Harmful element analysis method for blast furnace slag

Abstract

The invention is applicable to the technical field of blast furnace slag analysis, and provides a blast furnace slag harmful element analysis method which comprises the following steps: preparing a sample to be analyzed; collecting LIBS and XRF spectral data of a sample to be analyzed; the collected spectral data are preprocessed; training a machine learning model by using a slag sample with known components, and establishing a mapping relation between spectral data and element contents; inputting the preprocessed spectral data into the trained machine learning model, and outputting a content prediction value of each harmful element; according to the method for analyzing the harmful elements of the blast furnace slag, the LIBS and XRF spectrum technologies are combined, the coverage range of the elements is wider, real-time online analysis and multi-element synchronous analysis are achieved, the detection efficiency is improved, model parameters are automatically optimized through a machine learning algorithm, the method adapts to different slag components, and the method is suitable for popularization and application. The detection precision of trace elements is obviously improved, and automatic data analysis and report generation are supported.

Description

Technical Field

[0001] The invention belongs to the technical field of blast furnace slag analysis, and in particular relates to a method for analyzing harmful elements in blast furnace slag. Background Art

[0002] Blast furnace slag is a solid waste formed by the gangue in the ore, the ash in the fuel and the non-volatile components in the solvent (usually limestone) during the blast furnace ironmaking process. It mainly contains oxides of calcium, silicon, aluminum, magnesium, iron and a small amount of sulfide.

[0003] During the blast furnace smelting process, harmful elements in slag (such as sulfur, phosphorus, heavy metals, etc.) have a significant impact on steel quality and the environment. Existing analysis methods have the following problems: Insufficient precision: Traditional chemical analysis methods have limited detection accuracy for trace harmful elements and are difficult to meet high-precision requirements.

[0004] Inefficiency: Existing methods are time-consuming and cannot meet real-time monitoring needs. Summary of the invention

[0005] The present invention provides a method for analyzing harmful elements in blast furnace slag, aiming to solve the above-mentioned problem.

[0006] The present invention is achieved by a method for analyzing harmful elements in blast furnace slag, comprising the following steps: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0007] Preferably, the step of preparing the sample to be analyzed is specifically: Dry the collected samples to be analyzed. Specifically, the collected samples to be analyzed can be placed in an oven, the oven temperature is set to 90-120°C, and dried for 1-2 hours. The samples are taken out and cooled to room temperature. Free water and bound water in the samples are removed by heating to avoid moisture generating interference signals in spectral analysis. The dried slag sample is ground to a particle size of <100μm. Specifically, the dried slag sample can be placed in a grinding jar, and grinding balls are added, with the mass ratio of the slag sample to the grinding balls being 1:10-15; the grinding is performed for 20-40 minutes at a rotation speed of 300-400rpm. After the grinding is completed, the sample is taken out, and the slag sample is ground into a uniform powder (particle size <100μm) by mechanical force to eliminate the interference of the particle size on the spectral analysis; The ground slag sample is placed in a sample mixer for mixing. Specifically, the ground slag powder can be placed in the sample mixer for a mixing time of 10-20 minutes at a speed of 50-70 rpm. The sample powder is fully mixed through three-dimensional motion to ensure uniform distribution of chemical components. The mixed slag sample is pressed into tablets, and the mixed slag powder is placed in a tablet pressing mold with a pressure of 10-12t and a holding time of 25-35s. After the pressing is completed, the sample tablets are taken out and the powder sample is pressed into a solid tablet by high pressure to reduce the unevenness of the sample surface and improve the accuracy of spectral analysis.

[0008] Preferably, the step of preparing the sample to be analyzed is specifically: Microwave digestion technology is used to digest the samples to be analyzed; The digested slag samples are adsorbed by nanomaterials; The filter residue sample after nanomaterial adsorption is eluted, and the eluate is collected for subsequent analysis.

