A polymer identification method and electronic device based on spectral decomposition

Abstract

The present application discloses a polymer identification method and electronic device based on spectrum decomposition, which relates to the field of polymer identification technology. The method includes: obtaining a spectrum of a test sample and a spectrum of each known sample; performing peak fitting on the spectrum of the test sample to obtain a test spectrum; performing peak fitting on the spectrum of each known sample to obtain each standard spectrum, and establishing a standard spectrum library based on this; if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is judged to be a single-component polymer that is the same as the corresponding known sample; otherwise, an approximate spectrum with a similarity to the test spectrum greater than or equal to a preset threshold is obtained, the approximate spectrum is linearly combined by multiple standard spectra, and the test sample is judged to be a multi-component polymer that is a mixture of the corresponding known samples of the multiple standard spectra. The present application determines the type of a single-component polymer and the main components and contents of a multi-component polymer without human intervention.

Description

Technical Field

[0001] The present application relates to the technical field of polymer identification, and in particular to a polymer identification method and electronic equipment based on spectral decomposition. Background Art

[0002] Since the 1930s, polymer materials have experienced rapid development due to their superior properties, including high toughness, ease of modification, fatigue resistance, and resistance to corrosion and aging. From rubber to plastics, and from general applications to specialized applications, the types of polymer materials have gradually expanded. Currently, there are nearly 100 types of polymers in practical use, corresponding to tens of thousands of commercially available grades. Consequently, this vast array of polymers presents challenges in polymer identification.

[0003] In related technologies, polymer material identification methods primarily rely on spectroscopic analysis using instruments such as infrared spectroscopy, differential scanning calorimetry, thermogravimetric spectroscopy, and Raman spectroscopy. These instruments then apply a series of criteria and various processing methods to arrive at identification conclusions. However, these conclusions are often based on empirical judgment, i.e., judgments based on the polymer's spectra, combined with its application area and characteristics. This method requires specialized knowledge and is not only highly subjective, but also offers poor qualitative judgment accuracy and efficiency, and is difficult to quantify. Summary of the Invention

[0004] In response to one of the defects in the prior art, the purpose of this application is to provide a polymer identification method and electronic equipment based on spectral decomposition to solve the problem of difficulty in qualitative or quantitative identification of polymer materials in related technologies.

[0005] The first aspect of the present application provides a polymer identification method based on spectral decomposition, which comprises the steps of:

[0006] Obtaining a spectrum of the test sample and spectra of each known sample, wherein the spectra are additive;

[0007] Performing peak fitting on the spectrum of the test sample to obtain a fitted test spectrum;

[0008] The spectra of each known sample are subjected to peak fitting to obtain fitted standard spectra, and a standard spectrum library is established based on these fitted spectra;

[0009] A similarity analysis is performed on the test spectrum and each standard spectrum in the standard spectrum library; if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is judged to be a single-component polymer that is the same as the corresponding known sample; otherwise, an approximate spectrum is obtained whose similarity to the test spectrum is greater than or equal to the preset threshold, and the approximate spectrum is a linear combination of multiple standard spectra, and the test sample is judged to be a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra.

[0010] In some embodiments, obtaining an approximate spectrum having a similarity with the test spectrum greater than or equal to a preset threshold value specifically includes:

[0011] Each standard spectrum in the above standard spectrum library is inner-producted with the test spectrum and transformed into a matrix form;

[0012] Obtain a coefficient matrix formed by the inner products of each pair of standard spectrum in the standard spectrum library, and a constant vector of the inner products of each standard spectrum in the standard spectrum library and the test spectrum;

[0013] Based on the above matrix form, a combination coefficient vector is calculated, where the above combination coefficient vector is the product of the inverse matrix of the above coefficient matrix and the above constant vector;

[0014] An approximate spectrum to be verified is obtained based on the above-mentioned combined coefficient vector, and when the similarity between the above-mentioned test spectrum and the approximate spectrum to be verified is greater than or equal to a preset threshold, the approximate spectrum to be verified is determined to be valid.

[0015] In some embodiments, after determining that the approximate spectrum to be verified is valid, the method further includes:

[0016] The components and proportions of the multi-component polymer are analyzed from the components of the combined coefficient vector.

[0017] In some embodiments, when each standard spectrum in the standard spectrum library is inner-producted with the test spectrum, the test spectrum is represented by a linear combination of the standard spectra in the standard spectrum library.

[0018] In some embodiments, A is the coefficient matrix, is the above combination coefficient vector, is the above constant vector; calculate the combination coefficient vector, specifically including:

[0019] Introducing Tikhonov regularization, the product of the inverse matrix of the above coefficient matrix and the above constant vector is transformed into solving the cost function Get the minimum value At this time, the corresponding solution satisfies the equation: Where I is the identity matrix, λ is the Tikhonov regularization parameter;

[0020] The above equation is solved using the non-negative least squares method NNLS, and the obtained solution is normalized to obtain the above combination coefficient vector.

[0021] In some embodiments, the λ is determined based on the condition number of the coefficient matrix.

[0022] In some embodiments, when performing peak fitting, a peak finding algorithm based on continuous wavelet transform is used, and multiple additional peaks are randomly assigned as redundancy of the peak finding algorithm according to the complexity of the spectrum.

