A rock sample test analysis method and system based on spectral analysis

Abstract

The present application relates to the technical field of test analysis, in particular to a rock sample test analysis method and system based on spectral analysis.A rock sample test analysis system based on spectral analysis comprises a rock sample spectral test module, an element analysis module, a lithofacies end-member inversion decomposition module and a rock test analysis result construction module.The present application cuts and prepares a rock sample into an approximate plane, constructs a regular grid on an XY displacement table, collects and pre-processes EDXRF energy spectrum data by grid, converts the element analysis network into a unified dimension element / oxide content matrix, implements lithofacies end-member inversion decomposition under physical constraints and spatial neighborhood constraints to obtain lithofacies end-member standard components and the end-member mixing ratio of each grid, and further reconstructs a lithofacies end-member spatial distribution map and an overall element distribution vector based on end-member volume fraction weighting.

Description

A method and system for testing and analyzing rock samples based on spectral analysis Technical Field

[0001] This invention relates to the field of testing and analysis technology, and specifically to a method and system for testing and analyzing rock samples based on spectral analysis. Background Technology

[0002] The elemental composition and spatial distribution characteristics of rock samples are crucial foundational data for sedimentary environment identification, lithofacies classification, reservoir heterogeneity assessment, and mineralization enrichment analysis. Traditional whole-sample rock chemical analysis methods (such as whole-sample XRF, ICP-AES / ICP-MS) typically require sample pulverization, yielding only a single, average elemental composition result representing the entire sample. This process damages the sample and completely obscures the spatial structure information of elements within it. In actual geological bodies, rocks often consist of multiple lithofacies or mineral components, exhibiting high heterogeneity in composition and structure. Relying solely on the average chemical composition of the whole sample is insufficient to identify fine lithofacies units, lithofacies boundaries, and localized anomalous enrichment zones, thus failing to meet the demands of detailed lithofacies analysis and high-resolution reservoir evaluation. Summary of the Invention

[0003] This invention involves cutting rock samples and constructing an H×W grid on an XY displacement stage at a preset step size. Each grid is then subjected to EDXRF acquisition followed by baseline correction, smoothing, and normalization preprocessing before being fed into an elemental analysis network to uniformly output fixed-dimensional elemental analysis vectors. This approach enables the acquisition of high spatial resolution, low noise, and physically meaningful elemental analysis matrices without damaging the entire sample. Furthermore, the elemental analysis matrix is ​​considered as the product of a lithofacies endmember matrix and a lithofacies mixing ratio matrix. Under physical and spatial neighborhood constraints, both are iteratively optimized to automatically extract standard lithofacies endmember components that conform to petrological and mineralogy principles. The invention also obtains the mixing ratios of different lithofacies endmembers within each grid, thus enabling the analysis of the heterogeneity of the internal composition and the spatial structure of the lithofacies in the rock sample. The process involves finely characterizing the lithofacies mixing ratio matrix and reconstructing it into a distribution map of lithofacies units for each endmember based on the original scan rows and columns. This provides a visual representation of the spatial distribution characteristics of different lithofacies endmembers in the sample and can serve as high-resolution foundational data for subsequent lithofacies delineation, structural interpretation, and reservoir evaluation. Simultaneously, the average proportion of each lithofacies endmember in the sample is calculated using the lithofacies mixing ratio matrix, and then the lithofacies endmember matrix is ​​weighted using this proportion to obtain the overall element distribution vector. This is equivalent to performing a volume-weighted summary of the elements in the rock sample under the constraints of endmember decomposition and volume fraction. This approach preserves the information on lithofacies and compositional heterogeneity while effectively suppressing the interference of local anomalies on the overall results. Furthermore, it can provide accurate, robust, and spatially interpretable rock test analysis results in the context of multi-lithofacies mixing.

[0004] This invention provides a method for testing and analyzing rock samples based on spectral analysis, comprising:

[0005] Spectroscopic tests were performed on the rock samples to obtain rock spectral cube data. The scale of the rock spectral cube data is H×W×C, where H×W is the number of grids corresponding to the rock sample, and C is the dimension of the energy spectrum data collected for each grid.

[0006] Preprocessing is performed on the energy spectrum data corresponding to all grids, and then the preprocessed energy spectrum data is fed into the elemental analysis network for processing to output elemental analysis vectors. All elemental analysis vectors are then concatenated from top to bottom to form an elemental analysis matrix, following the grid traversal order from left to right and from top to bottom during the concatenation process.

[0007] The element analysis matrix is ​​subjected to lithofacies endmember inversion decomposition to obtain lithofacies endmember matrix and lithofacies mixing ratio matrix. The lithofacies endmember inversion decomposition operation regards the element analysis matrix as the product of lithofacies mixing ratio matrix and lithofacies endmember matrix, and iteratively optimizes the lithofacies endmember matrix and lithofacies mixing ratio matrix under physical constraints and spatial neighborhood constraints.

[0008] Each column vector of the lithofacies mixing ratio matrix is ​​reconstructed into a lithofacies unit distribution map corresponding to the lithofacies endmember. The lithofacies unit distribution map represents the distribution of different lithofacies endmembers on the rock sample. Based on the lithofacies mixing ratio matrix, the average proportion of each lithofacies endmember on the rock sample is calculated. Then, a weighted calculation is performed based on the lithofacies endmember matrix and the average proportion of each lithofacies endmember on the rock sample to obtain the overall element distribution vector corresponding to the rock sample.

[0009] Preferably, a lithofacies endmember inversion decomposition operation is performed on the elemental analysis matrix to obtain the lithofacies endmember matrix and the lithofacies mixing ratio matrix, specifically including the following steps:

[0010] Step S1: Obtain candidate lithofacies quantity data, which includes the quantity of several candidate lithofacies;

[0011] Step S2: Traverse the candidate lithofacies quantity dataset and perform the solution operations for the lithofacies endmember matrix and lithofacies mixing ratio matrix for each candidate lithofacies quantity;

[0012] Step S2.1: Perform cluster analysis on all row vectors of the element analysis matrix, divide all row vectors of the element analysis matrix into clusters corresponding to the number of candidate lithofacies, and use the cluster center of each cluster as the row vector of the lithofacies end-member matrix;

[0013] Step S2.2: Under the condition of fixed lithofacies end-member matrix, perform linear regression operation on lithofacies mixing ratio matrix based on all element analysis vectors, and project the solution into feasible region where each component is non-negative and the sum of each component is 1. Then perform spatial optimization based on Markov random field to obtain updated lithofacies mixing ratio matrix.

