Data analysis method based on rockburst tendency

Abstract

The invention relates to the field of intelligent analysis, and particularly discloses a data analysis method based on rockburst tendency, which comprises the following steps: acquiring ground stress, underground water pressure and seismic activity frequency of a plurality of points to be measured of a mine, and adopting a deep learning technology and a data processing method to analyze the rockburst tendency. And performing data association analysis on the geological states of the plurality of to-be-measured points of the mine to determine the rockburst tendency grade of the mine. Through the method, a visual rockburst tendency grade can be provided for a decision maker, so that the accuracy and reliability of rockburst tendency evaluation are improved, risk evaluation and decision making are facilitated, corresponding safety measures and management measures are adopted, and powerful support and guidance are provided for mine safety management.

Description

Technical Field

[0001] The present application relates to the field of intelligent analysis, and more specifically, to a data analysis method based on rockburst tendency. Background Art

[0002] Rockburst refers to the phenomenon that rock breaks, disintegrates or collapses, releasing huge energy and shock waves due to stress or other factors within the rock body or between the rock body and engineering excavation.

[0003] Rockburst propensity refers to the possibility or degree of tendency of a rock mass or rock mass to undergo rockburst under certain conditions. It is used to assess the stability and rockburst risk of a rock mass and help determine what measures to take in the rock mass to reduce the occurrence of rockburst. Therefore, by conducting data analysis on rockburst propensity, the possibility of rockburst in a rock mass can be predicted. This helps to identify potential rockburst danger areas in advance and take corresponding preventive measures to reduce the occurrence of accidents.

[0004] However, because traditional methods usually only consider a few factors, they cannot comprehensively and comprehensively evaluate rockburst tendency. Specifically, rockburst is a complex process involving the interaction of multiple factors. However, traditional methods often fail to fully consider the combined impact of these factors, and there will be problems of incompleteness and inaccuracy, which limits the accuracy and reliability of the evaluation. In addition, traditional methods often rely on the subjective judgment and experience of experts, lacking scientificity and objectivity. The evaluation results may be affected by personal subjective bias, resulting in inconsistency in the evaluation results.

[0005] Therefore, an optimized data analysis scheme based on rockburst tendency is expected. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a data analysis method based on rockburst tendency, which obtains the ground stress magnitude, groundwater pressure and seismic activity frequency of multiple test points in the mine, and adopts deep learning technology and data processing methods to perform data correlation analysis on the geological conditions of multiple test points in the mine to determine the rockburst tendency level of the mine. This method can provide decision makers with an intuitive rockburst tendency level, thereby improving the accuracy and reliability of rockburst tendency assessment, so as to conduct risk assessment and decision-making, and take corresponding safety measures and management measures, providing strong support and guidance for mine safety management.

[0007] According to one aspect of the present application, a data analysis method based on rockburst tendency is provided, which includes:

[0008] Obtain the ground stress magnitude, groundwater pressure and seismic activity frequency at multiple test points in the mine;

[0009] Arranging the ground stress magnitudes, the groundwater pressures and the seismic activity frequencies of the multiple test points into multiple geological state input vectors of the test points according to the sample dimensions, and then performing multi-scale feature extraction to obtain multi-scale geological state feature vectors of the multiple test points;

[0010] Arranging the plurality of multi-scale geological state feature vectors of the test points to obtain a multi-scale geological state feature matrix of the test points, and then performing an attention mechanism enhancement to obtain an attention-enhanced geological state feature matrix of the test points;

[0011] Based on the attention-enhanced geological state characteristic matrix of the test point, the rockburst tendency level of the mine is determined.

[0012] In the above-mentioned data analysis method based on rock burst tendency, the ground stress magnitudes, the groundwater pressure and the seismic activity frequency of the multiple test points are arranged according to the sample dimension as multiple geological state input vectors of the test points, and then multi-scale feature extraction is performed to obtain multiple multi-scale geological state feature vectors of the test points, including: arranging the ground stress magnitudes, groundwater pressure and seismic activity frequency of the multiple test points according to the sample dimension as the multiple geological state input vectors of the test points; and passing the multiple geological state input vectors of the test points through a multi-scale neighborhood feature extraction module to obtain the multiple multi-scale geological state feature vectors of the test points.

[0013] In the above-mentioned data analysis method based on rock burst tendency, the multi-scale domain feature extraction module includes: a first convolution layer, a second convolution layer parallel to the first convolution layer, and a cascade layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer uses a one-dimensional convolution kernel with a first scale, and the second convolution layer uses a one-dimensional convolution kernel with a second scale.

