Early warning method and system for mine slope monitoring

Abstract

The invention provides an early warning method and system for mine slope monitoring, and relates to the technical field of geographic information analysis systems. The method comprises the following steps: acquiring a real-time slope monitoring data set based on slope monitoring equipment deployed in a plurality of mine slope key areas; determining multi-attribute features of the monitoring data, and obtaining a plurality of slope section clusters by using a multi-attribute fuzzy clustering algorithm based on mixed data; screening a target risk slope section cluster according to the historical disaster recall rate of each slope section cluster; constructing slope section monitoring time sequence data according to the real-time slope monitoring data corresponding to each target risk cluster and the corresponding historical slope monitoring data set; acquiring a meteorological parameter sequence of the key area of the mine slope; and inputting the monitoring time sequence data of each slope section and the meteorological parameter sequence into a slope risk monitoring model to obtain a slope risk identification result, and further performing early warning. Therefore, multi-dimensional sensing equipment and an advanced data processing algorithm are comprehensively utilized, and real-time monitoring and early warning for the stability of the mine slope are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of geographic information analysis systems, and in particular to an early warning method and system for monitoring mine slopes. Background Art

[0002] Mine slope stability monitoring is a key component of mine safety production, involving multiple fields such as geology, engineering technology and environmental protection. Slope instability and sliding are common major safety accidents in mines, which may lead to serious consequences such as casualties, equipment damage and environmental damage.

[0003] At present, the monitoring of mine slope stability mainly relies on manual field geological survey and exploration, through which workers set up survey instruments on the mine slopes to monitor surface displacement, measure groundwater levels and conduct geological exploration. Although these methods can provide basic data for slope stability assessment, most of them rely on manual operation and have a long data collection cycle, and cannot achieve continuous or real-time monitoring of mine slopes, resulting in greater geological risk hazards.

[0004] In response to the above problems, the industry has not yet proposed a better technical solution. Summary of the invention

[0005] The embodiments of the present invention provide an early warning method, system, electronic device and storage medium for monitoring a mine slope, which are used to at least solve the problem in the prior art that the mine slope relies on manual exploration and cannot achieve continuous real-time monitoring.

[0006] In a first aspect, an embodiment of the present invention provides an early warning method for mine slope monitoring, comprising: obtaining a real-time slope monitoring data set based on slope monitoring equipment deployed in multiple key areas of mine slopes, the real-time slope monitoring data comprising InSAR images, GNSS parameters and underground stress distribution maps; determining multi-attribute features corresponding to each real-time slope monitoring data set, and clustering and analyzing each multi-attribute feature using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters, the multi-attribute features comprising slope segment location attribute features, slope segment type attribute features and slope segment numerical attribute features; calculating the historical disaster recall rate of each slope segment cluster, and adding the calculated historical disaster recall rate to each slope segment cluster. The slope segment clusters whose recall rate exceeds the preset disaster recall rate threshold are determined as the target risk slope segment clusters to be subjected to early warning risk assessment; for the real-time slope monitoring data corresponding to each target risk slope segment cluster, the slope segment monitoring time series data corresponding to the preset time period are constructed according to the real-time slope monitoring data and the historical slope monitoring data set that matches the slope segment position of the real-time slope monitoring data; the meteorological parameter sequence corresponding to the preset time period for each key area of ​​the mine slope is obtained; the slope segment monitoring time series data and the meteorological parameter sequence are input into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the early warning conditions. When the judgment result indicates yes, a warning signal is issued.

[0007] Optionally, determining the multi-attribute features corresponding to each real-time slope monitoring data in the real-time slope monitoring data set, and performing cluster analysis on each multi-attribute feature using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters, including: determining the real-time slope monitoring data set X = (x1, x2, x3, ..., x n ) in each real-time slope monitoring data x i The corresponding object feature matrix [x i,1 ,x i,2 ,…,x i,e ,x i,e+1 ,…,x i,m ,x i,m+1 ,…,x i,p ], where the first e features are the type features of the slope position, the middle m features are the type features of the slope classification, and the last p features are the type features of the slope value; the real-time slope monitoring data set X is divided into k clusters by minimizing the objective function of each attribute; the objective function adopts the following formula:

[0008] In the formula, is the fuzzy partitioning matrix on attribute f, satisfying the condition and is the set of cluster center prototypes on attribute f; is the data object x on attribute f i ; is the cluster center prototype of the j-th cluster on attribute f; α(1 < α < ∞) represents the fuzzy coefficient, n represents the total number of objects in the dataset, and k is the number of clusters into which the real-time dataset is partitioned; is the object x on attribute f i and the prototype Q j of the dissimilarity measure:

[0009] represents the a-th attribute of the data object x on attribute f i ; represents the cluster center of the numerical attribute on attribute f, e is the number of slope segment position type features, i and j represent the indices of data objects, and m is the number of slope segment classification type features, represents the cluster center of the categorical attribute on attribute f, and the objective function is rewritten as the following formula:

[0010]

[0011] where represents the central value of the j-th cluster corresponding to attribute 1, represents the sum of the fuzzy weights of the j-th cluster on attribute 1, is the cluster center prototype of the z-th cluster on attribute f, n represents the total number of objects in the dataset, e is the number of slope segment position type features, m is the number of slope segment classification type features, and k is the number of clusters into which the real-time dataset is partitioned; through continuous iteration, the center point of the class is moved to the average position of the included data, and then its internal data is re-partitioned until the distance from all data to the cluster center of the class to which it belongs is the closest, so that the real-time slope monitoring dataset X is classified into k (k < n) slope segment clusters S = (S1, S1, S1,..., S k ):

[0012] In the formula, μ i represents the mean of the slope segment cluster S i , x represents any item in the slope segment cluster S i , and VarS i represents the sum of the distances between all features in the cluster of S i and the mean, and argmin represents the independent variable value when the function obtains the minimum value within its domain.

[0013] Optionally, calculating the historical disaster recall rate of each slope segment cluster includes: obtaining historical disaster analysis data corresponding to each slope segment cluster; the historical disaster analysis data includes actual historical disaster events and historical disaster prediction results corresponding to the slope segment location of each slope segment cluster; calculating the historical disaster recall rate of each slope segment cluster by the following formula: In the formula, Recall i represents the historical disaster recall rate corresponding to the i-th slope segment cluster; TP i FN represents the number of events correctly predicted as disasters corresponding to the i-th slope segment cluster; i Represents the number of disasters that actually occurred and were not predicted.

