CN116595409B - Coal rock identification method based on principal component analysis, electronic equipment and medium - Google Patents
Coal rock identification method based on principal component analysis, electronic equipment and medium Download PDFInfo
- Publication number
- CN116595409B CN116595409B CN202310409157.9A CN202310409157A CN116595409B CN 116595409 B CN116595409 B CN 116595409B CN 202310409157 A CN202310409157 A CN 202310409157A CN 116595409 B CN116595409 B CN 116595409B
- Authority
- CN
- China
- Prior art keywords
- coal
- rock
- sample
- libs
- sample set
- 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
- 239000003245 coal Substances 0.000 title claims abstract description 451
- 239000011435 rock Substances 0.000 title claims abstract description 306
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000000513 principal component analysis Methods 0.000 title claims abstract description 22
- 238000013145 classification model Methods 0.000 claims abstract description 48
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000002536 laser-induced breakdown spectroscopy Methods 0.000 claims description 202
- 230000003595 spectral effect Effects 0.000 claims description 193
- 238000001228 spectrum Methods 0.000 claims description 192
- 238000005070 sampling Methods 0.000 claims description 99
- 238000003062 neural network model Methods 0.000 claims description 12
- 229910052717 sulfur Inorganic materials 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 239000000523 sample Substances 0.000 claims 24
- 239000000758 substrate Substances 0.000 claims 2
- 238000005520 cutting process Methods 0.000 abstract description 82
- 238000005065 mining Methods 0.000 description 55
- 238000004458 analytical method Methods 0.000 description 33
- 230000000694 effects Effects 0.000 description 26
- 238000012360 testing method Methods 0.000 description 21
- 230000008859 change Effects 0.000 description 20
- 238000004422 calculation algorithm Methods 0.000 description 18
- 239000000126 substance Substances 0.000 description 17
- 238000005516 engineering process Methods 0.000 description 16
- 238000013528 artificial neural network Methods 0.000 description 15
- 238000009826 distribution Methods 0.000 description 13
- 230000001276 controlling effect Effects 0.000 description 11
- 230000008569 process Effects 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 8
- 230000001419 dependent effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000002159 abnormal effect Effects 0.000 description 6
- 230000009467 reduction Effects 0.000 description 6
- 230000001788 irregular Effects 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 229910052757 nitrogen Inorganic materials 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 239000011777 magnesium Substances 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 238000000585 Mann–Whitney U test Methods 0.000 description 3
- 239000002817 coal dust Substances 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000003631 expected effect Effects 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000007619 statistical method Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000007704 transition Effects 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- RZVAJINKPMORJF-UHFFFAOYSA-N Acetaminophen Chemical compound CC(=O)NC1=CC=C(O)C=C1 RZVAJINKPMORJF-UHFFFAOYSA-N 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 239000003595 mist Substances 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 230000035939 shock Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 206010063385 Intellectualisation Diseases 0.000 description 1
- FYYHWMGAXLPEAU-UHFFFAOYSA-N Magnesium Chemical compound [Mg] FYYHWMGAXLPEAU-UHFFFAOYSA-N 0.000 description 1
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 1
- 238000013019 agitation Methods 0.000 description 1
- 229910052782 aluminium Inorganic materials 0.000 description 1
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000000559 atomic spectroscopy Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 239000000498 cooling water Substances 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 238000000921 elemental analysis Methods 0.000 description 1
- 238000000295 emission spectrum Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 229910052749 magnesium Inorganic materials 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052755 nonmetal Inorganic materials 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000010239 partial least squares discriminant analysis Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000001967 plate count agar Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000007873 sieving Methods 0.000 description 1
- 229910052710 silicon Inorganic materials 0.000 description 1
- 239000010703 silicon Substances 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention relates to a coal rock identification method based on principal component analysis, which comprises the following steps: step C1, acquiring a coal sample set and a rock sample set; step C2, determining the number D of principal components and D principal component extraction architectures based on all samples in a coal sample set and a rock sample set, wherein the D principal component extraction architectures have orthogonal characteristics, and the sum of corresponding interpretation degrees of the whole data of the samples is larger than a preset interpretation degree threshold; step C3, extracting D main components corresponding to each sample in the coal sample set and the rock sample set; step C4, taking D main components corresponding to each sample as input of a preset classification model, taking a coal-rock classification result as output of the preset classification model, and training to obtain a classification model; and C5, identifying the coal and rock based on the classification model. The invention greatly simplifies the complexity of sample input data, improves the accuracy of coal and rock classification, and further improves the accuracy and reliability of cutting path planning.
Description
Technical Field
The invention relates to the technical field of intelligent coal mining, in particular to a coal and rock identification method, electronic equipment and medium based on principal component analysis.
Background
The prior art has at least the following technical problems in the process of realizing the intelligent coal mining technology:
(1) The intelligent level is low, and the exploitation efficiency is low;
(2) Because laser focusing caused by irregular surfaces of coal walls of the coal mining working face is difficult, spectrum acquisition quality cannot be guaranteed, spectrum acquisition quality is low, and coal and rock identification effect is poor;
(3) The classification accuracy of the coal and rock is low, so that the definition of the coal and rock interface is inaccurate, the planning accuracy of a cutting path is low, and the reliability is poor;
(4) When the coal cutter advances based on a cutting path, the cutting motor and the traction motor are required to be finely coordinated and controlled by combining cutting parameters, and the coordinated control in the prior art is low in accuracy and poor in reliability.
Disclosure of Invention
The invention aims to provide a coal rock identification method, electronic equipment and medium based on principal component analysis, which are used for solving the technical problems in the background technology or at least partially solving the technical problems in the background technology.
In a first aspect, an embodiment of the present invention provides a coal rock identification method based on principal component analysis, including:
step C1, acquiring a coal sample set and a rock sample set, wherein each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities;
Step C2, determining the principal component quantity D based on the coal sample set and all samples in the rock sample set, and D principal component extraction architectures { z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D -wherein z (x) d Extracting architecture for D-th principal component, wherein D is in the range of 1 to D, N>>D，z(x) 2 ,…,z(x) d ,…,z(x) D With orthogonal properties, z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D The corresponding sum of interpretation degrees of the whole data of the sample is larger than a preset interpretation degree threshold value;
step C3, based on { z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D Extracting D principal components corresponding to each sample in the coal sample set and the rock sample set;
step C4, taking D main components corresponding to each sample in the coal sample set and the rock sample set as input of a preset classification model, taking a coal-rock classification result as output of the preset classification model, and training to obtain a classification model;
and C5, identifying the coal rock based on the classification model.
In a second aspect, an embodiment of the present invention provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being arranged to perform the method according to the embodiment of the first aspect of the invention.
In a third aspect, embodiments of the present invention provide a computer readable storage medium (the non-transitory computer readable storage medium storing computer instructions) for performing the method according to embodiments of the first aspect of the present invention.
According to the technical scheme, principal component analysis is performed on the coal sample set and the rock sample set, a plurality of principal components are extracted, the sum of the interpretation degrees of the whole data of the samples is larger than a preset interpretation degree threshold value, the principal components corresponding to each sample are used as input of a preset classification model, the coal-rock classification result is used as output of the preset classification model, the classification model is obtained through training, the complexity of the input data of the samples is greatly simplified, the accuracy of coal-rock classification is improved, and the accuracy and reliability of cutting path planning are further improved.
Drawings
The above and other objects, features and advantages of the present invention will become more apparent by describing embodiments of the present invention in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, and not constitute a limitation to the invention.
Fig. 1 is a schematic diagram of the basic principle of an intelligent coal mining system based on LIBS technology according to the first embodiment;
fig. 2 is a schematic diagram of different focusing modes of the intelligent sampling device according to the second embodiment;
FIG. 3 is a flow chart of a coal and rock identification method based on LIBS single spectral lines provided in the third embodiment;
Fig. 4 is a schematic diagram of the dimension reduction effect of PCA on LIBS data provided in embodiment four;
fig. 5 is a schematic diagram of coal rock identification with fixed identification resolution of a PCA coal rock identification model in embodiment five;
fig. 6 is a diagram showing a coal rock identification result of dynamic identification resolution of a PLS coal rock identification model in a fifth embodiment;
fig. 7 is an effect diagram of VIP-based selection of a target reduced spectral line wavelength in the sixth embodiment;
fig. 8 is a schematic diagram of an intelligent collaborative control subsystem of a coal mining machine in a seventh embodiment.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The Laser-induced breakdown spectroscopy (Laser-Induced Breakdown Spectroscopy, LIBS) technology is one of the atomic spectroscopy techniques, and can perform substance analysis and element detection on a sample. The principle of the technology is that a sample is irradiated by laser pulse with high energy density, high electric field breakdown is generated at a focal spot on the surface of a laser focusing area of the sample, so that 'laser plasma' close to electric neutrality is formed, and an external emission spectrum of plasma, namely a laser induced breakdown spectrum (LIBS spectrum), is formed, wherein each spectral line of the LIBS spectrum corresponds to a unique transition of a specific atom, ion or molecule. Because each element emits light with the wavelength of the dedicated spectral line, after being effectively collected by the optical detection system, the LIBS spectral information can be used as a fingerprint to identify the element composition of the sample, so that the qualitative and quantitative analysis of the elements contained in the sample is realized. The LIBS technology can effectively detect elements with low atomic numbers such as carbon (C), magnesium (Mg), aluminum (Al), silicon (Si) and the like which are very important for coal rock identification and are difficult to detect by other analysis technologies. Therefore, the LIBS technology is an ideal substance analysis technology for coal and rock identification, and the embodiment of the invention provides a plurality of embodiments based on the LIBS technology, thereby promoting the intellectualization and the unmanned operation of coal exploitation, improving the exploitation efficiency, and enabling miners to be far away from dangerous and severe working environments so as to achieve the aim of safe and efficient exploitation.
In order to improve the intelligence level and the mining efficiency of the coal mining machine, a first embodiment is provided.
Embodiment 1,
The first embodiment provides an intelligent coal mining system based on LIBS technology, which comprises an LIBS spectrum intelligent acquisition subsystem, an LIBS spectrum intelligent analysis subsystem and an intelligent coordination control subsystem.
The LIBS spectrum intelligent acquisition subsystem is used for acquiring LIBS spectrum data corresponding to a sample in a preset band interval and sending the LIBS spectrum data to the LIBS spectrum intelligent analysis subsystem. As an example, the preset band interval is [196.08nm,507.78nm ], and the reference spectral lines of [196.08nm,507.78nm ] which can contain metal and nonmetal elements playing a decisive role in coal rock analysis are selected, so that the accuracy of coal rock identification based on LIBS data is improved.
The LIBS spectrum intelligent analysis subsystem is used for generating a coal and rock identification model based on LIBS spectrum data training corresponding to the sample. The LIBS spectrum intelligent acquisition subsystem is also used for setting sampling points on a target working surface, acquiring LIBS spectrum data corresponding to the sampling points and sending the LIBS spectrum data to the coal rock identification model.
The coal rock identification model is used for generating a coal rock identification result of a sampling point based on the LIBS spectrum data corresponding to the sampling point, feeding back the coal rock identification result of the sampling point to the LIBS spectrum intelligent acquisition subsystem, generating a planned cutting path according to the coal rock identification result of the sampling point, and sending the planned cutting path to the intelligent coordinated control subsystem. It should be noted that the coal-rock identification result may be a coal-rock type, may be a classification result or a multi-classification result, where the classification result includes coal and rock, and the multi-classification result may further divide the type of coal and the type of rock, for example, the type of coal includes raw coal, clean coal, middlings, and the like; the types of rock include sandy mudstone, gangue, and the like.
