Intelligent discrimination method and apparatus for mine water inrush source, device, and storage medium

Abstract

An intelligent discrimination method and apparatus for a mine water inrush source, a device, and a storage medium. The method comprises: acquiring a water chemical index of a water source sample to be discriminated (S1), wherein the water chemical index comprises a plurality of inorganic indexes and a plurality of organic indexes; and inputting the water chemical index into a preset water source discrimination model to obtain a water source type of said water source sample (S2). In addition to inorganic indexes for discrimination, organic indexes are added, so that the problem of being difficult to accurately discriminate, by means of inorganic water chemical indexes, water samples with similar component is solved, thereby improving the accuracy of water source type discrimination.

Description

A method, device, equipment and storage medium for intelligently identifying the source of water inrush in a mine Technical Field

[0001] The present invention belongs to the technical field of mines and relates to a method, device, equipment and storage medium for intelligently identifying a source of water inrush in a mine. Background Art

[0002] Mine flooding is a major hidden danger affecting mine safety and production. In my country, almost all mines face varying degrees of flooding. Identifying the source of the water is crucial for implementing appropriate preventative measures and countermeasures. Currently, indicators used to identify water sources primarily rely on inorganic indicators such as the main anions and cations in the water and total dissolved solids (TDS). However, for water sources with similar composition, accurate identification based solely on inorganic water chemistry indicators is difficult.

[0003] Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method, device, equipment and storage medium for intelligent identification of the source of mine water, so as to solve the problem that inorganic water chemical indicators in the existing technology are difficult to accurately identify the source of mine water with similar components.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A method for intelligently identifying the source of water inrush in a mine, the method comprising the following steps:

[0007] S1, obtaining water chemical indicators of the water source sample to be identified; water chemical indicators include inorganic indicators and organic indicators; the inorganic indicators include anions and cations and TDS value in the water sample; organic indicators include total organic carbon, ultraviolet absorbance and fluorescence intensity of dissolved organic matter fluorescent components in the water sample;

[0008] S2, inputting the water chemical index into a preset water source discrimination model to obtain the water source type of the water source sample to be discriminated; the construction of the water source discrimination model includes:

[0009] S21, constructing a data set of water chemical indicators of multiple water source samples according to the water chemical indicators of multiple water source samples;

[0010] S22, reducing the dimension of the data set to obtain a data set after dimension reduction;

[0011] S23, training the preset model according to the data set after dimensionality reduction to obtain a water source discrimination model.

[0012] The present invention also includes the following technical features:

[0013] Specifically, the data set of water chemical indicators of the water source sample in S21 is expressed by the following formula (1):

[0014] Where X is the dataset, x is one of the water chemistry indicators, n is the number of water source samples, and p is the number of water chemistry indicators.

[0015] Specifically, the S22 includes:

[0016] Perform zero-mean normalization on the data set to obtain a standardized data matrix;

[0017] Perform principal component extraction on the standardized data matrix to obtain the principal component score matrix;

[0018] The dimension of the standardized data matrix is ​​reduced by the principal component score matrix to obtain the reduced dimension data set.

[0019] Specifically, the zero-mean normalization process for the data set is implemented by the following formula (3):

[0020] Among them, μ A is the mean value of a water chemical index A, σ A is the standard deviation of a water chemical index A, v is the normalized value of a certain value x of a water chemical index A;

[0021] The dataset after dimensionality reduction is expressed by formula (4): Y = XP (4)

[0022] Among them, Y is the dataset after dimensionality reduction, X is the dataset obtained by standardizing the hydrochemical indicators of multiple water source samples, and P is the principal component score matrix.

[0023] Specifically, in S23, the data set after dimensionality reduction is input into a preset model for training; the number and depth of subtrees of the random forest model are optimized by an intelligent swarm optimization algorithm to obtain a water source discrimination model.

[0024] A device for intelligently identifying the source of water inrush in a mine, which can implement the method for intelligently identifying the source of water inrush in a mine, comprises:

[0025] An acquisition module is used to obtain water chemical indicators of the water source sample to be identified; the water chemical indicators include inorganic indicators and organic indicators; the inorganic indicators include anions and cations in the water and the TDS value; the organic indicators include total organic carbon in the water sample, ultraviolet absorbance and fluorescence intensity of the fluorescent components of dissolved organic matter;

[0026] The discrimination module is used to input the water chemical indicators into a preset water source discrimination model to obtain the water source type of the water source sample to be discriminated; the construction of the water source discrimination model includes: constructing a data set of water chemical indicators of multiple water source samples based on the water chemical indicators of multiple water source samples; reducing the dimension of the data set to obtain a reduced-dimensional data set; and training the preset model based on the reduced-dimensional data set to obtain a water source discrimination model.

