2.1. Biodegradable AM Using CNF Tape

This section outlines the fabrication method for composite filaments and printing method for biodegradable AM using CNF tape.

First, the fabrication method of the CNF filament, which is composed of continuous CNFs and biodegradable resin, is described. Fig. 1 shows the filament fabrication process. A composite nozzle is heated and a biodegradable resin material such as PLA is fed from the upper motor. The resin is heated inside the composite nozzle using a heater, and the melt resin is combined with solid-state CNF sheets. The extruded CNF composite is cooled to the solid state by the ambient air and shaped into a filament, which is then pulled by a lower motor. The resulting filament is referred to as “CNF tape.”

Next, the FFF method is described. The fabrication process follows the same principle as the conventional FFF, as shown in Fig. 2. However, because the CNF tape has a higher rigidity than the molten resin, it is difficult to change its orientation during deposition. Therefore, the object is fabricated by repeatedly forming straight deposition paths and cutting filament at the ends.

2.2. Adjustment of the Fiber Content via Process Parameters

The process parameters for controlling the fiber content in the composite filament are illustrated in Fig. 3.

Let \(S_{\mathit{fiber}}\) be the cross-sectional area of the continuous CNFs, \(S_{\mathit{pla}}\) be the cross-sectional area of PLA, and \(S_{\mathit{com}}\) be the cross-sectional area of the composite filament. Let \(v_{\mathit{pla}}\) denote the feeding speed of the upper motor supplying the PLA and \(v_{\mathit{com}}\) denote the take-up speed of the lower motor.

If neither the PLA filament nor the CNF sheet contains air bubbles or cavities before mixing, it can be assumed that there is no change in mass or volume before and after mixing. Assuming incompressible materials and the conservation of mass, the following equation holds true:

The speed ratio \(\mathit{SR}\) is defined as

Because \(S_{\mathit{fiber}}\) remains constant before and after the composite process, the ratio of PLA to CNF in the filament can be controlled by adjusting the speed ratio \(\mathit{SR}\). It is important to note that the cross-sectional area of the composite does not depend on the material speed. After passing through the nozzle, the cross-sectional area becomes constant and is determined by the relationship between the input and output amounts of the nozzle. The dependence of the cross-sectional area on the speed ratio is the same for both mass and volume ratios.

If the PLA filament has a circular cross section with diameter \(d\), and the CNF has a rectangular cross section with width \(a\) and thickness \(b\), then

Substituting these into the previous equation yields

In this study, the fiber content is controlled by adjusting the speed ratio \(\mathit{SR}\), thereby increasing the fiber content by reducing the resin proportion.

2.3. Cross-Sectional Shape of the Composite Filament

Filament diameters of 1.75 mm are commonly used in many studies on composite materials. When continuous fibers with nearly circular cross sections, such as hemp fibers, are used as they can maintain their shape within the filament and can be extruded without problems. However, in the case of sheet-like continuous fibers such as CNF sheets, the fibers may twist inside the filament due to internal movement. This can cause nozzle clogging, fiber deformation, or damage. To address this issue, we proposed a filament with a rectangular cross section. This design suppressed fiber twisting and increased the fiber content.

4.1. CNF Mass Content

The fiber mass content was calculated using

To determine this value, both the mass of the continuous CNFs and the mass of the composite filament are required.

First, we employed a method to directly measure the weight . Five CNF samples, each 200 mm in length, were prepared. Their total mass was measured three times using an electronic balance, and the average value was divided by the number of samples to obtain the mass per fiber. Similarly, composite filaments of 200 mm in length were prepared (five samples per condition), and the average mass per filament was calculated. These values were substituted into the above equation to determine the fiber mass content.

The relationship between the speed ratio and fiber mass content is shown in Fig. 11. The results showed that the fiber mass content increased monotonically as the speed ratio \(\mathit{SR}\) increased, demonstrating the theoretical formula given by Eq. (3). As the \(\mathit{SR}\) increased, the take-up speed of the lower motor increased, reducing the amount of resin per unit length. Since the fiber length remained unchanged, the relative fiber content increased.

This confirmed that adjusting the speed ratio was effective for controlling the fiber content. Furthermore, the maximum fiber content achieved was \(\sim\)35%, which was a significant improvement compared with the typical fiber content of 20%.

4.2. CNF Cross-Sectional Area of the Composite Filament

During filament fabrication, defects such as bubbles may form inside the PLA, causing expansion of the filament volume.

To evaluate this, the theoretical cross-sectional area was calculated using Eq. (6) and compared with the experimentally measured values. The width \(a\) and thickness \(b\) of the CNFs were taken as the average values measured, as described in Section 2.2. Here, \(a=1.36\) and \(b=0.20\).

