Terahertz absorption spectrum-based insulating paperboard manufacturer traceability method and system

Abstract

The invention discloses an insulating paperboard manufacturer traceability method and system based on a terahertz absorption spectrum, and relates to the technical field of traceability, and the method comprises the steps: obtaining a sample insulating paperboard, and carrying out the spectrum test through a transmission-type terahertz time-domain spectrum test platform, and obtaining a spectrum signal; performing Fourier transform on each sample, and calculating a terahertz absorption coefficient to obtain an absorption spectrum; carrying out downsampling processing on the spectrum, and carrying out principal component analysis; taking N principal component features with the highest content as input, taking manufacturers as labels, and training a random forest classifier; and performing manufacturer traceability on the to-be-tested insulating paperboard by using the trained classifier. According to the method, the newly produced insulating cardboard can be detected before network access, and the insulating cardboard manufacturer is traced, so that the insulating cardboard is ensured to meet the manufacturer standard, and the phenomenon that the manufacturer spends the best is prevented.

Description

Technical Field

[0001] The present application relates to the field of traceability technology, and in particular to a method and system for tracing the manufacturer of insulating paperboard based on terahertz absorption spectroscopy. Background Art

[0002] As a commonly used insulating material, insulating cardboard is widely used in key power equipment such as transformers and bushings. Its quality directly determines the safety of the equipment and the stability of the power grid. However, there are many suppliers of insulating cardboard at present, and the quality of products from different manufacturers varies. It is still common for manufacturers to arbitrarily replace suppliers and pass off inferior products as good ones. The traditional method of tracing the source of insulating cardboard manufacturers is mainly to test whether the conductivity, degree of polymerization and other properties of the insulating cardboard meet the manufacturer's standards. However, the above methods are all destructive tests, which cause irreversible damage to the samples and accelerate the aging of the samples. At the same time, ultrasonic testing requires the coating of a coupling agent on the surface of the sample, which has high requirements on the environment and operators. Therefore, it is urgent to propose a new type of non-destructive testing technology to achieve rapid and non-destructive traceability of insulating cardboard manufacturers. Summary of the invention

[0003] The purpose of this application is to provide a method and system for tracing the manufacturer of insulation paperboard based on terahertz absorption spectroscopy, which can realize rapid and non-destructive tracing of new insulation paperboard manufacturers.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy, comprising:

[0006] Obtain several sample insulation paperboards;

[0007] Based on the transmission terahertz time-domain spectroscopy testing platform, a spectral test is performed on each sample insulation paperboard to obtain a terahertz time-domain spectral signal of each sample insulation paperboard;

[0008] For each sample insulation paperboard, Fourier transform is performed on the terahertz time-domain spectrum signal of the sample insulation paperboard to obtain the terahertz spectrum frequency domain form, and the terahertz absorption coefficient of the sample insulation paperboard is calculated based on the terahertz spectrum frequency domain form to obtain the terahertz absorption spectrum of the sample insulation paperboard;

[0009] Downsampling the terahertz absorption spectrum of each sample insulation paperboard to obtain downsampling data of each sample insulation paperboard;

[0010] Perform principal component analysis on the collected data to obtain the principal component analysis results of each sample insulation paperboard;

[0011] The N principal component features with the highest content in the principal component analysis results are used as the input feature quantity of the random forest classifier, and the insulation cardboard manufacturer is used as the label to train the insulation cardboard random forest classifier to obtain the trained insulation cardboard random forest classifier;

[0012] Based on the trained insulation cardboard random forest classifier, the insulation cardboard manufacturer of the insulation cardboard to be tested is traced.

