CN107341521A - A kind of method based on coal spectroscopic data to grade of coal - Google Patents

A kind of method based on coal spectroscopic data to grade of coal Download PDF

Info

Publication number
CN107341521A
CN107341521A CN201710557742.8A CN201710557742A CN107341521A CN 107341521 A CN107341521 A CN 107341521A CN 201710557742 A CN201710557742 A CN 201710557742A CN 107341521 A CN107341521 A CN 107341521A
Authority
CN
China
Prior art keywords
coal
spectroscopic data
grade
neutral net
learning machine
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.)
Pending
Application number
CN201710557742.8A
Other languages
Chinese (zh)
Inventor
黎霸俊
肖冬
毛亚纯
宋亮
刘善军
何大阔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeastern University China
Original Assignee
Northeastern University China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northeastern University China filed Critical Northeastern University China
Priority to CN201710557742.8A priority Critical patent/CN107341521A/en
Publication of CN107341521A publication Critical patent/CN107341521A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features

Abstract

The present invention relates to a kind of method based on coal spectroscopic data to grade of coal, comprise the following steps：The spectroscopic data of multiple coal samples to be sorted is obtained, wherein, the spectroscopic data of each coal sample includes s spectral signature；The spectroscopic data is handled using grade of coal model, obtain categories of coal corresponding to the spectroscopic data, wherein, what the grade of coal model was established to use the spectroscopic data of multiple coal samples known to categories of coal in advance, the grade of coal model is convolutional neural networks extreme learning machine neutral net, the grade of coal model has the optimal weights matrix after particle cluster algorithm optimizes and optimal bias vector, and the grade of coal model is used to handle the spectroscopic data for including s spectral signature to obtain categories of coal corresponding to spectroscopic data.Method provided by the invention based on coal spectroscopic data to grade of coal, efficiency high, cost are low and precision is higher.

