Multispectral combined method for online detection of slurry components in mineral processing process

Abstract

A multispectral combined method for online detection of slurry components in a mineral processing process, relating to the technical field of spectrum detection. The multispectral combined method comprises: collecting slurry samples having a gradient distribution of moisture content to serve as test samples, and using an NIR system to acquire near-infrared spectrum data of the test samples; performing spectral preprocessing on the near-infrared spectrum data, so as to reduce the interference of spectral noise on subsequent modeling, and establishing an NIR regression model for slurry moisture content; using an LIBS system to acquire plasma spectrum data of the test samples, and establishing an LIBS regression model for slurry concentration; substituting the NIR regression model for slurry moisture content into the LIBS regression model for slurry concentration, and optimizing an algorithm to realize model correction and compensation; and establishing a quantitative analysis model for slurry element content to realize LIBS and NIR multispectral combination. The multispectral combined method can effectively improve the online detection precision of slurry components in a mineral processing process and improve the accuracy of a regression model.

Description

A multi-spectral combined method for online detection of slurry components in mineral processing Technical Field

[0001] The present application relates to the field of spectral detection technology, and in particular to a multi-spectral combination method for online detection of slurry components in a mineral processing process. Background Art

[0002] The mineral processing process can be used to transform raw ore with lower element content into higher-grade concentrate, enabling rapid and accurate online detection of elemental composition in the mineral processing slurry. This is of great significance in resolving the global technical difficulties faced by mineral processing processes, where traditional analytical detection methods rely on manual sampling, are slow and time-consuming, and provide severely lagging process guidance. Currently, laser-induced breakdown spectroscopy (LIBS) is commonly used to detect elemental composition in mineral processing slurries. LIBS technology has the advantages of non-contact measurement, real-time online, simultaneous multi-element detection, and no need for manual sample preparation. Therefore, its application in the field of mineral processing slurry detection has attracted widespread attention.

[0003] When using LIBS technology to measure slurry composition online, the water content in the slurry significantly affects the plasma intensity during laser excitation. Furthermore, the spectral data acquired by LIBS only reflects the elemental concentrations in the slurry, whereas slurry characteristics during the mineral processing process encompass multiple aspects. Using only the spectral intensity data at each pixel as the independent variable in modeling limits the accuracy of the slurry element grade regression model.

[0004] The moisture content of slurries of different grades is different, and water content data has a compensatory effect on the development of LIBS algorithms. NIR technology has a high accuracy in detecting the moisture content of slurries. The multi-spectral combination method of LIBS and NIR technology can meet the requirements of high-accuracy detection of slurry components in the mineral processing process, and can integrate various characteristic information of the slurry into the modeling algorithm. However, the existing LIBS and NIR combined detection methods are mainly concentrated in the field of solid-phase detection such as coal, and the morphology and types of the detected samples are limited. At present, there is no relevant application in the detection of slurry components in the mineral processing process. Summary of the Invention

[0005] The purpose of this application is to provide a multi-spectral combination method for online detection of slurry components during mineral processing to solve the above problems.

[0006] To achieve the above objectives, the present application adopts the following technical solutions: A multi-spectral combination method for online detection of slurry components in the mineral processing process, comprising: S1. Collecting slurry samples with a gradient distribution of moisture content as test samples, and using an NIR system to obtain near-infrared spectral data of the test samples; S2. Performing spectral preprocessing on the near-infrared spectral data to reduce the interference of spectral noise on subsequent modeling; Establishing an NIR regression model for slurry moisture content; S3. Using a LIBS system to obtain the plasma spectral data of the test sample, and establishing a LIBS regression model for slurry concentration; S4. Substituting the NIR moisture content regression model of the slurry into the LIBS slurry concentration regression model, optimizing the algorithm, and realizing model correction and compensation; S5. Establishing a quantitative analysis model for the element content of the slurry, and realizing the combination of LIBS and NIR multi-spectroscopy.

[0007] Due to the large fluctuations and constant process changes in the mineral processing process, the moisture content of the ore pulp fluctuates significantly during the mineral processing process, which has a significant impact on LIBS detection and leads to large deviations in the test results. Secondly, current methods for measuring moisture content are mostly offline in the factory, which cannot integrate moisture content data into the LIBS detection process in real time. Based on this, we use the NIR method to measure the ore pulp moisture content in real time online, and then integrate this data into the LIBS algorithm to realize a multi-spectral combined method for mineral pulp.

[0008] Preferably, the S4 includes: S41. Splicing moisture content and LIBS data: adding a column of predicted moisture content variables to the LIBS data to obtain a new matrix of size m×(n1+1); S42. Correcting the intensity of the characteristic peak of the hydrogen element in the LIBS spectrum according to the moisture content value predicted by the near-infrared data; S43. Correcting the laboratory analysis value used in the LIBS modeling to obtain the element content in the slurry sample.

[0009] Preferably, the calibration in S42 includes: establishing a standard curve of moisture content and H-line intensity, with moisture content data as the M axis and H-line intensity as the N axis, establishing a standard curve of a linear equation M=KN+B, wherein K is the slope of the standard curve; when performing subsequent experiments on samples to be tested, substituting all moisture content data X matrix measured by NIR into the standard curve, and calculating the absolute H-line intensity value H of each spectrum. i0 , and obtain the Y matrix. For each variable point of each spectrum, the reference intensity correction is performed. The correction formula is as follows: where Q i,j is the jth variable point of the i-th spectrum, H i0 is the absolute H line intensity value of the i-th spectrum, H i is the actual raw data H line intensity of the ith spectrum.

[0010] Preferably, in S1, the near-infrared spectral data of the test sample is an average value calculated by collecting multiple sets of spectral data for each pulp sample.

[0011] Preferably, the spectrum preprocessing method is SG convolution smoothing filter and first-order derivative method, wherein the SG convolution smoothing filter is used to eliminate high-frequency noise in the spectrum, and the first-order derivative method is used for spectrum baseline correction.

[0012] Preferably, the NIR regression model is a partial least squares regression model, and the input variables of the NIR regression model are the spectral line intensity and the slurry moisture content value after the spectral preprocessing.

[0013] Multiple groups of NIR spectral data and moisture content data after spectral preprocessing are used as training set samples for the partial least squares method, the moisture characteristic spectral line intensity in the NIR spectral data is used as the independent variable matrix X, and the moisture content data is used as the dependent variable matrix Y, and finally a NIR regression model is established.

[0014] Preferably, the LIBS regression model is a partial least squares model, and the input variables of the LIBS regression model are the spectral line intensities and concentration values ​​of the elemental components after the spectral preprocessing; before establishing the LIBS regression model, spectral data normalization and baseline correction are used as spectral data preprocessing methods.

[0015] Multiple sets of LIBS spectral data and chemical analysis concentration data were used as training set samples for the partial least squares method, where the element characteristic line intensities in the spectral data were used as the independent variable matrix X, and the concentration data of each element in the chemical analysis were used as the dependent variable matrix Y. Finally, a LIBS regression model was established.

[0016] Preferably, the correction and compensation includes: using the slurry NIR moisture content data as an independent variable, and simultaneously inputting it into a quantitative analysis model of the slurry element content along with several other LIBS spectral data independent variables; the number of the LIBS spectral data independent variables is the total number of pixels in the LIBS system, and the total number of pixels in the LIBS system is the product of the number of spectral channels of the LIBS system and the number of pixels in a single channel.

[0017] Preferably, principal component analysis is performed when the partial least squares method is used to establish the NIR regression model, and the optimal number of principal components for modeling is output; principal component analysis is performed when the partial least squares method is used to establish the LIBS regression model, and the optimal number of principal components for modeling is output.

[0018] Preferably, the prediction effect of the quantitative analysis model on the elemental composition of the verification set slurry is evaluated using relative error as an indicator.

[0019] Compared with the existing technology, the beneficial effects of the present application include: The multi-spectral combination method provided in this application for online detection of slurry components in the mineral processing process, taking into account the different moisture contents of slurry samples, uses the NIR method to measure the slurry moisture content for the first time, and integrates the moisture content into the correction of the H-line intensity. The moisture content data is further spliced ​​with the LIBS data for modeling, realizing the fusion of the NIR moisture content regression model and the LIBS component concentration regression model, which can effectively improve the defects of the traditional LIBS data modeling method, improve the online detection accuracy of slurry components in the mineral processing process, and improve the accuracy of the regression model. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope of the present application.

