Comparisons of the Performance of Novel Lightweight Three-Dimensional Hybrid Composites against GLARE Fiber–Metal Laminate

Abstract

The objective of the work presented in this paper is to overcome several major shortcomings of the recently introduced 3D composites (3DCs) and their fiber–metal-laminate renditions (3DFMLs). A new class of lightweight, stiff, and resilient three-dimensional hybrid composites (3DHCs) is introduced in this study, referred to as “inserts enhanced 3D hybrid composites” (IE3DHCs). The performances of all configurations were characterised by conducting three-point flexural tests using a span-to-thickness ratio of 32:1. The flexural performance of 3DFMLs with different core thicknesses was first compared using GLARE-3/2-0.4 as a baseline, revealing the superior performance of 3DFMLs; the optimal performance was exhibited by 3DFML with a 3 mm 3DC core. However, the lower ductility of 3DFMLs, as well as their poorly controlled and time-consuming fabrication process were recognized. The newly developed materials (IE3DHCs) had a comparatively simpler fabrication processes with significantly higher quality control. More importantly, IE3DHCs exhibited an approximately 160% improvement in ductility and as much as a 250% improved design strength compared to 3DFMLs. These findings showcase the promising potential of IE3DHCs for future research and real-world applications. Additionally, robust finite element models were developed to simulate flexural tests and optimize future renditions of the IE3DHCs.

1. Introduction

Environmental concerns have created significant challenges in various industries, including the automotive and aerospace industry. In the past decades, engineers have worked extensively on improving the efficiency of combustion engines; however, the development of combustion engines is nearing its plateau. Moreover, unfortunately, it has been speculated that most fossil-fuel resources will be depleted soon. Furthermore, carbon emission has also been a significant concern in recent years, and many sectors and countries have set challenging goals for reducing emissions. For example, the International Civil Aviation Organization’s target is to half aviation emissions by 2050 [1]. As a result, new energy sources and low-emission structures have become critically important topics. It has been several decades since the aerospace industry recognized the importance of lightweight materials; one of the proofing points is the extensive use of advanced composite materials in the industry in recent years. It has been found that 20% weight savings of B787 can lead to an improvement in fuel efficiency by 10–12% [2]. Therefore, developing new lightweight materials with enhanced static and dynamic performances is highly relevant in the field.

Fibre–metal laminates (FMLs) are hybrid laminated material systems consisting of layers of fibre-reinforced plastics (FRPs) and thin metal sheets. FMLs were developed with the aim of taking advantage of the positive attributes of the metallic and FRP constituents. Since their introduction, various types of FMLs have been developed (e.g., ARALL, GLARE and CARALL). FMLs offer outstanding static and dynamic properties, impact resistance and fatigue resistance, and they have primarily been used in several applications in the aerospace industry (e.g., fuselage, cargo floor, radome, helicopter blades, etc.). Among the most recent examples is the use of glass-laminated aluminum-reinforced epoxy (GLARE), possibly the most widely used FML type, in the fuselage of the Airbus A380 airliner [3].

A few years ago, a novel class of FML was developed by our research group, referred to as 3D fibre–metal laminates (3DFMLs). In contrast to the planar (2D) FRP layers used in conventional FMLs, 3DFML incorporates a 3D composite (3DC) core, which is produced by impregnating a recently developed 3D E-glass fabric (3DFGF) with a resin. The 3DC is then laminated with metallic layers, rendering a 3DFML. Note that the metallic layer could be made of thin sheets of a relatively lightweight alloy (e.g., magnesium), or stiffer and relatively thinner metals (e.g., stainless steel). To further enhance the response of the 3DFML, lightweight polyurethane (PU) foam is injected into the empty channels of the 3DC, giving extra stability to the 3DC internal structures. The incorporation of the foam also significantly enhances the vibration characteristics of this class of FML. These 3DFMLs have excellent stiffness and strength-to-weight ratios due to their unique internal geometry and configuration, making them a strong candidate for applications in various industries.

Since the inception of the first rendition of 3DFML, it has undergone a systematic series of experimental and numerical investigations. Asaee et al. [4] investigated the low-velocity impact response of 3DFML and compared it to 2DFML configurations fabricated with 4, 7, and 16 layers of cross-ply FRP laminates. The specific flexural stiffness of 3DFMLs was found to be greater than that of all the 2DFMLs. Moreover, all the considered 2DFML configurations were heavier than 3DFML but exhibited a higher impact energy capacity than 3DFML. However, the specific energy absorption capacity of 3DFML was found to be on par with its heavier 2DFML counterpart made with four cross-ply FRP laminates.

Another investigation examined the low-velocity response of 3DFMLs configured with different stacking sequences and varying 3DFGF core thicknesses [5]. Specifically, 3DFMLs with 10 mm and 4 mm thick composite cores were compared, and the results indicated that 3DFML configurations consisting of 4 mm cores demonstrated a superior performance per unit weight. This was attributed to the more effective stability of the shorter-length pillars in the thinner 3DFGF core.

The axial-impact buckling response of 3DFML was also investigated in [6,7]. It was discovered that the magnesium-based 3DFML exhibited a 31% higher impact buckling capacity compared to its stainless-steel-based counterpart. Therefore, the authors recommended incorporating thicker metallic skins rather than stiffer but thinner skins. Yaghoobi et al. [8] investigated the buckling responses of stainless-steel-based 3DFML to determine the influence of its nanoparticle content, polyurethane (PU) foam density, and stainless-steel thickness. The study revealed that high-density PU foam with a 0.25 wt% graphene nanoplatelet content achieved the highest buckling capacity in a cost-effective manner. This finding suggests that a stiffer and stronger filler material would be desirable to optimize 3DFML performance.