[0023] In some embodiments, the functional forms of the test spectrum and the standard spectrum are both Gaussian functions or Lorentzian functions.

[0024] In some embodiments, when the similarity between the test spectrum and any standard spectrum is less than a preset threshold, and an approximate spectrum with a similarity to the test spectrum not less than the preset threshold cannot be obtained, standard spectra of other known samples are added or the preset threshold is adjusted and re-identification is performed.

[0025] A second aspect of the present application provides an electronic device for polymer identification, comprising a processor and a memory, wherein the processor executes a code in the memory to implement the method.

[0026] The beneficial effects of the technical solution provided by this application include:

[0027] The polymer identification method and electronic device based on spectral decomposition of the present application are characterized by additivity of the spectrum of the test sample and the spectra of each known sample, and the spectrum of the test sample can be fitted by peak fitting to obtain a fitted test spectrum, and the spectrum of each known sample can be fitted by peak fitting to obtain each fitted standard spectrum, and a standard spectrum library is established based on this. When the similarity analysis is performed on the test spectrum and each standard spectrum in the standard spectrum library, if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is judged to be the same single-component polymer as the corresponding known sample, otherwise the similarity is obtained. The test spectrum is an approximate spectrum with a similarity greater than or equal to a preset threshold, and the above-mentioned approximate spectrum is a linear combination of multiple standard spectra, and the test sample is judged to be a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra; therefore, all the information of the spectrum is used to qualitatively identify the composition of the unknown polymer without human intervention, confirm whether it is a single-component pure substance or a multi-component mixture, and determine the type of the single-component polymer and the main components and content of the multi-component polymer, so as to solve the current problems of difficulty in qualitative or quantitative identification of polymer mixtures and slow calculation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0029] Figure 1This is a first flow chart of the polymer identification method according to an embodiment of the present application;

[0030] Figure 2 This is a second flow chart of the polymer identification method according to the embodiment of the present application.

[0031] Figure 3 The infrared spectrum pretreatment results of the test sample in Example 1;

[0032] Figure 4 The infrared spectrum of the sample tested in Example 1 is compared with the original infrared spectrum.

[0033] Figure 5 The infrared spectrum pretreatment results of the test sample in Example 2;

[0034] Figure 6 The coefficient matrix A and the similarity vector s heat map obtained in Example 2;

[0035] Figure 7 The infrared spectrum of the sample tested in Example 2 is compared with the original infrared spectrum.

[0036] Figure 8 The DSC (Differential Scanning Calorimetry) spectrum pretreatment results of the test sample in Example 3;

[0037] Figure 9 The DSC spectrum of the sample tested in Example 3 is compared with the original DSC spectrum.

[0038] Figure 10 The results of the TG (Thermogravimetry) spectrum fitting and the comparison with the original TG spectrum of the test sample in Example 4 are shown;

[0039] Figure 11 This is an example of the interference spectrum in Example 5. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0041] like Figure 1 As shown, the embodiment of the present application provides a polymer identification method based on spectral decomposition, which includes the steps of:

[0042] S1. Obtain a spectrum of the test sample and spectra of each known sample, wherein the spectra are additive.

[0043] In this embodiment, the spectrum of the test sample is used as the unknown spectrum, and the spectra of a series of known samples are used as the standard spectra. In order to meet the requirements of spectrum decomposition, the spectra need to be additive, that is, the spectrum f of the mixture should be equal to the spectra f of its components. i The formula for point-by-point summation is: f = ∑ i f i .

[0044] For example, an infrared spectrum should be in the form of "wave number ν - absorbance A", and a thermogravimetric analysis spectrum should be in the form of "temperature T - first-order derivative of mass change with respect to temperature dm / dT", etc. The additivity of infrared spectra is guaranteed by the additivity of absorbance (Lambert-Beer law), and the additivity of thermogravimetric analysis spectra is guaranteed by the additivity of mass (law of conservation of mass).

[0045] In addition, spectra should undergo preprocessing such as baseline correction, smoothing, and normalization to remove noise and baseline drift. Preferably, baseline correction is applied to infrared spectra and differential scanning calorimetry spectra, while smoothing is applied to differential spectra such as thermogravimetric analysis spectra or spectra with high noise levels. This smoothing can be achieved using the Savitsky-Golay algorithm, and the smoothing window size can be adjusted appropriately based on the actual data.

[0046] S2. Perform peak fitting on the spectrum of the test sample to obtain a fitted test spectrum.

[0047] S3. Perform peak fitting on the spectra of each known sample to obtain fitted standard spectra, and use them to establish a standard spectrum library.

[0048] In this embodiment, the spectrum directly obtained by the instrument is actually a series of discrete data points, and the spacing and number of the points are related to the instrument settings. If the point spacing between different spectra is different, or there are too many data points, the subsequent processing will be very complicated. In order to simplify the spectrum form and speed up subsequent calculations while utilizing all the information of the spectrum, it is necessary to perform peak fitting on the spectrum obtained by the instrument. In principle, the spectrum f(x i ) can be represented by the sum of multiple continuous functions h(x; p)∑ k h k (x; p k ) approximation, where x is the independent variable, i is the data point number, and p k is a continuous function h k Optional parameters.