The intelligent coordinated control subsystem is used for controlling coal cutting of the coal mining machine based on the planned cutting path, acquiring the operation parameters of the coal mining machine, and adjusting the cutting path in real time based on the operation parameters of the coal mining machine.
The intelligent coordinated control subsystem can effectively control the coal mining machine to run according to the planned cutting path. In addition, under the abnormal condition, the information such as the coal rock distribution, the coal cutter operation parameters and the like can be synthesized to serve as an alarm signal for cutting the rock, and the operation of the coal cutter can be automatically stopped or restored under the manual intervention.
The intelligent coal mining system can realize the effective perception of coal and rock information and the effective control of the coal mining machine based on the coal and rock perception through the cooperative cooperation of the three subsystems, and lays a foundation for the realization of a remote intelligent control platform of an unmanned coal mining working face of a coal mine.
As shown in fig. 1, the LIBS spectrum intelligent acquisition subsystem (abbreviated as an intelligent acquisition system in fig. 1) acquires LIBS spectrum data of sampling points based on actual distribution of coal and rock on a working surface, generates a coal and rock identification result of each sampling point through the LIBS spectrum intelligent analysis subsystem (abbreviated as an intelligent analysis system in fig. 1), generates a coal and rock distribution interface based on the coal and rock identification result of each sampling point, and realizes cutting path planning based on the coal and rock distribution interface. An intelligent coordination control subsystem (called as an intelligent coordination control system in figure 1 for short) controls coal mining of the coal mining machine based on cutting path planning, meanwhile, the operation parameters of the coal mining machine are obtained according to actual distribution of coal and rock of a working face, and the cutting path is adjusted in real time based on the operation parameters of the coal mining machine.
As an example, the LIBS spectrum intelligent acquisition subsystem is further configured to dynamically adjust the sampling interval according to the coal-rock identification result of the sampling point, for example, determine a sensitive area and a non-sensitive area according to the coal-rock identification result of the sampling point, where the sensitive area is an area near a coal-rock boundary, and the non-sensitive area is an area including only coal or an area including only rock. For a non-sensitive area, a larger sampling interval is set, a lower recognition resolution is maintained, and the number of sampling points is saved so as to meet the speed requirement of coal rock recognition of a coal face. The recognition resolution ratio in the sensitive area is improved, the sampling interval is reduced, and the number of sampling points is increased so as to meet the accuracy requirement of the coal cutter cutting track on coal rock recognition. And the effective utilization of the sampling points is improved through intelligent dynamic coal rock identification resolution.
The control device is used for judging whether the object distance between the substance to be sampled and the LIBS spectrum acquisition device is equal to a preset fixed object distance, if not, the LIBS spectrum acquisition device is controlled to move along the object distance direction, so that the object distance between the substance to be sampled and the LIBS spectrum acquisition device is equal to the preset fixed object distance, the LIBS spectrum intelligent acquisition subsystem can overcome laser focusing difficulty caused by irregular surfaces of coal walls, intelligent focusing of a LIBS analyzer on sampling points is realized, and LIBS spectrum acquisition quality is guaranteed.
The LIBS spectrum acquisition device is used for generating laser pulses and focusing the laser pulses on the surface of a substance to be sampled when the object distance between the substance to be sampled and the LIBS spectrum acquisition device is equal to a preset fixed object distance, the surface of the substance to be sampled is ablated to form plasma and emit a spectrum, LIBS spectrum data corresponding to the substance to be sampled is acquired, and the substance to be sampled is a sampling point on a sample or target working surface. In order to improve the accuracy and reliability of the collected LIBS spectrum data, the LIBS spectrum collection device can collect the surface of the substance to be sampled for multiple times to obtain multiple collection data, and average the multiple collection data to generate corresponding LIBS spectrum data. As a preferred example, the LIBS spectrum acquisition device has a single pulse energy of no more than 100mJ.
As an example, each sample includes N LIBS spectral data (a n ,B n ）, A n For the spectral line wavelength corresponding to the nth LIBS spectral data corresponding to the sample, A n Is positioned in the preset wave band interval, B n For sample at A n The corresponding LIBS spectrum intensity is obtained by determining the value range of N from 1 to N based on the sampling precision of the LIBS spectrum intelligent acquisition subsystem, and the preset band interval is [196.08nm,507.78nm ]N has a value of 6074.
Because the coal samples are all in the form of coal dust, the necessary simple preparation can be carried out on the samples, so that the uniform material distribution and compact structure of the samples are ensured, the phenomenon that the pulse laser causes agitation on the surfaces of the samples is avoided, and the accuracy and the stability of LIBS spectrum acquisition quality are influenced. The sample preparation process is as follows: sieving and shaking all coal dust uniformly, and tabletting the coal dust by using a die and a tablet press to prepare a coal cake sample with compact structure and flat shape; the gangue maintains the original shape, and only the surface is cleaned simply. It should be noted that samples related to other examples may be prepared in this manner.
It should be noted that, all existing methods for training and generating the coal-rock identification model fall within the scope of the present invention, and the intelligent coal mining system may adopt any existing method for training and generating the coal-rock identification model.
As an example, the LIBS spectral intelligent analysis subsystem is used toBased on LIBS spectrum data corresponding to all samples, determining a target spectrum line wavelength corresponding to a target spectrum line, wherein the LIBS spectrum intensity difference of coal and rock corresponding to the target spectrum line is larger than a preset difference threshold value, and determining a target LIBS spectrum intensity confidence interval [ BI ] based on the target spectrum line wavelength 1 ,BI 2 ]Based on [ BI ] 1 ,BI 2 ]Generating a single spectral line coal rock identification model; the LIBS spectrum intelligent acquisition subsystem is used for transmitting LIBS spectrum intensity corresponding to the target spectral line wavelength of the sampling point to the single spectral line coal rock identification model, and if the LIBS spectrum intensity corresponding to the target spectral line wavelength of the sampling point is located [ BI ] 1 ,BI 2 ]If the identification result output by the single-spectral-line coal-rock identification model is coal, otherwise, the output identification result is rock. A specific method for generating a single-line coal rock identification model is described in the third embodiment.
As an example, the LIBS spectrum intelligent analysis subsystem is configured to perform principal component analysis based on LIBS spectrum data corresponding to all samples, and determine M principal component extraction modules { W 1 ,W 2 ,…,W M }，W m The M is the M-th main component extraction module, and the value range of M is 1 to M; the LIBS spectral intelligent analysis subsystem converts the N LIBS spectral data of each sample into M principal packet component variables (C 1 ,C 2 ,…,C M ) M main group component variables (C 1 ,C 2 ,…,C M ) Taking the type corresponding to the sample as an input, training a neural network model, and generating a neural network classification model; and combining the M principal component extraction modules and the neural network classification model to generate a coal rock identification model. The LIBS spectrum intelligent analysis subsystem determines M principal component extraction modules based on a PCA algorithm, and combines the M principal component extraction modules with a neural network classification model to generate a principal component analysis (Principal Components Analysis, PCA) coal and rock identification model. The LIBS spectrum intelligent analysis subsystem determines M principal component extraction modules based on a PLS algorithm, and combines the M principal component extraction modules with a neural network classification model to generate a Partial Least Squares (PLS) coal rock identification model Type (2). Specific technical details of principal component analysis coal rock identification are described in example four. Specific technical details based on partial least squares coal rock identification are described in embodiment five.
As an example, the intelligent coordination control subsystem includes a coordination control module and a motor control module, where the coordination control module is configured to obtain the planned cutting path and the coal cutter operation parameter, generate a coordination control instruction based on the planned cutting path and the coal cutter operation parameter, and control the motor control module based on the coordination control instruction to implement control of the coal cutter cutting motor and the traction motor.
According to the embodiment, the LIBS spectrum intelligent acquisition subsystem is arranged to realize LIBS spectrum data intelligent acquisition, the coal and rock identification model is built based on the LIBS spectrum intelligent analysis subsystem, the planned cutting path is generated based on the coal and rock identification model, the intelligent cooperative control subsystem controls the operation of the cutting motor and the traction motor based on the planned cutting path and the operation parameters of the coal mining machine, and intelligent coal mining is realized through cooperative control of the LIBS spectrum intelligent acquisition subsystem, the LIBS spectrum intelligent analysis subsystem and the intelligent cooperative control subsystem, so that the intelligent coal mining degree and the mining efficiency are improved.
The coal face is generally an irregularly-fluctuated coal wall, when the coal face is detected, the parameters of laser equipment must meet the safety requirements of underground coal mine, and laser focusing difficulty caused by the change of the coal rock height of the irregularly-fluctuated coal wall surface must be overcome, so that the LIBS spectrum acquisition stability is ensured. When in-situ sampling analysis is performed on the coal wall of the coal face, the most convenient and quick method is to directly emit laser to a region to be detected of the coal face by using an online analyzer, and cover the region to be identified of the coal wall by sampling points through movement of a plane formed by an X axis and a Y axis parallel to the coal face. However, this approach typically suffers from serious measurement errors and high uncertainties. The method is mainly characterized in that the coal wall of the coal face is not a regular plane, and due to different sizes of coal blocks or rock particles and the fact that the cutting drum cuts the coal wall, the surface of the coal wall has larger irregular fluctuation, so that the object distance (Z-axis direction) between the LIBS spectrum acquisition device and the sampling point has obvious fluctuation, the LIBS analyzer has poor focusing effect on the coal and rock identification sampling point, and further the LIBS spectrum data of the sampling point is unstable and has larger errors in measurement, and the coal and rock identification effect of the coal face is seriously affected. Therefore, when the LIBS spectrum acquisition device acquires a spectrum, the LIBS spectrum acquisition device not only needs to move on a X, Y axial plane to cover a coal face, but also needs to adjust the position in the Z axial direction in real time to realize accurate focusing of a sampling point.
Embodiment II,
The second embodiment provides an intelligent LIBS spectrum acquisition subsystem, which comprises an LIBS spectrum acquisition device and a control device, wherein the LIBS spectrum acquisition device comprises a laser, an LIBS spectrum acquisition device, an acquisition track and a mobile device, the laser and the LIBS spectrum acquisition device are arranged on the mobile device, the mobile device is arranged on the acquisition track, the mobile device can move along the acquisition track according to a preset sampling interval, the acquisition track is positioned parallel to a coal face, and the mobile device can also move along a direction perpendicular to the coal face.
When the LIBS spectrum intelligent acquisition subsystem samples the sampling points on the coal face, the control device is used for judging whether the object distance between the current sampling points and the LIBS spectrum acquisition device is equal to a preset fixed object distance, if not, the LIBS spectrum acquisition device is controlled to move along the direction perpendicular to the coal face, so that the object distance between the current sampling points and the LIBS spectrum acquisition device is equal to the preset fixed object distance, and the object distance is kept stable.
The laser is used for generating laser pulses and focusing the surface of the substance to be sampled when the object distance between the substance to be sampled and the LIBS spectrum acquisition device is equal to the preset fixed object distance, and the surface of the substance to be sampled is ablated to form plasma and emit a spectrum. As an example, according to the coal mine environment safety requirement and in order to ensure certain safety redundancy, the embodiment of the invention selects a compact nanosecond solid laser, the laser parameter is the spectral line wavelength of 532nm, the laser energy is 25mJ/pulse, and the invention has the advantages of small volume, compact structure, firmness and durability and high engineering degree.
The LIBS spectrum collector is used for acquiring LIBS spectrum data corresponding to a sampling point when the object distance between a substance to be sampled and the LIBS spectrum collection device is equal to a preset fixed object distance. As an example, the LIBS spectrum collector has high resolution and wide spectrum coverage, the resolution can be set to be 0.07nm, the spectrum coverage can be set to be 0-1500nm, and the LIBS spectrum collector has a calibration system, so that the LIBS system can obtain high-quality signal-to-noise ratio; by providing a high efficiency optical laser focusing and signal collection system, lower detection limits and shorter integration times can be achieved.
