Method for constructing identification model for recycled polycarbonate material and use thereof

Abstract

The present invention belongs to the field of plastic classification and identification, and provides a method for constructing an identification model for recycled polycarbonate material and the use thereof. In the present invention, by using the differences in various types of polycarbonates in terms of volatile compounds and using a volatile compound feature database, a prediction model is constructed, such that polycarbonates from different sample sources, recycled polycarbonate materials from different recycling sources, and virgin polycarbonate materials from different synthesis processes are accurately identified. The present invention is more accurate than ordinary comparison methods.

Description

A method for constructing an identification model for recycled polycarbonate materials and its application. Technical Field

[0001] This invention belongs to the field of plastic classification and identification, specifically relating to a method for constructing an identification model for recycled polycarbonate materials and its application. Background Technology

[0002] my country's 14th Five-Year Plan emphasizes strengthening the control of plastic pollution and encourages the recycling and reuse of waste plastic resources. Currently, approximately 99% of plastics are petroleum-based, and it is projected that by 2025, the global plastics industry will account for 16% of oil consumption and contribute 15% of carbon emissions. Research reports indicate that recycled plastics can significantly reduce the carbon emissions of products, and many developed countries are advocating that manufacturers add recycled plastics to their products.

[0003] Polycarbonate is an excellent engineering plastic with outstanding impact resistance and creep resistance, high tensile strength, flexural strength, elongation and rigidity, good heat and cold resistance, excellent electrical properties, low water absorption, and good light transmittance. In 2020, China's polycarbonate production capacity reached 1.79 million tons, accounting for approximately 30% of the global polycarbonate industry.

[0004] The main application markets for polycarbonate include electronics, sheet / film, automotive lighting, automotive windows, optics, home appliances, packaging, and medical fields. During the use, recycling, and reprocessing of plastics, pollutants, degradation products, organic impurities, oligomers, monomers, cleaning agents, and fragrances inevitably remain in recycled polycarbonate products. Furthermore, to restore the performance of recycled plastics to meet the application scenarios of the next life cycle, various modifying agents may be added to recycled polycarbonate plastics, such as antioxidants, chain extenders, transesterification agents, and crosslinking agents. Traditional physical property testing methods have limited applicability in distinguishing recycled plastics. Therefore, accurate identification of recycled polycarbonate plastics is an urgent need in this field.

[0005] However, the current industry certification of recycled plastic products is mostly done through document review, and there are few corresponding identification methods for whether the product contains recycled plastic. Most of the existing identification technologies rely on simple comparisons, which are rather crude and often have problems such as insufficient identification depth and low identification accuracy.

[0006] CN109870558 discloses a method for identifying recycled polycarbonate plastics. The method includes determining the peaks of the sample by headspace gas chromatography. However, judging the sample as recycled solely based on the peak time being greater than 14 minutes and the number of peaks being greater than 7 is prone to misjudgment and the accuracy of identification is relatively insufficient. Summary of the Invention

[0007] To address the technical problem of insufficient accuracy in identifying recycled polycarbonate plastics in existing technologies, this invention provides a method for constructing a model for identifying recycled polycarbonate materials and a method for identifying recycled polycarbonate materials.

[0008] The primary objective of this invention is to provide a method for constructing an identification model for recycled polycarbonate materials.

[0009] A secondary objective of this invention is to provide a method for identifying recycled polycarbonate materials.

[0010] The above-mentioned objective of the present invention is achieved through the following technical solution:

[0011] This invention protects a method for constructing an identification model for recycled polycarbonate materials, comprising the following steps:

[0012] Step S1: Obtain polycarbonate samples, which include polycarbonate recycled material samples from different recycling sources and polycarbonate virgin material samples from different synthesis processes. Classify and label the polycarbonate samples. The classification label includes at least one of sample source, recycling source, and synthesis process. The sample source includes virgin material and recycled material. The recycling source includes CDs, water bucket materials, car lights, black miscellaneous materials, small household appliances, and game console casings. The synthesis process includes phosgene and non-phosgene methods.