[0021] Figure 1 is a schematic flow chart of a multi-spectral combined method for online detection of slurry components in a mineral processing process provided in an embodiment; Figure 2 is an original NIR spectrum of moisture in the slurry in an embodiment; Figure 3 is a spectrum of the moisture spectrum in the slurry after SG convolution filtering and first-order derivative in an embodiment; Figure 4 is a diagram showing the relationship between a reference value and a predicted value of the moisture content of a predicted sample in an embodiment; and Figure 5 is a schematic diagram of a matrix used in an embodiment. DETAILED DESCRIPTION

[0022] The embodiments of the present application will be described in detail below in conjunction with specific examples, but it will be understood by those skilled in the art that the following examples are merely illustrative of the present application and should not be considered as limiting the scope of the present application. In the examples, if specific conditions are not specified, the conditions are carried out according to conventional conditions or manufacturer recommendations. The reagents or instruments used are not specified by the manufacturer and are conventional products that can be purchased commercially.

[0023] Example 1-3 As shown in Figure 1, this embodiment provides a multi-spectral combination method for online detection of slurry components in a mineral processing process, which specifically includes the following steps: S1. Collect 40 groups of slurry samples with a gradient distribution of moisture content as test samples, and use an NIR system to obtain near-infrared spectral data of different slurries; The moisture content data of the 40 groups of slurries to be tested are shown in Table 1 below: Table 1 Moisture content data of slurries to be tested After using the NIR system to obtain the near-infrared spectral data of the slurry, 800 sets of spectral data were collected for each slurry sample, and the average value of the 800 sets of data was finally taken as the original spectral data; the original NIR spectrum of water in the slurry is shown in Figure 2.

[0024] S2. Perform spectral preprocessing on the slurry near-infrared spectral data to reduce the interference of spectral noise on subsequent modeling, and establish a NIR regression model for slurry moisture content.

[0025] The NIR spectrum preprocessing method uses SG convolution smoothing filter and first-order derivative method. The smoothing filter mainly eliminates high-frequency noise in the spectrum, and the first-order derivative is used for spectral baseline correction. The NIR moisture spectrum after SG convolution filtering and first-order derivative is shown in Figure 3.

[0026] When the partial least squares method is used to establish a regression model, the principal component analysis method is used to output the optimal number of principal components for modeling and improve the modeling accuracy.

[0027] The established NIR slurry moisture content regression model is a partial least squares model. Multiple sets of NIR spectral data and moisture content data are used as training set samples of the partial least squares method. The moisture characteristic line intensity in the spectral data is used as the independent variable matrix X, and the moisture content data is used as the dependent variable matrix Y. Finally, the NIR regression model is established. The relationship between the reference value and the predicted value of the moisture content of 40 predicted samples is shown in Figure 4. The correlation coefficient Rp is 2 The prediction root mean square error (RMSEP) is 0.376, and the relative deviation of the prediction results is basically less than 2%, indicating a high prediction accuracy.

[0028] S3. Use the LIBS system to obtain slurry plasma spectrum data and establish a LIBS regression model for slurry concentration.

[0029] In step S3, before establishing the LIBS regression model, spectral data normalization and baseline correction are used as spectral data preprocessing methods.

[0030] The established LIBS regression model is a partial least squares regression model. Multiple sets of LIBS spectral data and chemical analysis concentration data are used as training set samples of the partial least squares method. The element characteristic line intensity in the spectral data is used as the independent variable matrix X, and the concentration data of each element in the chemical analysis is used as the dependent variable matrix Y. Finally, the LIBS regression model is established.

[0031] S4. Substitute the NIR moisture content regression model of the slurry into the LIBS slurry concentration regression model, optimize the algorithm, and achieve model correction and compensation.

[0032] Specifically, S4 includes: S41. Concatenating moisture content and LIBS data: adding a column of predicted moisture content variables to the LIBS data to obtain a new matrix of size m×(n1+1) (as shown in Figure 5); S42. Correcting the intensity of the characteristic peak of the hydrogen element in the LIBS spectrum based on the moisture content value predicted by the near-infrared data; S43. Correcting the laboratory analysis values ​​used in the LIBS modeling to obtain the element content in the slurry sample.

[0033] The calibration in S42 includes: establishing a standard curve of moisture content and H-line intensity, with moisture content data as the M axis and H-line intensity as the N axis, establishing a standard curve of a linear equation N=KM+B, where K is the slope of the standard curve; when performing subsequent experiments on the sample to be tested, substituting all moisture content data X matrix measured by NIR into the standard curve, and calculating the absolute H-line intensity value H of each spectrum. i0 , and obtain the Y matrix. For each variable point of each spectrum, the reference intensity correction is performed. The correction formula is as follows: where Q i,j is the jth variable point of the i-th spectrum, H i0 is the absolute H line intensity value of the i-th spectrum, H i is the actual raw data H line intensity of the i-th spectrum.

[0034] Among them, model correction and compensation require the slurry NIR moisture content data as an independent variable and input it into the quantitative analysis model of slurry element content simultaneously with several other LIBS spectral data independent variables, that is, adding moisture content data to the LIBS independent variable matrix X.

[0035] In this embodiment, the number of spectral channels of LIBS is 2, the number of pixels in a single channel is 2048, and the length of its independent variable matrix X is 4096. After the moisture content data is added, the length of the independent variable matrix X is 4097.

[0036] S5. Establish a quantitative analysis model for element content in slurry and realize the combination of LIBS and NIR multispectroscopy.

[0037] Table 3 below shows the reference value, predicted value, and relative error data of the element content in the slurry modeled using LIBS spectral data and moisture content data as independent variables in this embodiment. As can be seen from Table 3 below, compared with the modeling method that only uses the spectral intensity of the characteristic spectral line as the independent variable matrix X, the data analysis results formed by the modeling method after integrating the moisture content data show that the predicted results of the element content in the raw, refined, and tailings slurries are significantly improved, and the overall relative errors are all within 10%, with a high prediction accuracy.

[0038] Table 2 Prediction data of element content (%) of slurry to be tested after integrating moisture content data Comparative Examples 1-3 The uncorrected LIBS regression model was used to predict the element content in the raw, refined and tailings slurries. The reference value, predicted value and relative error of the element content in the slurry are shown in Table 3 below. It can be seen from the data that after modeling and prediction using the regression model without adding moisture content data, the overall relative error of the element prediction in the raw, refined and tailings slurries is large. The relative error of some elements is as high as more than 100%, and the prediction accuracy is poor.

[0039] Table 3 Prediction data of element content (%) of ore pulp to be tested By comparing the data of Examples 1-3 and Comparative Examples 1-3, it can be seen that the online detection accuracy of the examples of the present application is much higher than that of the comparative examples, indicating that the multi-spectral combination method provided by the present application for online detection of slurry components in the mineral processing process realizes the fusion of the NIR moisture content regression model and the LIBS component concentration regression model, which can effectively improve the defects of the traditional LIBS data modeling method, improve the online detection accuracy of slurry components in the mineral processing process, and improve the accuracy of the regression model.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-spectral combined method for online detection of slurry components in a mineral processing process, characterized in that: include: S1. Collecting a slurry sample with a gradient distribution of moisture content as a test sample, and using a NIR system to obtain near infrared spectral data of the test sample; S2. performing spectral preprocessing on the near infrared spectral data to reduce the interference of spectral noise on subsequent modeling; establishing a NIR regression model of slurry moisture content; S3. Using the LIBS system to obtain the plasma spectrum data of the test sample, and establishing a LIBS regression model of the slurry concentration; S4. Substitute the NIR moisture content regression model of the slurry into the LIBS slurry concentration regression model, optimize the algorithm, and realize model correction and compensation; S5. Establish a quantitative analysis model for the element content of slurry and realize the combination of LIBS and NIR multispectroscopy.