According to the method, the LIBS spectrum acquisition device is used for automatically and rapidly focusing on the irregular surface of the coal wall by means of the method that the object distance is kept stable, namely, the focal length of the original LIBS spectrum acquisition device is not changed, the LIBS spectrum acquisition device is loaded on the mobile device, the change of the object distance is counteracted by means of movement of the mobile device on a track, laser emitted by a laser can be always focused on an ideal position of a sampling point, and dynamic and precise focusing on the irregular surface is achieved.
Fig. 2 shows two different focusing modes, and a LIBS spectrum acquisition device (abbreviated as an intelligent sampling device in fig. 2) is shown in a dotted line frame. States 1 to 3 represent that focusing is achieved by focusing the optical path of the LIBS spectrum collector (abbreviated as LIBS instrument in the figure). States 1 to 2 represent that the intelligent sampling device is used to stabilize the fixed object distance (object distance D in fig. 2) between the LIBS analyzer and the sampling point of the coal wall, so that the object distance is always matched with the focal length of the LIBS analyzer to realize focusing. The intelligent sampling device can realize the focusing process through the high-precision position servo control of the mobile device (namely the mobile platform in fig. 2), the device has simple structure, convenient realization, response speed and precision which are better than the zooming of the light path, and the laser energy value of the laser source does not need to be regulated, so the method for assisting in focusing by using the intelligent sampling device is more feasible.
As an example, the control device is further configured to obtain a coal rock type of the sampling point based on LIBS spectrum data corresponding to the sampling point, and dynamically adjust the sampling interval according to a change of the coal rock type of the sampling point of the continuous V identification types, where V is greater than or equal to 2. The coal rock types comprise clean coal, coal slime, raw coal, middling, sandy mudstone, gangue and the like, the gangue content corresponding to different coal rock types is different, and the sampling interval is inversely proportional to the gangue content.
The accurate movement of the mobile device is the core of maintaining a fixed object distance, and the mobile device has a plurality of uncertain complex factors, such as instability of moment of a mobile motor, change of friction moment of a track, dead zone of an element, precision of a position sensor and the like, and has great difficulty in establishing an accurate mathematical model. When the control is performed by using a traditional PID control mode, the complex factors are often not fully considered and solved, and the actual control effect is not ideal. PID control refers to proportional, integral, derivative control.
To achieve accurate movement of a mobile device, which includes, as an example, a rangefinder, BP (Back Propagation) neural network model, PID controller, and mobile motor, how to accurately control the mobile motor is a key to achieving a fixed object distance. The distance meter is used for monitoring the current distance y (k) between the sampling point and the mobile device in real time. The PID controller is used for determining a current object distance difference e (k) based on a current distance y (k) and a preset fixed object distance u (k), and inputting the current object distance difference e (k) into the BP neural network. The BP neural network is used for outputting a mobile motor proportion control parameter K based on a current object distance difference e (K) p Integral control parameter K of mobile motor i And a differential control parameter K of the mobile motor d . The PID controller is also used for obtaining the K output by the BP neural network p 、K i 、K d Based on K p 、K i 、K d Outputting the target distance u (K) and converting K p 、K i 、K d Acting on the mobile motor, controlling the mobile device to move, acquiring an actual distance v (k), and inputting the actual distance v (k) and a target distance u (k) into the BP neural network. The BP neural network model alsoModel parameters for updating the BP neural network model based on the actual distance v (k), the target distance u (k), and the object distance difference e (k). By adopting the mode of combining the BP neural network model and the PID controller, the intelligent algorithm of the BP neural network model can adjust various parameters of a motor servo control system in real time according to the operation requirement of the LIBS spectrum acquisition device, so that the negative interference of various nonlinear factors is eliminated, the whole LIBS spectrum acquisition device always operates at a good level, and the control precision of the mobile motor is greatly improved. The combination of PID control and an intelligent algorithm promotes the dynamic and accurate focusing effect of the LIBS spectrum acquisition device.
As an example, the BP neural network model is:
wherein,e (k-1) is the previous object distance difference and e (k-2) is the previous two object distance difference, which is the difference between the target distance u (k) and the actual distance v (k).
As examples, the moving motor is a servo motor, an ac motor, a synchronous motor, an asynchronous motor, or the like.
According to the embodiment II, the LIBS spectrum acquisition device and the control device are arranged, the control device dynamically controls the position of the LIBS spectrum acquisition device based on the change of the coal face, so that LIBS spectrum acquisition is always realized under the condition that the object distance between the current sampling point and the LIBS spectrum acquisition device is equal to the preset fixed object distance, the accurate focusing of the LIBS spectrum acquisition device is ensured, and the spectrum acquisition quality and the coal rock identification effect are improved.
The core technology of the LIBS spectrum intelligent analysis subsystem is how to realize coal and rock identification. Spectral lines contained in the LIBS spectrum correspond to unique transitions of a specific atom, ion or molecule, and spectral information can be used as a fingerprint to identify the elemental composition of a sample after being effectively collected by an intelligent collector of the LIBS spectrum, so that the coal-rock classification of the sample can be determined by elemental analysis. However, the LIBS spectrum of the coal has the characteristics of obvious matrix effect, self-absorption effect and the like, so that the element analysis of the coal by utilizing the LIBS spectrum is complicated in calculation all the time and has an unsatisfactory effect. Therefore, a reasonable and effective LIBS coal rock identification model needs to be established, and quick and accurate identification of the coal rock is realized. Based on this, embodiment three is further proposed.
Third embodiment,
An embodiment III provides a coal rock identification method based on LIBS single spectral lines, as shown in FIG. 3, comprising:
step S1, acquiring a coal sample set and a rock sample set, wherein each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities.
The coal sample set and the rock sample set can come from the same coal face or can come from a plurality of different faces. If the coal sample set and the rock sample set come from the same coal face, the generated single-spectral-line coal-rock identification model is only applicable to the coal face and has no universality. If the coal sample set and the rock sample set can come from a plurality of different working surfaces, the single-spectral-line coal-rock identification model generated by the group has universality. It will be appreciated that the greater the number of samples, the greater the accuracy of the single line coal rock identification model.
And S2, determining a target spectral line wavelength capable of distinguishing coal and rock remarkably based on LIBS spectral intensities of all samples in the coal sample set and the rock sample set.
In order to achieve the goal of identifying coal and rock for unknown samples, it is necessary to find a target spectral line wavelength at which there is a significant difference between the samples of the "coal" and "rock" groups, and construct a coal and rock identification model at that target spectral line wavelength.
And S3, constructing a single-spectral-line coal rock identification model based on LIBS spectrum intensity data corresponding to the target spectral line wavelengths of all samples.
And S4, acquiring LIBS spectrum intensity of a target spectral line wavelength corresponding to the sampling point on the target working surface, inputting the LIBS spectrum intensity into the single spectral line coal rock identification model, and generating a coal rock identification result of the sampling point.
As a means ofAn example, the set of coal samples is { X ] 1 ,X 2 ,…,X r ,…,X R The rock sample set is { Y } 1 ,Y 2 ,…,Y s ,…Y S (wherein X is r The value range of R is 1 to R, and R is the total amount of the coal samples; y is Y s For the sample with the S type of rock, the value range of S is 1 to S, and S is the total amount of the rock sample; x is X r =[(A 1 Xr ,B 1 Xr ),(A 2 Xr ,B 2 Xr ),…,(A n Xr ,B n Xr ),…,(A N Xr ,B N Xr )],Y s =[(A 1 Ys ,B 1 Ys ),(A 2 Ys ,B 2 Ys ),…,(A n Ys ,B n Ys ),…,(A N Ys ,B N Ys )]；(A n Xr ,B n Xr ) Is X r N-th LIBS spectral data of A n Xr Is X r Corresponding nth spectral line wavelength, B n Xr Is X r At A n Xr The corresponding LIBS spectral intensity; (A) n Ys ,B n Ys ) Is Y s N-th LIBS spectral data of A n Ys Is Y s Corresponding nth spectral line wavelength, B n Ys Is Y s At A n Ys And the value range of N is 1 to N, N is the total number of LIBS spectrum data corresponding to each sample, and N is determined based on the sampling precision of the LIBS spectrum intelligent acquisition subsystem. A is that n Xr 、A n Ys All belong to the preset wave band interval, and the preset wave band interval is [196.08nm,507.78nm ]。
As an example, the step S2 includes:
step S21, all A n Xr Corresponding B n Xr All A n Ys Corresponding B n Ys Mix and follow LIBS spectral intensityOrder of small to large, each B is generated n Xr Corresponding rank T i Xrn Each B n Ys Corresponding rank T j Ysn The value range of i is 1 to R, and the value range of j is 1 to S.
Step S22, under the wavelength of the nth spectral line, B n Ys Greater than B n Xr Number W of (2) XY Based on T i Xrn Determining coal sample rank and statistic W X Based on T j Ysn Determining rock sample rank and statistic W Y ：
W X =；
W Y =
Step S23, based on W corresponding to the nth spectral line wavelength XY 、W X 、W Y And performing rank sum test to generate a test P value corresponding to the nth spectral line wavelength, wherein the test P value is used for representing a probability value that the LIBS spectral intensities of the coal sample and the rock sample are not different under the nth spectral line wavelength.
It should be noted that, the test P value corresponding to each spectral line wavelength may be obtained specifically based on Wilcoxon rank sum test, which is not described herein.
And S24, determining the spectral line wavelength with the check P value smaller than a preset check P value threshold as a candidate spectral line wavelength, and constructing a candidate spectral line wavelength set.
And S25, selecting a target spectral line wavelength from the candidate spectral line wavelength set.
It should be noted that, when the sample size is greater than or equal to the preset sample size threshold, the accuracy of the data can be ensured, at this time, any one spectral line wavelength randomly selected from the candidate spectral line wavelength set can be used as the target spectral line wavelength, and as a preferred example, when the sample size is greater than or equal to the preset sample size threshold, in the step S25, the spectral line wavelength with the smallest test P value in the candidate spectral line wavelength set is determined as the target spectral line wavelength, and under the condition that the accuracy of the test P value can be ensured, the smaller the test P value, the larger the difference exists between the LIBS spectral intensities of the corresponding coal sample and the rock sample, and the easier the distinction between the coal and the rock is made.
When the sample size is smaller than the preset sample size threshold, the middle may be affected by other factors, so that the accuracy of the P-value of the test corresponding to each spectral line wavelength cannot be guaranteed, and therefore, the spectral line wavelength capable of distinguishing the coal sample from the rock sample may not be selected by directly randomly selecting or selecting the spectral line wavelength with the smallest P-value of the test by the P-value of the test, as an example, the step S25 includes:
step S251, obtaining a preset element spectral line mapping table, wherein the element spectral line mapping table comprises H preset element spectral line wavelengths (AZ) 1 ,AZ 2 ,…,AZ h ,…,AZ H ），AZ h Is the h preset element spectral line wavelength and AZ 1 ,AZ 2 ,…,AZ h ,…,AZ H The significance of the target spectral line wavelength, which is able to distinguish between coal and rock, decreases in turn.
Wherein the predetermined elemental spectral line wavelength (AZ 1 ,AZ 2 ,…,AZ h ,…,AZ H ) May be specified directly by the user.
Step S252, set h=1.
Step S253, AZ is judged h Belongs to the candidate spectral line wavelength set, if so, AZ is added h And if the wavelength is determined to be the target spectral line wavelength, ending the flow, and if the wavelength does not belong to the target spectral line wavelength, executing the step S254.