[0049] Perform peak fitting on the spectrum of the test sample m to obtain the fitted test spectrum fm (x); Perform peak fitting on the spectra of n known samples, and establish a standard spectrum library through the obtained standard spectra

[0050] Peak finding can be achieved through common methods such as manual marking, derivative peak finding, or applied transformation peak finding. Good peak finding results can significantly improve the convergence speed and accuracy of the fitting process. Furthermore, the fitting process can be performed using the multi-peak fitting function of professional software such as Origin Pro, or by programming with open-source packages such as lmfit in Python or limma in R.

[0051] S4. Perform a similarity analysis on the test spectrum and each standard spectrum in the standard spectrum library; if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is determined to be a single-component polymer identical to the known sample corresponding to the standard spectrum. Otherwise, an approximate spectrum with a similarity greater than or equal to the preset threshold to the test spectrum is obtained based on the standard spectrum library. The approximate spectrum is a linear combination of multiple standard spectra, and the test sample is determined to be a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra. At this point, the components and proportions of the multi-component polymer can be resolved. The multiple standard spectra are at least two.

[0052] In the polymer identification method of this embodiment, since the spectrum of the test sample and the spectrum of each known sample are both additivity, and the spectrum of the test sample is subjected to peak fitting to obtain a fitted test spectrum, and the spectrum of each known sample is subjected to peak fitting to obtain fitted standard spectra, and a standard spectrum library is established based on this, and then when the test spectrum and each standard spectrum in the standard spectrum library are subjected to similarity analysis, if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is judged to be the same single-component polymer as the corresponding known sample, otherwise the test spectrum with a similarity greater than or equal to a preset threshold is obtained. An approximate spectrum of a threshold value is obtained, and the above-mentioned approximate spectrum is linearly combined by multiple standard spectra, and the test sample is judged to be a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra; therefore, the method realizes the use of all the information of the spectrum, and qualitatively identifies the composition of the unknown polymer without human intervention, confirms whether it is a single-component pure substance or a multi-component mixture, and determines the type of the single-component polymer and the main components and content of the multi-component polymer, so as to realize qualitative and quantitative analysis of unknown polymer materials, and solve the current problems of difficulty in qualitative or quantitative identification of polymer mixtures and slow calculation results.

[0053] Based on the above embodiment, in this embodiment, when performing peak fitting on the spectra of the above test samples and the spectra of each known sample, a continuous wavelet transform peak finding algorithm is adopted, and multiple additional peaks are randomly specified as redundancy of the peak finding algorithm according to the complexity of the spectrum.

[0054] This continuous wavelet transform enables automatic peak detection across the entire spectrum. It is not only robust against background noise but can also identify and separate overlapping peaks, such as shoulders, with high accuracy. However, during automatic peak detection, some peaks may be missed, resulting in a small number of continuous functions and poor fitting results. Therefore, introducing additional random peaks as redundancy improves the accuracy of the results, although this slightly slows convergence.

[0055] The peak-finding process can be implemented by programming with professional software such as Origin Pro or open-source packages such as signal.find_peaks_cwt in the scipy package in Python.

[0056] Preferably, the functional forms of the above-mentioned test spectrum and standard spectrum are both Gaussian functions or Lorentzian functions.

[0057] In this embodiment, the continuous function used for fitting is uniformly a Gaussian function g(x; H, μ, σ) or a Lorentzian function l(x; H, μ, γ), so as to accelerate the subsequent calculation process based on known analytical integral formulas and reasonable approximations.

[0058] In this embodiment, the Gaussian function g(x; H, μ, σ) is in the form of:

[0059]

[0060] Among them, the optional parameters are H, μ, and σ, which are related to the peak height, peak center position, and peak width, respectively.

[0061] The form of the Lorentz-type function l(x; H, μ, γ) is:

[0062]

[0063] Among them, the optional parameters are H, μ, and γ, which are related to the peak height, peak center position, and peak width, respectively.

[0064] Define the inner product of the function as:

[0065]

[0066] The integral formulas related to the two continuous functions mentioned above are:

[0067]

[0068]

[0069]

[0070]

[0071] In addition, for the inner product of two Gaussian functions, when the factors in the exponential term are When , it can be directly approximated to 0 to further improve the calculation speed.

[0072] In this embodiment, the test spectrum f is defined as m (x) and standard spectrum library The jth standard spectrum in The similarity is The similarity can be obtained by calculating the inner product of the function, and the formula is:

[0073]

[0074] The similarity value ranges from [0, 1], with larger similarity values ​​indicating higher similarity between the two spectra. In this embodiment, a preset threshold of 0.95 is given for single-component similarity. If the similarity between the test spectrum and a standard spectrum reaches this threshold, the unknown test sample is considered to be identical to the known sample corresponding to the standard spectrum, and the identification process for the unknown test sample ends. If the similarity between the test spectrum and all standard spectra fails to reach the preset threshold, the unknown test sample can be assumed to be a mixture of known samples in a certain proportion.

[0075] Based on the above embodiment, in this embodiment, obtaining an approximate spectrum whose similarity to the above test spectrum is greater than or equal to a preset threshold specifically includes the following steps:

[0076] First, each standard spectrum in the standard spectrum library is inner-producted with the test spectrum and transformed into a matrix form.