[0009] Preferably, the step of using microwave digestion technology to digest the sample to be analyzed is specifically: Grind the sample to be analyzed to a particle size of <100 μm; Weigh 0.4-0.6g of sample and put it into a microwave digestion tank with a microwave power of 700-900W; Add 10 mL of mixed acid, wherein the mixed acid is a mixture of nitric acid and hydrofluoric acid in a volume ratio of 2-4:1.

[0010] At a temperature of 160-200°C, react for 20-30 minutes, and cool to room temperature after digestion.

[0011] Microwave energy causes the acid solution to heat up rapidly, accelerating the decomposition of the mineral phase in the slag and releasing harmful elements. The mixed acid system of nitric acid and hydrofluoric acid can effectively dissolve insoluble minerals such as silicates.

[0012] Preferably, the step of adsorbing the digested slag sample by nanomaterials is specifically as follows: Dilute the microwave digested solution to 40-60 mL with deionized water and adjust the solution pH to 5.0; Add 8-12 mg of functionalized titanium dioxide nanotubes and stir at room temperature for 25-35 min to allow the nanomaterials to fully absorb harmful elements; The nanomaterials and the solution are separated by a centrifuge with a rotation speed of 4500-5500 rpm and a separation time of 10-15 minutes.

[0013] The thiol (-SH) groups on the surface of functionalized titanium dioxide nanotubes have highly selective adsorption capacity for harmful elements (such as lead, cadmium, mercury, etc.).

[0014] Preferably, the steps of eluting the filter residue sample after adsorption of the nanomaterial and collecting the eluate for subsequent analysis are specifically as follows: The nanomaterials that adsorb harmful elements are placed in a centrifuge tube; Add 3-7 mL of 0.1-0.2 M nitric acid solution and stir at room temperature for 10-15 min to elute the harmful elements from the nanomaterials; The eluent and the nanomaterials are separated by a centrifuge at a speed of 4500-5500 rpm and a separation time of 10-15 min; The eluate was collected for subsequent analysis.

[0015] Nitric acid solution can destroy the chemical bonds between nanomaterials and harmful elements, achieve efficient elution, and use a small volume of eluent, thereby concentrating harmful elements and improving analytical sensitivity.

[0016] Preferably, the step of collecting LIBS and XRF spectrum data of the sample to be analyzed is specifically: Place the sample in the LIBS instrument. If it is a sample sheet, place it on the LIBS sample stage. If it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the LIBS instrument to analyze the film. Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0017] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0018] Preferably, the step of preprocessing the collected spectral data is specifically: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0019] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0020] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0021] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0022] Preferably, the step of using a slag sample with known composition to train a machine learning model and establish a mapping relationship between spectral data and element content is specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0023] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0024] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0025] Preferably, the step of visually outputting the results is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0026] Compared with the prior art, the embodiments of the present application have the following beneficial effects: The harmful element analysis method of blast furnace slag provided by the present invention combines LIBS and XRF spectroscopy technology to cover a wider element detection range, real-time online analysis, multi-element synchronous analysis, improve detection efficiency, and automatically optimize model parameters with a machine learning algorithm to adapt to different slag compositions, significantly improve the detection accuracy of trace elements, and support automatic data analysis and report generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention provides a flow chart of a method for analyzing harmful elements in blast furnace slag. DETAILED DESCRIPTION

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0029] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0030] Example 1 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0031] The steps of preparing the sample to be analyzed are specifically as follows: Dry the collected samples to be analyzed. Specifically, the collected samples to be analyzed can be placed in an oven, the oven temperature is set to 90°C, and dried for 1 hour. The samples are taken out and cooled to room temperature. Free water and bound water in the samples are removed by heating to avoid moisture from generating interference signals in spectral analysis. The dried slag sample is ground to a particle size of <100μm. Specifically, the dried slag sample can be placed in a grinding jar, and grinding balls are added, with the mass ratio of the slag sample to the grinding balls being 1:10; the grinding is performed for 20 minutes at a rotation speed of 300rpm. After the grinding is completed, the sample is taken out, and the slag sample is ground into a uniform powder (particle size <100μm) by mechanical force to eliminate the interference of the particle size on the spectral analysis; The ground slag sample is placed in a sample mixer for mixing. Specifically, the ground slag powder can be placed in the sample mixer for 10 minutes at a speed of 50 rpm. The sample powder is fully mixed through three-dimensional motion to ensure uniform distribution of chemical components. The mixed slag sample is pressed into a tablet, and the mixed slag powder is placed in a tablet pressing mold with a pressure of 10t and a holding time of 25s. After the pressing is completed, the sample tablet is taken out and the powder sample is pressed into a solid tablet by high pressure to reduce the unevenness of the sample surface and improve the accuracy of spectral analysis.

[0032] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument and place the sample sheet on the LIBS sample stage; Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0033] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0034] Furthermore, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0035] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0036] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0037] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0038] Furthermore, the steps of using slag samples with known composition to train the machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0039] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0040] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0041] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0042] Example 2 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0043] The steps of preparing the sample to be analyzed are specifically as follows: Dry the collected samples to be analyzed. Specifically, the collected samples to be analyzed can be placed in an oven, the oven temperature is set to 105°C, and dried for 1.5 seconds. The samples are taken out and cooled to room temperature. Free water and bound water in the samples are removed by heating to avoid water from generating interference signals in spectral analysis. The dried slag sample is ground to a particle size of <100μm. Specifically, the dried slag sample can be placed in a grinding jar, and grinding balls are added, with the mass ratio of the slag sample to the grinding balls being 1:12; the grinding is performed for 30 minutes at a rotation speed of 350rpm. After the grinding is completed, the sample is taken out, and the slag sample is ground into a uniform powder (particle size <100μm) by mechanical force to eliminate the interference of the particle size on the spectral analysis; The ground slag sample is placed in a sample mixer for mixing. Specifically, the ground slag powder can be placed in the sample mixer for 15 minutes at a speed of 60 rpm. The sample powder is fully mixed through three-dimensional motion to ensure uniform distribution of chemical components. The mixed slag sample is pressed into a tablet, and the mixed slag powder is placed in a tablet pressing mold with a pressure of 11t and a holding time of 30s. After the pressing is completed, the sample tablet is taken out and the powder sample is pressed into a solid tablet by high pressure to reduce the unevenness of the sample surface and improve the accuracy of spectral analysis.

[0044] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument and place the sample sheet on the LIBS sample stage; Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0045] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0046] Furthermore, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0047] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0048] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0049] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0050] Furthermore, the steps of using slag samples with known composition to train the machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0051] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0052] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0053] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0054] Example 3 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0055] The steps of preparing the sample to be analyzed are specifically as follows: Dry the collected samples to be analyzed. Specifically, the collected samples to be analyzed can be placed in an oven, the oven temperature is set to 120°C, and dried for 2 hours. The samples are taken out and cooled to room temperature. Free water and bound water in the samples are removed by heating to avoid moisture from generating interference signals in spectral analysis. The dried slag sample is ground to a particle size of <100 μm. Specifically, the dried slag sample can be placed in a grinding jar, and grinding balls are added, with the mass ratio of the slag sample to the grinding balls being 1:15; the grinding is performed for 40 minutes at a rotation speed of 400 rpm. After the grinding is completed, the sample is taken out, and the slag sample is ground into a uniform powder (particle size <100 μm) by mechanical force to eliminate the interference of the particle size on the spectral analysis; The ground slag sample is placed in a sample mixer for mixing. Specifically, the ground slag powder can be placed in the sample mixer for 20 minutes at a speed of 70 rpm. The sample powder is fully mixed through three-dimensional motion to ensure uniform distribution of chemical components. The mixed slag sample is pressed into a tablet, and the mixed slag powder is placed in a tablet pressing mold with a pressure of 12t and a holding time of 35s. After the pressing is completed, the sample tablet is taken out and the powder sample is pressed into a solid tablet by high pressure to reduce the unevenness of the sample surface and improve the accuracy of spectral analysis.

[0056] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument and place the sample sheet on the LIBS sample stage; Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0057] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0058] Furthermore, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0059] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0060] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0061] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0062] Furthermore, the steps of using slag samples with known composition to train the machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0063] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0064] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0065] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0066] Example 4 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0067] The steps of preparing the sample to be analyzed are specifically as follows: Microwave digestion technology is used to digest the samples to be analyzed; The digested slag samples are adsorbed by nanomaterials; The filter residue sample after nanomaterial adsorption is eluted, and the eluate is collected for subsequent analysis.

[0068] Furthermore, the steps of using microwave digestion technology to digest the sample to be analyzed are specifically as follows: Grind the sample to be analyzed to a particle size of <100 μm; Weigh 0.4 g of sample and put it into a microwave digestion tank with a microwave power of 700 W; 10 mL of mixed acid was added, wherein the mixed acid was a mixture of nitric acid and hydrofluoric acid in a volume ratio of 2:1.

[0069] The mixture was reacted at 160°C for 20 minutes and cooled to room temperature after digestion.

[0070] Microwave energy causes the acid solution to heat up rapidly, accelerating the decomposition of the mineral phase in the slag and releasing harmful elements. The mixed acid system of nitric acid and hydrofluoric acid can effectively dissolve insoluble minerals such as silicates.

[0071] Furthermore, the step of adsorbing the digested slag sample by nanomaterials is specifically as follows: The microwave digested solution was diluted to 40 mL with deionized water, and the pH of the solution was adjusted to 5.0; Add 8 mg of functionalized titanium dioxide nanotubes and stir at room temperature for 25 min to allow the nanomaterials to fully absorb harmful elements; The nanomaterials and the solution were separated by a centrifuge with a rotation speed of 4500 rpm and a separation time of 10 min.

[0072] The thiol (-SH) groups on the surface of functionalized titanium dioxide nanotubes have highly selective adsorption capacity for harmful elements (such as lead, cadmium, mercury, etc.).

[0073] In a specific implementation, the steps of eluting the filter residue sample after the nanomaterial adsorption and collecting the eluate for subsequent analysis are specifically as follows: The nanomaterials that adsorb harmful elements are placed in a centrifuge tube; Add 3 mL of 0.1 M nitric acid solution and stir at room temperature for 10 min to elute the harmful elements from the nanomaterials; The eluent and the nanomaterials were separated by a centrifuge at a speed of 4500 rpm and a separation time of 10 min; The eluate was collected for subsequent analysis.

[0074] Nitric acid solution can destroy the chemical bonds between nanomaterials and harmful elements, achieve efficient elution, and use a small volume of eluent, thereby concentrating harmful elements and improving analytical sensitivity.

[0075] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument. If it is a sample sheet, place it on the LIBS sample stage. If it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the LIBS instrument to analyze the film. Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0076] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0077] Preferably, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0078] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0079] Normalization formula: ; Where I is the original strength, I min and Imax are the minimum and maximum values ​​of the spectral intensity, respectively.

[0080] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0081] In this embodiment, the steps of using a slag sample with known composition to train a machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0082] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0083] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0084] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0085] Example 5 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0086] The steps of preparing the sample to be analyzed are specifically as follows: Microwave digestion technology is used to digest the samples to be analyzed; The digested slag samples are adsorbed by nanomaterials; The filter residue sample after nanomaterial adsorption is eluted, and the eluate is collected for subsequent analysis.

[0087] Furthermore, the steps of using microwave digestion technology to digest the sample to be analyzed are specifically as follows: Grind the sample to be analyzed to a particle size of <100 μm; Weigh 0.5 g of sample and put it into a microwave digestion tank with a microwave power of 800 W; 10 mL of mixed acid was added, wherein the mixed acid was a mixture of nitric acid and hydrofluoric acid in a volume ratio of 3:1.