[0014] Step S2.3: Under the condition of a fixed lithofacies mixing ratio matrix, perform linear regression operation on the lithofacies end-member matrix based on the column vector of the element analysis matrix, and then perform element content constraints on several lithofacies units corresponding to the lithofacies end-member matrix to obtain the updated lithofacies end-member matrix.

[0015] Step S2.4: Repeat steps S2.2-S2.3 until convergence, and output the lithofacies end-member matrix and lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies.

[0016] Step S3: The product of the lithofacies mixing ratio matrix and the lithofacies end-member matrix corresponding to the number of candidate lithofacies is recorded as the corresponding theoretical element analysis matrix. Then, the mean square error between all row vectors in the theoretical element analysis matrix and the corresponding row vectors in the element analysis matrix is ​​calculated and recorded as the fitting error. Then, the spatial structure penalty value is calculated based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies. The fitting error and the spatial structure penalty value corresponding to the number of candidate lithofacies are weighted and summed to obtain the comprehensive evaluation value. The lithofacies end-member matrix and lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies with the smallest comprehensive evaluation value are selected for output.

[0017] Preferably, under the condition of a fixed lithofacies end-member matrix, a linear regression operation is performed on the lithofacies mixing ratio matrix based on the analysis vectors of all elements, and the solution is projected onto a feasible region where each component is non-negative and the sum of each component is 1. Then, spatial optimization based on Markov random fields is performed to obtain the updated lithofacies mixing ratio matrix. The specific steps include the following:

[0018] Iterate through the row vectors of all element analysis matrices. For each row vector of an element analysis matrix, perform the following operation: using the row vector of the element analysis matrix as the dependent variable and the row vector of the lithofacies mixing ratio matrix that is in the same row position as the row vector of the element analysis matrix as the independent variable, perform a linear regression operation on the row vector of the lithofacies mixing ratio matrix until all row vectors of all element analysis matrices have been traversed. Output the row vectors of all lithofacies mixing ratio matrices and denote them as the preliminary fitted lithofacies mixing ratio vector.

[0019] Iterate through all the preliminary fitted lithofacies mixing ratio vectors. For each preliminary fitted lithofacies mixing ratio vector, cut off the negative values ​​in the preliminary fitted lithofacies mixing ratio vector to 0. Then, perform a normalization operation on the cut preliminary fitted lithofacies mixing ratio vector to obtain a non-negative lithofacies mixing ratio vector.

[0020] For the non-negative lithofacies mixing ratio vector, the following operations are performed: Select the element analysis vector in the element analysis matrix that has the same row position as the non-negative lithofacies mixing ratio vector and denote it as the target element analysis vector. Calculate the similarity between the row vectors of the lithofacies end-member matrix and the target element analysis vector, and denote it as the fitted matching value. Select the index of the row vector of the lithofacies end-member matrix corresponding to the largest fitted matching value and denote it as h. Construct a convex update vector, where the h-th term in the convex update vector is 1 and the rest are set to 0. Add the output of the product of the non-negative lithofacies mixing ratio vector and the update coefficient to the convex update vector to obtain the updated non-negative lithofacies mixing ratio vector. Combine all the updated non-negative lithofacies mixing ratio vectors to form the updated lithofacies mixing ratio matrix.

[0021] Preferably, under the condition of a fixed lithofacies mixing ratio matrix, a linear regression operation is performed on the lithofacies endmember matrix based on the column vectors of the elemental analysis matrix, and then elemental content constraints are applied to several lithofacies units corresponding to the lithofacies endmember matrix to obtain an updated lithofacies endmember matrix. Specifically, the steps include:

[0022] Iterate through the column vectors of all element analysis matrices. For each column vector of an element analysis matrix, perform the following operation: using the column vector of the element analysis matrix as the dependent variable and the column vector of the lithofacies end-member matrix that is in the same column position as the column vector of the element analysis matrix as the independent variable, perform linear regression on the column vector of the lithofacies end-member matrix until all column vectors of all element analysis matrices have been traversed. Output the column vectors of all lithofacies end-member matrices and denote them as the preliminary fitted lithofacies end-member vectors.

[0023] All initially fitted lithofacies end-member vectors are concatenated into an initially fitted lithofacies end-member matrix. The row vectors of the initially fitted lithofacies end-member matrix are traversed, and the similarity between the row vectors of the initially fitted lithofacies end-member matrix and the lithofacies end-member standard vectors in the lithofacies end-member standard library is calculated sequentially. This similarity is denoted as the lithofacies end-member matching value. The lithofacies end-member standard library includes several lithofacies end-member standard vectors and their corresponding element content constraint ranges. The element content constraint range corresponding to the largest lithofacies end-member matching value is selected. Data in the row vectors of the initially fitted lithofacies end-member matrix that exceed the corresponding element content constraint range are pruned to obtain the updated lithofacies end-member matrix.

[0024] Preferably, the spatial structure penalty value is calculated based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies, specifically including the following steps: constructing a lithofacies label matrix, in which the positions of the lithofacies label matrix correspond one-to-one with the grid of the rock sample, and the data of each position in the lithofacies label matrix is ​​the lithofacies endmember label corresponding to the largest data in the row vector of the lithofacies mixing ratio matrix corresponding to the grid, and obtaining all grid neighborhood pairs, calculating the ratio between the number of grid neighborhood pairs with inconsistent lithofacies endmember labels corresponding to the grid and the number of all grid neighborhood pairs, which is recorded as the spatial structure penalty value.

[0025] Preferably, each column vector of the lithofacies mixing ratio matrix is ​​reconstructed into a lithofacies unit distribution map corresponding to the lithofacies endmember. Specifically, this includes the following steps: constructing a rock sample matrix diagram, where the position in the rock sample matrix diagram corresponds one-to-one with the grid of the rock sample; then filling the data in the column vectors of the lithofacies mixing ratio matrix into the rock sample matrix diagram according to the grid corresponding to the row number, thereby constructing a lithofacies unit distribution map corresponding to the lithofacies endmember of the column vector of the lithofacies mixing ratio matrix.

