CN107632010B - Method for quantifying steel sample by combining laser-induced breakdown spectroscopy - Google Patents
Method for quantifying steel sample by combining laser-induced breakdown spectroscopy Download PDFInfo
- Publication number
- CN107632010B CN107632010B CN201710865660.XA CN201710865660A CN107632010B CN 107632010 B CN107632010 B CN 107632010B CN 201710865660 A CN201710865660 A CN 201710865660A CN 107632010 B CN107632010 B CN 107632010B
- Authority
- CN
- China
- Prior art keywords
- vector machine
- wavelet packet
- data
- laser
- induced breakdown
- 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
Images
Abstract
The invention discloses a method for quantifying a steel sample by combining laser-induced breakdown spectroscopy, which comprises the following steps: 1) collecting spectral data to form two groups of data sets; 2) extracting the characteristics of the data set by using a wavelet packet transformation method, converting the spectrum peak intensity into wavelet energy for representation, and applying the wavelet packet energy as an input end to a vector machine regression model; 3) establishing a vector machine regression model, and modeling to predict content; the rapid quantitative detection of the steel sample is realized by combining the wavelet packet transformation and an improved correlation vector machine kernel function quantitative analysis method with the laser-induced breakdown spectroscopy.
Description
Technical Field
The invention relates to the field of spectral analysis technology and the like, covers wavelet packet transformation to extract original data, and quantitatively analyzes a steel sample through an improved correlation vector machine based on laser-induced breakdown spectroscopy, in particular to a method for quantifying the steel sample based on the wavelet packet transformation and the correlation vector machine in combination with the laser-induced breakdown spectroscopy.
Background
Laser Induced Breakdown Spectroscopy (LIBS) is a technology for rapidly detecting the types and contents of elements contained in substances based on atomic emission Spectroscopy, and is widely applied to the fields of metallurgical analysis, environmental monitoring, geological exploration, medical monitoring, materials and the like. The principle is that qualitative or quantitative analysis of chemical element components in a sample is realized through the one-to-one correspondence relationship between the characteristic spectral line intensity and specific elements in the sample. The elements in the measurement sample can thus be quantitatively analyzed by the quantitative relationship therein. In addition, the LIBS has real-time multi-element analysis capability, and does not need sample pretreatment, so the LIBS has great development value in the field of quality monitoring of steel samples.
The steel industry is used as a basic industry for sustainable development, and the industrial quantity is very large. The steel is rich in varieties, and the performance and the application are different according to different element contents. However, it is difficult to correctly determine different types of steel samples by naked eyes in shape, and particularly, in steel enterprises, sales markets and warehouses of various large ports, where a large amount of steel is gathered, the steel samples are easily mislabeled and misjudged. Moreover, for the same batch of national steel samples, the difference of different manufacturers in the production and manufacturing process is the difference of the components. The traditional steel sample quality detection and analysis method is long in consumption period, complex in operation and not suitable for on-site real-time analysis requirements, so that the LIBS analysis technology provides a solution for the dilemma.
Wavelet packet transform is an extension of wavelet transform in a high-frequency signal part, and not only can well represent low-frequency signals, but also can well decompose signals containing a large amount of detailed information (high-frequency). Moreover, the decomposition is more elaborate, has no redundancy and has no omission. The decomposition is performed with a wavelet packet for the laser induced breakdown spectroscopy. The method can adaptively perform better time-frequency domain local analysis on the medium-high frequency signals in the spectrogram, namely reflect the complete information of the original signals to the maximum extent.
The relevance vector machine is a Bayes frame-based machine learning algorithm, is similar to a support vector, and converts a low-dimensional space nonlinear problem into a high-dimensional space linear problem based on kernel function mapping. However, the correlation vector machine does not need a penalty factor, does not cause the problem of over-learning, and does not need to be restricted by the Meixi theorem on the kernel function construction. Therefore, the correlation vector machine has better sparsity and shorter test time, so that the method is more suitable for online detection, and the obtained result output not only has binary output, but also obtains probability output. Therefore, the correlation vector machine is a machine learning algorithm with stronger generalization capability and robustness, and is more suitable for online monitoring of steel products.
Disclosure of Invention
The invention aims to provide a method for quantifying a steel sample by combining laser-induced breakdown spectroscopy, which realizes the rapid quantitative detection of the steel sample by combining wavelet packet transformation and an improved correlation vector machine kernel function quantitative analysis method with the laser-induced breakdown spectroscopy.