Recently, the authors of the present study conducted a comparative analysis of the low-velocity impact and impact damage tolerance between a 3DFML with a 3 mm 3DC core and an equivalent GLARE [9]. The results indicated that the current 3DFML configuration had a comparable post-impact residual capacity to GLARE, but it underperformed in terms of its impact resistance. The authors attributed this to two factors: (i) the poor consolidation of the 3DC and (ii) the inability of the PU-foam filler material to adequately sustain the stability of the pillars under intensive loading conditions.

Despite the exemplary performance of 3DFMLs, their relatively time-consuming fabrication process has hindered their incorporation into industrial applications. Additionally, the foam injection or aspiration processes have been found to result in inconsistent distribution within the channels of the core constituent. This issue limits the competitiveness of 3DFMLs in high-performance applications, where a small discrepancy in performance margin is expected. Ironically, despite the systematic investigation of 3DFMLs’ performance under various sophisticated loading conditions, there is a clear lack of studies considering the quasi-static flexural performance of 3DFMLs.

The objective of the work presented in this manuscript is to develop lightweight 3D hybrid composites with an improved performance and a more robust and cost-effective fabrication process compared to the previously developed 3D fiber–metal laminates. As a result, the development of the coined “novel Insert Enhanced 3D Hybrid Composites” (IE3DHCs) is introduced. The work is structured into three main parts. Firstly, the flexural performances of 3DFMLs with varying 3DC core thicknesses are compared to that of GLARE-3/2-0.4. GLARE-3/2-0.4 is selected as a baseline since it has been widely used in various practical applications and research and has a similar area mass to the 3DFML. It also involves determining the optimal core thickness in the originally configured 3DFMLs. In the second phase, an effective and efficient fabrication process is developed for IE3DHCs, followed by a comparison of their flexural performance against the original 3DFMLs. This step also involves establishing their optimal core thickness and assessing their performance. Lastly, accurate numerical models are developed to evaluate the performance of the new hybrid composites, providing an effective and reliable foundation for future parametric studies.

Since a large number of material configurations and associated naming are used in this study, for the readers’ convenience, the configuration names and their relationships are displayed in Figure 1. Definitions of acronyms are listed in

Table 1. Acronyms and their definitions.

Acronyms	Full Names
GLARE	Glass-Reinforced Laminate
3DHC	3D Hybrid Composite
3DC	3D Composite
3DFML	3D Fiber–Metal Laminate
IE3DHC	Insert-Enhanced 3D Hybrid Composite
3DFML-PI	3D Fiber–Metal Laminate with Plastic Inserts
3DHC-PI	3D Hybrid Composite with Plastic Inserts

, and more detailed descriptions can be found in later parts of this manuscript.

2. Experimental Investigation

2.1. Flexural Testing

Three-point bending tests were conducted according to ASTM D7264 [10], using a span-to-thickness ratio of 32:1. This choice of a relatively large span-to-thickness ratio is in line with the characteristics of thin aircraft fuselage skins, which are prone to global instability. It allows the assessment of the materials’ flexural performances independently of through-the-thickness shear stress, providing a closer approximation to real-world structural stability. The tests were conducted on an MTS servo-hydraulic universal testing machine with a 250 kN load cell. The specimen cross-sections were homogenized (with respect to one of the materials) in order to make use of Equation (1):

$${\sigma }_f=\frac{3PL}{2b{t}^2}$$

(1)

In the above equation, ${\sigma }_f$ is the effective maximum flexural stress, $P$ is the applied load, $L$ is the span length, $b$ is the width of the beam, and $t$ is the thickness of the beam. ASTM D790 [11] suggests the use of the following equation to determine the corrected effective maximum flexural stress (${\sigma }_{fc}$) for potential large deformation, where $\delta$ is the mid-span deflection:

$${\sigma }_{fc}=\frac{3PL}{2b{t}^2}\left[1+6{\left(\frac{\delta }{L}\right)}^2-\frac{4\delta t}{L^2}\right]$$

(2)

Equation (3) can be used to calculate the maximum strain (${\epsilon }_{max}$).

$${\epsilon }_{max}=\frac{6\delta t}{L^2}$$

(3)

The flexural modulus of elasticity, $E_f$, can be determined by dividing the change in stress ($∆{\sigma }_{fc}$) by the change in strain ($∆\epsilon$) in the linear section of the stress–strain curve, as given in Equation (4):

$$E_f=\frac{∆{\sigma }_{fc}}{∆\epsilon }$$

(4)

Since the thicknesses of the material systems are specific to the configurations used in this study, flexural stiffness is used to effectively compare the flexural performances of the hybrid composites. The flexural stiffness per unit width ($k_f$) and the maximum moment per unit width, $M_{max}$, can be calculated by the following equations:

$$k_f=\frac{E_f\times t^3}{12}\mathrm{and}{M}_{max}=\frac{{\sigma }_{max}\times t^2}{6}$$

(5)

where ${\sigma }_{max}$ is the maximum flexural stress in the homogenized material.

