JP2005249624A - Polymer group discriminating method by optical spectrum - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a polymer group classification method by optical spectrum capable of precisely classifying unknown polymeric materials into respective material types. <P>SOLUTION: In the polymer group classification method by optical spectrum, the unknown polymer materials including various polymers in a mixed manner are discriminated into respective material types. When absorption strength of a specific wave length of near infrared spectrum is used to discriminate respective polymer groups by binary tree classification, discrimination of polymer groups is combined step by step and polymer discrimination is conducted so that incorrect discrimination ratio is extremely low in total by using a hierarchical discrimination flow starting from classification of a polymer group of which incorrect classification ratio is lowest and gradually proceeding classification of polymer groups having higher incorrect classification ratio. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a method for discriminating polymer groups from a spectral spectrum.

  Infrared absorption spectrum has been widely used as a simple and rapid analysis method for polymers. Therefore, qualitative methods for polymer materials from spectra have been well studied.

  On the other hand, in recent years, measurement methods using near-infrared spectra have been widely recognized for their ease of pre-measurement processing and the richness of measurement information that can be obtained. Interest has also been directed to. Along with this, studies on a method for discriminating polymers for the purpose of polymer separation for plastic recycling are also in progress.

  Conventionally, as a method for discriminating a polymer, a method using multivariate analysis such as a principal component analysis method or a multidimensional scaling method, or a method using a neural network has been used. In general, it is known that a method using a neural network capable of processing nonlinear data can perform more accurate discrimination than multivariate analysis such as principal component analysis or multidimensional scaling. However, in the method using the neural network, it is not always clear what the spectrum is and how it is viewed, and it is a black box. In addition, in the multivariate analysis, the spectrum pattern changes depending on other additives and the like, so the misclassification rate is not low (about 15 to 20%). However, even in a neural network with a low misclassification rate, a copolymer such as ABS has a plurality of components coexisting, so the misclassification rate exceeds about 15 to 20%.

  The prior art includes those disclosed below.

  (1) An example in which discrimination of five polymer groups by principal component analysis and data ternarization is applied to 14 plastic samples (see Non-Patent Document 1 below).

  (2) Reported that the average percentage of polymer group discrimination was about 80% in the spectrum of nearly 300 polymer samples by the neural network method (see Non-Patent Document 2 below).

In addition, the following non-patent documents 3 to 5 are disclosed.
“Everything about analytical chemistry that can be understood well” edited by Japan Analytical Instruments Industry Association, Nikkan Kogyo Shimbun (issued on October 25, 2001) p. 171-191 “Non-destructive discrimination of plastic waste by combining near infrared spectroscopy and neutral network analysis” analytical chemistry, No. 5, pp. 483-489, 1999 Norihisa Kusagawa, Plastic Age, Vol. 48, special issue, 29 (2002) Feldhoff, R.A. Wienke, D .; , Cammann, K .; , Fuchs, H .; , On-Line Post Consumer Package Identification by NIR Spectroscopy Combined with a Fuzzy ARTMAP Classifier in an Industrial Environmental, Applied Applied. 51, no. 3, 1997, p. 362-368 van den Broek, WHAM. Wienke, D .; , Melssen, WJ. , Buydens, LMC. , Plastic material identification with spectroscopic near infrared imaging and artificial neural networks, Analytical Chimera-Acta-Incubating Cumulative. 361, no. 1, 1998, p. 161-176

  In order to recycle waste plastics, it is necessary to first classify the collected waste materials by material type, and if they are correctly classified, the recycling process becomes easy, but the conventional classification method has a misclassification rate of 20 Because it was about%, it was limited to some uses such as bottle material identification.

  In the present invention, when classifying by a binary tree using the absorption intensity of a specific wavelength in the near-infrared absorption spectrum, if classification is performed according to a flow of separating from a grade with a low misclassification rate, the misclassification rate is at most. Focusing on the fact that it can be suppressed to about 2.5%, it is an object of the present invention to provide a method for discriminating polymer groups based on a spectroscopic spectrum capable of classifying unknown polymer materials into material types with high accuracy.