Description

A kind of method based on coal spectroscopic data to grade of coal
Technical field
The present invention relates to computer technology, particularly a kind of method based on coal spectroscopic data to grade of coal.
Background technology
With industrial expansion, the coal of high-quality plays a decisive role to production efficiency and environmental pollution, so as to improve Requirement to grade of coal.Traditional grade of coal method has two kinds.First, manual sort's method is utilized, although this method is classified Speed, but nicety of grading is often relatively low.Second, classified using the method for chemical analysis, although its precision is higher, The shortcomings that cost is high, time-consuming be present in this method.
Drawbacks described above is that those skilled in the art it is expected to overcome.
The content of the invention
(1) technical problems to be solved
In order to solve the above mentioned problem of prior art, the present invention provide it is a kind of based on coal spectroscopic data to grade of coal Method.Method provided by the invention based on coal spectroscopic data to grade of coal, efficiency high, cost are low and precision is higher.
(2) technical scheme
In order to achieve the above object, the main technical schemes that the present invention uses include：
A kind of method based on coal spectroscopic data to grade of coal, comprise the following steps：
The spectroscopic data of multiple coal samples to be sorted is obtained, wherein, the spectroscopic data of each coal sample includes s Individual spectral signature；
The spectroscopic data is handled using grade of coal model, obtains coal product corresponding to the spectroscopic data Kind,
Wherein, the grade of coal model is to be built in advance using the spectroscopic data of multiple coal samples known to categories of coal Vertical, the grade of coal model is convolutional neural networks-extreme learning machine neutral net, and the grade of coal model has warp The optimal weights matrix and optimal bias vector crossed after particle cluster algorithm optimization, the grade of coal model, which is used to handle, includes s The spectroscopic data of individual spectral signature is to obtain categories of coal corresponding to spectroscopic data.
Further, described method, before the use grade of coal model is handled the spectroscopic data, also Including：
Using the spectroscopic data of multiple coal samples known to categories of coal, using particle cluster algorithm, foundation is based on convolution The grade of coal model of neutral net-extreme learning machine neutral net.
Specifically, in described method, the spectroscopic data using multiple coal samples known to categories of coal, use Particle cluster algorithm, the grade of coal model based on convolutional neural networks-extreme learning machine neutral net is established, including：
The spectroscopic data of multiple coal samples known to categories of coal is obtained, wherein, the spectroscopic data of each coal sample Include s spectral signature, the quantity of categories of coal is t, and s, t are the integer more than 1；
The parameter of convolutional neural networks-extreme learning machine neutral net is set, wherein, the node number for setting input layer is S, the node for setting the output layer of convolutional neural networks-extreme learning machine neutral net is t；
Using the spectroscopic data of multiple coal samples known to the categories of coal as the convolutional neural networks-limit The input data of machine neural network is practised, using the coal sample of multiple coal samples known to the categories of coal accordingly as institute The output data of convolutional neural networks-extreme learning machine neutral net is stated, the convolution god is obtained using particle swarm optimization algorithm Optimal weights matrix and optimal bias vector through network-extreme learning machine neutral net；
The weight matrix for setting the convolutional neural networks-extreme learning machine neutral net is the optimal weights matrix, The bias vector for setting the convolutional neural networks-extreme learning machine neutral net is the optimal bias vector, obtained institute It is to be used to handle the spectroscopic data for including s spectral signature to obtain to state convolutional neural networks-extreme learning machine neutral net The grade of coal model of categories of coal corresponding to spectroscopic data.
Specifically, in described method, learn obtaining the convolutional neural networks-limit using particle swarm optimization algorithm When the optimal weights matrix of machine neural network and optimal bias vector, in each round iteration, inertia is obtained according to the first formula Weights omega, first formula are：
ω=rand (α, β), wherein, α=0.41, β=0.69, rand (α, β) are to seek the random number between [α, β].
Specifically, in described method, in each round iteration, current iteration wheel sequence number N is obtainedi, and according to second Formula obtains particle accelerator coefficient Ci1, population accelerator coefficient C is obtained according to the 3rd formulai2；
Second formula is：Wherein, C10=0.5, C11=2.5；
3rd formula is：Wherein, C20=3.5, C21=-3, NmaxTo be set in advance Greatest iteration wheel number.
Specifically, in described method, in the convolutional neural networks-extreme learning machine neutral net, convolutional layer swashs Function living is Sigmoid functions, and the sampling function of sample level is average sampling function.
Specifically, in described method, in the convolutional neural networks-extreme learning machine neutral net, extreme learning machine Activation primitive in neutral net is Sigmoid functions.
It should be appreciated that described method, in addition to the step of spectroscopic data of acquisition coal sample：
It is milled after carrying out cleaning treatment to coal sample, obtains coal dust sample strip；
Multiple spectrum test is carried out to coal dust sample strip, and by after the spectroscopic data progress arithmetic average of multiple spectrum test As the spectroscopic data of the coal sample, wherein, the spectral region in the spectrum test includes visible and near infrared light light Spectrum, the spectroscopic data include s spectral signature.
Specifically, in described method, categories of coal corresponding to the coal sample is any one of following：Anthracite Class, rich coal class, coking coal class and lignite class.
(3) beneficial effect
The beneficial effects of the invention are as follows：The method based on coal spectroscopic data to grade of coal of the present invention, using process There is convolutional neural networks-extreme learning machine neutral net pair of optimal weights matrix and optimal bias vector after particle group optimizing The spectroscopic data of coal sample is handled, and obtains that assortment result efficiency high, cost corresponding to coal sample be low and precision It is higher.