Furthermore, the nonlinear stress–strain response of the hybrid composites can be described by a Ramberg–Osgood-type relationship [12], as mathematically represented by Equation (6):

$$\epsilon =\frac{\sigma }{E}+\beta {\left(\frac{\sigma }{{\sigma }_{max}}\right)}^n+c$$

(6)

where $\beta$ and $n$ are material parameters and $c$ is the offset strain (if any). The offset strain was zero in the tests conducted in this study. The proportional limit stress was assumed to be the same as the yield strength of the material, ${\sigma }_y$, in this study. The proportional limit stress was selected at the intersection point of a straight line, with a slope $3\%$ lower than the elastic modulus, and the curve represented mathematically by the following equation:

$$\epsilon =\frac{\sigma }{0.97E}+c$$

(7)

By combining Equations (6) and (7), the yield strength of the material can be obtained as

$${\sigma }_y={\left(\frac{3{{\sigma }_{max}}^n}{97E\beta }\right)}^{1/\left(n-1\right)}$$

(8)

2.2. Internal Structure Inspection

A BRUKER Skyscan 1276 micro-computerized tomography (micro-CT) scanner (BRUKER, Billerica, MA, USA) was used to examine the internal structure of the hybrid composite. The micro-CT scan employs X-rays to scan 360 degrees around an object, producing detailed images of the internal structure. The resulting images can present cross-sectional views of the specimen without the need for sectioning the specimen, and the material can be scanned in any arbitrary orientation and position. Micro-CT has been proven to be effective in observing the internal damage of composite panels [13,14]. The emission parameters for the scans were set at 55 kV and 200 μA, using a 0.25 mm thick aluminum filter, and the voxel size was 42 μm. These parameters were determined as optimal for the specimens through several trials.

3. Performances of 3DFMLs and GLARE

3.1. Specifications of the GLARE and 3DFMLs

GLARE3/2-0.4 with a ${\left[Al/0/90/\overline{Al}\right]}_s$ layup sequence was developed in-house, consisting of three layers of 0.4 mm thick 2024 aluminum sheets and two layers of 16 Oz [0/90] E-glass fabric wetted with a cold-cured epoxy matrix (West System room-cured 105 resin and 206 slow hardener (West System, Bay City, MI, USA)). The aluminum sheets underwent several preparational steps to ensure proper surface treatment and cleanliness. These steps included degreasing with acetone, sandblasting, rinsing with deionized water, etching with sodium dichromate, rinsing and a final wipe with acetone. Each layer was hand-laid, and the entire assembly was vacuum-bagged and left to cure under a 1-bar vacuum.

The general architecture of the original 3DFML is illustrated in Figure 2a, with its 3DC core shown in Figure 2b. The 3DC core was offered in three distinct thicknesses ($tc$ in Figure 2d), leading to the examination of three distinct 3DFML configurations within this study: namely, 3DFML2, 3DFML3, and 3DFML4. These configurations corresponded to 3DFMLs with core thicknesses of 2 mm, 3 mm, and 4 mm respectively. The details of each constituent, the vendor providing them, and the complete fabrication procedure can be found in [9].

The dimensions of the specimens are shown in Figure 2c,d. A width of 16 mm was chosen to ensure the inclusion of balanced cells of empty channels. All GLARE specimens had a width of 13 mm, following the standard. The lengths of all specimens were 20% longer than their spans, which were calculated to be 32 times the thicknesses of each configuration. Bending tests were performed on two sets of specimens for each 3DFML and 3DHC configuration. In other words, specimens were cut along both their major (i.e., along the pillar channels) and minor principal material axes (i.e., along the pillar channels and orthogonal to them). A total of 10 specimens (five in each principal direction) were used for each configuration.

3.2. Results and Discussion

All specimens’ average areal density and thickness are listed in

Table 2. Specimen thicknesses and areal masses.

Material Configuration	GLARE	3DFML2	3DFML3	3DFML4
Thickness (mm)	2.27	4.25	5.15	5.17
Areal Mass (kg/m2)	4.64	4.40	4.45	4.46

. As seen, while all the 3DFML specimens had larger thicknesses than GLARE, they were lighter. Notice that 3DFML4 was expected to be approximately 1 mm thicker than 3DFML3 due to the added length of pillar yarns. However, the longer yarns did not rise properly during impregnation and subsequent curing, resulting in a similar thickness to 3DFML3.

The flexural stiffness per unit width of GLARE was compared to that of 3DFMLs with various thicknesses, as shown in Figure 3a. Although the 3DFMLs included magnesium sheets with a significantly lower stiffness than the aluminum sheets used in GLARE, they demonstrated significantly higher overall stiffness than GLARE. The flexural stiffness of 3DFMLs evaluated in the minor direction was typically about 15% lower than that in the major direction. As expected, 3DFML3 had a considerably higher stiffness than 3DFML2, while 3DFML4 showed almost the same stiffness as 3DFML3, with the minor direction even lower than that of 3DFML3. Ironically, despite having the same thickness as 3DFML3, 3DFML4 performed marginally worse than 3DFML3. This is believed to have been caused by the fact that the longer pillar yarns in 3DFML4 did not rise adequately during the composite’s consolidation. Therefore, longer and wavier sloped pillars would be less stable than the shorter and less-sloped pillars in 3DFML3, compromising the effective transfer of through-thickness normal and shear forces and resulting in a reduced performance.

