A method for rapidly determining the composition of rare earth ore by eliminating the interference of moisture and minerals

Abstract

The application provides a rare earth ore composition rapid value determination method for eliminating moisture and mineral interference, comprising the following steps: first, obtaining near-infrared diffuse reflection spectrum of a sample to be measured; extracting microphase state characteristics and reconstructing matrix physical parameters of a detection site through nonlinear inversion; then, performing physical simulation of a photon transmission path based on the parameters to dynamically synthesize a virtual theoretical energy spectrum matched with a current physical state; subsequently, constructing a feature residual error objective function of a measured X-ray fluorescence energy spectrum and the virtual theoretical energy spectrum; finally, performing collaborative feedback iteration of physical parameters and chemical concentrations based on the function to output rare earth element value determination results. The application establishes a deep correlation between near-infrared microfeatures and X-ray fluorescence physical paths, eliminates the enhancement-absorption effect of moisture and mineral matrix in ionic rare earth ore from a physical bottom dimension, and solves the industry pain points of large environmental interference, poor linearity and low value determination precision of traditional in-situ detection methods.

Description

Technical Field

[0001] This invention relates to the fields of nuclear analysis technology and geological environment monitoring, and more specifically, to a rapid method for determining the composition of rare earth minerals by eliminating the interference of moisture and minerals. Background Technology

[0002] Ion-adsorption rare earth minerals are highly valuable resources, and in-situ, rapid, and accurate compositional analysis is crucial for mine exploration and dynamic monitoring. Currently, X-ray fluorescence spectroscopy (XRF) is widely used for on-site detection of rare earth minerals due to its non-destructive and efficient characteristics. However, the complex geological environment of ion-adsorption rare earth mineral formations, significantly influenced by heterogeneous matrices, makes it difficult for traditional XRF detection accuracy to meet practical requirements.

[0003] Firstly, the matrix effect (enhancement-absorption effect) is the core factor limiting the accuracy of XRF. The characteristic X-rays of rare earth elements are easily absorbed by lighter elements in the formation (such as water and aluminosilicate minerals) during their propagation path. In actual mining areas, formation water content fluctuates drastically with depth and weather conditions, and the structural differences of clay minerals (such as kaolinite and halloysite) also alter the X-ray scattering background. Traditional empirical coefficient methods or calibration curve methods often rely on static, dry standard soil samples. When faced with high moisture content and dynamically changing composition in the field, these methods often produce severe negative deviations, leading to invalidated values.

[0004] Secondly, most existing in-situ compensation techniques only perform simple statistical corrections for single interference terms, lacking underlying physical modeling of photon propagation paths in complex media. When dealing with coupled interference from moisture and mineral components, the algorithms exhibit poor robustness and struggle to achieve adaptive correction in the absence of standard samples.

[0005] Therefore, there is an urgent need for a rapid method for determining the composition of rare earth minerals that can eliminate the interference of moisture and mineral matrix from the physical level and does not require a large number of physical standard samples. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a rapid method for determining the composition of rare earth minerals by eliminating interference from moisture and minerals. This method aims to solve the problems of low accuracy and poor environmental adaptability of existing X-ray fluorescence techniques for in-situ detection of ion-adsorption rare earth minerals due to moisture fluctuations and differences in the mineral matrix.

[0007] This invention is achieved through the following technical solution: This invention provides a rapid method for determining the composition of rare earth minerals by eliminating interference from moisture and minerals, comprising the following steps: The near-infrared diffuse reflectance spectrum of the sample to be tested is obtained, the microscopic phase characteristics characterizing the mineral framework and moisture state are extracted, and the matrix physical parameters of the detection site are reconstructed in real time. Based on the physical parameters of the substrate, a Monte Carlo simulation of the photon transmission path is performed to dynamically synthesize a virtual theoretical energy spectrum that matches the physical state of the current detection environment. Obtain the measured X-ray fluorescence energy spectrum of the same detection site, and construct the full-spectrum characteristic residual objective function between the measured energy spectrum and the virtual theoretical energy spectrum; Under the dual constraints of the physical framework formed by the physical parameters of the matrix and the energy spectrum background, a cooperative feedback iteration is performed. When the objective function of the characteristic residual satisfies the convergence condition, the fixed value result of the rare earth element to be measured is output.