[0077] Optionally, when each standard spectrum in the standard spectrum library is inner-producted with the test spectrum, the test spectrum f m It is approximately expressed by the linear combination of each standard spectrum in the standard spectrum library, that is:

[0078]

[0079] Among them, f a is the approximate spectrum obtained by linear combination of each standard spectrum, c j is the combination coefficient, which is proportional to the proportion of the known sample in the test sample. a The form of is unknown, so in order to find the combination coefficient c j , the standard spectrum library Each spectrum in f m Inner product, we can get the following n equations:

[0080]

[0081] Its matrix form is:

[0082]

[0083] Secondly, a coefficient matrix formed by the inner products of each pair of standard spectrum in the standard spectrum library and a constant vector of the inner product of each standard spectrum in the standard spectrum library and the test spectrum are obtained.

[0084] Preferably, the above matrix form is simplified as Among them, A is the coefficient matrix composed of the inner products of the two standard spectra, is the combination coefficient vector to be determined, is a constant vector consisting of the inner product of each standard spectrum and the test spectrum, where A and can be calculated directly.

[0085] Then, based on the above matrix form, the combined coefficient vector can be calculated. At this time, the combined coefficient vector is the product of the inverse matrix of the above coefficient matrix and the above constant vector, that is, Obtain.

[0086] Finally, the approximate spectrum to be verified is obtained based on the above-mentioned combined coefficient vector, and when the similarity between the above-mentioned test spectrum and the above-mentioned approximate spectrum to be verified is greater than or equal to a preset threshold, the approximate spectrum to be verified is judged to be a valid approximate spectrum.

[0087] Furthermore, after determining that the approximate spectrum to be verified is valid, the method further includes the following steps: parsing the components and proportions of the multi-component polymer from the components of the combined coefficient vector.

[0088] Among them, the approximate spectrum f to be verified can be obtained by solving the components of the combined coefficient vector a , then the test spectrum f can be calculated m The similarity S(f m ,f a ), the calculation formula is as follows:

[0089]

[0090] Optionally, the preset threshold of the above similarity is 0.95. If the similarity between the test spectrum and the approximate spectrum to be verified reaches the preset threshold, the unknown sample is considered to be a mixture of known samples corresponding to the approximate spectrum, and the composition ratio can be calculated by the combination coefficient vector. At this time, the identification process of the unknown sample is completed.

[0091] In addition, if the similarity between the test spectrum and the approximate spectrum to be verified does not reach the preset threshold, that is, the similarity between the above test spectrum and any standard spectrum is less than the preset threshold, and an approximate spectrum with a similarity to the above test spectrum that is not less than the preset threshold cannot be obtained, it indicates that the difference between the standard spectrum library and the test spectrum is too large, and it is necessary to supplement the standard spectra of other known samples or adjust the preset threshold and re-identify, or other methods can be used to assist in the judgment.

[0092] In this embodiment, generally, each combination coefficient can be obtained by Obtained, thus However, in some cases, there may be two problems: first, when the number of standard spectrum libraries is large, they may contain similar or even repeated spectra. In this case, A is nearly singular or even singular, resulting in inaccurate or impossible inverse matrix solution; second, the combination coefficient is related to the content of each substance. From a physical point of view, the result should be non-negative, but The solution may produce negative results. Therefore, in order to solve the above two problems, Tikhonov regularization can be introduced and the non-negative least squares (NNLS) method can be used for solution.

[0093] Preferably, A is the above coefficient matrix, is the above combination coefficient vector, is the above constant vector; calculating the combination coefficient vector specifically includes the following steps:

[0094] First, Tikhonov regularization is introduced to calculate the product of the inverse matrix of the above coefficient matrix and the above constant vector Transformed into solving the cost function Get the minimum value Right now

[0095]

[0096] At this time, the corresponding solution is the minimum value Satisfies the equation:

[0097]

[0098] Where I is the identity matrix and λ is the Tikhonov regularization parameter. For non-singular A, λ can be zero, which degenerates to For the case where A is singular or nearly singular, the matrix (A TA+λI) singularity, thereby obtaining a reliable least squares solution. In this embodiment, the regularization parameter λ can be reasonably selected according to the condition number cond(A) of the coefficient matrix A.

[0099] Preferably, whether to perform regularization and the size of the parameter λ are determined according to the condition number cond(A) of the coefficient matrix A. When cond(A) < 10 12 When the regularization parameter λ is zero, otherwise adjust λ until cond(A T A+λI)<10 12 .

[0100] Then, the non-negative least squares method NNLS is used to solve the above equation, and the obtained solution is normalized to obtain the above combination coefficient vector.

[0101] Among them, since the combination coefficient vector to be determined is It also needs to have the significance of the proportion of each component, so The components in should be non-negative. To meet this condition, NNLS is used to solve the equation The solution can be achieved by lsqnonneg in MATLAB or optimize.nnls in the scipy package in Python. Afterwards, Perform normalization processing.

[0102] In this embodiment, the combination coefficient vector The size of each component determines whether the corresponding component is set to zero. The jth component c j <10 -2 , that is, when the content of the corresponding component is less than 1%, c j Set it to zero to reduce the impact of numerical calculation errors on result judgment.

[0103] This embodiment can reduce the influence of similar spectra and other interference terms, and ensure that the obtained component content is non-negative and has a clear physical meaning.

[0104] like Figure 2 Specifically, the polymer identification method based on spectrum decomposition of this embodiment includes the following steps:

[0105] A1. Collect spectra, adjust them to an additive format, and perform preprocessing.