[0014] Optionally, the slope risk monitoring model adopts a convolutional long short-term memory network, which includes a first input layer, a second input layer, a third input layer, a fourth input layer, a spatial convolution layer, a preprocessing layer, a feature fusion layer, a spatial pyramid pooling layer, a ConvLSTM layer and an output layer; the first input layer is used to obtain InSAR image time series data in the slope monitoring time series data; the second input layer is used to obtain GNSS parameter time series data in the slope monitoring time series data; the third input layer is used to obtain underground stress map time series data in the slope monitoring time series data; the fourth input layer is used to obtain meteorological parameter sequence in the slope monitoring time series data; the spatial convolution layer is used to extract spatial features from each initial time series data, and use the spatial features to update the initial time series data to obtain corresponding time series-spatial features; the initial time series data includes InSAR Image time series data, GNSS parameter time series data or underground stress map time series data; based on the preprocessing layer, the meteorological parameter sequence is normalized to obtain the corresponding meteorological time series features; based on the feature fusion layer, each time series-spatial feature and the meteorological time series feature are feature fused to determine the corresponding meteorological-time series-spatial feature; based on the spatial pyramid pooling layer, the multi-scale spatial feature information in the meteorological-time series-spatial feature is extracted, and the meteorological-time series-spatial feature is updated to obtain the corresponding target meteorological-time series-multi-scale spatial feature; based on the ConvLSTM layer, the meteorological-time series-multi-scale spatial feature is processed to obtain the corresponding time series relationship dependency feature; based on the output layer, the time series relationship dependency feature is processed to determine the risk probability for each preset slope risk type, and obtain the corresponding slope risk identification result.

[0015] Optionally, the structure of the spatial convolution layer is: S type =ReLU(W conv *X type +b conv ), where X type Represents any type of initial time series data X InSAR ,X GNSS ,X stress , XInSAR ,X GNSS and X stress They represent the InSAR image time series data, GNSS parameter time series data and underground stress map time series data respectively; S type W represents the temporal-spatial features output by the spatial convolutional layer for any type of initial temporal data; conv represents the convolution kernel weight of the spatial convolution layer, b conv represents the bias term of the spatial convolution layer, * represents the convolution operation, and ReLU represents the RELU activation function; the structure of the preprocessing layer is: In the formula, X weather represents the meteorological parameter sequence, μ weather and σ weather Respectively represent the mean and standard deviation corresponding to the meteorological data, P weather Represents the meteorological time series features output by the preprocessing layer; the structure of the feature fusion layer is:

[0016] F combined =∑ type α type ·S type +α weather ·P weather , where F combined represents the meteorological-temporal-spatial features determined by the feature fusion layer through the feature fusion operation; α type represents the feature adjustment weights for various types of initial time series data, α weather represents the feature adjustment weight for meteorological time series features, which is dynamically generated by the weight generation network. type represents the temporal-spatial features output by the spatial convolutional layer for any type of initial temporal data; the structure of the spatial pyramid pooling layer is: F spp =concat(P1(F combined ),P2(F combined ),…,P r (F combined )), where P r represents the rth level pooling operation, which represents the maximum pooling operation of the corresponding level scale; concat represents the feature concatenation operation, F combined represents the meteorological-temporal-spatial features determined by the feature fusion layer through the feature fusion operation; the structure of the ConvLSTM layer is: i t =σ(W xi *F spp +W hi *H t-1 +b i ), f t =σ(W xf *F spp +Whf *H t-1 +b f ), o t =σ(W xo *F spp +W ho *H t-1 +b o ), In the formula, i t ,f t ,o t They represent the input gate, forget gate and output gate respectively, and C t Indicates the unit status; W xi ,W xf ,W xo ,W xc and b i ,b f ,b o ,b c They represent the weight parameters in the corresponding model layer modules, σ represents the Sigmoid activation function; H t represents the hidden state, t represents the number of ConvLSTM layers, represents the Hadamard product, tanh represents the hyperbolic tangent function; the structure of the output layer is: score = W y ·H t +b y , Where W y represents the weight matrix with dimension [hidden layer dimension, number of categories], b y represents a bias vector with dimension [number of categories], where the number of categories is the total number of types U corresponding to the preset slope risk types; score represents the original risk score generated for each preset slope risk type, Y risk Indicates the risk probability of the current slope for each preset slope risk type; represents the exponential function e applied to the original risk score of category i x The result after processing, Represents the sum of the results of indexing the raw risk scores of all categories.

[0017] Optionally, the loss function L of the slope risk monitoring model is: L = λ1L class +λ2L severity , Where, L class represents the main classification loss, L severity represents the mean square error loss, λ1 and λ2 represent the corresponding loss adjustment weights, N represents the total number of samples in the data sample set, U represents the total number of categories corresponding to the preset slope risk type, and yi,c The label of the i-th sample indicates whether it belongs to category u, p i,c represents the probability that the model predicts that the i-th sample belongs to category u, s i represents the true severity score, Represents the severity score predicted by the model.

[0018] In the second aspect, an embodiment of the present invention also provides an early warning system for mine slope monitoring, including: a data acquisition unit, configured to acquire a real-time slope monitoring data set based on slope monitoring equipment deployed in multiple key areas of mine slopes; the real-time slope monitoring data includes InSAR images, GNSS parameters and underground stress distribution maps; an attribute clustering unit, configured to determine the multi-attribute features corresponding to each real-time slope monitoring data in the real-time slope monitoring data set, and cluster analyze each multi-attribute feature using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters; the multi-attribute features include slope segment location attribute features, slope segment type attribute features and slope segment numerical attribute features; a risk clustering identification unit, configured to calculate the historical disaster recall rate of each slope segment cluster, and to identify the historical disaster recall rate corresponding to the calculated historical disaster recall rate in each slope segment cluster. The slope segment cluster whose recall rate exceeds the preset disaster recall rate threshold is determined as the target risk slope segment cluster to be subjected to early warning risk assessment; the time series data determination unit is configured to construct the slope segment monitoring time series data corresponding to the preset time period according to the real-time slope monitoring data and the historical slope monitoring data set matching the slope segment position of the real-time slope monitoring data for the real-time slope monitoring data corresponding to each target risk cluster; the meteorological data acquisition unit is configured to acquire the meteorological parameter sequence corresponding to the preset time period for each key area of ​​the mine slope; the slope risk identification unit is configured to input the slope segment monitoring time series data and the meteorological parameter sequence into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the early warning condition, and issue an early warning signal when the judgment result indicates yes.

[0019] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the above method.

[0020] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, the computer-readable storage medium including a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the steps of the above method.

[0021] In a fifth aspect, an embodiment of the present invention further provides a computer program product, the computer program product comprising a computer program stored on a storage medium, the computer program comprising program instructions, and when the program instructions are executed by a computer, the computer executes the steps of the above method.

[0022] The early warning method, system, electronic device and computer-readable storage medium for monitoring a mine slope provided by the present invention can produce at least the following technical effects:

[0023] (1) By deploying monitoring equipment for InSAR images, GNSS parameters, and underground stress distribution maps, real-time data collection of the slope is achieved. Compared with traditional manual surveys, this technical solution can continuously monitor the slope conditions, greatly shorten the data collection cycle, improve real-time monitoring, and ensure the timeliness of the data collection process.