[0013] Step S2: Perform volatile compound analysis on the polycarbonate sample and establish a volatile compound characteristic database;

[0014] Wherein, when the classification label includes the sample source, the volatile compound characteristic database includes at least one of dichloromethane, chlorobenzene, methyl methacrylate, and at least one of n-heptane, styrene, benzene, and phenol;

[0015] When the classification label includes a recycling source, the volatile compound characterization database includes at least one of benzaldehyde, 1,4-dichlorobenzene, and cyclohexanone; at least one of 5,8-diethyldodecane, methyl 2-acrylate, undecane, nonadecane, and dodecane; at least one of methyl isobutyrate, 3-methyleneheptane, ethylbenzene, 6-ethyl-2-methyldecane, and tetratetradecane; at least one of 1-methoxy-2-propanol, 1-butanone, and trans-1-butenyloxy-pentane; and at least one of benzofuran and dichloromethane.

[0016] When the classification label includes a synthesis process, the volatile compound characteristic database includes at least one of carbon tetrachloride, acetone, m-xylene, p-tert-butylphenol, chlorobenzene, butyl 2-acrylate, and at least one of p-xylene, 2,6-bis(1,1-dimethylethyl)phenol, phenol, heptadecane, 1-butanol, and nonadecane.

[0017] Step S3: Perform data preprocessing, dimensionality reduction, and feature screening on the volatile compound feature database, and train and optimize at least one machine learning model to obtain at least one single prediction model.

[0018] This invention provides a method for constructing an identification model for recycled polycarbonate. Polycarbonates from different sample sources, recycled polycarbonates from different sources, and polycarbonates from different synthesis processes exhibit differences in their volatile compounds. This method utilizes these differences in volatile compounds by measuring them to construct a characteristic database based on volatile compounds. A predictive model is then built using this database, enabling the identification of polycarbonates from different sample sources, recycled polycarbonate from different sources, and virgin polycarbonate from different synthesis processes.

[0019] For identifying polycarbonate from different sample sources: the characteristic substances of virgin polycarbonate are dichloromethane, chlorobenzene, and methyl methacrylate, while the characteristic substances of recycled polycarbonate are n-heptane, styrene, benzene, and phenol. Therefore, the volatile compound characteristic database for identifying polycarbonate from different recycled sources should include at least one of dichloromethane, chlorobenzene, and methyl methacrylate, and at least one of n-heptane, styrene, benzene, and phenol.

[0020] To identify polycarbonate from different recycling sources: The characteristic substances of recycled polycarbonate from optical discs are benzaldehyde, 1,4-dichlorobenzene, and cyclohexanone; the characteristic substances of recycled polycarbonate from water buckets are 5,8-diethyldodecane, methyl 2-acrylate, undecane, nonadecane, and dodecane; the characteristic substances of recycled polycarbonate from black scrap materials are methyl isobutyrate, 3-methyleneheptane, ethylbenzene, 6-ethyl-2-methyl-decane, and tetratetradecane; the characteristic substances of recycled polycarbonate from automotive lighting materials are 1-methoxy-2-propanol, 1-butanone, and trans-1-butenyloxy-pentane; and the characteristic substances of recycled polycarbonate from small household appliances and game console casings include benzofuran and dichloromethane. Therefore, the database of volatile compounds for identifying polycarbonate from different recycling sources should include at least one of benzaldehyde, 1,4-dichlorobenzene, and cyclohexanone. At least one of 8-diethyldodecane, methyl 2-acrylate, undecane, nonadecane, and dodecane; at least one of methyl isobutyrate, 3-methyleneheptane, ethylbenzene, 6-ethyl-2-methyl-decane, and tetratetradecane; at least one of 1-methoxy-2-propanol, 1-butanone, and trans-1-butenyloxy-pentane; and at least one of benzofuran and dichloromethane, can be used to identify polycarbonate recycled materials from different recycling sources.

[0021] For identifying polycarbonates synthesized using different processes: the characteristic substances in virgin polycarbonate synthesized by the phosgene method are carbon tetrachloride, acetone, m-xylene, p-tert-butylphenol, chlorobenzene, and butyl 2-acrylate; the characteristic substances in virgin polycarbonate synthesized by the non-phosgene method are p-xylene, 2,6-bis(1,1-dimethylethyl)phenol, phenol, heptadecane, 1-butanol, and nonadecane. Therefore, the database of volatile compounds for identifying polycarbonates synthesized using different processes should at least include at least one of carbon tetrachloride, acetone, m-xylene, p-tert-butylphenol, chlorobenzene, and butyl 2-acrylate, and at least one of p-xylene, 2,6-bis(1,1-dimethylethyl)phenol, phenol, heptadecane, 1-butanol, and nonadecane, thus enabling the identification of virgin polycarbonate synthesized using different processes.