In order to achieve the above object, the present invention provides
[1] In a method for discriminating individual polymer groups by spectral spectrum, in which unknown polymer materials in which various polymers are mixed are discriminated by material type, individual polymer groups are identified by the absorption intensity at a specific wavelength in the near-infrared absorption spectrum. When classifying by binary tree, the classification of individual polymer groups is combined in stages, starting with the classification of the polymer group with the lowest misclassification rate, and gradually classifying the polymer group with a large misclassification rate. A hierarchical discrimination flow to be performed is characterized in that a polymer having a very low misclassification rate as a total is discriminated.

  In the conventional method for classifying polymer groups based on a spectral spectrum, a polymer group with a high misclassification rate has a misclassification rate of about 15 to 20% (for example, in the case of ABS). According to the classification method, the level was about 2.5% at most, which was extremely low.

  Therefore, it is particularly effective as a practical, simple and rapid classification method for plastic material recycling.

  When classifying by binary tree using the absorption intensity of a specific wavelength in the near-infrared absorption spectrum to distinguish individual polymer groups, the polymers with the lowest misclassification rate are combined in stages. Start with group discrimination, sort from grades with small misclassification rates, gradually identify polymer groups with large misclassification rates, and create a hierarchical discrimination flow that gradually sorts grades with high misclassification rates. As a result, a polymer with a very low misclassification rate is classified as a total.

  Therefore, the misclassification rate can be reduced to approximately 1/10 of the conventional method in the method of discriminating the polymer group based on the spectral spectrum that discriminates the member in which various polymers are mixed.

  Hereinafter, embodiments of the present invention will be described in detail.

  FIG. 1 is a schematic diagram of a system for discriminating polymer groups based on a spectral spectrum showing an embodiment of the present invention.

  In this figure, A is a discriminating device, 1 is a high-frequency power source (30 to 80 MHz), 2 is a piezo element, 3 is an AOTF (automatic light conversion function), 4 is a damper, 5 is a light source, 6 is transmitted light, 7 Is reflected light (0.9 to 2.4 μm irradiation light), 8 is an unknown polymer material as an object to be measured, 9 is diffuse reflection light or transmitted reflection light, 10 is a light receiving element, and 11 is a polymer group discrimination device. , 12 is an absorption spectrum generation unit (monitor), 13 is a known data storage unit, 14 is a comparison unit between an absorption spectrum and known data, and 15 is a polymer group discrimination unit.

  Therefore, when the discrimination device A is pressed against an unknown polymer material 8 as a measurement object (sample) and a switch (not shown) is pressed, the reflected light 7 from the light source 5 via the AOTF 3 is high as the measurement object. The light is incident on the molecular material 8 and the absorption spectrum is displayed on the monitor 11 after about 3 seconds. The measurement wavelength region is 1200 to 2400 nm, and when the measurement object 8 is transparent, a ceramic plate (not shown) is placed on the opposite side so that the discrimination device A near infrared light can be obtained.

  Thus, according to the present invention, the measurement is simple. That is, there is little need for pretreatment of the sample (measurement object), and measurement can be performed instantaneously.

  As measurement samples, polymer groups (total 18 groups, total number of samples 280) in which three or more kinds of polymers belonging to each polymer group were collected were used.

  Table 1 shows polymer materials used as measurement samples, the number of samples, and misclassification rates by pretreatment.

以下、本ポリマーグループの判別システムにより、どのようにポリマーグループを判別するか詳述する。

Hereinafter, it will be described in detail how the polymer group is discriminated by this polymer group discrimination system.

  The inventors of the present invention collect and measure about 280 types of polymer samples shown in Table 1 to determine where and how in the near-infrared spectrum can be easily determined. And examined the discrimination. The goal is to devise simple guidelines for classification that even beginners of spectrum analysis can handle, and considering the simplicity of measurement, which is an advantage of near-infrared spectroscopy, operations such as preprocessing at the measurement stage We proceeded with an analysis-oriented attitude.

  Specifically, as shown in Table 1, about 280 kinds of polymer samples are divided into about 18 groups such as polystyrene, polyethylene, ABS resin, polypropylene, polycarbonate, and polyvinyl chloride. The decision tree (binary tree) was used to classify the characteristics of the other groups.

  Binary trees are known as exploratory information analysis methods, and their contents are divided into two parts, those that satisfy a specific level of explanatory variables and those that do not satisfy quantitative or qualitative objective variables. In the present invention, however, the misclassification rate of the qualitative objective variable is calculated at each stage, and the binary tree is successively grown while selecting the explanatory variable that minimizes it. The objective variable here is whether or not an unknown sample is a polymer to be determined, and the explanatory variable is the magnitude of a specific wavelength intensity in the near-infrared absorption spectrum of the sample.

