Performance analysis of recycled carbon fiber under recycling process parameters optimized using response surface methodology

The recycling of high-performance carbon fiber from carbon fiber-reinforced polymer (CFRP) wastes have great economic value and environmental significance. Based on the principle of thermally activated oxide semiconductors, the resin matrix decomposition of process model was established by response surface methodology (RSM). The morphology, elements, functional groups and mechanical properties of recycled carbon fiber were investigated. The results indicated that the degradation was positively correlated with time and temperature, and the degree of influence was significant, the effect of O2 flow rate and concentration were not significant. The influence of process parameters on resin decomposition was as follows shown: temperature > time > O2 concentration > O2 flow rate. The actual degradation was 96.12 wt.% under temperature of 520°C, time of 23 min, O2 concentration of 80% and flow rate of 180 mL/min. The recycled carbon fiber (rCF) was compared with the original carbon fiber the surface roughness of rCF was increased, and without carbon deposition on the surface. The content of C element and C–C bond on the surface of rCF were significantly increased, and functional group of COOH was produced. The tensile strength of rCF was maintained above 99%, the Young’s modulus was maintained at 92%, and the interfacial shear strength was maintained at 85%.


Introduction
Carbon fiber reinforced polymer (CFRP) has been widely used in transportation and aviation for their excellent performance, such as light weight and high specific strength. As the demand for CFRP continues to grow, so does the waste of CFRP. It was reported that the global CFRP wastes will reach 170,000 tons by 2025, and the growth was expected to continue. 1,2 The rapid growth of CFRP wastes have led to the great waste of resources, and the potential of carbon fiber (CF) have not been fully utilized.
At present, CF is mainly recycled by mechanical, pyrolysis and chemical recycling. The interface between the fiber and the resin-matrix was broken in the mechanical recycling by grinding CFRP wastes. The fiber surface obtained by mechanical recovery has a high content of resin matrix, and mechanical properties of the fiber is significantly reduced. The recycled carbon fiber (rCF) could only be used as a filling material and have little utilization value. 3 Chemical recycling is the decomposition of the resin-matrix in CFRP by solvent. Xing et al., 4 recycled single-layer CFRP at 160-220°C, and the flexural strength of CFRP plates preparation by rCF were maintained at 47%-89%. Pei et al. recycled CFRP in ammonium acetate solution at concentrations of 1.04-2.08 mol/L, and the interfacial shear strength (IFSS) of recycled carbon fiber reinforced polymer (rCFRP) was maintained at 87.23%-122.41%. 5 The pyrolysis recycling is to decompose the resin-matrix of CFRP on high temperature and certain atmosphere conditions. Ren et al. 6 recycled CFRP by microwave pyrolysis (500-550°C), and the degradation-matrix was maintained at 96.4%-97.9%. Matsuda et al. recycled CFRP by the pyrolysis at 600°C for 90-180 min, and the average mechanical strength of rCF was maintained at 43.13%-80.50%. 7 In summary, the parameters of recycling process have a significant effect on the degradation and the performance of rCF. To achieve efficient recycling of CF, we innovatively proposed a method for recycling CFRP waste using thermally activated oxide semiconductors. 8 Sommer et al. investigated the influence of process parameters on the degradation and rCF performance, and high-quality rCF was efficiently recovered by optimizing process parameters. 9 In this paper, the central composite experimental design (CCD) method was used to established the quantization relationship model in this investigation. The influence of multiple factors and multi-factor interaction on the degradation were analyzed. Based on the quantitative relationship model, the optimal recycle process parameters were solved. The CFRP waste of different shapes was recycled, and the morphology, elements, functional groups and mechanical properties of rCF were evaluated by Raman spectroscopy, X-ray photoelectron spectroscopy (XPS), mechanical property testing and other methods.

