edges in the SJs. This small deformation occurs due to the known stress concentrations at these sites. With the continued loading, a slightly deformation starts to occur in the adherend at the overlap edges. Furthermore, SDEG increases, and the damage propagates towards the central area, as shown in Figure 8b). By the end of the test, depending on which material is more resistant, either a cohesive failure of the adhesive (Figure 8c) or an adherend failure is obtained (Figure 9).

Figure 9 demonstrates two different types of adherend rupture by showing the deformed contour of joint at failure: outside the overlap (a) or in the centre of overlap (b). In the first case, there is necking near the clamping point. In the second case, there is significant plastic deformation at the taper region at the overlap. Regardless of the adhesive or adherend, the two different failures modes always occurred. The PLA joints showed the least adherend failures, due to the higher PLA mechanical properties. With Araldite^® 2015, only for L_O = 5 mm, a cohesive failure of the adhesive occurred. For L_O = 10 and 20 mm. the adherends deformed plastically at the centre of the overlap. With Sikaforce^® 7752, an adherend failure occurred only for L_O = 20 mm. In the ABS joints, either with Araldite^® 2015 or Sikaforce^® 7752, cohesive failures were obtained for L_O = 5 mm, while adherend failures occurred in the centre of the overlap for L_O = 10 and 20 mm. Using PETG joints, cohesive failure was only obtained for L_O = 5 mm with Sikaforce^® 7752. For other combinations, the adherend failed in the centre of the overlap, as shown in Figure 9b. Figure 9a shows the failure outside the adhesive layer that occurred with PETG and Araldite^® 2015 for L_O = 5 mm.

Table 4. Experimental/numerical failure mode comparison.

Adhesive	LO [mm]	PLA		ABS		PETG
	Analysis	EXP	NUM	EXP	NUM	EXP	NUM
Araldite® 2015	5	Coh	Coh	Adh	Coh	Adh	Adh
	10	Adh	Adh	Adh	Adh	Adh	Adh
	20	Adh	Adh	Adh	Adh	Adh	Adh
Sikaforce® 7752	5	Coh	Coh	Adh	Coh	Coh	Coh
	10	Coh	Coh	Adh	Adh	Adh	Adh
	20	Coh	Adh	Adh	Adh	Adh	Adh

Adh—adherend failure; Coh—cohesive failure in the adhesive.

compares the different failures modes obtained experimentally and numerically for all combinations of adherend, adhesives, and L_O. Numerically and experimentally, with Araldite^® 2015, all failures occurred in the adherend, except with PLA and ABS for L_O = 5 mm. With PLA, a cohesive failure was obtained. With ABS, a different result presented itself due to defects in the fusion between the layers during the 3D printing process, resulting in a premature experimental failure of the adherend. The same reason can explain the different results obtained with SJs with ABS and Sikaforce^® 7752 for L_O = 5 mm. With PLA and Sikaforce^® 7752 for L_O = 20 mm, numerically, an adherend failure was obtained that contradicted an experimentally cohesive failure. This difference can be explained by poor adhesion between the adhesive layer and the adherend.

3.2. Joint Strength

The experimental and numerical P-δ curves for the SJs with the Araldite^® 2015 for all evaluated adherends are presented in Figure 10, while Figure 11 compares the results for Sikaforce^® 7752. Both figures only show L_O = 10 mm as an example to illustrate the behaviours of the load curves. Experimentally, SJs with PLA adherends and L_O = 10 mm presented a P_m of 4.16 ± 0.18 kN with Araldite^® 2015 (Figure 10a) and 2.39 ± 0.11 kN with Sikaforce^® 7752 (Figure 11a). Numerically, P_m values presented 7.3% and 10.26% overpredictions of 4.47 kN and 2.64 kN, respectively. Using ABS adherends, with the same L_O, experimentally P_m = 2.54 ± 0.16 kN was obtained with Araldite^® 2015 (Figure 10b) and 2.38 ± 0.04 kN with Sikaforce^® 7752 (Figure 11b), while 2.79 kN and 2.63 kN, respectively, were numerically obtained. These values represent overestimations of 9.6% and 10.2%, respectively. For PETG with different adhesives, similar values of P_m were obtained experimentally, as a result of failures by the adherend. Experimentally, P_m was 2.20 ± 0.09 kN with Araldite^® 2015 (Figure 10c) and 2.19 ± 0.12 kN with the Sikaforce^® 7752 (Figure 11c). Using FEA models, P_m was estimated at 2.41 kN for both adhesives. These P_m values correspond to overestimated values of 9.5% for Araldite^® 2015 and 10.9% for Sikaforce^® 7752. It must be noted that, for all conditions and in both analyses, the failure occurred always after P_m was obtained. In general, the experimental data showed good consistency, demonstrating that the specimens were prepared properly. Across all conducted tests, the experimental stiffness was consistent, and there was also a strong correlation with the numerical results.

