Dynamic optimization arrangement method for high slope monitoring points

Abstract

The invention belongs to the technical field of geotechnical engineering safety monitoring, and particularly discloses a high slope monitoring point dynamic optimization arrangement method, which comprises the following steps: acquiring multi-source monitoring data of each monitoring point in a plurality of monitoring points on a high slope monitoring area; determining the neighborhood radius and the feature density of each monitoring point according to the multi-source monitoring data, and partitioning the monitoring area according to the neighborhood radius and the feature density; obtaining partition features of each partition, and inputting the partition features and the multi-source monitoring data into a deep learning model based on an attention mechanism to obtain an importance score and a stable state of each partition; and according to the importance score, the stable state and the feature density, determining an optimization strategy for arranging the monitoring points in each partition. According to the method, the monitoring effect can be maximized under the condition that the number of monitoring points is limited, real-time monitoring and early warning of deformation and stability of the high slope are achieved, the method has high scientificity, effectiveness and practicability, and the high slope monitoring efficiency and accuracy can be remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of geotechnical engineering safety monitoring, and discloses a method for dynamically optimizing the arrangement of high slope monitoring points, and in particular relates to a method for dynamically optimizing the arrangement of high slope monitoring points that integrates deep learning and a dynamic optimization algorithm, which is suitable for high slope stability monitoring in the fields of water conservancy, transportation, and mining. Background Art

[0002] In various types of engineering construction, high slopes are a common geological structure, and their stability is directly related to the safety and sustainable development of the project. With the continuous advancement of infrastructure construction, such as the vigorous development of large-scale projects such as roads, railways, water conservancy and hydropower, the scale and complexity of high slopes are increasing, and the demand for their stability monitoring is becoming more urgent.

[0003] The traditional method of arranging monitoring points on high slopes mainly relies on engineering experience and simple static arrangement, such as contour arrangement and equidistant arrangement, which has many limitations. On the one hand, the traditional method often fails to fully consider the influence of various factors such as complex geological conditions, topography and engineering construction on the stability of high slopes, resulting in the lack of pertinence and rationality in the arrangement of monitoring points. For example, in some areas with complex geological conditions, the arrangement of monitoring points F based solely on experience may miss key deformation areas, resulting in monitoring blind spots and failure to fully and accurately capture the deformation information of the slope. On the other hand, the traditional arrangement method lacks dynamic adjustment capabilities. During the construction and operation of high slopes, when the state of the slope changes, it is difficult to optimize and adjust the monitoring points in time according to the actual situation, thus affecting the accuracy and effectiveness of the monitoring data.

[0004] The method of high slope monitoring point arrangement based on numerical simulation mainly uses numerical simulation software to analyze the stress and strain distribution of the slope under different working conditions to determine the arrangement location of the monitoring points. The principle of this method is based on the basic theory of geotechnical mechanics. The slope is regarded as a complex mechanical system. The mechanical response of the slope under the action of self-weight, external load, groundwater and other factors is simulated by establishing a mathematical model. Although this method can comprehensively consider various influencing factors of the slope and provide a relatively scientific basis for the arrangement of monitoring points. However, the accuracy of numerical simulation is highly dependent on the accuracy of the model and the selection of parameters. If the physical and mechanical parameters of the rock and soil body are inaccurate, or the boundary conditions of the model are set unreasonably, there will be a large deviation between the simulation results and the actual situation, thereby affecting the accuracy of the arrangement of monitoring points. The numerical simulation process has a large amount of calculation, which requires a lot of time and computing resources. Especially for complex high slope models, the calculation time may be very long, which limits the application efficiency of this method to a certain extent.

[0005] The reasonable arrangement of monitoring points on high slopes directly affects the comprehensiveness of data collection and the timeliness of early warning. Existing monitoring point optimization methods mostly rely on cluster analysis (such as Affinity Propagation algorithm) combined with optimization algorithms (such as quantum particle swarm algorithm) to optimize monitoring points. Although they can reduce the number of monitoring points to a certain extent, they have the following problems: (1) Static optimization relies on historical data and is difficult to adapt to the dynamic deformation characteristics of complex geological conditions, resulting in monitoring blind spots; (2) Clustering parameters need to be adjusted manually, which is highly subjective, and monitoring points are redundant or insufficiently covered, resulting in resource waste or early warning delays; (3) The spatiotemporal correlation of historical monitoring data is not fully explored, and the spatiotemporal nonlinear relationship between monitoring points is difficult to fully express through traditional mathematical models, resulting in the inability to predict potential risk areas. Therefore, there is an urgent need for a method that can dynamically optimize the arrangement of monitoring points in combination with the dynamic change characteristics of high slopes. Summary of the invention

[0006] The purpose of the present invention is to provide a method for dynamically optimizing the arrangement of high slope monitoring points to solve the technical problems of existing arrangement methods, such as blind spots in monitoring, inability to predict potential risk areas, redundant monitoring points or insufficient coverage.