Materials
The type of oCF (Guangwei Composite Material Co., Ltd, China) with sizing was T300 with a diameter of 7 μm. The CFRP plates (Guangwei Composite Material Co., Ltd, China) with bisphenol A epoxy content was 40 wt.% (GB-T 3855-2005). The resin content of block CFRP and tubular CFRP waste was 20 wt.% and 30 wt.% (Guangwei Composite Material Co., Ltd, China, GB-T 3855-2005). The purity of powdered Cr 2 O 3 (Wako Corporation, Japan) is 99%, and its specific surface area is 3 m 2 /g.

Recycle CFRP experiments
The hot sulfuric acid was used as the medium to measure the percentage of carbon fiber mass in the composite and the specific operations are as follows: (1) Firstly, the CFRP waste was cut into small pieces with a mass of m 1 which is equal to 0.5000 ± 0.0200 g. Secondly, put them into a 100 mL flask and inject 50 mL of concentrated sulfuric acid with a mass fraction of 95%-98% Then the solution was heated to 220°C. (2) When the reaction was over, a hydrogen peroxide solution with a concentration of 30% was used for titration until the solution was clear. Therefore, the resin content in the composite was shown in equation (1).
Where the m 1 is the mass of fiber, the m 2 is the mass of fiber after reaction. The W t is resin content in CFRP, wt.%. Repeat above steps three times and take the average of W t .
The tube furnace (BTF-1200°C) was supplied by Anhui Beiyike, China, as shown in Figure 1. Firstly, evenly spread a layer of Cr 2 O 3 powder on the crucible. Secondly, the CFRP waste was cut into 0.2000 ± 0.0010 g (M 1 ) cubes and placed in a crucible covered with Cr 2 O 3 powder. Then, Cr 2 O 3 powder is evenly coated on the CFRP waste, and gentle pressure is applied to allow sufficient contact between the Cr 2 O 3 powder and the CFRP waste. Finally, crucibles with CFRP waste and Cr 2 O 3 powder were placed in tubular furnace (already heated to the target temperature).
After the reaction, the Cr 2 O 3 powder on the fiber surface were removed. The fibers were gently rinsed with distilled water until free of obvious impurities, and finally once with acetone solution. The obtained carbon fiber was dried in a drying oven at 120°C for 2 h, and then cooled to normal temperature in a vacuum drying oven, and then taken out for weighing (M 2 ).
The resin decomposition is calculated as follows: In the formula, η -CFRP resin decomposition, wt.%. M 1 -mass of CFRP before reaction. M 2the mass of fiber after the reaction. athe content of resin in CFRP, wt.% As is shown in Table 1, the effects of time, temperature, O 2 concentration and flow rate on the degradation were explored. The results of the central experimental design were shown in Table 2.

Analysis and characterization
(1) The Supra55 scanning electron microscope produced by Zeiss company in Germany is used to observe the cross-sectional morphology of recycled carbon fiber and composite materials, respectively. The electron microscope has a magnification of 10-900000 x, an accelerating voltage of 0.1-30 kV, and a resolution of 1.0 nm at 15 kV. (2) The exhaust gas components of the recycle process were analyzed by FTIR, and the volume of the gas chamber of the instrument was 6 mL, the heating temperature range is 25-1200°C, and the infrared wavenumber range is 6000-500 cm À1 . (3) The graphite structure on the surface of the rCF was detected by Raman spectrometer, with a laser source of 514 nm and an excitation power of 10 mW. (4) XPS is used to characterize the relative atomic concentration and functional group content of the surface atoms of the oCF and the rCF. Between length and width 0.5-3 cm, thickness between 0.1-1 cm, surface roughness does not exceed 5 μm. (6) The fiber tensile performance of the carbon fiber is tested by the fiber strength extensibility meter (XQ-1 type), and the tensile strength of the monofilament of the carbon fiber is calculated according to formula (2). The instrument load measurement range is 0-100 cN, the clamping distance is 10-50 mm, the gripper stroke is 100 mm, and the test speed is 0-200 mm/min.
2In the formula, n-the amount of carbon fiber. m-the Wechsler modulus, which characterizes the uniformity and reliability of the material. σ i = 4F i /πd 2 : the tensile strength of each carbon fiber, F ithe maximum tensile force that each carbon fiber can withstand, dthe average diameter of the measured carbon fiber. σ 0the Weibull scale parameter, that is the monofilament tensile strength of carbon fiber. The average rate of chuck descent is 2.4 mm/ min, the specimen gauge length is 25 mm, and the number of tests was 40 (typically 20+), and the sampling length of rCF and oCF >4 cm. (7) Composite material interface evaluation device (Model HM410) provided by Toei Industrial Co., Ltd. Is used to test the interface shear strength of original carbon fiber reinforced polylactic acid (oCF/PLA), recycled carbon fiber reinforced polylactic acid (rCF/PLA), original carbon fiber reinforced epoxy resin (oCF/EP) and recycled carbon fiber reinforced epoxy resin (rCF/EP) at room temperature and air atmosphere. In the test, the stretching distance, the tensile rate, the wrap length and the number  of tests were 500 μm, 60 μm/min, 300 μm and 30 respectively.

