CN108458989B - Terahertz multi-parameter spectrum-based coal rock identification method - Google Patents
Terahertz multi-parameter spectrum-based coal rock identification method Download PDFInfo
- Publication number
- CN108458989B CN108458989B CN201810403512.0A CN201810403512A CN108458989B CN 108458989 B CN108458989 B CN 108458989B CN 201810403512 A CN201810403512 A CN 201810403512A CN 108458989 B CN108458989 B CN 108458989B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- coal
- rock
- terahertz
- domain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001228 spectrum Methods 0.000 title claims abstract description 190
- 239000003245 coal Substances 0.000 title claims abstract description 188
- 239000011435 rock Substances 0.000 title claims abstract description 166
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000005065 mining Methods 0.000 claims abstract description 40
- 238000000411 transmission spectrum Methods 0.000 claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 28
- 238000010521 absorption reaction Methods 0.000 claims abstract description 23
- 238000004458 analytical method Methods 0.000 claims abstract description 23
- 238000005516 engineering process Methods 0.000 claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000004519 manufacturing process Methods 0.000 claims abstract description 9
- 230000009467 reduction Effects 0.000 claims abstract description 9
- 238000001328 terahertz time-domain spectroscopy Methods 0.000 claims abstract description 7
- 230000003287 optical effect Effects 0.000 claims abstract description 6
- 239000000523 sample Substances 0.000 claims description 70
- 230000003595 spectral effect Effects 0.000 claims description 41
- 239000000843 powder Substances 0.000 claims description 29
- 239000011159 matrix material Substances 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 15
- 238000000862 absorption spectrum Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 8
- 239000000463 material Substances 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 230000005540 biological transmission Effects 0.000 claims description 6
- -1 bituminous Substances 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- RHZUVFJBSILHOK-UHFFFAOYSA-N anthracen-1-ylmethanolate Chemical compound C1=CC=C2C=C3C(C[O-])=CC=CC3=CC2=C1 RHZUVFJBSILHOK-UHFFFAOYSA-N 0.000 claims description 4
- 239000003830 anthracite Substances 0.000 claims description 4
- 239000003077 lignite Substances 0.000 claims description 4
- JBRZTFJDHDCESZ-UHFFFAOYSA-N AsGa Chemical compound [As]#[Ga] JBRZTFJDHDCESZ-UHFFFAOYSA-N 0.000 claims description 3
- 229910001218 Gallium arsenide Inorganic materials 0.000 claims description 3
- 239000004698 Polyethylene Substances 0.000 claims description 3
- 230000008033 biological extinction Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 230000014759 maintenance of location Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 229920000573 polyethylene Polymers 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 3
- 229910052594 sapphire Inorganic materials 0.000 claims description 3
- 239000010980 sapphire Substances 0.000 claims description 3
- 229910052710 silicon Inorganic materials 0.000 claims description 3
- 239000010703 silicon Substances 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 238000005520 cutting process Methods 0.000 abstract description 10
- 238000012706 support-vector machine Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000001035 drying Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007873 sieving Methods 0.000 description 2
- 238000004939 coking Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003703 image analysis method Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000010117 shenhua Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000036346 tooth eruption Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3581—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation
- G01N21/3586—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation by Terahertz time domain spectroscopy [THz-TDS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Abstract
The invention discloses a terahertz time-domain spectroscopy-based coal-rock interface identification method, which is suitable for the technical fields of coal-rock interface identification, autonomous height adjustment and unmanned and intelligent mining of a coal mining machine. Collecting a terahertz time-domain spectrum of the coal rock sample by using a terahertz spectrometer, and converting the terahertz time-domain spectrum into a terahertz frequency-domain spectrum by using fast Fourier transform; extracting a transmittance spectrum, a refractive index spectrum and an absorption coefficient spectrum from the terahertz frequency domain spectrum; performing dimension reduction and feature extraction on the detection spectrum data by using an LDA technology, and performing deep neural network analysis modeling on the optical parameter spectrum; and (3) collecting a coal rock manufacturing sample of the detected region, obtaining a terahertz time-domain spectrum of the coal rock sample, carrying out treatment in the step (2) and then carrying out the coal rock interface identification in the model built in the steps (3) and (4). The method has the advantages that the coal rock medium can be rapidly, efficiently and accurately distinguished through detection, the cutting state of the coal mining machine can be accurately identified, and the problem of automatic height adjustment of the coal mining machine is solved.
Description
Technical Field
The invention relates to a terahertz multi-parameter spectrum-based coal-rock interface identification method, and belongs to the technical field of coal-rock interface identification, autonomous height adjustment and unmanned and intelligent mining of coal mining machines.
Background
The unmanned and intelligent working face mining technology is an efficient safe mining mode, can effectively reduce personnel accidents, ensures safe mining of coal mines, and is the development direction and trend of the coal mine industry. The coal science and technology 'thirteen five' clearly indicates that innovation-driven development strategies are comprehensively implemented, relevant key technologies in intelligent mines are overcome, and intelligent control technologies are brought into the main production links of coal. To achieve the purpose, the problem of intelligent coal rock identification is solved, which is a significant problem that the intelligent coal mining equipment is restricted at present.