[0027] Specifically, in the discrimination module, the data set of water source sample water chemical index is expressed by the following formula (1):

[0028] Where X is the dataset, x is one of the water chemistry indicators, n is the number of water source samples, and p is the number of water chemistry indicators.

[0029] Specifically, in the discrimination module, the data set is reduced in dimension to obtain the reduced-dimensional data set, including: performing zero-mean normalization processing on the data set to obtain a standardized data matrix; performing principal component extraction on the standardized data matrix to obtain a principal component score matrix; performing dimensionality reduction on the standardized data matrix through the principal component score matrix to obtain a reduced-dimensional data set; the zero-mean normalization processing on the data set is achieved by the following formula (3):

[0030] Among them, μ A is the mean value of a water chemical index A, σ A is the standard deviation of a water chemical index A, v is the normalized value of a certain value x of a water chemical index A;

[0031] The dataset after dimensionality reduction is expressed by formula (4): Y = XP (4)

[0032] Where Y is the dataset after dimensionality reduction, X is the dataset obtained by standardizing the hydrochemical indicators of multiple water source samples, and P is the principal component score matrix;

[0033] The reduced-dimensional data set is input into the preset model for training; the number and depth of subtrees in the random forest model are optimized through the intelligent swarm optimization algorithm to obtain a water source discrimination model.

[0034] A computer device includes a memory and a processor, wherein the memory stores a computer program and the processor implements the steps of the method for intelligently identifying the source of water inrush in a mine when executing the computer program.

[0035] A computer-readable storage medium is used to store program instructions, which can be executed by a processor to implement the steps of the method for intelligently distinguishing the source of mine water inrush.

[0036] Compared with the prior art, the present invention has the following technical effects:

[0037] In this invention, a water source discrimination model is used to identify the source type of a water sample based on the multiple inorganic and organic indicators contained in the sample's hydrochemical parameters. This approach, incorporating organic indicators into the identification of water samples based on inorganic indicators, addresses the difficulty of accurately identifying water samples with similar composition using inorganic hydrochemical indicators, thereby improving the accuracy of water source classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] FIG1 is a schematic flow chart of an embodiment of a method for identifying the source of water inrush in a mine provided by the present invention;

[0039] FIG2 is a fluorescence spectrum and loading diagram of an organic component;

[0040] FIG3 is a fluorescence spectrum and loading diagram of another organic component;

[0041] FIG4 is a fluorescence spectrum and loading diagram of another organic component;

[0042] FIG5 is a schematic diagram of a flow chart of an embodiment of obtaining a water source discrimination model in an embodiment of the present invention;

[0043] Figure 6 is a schematic diagram of the performance of the traditional RF model under different discriminant index data sets;

[0044] FIG7 is a performance diagram of the RF model improved based on AFSA under different discrimination index data sets in an embodiment of the present invention;

[0045] FIG8 is a schematic diagram showing the results of water source type identification using the traditional RF model;

[0046] FIG9 is a schematic diagram showing the result of water source type identification based on the RF model improved by AFSA in an embodiment of the present invention;

[0047] FIG10 is a schematic structural diagram of a device for identifying a source of water inrush in a mine according to an embodiment of the present invention;

[0048] FIG11 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] At present, the indicators used to identify water sources are mainly inorganic indicators such as the main anions and cations in water and total dissolved solids (TDS). However, in fact, the total organic carbon (TOC), ultraviolet absorbance (UV 254) and dissolved organic matter (DOM). At present, the methods for water source identification using hydrochemical methods are mainly concentrated in three categories: multivariate statistics, nonlinear analysis and machine learning. Compared with traditional methods, machine learning algorithms have more advantages in processing nonlinear and high-dimensional data, and have stronger adaptability. Among the commonly used machine algorithms, support vector machines (SVM), extreme learning machines (ELM), random forests (RF), BP neural networks (Back Propagation Neural Network), etc. have all been used in mine water source identification; for water sources with similar components, it is difficult to make an accurate judgment only through inorganic hydrochemical indicators.

[0050] The present invention proposes a method, device, equipment, and storage medium for intelligently identifying the source of mine water. The method comprises: obtaining hydrochemical indicators of a water source sample to be identified; the hydrochemical indicators include multiple inorganic and organic indicators; and inputting the hydrochemical indicators into a preset water source identification model to determine the source type of the water source sample to be identified. This approach, incorporating organic indicators in addition to inorganic indicators, solves the problem of inorganic hydrochemical indicators being difficult to accurately identify water samples with similar compositions, thereby improving the accuracy of water source classification.

