Systematic analysis method for thallium in ore by super-microwave digestion and ICPMS determination

Abstract

This invention provides a method and system for the supermicrowave digestion and determination of thallium in ores. Based on X-ray fluorescence spectroscopy, the matrix insolubility index is calculated to establish the digestion decision variables. Redundant points are eliminated using Sobol sequences combined with Thiessen polygons, and a non-uniform experimental design is constructed to obtain training data. On this basis, a hybrid response surface model is constructed, and kernel partial least squares regression is applied to fit the main trend, Gaussian process regression is used to process the residuals, and the interference index is introduced as a hyperparameter. The Grey Wolf optimization algorithm, incorporating Lévy flight, is used to optimize the ore while ensuring low residual organic carbon and maximizing the recovery rate. The obtained optimal parameters are used for supermicrowave digestion of the ore, and the thallium content is determined by ICP-MS, achieving efficient and accurate detection.

Description

Technical Field

[0001] This application belongs to the field of determination, and in particular relates to a systematic analytical method for the determination of thallium in ores by super microwave digestion and ICPMS. Background Technology

[0002] Thallium (Tl) is a typical rare and highly toxic heavy metal element, often found in sulfide deposits. Currently, inductively coupled plasma mass spectrometry (ICP-MS) has become the mainstream method for determining trace thallium in geological samples due to its high sensitivity and wide linear range. However, the accuracy of ICP-MS analysis is greatly limited by the effectiveness of sample pretreatment. The matrix composition of ore samples is complex and variable. Traditional hotplate wet digestion or conventional microwave digestion often faces problems such as incomplete decomposition of sparingly soluble minerals, high blank values ​​due to large acid dosage, and long processing times. Supermicrowave digestion technology, with its unique single-chamber pre-pressurization and ultra-high temperature and pressure advantages, can largely solve these problems. However, subsequent ICP-MS determination methods often struggle to ensure complete thallium release while effectively controlling residual organic carbon content, making it difficult to predict the degree of matrix interference and thus affecting the stability and recovery rate of subsequent mass spectrometry signals. Summary of the Invention

[0003] To address the aforementioned problems, this invention proposes a systematic analytical method for the super-microwave digestion and ICP-MS determination of thallium in ores by constructing modules for parameter setting, calculation, sample training, update optimization, and determination. The specific steps include: (1) Obtain X-ray fluorescence spectrum data of the ore, calculate the ratio of the total intensity of these major elements to the Compton scattering intensity based on the characteristic peak intensities of silicon, aluminum, iron and calcium, and obtain the matrix insolubility index; set the microwave heating power, the maximum digestion temperature, the pre-pressure and the volume ratio of nitric acid to hydrofluoric acid as decision variables and define the range of values. (2) The initial sample space is generated using the Sobol sequence, the volume of the Thiessen polygon of the sample points is calculated, highly correlated redundant points with excessive local space crowding are removed, a non-uniform space filling test design matrix is ​​constructed, a digestion test scheme is output, and the ore is digested by super microwave according to the scheme. The thallium element recovery rate and residual organic carbon content are determined to form a training sample set. (3) For the recovery rate of thallium and the content of residual organic carbon, the main trends of decision variables and response values ​​are fitted by kernel partial least squares regression, the residual components are fitted by Gaussian process regression, and the matrix insolubility index is mapped to the hyperparameter of the covariance function. The model weight coefficients are updated by maximizing the log-likelihood function to construct a hybrid response surface model. (4) Establish an objective function with the constraint of maximizing the thallium recovery rate and the residual organic carbon content being lower than the threshold. Use the gray wolf optimization algorithm that integrates the Lévy flight strategy to find the optimal solution. In the individual position update, superimpose the Lévy distribution random step size and iteratively calculate the output of the optimal digestion parameter combination. Use the optimal digestion parameter combination to perform super microwave digestion of the ore and determine the thallium content by ICP-MS.