The coal rock interface recognition means that the coal mining machine is used for mining along the interface of a coal bed and a rock stratum as far as possible when the coal mining machine cuts the coal bed, so that the maximum output is guaranteed, the height of a roller of the coal mining machine is adjusted in time when the rock stratum is cut, the over-cutting is avoided, a large amount of gangue is mixed into raw coal, the coal quality and the output rate are reduced, and cutting teeth of the roller of the coal mining machine are damaged.
The other problems of the coal and rock are always hot research at home and abroad, and various scholars also put forward a plurality of representative research methods, such as an image analysis method, a cutting data method, a radar detection method and the like. The above-described methods have achieved certain success but suffer from several drawbacks. If the hardness characteristics of the coal rock are relative, the cutting state of the coal mining machine is difficult to judge through the cutting data of the coal mining machine; on the other hand, different sensors need to be selected and the discrimination model needs to be adjusted according to different mining areas, mining environments and equipment.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a coal-rock interface identification method based on a terahertz spectrum, which is used for solving the problems. The terahertz is a special frequency band, and the terahertz photon energy level is equivalent to the vibration energy levels of a plurality of dielectric media and materials, so that the photon energy of terahertz waves can be absorbed after the terahertz waves irradiate the coal rock, the light intensity of the terahertz waves can be changed to form a terahertz spectrum, physical information (such as an absorption spectrum and a refractive index spectrum) of a detected substance can be obtained by researching the terahertz spectrum, the coal rock property is judged, and the drum of the coal mining machine is guided to be heightened. Compared with the coal rock hardness difference, the terahertz spectrum data can change along with the change of the frequency, and the frequency points with large difference can be found out. Compared with other spectrum technologies, the terahertz spectrum can directly obtain the phase of the electric field without using a Kramers-Kronig equation to solve the phase angle, and the calculation amount and complexity are greatly reduced. Even if the coal rock characteristics are similar, the problem of 'same foreign matter spectrum' when the coal rock characteristics are similar can be efficiently solved by jointly using the terahertz transmission spectrum, the absorption spectrum and the refractive index spectrum of the coal rock and reasonably using a deep learning technology, and the recognition accuracy which cannot be achieved by a Support Vector Machine (SVM) is realized. The invention is realized by the following technical scheme: a coal-rock interface identification method based on a terahertz time-domain spectrum is characterized by comprising the following steps:
1. a coal-rock interface identification method based on a terahertz multiparameter spectrum uses a terahertz time-domain spectrometer which comprises five parts, namely a femtosecond laser (3) for generating a light source, a transmitting antenna (1) for generating a terahertz pulse, a sample detection bin (4), a receiving antenna (2) for receiving the terahertz pulse and a delay line (5) for adjusting the time difference between the femtosecond laser and the terahertz pulse; the sample detection bin (4) is arranged between the transmitting antenna (1) and the receiving antenna (2), and terahertz light of the transmitting antenna vertically acts on a coal rock sample of the sample detection bin (4); the method is characterized by comprising the following steps:
and 5, collecting different coal rocks in the step 1 and manufacturing samples, obtaining terahertz time-domain spectrums of the coal and rock samples, carrying out processing in the step 2, then carrying out the step 3 and the step 4, and carrying out coal rock interface recognition on the built model.
The femtosecond laser Hertz time domain spectrum acquisition device is system equipment constructed by Tianjin university, the central wavelength of the femtosecond laser Hertz time domain spectrum acquisition device is 800nm, the bandwidth is more than 70nm, and the repetition frequency is 80 MHz; the transmitting antenna is made of high-resistance gallium arsenide, the receiving antenna is made of silicon on sapphire, and the system frequency range is as follows: 0.1-3.5 THz; the sample detection bin is made of polyethylene material.
3-5 different point positions are selected for collection when the terahertz time-domain spectrum of the coal rock sample is collected every time, and the spectrum is repeatedly collected for 3-4 times at each point position.
In the detection process, a transmission scanning module of a terahertz time-domain spectroscopy system is adopted to obtain a reference signal time-domain spectrum taking dry air as a background and a terahertz time-domain spectrum of a coal rock sample, wherein the humidity of the dry air is lower than 5%.
The terahertz frequency spectrum range after Fourier transform in the step 2 is 0.1-3.5 THz;
based on a THz optical parameter extraction model for extracting a transmission spectrum T (omega), a refractive index spectrum n (omega) and an absorption coefficient spectrum alpha (omega) based on a terahertz frequency domain spectrum, the calculation formula is as follows:
T(ω)＝Esam(ω)/Eref(ω)
wherein the transmission coefficient is T (omega), Eref(omega) and Esam(omega) is a reference signal terahertz frequency spectrum and a sample terahertz frequency spectrum respectively, the thickness of the sample is d, c is the speed of light, n (omega) is the refractive index of the sample, α (omega) is the absorption coefficient of the sample, rho (omega) is the ratio of the amplitude of the test sample signal to the reference signal,
the term is the phase difference between the test sample signal and the reference signal, and k (ω) is the imaginary part of the complex index of refraction of the test sample, also known as the extinction coefficient.
