3D printing of high-strength continuous carbon fiber-reinforced composites with customized layers: Failure mechanisms and manufacturing validation

Materials

The substrate is composed of Onyx, a hybrid short carbon fiber material produced by Mark Forged. When printed alone, Onyx exhibits high strength, toughness, and chemical resistance. Pure Onyx material samples have a tensile strength of 37 MPa, a bending strength of 71 MPa, and a strength and hardness that is 1.4 times greater than that of ABS. The density is 1.2 g/cm³, and the material can withstand high temperatures of 145°C.

The continuous fiber material is produced by Mark Forged (USA) using carbon fiber prepreg filaments and exhibits tensile strength of 700 MPa and flexural strength of 540 MPa. The material is capable of producing parts with strengths comparable to those of the 6061-T6 aluminum alloy. The material is characterized by an extremely high degree of stiffness and ultra-high strength, with a density of 1.4 g/cm³. It can be heated and pressurized to automatically lay down into various geometries with printing equipment. The diameter of the carbon fiber bundle is 0.35 mm, and according to scanning electron microscope (SEM) images, ²⁹ the diameter of a single fiber is 8 µm, with an estimated volume content of approximately 52% of the continuous fiber prepreg filament.

Onyx and Markforged continuous carbon fiber prepreg are selected for three main motivations. Onyx, a short carbon fiber-filled nylon composite, possesses low thermal shrinkage and high toughness, which can be printed as outer constraint frames to suppress specimen warpage and guarantee forming accuracy. The commercial carbon prepreg is pre-impregnated with homogeneous sizing agent and maintains a steady fiber volume fraction of around 52%, effectively avoiding inconsistent infiltration issues of self-made fiber filaments and securing repeatable mechanical test results. Besides, this dual-material system perfectly matches our self-developed dual-nozzle printing equipment, allowing the preparation of staggered matrix-fiber and full-fiber layup specimens to comparatively analyze interfacial bonding performance and mechanical reinforcement mechanisms.

Notably, continuous carbon fiber is an inorganic carbon material without polymer chains, and thus it has no glass transition temperature ( T g ),The polymer matrix of Markforged Onyx is nylon. According to the official material datasheet, its glass transition temperature ranges from 45 to 50°C, and the material can sustain a high temperature up to 145°C as mentioned above. Differential scanning calorimetry (DSC) tests were not carried out in the current work, and all thermal parameters of Onyx are quoted from the supplier’s material specifications.

Printing equipment and processes

A process study on the fabrication of CFRTC by FFF process revealed that the single nozzle fabrication method resulted in deterioration of the printing condition. To circumvent this limitation, a self-developed stand-alone dual-nozzle printing device was employed, whose structure and process principle are depicted in Figure 3. The equipment frame was processed by aluminum alloy CNC machining to guarantee structural rigidity during high-speed reciprocating motion. The three-axis motion platform adopted linear guide rails and closed-loop stepping motors to realize precise positioning of printheads. Two mutually independent extrusion printheads were mounted on the Z-axis sliding table: one printhead was configured for short-carbon-fiber Onyx matrix filament extrusion, while the other was equipped with a dedicated fiber traction cutter mechanism for continuous carbon fiber prepreg filaments. The feeding subsystem was assembled in the following sequence: feeding roller → tension sensor → traction gear → heating block → cutter mechanism → nozzle. Independent cooling fans were installed below each nozzle to control the cooling rate of extruded matrix and prepreg. The electrical control system integrated motion control unit, multi-channel temperature acquisition module, fan speed regulator and human-computer interaction unit; the slicing path data was transmitted to the controller via serial communication to realize automatic printing. All mechanical parts were deburred and assembled with pre-tightening torque standards to eliminate assembly clearance.

OPEN IN VIEWER

Figure 3.

Printing equipment and molding process. (a) Dual nozzle continuous fiber 3D printing equipment; (b) schematic diagram of molding process.

Prior to all specimen printing, comprehensive calibration was carried out on the self-developed dual-nozzle printer to eliminate assembly and motion errors. The printing platform was leveled with a 0.15 mm feeler gauge for uniform nozzle-substrate clearance, and X/Y/Z offset between the two nozzles was corrected via test lines with compensation parameters written into slicing software. Temperature sensors of heating blocks and hot bed were calibrated to keep temperature error within ±1°C. Fiber feeding tension and 1 mm retraction length listed in Table 1 were tuned to avoid fiber deviation and breakage during turning. Extrusion flow rates were matched to the layer thickness, pitch and printing speeds in Table 1 to guarantee stable fiber mass fraction, and motion acceleration compensation was applied for paths with radius below 1 mm to reduce fiber folding. After full calibration, standard test strips were printed for inspection, and only qualified equipment was used for subsequent sample and grille component manufacturing.