  That is, in the present invention, it is determined whether or not an unknown polymer material belongs to a specific group (for example, a polyethylene group) by determining several wavelengths of the spectrum and the magnitude of the spectrum intensity (the spectrum intensity is generally Standardization) is determined by a binary tree with a misclassification rate.

  Therefore, first, spectrum preprocessing (standardization) for each polymer group will be described.

  FIG. 2 is a diagram showing an overlay of spectral total spectra obtained by measurement, FIG. 3 is a diagram in which a polystyrene group is extracted from FIG. 2, and FIG. 4 is a diagram in which a polyethylene group is extracted from FIG. In these figures, the vertical axis represents absorption intensity and the horizontal axis represents wavelength.

  FIG. 2 shows that there are common spectral peaks in the vicinity of 1400 nm, 1650 nm, and 2400 nm. From these, as shown in FIG. 3, the polystyrene group and the polyethylene group as shown in FIG. 4 were taken out and compared.

  Then, as preprocessing for reducing the variance of each group spectrum, normalization of the raw spectrum (for example, the total is set to 100 or 1.00) and the second derivative spectrum is normalized. .

  Next, a binary tree was used to determine whether the unknown polymer material was polystyrene or other. As a result, in the case of polystyrene shown in FIG. 3, discrimination was performed using the normalized absorbance at 1258 nm, 2192.5 nm, and 2268.5 nm. ) Was the correct answer (misclassification rate 2.0%). Similarly, in the case of the polyethylene shown in FIG. 4, the misclassification rate was 0.4% as determined using the normalized absorbance magnitudes of 1288.5 nm, 1305.5 nm, and 1663 nm.

  In this way, it is determined whether or not an unknown polymer material belongs to a specific group (for example, a polyethylene group) by determining several wavelengths of the spectrum and the magnitude of the spectrum intensity (spectrum intensity is generally normalized). For example, the misclassification rate is small as compared with the conventional neural network method. For example, with respect to ABS, the misclassification rate of the present invention is about 9%, and the misclassification rate is small even compared to the misclassification rate by the neural network (about 20% or more). However, binary trees alone are insufficient for practical use. In the present invention, the concept of hierarchical tree hierarchical discrimination is greatly expanded, and discrimination by decision trees (for example, whether or not a polyethylene group is used). Combining a large number of discriminants), it was developed into discrimination based on a step-by-step flowchart. That is, the type of misclassified polymer in a group with a large misclassification rate, such as ABS, is investigated, and the misclassified polymer is discriminated before the ABS (step S28) determination, so that the total As a result, a very low misclassification rate was obtained.

  That is, the determination was made by creating a hierarchical determination flowchart from a low misclassification rate to a high misclassification rate.

  Therefore, as in the case of the polystyrene and polyethylene described above, the other 16 groups were examined. As a result, focusing on the normalized absorbance at wavelengths at about 2 to 4 points, each of the 18 groups shown in Table 1 was obtained. The misclassification rate for was obtained. From this result, polymers with a high misclassification rate were PC / ABS and ABS. Since these polymers are originally composed of a copolymer or the like, it can be considered as a natural result.

  Based on the result of this misclassification rate, we created a hierarchical classification flow of polymer groups based on the spectrum of unknown polymer materials. In this flow, it is configured to classify from a group with a small misclassification rate and to separate them.

  FIG. 5 is a flowchart showing an example of classification of polymer groups based on the spectrum of an unknown polymer material according to an embodiment of the present invention, and FIG. 6 is a diagram showing a misclassification rate according to the flowchart (the comparative example in FIG. This shows the misclassification rate by the discriminating method of classifying using a binary tree based on the normalized spectrum.)

  (1) First, a standardized spectrum of an unknown polymer material is taken into the polymer group discriminator 11 (step S1).

  (2) Next, when the normalized absorption intensity is smaller than 2.815 at the wavelength of 1999 nm of the spectrum, the process proceeds to the next step S3, and when it is larger, it is determined to be melamine (step S2).