[0007] The present invention provides a method for dynamically optimizing the arrangement of high slope monitoring points, comprising: Acquire multi-source monitoring data of each monitoring point in multiple monitoring points in the high slope monitoring area.

[0008] The neighborhood radius and feature density of each monitoring point are determined according to the multi-source monitoring data, and the monitoring area is partitioned according to the neighborhood radius and feature density.

[0009] The partition features of each partition are obtained, and the partition features and the multi-source monitoring data are input into a deep learning model based on the attention mechanism to obtain the importance score and stable state of each partition.

[0010] Based on the importance score, stable state and feature density, the optimal strategy for placing monitoring points in each partition is determined.

[0011] Preferably, the multi-source monitoring data includes displacement spatiotemporal data and geological data, and the neighborhood radius and feature density of each monitoring point are determined according to the multi-source monitoring data, specifically including: The deformation feature weight is determined according to the standard deviation and mean of the displacement spatiotemporal data; and the geological feature weight is determined according to the standard deviation and mean of the geological data.

[0012] The neighborhood radius of each monitoring point and the density of deformation features and the density of geological features within the neighborhood radius are determined.

[0013] The characteristic density of each monitoring point is determined according to the deformation characteristic weight, deformation characteristic density, geological characteristic weight and geological characteristic density.

[0014] Preferably, determining the neighborhood radius of each monitoring point specifically includes: The distances between each monitoring point and other monitoring points are sorted from small to large, and the first distance is recorded as the nearest neighbor distance of the corresponding monitoring point.

[0015] The average value of the nearest neighbor distances corresponding to all monitoring points is determined, and the standard deviation of the distances between the monitoring points is determined based on the average value.

[0016] Determine the neighborhood parameter based on the mean value and the standard deviation of the distance between monitoring points k , and the distance between each monitoring point and other monitoring points k The distance is taken as the neighborhood radius of the monitoring point.

[0017] Preferably, partitioning the monitoring area according to the neighborhood radius and feature density specifically includes: Determine whether each monitoring point is a core point and determine the adaptive neighborhood radius of the core point according to the neighborhood radius and the feature density.

[0018] Each core point and all monitoring points within its adaptive neighborhood radius constitute a normal partition of the monitoring area; monitoring points that do not belong to the normal partition are regarded as noise points, and the neighborhood radius of the noise point is regarded as an abnormal partition.

[0019] Preferably, determining whether each monitoring point is a core point according to the neighborhood radius and the feature density specifically includes: Determine the density threshold and the minimum point count threshold.

[0020] Determine whether the number of monitoring points within the neighborhood radius of each monitoring point is greater than or equal to the minimum point number threshold. If so, when the feature density of each monitoring point is greater than the density threshold, the corresponding monitoring point is considered to be a core point.

[0021] Preferably, determining the minimum point threshold specifically includes: The expected neighborhood radius is determined based on the average distance between all monitoring points.

[0022] Obtain the number of monitoring points within the expected neighborhood radius corresponding to each monitoring point, and record the median of the corresponding numbers of all monitoring points as the minimum point count threshold.

[0023] Preferably, determining the density threshold specifically includes: Determine the feature density of each monitoring point within the corresponding expected neighborhood radius and the mean and standard deviation of all feature densities.

[0024] The density threshold is determined based on the mean and standard deviation of all feature densities.

[0025] Preferably, determining the adaptive neighborhood radius of the core point specifically includes: The quotient of the average value of the characteristic density of all monitoring points and the characteristic density of the core point is recorded as the first value.

[0026] The adaptive neighborhood radius of the core point is determined according to the product of the neighborhood radius and the first value.

[0027] Preferably, the partition features and the multi-source monitoring data are input into a deep learning model based on an attention mechanism, specifically including: A fusion matrix is ​​constructed according to the partition characteristics and the multi-source monitoring data.

[0028] The fusion matrix is ​​input into the convolutional neural network layer to extract spatial features and obtain a feature map.

[0029] The feature map is input into the CNN-LSTM model based on the attention mechanism.

[0030] Preferably, according to the importance score, stable state and feature density, an optimization strategy for arranging monitoring points in each partition is determined, specifically including: The minimum and maximum score thresholds are determined based on the mean and standard deviation of the importance scores of all partitions.

[0031] When the importance score of a partition is less than the minimum score threshold and is unstable, the optimization strategy for the partition is to increase the number of monitoring points.

[0032] When the importance score of a partition is greater than the maximum score threshold and the feature density of the partition is greater than the partition density threshold, the optimization strategy for the partition is to reduce the number of monitoring points.

[0033] When a partition is an abnormal partition, the optimization strategy for the partition is to increase the number of monitoring points.

[0034] Compared with the prior art, the high slope monitoring point dynamic optimization arrangement method of the present invention has the following beneficial effects: The present invention combines the deformation characteristics of high slopes with deep learning technology to dynamically optimize the layout of monitoring points, maximize the monitoring effect with a limited number of monitoring points, and achieve real-time monitoring and early warning of high slope deformation and stability. This method is highly scientific, effective and practical, and can significantly improve the efficiency and accuracy of high slope monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1It is the overall framework of the method for dynamically optimizing the arrangement of high slope monitoring points in an embodiment of the present invention.

[0036] Figure 2 It is a flow chart of dynamic partitioning of monitoring areas in an embodiment of the present invention.

[0037] Figure 3 It is a flow chart of dynamic optimization of monitoring points in an embodiment of the present invention.

[0038] Figure 4 This is a monitoring partition result diagram of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0039] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0040] The embodiment of the present invention provides a method for dynamically optimizing the arrangement of high slope monitoring points. Figures 1 to 4 As shown, including: Step 1: Real-time data collection to obtain multi-source monitoring data of each of the multiple monitoring points in the high slope monitoring area.

[0041] In the embodiment of the present invention, the multi-source monitoring data includes displacement spatiotemporal data and geological data, and may further include environmental quantity data.

[0042] Exemplarily, an embodiment of the present invention uses sensors such as the Global Navigation Satellite System (GNSS), displacement meters, and earth pressure meters to collect real-time spatiotemporal displacement data of the high slope monitoring area; and uses sensors such as rain gauges and water level meters to collect real-time environmental quantity data. The two types of data together constitute multi-source monitoring data of high slope displacement.

[0043] Assume the monitoring data is , first standardize it , It is The mean of the feature dimensions, It is The standard deviation of each feature dimension is calculated to eliminate the influence of dimension and order of magnitude and ensure that different features of multi-source monitoring data are analyzed at the same scale.

[0044] According to the time step Monitoring Points Feature Dimension Construct the spatiotemporal feature matrix in the format of , each row of the spatiotemporal feature matrix represents a time step, and each column represents a feature of a monitoring point. Specifically, for Time steps , No. Monitoring points , No. Feature Dimensions , the matrix elements is the data after the feature standardization of the monitoring point at this time step, that is .

[0045] The time step is determined according to the collection frequency and analysis requirements of the high slope monitoring data, and can be set to 1 hour, 1 day, etc. Indicates the total time step data collected; Indicates the total number of monitoring points arranged. Each monitoring point can collect multiple types of data, such as displacement, stress, rainfall, etc.; analyze the data type collected by each monitoring point to determine the total number of feature dimensions .

[0046] When determining the location of monitoring points, the prior art often only focuses on the monitoring data itself and ignores the geological features, resulting in inaccurate results. To solve this problem, the embodiment of the present invention also integrates geological data into the multi-source monitoring data, such as the internal friction angle of the rock and soil body. , Cohesion The number of parameters is recorded as These parameters provide more physical meanings and constraints for the space-time characteristic matrix model. They are used as additional characteristic dimensions and combined with the established space-time characteristic matrix Splicing to form a spatiotemporal feature matrix with enhanced attributes ,in , No. Geological parameters In the The value of the monitoring area is , then for all time steps definition . For the enhanced spatiotemporal feature matrix Normalization is performed to ensure that all features are in the same scale range. The normalization formula is: ,in and are the minimum and maximum values ​​of the corresponding dimensions in the spatiotemporal feature matrix, respectively.

[0047] Step 2: Determine the neighborhood radius and feature density of each monitoring point based on the multi-source monitoring data, and divide the monitoring area into zones based on the neighborhood radius and feature density, including: Step 2.1: Determine the neighborhood radius and feature density of each monitoring point based on multi-source monitoring data, including: Step 2.1.1: Determine the deformation feature weights based on the standard deviation and mean of the displacement spatiotemporal data; determine the geological feature weights based on the standard deviation and mean of the geological data.

[0048] The embodiment of the present invention combines the geological conditions of the slope (such as strata, geological structures, which are reflected as geological data) and deformation characteristics (displacement, stress, etc., which are reflected as displacement spatiotemporal data), and assigns corresponding weights to different features when calculating the feature density, so that the subsequent clustering and partitioning results are more in line with the actual situation of the high slope.

[0049] For example, let the deformation feature weight be The weight of geological characteristics is The embodiment of the present invention first calculates the standard deviation of the displacement spatiotemporal data and mean and the standard deviation of the geological data and mean , and then determine the fluctuation coefficient of the corresponding deformation feature based on the above standard deviation and mean and the fluctuation coefficient of geological characteristics . Further according to the fluctuation coefficient of deformation characteristics and the fluctuation coefficient of geological characteristics Determine the corresponding deformation feature weights and geological feature weights, as shown in formula (1).

[0050] (1) Step 2.1.2: Determine the neighborhood radius of each monitoring point and the deformation feature density and geological feature density within the neighborhood radius.

[0051] The neighborhood radius of each monitoring point is determined, including: Step S1, sort the distances between each monitoring point and other monitoring points from small to large, and record the first distance as the nearest neighbor distance of the corresponding monitoring point.

[0052] The embodiment of the present invention uses Euclidean distance to calculate the monitoring point Distance to other monitoring points , generating the distance vector , is the data dimension, The monitoring point dimensional coordinates, for each data point, take the first distance value after sorting as the nearest neighbor distance .

[0053] Step S2: Determine the average value of the nearest neighbor distances corresponding to all monitoring points , and based on the average Determine the standard deviation of the distance between monitoring points , as shown in formula (2).

[0054] (2) In the formula, Represents data points The average of the distances to all other points.

[0055] Step S3: According to the average value and standard deviation Determining neighborhood parameters , and the distance between each monitoring point and other monitoring points The distance value is taken as the neighborhood radius of the monitoring point and recorded as the initial neighborhood radius.

[0056] The neighborhood parameters are determined above , as shown in formula (3).

[0057] (3) In the formula, The adjustment factor , determined according to the data distribution and slope zoning requirements , when the monitoring points are distributed relatively evenly, You can take a smaller value; if the data distribution is complex and there are many outliers, A larger value can be taken.

[0058] The neighborhood parameters are obtained using formula (3): After that, you can get the monitoring point Neighborhood radius .

[0059] The deformation feature density and geological feature density within the neighborhood radius of each monitoring point are determined as shown in formula (4).

[0060] (4) In the formula, For monitoring point The deformation feature density, For monitoring point Within the neighborhood radius of The number of monitoring points with similar deformation characteristics is determined based on experience. is the neighborhood volume, For monitoring point The density of geological features, For monitoring point Within the neighborhood radius of The number of monitoring points with similar geological characteristics is determined based on experience.

[0061] Step 2.1.3: Determine the characteristic density of each monitoring point according to the deformation characteristic weight, deformation characteristic density, geological characteristic weight and geological characteristic density, as shown in formula (5).

[0062] (5) In the formula, For monitoring point The characteristic density of is recorded as the monitoring characteristic density, is the deformation feature weight, For monitoring point The deformation feature density, is the geological feature weight, For monitoring point The density of geological features.

[0063] Step 2.2: Partition the monitoring area according to the neighborhood radius and feature density, including: Whether each monitoring point is a core point and the adaptive neighborhood radius of the core point are determined according to the neighborhood radius and feature density, where each core point and all the monitoring points within its adaptive neighborhood radius constitute the normal partition of the monitoring area; at the same time, the monitoring points that do not belong to the normal partition are regarded as noise points, and the neighborhood radius of the noise point is regarded as the abnormal partition.

[0064] The above determines whether each monitoring point is a core point based on the neighborhood radius and feature density, specifically including: determining the minimum point count threshold and density threshold ; Determine whether the number of monitoring points within the neighborhood radius of each monitoring point is greater than or equal to the minimum point count threshold (i.e. ), if so, determine whether the characteristic density of each monitoring point is greater than the density threshold (i.e. ), if yes, then the monitoring point is the core point.

[0065] The determination of the minimum point threshold specifically includes: based on the average value of the distance between all monitoring points Determine the desired neighborhood radius , and then obtain the expected neighborhood radius corresponding to each monitoring point The number of monitoring points in the system, and the median of the number of all monitoring points is recorded as the minimum point threshold .

[0066] For example, the embodiment of the present invention first calculates all the slope The average distance of monitoring points: , and then define a neighborhood radius , , minimum point threshold Take The median number of all monitoring points in the neighborhood .

[0067] Determine the density threshold value in the embodiment of the present invention Specifically, the characteristic density of each monitoring point within the corresponding expected neighborhood radius and the mean and standard deviation of all characteristic densities are determined; the density threshold is determined according to the mean and standard deviation of all characteristic densities.

[0068] For example, for each monitoring point Calculate is the characteristic density of the neighborhood: ,in, For monitoring point of In range and The number of monitoring points with similar deformation characteristics, For is the neighborhood volume of the neighborhood radius. Further calculation of all data points The mean and standard deviation , set the density adjustment coefficient according to the distribution of monitoring points ,but When the data distribution is relatively concentrated, A small value can be taken appropriately; when the data distribution is more dispersed, Take the larger value.

[0069] The embodiment of the present invention determines the adaptive neighborhood radius of the core point, specifically including: recording the quotient of the average value of the characteristic density of all monitoring points and the characteristic density of the core point as a first value; and determining the adaptive neighborhood radius of the core point according to the product of the neighborhood radius and the first value, as shown in formula (6).

[0070] (6) In the formula, is the adaptive neighborhood radius, For monitoring point The neighborhood radius of is the average density of all monitoring points, is the feature density of the core points.

[0071] The embodiment of the present invention partitions the monitoring area using a density-based spatial clustering algorithm (DBSCAN) with noise applications and extends it (EXDBSCAN), which divides the monitoring area with a sufficiently high density into clusters and determines the boundaries of the partitions according to the density of the clusters, such as Figure 2 The embodiment of the present invention introduces an adaptive neighborhood radius to improve the neighborhood radius. , thus better identifying clusters of different density areas.

[0072] The embodiment of the present invention further traverses the determined Core point set , for each core point Calculate the adaptive neighborhood radius The set of all points in the area with radius , the collection All points within (including ) is divided into a cluster If there is a point ,satisfy and , then cluster and merge.

[0073] After the cluster generation is completed, all data points are traversed. For a data point ,like does not belong to the neighborhood of any core point, then Marked as a noise point, the noise point is taken as the core point, and the neighborhood of the noise point is The area is marked as abnormal.

[0074] The stopping condition of clustering in the embodiment of the present invention is as follows: the dual constraints of no change of core points and maximum number of iterations are used as the stopping condition of partitioning. The core point set at the iteration , Representing a collection The number of elements in , when satisfied , and for any core point , For any absence At the same time, the maximum number of iterations is set according to the calculation conditions. , when the core point does not change and When , stop training and output the partition results.

[0075] Step 3: Obtain the partition features of each partition, and input the partition features and multi-source monitoring data into the deep learning model based on the attention mechanism to obtain the importance score and stable state of each partition, such as Figure 3 As shown in FIG, the method specifically includes: constructing a fusion matrix based on partition features and multi-source monitoring data; inputting the fusion matrix into a convolutional neural network layer (CNN layer) for spatial feature extraction, and obtaining a feature map ; The feature map Input into the CNN-LSTM model based on the attention mechanism.

[0076] Based on the results of high slope deformation zoning by the EXDBSCAN model, the embodiment of the present invention proposes a new monitoring point optimization method coupling EXDBSCAN-ATTCNN-LSTM, and further expands the ATTCNN-LSTM monitoring point dynamic optimization model. ATTCNN is used to extract spatial features in the spatiotemporal feature matrix and capture the spatial relationship between monitoring points; LSTM is used to process time series information and analyze the changing trend of monitoring data over time, thereby determining the optimal location and number of deployed monitoring points. The specific steps are as follows: (1) Partition feature encoding: The partition result output in step 2 is encoded using one-hot encoding. Assume that the output is For the partition category partitions, whose one-hot encoding The generation rules are: , Representation vector No. elements, Value range 1 to .

[0077] (2) Model input: The encoded partition features With the spatiotemporal feature matrix In the data feature dimension To expand the dimension, first convert the one-dimensional vector Expand to The matrix , then and exist Dimensions are concatenated to generate a fusion matrix , The number of feature dimensions is , input into the constructed improved ATTCNN-LSTM model feature extractor.

[0078] (3) Feature extraction: The spatiotemporal feature matrix is ​​extracted through the CNN layer Perform spatial extraction, After convolutional layers, we get the feature map , It contains information such as the spatial distribution relationship between monitoring points and the characteristic differences at different locations.

[0079] (4) Introducing the attention mechanism (ATTENTION) to improve the CNN-LSTM model (ATTCNN-LSTM): By calculating the attention weight of each spatiotemporal monitoring position, the model determines the degree of attention to the features of different monitoring positions. First, the feature map Input the fully connected layer for dimensionality reduction to obtain three key parameters in the attention mechanism: query vector , the key vector , value vector ,in , , is a learnable weight matrix. Then the attention score is calculated , is the transpose symbol, for and Finally, the attention score is normalized by Softmax: ,in is the total number of monitoring locations in the feature map.