The experimental joint behaviour observed as a function of the adherend type was as follows:

• In the joints with PLA and Araldite^® 2015, P_m increases by 66.5% for L_O = 5–10 mm, as a result of different types of failure, including cohesive and adherend failures, respectively. For L_O = 10–20 mm, the same failure was obtained, and no increase is noted. However, using Sikaforce^® 7752, P_m increased to 79.0% and 71.4% for L_O = 5–10 mm and L_O = 10–20 mm, respectively, as cohesive failures were always obtained. Comparing both adhesives with L_O = 5 mm, the SJs with Araldite^® 2015 were 87.0% more resistant than those with Sikaforce^® 7752. The same fact is obtained for L_O = 10 and 20 mm, with Araldite^® 2015 being more resistant by 74.0% and 0.8%. The minor percentile difference for L_O = 20 mm is related to the adhesive’s mechanical properties, such as stiffness, strength, and ductility. For higher L_O, Sikaforce^® 7752 increases its load capacity as it is a more flexible and ductile adhesive than Araldite^® 2015, which is stiffer and more brittle.
• For the SJs with ABS and Araldite^® 2015, the failure mode always occurred due to the adherend. For the different L_O values, no increase in P_m was noticed. For Sikaforce^® 7752 with L_O = 5–10 mm, a P_m increase of 81.6% was obtained, mainly due to different failure modes in each length. When the failure occurs in the adherend at L_O = 10–20 mm, the P_m increase was reduced to 7.2%. This slight increase results from the elastic properties of the ABS shown in Section 2.1. For the different L_O values, Araldite^® 2015 was always superior to Sikaforce^® 7752: 89.8% for L_O = 5 mm, 7.1% for L_O = 10 mm, and 0.6% for L_O = 20 mm.
• Using PETG and Araldite^® 2015, similar to the ABS results, no increase in P_m was detected due to identical adherend failures. Using Sikaforce^® 7752, P_m increased 65% for L_O = 5–10 mm. Despite the same failure mode, for L_O = 5–10 mm, the pull-out force transforms in shear force due to an increase in α, allowing the adherend and the adhesive to resist more forces. For L_O = 10–20 mm, the increase was nearly 0%. When comparing the different adhesives, Araldite^® 2015 was superior to Sikaforce^® 7752. For L_O = 5 mm, P_m for Araldite^® 2015 was 70.5% higher compared to Sikaforce^® 7752. This percentage reduces to nearly 0.5% for L_O = 10 and 20 mm.

Figure 12 compares experimental and numerical P_m between joints with different adherends and adhesives. PLA joints present the highest P_m, followed by ABS and PETG. As previously shown in Section 2.1, PLA revealed mostly elastic behaviour until failure and resisted the highest tensile load. Between adhesives, Araldite^® 2015 presents better results than Sikaforce^® 7752. For L_O = 20 mm, due to the adherend failure with both adhesives, the slight P_m difference can be explained by the behaviour and properties of each adhesive. Sikaforce^® 7752 is a more ductile and flexible adhesive than Araldite^® 2015, as shown in Section 2.2. Thus, by increasing L_O with Sikaforce^® 7752, it can better distribute stresses over the adhesive layer, while Araldite^® 2015 stiffness leads to higher stress concentrations at the overlap edges, promoting premature failures. With ABS and PETG, same conclusions can be obtained with both adhesives. Specifically, a superiority of P_m is noted with Araldite^® 2015, and P_m values become similar as L_O increases.

Both experimental and numerically, the values of P_m present the same behaviour as L_O increases for the different adherend/adhesive combinations. As the failure mode differs with the increase in L_O, namely for L_O = 5–10 mm (cohesive in the adhesive to adherend failure), the P_m increase is significant for the PLA joints. On the other hand, when the same adherend failure always occurs at L_O = 10–20 mm for the ABS and PETG joints, no evolution of P_m is noticed, showing that the full strength of the adherends was reached. The numerical P_m predictions for the SJs with PLA adherends and Araldite^® 2015 (Figure 12a) exceeded the experimental results by 9.9%, 7.3%, and 8.5% for L_O = 5, 10, and 20 mm, respectively. For the SJs using the Sikaforce^® 7752, the numerical P_m was overestimated between 9.0% and 10.3% across all evaluated L_O values. The data for ABS adherends and Araldite^® 2015 are presented in Figure 12b. The numerical results overestimated P_m by 10.7, 9.5, and 9.8% compared to the experimental values for L_O = 5, 10, and 20 mm, respectively. Using Sikaforce^® 7752, the ABS joints showed numerical P_m overestimations of 11.6, 10.9, and 10.4% for L_O = 5, 10, and 20 mm, respectively. Finally, in Figure 12c, for the SJs using PETG and Araldite^® 2015, numerical P_m values were 10.3, 9.6, and 8.8% higher than experimental values for L_O = 5, 10, and 20 mm, respectively. With Sikaforce^® 7752, the numerical P_m values were higher than the experimental values of 10.0, 10.2, and 9.5% for L_O = 5, 10, and 20 mm, respectively.