  (3) Next, at the wavelength of 1429.5 nm of the spectrum, if the normalized absorption intensity is smaller than 1.661, the process proceeds to the next step S4, and if larger, it is determined to be polyphenol (step S3). ).

  (4) Next, when the normalized absorption intensity is smaller than 3.098 at the wavelength of 2053.5 nm of the spectrum, the process proceeds to the next step S5, and when it is larger, it is determined that it is nylon 6 (step). S4).

  (5) Next, when the normalized absorption intensity is smaller than 0.001 at a wavelength of 1234.5 nm of the spectrum, it is determined that the resin is urea resin, and when it is larger, the process proceeds to the next step S6 (step S6). S5).

  (6) Next, when the absorption intensity after normalization is smaller than 1.069 at the wavelength 1533.5 nm of the spectrum, the process proceeds to the next step S7, and when it is larger, it is determined to be nylon 66 (step). S6).

  (7) Next, if the normalized absorption intensity is smaller than 3.389 at the wavelength 2192.5 nm of the spectrum, the process skips to step S10, and if larger, the process proceeds to the next step S8 (step S7). .

  (8) Next, when the normalized absorption intensity is smaller than 3.024 at a wavelength of 2268.5 nm of the spectrum, the process proceeds to the next step S9, and when larger, the process skips to step S10 (step S8). .

  (9) Next, if the absorption intensity after normalization is smaller than 2.403 at the wavelength of 1747 nm of the spectrum, it is determined to be polystyrene, and if larger, the process proceeds to the next step S10 (step S9). .

  (10) Next, when the absorption intensity after normalization is smaller than 0.002 at the wavelength of 1276 nm of the spectrum, it is determined that the spectrum is polyoxymethylene, and when it is larger, the process proceeds to the next step S11 (step S10). ).

  (11) Next, when the absorption intensity after normalization is smaller than 2.705 at the wavelength of 1899 nm of the spectrum, the process proceeds to the next step S12, and when it is larger, it is determined to be cellulose (step S11).

  (12) Next, when the absorption intensity after normalization is smaller than 0.390 at the wavelength of 1327 nm of the spectrum, the process proceeds to the next step S13, and when it is larger, it is determined to be polymethyl methacrylate (step). S12).

  (13) Next, when the normalized absorption intensity is smaller than 3.086 at a wavelength of 2138.5 nm of the spectrum, the process proceeds to the next step S14, and when it is larger, it is determined to be polycarbonate (step S13). ).

  (14) Next, at the wavelength of 1258 nm of the spectrum, when the normalized absorption intensity is smaller than 0.001, it is determined as polybutylene terephthalate, and when it is larger, the process proceeds to the next step S15 (step S14). ).

  (15) Next, when the normalized absorption intensity is smaller than 3.447 at a wavelength of 2261.5 nm of the spectrum, the process proceeds to the next step S16, and when it is larger, it is determined to be polyethylene terephthalate (step). S15).

  (16) Next, when the absorption intensity after normalization is smaller than 3.196 at the wavelength of 2150.2 nm of the spectrum, the process proceeds to the next step S17, and when it is larger, it is determined to be polycarbonate / ABS resin. (Step S16).

  (17) Next, if the absorption intensity after normalization is smaller than 0.762 at the wavelength of 1663 nm of the spectrum, the process proceeds to the next step S18, and if larger, the process skips to step S20 (step S17).

  (18) Next, when the normalized absorption intensity is smaller than 0.007 at a wavelength of 1288.5 nm of the spectrum, the process skips to step S20, and if larger, the process proceeds to the next step S19 (step S18). .

  (19) Next, when the absorption intensity after normalization is smaller than 4.908 at the wavelength of 2313.5 nm of the spectrum, it is determined as polyethylene, and when it is larger, the process proceeds to the next step S20 (step S19). ).

  (20) Next, at the wavelength of 1689 nm of the spectrum, when the normalized absorption intensity is smaller than 1.571, the process proceeds to step S21, and when larger, the process proceeds to step S23 (step S20).

  (21) Next, when the absorption intensity after normalization is smaller than 0.525 at the wavelength of 1382 nm of the spectrum, it is determined that the polyvinyl chloride is larger, and if larger, the process proceeds to step S23 (step S21).

  (22) Next, when the absorption intensity after normalization is smaller than 1.037 at a wavelength of 1658.5 nm of the spectrum, it is determined that it is polypropylene, and when it is larger, the process proceeds to the next step S23 (step S22). ).