[0008] Furthermore, the acquisition of the near-infrared diffuse reflectance spectrum of the sample to be tested, the extraction of microscopic phase characteristics characterizing the mineral framework and moisture state, and the real-time reconstruction of the matrix physical parameters of the detection site specifically include: Convert the original diffuse reflectance spectrum into an absorption index energy spectrum; The composite absorption band at the characteristic absorption band in the absorption index energy spectrum is deconvoluted and integrally decomposed to separate the spectral contributions of different physical phases. Physical feature parameters are extracted from the decomposed sub-peaks to construct high-dimensional physical feature vectors; The high-dimensional physical feature vector is input into a one-dimensional convolutional neural network, and the mass fraction of the target component is output. The comprehensive physical parameters of the current detection point are reconstructed based on the mass fraction.

[0009] Specifically, the conversion of the original diffuse reflectance spectral data into absorption index data employs the following mathematical transformation: F(R) = (1- R ) 2 / 2 R； Where R is the original diffuse reflectance and F(R) is the absorption index.

[0010] Specifically, the characteristic absorption bands include composite absorption bands with numerical fluctuations at 1400 nm and 1900 nm; the different physical phase components include kaolinite-structured hydroxyl groups, halloysite-structured hydroxyl groups, adsorbed water, and free water; the mass fraction of the target component includes the mass fraction of kaolinite, the mass fraction of halloysite, and the real-time water content.

[0011] Specifically, the deconvolution integral is based on nonlinear fitting using the Voigt function, which is composed of the convolution of a Gaussian term representing the broadening of inhomogeneous regions and a Lorentz term representing the broadening of natural energy levels; the physical feature parameters include the center position, peak intensity, Gaussian half-width, and Lorentz half-width of each sub-peak.

[0012] Furthermore, the step of performing a Monte Carlo physics simulation of the photon transmission path based on the substrate physical parameters, and dynamically synthesizing a virtual theoretical energy spectrum that matches the current physical state of the detection environment, specifically includes: Load the emission spectrum information of the excitation source and the geometric constraint variables of the instrument to establish a physical simulation scenario consistent with the actual hardware detection environment; Based on the physical parameters of the matrix, the mass attenuation coefficient is determined, photon step size sampling is performed, and single-photon transmission path tracing is performed within the set effective interaction volume of the sample. The interaction type of photon at the detection site is determined based on random numbers. The interaction type includes photoelectric effect, Rayleigh scattering and Compton scattering. The energy and vector direction of the photons are updated in real time based on the interaction type, and the energy spectrum of the photons escaping the sample boundary is reconstructed using the detector response function to obtain the virtual theoretical energy spectrum.

[0013] Specifically, the objective function of the full-spectrum characteristic residual includes: using the least squares method to compare the residual distribution of the measured energy spectrum and the virtual theoretical energy spectrum in terms of the morphology of the scattering background bag in the low-energy region and the intensity of the characteristic peaks in the high-energy region.

[0014] Furthermore, the collaborative feedback iteration specifically includes: The total residual between the measured energy spectrum and the virtual theoretical energy spectrum is decomposed into energy domains, and the dominant error sources of the residual are determined by using the sensitivity matrix. If the morphology of the scattering packet is determined to be mismatched, it is determined to be a deviation in the physical parameters of the matrix, and the execution path will prioritize the correction of the moisture content and component ratio parameters; If the characteristic peak intensity is determined to be mismatched, it is determined to be an element concentration deviation, and the path is executed to correct the mass fraction parameter of the rare earth element to be measured. The parameter vector is updated using a nonlinear optimization algorithm and fed back to the physical simulation step in a loop. The process is repeated until the measured energy spectrum and the virtual theoretical energy spectrum achieve full spectrum matching.

[0015] Specifically, the correction targets for the deviations in the matrix physical parameters include water content, equivalent atomic number, and matrix density.

[0016] Specifically, the convergence condition is that the value of the objective function of the feature residual is less than a preset convergence threshold; when the convergence condition is met, the fixed value result of the rare earth element to be tested is output, and the fixed value result includes the mass fraction of the rare earth element to be tested.

[0017] This invention also discloses a rapid rare earth mineral composition determination system that eliminates interference from moisture and minerals, comprising: The near-infrared spectroscopy acquisition module is used to acquire the near-infrared diffuse reflectance spectrum of the sample to be tested; The physical parameter reconstruction module is used to extract microscopic phase characteristics that characterize the mineral skeleton and moisture state, and to reconstruct the matrix physical parameters of the detection site in real time. The physics simulation module is used to perform Monte Carlo physics simulations of photon transmission paths based on the physical parameters of the substrate, and dynamically synthesize virtual theoretical energy spectra. The X-ray fluorescence acquisition module is used to acquire the measured X-ray fluorescence energy spectrum of the same detection site. The collaborative iterative correction module is used to construct the full-spectrum characteristic residual objective function of the measured energy spectrum and the virtual theoretical energy spectrum, and to perform collaborative feedback iteration under bidirectional constraints. When the convergence condition is met, it outputs the fixed value result of the rare earth element to be measured.

[0018] By adopting the above technical solution, the present invention can achieve the following technical effects: This invention establishes a deep correlation between near-infrared microscopic features and X-ray fluorescence physical paths, and uses real-time extracted mineral phase and moisture characteristics to drive underlying physical simulation. This compensates for energy loss during X-ray transmission from a physical perspective, completely eliminating matrix interference caused by moisture and mineral matrix in heterogeneous mining environments. Simultaneously, this invention uses virtual theoretical energy spectra generated by physical simulation to replace reliance on physical standard samples, solving the problem of lacking standard determination curves in the field and significantly improving the robustness of in-situ determination. Experiments demonstrate that this invention significantly increases the coefficient of determination of detection results, achieving high-precision and rapid determination in dynamic water-bearing environments, providing reliable technical support for the field exploration and dynamic monitoring of rare earth mines. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the overall process for a rapid determination method of rare earth mineral elements to eliminate the interference of moisture and minerals according to the present invention. Figure 2 This is a flowchart of mineral phase identification and parameter reconstruction based on near-infrared (NIR) spectroscopy in an embodiment of the present invention; Figure 3 This is a partial flowchart of the Voigt-CNN-based feature transformation and inversion in steps 103-105 of this embodiment of the invention; Figure 4 This is a flowchart of the full-energy spectral Monte Carlo virtual synthesis process in an embodiment of the present invention; Figure 5 This is a partial flowchart of the Monte Carlo photon interaction decision tree in steps 203-204 of an embodiment of the present invention; Figure 6 This is a flowchart of the dual-path constraint iterative correction process in an embodiment of the present invention; Figure 7 This is a partial flowchart of the multi-parameter collaborative feedback correction logic in step 305 of an embodiment of the present invention; Figure 8 The figures show a comparison of the effects of different methods of the present invention on the determination of the rare earth element yttrium; where (A) is the determination result obtained by the traditional detection method; and (B) is the determination result obtained by the detection method of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0022] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: like Figure 1As shown, this embodiment provides a rapid method for determining rare earth mineral elements to eliminate interference from moisture and minerals. This method eliminates matrix interference in heterogeneous layers by establishing a correlation between the microscopic phase characteristics extracted by near-infrared (NIR) imaging and the XRF physical transport path. The method includes the following core steps: Step 10: Obtain the NIR diffuse reflectance spectrum of the sample to be tested, extract the microscopic phase characteristics through nonlinear inversion, and reconstruct the matrix physical parameters of the detection site in real time, including the equivalent atomic number and matrix density; Step 20: Based on the physical parameters of the substrate, synthesize a virtual theoretical energy spectrum that matches the current physical state through physical simulation of the photon transmission path; Step 30: Obtain the measured XRF at the same detection site and construct the characteristic residual objective function of the measured energy spectrum and the virtual theoretical energy spectrum; Step 40: Perform collaborative feedback iteration under the bidirectional constraints of the physical framework and the energy spectrum background. When the residual objective function satisfies the convergence condition, output the fixed values ​​of the rare earth elements and characteristic pollutants to be measured in the model.

[0025] like Figure 2 As shown in the figure, this embodiment details the first-stage process for eliminating moisture and mineral interference, namely, the mineral phase identification and parameter reconstruction process based on NIR spectroscopy. This process provides accurate physical boundary constraints for subsequent XRF correction by analyzing the mineral microstructure. The specific steps are as follows: Step 101: The system acquires the raw diffuse reflectance spectral data of the ore sample at the detection point through the NIR detection module. R This data contains information on overfrequency and combined vibrations that reflect the characteristics of the clay mineral skeleton and the form of water presence. Step 102, perform physical feature enhancement processing. Convert the original reflectivity R to the absorption index F(R), whose mathematical expression is F(R) = (1- R ) 2 / 2 R, This step eliminates nonlinear interference caused by multiple scattering of mineral particles through physical transformation, thus establishing a quasi-linear correlation between the spectral signal and the component concentration. Step 103: For the coincidence absorption bands near 1400 nm and 1900 nm in the F(R) energy spectrum, the system calls the feature deconvolution algorithm. Using a preset line shape function, the overlapping spectral envelope is mathematically decomposed to separate the structural hydroxyl features and free water features related to the kaolinite and halloysite crystal structures. Step 104: The system extracts physical feature parameters from each of the disassembled sub-peaks. These feature parameters include, but are not limited to, the center position of each sub-peak. Peak intensity Gaussian half-width and Lorenz half-width This is used to construct a high-dimensional feature vector characterizing the current microstructure of the matrix; Step 105: Input the constructed feature vector into a one-dimensional convolutional neural network model. This model performs feature extraction and nonlinear mapping through multiple convolutional kernels, eliminating spectral overlap interference between components caused by hydrogen bonding coupling; Step 106: The system outputs the mass fraction results of the fixed-value components. Specifically, the results include the kaolinite mass fraction W. kaol Halloysite mass fraction W hall and real-time moisture content W water ; Step 107: Based on the mass fraction output above, the system reconstructs the comprehensive physical parameters of the current detection point in real time; specifically, this includes calculating the equivalent atomic number Z. eff and matrix density rho bulk ; Step 108: The reconstructed physical parameters are used as physical path correlation variables and output to the subsequent full-energy spectrum simulation synthesis module, thereby providing a physical basis for eliminating matrix effects in the XRF measurement process.

[0026] like Figure 2 and Figure 3 As shown, this implementation details the local processing flow of CNN-based feature transformation and inversion in steps 103 to 105. This flow performs feature analysis on complex energy spectra using physical line-type functions and combines a deep learning model to achieve accurate inversion of component content. Specifically: Step 1031: The microscopic phase identification module receives the absorption index energy spectrum F(R); Step 1032 involves nonlinear fitting of the system based on the Voigt function. This function consists of the convolution of a Gaussian term representing the inhomogeneous broadening and a Lorentz term representing the broadening of natural energy levels. In this step, the system sets the center position of each potential absorption peak. Peak intensity Gaussian half-width and Lorenz half-width The initial evolution parameters; Step 1033: The system performs physical phase feature stripping. Using a least-squares iterative algorithm, the composite absorption band near 1400 nm and 1900 nm is decomposed into multiple independent physical feature sub-peaks, thereby mathematically separating the contributions of different physical phases such as kaolinite-structured hydroxyl groups, halloysite-structured hydroxyl groups, adsorbed water, and free water. Step 1034: Construct a high-dimensional physical feature vector based on the parameters of each isolated independent sub-peak (physical phase separation). Each independent sub-peak includes: kaolinite peak (PeakA), halloysite peak (PeakB), adsorbed water peak (PeakC), and free water peak (PeakD); the eigenvectors arrange the center frequencies, peak intensities, and bivariate broadening parameters of each phase in tensor order according to wavelength. This step achieves physical dimensionality reduction of the original spectral data; Step 1051: The constructed physical feature vector is input into the convolution kernel of a one-dimensional convolutional neural network (1D-CNN) to perform sliding window operation on the physical feature parameters, capturing the interaction effect between adjacent feature peaks, in order to characterize the nonlinear cooperative interference between different phase states; Step 1052: The system performs deep fusion and nonlinear mapping. Feature compression is performed through the pooling layers of the neural network, and the latent features are enhanced by combining a nonlinear activation function. Step 1053: The final output is the fixed component mass fraction result, specifically including the kaolinite mass fraction W. kaol Halloysite mass fraction W hall Real-time moisture content W water .

[0027] like Figure 4 As shown, the second-stage process for eliminating moisture and mineral interference is illustrated in detail: the full-spectrum physical simulation energy spectrum synthesis process. This process reconstructs the interaction between photons and the heterogeneous matrix in digital space through a built-in Monte Carlo physical simulation, thereby obtaining a reference signal containing complex matrix effect characteristics. The specific steps are as follows: Step 201: The full-energy spectral physics simulation receives the physical parameters of the detection points output by the microscopic phase identification module. These physical parameters include the equivalent atomic number Z. eff Matrix density r hobulk Mass fraction of each component W j and real-time moisture content W water ; Step 202: Load the emission spectrum information of the excitation source and the geometric constraint variables of the instrument. The variables include, but are not limited to, X-ray tube voltage, target characteristic energy spectrum distribution, and detection solid angle parameters, in order to establish a physical simulation scene consistent with the actual hardware detection environment; Step 203: Perform Monte Carlo random sampling of the photon transmission path. The algorithm simulates the motion of incident X-ray photons in a mineral framework composed of kaolinite, halloysite, and free water, and samples the mean free path of the photons in real time according to the mass attenuation coefficient; Step 204: Solve for the characteristic fluorescence yield and scattering energy spectrum distribution. Calculate the probability of the target element producing characteristic fluorescence under stimulation, and simulate the Compton scattering envelope and Rayleigh scattering peak morphology generated by the collision of photons with the light element matrix; Step 205: Perform full-energy-axis signal reconstruction and Gaussian broadening. Convert discrete photon counts into continuous energy spectrum signals, broaden spectral lines through the detector response function, and establish a physical mapping profile of the energy spectrum interface; Step 206: Output the virtual theoretical spectrum I containing complete matrix effect information. theory This spectrum serves as the physical reference for subsequent dual-path constraint iterative correction.

[0028] like Figure 5 As shown, this implementation details the local processing flow of the photon interaction decision tree in steps 203 to 204. This flow simulates the trajectory of a single X-ray photon in the spatial dimension to characterize the enhancement-absorption effect of the heterogeneous matrix on the feature signal. The specific steps are as follows: Step 2031: Load initial photon parameters. These parameters include the initial photon energy E0, spatial position coordinates (x, y, z), and initial vector direction d, used to simulate the excitation process of the sample by primary X-rays. Step 2032 performs photon step-size sampling. The engine calls the equivalent atomic number Z output from the first stage. eff The matrix density ρ determines the cross-sectional data, and a random number generation algorithm is used to sample the path of free of photons in the current medium. This process simulates the physical constraints imposed by moisture and the mineral framework on the depth of X-ray penetration. Step 2033: Update the spatial coordinates of the photon according to the path of freedom L, and execute the boundary determination in step 2034. Determine whether the current photon coordinates are within the set effective interaction volume of the sample. If the photon coordinates are within the sample boundary, a photon collision is determined, and proceed to step 2041. If the photon coordinates are outside the sample boundary, a photon escape is determined, and proceed to step 2046. The system determines whether the escaped photon enters the detector's receiving solid angle. If it does, energy spectrum weighting is performed and recorded as the effective count value of the synthesized energy spectrum. Step 2041: Perform random sampling to determine the interaction type. Based on the physical cross-sectional ratio of the matrix composition, determine the interaction type of the photon. There are three interaction types: Branch A: Photoelectric effect (Step 2042). Determine if the photon is absorbed by matrix atoms, and calculate the characteristic fluorescence yield based on the elemental content ratio. If it excites a fluorescent photon, update the energy and return to step 2032 to continue tracking. Branch B: Rayleigh scattering (Step 2043). Determine if the photon undergoes an elastic collision. In this case, the photon's energy remains unchanged, only its scattering vector direction is updated, and return to step 2032. Branch C: Compton scattering (Step 2044). Determine if the photon undergoes an inelastic collision. Update the photon's scattering angle and energy loss using physical formulas, and return to step 2032. Step 2045: Before the end of each tracking round, the system performs an energy termination check. This checks whether the current photon energy is higher than the detection lower limit threshold E. cut If the energy is too low, the tracking of the photon is terminated, and the process returns to step 2031 to start the next photon simulation. Through the iterative process of the decision tree described above, the transmission path of the photon under specific water content and mineral phase is accurately reconstructed, thereby directly covering and eliminating matrix effect interference in the physical model.

[0029] like Figure 6 As shown, the third-stage process for eliminating moisture and mineral interference is illustrated in detail: the dual-path constraint collaborative iterative correction process. This process achieves dynamic determination of rare earth element values ​​through feature matching between measured energy spectra and physically simulated energy spectra. The specific steps are as follows: Step 301: Simultaneously receive the measured XRF data Ireal acquired by the detection hardware, and the virtual theoretical energy spectrum I generated by the full-spectrum simulation module. theory ; Step 302: Establish a dual-path constraint mechanism, consisting of physical constraints and energy spectrum constraints, to ensure that the calibration process conforms to the physical nature of the geological sample. Constraint 1: Using the rare earth mineral data obtained in step 10, constrain the mean free path of X-ray photons in the heterogeneous matrix and fix the matrix absorption terms. Constraint 2: Using the Compton and Rayleigh scattering background morphology in the measured energy spectrum, constrain the equivalent density and real-time moisture state of the sample and calibrate the scattering background. By establishing dual constraints, the algorithm's search space is limited to a reasonable range. Step 303: Calculate the least squares objective function R for the entire spectrum. This function evaluates the degree of matching between the current physical model and the actual detection target by comparing the morphological differences between the measured spectrum and the theoretical spectrum on the full energy axis. Step 304: Execute the judgment logic. Determine whether the value of the current objective function R is less than a preset convergence threshold ε. This convergence threshold is determined based on statistical laws and physical constraints related to hardware noise characteristics. The formula for calculating the convergence threshold ε is as follows: ; Where N is the total number of analysis channels of the instrument, typically 2048 or 1024 channels; I real,i α is the photon count at channel i; α is the instrument's background count; α is a dynamically adjusted weighting factor, with a value of 1.05~1.15 for in-situ analysis and 0.8~1.00 for detection in drier areas.

[0030] If the determination result is yes, it is determined that the current model parameters match the physicochemical state of the actual detection site, and proceeds to step 306. If the determination result is no, it is determined that the model has a deviation, and proceeds to step 305. Step 305: Perform parameter feedback iteration, automatically fine-tune the initial value of rare earth element concentration, moisture parameter and matrix correction variable according to residual characteristics, and feed the updated parameter set back to the second stage step 20, and re-execute Monte Carlo physical simulation. Step 306: The system finally outputs a high-precision value for the rare earth element content.

[0031] like Figure 7 The diagram illustrates in detail the local processing flow of the multi-parameter collaborative feedback correction logic in step 305. This flow achieves hierarchical correction and updating of fundamental physical parameters and chemical concentration parameters by analyzing the residual characteristics between the measured and simulated energy spectra. The specific steps are as follows: Step 3051: Perform parametric energy domain decomposition. Align and divide the total residual between the measured energy spectrum and the virtual theoretical energy spectrum along the energy axis, focusing on distinguishing the differences in scattering background morphology in the low-energy region and the differences in the intensity of elemental characteristic peaks in the high-energy region, thereby achieving a preliminary location of the energy spectrum differences; Step 3052: Perform parameter sensitivity matrix analysis. By calculating the partial derivatives of the parameters to be corrected (such as moisture content, elements, mineral content, element concentration, etc.) with respect to the residuals in different energy domains, a sensitivity matrix is ​​constructed to determine the source of the current energy spectrum deviation. This primarily determines whether it is caused by changes in the physical state of the matrix or changes in the content of the target elements. Step 3053: After error determination, based on the sensitivity analysis results, the process enters a classification feedback path, mainly of two types: physical matrix parameter correction path and chemical concentration parameter correction path. If it is determined that the scattering packet morphology is mismatched, proceed to step 3054, prioritizing the water content W output in step 10. water Equivalent atomic number Z eff and matrix density ρ bulk Corrections are made to compensate for absorption effect deviations caused by fluctuations in moisture or mineral types; if it is determined that the characteristic peak intensity is mismatched, proceed to step 3054. Step 3054: Correct the rare earth element content; Step 3055: Update the parameter set matrix. This step uses a nonlinear optimization algorithm to integrate the corrected physical parameters and chemical concentration parameters into a new parameter vector θ. new Calculate the descent vector for the next iteration to ensure that the parameter updates conform to the physical logic; Step 3056: Trigger a resimulation. Feed the updated parameter set back to the Monte Carlo simulation in step 20, resynthesize the virtual theoretical energy spectrum, and enter the next cycle.

[0032] Example: Rapid on-site determination of yttrium (Y) in ion-adsorption rare earth ores The application of this method will be illustrated below with examples. This example selects a typical ion-adsorption rare earth mining area as the research object. The weathering crust leaching type rare earth ore in this mining area exhibits significant heterogeneity. Twenty-four sets of original mineral samples were collected from the mining area profile. Based on moisture content testing, the moisture content of the samples ranged from 5.2% to 34.8%. The actual content of the rare earth element yttrium in each sample was determined using inductively coupled plasma mass spectrometry (ICP-MS) in the laboratory, serving as the standard value for evaluating the accuracy of the determination. The moisture content of the samples and the actual content of yttrium are shown in Table 1, which represents the standard test value of yttrium measured by laboratory ICP-MS.

[0033] Table 1 ; The detailed testing process is as follows, corresponding to Figures 1 to 7 The physical path. In-situ data acquisition was performed on 24 samples using a device integrating a near-infrared detector and an X-ray fluorescence detector, executing the following setpoint logic: Phase 1: Identification of the Physical State of the Substrate This stage corresponds to Figure 2 , Figure 3 Obtain the near-infrared spectrum of the sample, and then perform linear deconvolution ( Figure 3 The lattice-constrained structural water peak and environmentally influenced free water features were successfully separated. The real-time water content W of the sample was obtained using a pre-trained CNN model. water Based on the clay mineral ratio, the equivalent atomic number Z of the detection site was reconstructed. eff With matrix density ρ bulk .

[0034] Phase Two: Full-Process Physical Simulation This stage corresponds to Figure 4 and Figure 5 The physical parameters extracted in the first stage are input into the Monte Carlo simulation program, such as... Figure 5 As shown, the entire process of X-ray photon transmission, collision, and escape under the current water content and mineral framework was simulated. Through this physical simulation, the absorption interference of water on the characteristic peak (14.96 keV) of the rare earth element yttrium was predicted, and a virtual theoretical energy spectrum was generated that can match the current environment well.

[0035] Phase 3: Dual-path feedback iteration This stage corresponds to Figure 6 and Figure 7 The measured energy spectrum is compared with the virtual theoretical energy spectrum through multi-parameter collaborative feedback logic. Figure 7Under the dual constraints of scattering background and characteristic peaks, parameter fine-tuning was performed, and the net intensity of yttrium after interference elimination was finally calculated within the physical model, thus retrieving the elemental content. The customized results from the above 24 sets of samples were summarized and regression analysis was performed with the actual ICP-MS values. The results are as follows: Figure 8 As shown.

[0036] like Figure 8 As shown in Figure (A), when using the traditional single XRF regression method, the prediction results deviate significantly because it cannot identify and compensate for changes in moisture and mineral content in the matrix. The measured values ​​are mostly lower than the true values. (Coefficient of Determination R0) 2 =0.718, RMSE = 104.5 mg / kg. Considering that traditional methods are subject to strong moisture interference, rapid determination cannot be achieved in an in-situ environment.

[0037] like Figure 8 As shown in Figure (B), the performance of the setpoint is significantly improved after adopting the correction method of this invention. Changes in sample moisture content and mineral content have minimal impact on the setpoint, and all 24 data points converge uniformly near the ideal reference line y=x. The coefficient of determination R0 2 =0.996, RMSE = 4.2 mg / kg.

[0038] This invention also provides a rapid rare earth mineral composition determination system to eliminate interference from moisture and minerals, comprising: The near-infrared spectroscopy acquisition module is used to acquire the near-infrared diffuse reflectance spectrum of the sample to be tested; The physical parameter reconstruction module is used to extract microscopic phase characteristics that characterize the mineral skeleton and moisture state, and to reconstruct the matrix physical parameters of the detection site in real time. The physics simulation module is used to perform Monte Carlo physics simulations of photon transmission paths based on the physical parameters of the substrate, and dynamically synthesize virtual theoretical energy spectra. The X-ray fluorescence acquisition module is used to obtain the measured X-ray fluorescence energy spectrum of the same detection site; The collaborative iterative correction module is used to construct the full-spectrum characteristic residual objective function of the measured energy spectrum and the virtual theoretical energy spectrum, and to perform collaborative feedback iteration under bidirectional constraints. When the convergence condition is met, it outputs the fixed value result of the rare earth element to be measured.

[0039] The results of this embodiment strongly support the inventiveness of the present invention, demonstrating how physical laws can be used to simulate and eliminate the physical interference of moisture on X-ray transmission from the bottom layer. This invention eliminates the need to prepare numerous physical standard samples for different moisture contents; by combining NIR identification with physical simulation, it can achieve adaptive adaptation to unknown geological environments, thus significantly improving the exploration of rare earth elements in mining areas.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

Claims

1. A rapid method for determining the composition of rare earth minerals, eliminating interference from moisture and minerals, characterized in that, Includes the following steps: The near-infrared diffuse reflectance spectrum of the sample to be tested is obtained, the microscopic phase characteristics characterizing the mineral framework and moisture state are extracted, and the matrix physical parameters of the detection site are reconstructed in real time. Based on the physical parameters of the substrate, a Monte Carlo simulation of the photon transmission path is performed to dynamically synthesize a virtual theoretical energy spectrum that matches the physical state of the current detection environment. Obtain the measured X-ray fluorescence energy spectrum of the same detection site, and construct the full-spectrum characteristic residual objective function between the measured energy spectrum and the virtual theoretical energy spectrum; Under the dual constraints of the physical framework formed by the physical parameters of the matrix and the energy spectrum background, a cooperative feedback iteration is performed. When the objective function of the characteristic residual satisfies the convergence condition, the fixed value result of the rare earth element to be measured is output.

2. The method according to claim 1, characterized in that, The process of acquiring the near-infrared diffuse reflectance spectrum of the sample to be tested, extracting microscopic phase characteristics characterizing the mineral framework and moisture state, and reconstructing the matrix physical parameters of the detection site in real time specifically includes: Convert the original diffuse reflectance spectrum into an absorption index spectrum; The composite absorption band at the characteristic absorption band in the absorption index energy spectrum is deconvoluted and integrally decomposed to separate the spectral contributions of different physical phases. Physical feature parameters are extracted from the decomposed sub-peaks to construct high-dimensional physical feature vectors; The high-dimensional physical feature vector is input into a one-dimensional convolutional neural network, and the mass fraction of the target component is output. The comprehensive physical parameters of the current detection point are reconstructed based on the mass fraction.