2. The multi-spectral combination method for online detection of slurry components in the mineral processing process according to claim 1 is characterized in that: The S4 includes: S41. Concatenate moisture content and LIBS data: Add a column of predicted moisture content variables to the LIBS data to obtain a new matrix of size m×(n1+1); where m is the number of samples and n1 is the number of LIBS data; S42. Correcting the intensity of the characteristic peak of hydrogen element in the LIBS spectrum according to the water content value predicted by the near infrared data; S43. Correct the laboratory analysis values ​​used in LIBS modeling to obtain the element content in the slurry sample.

3. The multi-spectral combined method for online detection of slurry components in a mineral processing process according to claim 1 is characterized in that: The correction in S42 includes: Establish a standard curve of moisture content and H-line strength, with moisture content data as the M axis and H-line strength as the N axis, and establish a standard curve of a linear equation N=KM+B, where K is the slope of the standard curve; When conducting subsequent experiments on the samples to be tested, substitute all the moisture content data X matrix measured by NIR into the standard curve and calculate the absolute H line intensity value H of each spectrum. i0 , and obtain the Y matrix. For each variable point of each spectrum, the reference intensity correction is performed. The correction formula is as follows: Where Q i，j is the jth variable point of the i-th spectrum, H i0 is the absolute H line intensity value of the ith spectrum, H i is the actual raw data H line intensity of the ith spectrum.

4. The multi-spectral combined method for online detection of slurry components in a mineral processing process according to claim 1 is characterized in that: In S1, the near infrared spectrum data of the test sample is an average value calculated by collecting multiple groups of spectrum data for each pulp sample.

5. The multi-spectral combination method for online detection of slurry components in a mineral processing process according to claim 1, characterized in that: The spectrum preprocessing method is SG convolution smoothing filter and first-order derivative method, wherein the SG convolution smoothing filter is used to eliminate high-frequency noise in the spectrum, and the first-order derivative method is used for spectrum baseline correction.

6. The multi-spectral combination method for online detection of slurry components in a mineral processing process according to claim 1, characterized in that: The NIR regression model is a partial least squares regression model, and the input variables of the NIR regression model are the spectral line intensity and the slurry water content value after the spectral preprocessing; Multiple groups of NIR spectral data and moisture content data after spectral preprocessing are used as training set samples of partial least squares method, the moisture characteristic spectral line intensity in the NIR spectral data is used as the independent variable matrix X, and the moisture content data is used as the dependent variable matrix Y, and finally a NIR regression model is established.

7. The multi-spectral combination method for online detection of slurry components in a mineral processing process according to claim 1, characterized in that: The LIBS regression model is a partial least squares model, and the input variables of the LIBS regression model are the spectral line intensity and the concentration values ​​of the element components after the spectral preprocessing; Before establishing the LIBS regression model, spectral data normalization and baseline correction were used as spectral data preprocessing methods; Multiple groups of LIBS spectral data and chemical analysis concentration data were used as training set samples for the partial least squares method, where the element characteristic line intensity in the spectral data was used as the independent variable matrix X, and the concentration data of each element in the chemical analysis was used as the dependent variable matrix Y, and finally a LIBS regression model was established.

8. The multi-spectral combination method for online detection of slurry components in a mineral processing process according to claim 1 is characterized in that: The correction and compensation includes: taking the slurry NIR moisture content data as an independent variable, and simultaneously inputting it into a quantitative analysis model of slurry element content with other LIBS spectrum data independent variables; The number of the LIBS spectral data independent variables is the total number of pixels of the LIBS system, and the total number of pixels of the LIBS system is the product of the number of spectral channels of the LIBS system and the number of pixels of a single channel.

9. The multi-spectral combination method for online detection of slurry components in a mineral processing process according to claim 1, characterized in that: When establishing the NIR regression model using the partial least squares method, principal component analysis is performed to output the optimal number of principal components for modeling; When the partial least squares method is used to establish the LIBS regression model, principal component analysis is performed to output the optimal number of principal components for modeling.

10. The multi-spectral combined method for online detection of slurry components in a mineral processing process according to any one of claims 1 to 9, characterized in that: The prediction effect of the quantitative analysis model on the elemental composition of the verification set slurry is evaluated by relative error.

Images