[0051] Specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent modifications made on the basis of the technical solution of this application fall within the protection scope of the present invention.

[0052] Example:

[0053] This embodiment provides a method for intelligently identifying the source of water inrush in a mine. FIG1 is a flow chart of the method for intelligently identifying the source of water inrush in a mine according to the present invention. Referring to FIG1 , the method includes the following steps:

[0054] S1, obtaining water chemical indicators of the water source sample to be identified; wherein the water chemical indicators include multiple inorganic indicators and organic indicators;

[0055] Inorganic indicators include the main anions and cations in water and TDS value. The main ions in water include K + 、Na + , Ca 2+ Mg 2+ 、Cl - 、SO4 2- 、HCO3 -Etc. Refer to GB / T 14848-2017 Groundwater Quality Standard to test various inorganic ions and inorganic indicators such as TDS in water samples. K + 、Na + They are all common inorganic ions in water, but usually K + The content of K + 、Na + As a whole, K + +Na + .

[0056] Organic indicators include total organic carbon, UV absorbance and dissolved organic matter in the water layer. The differences between different organic components in dissolved organic matter can be intuitively reflected through fluorescence spectra. TOC can be tested using the multi N / C 2100 Expert Total Organic Carbon / Total Nitrogen Analyzer; UV can be tested using the Evolution 60 UV-Visible Photometer. 254 Detection; DOM fluorescence data was extracted using a fluorescence spectrophotometer (HITACHI F-7000). To quantitatively analyze the differences in DOM in different water samples, the DOM fluorescence data of each water source sample was quantified using mathematical methods. Specifically, the main components of the fluorescence spectrum were extracted and quantified using the Parallel Factors Analysis (PARAFAC) method. Specifically, the three-dimensional fluorescence data of DOM in all water samples were first compiled. Then, the Digital Object Identifier for Fluorescence Microscopy (DOMFluor) toolbox provided by Matrix Laboratory (MATLAB) was used to analyze and process the three-dimensional fluorescence data. The specific processing steps include:

[0057] (a1) Load data and draw excitation-emission matrices (EEMs);

[0058] (a2) cutting the spectral region affected by the scattering peak;

[0059] (a3) Outlier identification and load analysis of models with different component numbers;

[0060] (a4) Perform split-half test on the model with the selected number of components;

[0061] (a5) Export experimental data, including: fluorescence spectrum of each component, fluorescence intensity of each component in each sample and its corresponding emission load and excitation load.

[0062] Figures 2-4 are schematic diagrams of the fluorescence spectra and loads of the organic components in this embodiment, wherein Figures 2 to 4 are the fluorescence spectra and loads of three organic components, respectively. Referring to Figure 2, component 1 has two excitation peaks (250nm / 330nm) and one emission peak (405nm), which shows that component 1 contains hydrophobic organic acids and humic acid-like substances, with the latter being the majority. Referring to Figure 3, component 2 has three excitation peaks (230nm / 250nm / 280nm) and one emission peak (308nm), which shows that component 2 contains tryptophan-like substances and tryptophan-containing protein-like substances. Referring to Figure 4, component 3 has two excitation peaks (225nm / 280nm / 305nm) and one emission peak (340nm), which shows that component 3 contains tyrosine and tryptophan-containing protein-like substances, with the former being the majority.

[0063] S2, inputting the water chemical index into a preset water source identification model to obtain the water source type of the water source sample to be identified.

[0064] Water source types include Quaternary water, Cretaceous water, Zhiluo Formation water or Yan'an Formation water and other aquifer water samples.

[0065] Identifying the source of gushing water is crucial for preventing mine flooding. Therefore, in this embodiment of the present invention, prior to determining the source type, the target aquifer to be studied can be determined by studying the geological and hydrogeological conditions of the area being studied. In other words, the hydrochemical indicators of the water source sample to be identified can be samples obtained from the target aquifer.

