**3. Results**

#### *3.1. Sliver Characterization*

Throughout the textile process, the fiber length and mixing ratio were measured to continuously assess the individual process steps and changes in the material. During the production of the sliver, mechanical work was applied to open and align the fibers, resulting in a shortening of the carbon fibers, as shown in Figure 7. Despite ongoing optimization of all textile process steps, the carbon fibers were shortened by 20% from the raw material to the sliver regarding the median fiber length.

**Figure 7.** Change of the average and maximum carbon fiber length through the sliver production process.

To measure the single-fiber length, the examined carbon fibers were laid out on black velvet, covered by a glass plate, and placed under a magnifying glass with illumination. Using tweezers and fingers, the carbon fibers were carefully removed from the fiber agglomeration. The sliver in particular, had to be delicately dissolved by hand to remove the carbon fibers without damage. For this reason, the resulting single-fiber length is questionable. The tendency of a fiber to shorten after each processing step, however, was assumed to be correct.

Even slight mechanical processing of the fibers during pre-opening resulted in shortening, especially for the longest carbon fibers. A further reduction of the carbon fiber length was due to the mechanical processing of the fibers in the carding machine. Overall, the processing of carbon fibers results in a more uniform fiber length distribution, which can be beneficial for further processing and lead to stable process conditions.

Figure 8 shows the change in the mixing ratio over the sliver production process. The development of fiber dust in the carding process can be detected by measuring a change in the mixing ratio. While the pre-opening, which included measurement inaccuracies due to the sample size, did not lead to a change in the mixing ratio, the mechanical processing of the carbon fibers in the carding machine caused fiber breakage, and thus, fiber dust, decreasing the carbon fiber fraction. The further reduction of the carbon content was due to the removal of fiber dust from the sliver. Overall, a reduction of the carbon content by 7% resulted during manufacturing.

**Figure 8.** Change of the mixing ratio over the sliver production process.

#### *3.2. Preliminary Investigations of Fiber Content in Relation to the Process Settings*

In a previous study, comparatively strong results were achieved when processing glass fibers in a single-stage process via the direct fiber feed module [10], and the fiber content was predicted successfully. Hitherto, carbon fibers had never been processed in a single-stage process using a direct fiber feed module, so this study investigated the transferability of the results to the processing of recycled carbon fibers. Therefore, the settings of speed and mass throughput at the IMC for a direct fiber feed of carbon fibers had to be determined in preliminary tests. To ensure comparability with previous tests of glass-fiber-reinforced recycled polyamide [10], carbon fiber contents of 20 wt-% and 30 wt-% were targeted.

Contrary to the use of virgin carbon fiber rovings, it was not ye<sup>t</sup> known whether the processing settings of the recycled carbon fibers via the staple-fiber sliver could reliably be determined with the calculation model for the direct fiber feed module. Due to its hybrid structure, the carbon fiber content of the sliver is only roughly 60 wt-%. In addition, the low consolidated structure (low tex value) of the sliver presented di fficulty. Hence, a high sliver throughput was needed, requiring high screw speeds, which could increase fiber slippage and thus reduce the fiber content. Therefore, three setting approaches for the sliver were attempted in the preliminary tests. The selected test settings are shown in Table 3.


**Table 3.** Settings of the preliminary tests.

The results of the fiber content measurements of the preliminary tests in Figure 9 show that the calculation model is also applicable to the processing of (virgin) carbon fiber rovings, in general. For the sliver, in contrast, the mass throughput of . *m* = 3 kg/h and a screw speed of *n* = 300 min−<sup>1</sup> yielded a fiber content of approximately 20 wt-%, which is lower than predicted. Since the screw speed of *n* = 400 min−<sup>1</sup> induced a fiber content that was too high, the speed for the further investigation was set to *n* = 350 min−<sup>1</sup> for a mass throughput of . *m* = 3 kg/h.