[0088] The reaction was carried out at 180°C for 25 minutes and then cooled to room temperature after digestion.

[0089] Microwave energy causes the acid solution to heat up rapidly, accelerating the decomposition of the mineral phase in the slag and releasing harmful elements. The mixed acid system of nitric acid and hydrofluoric acid can effectively dissolve insoluble minerals such as silicates.

[0090] Furthermore, the step of adsorbing the digested slag sample by nanomaterials is specifically as follows: The microwave digested solution was diluted to 50 mL with deionized water, and the pH of the solution was adjusted to 5.0; Add 10 mg of functionalized titanium dioxide nanotubes and stir at room temperature for 30 min to allow the nanomaterials to fully absorb harmful elements; The nanomaterials and the solution were separated by a centrifuge with a rotation speed of 5000 rpm and a separation time of 12 min.

[0091] The thiol (-SH) groups on the surface of functionalized titanium dioxide nanotubes have highly selective adsorption capacity for harmful elements (such as lead, cadmium, mercury, etc.).

[0092] In a specific implementation, the steps of eluting the filter residue sample after the nanomaterial adsorption and collecting the eluate for subsequent analysis are specifically as follows: The nanomaterials that adsorb harmful elements are placed in a centrifuge tube; Add 5 mL of 0.15 M nitric acid solution and stir at room temperature for 12 min to elute the harmful elements from the nanomaterials; The eluent and the nanomaterials were separated by a centrifuge at a speed of 5000 rpm and a separation time of 12 min; The eluate was collected for subsequent analysis.

[0093] Nitric acid solution can destroy the chemical bonds between nanomaterials and harmful elements, achieve efficient elution, and use a small volume of eluent, thereby concentrating harmful elements and improving analytical sensitivity.

[0094] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument. If it is a sample sheet, place it on the LIBS sample stage. If it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the LIBS instrument to analyze the film. Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0095] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0096] Preferably, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0097] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0098] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0099] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0100] In this embodiment, the steps of using a slag sample with known composition to train a machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0101] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0102] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0103] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0104] Example 6 The embodiment of the present invention provides a method for analyzing harmful elements in blast furnace slag, such as Figure 1 As shown, the following steps are included: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

[0105] The steps of preparing the sample to be analyzed are specifically as follows: Microwave digestion technology is used to digest the samples to be analyzed; The digested slag samples are adsorbed by nanomaterials; The filter residue sample after nanomaterial adsorption is eluted, and the eluate is collected for subsequent analysis.

[0106] Furthermore, the steps of using microwave digestion technology to digest the sample to be analyzed are specifically as follows: Grind the sample to be analyzed to a particle size of <100 μm; Weigh 0.6 g of sample and put it into a microwave digestion tank with a microwave power of 900 W; Add 10 mL of mixed acid, wherein the mixed acid is a mixture of nitric acid and hydrofluoric acid in a volume ratio of 2-4:1.

[0107] The reaction was carried out at 200°C for 30 minutes and then cooled to room temperature after digestion.

[0108] Microwave energy causes the acid solution to heat up rapidly, accelerating the decomposition of the mineral phase in the slag and releasing harmful elements. The mixed acid system of nitric acid and hydrofluoric acid can effectively dissolve insoluble minerals such as silicates.

[0109] Furthermore, the step of adsorbing the digested slag sample by nanomaterials is specifically as follows: The microwave digested solution was diluted to 60 mL with deionized water, and the pH of the solution was adjusted to 5.0; Add 12 mg of functionalized titanium dioxide nanotubes and stir at room temperature for 35 min to allow the nanomaterials to fully absorb harmful elements; The nanomaterials and the solution were separated by a centrifuge with a rotation speed of 5500 rpm and a separation time of 15 min.