Brief description of the drawings
Fig. 1 is the schematic diagram based on coal spectroscopic data to the method for grade of coal of the embodiment of the present invention；
Fig. 2 is for the embodiment of the present invention based on coal spectroscopic data to establishing grade of coal model in the method for grade of coal Method schematic diagram；
Fig. 3 is the grade of coal data of test data in one embodiment of the invention and dissipating for categories of coal actual classification data Point distribution map.
Embodiment
In order to preferably explain the present invention, in order to understand, below in conjunction with the accompanying drawings, by embodiment, to this hair It is bright to be described in detail.
The method based on coal spectroscopic data to grade of coal of the embodiment of the present invention, comprises the following steps：
S1：The spectroscopic data of multiple coal samples to be sorted is obtained, wherein, wrapped in the spectroscopic data of each coal sample Include s spectral signature；
S2：The spectroscopic data is handled using grade of coal model, obtains coal corresponding to the spectroscopic data Kind,
Wherein, the grade of coal model is to be built in advance using the spectroscopic data of multiple coal samples known to categories of coal Vertical, the grade of coal model is convolutional neural networks-extreme learning machine neutral net, and the grade of coal model has warp The optimal weights matrix and optimal bias vector crossed after particle cluster algorithm optimization, the grade of coal model, which is used to handle, includes s The spectroscopic data of individual spectral signature is to obtain categories of coal corresponding to spectroscopic data.
Further, described method, before the use grade of coal model is handled the spectroscopic data, also Including：
Using the spectroscopic data of multiple coal samples known to categories of coal, using particle cluster algorithm, foundation is based on convolution The grade of coal model of neutral net-extreme learning machine neutral net.
It should be appreciated that establish the step of the grade of coal model based on convolutional neural networks-extreme learning machine neutral net Suddenly should be before being handled using grade of coal model the spectroscopic data.But obtain multiple coals to be sorted The spectroscopic data of sample can establish the step of the grade of coal model based on convolutional neural networks-extreme learning machine neutral net Before rapid, or afterwards.
Specifically, the spectroscopic data for obtaining multiple coal samples to be sorted is based on convolutional neural networks-limit in foundation Before the step of learning the grade of coal model of machine neural network, it is adapted in the modelling phase.Specifically, by the coal of acquisition Charcoal sample, its corresponding spectroscopic data is obtained using the parameter of indifference.It is and these coal samples are uniform according to categories of coal A part is selected as the training sample for establishing model in ground, the verification sample using remaining sample as checking model.
Specifically, the spectroscopic data for obtaining multiple coal samples to be sorted is based on convolutional neural networks-limit in foundation After the step of learning the grade of coal model of machine neural network, it is adapted to and carries out coal product using model after model is established The stage of kind prediction.Specifically, the grade of coal model based on convolutional neural networks-extreme learning machine neutral net is being established Afterwards, the coal sample unknown to the kind of acquisition, it is corresponding using its is obtained with training sample and checking sample identical parameter Spectroscopic data.Using the spectroscopic data of these prediction samples of the grade of coal model treatment of foundation, the product of prediction sample are obtained Kind classification.
By foregoing training step and verification step, predict that the accuracy rate of the assortment of sample is higher, can meet Produce reality needs.Specifically, convolutional neural networks (Convolutional Neural Networks, hereinafter referred to as CNN) are main Will be by convolutional layer and sample level come extracting object feature, and train the minimum for finding network global using gradient descent method Error.Convolutional neural networks can extract more preferable characteristic of division, but its perceptron is not a good grader.
Extreme learning machine (Extreme Learning Machine, hereinafter referred to as ELM) neutral net is that a list implies The feedforward neural network of layer, it the advantage is that network training speed is very fast, nicety of grading is high, Generalization Capability is good.The but limit Study machine neural network could obtain high accuracy rate on the premise of only having feature good enough in training data.
Generally in extreme learning machine neutral net, weight matrix and bias vector are given at random, cause part to be weighed Value and deviation are not up to optimum state, so the result exported each time has bigger difference.
And particle group optimizing (Particle Swarm Optimization, hereinafter referred to as PSO) algorithm is a kind of global excellent Change algorithm.Therefore, convolutional neural networks, extreme learning machine neutral net and particle swarm optimization algorithm are combined, overcomes it It is respective the shortcomings that, its advantage is made full use of, using particle cluster algorithm to convolutional neural networks-extreme learning machine neutral net Weight matrix and bias vector carry out global optimization, the convolutional Neural with optimal weights matrix and optimal bias vector of acquisition Network-extreme learning machine neutral net can be used for processing spectroscopic data to obtain the coal of categories of coal corresponding to spectroscopic data point Class model.
Specifically, in described method, the spectroscopic data using multiple coal samples known to categories of coal, use Particle cluster algorithm, the grade of coal model based on convolutional neural networks-extreme learning machine neutral net is established, including：
The spectroscopic data of multiple coal samples known to categories of coal is obtained, wherein, the spectroscopic data of each coal sample Include s spectral signature, the quantity of categories of coal is t, and s, t are the integer more than 1；
The parameter of convolutional neural networks-extreme learning machine neutral net is set, wherein, the node number for setting input layer is S, the node for setting the output layer of convolutional neural networks-extreme learning machine neutral net is t；