The fabricated composite filaments were used to measure the cross-sectional area. A 10-mm segment was cut from each filament, and its cross section was measured using image analysis, as shown in Fig. 12. Five samples were measured for each condition and the average values were used for comparison.

The measured and theoretical cross-sectional areas are plotted against the speed ratio in Fig. 13. At a speed ratio of 1, the measured cross-sectional area was significantly smaller than the theoretical value. This is attributed to the high resistance when pushing the resin, resulting in slippage between the filament and the drive gear or leakage of the molten resin. At a higher overall speed (\(\pi/9\)), voids were observed in the cross section, as shown in Fig. 14. Because the measured cross-sectional area included these voids, the values appeared to be larger than the actual material cross section.

Bubbles might form due to the impregnation of resin into the CNFs, caused by the excessive take-up speed of the lower motor. Consequently, the air in the CNFs was transferred into the resin and formed voids upon cooling. Because the resulting cross-sectional area was sometimes larger than that calculated using the formula (as shown in Fig. 13), we inferred that the voids expanded during mixing. Furthermore, because the void expanded in the radial direction, we perceived that it affected the cross-sectional area, but it did not affect the mass ratio.

These results indicated that reducing the overall process speed was necessary to suppress void formation and improve the adhesion between the CNFs and resin.

4.3. Surface Properties

Surface roughness measurements were conducted to evaluate the influence of the process parameters on the surface properties of the composite filaments. A complementary metal oxide semiconductor laser sensor (Keyence, ILS100) and positioning stage (COMS, PM80B-200XY) were used to acquire the 3D surface data. The arithmetic mean roughness \(R_a\) was calculated using 3D analysis software. The cut-off value was set to 2.5 mm.

A 50-mm segment was cut from each filament and the central 30-mm region was measured along the longitudinal direction. The measurement position is illustrated in Fig. 15. Five samples were analyzed for each condition.

The relationship between the speed ratio and surface roughness \(R_a\) is shown in Fig. 16.

It can be seen that the surface roughness increased as the speed ratio increased except for the lowest upper motor speed condition (\(\pi/36\)). This is attributed to bubble formation. At a high speed ratio (\(\mathit{SR}=4\)) and upper motor speed of \(\pi/9\), traces of burst voids were observed on the surface (Fig. 17). These voids are caused by insufficient resin impregnation, which leads to surface irregularities. In addition, the contact between the drive gear of the lower motor and the filament can deteriorate the surface condition. At higher take-up speeds, the cooling time is shorter, and the filament may still be soft when in contact with the gear, resulting in surface deformation.

By contrast, when the upper motor speed was low (\(\pi/36\)), we did not observe any significant changes in surface roughness, even at higher speed ratios. This is because lower speeds promote better impregnation and allow sufficient cooling before gear contact.

These results suggest that lower processing speeds are preferable for improving impregnation and surface quality.

4.4. Optimal Filament Fabrication Conditions

The optimal fabrication conditions must be selected based on fiber content, cross-sectional quality, surface conditions, and production speed. Each factor has a trade-off, and no single condition is optimal for all criteria. Therefore, a balanced condition (upper motor speed: \(\pi/18\), speed ratio: 4) was selected for subsequent fabrication experiments.

5.1. Effect of Tool Path on the Fabrication Accuracy

The FFF-AM system using a CNF tape is shown in Fig. 18. The system consisted of a desktop three-axis robot equipped with an extruder and a heated bed.

In this study, a drive gear modified for rectangular filaments was used, as shown in Fig. 19. A bearing (NTN, 694ZZ/5K) with an outer diameter of 11 mm was combined with a grooved guide ring and fabricated using a conventional FFF machine. A filament guide was also installed above the extruder to ensure stable feeding.

The fabrication procedure followed the procedure described in Section 2.1. The deposition paths are shown in Fig. 20, where extrusion was performed sequentially along paths \(\textrm{A} \to \textrm{B}\), \(\textrm{C} \to \textrm{D}\), and \(\textrm{E} \to \textrm{F}\). Straight-line tool paths were used. At the end of each path, extrusion was temporarily stopped and the filament was cut. The starting position was then shifted and the process was repeated. This procedure built an object consisting of three parallel paths in a single layer.

The tool path width \(W\) was varied from 1.0 mm to 2.0 mm in 0.2 mm increments. In addition, because gaps began to appear visually at \({W=2.0}\), two additional conditions (\({W= 2.1}\) and \({W=2.2}\)) were tested, resulting in eight types of specimens. After fabrication, both ends of each specimen were cut to a length of 50 mm. The dimensions of the specimens are shown in Fig. 21 and the process parameters used in the fabrication experiment are shown in Fig. 22. The fabrication conditions are summarized in Table 2.