[0014] In the above-mentioned data analysis method based on rock burst tendency, the multi-scale geological state feature vectors of the multiple test points are arranged to obtain the multi-scale geological state feature matrix of the test points, and then the attention mechanism is enhanced to obtain the attention-enhanced geological state feature matrix of the test points, including: arranging the multi-scale geological state feature vectors of the multiple test points to obtain the multi-scale geological state feature matrix of the test points; and the multi-scale geological state feature matrix of the test points is subjected to a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test points.

[0015] Compared with the prior art, the present application provides a data analysis method based on rockburst tendency, which obtains the ground stress magnitude, groundwater pressure and seismic activity frequency of multiple test points in the mine, and uses deep learning technology and data processing methods to perform data correlation analysis on the geological state of multiple test points in the mine to determine the rockburst tendency level of the mine. This method can provide decision makers with an intuitive rockburst tendency level, thereby improving the accuracy and reliability of rockburst tendency assessment, so as to conduct risk assessment and decision-making, and take corresponding safety measures and management measures, providing strong support and guidance for mine safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 The figure is a flow chart of a data analysis method based on rockburst tendency according to an embodiment of the present application.

[0018] Figure 2 Schematic diagram of data flow of a data analysis method based on rockburst tendency according to an embodiment of the present application.

[0019] Figure 3 A flowchart of a data analysis method based on rock burst tendency according to an embodiment of the present application, in which the ground stress magnitudes of the multiple test points, the groundwater pressure and the seismic activity frequency are arranged according to sample dimensions as multiple geological state input vectors of the test points, and then multi-scale feature extraction is performed to obtain multi-scale geological state feature vectors of the multiple test points.

[0020] Figure 4 This is a flow chart of arranging the multi-scale geological state feature vectors of the multiple test points to obtain the multi-scale geological state feature matrix of the test points, and then performing attention mechanism enhancement to obtain the attention-enhanced geological state feature matrix of the test points in the data analysis method based on rock burst tendency according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0022] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0023] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0024] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0025] Rockburst refers to the phenomenon of rock mass rupture and collapse in underground mines or rock engineering due to the influence of geological conditions, ground stress, rock mechanical properties and other factors. Rockburst is usually accompanied by huge energy release, causing serious damage and harm. Rockburst is an important disaster in mine safety, which may cause serious harm to miners and equipment, and even cause casualties and mine shutdown.

[0026] Rockburst tendency refers to an evaluation index of the possibility or potential risk of rockburst in underground mines or rock engineering. It is used to evaluate the influence of factors such as rock structure, ground stress, rock mechanical properties, groundwater pressure, etc. on rockburst occurrence to determine the possibility of rockburst. Rockburst tendency assessment is usually based on the collection and analysis of data such as geological surveys, underground exploration, rock mechanical tests and numerical simulations at the mine or engineering site. With these data, rockburst tendency can be evaluated and divided, usually using a hierarchy or quantitative scoring method to divide rockburst tendency into different grades or levels. This helps mine managers and engineers understand the potential risks of rockburst and take corresponding prevention and control measures to ensure the safe operation of mines or projects.

[0027] Traditional methods often only consider a few factors when evaluating rockburst tendency. Rockburst is a complex process involving the interaction of multiple factors such as geology, mechanics, and hydrology. Therefore, traditional methods cannot fully and comprehensively consider the interaction of multiple factors, resulting in incompleteness and inaccuracy in the evaluation results, which limits the accuracy and reliability of the evaluation. In addition, traditional methods rely on the subjective judgment and experience of experts, which may be subjectively biased and lack scientificity and objectivity, thus affecting the correctness of the evaluation results. Therefore, an optimized data analysis method based on rockburst tendency is desired.

[0028] Figure 1 The figure is a flow chart of a data analysis method based on rockburst tendency according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of data flow of a data analysis method based on rockburst tendency according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the data analysis method based on rockburst tendency according to the embodiment of the present application, the method includes: S110, obtaining the ground stress magnitude, groundwater pressure and seismic activity frequency of multiple test points in the mine; S120, arranging the ground stress magnitude, the groundwater pressure and the seismic activity frequency of the multiple test points into multiple geological state input vectors of the test points according to the sample dimension, and then performing multi-scale feature extraction to obtain multiple multi-scale geological state feature vectors of the test points; S130, arranging the multi-scale geological state feature vectors of the multiple test points to obtain a multi-scale geological state feature matrix of the test points, and then performing attention mechanism enhancement to obtain an attention-enhanced geological state feature matrix of the test points; and, S140, determining the rockburst tendency level of the mine based on the attention-enhanced geological state feature matrix of the test points.