Using the spectroscopic data of multiple coal samples known to the categories of coal as the convolutional neural networks-limit The input data of machine neural network is practised, using the coal sample of multiple coal samples known to the categories of coal accordingly as institute The output data of convolutional neural networks-extreme learning machine neutral net is stated, the convolution god is obtained using particle swarm optimization algorithm Optimal weights matrix and optimal bias vector through network-extreme learning machine neutral net；
The weight matrix for setting the convolutional neural networks-extreme learning machine neutral net is the optimal weights matrix, The bias vector for setting the convolutional neural networks-extreme learning machine neutral net is the optimal bias vector, obtained institute It is to be used to handle the spectroscopic data for including s spectral signature to obtain to state convolutional neural networks-extreme learning machine neutral net The grade of coal model of categories of coal corresponding to spectroscopic data.
It should be appreciated that the node number for the input layer established is s, the node of output layer is t based on convolutional Neural net The grade of coal model of network-extreme learning machine neutral net to the spectroscopic data of each coal sample suitable for including s light Spectrum signature, and the quantity of corresponding possible categories of coal is the classification of t coal sample.Once the spectrum of each coal sample The spectral signature quantity that data include changes, or the quantity of corresponding possible categories of coal changes, then needs Re-establish with change after input layer node number, have change after output layer node based on convolutional Neural The grade of coal model of network-extreme learning machine neutral net.
Convolutional neural networks are described as follows：
Convolutional neural networks include convolutional layer and sample level using data as matrix form as input, hidden layer, last defeated It is an one-dimensional data to go out layer.If l layers are convolutional layers, then j-th of Feature Mapping is：
Here, MjIt is the set of input data；F is nonlinear function such as Sigmoid functions, ReLU functions, Softplus Function etc.；The convolution kernel of j-th of data of i-th of data and l layers for l-1 layers；For biasing.
If l layers are sample levels, then j-th of Feature Mapping is：
Here, DO () is sampling function, has optional such as the stochastical sampling of many different sampling function schemes, averages, most Big value etc.；For weights；For biasing.
If l layers are output layer, then corresponding to j-th of Feature Mapping is：
V is the output vector of last layer；For weights；For biasing.
Extreme learning machine neutral net is described as follows：
For the sample (x that sample size is Ni,ti), wherein xi=[xi1,xi2,…,xin]T∈Rn, ti=[ti1,ti2,…, tim]T∈Rm, then the Single hidden layer feedforward neural networks of L hidden layer neuron of standard are：
Here, ai=[ai1,ai2,…,ain]T∈RnBetween expression i-th of neuron of input layer and hidden layer Input weights；βiTo export weights；biFor biasing；ai·xjRepresent aiAnd xjInner product.
Then ELM object function can be expressed as in another way：
H β=T (5)
Here：
H is the matrix of the hidden layer output of neutral net；T is desired output.
The algorithm of Huang professors is random selection input weights and hidden layer deviation, trains this network structure suitable In the least square solution for solving linear system H β=T
Minimize:||Hβ-T|| (8)
The minimum value of the last professor Huang verified least-square solution of linear system is：
Here：H+It is H Moore-Penrose generalized inverse matrix, and the minimum value of H β=T least square solution is only One.
Particle cluster algorithm is described as follows：
Population is to be inspired by a kind of global optimization approach of Kennedy and Eberhart propositions, research idea and come from fish Group and flock of birds foraging behavior.Because PSO algorithm structures are simple, easily realization, powerful etc., so being obtained in optimization problem Universal application.Particle swarm optimization algorithm implementation process is as follows：In a group, each bird is seen as a particle, other Have corresponding speed Vi=(Vi1,Vi2,...Vid)；With position Xi=(Xi1,Xi2,...,Xid) wherein i=1,2 ..., m.
In each generation, its population constantly updates position by equation below (10) and (11), to find out optimal location, And record.
Here, pBesti=(pBesti1,pBesti2,...,pBestid) it is the optimal location that particle undergoes in itself； gBesti=(gBesti1,gBesti2,...gBestid) it is the optimal location passed through in all populations, i=1,2 ..., m；K is Current iteration number；ω is inertia weight；c1,c2It is accelerator coefficient；r1,r2It is the uniform random number on [0,1].
In existing particle cluster algorithm, in every wheel iteration, inertia weight ω and accelerator coefficient c1,c2It is fixed value.This In the method for inventive embodiments, using the inertia weight and accelerator coefficient of dynamic renewal, so that Premature Convergence is more effectively avoided, Stability is increased to network.
Specifically, in described method, learn obtaining the convolutional neural networks-limit using particle swarm optimization algorithm When the optimal weights matrix of machine neural network and optimal bias vector, in each round iteration, inertia is obtained according to the first formula Weights omega, first formula are：
ω=rand (α, β), wherein, α=0.41, β=0.69, rand (α, β) are to seek the random number between [α, β].
Inventor obtains after many experiments Integrated comparative：In α values 0.41 and β values 0.69, particle group optimizing Algorithm can be effectively prevented from precocious namely restrain in advance, so as to increase the stability of network.
Specifically, in described method, in each round iteration, current iteration wheel sequence number N is obtainedi, and according to second Formula obtains particle accelerator coefficient Ci1, population accelerator coefficient C is obtained according to the 3rd formulai2；
Second formula is：Wherein, C10=0.5, C11=2.5；
3rd formula is：Wherein, C20=3.5, C21=-3, NmaxTo be set in advance Greatest iteration wheel number.
It should be noted that the particle accelerator coefficient C in the second formulai1The c being equal in above formula (10)1, the 3rd formula In population accelerator coefficient Ci2The c being equal in above formula (10)2。
Inventor obtains after many experiments Integrated comparative, in C10 values 0.5, C11 values 2.5, and C20 values During 3.5, C21 value -3, particle swarm optimization algorithm can be effectively prevented from precocious namely restrain in advance, so as to increase network Stability.
Specifically, in described method, in the convolutional neural networks-extreme learning machine neutral net, convolutional layer swashs Function living is Sigmoid functions, and the sampling function of sample level is average sampling function.
Specifically, in described method, in the convolutional neural networks-extreme learning machine neutral net, extreme learning machine Activation primitive in neutral net is Sigmoid functions.