Step S254, if H < H, set h=h+1, and return to step S253, and if h=h, determine the spectral line wavelength with the smallest test P value in the candidate spectral line wavelength set as the target spectral line wavelength.
As an example, the last determined target spectral line wavelength is 247.95nm and the corresponding element is C. It should be noted that, when the sample size is smaller than the preset sample size threshold, the middle may be affected by other factors, so that the accuracy of the inspection P value corresponding to each spectral line wavelength cannot be ensured, and the spectral line wave corresponding to the C element may be acquiredThe test P value corresponding to the length is larger and cannot be selected as the target spectral line wavelength, and AZ can be obtained through the steps S251-S254 1 The spectral line wavelength corresponding to the element C is set as the spectral line wavelength, and the spectral line wavelength corresponding to the element C can be determined as the target spectral line wavelength as long as the spectral line wavelength corresponding to the element C is also in the candidate spectral line wavelength set. It will be appreciated that the above is merely an example, and that the spectral line wavelength corresponding to element C is not necessarily set to AZ 1 。
As an example, the step S3 includes:
step S31, obtaining a light spectrum average value of LIBS spectrum intensities corresponding to target spectral line wavelengths of all coal samples in the coal sample setAnd standard deviation of spectral intensity->。
Step S32, based onAnd->Confidence interval for determining sample type as coal>The confidence coefficient of the confidence interval is b, wherein a is a preset coefficient, and R is the total number of the coal samples in the coal sample set.
As an example, a has a value of 1.96 and b has a value of 95%.
As an example, the step S4 includes:
and S41, acquiring LIBS spectrum intensity of a target spectral line wavelength corresponding to a sampling point on a target working surface, and inputting the LIBS spectrum intensity into the single spectral line coal rock identification model.
Specifically, the LIBS spectrum intelligent acquisition subsystem in the second embodiment may be used to acquire the LIBS spectrum intensity of the target spectral line wavelength corresponding to the sampling point. After the single spectral line coal and rock identification model is established, LIBS spectrum data are not required to be collected based on the preset band interval, the band interval can be reduced, only the target spectral line wavelength is required to be contained, and the data collection amount is reduced.
And S42, if the LIBS spectrum intensity of the target spectral line wavelength corresponding to the sampling point is in the confidence interval, determining the type of the sampling point as coal, otherwise, determining the type of the sampling point as rock.
According to the embodiment, the target spectral line wavelength capable of distinguishing the target spectral line wavelength of the coal and the rock is determined, the single spectral line coal rock identification model is constructed based on LIBS spectral intensity data corresponding to the target spectral line wavelength of all samples, coal rock type identification of the working face sampling point is realized based on the single spectral line, and under the condition that the sample size is enough, the accuracy of coal rock classification is improved, the accuracy and reliability of cutting path planning generation are further improved, and the calculation amount and cost of coal rock identification are reduced.
The single-spectral-line coal rock identification model only utilizes one spectral line of a plurality of spectral lines of an original LIBS spectrum of a coal rock sample to identify, the single-spectral-line coal rock identification model contains very limited data information, analysis is often isolated, in addition, due to the influences of matrix effect, self-absorption effect, inter-element influence, plasma parameters and the like, spectral line overlapping and inter-element interference are very common in the LIBS spectrum, the information contained in one spectral line can come from certain specific element and content thereof, and can come from interference information of other factors existing in plasma and even noise, under the condition of limited sample quantity, the analysis degree of the single-spectral-line coal rock identification model on the original LIBS spectrum data information is not very ideal, and the identification accuracy of the model identification can be improved only through the increase of modeling sample quantity. And most coal face needs the coal rock identification model to have small sample size modeling capability, if good identification accuracy is to be obtained under the condition of small sample size modeling, the coal rock classification rule is found out from as much spectral line information as possible, complex relations among spectral lines caused by various influencing factors can be processed, and the accuracy of the identification model is improved. A large dataset of multiple variables would certainly provide rich information for coal and rock classification, but there may be correlation between many variables, potentially increasing redundancy of the data. In addition, the classifying regression is carried out on a plurality of input variables, the calculated amount is extremely large, and the calculation time and the hardware cost for realizing the calculation are far beyond the allowable range of the actual coal face site for identifying the coal rock. Based on this, the present invention further proposes an embodiment four.
Fourth embodiment,
An embodiment IV provides a coal rock identification method based on principal component analysis, comprising:
step C1, acquiring a coal sample set and a rock sample set, wherein each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities.
The coal sample set and the rock sample set can come from the same coal face or can come from a plurality of different faces. If the coal sample set and the rock sample set come from the same coal face, the generated classification model is only applicable to the coal face and has no universality. The group generated classification model has versatility if the coal sample set and the rock sample set can come from a plurality of different working surfaces.
Step C2, determining the principal component quantity D based on the coal sample set and all samples in the rock sample set, and D principal component extraction architectures { z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D -wherein z (x) d Extracting architecture for D-th principal component, wherein D is in the range of 1 to D, N>>D，z(x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D With orthogonal properties, z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D The sum of the corresponding interpretation degrees of the whole data of the sample is larger than a preset interpretation degree threshold value.
Wherein, through z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D To interpret as much as possible of the LIBS raw data information, each z (x) d All the main components are linear changes of all original variables, have orthogonal characteristics, and maximally retain original data in the corresponding coordinate axis directionIs a piece of information of (a). In the subsequent process of constructing the two-classification model, z (x) is considered seriously 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D The data structure is greatly simplified within the allowable range of information loss.
Step C3, based on { z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D And extracting D principal components corresponding to each sample in the coal sample set and the rock sample set.
And C4, taking D main components corresponding to each sample in the coal sample set and the rock sample set as input of a preset classification model, taking a coal-rock classification result as output of the preset classification model, and training to obtain a classification model.
It is understood that the classification model is the principal component analysis coal rock identification model mentioned in embodiment one.
And C5, identifying the coal rock based on the classification model.
As an example, the step C5 includes:
step C51, acquiring N LIBS spectrum data corresponding to sampling points on a target working surface, respectively inputting the N LIBS spectrum data into D principal component extraction frameworks, and generating D principal component variables corresponding to the sampling points;
and step C52, inputting the D main component variables corresponding to the sampling points into the classification model to generate a coal rock identification result of the sampling points.
As an example, the set of coal samples is { X ] 1 ,X 2 ,…,X r ,…,X R The rock sample set is { Y } 1 ,Y 2 ,…,Y s ,…Y S (wherein X is r The value range of R is 1 to R, and R is the total amount of the coal samples; y is Y s For the sample with the S type of rock, the value range of S is 1 to S, and S is the total amount of the rock sample; x is X r =[(A 1 Xr ,B 1 Xr ),(A 2 Xr ,B 2 Xr ),…,(A n Xr ,B n Xr ),…,(A N Xr ,B N Xr )],Y s =[(A 1 Ys ,B 1 Ys ),(A 2 Ys ,B 2 Ys ),…,(A n Ys ,B n Ys ),…,(A N Ys ,B N Ys )]；(A n Xr ,B n Xr ) Is X r N-th LIBS spectral data of A n Xr Is X r Corresponding nth spectral line wavelength, B n Xr Is X r At A n Xr The corresponding LIBS spectral intensity; (A) n Ys ,B n Ys ) Is Y s N-th LIBS spectral data of A n Ys Is Y s Corresponding nth spectral line wavelength, B n Ys Is Y s At A n Ys The corresponding LIBS spectrum intensity is obtained, the value range of N is 1 to N, and N is the total LIBS spectrum data corresponding to each sample. A is that n Xr 、A n Ys All belong to the preset band interval, which is 196.08nm,507.78nm as an example]N is determined based on the sampling precision of the LIBS spectrum intelligent acquisition subsystem, and the value of N in the example is 6074.
In the embodiment, the effective dimension reduction is performed on the LIBS spectrum data through principal component analysis, and the principal component analysis is a statistical analysis method for achieving the dimension reduction effect by reducing the correlation between variables through orthogonal transformation. PCA can effectively simplify the data set on the premise of losing the original data information as much as possible, and converts multidimensional variables possibly with correlation into a smaller number of principal component variables which are not related with each other. In this embodiment, D has a value of 3, z (x) 1 The interpretation degree of the whole data of the sample is 0.616, z (x) 2 The interpretation degree of the whole data of the sample is 0.316, z (x) 3 The interpretation degree of the whole data of the sample is 0.0286, z (x) 1 、z(x) 2 、z(x) 3 The sum is 0.9606, and the first three main components can retain 96% of the spectrum information of the whole original data. The addition of four principal components does not significantly improve the overall data interpretation, and the first three principal components are taken in view of ensuring the effect of simplifying the data structure.
Fig. 4 shows a schematic diagram of PCA dimension reduction effect. The data structure of the LIBS spectrum was reduced from 6074 dimension to 3 dimension by PCA analysis. The PCA scatter plot in fig. 4 shows that the differences between the groups of the coal sample set and the rock sample set are clearly shown in spatial distribution, which indicates that the obvious differences between the coal and the rock in the LIBS spectrum statistics characteristics are shown, the expected effect of classifying the coal and the rock based on the LIBS spectrum statistics characteristics is achieved, and the expected effect is fully shown by a PCA analysis algorithm.
As an example, step C2 includes:
step C21, constructing an initial principal component extraction architecture: z (x) =l 1 x 1 +l 2 x 2 +…l n x n …+L N x N Wherein l is n Is the coefficient corresponding to the wavelength of the nth spectral line, x n Is LIBS spectral intensity variable corresponding to the nth spectral line wavelength.
Step C22, inputting all samples in the coal sample set and the rock sample set z (x) =l 1 x 1 +l 2 x 2 +…l n x n …+L N x N Adjust l 1 、l 2 、…l n …、L N Generating group D/ 1 d 、l 2 d 、…l n d …、L N d Simultaneously satisfies: condition one, each group l 1 d 、l 2 d 、…l n d …、L N d Satisfy the following requirements。
Condition two, each group l 1 d 、l 2 d 、…l n d …、L N d Satisfy z (x) of all samples d The variance of (2) is less than a preset variance threshold from the original variance of all samples in the coal sample set and the rock sample set.
Condition three, group D l 1 d 、l 2 d 、…l n d …、L N d Satisfy the following requirementsz(x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D The sum of the corresponding interpretation degrees of the whole data of the sample is larger than a preset interpretation degree threshold value. .
Step C23, based on group D 1 d 、l 2 d 、…l n d …、L N d Generate { z (x) 1 ,z(x) 2 ,…,z(x) d ,…,z(x) D }。
As an example, the preset classification model is a two-class neural network model. The neural network model may specifically include D inputs, 2 outputs, and 2 hidden layers, where the first hidden layer includes 20 neurons and the second hidden layer includes 10 neurons.
Through Matlab program, through analyzing the PCA coal rock identification model, the fitting degree of the PCA coal rock identification model to the modeling sample coal rock two-classification is about 1, and the fitting goodness is high. And randomly selecting samples from a modeling sample library by the Matlab program as a training set to train a coal and rock identification model, and verifying an experimental result to show that all the samples are correctly identified. In addition, when the samples selected by the training set of the neural network classifier are changed each time, the self-learning effect of the neural network is also changed, and due to the limitation of the number of modeling samples, multiple modeling and identification verification are required to be carried out by continuously replacing the sample data of the training set and the test set, so that the optimal identification accuracy which can be achieved by the established coal rock identification model and the stability of the identification performance of the observed coal rock identification model are found. And the repeated recognition verification experiments are carried out by continuously changing the sample composition of the training set and the test set, so that the recognition accuracy of the PCA coal rock recognition model on the coal rock two-classification can be stably kept at 100%.