[0106] A2. Select an appropriate continuous function and perform peak fitting on the test sample spectrum and the spectra of each known sample to obtain the fitted test spectrum and each standard spectrum;

[0107] A3. Calculate the unknown test spectrum f m and each standard spectrum s j The similarity S(f m ,s j );

[0108] A4. Determine the single component similarity S(f m ,s j ) is greater than a preset threshold; if so, go to A5, otherwise, go to A6.

[0109] A5. The unknown sample is determined to be consistent with the corresponding standard sample, and the sample identification is completed.

[0110] A6. Calculate the coefficient matrix A and the constant vector s, and determine the regularization parameter λ from the condition number of A.

[0111] A7. Construction equation (A T A+λI)c=A T s, use the non-negative least squares method to solve the combined coefficient vector c, set the components below the threshold in the vector c to zero and normalize them to obtain the approximate spectrum f a ;

[0112] A8. Calculate the unknown spectrum f m and approximate spectrum f a The similarity S(f m ,f a );

[0113] A9. Determine the similarity S(f m ,f a ) is greater than a preset threshold, if so, go to A10, otherwise go to A11.

[0114] A10. The unknown sample is determined to be a mixture of known samples corresponding to similar spectra, and the sample identification is completed.

[0115] A11. Determine whether to change the conditions and re-evaluate. If so, go to A1. Otherwise, end, that is, the sample cannot be identified using this method.

[0116] Example 1

[0117] The above-mentioned polymer identification method is used to identify an unknown test sample 1, thereby providing a method for identifying a single type of polymer using infrared spectra.

[0118] The infrared spectrum of an unknown polymer sample 1, along with 82 spectra of potentially related single-type polymer standard samples, is known. The material of the unknown sample needs to be determined. The following infrared spectrum acquisition parameters were used: a PerkinElmer Spectrum 100 Fourier transform infrared spectrometer was used for spectrum acquisition using the included Spectrum software. An InGaAs detector and an integrating sphere diffuse reflectance method were used, with a wavenumber range of 4000–450 cm -1 , scan times 16 times, resolution 4cm -1 .

[0119] (1) Spectrum preprocessing: The infrared spectra of unknown samples and standard samples are converted into the form of "wave number ν-absorbance A", and baseline correction and Savitsky-Golay smoothing are performed. The smoothing window width is 11 data points, and normalization is performed. Taking the unknown sample as an example, the processing results are as follows: Figure 3 shown.

[0120] (2) Spectrum peak fitting: Continuous wavelet transform is used to find peaks in the processed infrared spectrum, and 10 peaks are randomly selected as additional peaks. Gaussian function is used for fitting. The peak finding and fitting process is completed by programming with the help of Python language scipy package and lmfit package. Taking the unknown sample as an example, the fitting result is as follows Figure 4 As shown, the dotted line is the original spectrum, i.e., the infrared spectrum after preprocessing, the thick solid line is the fitting spectrum, i.e., the fitted test spectrum, and the thin solid lines are the peaks of the fitting spectrum.

[0121] (3) Similarity Analysis (Single Component): The threshold for single-component similarity was set at 0.95. The similarity S of the unknown spectrum and the 82 standard spectra was calculated and sorted in descending order. The results are shown in Table 1. As can be seen from the table, the five standard spectra with the highest similarity to the unknown spectrum were all from ABS material, and the similarity S exceeded the threshold of 0.95. At this point, it can be determined that the unknown sample is a single type of polymer, and the material is ABS.

[0122] Finally, the similarity S' calculated using the instrument's accompanying Spectrum software is listed for comparison. Comparing the last two columns of Table 1 reveals similarity results between the spectrum fitting and the software calculation. Both methods confirm that the unknown sample is ABS material, with consistent conclusions. This demonstrates that this identification method can accurately identify a single polymer species.

[0123] Table 1 Example 1 - Similarity calculation results and comparison

[0124]

[0125] Example 2

[0126] The above-mentioned polymer identification method is used to identify the unknown test sample 2, thereby providing a method for identifying multiple types of polymers using infrared spectra.

[0127] The infrared spectrum of an unknown polymer sample 2, along with 130 spectra of potentially related single-type polymer standard samples, is known. The material of the unknown sample needs to be determined. The following infrared spectrum acquisition parameters were used: a PerkinElmer Spectrum 100 Fourier transform infrared spectrometer was used for spectrum acquisition using the included Spectrum software. An InGaAs detector and an integrating sphere diffuse reflectance method were used, with a wavenumber range of 4000–450 cm -1 , scan times 16 times, resolution 4cm -1 .

[0128] (1) Spectrum preprocessing: The infrared spectra of unknown samples and standard samples are converted into the form of "wave number ν-absorbance A", and baseline correction and Savitsky-Golay smoothing are performed. The smoothing window width is 11 data points, and normalization is performed. Taking the unknown sample as an example, the processing results are as follows: Figure 5 shown.

[0129] (2) Spectral peak fitting: Continuous wavelet transform was used to find peaks in the processed infrared spectrum, and 10 additional peaks were randomly assigned. Gaussian functions were used for fitting. The peak finding and fitting process was completed by a self-written program using the scipy and lmfit packages in Python.