And 2, optimizing the transmission spectrum, the refractive index spectrum and the absorption spectrum by a windowing technology, extracting the spectral data of the 0.4-1THz frequency band with high signal-to-noise ratio, and processing a spectral curve by using a Savitzky-Golay smoothing method to eliminate system and environment noises.
The steps of reducing the dimension of the spectral data and extracting the spectral characteristic value by adopting a linear discriminant analysis method are as follows: combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the coal samples to construct and label a multi-parameter spectrum X1 of the coal, and combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the rock samples to construct and label a multi-parameter spectrum X2 of the rock; wherein the matrix structures of the multi-parameter spectrum X1 of the coal and the multi-parameter spectrum of the rock are as follows:
wherein n represents different coal types/rock types, p represents the spectral data dimension of the same coal/rock, and x represents all sample data, represented here in a matrix form;
the multi-parameter optical spectrum of the coal and rock samples is subjected to dimension reduction and feature extraction through linear discriminant analysis/LDA, and compared with single spectral data, the combined spectrum can obtain higher recognition rate through a linear discriminant analysis method.
In step 3, the dimension reduction process of the linear discriminant analysis method comprises the following steps:
1) calculating mean vectors of spectral data of different types of coal rocks from the coal rock spectral data set, wherein the vector dimension is set to be 200 dimensions;
2) using the formula:
calculating an inter-class divergence matrix SbIn which N isjDenotes the number of samples labeled class j, μ is the mean vector of all samples, μjIs the mean vector of all samples in category j;
3) using the formula:calculating an intra-class divergence matrix SwWherein X isjSet of spectral data for class j coal/rock, NjThe number of samples marked as j class;
4) calculating eigenvalues and eigenvectors corresponding to the inter-class divergence and the intra-class divergence, wherein the eigenvalue is lambda1，λ2，...，λkThe feature vectors are respectively W1, W2, … and Wk;
5) arranging the eigenvalues in a descending order, and selecting eigenvectors corresponding to the first d eigenvalues to form a matrix W;
6) and (3) projecting and mapping the raw coal rock spectrum data X1/X2 to a new space by using a matrix W to obtain a new coal spectrum data set Y1 and a new rock spectrum data set Y2, wherein the dimensionalities of the coal spectrum data set Y1 and the rock spectrum data set Y2 are reduced to 80 dimensionalities, and the data size and the calculation time are greatly reduced.
Respectively labeling different types of coal and rock, and then modeling by using a deep neural network method by using a coal spectrum data set Y1 and a rock spectrum data set Y2 which are subjected to linear discriminant analysis as input, wherein the modeling dimension is 80 dimensions, the number of neurons is 500, the output dimension is 2 dimensions, the maximum iteration number is 100, the learning rate is 0.001, the dropout retention node ratio is 0.9, and the loss function is determined to be the minimum cross entropy loss by using an AdamaOptizer optimizer.
When the coal mining machine works, a roller of the coal mining machine cuts a coal bed and a rock stratum to generate a blocky coal rock sample and a powdery coal rock sample, the powder collecting device is used for collecting material powder cut by the current coal mining machine, the powder is gathered in the sample detection bin through the conveying channel, and the terahertz time-domain spectrum of unknown powder is obtained through the terahertz spectrum collecting probe.
Has the advantages that: the method adopts a terahertz time-domain spectrometer to collect a reference signal and a terahertz time-domain spectrum of the coal rock; converting the terahertz time-domain spectrum of the coal rock into a frequency-domain spectrum by using fast Fourier transform; and extracting a transmission spectrum, a refractive index spectrum and an absorption coefficient spectrum of the coal rock from the terahertz frequency domain spectrum. The spectrum data are preprocessed by adopting technologies such as windowing, smoothing and LDA, and then classification modeling and recognition are carried out on the processed coal rock spectrum by utilizing a deep neural network, so that the coal rock interface recognition based on the terahertz spectrum technology is realized.
The method has the advantages that different coal rocks are distinguished by utilizing the terahertz spectrum technology and the mode recognition method, the terahertz spectrum data can change along with the change of frequency compared with the traditional detection method based on the coal rock hardness difference, the problem of foreign matter cospectrum when the coal rock characteristics are close can be efficiently solved by utilizing the deep learning technology, the problem that the recognition accuracy rate of the existing method is reduced after the quantity/types of coal and rock are increased is solved, the recognition effect is good, the deep neural network is used as the deep learning method, the problem that the recognition accuracy rate is reduced along with the increase of the quantity in the prior art, namely the model has poor robustness and expansibility is solved, and the recognition accuracy is improved. The method can rapidly, efficiently and accurately distinguish the coal rock medium and is applied to the field of unmanned/intelligent mining.
Drawings
FIG. 1 is a flow chart of a coal-rock interface identification method based on terahertz time-domain spectroscopy according to an embodiment of the invention;
FIG. 2 is a schematic structural diagram of a terahertz time-domain spectroscopy acquisition system of the present invention;
FIG. 3a is a portion of the terahertz spectrum of the present invention;
FIG. 3b is a partial transmission spectrum of the present invention;
FIG. 3c is a partial absorption spectrum of the present invention;
FIG. 3d is a partial refractive index spectrum of the present invention;
FIG. 4 is a graph of the effect of a single spectrum LDA feature extraction on accuracy after LDA feature extraction of a combined spectrum;
fig. 5 is a graph comparing the results of various coal rock classifications tested in accordance with the present invention with a prior art svm method.