Based on a statistical analysis algorithm and a PCA coal rock identification model constructed by combining a neural network algorithm, the analysis of a single variable of one spectral line variable is changed into the analysis of multiple variables of 3 main component variables, and the quality of the variables is improved, because each main component variable is actually obtained by linear transformation of all variables of an original spectrum, and a large amount of original spectrum information is contained. And when the classification regression is performed by utilizing a plurality of principal component variables, a neural network classifier is added. The neural network has the most outstanding advantage that a specific mathematical model of an object is not required to be established, and the mapping relation from an input space to an output space can be established only by knowing input and output information of the object. The model fully utilizes the characteristics of an artificial intelligence algorithm, not only can obtain a good recognition effect, but also endows the PCA coal rock recognition model with self-learning self-adaptability, so that the model becomes intelligent, and has the capability of coping with complex geological condition changes. Compared with the method for classifying and judging by directly using 6074 variables of the original spectrum, the input signal of the neural network classifier of the PCA coal rock identification model only has 3 main component variables, so that the network structure and the operation complexity of the neural network are greatly simplified, the algorithm pressure of the classifier is reduced, the space-time expenditure of algorithm realization is greatly reduced, and the software and hardware cost of the intelligent coal rock sensing system is reduced on the model level.
In the fourth embodiment, the main component analysis is performed on the coal sample set and the rock sample set, a plurality of main components are extracted, the sum of the interpretation degree of the whole data of the samples is larger than a preset interpretation degree threshold value, the main component corresponding to each sample is used as the input of a preset classification model, the coal-rock classification result is used as the output of the preset classification model, the classification model is obtained through training, the complexity of the input data of the samples is greatly simplified, the accuracy of the classification of the coal-rock is improved, and the planning accuracy and the reliability of a cutting path are further improved.
The coal rock identification method based on the LIBS technology is essentially a point identification method, namely, the coal rock distribution of the face is described by utilizing the coal rock identification results of sampling points scattered in the area to be measured of the coal face, and the final purpose is to draw a coal rock interface planning coal cutter cutting track. For a region to be measured of a coal rock of a coal mining working face, the more sampling points are, the higher the identification resolution is, and the more accurate the coal rock interface is depicted. But the number of sampling points also affects the time for the region to finish coal rock identification, and the identification resolution and the identification speed are mutually restricted. If it is desired to describe the coal-rock interface as quickly and accurately as possible under certain conditions. The LIBS coal rock identification model needs to be further optimized, the effective utilization rate of sampling points is improved, and the identification precision is improved as much as possible under the condition that the number of the sampling points is limited.
The recognition resolution near the coal-rock interface directly determines whether the shearer will cause "missed" and "overdischarge" of the coal due to excessive sampling spacing. Obviously, the higher the recognition resolution, the better the effect of avoiding "missed" and "overdriving". While the identification resolution of other areas is less important for coal rock identification. The area near the coal-rock interface is defined as a sensitive area, and other areas are non-sensitive areas. In order to enable dynamic sampling, how to efficiently identify sensitive areas becomes critical. The higher the impurity (gangue, mudstone on top and bottom, etc.) content in the coal, the more near the coal-rock interface, forming a coal-rock gradual change region with changed coal quality. The characteristic of coal quality change in the coal-rock gradient zone is utilized as the basis for judging the sensitive zone. Therefore, the coal rock identification model is required to accurately identify the coal quality change so as to realize the coal rock characterization of dynamic identification resolution of the area to be detected of the working surface. In summary, the LIBS coal and rock classification model has the capability of recognizing the coal quality change in addition to the coal and rock classification, so that the LIBS coal and rock multi-classification recognition model needs to be further generated.
When classifying coal and rock based on PCA, coal and rock can be distinguished, and when classifying the types of coal into a plurality of types and rock into a plurality of types, the spatial distinction between different types of coal obtained based on PCA is limited, and different types of coal samples may be mixed together, which indicates that under the PCA analysis method, the data difference in the coal group samples is not ideal, so that when performing multi-classification recognition based on the PCA coal and rock recognition model of PCA principal component analysis, the accuracy is also affected, namely the PCA coal and rock recognition model is not suitable for performing multi-classification coal and rock recognition. In the PCA analysis process, the dimension reduction is mainly carried out from the angle of reducing the correlation relation between the independent variable spectral line wavelengths, and the interpretation effect of the independent variable spectral line wavelengths on the dependent variable coal and rock identification result is ignored. Noise which may cause large variation is calculated as key information into the principal component, and variables which are actually in large relation to the identification result of the variable coal rock may be ignored. This is a cause of low recognition accuracy in multi-classification recognition of the coal rock recognition model based on PCA analysis. Based on this, the present invention further proposes an embodiment five.
Fifth embodiment (V),
The invention provides a coal rock identification method based on a partial least square method, which comprises the following steps:
e1, acquiring P1 coal sample sets and Q1 rock sample sets, wherein P1 is more than or equal to 2, Q1 is more than or equal to 1, the coal rock identification types corresponding to different sample sets are different, the gangue content corresponding to the samples in different sample sets is different, each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities.
The samples can be taken from the same coal face or from different coal faces. The P1 coal sample sets comprise a clean coal sample set, a middling coal sample set and a raw coal sample set; the Q1 rock sample sets comprise a sandy mudstone sample set and a gangue sample set, and the corresponding coal rock W classification result comprises clean coal, middlings, raw coal, sandy mudstone and gangue. Raw coal refers to coal from which only visible gangue is removed, and coal which has not been subjected to any other treatment. The clean coal is high-quality coal after coal washing process. The middlings are coals with gangue content between raw coals and clean coals. The quality of the clean coal, the middling coal and the raw coal is lower and lower, and the gangue content is sequentially increased. The three samples of clean coal, middling and raw coal can simulate the coal quality change of the coal rock transition zone of the coal face. Further, the P1 coal sample sets further comprise coal slime sample sets, and the corresponding coal rock W classification results further comprise coal slime. The slime is a semi-solid substance formed from aqueous coal fines, and a slime sample was used to simulate coal wetted with water mist used to cool the shearer. Because the coal slime contains water, the components and LIBS spectra of the coal slime are different from those of normal coal, the coal slime is accurately identified, the interference of cooling water mist on coal rock identification is reduced, and the effectiveness of coal rock identification is ensured.
Step E2, determining the principal component quantity F based on all samples in the P1 coal sample set and the Q1 rock sample set, and F principal componentsComponent extraction architecture { V (x) 1 ,V(x) 2 ,…,V(x) f ,…,V(x) F -wherein V (x) f Extracting architecture for F-th principal component, wherein F is in the range of 1 to F, N>>F，V(x) 1 ,V(x) 2 ,…,V(x) f ,…,V(x) F The corresponding sum of the interpretation degrees of the whole data of the sample is larger than a preset first interpretation degree threshold value; v (x) 1 ,V(x) 2 ,…,V(x) f ,…,V(x) F The sum of the interpretation degree of the corresponding independent variable spectral line wavelength to the dependent variable coal and rock classification result is larger than a preset second junction interpretation degree threshold value.
The method comprises the steps of determining the number F of principal components by adopting a partial least square method, and extracting the architecture of the F principal components. The partial least square method is a multi-element analysis method which simplifies the data structure and simultaneously gives consideration to the correlation analysis between independent variables and dependent variables, obtains the principal components of the independent variable spectral line wavelength and the dependent variable coal and rock identification result which are orthogonal to each other by projecting the independent variable spectral line wavelength and the high-dimensional space of the dependent variable coal and rock identification result into the corresponding low-dimensional space, and then establishes the linear regression relation between the independent variable spectral line wavelength and the principal components of the dependent variable coal and rock identification result. The detailed algorithm of the partial least squares method is not described in detail herein. As an example, F has a value of 3, V (x) 1 The interpretation degree of the whole data of the sample is 0.615, V (x) 2 The interpretation degree of the whole data of the sample is 0.316, V (x) 3 The interpretation degree of the whole data of the sample is 0.028, V (x) 1 、V(x) 2 、V(x) 3 The sum is 0.959, and the first three main components can retain 95.9% of the spectrum information of the whole original data. The addition of four principal components does not significantly improve the overall data interpretation, and the first three principal components are taken in view of ensuring the effect of simplifying the data structure.
Step E3, based on { V (x) 1 ,V(x) 2 ,…,V(x) f ,…,V(x) F And F principal components corresponding to each sample in the coal sample set and the rock sample set are extracted.
The { V (x) based 1 ,V(x) 2 ,…,V(x) f ,…,V(x) F Extracting principal components can emphasize argumentsThe interpretation of spectral line wavelength to the dependent variable coal rock recognition result can also reduce noise influence, is favorable to coal rock recognition classification.
And E4, taking F main components corresponding to each sample in the coal sample set and the rock sample set as input of a preset classification model, taking a coal rock W classification result as output of the preset classification model, and training to obtain a W classification model, wherein W is more than 2.
And E5, identifying the coal rock based on the W classification model.
When the partial least square method is used for classifying the coal and the rock, no matter between the coal group sample and the rock group sample or in each group of the coal group sample and the rock group sample, the recognition result is obviously improved compared with the method for recognizing the coal and the rock based on PCA,
As an example, the step E5 includes:
and E51, acquiring N LIBS spectrum data corresponding to sampling points on a target working surface, and respectively inputting the N LIBS spectrum data into F principal component extraction frameworks to generate F principal component variables corresponding to the sampling points.
And E52, inputting F main component variables corresponding to the sampling points into the W classification model to generate a coal rock identification result of the sampling points.
As an example, the step E5 further includes:
and E53, acquiring coal and rock identification results of continuous V sampling points on the target working surface, and dynamically adjusting sampling intervals of the sampling points according to the variation of coal and rock types of the continuous V sampling points, wherein the sampling intervals are inversely proportional to the gangue content.
Based on the step E53, intelligent distribution of sampling points can be realized, dynamic identification resolution is realized for different areas, the effective utilization rate of the sampling points is improved, and the coal-rock interface can be more accurately depicted under the condition of the same number of the sampling points.
Taking a coal-rock interface randomly distributed simulated coal face physical model as an example, the model is 1.7 m long and 1.2 m wide, and coal with different coal qualities is used for simulating a coal-rock gradient zone near the coal-rock interface of the physical model. Fig. 5 shows a coal rock recognition result of a fixed recognition resolution of the PCA coal rock recognition model, and since the PCA coal rock recognition model cannot effectively recognize a coal quality change, a recognition sensitive area cannot be determined, so that a sampling point interval cannot be dynamically adjusted in a recognition process, the whole recognition process is completed with the fixed recognition resolution, the sampling point interval is 0.1 meter, and the recognition result of the sampling point is used as the recognition result of a square where the sampling point is located. A total of 204 sampling points are uniformly distributed in the area to be identified, and the distribution situation of coal and rock of the simulated coal face is depicted by the 204 square grids.
Fig. 6 shows a coal rock recognition result of dynamic recognition resolution of the PLS coal rock recognition model, and since the PLS coal rock recognition model can effectively recognize a coal quality change, after determining a recognition sensitive area, a sampling point interval can be dynamically adjusted in a recognition process to complete the whole recognition process with the dynamic recognition resolution. Sample point identification is performed on the simulated coal face model in a progressive scanning manner from left to right. And in the identification process, the sampling interval in the horizontal direction is adjusted in real time according to the change of the coal quality, when the gangue content of the coal quality is detected to be large, the sampling interval distance is shortened, and otherwise, the sampling interval is increased. The distribution of the coal and rock of the final simulated coal face is characterized by 184 square grids, and compared with 204 sampling points with fixed recognition resolution, the distribution of the coal and rock of the final simulated coal face is reduced by about 10% of the number of sampling points. As can be seen from the recognition result graph, the sampling points of the PLS coal rock recognition model are distributed more intelligently, the coal rock interface is depicted more accurately, and the possibility of overdriving and missed mining of the coal mining machine can be effectively reduced when the cutting track of the coal mining machine is planned in practice.