[0066] FIG5 is a flow chart of obtaining a water source discrimination model according to an embodiment of the present invention. Referring to FIG5 , the process of obtaining a water source discrimination model includes the following steps:

[0067] S21, constructing a data set of water chemical indicators of the plurality of water source samples according to the water chemical indicators of the plurality of water source samples;

[0068] The water chemical index includes the above-mentioned inorganic indexes and organic indexes. Thus, for each sample water source, a set of data including the above-mentioned inorganic indexes and organic indexes can be obtained. By obtaining the water chemical indexes of multiple sample water sources, a data set of water chemical indexes of multiple sample water sources can be constructed. For example, the inorganic index includes K + +Na + , Ca 2+ Mg 2+ 、Cl - 、SO4 2- 、HCO3 -Inorganic ions such as total organic carbon, ultraviolet absorbance, and dissolved organic matter (TDS). Organic indicators include total organic carbon, ultraviolet absorbance, and dissolved organic matter. The specific organic components and content of dissolved organic matter can be determined by analyzing the fluorescence spectrum. For example, the organic components in the embodiment include the three components shown in Figures 2 to 4. In this way, a data set of water chemical indicators for each sample water source can be obtained, thereby constructing a data set of water chemical indicators for multiple sample water sources.

[0069] The number of water source samples is n, and each water source sample includes p water chemical indicators, which are expressed by x1, x2, ..., x p The data set of water chemical indicators of n water source samples is expressed by the following formula (1):

[0070] Where X is the dataset, x is one of the water chemistry indicators, n is the number of water source samples, and p is the number of water chemistry indicators.

[0071] S22, reducing the dimension of the data set by a preset data method to obtain a reduced-dimensional data set;

[0072] The identification of the source of mine water inrush is affected by the influence of multiple hydrochemical indicators, such as the influence between the inorganic ions and organic ions mentioned above. There is a certain correlation between the hydrochemical indicators, which greatly increases the complexity of the analysis. Therefore, it is necessary to reduce the dimensionality of the original data set and process the high-dimensional, correlated variables into low-dimensional, uncorrelated, new comprehensive variables that can represent most of the information of the original variables. Unnecessary calculations can be reduced by determining the correlation between the various hydrochemical indicators. The correlation between the various hydrochemical indicators is determined by the Pearson correlation coefficient, which is expressed by formula (2):

[0073] Among them, r xy is the Pearson correlation coefficient between indicator x and indicator y, n is the number of samples, x i is the actual value of the index x, y i is the actual value of the i-th indicator y.

[0074] Table 1 is a correlation coefficient matrix consisting of the correlation coefficients between various water chemical indicators in the sample water sources actually measured.

[0075] Table 1 Correlation coefficient matrix composed of the correlation coefficients between various water chemical indicators in the sample water source actually measured

[0076] Among them, water chemical indicators include inorganic indicators K + +Na + , Ca 2+ 、Cl - 、SO42- Inorganic ions and TDS, organic indicators include UV 254 , TOC and component 2 as shown in Figure 2.

[0077] As can be seen from Table 1, there is a clear correlation between the indicators, among which SO4 2- With K + +Na + , Ca 2+ The correlations of K and TDS were 0.916, 0.949 and 0.982 respectively; + +Na + With Ca 2+ The correlation coefficients of Ca and TDS were 0.825 and 0.969 respectively; 2+ The correlation coefficients with TDS were 0.908. + +Na + , Ca 2+ 、SO4 2- There is a significant negative correlation between the 8 indicators and TDS, indicating that there is information redundancy in the dataset. If these 8 indicators are used directly to identify water source types, the model complexity and computational effort will increase, affecting the efficiency and accuracy of the discrimination. Therefore, it is necessary to reduce the dimensionality of the dataset.

[0078] In the embodiment of the present invention, the dimension reduction of the data set of the water chemical indicators of the above-mentioned multiple sample water sources is achieved by principal component analysis (PCA).

[0079] Among them, PCA is a classic statistical method that extracts new variables that reflect the essence of things by performing linear transformation on the attribute characteristics of the original variables, while removing redundancy and reducing noise to achieve the purpose of dimensionality reduction.

[0080] The above step S22 includes the following steps:

[0081] S221, performing zero-mean normalization processing on the data set to obtain a standardized data matrix;

[0082] S222, performing principal component extraction on the standardized data matrix to obtain a principal component score matrix;

[0083] S223, reducing the dimension of the standardized data matrix using the principal component score matrix to obtain a reduced-dimensionality data set.

[0084] The zero-mean normalization process is performed on the dataset so that the mean of each attribute after the transformation of the data in the dataset is zero. This can be obtained by subtracting the mean of each attribute from the value of the attribute in the dataset.

[0085] Zero-mean normalization of the data set can be achieved using the following formula (3):

[0086] Among them, μ A is the mean value of a water chemical index A, σ A is the standard deviation of a water chemical index A, and v is the normalized value of a certain value x of a water chemical index A.