[0130] (3) Similarity analysis: The threshold for single-component similarity was set at 0.95. The similarities S of the unknown spectra and the 130 standard spectra were calculated and sorted in descending order. The results are shown in Table 2. As can be seen from the table, the similarities S of all unknown spectra with the standard spectra were less than the set threshold of 0.95, indicating that the unknown sample is not a single type of polymer but may be composed of multiple polymers.

[0131] Table 2 Example 2 - Similarity calculation results

[0132] Standard spectrum number Standard spectrum corresponding to material type Spectral similarity S IR_std_m39 PA66 0.8619 IR_std_m116 ABS 0.8600 IR_std_m82 PA6 0.8582 IR_std_m7 ABS 0.8393 IR_std_m123 ABS 0.8228 … … … IR_std_m3 PC 0.3353 IR_std_m110 PVC 0.2833 IR_std_m37 TPV 0.2628

[0133] (4) Composition analysis (multi-component): The threshold of multi-component similarity is set to 0.95. The coefficient matrix A is obtained by calculating the inner product of each of the 130 standard spectra, and the constant vector is obtained by calculating the inner product of each of the 130 standard spectra with the unknown spectra. A and The heat map results are shown in Figure 6 After obtaining the coefficient matrix A, we calculate its condition number cond(A) = 7.622×10 12Since cond(A) is greater than the threshold 10 12 , the Tikhonov regularization parameter λ=1 can be selected, and the condition number cond(A T A + λI) = 1.468 × 10 9 Below the threshold, the solution requirement is met.

[0134] Constructing equations Solving for the coefficient vector via NNLS This process is completed by writing a program by yourself using the optimize.nnls package in Python language scipy. The vector solved by NNLS is There are 130 components, 124 of which are zero, and the remaining 6 components are arranged in order of size, of which only 2 components reach 10 -2 threshold, set the remaining 4 components to zero and After normalization, the results are shown in Table 3.

[0135] Table 3 Example 2 - Results of NNLS solution

[0136]

[0137] (5) Result verification and processing (multi-component): And the standard spectrum IR_std_m82 and IR_std_m123 to obtain the approximate spectrum f a , the approximate spectrum f a 、compose f a The component spectra of each component and the original unknown spectrum f m The results are shown in Figure 7 .Depend on Figure 7 Visible approximate spectrum f a and the original unknown spectrum f m Similar, further calculation is performed to obtain the similarity S(f m ,f a ) = 0.9701, which is greater than the multi-component similarity threshold of 0.95, confirming that the unknown sample is PA6 (52.7%) + ABS (47.3%). The actual polymer composition of this unknown sample is PA6 (50%) + ABS (48%) + SMA (2%), with PA6 and ABS as the primary components and SMA as an additional compatibilizer, a minor component. Comparing the determination results with the actual values ​​shows that this method can determine the primary components of a polymer mixture from infrared spectra, with a content difference of less than 3%, enabling qualitative and quantitative determination of polymer mixtures.

[0138] Finally, direct Inverse Table 4 lists in order the corresponding standard spectra of some of the calculated spectra by this method. Component. The components contain negative values ​​and lack a clear physical meaning. Each component is nonzero and of comparable magnitude, making it impossible to determine the main components of the mixture. Furthermore, constructing PLS models or neural networks requires pre-training, making them difficult to apply when the polymer species are unknown. The results of the P-matrix method are similar to the direct inversion problem described above, and neither method can determine the composition of the polymer mixture.

[0139] Table 4 Example 2 - Results of Direct Solution

[0140]

[0141] Example 3

[0142] The above polymer identification method is used to identify the unknown test sample 3, thereby providing a method for identifying multiple types of polymers using differential scanning calorimetry (DSC) spectra.

[0143] Given the DSC spectrum of an unknown polymer sample (3) and 23 spectra of potentially related single-species polymer standards, the material of the unknown sample needs to be determined. The DSC spectra were acquired using the following parameters: a NETZSCH Maia DSC200 F3 instrument, DSC spectra acquired and preprocessed using the instrument's accompanying data recording and processing software, a perforated Al crucible for sample storage, a scan range of 40–300°C, and a heating rate of 10 K / min. During data processing, the initial heat flow instability region of the scan range was removed to improve fitting accuracy.

[0144] (1) Spectrum preprocessing: Convert the DSC spectra of unknown samples and standard samples into the form of "temperature T-unit heat flow power Δ" and perform baseline correction. Taking the unknown sample as an example, the processing results are as follows Figure 8 shown.

[0145] (2) Spectrum peak fitting: Peaks were found in the processed DSC spectrum using the method of solving high-order derivatives, and two additional peaks were randomly assigned. Gaussian functions were used for fitting. The peak finding and fitting process was completed with the professional software Origin Pro.

[0146] (3) Similarity Analysis: The threshold for single-component similarity was set at 0.95. The similarities S of the unknown sample and the 20 standard spectra were calculated and sorted in descending order. The results are shown in Table 5. As can be seen from the table, the spectrum of the unknown sample and the homopolymer POM is very similar, but all similarities S are less than the set threshold of 0.95, indicating that the unknown sample is not a single type of polymer, but may be composed of multiple polymers.