[0024] (2) Cluster analysis of monitoring data is performed using a multi-attribute fuzzy clustering algorithm based on mixed data. The various slope risk categories involved in mine slopes can be effectively identified based on multiple attributes (such as slope location, type, and value). The comprehensive multi-attribute analysis method in this technical solution can more accurately cluster areas with similar slope risks compared to a single data source or traditional clustering method, which helps to improve the pertinence and effectiveness of early warning.

[0025] (3) By calculating the historical recall rate and setting the threshold, high-risk slope clusters can be identified in time, so that the early warning system can issue early warnings for possible landslides and slope instability accidents. The analysis method based on the combination of historical data and real-time monitoring data can accurately identify the risky slope clusters, so as to facilitate the subsequent targeted risk identification. There is no need to conduct risk monitoring and analysis in ordinary areas, which reduces the system computing resources and improves the efficiency of real-time risk identification.

[0026] (4) When identifying the risk of the slope area in the target risk cluster, the meteorological parameter series is introduced, such as rainfall, temperature change and other meteorological factors, and input into the slope risk monitoring model together with the slope monitoring time series data. In this way, the impact of external environmental changes on slope stability can be comprehensively evaluated, and the adaptability and accuracy of the early warning system under different environmental conditions can be enhanced.

[0027] Through this technical solution, real-time monitoring of slopes, multi-dimensional data analysis, timely and accurate risk warnings and assessment of comprehensive environmental factors are achieved, which significantly improves the monitoring and early warning capabilities of mine slopes, effectively reduces the risks of casualties and environmental damage, ensures the continued stability of mine safety production, and solves the problem in existing technologies that mine slopes rely on manual exploration and cannot achieve continuous real-time monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 A flowchart showing an example of an early warning method for monitoring a mine slope according to an embodiment of the present invention is shown;

[0030] Figure 2 A structural block diagram of an example of a slope risk monitoring model according to an embodiment of the present invention is shown;

[0031] Figure 3 A structural block diagram showing an example of an early warning system for monitoring a mine slope according to an embodiment of the present invention;

[0032] Figure 4 It is a schematic structural diagram of an embodiment of an electronic device of the present invention.

[0033] The above drawings include the following reference numerals:

[0034] 211, first input layer; 212, second input layer; 213, third input layer; 214, fourth input layer; 221, spatial convolution layer; 222, preprocessing layer; 230, feature fusion layer; 240, spatial pyramid pooling layer; 250, ConvLSTM layer; 260, output layer; 410, processor; 420, memory; 430, input device; 440, output device. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one", "one" or "the" do not indicate a quantitative limitation, but indicate the existence of at least one. Words such as "include" or "comprise" mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

[0037] It should be noted that the terms "up", "down", "left", "right", "front", "back", etc. used in the present invention are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0038] Figure 1 A flowchart of an example of an early warning method for mine slope monitoring according to an embodiment of the present invention is shown.

[0039] Regarding the execution subject of the method of the embodiment of the present invention, it can be any controller or processor with computing or processing capabilities, so as to significantly improve the monitoring efficiency and early warning accuracy of slope stability by using advanced monitoring equipment and advanced data processing algorithms. In some examples, it can be integrated into the configuration server through software, hardware or a combination of software and hardware, which should not be limited here.

[0040] The following will take the mine slope risk warning platform as an exemplary implementation subject to introduce the details of the technical solution involved in the present invention. However, it should be understood that one or more steps involved in the following process can be implemented by one or more controllers or software installed and deployed in the client or server.

[0041] like Figure 1 As shown, in step S201, a real-time slope monitoring data set is obtained based on slope monitoring equipment deployed in multiple key areas of mine slopes.

[0042] Here, real-time slope monitoring data includes InSAR images, GNSS parameters and underground stress distribution maps. Specifically, GNSS monitoring, radar monitoring and underground displacement monitoring are used to integrate and centrally analyze the multi-dimensional parameters of ground displacement parameters, surface deformation parameters and underground stress parameters, so as to achieve efficient analysis of slope change data and timely warning of slope change conditions.

[0043] It should be noted that by deploying slope monitoring equipment and collecting real-time slope monitoring data (such as InSAR images, GNSS parameters, and underground stress distribution maps), real-time monitoring of the dynamic changes of mine slopes is achieved, creating conditions for continuous monitoring and analysis of slope risks. In addition, the acquisition of real-time data makes it possible to continuously track the state changes of slopes, providing the possibility of continuous assessment of slope stability, greatly improving the frequency of monitoring and the timeliness of data.

[0044] In some embodiments, GNSS receivers are typically installed at key points of the slope or monitoring area, such as the top and bottom of the slope. Ground radar equipment can use synthetic aperture radar (SAR) or ground-based radar interferometer (GB-InSAR), which is mainly used to monitor surface deformation. It can be set to aim at the slope or monitoring area to optimize the reflected signal and coverage. Underground displacement sensors can use ground stress sensors or underground deformation meters, which help the system understand the internal mechanism of the slope and identify landslide tendencies early, and can be distributed according to specific depths and patterns.

[0045] In step S202, the multi-attribute features corresponding to each real-time slope monitoring data in the real-time slope monitoring data set are determined, and each multi-attribute feature is clustered and analyzed using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters.

[0046] It should be noted that cluster analysis can identify clusters of slope sections with similar characteristics, which provides a scientific basis for subsequent risk assessment. By analyzing the clustering patterns of different slope sections, areas with similar risk characteristics can be more accurately located, thereby improving the efficiency and accuracy of monitoring.

[0047] Here, the multi-attribute features include slope location attribute features, slope type attribute features, and slope numerical attribute features. By performing fuzzy clustering analysis on the multi-attribute features of real-time monitoring data, the slope area can be classified according to geographical location, geological type, and numerical characteristics, thereby more accurately simulating the risk classification of each area in the slope.

[0048] By extracting multi-attribute features including slope location, type and numerical attributes from real-time slope monitoring data, the condition and potential risks of each slope segment can be fully described. The multi-attribute fuzzy clustering algorithm based on mixed data can not only process different types of data (such as categorical data and numerical data), but also enhance the robustness of clustering results through fuzzy logic and avoid information loss caused by hard partitioning.

[0049] In step S203, the historical disaster recall rate of each slope segment cluster is calculated, and the slope segment cluster whose corresponding calculated historical disaster recall rate exceeds a preset disaster recall rate threshold is determined as the target risk slope segment cluster for early warning risk assessment.

[0050] Here, the calculation of the recall rate is one of the key indicators for evaluating the performance of the monitoring system. By calculating the recall rate of each slope segment cluster and comparing it with the preset threshold, we can effectively screen out those key slope segments with recall rates higher than the threshold and higher risks, ensuring that resources can be concentrated on areas where disasters are most likely to occur, thereby improving the efficiency and accuracy of the early warning system.

[0051] By screening the recall rate, it is ensured that only those slope sections that really need attention will enter the next detailed analysis stage. The data-driven risk assessment method can minimize the possibility of false alarms and missed alarms, achieve targeted monitoring and early warning analysis, and effectively improve the early warning efficiency of risky mining areas while reducing system resource consumption.