[0029] In step S110, the magnitude of the ground stress, groundwater pressure and seismic activity frequency of multiple points to be measured in the mine are obtained. It should be understood that considering that the ground stress refers to the magnitude and direction of the force exerted on the underground rock body, the ground stress can be divided into stress components in different directions such as horizontal stress, vertical stress and shear stress. Understanding the magnitude of the ground stress can help evaluate the stability of the rock body and the potential risk of rock burst. The magnitude of the ground stress has an important influence on the deformation and fracture of the rock body. Groundwater pressure refers to the pressure exerted by groundwater on the rock body, that is, the height of the groundwater level and the distribution of groundwater pressure. The magnitude and change of groundwater pressure have an important influence on the stability of the rock body and the occurrence of rock burst. Specifically, high groundwater pressure may cause the fracture and instability of the rock body, increasing the risk of rock burst. The frequency of seismic activity refers to the frequency of earthquake events in a certain area. The frequency of seismic activity reflects the activity of geological structures and the accumulation and release of seismic stress. There is a certain correlation between the frequency of seismic activity and the tendency of rock burst, and high-frequency seismic activity may increase the risk of fracture and rock burst of the rock body. Because the fracture and deformation of the rock body may trigger an earthquake. Understanding the frequency of seismic activity can provide information about the dynamic changes of underground rock masses, and then evaluate and predict the risk of rock bursts. Based on this, in the technical solution of this application, obtaining the ground stress magnitude, groundwater pressure and seismic activity frequency of multiple test points in the mine can reveal the distribution characteristics of ground stress at different locations, understand the distribution, pressure magnitude and change trend of groundwater, and reveal the characteristics of seismic activity at different locations, so as to further evaluate the potential risk of rock bursts, infer the possibility of rock bursts, and provide a more comprehensive basis for the subsequent evaluation of rock burst tendency.

[0030] In particular, in a specific embodiment of the present application, the ground stress magnitude of the test point can be collected by a ground stress measuring instrument. Specifically, the ground stress measuring instrument is an instrument specially used to measure ground stress, and the ground stress measuring instrument can directly measure the magnitude and direction of the force exerted on the underground rock body. The groundwater pressure of the test point is collected by a groundwater pressure gauge or a water level gauge. Specifically, the groundwater pressure can be monitored by installing sensors such as pressure gauges or water level gauges in groundwater wells or observation holes with shallow groundwater levels. The frequency of seismic activity can be directly obtained by seismic monitoring agencies.

[0031] In step S120, the ground stress magnitudes, the groundwater pressures and the seismic activity frequencies of the multiple test points are arranged according to sample dimensions into multiple geological state input vectors of the test points, and then multi-scale feature extraction is performed to obtain multi-scale geological state feature vectors of the multiple test points.

[0032] Figure 3A flowchart of a method for analyzing rockburst tendency based on an embodiment of the present application is provided for arranging the geostress magnitudes, groundwater pressures, and seismic activity frequencies of the multiple test points according to sample dimensions as multiple test point geological state input vectors, and then extracting multi-scale features to obtain multi-scale geological state feature vectors of the multiple test points. Specifically, in the embodiment of the present application, Figure 3 As shown, the geostress magnitudes, groundwater pressures and seismic activity frequencies of the multiple test points are arranged according to sample dimensions as multiple geological state input vectors of the test points, and then multi-scale feature extraction is performed to obtain multi-scale geological state feature vectors of the multiple test points, including: S210, arranging the geostress magnitudes, groundwater pressures and seismic activity frequencies of the multiple test points according to sample dimensions as the multiple geological state input vectors of the test points; and, S220, passing the multiple geological state input vectors of the test points through a multi-scale neighborhood feature extraction module to obtain the multi-scale geological state feature vectors of the multiple test points.

