High slope cutting safety level evaluation method, device and equipment based on multi-source data

Abstract

The invention provides a high slope cutting safety level evaluation method, device and equipment based on multi-source data, and relates to the technical field of geological disaster prediction, and the method comprises the steps: evaluating a target high slope cutting safety level score corresponding to a to-be-monitored region according to multi-source high slope cutting monitoring data corresponding to the to-be-monitored region; performing correlation screening and contribution degree screening on a plurality of pre-configured alternative high-slope-cutting instability disaster-inducing factors to obtain a plurality of target high-slope-cutting instability disaster-inducing factors; constructing a sample data set based on the target high slope cutting safety level score and the target high slope cutting instability disaster-inducing factor; and training model parameters of a machine learning model by using the sample data set, wherein the trained machine learning model is used for evaluating a new target high slope cutting safety level score corresponding to the to-be-monitored area. According to the method, the precision of high slope cutting safety level evaluation can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster prediction, and in particular to a method, device and equipment for evaluating the safety level of a high-cut slope based on multi-source data. Background Art

[0002] High-cut slopes, also known as artificial slope-cutting projects, refer to slopes formed by modifying the terrain due to construction needs. Their maximum height is usually greater than or equal to 15 meters in rock slopes and greater than or equal to 8 meters in soil slopes. The main morphological characteristics of high-cut slopes are "steep" and "high", which makes them prone to geological disasters. The formation of high-cut slopes is mainly due to strong unloading effects, which will lead to reduced stability of the slopes and increase the risk of geological disasters such as landslides and collapses. There are many types of failure modes for high-cut slopes, including wedge-shaped plane sliding failure mode, collapse sliding failure mode, sliding tensile failure mode, and sliding fracturing failure mode. The analysis of these failure modes is crucial for the stability assessment and management of high-cut slopes.

[0003] For the treatment of high cut slopes, a series of engineering measures are usually required, such as support, drainage, reinforcement, etc. These measures are aimed at improving the stability of the slope and reducing the risk of geological disasters. In addition to engineering measures, it is also necessary to establish a complete monitoring and early warning system to monitor the stability of high cut slopes in real time, so as to timely discover and deal with potential geological disaster risks.

[0004] In recent years, with the rapid development of computer technology and the continuous progress of spatial information technology, especially the increasing application of remote sensing technology and geographic information system (GIS), more and more new technologies have been applied to the evaluation of the susceptibility of high-cut slope geological hazards. In the field of quantitative evaluation of high-cut slope geological hazards, machine learning and deep learning have become two mainstream and efficient methodological tools. However, the existing algorithms have relatively low prediction accuracy in specific applications, especially under sparse sample conditions. Summary of the invention

[0005] In view of this, an object of the present invention is to provide a method, device and equipment for evaluating the safety level of a high-cut slope based on multi-source data, which can significantly improve the accuracy of the safety level evaluation of a high-cut slope.

[0006] In a first aspect, the present invention provides a method for assessing the safety level of a high-cut slope based on multi-source data, comprising: According to the multi-source high-cut slope monitoring data corresponding to the monitored area, the safety level score of the target high-cut slope corresponding to the monitored area is evaluated; Perform correlation screening and contribution screening on multiple pre-configured candidate high-cut slope instability disaster factors to obtain multiple target high-cut slope instability disaster factors; Construct a sample data set based on the target high-cut slope safety grade score and the target high-cut slope instability disaster factor; The model parameters of the machine learning model are trained using the sample data set. The trained machine learning model is used to evaluate the safety level score of the new target high-cut slope corresponding to the monitored area according to the instability disaster factors of the new target high-cut slope corresponding to the monitored area.

[0007] In one embodiment, based on the multi-source high-cut slope monitoring data corresponding to the monitored area, the safety level score of the target high-cut slope corresponding to the monitored area is evaluated, including: For any monitoring data in the multi-source high-cut slope monitoring data corresponding to the area to be monitored, an initial high-cut slope safety level score corresponding to the area to be monitored is evaluated based on the monitoring data; The weight coefficient corresponding to each monitoring data is normalized, and the initial high-cut slope safety level score is fused using the normalized weight coefficient to obtain the target high-cut slope safety level score corresponding to the monitored area.

[0008] In one embodiment, the multi-source high-cut slope monitoring data includes group monitoring and group prevention data, professional monitoring data, real-time monitoring data, and InSAR deformation monitoring data; based on the monitoring data, the initial high-cut slope safety level score corresponding to the monitored area is evaluated, including: Based on the group monitoring and group prevention data, it is judged that there are abnormal events in the monitored area, and the abnormal events are judged to be cracks or abnormal damage of the retaining wall, and the initial high-cut slope safety level score is determined according to the judgment results; Based on professional monitoring data, real-time monitoring data and InSAR deformation monitoring data, the monthly displacement change rate of the monitored area is determined respectively. According to the mapping relationship between the interval of the monthly displacement change rate and the safety level score, the initial high-cut slope safety level score is determined.