[0022] On the one hand, the types of volatile compounds in the volatile compound feature database are obtained through the analysis of a large number of polycarbonates, which has both universality and accuracy; on the other hand, the volatile compound feature database is used to train a machine learning model to obtain a prediction model, and the judgment is made using the prediction model, which has higher accuracy than ordinary comparison methods.

[0023] The black impurities described in this invention are black recycled impurity fragments or particles.

[0024] Specifically, the methods used for the analysis of volatile compounds in step S2 include, but are not limited to, headspace-gas chromatography.

[0025] More specifically, the headspace-gas chromatography method has an equilibrium temperature of 100~200℃ and an equilibrium time of 200~400min.

[0026] Specifically, the chromatographic column used in the headspace-gas chromatography method is selected from low-polarity columns or non-polar columns, and further selected from 100% methyl polysiloxane or 5% phenyl-95% methyl polysiloxane, including but not limited to Mega-5MS columns.

[0027] Specifically, the headspace-gas chromatography temperature program is as follows: initial temperature 30~50℃, held for 4~6 min, then increased to 240~280℃ at 6~10℃ / min, and held for 6~10 min.

[0028] Specifically, the flow rate of the headspace-gas chromatography method is 1.0~1.5 mL / min.

[0029] Specifically, the injection port temperature of the headspace-gas chromatography method is 220~280℃.

[0030] Specifically, the split ratio of the headspace-gas chromatography method is 1 to 10:1.

[0031] Specifically, the detectors used in the headspace-gas chromatography method include, but are not limited to, mass spectrometry detectors, FID detectors, ECD detectors, TCD detectors, and FPD detectors.

[0032] More specifically, the MSD transmission line temperature of the mass spectrometer detector is 250~300°C.

[0033] Specifically, the ion source temperature of the headspace-gas chromatography method is 200~250℃.

[0034] Preferably, the construction method further includes the following steps:

[0035] Step S4: Assign weights to the prediction results of at least two individual prediction models based on the optimal weighting method, and obtain a combined prediction model after training and optimization.

[0036] To further improve the accuracy of identification, this invention further constructs a combined prediction model based on the optimal weighting method.

[0037] Specifically, the machine learning model is selected from at least one of Partial Least Squares Discriminant Analysis (PLS-DA), Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), Principal Component Analysis (PCA), Decision Tree (DT), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), Bayesian Network, Artificial Neural Network (ANN), and AdaBoosting. Preferably, when the classification label includes sample source, the machine learning model is Random Forest, Decision Tree, or AdaBoosting; when the classification label includes synthesis process, the machine learning model includes Decision Tree or AdaBoosting; when the classification label includes recycling source, the machine learning model includes Random Forest or AdaBoosting.

[0038] Specifically, the data preprocessing includes data aggregation and / or data cleaning; the data dimensionality reduction employs principal component analysis. To reduce the database complexity used for model training, testing, and optimization, this invention uses principal component analysis to perform dimensionality reduction on the volatile characteristic database.

[0039] This invention also protects a method for identifying recycled polycarbonate materials, comprising the following steps:

[0040] Step S1: Obtain the polycarbonate sample to be identified;

[0041] Step S2: Perform volatile compound analysis on the polycarbonate sample to be identified to obtain the dataset to be identified;

[0042] Step S3: Input the dataset to be identified into the single-item prediction model above to obtain the classification label of the polycarbonate sample to be identified.

[0043] Specifically, the methods used for the analysis of volatile compounds in step S2 include, but are not limited to, headspace-gas chromatography.

[0044] More specifically, the headspace-gas chromatography method has an equilibrium temperature of 100~200℃ and an equilibrium time of 200~400min.

[0045] Specifically, the chromatographic column used in the headspace-gas chromatography method is selected from low-polarity columns or non-polar columns, and further selected from 100% methyl polysiloxane or 5% phenyl-95% methyl polysiloxane, including but not limited to Mega-5MS columns.

[0046] Specifically, the headspace-gas chromatography temperature program is as follows: initial temperature 30~50℃, held for 4~6 min, then increased to 240~280℃ at 6~10℃ / min, and held for 6~10 min.

[0047] Specifically, the flow rate of the headspace-gas chromatography method is 1.0~1.5 mL / min.