As an example, each sample corresponds to a spectral line wavelength belonging to the preset band interval, and the preset band interval is [196.08nm,507.78nm ].
As an example, the preset classification model is a W-classification neural network model.
In the identification verification of the modeling sample library, the 6-classification identification accuracy of the PLS coal rock identification model can be kept above 94.4 percent and can reach 100 percent at most. The recognition accuracy of the PLS coal rock recognition model in classification recognition of the coal rock 6 is better than that of the PCA coal rock recognition model, which is consistent with the advanced algorithm and the strong spectral analysis capability of PLS. The PLS coal rock identification model can realize multi-classification identification of coal rocks, has the capability of identifying coal quality changes of a coal face, can provide theoretical basis and data support for judging sensitive areas, and realizes intelligent dynamic coal rock identification resolution. The PLS coal rock identification model can be suitable for coal face with large coal quality variation range and high requirements on the accuracy of the coal rock interface characterization.
The fifth embodiment is based on the partial least square method to analyze the main components of the coal sample set and the rock sample set, extracts a plurality of main components, not only realizes the dimension reduction processing of sample input data, but also covers the interpretation effect of spectral line wavelength on the coal and rock identification result, takes the main component corresponding to each sample as the input of a preset classification model, takes the coal and rock classification result as the output of the preset classification model, trains to obtain a multi-classification model, greatly simplifies the complexity of the sample input data, improves the accuracy of coal and rock classification, and further improves the planning accuracy and reliability of cutting paths.
Although the PCA and PLS algorithms can reduce LIBS spectrum data from 6074 spectrum variables to 3 principal component variables and greatly simplify the spectrum data structure while maintaining enough spectrum information, the analysis of the PCA and PLS algorithms is completed based on a 6074×60-scale data set, and each principal component is calculated from all variable data of an original spectrum, and all spectrum data contained in an expression of the principal component is calculated. The operation process of statistical analysis and calculation of the main components is still complex, and the engineering application has high requirements on the data processing chip and the data storage space. Based on this, the present invention further proposes embodiment six.
Embodiment six,
An embodiment six provides a LIBS spectrum data processing method based on variable projection importance, including:
f1, acquiring P1 coal sample sets and Q1 rock sample sets, wherein P1 is more than or equal to 2, Q1 is more than or equal to 1, the coal rock identification types corresponding to different sample sets are different, the gangue content corresponding to the samples in different sample sets is different, each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities.
And F2, generating a weight value VIP of influence of each spectral line wavelength on a final coal-rock identification result based on LIBS spectrum data of all samples in the P1 coal sample sets and the Q1 rock sample sets.
Step F3, sorting the values of all VIPs from large to small, and using spectral line wavelengths corresponding to the values of the L VIPs before selecting as candidate spectral line wavelength sets { A } 1 VIP ,A 2 VIP ,…,A l VIP ,…,A L VIP }，A l VI For the first candidate spectral line wavelength, L has a value ranging from 1 to L, A 1 VIP ,A 2 VIP ,…,A l VIP ,…,A L VIP The corresponding VIP values decrease in sequence.
Step F4, obtaining each A l VIP Corresponding elements for any two candidate spectral line wavelengths A l1 VIP And A l2 VIP If A l1 VIP And A l2 VIP Corresponding elements are identical, and A l1 VIP VIP value of greater than A l2 VIP VIP value of (c), will a l2 VIP Removing the candidate spectral line wavelength set to generate a target reduced spectral line wavelength set, wherein the number of spectral line wavelengths in the target reduced spectral line wavelength set is far smaller than N, and the spectral line wavelengths in the target reduced spectral line wavelength set are spectral line wavelengths which play a leading role in coal and rock identification, wherein A is as follows l1 VIP And A l2 VIP All belong to candidate spectral line wavelength sets, the value of L1 is 1 to L, and the value range of L2 is 2 to L.
Before the fourth embodiment and the fifth embodiment are implemented, the original spectrum can be analyzed and optimized, and the original spectrum is simplified into a 'dominant factor reduced spectrum' which is formed by a small number of spectral lines which have a dominant effect on the coal and rock identification result, so that the data quantity and the operation quantity in the process of calculating the main component can be reduced, the software and hardware cost of coal and rock identification is reduced, and the identification speed is improved.
The partial least squares discriminant analysis (Partial Least Squairs Discriminantion Analysis, PLS-DA) is a further application based on PLS, and is a regression analysis method specially used for classification problems, and when the dependent variables are classified data, the PLS-DA analysis method can find principal components obtained by linear combination of the variables, and can analyze and obtain a VIP (Variable Importance in Projection, variable projection importance) value of the influence of each variable in the original data on the final discriminant classification. In LIBS coal rock identification spectrum data analysis, the weight of each spectral line on the final coal rock identification result can be obtained through PLS-DA analysis. PLS-DA analysis and calculation of VIP values can be performed by data analysis software such as SIMCA-P. As an example, the step F2 includes:
and F21, importing all samples in the P1 coal sample sets and the Q1 rock sample sets into SIMCA-P software, and establishing a database, wherein the database comprises PQ1 sample information records, PQ1 is the total number of samples in the P1 coal sample sets and the Q1 rock sample sets, and the sample information records comprise a coal rock identification method, electronic equipment and medium fields, sample type fields and LIBS spectrum intensity fields corresponding to N spectral line wavelengths based on principal component analysis.
And F22, performing scaling processing on the sample data in the database.
The sample data in the database can be scaled by adopting a UV data scaling mode. The manner of scaling the UV data is a manner of scaling provided by SIMCA-P software, and the UV algorithm is most suitable for small peaks with small peak areas, and if there are many small peaks in the data, the UV algorithm can be used, so the manner of scaling the UV data is selected in this embodiment.
And F23, acquiring a VIP value corresponding to each spectral line wavelength based on the SIMCA-P software and the sample data after the scaling processing.
Because the VIP value of the spectral line wavelength represents the size of the spectral line playing a role in final coal and rock identification, the VIP value is used as a reference condition for screening dominant factors, and the spectral line with large VIP value is selected, and the spectral line with small VIP value which is not beneficial to coal and rock identification is deleted.
To further improve the accuracy and reliability of the generated target reduced spectral line wavelength set, as an example, the steps F2 and F3 further include:
and F30, generating a test P value corresponding to each spectral line wavelength based on LIBS spectrum data of all samples in the P1 coal sample set and the Q1 rock sample set, wherein the test P value is used for indicating a probability value that the LIBS spectrum intensities of the coal sample and the delay sample are not different under the nth spectral line wavelength, eliminating spectral line wavelengths with the test P value being greater than or equal to a preset test P value threshold, and then executing the step F3.
As an example, in the step F30, a Wilcoxon rank sum test algorithm is used to obtain a test P value corresponding to each spectral line wavelength. The specific Wilcoxon rank sum test algorithm is not described in detail herein.
The test P value is related to its intra-group variance and inter-group fold of change, and the test P value is related only to the change in the variable itself, irrespective of the change in the overall data. The VIP value examines both the change of the variable itself and the specific gravity of the variable in the overall data, with the VIP value being greater the change and specific gravity. In actual calculation, VIP looks more at the specific gravity of the weight variable in the overall data, and then looks at the change of the substance itself. In the example, the spectral line wavelength of VIP <1 has 4500 more, but the VIP value of the spectral line wavelength with the highest VIP ranking exceeds 6, which indicates that on one hand, the raw LIBS spectrum data of coal and rock has a large number of spectral lines which are useless for recognizing coal and rock, the data redundancy is very high, the raw spectrum has a very large optimization space, on the other hand, the raw spectrum has the spectral lines which have very large importance for recognizing coal and rock, the spectral line VIP value span is very large, and the distribution characteristic of the VIP value is very favorable for screening dominant factors, so that the expected effect is completely achieved. Taking VIP value=1, for example, the check P value threshold is 0.05, some variables have higher values, which may result in VIP greater than 1, but the check P value is likely to be greater than 0.05 (without significant differences). So in addition to VIP >1 as a limiting condition for screening variables for coal rock identification, the dominant factor is screened with the condition P < 0.05.
When the dominant factors are screened to construct a simplified spectrum, the actual physical meaning of the spectrum line is considered besides the screening condition, and the similar spectrum lines belonging to the similar element reference spectrum line range only take the highest VIP value so as to ensure the quality of the VIP value. For example, although the VIP value of the spectral line corresponding to element C is relatively ranked after the PLS-DA analysis in the sample library of the present study due to the difficulty in exciting the spectral line of element C at a lower laser intensity, the step F4 further includes, as an example, in consideration of the actual physical meaning that element C is a typical element contained in coal:
and F41, judging whether the VIP value corresponding to each preset spectral line wavelength in the preset spectral line wavelength set is larger than a preset VIP threshold value, and if so, executing the step F42, wherein the preset spectral line wavelength is a spectral line wavelength corresponding to a preset element which can distinguish coal from rock in a physical sense or a chemical sense and is difficult to excite.
Wherein the preset elements comprise a C element, an N element and the like. As an example, the preset VIP threshold is equal to 1.
And F42, judging whether the preset spectral line wavelength belongs to the candidate spectral line wavelength set, and if not, adding the preset spectral line wavelength into the target simplified spectral line wavelength set.
The spectral change between the wave bands can be described in more detail while the wave band range of the simplified spectrum is enlarged through the steps F41-F42, so that the data information useful for identifying the coal rock in the original LIBS spectrum of the sample is reserved to a greater extent.
As an example, the step F4 further includes:
and F5, inputting the generated coal and rock identification model based on LIBS spectral intensity corresponding to the target simplified spectral line wavelength in the sample target simplified spectral line wavelength set, and training to generate the coal and rock identification model.
The step F5 specifically may train and generate a PCA coal rock identification model by a manner described in the fourth embodiment or train and generate a PLS coal rock identification model by a manner described in the fifth embodiment, without directly acquiring a corresponding principal component extraction architecture based on N LIBS spectrum data of each sample, but generating N LIBS spectrum data into a target reduced spectral line wavelength set, acquiring a corresponding principal component extraction architecture based on LIBS spectrum intensities corresponding to the target reduced spectral line wavelengths in the sample target reduced spectral line wavelength set, and generating a principal component corresponding to each sample based on the corresponding principal component extraction architecture and the LIBS spectrum intensities corresponding to the target reduced spectral line wavelengths in the sample target reduced spectral line wavelength set, where the principal component is used as an input of a preset classification model, and the coal rock classification result is used as an output of the preset classification model to train and obtain a classification model or a W classification model, i.e., a PCA coal rock identification model or a PLS coal rock identification model. Under the condition that the calculation amount is greatly reduced, the PCA coal rock identification model or PLS coal rock identification model established by using the target simplified spectrum wavelength still has the best identification accuracy equivalent to the PCA coal rock identification model or PLS coal rock identification model established by using the original spectrum wavelength, and the main factor simplified spectrum model based on the VIP value can reasonably and effectively optimize the original spectrum.