In the figure: 1. transmitting antenna, 2, receiving antenna, 3, femtosecond laser, 4, sample detection bin, 5, delay line, 6 and beam splitter.
Detailed Description
As shown in fig. 1 and 2, a coal-rock interface identification method based on terahertz time-domain spectroscopy is characterized by comprising the following steps:
the terahertz time-domain spectrograph comprises five parts, namely a femtosecond laser 3 for generating a light source, a transmitting antenna 1 for generating terahertz pulses, a sample detection bin 4, a receiving antenna 2 for receiving the terahertz pulses and a delay line 5 for adjusting the time difference between the femtosecond laser and the terahertz pulses, and light beams are mutually transmitted through a beam splitter 6; the sample detection bin 4 is positioned between the transmitting antenna 1 and the receiving antenna 2, and terahertz light of the transmitting antenna vertically acts on a coal rock sample of the sample detection bin 4; the femtosecond laser Hertz time domain spectrum acquisition device is system equipment constructed by Tianjin university, the central wavelength of the femtosecond laser Hertz time domain spectrum acquisition device is 800nm, the bandwidth is more than 70nm, and the repetition frequency is 80 MHz; the transmitting antenna is made of high-resistance gallium arsenide, the receiving antenna is made of silicon on sapphire, and the system frequency range is as follows: 0.1-3.5 THz; the sample detection bin is made of polyethylene material; 3-5 different point positions are selected for collection when the terahertz time-domain spectrum of the coal rock sample is collected every time, and the spectrum is repeatedly collected for 3-4 times at each point position.
The method comprises the following steps:
the terahertz frequency spectrum range after Fourier transform in the step 2 is 0.1-3.5 THz;
based on a THz optical parameter extraction model for extracting a transmission spectrum T (omega), a refractive index spectrum n (omega) and an absorption coefficient spectrum alpha (omega) based on a terahertz frequency domain spectrum, the calculation formula is as follows:
T(ω)＝Esam(ω)/Eref(ω)
wherein the transmission coefficient is T (omega), Eref(omega) and Esam(ω) Respectively a reference signal terahertz frequency spectrum and a sample terahertz frequency spectrum, the thickness of the sample is d, c is the speed of light, n (omega) is the refractive index of the sample, α (omega) is the absorption coefficient of the sample, the term rho (omega) is the ratio of the amplitude of a test sample signal to the amplitude of the reference signal,
the term is the phase difference between the test sample signal and the reference signal, and k (omega) is the imaginary part of the complex refractive index of the test sample, also called extinction coefficient;
the steps of reducing the dimension of the spectral data and extracting the spectral characteristic value by adopting a linear discriminant analysis method are as follows: combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the coal samples to construct and label a multi-parameter spectrum X1 of the coal, and combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the rock samples to construct and label a multi-parameter spectrum X2 of the rock; wherein the matrix structures of the multi-parameter spectrum X1 of the coal and the multi-parameter spectrum of the rock are as follows:
wherein n represents different coal types/rock types, p represents the spectral data dimension of the same coal/rock, and x represents all sample data, represented here in a matrix form;
performing dimensionality reduction and feature extraction on multi-parameter spectra of coal and rock samples through linear discriminant analysis/LDA (linear discriminant analysis/linear discriminant analysis), wherein a higher recognition rate can be obtained after the combined spectrum is subjected to a linear discriminant analysis method relative to single spectrum data;
the dimension reduction process of the linear discriminant analysis method comprises the following steps:
1) calculating mean vectors of spectral data of different types of coal rocks from the coal rock spectral data set, wherein the vector dimension is set to be 200 dimensions;
2) using the formula:
calculating an inter-class divergence matrix SbIn which N isjDenotes the number of samples labeled class j, μ is the mean vector of all samples, μjIs the mean vector of all samples in category j;
3) using the formula:
calculating an intra-class divergence matrix SwWherein X isjSet of spectral data for class j coal/rock, NjThe number of samples marked as j class;
4) calculating eigenvalues and eigenvectors corresponding to the inter-class divergence and the intra-class divergence, wherein the eigenvalue is lambda1，λ2，...，λkThe feature vectors are respectively W1, W2, … and Wk;
5) arranging the eigenvalues in a descending order, and selecting eigenvectors corresponding to the first d eigenvalues to form a matrix W;
6) the method comprises the steps of utilizing a matrix W to project and map raw coal rock spectrum data X1/X2 to a new space to obtain a new coal spectrum data set Y1 and a new rock spectrum data set Y2, wherein the dimensionality of the coal spectrum data set Y1 and the dimensionality of the rock spectrum data set Y2 are reduced to 80 dimensions, and the data size and the computing time are greatly reduced;
respectively labeling different types of coal and rock, and then modeling by using a deep neural network method by using a coal spectrum data set Y1 and a rock spectrum data set Y2 subjected to linear discriminant analysis as input, wherein the modeling dimension is 80 dimensions, the number of neurons is 500, the output dimension is 2 dimensions, the maximum iteration number is 100, the learning rate is 0.001, the dropout retention node ratio is 0.9, and the loss function is determined to be the minimum cross entropy loss by using an AdamaOptizer optimizer;
FIG. 4 is the effect of the single spectrum LDA feature extraction on the accuracy of the combined spectrum after LDA feature extraction, where LDA (11) is the combined spectrum and the remainder is the single spectrum. The abscissa in the graph is the number of samples, 200 samples are included, 15 kinds of coal and 5 kinds of rock are included, 10 samples are included in each kind, and obviously, compared with single spectrum data, the joint spectrum can obtain higher recognition rate after being subjected to LDA;
and 5, collecting different kinds of coal rocks and manufacturing samples in the step 1, obtaining terahertz time-domain spectrums of the coal and rock samples, carrying out processing in the step 2, and then carrying out coal-rock interface recognition in the models built in the step 3 and the step 4.