[0026] Preferably, training the elemental analysis network includes the following steps: obtaining several elemental analysis training samples, which include energy spectrum data; labeling the elemental analysis training samples using elemental analysis vectors; forming an elemental analysis training set from all labeled elemental analysis training samples; and training the elemental analysis network using the elemental analysis training set.

[0027] This invention provides a rock sample testing and analysis system based on spectral analysis, comprising:

[0028] The rock sample spectral testing module is used to perform spectral testing on rock samples and obtain rock spectral cube data. The scale of the rock spectral cube data is H×W×C, where H×W is the number of grids corresponding to the rock sample, and C is the dimension of the energy spectrum data collected for each grid.

[0029] The elemental analysis module is used to perform preprocessing operations on the energy spectrum data corresponding to all grids, and then send the preprocessed energy spectrum data into the elemental analysis network for processing, outputting elemental analysis vectors; and concatenating all elemental analysis vectors from top to bottom into an elemental analysis matrix, following the grid traversal order from left to right and from top to bottom during the concatenation process;

[0030] The lithofacies endmember inversion decomposition module is used to perform lithofacies endmember inversion decomposition operation on the element analysis matrix to obtain the lithofacies endmember matrix and the lithofacies mixing ratio matrix. The lithofacies endmember inversion decomposition operation treats the element analysis matrix as the product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix, and iteratively optimizes the lithofacies endmember matrix and the lithofacies mixing ratio matrix under physical constraints and spatial neighborhood constraints.

[0031] The rock test analysis result construction module is used to reconstruct each column vector of the lithofacies mixing ratio matrix into a lithofacies unit distribution map corresponding to the lithofacies end-member. The lithofacies unit distribution map represents the distribution of different lithofacies end-members on the rock sample. Based on the lithofacies mixing ratio matrix, the average proportion of each lithofacies end-member on the rock sample is calculated. Then, based on the lithofacies end-member matrix and the average proportion of each lithofacies end-member on the rock sample, a weighted calculation is performed to obtain the overall element distribution vector corresponding to the rock sample.

[0032] The present invention has the following advantages:

[0033] This invention involves cutting rock samples and constructing an H×W grid on an XY displacement stage at a preset step size. Each grid is then subjected to EDXRF acquisition followed by baseline correction, smoothing, and normalization preprocessing before being fed into an elemental analysis network to uniformly output fixed-dimensional elemental analysis vectors. This approach enables the acquisition of high spatial resolution, low noise, and physically meaningful elemental analysis matrices without damaging the entire sample. Furthermore, the elemental analysis matrix is ​​considered as the product of a lithofacies endmember matrix and a lithofacies mixing ratio matrix. Under physical and spatial neighborhood constraints, both are iteratively optimized to automatically extract standard lithofacies endmember components that conform to petrological and mineralogy principles. The invention also obtains the mixing ratios of different lithofacies endmembers within each grid, thus enabling the analysis of the heterogeneity of the internal composition and the spatial structure of the lithofacies in the rock sample. The process involves finely characterizing the lithofacies mixing ratio matrix and reconstructing it into a distribution map of lithofacies units for each endmember based on the original scan rows and columns. This provides a visual representation of the spatial distribution characteristics of different lithofacies endmembers in the sample and can serve as high-resolution foundational data for subsequent lithofacies delineation, structural interpretation, and reservoir evaluation. Simultaneously, the average proportion of each lithofacies endmember in the sample is calculated using the lithofacies mixing ratio matrix, and then the lithofacies endmember matrix is ​​weighted using this proportion to obtain the overall element distribution vector. This is equivalent to performing a volume-weighted summary of the elements in the rock sample under the constraints of endmember decomposition and volume fraction. This approach preserves the information on lithofacies and compositional heterogeneity while effectively suppressing the interference of local anomalies on the overall results. Furthermore, it can provide accurate, robust, and spatially interpretable rock test analysis results in the context of multi-lithofacies mixing. Attached Figure Description

[0034] Figure 1 is a schematic diagram of the rock sample testing and analysis system based on spectral analysis used in an embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0036] Example 1: A method for testing and analyzing rock samples based on spectral analysis, comprising:

[0037] Spectroscopic testing was performed on the rock sample to acquire rock spectral cube data. The scale of the rock spectral cube data is H×W×C, where H×W is the number of grids corresponding to the rock sample, that is, dividing the plane corresponding to the rock sample into a network. The length of the grid plane composed of all grids is H, the width is W, and C is the dimension corresponding to the energy spectrum data collected for each grid. It should be noted that before performing spectral testing on the rock sample, it is necessary to first cut an approximate plane from the rock using a rock cutter, ensuring that the flatness of this approximate plane meets the expected standard. This cut rock is recorded as the rock sample, which is used for subsequent spectral testing. The rock sample is then fixed to the XRF instrument using a clamp. On the XY scanning displacement stage, set the scanning step distance in the X / Y directions, for example, 0.5-2 mm. Set the excitation conditions of the XRF instrument, including tube pressure, tube current, filter type, and test time. Control the rock sample to move along the preset path on the XY scanning displacement stage, and perform EDXRF acquisition once on each grid on the approximate plane through the XRF instrument to obtain the energy spectrum data on that grid. The energy spectrum data includes the energy count corresponding to each channel number. When the XRF instrument performs spectral analysis, it divides the energy range into several segments, with the energy range generally being 0-40 KeV. Each segment is regarded as a channel number, and the energy count is the X-ray photon count corresponding to the channel number.

[0038] Preprocessing operations are performed on the energy spectrum data corresponding to all grids. These operations include baseline correction, smoothing, and normalization. Baseline correction involves estimating the continuous background using polynomial and SNIP algorithms and removing it from the energy spectrum data. Smoothing is achieved through filtering functions. Normalization unifies the counts to between 0 and 1. These preprocessing operations reduce noise in the energy spectrum data. The preprocessed energy spectrum data is then fed into an elemental analysis network for further processing, outputting elemental analysis vectors. These vectors contain the content of several elemental substances or oxides, described as mass percentages, such as copper, iron, sulfur, and silicon dioxide. It is important to note that all elemental analysis vectors have the same dimension, and the elemental substance or oxide corresponding to each item in the vector is set by the operator. All elemental analysis vectors are then concatenated from top to bottom to form an elemental analysis matrix, following a left-to-right, top-to-bottom grid traversal order. The elemental analysis network employs an MLP (Multi-Level Processing) model.

[0039] Performing lithofacies endmember inversion decomposition on the elemental analysis matrix yields a lithofacies endmember matrix and a lithofacies mixing ratio matrix. Each row in the lithofacies endmember matrix represents a standard elemental analysis vector corresponding to a lithofacies endmember. The standard elemental analysis vector has the same dimension as the elemental analysis vector. Lithofacies endmembers represent different regions in the rock sample with significant differences in composition and structure, such as quartz and feldspar grain regions, carbonate cement, and clay matrix in carbonate rocks. Each row in the lithofacies mixing ratio matrix represents the proportion of different lithofacies endmembers in the corresponding grid. The number of rows in the lithofacies mixing ratio matrix is ​​the same as the number of rows in the elemental analysis matrix. The number of columns in the matrix is ​​the same as the number of rows in the lithofacies endmember matrix. The lithofacies endmember inversion decomposition operation treats the elemental analysis matrix as the product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix, and iteratively optimizes the lithofacies endmember matrix and the lithofacies mixing ratio matrix under physical constraints and spatial neighborhood constraints. It should be noted that in rock samples, the elements are not homogeneously distributed, but rather a mixed distribution of multiple lithofacies endmembers. By constructing the rock sample into a grid-like energy spectrum data distribution and implementing the lithofacies endmember inversion decomposition operation, a more accurate analysis of the elemental composition within the rock sample can be performed based on the lithofacies mixture.

[0040] Each column vector of the lithofacies mixing ratio matrix is ​​reconstructed into a lithofacies unit distribution map corresponding to the lithofacies endmember. The lithofacies unit distribution map represents the distribution of different lithofacies endmembers on the rock sample. Based on the lithofacies mixing ratio matrix, the average proportion of each lithofacies endmember in the rock sample is calculated. Then, a weighted calculation is performed based on the lithofacies endmember matrix and the average proportion of each lithofacies endmember in the rock sample to obtain the overall element distribution vector corresponding to the rock sample. The overall element distribution vector obtained by the weighted calculation of lithofacies endmembers can provide more accurate rock test and analysis results when considering the heterogeneity of the distribution of various elements in the rock sample.

[0041] This application constructs an H×W grid on an XY displacement stage by cutting the rock sample and performing EDXRF acquisition on each grid, followed by baseline correction, smoothing, and normalization preprocessing. The data is then fed into an elemental analysis network to uniformly output a fixed-dimensional elemental analysis vector. This approach achieves high spatial resolution, low noise, and physically meaningful elemental analysis matrices without damaging the entire sample. Furthermore, the elemental analysis matrix is ​​considered as the product of a lithofacies endmember matrix and a lithofacies mixing ratio matrix. Under physical and spatial neighborhood constraints, both are iteratively optimized to automatically extract standard lithofacies endmember components that conform to petrological and mineralogy principles. The mixing ratios of different lithofacies endmembers in each grid can also be obtained, enabling the analysis of the internal compositional heterogeneity and spatial structure of the rock sample. The process involves finely characterizing the lithofacies mixing ratio matrix and reconstructing it into a distribution map of lithofacies units for each endmember based on the original scan rows and columns. This provides a visual representation of the spatial distribution characteristics of different lithofacies endmembers in the sample and can serve as high-resolution foundational data for subsequent lithofacies delineation, structural interpretation, and reservoir evaluation. Simultaneously, the average proportion of each lithofacies endmember in the sample is calculated using the lithofacies mixing ratio matrix, and then the lithofacies endmember matrix is ​​weighted using this proportion to obtain the overall element distribution vector. This is equivalent to performing a volume-weighted summary of the elements in the rock sample under the constraints of endmember decomposition and volume fraction. This approach preserves the information on lithofacies and compositional heterogeneity while effectively suppressing the interference of local anomalies on the overall results. Furthermore, it can provide accurate, robust, and spatially interpretable rock test analysis results in the context of multi-lithofacies mixing.

[0042] Perform lithofacies endmember inversion decomposition on the elemental analysis matrix to obtain the lithofacies endmember matrix and lithofacies mixing ratio matrix. The specific steps include the following:

[0043] Step S1: Obtain candidate lithofacies quantity data. The candidate lithofacies quantity data includes several candidate lithofacies quantities. The candidate lithofacies quantity data is set in advance by the operator, usually set to 1-10, which is used to a priori set the number of lithofacies endmembers.

[0044] Step S2: Traverse the candidate lithofacies quantity dataset and perform the solution operations for the lithofacies endmember matrix and lithofacies mixing ratio matrix for each candidate lithofacies quantity;

[0045] Step S2.1: Perform cluster analysis on all row vectors of the element analysis matrix, dividing all row vectors of the element analysis matrix into clusters corresponding to the number of candidate lithofacies, and using the cluster center of each cluster as the row vector of the lithofacies end-member matrix. The cluster analysis here uses the K-means algorithm, and the cluster center is selected as the average value of all data in the cluster.

[0046] Step S2.2: Under the condition of fixed lithofacies end-member matrix, perform linear regression operation on lithofacies mixing ratio matrix based on all element analysis vectors, and project the solution into feasible region where each component is non-negative and the sum of each component is 1. Then perform spatial optimization based on Markov random field to obtain updated lithofacies mixing ratio matrix.

[0047] Step S2.3: Under the condition of a fixed lithofacies mixing ratio matrix, perform linear regression operation on the lithofacies end-member matrix based on the column vector of the element analysis matrix, and then perform element content constraints on several lithofacies units corresponding to the lithofacies end-member matrix to obtain the updated lithofacies end-member matrix.

[0048] Step S2.4: Repeat steps S2.2-S2.3 until convergence, and output the lithofacies end-member matrix and lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies; the convergence condition is that the difference between the product of the lithofacies mixing ratio matrix and the lithofacies end-member matrix corresponding to the number of candidate lithofacies and the element analysis matrix meets the expectation.

