Strip mine slope risk prediction method and system based on multi-source data

Abstract

The invention discloses an open-pit mine slope risk prediction method and system based on multi-source data, and relates to the technical field of intelligent geological engineering, and the method comprises the steps: collecting geological data, displacement monitoring data and environmental activity data of an open-pit mine slope, carrying out the preprocessing, and constructing a three-dimensional digital twinborn body; carrying out fusion calculation on the preprocessed displacement monitoring data and environmental activity data through a physical information neural network, predicting a displacement value and a stress value of the strip mine slope, and obtaining full-field physical quantity prediction data; according to the strip mine slope slip crack surface damage evolution path, the stability risk levels of the whole strip mine slope and different areas in the strip mine slope are judged, high-risk areas are positioned, and a strip mine slope risk prediction report is generated. According to the method, the damage evolution path of the slip crack surface of the strip mine slope is dynamically deduced, so that space-time continuous modeling of the damage evolution process, early recognition of the slope instability precursor and accurate insight of the risk evolution trend are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent geological engineering, and particularly relates to an open-pit mine slope risk prediction method and system based on multi-source data. BACKGROUND

[0002] Open-pit mine slope stability monitoring and risk prediction is a core link of mine safety production. At present, the field generally adopts an analysis method based on multi-source monitoring data, continuously collects slope surface displacement, deep deformation and vibration signal data through global navigation satellite system (GNSS) receivers, inclinometers and microseismic monitoring network sensors, and comprehensively judges in combination with rock mechanics parameters and a three-dimensional geological model obtained through geological exploration; a conventional method generally relies on a geographic information system (GIS) platform to integrate multi-source data, and sets a fixed threshold to perform early warning on a single or multiple monitoring indexes.

[0003] However, the existing method still has two limitations. Multi-source data lacks deep coupling and dynamic correlation analysis, and it is difficult to effectively mine the internal causal relationship of displacement, microseismic and hydrological monitoring data in the time and space dimensions, resulting in insufficient identification sensitivity and accuracy of risk precursor information; the existing numerical simulation method relies on a static constitutive relation, and cannot continuously deduce the slope damage accumulation and crack expansion evolution process, so that the prediction result lags behind the actual deformation development. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an open-pit mine slope risk prediction method based on multi-source data to solve the problems of insufficient early warning sensitivity and lagging prediction result.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a method for predicting the risk of an open-pit mine slope based on multi-source data, which comprises collecting and preprocessing geological data, displacement monitoring data and environmental activity data of the open-pit mine slope, and constructing a three-dimensional digital twin; performing fusion calculation on the preprocessed displacement monitoring data and environmental activity data through a physical information neural network, predicting the displacement value and stress value of the open-pit mine slope, and obtaining full-field physical quantity prediction data; mapping the displacement value and stress value in the full-field physical quantity prediction data to the spatial nodes of the three-dimensional digital twin, calculating the node damage index according to the full-field stress value, connecting the edges according to the spatial adjacency relationship of the nodes, and constructing a damage feature space-time graph sequence; inputting the damage feature space-time graph sequence into a space-time graph neural network, capturing the propagation law of the node damage index between the spatial nodes, and dynamically deducing the damage evolution path of the open-pit mine slope slip surface; determining the stability risk level of the overall and internal different regions of the open-pit mine slope and locating the high-risk area according to the damage evolution path of the open-pit mine slope slip surface, and generating a risk prediction report of the open-pit mine slope.

[0008] As a preferred scheme of the method for predicting the risk of an open-pit mine slope based on multi-source data, the geological data comprises the cohesion value, internal friction angle value, elastic modulus value of the rock-soil body, and the spatial coordinate point set of the rock layer and joint fissure.

[0009] The displacement monitoring data comprises surface displacement, deep displacement and deformation field data.

[0010] The environmental activity data comprises rainfall and blasting parameters.

[0011] The preprocessing comprises data cleaning, space-time alignment and normalization.

[0012] As a preferred scheme of the method for predicting the risk of an open-pit mine slope based on multi-source data, the construction of the three-dimensional digital twin is as follows.

[0013] The spatial coordinate point set of the rock layer and joint fissure in the preprocessed geological data is subjected to spatial interpolation operation through the Kriging spatial interpolation algorithm to generate a three-dimensional spatial framework of the geological body structure.

[0014] The cohesion value, internal friction angle value and elastic modulus value of the rock-soil body in the preprocessed geological data are taken as material properties and assigned to the corresponding rock layer spatial region in the three-dimensional spatial framework of the geological body structure to form a mechanical characterization body of the open-pit mine slope.

[0015] The preprocessed displacement monitoring data and environmental activity data are mapped to the mechanical characterization body of the open-pit mine slope to generate a three-dimensional digital twin.

[0016] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data provided by the application, wherein: the displacement monitoring data and the environmental activity data after preprocessing are aligned and spliced to generate a multi-dimensional feature vector;

[0017] The displacement monitoring data and the external environmental activity data after preprocessing are aligned and spliced to generate a multi-dimensional feature vector;

[0018] The multi-dimensional feature vector is input into the physical information neural network for forward propagation calculation, and the displacement value and stress value of the future open-pit mine slope rock-soil body are output;

[0019] The displacement value and stress value of the future open-pit mine slope rock-soil body are subjected to physical consistency verification and smoothing processing, and a full-field physical quantity prediction data set is generated.

[0020] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data provided by the application, wherein: the displacement value and stress value in the full-field physical quantity prediction data are mapped to the spatial nodes of the three-dimensional digital twin, and the node damage index is calculated according to the full-field stress value, and the edges are connected according to the spatial adjacency relationship of the nodes to construct a damage feature space-time graph sequence, and the steps are as follows,

[0021] The displacement value and stress value in the full-field physical quantity prediction data are mapped to the corresponding spatial nodes in the three-dimensional digital twin according to the spatial coordinates, and a node physical state data set is obtained;

[0022] According to the Mises yield criterion, the stress value of each spatial node is calculated to obtain the node damage index;

[0023] According to the adjacency relationship of the spatial nodes, all spatial nodes carrying the node damage index are connected by edges to obtain a static damage feature graph of the open-pit mine slope;

[0024] The static damage feature graphs at different time points are connected in time sequence to generate a damage feature space-time graph sequence.

[0025] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data provided by the application, wherein: the damage feature space-time graph sequence is input into the space-time graph neural network to capture the propagation law of the node damage index between the spatial nodes, and the steps are as follows,

[0026] The damage feature space-time graph sequence is input into the space-time graph neural network, and the connection relationship between each node and the neighbor node is calculated by the graph attention network layer to generate an edge weight matrix;

[0027] Based on the edge weight matrix, the damage indicators of the neighbor nodes of each node are weighted and summed and transformed to generate a node enhanced spatial feature sequence.

[0028] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data, the dynamic deduction of the open-pit mine slope sliding surface damage evolution path comprises the following steps,

[0029] The node enhanced spatial feature sequence is input into the time convolution layer of the space-time graph neural network to learn the trend of the node damage indicator evolution over time, and the node damage indicator prediction value is output through the full connection layer regression prediction;

[0030] The node damage indicator prediction value is mapped back to the spatial position of the three-dimensional digital twin, and the node damage indicator prediction value is threshold filtered and spatially clustered to form an initial sliding surface morphology;

[0031] The initial sliding surface morphology is filtered for noise and filled for cavities through a three-dimensional connected component analysis algorithm to generate an optimized three-dimensional morphology of the sliding surface;

[0032] The optimized three-dimensional morphologies of the sliding surface at different time points are connected in chronological order to form the open-pit mine slope sliding surface damage evolution path.

[0033] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data, the determination of the stability risk level of the open-pit mine slope as a whole and different regions inside the slope comprises the following steps,

[0034] According to the open-pit mine slope sliding surface damage evolution path, the expansion speed, area change rate and maximum depth of the sliding surface are calculated to obtain an evolution intensity index set;

[0035] The evolution intensity index set is weighted and fused and threshold judged through a preset comprehensive risk assessment rule to obtain the open-pit mine slope stability risk level result.

[0036] As a preferred scheme of the open-pit mine slope risk prediction method based on multi-source data, the positioning of the high-risk area and the generation of the open-pit mine slope risk prediction report comprise the following steps,

[0037] According to the expansion speed and the open-pit mine slope stability risk level result, the high-risk area exceeding the preset local speed threshold is positioned in the three-dimensional digital twin to generate a high-risk area spatial coordinate set;

[0038] The high-risk area spatial coordinate set, the open-pit mine slope stability risk level result and the evolution intensity index set are integrated to generate the open-pit mine slope risk prediction report.

[0039] In a second aspect, the present application provides a strip mine slope risk prediction system based on multi-source data, comprising a data acquisition module, a physical quantity prediction module, a damage space-time graph construction module, a damage evolution path deduction module, and a risk judgment module.

[0040] The data acquisition module is used to acquire geological data, displacement monitoring data and environmental activity data of the strip mine slope and perform preprocessing, and construct a three-dimensional digital twin.

[0041] The physical quantity prediction module is used to perform fusion calculation on the preprocessed displacement monitoring data and environmental activity data through a physical information neural network, predict the displacement value and stress value of the strip mine slope, and obtain full-field physical quantity prediction data.

[0042] The damage space-time graph construction module is used to map the displacement value and stress value in the full-field physical quantity prediction data to the spatial nodes of the three-dimensional digital twin, calculate the node damage index according to the full-field stress value, connect the edges according to the spatial adjacency relationship of the nodes, and construct a damage feature space-time graph sequence.

[0043] The damage evolution path deduction module is used to input the damage feature space-time graph sequence into a space-time graph neural network, capture the propagation law of the node damage index among the spatial nodes, and dynamically deduce the damage evolution path of the strip mine slope slip surface.

[0044] The risk judgment module is used to determine the stability risk level of the whole and different regions inside the strip mine slope according to the damage evolution path of the strip mine slope slip surface, locate the high-risk region, and generate a strip mine slope risk prediction report.

[0045] The present application has the following advantages: by predicting the displacement value and stress value of the strip mine slope, obtaining full-field physical quantity prediction data, realizing the collaborative fusion of multi-source monitoring information and geotechnical mechanics physical laws, and improving the reliability and generalization ability of slope deformation and stress response prediction, the damage evolution process is modeled in space-time continuously, and early identification of slope instability precursors and accurate insight into risk evolution trends are realized. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0047] Fig. 1 The flowchart of the strip mine slope risk prediction method based on multi-source data.

[0048] Fig. 2 A schematic diagram of a strip mine slope risk prediction system based on multi-source data.

[0049] Fig. 3 A flowchart of dynamically deducing a damage evolution path of a strip mine slope sliding surface.

[0050] Fig. 4 A flowchart of generating a strip mine slope risk prediction report. DETAILED DESCRIPTION

[0051] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0052] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0053] Secondly, the "one embodiment" or "embodiment" referred to herein means that a specific feature, structure or characteristic can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0054] REFERENCE Figs. 1-4 For one embodiment of the present application, the embodiment provides a strip mine slope risk prediction method based on multi-source data, comprising the following steps:

[0055] S1, collecting geological data, displacement monitoring data and environmental activity data of a strip mine slope and preprocessing, and constructing a three-dimensional digital twin;

[0056] The geological data includes cohesion value, internal friction angle value, elastic modulus value of rock-soil mass, and spatial coordinate point set of rock stratum and joint fissure;

[0057] It should be noted that the cohesion value and internal friction angle value of the rock-soil mass are parameters directly measured by mechanical testing of the rock-soil sample collected on site through indoor direct shear test, which represent the shear strength of the rock-soil mass; the elastic modulus value is a parameter calculated by measuring the stress-strain ratio of the rock-soil mass under stress through indoor uniaxial compression test, which represents the deformation characteristic; the spatial coordinate point set of the rock stratum and joint fissure is three-dimensional position data used to control the spatial form of the geological interface, which is obtained on the actual slope body through field geological surveying and drilling exploration.

[0058] The displacement monitoring data includes surface displacement, deep displacement and deformation field data;

[0059] It should be noted that the ground displacement is obtained by continuously measuring the three-dimensional coordinate changes of the open-pit slope through the GNSS receiver arranged on the ground surface of the open-pit slope; the deep displacement is obtained by measuring the horizontal offset along the depth point by point through the inclinometer probe installed in the drill hole; and the deformation field data is the large-scale millimeter-level deformation information of the slope surface obtained by solving the differential interference processing of the interferometric radar image obtained by the synthetic aperture radar satellite.

[0060] The environmental activity data includes rainfall and blasting parameters;

[0061] It should be noted that the rainfall is the precipitation data in a unit of time directly measured by the rain gauge sensor of the automatic weather station arranged in the mining area; and the blasting parameters are the total weight of explosives, the three-dimensional coordinates of the blast center and the peak vibration speed data of each blasting obtained by referring to the design records of mining blasting operation and the on-site vibration monitoring report.

[0062] The preprocessing includes data cleaning, time-space alignment and normalization;

[0063] It should be noted that the data cleaning refers to eliminating gross errors and data jumps, ensuring data sequence continuity and reliability by removing outliers based on statistical distribution (such as 3σ criterion) and filling missing values by spline interpolation method based on adjacent data points; the time-space alignment refers to unifying the geological data, displacement monitoring data and environmental activity data from different devices with different collection time points and coordinate systems to the same time interval through time stamp resampling, and to the local coordinate system of the open-pit mine area through coordinate conversion; and the normalization refers to linearly transforming the geological data, displacement monitoring data and environmental activity data with large differences in dimension and numerical range to the [0, 1] interval through the min-max scaling method, so as to eliminate the feature magnitude difference.

[0064] The spatial coordinate point set of the rock stratum and joint fissure in the preprocessed geological data is subjected to spatial interpolation operation through the Kriging spatial interpolation algorithm, so as to generate a three-dimensional spatial framework of the geological body structure;

[0065] Further, traverse all points in the spatial coordinate point set of the rock stratum and joint fissure in the preprocessed geological data, calculate the Euclidean distance and attribute value difference between any two points according to the principle of spatial statistics, take the Euclidean distance between any two points as the abscissa and the corresponding attribute value difference as the ordinate to form a two-dimensional data point, visualize the two-dimensional data point using a statistical plotting algorithm, and draw a variogram scatter plot of the experimental variogram; compare and fit the experimental variogram with the theoretical variogram (including spherical function, exponential function basic structure), adjust the three parameters of the range, base value and nugget value to make the curve of the theoretical variogram best match the variogram scatter plot of the experimental variogram, generate the parameters of the fitted theoretical variogram, and according to the parameters of the fitted theoretical variogram, establish a Kriging linear equation set containing all known spatial coordinate point sets in the effective neighborhood of each estimated node in the regular three-dimensional grid, and obtain the optimal weight distribution of each known spatial coordinate point set by solving the weight coefficient matrix in the Kriging equation set; perform a weighted average on the attribute values of the known spatial coordinate point set to calculate the spatial coordinate value of each estimated node, and connect and triangulate the spatial coordinate values of all nodes to generate a continuous geological body structure three-dimensional space framework.

[0066] The cohesion value, internal friction angle value and elastic modulus value of the rock-soil mass in the preprocessed geological data are taken as material attributes and assigned to the corresponding rock stratum space region in the geological body structure three-dimensional space framework to form an open pit slope mechanical representation body.

[0067] Further, a correspondence table of rock stratum identification and rock-soil mass mechanical parameters is established according to the geological survey results to clearly define the cohesion value, internal friction angle value and elastic modulus value corresponding to different rock stratum regions; each spatial position point in the geological body structure three-dimensional space framework is traversed, the rock stratum type to which each spatial position point belongs is determined through the positional relationship between the spatial position point and the rock stratum boundary (if the three-dimensional coordinates (x1, y1, z1) of the spatial position point fall within the three-dimensional space region A circumscribed by the rock stratum boundary coordinates, the spatial position point is determined to belong to rock stratum type A), and the cohesion value, internal friction angle value and elastic modulus value of the rock stratum are retrieved from the correspondence table according to the rock stratum type to obtain the material attribute parameters of the corresponding spatial position point; all spatial position points with assigned material attributes are integrated to form an open pit slope mechanical representation body.

[0068] It should be noted that the open pit slope mechanical representation body refers to a digital reconstruction entity of the open pit slope engineering geological body in three-dimensional space, which accurately presents the geometric shape and internal rock stratum structure distribution of the open pit slope, and completely maps the mechanical attribute parameters (cohesion value, internal friction angle value and elastic modulus value) of the rock-soil mass as material attributes to the corresponding rock stratum region in three-dimensional space to form a comprehensive data body with both geometric information and physical mechanical characteristics.

[0069] mapping the pre-processed displacement monitoring data and environmental activity data to the open-pit slope mechanical representation, to generate a three-dimensional digital twin;

[0070] Further, read the spatial coordinate information of each monitoring point in the pre-processed displacement monitoring data (such as the three-dimensional geodetic coordinates of GNSS monitoring points), and accurately match the spatial coordinates of each monitoring point to the corresponding spatial position of the open-pit slope mechanical representation through coordinate transformation algorithm (such as seven-parameter Bursa transformation); analyze the time stamp sequence of the pre-processed environmental activity data, and align it with the time axis of the pre-processed displacement monitoring data; dynamically assign the surface displacement, deep displacement and deformation field data in the pre-processed displacement monitoring data to the matched spatial position points in the open-pit slope mechanical representation, and real-time update the spatial coordinate state of the matched spatial position points; convert the rainfall and blasting parameters in the pre-processed environmental activity data into equivalent loads, and apply them to the corresponding regions of the open-pit slope mechanical representation according to the load position distribution, to generate a three-dimensional digital twin that is in real-time linkage with the open-pit slope.

[0071] It should be noted that the three-dimensional digital twin is a virtual entity that completely corresponds to the physical slope, and is based on the open-pit slope mechanical representation containing geometric shape and mechanical properties, through real-time access to pre-processed displacement monitoring data to update spatial position state, and fusion of pre-processed environmental activity data as external action condition, to form a digital instance that can synchronously reflect the actual condition, dynamic change process and mechanical behavior of the physical slope.

[0072] S2, fusion calculation of the pre-processed displacement monitoring data and environmental activity data through physical information neural network, to predict the displacement value and stress value of the open-pit slope, and obtain full-field physical quantity prediction data;

[0073] align and splice the pre-processed displacement monitoring data and external environmental activity data to generate a multi-dimensional feature vector;

[0074] Further, a unified time reference is established to align the time stamp of the time sequence of the surface displacement, deep displacement and deformation field data in the pre-processed displacement monitoring data with the time sequence of the rainfall, blasting parameter in the pre-processed external environmental activity data, for each aligned time point, extract the displacement component of the surface displacement, the offset of the deep displacement, the deformation of the deformation field data, the millimeter of the rainfall, the total weight of the explosive and the peak vibration velocity of the blasting parameter to splice into a one-dimensional array, and perform standardization processing on the spliced one-dimensional array to eliminate the influence of different physical quantities, to generate a multi-dimensional feature vector.

[0075] The multi-dimensional feature vector is input into the physical information neural network for forward propagation calculation, and displacement values and stress values of the future open-pit slope rock-soil mass are output.

[0076] Further, the multi-dimensional feature vector is input into the physical information neural network, and data is transmitted layer by layer according to the hierarchical structure: the input layer receives the multi-dimensional feature vector and distributes it to the first hidden layer, the first hidden layer performs weight matrix multiplication operation on the multi-dimensional feature vector, and after adding the bias vector, generates the first layer feature output through the ReLU activation function; each subsequent hidden layer performs the same linear transformation and nonlinear activation operation on the output result of the previous layer, and gradually extracts and converts the feature output of the previous layer; the final output layer performs linear transformation on the output of the last hidden layer to generate displacement values and stress values of each spatial position point of the future time period open-pit slope rock-soil mass.

[0077] It should be noted that the pre-training process of the physical information neural network is based on the feedforward neural network architecture, and in the specific operation, the preprocessed displacement monitoring data and external environmental activity data are cut into multiple training samples according to the time sequence, each sample contains a multi-dimensional feature vector of a historical time period as input, and an actual monitoring displacement value of a corresponding future time period; after initializing the weight and bias parameters of the feedforward neural network, the training sample is input into the physical information neural network to perform forward propagation calculation, and the predicted displacement value and the predicted stress value are output through the linear transformation and ReLU activation function of the hidden layer; the mean square error between the predicted displacement value and the actual monitoring displacement value is calculated as the data fitting loss, and the predicted displacement value and the predicted stress value are substituted into the geotechnical mechanics control equation to calculate the physical constraint residual as the physical loss; the data fitting loss and the physical loss are combined into the total loss, and the gradient descent algorithm is used to calculate the gradient of the total loss with respect to the parameters of each layer of the physical information neural network, and the weight and bias parameters of the physical information neural network are updated according to the gradient direction; the forward propagation, loss calculation and parameter update steps are repeated until the maximum training round (such as 5000 times) is reached, and the physical information neural network that can learn data features and comply with physical laws is obtained.

[0078] The displacement values and stress values of the future open-pit slope rock-soil mass are physically consistent and smoothed, and the full-field physical quantity prediction data set is integrated and generated;

[0079] Further, the displacement value and the stress value of the future open-pit mine slope rock-soil mass are substituted into the rock-soil mechanics balance equation and the rock-soil constitutive relation equation, the numerical difference between the left and right terms of the equation is calculated respectively, and the balance equation residual and the constitutive relation residual at each spatial point are obtained; the balance equation residual and the constitutive relation residual are compared with the preset allowable error range, the spatial point position whose residual exceeds the limit is identified, and the displacement value and the stress value corresponding to the spatial point position are marked as not satisfying the physical constraint (if the balance equation residual of the spatial point position is 0.15 MPa, and the preset allowable error range is 0.1 MPa, then the displacement value and the stress value of the spatial point position will be marked as not satisfying the physical constraint); for the displacement value and the stress value that do not satisfy the physical constraint, a smoothing algorithm based on the least square principle is used to locally correct the displacement value and the stress value, so as to ensure that the displacement value and the stress value satisfy the continuity and balance conditions; the displacement value and the stress value that have passed the verification and smoothing processing are reorganized according to the spatial coordinate points and time steps, the corresponding spatial position coordinates and time stamp information are added, and the full-field physical quantity prediction data set is integrated and generated.

[0080] It should be noted that the rock-soil mechanics balance equation is a differential equation describing the balance relationship between the stress field and the external force in the rock-soil body, which is based on the statics balance condition and requires that the internal force and the external force on any cross section of the rock-soil body are balanced with each other. For example, when analyzing a circular arc-shaped sliding surface, the external forces acting on the sliding surface include its own gravity, external loads (such as slope top equipment) and possible water pressure, and the internal force is the shear resistance provided by the rock-soil body on the sliding surface. The statics balance condition requires that on any imaginary sliding surface element, the internal force component along the tangential direction of the sliding surface element must be equal in size and opposite in direction to the external force component in the same direction, otherwise the sliding surface element will slide. The rock-soil mechanics balance equation ensures that the stress field satisfies the mechanics balance law by establishing the differential relationship between the stress component and the volume force and surface force. The rock-soil constitutive relation equation is a mathematical relationship representing the corresponding law between the stress and strain of the rock-soil material, which is based on the deformation characteristics of the rock-soil body under different stress states. The rock-soil constitutive relation equation reflects the elastic and plastic mechanics behavior of the rock-soil material by defining the functional relationship between the stress tensor and the strain tensor. For example, when the stress state (represented by the stress tensor) of a certain point in the slope does not reach the yield condition, the rock-soil body behaves elastically, and the stress and strain (represented by the strain tensor) are linearly related, i.e. the deformation can be recovered after unloading. Once the stress combination satisfies the yield criterion, the material enters the plastic state and produces irreversible plastic deformation. The allowable error range is set based on the material properties of the rock-soil body, the accuracy of the monitoring data and the safety requirements of the project, and the exemplary value range is 1%-5% of the numerical value of the corresponding physical quantity.

[0081] S3, mapping the displacement value and the stress value in the full-field physical quantity prediction data to the spatial nodes of the three-dimensional digital twin, and calculating the node damage index according to the full-field stress value, connecting the edges according to the spatial adjacency relationship of the nodes, and constructing a damage characteristic space-time graph sequence;

[0082] The displacement value and the stress value in the full-field physical quantity prediction data are mapped to the corresponding spatial nodes in the three-dimensional digital twin according to the spatial coordinates, and a node physical state data set is obtained;

[0083] Further, the coordinates of all spatial nodes in the three-dimensional digital twin are traversed, for the coordinates of each spatial node, a record completely matching the spatial position coordinates is searched in the full-field physical quantity prediction data, and the contained displacement value and stress value are respectively assigned to the corresponding spatial nodes in the three-dimensional digital twin, so that each spatial node carries complete mechanical state information (including displacement value and stress value), the displacement value and stress value attached to each spatial node are collected, and a node physical state data set is formed.

[0084] According to the Mises yield criterion, the stress value of each spatial node is calculated to obtain the node damage index.