For example, in the spectrum data in fig. 3a, fig. 3b, fig. 3c, and fig. 3d, the original spectrum data of the coal and the rock cannot be directly distinguished, but after windowing data selection and smoothing processing are performed, and LDA data dimension reduction and feature extraction are performed, feature values of the coal and rock samples are remarkably distinguished, and the combined spectrum is modeled through a deep neural network, so that identification of various kinds of coal and rock can be accurately achieved. The method also shows that the problem of 'same foreign matter and spectrum' when the coal and rock characteristics are similar can be efficiently solved, and the identification accuracy is improved.
When the coal mining machine works, the roller of the coal mining machine is used for cutting a coal bed and a rock stratum to generate a blocky coal rock sample and a powdery coal rock sample, the powder collecting device is used for collecting material powder cut by the current coal mining machine, the powder is gathered in the sample detection bin through the conveying channel, and the terahertz time-domain spectrum of unknown powder is obtained through the terahertz spectrum collecting probe.
The terahertz time-domain spectrograph is placed near a roller of a coal mining machine, when the coal mining machine starts to work, a coal layer/rock stratum is cut by a roller cutting tooth along with the rotation of the roller to form coal/rock powder or coal/rock blocks, a coal/rock powder collector is installed on a machine body of the coal mining machine below the roller, coal powder and rock powder generated when the roller of the coal mining machine cuts the coal layer/rock stratum are collected in a timing suction mode, and the coal powder is sent into a sample detection bin 4 through a pipeline channel to realize the filling of samples. The sample detection bin 4 can pour out the original powder in a timed rotating mode and then reset and load new coal rock powder. The sample detection chamber 4 is positioned on the terahertz emission light path, so that the terahertz light vertically acts on the sample.
In step 5, the spectral data processed by LDA is used as input and sent into the established deep neural network model, and the model outputs a result; when the result is true, the current powder is the coal powder, namely the coal mining machine is currently cutting the coal bed; when the output result is false, the current powder is indicated to be rock powder, namely the coal mining machine is cutting rock strata currently, the current top plate or bottom plate can be judged by combining the inclination angle of the rocker arm of the coal mining machine, the height of the roller needs to be adjusted downwards immediately when the top plate is cut, and the height of the roller needs to be adjusted upwards immediately when the bottom plate is cut.
Fig. 5 is a comparison of the results of the various coal rock classifications tested against the prior svm method. Wherein case1 is the method, and case2 is the SVM method. The abscissa is the number of samples, and the total number of the samples is 80, wherein the samples comprise 15 kinds of coal and 5 kinds of rocks, 4 kinds of samples of different coal rock samples are compared by utilizing the method and the SVM method, wherein the coal rock samples are from Shandong, Yangyuan mine of an ore group, Guoguang coal mine of a union group, Daihuang river mine of a Huaibei mine group, Heihua black Daigui ditch mine of a Shenhua group, Dongpong mine of a Chachentai mine group and the like, and the kinds of the coal samples comprise fine sandstone, shale and medium sandstone, anthracite, lean coal, 1/3/coking coal, gas coal, non-caking coal, long-flame coal, lignite and the like; the collected samples are tested according to the step 5, and the result shows that when the number of the coal types and the rock types is small, the method and the SVM method can keep high recognition rate, but when the number of the coal types and the rock types is increased, the recognition rate of the SVM is reduced, and the method can still accurately recognize various coal rocks, so that the cutting state of the coal mining machine can be accurately judged, and a basis is provided for automatic height adjustment of the coal mining machine.