[0033] Specifically, in step S210, the ground stress magnitudes, groundwater pressures, and seismic activity frequencies of the multiple test points are arranged according to the sample dimension as the geological state input vectors of the multiple test points. Accordingly, considering that each test point has different geological state test data, each geological test data is interrelated. Based on this, in the technical solution of the present application, the ground stress magnitudes, groundwater pressures, and seismic activity frequencies of the multiple test points are arranged according to the sample dimension as the geological state input vectors of the multiple test points, and the geological state information of each test point can be integrated together to form a complete data sample for subsequent unified processing and comparison. In other words, integrating the geological state information of multiple test points into one input vector can more comprehensively consider the mutual relationship and influence between different test points. This helps to comprehensively evaluate the tendency of rock bursts and predict potential risks, rather than just independently evaluating each test point.

[0034] Specifically, in step S220, the geological state input vectors of the multiple test points are passed through a multi-scale neighborhood feature extraction module to obtain the multi-scale geological state feature vectors of the multiple test points. Specifically, in an embodiment of the present application, the multi-scale neighborhood feature extraction module includes: a first convolution layer, a second convolution layer parallel to the first convolution layer, and a cascade layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer uses a one-dimensional convolution kernel with a first scale, and the second convolution layer uses a one-dimensional convolution kernel with a second scale. It should be understood that the geological state may have different characteristics and changes at different scales. For example, the local structure and deformation of the rock may be more obvious at a small scale, while groundwater pressure and seismic activity may have a wider impact at a larger scale. At the same time, the interaction between different test points has an important influence on the tendency of rock burst. Changes in the geological state at a test point may affect the geological state of the surrounding neighborhood. Therefore, in the technical solution of the present application, the geological state input vectors of the multiple test points are passed through a multi-scale neighborhood feature extraction module to obtain multi-scale geological state feature vectors of the multiple test points, which can capture geological information at different scales, and can take into account the interaction between different test points in the neighborhood to obtain more accurate geological state characteristics, and provide a more comprehensive feature representation for rock burst tendency assessment. In particular, the assessment of rock burst tendency requires comprehensive consideration of geological information at different scales. Through the multi-scale neighborhood feature extraction module, geological state characteristics of different scales can be integrated into a feature vector for comprehensive evaluation and analysis, so as to fully understand multiple aspects of the geological state and improve the accuracy and reliability of rock burst tendency assessment.

[0035] In particular, in another achievable embodiment, the geostress magnitudes, groundwater pressures and seismic activity frequencies of the multiple points to be tested are arranged as multiple geological state input vectors of the points to be tested according to the sample dimension, and then multi-scale feature extraction is performed to obtain multiple multi-scale geological state feature vectors of the points to be tested, which can be achieved by the following steps. First, the data of the geostress magnitudes, groundwater pressures and seismic activity frequencies of the multiple points to be tested are collected and sorted to ensure that the data have the same dimension and format. Then, the geostress magnitudes, groundwater pressures and seismic activity frequencies of each point to be tested are arranged according to the sample dimension to form an input vector, and each input vector represents the geological state information of a point to be tested. Next, neighborhood ranges of different scales are defined. Different neighborhood radii or neighborhood sizes can be selected to cover geological features at different scales. For each point to be tested, according to the defined neighborhood range, the geostress magnitudes, groundwater pressures and seismic activity frequency data in the neighborhood are extracted, and feature extraction is performed on the extracted neighborhood data, such as statistical features (mean, variance, etc.) or frequency domain features (Fourier transform, etc.). These features can capture the multi-scale information of the geological state in the neighborhood. The extracted features are then combined into a multi-scale geological state feature vector, which represents the geological features of the test point at different scales. Finally, for each test point, a multi-scale geological state feature vector is obtained, which contains the geological feature information of the test point at different scales. Through the above steps, the ground stress magnitude, groundwater pressure and seismic activity frequency data of multiple test points can be arranged according to the sample dimension, and the multi-scale geological state feature vectors of multiple test points can be obtained through the multi-scale feature extraction module. This processing method can more comprehensively consider the geological characteristics and interactions at different scales, and improve the accuracy and reliability of rock burst tendency assessment.

[0036] In step S130, the plurality of multi-scale geological state feature vectors of the test points are arranged to obtain a multi-scale geological state feature matrix of the test points, and then an attention mechanism is performed to enhance the feature matrix to obtain an attention-enhanced geological state feature matrix of the test points. Figure 4 The present invention is a flow chart of arranging the multi-scale geological state feature vectors of the plurality of test points to obtain the multi-scale geological state feature matrix of the test points and then performing attention mechanism enhancement to obtain the attention-enhanced geological state feature matrix of the test points in the data analysis method based on rockburst tendency according to the embodiment of the present application. Specifically, in the embodiment of the present application, if Figure 4As shown, the plurality of multi-scale geological state feature vectors of the test points are arranged to obtain a multi-scale geological state feature matrix of the test points, and then an attention mechanism is performed to enhance the attention of the test points to obtain the attention-enhanced geological state feature matrix of the test points, including: S310, arranging the plurality of multi-scale geological state feature vectors of the test points to obtain the multi-scale geological state feature matrix of the test points; and, S320, passing the multi-scale geological state feature matrix of the test points through a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test points.

[0037] Specifically, in step S310, the multi-scale geological state feature vectors of the plurality of test points are arranged to obtain the multi-scale geological state feature matrix of the test points. It should be understood that, considering that each of the multi-scale geological state feature vectors of the plurality of test points has different geological state feature information, each geological state feature information is mutually related and influenced. Based on this, in the technical solution of the present application, the multi-scale geological state feature vectors of the plurality of test points are arranged to obtain the multi-scale geological state feature matrix of the test points, which can facilitate unified data processing and analysis. That is to say, each row of the multi-scale geological state feature matrix of the test points corresponds to a test point, and each column corresponds to a geological state feature, so that the multi-scale geological state features of different test points can be intuitively compared and analyzed. In this way, the differences and similarities between them can be understood. At the same time, the relationship between the rows of the matrix can be used to study the geological state change trend and interaction between the test points. Such an arrangement method can better utilize the multi-scale geological state information and improve the accuracy and reliability of rock burst tendency assessment.

[0038] Specifically, in step S320, the multi-scale geological state feature matrix of the test point is passed through a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test point. It should be understood that considering that the convolutional neural network has advantages in processing spatial data. For the geological state feature matrix, the multi-scale features of each test point can be regarded as a local structure in space. By using a method based on a convolutional neural network, spatial features can be effectively extracted from the geological state feature matrix to capture the spatial relationship and local pattern between the test points. The bidirectional attention mechanism can help the model pay better attention to important geological features when processing the geological state feature matrix. It is worth mentioning that the bidirectional attention mechanism can automatically learn the correlation and weight distribution between the test points, so that important geological features get more attention and weight. At the same time, the bidirectional attention mechanism can simultaneously consider the self-attention and mutual attention of the test point, so that the model can perform feature weighting and interaction in the global and local range, pay more attention to important geological information, and reduce the attention to irrelevant or minor features, so as to more accurately capture the important information of the geological state and improve the accuracy of rock burst tendency assessment.

[0039] Specifically, in an embodiment of the present application, the multi-scale geological state feature matrix of the test point is subjected to a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test point, including: pooling the multi-scale geological state feature matrix of the test point along the horizontal direction and the vertical direction respectively to obtain a first-direction multi-scale geological state pooling vector of the test point and a second-direction multi-scale geological state pooling vector of the test point; associatively encoding the first-direction multi-scale geological state pooling vector of the test point and the second-direction multi-scale geological state pooling vector of the test point to obtain a bidirectional association matrix of the multi-scale geological state of the test point; inputting the bidirectional association matrix of the multi-scale geological state of the test point into a Sigmoid activation function to obtain a bidirectional association weight matrix of the multi-scale geological state of the test point; and calculating the position point multiplication between the bidirectional association weight matrix of the multi-scale geological state of the test point and the multi-scale geological state feature matrix of the test point to obtain the classification feature matrix.

[0040] In step S140, the rockburst tendency level of the mine is determined based on the attention-enhanced geological state characteristic matrix of the test point. It should be understood that, considering that the attention-enhanced geological state characteristic matrix of the test point is obtained by the geological state characteristic matrix of the test point after attention enhancement, it can simultaneously consider the local characteristics of the point and the global relationship with other test points, so that the details and overall characteristics of the geological state can be more comprehensively captured, which is helpful to accurately evaluate the rockburst tendency.

[0041] Specifically, in an embodiment of the present application, based on the attention-enhanced geological state characteristic matrix of the test point, the rockburst tendency level of the mine is determined, including: expanding the attention-enhanced geological state characteristic matrix of the test point to obtain a characteristic vector of the attention-enhanced geological state of the test point; performing feature optimization on the characteristic vector of the attention-enhanced geological state of the test point to obtain an optimized characteristic vector of the attention-enhanced geological state of the test point; and passing the optimized characteristic vector of the attention-enhanced geological state of the test point through a classifier to obtain a classification result, and the classification result is used to represent the rockburst tendency level of the mine.

