A method and device for short-term early warning of rock failure based on acoustic emission clustering analysis

Abstract

This invention provides a method and device for short-term early warning of rock failure based on acoustic emission clustering analysis, relating to the fields of rock mechanics and geotechnical engineering. The method includes: acquiring acoustic emission characteristic parameters during rock deformation and failure; performing clustering analysis on the acoustic emission characteristic parameters using the K-means++ algorithm to obtain cluster labels corresponding to the acoustic emission characteristic parameters; calculating the importance scores of the acoustic emission characteristic parameters based on the cluster labels using the random forest algorithm; constructing an early warning index set based on the acoustic emission characteristic parameters; constructing an initial CNN-LSTM model; establishing a sample dataset of conventional and precursor signals of rock failure based on the early warning index set; training the initial CNN-LSTM model using the sample dataset to obtain a trained CNN-LSTM model; acquiring real-time acoustic emission signals; and inputting the real-time acoustic emission signals into the trained CNN-LSTM model to provide early warning by identifying the acoustic emission signal category. This invention can provide early warning for complex rock failure processes.

Description

Technical Field

[0001] This invention relates to the fields of rock mechanics and geotechnical engineering, and in particular to a method and device for short-term early warning of rock failure based on acoustic emission clustering analysis. Background Technology

[0002] Mineral resources play an irreplaceable role in people's lives and social development. As mineral resource extraction progresses, engineering activities disrupt the original stress balance of rock masses, easily leading to stress concentration, inducing the initiation and propagation of rock cracks, and potentially causing rock fragmentation, tunnel spalling, and roof collapses. Especially in deep, high-stress environments, rockburst accidents in deep hard rock pose a serious threat to underground production safety. Therefore, research on precursor signals and early warning technologies for rock failure is of great significance for ensuring safe mine production.

[0003] Currently, acoustic emission (AE) technology is widely used to study rock failure. Before rock failure, AE signals increase significantly. AE ring counts, energy, and event rates all begin to increase rapidly during the rock crack propagation stage. At this stage, AE signals reflect the local release of strain energy stored within the rock, but they cannot accurately identify early warning points of macroscopic rock failure. Furthermore, using an abnormal increase in a single indicator as a warning is easily interfered with by other abnormal AE signals in the field environment. AE signal processing occurs after rock failure, making it difficult to determine the overall trend of AE signals during actual failure data collection. Therefore, there is a lack of effective nonparametric statistical methods and machine learning algorithms to analyze AE signal sequences and extract discriminative feature information. Existing research proposes using relevant mathematical analysis theories and methods to assign weights to early warning indicators to achieve comprehensive early warning of rock failure based on multiple indicators. However, the weight determination process involves a degree of subjectivity, leading to low accuracy in early warning results. Existing methods determine whether rock failure is imminent based on the changing trends of AE parameters. Before rock failure, the peak values ​​of AE parameters are unclear, and the magnitude of data growth is difficult to measure, making real-time and accurate identification impossible. Rock is a heterogeneous material, and its failure is inherently complex. Even rocks of the same lithology can exhibit vastly different macroscopic and microscopic manifestations of their failure processes. Therefore, existing models struggle to accurately describe the physical mechanisms underlying the complex failure processes of rocks and to provide timely and effective early warnings. Summary of the Invention

[0004] To address the limitations of existing technologies in accurately describing the physical mechanisms of complex rock failure processes and providing timely and effective early warnings, as well as the susceptibility of existing technologies to interference from abnormal signals, leading to low accuracy in early warning results, this invention provides a method and apparatus for short-term early warning of rock failure based on acoustic emission clustering analysis. The technical solution is as follows:

[0005] On the one hand, a short-term early warning method for rock damage based on acoustic emission clustering analysis is provided. This method is implemented by a short-term early warning device for rock damage based on acoustic emission clustering analysis, and includes:

[0006] S1. Obtain acoustic emission characteristic parameters during rock deformation and failure;

[0007] S2. The Kmeans++ algorithm is used to perform cluster analysis on the acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters;

[0008] S3. Based on the clustering labels, use the random forest algorithm to calculate the importance scores of the acoustic emission feature parameters; extract the acoustic emission feature parameters with an importance score greater than 0.1 to construct a set of early warning indicators;

[0009] S4. Construct the initial CNN-LSTM dual-branch feature fusion model;

[0010] S5. Based on the set of early warning indicators, establish a sample dataset of conventional rock damage signals and precursor signals; train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM dual-branch feature fusion model.

[0011] S6. Acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition, obtain the recognition results, and use the acoustic emission signals with a recognition result of 1 as early warning signals for rock damage.

[0012] Optionally, the acoustic emission characteristic parameters during the rock deformation and failure process include: acoustic emission event rate, number of events, ring count, cumulative ring count, energy, cumulative energy, rise time, duration, average rise time, average duration, and average frequency.

[0013] Optionally, step S2 employs the Kmeans++ algorithm to perform cluster analysis on the acoustic emission feature parameters to obtain cluster labels corresponding to the acoustic emission feature parameters, including:

[0014] S21. Standardize the acoustic emission characteristic parameters to obtain standardized acoustic emission characteristic parameters;

[0015] The standardization process is represented by the following formula (1):

[0016] (1)

[0017] Where, x i Represents the original value; X min X represents the minimum value among the characteristic parameters X;max The maximum value in the feature parameter X is represented by min; the lower bound of the scaled data range is represented by min; the upper bound of the scaled data range is represented by max; where min takes the value of -1 and max takes the value of 1.

[0018] S22. Calculate the sum of squared errors for clustering acoustic emission parameters corresponding to different K values, and determine the optimal number of clusters;

[0019] The process of determining the optimal number of clusters is represented by the following formula (2):

[0020] (2)

[0021] Where SSE represents the sum of squared errors; K represents the number of acoustic emission parameter categories; C i Let x represent the set of data points of class n; let x represent a data point in the set; let μ represent the acoustic emission characteristic parameters per second. i Represents set C i The center;

[0022] S23. Using the optimal number of clusters as the target number of categories for the Kmeans++ algorithm, perform cluster analysis on the standardized acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

[0023] Optionally, step S3 calculates the importance score of the acoustic emission feature parameters using a random forest algorithm based on the clustering labels, including:

[0024] S31. Based on the acoustic emission feature parameters, use the random forest algorithm to calculate the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree;

[0025] S32. Calculate the importance score of the acoustic emission feature parameters corresponding to the entire decision tree based on the importance contribution of each node's acoustic emission feature parameters.