In terms of the moment capacity per unit width, 3DFMLs also showed a better performance than GLARE, despite magnesium skins having a lower strength and stiffness compared to the aluminum skins in GLARE (see Figure 3b). Furthermore, 3DFML4 and 3DFML3 exhibited higher moment capacities than 3DFML2. The moment capacity in the minor direction for all configurations was about 85% of that in the major direction. However, although all 3DFML configurations showed higher moment capacities than GLARE-3/2, the advantage of 3DFMLs became less significant compared to their comparative stiffness. The representative effective maximum stress–strain curves (in the major direction) are plotted in Figure 4. Notice that GLARE greatly facilitates the plastic deformation of its metallic constituent, while 3DFMLs showed limited plastic responses.

Using the Ramberg–Osgood relationship, the stress–strain responses of GLARE and 3DFMLs of different thicknesses can be represented using the parameters presented in

Table 3. The Ramberg–Osgood parameters and the yield moment values for the GLARE and 3DFMLs.

Material	E (MPa)	${\mathit{\sigma }}_{\mathit{m}\mathit{a}\mathit{x}}$	$\mathit{\beta }$	$\mathit{n}$	Yield Moment Capacity (Nm/m)
GLARE	43.99	579.8	0.02824	4.855	165.8
3DFML2-D1	23.35	196.7	0.009580	7.376	341.2
3DFML2-D2	22.70	189.8	0.009786	6.622	296.2
3DFML3-D1	19.88	169.4	0.006303	8.124	486.8
3DFML3-D2	16.42	146.6	0.003737	7.021	457.4
3DFML4-D1	17.35	158.6	0.005522	4.906	377.4
3DFML4-D2	19.13	157.4	0.005162	5.941	347.1

, where D1 and D2 in the material names indicate the major direction and minor direction, respectively. The yield moment capacities of all configurations were also calculated and tabulated in Table 3. As seen, 3DFMLs showed significantly higher yield moments than GLARE. Additionally, the yield moment of each 3DFML configuration was slightly higher in the major direction compared to the minor direction. This difference was likely caused by the undesired fiber waviness in the upper and lower cross-plies of 3DC, which can be observed in the micro-CT image of a 3DFML shown in Figure 5. As explained earlier, these uneven profiles along the minor material direction were developed due to the absence of consolidation pressure. Consequently, a curved surface was formed between two neighbouring rows of pillars during the fabrication process of the 3DCs. Due to this waviness in the minor direction, the 3DFML core was more flexible along its minor direction than in its major direction. As a result, the magnesium skins take a relatively larger percentage of the load when the specimen undergoes bending, causing them to yield earlier. This compromise leads to a reduction in the yield moment of all 3DFML configurations in the minor direction.

Figure 6 illustrates the failure modes of GLARE and the 3DFMLs. At the point of failure, GLARE exhibited significantly more-extensive deformation, signifying good interlaminar bonding strength. However, 3DFMLs experienced premature failure due to delamination between the magnesium layer on the compression side and the core. This can be attributed to several factors. Firstly, as mentioned earlier, the current fabrication process of 3DHC does not allow for the application of externally applied consolidation pressure, compromising the consistency and quality of interface bonds between the 3DC core and magnesium sheets, as well as resulting in uneven surface morphology, as described above. Additionally, the magnesium layers had to be bonded to the 3DC-core surfaces in separate stages, resulting in a large interface bond thickness. Therefore, the fabrication method of 3DC and the resulting hybrid materials needed to be revisited, a task carried out in this investigation, as will be presented later.

4. Development of a New 3D Hybrid Composite

This study introduces the development of IE3DHCs. In this hybrid composite, the hollow channels of the 3DC were supported by lightweight square-profiled hollow plastic inserts rather than PU foam, as depicted in Figure 7. During the initial phase of the research, the plastic inserts were fabricated using 3D printing technology with PLA material; however, these inserts can be procured from commercial sources. Once the plastic inserts were manually inserted into the channels of 3DFGF, the skin reinforcements and 3DFGF were manually laid and stacked in a specific sequence. The assembly was covered with layers of peel-ply and then sandwiched between two flat metal plates and subjected to a uniform pressure of about 37.5 kPa, and left to cure for 36 h. After curing, the panel edges were trimmed. It should be noted that the plastic inserts were manually inserted to ensure the basic concept would work; subsequently, the inserts could be seamlessly integrated into the 3DFGF during the automated weaving process of the fabric, making the fabrication process more efficient. The automated approach also yielded a class of 3DFGF that could be efficiently transported in a conventional roll form.

The new IE3DHC introduced in this study offers several advantages over the original 3DFML. Firstly, the plastic inserts can provide uniform through-thickness support, thus allowing the application of external pressure during curing to generate more effective consolidation. In addition, since these inserts can be easily integrated into the 3D fabric during its weaving process, it eliminates the need for the injection of the more-expensive foam used in the original 3DFML. This in turn also saves a considerable amount of manpower, hence reducing the overall cost of fabrication. Additionally, contrary to the two-step fabrication of the original 3DFMLs, the bonding process of the surface reinforcement can be integrated into the same process by implementing them before consolidation. This remarkably enhances the efficiency of the fabrication and improves the bond strength between surface reinforcements and the 3DC interface. Moreover, the adoption of plastic inserts enables the production of IH3DHCs either using a resin infusion process or in an autoclave to maximize the efficacy of the products. Furthermore, the plastic inserts provide additional stability to the pillars of the fabric, thus improving their buckling resistance and the overall stability of such panels when subjected to through-thickness compressive loading. Therefore, the integration of the hollow plastic inserts optimizes the fabrication process, reducing it from a three-stage process to a single-stage. This single-stage procedure and the integral inclusion of plastic inserts during fabric weaving will substantially reduce the production cost of lightweight 3DHC panels and the 3DFMLs produced using the plastic-inserted core.