[0052] In step S204, for the real-time slope monitoring data corresponding to each target risk slope segment cluster, slope segment monitoring time series data corresponding to a preset time period is constructed based on the real-time slope monitoring data and a historical slope monitoring data set matching the slope segment position of the real-time slope monitoring data.

[0053] Specifically, by combining real-time slope monitoring data with matching historical slope monitoring data sets, and constructing slope monitoring time series data for a preset time period, it is possible to observe the dynamic changes in slope status over time, identify potential risk change trends, and provide a richer and more accurate data basis for the final risk type identification. Through the analysis of time series data, the early warning system does not only rely on data from a single point in time to judge risks, but also conducts trend analysis of long-term data, which can more effectively capture the subtle dynamics of slope changes and enhance the foresight of early warnings and the accuracy of predictions.

[0054] In step S205, a meteorological parameter sequence corresponding to a preset time period for each key area of ​​the mine slope is obtained.

[0055] Here, the preset time period may be one week, half a month, or other time lengths, and may be adjusted according to business requirements or system resources.

[0056] For example, by installing environmental sensors on the slopes of the mine, various meteorological parameters can be monitored, such as rainfall, temperature, humidity, etc. It should be understood that all monitoring equipment works synchronously to ensure that the collected data is consistent in time, which is convenient for subsequent analysis and processing.

[0057] In step S206, the monitoring time series data and meteorological parameter sequence of each slope section are input into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the warning condition. If the judgment result indicates yes, a warning signal is issued.

[0058] Here, the risk monitoring model can adopt various types of multi-objective classification models to realize the classification and identification of the slope risk types (for example, collapse, debris flow and landslide) corresponding to the slope section monitoring time series data in the risk area, so as to obtain the corresponding slope risk identification results, such as whether there is a specific slope risk on the slope.

[0059] Furthermore, when it is determined that there is a specific slope risk, an early warning operation is performed according to the specific slope risk. For example, the mine slope risk early warning platform triggers the terminal at the corresponding risk slope location to perform an audible and visual alarm operation to warn workers not to carry out construction at this risk slope location in the near future (corresponding to a preset time period), so as to ensure the safety of mine construction to the greatest extent through intelligent means.

[0060] Regarding the implementation details of the above step S202, in some implementations, the real-time slope monitoring data set X=(x1, x2, x3, ..., x n ) in each real-time slope monitoring data x i The corresponding object feature matrix [x i,1 ,x i,2 ,…,x i,e ,x i,e+1 ,…,x i,m ,x i,m+1 ,…,x i,p ], where the first e features are the type features of the slope position, the middle m features are the type features of the slope classification, and the last p features are the type features of the slope value.

[0061] The real-time slope monitoring dataset X is divided into k clusters by minimizing the objective function of each attribute; the objective function adopts the following formula:

[0062]

[0063] In the formula, is the fuzzy partitioning matrix on attribute f, satisfying the condition and is the prototype set of cluster centers on attribute f; is the data object x on the attribute f i ; is the cluster center prototype of the j-th cluster on attribute f; α (1 < α < ∞) represents the fuzzy coefficient, n represents the total number of objects in the dataset, and k is the number of clusters into which the real-time dataset is partitioned.

[0064] For object x on attribute f i and prototype Q j of the dissimilarity measure:

[0065]

[0066] represents the a-th attribute of data object x on attribute f i ; represents the cluster center of the numerical attribute on attribute f, represents the cluster center of the categorical attribute on attribute f, e is the number of slope section position type features, i and j represent the indices of data objects, m is the number of slope section classification type features, and the objective function is rewritten as the following formula:

[0067]

[0068] represents the central value of the j-th cluster corresponding to attribute 1, represents the sum of the fuzzy weights of the j-th cluster on attribute 1, is the cluster center prototype of the z-th cluster on attribute f, n represents the total number of objects in the dataset, e is the number of slope section position type features, m is the number of slope section classification type features, and k is the number of clusters into which the real-time dataset is partitioned. Through continuous iteration, the center point of the class is moved to the average position of the included data, and then its internal data is repartitioned until the distance from all data to the cluster center of the class to which it belongs is the closest, so that the real-time slope monitoring dataset X is classified into k (k < n) slope section clusters S = (S1, S1, S1, …, S k ):

[0069]

[0070] where μ i represents the mean of slope section cluster S i , VarS i represents the sum of the distances between all features in the cluster of S i and the mean, x represents any item in slope section cluster S i , and argmin represents the independent variable value when the function obtains the minimum value within its domain.

[0071] Therefore, by improving the multi-attribute fuzzy clustering algorithm of mixed data, three types of attributes are used to represent the mixed data objects, representing the slope position type elements, slope classification type elements and slope value type elements respectively. By dividing them into three attributes, the accuracy of the objective function when performing the multi-attribute fuzzy clustering algorithm is increased, and the iteration speed is made faster, thereby speeding up the algorithm.

[0072] Regarding the implementation details of the above step S203, in some implementations, the historical disaster analysis data corresponding to each slope segment cluster is obtained, and the historical disaster analysis data includes the actual historical disaster events and historical disaster prediction results corresponding to the slope segment location of each slope segment cluster. Then, the historical disaster recall rate of each slope segment cluster is calculated by the following formula:

[0073]

[0074] In the formula, Recall i represents the historical disaster recall rate corresponding to the i-th slope segment cluster; TP i FN represents the number of events correctly predicted as disasters corresponding to the i-th slope segment cluster; i Represents the number of disasters that actually occurred and were not predicted.

[0075] Here, the recall rate is the ratio of the number of real disaster events detected by the system to the total number of disaster events that actually occurred. The historical disaster recall rate of the risk cluster slope section is calculated by the above formula (8) and compared with the preset disaster recall rate threshold. If the historical disaster recall rate is greater than the critical disaster recall rate threshold, it means that there is no need to analyze and warn the risk of such ordinary slopes; in addition, if the historical disaster recall rate is greater than the critical disaster recall rate threshold, it means that slope risk monitoring, analysis and warning should be carried out for this cluster slope section.

[0076] Figure 2 A structural block diagram of an example of a slope risk monitoring model according to an embodiment of the present invention is shown.

[0077] like Figure 2 As shown, the slope risk monitoring model adopts a convolutional long short-term memory network (ConvLSTM), which includes a first input layer 211, a second input layer 212, a third input layer 213, a fourth input layer 214, a spatial convolution layer 221, a preprocessing layer 222, a feature fusion layer 230, a spatial pyramid pooling layer 240, a ConvLSTM layer 250 and an output layer 260.

[0078] It should be noted that ConvLSTM introduces the convolution structure into LSTM, so that it can not only handle the long-term dependency problem of time series data, but also handle the spatial characteristics of input data. In particular, various types of slope monitoring data (such as InSAR images, GNSS parameters, underground stress distribution maps, etc.) have obvious spatial structural characteristics. ConvLSTM can effectively process and analyze spatiotemporal data sequences, which is particularly critical for monitoring and predicting the stability of mine slopes with complex spatial characteristics and time dependence.