[0080] (5) Feature weighting: The attention weight matrix With feature map Perform weighted fusion to obtain the weighted feature map ,in is the attention weight matrix.

[0081] (6) Dynamically adjust the LSTM structure: the weighted feature map Input LSTM layer for time series analysis. Introduce Gradient Weighted Class Activation Mapping (Grad-CAM) method to output the importance score of monitoring point features: , the higher the score, the greater the contribution of the monitoring point to the judgment of the slope state. In the optimization of monitoring points, this score can be used to decide whether to retain, adjust or add monitoring points, so as to evaluate the importance of each monitoring point to the evaluation of the slope safety state. According to the real-time monitoring data of the slope and the results of the partitioning algorithm, it is judged whether the state of the slope has changed significantly. If the importance score of a monitoring point changes significantly, or the state of the slope changes significantly, the number of hidden units in the LSTM layer is dynamically adjusted according to the pre-set rules. , the adjustment range is , The adjustment step size is set according to the actual situation.

[0082] (7) Result output: Output the slope state prediction results, with the stable state marked as 1 and the unstable state marked as 0.

[0083] Step 4: Determine the optimization strategy for placing monitoring points in each partition based on the importance score, stable state and feature density, including: determining the minimum score threshold and the maximum score threshold based on the mean and standard deviation of the importance scores of all partitions; when the importance score of a partition is less than the minimum score threshold and is unstable, the optimization strategy for the partition is to increase the number of monitoring points; when the importance score of a partition is greater than the maximum score threshold and the feature density of the partition is greater than the partition density threshold, the optimization strategy for the partition is to reduce the number of monitoring points; when the partition is an abnormal partition, the optimization strategy for the partition is to increase the number of monitoring points.

[0084] For example, the importance score of the output Sort and calculate the mean of the scores and standard deviation , set the score threshold Then the feature density , calculate the average value of the density and standard deviation , set the density threshold .

[0085] Dynamic optimization of monitoring points is performed when the following three situations occur: 1) When , and the measurement points should be increased in partitions with unstable slope conditions.

[0086] 2) When but , retain the top 80% monitoring points with the highest scores and reduce the number of monitoring points.

[0087] 3) When a zone is identified as an abnormal area, the monitoring points should be encrypted and their locations optimized according to geological conditions.

[0088] In order to verify the effect of the embodiment of the present invention, the following embodiment is verified.

[0089] (1) Project Overview A large slope located on the west bank of the Yellow River in China was used, with an area of ​​about 0.52 km 2 The foot of the slope is about 1086m long, and the top of the slope is about 312m long. 32 monitoring points were initially deployed, and the monitoring period is 10 years from 2014 to 2024. Data are collected once a month, for a total of 12 time steps (T=12). The monitoring data includes two characteristic dimensions (D=2): displacement (M1) and stress (M2). At the same time, two geological characteristic dimensions (G1) and cohesion (G2) are introduced. ).

[0090] (2) Deformation partitioning results based on EXDBSCAN First, the displacement and stress monitoring data are standardized and the geological information is integrated to construct a normalized feature matrix with enhanced attributes. The constructed EXDBSCAN algorithm is used to deform and partition the monitoring points, traversing 32 monitoring points, and the training parameters are set as: neighborhood radius =50m, minimum number of samples = 5, =0.07 pieces / m 2 , the number of core points is 9. Taking the 8 core points as the starting point, the density connected points are expanded. After calculation, the slope is divided into 6 areas, namely P1, P2, P3, P4, P5, P6, and P7. Figure 4 As shown in Table 1, since the M15 measurement point does not belong to any cluster after calculation, the area where this point is located is marked as an abnormal area. The monitoring point numbers contained in each partition are shown in Table 1.

[0091] Table 1 Calculation partition and optimization results

[0092] (3) Dynamic optimization layout of high slope monitoring points by coupling EXDBSCAN-ATTCNN-LSTM The matrix that fuses partition features and spatiotemporal features The training set, validation set, and test set were divided into 8:1:1 ratios and input into the improved ATTCNN-LSTM model. The ATTCNN part contains 3 convolutional layers, with convolution kernel sizes of 3×3, 5×5, and 7×7, respectively, and the step size is 1; the LSTM part contains 2 LSTM layers, with the number of hidden units being 64 and 32, respectively. The cross entropy loss function and Adam optimizer were used in the training process, with a learning rate of 0.001. After 50 epochs of training, the accuracy of the model on the test set reached 87%. The output slope safety status is shown in Table 1, and the output monitoring point importance evaluation results are shown in Table 2.