Claims (10)
1. A coal-rock interface identification method based on a terahertz multiparameter spectrum uses a terahertz time-domain spectrometer which comprises five parts, namely a femtosecond laser (3) for generating a light source, a transmitting antenna (1) for generating a terahertz pulse, a sample detection bin (4), a receiving antenna (2) for receiving the terahertz pulse and a delay line (5) for adjusting the time difference between the femtosecond laser and the terahertz pulse; the sample detection bin (4) is positioned between the transmitting antenna (1) and the receiving antenna (2), and terahertz light of the transmitting antenna vertically acts on a coal rock sample of the sample detection bin (4); the method is characterized by comprising the following steps:
step 1, respectively selecting and manufacturing coal rock samples from different kinds of coal rocks, and acquiring a reference signal and terahertz time-domain spectral data of the coal rock samples by using a terahertz time-domain spectrometer, wherein the coal rock samples comprise: anthracite, bituminous, lignite, sandstone, and shale;
step 2, converting the terahertz time-domain spectrum into a terahertz frequency-domain spectrum by using fast Fourier transform; extracting a transmittance spectrum, a refractive index spectrum and an absorption coefficient spectrum of each coal rock sample from the terahertz frequency domain spectrum; respectively carrying out smoothing and windowing pretreatment on the transmittance spectrum, the refractive index spectrum and the absorption coefficient spectrum, and respectively selecting spectral data of 0.4-1.0THz frequency band in the transmittance spectrum, the refractive index spectrum and the absorption coefficient spectrum after comparison;
step 3, marking the spectral data of the transmittance spectrum, the refractive index spectrum and the absorption coefficient spectrum as the spectral data of coal and rock, and performing dimension reduction and feature extraction on the spectral data by utilizing a linear discriminant analysis technology;
step 4, performing combined analysis modeling on the transmittance spectrum, the refractive index spectrum and the absorption coefficient spectrum by using a deep neural network;
and 5, collecting different kinds of coal rocks and manufacturing samples in the step 1, obtaining terahertz time-domain spectrums of the coal and rock samples, carrying out processing in the step 2, and then carrying out coal-rock interface recognition in the models built in the step 3 and the step 4.
2. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: the terahertz time-domain spectrum acquisition device of the femtosecond laser is system equipment constructed by Tianjin university, the central wavelength of the terahertz time-domain spectrum acquisition device is 800nm, the bandwidth is more than 70nm, and the repetition frequency is 80 MHz; the transmitting antenna is made of high-resistance gallium arsenide, the receiving antenna is made of silicon on sapphire, and the system frequency range is as follows: 0.1-3.5 THz; the sample detection bin is made of polyethylene material.
3. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: 3-5 different point positions are selected for collection when the terahertz time-domain spectrum of the coal rock sample is collected every time, and the spectrum is repeatedly collected for 3-4 times at each point position.
4. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: in the detection process, a transmission scanning module of a terahertz time-domain spectroscopy system is adopted to obtain a reference signal time-domain spectrum taking dry air as a background and a terahertz time-domain spectrum of a coal rock sample, wherein the humidity of the dry air is lower than 5%.
5. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: the terahertz frequency spectrum range after Fourier transform in the step 2 is 0.1-3.5 THz;
based on a THz optical parameter extraction model for extracting a transmission spectrum T (omega), a refractive index spectrum n (omega) and an absorption coefficient spectrum alpha (omega) based on a terahertz frequency domain spectrum, the calculation formula is as follows:
T(ω)＝Esam(ω)/Eref(ω)
wherein the transmission coefficient is T (omega), Eref(omega) and Esam(omega) is a reference signal terahertz frequency spectrum and a sample terahertz frequency spectrum respectively, the thickness of the sample is d, c is the speed of light, n (omega) is the refractive index of the sample, α (omega) is the absorption coefficient of the sample, rho (omega) is the ratio of the amplitude of the test sample signal to the reference signal,
the term is the phase difference between the test sample signal and the reference signal, and k (ω) is the imaginary part of the complex index of refraction of the test sample, also known as the extinction coefficient.
6. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: and 2, optimizing the transmission spectrum, the refractive index spectrum and the absorption spectrum by a windowing technology, extracting the spectral data of the 0.4-1THz frequency band with high signal-to-noise ratio, and processing a spectral curve by using a Savitzky-Golay smoothing method to eliminate system and environment noises.
7. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein the steps of performing dimension reduction on the spectral data and extracting the spectral feature value by using a linear discriminant analysis method comprise: combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the coal samples to construct and label a multi-parameter spectrum X1 of the coal, and combining the absorption spectrum, the transmission spectrum and the refraction spectrum of all the rock samples to construct and label a multi-parameter spectrum X2 of the rock; wherein the matrix structures of the multi-parameter spectrum X1 of the coal and the multi-parameter spectrum of the rock are as follows:
wherein n represents different coal types/rock types, p represents the spectral data dimension of the same coal/rock, and x represents all sample data, represented here in a matrix form;
and performing dimension reduction and feature extraction on the multi-parameter optical spectrum of the coal and rock sample through linear discriminant analysis.
8. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein in step 3, the dimension reduction process of the linear discriminant analysis method is as follows:
1) calculating mean vectors of spectral data of different types of coal rocks from the coal rock spectral data set, wherein the vector dimension is set to be 200 dimensions;
2) using the formula:
calculating an inter-class divergence matrix SbIn which N isjDenotes the number of samples labeled class j, μ is the mean vector of all samples, μjIs the mean vector of all samples in category j;
3) using the formula:
calculating an intra-class divergence matrix SwWherein X isjSet of spectral data for class j coal/rock, NjThe number of samples marked as j class;
4) calculating eigenvalues and eigenvectors corresponding to the inter-class divergence and the intra-class divergence, wherein the eigenvalue is lambda1，λ2，...，λkThe feature vectors are respectively W1, W2, … and Wk;
5) arranging the eigenvalues in a descending order, and selecting eigenvectors corresponding to the first d eigenvalues to form a matrix W;
6) and (3) projecting and mapping the raw coal rock spectrum data X1/X2 to a new space by using a matrix W to obtain a new coal spectrum data set Y1 and a new rock spectrum data set Y2, wherein the dimensionality of the coal spectrum data set Y1 and the rock spectrum data set Y2 is reduced to 80 dimensionalities.
9. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1 or 8, wherein: respectively labeling different types of coal and rock, and then modeling by using a deep neural network method by using a coal spectrum data set Y1 and a rock spectrum data set Y2 which are subjected to linear discriminant analysis as input, wherein the modeling dimension is 80 dimensions, the number of neurons is 500, the output dimension is 2 dimensions, the maximum iteration number is 100, the learning rate is 0.001, the dropout retention node ratio is 0.9, and the loss function is determined to be the minimum cross entropy loss by using an AdamaOptizer optimizer.
10. The coal-rock interface identification method based on the terahertz multiparameter spectrum as claimed in claim 1, wherein: when the coal mining machine works, a roller of the coal mining machine cuts a coal bed and a rock stratum to generate a blocky coal rock sample and a powdery coal rock sample, the powder collecting device is used for collecting material powder cut by the current coal mining machine, the powder is gathered in the sample detection bin through the conveying channel, and the terahertz time-domain spectrum of unknown powder is obtained through the terahertz spectrum collecting probe.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810403512.0A CN108458989B (en) | 2018-04-28 | 2018-04-28 | Terahertz multi-parameter spectrum-based coal rock identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810403512.0A CN108458989B (en) | 2018-04-28 | 2018-04-28 | Terahertz multi-parameter spectrum-based coal rock identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108458989A CN108458989A (en) | 2018-08-28 |
| CN108458989B true CN108458989B (en) | 2020-10-09 |
Family
ID=63236397
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810403512.0A Active CN108458989B (en) | 2018-04-28 | 2018-04-28 | Terahertz multi-parameter spectrum-based coal rock identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108458989B (en) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109632693A (en) * | 2018-12-10 | 2019-04-16 | 昆明理工大学 | A kind of tera-hertz spectra recognition methods based on BLSTM-RNN |
| CN109886421B (en) * | 2019-01-08 | 2021-09-21 | 浙江大学 | Swarm intelligence coal-winning machine cutting pattern recognition system based on ensemble learning |
| CN109765191B (en) * | 2019-01-18 | 2023-11-10 | 中国矿业大学 | Motion coal rock translation type tracking hyperspectral identification device |
| CN110068543A (en) * | 2019-03-26 | 2019-07-30 | 昆明理工大学 | A kind of tera-hertz spectra recognition methods based on transfer learning |
| CN110080766B (en) * | 2019-04-30 | 2024-03-01 | 中国矿业大学 | Comprehensive mining working face coal-rock interface identification device and method |
| CN110082307B (en) * | 2019-04-30 | 2023-08-29 | 中国矿业大学 | Coal rock reflection spectrum identification experimental device and method for simulating underground coal mine environment |
| CN110068544B (en) * | 2019-05-08 | 2021-09-17 | 广东工业大学 | Substance identification network model training method and terahertz spectrum substance identification method |
| CN110751230A (en) * | 2019-10-30 | 2020-02-04 | 深圳市太赫兹科技创新研究院有限公司 | Substance classification method, substance classification device, terminal device and storage medium |
| CN111337883B (en) * | 2020-04-17 | 2022-02-08 | 中国矿业大学(北京) | Intelligent detection and identification system and method for mine coal rock interface |
| CN111982838A (en) * | 2020-08-25 | 2020-11-24 | 吉林大学 | Hyperspectrum-based coal rock identification and detection method |
| CN113029995B (en) * | 2021-03-10 | 2022-09-27 | 太原理工大学 | Linear frequency modulation coal rock radiation detection device and method |
| CN113295673B (en) * | 2021-04-29 | 2022-10-11 | 中国科学院沈阳自动化研究所 | Laser-induced breakdown spectroscopy weak supervision feature extraction method |
| CN113569664B (en) * | 2021-07-08 | 2023-12-26 | 太原理工大学 | Coal rock identification method based on acoustic-electric signal fusion |
| CN114937190B (en) * | 2022-05-31 | 2023-05-16 | 南京林业大学 | Method and system for judging seed cotton opening effectiveness |
| CN115165847A (en) * | 2022-07-07 | 2022-10-11 | 中煤科工集团上海有限公司 | Coal rock spectrum sensing device and coal mining machine comprising same |
| CN116383704A (en) * | 2023-04-17 | 2023-07-04 | 中煤科工集团上海有限公司 | LIBS single spectral line-based coal and rock identification method |
| CN116561620B (en) * | 2023-04-17 | 2024-05-03 | 中煤科工集团上海有限公司 | LIBS spectrum data processing method, device and medium based on variable projection importance |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105279379A (en) * | 2015-10-28 | 2016-01-27 | 昆明理工大学 | Terahertz spectroscopy feature extraction method based on convex combination kernel function principal component analysis |
| CN105569658A (en) * | 2015-12-12 | 2016-05-11 | 江苏师范大学 | Coalcutter coal rock distribution recognition device and method adopting terahertz imaging technology |
| CN106124435A (en) * | 2016-07-04 | 2016-11-16 | 江苏大学 | Based on visible ray, near-infrared, the rice new-old quality inspection device of Terahertz fusion spectral technique and detection method |
