Advanced Interface Modeling and Characterization of Thermoplastic Fusion Bonds for Sustainable Structural Applications: An In-Depth Review

Abstract

In the transition toward the circular economy and high-rate manufacturing, thermoplastic composites (TPCs) are increasingly outperforming conventional thermosets due to their superior fracture toughness, recyclability, and rapid processing capabilities. Among available joining techniques, fusion bonding stands as the main mechanism for structural integration, as it bypasses the fundamental limitations of traditional assembly: the weight penalties and stress concentrations inherent in mechanical fastening, as well as the long cycle times and interfacial weaknesses often associated with adhesive bonding. This paper provides a comprehensive evaluation of welded TPC joints through a dual-methodological approach: a historical narrative review tracing the evolution of fusion bonding principles, and an in-depth literature review of 25 key articles published since 2015. The analysis focuses on the intersection of experimental characterization—quantifying interfacial strength and fracture energy—and numerical modeling techniques, such as Cohesive Zone Modeling (CZM) and progressive damage analysis. By categorizing recent advancements into specific thematic pillars, this study correlates process-induced phenomena with macro-scale mechanical performance and virtual predictive accuracy. The findings synthesize decades of foundational knowledge with cutting-edge research trends, highlighting the transition from empirical testing to computational design. This work serves as a roadmap for achieving standardized, high-performance thermoplastic assemblies in safety-critical applications.

1. Introduction

1.1. General Background

Thermoplastics constitute a fundamental category of polymeric materials, distinguished by their ability to undergo reversible phase transitions upon the application of thermal energy. Unlike thermosets, which rely on permanent chemical cross-links, the molecular architecture of thermoplastics consists of long linear or branched hydrocarbon chains held together by relatively weak intermolecular forces, such as van der Waals interactions or hydrogen bonding. This structural configuration allows the material to soften and become viscous when heated above its glass transition temperature or melting point, subsequently regaining its solid-state properties upon cooling [1,2]. This thermal reversibility enables thermoplastic composites (TPCs) to be repeatedly melted, reshaped, and recycled, offering high versatility, durability, chemical resistance, and cost-effectiveness for diverse applications such as packaging, pipes, and auto parts. Moreover, their ability to be molded into complex shapes, combined with properties such as impact resistance and specific strength, makes them superior to metal in many applications [3,4,5].

However, the shift from individual components to complex structural assemblies requires robust joining strategies that can maintain the integrity of the polymer network. Traditional methods, such as mechanical fastening and adhesive bonding, are frequently employed but carry inherent limitations; mechanical fasteners typically introduce deleterious stress concentrations and add parasitic weight, while adhesive bonding demands rigorous, time-consuming surface preparation and is usually limited by chemical compatibility [6,7]. In response to these challenges, fusion bonding, or welding, has emerged as a highly efficient alternative that exploits the melt-reprocessability of the thermoplastic matrix to create a monolithic joint [8].

The fundamental principle of fusion bonding involves applying localized heat at the interface of two components to increase molecular mobility, combined with consolidation pressure. The success of this process counts on the phenomenon of autohesion, where polymer chains from the mating surfaces undergo intermolecular diffusion and entanglement across the interface, often referred to as “interface healing” [9,10]. Various technical approaches exist to deliver the necessary thermal energy, including ultrasonic, induction, resistance, and laser welding. While these techniques differ significantly in their physical mechanisms—ranging from high-frequency mechanical vibrations to electromagnetic excitation—they all converge on the objective of achieving a controlled melt state for optimal molecular coalescence [11,12,13,14]. Consequently, fusion bonding achieves a level of molecular continuity that can rival the strength of the parent material, establishing itself as a vital technology in modern manufacturing [15].

1.2. Problem Statement and Formulation of the Research Questions

The integration of TPCs and fusion bonding is a strategic priority for high-performance sectors seeking to meet future market demands while enhancing recyclability and overall performance [16]. In the aerospace sector, TPCs are candidate materials for primary structures and space exploration missions [17,18]. For instance, the NASA Game Changing Development project has assessed resistance, induction, and ultrasonic welding for large-scale truss structures, highlighting the potential to reduce the complexity of on-orbit manufacturing [18,19]. Similarly, the global market for ultrasonic welders and fusion bonding equipment is projected to grow steadily, driven by vehicle electrification, light-weighting, and the packaging industry’s transition toward mono-material films [20]. This adoption is a direct response to the escalating environmental crisis; as shown in Figure 1, the continued rise in global emissions—particularly within the transport and industrial sectors—necessitates a shift toward materials with circular lifecycles. In this context, the recyclability and process repeatability of thermoplastics afford a critical pathway to lower emissions compared to traditional thermosets or metals [21,22].

Over the past decades, extensive research has addressed the processing aspects of these materials. Ageorges et al. [17] and Yousefpour et al. [8] provided foundational reviews identifying intimate contact, molecular chain diffusion, and consolidation as the primary stages controlling joint formation. More recent studies emphasize the industrial relevance of these processes [24] and the move toward hybrid joining processes that combine ultrasonic energy with laser or thermal methods [25]. However, despite these advances in manufacturing, welded thermoplastic joints exhibit complex mechanical behavior that is strongly dependent on processing conditions and interfacial quality.

A remarkable gap remains in the mechanical characterization of the cohesive bond formed at the welded interface [26]. Most existing studies evaluate joint performance using global metrics such as lap shear or tensile strength, which afford limited insight into underlying failure mechanisms [27]. In particular, the cohesive properties of the fusion bond—specifically fracture toughness, damage initiation, and crack propagation—are not yet fully understood [28]. While Villegas et al. [12] demonstrated that failure modes vary with molecular interdiffusion, more studies are needed to quantitatively characterize cohesive zone behavior. Furthermore, the lack of standardized experimental methodologies for determining cohesive parameters hinders the development of reliable predictive models. These studies emphasize the need for accurate control of pressure and temperature histories, which are difficult to monitor directly, further complicating the structural assessment [19]. Finally, the most critical cohesive parameters for standardization are the critical energy release rates and the interfacial traction–separation law, as these properties are essential for ensuring an accurate prediction of the junction fracture behavior in industrial applications.

To tackle these problems and advance the structural certification of welded composites, this paper investigates the following Research Questions (RQs):

•

RQ1: How do variations in process parameters across different fusion bonding techniques dictate the resulting interfacial microstructure and the ultimate mechanical performance of TPC joints?

•

RQ2: In what ways can experimental characterization be integrated with numerical modeling to overcome the current limitations in determining the fracture toughness and damage initiation of welded interfaces?

•

RQ3: What are the primary challenges in testing and simulating these joints, and what are the future trajectories for standardizing their structural assessment?

1.3. Scopes and Contributions of This Work

In the context of structural thermoplastic applications, a comprehensive overview of fusion bonding techniques must extend beyond processing to encompass the ultimate structural behavior of the joints. As depicted by the exponential rise in research output over the last two decades (Figure 2), the field has reached a level of maturity where individual welding techniques are well documented. However, previous reviews [29,30,31] have predominantly focused on specific techniques or manufacturing parameters. Consequently, they fail to provide a generalized framework for the mechanical performance of the resulting bonds. Furthermore, the existing literature relies on global performance metrics that offer limited insight into underlying cohesive properties and failure mechanisms.

The primary objective of this review is to address this remarkable gap in the literature regarding the mechanical characterization of welded joints. Unlike previous investigations, this paper specifically bridges the gap between manufacturing processes and mechanical characterization by correlating process-induced phenomena with macro-scale structural integrity. To achieve this, this study integrates experimental testing procedures with advanced numerical modeling techniques to rigorously evaluate interfacial strength.

This work employs a dual-methodological framework that combines a historical narrative of foundational principles with a systematic and bibliometric analysis of contemporary research published from 2015 to the present. By identifying prevalent trends and patterns utilizing the Scopus database, this analysis enables readers to access the most pertinent information on the subject [32]. Ultimately, this review synthesizes decades of empirical knowledge into a unified guide, correlating processing parameters with mechanical performance and predictive accuracy to outline future trajectories for the computational design of high-performance thermoplastic assemblies.

1.4. Organization of the Manuscript

This paper is organized as follows. Section 1 introduced the topics of primary interest for this investigation. Section 2 will describe the transition from thermosetting to thermoplastic matrices, introducing TPCs and the classification of their joining techniques, with a focus on fusion bonding. This classification will precede the literature review to offer context and clarify the terminology used in this field. Section 3 will provide a historical overview of the topics in Section 2, utilizing a narrative approach to underline the evolution of TPCs over time. Section 4 will detail the methodology used for the in-depth literature review, the subsequent bibliometric analysis, and the results obtained from the selected papers. This review process will cover the period from 2015 to the present. Section 5 will explore the fundamental issues regarding welding techniques, experimental characterization, and numerical methods arising from the documents within the scope of the in-depth review. Section 6 will summarize the conclusions of this work, outlining prospective research avenues for TPCs and fusion bonding methodologies.

2. Thermoplastics and Fusion Bonding

2.1. The Evolutionary Transition from Thermosetting to Thermoplastic Matrices

The aerospace and automotive sectors are increasingly shifting from traditional Carbon Fiber Reinforced Thermoset (CFRTS) composites to advanced TPCs. While CFRTSs, predominantly epoxy-based, offer stability via irreversible cross-linking [33], TPCs introduce a paradigm shift with superior processing speed, recyclability, and fracture toughness [34,35]. This change stems from fundamental chemical differences. Thermosets form permanent cross-links, ensuring thermal stability but resulting in brittleness and non-recyclability. Conversely, thermoplastics have linear chains that allow repeated melting and reforming, aligning with circular economy goals [36]. High-performance thermoplastics like Polyether-Ether-Ketone (PEEK), Polyetherketoneketone (PEKK), and Polyphenylene Sulfide (PPS) have been developed to provide strength, durability, and resistance that meet or exceed those of traditional aerospace-grade epoxies [37]. Although high melt viscosity hinders impregnation compared to liquid thermosets, innovations in in situ polymerization and thermoplastic epoxies are closing this gap, offering better mechanical performance [38].