Fig. 7 shows an effect graph of VIP-based selection of target reduced spectral line wavelengths, each initially including 6074 LIBS data. The element corresponding to the spectral line with the VIP value arranged at the fourth is Mg, and the element corresponding to the spectral line with the VIP value arranged at the second is Mg, so that the spectral line arranged at the fourth is removed, and similarly, the spectral lines arranged at the seventh, eighth and ninth positions are removed, and then the spectral line corresponding to the C element is added based on the steps F41-F42, and as can be seen from the table, the VIP value of the C element is arranged at the 1176 position, because the excitation of the spectral line of the C element is difficult when the laser intensity is lower, the VIP value of the spectral line corresponding to the C element is relatively ranked later in PLS-DA analysis in the sample library of the study, but the C element is a typical element contained in coal, and has practical physical significance, so that the spectral line corresponding to the C element is also selected, and finally 7 spectral wavelengths are selected as target reduced wavelengths.
In the sixth embodiment, through obtaining the weight VIP of each spectral line wavelength in the P1 coal sample sets and the Q1 rock sample sets on the influence of the final coal rock identification result, a target simplified spectral line wavelength set is generated based on the VIP, the spectral line wavelength in the target simplified spectral line wavelength set is the spectral line wavelength which plays a leading role in coal rock identification, so that the LIBS spectrum data structure is greatly simplified, the empty cost condition of coal rock identification based on the LIBS spectrum data can be further reduced, and the accuracy of coal rock identification can be still maintained.
Development of high-yield and high-efficiency coal mine comprehensive supporting equipment is a mainstream of coal technology development. High-power, large-section electric traction drum shearer has been widely used. The motor is one of the main power source and the main executing mechanism of the electric traction coal mining machine, and mainly comprises a cutting motor and a traction motor. The cutting motor is responsible for driving the drum of the coal mining machine to cut coal, and the traction motor is responsible for driving the whole machine of the coal mining machine to move along the traction direction of the working face. The reliability and effectiveness of the motor driving control are related to the effect of the actual working of the coal mining machine, and the effect of planning and finishing the cutting path of the information sensing system such as intelligent sensing of the coal and the rock is guaranteed.
Embodiment seven,
The seventh embodiment provides a coal cutter intelligent cooperative control subsystem, as shown in fig. 8, which comprises a cooperative control module and a motor control module, wherein the cooperative control module is used for acquiring a planned cutting path and also used for acquiring coal cutter operation parameters in real time, and when the coal cutter operation parameters meet preset requirements, a first cutting motor control instruction and a first traction motor control instruction are generated based on the planned cutting path, and the first cutting motor control instruction and the first traction motor control instruction are sent to the motor control module.
The planned cutting path may be generated based on the coal-rock identification results of the third embodiment, the fourth embodiment and the fifth embodiment, and the acquisition of the LIBS data of the sampling point may also be realized through the LIBS spectrum intelligent acquisition subsystem described in the second embodiment, and specific technical details are described in the foregoing embodiments and are not repeated herein. As an example, the coal cutter operation parameters may include current, voltage and vibration values, and it is understood that as the gangue content increases, the current, voltage and vibration values of the harvester increase.
When the operation parameters of the coal mining machine do not meet the preset requirements, a second cutting motor control instruction and a second traction motor control instruction are generated based on the operation parameters of the current coal mining machine, the second cutting motor control instruction and the second traction motor control instruction are sent to the motor control module, and the priorities of the second cutting motor control instruction and the second traction motor control instruction are higher than those of the first cutting motor control instruction and the first traction motor control instruction.
The motor control module is used for controlling the traction motor and the cutting motor of the coal mining machine to operate based on the first cutting motor control instruction and the first traction motor control instruction under the condition that the second cutting motor control instruction and the second traction motor control instruction are not received, and executing coal mining operation according to the planned cutting path. And under the condition that the second cutting motor control instruction and the second traction motor control instruction are received, controlling the traction motor and the cutting motor of the coal mining machine to operate based on the second cutting motor control instruction and the second traction motor control instruction, and executing coal mining operation based on the current operation parameters of the coal mining machine. It can be understood that under the condition that the second cutting motor control instruction and the second traction motor control instruction are not received, the coal cutter can be explained to normally operate based on the planned cutting path, at this time, the coal cutter can directly execute the coal mining operation according to the planned cutting path, when the second cutting motor control instruction and the second traction motor control instruction are received, the explanation is that the planned cutting path is inaccurate, the acquisition machine cuts the rock or is about to cut the rock, or the coal cutter itself is abnormal, at this time, the second cutting motor control instruction and the second traction motor control instruction need to be generated based on the running parameters of the current coal cutter, and the coal cutter is enabled to execute the coal mining operation based on the second cutting motor control instruction and the second traction motor control instruction.
The intelligent coordinated control subsystem of the coal mining machine realizes the coordinated control of the cutting motor and the traction motor based on the comprehensive coal rock sensing information of the planned cutting path and the running parameters of the coal mining machine, and can timely slow down or even stop cutting and running when the cutting pick is about to cut rock and the cutting pick is already cutting the rock, and restart and speed-regulating running when the cutting pick is recovered to be normal. When other emergency abnormal conditions occur, the remote monitoring personnel can also perform manual intervention through the coordination control system.
As an example, the coordination control module is configured to determine that an operation parameter of the current coal mining machine meets a preset requirement when the current is less than or equal to a preset current threshold, the current voltage is less than or equal to a preset voltage threshold, and the current vibration value is less than or equal to a preset vibration threshold, and generate a first cutting motor control instruction and a first traction motor control instruction based on the planned cutting path.
As an example, the coordination control module is configured to, if at least one of the following occurs: if the current is greater than a preset current threshold, the current voltage is greater than a preset voltage threshold or the current vibration value is greater than a preset vibration threshold, determining that the operation parameters of the current coal mining machine do not meet the preset requirements, and generating a second cutting motor control instruction and a second traction motor control instruction based on at least one of the difference value between the current and the preset current threshold, the difference value between the current voltage and the preset voltage threshold or the difference value between the current vibration value and the preset vibration threshold. It should be noted that, as the gangue content in the mining area of the coal mining machine increases, the current, voltage and vibration values may increase. When the coal mining machine itself is abnormal or the environmental factors such as temperature are abnormal, the current, voltage and vibration value may be increased.
And controlling the operation of a traction motor and a cutting motor of the coal mining machine based on the first cutting motor control instruction and the first traction motor control instruction, or controlling the operation of the traction motor and the cutting motor of the coal mining machine based on the second cutting motor control instruction and the second traction motor control instruction, wherein the method specifically comprises the steps of controlling the starting and stopping of the traction motor and/or the cutting motor. Controlling the frequency of the traction motor and/or the cutting motor, and controlling the forward rotation and the overturning of the traction motor and/or the cutting motor; the rotational speed of the traction motor and/or the cutting motor is controlled, it being understood that the amount of parameters specifically controlled is determined in accordance with the planned cutting path and the operating parameters of the shearer.
As an example, if the current, voltage, and shock values all return to within the corresponding normal ranges, the coordination control module generates a first cutting motor control command and a first traction motor control command based on the planned cutting path. It will be appreciated that when the current, voltage and shock values all return to within the corresponding normal ranges, the abnormal condition is declared to disappear, or a planned cutting path with a large error has been bypassed, so that travel based on the planned cutting path can continue.
In order to meet the requirements of high yield, high efficiency and rapid cutting under different geological conditions of a fully mechanized coal face, the coal mining machine for various coal seam thicknesses is continuously increased in installed power and power supply voltage, and as an example, the motor control module is a three-level converter, and the three-level converter is a back-to-back NPC type three-level control system and can be used for driving and controlling a 1000kW high-power asynchronous motor for underground coal mine production with 1140V voltage level.
As an example, the three-level converter includes a core control board and an interface board, the core control board being connected to the interface board; the core control board comprises a DSP module and an FPGA module, the DSP module is used for realizing the communication of a main control program and a user interface of a traction motor or a cutting motor, the FPGA module is used for controlling logic coupling and pulse signal control, and AD chip selection, IO control and PWM signal output are completed based on instructions sent by the DSP module.
As an example, the DSP module includes a system control unit for implementing system control for implementing AD sampling, system protection, external I/O input output sampling, coordinate transformation, space Vector Pulse Width Modulation (SVPWM) modulation, and proportional and integral regulation (PI regulation) in an external interrupt of the DSP, an operation state monitoring unit, and a communication unit. The working state monitoring unit is used for monitoring the input of control words by a user and completing the switching of the state machine of the frequency converter system according to the related control words. The communication unit is used for receiving the instructions of the upper computer and the human-computer interface and finishing the setting of the control word of the system.
According to the technical scheme, the cutting motor and the traction motor of the coal mining machine are coordinately controlled to execute coal mining operation based on the planned cutting path and the operation parameters of the coal mining machine, so that the accuracy and the reliability of coordinative control of the coal mining machine are improved.
It should be noted that the same or related technical features between the format examples of the present invention may be used in combination with each other, which is not described herein in detail. In addition, the steps described in the third, fourth, fifth and sixth embodiments of the method of the present invention may be performed in a different order and/or performed in parallel. Furthermore, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the invention is not limited in this respect.
The embodiment of the invention also provides electronic equipment, which comprises: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being arranged to perform the method according to embodiments three, four, five, six of the present invention.
The embodiment of the invention also provides a computer readable storage medium, and the computer instructions are used for executing the methods described in the third embodiment, the fourth embodiment, the fifth embodiment and the sixth embodiment of the invention.
The above description is only illustrative of the preferred embodiments of the present invention and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the disclosure referred to in the present invention is not limited to the specific combinations of technical features described above, but also covers other technical features formed by any combination of the technical features described above or their equivalents without departing from the spirit of the disclosure. Such as the above-mentioned features and the technical features disclosed in the present invention (but not limited to) having similar functions are replaced with each other.
Claims (9)
1. The coal and rock identification method based on principal component analysis is characterized by comprising the following steps of:
step C1, acquiring a coal sample set and a rock sample set, wherein each sample comprises N LIBS spectrum data, and the LIBS spectrum data comprises spectral line wavelengths and corresponding LIBS spectrum intensities;
step C2, determining the number D of principal components based on all samples in the coal sample set and the rock sample set, and determining D principal component extraction architectures { z (x) 1, z (x) 2, …, z (x) D, …, z (x) D }, wherein z (x) D is the D-th principal component extraction architecture, the D has a value ranging from 1 to D, N > > D, z (x) 2, …, z (x) D, …, z (x) D has orthogonal characteristics between z (x) 1, z (x) 2, …, z (x) D, …, and the sum of the solutions of the z (x) D to the sample overall data is greater than a preset threshold of the interpretation degree;
Step C3, extracting D principal components corresponding to each sample in the coal sample set and the rock sample set based on { z (x) 1, z (x) 2, …, z (x) D, …, z (x) D };
step C4, taking D main components corresponding to each sample in the coal sample set and the rock sample set as input of a preset classification model, taking a coal-rock classification result as output of the preset classification model, and training to obtain a classification model;
step C5, carrying out coal rock identification based on the classification model;
step C2 includes:
step C21, constructing an initial principal component extraction architecture: z (x) =l1x1+l2x2+ … LNxN … +lnxn, where ln is the coefficient corresponding to the nth spectral line wavelength and xn is the LIBS spectral intensity variable corresponding to the nth spectral line wavelength;
step C22, inputting all samples in the coal sample set and the rock sample set z (x) =l1x1+l2x2+ … LNxN … +lnxn, adjusting l1, l2, … LN …, LN, generating D groups l1D, l 2D, … LN D …, LN D, satisfying simultaneously: condition one, each group l1d, l2 d, … LN d …, LN d meets；
The difference value between the variances of z (x) d of all samples and the original variances of all samples in the coal sample set and the rock sample set is smaller than a preset variance threshold value, wherein the variances of z (x) d of all samples are met by each group l1d, l2 d, … LN d …;
The third condition, the D groups l1D, l 2D, … LN D …, LN D, satisfies that the sum of the interpretation degrees of the whole data of the sample corresponding to z (x) 1, z (x) 2, …, z (x) D, …, z (x) D is larger than a preset interpretation degree threshold;
step C23, generate { z (x) 1, z (x) 2, …, z (x) D, …, z (x) D } based on D groups l1D, l 2D, … LN D …, LN D.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises,
the step C5 includes:
step C51, acquiring N LIBS spectrum data corresponding to sampling points on a target working surface, respectively inputting the N LIBS spectrum data into D principal component extraction frameworks, and generating D principal component variables corresponding to the sampling points;
and step C52, inputting the D main component variables corresponding to the sampling points into the classification model to generate a coal rock identification result of the sampling points.