[0013] In a second aspect, the present invention provides an insulation paperboard manufacturer traceability system based on terahertz absorption spectroscopy, comprising:

[0014] A sample acquisition module, used to acquire a number of sample insulating paperboards;

[0015] A testing module, used to perform a spectral test on each of the sample insulation paperboards based on a transmission terahertz time-domain spectroscopy testing platform to obtain a terahertz time-domain spectral signal of each sample insulation paperboard;

[0016] A calculation module is used to perform Fourier transform on the terahertz time-domain spectrum signal of each sample insulation paperboard to obtain a terahertz spectrum frequency-domain form, and calculate the terahertz absorption coefficient of the sample insulation paperboard based on the terahertz spectrum frequency-domain form to obtain a terahertz absorption spectrum of the sample insulation paperboard;

[0017] A downsampling module is used to downsample the terahertz absorption spectrum of each sample insulation paperboard to obtain downsampled data of each sample insulation paperboard;

[0018] An analysis module is used to perform principal component analysis on the collected data to obtain the principal component analysis results of each sample insulation paperboard;

[0019] A training module is used to use the N principal component features with the highest content in the principal component analysis results as input feature quantities of the random forest classifier, and use the insulation cardboard manufacturer as a label to train the insulation cardboard random forest classifier to obtain a trained insulation cardboard random forest classifier;

[0020] The traceability module is used to trace the insulation cardboard manufacturer of the insulation cardboard to be tested based on the trained insulation cardboard random forest classifier.

[0021] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0022] The present application provides a method and system for tracing the manufacturer of insulating cardboard based on terahertz absorption spectroscopy. The method includes: first, the system collects multiple insulating cardboard samples as analysis objects. Subsequently, a transmission terahertz time-domain spectroscopy test platform is used to perform spectral testing on each sample to obtain its terahertz time-domain spectral signal. Then, for each sample insulating cardboard, Fourier transform technology is used to convert its terahertz time-domain spectral signal into a terahertz spectral frequency domain form, and the terahertz absorption coefficient of the sample is calculated based on this, thereby obtaining its terahertz absorption spectrum. In order to reduce the data dimension and extract key information, the system performs downsampling processing on the terahertz absorption spectrum of each sample to generate downsampled data. Afterwards, the principal component analysis (PCA) method is used to perform an in-depth analysis of the downsampled data to extract the principal component features of each sample. On this basis, the N principal component features with the highest contribution in the principal component analysis results are selected as the input features of the random forest classifier. Subsequently, the random forest classifier is trained with the manufacturer of the insulating cardboard as a label until it achieves satisfactory classification performance, thereby obtaining a trained random forest classifier for insulating cardboard. Finally, the trained classifier is used to perform traceability analysis on the insulation paperboard to be tested from an unknown manufacturer to determine its manufacturer information. This application uses terahertz and machine learning emerging technologies to achieve rapid and non-destructive traceability of insulation paperboard manufacturers. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 paying creative work.

[0024] Figure 1 This is a flow chart of a method for tracing the source of an insulation paperboard manufacturer based on terahertz absorption spectroscopy in one embodiment of the present application;

[0025] Figure 2 A flowchart of a method for tracing the source of an insulating paperboard manufacturer provided in one embodiment of the present application;

[0026] Figure 3 A system block diagram of a transmission terahertz time-domain spectroscopy testing platform provided in one embodiment of the present application;

[0027] Figure 4 Terahertz absorption spectra of insulating paperboards from different manufacturers provided in one embodiment of the present application;

[0028] Figure 5 The traceability results of 8 types of insulation paperboard manufacturers are provided in one embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] Terahertz waves are electromagnetic waves between microwaves and infrared, with a frequency range of 0.1 to 10 THz. Terahertz waves can pass through many non-polar substances and dielectric materials, and the energy of terahertz photons is low, and the pulse width is at the picosecond level, which can achieve rapid and non-destructive testing of samples. In addition, the rotation and vibration of molecules are also within the terahertz band. Therefore, terahertz is widely used in the field of material identification and recognition.

[0031] The purpose of this application is to provide a method and system for tracing the manufacturer of insulation paperboard based on terahertz absorption spectroscopy, which can realize rapid and non-destructive tracing of the manufacturer of insulation paperboard.