Distinct operational advantages drive TPC adoption. Their entangled polymer chains yield high fracture toughness and low toxicity ratings compared to brittle thermosets [37]. Manufacturing is considerably faster; melt-processing occurs in minutes, contrasting with the hours needed for curing epoxies [37]. Economically, TPCs exhibit an extended ambient shelf life without the cold-storage requirements of thermosets and reduce part costs through energy efficiency and autoclave elimination, despite higher raw material prices [16,37]. Still, the fiber–matrix interface remains a critical challenge, where poor adhesion can compromise load transfer. To address this, various surface treatment methods are employed to improve fiber/matrix bonding. These include plasma etching, ozone-based chemical treatment, and the inclusion of nanoparticles such as carbon nanotubes (CNTs) or graphene oxide [39]. For example, CNT deposition onto carbon fibers via chemical vapor deposition can enhance tensile and impact strength by 30% and 29%, respectively, through mechanical interlocking [40]. While polymer sizing is standard, advanced techniques like gamma and electron beam irradiation are being explored to induce covalent bonding [39].

2.2. Classification of TPC Joining Methodologies

The ability to join individual composite parts into integrated structures is essential for the fabrication of complex geometries. For thermoplastic composites, joining methods are generally categorized into three primary groups: mechanical fastening, adhesive bonding, and fusion bonding [8,41]. Mechanical fastening, involving rivets, bolts, and clamps, is a mature technology but presents several drawbacks for composites, including high stress concentrations at drilled holes and the weight penalty of the fasteners. Drilling holes also risks damaging the fibers and introducing delamination [42]. Adhesive bonding, while providing more uniform stress distribution, is particularly problematic for TPCs. Many high-performance thermoplastics, such as carbon fiber reinforced Polyaryletherketone laminates (PAEKs), are chemically inert and possess low surface energy, making them difficult to bond with conventional adhesives [43,44,45]. Achieving a strong bond often requires harsh surface treatments, such as grit-blasting or ultraviolet irradiation, to increase surface roughness and chemical activity. Furthermore, in this kind of joint, the weak element is typically the adhesive, regardless of the adherend materials [46,47,48,49,50]. Fusion bonding (Figure 3) utilizes the unique ability of thermoplastics to remelt by heating the polymer matrix at the joint interface to a viscous state and applying pressure to facilitate surface contact before cooling the material for solidification and consolidation under pressure [8,51,52,53]. This technique provides substantial benefits, notably rapid production rates, reduced need for surface preparation, and structural recyclability. Moreover, avoiding stress concentrations inherently extends fatigue life and mitigates the need for overdesigned laminates. Consequently, fusion bonding exhibits vastly improved lifecycle performance, resulting in a more favorable Lifecycle Assessment relative to conventional joining methods.

2.3. Theoretical Mechanisms Governing the Fusion Bonding Process

In fusion bonding, the development of bond strength is governed by transport phenomena that transform two distinct surfaces into a unified bulk material. This process is phenomenologically described by Wool & O’Connor [54] as occurring in five sequential stages: surface rearrangement, surface approach, wetting, diffusion, and randomization [17,55]. For process modeling purposes, these stages are aggregated into two primary governing mechanisms: the development of intimate contact (covering surface rearrangement and wetting) and autohesion or healing (covering diffusion and randomization) [56,57,58].

2.3.1. Intimate Contact Development

The surfaces of thermoplastic composite plies have inherent roughness. For intermolecular diffusion to occur, these surfaces must be brought into physical proximity at the molecular level. The development of intimate contact describes the deformation of surface asperities under applied pressure and elevated temperature, thereby eliminating gaps at the interface [56]. The degree of intimate contact, $D_{ic}$, is defined as the ratio of the actual contact area to the total nominal surface area, where $D_{ic}=1$ represents perfect contact. The evolution of $D_{ic}$ is modeled based on the squeeze flow of surface asperities, represented geometrically as a series of non-uniform rectangles (Dara & Loos model [59]) or identical rectangles (Lee & Springer model [60]) [17,56,61]. The time required to achieve intimate contact ($t_{ic}$) is inversely proportional to the applied pressure ($P_{app}$) and proportional to the temperature-dependent fiber–matrix viscosity (${\mu }_{mf}$) [56].

The evolution of the degree of intimate contact under non-isothermal conditions is described mathematically as [57,62]:

$$D_{ic}\left(t\right)=R_c{\left[{\int }_0^{t_c}{\displaystyle \frac{P_{app}}{{\mu }_{mf}}}dt\right]}^{1/5},$$

(1)

where $R_c$ is a geometric parameter representing the initial surface roughness, and $t_c$ is the contact time. This formulation indicates that variations in viscosity, driven by local temperature gradients, significantly influence the rate at which intimate contact is established.

2.3.2. Autohesion

Once intimate contact is established, the polymer chains diffuse across the interface to entangle with chains from the opposing surface, a process known as autohesion or healing. This phenomenon is governed by the reptation theory introduced by de Gennes [63], which postulates that a polymer chain is confined within a virtual tube formed by the constraints of neighboring chains [17,56,57]. The chain moves via Brownian motion, and the time needed for the chain to completely escape its initial tube configuration is defined as the reptation time, $t_r$.

The degree of healing, $D_h$, is defined as the ratio of the instantaneous interfacial bond strength ($\sigma$) to the ultimate bulk strength (${\sigma }_{\infty }$). For non-isothermal processes typical of welding, the degree of healing is determined by integrating over the thermal history [56]:

$$D_h\left(t\right)={\displaystyle \frac{\sigma \left(t\right)}{{\sigma }_{\infty }}}={\left({\int }_0^t{\displaystyle \frac{1}{t_w\left(T\left(t\right)\right)}}dt\right)}^{1/4}.$$

(2)

In this expression, T represents the absolute temperature, while $t_w$ denotes the temperature-dependent welding time. The latter typically follows an Arrhenius relationship [17,56,57]:

$$t_w\left(T\right)=B_r\exp \left({\displaystyle \frac{A_r}{T}}\right),$$

(3)

where $A_r$ and $B_r$ are experimentally determined material constants.

While the reptation model is standard for amorphous polymers, healing in semi-crystalline polymers (e.g., PEEK, polycaprolactam) is more complex. Below the melting temperature ($T_m$), crystalline structures restrict chain mobility, hindering reptation [56]. Recent approaches suggest modeling this process based on the “degree of melting”: chain mobility is activated only after the crystalline fraction melts, meaning the degree of healing is proportional to the peak temperature reached relative to the melting curve [56].

2.3.3. Coupled Bonding Model

The mechanisms of intimate contact and healing are physically coupled; healing can only proceed in regions where intimate contact has already been established [57,58]. Therefore, the total degree of bonding, $D_b$, evolves as a convolution of the wetting rate and the healing degree:

$$D_b\left(t\right)=D_{ic}\left(0\right)D_h\left(t\right)+{\int }_0^t{D}_h(t-t^{\prime }){\displaystyle \frac{d{D}_{ic}\left(t^{\prime }\right)}{d{t}^{\prime }}}d{t}^{\prime }.$$

(4)

This equation accounts for the fact that healing occurs over new contact areas established at time $t^{\prime }$ [57]. A dimensionless parameter, $\Omega$, representing the ratio of the time scale for full healing to that for intimate contact, is used to determine the rate-controlling mechanism.

$$\Omega ={\left({\displaystyle \frac{P{t}_R}{\mu \left(T\right)}}\right)}^{1/5}{R}_c$$

(5)

If $\Omega <1$, the process is controlled by intimate contact; if $\Omega >1$, healing is the limiting factor [57].

2.4. Classification of Fusion Bonding Techniques

Building upon the foundational classification established by Ageorges et al. [17], which categorizes fusion bonding techniques into four primary classes based on heat generation (bulk heating, frictional heating, electromagnetic heating, and two-stage techniques), this framework is herein expanded to encompass recent technological advancements and novel methodologies (

Table 1. Classification and comparison of fusion bonding techniques for thermoplastic composites.

Technique	Advantages	Disadvantages
Bulk heating
Co-curing	No foreign material added; joint strength equal to parent laminate.	Complex/costly tooling to prevent de-consolidation.
Hot-melt adhesives	Improved gap filling for part mismatch.	Introduces a separate layer that may affect bond line thickness.
Dual resin bonding	Lower processing temperature (preserves structural integrity); reduced scatter in strength.	Limited to specific compatible resin pairs (e.g., PEI/PEEK).
Frictional heating
Spin welding	High weld quality and reproducibility; simple equipment (e.g., lathes).	Limited to circular mating surfaces; non-uniform heat distribution in solid parts.
Vibration welding	Rapid cycle times; high production rates; suitable for small/medium parts.	Risk of fiber distortion/disruption at the interface due to reciprocating motion.
Ultrasonic welding	Ultra-fast (seconds); clean and non-contact; highly automated.	Thickness limits (≈3 mm); requires EDs for focus; risk of audible noise.
Electromagnetic heating
Induction welding	Non-contact heating; suitable for long continuous welds; no susceptor needed for carbon fibers.	Challenges with edge effects (overheating); sensitive to fiber type and layup.
Microwave welding	Fast heating; capability to illuminate and join complex 3D structures.	Composites act as electromagnetic shields (shielding effect); requires susceptible materials at the interface.
Dielectric welding	Fast processing for thin structures.	Risk of bulk heating/de-consolidation of the entire joint area; weak with conductive fibers.
Resistance welding	Heat applied directly at the bond line; independent of part thickness or layup.	Resistive element stays in the weld (potential stress riser); current leakage risk in conductive parts.
Two-stage techniques
Hot plate welding	Simple, reliable, and economical; can handle complex joint geometries.	Slow cycle times; molten polymer may stick to the heated tool.
Hot gas welding	Portable equipment; flexible for large/complex structures.	Very slow process; highly operator-dependent quality.
Infrared welding	Non-contact; fast heating (≈5 s); suitable for large flat or curved areas.	Deep heat penetration may cause warpage or laminate de-consolidation.
Laser welding	Precision and flexibility; localized heat (limited Heat-Affected Zone [HAZ]); no mechanical stress.	High equipment cost; requires a laser-transparent upper part.
Solar energy	Renewable energy source usage.	Highly dependent on weather conditions and concentration optics.
Layer-based techniques
FFF	Creates complex geometries without molds/tooling; handles short and continuous fibers.	Mechanical properties are typically orthotropic; quality is highly dependent on build parameters.
FGF	Cost efficiency; high throughput; reduces waste; versatile material range.	Properties heavily reliant on precise fiber dispersion control; requires specialized granular feeding.

).

•

Bulk heating. These techniques involve bringing the entire component, or large sections of it, to its melting temperature. While effective, this usually requires complex tooling to maintain pressure on the structure and prevent de-consolidation [17].

–

Co-curing. Regarded as an ideal method because no weight is added and no foreign material is introduced, the resulting bond strength can equal that of the parent laminate [8,17].

–

Hot-melt adhesives. These involve the insertion of thermoplastic films at the bond line to improve the filling of gaps caused by part mismatch [8,17].

–

Dual resin bonding. Also known as amorphous bonding, this involves co-molding an amorphous film (e.g., Polyetherimide [PEI]) onto a semi-crystalline laminate (e.g., PEEK), allowing fusion at temperatures above the film’s $T_g$ but below the matrix’s $T_m$ to preserve structural integrity [8,17].

•

Frictional heating. These methods generate localized heat at the joint interface through mechanical work performed under pressure [17,52].

–

Spin welding. Specifically suited for components with circular mating surfaces, where one part is rotated at high speed against a fixed part [8].

–

Vibration welding. This technique generates heat through rapid linear reciprocating motion between parts held under load [8,51].

–

Ultrasonic welding. An ultra-fast process using high-frequency (20–40 kHz) and low-amplitude vibrations applied perpendicular or parallel to the interface [8,41,64]. It typically requires Energy Directors (EDs) to concentrate heat and initiate melting [8,17,64].

•

Electromagnetic heating. These techniques utilize electromagnetic fields to excite conductive, magnetic, or dielectric materials positioned at the interface [8,17].

–

Induction welding. An alternating magnetic field induces eddy currents in carbon fiber reinforcements or metallic susceptors, generating heat through Joule losses [8,17,18].

–

Microwave welding. Employs energy in the GHz range and often requires microwave-susceptible materials at the interface, as composites can act as electromagnetic shields [8,17].

–

Dielectric welding. Heat is generated by the rapid polarity reversal of a high-frequency electric field acting on polymers with a high dielectric loss factor [8].

–

Resistance welding. Based on Joule heating produced by an electric current passing through a resistive element (such as a metal mesh or carbon strip) trapped between the adherends [8,18,52].

•

Two-stage techniques. They are defined by the temporal separation between the surface heating stage and the subsequent forging/consolidation stage [8,17].

–

Hot plate welding. Surfaces are brought into contact with a heated tool (often Teflon-coated) until melted, then joined under pressure after the tool is removed [8,25,65].

–

Hot gas welding. A stream of heated gas is aimed at the joint to melt the surfaces, and a filler rod of the same polymer [8,51].

–

Radiant welding. Utilizes non-contact radiant energy sources to melt the interface [8,17]. This category includes:

*

Infrared welding. Uses high-intensity quartz lamps for rapid, non-contact heating [8,25]. A variant of this method adopts lenses to concentrate radiation for more precise melting [17,66].

*

Laser welding. The laser beam penetrates a transparent part and is absorbed by an opaque part or additive at the interface to generate localized heat [8,16,17].

*

Solar energy. Direct or concentrated solar radiation is used as a thermal source for fusion [17].

•

Layer-based techniques. These methods rely on the localized melting and deposition of thermoplastics to build structures layer by layer [67,68].

–

Fused Filament Fabrication (FFF). Also known as Fused Deposition Modeling (FDM), this is a material extrusion process in which a spooled thermoplastic filament is heated to a flowable state in a hot extruder and deposited through a small nozzle onto a build platform to construct parts layer by layer [67,69].

–

Fused Granular Fabrication (FGF). It is an advanced 3D printing technology that directly processes granular or pelletized feedstock rather than filaments, offering cost efficiency and material versatility [68].

3. Historical Perspective

3.1. Overview

Joining plastics and composites is critical to replace metallic assemblies in aerospace and automotive applications [6,8]. While eliminating joints would be ideal to avoid potential failure points and weight penalties, practical realities impose a limit on component size. Driven by manufacturing constraints and the need for accessibility, repair, and assembly, load-bearing joints remain a basic requirement in engineering applications [14,70,71].

3.2. Mid-to-Late 20th Century

The formal documentation of fusion bonding techniques began with the first reports of hot gas welding in 1940 [6], followed by the publication of foundational texts on general practice by Haim & Zade (1947) [72] and Neumann & Bockhoff (1959) [73]. By the 1960s, research in this field intensified, driven by a growing demand for high-performance structural materials [6,8,12].

A remarkable turning point occurred in 1965 with Soloff’s patent on ultrasonic welding [74]. By enabling sub-second cycle times, this technology revolutionized small-component assembly—first in the toy and electronics sectors, then in automotive applications during the 1980s. This era also saw the development of electromagnetic methods, such as dielectric and induction bonding, alongside the widespread adoption of hot plate processes for large-diameter piping and vibration welding for flat-seamed parts [6,8].

As these industrial processes scaled, theoretical breakthroughs established the physical governing laws of polymer dynamics. In 1971, de Gennes [63] introduced the reptation model, describing the thermal motion of entangled long-chain molecules. This framework was later formalized into the concept of crack healing by Wool & O’Connor in 1981 [54]. Their model enabled the prediction of the recovery ratio of mechanical properties as a function of time, temperature, and molecular weight, with initial results demonstrating that fracture stress recovery is proportional to $t^{1/4}$ [54] (see Section 2.3.2).

In the final decades of the century, focus shifted toward specialized applications and rigorous process modeling. Resistance welding was extensively scrutinized for thermoplastic composites, particularly graphite/PEEK systems [9,10]. These investigations characterized critical parameters like power density and clamping pressure while identifying constraints such as thermal non-uniformity and current leakage [9]. Parallel advancements culminated in the 1996 proposal of the Benzeggagh & Kenane (BK) criterion, which is still the industry standard for evaluating delamination under mixed-mode loading [75,76].

3.3. From 2000s to 2015

Recent decades witnessed a systematic push to optimize, model, and monitor high-performance thermoplastic composites [8,12]. Experimental studies on resistance welding of PEI laminates defined optimal processing windows and heat transfer limitations [8,9,10]. In modeling, thermodynamically consistent damage mechanics and decohesion were developed to simulate progressive delamination, using criteria derived from BK for mixed-mode loading [76]. Additionally, process control for ultrasonic welding has advanced significantly; modern systems use power and sonotrode displacement feedback to monitor physical interface transformations, enabling robust real-time optimization [12].

4. In-Depth Literature Review Methodology

4.1. Generality

Literature reviews are crucial for presenting an up-to-date understanding of a research area, especially in light of ongoing advancements in scientific and technological research [77,78]. While researchers often count on historical reviews and expert opinions to update their knowledge of a topic or issue [79], such narrative approaches may be shaped by the reviewer’s prior knowledge, assumptions, and subjective judgment [80]. This subjectivity can lead to an incomplete portrayal of the field or introduce selection bias [81].

In contrast, as highlighted by [81,82,83], the systematic review approach utilizes a replicable, scientific, and transparent procedure. By employing exhaustive searches and reliable extraction techniques, this method reduces bias and ensures scientific integrity [79,81,84]. Building on the historical background reported in Section 3, this section details the rigorous methodology applied to studies from 2015 onward. It outlines the standardized procedure for collecting, filtering, and analyzing publications on TPCs and welding techniques, complemented by a bibliometric analysis to quantitatively identify key authors, influential works, central themes, and research groups.

4.2. Implementation of the Methodology

Following the systematic review approach described by [77], the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standard [85] was employed to structure the data collection. This methodology, illustrated in Figure 4, provides a rigorous framework for identifying leading publications from 2015 to the present. The study selection followed PRISMA guidelines, beginning with an identification phase across Web of Science (WoS/WoK), Scopus, and Google Scholar using a Boolean search string (

Table 2. Search criteria logic combining manufacturing techniques and thermoplastic composite materials with fracture mechanics methodologies, excluding non-polymeric systems.