3. The method of claim 1, wherein the step of determining the position of the substrate comprises,
the coal sample set is { X1, X2, …, xr, …, XR }, the rock sample set is { Y1, Y2, …, ys, … YS }, wherein Xr is the sample of the type R of coal, the value range of R is 1 to R, and R is the total amount of the coal samples; ys is the sample of which the S type is rock, the value range of S is 1 to S, and S is the total amount of the rock sample; xr= [ (A1 Xr, B1 Xr), (A2 Xr, B2 Xr), …, (AnXr, bnXr), …, (AnXr, bnXr) ], ys= [ (A1 Ys, B1 Ys), (A2 Ys, B2 Ys), …, (AnYs, bnYs), …, (AnYs, bnYs) ]; (AnXr, bnXr) is the n LIBS spectral data of Xr, anXr is the n spectral line wavelength corresponding to Xr, and BnXr is the LIBS spectral intensity corresponding to Xr on AnXr; (AnYs, bnYs) is the nth LIBS spectrum data of Ys, anYs is the nth spectral line wavelength corresponding to Ys, bnYs is the LIBS spectrum intensity corresponding to Ys on AnYs, N is the total LIBS spectrum data corresponding to each sample, and the value range of N is 1 to N.
4. The method of claim 3, wherein the step of,
n has a value of 6074.
5. The method of claim 3, wherein the step of,
d has a value of 3, and the preset interpretation threshold value is 96%.
6. The method of claim 5, wherein the step of determining the position of the probe is performed,
the preset classification model is a classification neural network model.
7. The method of claim 6, wherein the step of providing the first layer comprises,
the neural network model includes D inputs, 2 outputs, and 2 hidden layers, where a first hidden layer includes 20 neurons and a second hidden layer includes 10 neurons.
8. An electronic device, comprising:
at least one processor;
and a memory communicatively coupled to the at least one processor;
wherein the memory stores instructions executable by the at least one processor, the instructions being arranged to perform the method of any of the preceding claims 1-7.
9. A computer readable storage medium, characterized in that computer executable instructions are stored for performing the method of any of the preceding claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310409157.9A CN116595409B (en) | 2023-04-17 | 2023-04-17 | Coal rock identification method based on principal component analysis, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310409157.9A CN116595409B (en) | 2023-04-17 | 2023-04-17 | Coal rock identification method based on principal component analysis, electronic equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116595409A CN116595409A (en) | 2023-08-15 |
| CN116595409B true CN116595409B (en) | 2024-01-09 |
Family
ID=87599817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310409157.9A Active CN116595409B (en) | 2023-04-17 | 2023-04-17 | Coal rock identification method based on principal component analysis, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595409B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2532790A1 (en) * | 2003-07-18 | 2005-01-27 | Bio-Rad Laboratories, Inc. | System and method for multi-analyte detection |
| WO2008103865A2 (en) * | 2007-02-23 | 2008-08-28 | Thermo Niton Analyzers Llc | Fast and precise time-resolved spectroscopy with linear sensor array |
| CN103076310A (en) * | 2012-12-28 | 2013-05-01 | 深圳大学 | Spectrum detection system for material component analysis and detection method thereof |
| CN103488874A (en) * | 2013-09-01 | 2014-01-01 | 西北大学 | Improved support vector machine-LIBS (laser-induced breakdown spectroscopy) combined sorting method for steel materials |
| IL236114A (en) * | 2014-12-07 | 2016-04-21 | Yoav Grauer | Object detection enhancement of reflection-based imaging unit |
| WO2016090439A1 (en) * | 2014-12-09 | 2016-06-16 | Universidade Estadual De Campinas - Unicamp | Method for detecting the brightness of fingerprints using convolutional networks |
| CN106780266A (en) * | 2017-02-06 | 2017-05-31 | 重庆大学 | The analysis of accident focus internal characteristic and traveling guide method based on principal component contribution degree parameter |
| CN107764773A (en) * | 2017-10-13 | 2018-03-06 | 吉林大学 | A kind of plastic sample sorting technique based on LIBS |
| CN109872061A (en) * | 2019-01-30 | 2019-06-11 | 深圳供电局有限公司 | Power grid infrastructure improvement and promotion decision-making method |
| KR20190084479A (en) * | 2018-01-08 | 2019-07-17 | 주식회사 엘지화학 | Analysis method for metal component of battery negative electrode surface |
| CN111461243A (en) * | 2020-04-08 | 2020-07-28 | 中国医学科学院肿瘤医院 | Classification method, classification device, electronic equipment and computer-readable storage medium |
| CN114219531A (en) * | 2021-12-15 | 2022-03-22 | 北京工业大学 | Waste mobile phone dynamic pricing method based on M-WU concept drift detection |
| CN114782763A (en) * | 2022-05-19 | 2022-07-22 | 哈尔滨工业大学 | Sparse principal component alignment method for multi-view high-resolution remote sensing image |
| CN115375941A (en) * | 2022-08-22 | 2022-11-22 | 海南大学 | Multi-feature fusion hyperspectral image classification method based on GAT and 3D-CNN |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0700450D0 (en) * | 2007-01-10 | 2007-02-21 | Radiation Watch Ltd | The RWL threat engine |
| US20190285611A1 (en) * | 2015-07-30 | 2019-09-19 | The Research Foundation For The State University Of New York | Gender and race identification from body fluid traces using spectroscopic analysis |
| US20210256538A1 (en) * | 2020-02-14 | 2021-08-19 | Actimize Ltd. | Computer Methods and Systems for Dimensionality Reduction in Conjunction with Spectral Clustering of Financial or Other Data |
| JP2023020671A (en) * | 2021-07-30 | 2023-02-09 | 株式会社キーエンス | Laser-induced breakdown spectroscope |
-
2023
- 2023-04-17 CN CN202310409157.9A patent/CN116595409B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2532790A1 (en) * | 2003-07-18 | 2005-01-27 | Bio-Rad Laboratories, Inc. | System and method for multi-analyte detection |
| WO2008103865A2 (en) * | 2007-02-23 | 2008-08-28 | Thermo Niton Analyzers Llc | Fast and precise time-resolved spectroscopy with linear sensor array |
| CN103076310A (en) * | 2012-12-28 | 2013-05-01 | 深圳大学 | Spectrum detection system for material component analysis and detection method thereof |
| CN103488874A (en) * | 2013-09-01 | 2014-01-01 | 西北大学 | Improved support vector machine-LIBS (laser-induced breakdown spectroscopy) combined sorting method for steel materials |
| IL236114A (en) * | 2014-12-07 | 2016-04-21 | Yoav Grauer | Object detection enhancement of reflection-based imaging unit |
| WO2016090439A1 (en) * | 2014-12-09 | 2016-06-16 | Universidade Estadual De Campinas - Unicamp | Method for detecting the brightness of fingerprints using convolutional networks |
| CN106780266A (en) * | 2017-02-06 | 2017-05-31 | 重庆大学 | The analysis of accident focus internal characteristic and traveling guide method based on principal component contribution degree parameter |
| CN107764773A (en) * | 2017-10-13 | 2018-03-06 | 吉林大学 | A kind of plastic sample sorting technique based on LIBS |
| KR20190084479A (en) * | 2018-01-08 | 2019-07-17 | 주식회사 엘지화학 | Analysis method for metal component of battery negative electrode surface |
| CN109872061A (en) * | 2019-01-30 | 2019-06-11 | 深圳供电局有限公司 | Power grid infrastructure improvement and promotion decision-making method |
| CN111461243A (en) * | 2020-04-08 | 2020-07-28 | 中国医学科学院肿瘤医院 | Classification method, classification device, electronic equipment and computer-readable storage medium |
| CN114219531A (en) * | 2021-12-15 | 2022-03-22 | 北京工业大学 | Waste mobile phone dynamic pricing method based on M-WU concept drift detection |
| CN114782763A (en) * | 2022-05-19 | 2022-07-22 | 哈尔滨工业大学 | Sparse principal component alignment method for multi-view high-resolution remote sensing image |
| CN115375941A (en) * | 2022-08-22 | 2022-11-22 | 海南大学 | Multi-feature fusion hyperspectral image classification method based on GAT and 3D-CNN |
Non-Patent Citations (3)
| Title |
|---|
| Band Selection Using Improved Sparse Subspace Clustering for Hyperspectral Imagery Classification;Weiwei Sun;《IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing》;全文 * |
| 基于空谱一体化的农田高光谱图像分类;苗荣慧;黄锋华;杨华;邓雪峰;陈晓倩;;江苏农业学报(第04期);全文 * |
| 激光纳米粒子粒径、浓度、成分三合一在线监测仪原理介绍;朱晓阳;邹伟栋;;分析仪器(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116595409A (en) | 2023-08-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116383704B (en) | LIBS single spectral line-based coal and rock identification method | |
| CN116561620B (en) | LIBS spectrum data processing method, device and medium based on variable projection importance | |
| CN111504980B (en) | LIBS online monitoring device and method for metal additive manufacturing process | |
| CN110135468A (en) | A kind of recognition methods of gangue | |
| CN105378219B (en) | Wellbore depth interval is characterized from landwaste | |
| CN110795510A (en) | Spacecraft system health state evaluation method based on high-dimensional data association mining | |
| CN109396576A (en) | Stability of EDM and power consumption state Optimal Decision-making system and decision-making technique based on deep learning | |
| Wei et al. | Tool wear state recognition based on feature selection method with whitening variational mode decomposition | |
| CN116307123B (en) | Knowledge graph driving-based mineral resource prediction method and storage medium | |
| CN116595409B (en) | Coal rock identification method based on principal component analysis, electronic equipment and medium | |
| CN116401584A (en) | Coal rock identification method based on partial least square method, electronic equipment and medium | |
| CN116498316A (en) | Intelligent coal mining system based on LIBS technology | |
| CN116804365A (en) | Intelligent coordinated control subsystem of coal mining machine | |
| CN116539527A (en) | LIBS spectrum intelligent acquisition subsystem | |
| CN117541847B (en) | Detection method and system for tunneling environment | |
| Huang et al. | Key Technologies of Intelligentized Welding Manufacturing | |
| CN117079753A (en) | Prediction method for shear modulus and bulk modulus based on interpretability and Matminer feature generator | |
| WO2024217364A1 (en) | Coal-rock identification method based on libs single spectral line | |
| WO2024217362A1 (en) | Intelligent coal mining system based on libs technology | |
| WO2024217367A1 (en) | Coal and rock identification method based on partial least squares, electronic device, and medium | |
| WO2024217180A1 (en) | Coal and rock identification method based on principal component analysis, electronic device, and medium | |
| WO2024217368A1 (en) | Libs data processing method and device based on variable importance in projection, and medium | |
| WO2024217363A1 (en) | Libs spectrum intelligent acquisition subsystem | |
| Yan et al. | Transformer fault diagnosis research based on LIF technology and IAO optimization of LightGBM | |
| CN109765878A (en) | The aided analysis method of new-energy automobile CAN bus network failure |
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 |