[0042] Specifically, in an embodiment of the present application, the feature optimization of the attention-enhanced geological state feature vector of the test point to obtain the optimized attention-enhanced geological state feature vector of the test point includes: the metric data-guided optimization of the attention-enhanced geological state feature vector of the test point based on the decoupling of the category weight information to obtain the optimized attention-enhanced geological state feature vector of the test point. In particular, in the technical solution of the present application, considering that there may be some extreme values ​​or outliers in the actual geological data, these data points have significantly different characteristics compared with most data points. This local abnormal distribution may be caused by a sudden change in geological conditions, measurement errors or other atypical factors. In the machine learning model, these abnormal points may have a negative impact on the training and prediction of the model. When the geological state feature vector of the test point has a local abnormal distribution, these abnormal points may be far away from most points in the feature space, forming outliers. When the classifier is classified, these outliers may be misclassified or affect the decision boundary of the classifier, causing the classification result to deviate from the predetermined class label. This phenomenon is called category outlier offset. If the classifier is affected by these outliers during the training process, it may cause the generalization ability of the model to decrease, affecting the prediction effect of the model on unseen data. In addition, the existence of outliers may also affect the iterative optimization process of the model, causing the model to overfit these abnormal points while pursuing the minimization of the overall error, thereby reducing the classification accuracy of the model for normal data points. In order to solve this problem, in the technical solution of the present application, the attention-enhanced geological state feature vector of the test point is optimized based on the metric data guided by the decoupling of the category weight information to obtain the optimized attention-enhanced geological state feature vector of the test point.

[0043] More specifically, in an embodiment of the present application, the attention-enhanced geological state feature vector of the test point is optimized based on the metric data guided optimization of the decoupling of the category weight information to obtain the optimized attention-enhanced geological state feature vector of the test point, including: determining the category decoding weight matrix of the classifier; performing matrix decoupling of the category decoding weight matrix in units of row vectors to obtain a set of category decoding weight vectors; calculating the category metric learning loss information item of the attention-enhanced geological state of the test point between each category decoding weight vector in the set of category decoding weight vectors and the attention-enhanced geological state feature vector of the test point to obtain the attention-enhanced geological state of the test point. The invention relates to a method for obtaining a probabilistic attention-enhanced geological state category metric learning compensation representation vector of the attention-enhanced geological state category metric learning compensation representation vector of the test point; inputting the probabilistic attention-enhanced geological state category metric learning compensation representation vector of the test point into a softmax activation function to obtain a probabilistic attention-enhanced geological state category metric learning compensation weight vector of the test point; calculating the point-by-point multiplication of the probabilistic attention-enhanced geological state category metric learning compensation weight vector and the attention-enhanced geological state feature vector of the test point to obtain a point-by-point multiplication feature vector, and calculating the weighted sum of the point-by-point multiplication feature vector and the attention-enhanced geological state feature vector of the test point to obtain the optimized attention-enhanced geological state feature vector of the test point.

[0044] More specifically, in an embodiment of the present application, the following category metric learning compensation formula is used to calculate the category metric learning loss information item of the attention-enhanced geological state of the test point between each category decoding weight vector in the set of category decoding weight vectors and the attention-enhanced geological state feature vector of the test point to obtain the category metric learning compensation representation vector of the attention-enhanced geological state of the test point;

[0045] Among them, the category metric learning compensation formula is:

[0046]

[0047] in, represents the eigenvalue of the i-th position of the j-th category decoding weight vector in the set of category decoding weight vectors, x i represents the eigenvalue of the i-th position of the attention-enhanced geological state feature vector of the test point, L represents the length of the feature vector, D j Represents the eigenvalue of the j-th position of the attention-enhanced geological state category metric learning compensation representation vector of the test point.