| CN106706552A (en) * | 2016-11-25 | 2017-05-24 | 河南工业大学 | Method for detecting wheat quality based on multiple-information integration |
| CN106778815A (en) * | 2016-11-23 | 2017-05-31 | 河南工业大学 | Wheat quality THz spectral classification methods based on DS evidence theories |
| CN107328735A (en) * | 2017-08-30 | 2017-11-07 | 浙江大学 | Rape species discrimination method based on terahertz light spectral technology |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004077885A (en) * | 2002-08-20 | 2004-03-11 | Fuji Xerox Co Ltd | Photonic crystal and its manufacturing method, and functional element |
-
2018
- 2018-04-28 CN CN201810403512.0A patent/CN108458989B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105279379A (en) * | 2015-10-28 | 2016-01-27 | 昆明理工大学 | Terahertz spectroscopy feature extraction method based on convex combination kernel function principal component analysis |
| CN105569658A (en) * | 2015-12-12 | 2016-05-11 | 江苏师范大学 | Coalcutter coal rock distribution recognition device and method adopting terahertz imaging technology |
| CN106124435A (en) * | 2016-07-04 | 2016-11-16 | 江苏大学 | Based on visible ray, near-infrared, the rice new-old quality inspection device of Terahertz fusion spectral technique and detection method |
| CN106778815A (en) * | 2016-11-23 | 2017-05-31 | 河南工业大学 | Wheat quality THz spectral classification methods based on DS evidence theories |
| CN106706552A (en) * | 2016-11-25 | 2017-05-24 | 河南工业大学 | Method for detecting wheat quality based on multiple-information integration |
| CN107328735A (en) * | 2017-08-30 | 2017-11-07 | 浙江大学 | Rape species discrimination method based on terahertz light spectral technology |
Non-Patent Citations (4)
| Title |
|---|
| Characterization and Classification of Coals and Rocks Using Terahertz Time-Domain Spectroscopy;Xin Wang 等;《J Infrared Milli Terahz Waves》;20160928;第38卷;第248-260页 * |
| 基于太赫兹光谱技术的煤岩识别方法;王昕 等;《煤矿开采》;20180228;第23卷(第1期);第13-17页 * |
| 基于太赫兹时域光谱技术材料参数提取方法的研究;季琲琲;《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》;20180315(第03期);第30-53页 * |
| 基于电磁波技术的煤岩识别方法研究;王昕;《中国博士学位论文全文数据库 工程科技Ⅰ辑》;20180115(第01期);第53-94页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108458989A (en) | 2018-08-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108458989B (en) | Terahertz multi-parameter spectrum-based coal rock identification method | |
| CN108279217B (en) | Coal rock distinguishing method based on terahertz time-domain spectroscopy | |
| CN107727592B (en) | Coal rock interface identification method based on high spectral reflectance characteristics of coal rock | |
| Burton et al. | Lidar intensity as a remote sensor of rock properties | |
| CN108801934A (en) | A kind of modeling method of soil organic carbon EO-1 hyperion prediction model | |
| CN105938099A (en) | Rock character judging method and system based on laser-induced breakdown spectroscopy | |
| WO2018170035A1 (en) | Collaborative sensing and prediction of source rock properties | |
| CN102004088B (en) | Method for measuring coal property on line based on neural network | |
| Mahdavi et al. | Application of the local maximum synchrosqueezing transform for seismic data | |
| CN110967311B (en) | Porphyry deposit alteration zone identification method based on infrared spectrum and magnetic susceptibility measurement | |
| CN104050252A (en) | Hyperspectral remote sensing alteration information extracting method | |
| Acosta et al. | Drill-core hyperspectral and geochemical data integration in a superpixel-based machine learning framework | |
| CN104134074B (en) | A kind of Coal-rock identification method based on laser scanning | |
| CN114646682A (en) | Ore finding method based on trace elements of green cord stone | |
| Walton et al. | An approach for automated lithological classification of point clouds | |
| CN106199725A (en) | A kind of coal petrography thickness prediction method and device based on positive amplitude summation attribute | |
| CN106707355A (en) | Automatic lithology identification method of sedimentary rock | |
| Yang et al. | Research on CNN Coal and Rock Recognition Method Based on Hyperspectral Data | |
| CN105784633A (en) | Terahertz time-domain spectroscopy based method for discriminating mineral evolution | |
| CN105067556A (en) | Optical device and optical method for quickly detecting biological liquid sample | |
| Germay et al. | High-resolution core data and machine learning schemes applied to rock facies classification | |
| Lim et al. | Machine learning derived AVO analysis on marine 3D seismic data over gas reservoirs near South Korea | |
| Liu et al. | Advances in automatic identification of road subsurface distress using ground penetrating radar: State of the art and future trends | |
| Wang et al. | Rocks/minerals information extraction from EO-1 Hyperion data base on SVM | |
| Beriro et al. | Infrared spectroscopy and chemometric modelling of organic carbon measured by Rock-Eval pyrolysis of UK shale rock |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |