A Laser-Induced Breakdown Spectroscopy Standardization Method for Correcting the Influence of Moisture

Abstract

The present invention relates to the field of spectral standardization, and specifically to a method for laser-induced breakdown spectroscopy standardization that corrects the influence of moisture. It is used to solve the problem of the influence of the moisture level in the substance to be measured on the spectral intensity and the accuracy and stability of the quantitative analysis results. The specific steps are as follows: (1) Calculate the absolute value |θ i，j | of the correlation between each dimension of the non-standard spectral matrix and each dimension of the standard spectral matrix; (2) Select m-dimensional non-standard spectral lines to fit the standard spectral lines according to the magnitude of |θ i，j |; (3) Extract the coefficients before each dimension of the spectral lines participating in the fitting to form a transformation matrix; (4) Perform quantitative analysis on the standardized spectra in the validation set, and determine the optimal number of modeling samples, the number of spectral lines participating in the fitting, and the number of principal components during the modeling process through the RMSE of the validation set.

Description

Technical Field

[0001] The present invention relates to the field of spectral standardization, and specifically to a method for laser-induced breakdown spectroscopy standardization that corrects the influence of moisture. Background Art

[0002] Currently, for the analysis of the composition of substances, an off-line sampling method is generally adopted. After complex sample preparation, it is measured by chemical titration or X-ray fluorescence spectroscopy analysis. There are also some that use neutron activation technology to achieve on-line measurement, but due to its radioactivity, it cannot be widely applied. Laser-induced breakdown spectroscopy (LIBS), due to its characteristics of being able to perform real-time, on-line, in-situ, non-radiation pollution, and full-element analysis, has currently been applied in many fields.

[0003] In the quantitative analysis of LIBS, the water content in the substance will affect the intensity and stability of the LIBS spectrum. Currently, the research on the influence of the water content of the sample mainly focuses on aspects such as the influence of moisture on the characteristics of the laser-induced plasma, spectral line intensity and stability, and quantitative analysis. There are relatively few studies on correcting the influence of moisture. The present invention provides a method for laser-induced breakdown spectroscopy standardization that corrects the influence of moisture, which improves the spectral stability and the accuracy of quantitative analysis under the condition of moisture change by correcting the spectral differences under different moisture levels. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the purpose of the present invention is to solve the problem of the influence of the water content in the sample on the strength of the LIBS spectrum and the accuracy and stability of quantitative analysis, and to provide a method for laser-induced breakdown spectroscopy standardization that corrects the influence of moisture to improve the spectral stability and the accuracy of quantitative analysis under the condition of moisture change.

[0005] The present invention proposes the following technical solutions to achieve the above purpose:

[0006] A method for laser-induced breakdown spectroscopy standardization that corrects the influence of moisture, comprising the following steps:

[0007] Step 1: For the samples of the substance to be measured in different moisture content ranges, collect the original data of the laser-induced breakdown spectrum multiple times, perform normalization processing, and obtain a multi-dimensional spectral line sample data set;

[0008] Step 2: Establish the relationship between different moisture content ranges of the samples of the substance to be measured and the corresponding spectral means, and obtain the standard moisture content range;

[0009] Step 3: For the spectral lines measured multiple times for the same sample, average the spectra within the standard moisture content range and use it as the standard spectrum of the sample, and use the spectra outside the standard moisture content range as the non-standard spectra of the sample;

[0010] Step 4: Divide the multi-dimensional spectral line sample data set into a training set, a validation set, and a test set; both the training set, the validation set, and the test set include the standard spectra and non-standard spectra of the samples.

[0011] Step 5: Select some samples of the substance to be measured from the training set. Using the non-standard spectral data of each sample as the input and the standard spectral data as the output, perform regression fitting using a regression method to establish a spectral normalization model; optimize the model parameters according to the concentration quantitative analysis results of the validation set to obtain an optimized spectral normalization model.

[0012] Step 6: Use the non-standard spectrum of any sample of the substance to be measured in the test set as the input of the normalization model, and input it into the optimized spectral normalization model to obtain the normalized spectrum.

[0013] The samples of the substance to be measured are samples of the substance to be measured with continuously gradient moisture content configured in advance.

[0014] The original laser-induced spectral data of the collected samples is collected using a laser-induced breakdown spectrometer.

[0015] The relationship between different moisture content intervals and the corresponding spectral means is a moisture-spectral mean relationship line, which is linear; according to the moisture-spectral mean relationship line, the spectral mean range corresponding to the standard moisture interval is inversely deduced, and the numerical value of this mean range is used to replace the standard moisture content interval of the substance to be measured.

[0016] The parameters of the spectral normalization model include: the number of modeling samples, the number of fitting spectral lines m, and the number of principal components.

[0017] The establishment of the spectral normalization model specifically includes the following steps:

[0018] Step a: For the training set data, calculate the correlation coefficient θ between each dimension spectral line Y(:,i) of the standard spectral matrix and each dimension spectral line X(:,j) of the non-standard spectral matrix i,j ;

[0019] Step b: Sort the absolute value of the correlation coefficient |θ i,j | from large to small, and extract the first m-dimensional non-standard spectral lines X(:,j1)...X(:,j i,j | corresponding to |θ m ) as the spectral lines participating in fitting Y(:,i) in step c, and save the indices of these m-dimensional spectral lines in the non-standard spectral matrix.

[0020] Step c: Use the partial least squares method to calculate the regression coefficients β between Y(:,i) and X(:,j1)....X(:,j m ) 1,i …βm,i ;

[0021] Step d: Define a vector with the same number of columns as the non-standard spectral matrix. Fill in the index positions saved in step b according to the corresponding regression coefficients β 1,i …β m,i . The index positions of the spectral lines not participating in the fitting are 0, forming a vector α i =(β 1,i ,...0,...β m,i ,...) T ;

[0022] Step e: Use α i to form a transformation matrix F, i.e., F=(α1,α2,...,α i ,...α n ). Multiply the non-standard spectral matrix X by the transformation matrix F, Y = X·F, to achieve the conversion from non-standard spectra to standard spectra.

[0023] Optimizing the model parameters according to the quantitative analysis results of the validation set means inputting the non-standard spectral data of the validation set into Y = X·F to obtain the predicted standard spectra; determining the optimal model parameters based on the root mean square error between the predicted standard spectra and the elemental concentrations of the standard spectra in the validation set.

[0024] The finally obtained optimal spectral normalization model is:

[0025] Y(:,i)=β 1,i X(:,j1)+...+β m,i X(:,j m )

[0026] where Y is the standard spectral matrix, Y(:,i) is the target spectral line to be fitted in the i-th dimension of the standard spectral matrix, X is the non-standard spectral matrix, X(:,j1)....X(:,j m ) are the m-dimensional spectral lines participating in the fitting in the non-standard spectral matrix, and β 1,i ....β m,i are the regression coefficients before each dimension of the spectral lines participating in the fitting.

[0027] A laser-induced breakdown spectroscopy normalization system for correcting the influence of moisture includes a spectral acquisition device, a processor, and a memory; the spectral acquisition device is used to acquire the laser-induced original spectral data of the pulp sample; the memory stores the following program modules, and the processor reads the program to execute the above method steps to achieve the correction of the spectral moisture content of the current pulp sample to obtain the concentration-normalized spectra of the specified elements;

[0028] A spectral acquisition module, which is used to acquire the original data of the laser-induced breakdown spectra of the substance to be measured in different moisture ranges;

[0029] A data preprocessing module, which is used to normalize the original data of the laser-induced breakdown spectroscopy collected for the substance to be measured in different moisture ranges, so as to obtain a multi-dimensional spectral line data set;

[0030] A standard spectrum selection module, which is used to select the spectrum of the substance to be measured within the standard moisture range and the spectrum outside the standard moisture range;

[0031] A data set division module, which is used to divide the multi-dimensional spectral line data set into a training set, a validation set and a test set;

[0032] A spectral standardization model establishment and optimization module, which is used to establish a model by using the training set data and the validation set data, and obtain the optimal number of modeling samples, the number of principal components and the number of spectral lines participating in fitting;

[0033] A test module, which is used to input the test set data into the established spectral standardization model to obtain the standardized spectral data.

[0034] The present invention has the following beneficial effects and advantages:

[0035] The present invention proposes a method for laser-induced breakdown spectroscopy standardization to correct the influence of moisture, specifically a spectral standardization method based on spectral line intensity correlation matching (SICMS). The core idea is to select the non-standard spectral lines that meet the conditions to fit each dimension of the standard spectrum according to the absolute value |θ i,j | of the correlation between the non-standard spectral lines and the standard spectral lines. Compared with other spectral standardization methods with window sliding, the SICMS spectral standardization method can break through the window limitation. By using the SICMS spectral standardization method to correct the spectra under different moisture levels and standardize them to the spectra under the standard moisture level, the influence of moisture on the stability and accuracy of the quantitative analysis results can be corrected. The SICMS spectral standardization method can be actually applied to correct the influence of the moisture in the sintered ore mixture, and also has reference significance for the correction of the influence of the target moisture in other types of applications. Description of the Drawings

[0036] Figure 1 It is a flow chart of the present invention.

[0037] Figure 2 It is a quantitative analysis diagram of a four-element model.

[0038] Figure 3 It is a diagram of the influence of moisture on the quantitative analysis results.

[0039] Figures 4 to 6 It is a parameter optimization diagram for establishing a spectral standardization model.

[0040] Figure 7It is a spectral comparison chart before and after adopting the spectral normalization method.

[0041] Figure 8 It is an error bar chart of the predicted values of the concentrations of four elements before and after adopting the spectral normalization method.

[0042] Figure 9 It is an RSD comparison chart of the predicted values of the concentrations of four elements before and after adopting the spectral normalization method. Specific implementation manners

[0043] To make the above objects, features and advantages of the present invention more obvious and understandable, the technical solution of the present invention will be further described below in conjunction with examples of sintered ore mixture composition analysis and moisture influence correction. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the invention. Therefore, the present invention is not limited by the specific implementations disclosed below.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0045] Example: A laser-induced breakdown spectroscopy normalization method for correcting the influence of moisture. The flowchart is as Figure 1 shown, and specifically includes the following steps:

[0046] (1) Pre-configure sintered ore mixtures in different moisture ranges: They are samples of substances to be measured with a continuous gradient moisture content range, and the configured continuous gradient moisture range can cover the standard moisture range of the substances to be measured itself. Collect spectral data through a laser-induced breakdown spectrometer, perform full-spectrum and normalization, and take the spectra with the spectral line mean between 50 and 70 for each sample for averaging. The generated average spectrum is used as the standard spectrum of the sample, and the spectra outside the range of 50 to 70 are used as non-standard spectra under this sample.

[0047] Among them, the numerical range of "50-70" is obtained as follows: According to the measured moisture content and spectral mean data, establish the relationship between different moisture content ranges and the corresponding spectral means of each sample to be measured in advance: a moisture-spectral mean relationship line, which is linear. According to the moisture-spectral mean relationship line, inversely deduce the spectral mean range corresponding to the standard moisture range, and use the numerical range "50-70" of this mean range to replace the standard moisture content range of the substance to be measured.

[0048] (2) Divide the training set, validation set, and test set. The total number of samples is 92. Select 11 samples as the validation set, 16 samples as the test set, and 65 samples as the training set.

[0049] Obtain the concentration contents of four elements, namely Fe, Si, Ca, and Mg, according to the standard spectra of each sample in the training set, validation set, and test set. Among them, only taking the quantitative analysis of the concentration contents in the test set as an example, the concentration contents are shown in Figure 2 .

[0050] (3) Put the non-standard spectra of the 16 samples in the test set into the above quantitative analysis model, draw the error bars of the predicted values of the four element concentrations respectively, and explore the influence of moisture on the quantitative analysis results, as shown in Figure 3 (The dots in the figure are the means of the predicted values of multiple non-standard spectra of each sample in the test set, and the squares are the predicted values of the standard spectra of each sample in the test set. It can be seen from the figure that there is a certain deviation between the dots and the squares, and the error bars of each sample are also large, indicating that moisture has a serious impact on the accuracy and stability of the quantitative analysis results).

[0051] (4) Suppose the number of modeling samples selected from the training set is {22, 28, 34, 43, 49} respectively, the spectral line dimensions involved in fitting are {150, 200, 250, 300, 500, 1000}, and the number of principal components is {5, 6, 7, 8, 9}.

[0052] The training steps are as follows:

[0053] Step a: Calculate the correlation coefficient θ between each spectral line Y(:,i) of the standard spectral matrix and each spectral line X(:,j) of the non-standard spectral matrix i,j ;

[0054] Step b: Sort the absolute values of the correlation coefficients |θ i,j | from large to small, and extract the first m non-standard spectral lines X(:,j1)...X(:,j m ) corresponding to |θ i,j | as the spectral lines participating in fitting Y(:,i) in step c, and save the indices of these m spectral lines in the non-standard spectral matrix;

[0055] Step c: Calculate the regression coefficients β m ) between Y(:,i) and X(:,j1)....X(:,j 1,i …β m,i ;

[0056] That is, for each dimension Y(:,i) of the standard spectrum, the eligible X(:,j1)....X(:,j m) A total of m spectral lines are involved in the fitting. To prevent overfitting, the partial least squares method is used for regression, and the regression model is shown in Equation (1).

[0057] Y(:,i) = β 1,i X(:,j1)+...+β m,i X(:,j m ) (1)

[0058] In Equation (1), Y is the standard spectral matrix, Y(:,i) is the target spectral line to be fitted in the i-th dimension of the standard spectral matrix, X is the non-standard spectral matrix, and X(:,j1)....X(:,j m ) are the m-dimensional spectral lines involved in the fitting in the non-standard spectral matrix. β 1,i ....β m,i are the regression coefficients before each dimension of the spectral line involved in the fitting.

[0059] Step d: Arrange β 1,i …β m,i in column order according to the index positions saved in Step b. The index positions of the spectral lines not involved in the fitting are 0, and a vector α i =(β 1,i ,...0,...β m,i ,...) T ;

[0060] Step e: Combine α i to form a transformation matrix F, that is, F=(α1,α2,...,α i ,...α 6133 );

[0061] Multiply the non-standard spectral matrix X by the transformation matrix F to achieve the conversion from the non-standard spectrum to the standard spectrum, that is, Y = X·F.

[0062] The steps to optimize the model parameters according to the validation set data are as follows:

[0063] Input the non-standard spectral data of the validation set into Y = X·F to obtain the predicted standard spectrum; perform quantitative analysis on the standardized spectra in the validation set: obtain the element concentrations, and determine the optimal parameters for establishing the spectral standardization model according to the root mean square error (RMSE) of the quantitative analysis results. First, fix the number of spectral lines involved in the fitting at 200 dimensions and the number of principal components at 6, and explore the corresponding relationship between the number of modeling samples and RMSE. It can be seen that when the number of modeling samples is 43, the RMSE corresponding to the predicted values of each element concentration is the smallest (see Figure 4 ). Then, fix the number of modeling samples at 43 and the number of spectral lines involved in the fitting at 200, and explore the relationship between the number of principal components and RMSE. It can be known that when the number of principal components is 6, the RMSE corresponding to the predicted values of each element concentration is the smallest (see Figure 5)。 Then, take the number of modeling samples as 43 and the number of principal components as 6 to explore the relationship between the spectral line dimension involved in fitting and RMSE. It can be seen that when the spectral line dimension involved in fitting is 200, the RMSE of the predicted values of each element concentration is the smallest (see Figure 6 ).