OPEN IN VIEWER

Table 1.

Printing parameter Configuration Tab1.

Items	Technical specifications
Matrix pitch	0.4 mm
Fiber pitch	0.7 mm
Matrix layer thickness	0.15 mm
Fiber layer thickness	0.12 mm
Substrate print speed	1800 mm/min
Fiber print speed	600 mm/min
Minimum fiber length	50 mm
Inclined exit height	8 mm
Fiber retraction length	1 mm
Nozzle diameter	0.4 mm
Printing accuracy	0.2 mm
Minimum turning arc radius for fiber path	0.1 mm
Printhead temperature	240°C
Carbon fiber prepreg nozzle temperature	250°C
Cooling fan power	60% output
Hot bed temperature	60°C

Onyx staple fiber blends and carbon fiber pre-preg filaments were fed into the two printheads and extruded and deposited from the nozzle outlets, respectively, on the printing platform. Onyx is employed to print the exterior shell of the component, thereby enhancing the precision of the printing process. Carbon fiber prepreg filament is utilized to print the internal reinforcement path. During the printing of carbon fiber prepreg filament, the heated prepreg filament adheres to the previous layer in the air after heating, bonds with the previous layer under air cooling, and exerts tensile force on the fibers in the subsequently extruded fiber filament, thereby enabling the continuous pull-out of fibers. During the manufacturing process, the intensity of the fan at the fiber exit can be adjusted to achieve the desired printing effect. The three-dimensional print head, under the control of the operating program, extrudes the composite material from line to surface and from surface to body in a gradual process, thus forming three-dimensional parts.

All process parameters summarized in Table 1 were optimized through single-factor exploratory printing trials. The optimization criteria included continuous fiber extrusion without breakage, low internal void content and reliable interlayer interfacial bonding. Multiple groups of trial prints were conducted to adjust nozzle temperature, printing speed, layer thickness and fiber spacing, and the optimal parameter combination was confirmed and listed in Table 1.

Slice path

In continuous fiber additive manufacturing, matrix material path planning has been relatively mature, but the fiber laying process faces two core challenges:

(1) Vector characteristics: The resistance generated by the high-speed movement of the nozzle can easily cause the fibers to deviate from the preset trajectory, especially in small-size structure printing, the slight deviation will significantly affect the continuity of the internal fiber distribution.

To alleviate fiber fracture caused by large-curvature corners, arc paths are used instead of sharp turns to distribute bending stress. Based on vector characteristics, directional path planning is adopted to customize fiber orientation, interlayer angle variation, and filling modes (transverse and diagonal). As shown in Figure 4, transverse filling and 60° incremental diagonal filling are applied. Customized paths optimize fiber placement, reinforce weak areas, align fibers with principal stress in thin-walled grille models, and enhance load-bearing efficiency. When the nozzle turns with a curvature radius below 1 mm, fibers are prone to folding and breakage. Due to the large amount of curvature at the toughness of carbon fibers, folding at the edges of local stress concentrations often leads to fibrous filament breakage. ³⁰ Therefore, a slice transition algorithm was developed to facilitate transition transitions at sharp turns and small angles of large curvature, thereby reducing the incidence of internal fiber detects.

OPEN IN VIEWER

Figure 4.

Fiber oriented layup (a) transverse fiber layup; (b) incremental 60° diagonal fiber layup.

Furthermore, the fiber laying path can be customized. The filling mode is illustrated in Figure 5, which depicts three filling modes: parallel lines, contours, and grids.

OPEN IN VIEWER

Figure 5.

Three path filling methods (a) parallel lines; (b) contours; (c) grids.

Mechanical test characterization

The tensile samples are tested in accordance with the standard ASTM D638 (Standard Test Method for Tensile Properties of Plastics). The dimensions of the rectangular samples are shown in Figure 6(a), with a marking length of 50 mm. Nevertheless, certain dimensions (e.g., thickness and length) have been slightly modified from those recommended by the standard in order to accommodate the dimensions of the print platform. The fibers were printed in a unidirectional orientation at an angle of 0°.

OPEN IN VIEWER

Figure 6.

(a) Tensile specimen; (b) bending specimen.