COMBINATION (use of AND between blocks): topics chosen for the search query
BLOCK 1: Joining Techniques	BLOCK 2: Selected Materials	BLOCK 3: Testing/Analyses
Search Fields: TITLE-ABS-KEY	Search Fields: TITLE-ABS-KEY	Search Fields: TITLE-ABS-KEY
“Fusion” W/3 “Bond*” “Thermoplastic” W/3 “Weld*” “Ultrasonic” W/3 “Weld*” “Induction” W/3 “Weld*” “Resistance” W/3 “Weld*” “Laser” W/3 “Weld*” “Friction” W/3 “Weld*” “Vibration” W/3 “Weld*” “Spin” W/3 “Weld*” “Conduction” W/3 “Weld*” “Hot plate” W/3 “Weld*” “Autohesion” “Inter” W/3 “Bond*” “Layer” W/3 “Adhesion”	“FRP” “CFRP” “GFRP” “Carbon fiber” “Carbon fibre” “Glass fiber” “Glass fibre” “Thermoplastic composite*”	“End*Notch Flexure” “ENF” “Cohesive Zone Model*” “CZM” “Digital Image Correlation” “DIC”
EXCLUSIONS: materials not included in the research
Search Fields: NOT TITLE-ABS-KEY
“Concrete” OR “Cement” OR “Steel” OR “Aluminum” OR “Aluminium” OR “Copper” OR “Iron” OR “Titanium” OR “Zinc” OR “Metal” OR “Hybrid”
LIMITATIONS: time frame publication year limit, 10 years
Search Fields: LIMIT-TO (PUBYEAR, “x”)
“x”: 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024, 2025, 2026

Scopus query search: (BLOCK 1) AND (BLOCK 2) AND (BLOCK 3) AND NOT (EXCLUSIONS) AND LIMIT TO (LIMITATIONS).

). Before the formal screening phase, a total of 146 records were removed, which included 115 duplicate records and 31 records identified as ineligible by automation tools; no records were removed for other reasons at this stage. This resulted in 1421 unique records identified for the screening of titles and abstracts. During this screening stage, 1367 records were excluded for failing to meet this study’s thematic scope, leaving 54 records deemed potentially relevant for inclusion. To transition from preliminary screening to a detailed eligibility assessment, these 54 records entered the “sought for retrieval” phase, which represents the formal effort to acquire the full-text manuscripts for analysis. Of these, 44 reports were successfully sought and retrieved, while 10 reports were not retrieved due to a lack of institutional access or availability. The 44 retrieved articles then underwent a meticulous eligibility assessment to ensure they met all predefined criteria. This final evaluative stage led to the exclusion of 19 reports: 18 were removed through manual keyword exclusion because their content did not align with the research objectives, and one report was excluded for not being published in English. Following this systematic filtering process, a final total of 25 articles were included in the literature review.

4.3. Bibliographic Analysis

The bibliographic analysis of the retrieved papers indexed in Web of Science and Scopus was performed using the Bibliometrix tool in R software 4.5.2 [86].

Table 3. Main information about the collected documents, including document types, contents, authors, and authors’ collaboration.

Main Information About Data
Timespan	2015:2026
Sources (Journals, Books, etc.)	16
Documents	25
Annual Growth Rate %	12.98
Document Average Age	3.60
Average Citations per Doc	10.04
Document Types & Contents
Articles	19
Conference Papers	6
Keywords Plus (ID)	314
Author’s Keywords (ID)	102
Authors & Collaboration
Total Authors	104
Co-Authors per Doc	4.80
International Co-Authorships %	28.0

reports preliminary data, including the number of journals, average publications per year, average citations per document, and the total number of authors, among other metrics, whereas

Table 4. The top 10 most cited documents in this review collection.

Most Cited Documents	Ref.	Author	Year	TC †	TC/y †
Fracture behavior of 3D-printed carbon fiber reinforced polymer composites	[67]	Yavas D	2021	97	16.17
Inter-layer bonding characterization between materials with different degrees of stiffness processed by Fused Filament Fabrication	[87]	Khudiakova A	2019	50	6.25
Performance prediction for ultrasonic spot welds of short carbon fiber-reinforced composites under shear loading	[88]	Wang K	2017	33	3.30
Design, analysis and testing of thermoplastic welded stiffened panels to investigate skin–stringer separation in post-buckling	[89]	Van Dooren KS	2023	18	4.50
Characterization and analysis of conduction welded thermoplastic composite joints considering the influence of manufacturing	[90]	Tijs B	2024	7	2.33
Influence of environmental effect on thermal and mechanical properties of welded PPS/carbon fiber laminates	[91]	Araújo IG	2019	7	0.88
Bending of fusion-bonded thermoplastic single lap joints	[92]	Liechti KM	2024	6	2.00
Restoration of strength in polyamide woven glass fiber organosheets by hot pressing: case study of impact and Compression After Impact	[93]	Saquib MN	2024	5	1.67
Effect of fiber concentration on the mechanical properties of welded reinforced polypropylene	[94]	Mofakhami E	2024	5	1.67
Towards ultra-fast and high strength structural repair of damaged thermoplastic composites: ultrasonic welding	[95]	Zhao T	2025	4	2.00

^† TC = total citations; TC/y = total citations per year.

highlights the most relevant articles selected through the systematic process.

Figure 5 displays the scientific output of the top 15 authors over the covered period. Complementing this, a factorial analysis was conducted using Multiple Correspondence Analysis (MCA) based on the authors’ keywords. In the visualization, bubble size corresponds to publication volume—ranging from single papers to larger clusters—while color intensity represents the total citations accrued per year.

Figure 6 summarizes the temporal distribution and frequency of key research topics. For each thematic item, three temporal indicators are reported, collectively affording an overview of the evolution and concentration of research activity over the chosen period. In addition, each item is associated with a frequency value indicating how many times the topic occurs in the analyzed literature. Overall, the data reveal clear differences in both research intensity and temporal trends across the issues. Themes such as “reinforced plastics” and “thermoplastic composite/(s)” exhibit the highest frequencies, indicating substantial and sustained research attention. Their corresponding temporal metrics suggest that these topics have gained prominence in the most recent phase of the literature. Other frequently occurring terms, including “welding”, “welds” and “spot weldings”, evidence their continued relevance in contemporary studies. The dataset highlights a general trend toward increased research attention from 2018 to 2024 across nearly all topics, with particularly strong growth in studies on thermoplastic composites, reinforced plastics, and modern characterization techniques.

Another remarkable result is shown in Figure 7. The factorial analysis reveals the conceptual structure of the examined literature by positioning recurrent technical terms across two principal dimensions and grouping them into coherent clusters. In more detail, the figure depicts the relationships between keywords, indicating their similarity [77,78]. The origin of the axes represents the center of the research. The axes indicate the average position of all column profiles [86]. The percentages on the Dimension 1 (43.70%) and Dimension 2 (20.22%) axes represent the data variation resulting from dimensionality reduction. The K-means algorithm was set to identify three main clusters, each shown in a different color. The first dimension primarily separates general composite processing and simulation terms from additive manufacturing applications, whereas the second dimension distinguishes between welding/joining technologies and fracture mechanics phenomena.

Cluster 1, which represents the central core of the research, involves terms associated with material types and joining processes (“thermoplastics”, “welding”, “ultrasonic welding”), as well as computational approaches such as “finite element method”, “numerical methods”, and “cohesive zone model”. This heterogeneous cluster underscores a strong focus on the intersection of welding technologies, material characterization, and numerical simulation. Cluster 2, positioned on the far right-hand side of the map, contains terms such as “3d-printing”, “tensile testing”, and “carbon fiber reinforced plastics”, reflecting specific developments in additive manufacturing and the mechanical testing of high-performance reinforced systems. Cluster 3, located in the upper quadrant of the map, includes topics related to structural integrity and failure analysis, such as “interlaminar fracture”, “fracture toughness” and “laminated composites”. Overall, the spatial distribution of clusters highlights three dominant thematic areas within the field—numerical modeling and welding processes, additive manufacturing, and fracture mechanics analysis—demonstrating how these domains collectively shape the scientific landscape of current research.

The last word analysis is shown in the thematic map (Figure 8), where the topic’s importance and the research’s relevance are more clearly evident. The thematic map places key research topics according to their Relevance Degree (centrality) on the X-axis and Development Degree (density) on the Y-axis. The Motor Themes quadrant (upper-right) is defined by highly developed and interconnected topics, notably reinforced plastics and thermoplastics, which, alongside practical applications like spot welding and theoretical frameworks such as the CZM, form the core and driving force of current research. On the other hand, the Niche Themes (Upper-Left), exemplified by fracture toughness and laminated composites, exhibit high internal cohesion but are peripheral to the broader field, representing specialized knowledge areas. The Basic Themes (Lower-Right), which include emerging, high-centrality methodologies like 3D printing and DIC, are fundamental cross-linkers that require further internal development to mature. Finally, topics in the Emerging or Declining Themes quadrant (lower-left), such as fusion bonding and fracture mechanics, currently demonstrate both low cohesion and low relevance, potentially indicating nascent areas or topics that are being superseded or absorbed by the Motor Themes. This map emphasizes the importance of the present work by illustrating coverage of both emerging and well-developed themes, evidenced by the concentration of keywords in the upper-right and lower-left quadrants.

Finally, Figure 9 illustrates the social structure of the field under study through a co-authorship network, pinpointing the associations among authors of the collected documents in this review. Each author is mentioned using the format “surname + first initial letters of their names”. The Kamada–Kawai network layout was used, and normalization was carried out using the Salton index, with clustering performed using the Louvain algorithm. The thickness of the edges is proportional to the number of shared works between the connecting authors. The colors indicate the identified clusters of common co-authorship. This visualization is helpful to identify influential research groups and authors.

5. Fundamental Issues

5.1. Summary

This section provides a rigorous evaluation of the primary challenges related to thermoplastics and fusion bonding methods highlighted in the earlier literature review. A more detailed analysis of the collected documents has been conducted to accomplish this goal. The results of this review have uncovered numerous research directions, and a conceptual map has been created utilizing the MCA and K-Means clustering algorithms. The main concerns underscored in this in-depth review and explored in this paper relate to the influence of process parameters on junction properties, the characterization of TPC welded joints, and the numerical modeling of damage and fracture.