[0009] In one embodiment, correlation screening and contribution screening are performed on a plurality of pre-configured candidate high-cut slope instability disaster factors to obtain a plurality of target high-cut slope instability disaster factors, including: Determine the correlation between a plurality of pre-configured candidate high-cut slope instability disaster factors, and select a plurality of intermediate high-cut slope instability disaster factors from the candidate high-cut slope instability disaster factors according to the correlation; The contribution of each intermediate high shear slope instability hazard factor to the high shear slope risk prediction is evaluated, and multiple target high shear slope instability hazard factors are screened out from the intermediate high shear slope instability hazard factors in descending order of contribution.

[0010] In one embodiment, the machine learning model adopts a decision tree model, and the model parameters of the decision tree model include a learning rate, a number of decision trees, a maximum depth of the decision tree, a depth required for internal node re-division, a minimum leaf node sample weight and a feature sampling ratio of each tree; the machine learning model is trained using a sample data set, including: Determine different parameter combinations consisting of learning rates, number of decision trees, maximum depth of decision trees, depth required for internal node re-division, minimum leaf node sample weight and feature sampling ratio of each tree. For each parameter combination, use the parameter combination to initialize the decision tree model, use the sample data set to train the initialized decision tree model, determine the performance indicators of the decision tree model after training based on the parameter combination, and determine the target parameter combination based on the performance indicators.

[0011] In one embodiment, the method further comprises: If new multi-source high-cut slope monitoring data corresponding to the monitored area is detected, the safety level score of the new target high-cut slope corresponding to the monitored area is evaluated based on the new multi-source high-cut slope monitoring data; The machine learning model is optimized based on the new target high-cut slope safety level score and target high-cut slope instability disaster factor.

[0012] In one embodiment, the target high-cut slope instability disaster factors include land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall and surface waters; the method also includes: Obtaining the rainfall in the target time period of the monitored area, and using the rainfall in the target time period of the monitored area to replace the rainfall in the target high-cut slope instability disaster factor to obtain a new target high-cut slope instability disaster factor combination in the target time period of the monitored area; and / or obtaining a new target high-cut slope instability disaster factor combination in the target time period of a new monitored area; The new combination of target high-cut slope instability disaster factors is input into the trained machine learning model so that the trained machine learning model outputs a new target high-cut slope safety level score.

[0013] In a second aspect, the present invention further provides a high slope safety level assessment device based on multi-source data, comprising: A grade assessment module is used to assess the safety grade score of the target high-cut slope corresponding to the monitored area based on the multi-source high-cut slope monitoring data corresponding to the monitored area; A factor screening module is used to screen the correlation and contribution of multiple pre-configured candidate high-cut slope instability disaster factors to obtain multiple target high-cut slope instability disaster factors; A data set construction module is used to construct a sample data set based on the target high-cut slope safety grade score and the target high-cut slope instability disaster factor; The model training module is used to train the model parameters of the machine learning model using the sample data set. The trained machine learning model is used to evaluate the safety level score of the new target high-cut slope corresponding to the monitored area based on the instability disaster factors of the new target high-cut slope corresponding to the monitored area.

[0014] In a third aspect, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement any one of the methods provided in the first aspect.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the first aspect.

[0016] The present invention provides a high-cut slope safety level assessment method, device and equipment based on multi-source data. First, according to the multi-source high-cut slope monitoring data corresponding to the monitored area, the target high-cut slope safety level score corresponding to the monitored area is assessed; and, a plurality of pre-configured candidate high-cut slope instability disaster factors are screened for correlation and contribution to obtain a plurality of target high-cut slope instability disaster factors; then, a sample data set is constructed based on the target high-cut slope safety level score and the target high-cut slope instability disaster factor; finally, the model parameters of the machine learning model are trained using the sample data set, and the trained machine learning model is used to assess the new target high-cut slope safety level score corresponding to the monitored area according to the new target high-cut slope instability disaster factor corresponding to the monitored area. The above method combines multi-source high-cut slope monitoring data to evaluate the target high-cut slope safety level score corresponding to the monitored area, ensuring the diversity and comprehensiveness of the data, and at the same time screening out the target high-cut slope instability disaster factors, and then constructs a sample data set for training the machine learning model based on the target high-cut slope safety level score and the target high-cut slope instability disaster factors, so as to train the model parameters of the machine learning model, which can significantly improve the performance and efficiency of the machine learning model in high-cut slope safety level assessment, and comprehensively improve the accuracy of high-cut slope safety level assessment.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic flow chart of a high-cut slope safety level assessment method based on multi-source data provided by an embodiment of the present invention; Figure 2 A technical framework diagram of a high-cut slope safety level assessment method based on multi-source data provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a high slope cutting safety level assessment device based on multi-source data provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0022] At present, the existing algorithms have relatively low prediction accuracy in specific applications, especially under sparse sample conditions. Based on this, the present invention provides a method, device and equipment for evaluating the safety level of high-cut slopes based on multi-source data. Under sparse sample conditions, the prediction accuracy is relatively high, which can effectively solve the problems in the background technology.