[0048] Specifically, the injection port temperature of the headspace-gas chromatography method is 220~280℃.

[0049] Specifically, the split ratio of the headspace-gas chromatography method is 1 to 10:1.

[0050] Specifically, the detectors used in the headspace-gas chromatography method include, but are not limited to, mass spectrometry detectors, FID detectors, ECD detectors, TCD detectors, and FPD detectors.

[0051] More specifically, the MSD transmission line temperature of the mass spectrometer detector is 250~300°C.

[0052] Specifically, the ion source temperature of the headspace-gas chromatography method is 200~250℃.

[0053] Compared with existing technologies, the beneficial effects of this technical solution are as follows:

[0054] This invention provides a method for constructing a polycarbonate recycled material identification model. By utilizing the differences in volatile compounds among various types of polycarbonate, and constructing a predictive model using a volatile compound characteristic database, it is possible to identify polycarbonate from different sample sources, polycarbonate recycled materials from different recycling sources, and polycarbonate virgin materials from different synthesis processes.

[0055] This invention obtains the volatile characteristic substances in various types of polycarbonate by analyzing a large number of polycarbonates, which has both universality and accuracy. Moreover, the use of predictive models for judgment has higher accuracy than ordinary comparison methods. Attached Figure Description

[0056] Figure 1 shows the confusion matrix results of various machine learning models for virgin materials and recycled materials in Example 3;

[0057] Figure 2 shows the PLS-DA score diagram (a), three-dimensional scatter plot (b), load diagram (c), and bi-label diagram (d) of volatile substances in polycarbonate virgin materials for different synthesis processes in Example 4. Embodiments of the present invention

[0058] The method will be described below with reference to specific examples.

[0059] Example 1: A method for constructing an identification model for recycled polycarbonate materials

[0060] This embodiment provides a method for constructing an identification model for recycled polycarbonate materials, including the following steps:

[0061] Step S1: Obtain polycarbonate samples, which include polycarbonate recycled material samples from different recycling sources and polycarbonate virgin material samples from different synthesis processes. Classify and label the polycarbonate samples. The classification labels include sample source, recycling source, and synthesis process. The sample source includes virgin material and recycled material. The recycling sources include CDs, buckets, bucket sheets, car lights, black miscellaneous materials, small household appliances, and game console casings. The synthesis processes include phosgene and non-phosgene methods.

[0062] Step S2: Perform volatile compound analysis on the polycarbonate sample and establish a volatile compound characteristic database;

[0063] The analysis of volatile compounds employs headspace gas chromatography, and the headspace gas chromatography test method is as follows:

[0064] Weigh 2 g (accurate to 0.001 g) of sample into a 20 mL headspace vial for testing. Seal immediately after weighing to minimize loss of volatile substances. Equilibrate the vial at 150 °C for 300 min. Then, aspirate 1000 μL of the volatile substance using a headspace gas chromatography-mass spectrometry (GCMS 1000, China Hexin Mass Spectrometry Co., Ltd.) for analysis.

[0065] Chromatographic conditions: Mega-5ms column, dimensions 30 m × 0.25 mm × 0.25 μm; carrier gas: helium, flow rate: 1.2 mL / min; injection port temperature: 250 ℃; column oven temperature program: initial temperature 40 ℃, hold for 5 min, increase to 260 ℃ at 8 ℃ / min, hold for 8 min; split ratio: 5:1.

[0066] Mass spectrometry conditions: MSD transfer line temperature: 280 ℃; acquisition mode: full scan; scan mass range: 35~550 amu; ion source temperature: 220 ℃; solvent delay: 0.1 min;

[0067] The volatile compound database includes the compounds shown in Table 1:

[0068] Table 1. Volatile Compound Database

[0069]

[0070]

[0071]

[0072] Continued from Table 1: Volatile Compound Database

[0073]

[0074]

[0075] Continued from Table 1: Volatile Compound Database

[0076]

[0077]

[0078]

[0079]

[0080] Continued from Table 1: Volatile Compound Database

[0081]

[0082]

[0083] Continued from Table 1: Volatile Compound Database

[0084]

[0085] Step S3: Perform data preprocessing, dimensionality reduction, and feature screening on the volatile compound database, and train and optimize at least one machine learning model to obtain at least one single prediction model.