[0004] This invention quantifies the potential interference of complex ore matrix background on thallium determination by acquiring X-ray fluorescence spectral data of the ore and calculating the matrix insolubility index. By using Sobol sequences combined with Thiessen polygon volumes to screen sample points and eliminate redundant data with excessive spatial crowding, an efficient non-uniform space-filling experimental design is constructed, reducing experimental costs and time. A hybrid strategy of kernel partial least squares regression fitting the main trend and Gaussian process regression fitting the residual components is adopted, and the matrix insolubility index is mapped as a covariance function hyperparameter, greatly improving the model's ability to fit nonlinear relationships and its generalization performance. The gray wolf optimization algorithm, incorporating the Lévy flight strategy, utilizes the long-tail characteristics of the Lévy distribution to enhance global search capabilities, ensuring the acquisition of the optimal digestion process that maximizes thallium recovery while meeting the constraint of low residual organic carbon content. This achieves complete digestion and high-sensitivity, high-accuracy detection of trace thallium in the ore. Attached Figure Description

[0005] Figure 1 This is a flowchart of a specific embodiment. Detailed Implementation

[0006] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0007] See Figure 1 The systematic analytical method for super microwave digestion and ICPMS determination of thallium in ore, as shown, includes the following steps: S1. Obtain X-ray fluorescence spectrum data of the ore, extract the characteristic peak intensities of major elements such as silicon, aluminum, iron and calcium, calculate the ratio of the total intensity of major elements to the Compton scattering intensity to obtain the matrix insolubility index; set microwave heating power, maximum digestion temperature, pre-pressure and volume ratio of nitric acid to hydrofluoric acid as decision variables and define the range of values.

[0008] The ore sample to be tested was ground until it passed through a 200-mesh sieve. 4 grams of sample powder were weighed and placed into a tablet press mold. The tablet was pressed into a disc under a pressure of 30 MPa for 30 seconds. The disc was placed in the sample chamber of an X-ray fluorescence spectrometer. The X-ray tube voltage was set to 50 kV and the current to 50 mA. A rhodium target was selected as the target material. The Ka characteristic spectral line intensities of silicon, aluminum, iron, and calcium were measured. At the same time, the Compton scattering peak intensity generated by the characteristic radiation of the rhodium target was measured. The dimensionless ratio obtained by summing the characteristic spectral line intensities of silicon, aluminum, iron, and calcium and dividing it by the Compton scattering peak intensity was used as the matrix insolubility index. Based on the limitations of the microwave digester's equipment parameters and the characteristics of the chemical reagents, the microwave heating power is set to a range of 1000 watts to 1500 watts, the maximum digestion temperature to a range of 200 degrees Celsius to 260 degrees Celsius, the pre-pressure to a range of 30 to 50 bar, and the volume ratio of nitric acid to hydrofluoric acid to a range of 2:1 to 10:1.

[0009] In some embodiments, the step of acquiring X-ray fluorescence spectral data of the ore, extracting the characteristic peak intensities of major elements such as silicon, aluminum, iron, and calcium, and calculating the ratio of the total intensity of the major elements to the Compton scattering intensity to obtain the matrix insolubility index includes: In X-ray fluorescence spectroscopy, the Ka characteristic peak intensities of silicon, aluminum, iron, and calcium are read and their sum is calculated, which is recorded as the total intensity of the major elements. The intensity of the Compton scattering peak of the X-ray tube target was read in the high-energy region of X-ray fluorescence spectroscopy. The matrix insolubility index is the ratio of the total intensity of the major elements to the intensity of the Compton scattering peak.

[0010] The crushed and compressed ore samples were subjected to a full-spectrum scan using a wavelength dispersive X-ray fluorescence spectrometer. In the spectral analysis software, the characteristic energy positions corresponding to the Ka lines of silicon (Si), aluminum (Al), calcium (Ca), and iron (Fe) were identified. For example, the Ka line for silicon is located at 1.74 keV, and the Ka line for aluminum is located near 1.49 keV. The spectral peaks at these energy positions were automatically integrated, and the integrated intensities of each element were summed to obtain the total intensity of the major elements. For example, for a given sample, the sum of the intensities of silicon, aluminum, iron, and calcium was 50,000 count units.