ANOVA results analysis
Based on model reliability analysis, analysis of variance (ANOVA) and a lack of fit test (LOF) were performed. [10][11][12] Through the ANOVA results, the plots of predicted and actual were plotted to evaluate the actual value and the predicted response value (resin decomposition). As shown in Table 3, the model regression coefficient (R 2 ) of the decomposition was 0.9866, which is consistent with the experimental results, and the variability of 98.66% could be explained. The F value was not significant which meant that the reliability of ANOVA model exceeded 95%. The prediction results are based on formula (3) 13 : In the formula, β 0 -the fixed response value, and β i , β ij , and β ii represent the coefficients of linear, interaction, and quadratic parameter. As shown in Fig. 2(a), the actual values were evenly distributed on both sides of the fitted line and all the actual values were close to the predicted values. From Fig. 2(b), the normal probability was scattered on both sides of the fitted straight line without large error points. Through Fig. 2(c), the residuals were normally distributed, and the actual residuals were uniformly distributed on both sides of the predicted value as a whole without large error points. As shown in Fig. 2(d), the external residual points deviate from the predicted value in the lower range, the RSM model run efficiently without significant underfitting. In conclusion, the effectiveness of the model of Response Surface Methodology (RSM) was fully verified.

Analysis of the degree of factors disturbance
In order to explore the influence of different factors on the degradation, perturbation at the center point were plotted (Figure 3). The degree of influence of each factor: temperature > time > O 2 concentration > O 2 flow rate. Perturbation plots visualize the influence of individual factors, but it was only sensitive to independent variables and lacks analysis of the interactions. Therefore, Pareto chart was used to assess the impact of multiple factors on the final result. 14 The chart fully displayed the proportion of disturbances of each factor, excluded the influence of the least correlation factor, and identify the importance of the main factors to the results. As shown in Figure 4, the impact of time (A) was highest, the influence of temperature (B) and square of time (A 2 ) were secondary, and O 2 flow and concentration had a small percentage.

Establish surfaces plots of 3D models response
As shown in Figure 5, the 3D model was established to investigate the impact of the interaction between various factors on the response surface regression results. The results indicated that the reaction time and temperature had a significant effect on the degradation, and positively correlated with the decomposition (Fig. 5(a)).

Optimize and solve the optimal recycle parameters
The effects of process parameters on the degradation were investigated, such as time, temperature, O 2 flow rate and concentration. Firstly, the CCD was used to establish the model of RSM, and the response values obtained by the experiment (Table 4). Secondly, response functions of Second-order polynomial, independent variables and corresponding output datasets were used to evaluate the model coefficients. Then, the effect of the interaction of independent variables on the decomposition were combined. Finally, the resin decomposition of rate was predicted by optimizing processes and suitable models. The model had four independent variables and corresponding response, and the final quadratic equation was Equation in Terms of Coded Factors, combined with the Pareto chart (Figure 4) of important factors, minimal error terms were eliminated, and the resin decomposition under the optimal recycle process parameters is calculated in the form of independent variable code. Y 1 in code was the code resolution rate.
Considering the influence of the p value (less than 0.05), 12 the argument code form was expressed as follows: The response surface model shown that optimal recycling process parameters were 518.32°C, 22.59 min, 78.19% O 2 , 182.41 mL/min O 2 , and the degradation was 97.84 wt.%. Combined with the actual situation, the optimal recycle process parameters were corrected to 520°C, 23 min, 80% O 2 , 180 mL/min O 2 , the results of multiple recycles show that the error between the actual value and the predicted value was between ±1% Table 5.