It should be appreciated that described method, in addition to the step of spectroscopic data of acquisition coal sample：
It is milled after carrying out cleaning treatment to coal sample, obtains coal dust sample strip；
Multiple spectrum test is carried out to coal dust sample strip, and by after the spectroscopic data progress arithmetic average of multiple spectrum test As the spectroscopic data of the coal sample, wherein, the spectral region in the spectrum test includes visible and near infrared light light Spectrum, the spectroscopic data include s spectral signature.
Specifically, in described method, categories of coal corresponding to the coal sample is any one of following：Anthracite Class, rich coal class, coking coal class and lignite class.
Below in conjunction with concrete application, the embodiment of the present invention is carried out based on coal spectroscopic data to the method for grade of coal Explanation.
The coal sample ore product from Fushun, Yi Min and Henan Jia Jinkou coal fields of collection, including 31, anthracite, fertilizer Coal 57,60, coking coal and 58, lignite, share 206 samples.Using the SVC HR- of Spectra Vista companies of the U.S. 1024 portable field spectroradiometers are as laboratory apparatus.The instrument spectral scope：350-2500nm.Internal memory： 500scans (is swept).Weight：3kg, port number：1024.Spectral resolution (FWHM≤8.5nm) 1000~1850nm.Minimal product Between timesharing：1 millisecond.
Sample is cleaned first, then carries out milling processing.When obtaining the spectroscopic data of coal sample, instrument is swept The time is retouched as 1 second/time, the probe of spectrometer is away from coal dust sample strip surface 300mm, and perpendicular to sample strip surface.Testing Cheng Zhongwei reduces interference of the surrounding environment to spectrum test, it is desirable to which experiment is all completed in a sealing chamber, avoids the sun and its The irradiation of his light source, experimenter must not walk about and wear dark clothes.
Five spectrum tests are carried out to each sample using spectrometer (SVC HR-1024), then averaged as the sample The spectroscopic data of product.Carry out within every 10 minutes in experiment a blank measurement calibration..Because coal is black in itself, especially visible Most of energy in optical range is all absorbed,.Spectrum is mixed in together, therefore coal sample dulling luminosity ratio is higher, spectrum it is effective Information is few.Therefore feature is extracted by CNN, ELM is classified, and can effectively improve nicety of grading.
Specifically, the coal sample of acquisition and spectroscopic data share 206 groups, wherein choosing 120 groups as training data, 86 Group is used as test data.
For the ease of realizing digitized processing, the corresponding numeral 1 of anthracite class, the corresponding numeral 2 of rich coal class, coking coal class pair are set Should numeral 3, lignite class corresponding label 4.
Convolutional neural networks-pole with optimal weights matrix and optimal bias vector is established using 120 groups of training datas The grade of coal model of limit study machine neural network.As shown in Fig. 2 the embodiment of the present invention based on coal spectroscopic data to coal The method that grade of coal model is established in the method for classification, step are as follows：
(1) spectroscopic data is inputed to CNN neutral nets to extract feature, selects activation letters of the Sigmoid for convolutional layer Number, sample level selection average sampling function.Each spectroscopic data has 1024 spectral signatures, after feature extraction, exports number According to for 350 features.
(2) population and ELM neutral nets, PSO selections number of iterations=50, population quantity=40 are initialized；Selection Activation primitive of the Sigmoid functions as ELM neutral nets, node in hidden layer 20.
(3) spectrum characteristic data is trained using ELM algorithms, verified to obtain PSO fitness values, fitness value is carried out Judge, then preserve optimal value and PSO speed and position.
(4) optimizing is just exited when network reaches greatest iteration wheel number, convolutional neural networks-extreme learning machine god at this moment There is optimal weights matrix and optimal bias vector through network.
Using 86 groups of test datas of the grade of coal model treatment, obtained categories of coal classification results are as shown in Figure 3.
As shown in figure 3, using in 86 groups of test datas of the grade of coal model treatment, the classification knot of the first kind and the 4th class The no mistake of fruit point；The classification results of second class have 2 mistakes, and the classification results of the 3rd class have 1 mistake.The grade of coal mould When type handles 86 groups of test datas, nicety of grading is 96.51%, illustrates that the nicety of grading of the grade of coal model is higher, is met Actual use demand.
Table 1 is the throwing using artificial experience method, chemical method and the classification of CNN-ELM network models to 206 coal samples Rate are used and the time consuming table of comparisons of institute.Although it can be seen that using traditional low but other precision of artificial experience method expense not It is high.Using chemical method in addition to needing to buy experimental drug, some chemical experimental instrument input costs are more than 1,000,000. Although chemical method precision is high but cost is very high, time length.
The throwing for including spectrometer and computer to the method for grade of coal based on coal spectroscopic data of the embodiment of the present invention Enter cost within 350,000, it is less than Traditional Man method and chemical method input, nicety of grading is high, time-consuming shorter.
The different recognition methods table of comparisons of table 1
In summary, compared with traditional grade of coal, the embodiment of the present invention is divided coal based on coal spectroscopic data The method of class, use is visible, near infrared spectrum, optimizes the grade of coal model of CNN-ELM neural networks in coal using PSO Charcoal is accurate when classifying, efficiently.The embodiment of the present invention can be in economic, speed to the method for grade of coal based on coal spectroscopic data Degree, accuracy have without comparable advantage and important actual application value.
Finally it should be noted that：Above-described embodiments are merely to illustrate the technical scheme, rather than to it Limitation；Although the present invention is described in detail with reference to the foregoing embodiments, it will be understood by those within the art that： It can still modify to the technical scheme described in previous embodiment, or which part or all technical characteristic are entered Row equivalent substitution；And these modifications or substitutions, the essence of appropriate technical solution is departed from various embodiments of the present invention technical side The scope of case.