[0032] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0033] Embodiment 1

[0034] like Figure 1 As shown, this embodiment provides a method for tracing the manufacturer of insulating paperboard based on terahertz absorption spectroscopy, including:

[0035] Step 101: Obtain several sample insulating paperboards;

[0036] Step 102: Based on the transmission terahertz time-domain spectroscopy testing platform, perform spectral testing on each of the sample insulation paperboards to obtain a terahertz time-domain spectral signal of each sample insulation paperboard;

[0037] Step 103: for each sample insulation paperboard, perform Fourier transform on the terahertz time-domain spectrum signal of the sample insulation paperboard to obtain a terahertz spectrum frequency-domain form, and calculate the terahertz absorption coefficient of the sample insulation paperboard based on the terahertz spectrum frequency-domain form to obtain a terahertz absorption spectrum of the sample insulation paperboard;

[0038] Step 104: downsampling the terahertz absorption spectrum of each sample insulation paperboard to obtain downsampling data of each sample insulation paperboard;

[0039] Step 105: performing principal component analysis on the collected data to obtain principal component analysis results of each sample insulation paperboard;

[0040] Step 106: using the N principal component features with the highest content in the principal component analysis results as input feature quantities of the random forest classifier, using the insulation cardboard manufacturer as a label, training the insulation cardboard random forest classifier, and obtaining a trained insulation cardboard random forest classifier;

[0041] Step 107: Based on the trained insulation cardboard random forest classifier, the insulation cardboard manufacturer of the insulation cardboard to be tested is traced.

[0042] In some embodiments, before executing step 101, the method further includes:

[0043] Prepare sample insulation paperboard, including:

[0044] Collect insulation cardboard from several insulation cardboard manufacturers;

[0045] Cutting the insulating paperboard into a set size to obtain a cut insulating paperboard sample;

[0046] The cut sample insulation paperboards are placed in a vacuum drying oven for drying to obtain the prepared sample insulation paperboards.

[0047] Specifically, in terms of material selection, this embodiment selected insulation paper from eight manufacturers, including ABB, Dongfang, Guangxin, Weidmann, Xingrui, Xinyuan, AUO and Yingzhong, for testing. First, the insulation paper was cut into 100mm×100mm samples with a sample thickness of 1mm. The cut insulation paperboard samples were placed in a vacuum drying oven and dried for 48 hours to remove the influence of moisture on the terahertz test results.

[0048] Wherein, in some embodiments, when executing step 102, the transmission terahertz time-domain spectroscopy test platform includes: a femtosecond laser, a beam splitter, a time delay line, a reflector, a THz wave transmitter, a THz wave detector, a parabolic mirror, a phase-locked amplifier and a PC control end; the femtosecond laser is used to generate ultrashort pulse femtosecond laser; the beam splitter is used to divide the femtosecond laser into two beams, one of which is used to excite terahertz waves through the THz wave transmitter, and the other is used to pass through the time delay line, the reflector and the THz wave detector to detect the terahertz waves after passing through the sample insulating cardboard; the parabolic mirror is used to collimate and focus the terahertz waves so that the terahertz waves are concentrated on the sample insulating cardboard; the PC control end is connected to the THz wave detector signal to record the time domain signal of the terahertz pulse.

[0049] Specifically, the main components of the transmission terahertz time-domain spectroscopy test platform include: femtosecond laser, beam splitter, time delay line, reflector, THz wave transmitter, THz wave detector, parabolic mirror, phase-locked amplifier, and PC control terminal. The system connection relationship block diagram is as follows Figure 3 The main functions of each part are as follows:

[0050] 1) Femtosecond laser: a femtosecond laser that produces ultrashort pulses (the pulse width is usually between tens and hundreds of femtoseconds).

[0051] 2) Beam splitter: splits the femtosecond laser into two beams, one for exciting terahertz waves (pump light) and the other for detection (probe light).

[0052] 3) Pump light: transmitted to the photoconductive antenna, passing through the terahertz emitter, and used to excite terahertz waves.

[0053] 4) Detection light: transmitted to the time delay line, passes through the reflector, and then reaches the terahertz detector.

[0054] 5) Time delay line: adjust the optical path of the detection light, thereby controlling the relative time delay between the detection light and the terahertz wave in the detector to achieve time-resolved measurement.

[0055] 6) Parabolic mirror: collimates and focuses the terahertz wave so that it can be concentrated and effectively irradiated on the sample or transmitted to the detector.

[0056] 7) THz wave generator: used to emit terahertz waves.