[0093] Table 2 Results of importance assessment of monitoring points

[0094] According to the optimization results and the importance evaluation results of monitoring points, the following dynamic optimization strategies are given: 1) Add one more monitoring point within a radius of 50m around the M15 measuring point in the abnormal area to collect data more intensively and gain a deeper understanding of the deformation trend and mechanical response of the area.

[0095] 2) Two more monitoring points were added near the M16 measuring point with a low importance score (below the threshold of 0.5) in the deformation-active P6 area to make up for the lack of monitoring data in this area and improve the accuracy of deformation monitoring in this area.

[0096] 3) Remove the M measuring points from the adjacent M4 and M5 measuring points in area P1 where the importance score is high (higher than the threshold 0.8) but the monitoring point density exceeds the density threshold (0.07 / m²).

[0097] After completing the dynamic optimization and adjustment of the monitoring points, the monitoring data is collected again to construct a new matrix of fused partition features and spatiotemporal features. It is still divided into training set, validation set and test set in the ratio of 8:1:1, and input into the EXDBSCAN-ATTCNN-LSTM model for training according to the same parameters. The training results show that the accuracy of the optimized model is increased to 91%, indicating that the optimization strategy is effective and verifies the effectiveness of the method proposed in this invention.

[0098] The present invention combines the deformation characteristics of high slopes with deep learning technology to dynamically optimize the layout of monitoring points. It can maximize the monitoring effect with a limited number of monitoring points and achieve real-time monitoring and early warning of high slope deformation and stability. This method is highly scientific, effective and practical, and can significantly improve the efficiency and accuracy of high slope monitoring. The specific effects are as follows: (1) Improving monitoring accuracy: The proposed EXDBSCAN algorithm can be used to partition the monitoring points for deformation, and the slope can be divided into regions according to stability characteristics, making subsequent analysis more targeted. ATTCNN is responsible for extracting the spatial characteristics of the monitoring data, accurately capturing the spatial relationship between the monitoring points, and identifying the correlation between deformations at different positions of the slope; LSTM focuses on time series analysis and effectively mines the changing trend of monitoring data over time. The combination of the three processes the monitoring data in an all-round way, making the model more accurate in judging the state of the slope and greatly improving the prediction accuracy. The embodiment of the present invention combines deep learning with a dynamic optimization algorithm to provide a new idea and method for the layout of high slope monitoring points. This innovative method can give full play to the advantages of the two technologies, realize the intelligent and dynamic optimization layout of high slope monitoring points, improve the accuracy and efficiency of monitoring, and provide more reliable technical support for the safety monitoring of high slopes.

[0099] (2) Optimizing the layout of monitoring points: In the existing technology, the layout of monitoring points is mostly static, which is difficult to adjust in real time according to the slope status. The method of the present invention reasonably adjusts the number and location of monitoring points based on the importance scores of monitoring points output by the model. In areas with low importance scores and unstable slopes, the number of monitoring points is increased, and in areas with high scores but dense monitoring points, the number of monitoring points is reduced or adjusted to ensure a more reasonable distribution of monitoring points, achieve an optimized layout of monitoring points, improve monitoring efficiency and accuracy, avoid resource waste, and obtain the most critical monitoring information with the least number of monitoring points.

[0100] (3) Enhanced model adaptability: The EXDBSCAN-ATTCNN-LSTM coupling model can comprehensively consider multiple factors such as the spatial distribution of monitoring data, temporal changes, and geological characteristics of the slope. Whether facing complex terrain or different monitoring cycles and data types, it can be dynamically adjusted according to actual conditions, and has stronger adaptability.

[0101] (4) Early risk warning: The constructed model can more accurately analyze the slope status and predict the change trend. It can detect potential disaster risks such as landslides and collapses in advance, buy more time for taking protective measures, and effectively protect the safety of people and property around the slope.

[0102] These are only a few embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any technician familiar with the profession, without departing from the scope of the technical solution of the present invention, using the above-disclosed technical contents to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A method for dynamically optimizing the arrangement of high slope monitoring points, characterized in that: include: Acquire multi-source monitoring data of each monitoring point in a plurality of monitoring points in a high slope monitoring area; Determine the neighborhood radius and feature density of each monitoring point according to the multi-source monitoring data, and partition the monitoring area according to the neighborhood radius and feature density; Obtaining partition features of each partition, and inputting the partition features and the multi-source monitoring data into a deep learning model based on an attention mechanism to obtain an importance score and a stable state of each partition; Based on the importance score, stable state and feature density, the optimal strategy for placing monitoring points in each partition is determined.