[0087] After the data set is normalized to zero mean using the above formula (3), a standardized data matrix can be obtained. Then, principal component extraction is performed on the standardized data matrix to obtain a principal component score matrix.

[0088] Specifically, the above step S222 includes the following steps:

[0089] Step b1, determining the covariance matrix of the standardized data matrix;

[0090] Step b2, determining the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues;

[0091] Step b3, selecting N eigenvalues ​​that meet preset conditions based on the size of the eigenvalues ​​to obtain an eigenvalue set;

[0092] Step b4: construct a principal component score matrix based on the eigenvector corresponding to each eigenvalue in the eigenvalue set.

[0093] When the hydrochemical indicators for each sample water source contain p indicators, solving the covariance matrix yields p eigenvalues. Arranging the eigenvalues ​​from largest to smallest results in a gradually decreasing amount of characteristic information contained in the corresponding data, meaning that the contribution rate of each eigenvalue decreases. Therefore, to reduce redundant calculations during the calculation process, only the eigenvectors corresponding to certain eigenvalues ​​can be used as principal components based on their contribution rates, thereby constructing a principal component score matrix.

[0094] The contribution rate of an eigenvalue is the proportion of the eigenvalue to the sum of all eigenvalues ​​in the covariance matrix, that is, the proportion of the variance corresponding to the eigenvalue in the principal component to the sum of all variances. The greater the contribution rate of the eigenvalue, the more information the original water chemical index in the corresponding principal component contains. Therefore, the value of N can be determined by the contribution rate corresponding to the eigenvalue of the covariance matrix.

[0095] The preset condition may be that the characteristic value is greater than a first preset threshold value, wherein the value of the first preset threshold value may be a preset fixed threshold value, or may be specifically determined based on application requirements in an actual calculation process.

[0096] When an eigenvalue is less than a certain threshold, its corresponding contribution rate is relatively small, meaning it contains less information. Therefore, these eigenvalues ​​can be disregarded. By selecting eigenvalues ​​greater than a first preset threshold and excluding those with less information, redundant calculations can be reduced, improving computational efficiency.

[0097] N eigenvalues ​​meeting preset conditions are selected according to the size of the eigenvalue to obtain a eigenvalue set. Each eigenvalue can be compared with a first preset threshold value, and eigenvalues ​​greater than the first preset threshold value are selected to construct a eigenvalue set.

[0098] In other embodiments, the preset condition may be that the cumulative contribution rate of each characteristic value is greater than a second preset threshold.

[0099] Likewise, the value of the second preset threshold may also be a preset fixed threshold, or may be specifically determined based on application requirements in an actual calculation process.

[0100] In some embodiments, taking the second threshold of 85% as an example, the eigenvalues ​​are sorted from large to small according to their size. The eigenvalues ​​with a cumulative contribution rate exceeding 85% can actually represent the vast majority of the information. Therefore, the value of N can be the first N eigenvalues ​​with a contribution rate exceeding 85%. The specific contribution rate can also be selected based on the needs of the actual application process, and the embodiments of the present invention do not specifically limit this.

[0101] After completing steps b1 to b4 above, a principal component score matrix can be obtained, and the dimension of the standardized data matrix can be reduced by using the principal component score matrix, thereby obtaining a data set after dimension reduction.

[0102] The dataset after dimensionality reduction is expressed by formula (4): Y = XP (4)

[0103] Among them, Y is the dataset after dimensionality reduction, X is the dataset obtained after the hydrochemical indicators of multiple water samples are standardized, and P is the principal component score matrix.

[0104] Among them, each principal component in the principal component score matrix is ​​expressed by formula (5): F m =p m1 X1+p m2 X2+…+p mp X p (5)

[0105] Among them, F m is the mth principal component, p m1 to p mpThe mth principal component corresponds to the principal component score of each indicator (i.e., the mth row of the principal component matrix), X1 to X p are the standardized values ​​of the 1st to pth indicators of the n samples, that is, the 1st to pth columns of the matrix X.

[0106] S23, training the preset model according to the data set after dimensionality reduction to obtain a water source discrimination model.

[0107] Among them, the preset model can be a random forest model;

[0108] In some embodiments, the above step S23 may further include the following steps:

[0109] S231, inputting the reduced-dimensionality dataset into a preset model for training;

[0110] S232, optimize the number and depth of subtrees of the random forest model through an intelligent swarm optimization algorithm to obtain a water source discrimination model.