[0026] S33. Based on the importance scores of the acoustic emission feature parameters corresponding to the entire decision tree, calculate the final importance scores of the acoustic emission feature parameters corresponding to all trees in the random forest.

[0027] Optionally, the process of calculating the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree is represented by the following formula (3):

[0028] (3)

[0029] in, I(parent) represents the importance contribution of acoustic emission feature parameter j in decision tree node n; I(parent) represents the impurity of the parent node; I(left child) represents the impurity of the left child node; I(right child) represents the impurity of the right child node.

[0030] The process of calculating the importance score of the acoustic emission feature parameters corresponding to the entire decision tree is represented by the following formula (4):

[0031] (4)

[0032] in, Let represent the importance score of acoustic emission feature parameter j in the entire decision tree; N represents the set of nodes in decision tree T;

[0033] The process of calculating the final importance score of the acoustic emission feature parameters corresponding to all trees in the random forest is expressed by the following formula (5):

[0034] (5)

[0035] In this random forest, all trees are represented as {T1, T2, ..., T}. m}; represents the average importance score of all trees in the random forest for the acoustic emission feature parameter j; m represents the number of decision trees in the random forest, which gradually increases during the calculation until the importance ranking of the feature parameter remains basically unchanged.

[0036] Optionally, the initial CNN-LSTM dual-branch feature fusion model of S4 includes: a CNN convolutional neural network and an LSTM bidirectional long short-term memory network, a feature concatenation layer, a fully connected layer, and an output layer;

[0037] The CNN convolutional neural network includes: convolutional layers, batch normalization layers, max pooling layers, and flattening layers;

[0038] The convolutional layer has 64 filters, a kernel size of 3, and a ReLU activation function; the max pooling layer has a pooling window size of 2; and the LSTM bidirectional long short-term memory network consists of a single bidirectional LSTM layer with 64 units.

[0039] The CNN convolutional neural network is used to extract local features of the feature matrix.

[0040] The LSTM bidirectional long short-term memory network is used to extract global temporal features of the feature matrix.

[0041] Optionally, step S5 involves training the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM dual-branch feature fusion model, including:

[0042] S51. Input the sample dataset into the initial CNN-LSTM dual-branch feature fusion model, and perform parallel processing through the CNN convolutional neural network and the LSTM bidirectional long short-term memory network. Local feature extraction is performed through the convolutional layer to obtain local feature representation; temporal feature processing is performed through the LSTM bidirectional long short-term memory network to obtain temporal feature representation.

[0043] S52. Input the local features into the batch normalization layer for processing to obtain the batch normalized feature representation; input the batch normalized feature representation into the max pooling layer, and process it through the flattening layer to obtain the feature representation of a one-dimensional vector;

[0044] S53. Input the feature representation of the one-dimensional vector and the temporal feature representation into the feature concatenation layer for feature concatenation to obtain the comprehensive feature representation;

[0045] S54. Input the comprehensive feature representation into the fully connected layer, and perform nonlinear transformation on the comprehensive feature representation through the ReLU activation function to obtain the final feature representation; input the final feature representation into the output layer, and use the Sigmoid activation function to map the final feature to binary classification probability; based on the binary classification probability, use the binary cross-entropy loss function for training to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0046] On the other hand, a short-term rock failure early warning device based on acoustic emission clustering analysis is provided. This device is applied to a short-term rock failure early warning method based on acoustic emission clustering analysis. The device includes:

[0047] The acquisition unit is used to acquire acoustic emission characteristic parameters during rock deformation and failure.

[0048] The clustering analysis unit is used to perform clustering analysis on the acoustic emission feature parameters using the Kmeans++ algorithm to obtain the clustering labels corresponding to the acoustic emission feature parameters;

[0049] The first construction unit is used to calculate the importance score of the acoustic emission feature parameters according to the clustering labels using the random forest algorithm; extract the acoustic emission feature parameters with an importance score greater than 0.1, and construct a set of early warning indicators.

[0050] The second building unit is used to build the initial CNN-LSTM dual-branch feature fusion model;

[0051] The training unit is used to establish a sample dataset of conventional and precursor signals of rock damage based on the set of early warning indicators; and to train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0052] The early warning unit is used to acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition, obtain the recognition result, and use the acoustic emission signal with a recognition result of 1 as the early warning signal of rock damage.

[0053] Optionally, the acoustic emission characteristic parameters during the rock deformation and failure process include: acoustic emission event rate, number of events, ring count, cumulative ring count, energy, cumulative energy, rise time, duration, average rise time, average duration, and average frequency.

[0054] Optionally, the clustering analysis unit is used for:

[0055] The acoustic emission characteristic parameters are standardized to obtain standardized acoustic emission characteristic parameters;

[0056] The standardization process is represented by the following formula (1):

[0057] (1)

[0058] Where, x i Represents the original value; X min X represents the minimum value among the characteristic parameters X; max The maximum value in the feature parameter X is represented by min; the lower bound of the scaled data range is represented by min; the upper bound of the scaled data range is represented by max; where min takes the value of -1 and max takes the value of 1.

[0059] Calculate the sum of squared errors for clustering acoustic emission parameters corresponding to different K values, and determine the optimal number of clusters;

[0060] The process of determining the optimal number of clusters is represented by the following formula (2):

[0061] (2)

[0062] Where SSE represents the sum of squared errors; K represents the number of acoustic emission parameter categories; C i Let x represent the set of data points of class n; let x represent a data point in the set; let μ represent the acoustic emission characteristic parameters per second. i Represents set C i The center;

[0063] The optimal number of clusters is used as the target number of categories in the Kmeans++ algorithm. Cluster analysis is then performed on the standardized acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

[0064] Optionally, the first building unit is used for:

[0065] Based on the acoustic emission feature parameters, the random forest algorithm is used to calculate the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree;

[0066] Based on the importance contribution of the acoustic emission feature parameters of each node, calculate the importance score of the acoustic emission feature parameters of the entire decision tree;

[0067] Based on the importance scores of the acoustic emission feature parameters corresponding to the entire decision tree, the final importance scores of the acoustic emission feature parameters corresponding to all trees in the random forest are calculated.