[0085] Further, the stress value corresponding to each spatial node is read from the node physical state data set, and substituted into the mathematical expression of the Mises yield criterion, and the second invariant of the stress value is calculated to determine the equivalent stress value; the equivalent stress value is taken as the node damage index representing the yield degree of the material.

[0086] The expression for calculating the node damage index is:

[0087] ;

[0088] wherein, is the node damage index (i.e. von Mises equivalent stress); is the first principal stress, which is the normal stress component of the stress value in the maximum principal direction, representing the maximum tensile stress of the material in the maximum principal direction; is the second principal stress, which is the normal stress component of the stress value in the intermediate principal direction, representing the stress state of the material in the intermediate principal direction; is the third principal stress, which is the normal stress component of the stress value in the minimum principal direction, representing the minimum tensile stress of the material in the minimum principal direction;

[0089] It should be noted that the Mises yield criterion (also known as the maximum distortion energy criterion) is a theory for determining whether a material has entered a plastic state from an elastic state, and is used to predict the yield behavior of ductile materials under complex stress states by converting the complex three-dimensional stress state into an equivalent stress (i.e., Mises equivalent stress) and comparing the equivalent stress with the uniaxial tensile yield strength to determine whether yielding has occurred.

[0090] According to the proximity relationship of the spatial nodes, all spatial nodes carrying the node damage indicators are connected by edges to obtain a static damage feature map of the open-pit slope;

[0091] Further, the Euclidean distances between all spatial nodes in the three-dimensional digital twin of the open-pit slope are calculated by the Euclidean distance calculation formula, and spatial node pairs with a distance less than a preset spatial adjacency distance threshold are marked as adjacent nodes. An undirected edge is established for each adjacent node pair, and the weight of the undirected edge is set to the inverse of the Euclidean distance between the nodes. All spatial nodes carrying the node damage indicators are taken as vertices, and the undirected edges are taken as connection relationships to form a static damage feature map of the open-pit slope.

[0092] It should be noted that the spatial adjacency distance threshold is set based on the geological structure characteristics of the rock mass of the open-pit slope and the predicted development scale of the potential slip surface. The exemplary value range is 5-20 meters. In order to ensure that the minimum spatial range of the key mechanical interaction within the rock mass can be captured, less than 5 meters can easily lead to insufficient node connection and fail to reflect the effective damage propagation path, and more than 20 meters can result in irrelevant long-range connection, leading to distortion of the damage propagation pattern.

[0093] The static damage feature maps at different time points are connected in time sequence to generate a damage feature space-time graph sequence;

[0094] Further, the static damage feature maps at multiple time points are sorted according to the chronological order of the time stamps, and the sorted static damage feature maps are arranged in sequence, with each static damage feature map serving as a graph structure snapshot at a time step. The node damage indicators in each static damage feature map at a time step are taken as the node features at the time step, and the graph structure (connection relationship between nodes and edges) between adjacent time steps is kept consistent. The static damage feature maps with uniform graph structure and time sequence node features are combined to generate a damage feature space-time graph sequence.

[0095] It should be noted that the damage feature space-time graph sequence is an integrated data composed of static damage feature maps arranged in time sequence, which records the damage indicator distribution and spatial connection relationship of all monitoring nodes in the three-dimensional space of the open-pit slope at a specific time point, and is used to upgrade the traditional isolated point monitoring or static spatial analysis to a dynamic analysis paradigm coupled in space and time.

[0096] S4, inputting the damage feature space-time graph sequence into a space-time graph neural network to capture the propagation law of the node damage index between the spatial nodes and dynamically deducing the damage evolution path of the open-pit mine slope sliding surface;

[0097] The damage feature space-time graph sequence is input into a space-time graph neural network (GAT), and the connection relationship between each node and the neighbor node is calculated by the graph attention network layer to generate an edge weight matrix.

[0098] Further, the static damage feature graph of each time step is extracted from the damage feature space-time graph sequence, and the node damage index carried by each node in the static damage feature graph is taken as the initial feature of the node. The two independent trainable query weight matrices and key weight matrices contained in the graph attention network layer of the space-time graph neural network are used to perform linear transformation on the initial features of the current node and the neighbor nodes respectively to generate corresponding query vectors and key vectors. The dot product operation is performed on the query vector of the current node and the key vectors of all neighbor nodes to obtain an attention score. The attention scores of all neighbor nodes of each node are normalized to obtain normalized attention coefficients, and the normalized attention coefficients between all nodes and neighbor nodes are organized into a matrix form according to the edge connection relationship to generate an edge weight matrix.

[0099] It should be noted that the space-time graph neural network is pre-trained, and in the specific operation, the historical displacement monitoring data, external environmental activity data and corresponding actual slope deformation observation records are taken as training samples, and the subsequent actual slope state changes are taken as verification samples. After initializing the weight parameters of the space-time graph neural network, the training samples are input into the space-time graph neural network for forward propagation: the graph attention network layer processes the static damage feature graph of each time step to capture the spatial dependence relationship, the time convolution layer processes the time sequence of the node features to capture the time evolution mode, and the final output layer generates the prediction result. The mean square error between the prediction result and the verification sample is taken as the loss function, the gradient of the loss function with respect to the weight of each layer of the network is calculated by the back propagation algorithm, and the weight parameters of the space-time graph neural network are updated using the gradient descent method. The forward propagation, loss calculation and parameter updating steps are repeated until the loss function converges to a stable value, and the trained space-time graph neural network is output.

[0100] It should be noted that the graph attention network layer is a neural network layer for processing graph structure data, and the principle is to dynamically learn the importance weight of the connection relationship between each node and the neighbor node in the graph through the attention mechanism.

[0101] Based on the edge weight matrix, the damage indexes of the neighbor nodes of each node are weighted and summed and transformed to generate a node enhanced spatial feature sequence.

[0102] Further, the normalized attention coefficients between each spatial node and all neighbor nodes are found from the edge weight matrix, the damage index values of each neighbor node are weighted and summed with the corresponding normalized attention coefficients, and the damage index value of the spatial node itself is spliced to generate an enhanced spatial feature of the spatial node at the current time step; the enhanced spatial features of all spatial nodes generated at each time step are arranged in time step order to form a node enhanced spatial feature sequence.