The three-point bending test is conducted in accordance with the ASTM D790 standard (Standard Test Method for Flexural Properties of Reinforced Plastics). The dimensions of the ASTM D790 specimens are shown in Figure 6(b), with thicknesses ranging from 3.2 to 1.6 mm, and with a span defined as 16 times the thickness of the specimen. The fibers were printed in a unidirectional manner at an angle of 0°.

The fiber content of the 3D-printed filaments was calculated. Mechanical characterization was conducted using an electromechanical universal testing machine (MTS Systems Ltd, Shenzhen, China) to obtain the mechanical properties of the standardized samples. The fiber-matrix interface and section morphology in the preformed and printed samples were observed using a scanning electron microscope (SEM, S-0N, Hitachi, Tokyo, Japan). The overall experimental flow is depicted in Figure 7.

OPEN IN VIEWER

Figure 7.

Printing and testing process.

Experimental design

Class I samples use a staggered lamination printing of matrix/fiber layers, employing a “inner-outer layer + alternating lamination” composite structure. The samples are divided into outer and inner layers: the outer layer is an Onyx matrix, which serves as encapsulation and protection; the inner layer is the core functional layer, designed with an alternating lamination of “Onyx matrix layer + CCF prepreg fiber layer.” Class II samples, on the other hand, use lamination printing of all fiber layers, with no clear distinction between inner and outer layers, and no matrix layer spacing between fiber layers. Four parallel specimens were tested for each group, and mechanical data are averaged. The sample number are shown in Figures 8 and 9. The CCF prepreg fibers are deposited layer by layer through the fiber nozzle throughout the process, with Onyx only used as an initial printing base support (less than 1%, negligible). Each printed layer consists of CCF prepreg material, and adhesion between layers is achieved through melted resin. The tensile and three-point bending test samples were printed using the printing device, and the printing experiment and test flow are shown in Figure 7. Parallel lines were employed to delineate the paths. The fiber volume fractions were calculated based on the prepreg SEM images and software slicing layers. As mentioned above, by observing scanning electron microscope (SEM) images and performing mathematical calculations, the fiber volume fraction can be obtained. For the prepreg’s inherent fiber volume fraction: The cross-sectional area of single carbon fiber was calculated using the measured monofilament diameter of 8 μm. The total cross-sectional area of all monofilaments within one fiber bundle was divided by the cross-sectional area of the 0.35 mm-diameter prepreg bundle, yielding an intrinsic fiber volume fraction of approximately 52%. For Class I staggered samples: The slicing thickness ratio of Onyx matrix to prepreg fiber layer was fixed at 1.25:1. Taking a unit printing area, the total mass of one matrix-fiber layer cycle was calculated using the densities of Onyx (1.2 g/cm³) and prepreg (1.4 g/cm³). The total carbon fiber mass in the prepreg layer was determined by multiplying the prepreg layer mass by its internal fiber volume fraction of 52%. The overall fiber mass fraction of staggered specimens was calculated as the ratio of carbon fiber mass to the total mass of the layer cycle, which was approximately 26%. All dimensional and density parameters required for calculation are provided in Section 4.1 for reproducibility. It’s worth noting that during early test prints, slight edge warping appeared on the surface of large thin parts. To suppress thermal deformation, the hot bed was kept at 60°C to lower thermal residual stress, and all specimens were wrapped with Onyx outer frames to constrain shrinkage. No obvious warpage occurred on all formal test samples and grille models under the optimized printing parameters. The specific parameters of the software slices and hardware configuration are presented in Table 1 below.

OPEN IN VIEWER

Figure 8.

Fracture mode and strength profile of CCF/Onyx tensile samples. (a) CCF/Onyx tensile sample fracture pattern (b) strength curve.

OPEN IN VIEWER

Figure 9.

Fracture pattern of the bending sample and its strength profile. (a) Fracture pattern of CF/Onyx bent sample (b) strength curve.

Sample mechanical test results

The total fiber content of the printed material was calculated from prepreg filament fiber content and fiber layup thickness (Figure 8(a)). The two additional curves in Figure 8(b) are parallel tensile test replicates for 26 wt% cross-stacked and 52 wt% full-fiber laminates, eliminating single-test randomness to guarantee reliable mechanical data. Fracture occurred at specimen midpoints; the 52 wt% CCF/Onyx composite achieved an ultimate tensile strength of 581.1 MPa—312% higher than the 26 wt% counterpart (185.8 MPa) and over tenfold that of pure Onyx (37 MPa). Even so, its tensile strength remains inferior to reported thermoset printed composites (1476 MPa), ³¹ a gap arising from the near-full crosslinking (99%) of thermoset matrices under high-temperature curing that yields stiffer composite parts. This process involves the penetration and filling of the gaps between the carbon fiber tows by a mixture of low-viscosity epoxy resin and dicyandiamide.