For this study, three different configurations of the IE3DHC were fabricated. The first configuration, 3DFML-PI, utilized magnesium skins (where PI designates the inclusion of plastic inserts), making it a direct counterpart to the original 3DFML3. The other two configurations consisted of an E-glass/epoxy surface reinforcement (3DHC-PI), aimed at assessing the effect of skin reinforcement. Specifically, the 3DHC-PI1 (${[90/0/\overline{3DC}]}_s$) and 3DHC-PI2 (${[90/0/0/90/\overline{3DC}]}_s$) configurations had one and two layers of cross-ply (0/90) crimped E-glass fabric reinforcement on each surface. Preliminary experimental and numerical results revealed that the plies of the 3DC were prone to buckling and developed significant damage in the middle of a channel. Therefore, buckling of the 3D plies in the minor direction could be delayed or prevented when skin reinforcements with adequate flexural rigidity were incorporated. The 3DHC-PI configurations were designed with this consideration. Bending tests were conducted along both the weft and warp directions to evaluate the mechanical responses of these 3DHCs, as performed for the original 3DFMLs.

5. Finite Element Model of IE3DHCs

While several researchers have numerically assessed the responses of GLAREs and 3DFMLs [7,8,9,15], this study focuses on developing a robust numerical model to simulate the responses of the newly developed 3DHCs. The numerical models were developed and analyzed using LS-DYNA, a commercial finite element (FE) software, SMP R13.1 (Livermore, CA, USA). Due to the two-plane symmetry in geometry and boundary, only a quarter of the geometry was modelled.

5.1. Geometry and Mesh

The model is presented in Figure 8. The IE3DHC’s plies, pillars, and surface reinforcements were modelled using the thick-shell element of LS-DYNA (ELFORM = 3), while the inserts, support, and indenter were modelled using solid elements (ELFORM = −1). The mesh density along the major (weft) direction was progressively reduced towards the support side, while it was kept uniform for the specimen in the minor (warp) bending direction.

5.2. Material Models

The support and indenter were modelled using the rigid material of LS-DYNA (MAT_20). The elastic–plastic behaviour of the metallic layers was modelled using the piecewise plasticity model (MAT_24), and element erosion (deletion) was invoked using a plastic strain-based failure criterion. The corresponding material properties and stress–strain relations after yielding are provided in

Table 4. Material properties for AZ31B-H24 magnesium.

Density (kg/m3)	Elastic Modulus (Pa)	Poisson’s Ratio	Yield Strength (Pa)	Effective Plastic Strain at Failure
1722	4.48 × 1010	0.35	1.36 × 108	0.1411

and Figure 9.

Composite layers were modelled using MAT_54, which effectively models the response of an arbitrary orthotropic material in conjunction with the Chang–Chang failure criterion. Element erosion in MAT_54 follows a sophisticated algorithm and is not purely a failure-base criterion; instead, the deletion or erosion of an element is declared when one of several criteria coded in the material model is met. For an in-depth discussion of this material model, the reader is directed to [16] and LS-DYNA’s user manual and theory manual [17,18]. The mechanical properties of the FRP constituents are outlined in

Table 5. Mechanical properties of the composite materials.

Properties	3DC Ply	3DC Pillar	Unidirectional Glass/Epoxy *
$\rho \left(\mathrm{k}\mathrm{g}/{\mathrm{m}}^3\right)$	1750	1750	1750
$E_{11}\left(\mathrm{P}\mathrm{a}\right)$	1.13 × 1010	3.00 × 109	2.93 × 1010
$E_{22}\left(\mathrm{P}\mathrm{a}\right)$	1.13 × 1010	1.00 × 109	8.93 × 109
$E_{33}\left(\mathrm{P}\mathrm{a}\right)$	3.19 × 109	1.00 × 109	8.93 × 109
${\nu }_{21}$	0.05	0.05	0.0944
${\nu }_{31}$	0.05	0.05	0.0944
${\nu }_{32}$	0.05	0.05	0.42
$G_{12}\left(\mathrm{P}\mathrm{a}\right)$	1.25 × 109	1.00 × 109	2.56 × 109
$G_{23}\left(\mathrm{P}\mathrm{a}\right)$	1.25 × 109	1.00 × 109	1.70 × 109
$G_{31}\left(\mathrm{P}\mathrm{a}\right)$	1.25 × 109	1.00 × 109	2.56 × 109
${\epsilon }_{22}^{ult}$	0.080	0.12	0.0320
${\gamma }_{12}^{ult}$	0.120	0.012	0.0925
${\epsilon }_{11t}^{ult}$	0.080	0.0108	0.0696
${\epsilon }_{11c}^{ult}$	−0.080	−0.0108	−0.0417
${\sigma }_{11c}\left(\mathrm{P}\mathrm{a}\right)$	1.73 × 108	8.00 × 107	8.75 × 108
${\sigma }_{11t}\left(\mathrm{P}\mathrm{a}\right)$	3.46 × 108	8.00 × 107	1.46 × 109
${\sigma }_{22c}\left(\mathrm{P}\mathrm{a}\right)$	1.73 × 108	8.00 × 107	2.04 × 108
${\sigma }_{22t}\left(\mathrm{P}\mathrm{a}\right)$	3.46 × 108	8.00 × 107	6.67 × 107
${\tau }_{11c}\left(\mathrm{P}\mathrm{a}\right)$	6 × 107	3.00 × 107	8.46 × 107

* Each unidirectional layer in the 0/90 cross-ply glass/epoxy composite.