Claims (9)

  1. A kind of 1. method based on coal spectroscopic data to grade of coal, it is characterised in that comprise the following steps：
    The spectroscopic data of multiple coal samples to be sorted is obtained, wherein, the spectroscopic data of each coal sample includes s light Spectrum signature；
    The spectroscopic data is handled using grade of coal model, obtains categories of coal corresponding to the spectroscopic data,
    Wherein, the grade of coal model is established for the spectroscopic data in advance using multiple coal samples known to categories of coal , the grade of coal model is convolutional neural networks-extreme learning machine neutral net, and the grade of coal model, which has, to be passed through Optimal weights matrix and optimal bias vector after particle cluster algorithm optimization, the grade of coal model, which is used to handle, includes s The spectroscopic data of spectral signature is to obtain categories of coal corresponding to spectroscopic data.
  2. 2. according to the method for claim 1, it is characterised in that described that the spectroscopic data is entered using grade of coal model Before row processing, in addition to：
    Using the spectroscopic data of multiple coal samples known to categories of coal, using particle cluster algorithm, foundation is based on convolutional Neural The grade of coal model of network-extreme learning machine neutral net.
  3. 3. according to the method for claim 2, it is characterised in that described using multiple coal samples known to categories of coal Spectroscopic data, using particle cluster algorithm, establish the grade of coal mould based on convolutional neural networks-extreme learning machine neutral net Type, including：
    The spectroscopic data of multiple coal samples known to categories of coal is obtained, wherein, wrapped in the spectroscopic data of each coal sample S spectral signature is included, the quantity of categories of coal is t, and s, t are the integer more than 1；
    The parameter of convolutional neural networks-extreme learning machine neutral net is set, wherein, the node number for setting input layer is s, if The node for putting the output layer of convolutional neural networks-extreme learning machine neutral net is t；
    Using the spectroscopic data of multiple coal samples known to the categories of coal as the convolutional neural networks-extreme learning machine The input data of neutral net, using the coal sample of multiple coal samples known to the categories of coal accordingly as the volume The output data of product neutral net-extreme learning machine neutral net, the convolutional Neural net is obtained using particle swarm optimization algorithm The optimal weights matrix and optimal bias vector of network-extreme learning machine neutral net；
    The weight matrix for setting the convolutional neural networks-extreme learning machine neutral net is the optimal weights matrix, is set The bias vector of the convolutional neural networks-extreme learning machine neutral net is the optimal bias vector, the obtained volume Product neutral net-extreme learning machine neutral net is to be used to handle the spectroscopic data for including s spectral signature to obtain spectrum The grade of coal model of categories of coal corresponding to data.
  4. 4. according to the method for claim 3, it is characterised in that
    The optimal weights matrix of the convolutional neural networks-extreme learning machine neutral net is being obtained using particle swarm optimization algorithm During with optimal bias vector, in each round iteration, inertia weight ω is obtained according to the first formula, first formula is：
    ω=rand (α, β), wherein, α=0.41, β=0.69, rand (α, β) are to seek the random number between [α, β].
  5. 5. according to the method for claim 4, it is characterised in that
    In each round iteration, current iteration wheel sequence number N is obtainedi, and particle accelerator coefficient C is obtained according to the second formulai1, root Population accelerator coefficient C is obtained according to the 3rd formulai2；
    Second formula is：Wherein, C10=0.5, C11=2.5；
    3rd formula is：Wherein, C20=3.5, C21=-3, NmaxChanged for maximum set in advance Generation wheel number.
  6. 6. according to the method for claim 3, it is characterised in that
    In the convolutional neural networks-extreme learning machine neutral net, the activation primitive of convolutional layer is Sigmoid functions, sampling The sampling function of layer is average sampling function.
  7. 7. according to the method for claim 3, it is characterised in that
    In the convolutional neural networks-extreme learning machine neutral net, the activation primitive in extreme learning machine neutral net is Sigmoid functions.
  8. 8. according to the method for claim 1, it is characterised in that the step of also including obtaining the spectroscopic data of coal sample：
    It is milled after carrying out cleaning treatment to coal sample, obtains coal dust sample strip；
    Multiple spectrum test is carried out to coal dust sample strip, and after the spectroscopic data of multiple spectrum test is carried out into arithmetic average as The spectroscopic data of the coal sample, wherein, the spectral region in the spectrum test includes visible and near infrared light spectrum, institute Stating spectroscopic data includes s spectral signature.
  9. 9. according to the method for claim 1, it is characterised in that
    Categories of coal corresponding to the coal sample is any one of following：Anthracite class, rich coal class, coking coal class and lignite Class.
CN201710557742.8A 2017-07-10 2017-07-10 A kind of method based on coal spectroscopic data to grade of coal Pending CN107341521A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710557742.8A CN107341521A (en) 2017-07-10 2017-07-10 A kind of method based on coal spectroscopic data to grade of coal