[0111] Specifically, after the dimensionality-reduced dataset is input into the random forest model, K-fold cross-validation can be used to partition the dataset, with the K-th fold used as the test set, where the value of K can be any positive integer less than the amount of data in the dataset. After taking the K-th fold as the test set, the remaining K-1 folds can be used as the training set, and this is repeated K times. In the classifier of the random forest model, bootstrap sampling is performed on the training set. During the feature selection process, the Gini index can be used to select the best partitioning features. The Classification and Regression Tree (CART) algorithm is used to construct a decision tree, and the test set is predicted based on the voting results of the decision tree.

[0112] At the same time, the number of random forest subtrees n_estimators and depth depth can be optimized by intelligent swarm optimization algorithm. Here we use Artificial Fish Swarm Algorithm (AFSA). In AFSA, the state (current position) of each artificial fish is X i =(n_estimators, depth), where the food concentration (fitness) at the current position of the artificial fish is the average accuracy of the RF classifier after the K-fold cross-validation. First, the positions of all artificial fish in AFSA are initialized. Then, based on the current fitness, the algorithm executes four behaviors: foraging, flocking, chasing, and random. The algorithm terminates when the maximum number of iterations is reached.

[0113] Figure 4 is a schematic diagram of the performance comparison of the RF model actually measured under different indicator systems. Performance tests were carried out under four different indicator data sets, including accuracy (Accuracy), precision (Precision), recall rate (Recall) and f1 index (f1_score). The performance of different models for water source type identification is compared when organic indicators are added to the inorganic indicator discrimination provided in the embodiment of the present invention; and the performance of different models for water source type identification is compared when traditional inorganic indicator discrimination is used. Figure 6 is a schematic diagram of the performance of the traditional RF model under different indicator data sets. Figure 7 is a schematic diagram of the performance of the RF model improved based on AFSA in the embodiment of the present invention under different indicator data sets.

[0114] As shown in Figure 4, combining inorganic and organic indicators can significantly improve the accuracy of water source identification. Furthermore, AFSA can further improve the accuracy of the RF model.

[0115] Figure 5 is a schematic diagram showing the results of actual testing using different models to discriminate water source types according to an embodiment of the present invention. The sample water sources include 74. Figure 8 is a schematic diagram showing the results of water source type identification using a traditional RF model, and Figure 9 is a schematic diagram showing the results of water source type identification using an improved RF model based on AFSA according to an embodiment of the present invention.

[0116] As shown in Figures 8 and 9, the traditional RF model misclassified 74 water source types, resulting in two false positives. However, the AFSA-enhanced RF model provided by the present invention only misclassified one, fully demonstrating the performance advantages of the AFSA-enhanced RF model.

[0117] In this embodiment of the present invention, a water source discrimination model is used to identify the source type of a water sample based on the multiple inorganic and organic indicators contained in the sample's hydrochemical parameters. This approach, incorporating organic indicators into the inorganic hydrochemical parameters, addresses the difficulty of accurately distinguishing water samples with similar composition using inorganic hydrochemical parameters, thereby improving the accuracy of water source classification. Furthermore, RF is optimized using AFSA. The improved algorithm exhibits superior global search capabilities and convergence, enhancing the efficiency of model parameter tuning.

[0118] In some embodiments, after obtaining the water source discrimination model in step S23, the method may further include the following steps:

[0119] (c1) obtaining water chemical indicators of a plurality of test sample water sources;

[0120] (c2) inputting water chemical indicators of the plurality of test sample water sources into a water source discrimination model to obtain water source type discrimination results of the plurality of test sample water sources;

[0121] (c3) determining whether the accuracy of the judgment result is greater than a preset threshold;

[0122] (c4) when the accuracy is greater than or equal to a preset threshold, outputting a water source discrimination model;

[0123] (c5) When the accuracy is less than the preset threshold, the water source discrimination model is iteratively trained again.

[0124] The test sample water source may be a portion of samples selected from a plurality of sample water sources, or may be a separately obtained water source sample of a known water source type, which is not specifically limited in the embodiment of the present invention.

[0125] Due to factors such as the number of samples and the number of iterations during model training, the performance of a water source discrimination model obtained through a single training session may not meet the requirements of actual applications. Therefore, after obtaining the water source discrimination model, its performance can be tested using the aforementioned method. The accuracy of the water source discrimination results output by the model can be used to assess the model's performance. Only when the water source discrimination model's performance meets the requirements should the model be output for practical use.

[0126] In this embodiment of the present invention, a water source discrimination model is used to identify the source type of a water sample based on the multiple inorganic and organic indicators contained in the sample's hydrochemical parameters. This approach, incorporating organic indicators into the inorganic indicator approach, addresses the difficulty of accurately distinguishing water samples with similar composition using inorganic hydrochemical indicators, thereby improving the accuracy of water source classification. Furthermore, RF is optimized using AFSA. The improved algorithm boasts enhanced global search capabilities and convergence, improving the efficiency of model parameter tuning.