. Material properties within the numerical models were obtained through a combination of sources. Some were derived from experimental tests conducted specifically for this research. Others were acquired from the existing literature and well-established data sources. Additionally, certain properties were calibrated to ensure accuracy.

The plastic inserts were represented using an elastic material model (MAT_ELASTIC) characterized by a density of 1210 kg/m³, an Elastic Modulus of 1.8 GPa and a Poisson’s ratio of 0.36. Additionally, a failure strain of 0.07 was defined in ADD_EROSION.

5.3. Delamination Modelling

The interfacial delamination growth in this study was modelled using the CZM model MAT_138 (a bilinear mixed-mode relative-displacement CZM model) in conjunction with the surface-to-surface tiebreak contact by invoking OPTION 9 [19]. In this work, the bilinear cohesive model was utilized, combining both modes I and II, as shown schematically in Figure 10. The initial slopes for Mode I and Mode II were defined by the parameters $CN$ and $CN\times CT2CN,\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{l}\mathrm{y}$, in which $CT2CN$ represents the ratio of the stiffness of Mode I to that of Mode II. In this approach, once the displacement reaches a critical value (${\delta }^0$), the traction decreases linearly, simulating the crack growth and the resulting reduction in stiffness. In the tiebreak contact card, the peak tractions for Modes I and II were defined using the normal and shear failure stresses, $NFLS$ and $SFLS$, respectively. The critical displacements for mode I (${\delta }_I^0$) and Mode II (${\delta }_{II}^0$) were evaluated as ${\delta }_I^0=NFLS/CN$ and ${\delta }_{II}^0=SFLS/CN/CT2CN$, respectively. The complete debonding of the interface was declared when the traction reduced to zero, representing the ultimate displacement. The energy release rates in the corresponding modes were defined by the parameters ERATEN and ERATES, which represent the enclosed areas under the bilinear curves for Modes I and II, respectively (see Figure 10b). The mixed-mode critical displacement, ${\delta }_M^0$, and the ultimate displacement, ${\delta }_M^F$, were calculated using Equations (9) and (10), respectively.

$${\delta }_M^0={\delta }_I^0{\delta }_{II}^0\sqrt{\frac{1+{\beta }^2}{{{\left(\beta {\delta }_I^0\right)}^2+\left({\delta }_{II}^0\right)}^2}}$$

(9)

$${\delta }_M^F=\frac{2(1+{\beta }^2)}{{\delta }_M^0}{\left[{\left(\frac{CN}{ERATEN}\right)}^{PARAM}+{\left(\frac{CN\times CT2CN\times {\beta }^2}{ERATES}\right)}^{PARAM}\right]}^{-\frac{1}{PARAM}}$$

(10)

The CZM tiebreak contacts were utilized to simulate the response of interfaces between the metallic and FRP layers, as delamination is known to occur at these interfaces based on the literature and our experience. This technique was also employed to model the interfaces between the 3DC and plastic inserts, between the 3DC plies and surface reinforcements, and between the cross-ply fabrics in 3DHC-PIs. The parameters used in this model are summarized in

Table 6. CZM Tiebreak Contact Parameters.

NFLS (Pa)	SFLS (Pa)	ERATEN (J/m2)	ERATES (J/m2)
5.90 × 107	2.30 × 107	1500	2000
PARAM	CT2CN	CN (Pa)
1	0.4286	3.5 × 1012

, which are referenced from [20].

5.4. Boundary Conditions and Analysis Control

The rigid support (refer to Figure 8a) was fully restrained, and the downward motion of the indenter was defined using PRESCRIBED_MOTION_RIGID. The contacts between the specimen and the indenter, as well as between the specimen and support, were modelled using the surface-to-surface contact algorithm.

The nonlinear numerical analysis was conducted explicitly, with a crosshead rate of 1 m/s and a ramp-up stage for the initial 8% of the analysis timeframe.

6. Results and Discussion

The thickness and areal mass of all configurations were measured after fabrication and are listed in

Table 7. Thickness and areal-mass comparison of the conventional 3DFML and hybrid composites produced using 3DHC.

Configurations	3DFML3	3DFML-PI	3DHC-PI1	3DHC-PI2
Thickness (mm)	5.17	5.12	4.72	5.47
Areal mass (kg/m2)	4.45	5.88	5.26	6.32

. Despite the different manufacturing processes for 3DFML-PI, its thickness was comparable to that of 3DFML3, indicating that the plastic inserts were accurately dimensioned to fully expand the 3D fabric. The configurations with plastic inserts showed a slightly larger areal mass than the original 3DFMLs due to some epoxy leakage into the hollow channels of the inserts. This issue will be resolved when using commercially sourced injection-moulded inserts in the future, as they will be impermeable. Since the leaked epoxy was somewhat loose and mainly located near the neutral axis of the specimens, its effect on the measured results was considered insignificant. The flexural stiffnesses and moment capacities of 3DFML3, 3DFML-PI, and 3DHC-PIs are illustrated in Figure 11.