[0048] That is, the characteristic vector of the geological state of the attention enhancement of the test point is optimized based on the metric data guided by the decoupling of the category weight information. The category decoding weight matrix of the classifier is first determined to understand the classifier's distinction between different category features. Subsequently, the category decoding weight matrix is ​​decoupled by row vectors to obtain a set of category decoding weight vectors to allow the model to independently analyze the weight of each category and provide more refined control for feature compensation. Next, the category metric learning loss information item of the geological state of the attention enhancement of the test point is calculated. This process involves quantitative analysis of the interaction between each category weight vector and the characteristic vector of the geological state of the attention enhancement of the test point. Through the calculation of a specific loss function, the inconsistency between the feature vector and the category weight is captured, providing a quantitative indicator for compensation. Next, the compensation representation vector of the geological state category metric learning of the attention enhancement of the test point is input into the activation function, and a nonlinear transformation is performed to convert it into a probabilistic form. The application of the activation function introduces nonlinearity, enhances the model's ability to interpret the compensation weight, and makes the compensation weight more consistent with the actual distribution characteristics. Then, the compensation representation vector of the attention-enhanced geological state category metric learning of the test point after the activation function is used to calculate the probabilistic compensation weight vector of the attention-enhanced geological state category metric learning of the test point. This step provides probabilistic guidance for the adjustment of the feature vector in a probabilistic form, making the optimization process more in line with statistical characteristics, thereby enhancing the accuracy of the model's feature adjustment. Finally, by calculating the dot product of the probabilistic attention-enhanced geological state category metric learning compensation weight vector and the attention-enhanced geological state feature vector of the test point, and then calculating the weighted sum with the original attention-enhanced geological state feature vector of the test point, the optimized attention-enhanced geological state feature vector of the test point is obtained. This step combines the original feature information and the compensation information to generate an adjusted feature vector, which reduces the impact of abnormal distribution and improves the accuracy of classification.

[0049] Specifically, in an embodiment of the present application, the optimized attention-enhanced geological state feature vector of the test point is subjected to a classification result through a classifier, and the classification result is used to represent the rockburst tendency level of the mine. That is to say, the optimized attention-enhanced geological state feature vector of the test point is classified and processed by a classifier, and the classification result is used to represent the rockburst tendency level of the mine. Specifically, the rockburst tendency level may include low rockburst tendency, medium-low rockburst tendency, medium-high rockburst tendency, high rockburst tendency, or level 1, level 2, level 3, etc. The division of rockburst tendency levels helps to identify and evaluate rockburst risks in mines, and provides guidance and basis for mine safety management and decision-making. It should be understood that the classifier can map the input geological state feature matrix to different categories or probability distributions to represent different rockburst tendency levels. The output result of the classifier is a discrete rockburst tendency level, which can provide a standardized evaluation index for the rockburst tendency of the mine. This can facilitate comparison and evaluation between different mines and provide a unified reference for decision makers.

[0050] In summary, the data analysis method based on rockburst tendency according to the embodiment of the present application is explained, which obtains the ground stress magnitude, groundwater pressure and seismic activity frequency of multiple test points in the mine, and uses deep learning technology and data processing methods to perform data association analysis on the geological state of multiple test points in the mine to determine the rockburst tendency level of the mine. This method can provide decision makers with an intuitive rockburst tendency level, thereby improving the accuracy and reliability of rockburst tendency assessment, so as to conduct risk assessment and decision-making, and take corresponding safety measures and management measures, providing strong support and guidance for mine safety management.

[0051] Specifically, the present application also provides a kind of computer readable storage medium.Computer readable storage medium can be a tangible device that can keep and store instructions used by instruction execution device.Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above.More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination of the above.The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0052] The above is merely an example of the principles of the present disclosure, and various modifications may be made by those skilled in the art without departing from the scope of the present disclosure. The above embodiments are presented for the purpose of illustration rather than limitation. The present disclosure may also take many forms other than those explicitly described herein.

Claims

1. A data analysis method based on rockburst tendency, characterized in that: include: Obtain the ground stress magnitude, groundwater pressure and seismic activity frequency at multiple test points in the mine; Arranging the ground stress magnitudes, the groundwater pressures and the seismic activity frequencies of the multiple test points into multiple geological state input vectors of the test points according to the sample dimensions, and then performing multi-scale feature extraction to obtain multi-scale geological state feature vectors of the multiple test points; Arranging the plurality of multi-scale geological state feature vectors of the test points to obtain a multi-scale geological state feature matrix of the test points, and then performing an attention mechanism enhancement to obtain an attention-enhanced geological state feature matrix of the test points; Based on the attention-enhanced geological state characteristic matrix of the test point, the rockburst tendency level of the mine is determined.

2. The data analysis method based on rockburst tendency according to claim 1 is characterized in that: Arranging the geostress magnitudes, groundwater pressures and seismic activity frequencies of the multiple test points into multiple geological state input vectors of the test points according to the sample dimensions and then performing multi-scale feature extraction to obtain multi-scale geological state feature vectors of the multiple test points, including: Arrange the ground stress magnitudes, groundwater pressures, and seismic activity frequencies of the multiple test points according to sample dimensions as geological state input vectors of the multiple test points; The geological state input vectors of the multiple test points are passed through a multi-scale neighborhood feature extraction module to obtain multi-scale geological state feature vectors of the multiple test points.