Establish linear models
In order to further evaluate the influence of process parameters on the degradation, a linear model of the equilibrium relationship between time, temperature, O 2 concentration and flow rate was established. Based on different linear models, the reliability of the response surface model was further tested. The set temperature range, the time range, the O 2 concentration range and the O 2 flow range were 460-540°C, 10-30 min, 40%-100% and 100-300 mL/min respectively. As shown in Figure 6, the theoretical decomposition was 85.96 wt.% in the linear model 1. From the linear model 1, the reaction temperature, the reaction time, the O 2 concentration, the O 2 flow rate and the predicted decomposition were 478.97°C, 26.22 min, 92.74%, 161.73 mL/min and 85.96%. From Table 6, the parameters were optimized as 480°C, 26 min, 90% O 2 and 160 mL/min O 2 . The results of multiple recycles showed that the error between the actual value and the predicted value was between ±5%.    200 mL/min O 2 . Compared with the actual decomposition and the theoretical decomposition, the error was between ±2%, as shown in Table 7. The confidence coefficient of temperature was further improved.
The linear model 3 was established to explore the reliability of temperature on the degradation. The time parameter range was 5-35 min, and the rest of the parameter range remains unchanged. From Figure 8, the predicted decomposition was 76.33 wt.% under temperature of 540.54°C, reaction time of 12.49 min, O 2 concentration of 52.40%, O 2 flow rate of 108.24 mL/min. Combined with the actual situation, the process conditions were optimized to 540°C, 12.5 min, 50% O 2 and 110 mL/min O 2 . The error between the actual decomposition and the predicted decomposition was between ±5% Table 8.
From the 3D model and Pareto chart, the time and temperature were significant to the degradation. The influence of process parameters on the degradation were       analyzed by establishing three linear models, and the credibility of the two main factors (time and temperature) were supplemented and verified. The error was within 5% which was determined by comparing the actual and theoretical values. It showed that the reliability of the RSM model was enhanced.

Characterization of microscopic morphology
There were many types of CFRP waste, including expired prepregs, offcuts and end-of-life products.
However, most of the researches on CFRP wastes recycling for application had focused on expired prepreg. 15 Common CFRP wastes at the end-of-life cycle were selected in this dissertation, such as blocky CFRP and tubular CFRP. As Figure 9 shown, the filamentous rCF was recycled at 520°C, 23 min, 80% O 2 and 180 mL/min O 2 . The resin decomposition was shown in Figure 10. Compared with the oCF (Figure 11), the surface of rCF was smooth, and the resin-matrix was basically decomposed.

Gaseous product analysis
The CFRP gaseous products under the optimal parameters of recycle process were collected and analyzed by Fourier Transform Infrared (FTIR) Spectrometer instruments, and the main products were H 2 O and CO 2 ( Figure 12). 16 Compared with CO, CH 4 , benzene, ethylbenzene and other pollutants produced by pyrolysis and the acids and alcohols contained in the chemical recycle. [4][5][6][7] Thus, the recycled of CFRP by the process of thermally activated oxide semiconductors would not cause secondary pollution to the environment.