[0057] 8) THz wave detector: detects the terahertz wave after passing through the sample, and converts the terahertz signal into an easily measurable electrical signal through the electro-optic effect or photoconductivity effect.

[0058] 9) PC control terminal: data acquisition and processing, recording the time domain signal of terahertz pulse.

[0059] Specifically, the femtosecond laser generates two mutually perpendicular femtosecond pulse lasers under the action of the beam splitter, and the laser beam is divided into pump light and probe light by the beam splitter. The terahertz pulse generated by the pump light of the terahertz generator is reflected by the mirror and penetrates the sample to be tested. After the probe light passes through the time delay line, it is shot onto the terahertz detector together with the terahertz pulse that has penetrated the sample and reflected by the mirror. The two detected electrical signals are converted from digital to analog and transmitted to the computer in digital form, and the internal information of the test sample is obtained through computer data analysis and processing.

[0060] In some embodiments, when executing step 103, the specific steps may be as follows:

[0061] For each sample insulation paperboard, the terahertz time domain spectrum signal of the sample insulation paperboard is Fourier transformed to obtain the terahertz spectrum frequency domain form, and the terahertz absorption coefficient of the sample insulation paperboard is calculated based on the terahertz spectrum frequency domain form to obtain the terahertz absorption spectrum of the sample insulation paperboard, specifically including:

[0062] By performing Fourier transform on the time domain signal obtained by the terahertz time domain spectrometer, the measured terahertz time domain electric field can be converted into frequency domain form. The calculation formula is as follows:

[0063]

[0064] Among them, ω is the angular frequency, E(ω) is the terahertz frequency domain electric field, E(t) is the terahertz time domain electric field, A(ω) is the frequency domain amplitude, and φ(ω) is the frequency domain phase.

[0065] According to the frequency domain form of the terahertz spectrum, the sample absorption coefficient can be calculated using the following formula:

[0066]

[0067] In the formula, c is the speed of light in vacuum, k(ω) is the extinction coefficient, and the calculation formula is:

[0068]

[0069] Where d is the thickness of the insulating oil sample, A s (ω) is the frequency domain amplitude information of the frequency domain signal of the sample, A r (ω) is the frequency domain amplitude information of the reference frequency domain signal. n(ω) is the refractive index, and the calculation formula is:

[0070]

[0071] In the formula, φ s (ω) is the frequency domain phase information of the frequency domain signal of the sample, φ r (ω) is the frequency domain phase information of the reference frequency domain signal.

[0072] In some embodiments, when executing steps 104-106, the specific steps may be as follows:

[0073] like Figure 2 As shown in the figure, the terahertz test was carried out on 8 kinds of insulation paperboards from different manufacturers using the transmission terahertz time-domain spectroscopy test platform. 30 sets of data were tested on samples from each manufacturer, totaling 240 sets of data. The terahertz time-domain spectrum was Fourier transformed using formula (1) to obtain the terahertz frequency-domain spectrum (the frequency-domain form of the terahertz spectrum). The terahertz absorption coefficient of the sample was calculated using formulas (2) to (4) to obtain the terahertz absorption spectrum. Since the dynamic range of the sample terahertz frequency-domain signal amplitude is greatly reduced after 3THz, it is difficult to distinguish effective information, so the absorption spectrum in the 0.1THz-3THz frequency band was selected as the research object.

[0074] Among them, the terahertz absorption spectra of insulating papers from different manufacturers are downsampled, and the downsampling process is as follows:

[0075] Taking the data in this embodiment as an example, the absorption spectrum is composed of 164 frequency points in total, one frequency point is selected every 3 frequency points, and a total of 41 frequency points are selected as terahertz absorption spectrum downsampling data.

[0076] In order to further reduce the dimension of the terahertz absorption spectrum, principal component analysis (PCA) is performed on the collected data. Principal component analysis (PCA) is a linear statistical method used for dimensionality reduction and data analysis. It linearly combines the original variables into new, uncorrelated principal component variables, which can maximize the retention of information in the original data.

[0077] Taking this embodiment as an example, after downsampling, there are 8*30=240 groups of spectral data, and each group of data has p=41 frequency points, that is, 41 eigenvalues.