[0068] Optionally, the process of calculating the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree is represented by the following formula (3):

[0069] (3)

[0070] in, I(parent) represents the importance contribution of acoustic emission feature parameter j in decision tree node n; I(parent) represents the impurity of the parent node; I(left child) represents the impurity of the left child node; I(right child) represents the impurity of the right child node.

[0071] The process of calculating the importance score of the acoustic emission feature parameters corresponding to the entire decision tree is represented by the following formula (4):

[0072] (4)

[0073] in, Let represent the importance score of acoustic emission feature parameter j in the entire decision tree; N represents the set of nodes in decision tree T;

[0074] The process of calculating the final importance score of the acoustic emission feature parameters corresponding to all trees in the random forest is expressed by the following formula (5):

[0075] (5)

[0076] In this random forest, all trees are represented as {T1, T2, ..., T}. m}; represents the average importance score of all trees in the random forest for the acoustic emission feature parameter j; m represents the number of decision trees in the random forest, which gradually increases during the calculation until the importance ranking of the feature parameter remains basically unchanged.

[0077] Optionally, the initial CNN-LSTM dual-branch feature fusion model includes: a CNN convolutional neural network and an LSTM bidirectional long short-term memory network, a feature concatenation layer, a fully connected layer, and an output layer;

[0078] The CNN convolutional neural network includes: convolutional layers, batch normalization layers, max pooling layers, and flattening layers;

[0079] The convolutional layer has 64 filters, a kernel size of 3, and a ReLU activation function; the max pooling layer has a pooling window size of 2; and the LSTM bidirectional long short-term memory network consists of a single bidirectional LSTM layer with 64 units.

[0080] The CNN convolutional neural network is used to extract local features of the feature matrix.

[0081] The LSTM bidirectional long short-term memory network is used to extract global temporal features of the feature matrix.

[0082] Optionally, the training unit is used for:

[0083] The sample dataset is input into the initial CNN-LSTM dual-branch feature fusion model, and parallel processing is performed through the CNN convolutional neural network and the LSTM bidirectional long short-term memory network. Local feature extraction is performed through the convolutional layer to obtain local feature representation; temporal feature processing is performed through the LSTM bidirectional long short-term memory network to obtain temporal feature representation.

[0084] Local features are input into a batch normalization layer for processing to obtain a batch normalized feature representation; the batch normalized feature representation is then input into a max pooling layer and processed by a flattening layer to obtain a one-dimensional vector feature representation.

[0085] The feature representations of one-dimensional vectors and temporal features are input into the feature concatenation layer for feature concatenation to obtain a comprehensive feature representation.

[0086] The comprehensive feature representation is input into the fully connected layer, and the ReLU activation function is used to perform nonlinear transformation on the comprehensive feature representation to obtain the final feature representation. The final feature representation is then input into the output layer, and the Sigmoid activation function is used to map the final feature to binary classification probabilities. Based on the binary classification probabilities, the binary cross-entropy loss function is used for training to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0087] On the other hand, a short-term rock damage early warning device based on acoustic emission clustering analysis is provided. The short-term rock damage early warning device based on acoustic emission clustering analysis includes: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, any one of the methods described above for short-term rock damage early warning based on acoustic emission clustering analysis is implemented.

[0088] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described methods for short-term early warning of rock damage based on acoustic emission clustering analysis.

[0089] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0090] This invention first acquires acoustic emission characteristic parameters during rock deformation and failure; then, it uses the K-means++ algorithm to perform cluster analysis on the acoustic emission characteristic parameters to obtain cluster labels corresponding to the acoustic emission characteristic parameters; based on the cluster labels, it uses the random forest algorithm to calculate the importance scores of the acoustic emission characteristic parameters; it extracts acoustic emission characteristic parameters with importance scores greater than 0.1 to construct a set of early warning indicators; secondly, it constructs an initial CNN-LSTM two-branch feature fusion model; based on the set of early warning indicators, it establishes a sample dataset of conventional and precursor signals of rock failure; it trains the initial CNN-LSTM two-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM two-branch feature fusion model; finally, it acquires real-time acoustic emission signals; it inputs the real-time acoustic emission signals into the trained CNN-LSTM two-branch feature fusion model for recognition, obtains the recognition results, and uses acoustic emission signals with a recognition result of 1 as early warning signals for rock failure.

[0091] This invention uses multiple acoustic emission parameters to predict rock failure and can automatically identify acoustic emission precursor signals of rock failure for early warning. By implementing this invention, the physical mechanisms of complex rock failure processes can be accurately described and timely and effective early warnings can be provided. Attached Figure Description

[0092] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0093] Figure 1 This is a flowchart of a short-term early warning method for rock damage based on acoustic emission clustering analysis provided in an embodiment of the present invention;

[0094] Figure 2 This is a schematic diagram of the stress-strain curve and acoustic emission signal clustering results of rock failure under uniaxial loading provided in an embodiment of the present invention;

[0095] Figure 3 This is a schematic diagram of an embodiment of the present invention for calculating the sum of squared errors of different cluster numbers K and using the elbow method to find the optimal cluster number K;

[0096] Figure 4 This is a schematic diagram provided by an embodiment of the present invention, which uses the random forest algorithm to calculate the contribution of each acoustic emission feature parameter to the signal category and obtain the importance score of each parameter.

[0097] Figure 5 This is a schematic diagram illustrating a process of obtaining signal category labels by performing cluster analysis using the Kmeans++ algorithm according to an embodiment of the present invention;

[0098] Figure 6 This is a schematic diagram of the structure of a CNN-LSTM dual-branch feature fusion model provided in an embodiment of the present invention;

[0099] Figure 7 This is a schematic diagram of an early warning system provided by an embodiment of the present invention, which uses a trained CNN-LSTM dual-branch feature fusion model to identify acoustic emission signals throughout the entire process of rock deformation and failure.

[0100] Figure 8 This is a block diagram of a short-term early warning device for rock damage based on acoustic emission clustering analysis provided in an embodiment of the present invention;

[0101] Figure 9 This is a schematic diagram of the structure of a short-term early warning device for rock damage based on acoustic emission clustering analysis provided in an embodiment of the present invention. Detailed Implementation

[0102] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0103] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0104] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0105] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0106] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0107] This invention provides a short-term rock damage early warning method based on acoustic emission clustering analysis. This method can be implemented by a short-term rock damage early warning device based on acoustic emission clustering analysis, which can be a terminal or a server. Figure 1 The flowchart shown is for a short-term early warning method for rock damage based on acoustic emission clustering analysis. The processing flow of this method may include the following steps:

[0108] S1. Obtain acoustic emission characteristic parameters during rock deformation and failure.