[0103] The node enhanced spatial feature sequence is input into the time convolution layer of the spatio-temporal graph neural network to learn the trend of the node damage index over time, and the node damage index prediction value is output through regression prediction by the fully connected layer;

[0104] Further, the node enhanced spatial feature sequence is input into the time convolution layer of the spatio-temporal graph neural network, the time convolution layer uses a one-dimensional convolution kernel to perform sliding window convolution operation on the node enhanced spatial feature sequence in the time dimension, captures the local mode and trend of the node damage index over time (for example, assuming that the sequence of the node damage index at the past 5 consecutive time steps is [0.2, 0.5, 0.8, 1.1, 1.4] MPa), the time convolution layer uses a one-dimensional convolution kernel with a width of 3 to slide and calculate on the sequence, first performs weighted summation on the sub-sequence [0.2, 0.5, 0.8] to obtain the first local feature, then the window slides, and the calculation is performed in turn, according to the calculation result, the stable rising trend (local mode and accelerating growth momentum (trend)) of the node damage index is identified, high-level time sequence features are obtained; the fully connected layer linearly combines the high-level time sequence features, maps the feature dimension to be consistent with the number of future time steps to be predicted (for example, if the node damage index at the future 10 time steps is to be predicted, the fully connected layer will transform the feature dimension to 10, each output node corresponds to a prediction value of a future time step), and the node damage index prediction value is obtained.

[0105] The node damage index prediction value is mapped back to the spatial position of the three-dimensional digital twin, and the node damage index prediction value is threshold filtered and spatially clustered to form an initial slip surface morphology.

[0106] Further, according to the spatial node corresponding to the node damage index prediction value, the three-dimensional coordinates of each spatial node are found and located in the three-dimensional digital twin, and the node damage index prediction value of each spatial node at a specific future time step is taken as attribute data and given to the coordinate position of the spatial node; the node damage index prediction value is filtered according to a preset damage risk threshold, and the spatial node corresponding to the node damage index prediction value higher than the damage risk threshold is marked as a high-risk node; based on the spatial coordinates of the high-risk node, a spatial clustering algorithm (such as DBSCAN) is used to group the high-risk nodes that are adjacent to each other in space, form multiple high-risk node clusters, and extract the external boundary point set according to the spatial coordinates of each high-risk node cluster and connect to form a closed polygon contour, outlining the spatial distribution range of the high-risk node cluster; the spatial distribution range of the high-risk node cluster is converted into a continuous three-dimensional grid surface by a triangular mesh generation algorithm (such as Delaunay triangulation), forming an initial slip surface morphology.

[0107] It should be noted that the damage risk threshold is set based on the yield strength of the rock-soil material and the engineering safety factor, and the exemplary value range is 0.7-0.9 times the yield strength of the material, in order to ensure that the risk state is identified before the material actually yields.

[0108] The initial slip surface morphology is filtered for noise and filled with holes by a three-dimensional connected component analysis algorithm to generate an optimized three-dimensional morphology of the slip surface;

[0109] Further, all spatial position points in the initial slip surface morphology are marked as foreground points, the initial slip surface morphology is scanned, and all sets of spatially connected foreground points are identified and marked, each set being a connected region; the number of foreground points contained in each connected region is calculated, and the connected region whose number of foreground points is less than a preset minimum connected region size threshold is determined as noise and deleted, and the connected region whose number of foreground points is greater than the minimum connected region size threshold is subjected to a morphological closing operation processing, a three-dimensional dilation operation is used to fill small holes in the region (for example, if there is a spherical hole with a diameter of 4mm in the initial slip surface morphology, after using a spherical structure element with a diameter of 5mm for dilation, the spherical hole will be completely filled), and a three-dimensional erosion operation is used to restore the approximate original boundary of the region (for example, after dilation, the slip surface boundary expands outward by about 5mm, and the original boundary profile can be basically restored by the same size of erosion operation, while the filled holes are retained); all connected regions after the morphological closing operation processing are merged to generate an optimized three-dimensional morphology of the slip surface.

[0110] It should be noted that the minimum connected region size threshold is set based on the typical size range (e.g., 5-20 spatial location points) of the cluster of isolated points caused by noise in the initial slip surface morphology, and the exemplary value range is 5-20 spatial location points, in order to filter out extremely small isolated noise caused by data fluctuations and avoid mistakenly deleting potential effective connected regions that may represent tiny crack initiation; the morphological closing operation processing is a basic morphological operation for image or three-dimensional body data processing, which is a combination of inflation operation followed by erosion operation, which can effectively smooth the object boundary, connect narrow broken places and fill small internal holes, and generate a more continuous and complete slip surface optimized three-dimensional morphology.

[0111] The slip surface optimized three-dimensional morphologies at different time points are connected in time sequence to form the open-pit slope slip surface damage evolution path.

[0112] Further, according to the time stamps corresponding to each slip surface optimized three-dimensional morphology, all slip surface optimized three-dimensional morphologies at different time points are sorted in chronological order. For the slip surface optimized three-dimensional morphologies at adjacent time points, an intermediate transition morphology is generated by linear interpolation between the corresponding spatial positions, realizing smooth transition of the morphology. The original slip surface optimized three-dimensional morphology and the smooth transition morphology are seamlessly spliced in time sequence to form the open-pit slope slip surface damage evolution path that continuously displays the whole process from initiation, expansion to complete penetration of the slip surface.

[0113] S5, according to the open-pit slope slip surface damage evolution path, determining the stability risk level of the whole and different regions inside the open-pit slope and locating the high-risk area, and generating an open-pit slope risk prediction report;

[0114] According to the open-pit slope slip surface damage evolution path, the expansion speed, area change rate and maximum depth of the slip surface are calculated to obtain the evolution intensity index set.

[0115] Further, according to the open-pit slope slip surface damage evolution path, the slip surface optimized three-dimensional morphologies at adjacent time points are extracted, the maximum value of the shortest distance from the point on the most forward profile of the slip surface at the next time to the slip surface profile at the previous time is calculated as the most forward advancing distance, and according to the ratio of the most forward advancing distance to the time interval, the expansion speed of the slip surface is obtained. The ratio of the change amount of the surface area of the slip surface between adjacent time points to the time interval is calculated to obtain the area change rate of the slip surface. The maximum vertical distance from each time point slip surface surface point to the preset open-pit slope top reference plane is measured to obtain the maximum depth of the slip surface. The expansion speed, area change rate and maximum depth at all time points are collected to form the evolution intensity index set.