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710557742.8A CN107341521A (en) 2017-07-10 2017-07-10 A kind of method based on coal spectroscopic data to grade of coal

Publications (1)

Publication Number Publication Date
CN107341521A true CN107341521A (en) 2017-11-10

Family

ID=60219711

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710557742.8A Pending CN107341521A (en) 2017-07-10 2017-07-10 A kind of method based on coal spectroscopic data to grade of coal

Country Status (1)

Country Link
CN (1) CN107341521A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108489912A (en) * 2018-05-11 2018-09-04 东北大学 A kind of coal constituent analysis method based on coal spectroscopic data
CN108983127A (en) * 2018-05-29 2018-12-11 东北大学 A kind of bloodstone magnetism rate detection method based on spectroscopic data
CN111582387A (en) * 2020-05-11 2020-08-25 吉林大学 Rock spectral feature fusion classification method and system
CN112040174A (en) * 2020-07-20 2020-12-04 西安科技大学 Underground coal flow visual detection method
CN115646848A (en) * 2022-10-21 2023-01-31 微山金源煤矿 Intelligent classification method and system for coal mine

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102033043A (en) * 2010-10-19 2011-04-27 浙江大学 Grain moisture content detecting method based on hyperspectral image technology
CN102590211A (en) * 2011-01-11 2012-07-18 郑州大学 Method for utilizing spectral and image characteristics to grade tobacco leaves