[0011] The Compton scattering peak intensity of the rhodium target in the X-ray tube was read in the high-energy region. This peak is typically located on the low-energy side of the rhodium Kα characteristic peak. Assuming the measured Compton scattering peak intensity is 5000 count units, the matrix insolubility index was calculated to be 10. A higher value indicates a higher content of insoluble components such as silicates and aluminosilicates that form the mineral framework in the ore, and a stronger scattering effect relative to a heavy matrix. This invention utilizes the ratio of the sum of major elements to Compton scattering to construct the matrix insolubility index. A higher index indicates a more stable ore lattice structure, such as granite or basalt matrices, requiring higher digestion temperatures, pressures, and a higher proportion of hydrofluoric acid to disrupt the lattice; conversely, a lower index indicates a more porous matrix, such as a carbonate matrix, thus precisely guiding the generation of digestion schemes.

[0012] In some embodiments, setting the microwave heating power, maximum digestion temperature, pre-pressure, and volume ratio of nitric acid to hydrofluoric acid as decision variables and defining their value ranges includes: microwave heating power The value range is set to ; The highest digestion temperature The value range is set to ; Pre-pressurization The value range is set to ; The volume ratio of nitric acid to hydrofluoric acid The value range is set to .

[0013] During the experimental design phase, the operator first starts the control system of the microwave digester and enters the parameter setting interface. For the thallium-containing ore sample to be processed, four key process parameters are defined as input decision variables for the optimization algorithm. In the power setting section, the lower limit of the microwave heating power (1200 watts) and the upper limit (1600 watts) are entered to ensure effective destruction of the ore lattice while avoiding excessive energy consumption. In the temperature control module, the maximum digestion temperature is set between 220°C and 260°C; for example, the initial screening point can be set to 240°C to ensure complete dissolution of the insoluble silicate matrix. The pre-pressure parameter of the reaction vessel is adjusted to be within the range of 30 to 60°C to prevent the acid from boiling over at high temperatures.

[0014] The volume ratio of nitric acid to hydrofluoric acid is set to vary from 5:1 to 10:1. For example, if the total acid volume is fixed at 6 ml, the experimental scheme generated by the algorithm might use a ratio of 5 ml nitric acid to 1 ml hydrofluoric acid. These ranges constitute the multidimensional search space of the subsequent optimization algorithm. All generated experimental design points must fall within this four-dimensional hypercube to ensure the safety of experimental operations and the feasibility of the chemical reaction.

[0015] S2 uses the Sobol sequence to generate the initial sample space, calculates the Thiessen polygon volume of the sample points, removes highly correlated redundant points with excessive local space crowding, constructs a non-uniform space filling experimental design matrix, and outputs the elimination experimental scheme.

[0016] In a computer, a Sobol sequence generator is used to generate 100 initial sample points within a four-dimensional unit hypercube. Each sample point corresponds to a set of normalized decision variables. Thiessen polygons are constructed for the 100 sample points, and the geometric volume of each Thiessen polygon is calculated. The matrix insolubility index is used as a weighting factor, and the matrix insolubility index is multiplied by the geometric volume to obtain the weighted volume. The weighted volume threshold is set to 0.5 times the average volume of all polygons. All sample points are traversed. When the weighted volume of a sample point is less than the threshold, the Euclidean distance between the sample point and its nearest neighbor sample point is calculated. If the distance is less than 0.1, it is determined to be a highly correlated redundant point and is removed. The normalized coordinates of the remaining sample points are denormalized to restore the actual physical quantities, forming an experimental design matrix containing several sets of experimental conditions. Each row of the matrix is ​​output as an independent elimination experimental scheme.