Generally, P_m differs from 7% to 10% between numerically and experimentally results. In more detail, the difference in intervals are 7–9% for PLA and 9–10% for ABS and PETG. It is safe to say that the tests were precise, considering possible errors in the fabrication of the joints and misalignments during traction tests during the experimental campaign. Experimentally, the best P_m values were found for the PLA and Araldite^® 2015 for L_O = 10 mm (4.16 ± 0.18 kN) and numerically for the same conditions except for L_O = 20 mm (4.48 kN). The P_m difference between L_O = 10 and 20 mm was only 0.8% (experimentally) and 0.2% (numerically).

3.3. Assembly Stiffness

The assembly stiffness (k) can be approximated by determining the gradient of the P-δ curve during the early phases of loading. To obtain the maximum k (K_m), it is necessary to calculate k using the initial experimental data, where the force differences are significant, and the deformations are minimal. Figure 13 presents a comparison between the experimental and numerical K_m values obtained for the SJs using various adherends and adhesives. The observed increase in stiffness is attributed to the increased joint angle. A decrease in the joint angle corresponds an increase in the adhesive area, which allows for a more uniform distribution of stresses over the adhesive layer and enhances its resistance to shear stresses. This occurs while the adherends remain intact prior to any plastic deformation. Empirically, the collective behaviour varied depending on the type of adherend in the following manner:

• For joints with PLA and adhesive Araldite^® 2015, it is observed that K_m values are higher with L_O = 5 mm than with L_O = 10 mm, specifically, 0.5% higher. For L_O = 10–20 mm, the stiffness increases by 1.6%. Using the adhesive Sikaforce^® 7752, K_m was higher at L_O = 5 mm than at L_O = 10 mm, possibly due to better fusion of the layers, allowing it to withstand more load without deforming. The difference between L_O = 5 and 10 mm was 0.9%. Between L_O = 10 and 20 mm, a stiffness increase of 2.8% was recorded. In the joints with PLA, a higher K_m was recorded with the adhesive Araldite^® 2015 when compared to Sikaforce^® 7752 by 1.8%, 2.2%, and 1.1% for L_O = 5, 10, and 20 mm, respectively.
• In the joints with ABS adherends and adhesive Araldite^® 2015, a higher stiffness was recorded with L_O = 5 mm than with L_O = 10 mm, specifically, 0.6% higher. Between L_O = 10–20 mm, an increase of 1.7% was recorded. For the adhesive Sikaforce^® 7752, the increase in K_m was gradual at 5.8% between L_O = 5 and 10 mm and 1.4% between L_O = 10 and 20 mm. With ABS, the stiffness of the adhesive joints with Araldite^® 2015 was 8.0, 1.5, and 1.8% higher when compared to the stiffness of the joints with Sikaforce^® 7752 for L_O = 5, 10, and 20 mm, respectively.
• The joints with PETG adherends recorded a higher stiffness for both adhesives with L_O = 5 mm than with L_O = 10 mm. For the adhesive Araldite^® 2015, the difference was 0.7%, and it was 0.8% for the adhesive Sikaforce^® 7752. Between L_O = 10 and 20 mm, a 3.9% increase in stiffness was recorded with the adhesive Araldite^® 2015 and 1.3% with the adhesive Sikaforce^® 7752. For PETG, the percentage differences were 2.0%, 2.2%, and 4.9% higher with adhesive Araldite^® 2015 than with Sikaforce^® 7752 for L_O = 5, 10, and 20 mm, respectively.
• In terms of the CZM stiffness numerical model results, it can be observed (in Figure 13) that the values are overestimated for all investigated scenarios but faithfully predict the trend of the stiffness. Here, the PETG samples have the highest value, followed by the ABS and PLA samples. The overestimated values are 18–21% for PETG, 20–26% for ABS, and 20–22% for PLA. Also, similar to the experimental results, within each group, the joint size and adhesive seem to be irrelevant, whereas the adherend is the biggest contributor to the stiffness value.