[0147] Table 5 Example 3-Similarity calculation results

[0148] Standard spectrum number Standard spectrum corresponding to material type Spectral similarity S DSC_std_m3 Homopolymer POM 0.9382 DSC_std_m5 Homopolymer POM 0.9307 DSC_std_m14 Homopolymer POM 0.9294 DSC_std_m1 Copolymer POM 0.5991 DSC_std_m18 Copolymer POM 0.5450 … … … DSC_std_m6 PA6 0.0104 DSC_std_m23 PA66 0.0062 DSC_std_m21 PA66 0.0001

[0149] (4) Composition analysis (multi-component): Set the threshold of multi-component similarity to 0.95, vector The component threshold is 0.01. The coefficient matrix A is obtained by calculating the inner product of each of the 20 standard spectra, and the constant vector is obtained by calculating the inner product of each of the 23 standard spectra with the unknown spectra. After obtaining the coefficient matrix A, its condition number cond(A) = 3.390 × 10 6 , less than the threshold of 10 12 , we can directly use the equation Perform NNLS solution. This solution is completed by writing a program with the help of professional software MATLAB. Vector Among the 23 components, only 2 are non-zero, and their values ​​are all greater than the component threshold of 0.01. They are directly arranged by component size and normalized. The results are shown in Table 6.

[0150] Table 6 Example 3 - Results of NNLS solution

[0151]

[0152] (5) Result verification and processing (multi-component): Calculate the approximate spectrum f by combining the standard spectra DSC_std_m3 and DSC_std_m1 a , the approximate spectrum f a , the component spectra of fa and the original unknown spectrum f m The results are shown in Figure 9 . Further calculations yield the approximate spectrum f a and the original unknown spectrum f m The similarity S(f m ,f a ) = 0.9782, which is greater than the multi-component similarity threshold of 0.95. The unknown sample can be identified as homopolymer POM (90.6%) + copolymer POM (9.4%). The actual polymer composition of this unknown sample is homopolymer POM (90%) + copolymer POM (10%). Comparing the determination result with the actual value shows that this method can determine the main components in a polymer mixture through DSC spectra, and the content difference is less than 1%, which is capable of qualitative and quantitative determination of polymer mixtures.

[0153] Example 4

[0154] The above polymer identification method is used to identify the unknown test sample 4, thereby providing a method for identifying a single type of polymer using thermogravimetric (TG) spectra.

[0155] The TG spectrum of an unknown polymer sample, along with 20 spectra of potentially related single-species polymer standards, is known. The material of the unknown sample needs to be determined. The TG spectrum acquisition parameters are as follows: the instrument is a NETZSCH TarsusTG209 F3, TG spectrum acquisition and preprocessing are performed using the instrument's accompanying data recording and processing software, the sample is held in an aluminum oxide crucible, the scanning range is 40-600°C, and the heating rate is 10K / min.

[0156] (1) Spectrum preprocessing: The TG spectra of unknown samples and standard samples were converted into the form of “temperature T—first-order derivative of mass change with respect to temperature dm / dT”, and baseline correction and Savitsky-Golay smoothing were performed with a smoothing window width of 31 data points.

[0157] (2) Spectrum peak fitting: The processed TG spectrum is peak-finding using continuous wavelet transform, and 5 peaks are randomly assigned as additional peaks. Gaussian functions are uniformly used for fitting. The peak-finding and fitting processes are completed by programming with the help of Python's scipy package and lmfit package. Taking the unknown sample as an example, the fitting results are as follows: Figure 10 As shown, the dotted line is the original spectrum, the thick solid line is the fitting spectrum, and the thin solid lines are the peaks of the fitting spectrum.

[0158] (3) Similarity Analysis (Single Component): The threshold for single-component similarity was set to 0.95. The similarity S of the unknown spectrum and the 20 standard spectra was calculated and sorted in descending order. The results are shown in Table 7. As can be seen from the table, the standard spectrum with the highest similarity to the unknown spectrum is from TPO material, and the similarity S exceeds the threshold of 0.95. By comparing the unknown sample and the standard spectrum TG_std_s13, it can be determined that the unknown sample is a single type of polymer, made of TPO.

[0159] Table 7 Example 4 - Similarity calculation results

[0160] Standard spectrum number Standard spectrum corresponding to material type Spectral similarity S TG_std_s13 TPO 0.9786 TG_std_s4 TPE-S 0.7172 TG_std_s5 TPE-S 0.6990 TG_std_s8 TPV 0.5038 … … … TG_std_s12 PVC 0.2203 TG_std_s10 PVC 0.1911

[0161] Example 5

[0162] The purpose of this example is to illustrate that this method has a strong anti-interference ability and can still perform identification even when there are many irrelevant spectra.

[0163] On the basis of Example 2, 20 randomly generated interference spectra were added to the standard spectrum library. These interference spectra were composed of several Gaussian functions superimposed. The number of functions and the height, center and standard deviation of each Gaussian function were randomly generated. One of the interference spectra was as follows: Figure 11 shown.

[0164] After adding the random spectra, the same process as in Example 2 was followed. The final results are shown in Table 8. The types of the corresponding components did not change, and the relative contents were only slightly different, indicating that this method has good anti-interference ability.