[0086] Example 2: A method for identifying recycled polycarbonate materials

[0087] This embodiment provides a method for identifying recycled polycarbonate materials, including the following steps:

[0088] Step S1: Obtain the polycarbonate sample to be identified;

[0089] Step S2: Perform volatile compound analysis on the polycarbonate sample to be identified to obtain the dataset to be identified;

[0090] Step S3: Input the dataset to be identified into the single-item prediction model above to obtain the classification label of the polycarbonate sample to be identified;

[0091] The analysis of volatile compounds employs headspace gas chromatography, and the headspace gas chromatography test method is as follows:

[0092] Weigh 2 g (accurate to 0.001 g) of sample into a 20 mL headspace vial for testing. Seal immediately after weighing to minimize loss of volatile substances. Equilibrate the vial at 150 °C for 300 min. Then, aspirate 1000 μL of the volatile substance using a headspace gas chromatography-mass spectrometry (GCMS 1000, China Hexin Mass Spectrometry Co., Ltd.) for analysis.

[0093] Chromatographic conditions: Mega-5ms column, dimensions 30 m × 0.25 mm × 0.25 μm; carrier gas: helium, flow rate: 1.2 mL / min; injection port temperature: 250 ℃; column oven temperature program: initial temperature 40 ℃, hold for 5 min, increase to 260 ℃ at 8 ℃ / min, hold for 8 min; split ratio: 5:1.

[0094] Mass spectrometry conditions: MSD transfer line temperature: 280 ℃; acquisition mode: full scan; scan mass range: 35~550 amu; ion source temperature: 220 ℃; solvent delay: 0.1 min.

[0095] Example 3: A method for constructing an identification model for recycled polycarbonate materials

[0096] Based on the construction method of Example 1, this embodiment uses machine learning models such as SVM, RF, KNN, DT, and adaboosting to construct five single prediction models for the source of polycarbonate samples.

[0097] The validation results of the five single-item prediction models are shown in Figure 1 and Table 2. Characteristic substances were screened based on a VIP value > 1. The characteristic substances for virgin polycarbonate included dichloromethane, chlorobenzene, and methyl methacrylate, while the characteristic substances for recycled polycarbonate included n-heptane, styrene, benzene, and phenol.

[0098] Table 2 Evaluation results of various machine learning models for virgin and recycled materials

[0099]

[0100] For recycled polycarbonate plastics, the RF, DT, and adaboosting algorithms can achieve 100% accuracy in identifying volatile components in polycarbonate samples. Combining chemical characterization analysis with machine learning algorithms can effectively guide the qualitative identification of recycled plastics.

[0101] Example 4: A method for constructing an identification model for recycled polycarbonate materials

[0102] This embodiment is based on the construction method of Embodiment 1, and uses the PLS-DA machine learning model to construct a single prediction model for the synthesis process.

[0103] The validation results of the above single-item prediction model are shown in Figure 2 and Table 3. Characteristic substances were screened based on a VIP value > 1. The characteristic substances in virgin polycarbonate synthesized by the phosgene method include carbon tetrachloride, acetone, m-xylene, p-tert-butylphenol, chlorobenzene, and butyl 2-acrylate; the characteristic substances in virgin polycarbonate synthesized by the non-phosgene method include p-xylene, 2,6-bis(1,1-dimethylethyl)phenol, phenol, heptadecane, 1-butanol, and nonadecane.

[0104] Table 3 Evaluation results of various machine learning models for PC based on phosgene and non-phosgene methods.

[0105]

[0106] As shown in Table 3, the single-item identification models constructed using RF, DT, and adaboosting algorithms can accurately identify polycarbonate virgin materials from different synthesis processes. Among them, the single-item identification models constructed using DT and adaboosting algorithms can achieve an accuracy of 100%.

[0107] Example 5: A method for constructing an identification model for recycled polycarbonate materials

[0108] This embodiment is based on the construction method of Embodiment 1, and uses the PLS-DA machine learning model to construct a single prediction model for recycling sources.

[0109] The validation results of the above single-item prediction models are shown in Table 4. Characteristic substances were screened based on a VIP value > 1. Specifically, the characteristic substances of recycled polycarbonate from CD / DVD sources include benzaldehyde, 1,4-dichlorobenzene, and cyclohexanone; those from water bucket sources include 5,8-diethyldodecane, methyl 2-acrylate, undecane, nonadecane, and dodecane; those from black mixed materials sources include methyl isobutyrate, 3-methyleneheptane, ethylbenzene, 6-ethyl-2-methyl-decane, and tetratetradecane; those from automotive lighting materials include 1-methoxy-2-propanol, 1-butanone, and trans-1-butenyloxy-pentane; and those from small household appliances and game console casings include benzofuran and dichloromethane.