5.2. Influence of Process Parameters on Junction Properties

Understanding the relationship between process parameters and mechanical outcomes is critical to optimize the structural integrity of TPC joints. Consequently, the following analysis focuses exclusively on the fusion bonding techniques and process parameters identified as significant within the papers selected during the literature review in Section 4. By synthesizing data from this specific corpus, this section examines how key factors—such as temperature, pressure, and welding time—determine junction properties, including bond strength and interface porosity. For a visual overview of the specific fusion bonding schemes extracted from the reviewed literature, refer to Figure 10.

5.2.1. Resistance Welding

The efficacy of resistance welding in thermoplastic composites, such as PPS and PEI carbon fiber laminates, is fundamentally governed by the precise calibration of electrical current, welding duration, and consolidation pressure [91,98]. These parameters regulate the energy dissipation needed to keep the material within a designated thermal processing window. For instance, PEI necessitates temperatures up to 378 °C [98] to ensure that the thermoplastic matrix exceeds its glass transition temperature ($T_g$) for molecular diffusion while remaining below the onset of thermal degradation.

The inclusion of an embedded metallic mesh serves as a functional reinforcement; studies indicate it can enhance the ILSS by up to 32% and elevate the $T_g$ by approximately 15% in welded PEI specimens due to the restriction of polymer chain mobility [98]. However, this heating element simultaneously introduces a localized brittle phase within the weld line, which may compromise fracture mechanics. Specifically, a reduction of approximately 28% in Mode II fracture toughness has been observed in welded PPS/carbon fiber laminates [91]. Consequently, the optimization of these variables is crucial to achieve a robust interfacial bond that preserves mechanical integrity under hygrothermal conditioning, effectively balancing polymer chain entanglement against the structural knock-down effects inherent to the resistive element [91].

5.2.2. Induction Welding

The optimization of process parameters such as energy input, current, and pressure is critical to guarantee uniform heat distribution and joint integrity in induction welding. Large-scale applications often face non-uniform temperature profiles that can compromise weld quality [99]. While preliminary studies on carbon fiber composites using Elium 188-O resin show promise, induction-welded joints currently exhibit lower shear strength than adhesive alternatives [100]. Additionally, in hybrid processes like Induction Low Shear Friction Stir Riveting, parameters are highly interdependent, requiring considerable scaling of current and pressure when transitioning from simple coupons to complex mono-stiffener elements to achieve target temperatures [101].

5.2.3. Conduction Welding

A fundamental trade-off between the applied thermal history and the resulting structural integrity characterizes the conduction welding of thermoplastic composites. Interface temperature is the primary determinant of joint quality; research on AS4D/CF-PEKK composites [90] shows that reaching a certain temperature threshold is essential to prevent the formation of “kissing bonds” and promote fiber bridging and nesting, which significantly enhance fracture toughness and failure load. Conversely, stamp temperatures need precise regulation to mitigate localized surface degradation. The subsequent consolidation phase is equally important, as the cooling rate dictates the final mechanical morphology. Rapid free cooling suppresses complete crystallization in semi-crystalline polymers, thereby promoting greater ductility and plastic deformation. This contrasts with the characteristic brittle behavior observed in slow-cooled, fully crystalline benchmarks [90].

Despite the benefits of robust interfacial bonding, the concentration of localized heat and pressure introduces residual thermal stresses. These stresses manifest as transverse inward curvature and out-of-plane imperfections, which may alter the buckling response of the assembly, as evidenced in investigations of stiffened panels [89,102]. Therefore, the optimization of contact time is mandatory; the duration must be sufficient to facilitate polymer chain diffusion across the interface while remaining short enough to prevent thermal degradation of the polymer’s modulus and tensile strength [93].

5.2.4. Vibration Welding

The quality of vibration-welded TPC joints depends on the synergistic effects of vibration amplitude, welding pressure, and time. Higher amplitudes accelerate both frictional heat generation and shear plastic deformation; however, when paired with low pressure, they can result in a widened HAZ and diminished crystal orientation [103]. While clamping pressure is vital for consolidation, it induces a squeeze flow that preferentially displaces the polymer matrix, increasing fiber density within the weld zone [94]. Although this pressure ensures intimate contact, excessive fiber confinement may trigger void nucleation and subsequent joint embrittlement [94,103].

The interplay between weld duration and oscillatory kinematics further governs melt penetration and microstructural development. Specifically, the resulting melt flow frequently reorients reinforcing fibers perpendicular to the principal loading axis—a configuration that undermines structural integrity. In studies of polypropylene composites, Mofakhami et al. [94] observed that approximately two-thirds of the fibers were oriented within the weld plane. This distribution is generally unfavorable for mechanical strength compared to the base material. Notably, although these processing parameters profoundly alter fiber morphology, they typically do not induce appreciable chemical degradation or changes in matrix crystallinity [94].

5.2.5. Ultrasonic Welding

The mechanical integrity of ultrasonic welds in thermoplastic composites relies on the combined impact of welding time, energy, pressure, and vibration amplitude, which influence interfacial heat generation and material flow. The selection of an Energy Director is pivotal; while low-melting-point EDs (e.g., TecaPEI) facilitate rapid wetting with minimal adherend damage, they may yield lower lap shear strength compared to matrix-identical EDs (e.g., neat PEI) due to thermal degradation or lower intrinsic strength [104]. In repair scenarios, resin-film EDs are normally sized slightly larger than the bond area to guarantee comprehensive heat concentration [95].

Welding energy is a key factor for joint quality, dictating the transition between failure regimes. In short carbon fiber reinforced Nylon 6, increasing energy shifts failure modes from interfacial separation to nugget shear, and ultimately to nugget pull-out [88]. While insufficient energy limits fiber flow into the weld layer, excessive energy induces matrix porosity and degradation [88]. Similarly, welding time must be optimized; for CF-PPS repairs, strength initially increases as the weldline thins, but exceeding an optimal threshold (e.g., 1.5 s) results in voids and fiber distortion [95].

Consolidation is further controlled by welding and maintenance pressure, which ensures intimate contact. For glass fiber reinforced polypropylene, higher maintenance pressures correlate with increased steady-state fracture toughness [105]. This interacts with temporal parameters, where toughness peaks at a specific duration (e.g., 0.8 s) before declining [105]. Furthermore, vibration amplitude affects cohesive properties, including bridging stress and critical slip displacement [105].

Notably, optimal parameters are not transferable from neat resins to their composite counterparts. Research indicates that parameters effective for pure polypropylene lead to suboptimal bond areas and failure loads in glass fiber reinforced variants [53]. This discrepancy arises from the fibers’ influence on melt viscosity and energy dissipation, necessitating a precise balance of energy and pressure to promote polymer diffusion without causing thermal degradation.

5.2.6. Co-Consolidation/Co-Curing

The performance of co-consolidated TPCs depends on the interplay of temperature, pressure, and cooling rates, which dictate interfacial bonding and geometric accuracy. For CF-PAEK systems, optimal bonding occurs at 350 °C for 20 min under 6 bars, producing a uniform interface with superior mechanical properties [101,106]. In comparison, studies on PEI-based joints by Liechti et al. [92] demonstrate successful fabrication at a lower temperature of 260 °C for 10 min under 5 MPa, spotlighting how optimal processing windows shift considerably based on the particular polymer matrix used.

However, excessive pressure and mold constraints can introduce geometric defects that significantly compromise structural integrity. For instance, PEI joints may develop a 45° notch at the overlap terminus during processing [92]. Under bending, this defect acts as a stress raiser; tensile loading of the notched edge reduces peak capacity by 30% and failure-initiation displacement by 33% compared to baseline configurations [92]. Thus, while pressure is essential to prevent de-consolidation [98], it must be precisely calibrated to avoid deleterious stress concentrations.

Thermal history further affects fracture toughness through matrix morphology. Slow cooling during autoclave consolidation increases crystallinity in matrices like PEKK, which enhances stiffness but may induce brittleness. Consequently, the fracture toughness of autoclave-consolidated joints can be 2.5 times lower than that of rapidly cooled conduction-welded joints, which retain a more ductile microstructure [90]. Optimization thus requires a trade-off between achieving full consolidation and avoiding geometric or microstructural embrittlement.

5.2.7. Layer-Based Techniques

In layer-based techniques, such as FGF and FFF/FDM, the mechanical performance and fracture behavior of the resulting composites are fundamentally dictated by process parameters including raster orientation, layer height, and processing temperature. Research on Flax/PP fabricated via FGF [68] and CFRPs printed via FDM [67] highlights remarkable anisotropy, where a $\pm {45}^{\circ }$ raster orientation optimizes stress distribution, whereas ${90}^{\circ }$ orientations exhibit poor performance due to weak inter-path adhesion. Fracture morphology confirms these mechanisms, with $0^{\circ }$ orientations failing via fiber pull-out and ${90}^{\circ }$ orientations through delamination. Additionally, reducing layer height and increasing processing temperature improve ILSS in materials like CF-PEEK by promoting molecular diffusion at the interface [107]. This anisotropy extends to fracture toughness; in CF-nylon, cross-layer toughness is approximately three times higher than inter-layer toughness due to fibril bridging and larger Fracture Process Zones (FPZs) [67]. Finally, in multi-material Polylactic Acid (PLA) systems, high interfacial cohesion can deflect cracks into the stiffer reinforced layers rather than propagate along the interface [87].

5.3. Characterization of TPC Welded Joints

The evaluation of mechanical performance in composite materials and their joints requires a rigorous application of static and cyclic testing protocols to determine properties such as fracture toughness, shear strength, and fatigue life.