**Figure 9.** Fiber content in preliminary tests with carbon fiber roving and recycled staple fiber sliver.

In general, the deviations of the measured values were quite large. This could have been due to the unsteady sliver formation process. High fluctuations were observed during the single-stage process with the glass fibers as well, which could be attributed to the IMC [10].

#### *3.3. Fiber Content*

Figure 10 shows the averaged fiber contents of the different mixtures from the single-stage process. When processing the sliver, the calculated settings did not lead to the desired results. Nearly all mixtures exhibited higher fiber content values than the calculated values according to Equation (1). In particular, for the calculated fiber content of 20 wt-%, all measured values were above 30 wt-%. In some cases, this discrepancy was even larger at higher speeds. Overall, the fiber contents showed high deviations and inconsistent behavior.

**Figure 10.** Averaged fiber content of the recycled staple fiber sliver with varying recycling content of the matrix.

The deviation of fiber content was due to the sliver formation process. Further, the feeding of the sliver into the IMC led to difficulties during processing. The sliver did not have a homogeneous width (see Figure 6) and was stretched irregularly when entering the extruder because of its unconsolidated texture. As a result, the sliver was not drawn in continuously and sometimes even tore during processing, resulting in an uneven fiber distribution in the melt. Further, varying process parameters can additionally cause small fluctuations in the fiber content. For example, friction occurs during fiber draw-in at the extruder, which counteracts the fiber draw-in force and can thus have effects on fiber retention and length. The friction is influenced by the roving and roving guide type, bobbin winding, screw geometry, and physical properties of the melt.

An analysis of the recycling content in the matrix material shows no significant influence on the fiber content. The scatter of the measured fiber content values was large. Therefore, the influence of process fluctuations on fiber content is assumed to be strong. In previous work with glass fibers [10], it was also shown that there is no clear correlation between fiber content and recycling rate. Only in the case of twice-recycled material could a dependency be determined. The values obtained here, however, show only small deviations with no statistical significance, even for twice-recycled material.

In summary, the settings calculated according to Equation (1) in the single-stage process with the sliver did not lead to the desired results. The calculation model for the direct fiber feed in a single-stage process cannot, therefore, be applied to the staple fiber sliver made of recycled carbon fibers.

#### *3.4. Fiber Length*

The fiber lengths of two test specimens were measured: specimen one with a fiber content of 27 wt-% ( .*m* = 3 kg/h, *n* = 300 min−1) and specimen two with a fiber content of 31 wt-% ( .*m* = 3 kg/h, *n* = 350 min−1). The results were compared to a specimen made of virgin fiber roving with a fiber content of 26 wt-%. Analyses of the weight-averaged fiber lengths their distributions were based on the ISO 22,314 standard.

Table 4 shows the average fiber lengths of the different mixtures of the single-stage process. Overall, fiber lengths between 200 and 330 μm were obtained. When comparing the different types of carbon fibers used with an analogous adjustment of the fiber content, longer fibers were observed in the test specimens from rovings. The average fiber length values were about 35% higher.


**Table 4.** Settings of the preliminary tests.

When comparing the fiber contents of the two test specimens made of sliver, the higher fiber content of 31 wt-% yielded significantly longer fibers compared to the lower fiber content of 27 wt-%. The comparison of the fiber length distribution also shows that a fiber content of 31 wt-% leads to significantly longer fibers of up to 5 mm. At a fiber content of 27 wt-%, however, a maximum fiber length of 0.8 mm was observed.

Investigations with glass fibers utilizing a direct fiber feed module in the single-stage process, on the other hand, have shown that a higher fiber content increases the fiber degradation [10]. This effect is explained by the associated higher screw speed and the resulting increase in shear forces in the melt. The reduction of the mass flow rate led to an increased shear energy input, and thus, additional stress to the fibers. Additionally, an increase in the material residence time in the mixing zone could lead to a further reduction in fiber length. Likewise, as the fiber content increases, the fiber length decreases owing to the intensive fiber–fiber interaction [10]. Nonetheless, findings in the literature are contradictory in terms of the influence of the screw speed on fiber length. Higher screw speeds can also result in longer fibers because of the reduced melt viscosity and dwell time, as described in [37,38].