[0110] Table 4. Evaluation results of various machine learning models for recycled PC from different recycling sources.

[0111]

[0112] As shown in Table 4, the single-item identification models constructed using RF, DT, and adaboosting algorithms can all accurately identify polycarbonate recycled materials from different recycling sources. Among them, the single-item identification models constructed using RF and adaboosting algorithms can achieve an accuracy of 100%.

[0113] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for constructing an identification model for recycled polycarbonate materials, comprising the following steps: Step S1: Obtain polycarbonate samples, which include polycarbonate recycled material samples from different recycling sources and polycarbonate virgin material samples from different synthesis processes. Classify and label the polycarbonate samples. The classification label includes at least one of sample source, recycling source, and synthesis process. The sample source includes virgin material and recycled material. The recycling source includes CDs, water bucket materials, car lights, black miscellaneous materials, small household appliances, and game console casings. The synthesis process includes phosgene and non-phosgene methods. Step S2: Perform volatile compound analysis on the polycarbonate sample and establish a volatile compound characteristic database; Wherein, when the classification label includes the sample source, the volatile compound characteristic database includes at least one of dichloromethane, chlorobenzene, methyl methacrylate, and at least one of n-heptane, styrene, benzene, and phenol; When the classification label includes a recycling source, the volatile compound characterization database includes at least one of benzaldehyde, 1,4-dichlorobenzene, and cyclohexanone; at least one of 5,8-diethyldodecane, methyl 2-acrylate, undecane, nonadecane, and dodecane; at least one of methyl isobutyrate, 3-methyleneheptane, ethylbenzene, 6-ethyl-2-methyldecane, and tetratetradecane; at least one of 1-methoxy-2-propanol, 1-butanone, and trans-1-butenyloxy-pentane; and at least one of benzofuran and dichloromethane. When the classification label includes a synthesis process, the volatile compound characteristic database includes at least one of carbon tetrachloride, acetone, m-xylene, p-tert-butylphenol, chlorobenzene, butyl 2-acrylate, and at least one of p-xylene, 2,6-bis(1,1-dimethylethyl)phenol, phenol, heptadecane, 1-butanol, and nonadecane. Step S3: Perform data preprocessing, dimensionality reduction, and feature screening on the volatile compound feature database, and train and optimize at least one machine learning model to obtain at least one single prediction model.

2. The construction method according to claim 1, characterized in that, The methods used for analyzing volatile compounds in step S2 include headspace gas chromatography or solid-phase microextraction-gas chromatography.

3. The construction method according to claim 2, characterized in that, The equilibrium temperature for the headspace-gas chromatography method is 100~200℃, and the equilibrium time is 200~400min.

4. The construction method according to claim 2, characterized in that, The chromatographic column used in the headspace-gas chromatography method is selected from low-polarity columns or non-polar columns.

5. The construction method according to claim 2, characterized in that, The headspace-gas chromatography temperature program is as follows: initial temperature 30~50℃, hold for 4~6 min, increase to 240~280℃ at 6~10℃ / min, and hold for 6~10 min.

6. The construction method according to claim 1, characterized in that, The construction method further includes the following steps: Step S4: Assign weights to the prediction results of at least two individual prediction models based on the optimal weighting method, and obtain a combined prediction model after training and optimization.

7. The construction method according to claim 1, characterized in that, The machine learning model is selected from at least one of PLS-DA, OPLS-DA, PCA, decision tree, support vector machine, random forest, K-nearest neighbors, Bayesian network, artificial neural network, and AdaBoosting.

8. The construction method according to claim 1, characterized in that, The data preprocessing includes data aggregation and / or data cleaning.

9. The construction method according to claim 1, characterized in that, The data dimensionality reduction was achieved using principal component analysis.

10. A method for identifying recycled polycarbonate materials, characterized in that, Includes the following steps: Step S1: Obtain the polycarbonate sample to be identified; Step S2: Perform volatile compound analysis on the polycarbonate sample to be identified to obtain the dataset to be identified; Step S3: Input the dataset to be identified into any of the single prediction models described in claims 1 to 9 to obtain the classification label of the polycarbonate sample to be identified.

Images