The invention is realized by the following technical scheme: a method for quantifying a steel sample by combining laser-induced breakdown spectroscopy comprises the following steps:
1) collecting spectral data to form two groups of data sets;
2) extracting the characteristics of the data set by using a wavelet packet transformation method, converting the spectrum peak intensity into wavelet energy for representation, and applying the wavelet packet energy as an input end to a vector machine regression model;
3) and establishing a vector machine regression model for modeling and content prediction.
In order to further realize the invention, the following arrangement mode is adopted: the step 1) specifically comprises the following steps:
1.1) respectively carrying out spectrum data acquisition on steel samples with different steel grades at different measurement sites by utilizing a laser-induced breakdown spectroscopy device;
1.2) randomly and averagely dividing the collected spectral data into two groups, wherein one group is used as a training set, and the other group is used as a testing set.
In order to further realize the invention, the following arrangement mode is adopted: the step 2) comprises the following specific steps:
2.1) carrying out data normalization pretreatment on the characteristic spectral lines of the training set and the characteristic spectral lines of the test set by wavelet packet transformation respectively;
2.2) carrying out three-layer decomposition on the training set characteristic spectral line and the test set characteristic spectral line by adopting dB4 as a wavelet basis function in wavelet packet transformation;
2.3) respectively obtaining the three-layer wavelet packet coefficient of the training set and the three-layer wavelet packet coefficient of the testing set after the step 2.2); the third layer comprises 8 nodes, and each node represents the characteristics of a corresponding section of frequency band signals;
2.4) extracting the sum of the squares of the wavelet packet coefficients of each node as the wavelet packet energy of the characteristic spectral line according to the law of conservation of energy;
2.5) carrying out logarithmic transformation on the extracted wavelet packet energy value, and then using the wavelet packet energy value as a spectral line intensity value as input data to be applied to a vector machine regression model.
In order to further realize the invention, the following arrangement mode is adopted: full spectrum normalization is used in the normalization preprocessing.
In order to further realize the invention, the following arrangement mode is adopted: when the three-layer decomposition is carried out, formula (1) and formula (2) are adopted:
In the formula (1) and the formula (2), j, k and n are positive integers, and h is0，h1Is a multi-resolution filter.
In order to further realize the invention, the following arrangement mode is adopted: the step 3) comprises the following specific steps:
3.1) after determining kernel functions and training set model input data, starting to establish a vector machine regression model, wherein the vector machine regression model comprises a support vector machine, a correlation vector machine and a correlation vector machine combined with an improved Laplace kernel function;
3.2) carrying out accuracy and precision analysis on the established vector machine regression model by using test set data, wherein the specific method comprises the following steps: RMSEP and RE between the predicted and true values are calculated, respectively.
In order to further realize the invention, the following arrangement mode is adopted: the improved Laplace kernel function is obtained by performing an evolution on an original Laplace kernel function.
In order to further realize the invention, the following arrangement mode is adopted: the quantitative method further comprises robustness verification: and verifying the stability of the regression model of the vector machine by using the prediction set data, and judging the robustness of the correlation vector machine combined with the improved Laplace kernel function.
In order to further realize the invention, the following arrangement mode is adopted: the quantitative method further comprises generalization capability verification: and verifying the stability of the regression model of the vector machine by utilizing the data of the prediction set, and judging the generalization capability of the correlation vector machine combined with the improved Laplace kernel function.
The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy comprises the following steps:
s1, performing multi-site test on steel samples with different steel grades by using a laser-induced breakdown spectroscopy device, and collecting and recording test data;
s2 randomly and averagely dividing the test data into two parts, wherein one part is used as a training set, and the other part is used as a test set;
s3, decomposing the training set and the test set characteristic spectral lines respectively by wavelet packet transformation, adopting dB4 as a wavelet basis function for three-layer decomposition, and extracting wavelet packet energy as input quantity of a characteristic spectral line intensity model;
s4, the correlation vector machine kernel function prototype in the invention is Laplace kernel function; squaring a prokaryotic function (Laplace kernel function) to reduce the size of the kernel scale and enlarge the linear range of the data;
s5, the improved kernel function correlation vector machine is combined with the training set to establish a regression model (the correlation vector machine is combined with the improved Laplacian kernel function), and the test set is used to verify the quality of the model to obtain a quantitative result.