Figure 13 presents the values of K_m obtained numerically and experimentally in SJs for the different adhesives and adherends as a function of L_O. SJs exhibit very similar stiffness values under different conditions. As the inclination of the adhesive face always presents a peeling stress component, this component will be responsible for the initial slope in the elastic zone of the P-δ curves.

In the joints with PLA, the slopes of the experimental and numerical P-δ curves (Figure 10a and Figure 11a) in the elastic zone remain unchanged with the increase in L_O; however, the stiffness obtained numerically is higher than that obtained experimentally. With the adhesive Araldite^® 2015, the numerical stiffness is 24% higher than the experimental value. With the adhesive Sikaforce^® 7752, the numerical superiority is 26%.

ABS joints show that the increase in L_O does not change the slope of the elastic regions of the experimental and numerical P-δ curves (Figure 10b and Figure 11b) for both adhesives used. The numerical stiffness value was always higher than the experimentally obtained value. For the adhesive Araldite^® 2015, the increase was 26.0%, and the increase was 28.7% with the adhesive Sikaforce^® 7752.

For the PETG adhesive, the phenomenon persists, and the slopes of the elastic region of the experimental and numerical P-δ curves (Figure 10c and Figure 11c) follow a similar trend to those observed with PLA and ABS. The numerically obtained stiffness is also higher than the experimental values for both adhesives. For the adhesive Araldite^® 2015, the numerical stiffness increase was 24.3% compared to the experimentally obtained values. For Sikaforce^® 7752, the numerical stiffness showed an increase of 26.1% compared to its experimental data.

In Figure 13, it can be observed that among the adhesives, the stiffness of the joints with adhesive Araldite^® 2015 is slightly higher than those obtained with Sikaforce^® 7752. The differences between the adhesives did not exceed 2.5% in the numerical analysis and did not exceed 0.3% experimentally. Among the different adhesives, it is observed that the maximum stiffness is achieved with the PLA adhesive samples, followed by ABS and PETG.

In the scarf tensile tests, two types of stresses are applied: shear and peel stress. For L_O = 5 mm, a configuration that results in a face perpendicular to the loading, the only effort to be applied is pull-off. This effort causes the adhesive layer to deform only in the direction of the loading, which can lead to faster failures compared to other L_O sizes/geometries. With the increase in L_O, the inclination of the adhesive layer will cause the shear force component to rise. The transformation occurs gradually, and this effort promotes an increase in resistance.

3.4. Failure Energy

Figure 14 shows the maximum energy (E_max) for the scarf geometries. It is observed that for joints with PLA adherends, the increase in L_O promotes the increase in E_max regardless of the adhesive used, with the E_max value for the adhesive Araldite^® 2015 being higher for all L_O values. The joints with ABS adherends also show an increase in E_max with the increase in L_O, with the adhesive Sikaforce^® 7752 presenting higher values than Araldite^® 2015 for L_O = 20 mm. The case of joints with PETG adherends behaves similarly to the PLA joints.