[0110] The thiol (-SH) groups on the surface of functionalized titanium dioxide nanotubes have highly selective adsorption capacity for harmful elements (such as lead, cadmium, mercury, etc.).

[0111] In a specific implementation, the steps of eluting the filter residue sample after the nanomaterial adsorption and collecting the eluate for subsequent analysis are specifically as follows: The nanomaterials that adsorb harmful elements are placed in a centrifuge tube; Add 7 mL of 0.2 M nitric acid solution and stir at room temperature for 15 min to elute the harmful elements from the nanomaterials; The eluent and the nanomaterials were separated by a centrifuge at a speed of 5500 rpm and a separation time of 15 min; The eluate was collected for subsequent analysis.

[0112] Nitric acid solution can destroy the chemical bonds between nanomaterials and harmful elements, achieve efficient elution, and use a small volume of eluent, thereby concentrating harmful elements and improving analytical sensitivity.

[0113] In this embodiment, the steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument. If it is a sample sheet, place it on the LIBS sample stage. If it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the LIBS instrument to analyze the film. Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra. Collect spectral data 10 times for each sample and take the average value. Use high-energy laser pulses to excite the slag sample to generate plasma. Determine the type and content of harmful elements by analyzing the plasma emission spectrum. Place the sample in the XRF instrument. If it is a sample sheet, place it in the XRF sample chamber; if it is an eluent, drop it on a glass sheet and form a uniform film after drying. Use the XRF instrument to analyze the film. Start the X-ray source, excite the sample and collect characteristic X-ray fluorescence spectra; Spectral data were collected three times for each sample and the average value was taken. The sample was excited by X-rays, and the characteristic X-ray fluorescence was detected to determine the type and content of the element.

[0114] Combining LIBS and XRF, multiple harmful elements can be analyzed simultaneously to improve detection efficiency.

[0115] Preferably, the step of preprocessing the collected spectral data is specifically as follows: Data import and organization: Import spectral data collected by LIBS and XRF into data processing software (such as Python or MATLAB); Organize the data to ensure that the spectral data of each sample corresponds to its number and element content information; Denoise and normalize the spectral data: Specific: Denoising processing: Use wavelet transform to denoise the LIBS spectral data, select the "db4" wavelet basis function, and set the decomposition level to 5; The spectral curves were smoothed using a Savitzky-Golay filter (window size of 11 and polynomial order of 3); The XRF spectrum data were smoothed using a Gaussian filter with a filter width of 5.

[0116] Normalization: The LIBS and XRF spectra were normalized and the intensity values ​​were scaled to the range of [0, 1].

[0117] Normalization formula: ; Where I is the original strength, I min and I max are the minimum and maximum values ​​of the spectral intensity, respectively.

[0118] Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements: Specifically, the characteristic peaks in the LIBS spectrum (such as 180.7 nm for sulfur, 213.6 nm for phosphorus, 405.8 nm for lead, etc.) are extracted; the characteristic energy peaks in the XRF spectrum (such as the Kα line of sulfur 2.31 keV, the Kα line of phosphorus 2.01 keV, etc.) are extracted.

[0119] In this embodiment, the steps of using a slag sample with known composition to train a machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set (≥500 groups) was constructed using the spectral data of slag samples with known composition; each group of data included spectral features and corresponding element contents (calibrated by traditional chemical analysis methods).

[0120] Model training: Spectral features are used as input and element content as output to train support vector machine (SVM) and random forest models. Cross-validation method is used to optimize model parameters to ensure model generalization ability.

[0121] Model evaluation: The model accuracy was evaluated using the test set, with a target mean square error (MSE) < 0.01 and R² > 0.98.