Compared with the prior art, the invention has the following advantages and beneficial effects:
(1) the invention adopts wavelet packet transformation to carry out feature extraction on spectral data, retains the feature variable of the original spectrum to the maximum extent and plays a certain role in inhibiting spectral noise. And finally, the obtained wavelet packet energy is used as an input variable of the regression model, so that the data dimension is reduced, and the calculation cost is greatly reduced.
(2) After the correlation vector machine transforms the kernel function (i.e. adopts the improved laplacian kernel function), the prediction precision is obviously improved compared with the traditional vector machine. In the steel sample prediction experiment, the generalization capability and the robustness test are obviously improved.
(3) According to the invention, the wavelet packet transform is used for decomposing the characteristic spectrum in the laser-induced breakdown spectrum, so that the data dimension is reduced.
(4) The method for predicting the wavelet packet based on the Laplace kernel function includes the steps of obtaining roots of the Laplace kernel function, reducing the kernel scale, enlarging the linear range among data, and matching with the subsequent logarithmic transformation of the wavelet packet, so that accuracy of a model prediction result is effectively improved, and better generalization capability and robustness are obtained.
Drawings
FIG. 1 is a schematic diagram of the LIBS experiment.
FIG. 2 is a schematic diagram of a wavelet packet decomposition tree.
FIG. 3 is LIBS spectrum of different steel grades.
FIG. 4 is a flow chart of wavelet packet correlation vector machine quantification.
FIG. 5 is a graph of content prediction of a real-time relevance vector machine.
Detailed Description
The following examples are given to illustrate the present invention and it is necessary to point out here that the following examples are given only for the purpose of further illustration and are not to be construed as limiting the scope of the invention, which is susceptible to numerous insubstantial modifications and adaptations by those skilled in the art in light of the present disclosure.
Example 1:
a method for quantifying a steel sample by combining a laser-induced breakdown spectroscopy realizes the rapid quantitative detection of the steel sample by combining the wavelet packet transformation and an improved correlation vector machine kernel function quantitative analysis method with the laser-induced breakdown spectroscopy, and comprises the following steps:
1) collecting spectral data to form two groups of data sets;
2) extracting the characteristics of the data set by using a wavelet packet transformation method, converting the spectrum peak intensity into wavelet energy for representation, and applying the wavelet packet energy as an input end to a vector machine regression model;
3) and establishing a vector machine regression model for modeling and content prediction.
Example 2:
the embodiment is further optimized on the basis of the above embodiment, and in order to further better implement the invention, the following setting modes are particularly adopted: the step 1) specifically comprises the following steps:
1.1) respectively carrying out spectrum data acquisition on steel samples with different steel grades at different measurement sites by utilizing a laser-induced breakdown spectroscopy device;
1.2) randomly and averagely dividing the collected spectral data into two groups, wherein one group is used as a training set, and the other group is used as a testing set; the selected training set data comprises the maximum and minimum value of the content change of the whole steel sample, and the test set data does not contain the steel content related in the training set.
Example 3:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: the step 2) comprises the following specific steps:
2.1) carrying out data normalization pretreatment on the characteristic spectral lines of the training set and the characteristic spectral lines of the test set by wavelet packet transformation respectively;
2.2) carrying out three-layer decomposition on the training set characteristic spectral line and the test set characteristic spectral line by adopting dB4 as a wavelet basis function in wavelet packet transformation;
2.3) respectively obtaining the three-layer wavelet packet coefficient of the training set and the three-layer wavelet packet coefficient of the testing set after the step 2.2); the third layer comprises 8 nodes, and each node represents the characteristics of a corresponding section of frequency band signals;
2.4) extracting the sum of the squares of the wavelet packet coefficients of each node as the wavelet packet energy of the characteristic spectral line according to the law of conservation of energy;
2.5) carrying out logarithmic transformation on the extracted wavelet packet energy value, and then using the wavelet packet energy value as a spectral line intensity value as input data to be applied to a vector machine regression model.
Example 4:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: full spectrum normalization is used in the normalization preprocessing.
Example 5:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: when the three-layer decomposition is carried out, formula (1) and formula (2) are adopted:
wherein j, k and n are positive integers, h0，h1Is a multi-resolution filter.
Example 6:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: the step 3) comprises the following specific steps:
3.1) after determining kernel functions and training set model input data, starting to establish a vector machine regression model { support vector machine, correlation vector machine combined with improved Laplace kernel functions };
3.2) carrying out accuracy and precision analysis on the established vector machine regression model by using test set data, wherein the specific method comprises the following steps: RMSEP and RE between the predicted and true values are calculated, respectively.
