Dangerous rock monitoring system and risk assessment method thereof

Abstract

The invention relates to the technical field of dangerous rock monitoring, in particular to a dangerous rock monitoring system and a risk assessment method thereof. According to the method, environmental factors such as wind speed and vegetation are accurately modeled by introducing the graph convolutional neural network, the limitation of a traditional wind power calculation model is overcome, and the influence of wind power tension is corrected, so that the stability of the rock mass is more accurately evaluated; by introducing a dynamic adjustment mechanism of an environment change coefficient and a vibration propagation coefficient, changes of environment factors such as temperature, humidity and rainfall can be adapted in real time, and the adaptability and precision of the model are improved; besides, spatial interaction and dynamic effect among rock mass units are considered, an instability propagation equation is established, and a rock mass instability propagation process is captured, so that physical attributes, spatial interaction and environmental change of the rock mass are comprehensively considered, and more accurate and real-time instability prediction is provided. And effective risk early warning and emergency decision-making assistance are provided for projects such as mining and slope construction.

Description

Technical Field

[0001] The present invention relates to the technical field of dangerous rock monitoring, and in particular to a dangerous rock monitoring system and a risk assessment method thereof. Background Art

[0002] Monitoring methods for rock mass instability often rely on physical monitoring methods, such as crack observation and settlement measurement. These methods can provide a certain degree of reflection on the rock mass state, but they are usually unable to fully capture the dynamic characteristics of the rock mass instability process, especially the interactions between rock mass units and the combined impact of external environmental factors on rock mass stability. Moreover, existing monitoring methods often fail to reflect the long-term cumulative effects of external environmental factors (such as precipitation and temperature changes) on rock mass unit stability in real time, resulting in low prediction accuracy and timeliness.

[0003] Secondly, current rock mass instability assessment models ignore the combined effects of multiple factors such as topography, vegetation growth, and wind speed. For example, conventional wind impact models fail to fully consider the acceleration or retardation effects of local topography or vegetation on wind speed, resulting in inaccurate wind force calculations, which in turn affects the accuracy of rock mass stability predictions.

[0004] In addition, rock mass instability is often a complex dynamic process, affected by multiple factors such as the historical behavior of the rock mass itself, changes in the external environment, and the mutual influence of adjacent rock units. As a result, it is currently difficult to fully capture the spatial and temporal interactions and nonlinear relationships, resulting in the inability to improve the accuracy of rock mass monitoring and prediction of instability risks. Summary of the Invention

[0005] In response to the above-mentioned shortcomings of the existing technology, the present invention provides a dangerous rock monitoring system and a risk assessment method thereof, which can effectively solve the problem that the existing technology does not fully consider the actual environmental characteristics and environmental characteristics of the rock mass, thereby failing to improve the accurate monitoring of the risk of rock instability.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] The present invention provides a dangerous rock monitoring system, which at least comprises:

[0008] Regional positioning module, collects dangerous rock areas and establishes a three-dimensional geometric model to construct a rock unit set;

[0009] Dangerous rock mass judgment module, screening dangerous rock mass units based on joint surfaces or terrain;

[0010] The dangerous rock index analysis module calculates environmental indicators based on the dangerous rock mass units and determines the wind load in the environmental indicators:

[0011] A graph convolutional neural network is introduced to predict local surface wind speed gain, node feature vectors and edge features are established, and multi-layer graph convolution is performed to output graph-level feature vectors. The wind amplification factor is calculated based on the graph-level feature vector, and the wind load is established based on the wind amplification factor and environmental data.

[0012] A rock mass instability prediction module is configured to input characteristic indicators of the environmental indicators to establish exogenous variables and preliminarily predict the instability probability of the rock mass unit based on a time series model;

[0013] An adjacency matrix is ​​established in response to the spatial relationship between rock and soil units. An instability propagation equation is established based on the adjacency matrix to simulate the instability propagation process and re-predict the instability probability of the rock unit.

[0014] Furthermore, the adjacency matrix simulates the mutual influence and propagation effect between rock mass unit instabilities based on the physical properties, spatial distance, vibration propagation coefficient, geological structure and external environmental factors between rock mass units.

[0015] Furthermore, when the vibration propagation coefficient is determined:

[0016] Considering that the vibration propagation of rock mass units is affected by the environment and affects the changes in the physical properties of the rock mass, the vibration propagation coefficient is dynamically constructed by introducing the environmental variation coefficient.

[0017] Furthermore, when the vibration propagation coefficient is determined:

[0018] Considering the changes in vibration propagation of rock mass units caused by environmental changes, a nonlinear exponential function is introduced to reflect the sudden effect of environmental changes and define the vibration propagation coefficient.

[0019] Furthermore, when the vibration propagation coefficient is determined:

[0020] Considering that in arid environments, cracks in the rock mass shrink and the vibration transmission intensity decreases, in a humid environment, the rock mass may become more fragile and the vibration transmission intensity increases;

[0021] The vibration propagation coefficient is comprehensively defined by introducing adaptive mechanism, physical properties of rock mass and mutation effect.

[0022] Furthermore, the method for determining dangerous rock units by terrain is as follows:

[0023] The risk assessment method, applied to the dangerous rock monitoring system, includes the following steps:

[0024] Collect dangerous rock areas and establish a three-dimensional geometric model to construct a rock unit set;

[0025] Dangerous rock mass judgment module, screening dangerous rock mass units based on joint surfaces or terrain;

[0026] The dangerous rock index analysis module calculates environmental indicators based on the dangerous rock units.

[0027] Establish exogenous variables and preliminarily predict the instability probability of rock mass units based on the time series model;

[0028] An adjacency matrix is ​​established in response to the spatial relationship between rock and soil units. An instability propagation equation is established based on the adjacency matrix to simulate the instability propagation process, and the instability probability of the rock unit is re-predicted to achieve risk assessment of the rock unit.

[0029] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0030] By introducing a graph convolutional neural network to accurately model environmental factors such as wind speed and vegetation, the limitations of traditional wind calculation models are overcome, the influence of wind pull is corrected, and the stability of rock masses is more accurately assessed.

[0031] By introducing a dynamic adjustment mechanism for the environmental variation coefficient and the vibration propagation coefficient, it is possible to adapt to changes in environmental factors such as temperature, humidity, and precipitation in real time, thereby improving the adaptability and accuracy of the model. In addition, by considering the spatial interaction and dynamic effects between rock units, an instability propagation equation is established to capture the propagation process of rock instability, so as to comprehensively consider the physical properties, spatial interactions, and environmental changes of the rock mass, thereby providing more accurate and real-time instability predictions, and providing effective risk warnings and emergency decision-making assistance for projects such as mining and slope construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0033] Figure 1 It is the overall module block diagram of the present invention. DETAILED DESCRIPTION

[0034] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] The present invention will be further described below with reference to the embodiments.

[0036] Example 1 (see Figure 1 ): Dangerous rock monitoring system, including at least:

[0037] The regional positioning module inputs high-resolution satellite orthophotos of the monitoring area and defines dangerous rock areas. Dangerous rock areas refer to the locations of rock masses with signs of instability (such as cracks, cliffs, steep slopes, etc.). The geometric information of the dangerous rock areas can be obtained, and the corresponding three-dimensional geometric model M0 can be established to achieve spatial reference positioning of the dangerous rock areas.

[0038] Random sampling of surface point cloud is performed on the 3D geometric model, and the plane parameters (n j ,d j ), nj represents the surface normal vector, d j Representing the plane constant, the joint (fault) surface set {S j}, S j Represents the jth joint surface, that is, a single plane, each S j Corresponding to a plane equation:

[0039] S j :n j ·X+d j = 0, X represents the coordinate vector of any point in three-dimensional space, d j Indicates S j The offset of the plane S is a constant term. j Contains all j :n j ·X+d j = 0 three-dimensional coordinate vector;

[0040] Take two planes S j 、S j′ Intersection line, S j′ represents the j′th plane, belonging to {S j}, combined with the elevation model (DEM), topological segmentation is performed on the three-dimensional geometric model, and the three-dimensional geometric model is divided into a number of relatively independent rock unit sets. A rock unit is a polyhedron surrounded by triangular mesh surfaces.

[0041] Furthermore, the dangerous rock mass judgment module selects dangerous rock mass units k from the rock mass unit set and establishes a dangerous rock mass set Methods include:

[0042] 1) Judging by joint surface:

[0043] Identify all joint strips belonging to the rock mass unit on the surface or internal point cloud of the rock mass unit N k Indicates the total number of joints;

[0044] If the first Joints satisfy:

[0045]

[0046] Among them, φ min 、φ max Respectively represent the minimum and maximum tilt judgment thresholds, usually 30° and 70°, Representing joints The inclination angle, Representing joints The trend of α k represents the main slope direction of the rock mass unit, Δθ max Indicates the threshold for judging the direction difference, usually 20°;

[0047] Therefore, if there is a condition φ min ≤φ k , The rock mass unit is determined to be a dangerous rock mass unit;

[0048] 2) Judging by terrain:

[0049] Determine the topographic characteristics of each rock mass unit, specifically:

[0050] slope

[0051] in, Represents the DEM gradient at point p, reflecting the slope of the steepest direction of the terrain at that point.

[0052] It represents the calculated slope angle, therefore, the steepest point in the rock mass unit is taken to measure the overall steepness;

[0053] Maximum normal curvature

[0054] Among them, κ n (p) represents the curvature along the normal direction at point p, reflecting the bending or turning strength of the terrain surface. It should be noted that high normal curvature is usually found at cliff corners or protrusions, and is used to distinguish between cliffs and steep walls;

[0055] Secondly, root growth will generate wedging pressure, which has a significant impact on crack propagation. The evaluation of the vegetation status and root growth potential of the rock unit includes:

[0056] NDVI is calculated for rock mass units to measure vegetation activity and coverage, with higher NDVI values ​​indicating denser vegetation;

[0057] Determine the root growth rate category ρ based on NDVI value k (It can include multiple interval ranges, such as low, medium, and high). Usually, a root rate threshold is set and compared with the NDVI value, which will not be described here;

[0058] If there is a condition s k ≥s env ∨c k ≥c env ∨ρ k = high, the rock mass unit is determined to be a dangerous rock mass unit.

[0059] The dangerous rock index analysis module deploys corresponding sensors for dangerous rock units to collect time series data. It calculates environmental indicators based on the time series data to construct feature vectors for subsequent instability prediction of dangerous rock units. The details are as follows:

[0060] Calculation of environmental indicators:

[0061] Accumulated rainfall r i represents the rainfall intensity at the i-th moment, and R represents the length of the time window;

[0062] Maximum hourly rainfall intensity r t max =max i∈[t-L+1] r i ;

[0063] Daily temperature difference ΔT t =max i∈[t-1,t] T i -min i∈[t-1,t] T i , T i represents the temperature at the i-th moment;

[0064] Number of freeze-thaw cycles 1(·) represents the indicator function;

[0065] Heat stress index α represents the thermal expansion coefficient, d represents the thermal diffusion depth, I t Indicates light irradiance;

[0066] Dynamic root pressure index P root,k (t) = P 0,k +β k,1 t+β k,2 SM k,t+ β k,3 Sap k,t , P 0,k represents the initial wedging pressure of the root system, β k,1 represents the time sensitivity coefficient, β k,2represents the soil moisture coefficient, β k,3 Represents the sap activity coefficient, SM k,t Indicates soil moisture content, Sap k,t Indicates sap flow rate;

[0067] Deformation rate P k,t represents the three-dimensional space coordinate vector of the k-th rock mass unit at time t, and Δt represents the sampling time interval;

[0068] Displacement acceleration v k,t represents the displacement rate at time t;

[0069] When calculating wind pull:

[0070] Considering that in actual bare rock monitoring, wind force does not act uniformly on the entire rock surface. Local terrain (such as grooves and protrusions) and vegetation (such as shrubs and vines) will produce "wind speed acceleration" or "blocking" effects. Traditional mechanical models that use full-field wind speed calculations are difficult to capture this effect, resulting in prediction distortion and, in turn:

[0071] Deploy multi-point wind speed or direction sensor arrays around each dangerous rock unit, along with structured light scanning, to restore the instantaneous turbulence structure between wind and vegetation;

[0072] The physics-based graph convolutional neural network combines sensor array data with a three-dimensional grid input to predict the local surface wind speed gain, which is:

[0073] Input the 3D geometric model M0 and dangerous rock mass units;

[0074] According to the dangerous rock mass unit grid subset g k =M 0|k , retaining the triangular faces and their vertices constituting the dangerous rock mass unit;

[0075] Construct graph structure: nodes are grid vertices n=1,…,F k , corresponding to the three-dimensional coordinate X n , F k represents the number of vertices. For edges, if vertices n and m share an edge on the grid, then there is an undirected edge (n, m) in the graph;

[0076] Predicting amplification factors via graph convolutional neural networks:

[0077] Node feature vector n n represents the mesh vertex normal vector, Respectively represent the average wind speed and average wind direction around n (radius of 5, 10 or 15m)

[0078] Edge features: You can choose the geometric distance between vertices or the area of ​​the surface;

[0079] Multi-layer graph convolution: γ=0,…,L-1

[0080] in, When representing layer γ+1, the feature vector of node n in the graph is That is, the initial node feature vector, When representing layer γ, the feature vector of neighbor node m, b (γ) represents the bias vector of the γth layer, L represents the total number of network layers, σ(·) represents the activation function, such as ReLU, sigmoid, etc., which performs nonlinear mapping on the linear transformation results, acting element by element, N(n) represents the neighbor set of node n, and W (γ) represents the weight matrix of the γth layer, which is used to linearly transform the feature vectors from all neighbors;

[0081] Therefore, after L layers of graph convolution, each node n obtains a high-order feature vector

[0082] The node-level high-order feature vectors are pooled by global pooling (usually using maximum pooling or average pooling). Aggregate into a graph-level feature vector h graph ;

[0083] Graph-level feature vector h graph Output wind amplification factor α based on the multi-layer perceptron network MLP (two-layer MLP) k,t ,in:

[0084] If α k,t If it is greater than 1, it means that the local terrain or vegetation causes the wind speed to accelerate. If it is less than 1, it means that the obstacles or grooves cause the wind speed to decrease. If it is equal to 1, it means there is no significant attenuation.

[0085] Then, the wind force correction calculation is performed:

[0086] F wind,k (t) represents the modified wind load of the kth dangerous rock mass unit at time t, ρ air Indicates the air density, C d Indicates the drag coefficient, A k represents the windward area of ​​the dangerous rock mass unit, v t Indicates the wind speed recorded by the weather station (environmental data: i.e. the aforementioned air density, drag coefficient, etc.).

[0087] Rock mass instability prediction module, constructs characteristic vector x based on environmental indicators k,t , by in the feature vector x k,tCorresponding characteristic indicators, such as cumulative rainfall, daily temperature difference, number of freeze-thaw cycles, etc., are extracted as exogenous variables to construct a time series model to predict the instability probability of dangerous rock units in the future, as follows:

[0088] Obtain time series data that influences the probability of rock mass unit failure, including:

[0089] Instability event data (the probability of instability of a rock unit or whether an actual instability event has occurred), displacement data, deformation rate, wind speed, temperature, stress, strain, crack width and other data of the rock unit can be aligned with timestamps to construct a complete time series data of a rock unit;

[0090] Introduce the characteristic indicators calculated above to build a time series model:

[0091]

[0092] in, represents the instability probability of rock mass unit k at time t (target variable), c represents the constant term, φ i Represents the autoregressive coefficient, which indicates the impact of the historical value of the target variable on the current value, p represents the coefficient of the autoregressive term, θ j represents the sliding mean coefficient, ε t -j represents the error term at time tj, q represents the order of the sliding average term, β u represents the coefficient of the exogenous variable, represents the exogenous variable u related to rock mass unit k at time t, such as cumulative rainfall, daily temperature difference, number of freeze-thaw cycles, etc. U represents the number of exogenous variables, ε t represents the noise term;

[0093] By combining time series data with exogenous variables to establish a time series model, the probability of rock mass unit instability at future moments can be accurately predicted, thereby providing effective early warning for rock mass stability monitoring. Through the autoregressive part of historical time series data, the model captures the temporal pattern of rock mass instability, and the introduction of external environmental factors (such as accumulated rainfall, daily temperature difference, wind speed, etc.) as exogenous variables enables the model to adapt to environmental changes and reflect the real-time impact of environmental factors such as accumulated rainfall and daily temperature difference on rock mass stability.

[0094] Secondly, the addition of the sliding average part further optimizes the accuracy of the model, effectively eliminating short-term fluctuations and errors, thereby improving the accuracy of the prediction and realizing accurate monitoring of the risk status of mining rock masses, slopes and other areas.

[0095] It should be noted that:

[0096] The above results show the probability of instability of each dangerous rock mass unit at different times. However, in the actual rock mass environment, the following points need to be considered:

[0097] Interaction of Space:

[0098] The instability of a rock mass may be affected by the state of adjacent rock units. The instability of a rock unit is not only affected by its own historical behavior, but also closely related to the stability of the surrounding rock mass. Therefore, the instability of rock units may be spatially correlated.

[0099] Time delay and dynamic effects:

[0100] The instability of rock mass units is often a dynamic process and is affected by time delays. For example, the instability of a rock mass unit may lag behind the instability of its neighboring units in time. Such time delays and cumulative effects are difficult to capture and reflect through time series models.

[0101] Nonlinear relationships and complex propagation states:

[0102] The propagation of instability between rock mass units is usually not linear but highly complex. For example, the propagation of instability between rock mass units may have threshold effects and superposition effects. Time series models are difficult to capture these nonlinear relationships.

[0103] Therefore, the following steps are also included:

[0104] Considering the spatial relationship between multiple rock mass units, the adjacency matrix A is established k,k ′ , describing which rock mass units are spatially adjacent or may interact with each other:

[0105]

[0106] Among them, ω k,k′ represents the physical attribute weights between rock mass units, including hardness, cracks, and rock types, λ k,k′ Indicates the influence of geological transmission paths, such as cracks, faults, etc., α k,k′ represents the vibration propagation coefficient, β k represents the weighting factor of external environmental factors (daily temperature difference, wind speed, etc.), α1 represents the adjustment parameter, d k,k ′ represents the spatial distance between rock mass units k and k′ (calculated by Euclidean distance). By introducing the adjacency matrix, multiple factors such as the physical properties, spatial distance, vibration effect, geological structure and external environment between rock mass units are considered to more accurately simulate the mutual influence and propagation effect between rock mass unit instability.

[0107] Define the instability propagation equation:

[0108]

[0109] Among them, y k,t It represents the instability probability after propagation, reflecting the instability of rock mass unit k at time t, taking into account the influence of adjacent rock mass units and environmental factors. Indicates the dynamic influence of the external environment on the stability of the rock mass unit, such as daily temperature difference, etc. k,t 、x k′,t denote the characteristic vectors of rock mass units k and k′ at time t, such as accumulated rainfall, diurnal temperature difference, etc., respectively. f(·) denotes the propagation function, which can be linearly weighted (established by weighted summation of the instability probabilities and physical characteristic differences between rock mass units k and k′, such as stress) or nonlinear, and is not specifically limited.

[0110] Furthermore, considering that the instability of multiple rock units may affect each other, the instability propagation process is simulated by the instability propagation equation. The influence of the instability of one rock unit on the adjacent rock units is transmitted through the weighted matrix. The instability probability of each rock unit is affected not only by its own historical data, but also by the instability probability of the adjacent rock units. Then, the instability probability of the rock unit is corrected.

[0111] Furthermore, based on the corrected instability probability, it can be determined whether the instability threshold is reached. If the instability threshold is reached, it is determined that the rock unit has a greater instability risk. Subsequently, early warning and implementation of instability-related emergency control measures can be carried out for the rock unit to improve the accurate monitoring and control of instability risks in the rock area, and at the same time, adaptive resource management planning can be implemented for rock units with urgent or general risk levels, further balancing the difficulty of emergency disposal of rock units and reducing the risk expansion degree when subsequent mine rock disasters occur.