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102033043A (en) * 2010-10-19 2011-04-27 浙江大学 Grain moisture content detecting method based on hyperspectral image technology
CN102590211A (en) * 2011-01-11 2012-07-18 郑州大学 Method for utilizing spectral and image characteristics to grade tobacco leaves

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
ASANGA R. 等: "Self-Organizing Hierarchical Particle Swarm Optimizer With Time-Varying Acceleration Coefficients", 《IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION》 *
余丹 等: "一种卷积神经网络和极限学习机相结合的人脸识别方法", 《数据采集和处理》 *
李明 等: "基于LVQ与SVM算法的近红外光谱煤产地鉴别", 《光谱学与光谱分析》 *
王杰 等: "一种基于粒子群优化的极限学习机", 《郑州大学学报（理学版）》 *
黄少荣: "基于随机参数的粒子群优化算法", 《微电子学与计算机》 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108489912A (en) * 2018-05-11 2018-09-04 东北大学 A kind of coal constituent analysis method based on coal spectroscopic data
CN108489912B (en) * 2018-05-11 2019-08-27 东北大学 A kind of coal constituent analysis method based on coal spectroscopic data
CN108983127A (en) * 2018-05-29 2018-12-11 东北大学 A kind of bloodstone magnetism rate detection method based on spectroscopic data
CN108983127B (en) * 2018-05-29 2019-10-29 东北大学 A kind of bloodstone magnetism rate detection method based on spectroscopic data
CN111582387A (en) * 2020-05-11 2020-08-25 吉林大学 Rock spectral feature fusion classification method and system
CN112040174A (en) * 2020-07-20 2020-12-04 西安科技大学 Underground coal flow visual detection method
CN115646848A (en) * 2022-10-21 2023-01-31 微山金源煤矿 Intelligent classification method and system for coal mine

Similar Documents

Publication Publication Date Title
CN107341521A (en) A kind of method based on coal spectroscopic data to grade of coal
CN105243398B (en) The method of improvement convolutional neural networks performance based on linear discriminant analysis criterion
CN107230113A (en) A kind of house property appraisal procedure of multi-model fusion
CN111914728B (en) Hyperspectral remote sensing image semi-supervised classification method and device and storage medium
CN107044976A (en) Heavy metal content in soil analyzing and predicting method based on LIBS Yu stack RBM depth learning technologies
CN106874688A (en) Intelligent lead compound based on convolutional neural networks finds method
CN103605711B (en) Construction method and device, classification method and device of support vector machine
CN105431854B (en) Method and apparatus for analyzing biological sample
CN108108762A (en) A kind of random forest classification method based on core extreme learning machine and parallelization for the classification of coronary heart disease data
CN108596274A (en) Image classification method based on convolutional neural networks
CN112881306A (en) Hyperspectral image-based method for rapidly detecting ash content of coal
CN109030388A (en) A kind of iron ore all iron content detection method based on spectroscopic data
CN109214591B (en) Method and system for predicting aboveground biomass of woody plant
CN110082310A (en) A kind of near infrared band EO-1 hyperion diagnostic method of rubber tree LTN content
Liu et al. Estimating leaf chlorophyll contents by combining multiple spectral indices with an artificial neural network
Wang et al. Digital image processing technology under backpropagation neural network and K-Means Clustering algorithm on nitrogen utilization rate of Chinese cabbages
CN112766161A (en) Hyperspectral target detection method based on integrated constraint multi-example learning
Sugiharti et al. Integration of convolutional neural network and extreme gradient boosting for breast cancer detection
Liu et al. Estimation of Potato Above-Ground Biomass Based on Vegetation Indices and Green-Edge Parameters Obtained from UAVs
Qi et al. SAM-GAN: An improved DCGAN for rice seed viability determination using near-infrared hyperspectral imaging
Kahraman et al. Dynamic-fitness-distance-balance stochastic fractal search (dFDB-SFS algorithm): an effective metaheuristic for global optimization and accurate photovoltaic modeling
Jiang et al. Detecting dark spot eggs based on CNN GoogLeNet model
Li et al. A technique system for the measurement, reconstruction and character extraction of rice plant architecture
CN103761530B (en) Hyperspectral image unmixing method based on relevance vector machine
Lu et al. Image classification and identification for rice leaf diseases based on improved WOACW_SimpleNet

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20171110

RJ01 Rejection of invention patent application after publication