5.3.1. Fracture Toughness Characterization

Fracture mechanics tests are necessary for assessing the delamination resistance of bonded interfaces. To characterize Mode I interlaminar fracture toughness, the Double Cantilever Beam (DCB) test (Figure 11a) is widely employed [87,90]. In this configuration, the critical strain energy release rate ($G_{Ic}$) is normally evaluated using the Corrected Beam Theory (CBT) [90], which accounts for large displacements and stiffening effects caused by loading blocks. The governing equation for $G_{Ic}$ is expressed as:

$$G_{Ic}={\displaystyle \frac{3P\delta }{2b(a+\Delta )}}\left({\displaystyle \frac{F}{N}}\right),$$

(6)

where P is the load, $\delta$ is the displacement, b is the specimen width, a is the crack length, and $\Delta$ is the crack length correction factor determined experimentally. The factors F and N correct for large displacements and load block stiffening, respectively (see Tijs et al. [90]).

In cases involving additive manufacturing where layer adhesion is crucial, the Cracked Round Bar (CRB) test (Figure 11b) offers an alternative for determining the critical stress intensity factor ($K_{Ic}$) and energy release rate, particularly to avoid mixed-mode failure commonly encountered in asymmetric DCB specimens [87]. The stress intensity factor for the CRB test is calculated based on the maximum load P, specimen radius R, ligament radius b, and geometric function $f(b/R)$ (see Khudiakova et al. [87]).

$$K_{Ic}={\displaystyle \frac{P}{\pi b^2}}\sqrt{{\displaystyle \frac{\pi ab}{R}}}f\left({\displaystyle \frac{b}{R}}\right)$$

(7)

Under plane strain state assumption, $G_{Ic}$ can be computed as:

$$G_{Ic}={\displaystyle \frac{(1-{\nu }^2)K_{Ic}^2}{E}}.$$

(8)

Experimental comparisons conducted by Khudiakova et al. [87] between DCB and CRB tests on 3D-printed PLA and CF-PLA have shown good correlation, especially for brittle reinforced materials. However, for tougher neat polymers, the CRB test may yield slightly lower toughness values because the circumferential notch precludes the formation of a plane stress state at the specimen surface, which is present in DCB specimens.

Alternatively, for 3D-printed short carbon fiber reinforced polymer composites, Yavas et al. [67] have characterized the fracture behavior through the J-integral approach, considering both elastic and plastic energy dissipation during crack propagation. This method is particularly relevant because matrix materials, such as nylon, exhibit ductility that linear elastic fracture mechanics cannot fully capture. To determine the Mode I strain energy release rate, Compact Tension (CT) tests are performed on specimens printed with specific build orientations to isolate cross-layer and inter-layer fracture modes. The total strain energy release rate J is calculated as the sum of elastic ($J_{el}$) and plastic ($J_{pl}$) components:

$$J=J_{el}+J_{pl}={\displaystyle \frac{K_I^2}{E^{\prime }}}+J_{pl}.$$

(9)

In this formulation, $E^{\prime }$ denotes the plane strain elastic modulus, which depends on the orthotropy of the printed sample. For cross-layer fracture, where the crack propagates through the filaments, the material can be treated as isotropic on the plane transverse to the print direction. However, for inter-layer fracture, the material shows transverse anisotropy. The instantaneous stress intensity factor, $K_I^{\left(i\right)}$, for the CT geometry is derived from the applied load $P_i$, specimen thickness B, width W, and instantaneous crack length $a_i$:

$$K_I^{\left(i\right)}={\displaystyle \frac{P_i}{B\sqrt{W}}}f\left({\displaystyle \frac{a_i}{W}}\right).$$

(10)

The geometric function $f(a_i/W)$ is a polynomial that accounts for the changing compliance of the specimen as the crack grows (see Yavas et al. [67]). The plastic energy component $J_{pl}$ is calculated incrementally, considering the plastic work done for the advancing crack.

$$J_{pl}^{\left(i\right)}=\left[J_{pl}^{(i-1)}+{\displaystyle \frac{{\eta }_{pl}^{\left(i\right)}}{b_i}}{\displaystyle \frac{A_{pl}^{\left(i\right)}-A_{pl}^{(i-1)}}{B}}\right]\left[1-{\gamma }_{pl}^{\left(i\right)}{\displaystyle \frac{a_i-a_{i-1}}{b_{i-1}}}\right]$$

(11)

Here, $b_i$ is the instantaneous uncracked ligament length, defined as $b_i=W-a_i$. The geometric parameters ${\eta }_{pl}^{\left(i\right)}$ and ${\gamma }_{pl}^{\left(i\right)}$ are functions of the instantaneous uncracked ligament length (see Yavas et al. [67]). The term $[A_{pl}^{\left(i\right)}-A_{pl}^{(i-1)}]$ is the increment of the plastic area under the force–displacement curve as the crack advances from $a_{i-1}$ to $a_i$.

For Mode II fracture toughness, the ENF test (Figure 12) constitutes a standard procedure [99,105]. Evaluating the Mode II strain energy release rate ($G_{\mathit{II}}$) usually involves data reduction schemes such as the Compliance Calibration Method [90,105]. Following this approach, $G_{\mathit{II}}$ is derived from the slope m of the compliance versus cubed crack length curve ($C=C_0+m{a}^3$):

$$G_{II}={\displaystyle \frac{3m{P}^2{a}^2}{2b}}.$$

(12)

A more advanced method, i.e., the Compliance-Based Beam Method (CBBM), allows for the determination of fracture energy without direct crack length monitoring by using an equivalent crack length ($a_e$) based on specimen compliance [99,105]. For standard crack propagation ($a_e<L$, where L is the half-span length), $G_{\mathit{II}}$ is calculated as:

$$G_{II}={\displaystyle \frac{9P^2{a}_e^2}{16B^2{E}_1{h}^3}}$$

(13)

where B is the width, $E_1$ is the longitudinal modulus and h is the thickness.

5.3.2. Shear and Tensile Strength Evaluation

Beyond fracture toughness, the global strength of joints is assessed using shear and flexural tests. The single lap shear (SLS) test (Figure 13) is the standard for evaluating apparent shear strength in welded joints, although the results are often affected by peel stresses and bending moments [88]. For a more pure shear evaluation, the Iosipescu shear test (Figure 14) and the Interlaminar Short Beam test are utilized [98,107]. Flexural properties are determined via three-point bending tests, where the flexural stress (${\sigma }_{Flex}$) and strain (${ϵ}_{Flex}$) are calculated based on the applied load $F_z$, span length $l_s$, width w, thickness t and deflection ${\delta }_z$ [93]:

$${\sigma }_{\mathrm{Flex}}={\displaystyle \frac{3F_z{l}_s}{2w{t}^2}},{ϵ}_{\mathrm{Flex}}={\displaystyle \frac{6{\delta }_{z}t}{l_s^2}}$$

(14)

5.3.3. Fatigue and Durability Analysis

Fatigue and durability analysis characterize the degradation of materials under cyclic loading. High-cycle and low-cycle fatigue tests monitor parameters such as the dynamic elastic modulus, secant modulus, and hysteresis loop area, which represents dissipated energy [110]. The evolution of fatigue damage, particularly delamination growth, can be modeled using a modified Paris law that correlates the crack growth rate ($da/dN$) with the maximum energy release rate per cycle ($G_{max}$):

$${\displaystyle \frac{da}{dN}}=c{\left(G_{max}\right)}^m.$$

(15)

Here, c and m are intermediate mixed-mode parameters derived from pure Mode I and Mode II fatigue data (see Sioutis & Tserpes [101]).

In scenarios involving viscoelastic heating, such as ultrasonic welding, analysis of cyclic creep strain provides further insight into the irreversible deformation and heat dissipation mechanisms that contribute to failure. Cyclic creep strain (${ϵ}_{cc}$) can be determined by the change in the mean strain after N cycles [110]:

$${ϵ}_m={\displaystyle \frac{{ϵ}_{\max }+{ϵ}_{\min }}{2}},{ϵ}_{cc}={ϵ}_m^{\left(N\right)}-{ϵ}_m^{\left(1\right)}.$$

(16)

Another critical durability metric is the residual strength after impact, typically evaluated via Compression After Impact (CAI) tests (Figure 15). The residual strength (${\sigma }_r$) is normalized against the pristine strength (${\sigma }_0$) and can be modeled as a function of impact energy ($E_i$) using a curve-fitting power law:

$${\displaystyle \frac{{\sigma }_r}{{\sigma }_0}}={\left({\displaystyle \frac{E_i^{\left(0\right)}}{E_i}}\right)}^{\beta },$$

(17)

where $E_i^{\left(0\right)}$ represents the impact energy threshold above which strength reduction begins, and $\beta$ is a curve-fitting parameter [93].

5.3.4. Strain Measurement and Monitoring Techniques

The characterization of deformation and damage evolution in TPCs and their welded joints increasingly relies on non-contact optical methods to overcome the limitations of traditional contacting sensors. While strain gauges are a standard tool, they can introduce reinforcement effects when applied to compliant polymer composites, leading to measurements that are lower than the strains actually experienced locally by the test sample. To mitigate these issues and capture full-field deformation data, DIC has become a primary technique, often complemented by Acoustic Emission (AE) for monitoring internal damage.

DIC is a non-contact optical technique that measures full-field displacement and strain distributions by tracking the movement of a random speckle pattern applied to the specimen surface [112]. This method allows for the determination of both global and local mechanical properties, which is particularly valuable for analyzing heterogeneous regions such as welds. Leveraging this ability to isolate local properties, specific volumes of interest can be defined over the base material and the weld zone. This segmentation enables the extraction of local engineering strains necessary to calculate the volumetric strain ($\Delta V/V_0$) within the weld—a crucial indicator of void nucleation and embrittlement. The local Poisson’s ratio ($\nu$) and volume change are then formulated based on the local longitudinal ($\delta l/l_0$) and transverse ($\delta w/w_0$) strains:

$$\nu =-{\displaystyle \frac{\delta w/w_0}{\delta l/l_0}},{\displaystyle \frac{\Delta V}{V_0}}={\displaystyle \frac{\delta l}{l_0}}+2{\displaystyle \frac{\delta w}{w_0}}.$$

(18)

In this context, Mofakhami et al. [94] observed that an increase in fiber content leads to strain amplification and volume increase within the weld zone.