[0049] Step S3: The product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix corresponding to the number of candidate lithofacies is recorded as the corresponding theoretical element analysis matrix. Then, the mean square error between all row vectors in the theoretical element analysis matrix and the corresponding row vectors in the element analysis matrix is ​​calculated and recorded as the fitting error. Then, the spatial structure penalty value is calculated based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies. The spatial structure penalty value represents the connectivity of the lithofacies ports of each grid on the approximate plane mapped by the lithofacies mixing ratio matrix. The smaller the spatial structure penalty value, the greater the connectivity of the approximate plane mapped by the lithofacies mixing ratio matrix, which is more consistent with the actual lithofacies endmember distribution. The fitting error and the spatial structure penalty value corresponding to the number of candidate lithofacies are weighted and summed to obtain the comprehensive evaluation value. The weights in the weighted summation operation are set by the operator. The weights corresponding to the fitting error and the spatial structure penalty value are generally set to 0.6 and 0.4, respectively. The lithofacies endmember matrix and the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies with the smallest comprehensive evaluation value are selected for output.

[0050] When calculating the lithofacies endmember matrix and lithofacies mixing ratio matrix, considering the number of multiple lithofacies endmembers makes the lithofacies endmember inversion decomposition operation of the elemental analysis matrix more adaptive to the rock sample. This avoids the problems of mixing different lithofacies due to too few lithofacies endmembers and overfitting due to too many lithofacies endmembers. This makes the final determined number of lithofacies and the corresponding lithofacies endmember matrix and lithofacies mixing ratio matrix more reasonable in a statistical sense. Furthermore, when selecting the number of lithofacies, the fitting error and spatial structure penalty value are considered. While ensuring the fitting accuracy, spatial geological constraints are also introduced, further improving the geological interpretability of lithofacies classification.

[0051] Under the condition of a fixed lithofacies end-member matrix, a linear regression operation is performed on the lithofacies mixing ratio matrix based on the analysis vectors of all elements. The solution is then projected onto a feasible region where each component is non-negative and the sum of each component is 1. Spatial optimization based on Markov random fields is then performed to obtain the updated lithofacies mixing ratio matrix. The specific steps include the following:

[0052] Traverse the row vectors of all element analysis matrices. For each row vector of an element analysis matrix, perform the following operation: using the row vector of the element analysis matrix as the dependent variable and the row vector of the lithofacies mixing ratio matrix that is in the same row position as the row vector of the element analysis matrix as the independent variable, perform a linear regression operation on the row vector of the lithofacies mixing ratio matrix. In the linear regression operation, the product of the row vector of the lithofacies mixing ratio matrix and the fixed lithofacies end-member matrix is ​​the theoretical output value. The goal of linear fitting is to reduce the difference between the theoretical output value and the row vector of the element analysis matrix. Set the constraint that the sum of all data in the row vector of the lithofacies mixing ratio matrix is ​​1. Continue until the row vectors of all element analysis matrices have been traversed. Output the row vectors of all lithofacies mixing ratio matrices and denote them as the preliminary fitted lithofacies mixing ratio vectors.

[0053] Iterate through all the preliminary fitted lithofacies mixing ratio vectors. For each preliminary fitted lithofacies mixing ratio vector, cut off the negative values ​​in the preliminary fitted lithofacies mixing ratio vector to 0. Then, perform a normalization operation on the cut preliminary fitted lithofacies mixing ratio vector to obtain a non-negative lithofacies mixing ratio vector.

[0054] For the non-negative lithofacies mixing ratio vector, the following operations are performed: Select the element analysis vector in the element analysis matrix that has the same row position as the non-negative lithofacies mixing ratio vector and denote it as the target element analysis vector. Calculate the similarity between the row vectors of the lithofacies end-member matrix and the target element analysis vector, and denote it as the fitted matching value. The similarity calculation can use cosine similarity. Select the index of the row vector of the lithofacies end-member matrix corresponding to the largest fitted matching value and denote it as h. Construct a convex update vector, where the h-th term in the convex update vector is 1 and the rest are set to 0. Add the output of the product of the non-negative lithofacies mixing ratio vector and the update coefficient to the convex update vector to obtain the updated non-negative lithofacies mixing ratio vector. Combine all the updated non-negative lithofacies mixing ratio vectors to form the updated lithofacies mixing ratio matrix.

[0055] Under the condition of a fixed lithofacies mixing ratio matrix, a linear regression operation is performed on the lithofacies endmember matrix based on the column vectors of the elemental analysis matrix, and then elemental content constraints are applied to several lithofacies units corresponding to the lithofacies endmember matrix to obtain an updated lithofacies endmember matrix. The specific steps include the following:

[0056] Iterate through the column vectors of all element analysis matrices. For each column vector of an element analysis matrix, perform the following operation: using the column vector of the element analysis matrix as the dependent variable and the column vector of the lithofacies end-member matrix that is in the same column position as the column vector of the element analysis matrix as the independent variable, perform a linear regression operation on the column vector of the lithofacies end-member matrix. In the linear regression operation, the product of the fixed lithofacies mixing ratio matrix and the column vector of the lithofacies end-member matrix is ​​the theoretical output value. The goal of linear fitting is to reduce the difference between the theoretical output value and the column vector of the element analysis matrix. The goal is to solve for the content values ​​of different elements for each lithofacies end-member, so that the product of the lithofacies mixing ratio matrix and the lithofacies end-member matrix is ​​closer to the measured value. Continue until the column vectors of all element analysis matrices have been traversed. Output the column vectors of all lithofacies end-member matrices and denote them as the preliminary fitted lithofacies end-member vectors.