Example 7:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: the improved Laplace kernel function is obtained by performing an evolution on an original Laplace kernel function.
Example 8:
the present embodiment is further optimized on the basis of any of the above embodiments, and in order to further better implement the present invention, the following setting modes are particularly adopted: the quantitative method further comprises robustness and generalization capability verification: and verifying the stability of the regression model of the vector machine by utilizing the data of the prediction set, and judging the robustness and the generalization capability of the correlation vector machine combined with the improved Laplace kernel function.
Example 9:
the embodiment is further optimized on the basis of any one of the embodiments, and the method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy comprises the following specific operation steps:
(1) respectively collecting and storing spectral data of different sites of the selected steel samples with different steel grades by using a laser-induced breakdown spectroscopy device;
(2) randomly dividing the collected spectral data into two types, wherein one type is used as a model training set, and the other type is used as a model testing set;
(3) the prokaryotic function of the correlation vector machine used in the invention is a Laplace kernel function;
(4) the laser induced breakdown spectrum is subjected to normalization processing by adopting a full spectrum normalization method, so that subsequent data analysis and modeling are facilitated;
(5) and extracting a characteristic peak curve of the element to be detected, carrying out wavelet packet transformation three-layer decomposition on the characteristic peak according to a formula (1) and a formula (2), and obtaining 8 nodes on the third layer by adopting dB4 as a wavelet basis function. Each node corresponds to a segment of frequency band signal characteristics. According to the energy conservation law, extracting the sum of squares of wavelet packet coefficients of all nodes as wavelet packet energy of the whole characteristic peak, wherein the log logarithm of the energy value is used as a training data input value of a regression model of a correlation vector machine;
wherein j, k and n are positive integers, h0，h1A multi-resolution filter;
the theoretical derivation of the regression model of the correlation vector machine comprises the following steps (step (6) to step (14)):
(6) for a given certain sample content C, it is assumed that there is a relationship with the characteristic peak intensity I of the laser-induced breakdown spectrum as follows:
C＝y(I，ω)+ (3)；
where it is the noise added throughout the experiment and satisfies the Gaussian distribution N (0, σ)2)；
(7) The regression function y (I, ω) can be obtained after training by a correlation vector machine in combination with a training set:
wherein, K (I, I)i) Is a kernel function, ωiA weight vector (i is 1, 2, 3, …, n, which is the number of spectral data in the training sample);
(8) derived from equations 1 and 2:
where Φ is an n × (n +1) structural matrix composed of kernel functions (laplace kernel functions), that is:
wherein each row that makes up Φ:
φ(Ii)＝[1，K(Ii，I1)，...，K(Ii，In)]T(7)；
(9) to prevent in evaluating sigma2And (3) an over-adaptation phenomenon in omega maximum likelihood estimation, so that the prior probability distribution of the autocorrelation judgment theory is defined as follows:
wherein α ═ α0，α1，...，αn]TIs a (n +1) vector consisting of the hyper-parameters;
(10) then, through Bayesian theory, (omega, α, sigma)2) The posterior probability can be solved as follows:
(11) to solve the prior distribution problem, the posterior probability distribution of formula (9) can be calculated first,
the decomposition form is as follows:
P(ω，α，σ2|C)＝P(ω|C,α,σ2)P(α,σ2|C) (10)；
where the posterior probability distribution of ω can be decomposed into:
wherein, P (C | α, σ)2)＝∫P(C|ω,σ2)P(ω|α)dω (12)，
∑＝(σ-2ΦTΦ+A)-1(13)，
μ＝σ-2∑ΦTC (14)；
Wherein the matrix A ═ diag (α)0，α1，...，αn)；
(12) Jointly solving to obtain:
P(α，σ2|C)∝P(C|α,σ2)P(α)P(σ2) (15)，
(13) thus P (α, σ)2I C) can be converted into the (13) form of the maximum likelihood estimation, i.e.
And σ2α can be iteratively solved by:
Niiis the ith diagonal element of the covariance matrix of the multiple pairs of posterior weighted values;
(14) finally for a given laser induced breakdown Spectroscopy intensity (test set) I*The predicted content value is:
it can be seen that it satisfies the Gaussian distribution of μ*＝μTφ(I*) (19)；
(15) The above steps (step (6) -step (14)) are derived from the regression model theory of the relevance vector machine, and the original Laplace kernel function is
And then, squaring the Laplace kernel function to obtain an improved Laplace kernel function
The improved Laplace kernel function can reduce the kernel scale and enlarge the data interval.