[0122] Preferably, the step of visualizing the result output is specifically as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

[0123] It should be noted that, for the above embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, some steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0124] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in this field can still combine, add, delete or make other adjustments to the features in the various embodiments of the present invention according to the circumstances without conflict, without making creative work, so as to obtain different other technical solutions that do not deviate from the concept of the present invention in essence, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A method for analyzing harmful elements in blast furnace slag, characterized in that: The steps include: Prepare samples for analysis; Collect LIBS and XRF spectral data of the samples to be analyzed; Preprocessing the collected spectral data; Use slag samples with known composition to train machine learning models and establish a mapping relationship between spectral data and elemental content; The preprocessed spectral data is input into the trained machine learning model to output the predicted value of the content of each harmful element; Result visualization output.

2. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of preparing the sample to be analyzed are specifically: Drying the collected samples to be analyzed; The dried slag samples were ground to a particle size of <100 μm; The ground slag sample is placed in a sample mixer for mixing; The mixed slag samples are pressed into tablets.

3. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of preparing the sample to be analyzed are specifically: Microwave digestion technology is used to digest the samples to be analyzed; The digested slag samples are adsorbed by nanomaterials; The filter residue sample after nanomaterial adsorption is eluted, and the eluate is collected for subsequent analysis.

4. The method for analyzing harmful elements in blast furnace slag according to claim 3, characterized in that: The steps of using microwave digestion technology to digest the sample to be analyzed are specifically as follows: Grind the sample to be analyzed to a particle size of <100 μm; Weigh 0.4-0.6g of sample and put it into a microwave digestion tank with a microwave power of 700-900W; Add 10 mL of mixed acid, wherein the mixed acid is a mixture of nitric acid and hydrofluoric acid in a volume ratio of 2-4:1; At a temperature of 160-200°C, react for 20-30 minutes, and cool to room temperature after digestion.

5. The method for analyzing harmful elements in blast furnace slag according to claim 3, characterized in that: The step of adsorbing the digested slag sample by nanomaterials is specifically as follows: Dilute the microwave digested solution to 40-60 mL with deionized water and adjust the solution pH to 5.0; Add 8-12 mg of functionalized titanium dioxide nanotubes and stir at room temperature for 25-35 min to allow the nanomaterials to fully absorb harmful elements; The nanomaterials and the solution are separated by a centrifuge with a rotation speed of 4500-5500 rpm and a separation time of 10-15 minutes.

6. The method for analyzing harmful elements in blast furnace slag according to claim 3, characterized in that: The steps of eluting the filter residue sample after the nanomaterial adsorption and collecting the eluate for subsequent analysis are specifically as follows: The nanomaterials that adsorb harmful elements are placed in a centrifuge tube; Add 3-7 mL of 0.1-0.2 M nitric acid solution and stir at room temperature for 10-15 min; Separating the eluate from the nanomaterials by centrifugation; The eluate was collected for subsequent analysis.

7. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of collecting LIBS and XRF spectrum data of the sample to be analyzed are specifically as follows: Place the sample in the LIBS instrument; Adjust the laser focus to the sample surface to ensure that the laser energy density reaches the sample breakdown threshold; Collect plasma emission spectra, collect spectral data 10 times for each sample and take the average value; Place the sample in the XRF instrument; Start the X-ray source, excite the sample and collect the characteristic X-ray fluorescence spectrum.

8. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of preprocessing the collected spectral data are specifically as follows: Data import and organization; De-noising and normalizing the spectral data; Extract spectral characteristic peaks and identify characteristic wavelengths of harmful elements.

9. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of using a slag sample with known composition to train a machine learning model and establish a mapping relationship between spectral data and element content are specifically as follows: Dataset construction: The training set is constructed using the spectral data of slag samples with known composition; each set of data includes spectral features and corresponding element content; Model training: Use spectral features as input and element content as output to train support vector machine and random forest models, and use cross-validation to optimize model parameters to ensure model generalization ability; Model evaluation: Use the test set to evaluate the model accuracy.

10. The method for analyzing harmful elements in blast furnace slag according to claim 1, characterized in that: The steps of visualizing the results are as follows: The prediction results are displayed in a graphical form, including element types, contents and confidence intervals; Supports data export and report generation.

Images