• In the joints with PLA and Araldite^® 2015 adhesive, an increase in E_max of 214.8% is observed between L_O = 5 and 10 mm and 64.9% between L_O = 10 and 20 mm. For L_O = 5 mm, the rupture occurred in the adhesive with a P_max relatively lower than that obtained with L_O =10 mm where the rupture occurred in the adherend. Between L_O = 10 and 20 mm, the increase is significantly reduced due to the fracture that occurs in the adherend, resulting in a similar P_max. However, with L_O = 20 mm, there is a slight plasticisation of the adherend prior to rupture, which causes the displacements at rupture to be higher with L_O = 20 mm. Thus, a higher E_max is obtained as well. For the Sikaforce^® 7752 adhesive, the fractures always occurs in the adhesive, which means that there is not much difference in the increase in E_max between the different L_O values. Between L_O = 5 and 10 mm, E_max increased by 220.0%. Between L_O = 10 and 20 mm, it increased by 197.4%. With the PLA adhesive, the Araldite^® 2015 adhesive shows a higher E_max than the Sikaforce^® 7752 adhesive with increases of 197.3%, 192.4%, and 62.1% for L_O = 5, 10, and 20 mm, respectively. These differences are due to the failures in the adhesive with Sikaforce^® 7752, which resulted in P_max values that were relatively lower than those obtained with the Araldite^® 2015 adhesive.
• In the joints with the ABS adherend, the increases in E_max values that were observed with the use of the adhesive Araldite^® 2015 were 53.5% and 31.8% for L_O = 5–10 mm and L_O = 10–20 mm, respectively. With the Sikaforce^® 7752 adhesive, increases in E_max of 322.3% and 95.1% were observed for the intervals of L_O = 5 to 10 mm and L_O = 10 to 20 mm. The significant increase in E_max between L_O = 5–10 mm is due to the P_max and displacements recorded with L_O = 10 mm being much higher than those recorded with L_O = 5 mm. With ABS, the Araldite^® 2015 adhesive was superior to Sikaforce^® 7752 for L_O = 5 and 10 mm, with superiority values of 234.2% and 21.5%. The difference of 234.2% is due to the P_max obtained with the adhesive Sikaforce^® 7752 being relatively lower because of the rupture occurring practically in the adhesive. However, for L_O = 20 mm, the trend is reversed, and there is a clear improvement in the Sikaforce^® 7752 adhesive joint when compared to Araldite^® 2015.
• In the joints with PETG adherends using the Araldite^® 2015 adhesive, the increase in E_max was 32.9% and 175.0% for L_O = 5–10 mm and L_O = 10–20 mm, respectively. The increase of 175.0% is due to the plasticisation observed in the adherends, which caused a significant increase in the displacement at rupture. With the adhesive Sikaforce^® 7752, the increases in E_max observed were 186.4% and 17.9% for L_O = 5–10 mm and L_O = 10–20 mm, respectively. The increase of 186.4% is due to the different modes of failure, including cohesive failure in the adhesive with L_O = 5 mm and adhesive failure with L_O = 10 mm, which cause substantial increases in the P_max and displacements in the adhesive failure. The joints with PETG show a superior performance of E_max with the adhesive Araldite^® 2015 over Sikaforce^® 7752 with increases of 189.3%, 34.2%, and 213.0% for L_O = 5, 10, and 20 mm, respectively. Among the different adherend materials used in the applications of adhesive Araldite^® 2015, PLA showed the highest energy absorption, followed by PETG and ABS. For the Sikaforce^® 7752 adhesive, the order of performance was ABS, followed by PLA and PETG.
• The numerical CZM fracture energy model predictions show that, for all adherends and adhesives, as the SJ angle increases, so does the fracture energy. Overall, the numerical predictions underestimate the fracture energy when compared with the experimental data. PLA samples show differences of 21–44%, ABS samples show differences of 22–59%, and PETG samples show differences of 12–75%.

4. Conclusions

This work investigated the tensile performance of scarf adhesive joints formed with 3D-printed polymer adherends (PLA, PETG, and ABS) bonded using Araldite^® 2015 and Sikaforce^® 7752 adhesives. Experimental tests and numerical CZM analyses were performed to evaluate joint strength, stiffness, and failure mechanisms for three scarf angles corresponding to bond lengths of 5 mm, 10 mm, and 20 mm. The main findings, which provide guidelines for the future design of larger and mechanically robust structures in additive manufacturing applications, are as follows:

• Joint failure modes: Cohesive failures in the adhesive layer predominantly occurred in joints with PLA adherends and Sikaforce^® 7752, while adherend failures were more common in joints with ABS and PETG, regardless of the adhesive.
• Peak load numerical results showed strong agreement with experimental data, with peak load deviations of 7–10%, demonstrating the reliability of the CZM approach to predict the peak load of SJs.
• Stiffness values are similar when comparing the performance of the same adherend regardless of the L_O value or the adhesive. This shows the insensitivity of the SJ geometry and adhesive to the stiffness results. The PETG adherend shows the highest SL joint stiffness, followed by ABS and then PLA.
• Stiffness values are numerically overpredicted by 18–26% for all joint scenarios. However, the trend of the experimental results is captured, showing that the SJs depend more on the adherend rather than on the adhesive or joint angle.
• The joints with the Araldite^® 2015 adhesive exhibit higher energy than the Sikaforce^® 7752 overall. Among the adherends, PLA was the material that allowed the joints to absorb more energy.
• The numerical CZM models underpredict the fracture energy for all scenarios with predictions falling between 12% and 75%. This result shows that the current model does not perform well to predict the fracture energy and that more research is required for the prediction of this characteristic.
• A major advancement in the state-of-the-art described in Section 1 is presented, namely by validating the CZM approach on a new geometry and performing a systematic study on the effect of adherend and adhesive effect, including both failure modes and mechanical behaviour.
• Overall, the present investigation shows that the application of the CZM method, as implemented herein, to predict adhesive joint performance is excellent for determining the peak load at failure and still acceptable to analyse the fracture stiffness but falls very short on the prediction of the failure energy. Further investigation is required for the implementation of this methodology to properly predict all the relevant fracture properties.