6.1. Effect of Plastic Inserts

The effects of the new proposed fabrication method can be observed by comparing the results for 3DFML-PI to 3DFML3, as shown in Figure 11. Due to their similar thicknesses, they exhibited very similar stiffnesses. Plastic inserts, similar to PU foam, have a limited contribution to the global stiffness of the composites due to their significantly lower mechanical properties compared to glass–epoxy and magnesium constituents. However, as seen in Figure 11, plastic inserts reduced the difference between the major and minor direction responses. Moreover, 3DFML-PI exhibited significant ductility, with an approximately 160% higher failure strain than 3DFML3, as seen in Figure 11c. This indicates that integrating plastic inserts had a major influence on consolidating the 3DC and enhancing the metal–3DC interfacial bond strength. The inclusion of plastic inserts in 3DFML also greatly improved the FML’s ductility, enhancing the effective plastic strain of the original 3DFML by about 460%. Furthermore, the stress–strain curves of 3DFML-PI in both the major and minor directions were modelled using the Ramberg–Osgood approach, with the model’s parameters reported in

Table 8. Ramberg and Osgood parameters for 3DFMLs and GLARE.

Configurations	E (GPa)	${\mathit{\sigma }}_{\mathit{m}\mathit{a}\mathit{x}}$	$\mathit{\beta }$	$\mathit{n}$
3DFML-PI-D1	21.08	227.2	0.03060	6.828
3DFML-PI-D2	21.21	198.6	0.01153	6.550

.

Figure 12 illustrates the yield moments of 3DFML-PI and 3DFML3. It can be observed that the yield moment capacities of 3DFML-PI were slightly lower than those of 3DFML3, which can be attributed to the slight decrease in the core thickness of 3DFML-PI. Additionally, the difference in the major and minor yield moments was eliminated in the case of 3DFML-PI. This is attributed to the improved consolidation achieved in 3DFML-PI due to the presence of plastic inserts, which facilitated the application of through-thickness consolidation pressure. By comparing micro-CT scan images of 3DFML-PI and 3DFML (Figure 13a and Figure 5), one can observe that the plies of 3DC in 3DFML-PI were much more even and flatter. This enabled the 3DC plies to exhibit similar stiffnesses in both the major and minor directions. It is believed that the flatter surfaces of the 3DC also facilitated a more uniform resin thickness between the 3DC and magnesium sheet interfaces, which likely contributed to enhancing the interlaminar bonding strength. Figure 13b presents a 3D-rendition micro-CT image of the 3DFML-PI, illustrating the evenness of the upper and lower cross-plies in contact with the magnesium sheets. Please note that the plastic inserts, due to their material type, were not clearly detected by the micro-CT scan.

More evidence of the plastic inserts’ influence on the improvement of the responses of the new IE3DHC can be observed in Figure 14. In Figure 6c, it is evident that the failure of the original 3DFMLs was primarily dominated by the delamination and local buckling of the magnesium skin on the compressive side of the FML. In contrast, as depicted in Figure 14, the failure of 3DHC involved delamination buckling of one of the magnesium skins and its adjacent plies, followed by fiber breakage and insert failure. This observation provides evidence of improved magnesium–3DC interface bond strength resulting from the applied pressure during the curing process, facilitated by the presence of the inserts.

6.2. Effect of Skin Reinforcement Material

The performance of two 3DHC-PIs (3DHC-PI1 and 3DHC-PI2), both with glass-fiber surface reinforcements, was also compared to that of 3DFML-PI with magnesium skin reinforcements. Despite 3DHC-PIs having a comparable weight to 3DFML-PI (see Table 7), 3DHC-PI1 exhibited approximately half the stiffness of 3DFML-PI. On the other hand, 3DHC-PI2 demonstrated a similar flexural stiffness to 3DFML-PI while achieving a significantly higher moment capacity.

Furthermore, as shown in Figure 15, both 3DHC-PIs exhibited an almost-purely elastic response. Additionally, 3DHC-PI1 and 3DHC-PI2 yielded similar failure strains; however, the ultimate strain values were lower than the failure strain of 3DFML-PI (see Figure 11c). This difference is believed to have been due to the more ductile nature of the magnesium skins in 3DFML-PI. Comparing the curves in Figure 15 to the curve of 3DFML3 in Figure 11c, it can be observed that 3DHC-PIs exhibited approximately a 90% increase in failure strain. If the conventional method had been used to fabricate these 3DHCs, where the surface reinforcements were bonded to the 3DC core in a separate process, the failure strain might have been comparable to that of 3DFML3. Given the elastic behaviour of 3DHC-PIs, considering the results presented in Figure 15, it is hypothesized that incorporating the original fabrication method could have compromised the ultimate moment capacity of 3DHC-PIs by approximately 47%, further confirming the effectiveness of the new fabrication method.

The failure modes of the two 3DHC-PIs are shown in Figure 16. The failure modes were primarily characterized by fiber fracture on the compression side, with no evidence of ply delamination. Interestingly, fiber failure was also observed on the tension side of 3DHC-PI1 when the hybrid system was bent in its minor direction. This occurred after the initial drop in load caused by fiber breakage on the compression side. Subsequently, the inserts provided compressive resistance, supporting the load-carrying capacity for a short while until the applied load eventually led to failure. The fiber breakage on the tension side resulted in significant load transfer to the pillars, causing debonding at the pillar/insert interfaces, as shown in Figure 16b. Despite extensive failure initiation in the 3DC skins and skin reinforcements, the plastic inserts remained intact (see Figure 16b), indicating that the skin reinforcements were relatively weak. In contrast, when 3DHC-PI2 was subjected to bending in its minor direction (see Figure 16d), a crack developed on the compression face of the surface reinforcement and propagated through the thickness of the 3DHC ply and into the plastic insert. In this case, failure occurred in all the constituents that were initially under compression. Therefore, once the failure was initiated on the compression side, the additional load did not cause failure on the tension side (within a reasonable loading range) but rather continued to crush the compression side. This mechanism resulted in the deformed specimen springing back to its nearly original shape upon release of the bending load.