  (23) Next, if the absorption intensity after normalization is smaller than 0.001 at the wavelength of 1267.5 nm of the spectrum, the process proceeds to step S24, and if larger, the process proceeds to step S25 (step S23).

  (24) Next, if the absorption intensity after normalization is smaller than 2.967 at the wavelength of 2271 nm of the spectrum, the process proceeds to step S26, and if it is larger, it is determined to be an acrylonitrile styrene copolymer (step S24). ).

  (25) Next, when the absorption intensity after normalization is smaller than 0.001 at a wavelength of 1552.5 nm of the spectrum, it is determined that the copolymer is acrylonitrile styrene and a copolymer, and if larger, the next step S26. (Step S25).

  (26) Next, when the absorption intensity after normalization is smaller than 2.525 at the wavelength of 2152.5 nm of the spectrum, it is determined that it is other, and when it is larger, the process proceeds to the next step S27 (step S26). ).

  (27) Next, when the normalized absorption intensity is smaller than 3.281 at a wavelength of 2160.5 nm of the spectrum, the process proceeds to the next step S28, and when it is larger, it is determined to be other (step S27). ).

  (28) Finally, at the wavelength of 2124 nm of the spectrum, if the normalized absorption intensity is smaller than 2.657, it is determined that the resin is ABS resin, and if it is larger, it is determined that it is other (step S28). .

  As a result of the discrimination using the above flowchart, as shown in FIG. 6, in the case of belonging to an individual group (comparative example), the misclassification rate (PC / ABS) is about 13% at the maximum. However, according to the present invention, by using a hierarchical discrimination flowchart, it was possible to finally achieve a misclassification rate of 2.5% or less.

  Thus, according to the method for classifying polymer groups based on the spectral spectrum of the present invention, compared to the conventional method (multivariate analysis, neural network, etc.), the intensity of which wavelength of the spectrum is used as the determination condition. It is characterized by the fact that it is clear, the influence of other coexisting substances such as additives is small, and the misclassification rate is extremely small (at a level of 2.5% or less in the classification of 18 groups).

  In the above embodiment, the determination based on the binary tree is used. However, the determination is not limited to this, and the determination based on the decision tree may be used.

  Further, the present invention is not limited to the above-described embodiments, and various modifications can be made based on the spirit of the present invention, and these are not excluded from the scope of the present invention.

  The method for classifying a polymer group based on a spectral spectrum of the present invention is a practical, simple and quick separation method for recycling plastics, and is particularly suitable for plastic production using recycled plastics and the synthetic fiber field.

It is a schematic diagram of the classification system of the polymer group by the spectrum which shows the Example of this invention. It is a figure which shows the overwriting of all the spectral spectra obtained by measurement. It is the figure which took out the polystyrene group from FIG. It is the figure which took out the polyethylene group from FIG. It is a flowchart which shows the example of a classification | category of the polymer group by the spectrum of the unknown polymeric material which shows the Example of this invention. It is a figure which shows the misclassification rate by the flowchart of FIG.

Explanation of symbols

A discriminator 1 High frequency power supply (30-80MHz)
2 Piezo elements 3 AOTF (automatic light conversion function)
4 Damper 5 Light source 6 Transmitted light 7 Reflected light (0.9-2.4 μm irradiation light)
8 Unknown Polymer Material as Measurement Object 9 Diffuse Reflected Light or Transmitted Reflected Light 10 Light Receiving Element 11 Polymer Group Discriminating Device 12 Absorption Spectrum Generation Unit (Monitor)
13 Known Data Storage Unit 14 Absorption Spectrum and Known Data Verification Unit 15 Polymer Group Discrimination Unit

Claims (1)

In the method of discriminating polymer groups by spectroscopic spectrum that discriminates unknown polymer materials mixed with various polymers by material type,
When classifying with a decision tree using the absorption intensity of a specific wavelength in the near-infrared absorption spectrum to distinguish individual polymer groups, polymer groups with the smallest misclassification rate are combined in stages. Based on the spectral spectrum, which is characterized by the classification of polymers with a very low misclassification rate as a total by starting with a classification and gradually classifying a polymer group with a high misclassification rate. How to distinguish polymer groups.

Images