[0064] (5) Take the non-standard spectrum under each sample as the input and the standard spectrum as the output. Select the optimal parameters in (4) to establish a spectral normalization model, and normalize the non-standard spectrum under each sample in the test set. The comparison chart of the normalized spectra is shown in Figure 7 (It can be seen from the figure that after normalizing the non-standard spectrum, the difference in spectral line intensity between it and the standard spectrum is significantly reduced).

[0065] (6) Conduct quantitative analysis on the normalized spectra in the test set, draw the error bars and RSD of the predicted values of the concentrations of the four elements after spectral normalization respectively, and compare them with the results without spectral normalization in (3) to verify the correction effect of the spectral normalization method on the influence of moisture (It can be seen from Figure 8 that after spectral normalization, the deviation between the round dots and the square dots becomes smaller, and the error bars of each sample are also significantly reduced; it can be seen from Figure 9 that when conducting quantitative analysis on the normalized spectra, the RSD of the quantitative analysis results of the four elements Fe, Ca, Si, and Mg generally drops by more than half, indicating that the SICMS spectral normalization method can largely correct the influence of moisture on quantitative analysis.).

[0066] Result verification: Figure 7 is the comparison chart of the spectra before and after adopting the spectral normalization method, Figure 8 is the error bar chart of the predicted values of the concentrations of the four elements after adopting the spectral normalization method, Figure 9 is the RSD comparison chart of the predicted values of the concentrations of the four elements before and after adopting the spectral normalization method.

[0067] In this example, sintered ore mixture is used, which is only a preferred embodiment. During specific implementation, analysis can be carried out according to different application objects, and the modeling quantity, the number of fitting spectral lines, and the number of principal components for establishing the spectral normalization model can be adjusted.

[0068] The above specific implementation manners are used to explain the present invention, which are only preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and protection scope of the claims of the present invention all fall within the protection scope of the present invention.

Claims

1. A method for standardizing laser-induced breakdown spectroscopy to correct the influence of moisture, characterized in that, It includes the following steps: Step 1: For the samples of the substance to be measured in different moisture content ranges, collect the original data of laser-induced breakdown spectroscopy multiple times, perform normalization processing, and obtain a multi-dimensional spectral line sample data set; Step 2: Establish the relationship between different moisture content ranges of the samples of the substance to be measured and the corresponding spectral means, and obtain the standard moisture content range; Step 3: For the spectral lines measured multiple times for the same sample, average the spectra within the standard moisture content range and use it as the standard spectrum of the sample, and use the spectra outside the standard moisture content range as the non-standard spectra of the sample; Step 4: Divide the multi-dimensional spectral line sample data set into a training set, a validation set, and a test set; both the training set, the validation set, and the test set include the standard spectra and non-standard spectra of the samples; Step 5: Select some samples of the substance to be measured from the training set, use the non-standard spectral data of each sample as the input and the standard spectral data as the output, perform regression fitting using the regression method, and establish a spectral standardization model; optimize the model parameters according to the concentration quantitative analysis results of the validation set to obtain an optimized spectral standardization model; Step 6: Use the non-standard spectrum of any sample of the substance to be measured in the test set as the input of the standardization model, and input it into the optimized spectral standardization model to obtain the standardized spectrum.

2. A method for standardizing laser-induced breakdown spectroscopy for correcting the influence of moisture, according to claim 1, characterized in that, The samples of the substance to be measured are pre-configured samples of the substance to be measured with continuously gradient moisture content.

3. A method for standardizing laser-induced breakdown spectroscopy for correcting the influence of moisture, as claimed in claim 1, wherein The original spectral data of the laser-induced samples is collected using a laser-induced breakdown spectrometer.

4. A method for standardizing laser-induced breakdown spectroscopy for correcting the influence of moisture, as claimed in claim 1, wherein The relationship between the different moisture content ranges and the corresponding spectral means is a moisture-spectral mean relationship line, which is linear; according to the moisture-spectral mean relationship line, inversely deduce the spectral mean range corresponding to the standard moisture range, and use the numerical value of this mean range to replace the standard moisture content range of the substance to be measured.