6.3. Validation of the Numerical Models

The numerical results are compared with the experimental results in Figure 11 and Figure 15, showing that the developed numerical models accurately predicted the flexural stiffness, moment capacity per unit, and the elastic–plastic behaviour of 3DFML-PI and the elastic response of 3DHC-PIs. The largest discrepancy can be seen after the initiation of failure in Figure 11c and Figure 15a. Several reasons could cause these discrepancies. Firstly, some internal contacts were not defined in the model, causing the models to collapse without interference after failure. Furthermore, the quarter models used in this study assumed symmetry in deformation and failure, which may not have been the case in the actual specimens, where any fabrication defect can cause asymmetric results. Lastly, it is recognized that the material model used for the inserts cannot simulate the behaviour of 3D-printed PLA plastic accurately; this shortfall will be addressed in a future study.

Figure 17 illustrates the numerically predicted failure modes for all the investigated IE3DHC configurations, with different failure modes identified by colour schemes in the figure. The numerical predictions successfully captured the failure of the magnesium skin, plastic inserts, and pillars immediately after the peak load, closely resembling the actual failure modes depicted in Figure 14. Although pillar failure cannot be observed in the experiments, it can be implicitly understood by considering the skin failure mode and the subsequent shortening on the compression side. However, the clear delamination observed in 3DFML (as shown in Figure 14) could not be predicted by the numerical model, which can be attributed to two reasons. Firstly, the images in Figure 14 were taken after unloading, allowing the structure to partially spring back while the magnesium skins carried residual stresses. Secondly, the specimens underwent more extensive deformation in the experimental tests since the tests were terminated only after a significant decrease in the applied load was observed, causing the locally buckled magnesium skin to be further compressed and delaminated from its adjacent 3DC ply. The images in Figure 17c–f exhibit good agreement with the experimentally captured failure modes shown in Figure 16, except for the absence of debonding between the insert and the pillar in the numerically predicted images. The experimental observations revealed that the failure of the specimens was initiated on the compression side (top side of the specimens). As the applied load increased, the locally compromised segments were further compressed, straining the tension side. The combined bending stress and tensile stress led to the failure of the lower (tension) side of the specimens, creating excessive peeling stress at the pillar–insert interfaces. However, since the contacts between the inserts and neighbouring constituents were not accounted for in the numerical model, the compromised compression side was further shortened and crushed without resistance after the onset of compression failure. Consequently, no additional strain was transferred to the tension (lower) side in the numerical model. The failure on the lower side was solely caused by extensive bending deformation, explaining the absence of excessive peeling stress between the inserts and pillars in the numerical model. This difference also explains why the numerically predicted stress–strain curves abruptly decreased to zero shortly after the failure occurred, whereas the experimental stress–strain curves showed some residual capacity in the specimens after reaching the ultimate load.

7. Summary and Conclusions

The flexural performances of 3DFMLs with varying thicknesses were compared experimentally. The results showed that all 3DFMLs performed significantly better than the GLARE (GLARE-3/2-0.4). However, premature delamination failure in the 3DFML compromised the full effectiveness of its ductile metallic skins. To address this issue, a new class of 3DHC, IE3DHC, was developed by incorporating lightweight hollow square-section plastic inserts in the channels of the 3DFGF. Three IE3DHC configurations were prepared and compared to the optimal 3DFML in this study: they are 3DFML-PI, 3DHC-PI1 and 3DHC-PI2. 3DFML-PI was a direct counterpart of 3DFML, except that plastic inserts were adopted instead of PU foam in the 3DFML. The configurations 3DHC-PI1 and 3DHC-PI2 incorporated one and two layers of biaxial E-glass/epoxy surface reinforcement. Finally, a set of numerical models was developed to simulate the performance of these complex hybrid composites. A summary of the findings is presented below:

• The plastic inserts used instead of the originally used foam provided through-thickness support to the resin-wetted 3DFGF during fabrication. This allowed the application of external pressure, facilitating more optimal consolidation and resulting in improved interfacial strength.
• The improved fabrication process significantly enhanced the failure strain of the 3DFML-PI more than two-fold compared to the original 3DFML, along with a 460% increase in its effective plastic strain. These significant improvements were achieved by leveraging the ductile behaviour of the metallic constituents.
• Incorporating the hollow plastic inserts optimized the fabrication process, reducing it from a three-stage process to a single-stage one. This single-stage procedure, coupled with the integral inclusion of plastic inserts during fabric weaving, will substantially reduce the production cost of lightweight 3DHC panels.
• Comparing the performances of 3DHC-PI1 and 3DHC-PI2 against 3DFML-PI revealed that 3DHC-PI1 had a similar load capacity to 3DFML-PI but was less stiff. On the other hand, 3DHC-PI2 had a similar stiffness to 3DFML-PI but was significantly stronger, making it suitable for high-performance applications.
• Nonlinear finite element analysis conducted using models in the LS-DYNA environment successfully predicted the response and primary failure modes of IE3DHCs, aligning well with the experimental results.
• The developed model can be confidently and reliably used to investigate the performance of the newly developed IE3DHCs under various loading conditions and further optimize their performance.