[0057] All initially fitted lithofacies endmember vectors are concatenated into an initially fitted lithofacies endmember matrix. The row vectors of this matrix are traversed, and the similarity between each row vector and the standard vectors in the lithofacies endmember standard library is calculated sequentially. This similarity is recorded as the lithofacies endmember matching value. The lithofacies endmember standard library includes several lithofacies endmember standard vectors and their corresponding elemental content constraints, which are statistically calculated by the operator based on high-purity rock samples. For example, for quartz, high-purity quartz rock samples (with internal quartz lithofacies endmembers accounting for more than 95%) are selected. Energy dispersive spectroscopy (EDS) data is acquired using an XRF spectrometer, and then elemental analysis vectors are determined through an elemental analysis network. Further calculations are then performed on several quartz rock samples. The average value of the elemental analysis vectors corresponding to the samples is used as the standard vector of the corresponding lithofacies endmembers. The corresponding elemental content constraint range is determined according to the value range of different terms in the elemental analysis vectors corresponding to several quartz rock samples. The elemental content constraint range corresponding to the largest lithofacies endmember matching value is selected. Data in the row vectors of the initially fitted lithofacies endmember matrix that exceed the corresponding elemental content constraint range are pruned. The pruning method is generally as follows: if the data is higher than the upper limit of the corresponding elemental content constraint range, the corresponding data is set to the upper limit of the corresponding elemental content constraint range; if the data is lower than the lower limit of the corresponding elemental content constraint range, the corresponding data is set to the lower limit of the corresponding elemental content constraint range. The updated lithofacies endmember matrix is ​​obtained.

[0058] The spatial structure penalty value is calculated based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies, specifically including the following steps:

[0059] A lithofacies label matrix is ​​constructed, in which the positions of the lithofacies labels correspond one-to-one with the grids of the rock samples. The data at each position in the lithofacies label matrix is ​​the lithofacies endmember label corresponding to the largest data in the row vector of the lithofacies mixing ratio matrix corresponding to the grid. All grid neighborhood pairs are obtained, which are data composed of adjacent grids. The ratio between the number of grid neighborhood pairs with inconsistent lithofacies endmember labels corresponding to the grid and the total number of all grid neighborhood pairs is calculated and denoted as the spatial structure penalty value. The smaller the number of grid neighborhood pairs with inconsistent lithofacies endmember labels corresponding to the grid, the stronger the continuity of lithofacies endmember labels on the approximate plane of the mapping.

[0060] Reconstructing each column vector of the lithofacies mixing ratio matrix into a lithofacies unit distribution map corresponding to the lithofacies endmember involves the following steps:

[0061] Construct a rock sample matrix diagram, in which the positions of the rock sample matrix diagram correspond one-to-one with the grid of the rock sample. Then, fill the data in the column vector of the lithofacies mixing ratio matrix into the rock sample matrix diagram according to the grid corresponding to the row number, and construct the distribution map of lithofacies units corresponding to the lithofacies end elements of the column vector of the lithofacies mixing ratio matrix.

[0062] Training an elemental analysis network involves the following steps:

[0063] Several elemental analysis training samples are obtained, including energy dispersive spectroscopy (EDS) data collected by operators from actual rock sample testing experiments. These training samples are labeled with elemental analysis vectors obtained by performing chemical analysis on the rock samples corresponding to the EDS data. All labeled training samples are combined into an elemental analysis training set. The elemental analysis network is trained using this training set. During training, a loss value is constructed based on the difference between the network's output and the labeled elemental analysis vectors (MSE can be used). Based on this loss value, backpropagation of parameters is performed using gradient descent to adjust the network's parameters. If the network's accuracy meets expectations, the trained network is output; otherwise, training continues using the training set.

[0064] Example 2, a rock sample testing and analysis system based on spectral analysis, as shown in Figure 1, includes:

[0065] The rock sample spectral testing module is used to perform spectral testing on rock samples and acquire rock spectral cube data. The scale of the rock spectral cube data is H×W×C, where H×W is the number of grids corresponding to the rock sample, that is, dividing the plane corresponding to the rock sample into a network. The length of the grid plane composed of all grids is H, the width is W, and C is the dimension corresponding to the energy spectrum data collected by each grid. It should be noted that when performing spectral testing on the rock sample, it is necessary to first cut an approximate plane from the rock using a rock cutting machine, ensuring that the flatness of this approximate plane meets the expectations. This cut rock is recorded as the rock sample, which is used to facilitate subsequent spectral testing. The rock sample is then fixed in place using a clamp. On the XY scanning stage of the XRF instrument, set the scanning step size in the X / Y directions, for example, 0.5-2 mm. Set the excitation conditions of the XRF instrument, including tube pressure, tube current, filter type, and test time. Control the rock sample to move along the preset path on the XY scanning stage. Perform EDXRF acquisition once on each grid on the approximate plane through the XRF instrument to obtain the energy spectrum data on that grid. The energy spectrum data includes the energy count corresponding to each channel number. When the XRF instrument performs spectral analysis, it divides the energy range into several segments, with the energy range generally being 0-40 KeV. Each segment is regarded as a channel number, and the energy count is the X-ray photon count corresponding to the channel number.

[0066] The elemental analysis module performs preprocessing operations on the energy spectrum data corresponding to all grids. Preprocessing operations include baseline correction, smoothing, and normalization. Baseline correction estimates the continuous background using polynomial and SNIP algorithms and removes it from the energy spectrum data. Smoothing is performed using a filtering function. Normalization unifies the counts to between 0 and 1. These preprocessing operations reduce noise in the energy spectrum data. The preprocessed energy spectrum data is then fed into the elemental analysis network for further processing, outputting elemental analysis vectors. These vectors contain the content of several elemental substances or oxides, described as mass percentages, such as copper, iron, sulfur, and silicon dioxide. It's important to note that all elemental analysis vectors have the same dimension, and the elemental substance or oxide corresponding to each item in the vector is set by the operator. All elemental analysis vectors are then concatenated from top to bottom into an elemental analysis matrix, following a left-to-right, top-to-bottom grid traversal order. The elemental analysis network uses an MLP (Multi-Level Processing) model.

[0067] The lithofacies endmember inversion decomposition module is used to perform lithofacies endmember inversion decomposition operations on the elemental analysis matrix, obtaining a lithofacies endmember matrix and a lithofacies mixing ratio matrix. Each row in the lithofacies endmember matrix is ​​a standard elemental analysis vector corresponding to a lithofacies endmember. The standard elemental analysis vector has the same dimension as the elemental analysis vector. Lithofacies endmembers represent different regions in a rock sample with significant differences in composition and structure, such as quartz and feldspar grain regions, carbonate cement, and clay matrix in carbonate rocks. Each row in the lithofacies mixing ratio matrix represents the proportion of different lithofacies endmembers in the corresponding grid. The number of rows in the lithofacies mixing ratio matrix is ​​the same as the number of rows in the elemental analysis matrix. The number of columns in the facies mixing ratio matrix is ​​the same as the number of rows in the lithofacies endmember matrix. The lithofacies endmember inversion decomposition operation treats the elemental analysis matrix as the product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix, and iteratively optimizes the lithofacies endmember matrix and the lithofacies mixing ratio matrix under physical constraints and spatial neighborhood constraints. It should be noted that in rock samples, the elements are not homogeneously distributed, but rather a mixed distribution of multiple lithofacies endmembers. By constructing the rock sample into a grid-like energy spectrum data distribution and implementing the lithofacies endmember inversion decomposition operation, a more accurate analysis of the elemental composition within the rock sample can be performed based on lithofacies mixing.

[0068] The rock test analysis result construction module is used to reconstruct each column vector of the lithofacies mixing ratio matrix into a lithofacies unit distribution map corresponding to the lithofacies end-member. The lithofacies unit distribution map represents the distribution of different lithofacies end-members on the rock sample. Based on the lithofacies mixing ratio matrix, the average proportion of each lithofacies end-member in the rock sample is calculated. Then, a weighted calculation is performed based on the lithofacies end-member matrix and the average proportion of each lithofacies end-member in the rock sample to obtain the overall element distribution vector corresponding to the rock sample. The overall element distribution vector obtained by the weighted calculation of lithofacies end-members can obtain more accurate rock test analysis results when considering the heterogeneity of the distribution of various elements in the rock sample.

[0069] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for testing and analyzing rock samples based on spectral analysis, characterized in that, include: Spectroscopic tests were performed on the rock samples to obtain rock spectral cube data. The scale of the rock spectral cube data was H×W×C, where H×W is the number of grids corresponding to the rock sample, and C is the dimension of the energy spectrum data collected for each grid. Preprocessing was performed on the energy spectrum data corresponding to all grids, and then the preprocessed energy spectrum data was sent to the elemental analysis network for processing to output elemental analysis vectors. All elemental analysis vectors were then concatenated from top to bottom to form an elemental analysis matrix, following the grid traversal order from left to right and from top to bottom during the concatenation process. A lithofacies endmember inversion decomposition operation is performed on the element analysis matrix to obtain the lithofacies endmember matrix and the lithofacies mixing ratio matrix. The lithofacies endmember inversion decomposition operation treats the element analysis matrix as the product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix. Under physical constraints and spatial neighborhood constraints, the lithofacies endmember matrix and the lithofacies mixing ratio matrix are iteratively optimized. Each column vector of the lithofacies mixing ratio matrix is ​​reconstructed into a lithofacies unit distribution map corresponding to the lithofacies endmember. The lithofacies unit distribution map represents the distribution of different lithofacies endmembers on the rock sample. Based on the lithofacies mixing ratio matrix, the average ratio of each lithofacies endmember on the rock sample is calculated. Then, a weighted calculation is performed based on the lithofacies endmember matrix and the average ratio of each lithofacies endmember on the rock sample to obtain the overall element distribution vector corresponding to the rock sample. The lithofacies endmember inversion decomposition operation is performed on the element analysis matrix to obtain the lithofacies endmember matrix and the lithofacies mixing ratio matrix. Specifically, the steps are as follows: Step S1: Obtain candidate lithofacies quantity data, which includes several candidate lithofacies quantities; Step S2: Traverse the candidate lithofacies quantity dataset and solve for the lithofacies endmember matrix and the lithofacies mixing ratio matrix for each candidate lithofacies quantity; Step S2.1: Perform cluster analysis on all row vectors of the element analysis matrix, dividing all row vectors of the element analysis matrix into clusters corresponding to the number of candidate lithofacies, and using the cluster center of each cluster as the row vector of the lithofacies endmember matrix; Step S2.2: Under the condition of a fixed lithofacies endmember matrix, perform linear regression on the lithofacies mixing ratio matrix based on all element analysis vectors, project the solution onto a feasible region where each component is non-negative and the sum of each component is 1, and then perform spatial optimization based on a Markov random field to obtain the updated lithofacies mixing ratio matrix. Step S2.3: Under the condition of a fixed lithofacies mixing ratio matrix, perform linear regression operation on the lithofacies end-member matrix based on the column vector of the element analysis matrix, and then perform element content constraints on several lithofacies units corresponding to the lithofacies end-member matrix to obtain the updated lithofacies end-member matrix. Step S2.4: Repeat steps S2.2-S2.3 until convergence, and output the lithofacies end-member matrix and lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies; Step S3: Record the product of the lithofacies mixing ratio matrix and lithofacies end-member matrix corresponding to the number of candidate lithofacies as the corresponding theoretical element analysis matrix, then calculate the mean square error between all row vectors in the theoretical element analysis matrix and all corresponding row vectors in the element analysis matrix and record it as the fitting error, then calculate the spatial structure penalty value based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies, perform a weighted summation operation on the fitting error and spatial structure penalty value corresponding to the number of candidate lithofacies to obtain the comprehensive evaluation value, and select the lithofacies end-member matrix and lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies with the smallest comprehensive evaluation value for output.

2. The method for testing and analyzing rock samples based on spectral analysis according to claim 1, characterized in that, Under the condition of a fixed lithofacies end-member matrix, a linear regression operation is performed on the lithofacies mixing ratio matrix based on all element analysis vectors. The solution is then projected onto a feasible region where each component is non-negative and the sum of all components is 1. Spatial optimization based on a Markov random field is then performed to obtain the updated lithofacies mixing ratio matrix. Specifically, the following steps are included: Traversing the row vectors of all element analysis matrices, performing the following operation on each element analysis matrix row vector: using the row vector of the element analysis matrix as the dependent variable and the row vector of the lithofacies mixing ratio matrix that is in the same row position as the element analysis matrix as the independent variable, performing a linear regression operation on the row vector of the lithofacies mixing ratio matrix until all element analysis matrix row vectors have been traversed, outputting the row vectors of all lithofacies mixing ratio matrices, and denoting them as the preliminary fitted lithofacies mixing ratio vectors; traversing all the preliminary fitted lithofacies mixing ratio vectors, and performing a linear regression operation on each preliminary fitted lithofacies mixing ratio vector... The proportion vector is obtained by trimming negative values ​​to 0 in the initial fitted lithofacies mixing proportion vector, and then normalizing the trimmed initial fitted lithofacies mixing proportion vector to obtain a non-negative lithofacies mixing proportion vector. For the non-negative lithofacies mixing proportion vector, the following operations are performed: the element analysis vector in the element analysis matrix that has the same row position as the non-negative lithofacies mixing proportion vector is selected and denoted as the target element analysis vector. The similarity between the row vectors of the lithofacies end-member matrix and the target element analysis vector is calculated sequentially and denoted as the fitting match value. The index of the row vector of the lithofacies end-member matrix corresponding to the largest fitting match value is selected and denoted as h. A convex update vector is constructed, where the h-th term in the convex update vector is 1, and the rest are set to 0. The output of the product of the non-negative lithofacies mixing proportion vector and the update coefficient is added to the convex update vector to obtain the updated non-negative lithofacies mixing proportion vector. All updated non-negative lithofacies mixing proportion vectors are combined to form the updated lithofacies mixing proportion matrix.