[0078] The principal component analysis of the collected data was performed to obtain the principal component analysis results of each sample insulation paperboard, which can be specifically as follows:

[0079] Get the collected data;

[0080] Performing standardization processing on the collected data to obtain a standardized matrix of the collected data;

[0081] Calculating a standardized sample covariance matrix according to the standardized matrix of the collected data;

[0082] According to the eigenvalues ​​and eigenvectors of the standardized matrix, the principal component contribution rate and the cumulative contribution rate are calculated.

[0083] Specifically, firstly, the collected data is standardized and the mean is calculated by column. Standard Deviation: Normalize data: The original sample is standardized and the standardized matrix is ​​obtained as follows:

[0084]

[0085] Where n is the number of groups of data collected, n = 240, p is the frequency point of each group of data collected, p = 41, X np The pth frequency point of the nth group of collected data.

[0086] Then calculate the standardized sample covariance matrix R, the calculation formula is as follows:

[0087]

[0088] Finally, the standardized sample covariance matrix R is obtained:

[0089]

[0090] Then the standardized matrix eigenvalue and eigenvector are calculated, and the eigenvalue λ and the eigenvector a satisfy the following relationship: Ra=λa.

[0091] Eigenvalue: λ1≥λ2≥…≥λ p ≥0.

[0092] The eigenvectors are:

[0093]

[0094] Calculate the principal component contribution rate a and the cumulative contribution rate b. The calculation formula is as follows:

[0095]

[0096] In the formula, a is the principal component contribution rate, b is the cumulative contribution rate, k is the specific principal component number, i is the first i principal components selected, and λ k is the kth principal component, and P is the total dimension of each set of data.

[0097] The first three principal components corresponding to the eigenvalues ​​with cumulative contribution rates exceeding 80% are taken as the input feature quantities of the random forest algorithm. The expression of the i-th principal component is as follows:

[0098] F i =a 1i X1+a 2i X2+…+a pi X p (i=1,2,3) (10).

[0099] The top three eigenvectors in the PCA calculation results were extracted as input features of the random forest classifier to classify the eight insulation paper manufacturers.

[0100] Among the 240 groups of sample data, 70% were selected for training, that is, to build a random forest classifier for insulating cardboard, and the remaining 30% of the data were selected for verification, that is, to trace the manufacturer of insulating cardboard.

[0101] The random forest classifier construction process is as follows:

[0102] First, the three principal component features and manufacturer labels obtained by PCA are input into the random forest algorithm, and the input data set (principal component features and manufacturer labels) is randomly sampled with replacement, and a subset is generated each time. Each subset contains 70% of the original number of samples, and the remaining 30% of the samples are used to verify the model performance.

[0103] Next, a decision tree is constructed. For each sample subset, the random forest generates a decision tree. Unlike ordinary decision trees, when each node is divided, the random forest will randomly select a part of the features (rather than using all the features) to divide the node. This can increase the diversity of the model and avoid the overfitting problem of the decision tree. Each decision tree is generated based on the different divisions of the three PCA eigenvalues ​​(principal components). Each node will select a feature (or a combination of multiple features) for data division. The goal is to maximize the purity of the data set, that is, the samples in each node belong to the same manufacturer as much as possible.

[0104] Each tree will eventually generate multiple leaf nodes, and each leaf node corresponds to a classification result (manufacturer). Since the training data set of each tree is obtained through random sampling, they will have different division methods, so each tree may produce different classification results. Finally, the random forest determines the final classification result through "majority voting". When new data is input (that is, the PCA features of the new insulating paper spectrum to be traced to the manufacturer), each decision tree will give a predicted manufacturer. The random forest will summarize the prediction results of all decision trees and use a voting mechanism. The manufacturer with the most votes will be selected as the final classification result. This method increases the stability and accuracy of the model and avoids the problem of overfitting of a single decision tree.

[0105] Training: 70% of the principal component eigenvalues ​​and their manufacturer labels are used as the training set and input into the random forest algorithm for training.