Moreover, DIC data processing may require smoothing to reduce noise, especially when analyzing displacement field derivatives to detect damage edges. In their work, Smeets et al. [113,114] applied a half-cosine smoothing function to the strain fields ${ϵ}_{xx}$ and ${ϵ}_{yy}$, represented mathematically as:

$$f=\sum _{j=0}^n\sum _{i=0}^m\cos \left({\displaystyle \frac{i\pi x}{L}}\right)\left[A_{ij}\cos \left(j\theta \right)+B_{ij}\sin \left(j\theta \right)\right],$$

(19)

where L is the width of the specimen, and $A_{ij}$ and $B_{ij}$ are the amplitudes of the shape functions determined via linear least-squares algorithms.

Beyond basic constitutive properties, DIC is employed to evaluate fracture mechanics parameters. In fracture testing, the crack tip strain fields are visualized to estimate the size of the FPZ. This is achieved by identifying the region ahead of the crack tip where the opening strain exceeds a threshold, corresponding to the material’s critical cohesive stress [67]. Additionally, virtual extensometers can be implemented in DIC software by tracking target markers on the specimen, affording a cost-effective alternative to physical video extensometers for measuring longitudinal and transverse strains [115].

Complementary to surface strain measurements, AE monitoring detects transient elastic waves generated by the rapid release of energy from localized sources within the material, such as matrix cracking, fiber breakage, or delamination [95]. Signal intensity and duration help classify damage; for instance, low-amplitude signals suggest fiber breakage, while high-amplitude signals indicate debonding or matrix cracking. Moreover, cumulative AE energy serves as a powerful indicator of failure progression; in CF-PPS specimens, Zhao et al. [95] identified three distinct stages of degradation: an initial quiescent phase with minimal activity, followed by a gradual increase representing the steady accumulation of damage, and culminating in a sharp energy rise that signals imminent catastrophic failure.

Ultrasonic C-scan is a primary non-destructive evaluation technique for detecting internal defects in opaque thermoplastic composite joints. It is widely used to monitor damage progression during fatigue via interrupted testing, enabling mapping of crack evolution. For instance, Sioutis & Tserpes [101] used C-scans to measure crack length and symmetry in Cracked Lap Shear specimens for model validation. Post-failure, C-scans reveal failure mechanisms, as shown by van Dooren & Bisagni [102] when characterizing skin–stringer separation in omega-stiffened panels.

While effective for spatial damage mapping, ultrasonic C-scanning is considered less practical for continuous in situ monitoring during prolonged fatigue tests because it requires interrupting the test to perform the scan. This limitation has driven the development of alternative indirect measurement techniques, such as tracking global specimen compliance or using DIC to measure local surface strains, which can be correlated to internal damage growth without pausing the experiment [114]. Furthermore, in fracture toughness testing like the ENF, advanced data reduction schemes such as the CBBM have been developed to overcome the difficulties and inaccuracies associated with physical crack length monitoring via visual or C-scan methods during unstable crack propagation [99].

5.3.5. Thermal Characterization

Thermal characterization is indispensable for establishing the processing limits, service temperatures, and degradation characteristics of thermoplastic composites, utilizing Thermogravimetric Analysis (TGA), Thermomechanical Analysis (TMA), and Dynamic Mechanical Analysis (DMA) to provide complementary data. TGA measures mass change as a function of temperature to determine thermal stability and the onset of decomposition, which defines the maximum safe processing temperature [98]. It is also used to compare material stability across different forms and to confirm the complete removal of volatile components [104]. TMA measures dimensional changes under a constant load as temperature varies, yielding the coefficient of linear thermal expansion and the glass transition temperature ($T_g$), thereby delivering crucial data on the material’s dimensional stability [91,98]. DMA characterizes viscoelastic behavior by applying an oscillating force, allowing for the quantification of the storage modulus ($E^{\prime }$—elastic), loss modulus ($E^{″}$—viscous), and the loss factor ($\tan \delta$), which is highly sensitive for determining $T_g$ and detecting subtle microstructural changes induced by processing or environmental aging [91]. These three techniques collectively ensure that composite processing and application remain within the material’s optimal thermal and mechanical performance envelope.

5.4. Numerical Modeling of Damage and Fracture

Beyond the above-mentioned experimental techniques for TPC joints characterization, the design, analysis, and certification of TPC structures rely heavily on advanced computational models. More specifically, the macroscopic load–displacement responses and strain measurements obtained from the fracture tests, presented in the previous section, are processed to extract the critical strain energy release rates and peak interfacial strength. These experimental outputs directly dictate the shape and the area under the traction–separation curve, thereby serving as the fundamental constitutive inputs required to define structural analysis approaches. These ones provide indispensable insights into complex phenomena, such as internal damage progression and stress distribution, which may be difficult or impossible to observe directly. They enable engineers to predict performance, optimize designs, and validate the integrity of TPC joints and structures with greater confidence.

5.4.1. Cohesive Zone Modeling

CZM is a prevailing method for simulating the complete fracture sequence—encompassing damage initiation through to propagation—within TPC interfaces [88,90]. This technique idealizes the bonded region by applying traction–separation laws (TSLs), which relate cohesive traction across the interface to the corresponding separation displacement.

The predictive accuracy of this method strongly depends on the mathematical form of the TSL chosen to represent the material’s constitutive behavior. While bilinear TSLs are often employed for simulating brittle fracture [67,90,101,106], evidence suggests that trapezoidal TSLs deliver superior fidelity in capturing the ductile behavior and FPZ characteristic of welded joints [105]. For example, in ENF testing of conduction-welded joints [90], the terminal portion of the cohesive law has been identified as the principal mechanism driving the gradual nonlinearity in the load–displacement response, requiring a shift from linear softening to more complex shapes to accurately predict stable crack propagation.

A trapezoidal traction–separation response is frequently adopted to define the deformation and damage growth process in thermoplastic welding layers, offering a more comprehensive representation than simple bilinear models by incorporating both bridging and cohesive zones. In the work by Ahmadi et al. [105], a trapezoidal model is formulated to overcome stress singularities at zero displacement by introducing an initial linear stiffness K. The constitutive equation for stress $\tau$ is defined as:

$$\tau =(1-D)K\delta ,$$

(20)

where $\delta$ is the displacement and D is the damage. Denoting with ${\delta }_{0,\mathit{II}}$ the displacement at the end of the linear elastic phase, ${\delta }_{1,\mathit{II}}$ the displacement at the end of the stress plateau region, and ${\delta }_{C,\mathit{II}}$ the critical displacement at which complete failure occurs; the damage D evolves through distinct stages according to the following relation:

$$D\left(\delta \right)=\left\{\begin{array}{cc}0\hfill & \mathrm{if}0<\delta \le {\delta }_{0,\mathit{II}}\hfill \\ 1-{\displaystyle \frac{{\delta }_{0,\mathit{II}}}{\delta }}\hfill & \mathrm{if}{\delta }_{0,\mathit{II}}<\delta \le {\delta }_{1,\mathit{II}}\hfill \\ 1-{\displaystyle \frac{{\delta }_{0,\mathit{II}}}{\delta }}\left({\displaystyle \frac{{\delta }_{C,\mathit{II}}-\delta }{{\delta }_{C,\mathit{II}}-{\delta }_{1,\mathit{II}}}}\right)\hfill & \mathrm{if}{\delta }_{1,\mathit{II}}<\delta <{\delta }_{C,\mathit{II}}\hfill \end{array}\right..$$

(21)

To model the interface of 3D-printed composites, Wang et al. [68] utilize a bilinear constitutive model defined by zero-thickness cohesive elements. The tension–displacement relationship is governed by normal and shear strains (${\epsilon }_n,{\epsilon }_s$) derived from crack opening displacements (${\delta }_n,{\delta }_s$) and initial thickness $T_0$, such that ${\epsilon }_n={\delta }_n/T_0$ and ${\epsilon }_s={\delta }_s/T_0$. The stress tensor t is related to these strains via the stiffness matrix, $t=K·\epsilon$. The damage is represented by a scalar variable $D_1$, which evolves according to a linear softening law:

$$D_1={\displaystyle \frac{{\delta }_m^f({\delta }_m^{max}-{\delta }_m^0)}{{\delta }_m^{max}({\delta }_m^f-{\delta }_m^0)}},$$

(22)

where ${\delta }_m^{max}$ is the maximum effective displacement attained, ${\delta }_m^0$ is the effective displacement at damage initiation, and ${\delta }_m^f$ is the effective displacement at complete failure.

In the context of ultrasonic welding performance, Wang et al. [88] describe a mixed-mode failure criterion for surface-based cohesive layers. This criterion combines Mode I and Mode II energy release rates, $g_I$ and $g_{\mathit{II}}$, normalized by their respective toughness values $U_I$ and $U_{\mathit{II}}$. The failure condition is met when the sum of these ratios equals unity:

$${\displaystyle \frac{g_I}{U_I}}+{\displaystyle \frac{g_{\mathit{II}}}{U_{\mathit{II}}}}=1.$$

(23)

The energy release rates are calculated by integrating the traction–separation curves, where $g_I=\int \sigma \left({\delta }_n\right)d{\delta }_n$ and $g_{II}=\int \tau \left({\delta }_t\right)d{\delta }_t$.