The fibers of the (virgin) glass-fiber-reinforced test specimens produced in the single-stage process were significantly longer (an average of 0.74 mm) than the fibers of the carbon-fiber-reinforced test specimens made of staple fiber sliver (Table 4). The shorter fiber lengths of the recycled carbon fiber test specimens can be attributed to the carbon fibers being thinner and more brittle than glass fibers. It has also been observed in the literature that when comparing short-fiber-reinforced composites with a fiber content of 30 wt-%, the glass-fiber-reinforced test specimens have, on average, significantly longer fibers [39]. Dependence of the recycling fraction and fiber length cannot be determined owing to the small number of samples investigated.

In summary, the average fiber length of the carbon fibers was 0.3 mm. Shorter fibers were measured at a lower fiber content of 27 wt-%, whereas a higher fiber content of 31 wt-% resulted in longer fibers. The influence of the recycled content of the matrix on carbon fiber length could not be determined due to the small number of samples measured.

Since only representative samples were evaluated, the actual fiber length of each test specimen is unknown. The results are intended to test the possibility of fiber length measurement to this extent and to show a trend in fiber length and fiber-length distribution.

## *3.5. Mechanical Properties*

Figure 11 shows the average Young's moduli of the test specimens from the single-stage process. The test specimens with a fiber content of 27 wt-% had similar values to those of the test specimens with a fiber content of 30 wt-%, roughly 25,000 MPa. This suggests an overall good performance of the material compared to the recycling routes summarized by Pimenta [2], especially considering the advantages of the injection molding process compared to bulk molding compound compression or compression molding of non-wovens, which are other solutions for discontinuous fiber processing described in the literature. Pimenta stated that Young's moduli of 20 GPa, 5–30 GPa, and 14–16 GPa were reached for bulk molding compounds, non-wovens, and injection molded parts, respectively [2].

**Figure 11.** Young's moduli of specimens produced in the single-stage process with varying recycling content of the matrix.

When the results of the mechanical test are compared with those of the fiber content measurements presented above, a certain correlation can be established. Figure 12 shows the average Young's moduli of the test specimens and their average fiber contents. Overall, parallels can be seen between Young's modulus and fiber content, as expected.

**Figure 12.** Young's moduli of specimens produced in the single-stage process with varying recycling content and corresponding fiber content of the matrix.

However, the dependence of Young's modulus on the recycled content of the matrix cannot be determined. The double reuse of PA6 does not lead to any significant effects. This means that the recycling-related impairments of the plastic have no apparent effect on the tensile modulus. Instead, the material parameter is primarily determined by the fiber content.

The standard deviations of Young's modulus are very high, up to 13%, due to processing instabilities caused by the IMC and the resulting scattering of fiber content within a batch. As previously explained, there were irregularities in the feeding of sliver, which led to fluctuations of the fiber content in the melt.

When Young's moduli of the carbon-fiber-reinforced recycled parts tested are compared to those of the glass fiber-reinforced plastic parts produced under almost identical conditions, an influence of the type of reinforcing fiber on the part can be clearly established. In Figure 13, glass-fiber-reinforced parts produced in the same process exhibit Young's moduli of approximately 10,000 MPa, nearly half of that of recycled carbon-fiber-reinforced parts. The use of carbon fibers, therefore, leads to significantly higher stiffness of the part, even though the length of the recycled carbon fibers is significantly shorter than the length of the glass fibers.

**Figure 13.** Young's moduli of specimens produced in the single-stage process with varying recycling content and corresponding fiber content of the matrix compared to parts produced with virgin glass fiber rovings and similar fiber content.