[0165] Table 8 Example 5 - Comparison of results using NNLS before and after adding interference spectra

[0166]

[0167] The polymer identification method of this embodiment preprocesses the spectrum and performs peak fitting, calculates the spectrum similarity by function inner product, and decomposes the spectrum of the mixture sample through regularization method and non-negative least squares method for multi-component mixtures, so as to qualitatively and quantitatively judge the composition of the mixture. Not only does it use all the information of the spectrum, but it also uses the spectrum peak fitting to unify the data format and accelerate the calculation. The judgment process does not require human intervention, has high accuracy and reproducibility, and can handle both single-component pure substances and multi-component mixtures. It is insensitive to irrelevant spectra or peaks, that is, it has relatively low quality requirements for the spectrum.

[0168] An embodiment of the present application further provides an electronic device for polymer identification, which includes a processor and a memory. The processor executes the code in the memory to implement the polymer identification method.

[0169] Specifically, the processor executes the code in the memory to implement the following polymer identification method:

[0170] Obtaining a spectrum of the test sample and spectra of each known sample, wherein the spectra are additive;

[0171] Performing peak fitting on the spectrum of the test sample to obtain a fitted test spectrum;

[0172] The spectra of each known sample are subjected to peak fitting to obtain fitted standard spectra, and a standard spectrum library is established based on these fitted spectra;

[0173] A similarity analysis is performed on the test spectrum and each standard spectrum in the standard spectrum library; if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, the test sample is judged to be a single-component polymer that is the same as the corresponding known sample; otherwise, an approximate spectrum is obtained whose similarity to the test spectrum is greater than or equal to the preset threshold, and the approximate spectrum is a linear combination of multiple standard spectra, and the test sample is judged to be a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra.

[0174] Preferably, the processor executing the code in the memory can also implement other steps in the aforementioned macromolecular identification method.

[0175] In the description of this application, it should be noted that the terms "upper" and "lower" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be internal communication between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances.

[0176] It should be noted that, in this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0177] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand and implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.

Claims

1. A polymer identification method based on spectral decomposition, characterized in that: It includes the steps of: Obtaining a spectrum of the test sample and spectra of each known sample, wherein the spectra are additive; Performing peak fitting on the spectrum of the test sample to obtain a fitted test spectrum; The spectra of each known sample are subjected to peak fitting to obtain fitted standard spectra, and a standard spectrum library is established based on these fitted spectra; Performing a similarity analysis on the test spectrum and each standard spectrum in the standard spectrum library; if the similarity between the test spectrum and any standard spectrum is greater than or equal to a preset threshold, determining that the test sample is a single-component polymer identical to the corresponding known sample; otherwise, obtaining an approximate spectrum having a similarity to the test spectrum greater than or equal to the preset threshold, the approximate spectrum being a linear combination of multiple standard spectra, and determining that the test sample is a multi-component polymer mixed with the known samples corresponding to the multiple standard spectra; Obtaining an approximate spectrum whose similarity to the test spectrum is greater than or equal to a preset threshold, specifically comprising: Each standard spectrum in the standard spectrum library is inner-producted with the test spectrum, and the inner product is converted into a matrix form; Obtain a coefficient matrix formed by the inner products of each pair of standard spectrum in the standard spectrum library, and a constant vector of the inner products of each standard spectrum in the standard spectrum library and the test spectrum; Based on the matrix form, calculating a combined coefficient vector, the combined coefficient vector being the product of the inverse matrix of the coefficient matrix and the constant vector; An approximate spectrum to be verified is obtained based on the combination coefficient vector, and when the similarity between the test spectrum and the approximate spectrum to be verified is greater than or equal to a preset threshold, the approximate spectrum to be verified is determined to be valid.

2. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: After determining that the approximate spectrum to be verified is valid, the following steps are also included: The components and proportions of the multi-component polymer are analyzed from the components of the combination coefficient vector.

3. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: When each standard spectrum in the standard spectrum library is inner-producted with the test spectrum, the test spectrum is represented by a linear combination of the standard spectra in the standard spectrum library.

4. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: Let A be the coefficient matrix, is the combination coefficient vector, is the constant vector; calculating the combination coefficient vector, specifically including: Tikhonov regularization is introduced to convert the product of the inverse matrix of the coefficient matrix and the constant vector into a solution to make the cost function cost Get the minimum value , at this time, the corresponding solution satisfies the equation: ;in, is the identity matrix, λ is the Tikhonov regularization parameter; The equation is solved using the non-negative least squares method NNLS, and the obtained solution is normalized to obtain the combined coefficient vector.

5. The polymer identification method based on spectral decomposition according to claim 4, characterized in that: described λ The determination is made based on the condition number of the coefficient matrix.

6. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: When performing peak fitting, a peak-finding algorithm based on continuous wavelet transform is used, and multiple extra peaks are randomly assigned according to the complexity of the spectrum as redundancy of the peak-finding algorithm.

7. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: The functional forms of the test spectrum and the standard spectrum both adopt Gaussian functions or Lorentzian functions.

8. The polymer identification method based on spectral decomposition according to claim 1, characterized in that: When the similarity between the test spectrum and any standard spectrum is less than a preset threshold, and an approximate spectrum with a similarity to the test spectrum not less than the preset threshold cannot be obtained, standard spectra of other known samples are supplemented or the preset threshold is adjusted before re-identification.

9. An electronic device for polymer identification, characterized in that: The method comprises a processor and a memory, wherein the processor executes the code in the memory to implement the method according to any one of claims 1 to 8.

Images