[0017] In some embodiments, calculating the Thiessen polygon volume of the sample points includes: Construct Thiessen polygons based on the spatial coordinates of all sample points, and calculate the geometric volume of the Thiessen polygon corresponding to the i-th sample point. ; The geometric volume is defined as the Thiessen polygon volume. .

[0018] All experimental design sample points are treated as seed points in a multidimensional space, and a Thiessen polygon graph is constructed using computational geometry algorithms. For each sample point... The set of all points in space that are closer to a given point than to any other sample point constitutes its corresponding Thiessen polygon region. The geometric volume of this region in the decision variable space is calculated using Monte Carlo integration or analytical geometry methods. For example, in a four-dimensional parameter space, if a sample point is located in a region where the parameter distribution is relatively sparse, the volume of the Thiessen polygon it controls will be relatively large, assuming the calculated result is 0.05 cubic units.

[0019] In this embodiment, the geometric volume is defined as the Thiessen polygon volume. A larger sample point represents a wider area of ​​unknown exploration around it, and such a sample point is of greater importance in maintaining the spatial fill and representativeness of the design space. By controlling the volume of each sample point, a basis is provided for eliminating redundant points in crowded areas, ensuring that the retained sample points can cover the experimental design space to the greatest extent.

[0020] In some embodiments, the process of eliminating highly correlated redundant points with excessive local spatial congestion includes: Map all sample points to In the normalized space, calculate any two sample points and Euclidean distance between d is the dimension of the normalized space; Using the matrix sparing solubility index Set dynamic minimum distance threshold :

[0021] Iterate through all sample point pairs, and when a match is detected... At times, comparison and The volume of the Thiessen polygon; like Then retain the sample points. , sample points Remove it from the design matrix, where c is a coefficient.

[0022] The minimum-maximum normalization method is used to map the four dimensions of microwave power, temperature, pressure, and acid ratio to a dimensionless interval of 0 to 1, thus eliminating the influence of dimensional differences on distance calculation. The value of d is 4. The Euclidean distance between any two sample points in the calculation space is then calculated, for example, the distance between sample A and sample B in the normalized space. The calculated value is 2.12. Simultaneously, the matrix insolubility index obtained in the previous steps is used to dynamically set the screening criteria. Here, the adjustment coefficient c is set to 0.8, assuming the measured matrix insolubility index... The value is 4.0, and the dynamic minimum distance threshold is calculated to be 3.2 using the formula.

[0023] After a thorough screening, since the distance between samples A and B (2.12) is less than the threshold of 3.2, these two points are determined to be highly correlated and redundant. The Thiessen polygon volumes of the two points are then compared; assuming sample A has a volume of 0.08 and sample B has a volume of 0.03. Based on the principle of retaining larger volume points, the region where sample A is located is sparser and better represents the spatial characteristics of that region. Therefore, sample A is retained, while sample B is marked as redundant and removed from the experimental design matrix.

[0024] S3. The ore is subjected to super microwave digestion according to the digestion test plan, and the thallium recovery rate and residual organic carbon content are determined to form a training sample set.

[0025] Weigh 0.1 g of ore sample into a polytetrafluoroethylene digestion tube. Add a mixed solution of nitric acid and hydrofluoric acid according to the volume ratio specified in the digestion test protocol. Add a thallium standard solution of known concentration to one group of samples as a spiked sample. After removing the acid, place the digestion tube into the reaction chamber of a super microwave digester. Purge with high-purity nitrogen gas according to the pre-pressurization set in the protocol. Start the microwave heating program and linearly raise the temperature to the highest digestion temperature set in the protocol within 10 minutes and maintain it for 20 minutes. After digestion, cool to room temperature and depressurize. Transfer the digestion solution to a 50 mL volumetric flask and dilute to volume with water. Use inductively coupled plasma mass spectrometry to determine the thallium content in the spiked and unspecified samples. Calculate the thallium element recovery rate using the spike recovery rate formula. Simultaneously, inject 20 mL of the digestion solution into a total organic carbon analyzer and determine the residual organic carbon content in the solution using a high-temperature catalytic oxidation method. Use the combination of decision variables as input features and the corresponding thallium element recovery rate and residual organic carbon content as output responses. Summarize all experimental data to form a training sample set.