To further evaluate the mechanical properties, the tensile strength was determined. The results are shown in Figure 14. As expected, the tensile strength is also dependent on the fiber content. Similar to the modulus of elasticity, a correlation with the fiber content can be observed for the tensile strength as well.

**Figure 14.** Young's moduli of specimens produced in the single-stage process with varying recycling content and corresponding fiber content of the matrix.

For the comparison of the two different fiber types, the average tensile strengths of the glass fiber specimens and the sliver specimens are shown in Figure 15. As with Young's modulus, the test specimens from the recycled carbon fiber sliver showed significantly higher tensile strengths compared to the glass fiber specimens, confirming previous results.

**Figure 15.** Tensile strengths of specimens produced in the single-stage process with varying recycling content and corresponding fiber content of the matrix compared to parts produced with virgin glass fiber rovings and similar fiber content.

The analyses of the elongation at break, the impact strength, and the flexural modulus and strength showed similar results and will not be presented in this study to avoid redundancy.

#### **4. Summary and Outlook**

In this study, a novel method to recycle carbon fibers efficiently while preserving fiber length was investigated. A hybrid sliver with 60 vol-% carbon fiber content was produced from recycled carbon and PA6 fibers. A shortening of the carbon fiber of about 50% resulted after the entire process. At the same time, a narrowing of the fiber length distribution could be observed. The sliver was then used for the production of test specimens on an IMC utilizing a direct fiber feed module.

The feeding of the staple fiber sliver into the melt was challenging. The sliver stretched unevenly and sometimes tore. The predicted fiber content could not be achieved in the single-stage process. In fact, the fiber contents were higher than the calculated values and had high fluctuations. The fiber content measurements do not show a clear dependency on the recycled content of the matrix.

A fiber length measurement was performed on three random samples. The results show an increasing fiber length with increasing fiber content. Fibers of parts made of virgin carbon fiber rovings were longer than those of the recycled carbon fiber sliver.

Tensile tests were carried out to characterize the mechanical properties. Overall, very good mechanical characteristics could be achieved. Correlations between the actual fiber content and the mechanical properties were determined. The higher the fiber content, the better the mechanical properties. Large deviations of the measured values were observed. Furthermore, it can be stated that the recycling-related impairment of the matrix has no effect on the modulus of elasticity, tensile strength, flexural strength, or impact strength.

When comparing the recycled carbon-fiber-reinforced test specimens with the glass fiber test specimens, a clear dependence of the mechanical characteristics on the type of reinforcement fiber was determined. The sliver test specimens all achieved significantly higher results, although the recycled carbon fibers were significantly shorter than the glass fibers.

The results show that desirable mechanical properties can be achieved in principle when recycled carbon fibers are processed in the form of a staple fiber sliver. However, it is questionable whether the processing of staple fiber sliver allows a reproducible adjustment of the fiber content, and thus, predictability of the properties. The loose structure of the sliver causes difficulties with drawing the fibers in the process evenly. The use of the hybrid yarn could simplify the drawing of the recycled fibers, as the strength of the yarn is considerably higher. Thus, a yarn spinning process was developed to further process the sliver before injection molding. This further reduces the fiber content but improves the fiber distribution. Fewer process fluctuations might be reached this way.

**Author Contributions:** Conceptualization, J.W. and S.B.; methodology, J.W. and J.R.; validation, J.R.; investigation, J.W. and J.R.; resources, G.T.G. and C.B.; writing—original draft preparation, J.W. and J.R.; writing—review and editing, S.B., G.T.G. and C.B.; visualization, J.R. and S.B.; supervision, G.T.G. and C.B.; project administration, J.W. and S.B.; funding acquisition, G.T.G. and C.B.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Baden-Württemberg-Stiftung, gran<sup>t</sup> number MAT0005.

**Conflicts of Interest:** The authors declare no conflict of interest.