Raman spectral analysis
The graphite structure on the carbon fiber surface was analyzed by Raman spectroscopy, and the G peak and D peak of carbon element were respectively represented near 1,600 cm À1 and 1,350 cm À1 in the map. I G /I D represents the peak ratio of G and D, and the smaller the value, the more carbon atom crystal defects on the surface of the fiber, and it meant the graphite structure was affected in a certain extent. As shown in Figure 13 and Table 9, after treatment of high temperature and O 2 , the I G /I D value of rCF was reduced by 5.61%, and the graphite structure on the surface of carbon atoms was changed, but it did not affect the performance of rCF.

XPS analysis
XPS tests on oCF and rCF surfaces were shown in Figure 14 and Figure 15. The binding energies of C, N and O elements were in 248.6 eV, 532.0 eV and 399.5 eV, and the peak values of C-C bond, C = O bond, functional groups of C-OH and COOH were respectively around 284.8 eV, 286 eV, 287.5eV and 288.6 eV. The content of carbon fiber surface elements and functional groups were analyzed and quantified by Casa XPS software, as shown in Table 10.
Compared with the oCF, the C element and C-C bond on the     surface of the rCF increased significantly, and O/C, functional group of C-OH, C = O bond were reduced, and functional group of COOH appeared in the product.
Combined with the analysis of major gas-phase products (CO 2s ), C and O element react with O 2 in a certain proportion at high temperature to generate CO 2 . In addition, functional group of C-OH and O 2 react form functional group of COOH under high temperature conditions, and the production of COOH can improved the surface wettability of rCF, which was conducive to the interface bonding of rCF surface and resin-matrix. 17

AFM analysis
AFM and SEM were used to observed the surface roughness of oCF and rCF, as showed in Figure 16 and Figure 17. From the SEM image, we found that the fiber in few areas had residual resin on the surface. Therefore, SEM images were compared, and we found that the surface roughness of rCF was increased. It may be due to the remover of surface sizing of the fiber. Through SEM and AFM analysis, the rCF surface was clean and free of significant resin residue in most areas. It should be noted that the diameter of oCF and rCF were average value (GB-T 29,762-2013), and fiber measurement number more than 100.

Tensile strength and interfacial shear strength of oCF and rCF
The composite material interface evaluation device was shown in Figure 19. After testing, the tensile strength    was maintained at 99%, as well as the Young's modulus was maintained at 92% ( Figure 18). There were problems such as carbon deposits on the surface and obvious degradation of mechanical properties in pyrolysis recycle. However, there were no carbon deposition on the surface of rCF after disposing by thermal activation oxide semiconductor, and the mechanical property was excellent (The number of tests was 40). As the change of graphite structure and surface roughness, the surface interface bonding strength of rCF will be affected. From Figure 19, the carbon fiber reinforced epoxy resin (CF/EP), carbon fiber reinforced polylactic acid (CF/PLA), recycled carbon fiber reinforced epoxy resin (rCF/EP), and recycled carbon fiber reinforced polylactic acid (rCF/PLA) samples were prepared by microcapsule experiment. As shown in

Conclusion
(1) In this investigation, the model of the RSM was established considering perturbation diagram, Pareto diagram and 3D model. The model showed that the degradation was positively correlated with time and temperature, and the degree of influence was significant, the effect of O 2 flow and concentration were not significant. The influence degree of each factor: temperature > time > O 2 concentration > O 2 flow rate. (2) The three linear models indicated that the actual value error was less than 5%, and under the parameter of optimal recycling process, the resin-matrix decomposition of different CFRP wastes were higher than 96%, and the error was less than 2%. The model of RSM and linear models were highly reliable and could accurately predict the actual degradation. (3) Compared with the oCF, the surface roughness of rCF was increased, and the emergence of hydrophilic functional group (COOH) was conducive to new interfacial bonding. Furthermore, the increase of graphite structure defect on rCF surface is not obvious, and the I G /I D ratio only decreased by 5.61%. The tensile strength was maintained above 99%, the Young's modulus was maintained at 92%. Compared with oCF/PLA and oCF/EP, the shear strength of rCF/ PLA and rCF/EP interface were maintained at 85%.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work