Sioutis & Tserpes [101] extend CZM to simulate fatigue crack growth in co-consolidated thermoplastic joints using the modified Paris law (Equation (15)) presented in Section 5.3.3. Numerical degradation is implemented through a cumulative damage parameter, $d_{tot}$, which is the sum of static ($d_s$) and fatigue ($d_f$) damage variables. The stress state of an element is then degraded according to:

$$\sigma =(1-d_{tot}){\sigma }_{max}.$$

(24)

While frequently associated with the Virtual Crack Closure Technique (VCCT), the BK criterion is also relevant to the definition of critical energy release rates in cohesive modeling (see Tijs et al. [90]). Van Dooren & Bisagni [89,102] describe the BK criterion for calculating the critical equivalent strain energy release rate, $G_{equivC}$. This is formulated as:

$$G_{equivC}=G_{\mathit{IC}}+(G_{\mathit{IIC}}-G_{IC})·{\left({\displaystyle \frac{G_{\mathit{II}}+G_{\mathit{III}}}{G_I+G_{\mathit{II}}+G_{\mathit{III}}}}\right)}^{\eta },$$

(25)

where $\eta$ is a material-dependent interaction parameter. Fracture or node release occurs when the calculated equivalent energy release rate $G_{equiv}$ exceeds $G_{equivC}$.

A considerable challenge inherent to CZM is the precise determination of cohesive parameters—specifically, initial stiffness, cohesive strength, and fracture energy. These parameters are typically ascertained through inverse methods, whereby numerical simulations are iteratively fitted to match experimental data [67,116,117,118,119,120,121,122,123,124]. The model demonstrates a high degree of sensitivity to the penalty stiffness: an excessively high value can compromise numerical convergence, whereas a value that is too low may artificially delay the predicted onset of damage [99,105]. Furthermore, simplified CZM treatments often condense all energy-dissipating mechanisms (e.g., plasticity, fiber bridging) into a singular interfacial element. This simplification can introduce discrepancies between the predicted and experimentally observed FPZ dimensions if the contribution of plasticity within the adjacent bulk material is not explicitly accounted for [90].

5.4.2. Virtual Crack Closure Technique

The VCCT is a computationally efficient method utilized primarily for the analysis of progressive failure in large-scale structures, such as the prediction of skin–stringer separation in post-buckling stiffened panels [125,126]. The VCCT determines the strain energy release rate at the crack tip by calculating the work required to close the crack infinitesimally. This is achieved by using the forces acting on the crack tip node and the relative displacements of the nodes immediately behind it. The method commonly incorporates mixed-mode failure criteria, like the BK law (Equation (25)), to account for combinations of Mode I (opening), Mode II (sliding), and Mode III (tearing) fracture [89,102].

Unlike CZM, a fundamental requisite of the VCCT is the definition of a pre-existing crack. In certain applications, such as the analysis of conduction-welded omega-stringers, this limitation is managed by designating the inherent unwelded regions adjacent to the weld as the initial crack front [89,102]. While the VCCT affords the advantage of operating with comparatively coarser finite element meshes than other techniques, it is inherently incapable of predicting damage initiation in pristine material, thereby restricting its application to structural components with known defects or initial damage [89].

5.5. Limitations in the Characterization and Modeling of Thermoplastic Composite Joints

The continued advancement of thermoplastic composite joints is currently impeded by a confluence of methodological, analytical, and characterization limitations. A primary obstacle is the distinct lack of dedicated testing standards for composite welded joints. Consequently, researchers are compelled to adapt standards originally formulated for metallic [127] or laminated [128,129] adhesively bonded structures [90]. This adaptation means that experimental procedures, such as single lap shear tests, primarily yield apparent strength values rather than the fundamental material properties requisite for generalized structural predictive modeling. Standardized mechanical evaluations introduce further complications; for instance, the Interlaminar Shear Strength test produces a state of combined stress rather than a state of pure shear [98]. Similarly, Mode I and Mode II fracture characterization—typically executed via Double Cantilever Beam [130] and End-Notched Flexure [131] tests—frequently exhibit unstable crack growth, which severely limits the rigorous monitoring of fracture toughness during propagation [98]. Crucially, these established fracture mechanics standards are optimized for the interlaminar properties of continuous laminates and fail to account for the unique interface physics of fusion-bonded joints, where phenomena such as polymer chain reptation and transcrystallinity govern structural integrity.

Beyond mechanical testing, structural inspection and material characterization methodologies have not progressed concurrently with thermoplastic joining technologies. Evaluative techniques often default to those utilized for conventional mechanical fasteners (such as rivets, bolts, and nuts), fundamentally failing to capture the behavior of modern composite joints [98]. In situ monitoring methods similarly present inherent disadvantages. Standard visual inspections are incapable of detecting internal damage within circular spot welds, while Acoustic Emission techniques lack the precision required for accurate spatial damage localization [114]. Although Digital Image Correlation provides high-resolution strain mapping, it is inherently restricted to surface deformations—thereby neglecting volumetric changes—and requires a laborious speckle painting process that renders it largely impractical for large-scale or in-service applications outside the laboratory [94,114]. Furthermore, characterizing the localized material properties of the exceptionally thin cohesive layers formed during welding remains highly problematic. Standard techniques, including nanoindentation and surface strain gauges, are frequently unsuitable for highly compliant thermoplastic polymers. Nanoindentation is hindered by load resolution limits and curve-fitting inaccuracies, whereas strain gauges can artificially stiffen the compliant composite specimens, leading to erroneously low strain readings [53]. Environmental testing is similarly restricted; accelerated aging techniques fail to accurately replicate the complex variability of natural environments, resulting in potential divergences between empirical laboratory data and actual in-service degradation [103].

From a computational and numerical perspective, establishing robust predictive models for thermoplastic welding is complicated by multi-scale temporal and thermal challenges that often incur prohibitively long simulation times [88]. To ensure computational tractability, analysts frequently employ simplifications that compromise predictive fidelity, such as neglecting residual manufacturing stresses [89] or restricting simulations to a limited number of layers, which fails to accurately replicate the complex inter-layer interactions and realistic fracture profiles observed in physical multi-layer specimens [68]. Furthermore, prevalent analytical frameworks possess distinct theoretical limitations. The cohesive zone model, while widely utilized, is heavily reliant on comprehensive material design data and tends to collapse all energy-dissipating fracture mechanisms into a single interface, thereby ignoring plasticity and complex damage evolution within the surrounding bulk material [90]. Conversely, the Virtual Crack Closure Technique necessitates the presence of a pre-existing crack, largely restricting its utility to the analysis of structures with initial damage [102]. Predicting crack propagation is further destabilized by post-failure energy dynamics that induce load oscillations within the model, irrespective of mesh refinement [101]. Compounding these issues, models consistently struggle to predict the precise influence of initial manufacturing or installation damage on structural buckling, primarily because such defects are often approximated as perfectly sharp cracks rather than evaluated based on their true geometric complexity [102]. Consequently, many current numerical models and experimental investigations remain confined to quasi-static loading scenarios or simplified single-lap configurations, ultimately failing to represent the intricate fatigue conditions and realistic damage morphologies encountered by composite structures during actual service life [95,101].

6. Conclusions and Research Perspective

The synthesized evidence from this in-depth review underscores the crucial role of fusion-bonded TPC joints in the evolution toward a circular economy and high-rate manufacturing. As the industry moves away from the weight penalties and stress concentrations inherent in mechanical fastening, as well as the protracted curing cycles of thermoset adhesives, TPCs offer a compelling alternative characterized by superior recyclability and fracture toughness. Recent bibliometric trends highlight a significant transition from purely empirical methodologies toward computational design, with a marked increase in research intensity regarding numerical modeling and welding processes since 2018. However, the mechanical integrity of these joints remains inextricably linked to process-induced phenomena. Optimizing temperature, pressure, and time needs precise calibration to promote molecular diffusion and interfacial healing while preventing polymer degradation, de-consolidation, or void nucleation. Moreover, the semi-crystalline morphology of high-performance matrices such as PEEK and PEKK is highly sensitive to thermal history, with cooling rates dictating the balance between ductility and stiffness.

Structural integrity is further challenged by geometric and microstructural defects, such as fiber reorientation and notch formation, which often arise from the pressures required to achieve intimate contact. To address these complexities, the researchers have increasingly adopted fracture mechanics approaches to determine critical strain energy release rates through DCB and ENF testing, also based on non-contact optical methods like Digital Image Correlation for localized strain field isolation. From a computational point of view, Cohesive Zone Modeling has become the primary framework for simulating interface failure, with researchers increasingly favoring trapezoidal traction–separation laws over bilinear models for their higher-fidelity representation of the Fracture Process Zone. For larger structural assemblies, the Virtual Crack Closure Technique remains a viable, computationally efficient option, despite its reliance on predefined crack geometries.

To establish fusion bonding as a standardized industrial process, future efforts must focus on defining specific test methods for junction mechanical characterization. An independent experimental protocol for determining cohesive parameters, reducing the current reliance on inverse modeling, is required. Numerical fidelity must also be enhanced by developing constitutive models that explicitly account for the plasticity and fiber bridging characteristic of tough thermoplastic matrices. Furthermore, the integration of in situ process monitoring and real-time control is essential to ensure reliability in safety-critical applications, particularly as research scales from coupon-level testing to complex structural validation. Finally, a comprehensive understanding of long-term fatigue behavior and environmental durability—especially regarding the interaction between hygrothermal conditioning and localized brittle phases—remains a prerequisite for full-scale aerospace certification. By bridging the gap between manufacturing feasibility and structural predictability, the field can solidify fusion bonding as the definitive joining technology for next-generation sustainable composite structures.