[0109] In one feasible implementation, a single-axis compression test is conducted on a limestone sample with a specification of Φ50×100mm under displacement control mode and a loading rate of 0.002mm / s. An acoustic emission sensor is used to monitor the deformation and failure process in real time, and the acoustic emission signal is recorded at a sampling rate of 1MHz to obtain the acoustic emission characteristic parameters every 1 second.

[0110] Optionally, acoustic emission characteristic parameters during rock deformation and failure include: acoustic emission event rate, number of events, ring count, cumulative ring count, energy, cumulative energy, rise time, duration, average rise time, average duration, and average frequency.

[0111] The acquired acoustic emission feature parameters are constructed into a matrix, with each row corresponding to the acoustic emission feature parameters for each second. The K-means++ algorithm is used to perform cluster analysis on the acoustic emission feature parameters collected throughout the entire limestone failure process. Each row of the matrix returns a signal category, thus obtaining the acoustic emission signal category for each second. Figure 2 The figure shown is a schematic diagram of the stress-strain curve and acoustic emission signal clustering results of rock failure under uniaxial loading provided by an embodiment of the present invention.

[0112] In the cluster analysis process, it is necessary to determine the optimal cluster value K. The initial K value range is set to (1, 10). By calculating the sum of squared errors corresponding to different K values, a curve showing the change of the sum of squared errors with the K value is plotted. When the sum of squared errors does not decrease significantly with the increase of the K value, the corresponding K value is the optimal cluster. Figure 3 The diagram shown is a schematic diagram of calculating the sum of squared errors of different cluster numbers K values ​​and using the elbow method to find the optimal cluster number K value according to an embodiment of the present invention; the K value of the limestone sample in the embodiment of the present invention is calculated to be 4.

[0113] S2. The Kmeans++ algorithm is used to perform cluster analysis on the acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

[0114] Optionally, the specific implementation process of S2 includes S21-S23:

[0115] S21. Standardize the acoustic emission characteristic parameters to obtain the standardized acoustic emission characteristic parameters;

[0116] The standardization process is represented by the following formula (1):

[0117] (1)

[0118] Where, x i Represents the original value; X min X represents the minimum value among the characteristic parameters X; max The maximum value in the feature parameter X is represented by min; the lower bound of the scaled data range is represented by min; the upper bound of the scaled data range is represented by max; where min takes the value of -1 and max takes the value of 1.

[0119] S22. Calculate the sum of squared errors for clustering acoustic emission parameters corresponding to different K values, and determine the optimal number of clusters;

[0120] The process of determining the optimal number of clusters is represented by the following formula (2):

[0121] (2)

[0122] Where SSE represents the sum of squared errors; K represents the number of acoustic emission parameter categories; C i Let x represent the set of data points of class n; let x represent a data point in the set; let μ represent the acoustic emission characteristic parameters per second. i Represents set C i The center;

[0123] When K increases, the number of data points in each category decreases, thus reducing the distance between the data points in the set and the set center, leading to a decrease in the sum of squared errors. When the sum of squared errors does not decrease significantly with increasing K, the corresponding K value is the optimal number of categories for acoustic emission feature parameter clustering analysis.

[0124] S23. Using the optimal number of clusters as the target number of categories for the Kmeans++ algorithm, perform cluster analysis on the standardized acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

[0125] Among them, such as Figure 4 This is a schematic diagram illustrating how the random forest algorithm is used to calculate the contribution of each acoustic emission characteristic parameter to the signal category and obtain the importance score of each parameter, according to an embodiment of the present invention.

[0126] S3. Based on the clustering labels, use the random forest algorithm to calculate the importance scores of the acoustic emission feature parameters; extract the acoustic emission feature parameters with an importance score greater than 0.1 to construct a set of early warning indicators.

[0127] Optionally, S3 uses a random forest algorithm based on clustering labels to calculate importance scores for acoustic emission feature parameters, including:

[0128] The random forest algorithm takes the acquired acoustic emission feature parameters as input and outputs the clustering labels corresponding to the acoustic emission feature parameters acquired every second.

[0129] S31. Based on the acoustic emission feature parameters, use the random forest algorithm to calculate the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree;

[0130] Among them, the importance score of each acoustic emission feature parameter is the cumulative impurity reduction of that parameter across all nodes of all decision trees in the random forest.

[0131] For example, a decision tree T has a set of N nodes. For each node n∈N, it is assumed that node n is split using acoustic emission feature parameter j to obtain the importance contribution of the acoustic emission feature parameter corresponding to each node.

[0132] Optionally, the process of calculating the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree is expressed by the following formula (3):

[0133] (3)

[0134] in, I(parent) represents the importance contribution of acoustic emission feature parameter j in decision tree node n; I(parent) represents the impurity of the parent node; I(left child) represents the impurity of the left child node; I(right child) represents the impurity of the right child node.

[0135] S32. Calculate the importance score of the acoustic emission feature parameters corresponding to the entire decision tree based on the importance contribution of each node's acoustic emission feature parameters.

[0136] The process of calculating the importance score of the acoustic emission feature parameters corresponding to the entire decision tree is represented by the following formula (4):

[0137] (4)

[0138] in, Let represent the importance score of acoustic emission feature parameter j in the entire decision tree; N represents the set of nodes in decision tree T;

[0139] S33. Based on the importance scores of the acoustic emission feature parameters corresponding to the entire decision tree, calculate the final importance scores of the acoustic emission feature parameters corresponding to all trees in the random forest.

[0140] The process of calculating the final importance score of the acoustic emission feature parameters corresponding to all trees in the random forest is expressed by the following formula (5):

[0141] (5)

[0142] In this random forest, all trees are represented as {T1, T2, ..., T}. m}; denoted as the average importance score of all trees in the random forest for the acoustic emission feature parameter j; m represents the number of decision trees in the random forest, which gradually increases during the calculation until the importance ranking of the feature parameter remains basically unchanged.

[0143] S4. Construct the initial CNN-LSTM dual-branch feature fusion model.