[0079] The first input layer 211 is used to obtain InSAR image time series data in the slope monitoring time series data.

[0080] The second input layer 212 is used to obtain the GNSS parameter time series data in the slope monitoring time series data.

[0081] The third input layer 213 is used to obtain underground stress map time series data in the slope monitoring time series data.

[0082] The fourth input layer 214 is used to obtain the meteorological parameter sequence in the slope monitoring time series data.

[0083] The spatial convolution layer 221 is used to extract spatial features from each initial time series data, and use the spatial features to update the initial time series data to obtain corresponding time series-space features. The initial time series data includes InSAR image time series data, GNSS parameter time series data or underground stress map time series data.

[0084] Therefore, spatial features in each time series data are extracted through spatial convolution, so as to supplement the spatial features in the time series data and obtain corresponding spatiotemporal features.

[0085] More specifically, the structure of the spatial convolution layer 221 is:

[0086] S type =ReLU(W conv *X type +b conv ), Formula (9)

[0087] In the formula, X type Represents any type of initial time series data X InSAR ,X GNSS ,X stress , X InSAR ,X GNSS and X stress They represent the InSAR image time series data, GNSS parameter time series data and underground stress map time series data respectively; S type W represents the temporal-spatial features output by the spatial convolutional layer for any type of initial temporal data; convrepresents the convolution kernel weight of the spatial convolution layer, b conv Represents the bias term of the spatial convolution layer, * represents the convolution operation, and ReLU represents the RELU activation function.

[0088] Based on the preprocessing layer 222, the meteorological parameter sequence is normalized and standardized to obtain corresponding meteorological time series features.

[0089] More specifically, the structure of the pretreatment layer 222 is:

[0090]

[0091] In the formula, X weather represents the meteorological parameter sequence, μ weather and σ weather Respectively represent the mean and standard deviation corresponding to the meteorological data, P weather Represents the meteorological time series characteristics output by the preprocessing layer.

[0092] Based on the feature fusion layer 230, each time series-space feature and the meteorological time series feature are fused to determine the corresponding meteorological-time series-space feature.

[0093] More specifically, the structure of the feature fusion layer 230 is:

[0094] F combined =∑ type α type ·S type +α weather ·P weather , Formula (11)

[0095] In the formula, F combined represents the meteorological-temporal-spatial features determined by the feature fusion layer through the feature fusion operation, α type represents the feature adjustment weights for various types of initial time series data, α weather represents the feature adjustment weight for meteorological time series features, which is dynamically generated by the weight generation network. type Represents the time-space features output by the spatial convolution layer for any type of initial time series data.

[0096] Therefore, the importance of different features is learned through the weight generation network, and the weighted feature fusion strategy is used to dynamically adjust the contribution of features from different sources using the learned weights.

[0097] In some embodiments, the weight generation network may use a small neural network layer to calculate the weight α of each time series data feature. type and α weather, which can adopt a fully connected layer structure, and its input is the feature S obtained from the convolution layer or other feature processing layer type and P weather , the output is the weight score of this type of feature.

[0098] Let W fusion1 ,W fusion2 is the weight of the first and second fully connected layers, b fusion1 ,b fusion2 These parameters are learned through the training process so as to most effectively combine the contributions of different features to the final prediction results.

[0099] More specifically, for each feature, its weight score is calculated by the following formula:

[0100] z type =W fusion1 ·S type +b fusion1 , Formula (12)

[0101] z weather =W fusion2 ·P weather +b fusion2 , Formula (13)

[0102] In the formula, z type represents the weight score of various spatiotemporal series features (corresponding to InSAR image time series data, GNSS parameter time series data and underground stress map time series data), z weather Represents the weight score of the meteorological time series feature. Here, the weight score is an unnormalized weight score, which indicates the importance of this feature alone to the prediction result without considering other features.

[0103] The softmax function is used to normalize the weight scores of all features, ensuring that the sum of all weights is 1, so that they can be interpreted as probabilities or relative importance:

[0104]

[0105] Where g traverses all feature types (InSAR features, GNSS features, subsurface stress features, and meteorological features) to ensure that the weight of each feature is automatically adjusted based on its importance relative to other features, and z g Represents the weight scores of all feature categories; and Respectively represent z type 、z weather and z g The result of performing an exponential operation.

[0106] Therefore, the feature fusion layer does not simply stack or concatenate features, but intelligently adjusts the contribution of each feature based on the current data and training results. The above dynamic adjustment makes the model more flexible and adaptable, especially when processing multi-source data, it can more effectively utilize various information and improve the accuracy and robustness of prediction.

[0107] The multi-scale spatial feature information in the meteorological-temporal-spatial features is extracted based on the spatial pyramid pooling layer 240, and the meteorological-temporal-spatial features are updated to obtain the corresponding target meteorological-temporal-multi-scale spatial features.

[0108] Here, the spatial pyramid pooling layer can capture spatial information at different scales by performing pooling operations at multiple scales and then concatenating the pooling results to form a fixed-length output vector.

[0109] More specifically, the structure of the spatial pyramid pooling layer 240 is:

[0110] F spp =concat(P1(F combined ),P2(F combined ),…,P r (F combined )) , Formula (16)

[0111] Where P r Represents the r-th level pooling operation, which represents the maximum pooling operation of the corresponding level scale; concat represents the feature concatenation operation.

[0112] The meteorological-time series-multi-scale spatial features are processed based on the ConvLSTM layer 250 to obtain the corresponding time series relationship dependency features.

[0113] More specifically, the structure of the ConvLSTM layer 250 is:

[0114] i t =σ(W xi *F spp +W hi *H t-1 +b i ), Formula (17)

[0115] f t =σ(W xf *F spp +W hf *H t-1 +b f ), Formula (18)

[0116] o t =σ(W xo *Fspp +W ho *H t-1 +b o ), Formula (19)

[0117]

[0118] In the formula, i t ,f t ,o t They represent the input gate, forget gate and output gate respectively, and C t Indicates the unit status; W xi ,W xf ,W xo ,W xc and b i ,b f ,b o ,b c They represent the weight parameters in the corresponding model layer modules, σ represents the Sigmoid activation function; H t represents the hidden state, t represents the number of ConvLSTM layers, represents the Hadamard product, and tanh represents the hyperbolic tangent function.

[0119] Processing spatiotemporal series data through the ConvLSTM layer not only captures the spatial characteristics of a single time point, but also learns and memorizes long-term temporal dependencies through the time gating mechanism, which is particularly important for predicting slope stability based on time changes, such as coping with short-term geological changes caused by rainfall.

[0120] The temporal relationship dependency features are processed based on the output layer 260 to determine the risk probability for each preset slope risk type and obtain the corresponding slope risk identification result.