Figure 8 illustrates the fracture pattern observed in the tensile test for the CCF/Onyx samples. First, the thermoplastic matrix experiences tensile failure, with the fibers breaking in alignment with the matrix at the point of specimen failure. Subsequently, while the fibers are detached from the surrounding thermoplastic matrix, the fibers begin to break almost entirely within the matrix. Finally, the broken fibers detach from the CCF/Onyx sample.

The bending fracture pattern is illustrated in Figure 9(a), with the bending fracture occurring at the center. Similarly, the curves in Figure 9(b) show the repeated bending test data of the two above-mentioned layup types, verifying the stability of bending strength under different layup structures through multiple experiments. The strength profile is depicted in Figure 9(b), where the bending strength of the 52 Wt.% CCF/Onyx sample reaches 351.1 MPa, which is 228% higher than that of the 26 Wt.% CCF/Onyx sample (153.9 MPa). The strength is approximately four times that of the latter and nearly 5 times that of the pure Onyx sample (71 MPa).

In terms of the modulus, the tensile elastic modulus of 26 wt.% CCF/Onyx samples is 5.3 GPa, while that of 52 wt.% CCF/Onyx samples is 11.8 GPa. It is noteworthy that the flexural elastic modulus of the 26Wt.% CCF/Onyx samples is 13.206 GPa, which is 4.1 times that of the flexural modulus of pure Onyx samples (3.0 GPa). The flexural elastic modulus of the 52 Wt.% CCF/Onyx samples is 83.7 GPa, which is 28 times higher than the flexural modulus of the pure Onyx samples (3.0 GPa).

Microinterfacial structure

The results of the mechanical experiments indicate that the CCF/Onyx samples exhibit a notable enhancement in tensile and flexural strengths relative to the pure Onyx samples. This suggests that the samples fabricated by the FFF process have the potential to be utilized as load-bearing structures in mechanical applications. However, the FFF process is constrained by its own principles, and there are issues with weak bonding between wires and layers and poor fiber-substrate interface performance during the layered fabrication process. Therefore, it is essential to observe and analyze the fracture morphology based on the SEM images and investigate the damage failure mechanism at the micro-scale level in depth.

Fracture morphology analysis

Tensile fracture of FFF-printed CFRTC specimens is characterized by fiber breakage and fiber–matrix interfacial debonding, while flexural fracture shows fiber fracture, buckling, and interface failure. The microscale interfaces between fiber–matrix and between deposited layers are the dominant sites of damage initiation and propagation.

In the scanning electron microscopy (SEM) images of the stretch-damaged samples, for the full continuous carbon fiber (CCF) layer layup (52 Wt.% CCF/Onyx sample), there is a significant gap between the continuously deposited CCF layers and the upper and lower bottom surfaces, as seen in Figure 10(a1). This figure also shows a clear triple splitting of the stretched sample. Additionally, there are many tiny gaps between the continuously deposited CCF layers (see Figure 10(a2)). Figure 10(a3) illustrates the presence of a minor quantity of fiber extraction at the periphery of the gaps. In contrast, Figure 26 Wt.% CCF/Onyx sample depicts the staggered layup sample, which reveals that the CCF-substrate sample exhibits a greater degree of fiber extraction at the edge of the gaps. Figure 10(b1) illustrates that the interlayer voids of the stretched sample are increased and more uniformly distributed. They are primarily observed between the CCF layer and the substrate layer, indicating that the bonding between the substrate and the CCF layer may be less robust than that of the interlayer of the CCF layer. This is caused by the high-viscosity sizing agent in CCF filaments, which improves infiltration between CCF layers. Small voids in CCF layers result from insufficient infiltration of macromolecular chains during printing, which can be improved by extra heat and pressure. Moreover, the large difference in elongation at break (CCF: ∼1%, Onyx: 25%, composite: 5%) causes strong interfacial stress and induces delamination. As shown in Figure 10(b3), the staggered layup shows more obvious fiber pull-out due to the larger fiber–matrix contact area.

OPEN IN VIEWER

Figure 10.