[0106] Application: For insulation paperboard samples with unknown manufacturers, terahertz testing is performed on them, and the absorption spectrum is downsampled and principal component analysis is performed. The three eigenvalues ​​obtained by principal component analysis are input into the random forest classifier to obtain the corresponding insulation paperboard manufacturer name.

[0107] Through terahertz testing, terahertz time-domain spectra of insulation paperboard samples from 8 manufacturers were obtained. The time-domain spectra were Fourier transformed, and the absorption coefficients were calculated according to formulas (2) to (4). The terahertz absorption spectra of insulation paperboard samples from 8 manufacturers in the 0.1THz-3THz frequency band were obtained as follows: Figure 4 The insulation paper manufacturer traceability random forest classifier constructed in the previous article was used to test the manufacturer traceability of the insulation paperboard. The model prediction results were compared with the actual results. Figure 5 shown. Figure 5 The horizontal axis is the sample number, and the vertical axis is the insulation paperboard manufacturer category. The eight insulation paperboard manufacturers selected in this embodiment, including ABB, Dongfang, Guangxin, Weidmann, Xingrui, Xinyuan, AUO and Yingzhong, are numbered 1-8 in sequence. Figure 5It can be seen that among the 72 groups of sample data, there are 2 groups of data with prediction errors, and the model accuracy is 97.2%. The test results show that the method based on terahertz absorption spectroscopy combined with random forest classifier proposed in this embodiment can realize the traceability of 8 types of insulation paperboard manufacturers.

[0108] Embodiment 2

[0109] This embodiment provides an insulation paperboard manufacturer traceability system based on terahertz absorption spectroscopy, including:

[0110] The sample acquisition module is used to acquire a number of sample insulating paperboards.

[0111] The testing module is used to perform spectral testing on each sample insulation paperboard based on a transmission terahertz time-domain spectroscopy testing platform to obtain a terahertz time-domain spectral signal of each sample insulation paperboard.

[0112] The calculation module is used to perform Fourier transform on the terahertz time domain spectrum signal of each sample insulation paperboard to obtain the terahertz spectrum frequency domain form, and calculate the terahertz absorption coefficient of the sample insulation paperboard based on the terahertz spectrum frequency domain form to obtain the terahertz absorption spectrum of the sample insulation paperboard.

[0113] The downsampling module is used to downsample the terahertz absorption spectrum of each sample insulation paperboard to obtain the downsampled data of each sample insulation paperboard.

[0114] The analysis module is used to perform principal component analysis on the collected data to obtain the principal component analysis results of each sample insulation paperboard.

[0115] The training module is used to use the N principal component features with the highest content in the principal component analysis results as the input feature quantity of the random forest classifier, and use the insulation cardboard manufacturer as a label to train the insulation cardboard random forest classifier to obtain a trained insulation cardboard random forest classifier.

[0116] The traceability module is used to trace the insulation cardboard manufacturer of the insulation cardboard to be tested based on the trained insulation cardboard random forest classifier.

[0117] In summary, this application has the following technical effects:

[0118] This embodiment can trace the manufacturer of 8 kinds of insulating paperboards produced by different manufacturers before they are put into use in electrical equipment. After the insulating paperboards arrive, the power grid company can use the method of this embodiment to detect the new insulating paperboards that are not soaked in oil before entering the grid, and trace the manufacturers of the insulating paperboards to ensure that the insulating paperboards meet the manufacturer's standards and prevent the occurrence of manufacturers selling inferior products as good ones.

[0119] Since the density of cellulose molecular chains in insulating cardboards from different manufacturers is different, after being immersed in dry insulating oil, under the excitation of terahertz waves, the cellulose molecular chains will drive the surrounding insulating oil molecules to vibrate specifically, such as characteristic absorption peaks at 1.4THz and 1.8THz. The amplitude of terahertz characteristic absorption peaks of insulating cardboards produced by different manufacturers after oil immersion is different. The manufacturers of insulating cardboards after oil immersion can be traced according to the amplitude of characteristic absorption peaks.