[0116] It should be noted that the open-pit slope top reference plane is the average fitting plane of the original topography of the slope top determined by geological survey and topographic survey.

[0117] The evolution intensity index set is weighted and fused and threshold judged by a preset comprehensive risk assessment rule to obtain a result of the open-pit mine slope stability risk level;

[0118] Further, the extension speed, area change rate and maximum depth value in the evolution intensity index set are weighted and summed with the weight coefficient in the comprehensive risk assessment rule to obtain a comprehensive risk index, and the overall stability risk level of the open-pit mine slope is determined according to a preset risk level threshold and the comprehensive risk index: if the comprehensive risk index is less than the risk level threshold, it is determined as low risk, if the comprehensive risk index is greater than the risk level threshold, it is determined as high risk, and if the comprehensive risk index is within the risk level threshold, it is determined as medium risk, and a result of the open-pit mine slope stability risk level is generated.

[0119] It should be noted that the comprehensive risk assessment rule is a mathematical decision framework based on the principles of geotechnical mechanics and the evolution law of slope instability, including the weight coefficients of extension speed, area change rate and maximum depth, risk level threshold and corresponding determination logic; the risk level threshold is set based on the statistical relationship between the evolution intensity index and the instability event occurrence probability in historical slope instability cases, and the exemplary value range is 0.3-0.7.

[0120] According to the extension speed and the result of the open-pit mine slope stability risk level, a high-risk area exceeding a preset local speed threshold is located in the three-dimensional digital twin, and a high-risk area spatial coordinate set is generated;

[0121] Further, the evaluation period of the open-pit mine slope stability risk level result as high risk is extracted, the spatial range of the three-dimensional digital twin is divided into uniform local grids in the high-risk period, the extension speed of the slip surface in each local grid is calculated, and compared with the preset local speed threshold, all local grids with extension speed exceeding the local speed threshold are identified; the center point three-dimensional coordinates of all identified local grids are extracted and integrated according to the spatial position of the local grid, and a high-risk area spatial coordinate set is generated.

[0122] It should be noted that the local speed threshold is set based on the critical speed statistical value of the accelerated expansion of the local slip band in historical slope instability cases, and the exemplary value range is 5-20 mm, in order to distinguish the critical state between the stable creep stage and the accelerated destruction stage.

[0123] The high-risk area spatial coordinate set, the result of the open-pit mine slope stability risk level and the evolution intensity index set are integrated to generate an open-pit mine slope risk prediction report;

[0124] Further, the open-pit mine slope stability risk level result (such as "high risk") is taken as the overall risk conclusion of the report, and the numerical data of the expansion speed sequence, the area change rate sequence and the maximum depth sequence in the evolution intensity index set are taken as the dynamic evolution evidence supporting the overall risk conclusion; the three-dimensional coordinate list of all high-risk areas is extracted from the high-risk area spatial coordinate set as the specific distribution information of the risk in space; and the overall risk conclusion, the dynamic evolution evidence and the specific distribution information are organized according to the predefined report template to generate the open-pit mine slope risk prediction report.

[0125] It should be noted that the report template is a standardized document framework containing fixed chapters (such as risk level, evolution data, spatial positioning) that is designed in advance according to the requirements of the mine safety monitoring industry and the actual needs of open-pit mine slope risk assessment.

[0126] The embodiment also provides an open-pit mine slope risk prediction system based on multi-source data, comprising a data acquisition module, a physical quantity prediction module, a damage spatiotemporal graph construction module, a damage evolution path deduction module and a risk judgment module; the data acquisition module is used for acquiring and preprocessing geological data, displacement monitoring data and environmental activity data of the open-pit mine slope, and constructing a three-dimensional digital twin; the physical quantity prediction module is used for performing fusion calculation on the preprocessed displacement monitoring data and environmental activity data through a physical information neural network, predicting the displacement value and stress value of the open-pit mine slope, and obtaining full-field physical quantity prediction data; the damage spatiotemporal graph construction module is used for mapping the displacement value and stress value in the full-field physical quantity prediction data to the spatial nodes of the three-dimensional digital twin, calculating the node damage index according to the full-field stress value, connecting edges according to the spatial adjacency relationship of the nodes, and constructing a damage feature spatiotemporal graph sequence; the damage evolution path deduction module is used for inputting the damage feature spatiotemporal graph sequence into a spatiotemporal graph neural network, capturing the propagation law of the node damage index between the spatial nodes, and dynamically deducing the damage evolution path of the open-pit mine slope slip surface; and the risk judgment module is used for judging the stability risk level of the whole and different regions inside the open-pit mine slope according to the damage evolution path of the open-pit mine slope slip surface, locating the high-risk area, and generating an open-pit mine slope risk prediction report.

[0127] To sum up, the present application realizes the collaborative fusion of multi-source monitoring information and geotechnical mechanics physical laws by predicting the displacement value and stress value of the open-pit mine slope, obtaining full-field physical quantity prediction data, and improves the reliability and generalization ability of the slope deformation and stress response prediction; and realizes the spatiotemporal continuous modeling of the damage evolution process, early identification of the slope instability precursor and accurate insight into the risk evolution trend by dynamically deducing the damage evolution path of the open-pit mine slope slip surface.

[0128] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for predicting slope risk in open-pit mines based on multi-source data, characterized in that: include, Collect geological data, displacement monitoring data, and environmental activity data of open-pit mine slopes, preprocess them, and construct a three-dimensional digital twin; By fusing and calculating the preprocessed displacement monitoring data and environmental activity data through a physical information neural network, the displacement and stress values ​​of the open-pit mine slope are predicted, and the prediction data of physical quantities of the entire field are obtained. The displacement and stress values ​​in the full-field physical quantity prediction data are mapped to the spatial nodes of the three-dimensional digital twin, and the node damage index is calculated based on the full-field stress value. At the same time, the edges are connected according to the spatial adjacency relationship of the nodes to construct a spatiotemporal map sequence of damage characteristics. The damage feature spatiotemporal map sequence is input into the spatiotemporal map neural network to capture the propagation law of node damage indicators between spatial nodes and dynamically deduce the damage evolution path of the slip surface of the open-pit mine slope. Based on the damage evolution path of the slip surface of the open-pit mine slope, the stability risk level of the overall open-pit mine slope and different areas inside is determined, and high-risk areas are located, generating an open-pit mine slope risk prediction report.

2. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The geological data includes the cohesion value, internal friction angle value, elastic modulus value, and spatial coordinate point set of rock strata and joints / fractures; The displacement monitoring data includes surface displacement, deep displacement, and deformation field data; The environmental activity data includes rainfall and blasting parameters; The preprocessing includes data cleaning, spatiotemporal alignment, and normalization.

3. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps for constructing a three-dimensional digital twin are as follows: The spatial interpolation algorithm of Kriging is used to perform spatial interpolation on the set of spatial coordinate points of rock strata and joints in the preprocessed geological data to generate a three-dimensional spatial framework of the geological body structure. The cohesion, internal friction angle and elastic modulus of the soil and rock mass in the preprocessed geological data are used as material properties to be assigned to the corresponding rock strata spatial regions in the three-dimensional spatial framework of the geological mass structure, forming a mechanical characterization body for open-pit mine slopes. The preprocessed displacement monitoring data and environmental activity data are mapped onto the mechanical characterization body of the open-pit mine slope to generate a three-dimensional digital twin.

4. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The process involves fusing preprocessed displacement monitoring data and environmental activity data using a physical information neural network to predict the displacement and stress values ​​of the open-pit mine slope, thereby obtaining predicted physical quantity data for the entire field. The steps are as follows: The preprocessed displacement monitoring data and external environmental activity data are aligned and stitched together to generate a multi-dimensional feature vector. The multidimensional feature vector is input into the physical information neural network for forward propagation calculation, and the displacement and stress values ​​of the rock and soil mass of the future open-pit mine slope are output. The displacement and stress values ​​of the rock and soil on the slope of the future open-pit mine are physically consistent and smoothed, and then integrated to generate a dataset of predicted physical quantities for the entire field.

5. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps are as follows: mapping displacement and stress values ​​from the full-field physical quantity prediction data to spatial nodes of a three-dimensional digital twin, calculating node damage indices based on the full-field stress values, and connecting edges according to the spatial adjacency relationships of the nodes to construct a spatiotemporal map sequence of damage characteristics. The displacement and stress values ​​in the full-field physical quantity prediction data are mapped to the corresponding spatial nodes in the three-dimensional digital twin according to the spatial coordinates to obtain the node physical state dataset. Based on the Mises yield criterion, scalar equivalent stress is calculated for the stress value of each spatial node to obtain the node damage index; Based on the proximity of spatial nodes, all spatial nodes carrying node damage indicators are connected by edges to obtain a static damage feature map of open-pit mine slope. Static damage feature maps at different time points are concatenated in chronological order to generate a spatiotemporal sequence of damage feature maps.

6. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps for inputting the spatiotemporal graph sequence of damage features into a spatiotemporal graph neural network to capture the propagation pattern of node damage indicators between spatial nodes are as follows. The damage feature spatiotemporal graph sequence is input into the spatiotemporal graph neural network. The graph attention network layer calculates the weights of the connection relationships between each node and its neighboring nodes to generate an edge weight matrix. Based on the edge weight matrix, the damage indices of each node's neighboring nodes are weighted, summed, and transformed to generate a node-enhanced spatial feature sequence.

7. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps for dynamically extrapolating the damage evolution path of the slip surface on the open-pit mine slope are as follows. The node augmentation spatial feature sequence is input into the temporal convolutional layer of the spatiotemporal graph neural network to learn the trend of node damage index evolution over time, and regression prediction is performed through a fully connected layer to output the predicted value of node damage index. The predicted values ​​of node damage indicators are mapped back to the spatial location of the three-dimensional digital twin, and the predicted values ​​of node damage indicators are subjected to threshold screening and spatial clustering to form the initial slip surface morphology. The initial slip surface morphology is noise-filtered and void-filled by a three-dimensional connected component analysis algorithm to generate an optimized three-dimensional slip surface morphology. By connecting the slip surfaces at different time points in chronological order, the three-dimensional morphology is optimized to form the damage evolution path of the slip surface on the open-pit mine slope.

8. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps for determining the stability risk level of the open-pit mine slope as a whole and its different internal areas are as follows. Based on the damage evolution path of the slip surface on the open-pit mine slope, the expansion rate, area change rate and maximum depth of the slip surface are calculated to obtain a set of evolution intensity indicators. By using pre-set comprehensive risk assessment rules to weight and fuse the set of evolution intensity indicators and make threshold judgments, the results of the open-pit mine slope stability risk level are obtained.

9. The open-pit mine slope risk prediction method based on multi-source data as described in claim 1, characterized in that: The steps for locating high-risk areas and generating an open-pit mine slope risk prediction report are as follows. Based on the expansion speed and the open-pit mine slope stability risk level, high-risk areas exceeding the preset local velocity threshold are located in the three-dimensional digital twin, and a set of spatial coordinates of the high-risk areas is generated. By integrating the spatial coordinate set of high-risk areas, the results of open-pit mine slope stability risk level, and the set of evolution intensity indicators, an open-pit mine slope risk prediction report is generated.

10. A slope risk prediction system for open-pit mines based on multi-source data, based on the slope risk prediction method for open-pit mines based on multi-source data as described in any one of claims 1 to 9, characterized in that: It includes a data acquisition module, a physical quantity prediction module, a damage spatiotemporal map construction module, a damage evolution path deduction module, and a risk assessment module; The data acquisition module is used to collect geological data, displacement monitoring data, and environmental activity data of open-pit mine slopes and preprocess them to construct a three-dimensional digital twin. The physical quantity prediction module is used to fuse and calculate the preprocessed displacement monitoring data and environmental activity data through a physical information neural network to predict the displacement and stress values ​​of the open-pit mine slope and obtain the physical quantity prediction data for the entire field. The damage spatiotemporal map construction module is used to map the displacement and stress values ​​in the full-field physical quantity prediction data to the spatial nodes of the three-dimensional digital twin, calculate the node damage index based on the full-field stress value, and connect the edges according to the spatial adjacency relationship of the nodes to construct a sequence of damage feature spatiotemporal maps. The damage evolution path deduction module is used to input the spatiotemporal map sequence of damage features into the spatiotemporal map neural network, capture the propagation law of node damage indicators between spatial nodes, and dynamically deduce the damage evolution path of the slip surface of the open-pit mine slope. The risk assessment module is used to determine the overall stability risk level of the open-pit mine slope and different areas within it based on the damage evolution path of the slip surface, locate high-risk areas, and generate an open-pit mine slope risk prediction report.

Images