[0026] S4. Construct a hybrid response surface model. For thallium recovery rate and residual organic carbon content, kernel partial least squares regression is used to fit the main trends of decision variables and response values, respectively. Gaussian process regression is used to fit the residual components. The matrix insolubility index is mapped to the hyperparameter of the covariance function. The model weight coefficients are updated by maximizing the log-likelihood function.

[0027] The input features and output responses in the training sample set are centered. A kernel partial least squares regression model is constructed using a radial basis function as the kernel function. The number of principal components is determined through cross-validation. The predicted values ​​of thallium recovery rate and residual organic carbon content are calculated. The predicted values ​​are subtracted from the actual measured values ​​to obtain the residual components. A Gaussian process regression model is then established to fit these residual components. The squared exponential covariance function is selected, and the characteristic length scale parameter in the covariance function is set as the product of the inverse of the matrix insolubility index and the basic length scale. The hyperparameters of the Gaussian process regression model are iteratively optimized using the conjugate gradient method. The logarithmic marginal likelihood function value of the training data is calculated. The logarithmic likelihood function value is maximized by adjusting the model weight coefficients. The predicted values ​​of the kernel partial least squares regression model are added to the predicted residuals of the Gaussian process regression model to obtain the complete mixed response surface model.

[0028] In some embodiments, the step of fitting the principal trends of the decision variables and response values ​​using kernel partial least squares regression includes: The kernel matrix K is constructed using Gaussian radial basis functions as kernel functions; Perform eigenvalue decomposition on the centered kernel matrix, select the top h potential principal components with the largest eigenvalues, and ensure that the cumulative variance contribution rate of the sum of the top h eigenvalues ​​to the sum of all eigenvalues ​​is not less than 95%. A regression model was established based on the extracted h potential principal components to determine the decision variables and the recovery rate of thallium and the residual organic carbon content.

[0029] The Gaussian radial basis function kernel is used to map the original input space to a high-dimensional feature space. For N samples in the experimental design, the kernel function values ​​are calculated pairwise. For example, the kernel parameters are taken. If the squared Euclidean distance between sample i and sample j is 0.5, then the corresponding kernel matrix elements are... An N×N kernel matrix is ​​constructed in this manner, and then centered to eliminate the influence of the intercept.

[0030] Perform eigenvalue decomposition on the kernel matrix to obtain a series of eigenvalues. The eigenvalues ​​and their corresponding eigenvectors are then analyzed. The eigenvalues ​​are arranged in descending order, and the cumulative variance contribution rate is calculated. Assuming the sum of the eigenvalues ​​of the first four principal components accounts for 96% of the total sum of eigenvalues, satisfying the threshold requirement of greater than or equal to 95%, the number of potential principal components, h=4, is determined. Using these four extracted principal components as new independent variables, linear regression equations are established for the two response variables: thallium recovery rate and residual organic carbon content, respectively, thus completing the mathematical modeling of the main trend.

[0031] In some embodiments, mapping the matrix insolubility index to a hyperparameter of the covariance function includes: Using the squared exponential covariance function Constructing a Gaussian process regression model: ; in, For signal variance, For length scale parameters; Using the matrix sparing solubility index Set length scale parameters initial value The formula is as follows:

[0032] by Starting with the log-likelihood function, we maximize the hyperparameters. Perform iterative updates, where k is a coefficient.

[0033] When modeling the residuals of kernel partial least squares using Gaussian process regression, the squared exponential covariance function is chosen to describe the correlation between samples. The length scale parameter *l* determines the smoothness of the function's fluctuations. To avoid getting trapped in local optima, this embodiment utilizes the matrix insolubility index to provide a physical prior for hyperparameter optimization. It is assumed that the matrix insolubility index is measured by X-ray fluorescence spectroscopy. If the value is 5.0, the matrix is ​​relatively simple, and the data fluctuations may be relatively mild.