SEM micrographs of tensile specimens with different layups; (a1, a2, a3) full CCF layer layup tensile specimens; (b1, b2, b3) CCF substrate staggered layup tensile specimens; (c1, c2, c3) full CCF layer layup flexural specimens; (d1, d2, d3) CCF substrate staggered layup flexural specimens.

Further qualitative comparison of voids and interlayer defects is supplemented based on Figure 10. Staggered specimens exhibit abundant interlayer gaps and scattered voids at matrix-fiber interfaces with severe fiber pull-out and matrix tearing. In contrast, full-fiber samples only contain sparse tiny internal voids and minimal interlayer separation. Such differences stem from poor compatibility between fiber sizing agent and Onyx matrix. Massive interfacial voids and gaps in staggered samples act as crack initiation sources under load, accounting for their much lower tensile and flexural strengths.

In the scanning electron microscopy (SEM) images of the bending-damaged samples, the staggered layup samples (26 wt.% CCF/Onyx samples) exhibited more pronounced gaps between the successively deposited matrix-CCF layers, as illustrated in Figure 10(d1). Additionally, the bending samples exhibited distinct multilayer splits, which were attributed to the shear force at the internal fiber-matrix interface when the samples were subjected to bending, resulting in shear slip. Furthermore, the presence of white lines, indicative of pulled-out matrix, was observed among the fiber layers. Observe the white line that has been pulled off the matrix, as seen in Figure 10(d2, d3). It can be seen that the CCF-matrix staggered layup has a significant number of matrix tears and cluster fiber extraction phenomena. In contrast, for the full CCF layer layup samples (52 wt.% CCF/Onyx samples), as seen in Figure 10(c1), the existence of such shear slippage is not obvious. This is due to the printing process of layup of fibers, which will be in the fiber surface. The fiber surface leaves a thicker sizing agent, which acts as a bond enhancer between the infiltration. This prevents delamination, although a small amount of CCF layer gap persists. This is due to the delamination of molding after the accumulation of sizing agent failed to fully infiltrate, resulting in some of the defective areas.

The interfacial adhesion between carbon fiber and Onyx matrix was evaluated indirectly via mechanical performance and SEM fractography. The sizing agent on carbon fiber enhances inter-fiber bonding, while the fiber-matrix interface exhibits weaker adhesion, accompanied by obvious fiber pull-out and delamination in staggered specimens. Tensile and flexural moduli serve as indirect quantitative indicators of interfacial load transfer capacity, and fracture morphology further distinguishes the difference in interfacial bonding quality between two layup structures. Direct quantitative interfacial tests like single fiber pull-out will be conducted in follow-up research.

Experimental results of the grid model

With the self-developed slicing algorithm enabling adjustable fiber layout, the dual-nozzle FFF process and CFRTC material possess great engineering application potential. Mold-free printed components can serve aerospace lightweight load-bearing parts including aircraft fairings, UAV ribs and satellite supports, whose mechanical performance matches 6061-T6 aluminum alloy. This composite also fits small-batch automotive lightweight parts like e-vehicle brackets and bicycle frames, with directional fiber paths reinforcing stress-concentrated zones for weight reduction. Its corrosion-resistant thermoplastic matrix is suitable for customized medical orthoses. Unlike costly mold-based techniques such as RTM and automated fiber placement, this additive manufacturing method excels at fast custom fabrication of complex thin-walled high-strength composite parts. As illustrated in Figure 11, the test model of the thin-walled grille part exhibited satisfactory printing results, indicating that the transverse/diagonal fiber layup path can be practically applied and enhance the comprehensive mechanical load-bearing performance of the thin-walled model.

OPEN IN VIEWER

Figure 11.

Transverse/diagonal laying paths and grill model printing (a and b) Transverse and diagonal fiber laying paths printing process; (c) finished printed product after rough surface sanding.

We used morphological comparisons to show the benefits of our directional fiber path planning method. Compared to simple transverse fiber filling, a 60°diagonal fiber layout places continuous carbon fibers along the main stress directions of the thin-walled grid, which helps reduce local stress concentrations and cut down on fiber bending defects and voids between layers. Smart fiber path design improves how loads are transferred internally, boosting the overall load capacity of complex thin-walled parts, and giving early proof that our custom fiber placement algorithm works. Next up, we’ll run a complete set of mechanical tests like tension and compression on the grid parts to measure how well this planning approach performs.

The stress-matching layout principle is verified indirectly via mechanical comparison of two layup structures; finite element stress simulation will be conducted in future research for quantitative stress distribution analysis.