To further illustrate the operation flow of the present invention and demonstrate the generalization ability and robustness of the regression model of the relevance vector machine, the following embodiments are further illustrated with reference to the accompanying drawings and examples:
example 10:
the LIBS system involved in the present invention mainly consists of a laser (preferably a Q-switched pulsed Nd: YAG laser), a spectrometer (preferably an echelle spectrometer (German, LTB200, wavelength 200-. The laser energy is 85mJ, the repetition frequency is 5Hz, and the delay time is 2 mus.
This example uses 14 different steel grades provided by Pan Steel group, Inc: 45#, 20#, 27CrMnMo, 15CrMoG, 15NiCuMoNb, CS-Q780 and X52Q, wherein each steel grade of the seven steel samples comprises two types of samples with different contents. Each type of sample is a 6mm high steel column, is placed on a sample table after being polished and wiped, and is measured by using a laser induced breakdown spectroscopy system to obtain various types of spectral data, wherein the outline is shown in fig. 3.
4 measuring points are randomly selected on each section of seven steel samples, and a measuring spectrum is obtained by 30 times of laser pulse on each measuring point, so that 56 analysis spectra (measuring spectra) are obtained.
The spectral data of the first seven steel samples are randomly divided into two groups, one group is a training set, and the other group is a testing set. The P280GH spectral data will be used for regression model robust contrast analysis.
And carrying out full spectrum normalization on all the acquired spectrums, decomposing the characteristic peak by using wavelet packet transformation, and extracting wavelet packet energy as spectrum intensity characteristics. For comparison, the test data is subjected to different model (vector machine regression models (SVM, RVM, MRVM)) to model and predict the content, and the optimal kernel scale parameter of the vector machine is obtained by optimizing through a grid method, and a specific flow chart is shown in fig. 4. The final root mean square prediction error (RMSEP) and Relative Error (RE) are calculated.
Table one: predicted outcome RMSEP comparison
Table two: prediction result RE comparison
From the results of the two tables, it can be known that the method based on wavelet packet transformation and through improving the correlation vector machine of the kernel function (the correlation vector machine is combined with the improved laplace kernel function), which is provided by the invention, can show better generalization capability on the quantitative level of the steel sample, and has higher precision and smaller relative error on the prediction of the test set.
Example 11:
example 10 demonstrates that the improved kernel function model (correlation vector machine combined with improved laplacian kernel) has better generalization capability, and this example will further demonstrate the robustness of the improved kernel function model, and the LIBS system for data acquisition is the same as example 10. Test steels steel is also provided by Pan Steel group, Inc. as Steel No. P280 GH. For a sample of steel No. P280GH, the height is the same as 6mm, the sample is placed on a sample platform after being polished and wiped, 150 points are randomly selected, and other measurement conditions are the same, so that 150 pieces of analysis spectrum data are obtained.
The P280GH laser induced breakdown spectroscopy data was subjected to model quantitative analysis according to the flow chart shown in fig. 4, and the P280GH laser induced breakdown spectroscopy data served as prediction set data. The prediction results of the content of the P280GH sample are shown in FIG. 5 and Table III:
table three:
the result shows that the improved kernel function model has outstanding characteristics in both generalization capability and robustness test, and the accurate quantitative capability of the improved kernel function model lays a foundation for the LIBS technology to realize rapid real-time analysis on steel detection.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and all simple modifications and equivalent variations of the above embodiments according to the technical spirit of the present invention are included in the scope of the present invention.