[0034] Calculate the initial guess value of the length scale parameter according to the formula. .Will The initial signal variance is then substituted into the log-marginal likelihood function of the Gaussian process. Optimization algorithms such as the conjugate gradient method are used to iteratively search for the hyperparameter combination that maximizes the log-likelihood function, starting from this initial value.

[0035] S5 establishes an objective function constrained by maximizing thallium recovery rate and keeping residual organic carbon content below a threshold. The gray wolf optimization algorithm, which incorporates the Lévy flight strategy, is used to find the optimal solution. The Lévy distribution random step size is superimposed during individual position updates, and the optimal solution parameter combination is iteratively calculated and output.

[0036] The objective function is defined as the negative predicted thallium recovery rate, and a penalty function term is introduced. When the predicted residual organic carbon content is greater than 0.1%, a penalty value of 1000 is added to the objective function. The gray wolf population is initialized with a population size of 30 and a maximum number of iterations of 100. The fitness value of each gray wolf is calculated, and the three wolves with the best fitness are selected and labeled as α, β, and δ, respectively, with the rest labeled as ω. When updating the gray wolf position in each iteration, the movement vectors toward α, β, and δ are first calculated according to the basic formula of the gray wolf optimization algorithm. Then, a random step size following a Lévy distribution is generated using the Mantegna algorithm. This random step size is multiplied by the step size control factor and superimposed on the movement vector to obtain the new individual position. It is checked whether the new position exceeds the range of the decision variables. If it does, boundary projection is performed. The iteration is repeated until the maximum number of iterations is reached. The position coordinates corresponding to the α wolf with the best fitness are output, which are the optimal microwave heating power, maximum digestion temperature, pre-pressure, and volume ratio of nitric acid to hydrofluoric acid.

[0037] S6. The ore is subjected to super microwave digestion using the optimal digestion parameter combination, and the thallium content is determined by inductively coupled plasma mass spectrometry.

[0038] Weigh 0.05 g of the ore sample to be tested and place it in a digestion tube. Add nitric acid and hydrofluoric acid according to the volume ratio in the optimal digestion parameter combination. Place the tube in a super microwave digester, set the microwave power to the optimal microwave heating power, and the pre-pressure to the optimal pre-pressure. Run the heating program to the optimal maximum digestion temperature for digestion. After cooling and making up to volume, adjust the radio frequency power of the inductively coupled plasma mass spectrometer to 1550 W and the carrier gas flow rate to 1.0 L / min. Select 205Tl as the isotope to be determined and 103Rh as the internal standard element. Add the internal standard solution online and measure the signal intensity of the sample solution and the standard series solutions. Plot the standard curve and calculate the mass fraction of thallium in the ore.

[0039] In an optional embodiment, superimposing the Lévy distribution random step size in the individual location update includes: Generating Lévy random step size using the Mantegna algorithm : ; in, , Levy Index , for: ; Calculate the current gray wolf individual Encirclement step size vectors for Alpha, Beta, and Delta wolves in the population , , : ; in This represents the coefficient vector of the Grey Wolf optimization algorithm; Calculate the number using the following formula. The location of the gray wolf in the next iteration Add Levy flight perturbation to the base position update: ; Among them, step size factor .

[0040] To enhance the algorithm's ability to escape local optima when using the Grey Wolf optimization algorithm to find the optimal combination of elimination parameters, a Lévy flight mechanism is introduced. The program calls the Gamma function from the standard mathematical library to calculate the parameters. Set the Levy index Generate random numbers u and v that follow a normal distribution, and use the Mantegna algorithm to synthesize a random step size with long-tail characteristics. During the position update phase of the gray wolf population, the approximation vectors of ordinary gray wolves relative to the three best wolves in the population (Alpha, Beta, Delta) are calculated. This represents the guiding direction of collective wisdom.