[0121] More specifically, the structure of the output layer 260 is:

[0122] score=W y ·H t +b y , Formula (22)

[0123]

[0124] Where W y represents the weight matrix with dimension [number of hidden layer dimensions, number of categories], b y represents a bias vector with dimension [number of categories], where the number of categories is the total number of types U corresponding to the preset slope risk types; score represents the original risk score generated for each preset slope risk type, Y risk Indicates the risk probability of the current slope for each preset slope risk type; represents the exponential function e applied to the original risk score of category i x The result after processing, Represents the sum of the results of indexing the raw risk scores of all categories.

[0125] Based on the specific structure of the above-mentioned slope risk monitoring model provided by the embodiment of the present invention, through the combined use of convolutional layers and spatial pyramid pooling layers, the model can capture spatial features from different scales and realize multi-level spatial feature extraction, which not only enhances the model's perception of terrain and geological changes, but also enables the model to better understand and process slope monitoring data with complex spatial structures, such as InSAR images and underground stress maps. The feature fusion layer dynamically combines information from different sources (such as meteorological data and geological monitoring data) through weighting strategies (such as attention mechanisms). This not only improves the flexibility and adaptability of the model, but also ensures that the model can pay attention to more critical information when making decisions, improving the accuracy and efficiency of predictions. Through the application of the SPP layer, the model can process and analyze data at multiple resolution levels, thereby providing a comprehensive perspective on the microscopic and macroscopic features of the slope, which is particularly critical for processing and analyzing geological changes caused by multiple factors, such as underground stress and surface movement at different depths. Through the softmax classification processing of the output layer, the model can not only predict the risk level of the slope, but also quantify the probability of different risk levels. By outputting the risk probability, risk management is more accurate, providing decision makers with quantitative risk assessment and supporting the formulation of more targeted preventive measures.

[0126] In some examples of the embodiments of the present invention, in the slope risk monitoring model, a loss function combined with multi-task learning can be used to not only classify different types of slope risks, but also evaluate the severity of each type of risk. More specifically, the loss function L of the slope risk monitoring model is:

[0127] L=λ1L class +λ2L severity , Formula (24)

[0128]

[0129] Where, L class represents the main classification loss (which uses multi-classification cross entropy loss), L severity represents the mean square error loss, λ1 and λ2 represent the corresponding loss adjustment weights, N represents the total number of samples in the data sample set, U represents the total number of categories corresponding to the preset slope risk type, and y i,c The label of the i-th sample indicates whether it belongs to category u, p i,c represents the probability that the model predicts that the i-th sample belongs to category u, si represents the true severity score, Represents the severity score predicted by the model.

[0130] Here, by adjusting the values ​​of λ1 and λ2, the classification accuracy or accurate prediction of risk severity can be prioritized according to actual application requirements and risk management policies. Therefore, by simultaneously performing type classification and severity assessment model losses, the model can more comprehensively understand the complexity of slope risks and improve the model's understanding of the various characteristics of slope risks, thereby improving the accuracy and timeliness of early warnings, reducing the probability of major slope safety accidents, and greatly reducing the impact of mine safety risk events.

[0131] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of combined actions, but those skilled in the art should be aware that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0132] like Figure 3 The figure is a schematic diagram of the structure of an early warning system for monitoring a mine slope provided by an embodiment of the present invention. The system can execute the early warning method for monitoring a mine slope of any of the above embodiments and is configured in a terminal.

[0133] A mine slope monitoring and early warning system 300 provided in this embodiment includes a data acquisition unit 310 , an attribute clustering unit 320 , a risk clustering identification unit 330 , a time series data determination unit 340 , a meteorological data acquisition unit 350 and a slope risk identification unit 360 .

[0134] The data acquisition unit 310 is configured to acquire a real-time slope monitoring data set based on slope monitoring equipment deployed in multiple key areas of mine slopes; the real-time slope monitoring data includes InSAR images, GNSS parameters and underground stress distribution maps.

[0135] The attribute clustering unit 320 is configured to determine the multi-attribute features corresponding to each real-time slope monitoring data in the real-time slope monitoring data set, and cluster analyze each multi-attribute feature using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters; the multi-attribute features include slope segment location attribute features, slope segment type attribute features and slope segment numerical attribute features.

[0136] The risk cluster identification unit 330 is configured to calculate the historical disaster recall rate of each slope segment cluster, and determine the slope segment cluster whose corresponding calculated historical disaster recall rate exceeds a preset disaster recall rate threshold in each slope segment cluster as the target risk slope segment cluster for early warning risk assessment.

[0137] The time series data determination unit 340 is configured to construct slope section monitoring time series data corresponding to a preset time period based on the real-time slope monitoring data corresponding to each target risk cluster and the historical slope monitoring data set matching the slope section position of the real-time slope monitoring data.

[0138] The meteorological data acquisition unit 350 is configured to acquire a meteorological parameter sequence corresponding to a preset time period for each key area of ​​the mine slope.

[0139] The slope risk identification unit 360 is configured to input the monitoring time series data of each slope section and the meteorological parameter sequence into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the warning condition. If the judgment result indicates yes, a warning signal is issued.

[0140] In some embodiments, the attribute clustering unit includes a first sub-determination unit and a second sub-determination unit, wherein the first sub-determination unit is used to determine the real-time slope monitoring data set X=(x1, x2, x3, ..., x n ) in each real-time slope monitoring data x i The corresponding object feature matrix [x i,1 ,x i,2 ,…,x i,e ,x i,e+1 ,…,x i,m ,x i,m+1 ,…,x i,p ], where the first e features are the type features of the slope position, the middle m features are the type features of the slope classification, and the last p features are the type features of the slope value; the second sub-determination unit is used to divide the real-time slope monitoring data set X into k clusters by minimizing the objective function of each attribute, and the objective function adopts the following formula: In the formula, is the fuzzy partitioning matrix on attribute f, satisfying the condition and is the cluster center prototype set on attribute f, is the data object x on the attribute f i , is the cluster center prototype of the jth cluster on attribute f; α (1<α<∞) represents the fuzzy coefficient, n represents the total number of objects in the data set, and k is the number of clusters into which the real-time data set is divided; For object x on attribute f i and prototype Q j of the measure of distinctiveness: Denote the a-th attribute of data object x on attribute f i , where i represents the index of the data object, Denote the cluster center of the numerical attribute on attribute f Denote the cluster center of the categorical attribute on attribute f. e is the number of slope segment position type features, i and j represent the indices of data objects, m is the number of slope segment classification type features. Rewrite the objective function into the following formula: where, Denote the central value of the j-th cluster corresponding to attribute 1 Denote the sum of the fuzzy weights of the j-th cluster on attribute 1 is the cluster center prototype of the z-th cluster on attribute f. n represents the total number of objects in the dataset, e is the number of slope segment position type features, m is the number of slope segment classification type features, k is the number of clusters into which the real-time dataset is partitioned; Through continuous iteration, move the center point of the class to the average position of the included data, and then re-partition the internal data until the distance from all data to the cluster center of the class to which they belong is the closest, so as to classify the real-time slope monitoring dataset X into k (k < n) slope segment clusters S=(S1, S1, S1, …, S k ): In the formula, μ i Denote the mean of slope segment cluster S i , VarS i Denote the sum of the distances between all features in the cluster of S i and the mean. x represents any item in slope segment cluster S i , argmin represents the independent variable value when the function obtains the minimum value within its domain of definition.