5. A method for standardizing laser-induced breakdown spectroscopy corrected for moisture influence according to claim 1, characterized in that, The parameters of the spectral standardization model include: the number of modeling samples, the number of fitting spectral lines, and the number of principal components.

6. A method for standardizing laser-induced breakdown spectroscopy for correcting the influence of moisture, according to claim 1, characterized in that, The establishment of the spectral standardization model specifically includes the following steps: Step a: For the training set data, calculate the correlation coefficient θ between each spectral line Y(:,i) of the standard spectral matrix and each spectral line X(:,j) of the non-standard spectral matrix i,j ; Step b: Sort the absolute values |θ i,j | in descending order, and extract the first m non-standard spectral lines X(:,j1)...X(:,j i,j | corresponding to |θ m ) as the spectral lines participating in the fitting of Y(:,i) in Step c, and save the indices of these m spectral lines in the non-standard spectral matrix; Step c, use the partial least squares method to calculate the regression coefficients β m ) between Y(:,i) and X(:,j1)....X(:,j 1,i …β m,i ; Step d: Define a vector with the same number of columns as the non-standard spectral matrix, and fill in the index positions saved in step b according to the corresponding regression coefficients β 1,i …β m,i Fill in the spectral index positions that did not participate in the fitting with 0 to form a vector α i =(β 1,i ,...0,...β m,i ,...) T ; Step e: Take α i to form the transformation matrix F, i.e., F = (α1, α2,..., α i ,... α n ). Multiply the non-standard spectral matrix X by the transformation matrix F, Y = X·F, to achieve the conversion from the non-standard spectrum to the standard spectrum.

7. A method for standardizing laser-induced breakdown spectroscopy for correcting the influence of moisture, according to claim 1 or 5, characterized in that The optimization of the model parameters according to the concentration quantitative analysis results of the validation set is to input the non-standard spectral data of the validation set into Y = X·F to obtain the predicted standard spectrum; determine the optimal model parameters according to the root mean square error of the element concentrations between the predicted standard spectrum and the standard spectrum of the validation set.

8. A method for calibrating laser-induced breakdown spectroscopy normalization affected by moisture according to claim 7, characterized in that The finally obtained optimal spectral standardization model is: Y(:, i) = β 1,i X(:, j1) +... + β m,i X(:, j m ) where Y is the standard spectral matrix, Y(:,i) is the target spectral line to be fitted in the i-th dimension of the standard spectral matrix, X is the non-standard spectral matrix, and X(:,j1)....X(:,j m ) are the m-dimensional spectral lines participating in the fitting in the non-standard spectral matrix, and β 1,i ....β m,i are the regression coefficients before each dimension of the spectral lines participating in the fitting.

9. A laser-induced breakdown spectroscopy normalization system for correcting the influence of moisture, characterized in that, It includes a spectral acquisition device, a processor, and a memory; the spectral acquisition device is used to collect the original spectral data of laser-induced breakdown of the pulp sample; the memory stores the following program modules, and the processor reads the program to execute the method steps described in any one of claims 1-8 to realize the correction of the spectral moisture content of the current pulp sample to obtain the concentration standardized spectrum of the specified element; A spectral acquisition module, which is used to collect the original data of laser-induced breakdown spectroscopy of the substance to be measured in different moisture ranges; A data preprocessing module, which is used to perform normalization processing on the original data of laser-induced breakdown spectroscopy of the substance to be measured in different moisture ranges to obtain a multi-dimensional spectral data set; A standard spectrum selection module, which is used to select the spectra of the substance to be measured within the standard moisture range and the spectra outside the standard moisture range; A data set division module, which is used to divide a multi-dimensional spectral line data set into a training set, a validation set and a test set; A spectral normalization model establishment and optimization module, which is used to build a model using the training set data and the validation set data and obtain the optimal number of modeling samples, the number of principal components, and the number of spectral lines dimensions involved in fitting; A test module, which is used to input the test set data into the established spectral normalization model to obtain the normalized spectral data.

Images