[0112] In addition, this scheme also considers the actual physical environment of the rock mass unit to analyze the vibration propagation coefficient α k,k′ The following describes the effects of different situations on the vibration propagation coefficient α through different embodiments. k,k′ Impact:

[0113] Example 2: The vibration propagation capability of the rock mass unit is affected by the environment. In particular, factors such as temperature, humidity, and precipitation can cause changes in the physical properties of the rock mass (such as hardness and brittleness), thereby affecting the propagation of vibration. Therefore, an environmental variation coefficient is introduced to dynamically construct the vibration propagation coefficient α. k,k′ , then:

[0114] α k,k′ (t) = α0·exp(γ·ΔE(t))

[0115] α0 represents the vibration propagation coefficient when there is no external environmental influence, γ represents the environmental change adjustment factor, ΔE(t) represents the environmental change at time t, such as the change in temperature difference, precipitation change, etc., and exp(·) represents the indicator function.

[0116] Example 3: Environmental changes may cause a sharp change in the vibration propagation capacity of rock units near certain critical values. For example, temperature changes may cause cracks in rock units to expand, thereby increasing the vibration propagation capacity. A nonlinear exponential function is introduced to reflect the sudden effect of environmental changes, and the vibration propagation coefficient α is defined as k,k′ , then:

[0117]

[0118] Example 4: In arid environments, cracks in the rock mass may shrink, thereby reducing the vibration transmission capability. In a humid environment, the rock mass may become more fragile, increasing the efficiency of vibration transmission. An adaptive mechanism, the physical properties of the rock mass, and the mutation effect are introduced to comprehensively define the vibration transmission coefficient, ensuring that an accurate and comprehensive vibration transmission coefficient can be established based on the actual impact on the rock mass. The details are as follows:

[0119]

[0120] Among them, E k and E k′ denote the elastic modulus of rock mass elements k and k′, respectively, and FRF (f k ,ω k ,ζ k ) represents the response of rock mass unit k to vibrations of different frequencies during vibration propagation. It is used to consider the response characteristics of rock mass unit k to external vibrations, thereby affecting the intensity of vibration propagation. k represents the frequency response, ω k represents the angular frequency of vibration, ζ k represents the damping coefficient of the rock unit, φ(t) represents a time-varying factor, describing the adaptability of the rock unit k at time t. For example, changes in the adaptability of the rock mass over time may lead to changes in the vibration propagation coefficient. Then, the vibration propagation coefficient is adjusted by multiple factors, combining the physical properties of the rock mass (such as elastic modulus, distance), the frequency response of vibration, environmental changes (temperature, humidity, etc.) and time-related adaptability factors, so as to more accurately describe the vibration propagation process between rock units. Through these comprehensive factors, the complexity of the mutual influence between rock units can be captured, and a more dynamic vibration propagation coefficient related to changes in environmental and physical conditions can be provided.

[0121] Finally, the present invention also provides:

[0122] A risk assessment method is implemented based on the dangerous rock monitoring system;

[0123] A computer device comprising a memory and a processor, wherein the memory stores a computer program and the processor implements the above system when executing the computer program;

[0124] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the system is implemented; the specific reference is made to the dangerous rock monitoring system mentioned above, which will not be repeated here.

[0125] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. 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 will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the various embodiments of the present invention.

Claims

1. Dangerous rock monitoring system, characterized by: include: Regional positioning module, collects dangerous rock areas and establishes a three-dimensional geometric model to construct a rock unit set; Dangerous rock mass judgment module, screening dangerous rock mass units based on joint surfaces or terrain; The dangerous rock index analysis module calculates environmental indicators based on the dangerous rock mass units and determines the wind load in the environmental indicators: A graph convolutional neural network is introduced to predict local surface wind speed gain, node feature vectors and edge features are established, and multi-layer graph convolution is performed to output graph-level feature vectors. The wind amplification factor is calculated based on the graph-level feature vector, and the wind load is established based on the wind amplification factor and environmental data. A rock mass instability prediction module is configured to input characteristic indicators of the environmental indicators to establish exogenous variables and preliminarily predict the instability probability of the rock mass unit based on a time series model; An adjacency matrix is ​​established in response to the spatial relationship between rock and soil units. An instability propagation equation is established based on the adjacency matrix to simulate the instability propagation process and re-predict the instability probability of the rock unit.