[0141] In some embodiments, the risk cluster identification unit includes a first sub-acquisition unit and a third sub-determination unit. Among them, the first sub-acquisition unit is used to acquire the historical disaster analysis data corresponding to each slope segment cluster, and the historical disaster analysis data includes the actual historical disaster events and disaster historical prediction results of the slope segment positions corresponding to each slope segment cluster; The third sub-determination unit is used to calculate the historical disaster recall rate of each slope segment cluster through the following formula: In the formula, Recall i Denote the historical disaster recall rate corresponding to the i-th slope segment cluster; TP i Denote the number of events correctly predicted as disasters corresponding to the i-th slope segment cluster; FN i Denote the number of events that are actually disasters and not predicted.

[0142] In some embodiments, an embodiment of the present invention provides a computer-readable storage medium, which stores one or more programs including execution instructions. The execution instructions can be read and executed by electronic devices (including but not limited to computers, servers, or network devices, etc.) to execute the above-mentioned early warning method for mine slope monitoring of the present invention.

[0143] In some embodiments, an embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the above-mentioned early warning method for mine slope monitoring.

[0144] In some embodiments, an embodiment of the present invention also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute an early warning method for mine slope monitoring.

[0145] Figure 4 FIG. 1 is a schematic diagram of the hardware structure of an electronic device for executing an early warning method for monitoring a mine slope provided by another embodiment of the present invention. Figure 4 As shown, the device includes:

[0146] One or more processors 410 and memory 420, Figure 4 A processor 410 is taken as an example.

[0147] The device for executing the early warning method for mine slope monitoring may further include: an input device 430 and an output device 440 .

[0148] The processor 410, the memory 420, the input device 430 and the output device 440 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0149] The memory 420, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the early warning method for monitoring a mine slope in the embodiment of the present invention. The processor 410 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 420, that is, realizing the early warning method for monitoring a mine slope in the above method embodiment.

[0150] The memory 420 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 420 may optionally include a memory remotely arranged relative to the processor 410, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0151] The input device 430 may receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 440 may include a display device such as a display screen.

[0152] One or more modules are stored in the memory 420, and when executed by one or more processors 410, the early warning method for mine slope monitoring in any of the above method embodiments is executed.

[0153] The above product can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0154] The electronic device of the embodiment of the present invention exists in various forms, including but not limited to:

[0155] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0156] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0157] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0158] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0159] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0160] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An early warning method for mine slope monitoring, characterized in that: The method comprises: Based on the slope monitoring equipment deployed in key areas of multiple mine slopes, a real-time slope monitoring data set is obtained, wherein the real-time slope monitoring data includes InSAR images, GNSS parameters and underground stress distribution maps; Determine the multi-attribute features corresponding to each of the real-time slope monitoring data in the real-time slope monitoring data set, and perform cluster analysis on each of the multi-attribute features using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters, wherein the multi-attribute features include slope segment location attribute features, slope segment type attribute features, and slope segment numerical attribute features; Calculating the historical disaster recall rate of each of the slope segment clusters, and determining the slope segment clusters whose corresponding calculated historical disaster recall rate exceeds a preset disaster recall rate threshold value in each of the slope segment clusters as target risk slope segment clusters to be subjected to early warning risk assessment; For the real-time slope monitoring data corresponding to each of the target risk slope segment clusters, construct slope segment monitoring time series data corresponding to a preset time period based on the real-time slope monitoring data and a historical slope monitoring data set matching the slope segment position of the real-time slope monitoring data; Obtaining a meteorological parameter sequence corresponding to the preset time period for each of the key areas of the mine slope; The monitoring time series data of each slope section and the meteorological parameter sequence are input into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the warning condition. When the judgment result indicates yes, a warning signal is issued.

2. The method according to claim 1, characterized in that The determining of the multi-attribute features corresponding to each of the real-time slope monitoring data in the real-time slope monitoring data set, and clustering analysis of each of the multi-attribute features using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters, includes: Determine the real-time slope monitoring data set X = (x1, x2, x3, ..., x n ) in each of the real-time slope monitoring data x i The corresponding object feature matrix [x i,1 ,x i,2 ,…,x i,e ,x i,e+1 ,…,x i,m ,x i,m+1 ,…,x i,p ], where the first e features are the type features of the slope position, the middle m features are the type features of the slope classification, and the last p features are the type features of the slope value; The real-time slope monitoring dataset X is divided into k clusters by minimizing the objective function of each attribute. The objective function adopts the following formula: In the formula, is the fuzzy partitioning matrix on attribute f, satisfying the condition and is the prototype set of cluster centers on attribute f; is the data object x on the attribute f i ; is the cluster center prototype of the jth cluster on attribute f; α (1<α<∞) represents the fuzzy coefficient, n represents the total number of objects in the data set, and k is the number of clusters into which the real-time data set is divided; is the property f on object x i With prototype Q j The mutual dissimilarity measure of is: Represents the data object x on attribute f i The ath attribute of represents the cluster center of the numerical attribute on attribute f, represents the cluster center of the classification attribute on attribute f, e is the number of slope position type features, i and j represent the index of the data object, m is the number of slope classification type features, and the objective function is rewritten as the following formula: in, Indicates that attribute 1 corresponds to the center value of the jth cluster, represents the sum of the fuzzy weights of the jth cluster on attribute 1, is the cluster center prototype of the zth cluster on attribute f, n represents the total number of objects in the dataset, e is the number of slope location type features, m is the number of slope classification type features, and k is the number of clusters into which the real-time dataset is divided; Through continuous iteration, move the center point of the class to the average position of the contained data, and then re-divide the internal data until the distance from all data to the cluster center of the class it belongs to is the closest, so as to classify the real-time slope monitoring data set X into k (k < n) slope segment clusters S = (S1, S1, S1, …, S k ): In the formula, μ i Represents the slope segment cluster S i The mean of x represents the slope segment cluster S i Any of VarS i Indicates S i The sum of the distances between all features in the cluster and the mean, argmin represents the value of the independent variable when the function achieves the minimum value within its domain.

3. The method according to claim 1, characterized in that The calculating of the historical disaster recall rate of each slope segment cluster includes: Acquire historical disaster analysis data corresponding to each of the slope segment clusters, wherein the historical disaster analysis data includes actual historical disaster events and historical disaster prediction results corresponding to the slope segment locations of each of the slope segment clusters; The historical disaster recall rate of each slope segment cluster is calculated by the following formula: In the formula, Recall i represents the historical disaster recall rate corresponding to the i-th slope segment cluster; TP i FN represents the number of events correctly predicted as disasters corresponding to the i-th slope segment cluster; i Represents the number of disasters that actually occurred and were not predicted.