[0144] Among them, such as Figure 6 The diagram shown is a structural schematic of a CNN-LSTM dual-branch feature fusion model provided in an embodiment of the present invention.

[0145] Optionally, the initial CNN-LSTM dual-branch feature fusion model of S4 includes: a CNN convolutional neural network and an LSTM bidirectional long short-term memory network, a feature concatenation layer, a fully connected layer, and an output layer;

[0146] The CNN convolutional neural network includes: convolutional layers, batch normalization layers, max pooling layers, and flattening layers;

[0147] The convolutional layer has 64 filters, a kernel size of 3, and the activation function is ReLU; the max pooling layer has a pooling window size of 2; the LSTM bidirectional long short-term memory network consists of a single bidirectional LSTM layer with 64 units.

[0148] Among them, the CNN convolutional neural network is used to extract local features of the feature matrix;

[0149] Among them, the LSTM bidirectional long short-term memory network is used to extract global temporal features of the feature matrix.

[0150] The fully connected layer is a fully connected layer containing 128 neurons.

[0151] The output layer contains one neuron.

[0152] S5. Based on the set of early warning indicators, establish a sample dataset of conventional and precursor signals of rock damage; train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0153] Among them, such as Figure 5 The diagram shown is a schematic representation of a process for obtaining signal category labels using the Kmeans++ algorithm for cluster analysis, as provided in an embodiment of the present invention.

[0154] In one feasible implementation, the process of establishing a sample dataset of conventional and precursor signals of rock failure based on a set of early warning indicators includes:

[0155] (1) The set of early warning indicators is used as the feature matrix. The Kmeans++ algorithm is used to re-cluster the feature matrix to obtain the clustering results. The classification labels corresponding to each row of the feature matrix in the clustering results are saved to form a label column.

[0156] (2) The peak stress point of the stress-strain curve is taken as the rock failure point. One type of signal near the failure point is taken as the rock failure precursor signal and marked as 1. The other signals are all normal signals and marked as 0. The results are saved in the feature matrix and marked as the label column to form a sample dataset of normal signals and precursor signals of rock failure process.

[0157] The established sample datasets of conventional and precursor signals of rock damage were divided into training, validation, and test sets. The training set accounted for 0.6% of the total samples, the validation set accounted for 0.2%, and the test set accounted for 0.2%. The CNN-LSTM dual-branch feature fusion model was trained using the training set. When the loss rate of the validation set did not decrease after 10 iterations, the model parameters with the lowest loss on the validation set were returned and the optimal model was saved. The performance of the CNN-LSTM dual-branch feature fusion model on the test set was then observed.

[0158] Optionally, S5, the process of training the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain the trained CNN-LSTM dual-branch feature fusion model, includes S51-S54:

[0159] S51. Input the sample dataset into the initial CNN-LSTM dual-branch feature fusion model, and process it in parallel through the CNN convolutional neural network and the LSTM bidirectional long short-term memory network. Local feature extraction is performed through the convolutional layer to obtain local feature representation; temporal feature processing is performed through the LSTM bidirectional long short-term memory network to obtain temporal feature representation.

[0160] S52. Input the local features into the batch normalization layer for processing to obtain the batch normalized feature representation; input the batch normalized feature representation into the max pooling layer, and process it through the flattening layer to obtain the feature representation of a one-dimensional vector;

[0161] S53. Input the feature representation of the one-dimensional vector and the temporal feature representation into the feature concatenation layer for feature concatenation to obtain the comprehensive feature representation;

[0162] The feature concatenation layer is used to concatenate and fuse the features extracted by the CNN branch and the LSTM branch. This method not only preserves the independence of the two types of features and avoids information loss caused by early fusion, but also provides a richer joint representation for subsequent processing.

[0163] S54. Input the comprehensive feature representation into the fully connected layer, and perform nonlinear transformation on the comprehensive feature representation through the ReLU activation function to obtain the final feature representation; input the final feature representation into the output layer, and use the Sigmoid activation function to map the final feature to binary classification probability; based on the binary classification probability, use the binary cross-entropy loss function for training to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0164] Among them, the binary classification output corresponds to the rock failure state represented by the acoustic emission signal parameters.

[0165] In this method, a Dropout layer with a ratio of 0.5 is added after the fully connected layer to randomly drop 50% of the neuron outputs during the training phase, thereby reducing the risk of overfitting during model training.

[0166] The fully connected layer has 128 neurons and uses ReLU as the activation function, which can perform nonlinear transformation and dimensionality reduction on the concatenated high-order features.

[0167] In one feasible implementation, a binary cross-entropy loss function is used, the gradient descent employs the Adam optimizer, and the number of iterations is set to 50. During training, the learning rate is dynamically adjusted and an early stopping mechanism is set to prevent overfitting. In this embodiment of the invention, the validation set loss is used as the monitoring metric, and the learning rate adjustment is expressed by the following formula (6) when the validation set loss does not decrease during 5 consecutive training iterations:

[0168] (6)

[0169] in, This is the current learning rate; is the learning rate of the previous stage; f is the learning rate decay coefficient, which is set to 0.2 in this embodiment of the invention.

[0170] The early stopping counter threshold is set to 10. If the cumulative loss on the validation set does not decrease during 10 training iterations, the early stopping mechanism is triggered, the model stops training, and the model weights with the lowest validation set loss during the entire training process are returned.

[0171] S6. Acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition, obtain the recognition results, and use the acoustic emission signals with a recognition result of 1 as early warning signals for rock damage.

[0172] In one feasible implementation, the real-time acoustic emission signal is input into a trained CNN-LSTM dual-branch feature fusion model to obtain the prediction probability. When the prediction probability is <0.5, the corresponding acoustic emission signal category is 0, which is a normal signal of rock fracture, indicating that the rock has not entered the stage of accelerated damage, and monitoring and analysis continue. When the prediction probability is ≥0.5, the corresponding acoustic emission signal category is 1, which is a precursor signal of rock damage, indicating that the rock is close to damage, and an early warning is issued immediately.

[0173] Among them, such as Figure 7The diagram shown is an early warning illustration of identifying acoustic emission signals throughout the rock deformation and failure process using a trained CNN-LSTM dual-branch feature fusion model, as provided in an embodiment of the present invention. The rock failure precursor signal, as determined by the clustering results, appears around 214s, while the model predicts the danger signal at 218s, and the rock failure time point is 258s. The model's early warning time lags behind the precursor signal by 4s, but advances the rock failure time by nearly 40s. The method provided by the present invention achieves early warning and forecasting of rock failure.