2. The method for dynamic optimization arrangement of high slope monitoring points according to claim 1 is characterized in that: The multi-source monitoring data includes displacement spatiotemporal data and geological data. The neighborhood radius and feature density of each monitoring point are determined according to the multi-source monitoring data, specifically including: Determine the deformation feature weight according to the standard deviation and mean of the displacement spatiotemporal data; determine the geological feature weight according to the standard deviation and mean of the geological data; Determine the neighborhood radius of each monitoring point and the density of deformation features and the density of geological features within the neighborhood radius; The characteristic density of each monitoring point is determined according to the deformation characteristic weight, deformation characteristic density, geological characteristic weight and geological characteristic density.

3. The method for dynamic optimization arrangement of high slope monitoring points according to claim 1 or 2, characterized in that: Determine the neighborhood radius of each monitoring point, including: Sort the distances between each monitoring point and other monitoring points from small to large, and record the first distance as the nearest neighbor distance of the corresponding monitoring point; Determine the average value of the nearest neighbor distances corresponding to all monitoring points, and determine the standard deviation of the distances between the monitoring points based on the average value; Determine the neighborhood parameter based on the mean value and the standard deviation of the distance between monitoring points k , and the distance between each monitoring point and other monitoring points k The distance is taken as the neighborhood radius of the monitoring point.

4. The method for dynamic optimization arrangement of high slope monitoring points according to claim 1 is characterized in that: Partitioning the monitoring area according to the neighborhood radius and feature density specifically includes: Determine whether each monitoring point is a core point and determine the adaptive neighborhood radius of the core point according to the neighborhood radius and feature density; Each core point and all monitoring points within its adaptive neighborhood radius constitute a normal partition of the monitoring area; monitoring points that do not belong to the normal partition are regarded as noise points, and the neighborhood radius of the noise point is regarded as an abnormal partition.

5. The method for dynamic optimization arrangement of high slope monitoring points according to claim 4 is characterized in that: Determining whether each monitoring point is a core point according to the neighborhood radius and feature density specifically includes: Determine the density threshold and the minimum point count threshold; Determine whether the number of monitoring points within the neighborhood radius of each monitoring point is greater than or equal to the minimum point number threshold. If so, when the feature density of each monitoring point is greater than the density threshold, the corresponding monitoring point is considered to be a core point.

6. The method for dynamic optimization arrangement of high slope monitoring points according to claim 5 is characterized in that: Determine the minimum point threshold, including: The expected neighborhood radius is determined based on the average distance between all monitoring points; Obtain the number of monitoring points within the expected neighborhood radius corresponding to each monitoring point, and record the median of the corresponding numbers of all monitoring points as the minimum point count threshold.

7. The method for dynamic optimization arrangement of high slope monitoring points according to claim 6 is characterized in that: Determine the density threshold, including: Determine the characteristic density of each monitoring point within the corresponding expected neighborhood radius and the mean and standard deviation of all characteristic densities; The density threshold is determined based on the mean and standard deviation of all feature densities.

8. The method for dynamic optimization arrangement of high slope monitoring points according to claim 4 is characterized in that: Determining the adaptive neighborhood radius of the core point specifically includes: The quotient of the average value of the characteristic density of all monitoring points and the characteristic density of the core point is recorded as a first value; The adaptive neighborhood radius of the core point is determined according to the product of the neighborhood radius and the first value.

9. The method for dynamic optimization arrangement of high slope monitoring points according to claim 1, characterized in that: Inputting the partition features and the multi-source monitoring data into a deep learning model based on an attention mechanism specifically includes: Constructing a fusion matrix according to the partition characteristics and the multi-source monitoring data; Inputting the fusion matrix into a convolutional neural network layer to extract spatial features and obtain a feature map; The feature map is input into the CNN-LSTM model based on the attention mechanism.

10. The method for dynamic optimization arrangement of high slope monitoring points according to claim 1, characterized in that: According to the importance score, stable state and feature density, the optimization strategy for placing monitoring points in each partition is determined, including: Determine the minimum score threshold and the maximum score threshold based on the mean and standard deviation of the importance scores of all partitions; When the importance score of a partition is less than the minimum score threshold and is unstable, the optimization strategy for the partition is to increase the number of monitoring points; When the importance score of a partition is greater than the maximum score threshold and the feature density of the partition is greater than the partition density threshold, the optimization strategy for the partition is to reduce the number of monitoring points; When a partition is an abnormal partition, the optimization strategy for the partition is to increase the number of monitoring points.

Images