4. The method according to any one of claims 1 to 3, characterized in that The slope risk monitoring model adopts a convolutional long short-term memory network, which includes a first input layer, a second input layer, a third input layer, a fourth input layer, a spatial convolution layer, a preprocessing layer, a feature fusion layer, a spatial pyramid pooling layer, a ConvLSTM layer and an output layer; The first input layer is used to obtain the InSAR image time series data in the slope monitoring time series data; The second input layer is used to obtain the GNSS parameter time series data in the slope monitoring time series data; The third input layer is used to obtain underground stress map time series data in the slope monitoring time series data; The fourth input layer is used to obtain the meteorological parameter sequence in the slope monitoring time series data; The spatial convolution layer is used to extract spatial features from each initial time series data, and update the initial time series data using the spatial features to obtain corresponding time series-space features, wherein the initial time series data includes the InSAR image time series data, the GNSS parameter time series data or the underground stress map time series data; Based on the preprocessing layer, the meteorological parameter sequence is normalized to obtain corresponding meteorological time series features; Based on the feature fusion layer, each of the time series-spatial features and the meteorological time series features are subjected to feature fusion to determine a corresponding meteorological-time series-spatial feature; Extracting multi-scale spatial feature information in the meteorological-temporal-spatial features based on the spatial pyramid pooling layer, and updating the meteorological-temporal-spatial features to obtain corresponding target meteorological-temporal-multi-scale spatial features; Processing the meteorological-time series-multi-scale spatial features based on the ConvLSTM layer to obtain corresponding time series relationship dependency features; The temporal relationship dependency features are processed based on the output layer to determine the risk probability for each of the preset slope risk types, and obtain the corresponding slope risk identification result.

5. The method according to claim 4, characterized in that The structure of the spatial convolutional layer is: S type =ReLU(W conv *X type +b conv ) Where, X type Represents any type of initial time series data X InSAR ,X GNSS ,X stress , X InSAR ,X GNSS and X stress respectively represent the InSAR image time series data, the GNSS parameter time series data and the underground stress map time series data; S type Represents the time-space features output by the spatial convolution layer for any type of the initial time series data; W conv represents the convolution kernel weight of the spatial convolution layer, b conv represents the bias term of the spatial convolutional layer, * represents the convolution operation, and ReLU represents the RELU activation function; The structure of the pretreatment layer is: Where, X weather represents the meteorological parameter sequence, μ weather and σ weather Respectively represent the mean and standard deviation corresponding to the meteorological data, P weather Represents the meteorological time series characteristics output by the preprocessing layer; The structure of the feature fusion layer is: In the formula, F combined represents the meteorological-temporal-spatial features determined by the feature fusion layer through the feature fusion operation; α type represents the feature adjustment weights for various types of the initial time series data, α weather represents the feature adjustment weight for the meteorological time series feature, which is dynamically generated by the weight generation network, S type Represents the time-space features output by the spatial convolution layer for any type of initial time series data; The structure of the spatial pyramid pooling layer is: F spp =concat(P1(F combined ),P2(F combined ),…,P r (F combined )) Where P r represents the rth level pooling operation, which represents the maximum pooling operation of the corresponding level scale, concat represents the feature concatenation operation, F combined represents the meteorological-temporal-spatial features determined by the feature fusion layer through the feature fusion operation; The structure of the ConvLSTM layer is: i t =σ(W xi *F spp +W hi *H t-1 +b i ) f t =σ(W xf *F spp +W hf *H t-1 +b f ) o t =σ(W xo *F spp +W ho *H t-1 +b o ) In the formula, i t ,f t ,o t They represent the input gate, forget gate and output gate respectively, and C t Indicates the unit state, H t represents the hidden state, W xi ,W xf ,W xo ,W xc and b i ,b f ,b o ,b c They represent the weight parameters in the corresponding model layer modules, σ represents the Sigmoid activation function, t represents the number of ConvLSTM layers, ° represents the Hadamard product, and tanh represents the hyperbolic tangent function; The structure of the output layer is: score=W y ·H t +b y Where W y represents the weight matrix with dimension [number of hidden layer dimensions, number of categories], b y represents a bias vector with dimension [number of categories], where the number of categories is the total number of types U corresponding to the preset slope risk types; score represents the original risk score generated for each of the preset slope risk types, Y risk Indicates the risk probability of the current slope for each of the preset slope risk types; represents the application of exponential function e to the original risk score of category i x The result after processing, It represents the sum total of the results after the original risk scores of all categories are indexed.

6. The method according to claim 5, characterized in that The loss function L of the slope risk monitoring model is: L=λ1L class +λ2L severity Where, L class represents the main classification loss, L severity represents the mean square error loss, λ1 and λ2 represent the corresponding loss adjustment weights, N represents the total number of samples in the data sample set, U represents the total number of categories corresponding to the preset slope risk type, y i,c The label of the i-th sample indicates whether it belongs to category u, p i,c represents the probability that the model predicts that the i-th sample belongs to category u, s i represents the true severity score, Represents the severity score predicted by the model.

7. An early warning system for monitoring mine slopes, characterized in that: The system comprises: A data acquisition unit is configured to acquire a real-time slope monitoring data set based on slope monitoring equipment deployed in a plurality of key areas of the mine slopes, wherein the real-time slope monitoring data includes InSAR images, GNSS parameters, and underground stress distribution maps; An attribute clustering unit is configured to determine the multi-attribute features corresponding to each of the real-time slope monitoring data in the real-time slope monitoring data set, and perform cluster analysis on each of the multi-attribute features using a multi-attribute fuzzy clustering algorithm based on mixed data to obtain a corresponding preset number of slope segment clusters, wherein the multi-attribute features include slope segment location attribute features, slope segment type attribute features, and slope segment numerical attribute features; A risk cluster identification unit is configured to calculate the historical disaster recall rate of each of the slope segment clusters, and determine the slope segment clusters whose corresponding calculated historical disaster recall rate exceeds a preset disaster recall rate threshold value in each of the slope segment clusters as target risk slope segment clusters to be subjected to early warning risk assessment; A time series data determination unit is configured to construct slope monitoring time series data corresponding to a preset time period according to the real-time slope monitoring data corresponding to each of the target risk slope segment clusters and a historical slope monitoring data set matching the slope segment position of the real-time slope monitoring data; A meteorological data acquisition unit is configured to acquire a meteorological parameter sequence corresponding to the preset time period for each of the key areas of the mine slope; The slope risk identification unit is configured to input the monitoring time series data of each slope section and the meteorological parameter sequence into the slope risk monitoring model to identify each preset slope risk type, obtain the corresponding slope risk identification result, and judge whether the risk identification result meets the warning condition. If the judgment result indicates yes, a warning signal is issued.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the early warning method for mine slope monitoring according to any one of claims 1 to 6.

9. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by the processor, the early warning method for monitoring mine slopes according to any one of claims 1 to 6 is implemented.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include an early warning method for executing mine slope monitoring as described in any one of claims 1 to 6.

Images