Claims (7)
1. A method for quantifying a steel sample by combining laser-induced breakdown spectroscopy is characterized by comprising the following steps: the method comprises the following steps:
1) collecting spectral data to form two groups of data sets, and specifically comprising the following steps:
1.1) respectively carrying out spectrum data acquisition on steel samples with different steel grades at different measurement sites by utilizing a laser-induced breakdown spectroscopy device;
1.2) randomly and averagely dividing the collected spectral data into two groups, wherein one group is used as a training set, and the other group is used as a testing set;
2) the method comprises the following specific steps of extracting features of a data set by using a wavelet packet transformation method, converting the spectrum peak intensity into wavelet energy for representation, and applying the wavelet packet energy as an input end to a vector machine regression model:
2.1) carrying out data normalization pretreatment on the characteristic spectral lines of the training set and the characteristic spectral lines of the test set by wavelet packet transformation respectively;
2.2) carrying out three-layer decomposition on the training set characteristic spectral line and the test set characteristic spectral line by adopting dB4 as a wavelet basis function in wavelet packet transformation;
2.3) respectively obtaining the three-layer wavelet packet coefficient of the training set and the three-layer wavelet packet coefficient of the testing set after the step 2.2); the third layer comprises 8 nodes, and each node represents the characteristics of a corresponding section of frequency band signals;
2.4) extracting the sum of the squares of the wavelet packet coefficients of each node as the wavelet packet energy of the characteristic spectral line according to the law of conservation of energy;
2.5) carrying out logarithmic transformation on the extracted wavelet packet energy value, and taking the transformed wavelet packet energy value as a spectral line intensity value as input end data to be applied to a vector machine regression model;
3) and establishing a vector machine regression model for modeling and content prediction.
2. The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy as recited in claim 1, wherein the method comprises the following steps: full spectrum normalization is used in the normalization preprocessing.
3. The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy as claimed in claim 1, wherein the method comprises the following steps: when the three-layer decomposition is carried out, formula (1) and formula (2) are adopted:
in the formula (1) and the formula (2), j, k and n are positive integers, and h is0，h1Is a multi-resolution filter.
4. A method for quantifying steel samples in combination with laser induced breakdown spectroscopy according to claim 1 or 2 or 3, wherein: the step 3) comprises the following specific steps:
3.1) after determining kernel functions and training set model input data, starting to establish a vector machine regression model, wherein the vector machine regression model comprises a support vector machine, a correlation vector machine and a correlation vector machine combined with an improved Laplace kernel function;
3.2) carrying out accuracy and precision analysis on the established vector machine regression model by using test set data, wherein the specific method comprises the following steps: RMSEP and RE between the predicted and true values are calculated, respectively.
5. The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy as claimed in claim 4, wherein the method comprises the following steps: the improved Laplace kernel function is obtained by performing an evolution on an original Laplace kernel function.
6. The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy as claimed in claim 4, wherein the method comprises the following steps: the quantitative method further comprises robustness verification: and verifying the stability of the regression model of the vector machine by using the prediction set data, and judging the robustness of the correlation vector machine combined with the improved Laplace kernel function.
7. The method for quantifying the steel sample by combining the laser-induced breakdown spectroscopy as claimed in claim 4, wherein the method comprises the following steps: the quantitative method further comprises generalization capability verification: and verifying the stability of the regression model of the vector machine by utilizing the data of the prediction set, and judging the generalization capability of the correlation vector machine combined with the improved Laplace kernel function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710865660.XA CN107632010B (en) | 2017-09-22 | 2017-09-22 | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710865660.XA CN107632010B (en) | 2017-09-22 | 2017-09-22 | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107632010A CN107632010A (en) | 2018-01-26 |
| CN107632010B true CN107632010B (en) | 2020-09-18 |
Family
ID=61103392
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710865660.XA Active CN107632010B (en) | 2017-09-22 | 2017-09-22 | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107632010B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110579467B (en) * | 2018-06-11 | 2021-09-10 | 中国农业机械化科学研究院 | Time-resolved laser-induced breakdown spectroscopy quantification method |