[0127] An embodiment of the present invention also provides a device for identifying the source of water inrush in a mine. Figure 10 is a schematic structural diagram of a device for identifying the source of water inrush in a mine according to an embodiment of the present invention. Referring to Figure 10 , the device comprises: an acquisition module for acquiring water chemical indicators of a water source sample to be identified; the water chemical indicators include multiple inorganic indicators and multiple organic indicators; and a discrimination module for inputting the water chemical indicators into a preset water source discrimination model to determine the water source type of the water source sample to be identified.

[0128] The preset water source discrimination model is obtained by the following processing: obtaining a data set of water chemical indicators of multiple water source samples based on the water chemical indicators of multiple water source samples; reducing the dimension of the data set by a preset data processing method to obtain a reduced-dimensional data set; training the preset model based on the reduced-dimensional data set to obtain a water source discrimination model.

[0129] The data set is reduced in dimension by a preset data processing method to obtain a reduced-dimensional data set, including: performing zero-mean normalization on the data set to obtain a standardized data matrix; performing principal component extraction on the standardized data matrix to obtain a principal component score matrix; and reducing the dimension of the standardized data matrix using the principal component score matrix to obtain a reduced-dimensional data set.

[0130] The principal component extraction is performed on the standardized data matrix to obtain the principal component score matrix, including: determining the covariance matrix of the standardized data matrix; determining the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; selecting N eigenvalues ​​that meet preset conditions according to the size of the eigenvalues ​​to obtain an eigenvalue set; and constructing the principal component score matrix according to the eigenvector corresponding to each eigenvalue in the eigenvalue set.

[0131] The preset model is a random forest model; the preset model is trained according to the reduced-dimensionality data set to obtain a water source discrimination model, including: inputting the reduced-dimensionality data set into the preset model for training; optimizing the number and depth of subtrees of the random forest model through an intelligent swarm optimization algorithm to obtain a water source discrimination model.

[0132] The discrimination device also includes a testing module for obtaining water chemical indicators of multiple test sample water sources; inputting the water chemical indicators of the multiple test sample water sources into a water source discrimination model to obtain water source type judgment results of the multiple test sample water sources; determining whether the accuracy of the judgment result is greater than a preset threshold; when the accuracy is greater than or equal to the preset threshold, outputting the water source discrimination model; when the accuracy is less than the preset threshold, re-iteratively training the water source discrimination model.

[0133] An embodiment of the present invention provides an electronic device that can be used in accordance with the mine water source identification method described in one or more of the above embodiments. FIG11 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. As shown in FIG11 , the electronic device can utilize general-purpose computer hardware, including a processor and memory.

[0134] At least one processor can constitute any physical device having a circuit that performs a logical operation on one or more inputs. For example, at least one processor may include one or more integrated circuits (ICs), including application specific integrated circuits (ASICs), microchips, microcontrollers, microprocessors, all or part of a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a field programmable gate array (FPGA), or other circuits suitable for executing instructions or performing logical operations. The instructions executed by at least one processor can be preloaded into a memory integrated with or embedded in a controller, or can be stored in a separate memory. The memory can include a random access memory (RAM), a read-only memory (ROM), a hard disk, an optical disk, a magnetic medium, a flash memory, other permanent, fixed or volatile memory, or any other mechanism capable of storing instructions. In some embodiments, at least one processor may include more than one processor. Each processor may have a similar structure, or the processors may have different configurations that are electrically connected or disconnected from each other. For example, the processors may be separate circuits or integrated into a single circuit. When more than one processor is used, the processors may be configured to operate independently or collaboratively. The processors may be coupled electrically, magnetically, optically, acoustically, mechanically, or by other means that allow them to interact.

[0135] The present application provides a computer storage medium storing computer executable instructions. After the computer executable instructions are executed by a processor, a method for identifying a source of water inrush in a mine as described in one or more of the above embodiments can be implemented.

[0136] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments.

[0137] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present application.