In continuous-fiber AM, it is necessary to match filament feeding and tool speeds to prevent fiber bending or breakage. The tool speed \(v_{\mathit{tool}}\) is calculated as

In FFF, an “ironing effect” is often used to smooth surfaces by pressing the nozzle against the extruded material ¹⁷. This can reduce voids ¹⁸ and internal defects ¹⁹, which degrade the filament strength. However, when continuous fibers are used, it is difficult to select appropriate tool paths to avoid fiber damage and internal gaps ²⁰. In this study, the fabricated specimens were evaluated based on the relationship between the ironing effect and tool path width.

Three-dimensional surface data were obtained using the same method described in Section 4.3. The surface roughness \(R_a\) was measured along the \(x\)-axis direction at three positions in the central 30 mm region of the specimen, as shown in Fig. 23, and the geometric profiles were evaluated along the \(y\)-axis direction at five equally spaced positions by measuring the difference between the maximum and minimum heights, as shown in Fig. 24.

Photographs of the fabricated specimens at \({W=1.4}\) and \({W=2.2}\) are shown in Fig. 25. An example of the \(y\)-axis direction profile measurement is shown in Fig. 26. The relationship between the tool path width \(W\) and surface roughness \(R_a\) is shown in Fig. 27. The minimum roughness was observed at \({W=1.6}\), and the surface roughness increased as \(W\) decreased or increased. However, the variations were \(\sim\)1.5–4.0 μm, which were relatively small. Here, the surface roughness of the fabricated object was compared with that of the filament. The surface roughness \(R_a\) of the composite filament under an upper motor speed of \(\pi/18\) and speed ratio of 4 was \(\sim\)9.0 μm, as shown in Fig. 16. By contrast, the \(R_a\) of the fabricated object was evidently lower, indicating a smoother surface. This is likely due to the ironing effect.

The relationship between the tool path width and height difference is shown in Fig. 28. As \(W\) increased, the height difference increased, indicating the formation of grooves and gaps. At \({W=2.2}\), the height difference exceeded 500 μm, which was similar to the layer height, indicating complete gaps. By contrast, when the width of the deposition path was small, the height difference was small. However, when \({W=1.0}\) and \({W=1.2}\), the spacing was smaller than the width of the continuous CNF (1.36 mm), resulting in overlapping fibers. Although a flat surface was obtained under these conditions, the reason for this phenomenon remains unclear and further investigation is required in future work.

Based on these results, \({W=1.4}\) was selected as the most appropriate value to achieve a stable geometry and was used as the design parameter for the specimens in subsequent tensile tests.

5.2. Effect of Fiber Content on the Mechanical Strength

The specimens were fabricated using the tool path width \({W=1.4}\) selected in the previous section and a CNF tape having a higher fiber content. Two types of specimens with different fiber contents were fabricated (four specimens for each type), and tensile tests were conducted to evaluate the effect of the fiber content. The fabrication conditions are tabulated in Table 3. The specimens fabricated using filaments with speed ratios of 3 and 4 were designated as SR3 and SR4, respectively. The dimensions of the specimens are shown in Fig. 29.

The cross-sectional area used to calculate the nominal stress was obtained from the image measurements, as shown in Fig. 30. To prevent damage during the tensile tests, acrylic ultraviolet resin was applied to the gripping sections, as shown in Fig. 31. The distance between the grips was used as the gauge length.

The stress–strain curves for the SR3 and SR4 specimens are shown in Fig. 32. The average maximum tensile strength was \(\sim\)75 and 86 MPa for the SR3 and SR4 specimens, respectively. For comparison, the PLA-only specimens exhibited a tensile strength of \(\sim\)49 MPa ¹⁶. These results confirmed that CNF reinforcement improved the strength of PLA. An increase of \(\sim\)5% in fiber content increased the tensile strength by \(\sim\)11 MPa. This indicated that increasing the fiber content by reducing the resin content was effective in enhancing the mechanical properties. When compared with the experimental results for the CNF strips, the difference between the maximum and minimum tensile strengths of the CNF strips in Fig. 7 was \(\sim\)60 MPa, whereas the corresponding differences for the composites in Fig. 32 were limited to \(\sim\)10 MPa for both the SR3 and SR4 specimens.

Figure 33 shows the relationship between the fiber mass content and tensile strength and Young’s modulus. The data for the PLA-only specimen in the figure were obtained from the literature ¹⁶ and did not have error bars. The other data represent the average and standard deviations obtained from the experiments in this study. To compare the results with the rule of mixtures, the experimental results for PLA-only and CNF sheet-only specimens were connected by a straight line. The tensile strength of the printed specimens approximately followed this trend, whereas the Young’s modulus was lower than expected. This can be attributed to the delamination of the continuous CNFs. As shown in Fig. 34, delamination occurred in the continuous CNFs inside the specimen. The resulting voids caused the fibers to solidify in a bent state, which led to a reduction in the Young’s modulus.