| CN109060675A (en) * | 2018-09-05 | 2018-12-21 | 东北大学 | One kind is for iron content detection method and device in iron ore |
| CN109975273A (en) * | 2019-03-07 | 2019-07-05 | 四川大学 | A kind of petrographic classification method based on laser induced breakdown spectroscopy |
| CN111398254B (en) * | 2020-03-26 | 2021-09-24 | 浙江大学 | Rice leaf heavy metal copper element quantitative detection method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103983617A (en) * | 2014-05-04 | 2014-08-13 | 华中科技大学 | Improved laser probe quantitative analysis method based on wavelet transform |
| CN104697966A (en) * | 2015-03-10 | 2015-06-10 | 西北大学 | Method for quantitatively analyzing chromium and manganese in steel based on least square support vector machine algorithm combined with laser-induced breakdown spectroscopy |
| CN105938098A (en) * | 2016-07-07 | 2016-09-14 | 四川大学 | Rock soil ignition loss prediction method and system based on laser-induced breakdown spectroscopy |
| CN106770194A (en) * | 2017-01-20 | 2017-05-31 | 华中科技大学 | Cereal crops place of production discrimination method based on wavelet transformation LIBS |
-
2017
- 2017-09-22 CN CN201710865660.XA patent/CN107632010B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103983617A (en) * | 2014-05-04 | 2014-08-13 | 华中科技大学 | Improved laser probe quantitative analysis method based on wavelet transform |
| CN104697966A (en) * | 2015-03-10 | 2015-06-10 | 西北大学 | Method for quantitatively analyzing chromium and manganese in steel based on least square support vector machine algorithm combined with laser-induced breakdown spectroscopy |
| CN105938098A (en) * | 2016-07-07 | 2016-09-14 | 四川大学 | Rock soil ignition loss prediction method and system based on laser-induced breakdown spectroscopy |
| CN106770194A (en) * | 2017-01-20 | 2017-05-31 | 华中科技大学 | Cereal crops place of production discrimination method based on wavelet transformation LIBS |
Non-Patent Citations (4)
| Title |
|---|
| Laser-induced Breakdown spectroscopy quantitative analysis method via adaptive analytical line selection and relevance vector machine Regression model;Jianhong Yang et al.;《Spectrochimica Acta Part B》;20150303;第107卷;第45-55页 * |
| 共识别建模方法在光谱分析中的应用研究;刘科;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160315(第03期);摘要，正文第22-25页 * |
| 化学计量学在激光诱导击穿光谱分析中的研究进展;张天龙 等;《分析化学评述与进展》;20150630;第43卷(第6期);第939-948页 * |
| 舰用发动机某关键部件故障预测方法与应用研究;王双双;《中国优秀硕士学位论文全文数据库 工程科技II辑》;20170315(第03期);正文第27-31页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107632010A (en) | 2018-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107632010B (en) | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy | |
| CN101915744B (en) | Near infrared spectrum nondestructive testing method and device for material component content | |
| CN103983617A (en) | Improved laser probe quantitative analysis method based on wavelet transform | |
| CN103913432A (en) | Near infrared spectrum wavelength selecting method based on particle swarm optimization | |
| CN105388123A (en) | Method for predicting crude oil characteristic through near infrared spectrum | |
| CN101430276A (en) | Wavelength variable optimization method in spectrum analysis | |
| CN104596957A (en) | Estimation method for content of copper in soil on basis of visible-light near-infrared spectrum technology | |
| CN103487411A (en) | Method for recognizing steel grade by combining random forest algorithm with laser-induced breakdown spectroscopy | |
| CN103543132B (en) | A kind of coal characteristic measuring method based on wavelet transformation | |
| CN110321518B (en) | Method for judging trend type of hydrological time series | |
| CN104655583A (en) | Fourier-infrared-spectrum-based rapid coal quality recognition method | |
| CN104020135A (en) | Calibration model establishing method based on near infrared spectrum | |
| CN105069309A (en) | Method for identifying hydrologic time sequence nonlinear trend | |
| CN102937575B (en) | Watermelon sugar degree rapid modeling method based on secondary spectrum recombination | |
| CN102305792A (en) | Nonlinear partial least square optimizing model-based forest carbon sink remote sensing evaluation method | |
| CN103398971A (en) | Chemometrics method for determining cetane number of diesel oil | |
| CN108827909B (en) | Rapid soil classification method based on visible near infrared spectrum and multi-target fusion | |
| CN109142251A (en) | The LIBS quantitative analysis method of random forest indirect labor's neural network | |
| CN111191502B (en) | Stick-slip and jump drill abnormal working condition identification method based on drill string vibration signal | |
| CN108398400B (en) | Method for nondestructive testing of fatty acid content in wheat by terahertz imaging | |
| CN115728290A (en) | Method, system, equipment and storage medium for detecting chromium element in soil | |
| CN114354666B (en) | Soil heavy metal spectral feature extraction and optimization method based on wavelength frequency selection | |
| Xie et al. | Quantitative analysis of steel samples by laser-induced-breakdown spectroscopy with wavelet-packet-based relevance vector machines | |
| CN113295674B (en) | Laser-induced breakdown spectroscopy characteristic nonlinear processing method based on S transformation | |
| CN113740295B (en) | Method and device for detecting and analyzing finished oil additive based on near infrared modeling |
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 |