[0041] This embodiment adds a perturbation term to the average position. This perturbation term is determined by the step size factor. Levy random step size Together with the distance difference between the current wolf and the Alpha wolf, it constitutes the solution. For example, when the program stalls in the later stages, the long jump steps occasionally generated by Levi's flight can cause a drastic change in the current solution X(t+1), which may lead to a jump to a better region in the search space that has not been explored.

[0042] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A systematic analytical method for the super microwave digestion and ICPMS determination of thallium in ore, characterized in that, Includes the following steps: X-ray fluorescence spectral data of the ore were obtained, the characteristic peak intensities of major elements such as silicon, aluminum, iron and calcium were extracted, and the matrix insolubility index was obtained by calculating the ratio of the total intensity of major elements to the Compton scattering intensity. Microwave heating power, maximum digestion temperature, pre-pressure and volume ratio of nitric acid to hydrofluoric acid were set as decision variables and their value ranges were defined. The initial sample space is generated using the Sobol sequence. The Thiessen polygon volume of the sample points is calculated. Highly correlated redundant points with excessive local spatial crowding are removed. A non-uniform space-filling experimental design matrix is ​​constructed, and a solution experimental scheme is output. The ore was subjected to super microwave digestion according to the digestion test plan described above. The thallium recovery rate and residual organic carbon content were determined to form a training sample set. A hybrid response surface model was constructed. For thallium recovery rate and residual organic carbon content, kernel partial least squares regression was used to fit the main trends of decision variables and response values, Gaussian process regression was used to fit the residual components, and the matrix insolubility index was mapped to the hyperparameter of the covariance function. The model weight coefficients were updated by maximizing the log-likelihood function. An objective function is established with the constraint of maximizing the thallium recovery rate and keeping the residual organic carbon content below a threshold. The gray wolf optimization algorithm, which incorporates the Lévy flight strategy, is used to find the optimal solution. The Lévy distribution random step size is superimposed in the individual position update, and the optimal solution parameter combination is output through iterative calculation. The ore was subjected to super microwave digestion using the optimal digestion parameter combination, and the thallium content was determined by inductively coupled plasma mass spectrometry.

2. The method according to claim 1, characterized in that, The process of setting microwave heating power, maximum digestion temperature, pre-pressure, and the volume ratio of nitric acid to hydrofluoric acid as decision variables and defining their value ranges includes: microwave heating power The value range is set to ; The highest digestion temperature The value range is set to ; Pre-pressurization The value range is set to ; The volume ratio of nitric acid to hydrofluoric acid The value range is set to .

3. The method according to claim 1, characterized in that, The process of obtaining X-ray fluorescence spectral data of the ore, extracting the characteristic peak intensities of major elements such as silicon, aluminum, iron, and calcium, and calculating the ratio of the total intensity of the major elements to the Compton scattering intensity to obtain the matrix insolubility index includes: In X-ray fluorescence spectroscopy, the intensity of the Kα characteristic peak of silicon, aluminum, iron and calcium is read and their sum is calculated, which is recorded as the total intensity of the major elements. The intensity of the Compton scattering peak of the X-ray tube target was read in the high-energy region of X-ray fluorescence spectroscopy. The matrix insolubility index is the ratio of the total intensity of the major elements to the intensity of the Compton scattering peak.

4. The method according to claim 1, characterized in that, The calculation of the Thiessen polygon volume of the sample points includes: Construct Thiessen polygons based on the spatial coordinates of all sample points, and calculate the geometric volume of the Thiessen polygon corresponding to the i-th sample point. ; The geometric volume is defined as the Thiessen polygon volume. .

5. The method according to claim 3, characterized in that, The process of eliminating highly correlated redundant points with excessive local spatial congestion includes: Map all sample points to In the normalized space, calculate any two sample points and Euclidean distance between d is the dimension of the normalized space; Using the matrix sparing solubility index Set dynamic minimum distance threshold : ； Iterate through all sample point pairs, and when a match is detected... At times, comparison and The volume of the Thiessen polygon; like Then retain the sample points. , sample points Remove it from the design matrix, where c is a coefficient.