[0120] However, for newly produced insulating cardboard, immersion in insulating oil will affect the secondary use of the insulating cardboard, so it is necessary to explore how to trace the terahertz manufacturer of new insulating cardboard that has not been immersed in insulating oil. The terahertz absorption spectra of insulating cardboard produced by different manufacturers when not immersed in oil are as follows: Figure 4 As shown in the figure, since there is no insulating oil between the cellulose molecular chains of the insulating paperboard, the terahertz characteristic absorption peaks of different manufacturers are different. Although the insulating paperboards of different manufacturers all have characteristic absorption peaks at 1.48THz, except for the paperboard produced by Xinyuan, the absorption peak amplitudes of the paperboards of other manufacturers have no obvious regularity, so it is impossible to use the characteristic absorption peaks to trace the manufacturer of the insulating paperboard that is not soaked in oil.

[0121] In this embodiment, the problem that the insulation paperboards of different manufacturers cannot be traced by using characteristic absorption peaks when they are not soaked in oil is solved, and the terahertz absorption spectrum is used to trace the manufacturers. By downsampling the absorption spectrum and performing principal component analysis, the three principal components carrying the most absorption spectrum information are extracted as the input of the random forest classifier, and the random forest classifier is used to trace the manufacturers of the insulation paperboards of the eight mainstream insulation paper manufacturers on the market (ABB, Dongfang, Guangxin, Weidmann, Xingrui, Xinyuan, AUO, Yingzhong), with an identification accuracy of 97.2%. Compared with the traditional method, the insulation paperboard manufacturer tracing method based on terahertz absorption spectroscopy proposed in this embodiment can trace the manufacturer of the insulation paperboard that is not soaked in oil before entering the network, and the detection speed is fast.

[0122] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0123] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy, characterized in that: The insulating paperboard manufacturer tracing method based on terahertz absorption spectroscopy includes: Obtain several sample insulation paperboards; Based on the transmission terahertz time-domain spectroscopy testing platform, a spectral test is performed on each sample insulation paperboard to obtain a terahertz time-domain spectral signal of each sample insulation paperboard; For each sample insulation paperboard, Fourier transform is performed on the terahertz time-domain spectrum signal of the sample insulation paperboard to obtain the terahertz spectrum frequency domain form, and the terahertz absorption coefficient of the sample insulation paperboard is calculated based on the terahertz spectrum frequency domain form to obtain the terahertz absorption spectrum of the sample insulation paperboard; Downsampling the terahertz absorption spectrum of each sample insulation paperboard to obtain downsampling data of each sample insulation paperboard; Perform principal component analysis on the collected data to obtain the principal component analysis results of each sample insulation paperboard; The N principal component features with the highest content in the principal component analysis results are used as the input feature quantity of the random forest classifier, and the insulation cardboard manufacturer is used as the label to train the insulation cardboard random forest classifier to obtain the trained insulation cardboard random forest classifier; Based on the trained insulation cardboard random forest classifier, the insulation cardboard manufacturer of the insulation cardboard to be tested is traced.

2. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 1 is characterized in that: Before getting several sample insulation paperboards, also include: Prepare sample insulation paperboard, including: Collect insulation cardboard from several insulation cardboard manufacturers; Cutting the insulating paperboard into a set size to obtain a cut insulating paperboard sample; The cut sample insulation paperboards are placed in a vacuum drying oven for drying to obtain the prepared sample insulation paperboards.

3. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 1 is characterized in that: The transmission-type terahertz time-domain spectroscopy test platform comprises: a femtosecond laser, a beam splitter, a time delay line, a reflector, a THz wave transmitter, a THz wave detector, a parabolic mirror, a phase-locked amplifier and a PC control terminal; the femtosecond laser is used to generate ultrashort pulse femtosecond laser; the beam splitter is used to split the femtosecond laser into two beams, one of which is used to excite terahertz waves through the THz wave transmitter, and the other is used to pass through the time delay line, the reflector and the THz wave detector to detect the terahertz waves after passing through the sample insulating paperboard; the parabolic mirror is used to collimate and focus the terahertz waves so that the terahertz waves are concentratedly irradiated on the sample insulating paperboard; the PC control terminal is connected to the THz wave detector signal and is used to record the time domain signal of the terahertz pulse.

4. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 1 is characterized in that: For each sample insulation paperboard, the terahertz time domain spectrum signal of the sample insulation paperboard is Fourier transformed to obtain the terahertz spectrum frequency domain form, and the terahertz absorption coefficient of the sample insulation paperboard is calculated based on the terahertz spectrum frequency domain form to obtain the terahertz absorption spectrum of the sample insulation paperboard, specifically including: According to the formula Calculate the electric field in the terahertz frequency domain; According to the formula Calculate the terahertz absorption coefficient of the sample insulation paperboard; Among them, ω is the angular frequency, E(ω) is the terahertz frequency domain electric field, E(t) is the terahertz time domain electric field, A(ω) is the frequency domain amplitude, φ(ω) is the frequency domain phase, c is the speed of light in vacuum, and k(ω) is the extinction coefficient.

5. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 1 is characterized in that: The frequency band to be studied of the terahertz absorption spectrum is set to the 0.1THz-3THz frequency band.

6. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 1 is characterized in that: The principal component analysis of the collected data was performed to obtain the principal component analysis results of each sample insulation paperboard, including: Get the collected data; Performing standardization processing on the collected data to obtain a standardized matrix of the collected data; Calculating a standardized sample covariance matrix according to the standardized matrix of the collected data; According to the eigenvalues ​​and eigenvectors of the standardized matrix, the principal component contribution rate and the cumulative contribution rate are calculated.

7. The method for tracing the manufacturer of insulating paperboard based on terahertz absorption spectroscopy according to claim 6 is characterized in that: The standardized matrix of the collected data is specifically: Among them, n is the number of groups of down-sampled data, p is the frequency point of each group of down-sampled data, X np The pth frequency point of the nth group of collected data.

8. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 6 is characterized in that: According to the standardized matrix of the collected data, the standardized sample covariance matrix is ​​calculated, which specifically includes: According to the formula Calculate each element in the standardized sample covariance matrix R to obtain the standardized sample covariance matrix Where n is the number of groups of down-sampled data, p is the frequency point of each group of down-sampled data, X is the normalization matrix, X' is the transposed matrix of the normalization matrix, R np Collect the p-th element in the covariance matrix of the data for the n-th group.

9. The method for tracing the source of insulating paperboard manufacturers based on terahertz absorption spectroscopy according to claim 6, characterized in that: According to the eigenvalues ​​and eigenvectors of the standardized matrix, the principal component contribution rate and the cumulative contribution rate are calculated, including: According to the formula Calculate the principal component contribution rate; According to the formula Calculate the cumulative contribution rate; Among them, a is the principal component contribution rate, b is the cumulative contribution rate, k is the specific principal component number, i is the first i principal components selected, λ k is the kth principal component, and P is the total dimension of each set of data.

10. An insulation paperboard manufacturer traceability system based on terahertz absorption spectroscopy, characterized in that: include: A sample acquisition module, used to acquire a number of sample insulating paperboards; A testing module, used to perform a spectral test on each of the sample insulation paperboards based on a transmission terahertz time-domain spectroscopy testing platform to obtain a terahertz time-domain spectral signal of each sample insulation paperboard; A calculation module is used to perform Fourier transform on the terahertz time-domain spectrum signal of each sample insulation paperboard to obtain a terahertz spectrum frequency-domain form, and calculate the terahertz absorption coefficient of the sample insulation paperboard based on the terahertz spectrum frequency-domain form to obtain a terahertz absorption spectrum of the sample insulation paperboard; A downsampling module is used to downsample the terahertz absorption spectrum of each sample insulation paperboard to obtain downsampled data of each sample insulation paperboard; An analysis module is used to perform principal component analysis on the collected data to obtain the principal component analysis results of each sample insulation paperboard; A training module is used to use the N principal component features with the highest content in the principal component analysis results as input feature quantities of the random forest classifier, and use the insulation cardboard manufacturer as a label to train the insulation cardboard random forest classifier to obtain a trained insulation cardboard random forest classifier; The traceability module is used to trace the insulation cardboard manufacturer of the insulation cardboard to be tested based on the trained insulation cardboard random forest classifier.

Images