3. The data analysis method based on rockburst tendency according to claim 2 is characterized in that: The multi-scale neighborhood feature extraction module includes: a first convolutional layer, a second convolutional layer parallel to the first convolutional layer, and a cascade layer connected to the first convolutional layer and the second convolutional layer, wherein the first convolutional layer uses a one-dimensional convolutional kernel with a first scale, and the second convolutional layer uses a one-dimensional convolutional kernel with a second scale.

4. The data analysis method based on rockburst tendency according to claim 3 is characterized in that: Arranging the plurality of multi-scale geological state feature vectors of the test points to obtain a multi-scale geological state feature matrix of the test points, and then performing attention mechanism enhancement to obtain an attention-enhanced geological state feature matrix of the test points, including: Arranging the plurality of multi-scale geological state feature vectors of the test points to obtain a multi-scale geological state feature matrix of the test points; The multi-scale geological state feature matrix of the test point is passed through a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test point.

5. The data analysis method based on rockburst tendency according to claim 4 is characterized in that: The multi-scale geological state feature matrix of the test point is subjected to a bidirectional attention mechanism based on a convolutional neural network to obtain the attention-enhanced geological state feature matrix of the test point, including: Pooling the multi-scale geological state feature matrix of the test point in the horizontal direction and the vertical direction respectively to obtain a first-direction multi-scale geological state pooling vector of the test point and a second-direction multi-scale geological state pooling vector of the test point; Performing association coding on the first-direction multi-scale geological state pooling vector of the test point and the second-direction multi-scale geological state pooling vector of the test point to obtain a bidirectional association matrix of the multi-scale geological state of the test point; Inputting the bidirectional correlation matrix of the multi-scale geological state of the test point into the Sigmoid activation function to obtain the bidirectional correlation weight matrix of the multi-scale geological state of the test point; The classification feature matrix is ​​obtained by calculating the position point multiplication between the bidirectional association weight matrix of the multi-scale geological state of the test point and the multi-scale geological state feature matrix of the test point.

6. The data analysis method based on rockburst tendency according to claim 5 is characterized in that: Based on the attention-enhanced geological state characteristic matrix of the test point, the rockburst tendency level of the mine is determined, including: Expanding the attention-enhanced geological state feature matrix of the test point to obtain the attention-enhanced geological state feature vector of the test point; Performing feature optimization on the attention-enhanced geological state feature vector of the test point to obtain an optimized attention-enhanced geological state feature vector of the test point; The optimized attention-enhanced geological state feature vector of the test point is passed through a classifier to obtain a classification result, and the classification result is used to represent the rock burst tendency level of the mine.

7. The data analysis method based on rockburst tendency according to claim 6 is characterized in that: The feature optimization of the attention-enhanced geological state feature vector of the test point to obtain an optimized attention-enhanced geological state feature vector of the test point includes: The attention-enhanced geological state feature vector of the test point is optimized guided by metric data based on the decoupling of category weight information to obtain the optimized attention-enhanced geological state feature vector of the test point.

8. The data analysis method based on rockburst tendency according to claim 7 is characterized in that: The method of performing a metric data guided optimization based on the decoupling of category weight information on the attention-enhanced geological state feature vector of the test point to obtain the optimized attention-enhanced geological state feature vector of the test point includes: Determining a class decoding weight matrix for the classifier; Decoupling the class decoding weight matrix in units of row vectors to obtain a set of class decoding weight vectors; Calculating the category metric learning loss information item of the attention-enhanced geological state of the test point between each category decoding weight vector in the set of category decoding weight vectors and the attention-enhanced geological state feature vector of the test point to obtain the category metric learning compensation representation vector of the attention-enhanced geological state of the test point; Input the attention-enhanced geological state category metric learning compensation representation vector of the test point into a softmax activation function to obtain a probabilistic attention-enhanced geological state category metric learning compensation weight vector of the test point; Calculate the positional point multiplication of the probabilistic attention-enhanced geological state category metric learning compensation weight vector of the test point and the attention-enhanced geological state feature vector of the test point to obtain a point product feature vector, and calculate the weighted sum of the point product feature vector and the attention-enhanced geological state feature vector of the test point to obtain the optimized attention-enhanced geological state feature vector of the test point.

Images