[0174] This invention first acquires acoustic emission characteristic parameters during rock deformation and failure; then, it uses the K-means++ algorithm to perform cluster analysis on the acoustic emission characteristic parameters to obtain cluster labels corresponding to the acoustic emission characteristic parameters; based on the cluster labels, it uses the random forest algorithm to calculate the importance scores of the acoustic emission characteristic parameters; it extracts acoustic emission characteristic parameters with importance scores greater than 0.1 to construct a set of early warning indicators; secondly, it constructs an initial CNN-LSTM two-branch feature fusion model; based on the set of early warning indicators, it establishes a sample dataset of conventional and precursor signals of rock failure; it trains the initial CNN-LSTM two-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM two-branch feature fusion model; finally, it acquires real-time acoustic emission signals; it inputs the real-time acoustic emission signals into the trained CNN-LSTM two-branch feature fusion model for recognition, obtains the recognition results, and uses acoustic emission signals with a recognition result of 1 as early warning signals for rock failure.

[0175] This invention uses multiple acoustic emission parameters to predict rock failure and can automatically identify acoustic emission precursor signals of rock failure for early warning. By implementing this invention, the physical mechanisms of complex rock failure processes can be accurately described and timely and effective early warnings can be provided.

[0176] Figure 8 This is a block diagram illustrating a short-term rock failure early warning device based on acoustic emission clustering analysis, according to an exemplary embodiment. The device is used in a short-term rock failure early warning method based on acoustic emission clustering analysis. (Refer to...) Figure 8 The device includes an acquisition unit 810, a clustering analysis unit 820, a first construction unit 830, a second construction unit 840, a training unit 850, and an early warning unit 860. Wherein:

[0177] Acquisition unit 810 is used to acquire acoustic emission characteristic parameters during rock deformation and failure.

[0178] Clustering analysis unit 820 is used to perform clustering analysis on the acoustic emission feature parameters using the Kmeans++ algorithm to obtain the clustering labels corresponding to the acoustic emission feature parameters;

[0179] The first construction unit 830 is used to calculate the importance score of the acoustic emission feature parameters according to the clustering labels using the random forest algorithm; extract the acoustic emission feature parameters with an importance score greater than 0.1, and construct a set of early warning indicators.

[0180] The second building unit 840 is used to build the initial CNN-LSTM dual-branch feature fusion model;

[0181] Training unit 850 is used to establish a sample dataset of conventional rock damage signals and precursor signals based on the set of early warning indicators; and to train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM dual-branch feature fusion model.

[0182] The early warning unit 860 is used to acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition, obtain the recognition result, and use the acoustic emission signal with a recognition result of 1 as the early warning signal for rock damage.

[0183] Optionally, the acoustic emission characteristic parameters during the rock deformation and failure process include: acoustic emission event rate, number of events, ring count, cumulative ring count, energy, cumulative energy, rise time, duration, average rise time, average duration, and average frequency.

[0184] Optionally, the clustering analysis unit 820 is used for:

[0185] The acoustic emission characteristic parameters are standardized to obtain standardized acoustic emission characteristic parameters;

[0186] The standardization process is represented by the following formula (1):

[0187] (1)

[0188] Where, x i Represents the original value; X min X represents the minimum value among the characteristic parameters X; max The maximum value in the feature parameter X is represented by min; the lower bound of the scaled data range is represented by min; the upper bound of the scaled data range is represented by max; where min takes the value of -1 and max takes the value of 1.

[0189] Calculate the sum of squared errors for clustering acoustic emission parameters corresponding to different K values, and determine the optimal number of clusters;

[0190] The process of determining the optimal number of clusters is represented by the following formula (2):

[0191] (2)

[0192] Where SSE represents the sum of squared errors; K represents the number of acoustic emission parameter categories; C i Let x represent the set of data points of class n; let x represent a data point in the set; let μ represent the acoustic emission characteristic parameters per second. i Represents set C i The center;

[0193] The optimal number of clusters is used as the target number of categories in the Kmeans++ algorithm. Cluster analysis is then performed on the standardized acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

[0194] Optionally, the first building unit 830 is used for:

[0195] Based on the acoustic emission feature parameters, the random forest algorithm is used to calculate the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree;

[0196] Based on the importance contribution of the acoustic emission feature parameters of each node, calculate the importance score of the acoustic emission feature parameters of the entire decision tree;

[0197] Based on the importance scores of the acoustic emission feature parameters corresponding to the entire decision tree, the final importance scores of the acoustic emission feature parameters corresponding to all trees in the random forest are calculated.

[0198] Optionally, the process of calculating the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree is represented by the following formula (3):

[0199] (3)

[0200] in, I(parent) represents the importance contribution of acoustic emission feature parameter j in decision tree node n; I(parent) represents the impurity of the parent node; I(left child) represents the impurity of the left child node; I(right child) represents the impurity of the right child node.

[0201] The process of calculating the importance score of the acoustic emission feature parameters corresponding to the entire decision tree is represented by the following formula (4):

[0202] (4)

[0203] in, Let represent the importance score of acoustic emission feature parameter j in the entire decision tree; N represents the set of nodes in decision tree T;

[0204] The process of calculating the final importance score of the acoustic emission feature parameters corresponding to all trees in the random forest is expressed by the following formula (5):

[0205] (5)

[0206] In this random forest, all trees are represented as {T1, T2, ..., T}. m}; denoted as the average importance score of all trees in the random forest for the acoustic emission feature parameter j; m represents the number of decision trees in the random forest, which gradually increases during the calculation until the importance ranking of the feature parameter remains basically unchanged.

[0207] Optionally, the initial CNN-LSTM dual-branch feature fusion model includes: a CNN convolutional neural network and an LSTM bidirectional long short-term memory network, a feature concatenation layer, a fully connected layer, and an output layer;

[0208] The CNN convolutional neural network includes: convolutional layers, batch normalization layers, max pooling layers, and flattening layers;

[0209] The convolutional layer has 64 filters, a kernel size of 3, and a ReLU activation function; the max pooling layer has a pooling window size of 2; and the LSTM bidirectional long short-term memory network consists of a single bidirectional LSTM layer with 64 units.