2. The dangerous rock monitoring system according to claim 1, characterized in that: The method of multi-layer graph convolution to output graph-level feature vectors is as follows: in, represents the feature vector of node n in the layer γ+1 graph, When representing layer γ, the feature vector of neighbor node m, b (γ) represents the bias vector of the γth layer, L represents the total number of network layers, σ(·) represents the activation function, N(n) represents the neighbor set of node n, and W (γ) represents the weight matrix of the γth layer; After L layers of graph convolution, each node n obtains a high-order feature vector The node-level high-order feature vectors are pooled globally Aggregate into graph-level feature vector h graph .

3. The dangerous rock monitoring system according to claim 2, characterized in that: The method for determining the wind load is: Graph-level feature vector h graph According to the multi-layer perceptron network MLP output wind amplification factor α k,t ; Perform the calculation: F wind,k (t) represents the modified wind load of the kth dangerous rock mass unit at time t, ρ air Indicates the air density, C d Indicates the drag coefficient, A k represents the windward area of ​​the dangerous rock mass unit, v t Indicates the wind speed recorded by the weather station.

4. The dangerous rock monitoring system according to claim 1, characterized in that: The method for preliminarily predicting the instability probability of rock mass units using the time series model is: Obtain time series data and exogenous variables that affect the probability of rock unit instability; Build a time series model: in, represents the instability probability of rock mass unit k at time t, c represents the constant term, φ i represents the autoregressive coefficient, p represents the coefficient of the autoregressive term, θ j represents the sliding mean coefficient, ε t-j represents the error term at time tj, q represents the order of the sliding average term, β u represents the coefficient of the exogenous variable, represents the exogenous variable u associated with rock mass unit k at time t, U represents the number of exogenous variables, and ε t represents the noise term.

5. The dangerous rock monitoring system according to claim 1, characterized in that: The adjacency matrix simulates the mutual influence and propagation effect between rock mass unit instabilities based on the physical properties, spatial distance, vibration propagation coefficient, geological structure and external environmental factors between rock mass units.

6. The dangerous rock monitoring system according to claim 5, characterized in that: When the vibration propagation coefficient is determined: Considering that the vibration propagation of rock mass units is affected by the environment and affects the changes in the physical properties of the rock mass, the vibration propagation coefficient is dynamically constructed by introducing the environmental variation coefficient.

7. The dangerous rock monitoring system according to claim 5, characterized in that: When the vibration propagation coefficient is determined: Considering the changes in vibration propagation of rock mass units caused by environmental changes, a nonlinear exponential function is introduced to reflect the sudden effect of environmental changes and define the vibration propagation coefficient.

8. The dangerous rock monitoring system according to claim 5, characterized in that: When the vibration propagation coefficient is determined: Considering that in arid environments, cracks in the rock mass shrink and the vibration transmission intensity decreases, in a humid environment, the rock mass may become more fragile and the vibration transmission intensity increases; The vibration propagation coefficient is comprehensively defined by introducing adaptive mechanism, physical properties of rock mass and mutation effect.

9. The dangerous rock monitoring system according to claim 3, characterized in that: The method for judging dangerous rock units by terrain is: Determine the topographic characteristics of the rock mass unit: slope in, represents the DEM gradient at point p, Indicates the calculated slope angle; Maximum normal curvature Among them, κ n (p) represents the curvature along the normal direction at point p; Calculate NDVI for rock mass units; Determine the root growth rate category ρ based on NDVI value k ; If there is a condition s k ≥s env ∨c k ≥c env ∨ρ k = high, the rock mass unit is determined to be a dangerous rock mass unit.

10. A risk assessment method, applied to the dangerous rock monitoring system according to any one of claims 1 to 9, characterized in that: The method comprises the following steps: Collect dangerous rock areas and establish a three-dimensional geometric model to construct a rock unit set; Dangerous rock mass judgment module, screening dangerous rock units based on joint surfaces or terrain; The dangerous rock index analysis module calculates environmental indicators based on the dangerous rock units. Establish exogenous variables and preliminarily predict the instability probability of rock mass units based on the time series model; An adjacency matrix is ​​established in response to the spatial relationship between rock and soil units. An instability propagation equation is established based on the adjacency matrix to simulate the instability propagation process, and the instability probability of the rock unit is re-predicted to achieve risk assessment of the rock unit.

Images