3. The method for testing and analyzing rock samples based on spectral analysis according to claim 2, characterized in that, Under the condition of a fixed lithofacies mixing ratio matrix, a linear regression operation is performed on the lithofacies end-member matrix based on the column vectors of the element analysis matrix, and then element content constraints are applied to several lithofacies units corresponding to the lithofacies end-member matrix to obtain an updated lithofacies end-member matrix. The specific steps include: traversing the column vectors of all element analysis matrices, and performing the following operation on the column vectors of each element analysis matrix, using the column vectors of the element analysis matrix as the dependent variable and the column vectors of the lithofacies end-member matrix that are in the same column position as the column vectors of the element analysis matrix as the independent variable, performing a linear regression operation on the column vectors of the lithofacies end-member matrix until all column vectors of the element analysis matrix have been traversed, outputting the column vectors of all lithofacies end-member matrices, and recording them as the preliminary fitted lithofacies end-member vectors; All initially fitted lithofacies end-member vectors are concatenated into an initially fitted lithofacies end-member matrix. The row vectors of the initially fitted lithofacies end-member matrix are traversed, and the similarity between the row vectors of the initially fitted lithofacies end-member matrix and the lithofacies end-member standard vectors in the lithofacies end-member standard library is calculated sequentially. This similarity is denoted as the lithofacies end-member matching value. The lithofacies end-member standard library includes several lithofacies end-member standard vectors and their corresponding element content constraint ranges. The element content constraint range corresponding to the largest lithofacies end-member matching value is selected. Data in the row vectors of the initially fitted lithofacies end-member matrix that exceed the corresponding element content constraint range are pruned to obtain the updated lithofacies end-member matrix.

4. The method for testing and analyzing rock samples based on spectral analysis according to claim 3, characterized in that, The spatial structure penalty value is calculated based on the lithofacies mixing ratio matrix corresponding to the number of candidate lithofacies. The specific steps include: constructing a lithofacies label matrix, in which the positions of the lithofacies label matrix correspond one-to-one with the grid of the rock sample, and the data of each position in the lithofacies label matrix is ​​the lithofacies endmember label corresponding to the largest data in the row vector of the lithofacies mixing ratio matrix corresponding to the grid. All grid neighborhood pairs are obtained, and the ratio between the number of grid neighborhood pairs with inconsistent lithofacies endmember labels corresponding to the grid and the total number of all grid neighborhood pairs is calculated. This ratio is recorded as the spatial structure penalty value.

5. The method for testing and analyzing rock samples based on spectral analysis according to claim 4, characterized in that, Reconstructing each column vector of the lithofacies mixing ratio matrix into a lithofacies unit distribution map corresponding to the lithofacies endmember involves the following steps: constructing a rock sample matrix diagram, where the positions in the rock sample matrix diagram correspond one-to-one with the grid of the rock sample; then filling the data in the column vectors of the lithofacies mixing ratio matrix into the rock sample matrix diagram according to the grid corresponding to the row number, thereby constructing a lithofacies unit distribution map corresponding to the lithofacies endmember of the column vector of the lithofacies mixing ratio matrix.

6. The method for testing and analyzing rock samples based on spectral analysis according to claim 5, characterized in that, Training an elemental analysis network involves the following steps: obtaining several elemental analysis training samples, which include energy spectrum data; labeling the elemental analysis training samples using elemental analysis vectors; forming an elemental analysis training set from all labeled elemental analysis training samples; and training the elemental analysis network using the elemental analysis training set.

7. A rock sample testing and analysis system based on spectral analysis, characterized in that, The system employs a rock sample testing and analysis method based on spectral analysis as described in any one of claims 1-6, comprising: a rock sample spectral testing module for performing spectral testing on the rock sample and acquiring rock spectral cube data, wherein the scale of the rock spectral cube data is H×W×C, where H×W is the number of grids corresponding to the rock sample, and C is the dimension corresponding to the energy spectrum data collected by each grid; an elemental analysis module for performing preprocessing operations on the energy spectrum data corresponding to all grids, then sending the preprocessed energy spectrum data into an elemental analysis network for processing, outputting elemental analysis vectors; and concatenating all elemental analysis vectors from top to bottom into an elemental analysis matrix, following a grid traversal order from left to right and from top to bottom during the concatenation process; and a lithofacies endmember inversion decomposition module for performing elemental analysis on the rock sample. The analysis matrix performs a lithofacies endmember inversion decomposition operation to obtain a lithofacies endmember matrix and a lithofacies mixing ratio matrix. The lithofacies endmember inversion decomposition operation treats the element analysis matrix as the product of the lithofacies mixing ratio matrix and the lithofacies endmember matrix, and iteratively optimizes the lithofacies endmember matrix and the lithofacies mixing ratio matrix under physical constraints and spatial neighborhood constraints. The rock test analysis result construction module is used to reconstruct each column vector of the lithofacies mixing ratio matrix into a lithofacies unit distribution map corresponding to the lithofacies endmember. The lithofacies unit distribution map represents the distribution of different lithofacies endmembers on the rock sample. Based on the lithofacies mixing ratio matrix, the average ratio of each lithofacies endmember on the rock sample is calculated. Then, based on the lithofacies endmember matrix and the average ratio of each lithofacies endmember on the rock sample, a weighted calculation is performed to obtain the overall element distribution vector corresponding to the rock sample.

Images