[0210] The CNN convolutional neural network is used to extract local features of the feature matrix.

[0211] The LSTM bidirectional long short-term memory network is used to extract global temporal features of the feature matrix.

[0212] Optionally, the training unit 850 is used for:

[0213] The sample dataset is input into the initial CNN-LSTM dual-branch feature fusion model, and parallel processing is performed through the CNN convolutional neural network and the LSTM bidirectional long short-term memory network. Local feature extraction is performed through the convolutional layer to obtain local feature representation; temporal feature processing is performed through the LSTM bidirectional long short-term memory network to obtain temporal feature representation.

[0214] Local features are input into a batch normalization layer for processing to obtain a batch normalized feature representation; the batch normalized feature representation is then input into a max pooling layer and processed by a flattening layer to obtain a one-dimensional vector feature representation.

[0215] The feature representations of one-dimensional vectors and temporal features are input into the feature concatenation layer for feature concatenation to obtain a comprehensive feature representation.

[0216] The comprehensive feature representation is input into the fully connected layer, and the ReLU activation function is used to perform nonlinear transformation on the comprehensive feature representation to obtain the final feature representation. The final feature representation is then input into the output layer, and the Sigmoid activation function is used to map the final feature to binary classification probabilities. Based on the binary classification probabilities, the binary cross-entropy loss function is used for training to obtain the trained CNN-LSTM dual-branch feature fusion model.

[0217] This invention first acquires acoustic emission characteristic parameters during rock deformation and failure; then, it uses the K-means++ algorithm to perform cluster analysis on the acoustic emission characteristic parameters to obtain cluster labels corresponding to the acoustic emission characteristic parameters; based on the cluster labels, it uses the random forest algorithm to calculate the importance scores of the acoustic emission characteristic parameters; it extracts acoustic emission characteristic parameters with importance scores greater than 0.1 to construct a set of early warning indicators; secondly, it constructs an initial CNN-LSTM two-branch feature fusion model; based on the set of early warning indicators, it establishes a sample dataset of conventional and precursor signals of rock failure; it trains the initial CNN-LSTM two-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM two-branch feature fusion model; finally, it acquires real-time acoustic emission signals; it inputs the real-time acoustic emission signals into the trained CNN-LSTM two-branch feature fusion model for recognition, obtains the recognition results, and uses acoustic emission signals with a recognition result of 1 as early warning signals for rock failure.

[0218] This invention uses multiple acoustic emission parameters to predict rock failure and can automatically identify acoustic emission precursor signals of rock failure for early warning. By implementing this invention, the physical mechanisms of complex rock failure processes can be accurately described and timely and effective early warnings can be provided.

[0219] Figure 9 This is a schematic diagram of a short-term early warning device for rock damage based on acoustic emission clustering analysis provided in an embodiment of the present invention, as shown below. Figure 9 As shown, a short-term early warning device for rock damage based on acoustic emission clustering analysis may include the above-mentioned... Figure 8 The illustrated rock damage short-term early warning device is based on acoustic emission clustering analysis. Optionally, the rock damage short-term early warning device 910 based on acoustic emission clustering analysis may include a first processor 2001.

[0220] Optionally, the rock damage short-term early warning device 910 based on acoustic emission clustering analysis may also include a memory 2002 and a transceiver 2003.

[0221] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0222] The following is combined with Figure 9A detailed introduction to each component of the 910 short-term rock damage early warning device based on acoustic emission clustering analysis is provided below:

[0223] The first processor 2001 is the control center of the rock damage short-term early warning device 910 based on acoustic emission clustering analysis. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0224] Optionally, the first processor 2001 can perform various functions of the rock damage short-term early warning device 910 based on acoustic emission clustering analysis by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0225] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 9 CPU0 and CPU1 are shown in the diagram.

[0226] In a specific implementation, as one example, the rock failure short-term early warning device 910 based on acoustic emission clustering analysis may also include multiple processors, for example... Figure 9 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0227] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0228] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the rock damage short-term early warning device 910 based on acoustic emission clustering analysis. Figure 9 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0229] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0230] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 9 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0231] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected to the interface circuit of the rock damage short-term early warning device 910 based on acoustic emission clustering analysis. Figure 9 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0232] It should be noted that, Figure 9 The structure of the short-term rock damage early warning device 910 based on acoustic emission clustering analysis shown in the figure does not constitute a limitation on the router. The actual knowledge structure identification device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0233] Furthermore, the technical effect of the short-term rock damage early warning device 910 based on acoustic emission clustering analysis can be referred to the technical effect of the short-term rock damage early warning method based on acoustic emission clustering analysis described in the above method embodiments, and will not be repeated here.

[0234] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0235] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0236] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0237] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0238] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0239] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0240] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0241] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0242] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0243] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0244] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0245] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0246] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A short-term early warning method for rock failure based on acoustic emission clustering analysis, characterized in that, The method includes: S1. Obtain acoustic emission characteristic parameters during rock deformation and failure; S2. The Kmeans++ algorithm is used to perform cluster analysis on the acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters; S3. Based on the clustering labels, use the random forest algorithm to calculate the importance scores of the acoustic emission feature parameters; extract the acoustic emission feature parameters with an importance score greater than 0.1 to construct a set of early warning indicators; S4. Construct the initial CNN-LSTM dual-branch feature fusion model; S5. Based on the set of early warning indicators, establish a sample dataset of conventional rock damage signals and precursor signals; train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM dual-branch feature fusion model. S6. Acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition and obtain recognition results; use acoustic emission signals with a recognition result of 1 as early warning signals for rock damage.

2. The short-term early warning method for rock damage based on acoustic emission clustering analysis according to claim 1, characterized in that, The acoustic emission characteristic parameters during the rock deformation and failure process include: acoustic emission event rate, number of events, ring count, cumulative ring count, energy, cumulative energy, rise time, duration, average rise time, average duration, and average frequency.