6. The method according to claim 1, characterized in that, The main trends of the decision variables and response values ​​are respectively fitted using kernel partial least squares regression, including: The kernel matrix K is constructed using Gaussian radial basis functions as kernel functions; Perform eigenvalue decomposition on the centered kernel matrix, select the top h potential principal components with the largest eigenvalues, and ensure that the cumulative variance contribution rate of the sum of the top h eigenvalues ​​to the sum of all eigenvalues ​​is not less than 95%. A regression model was established based on the extracted h potential principal components to determine the decision variables and the recovery rate of thallium and the residual organic carbon content.

7. The method according to claim 1, characterized in that, The hyperparameter that maps the matrix insolubility index to a covariance function includes: Using the squared exponential covariance function Constructing a Gaussian process regression model: ； in, For signal variance, For length scale parameters; Using the matrix sparing solubility index Set length scale parameters initial value The formula is as follows: ； by Starting with the log-likelihood function, we maximize the hyperparameters. Perform iterative updates, where k is a coefficient.

8. A super microwave digestion and determination system for thallium in ore, characterized in that, Includes the following modules: The module is used to acquire X-ray fluorescence spectral data of ores, extract the characteristic peak intensities of major elements such as silicon, aluminum, iron and calcium, calculate the ratio of the total intensity of major elements to the Compton scattering intensity to obtain the matrix insolubility index; and set the microwave heating power, the maximum digestion temperature, the pre-pressure and the volume ratio of nitric acid to hydrofluoric acid as decision variables and define the range of values. The calculation module is used to generate an initial sample space using the Sobol sequence, calculate the Thiessen polygon volume of the sample points, remove highly correlated redundant points with excessive local spatial crowding, construct a non-uniform space-filling experimental design matrix, and output a solution experimental scheme. The training sample construction module is used to perform super microwave digestion of ore according to the digestion test scheme, determine the thallium element recovery rate and residual organic carbon content, and form a training sample set; The update module is used to construct a hybrid response surface model. For thallium recovery rate and residual organic carbon content, kernel partial least squares regression is used to fit the main trends of decision variables and response values, Gaussian process regression is used to fit the residual components, and the matrix insolubility index is mapped to the hyperparameter of the covariance function. The model weight coefficients are updated by maximizing the log-likelihood function. The optimization module is used to establish an objective function constrained by maximizing the thallium recovery rate and keeping the residual organic carbon content below a threshold. It employs a gray wolf optimization algorithm that incorporates the Lévy flight strategy to find the optimal solution. The Lévy distribution random step size is superimposed during individual position updates, and the optimal combination of resolution parameters is calculated iteratively. The determination module is used to perform super microwave digestion of the ore using the optimal digestion parameter combination, and then determine the thallium content using an inductively coupled plasma mass spectrometer.

9. The system according to claim 8, characterized in that, The process of setting microwave heating power, maximum digestion temperature, pre-pressure, and the volume ratio of nitric acid to hydrofluoric acid as decision variables and defining their value ranges includes: microwave heating power The value range is set to ; The highest digestion temperature The value range is set to ; Pre-pressurization The value range is set to ; The volume ratio of nitric acid to hydrofluoric acid The value range is set to .

10. The system according to claim 8, characterized in that, The process of obtaining X-ray fluorescence spectral data of the ore, extracting the characteristic peak intensities of major elements such as silicon, aluminum, iron, and calcium, and calculating the ratio of the total intensity of the major elements to the Compton scattering intensity to obtain the matrix insolubility index includes: In X-ray fluorescence spectroscopy, the intensity of the Kα characteristic peak of silicon, aluminum, iron and calcium is read and their sum is calculated, which is recorded as the total intensity of the major elements. The intensity of the Compton scattering peak of the X-ray tube target was read in the high-energy region of X-ray fluorescence spectroscopy. The matrix insolubility index is the ratio of the total intensity of the major elements to the intensity of the Compton scattering peak.

Images