3. The method according to claim 2, characterized in that, The conversion of the original diffuse reflectance spectral data into absorption index data employs the following mathematical transformation: F(R)=(1- R ) 2 / 2 R； Where R is the original diffuse reflectance and F(R) is the absorption index.

4. The method according to claim 2, characterized in that, The characteristic absorption bands include composite absorption bands with numerical fluctuations at 1400 nm and 1900 nm; the different physical phase components include kaolinite-structured hydroxyl groups, halloysite-structured hydroxyl groups, adsorbed water, and free water; the mass fraction of the target component includes the mass fraction of kaolinite, the mass fraction of halloysite, and the real-time water content.

5. The method according to claim 2, characterized in that, The deconvolution integral is based on nonlinear fitting using the Voigt function, which is composed of the convolution of a Gaussian term representing the broadening of inhomogeneous regions and a Lorentz term representing the broadening of natural energy levels. The physical feature parameters include the center position, peak intensity, Gaussian half-width, and Lorentz half-width of each sub-peak.

6. The method according to claim 1, characterized in that, The process of performing Monte Carlo physics simulations of photon transmission paths based on the substrate physical parameters, and dynamically synthesizing a virtual theoretical energy spectrum that matches the current physical state of the detection environment, specifically includes: Load the emission spectrum information of the excitation source and the geometric constraint variables of the instrument to establish a physical simulation scenario consistent with the actual hardware detection environment; Based on the physical parameters of the matrix, the mass attenuation coefficient is determined, photon step size sampling is performed, and single-photon transmission path tracing is performed within the set effective interaction volume of the sample. The interaction type of photon at the detection site is determined based on random numbers. The interaction type includes photoelectric effect, Rayleigh scattering and Compton scattering. The energy and vector direction of photons are updated in real time based on the interaction type, and the energy spectrum of photons escaping the sample boundary is reconstructed using the detector response function to obtain the virtual theoretical energy spectrum.

7. The method according to claim 1, characterized in that, The objective function for the full-spectrum characteristic residuals includes: using the least squares method to compare the residual distribution of the measured energy spectrum and the virtual theoretical energy spectrum in terms of the morphology of the scattering background bag in the low-energy region and the intensity of the characteristic peaks in the high-energy region.

8. The method according to claim 1, characterized in that, The collaborative feedback iteration specifically includes: The total residual between the measured energy spectrum and the virtual theoretical energy spectrum is decomposed into energy domains, and the dominant error sources of the residual are determined by using the sensitivity matrix. If the scattering packet morphology is determined to be mismatched, it is determined to be a deviation in matrix physical parameters. The execution path prioritizes correcting the water content and component ratio parameters. The correction targets for the deviation in matrix physical parameters include water content, equivalent atomic number, and matrix density. If the characteristic peak intensity is determined to be mismatched, it is determined to be an element concentration deviation, and the path is executed to correct the mass fraction parameter of the rare earth element to be measured. The parameter vector is updated using a nonlinear optimization algorithm and fed back to the physical simulation step in a loop. The process is repeated until the measured energy spectrum and the virtual theoretical energy spectrum achieve full spectrum matching.

9. The method according to claim 1, characterized in that, The convergence condition is that the value of the objective function of the characteristic residual is less than a preset convergence threshold; when the convergence condition is met, the fixed value result of the rare earth element to be tested is output, and the fixed value result includes the mass fraction of the rare earth element to be tested.

10. A rapid rare earth mineral composition determination system that eliminates interference from moisture and minerals, characterized in that, include: The near-infrared spectroscopy acquisition module is used to acquire the near-infrared diffuse reflectance spectrum of the sample to be tested; The physical parameter reconstruction module is used to extract microscopic phase characteristics that characterize the mineral skeleton and moisture state, and to reconstruct the matrix physical parameters of the detection site in real time. The physics simulation module is used to perform Monte Carlo physics simulations of photon transmission paths based on the physical parameters of the substrate, and dynamically synthesize virtual theoretical energy spectra. The X-ray fluorescence acquisition module is used to acquire the measured X-ray fluorescence energy spectrum of the same detection site. The collaborative iterative correction module is used to construct the full-spectrum characteristic residual objective function of the measured energy spectrum and the virtual theoretical energy spectrum, and to perform collaborative feedback iteration under bidirectional constraints. When the convergence condition is met, it outputs the fixed value result of the rare earth element to be measured.

Images