Claims

1. An intelligent discrimination method for the water source of mine water inrush, characterized in that The method includes the following steps: S1. Obtain the hydrochemical indexes of the water source sample to be discriminated; the hydrochemical indexes include inorganic indexes and organic indexes; the inorganic indexes include anions, cations and TDS value in the water sample; the organic indexes include total organic carbon, ultraviolet absorbance and fluorescence intensity of the fluorescence components of dissolved organic matter in the water sample; S2. Input the hydrochemical indexes into a preset water source discrimination model to obtain the water source type of the water source sample to be discriminated; the construction of the water source discrimination model includes: S21. Construct a data set of the hydrochemical indexes of multiple water source samples according to the hydrochemical indexes of multiple water source samples; S22. Reduce the dimension of the data set to obtain a data set after dimension reduction; S23. Train a preset model according to the data set after dimension reduction to obtain a water source discrimination model.

2. The intelligent discrimination method for mine water inrush water source according to claim 1, characterized in that, The data set of the hydrochemical indexes of the water source sample in S21 is represented by the following formula (1): Wherein, X is the data set, x is one of the hydrochemical indexes, n is the number of water source samples, and p is the number of hydrochemical indexes.

3. The intelligent discrimination method for mine water inrush source according to claim 2, characterized in that, The S22 includes: Perform zero-mean normalization processing on the data set to obtain a standardized data matrix; Extract the principal components from the standardized data matrix to obtain a principal component score matrix; Reduce the dimension of the standardized data matrix through the principal component score matrix to obtain a data set after dimension reduction.

4. The intelligent discrimination method for mine water inrush water source according to claim 3, characterized in that The zero-mean normalization process for the data set is implemented through the following formula (3): Among them, μ A is the mean value of a hydrochemical index A, σ A is the standard deviation of a hydrochemical index A, and v is the value after normalization of a certain value x of a hydrochemical index A; The data set after dimension reduction is represented by formula (4): Y = XP (4) Wherein, Y is the data set after dimension reduction, X is the data set obtained by standardizing the hydrochemical indexes of multiple water source samples, and P is the principal component score matrix.

5. The intelligent discrimination method for mine water inrush water source according to claim 1, characterized in that, In the S23, input the data set after dimension reduction into a preset model for training; optimize the number and depth of the subtrees of the random forest model through an intelligent swarm optimization algorithm to obtain a water source discrimination model.

6. An intelligent discriminator for the water inrush source in a mine, characterized in that, The device can implement the intelligent discrimination method for the water inrush source of the mine as described in claim 1, including: An acquisition module, configured to acquire the hydrochemical indexes of the water source sample to be discriminated; the hydrochemical indexes include inorganic indexes and organic indexes; the inorganic indexes include anions, cations and TDS value in the water; the organic indexes include total organic carbon, ultraviolet absorbance and fluorescence intensity of the fluorescence components of dissolved organic matter in the water sample; A discrimination module, configured to input the hydrochemical indexes into a preset water source discrimination model to obtain the water source type of the water source sample to be discriminated; the construction of the water source discrimination model includes: constructing a data set of the hydrochemical indexes of multiple water source samples according to the hydrochemical indexes of multiple water source samples; reducing the dimension of the data set to obtain a data set after dimension reduction; training a preset model according to the data set after dimension reduction to obtain a water source discrimination model.

7. The intelligent discriminator for mine water inrush source according to claim 6, characterized in that In the discrimination module, the data set of the hydrochemical indexes of the water source sample is represented by the following formula (1): Wherein, X is the data set, x is one of the hydrochemical indexes, n is the number of water source samples, and p is the number of hydrochemical indexes.

8. The intelligent discrimination device for mine water inrush source according to claim 6, wherein In the discrimination module, dimensionality reduction is performed on the data set to obtain the data set after dimensionality reduction, including: performing zero-mean normalization processing on the data set to obtain a standardized data matrix; performing principal component extraction on the standardized data matrix to obtain a principal component score matrix; performing dimensionality reduction on the standardized data matrix through the principal component score matrix to obtain the data set after dimensionality reduction; the zero-mean normalization processing of the data set is implemented by the following formula (3): Among them, μ A is the mean of a hydrochemical index A, σ A is the standard deviation of a hydrochemical index A, and v is the value after normalization of a certain value x of a hydrochemical index A; The data set after dimension reduction is represented by formula (4): Y = XP (4) Wherein, Y is the data set after dimension reduction, X is the data set obtained by standardizing the hydrochemical indexes of multiple water source samples, and P is the principal component score matrix; Input the data set after dimension reduction into a preset model for training; optimize the number and depth of the subtrees of the random forest model through an intelligent swarm optimization algorithm to obtain a water source discrimination model.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent discrimination method for mine water inrush sources according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program instructions, and the program instructions can be executed by the processor to implement the steps of the intelligent discrimination method for mine water inrush sources according to any one of claims 1 to 5.

Images