3. The short-term early warning method for rock damage based on acoustic emission clustering analysis according to claim 1, characterized in that, The S2 method uses the Kmeans++ algorithm to perform cluster analysis on the acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters, including: S21. Standardize the acoustic emission characteristic parameters to obtain standardized acoustic emission characteristic parameters; The standardization process is represented by the following formula (1): （1） Where, x i Represents the original value; X min X represents the minimum value among the characteristic parameters X; max The maximum value in the feature parameter X is represented by min; the lower bound of the scaled data range is represented by min; the upper bound of the scaled data range is represented by max; where min takes the value of -1 and max takes the value of 1. S22. Calculate the sum of squared errors for clustering acoustic emission parameters corresponding to different K values, and determine the optimal number of clusters; The process of determining the optimal number of clusters is represented by the following formula (2): （2） Where SSE represents the sum of squared errors; K represents the number of acoustic emission parameter categories; C i Let x represent the set of data points of class n; let x represent a data point in the set; let μ represent the acoustic emission characteristic parameters per second. i Represents set C i The center; S23. Using the optimal number of clusters as the target number of categories for the Kmeans++ algorithm, perform cluster analysis on the standardized acoustic emission feature parameters to obtain the cluster labels corresponding to the acoustic emission feature parameters.

4. The short-term early warning method for rock failure based on acoustic emission clustering analysis according to claim 1, characterized in that, S3 calculates the importance score of the acoustic emission feature parameters based on the clustering labels using a random forest algorithm, including: S31. Based on the acoustic emission feature parameters, use the random forest algorithm to calculate the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree; S32. Calculate the importance score of the acoustic emission feature parameters corresponding to the entire decision tree based on the importance contribution of each node's acoustic emission feature parameters. S33. Based on the importance scores of the acoustic emission feature parameters corresponding to the entire decision tree, calculate the final importance scores of the acoustic emission feature parameters corresponding to all trees in the random forest.

5. The short-term early warning method for rock damage based on acoustic emission clustering analysis according to claim 4, characterized in that, The process of calculating the importance contribution of the acoustic emission feature parameters corresponding to each node in each decision tree is expressed by the following formula (3): （3） in, I(parent) represents the importance contribution of acoustic emission feature parameter j in decision tree node n; I(parent) represents the impurity of the parent node; I(left child) represents the impurity of the left child node; I(right child) represents the impurity of the right child node. The process of calculating the importance score of the acoustic emission feature parameters corresponding to the entire decision tree is represented by the following formula (4): （4） in, Let represent the importance score of acoustic emission feature parameter j in the entire decision tree; N represents the set of nodes in decision tree T; The process of calculating the final importance score of the acoustic emission feature parameters corresponding to all trees in the random forest is expressed by the following formula (5): （5） In this random forest, all trees are represented as {T1, T2, ..., T}. m }; represents the average importance score of all trees in the random forest for the acoustic emission feature parameter j; m represents the number of decision trees in the random forest, which gradually increases during the calculation until the importance ranking of the feature parameter remains basically unchanged.

6. The short-term early warning method for rock damage based on acoustic emission clustering analysis according to claim 1, characterized in that, The initial CNN-LSTM dual-branch feature fusion model of S4 includes: a CNN convolutional neural network and an LSTM bidirectional long short-term memory network, a feature concatenation layer, a fully connected layer, and an output layer; The CNN convolutional neural network includes: convolutional layers, batch normalization layers, max pooling layers, and flattening layers; The convolutional layer has 64 filters, a kernel size of 3, and a ReLU activation function; the max pooling layer has a pooling window size of 2; and the LSTM bidirectional long short-term memory network consists of a single bidirectional LSTM layer with 64 units. The CNN convolutional neural network is used to extract local features of the feature matrix. The LSTM bidirectional long short-term memory network is used to extract global temporal features of the feature matrix.

7. The short-term early warning method for rock damage based on acoustic emission clustering analysis according to claim 1, characterized in that, Step S5 involves training the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain a trained CNN-LSTM dual-branch feature fusion model, including: S51. Input the sample dataset into the initial CNN-LSTM dual-branch feature fusion model, and perform parallel processing through the CNN convolutional neural network and the LSTM bidirectional long short-term memory network. Local feature extraction is performed through the convolutional layer to obtain local feature representation; temporal feature processing is performed through the LSTM bidirectional long short-term memory network to obtain temporal feature representation. S52. Input the local features into the batch normalization layer for processing to obtain the batch normalized feature representation; input the batch normalized feature representation into the max pooling layer, and process it through the flattening layer to obtain the feature representation of a one-dimensional vector; S53. Input the feature representation of the one-dimensional vector and the temporal feature representation into the feature concatenation layer for feature concatenation to obtain the comprehensive feature representation; S54. Input the comprehensive feature representation into the fully connected layer, and perform nonlinear transformation on the comprehensive feature representation through the ReLU activation function to obtain the final feature representation; input the final feature representation into the output layer, and use the Sigmoid activation function to map the final feature to binary classification probability; based on the binary classification probability, use the binary cross-entropy loss function for training to obtain the trained CNN-LSTM dual-branch feature fusion model.

8. A short-term rock damage early warning device based on acoustic emission clustering analysis, wherein the short-term rock damage early warning device based on acoustic emission clustering analysis is used to implement the short-term rock damage early warning method based on acoustic emission clustering analysis as described in any one of claims 1-7, characterized in that, The device includes: The acquisition unit is used to acquire acoustic emission characteristic parameters during rock deformation and failure. The clustering analysis unit is used to perform clustering analysis on the acoustic emission feature parameters using the Kmeans++ algorithm to obtain the clustering labels corresponding to the acoustic emission feature parameters; The first construction unit is used to calculate the importance score of the acoustic emission feature parameters according to the clustering labels using the random forest algorithm; extract the acoustic emission feature parameters with an importance score greater than 0.1, and construct a set of early warning indicators. The second building unit is used to build the initial CNN-LSTM dual-branch feature fusion model; The training unit is used to establish a sample dataset of conventional and precursor signals of rock damage based on the set of early warning indicators; and to train the initial CNN-LSTM dual-branch feature fusion model based on the sample dataset to obtain the trained CNN-LSTM dual-branch feature fusion model. The early warning unit is used to acquire real-time acoustic emission signals; input the real-time acoustic emission signals into the trained CNN-LSTM dual-branch feature fusion model for recognition, obtain the recognition result, and use the acoustic emission signal with a recognition result of 1 as the early warning signal of rock damage.

9. A short-term early warning device for rock damage based on acoustic emission clustering analysis, characterized in that, The rock damage short-term early warning device based on acoustic emission clustering analysis includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images