[0023] To facilitate understanding of this embodiment, firstly, a high slope safety level assessment method based on multi-source data disclosed in an embodiment of the present invention is described in detail. Figure 1 The flowchart of a method for evaluating the safety level of a high-cut slope based on multi-source data is shown in FIG. 1 , and the method mainly includes the following steps S102 to S108: Step S102, evaluating the safety grade score of the target high-cut slope corresponding to the area to be monitored based on the multi-source high-cut slope monitoring data corresponding to the area to be monitored.

[0024] Among them, multi-source high-cut slope monitoring data includes group monitoring and group prevention data, professional monitoring data, real-time monitoring data and InSAR deformation monitoring data. The target high-cut slope safety level score is used to describe the safety level of the high-cut slope in the monitored area. For example, the higher the score, the higher the risk (worse safety) of the monitored area. In one example, a safety level evaluation model is established in combination with the multi-source high-cut slope monitoring data corresponding to the monitored area to evaluate the target high-cut slope safety level score corresponding to each high-cut slope in the monitored area.

[0025] Step S104, performing correlation screening and contribution screening on a plurality of pre-configured candidate high-cut slope instability disaster factors to obtain a plurality of target high-cut slope instability disaster factors.

[0026] Among them, correlation screening can be understood as a screening operation based on the correlation between high-cut slope instability disaster factors, and contribution screening can be understood as a screening operation based on the contribution of high-cut slope instability disaster factors to high-cut slope risk prediction. The target high-cut slope instability disaster factors finally screened include land use type, slope aspect map, support structure / whether secondary reinforcement is required / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall and surface water area.

[0027] Step S106: construct a sample data set based on the target high-cut slope safety grade score and the target high-cut slope instability disaster factor.

[0028] In one example, the target high-cut slope safety grade score is used as a training label, and the target high-cut slope instability disaster factor is used as a model input to obtain a sample data set.

[0029] Step S108, using the sample data set to train the model parameters of the machine learning model, the trained machine learning model is used to evaluate the safety level score of the new target high cut slope corresponding to the monitored area according to the instability disaster factor of the new target high cut slope corresponding to the monitored area.

[0030] The machine learning model may adopt a decision tree (XGBoost) model. In one example, the sample data set may be divided into a training data set and a validation data set, and the parameters of the machine learning model may be iteratively optimized together.

[0031] In one embodiment, the rainfall in the target time period of the monitored area is obtained, and the rainfall in the original target high-cut slope instability disaster factor is replaced to obtain a new target high-cut slope instability disaster factor, which is input into the trained machine learning model, and the model will output a new target high-cut slope safety level score corresponding to each high-cut slope in the monitored area. In another embodiment, for a new monitored area, a new target high-cut slope instability disaster factor can be directly obtained, and the model can be used to output a new target high-cut slope safety level score corresponding to each high-cut slope in the new monitored area, so as to achieve a high-cut slope safety level assessment for the new monitored area in the absence of monitoring data for the new monitored area.

[0032] The high-cut slope safety level assessment method based on multi-source data provided by an embodiment of the present invention combines multi-source high-cut slope monitoring data to assess the target high-cut slope safety level score corresponding to the monitored area, thereby ensuring the diversity and comprehensiveness of the data, while screening out the target high-cut slope instability disaster factors, and then constructing a sample data set for training a machine learning model based on the target high-cut slope safety level score and the target high-cut slope instability disaster factors, thereby realizing the training of the model parameters of the machine learning model, which can significantly improve the performance and efficiency of the machine learning model in high-cut slope safety level assessment, and comprehensively improve the accuracy of high-cut slope safety level assessment.

[0033] For easier understanding, see Figure 2The technical framework diagram of a high-cut slope safety level assessment method based on multi-source data is shown, including: combining monitoring data (group monitoring and prevention, professional monitoring, and automated monitoring), InSAR satellite data, and historical high-cut slope early warning data to generate a historical monthly high-cut slope risk assessment table by month, wherein automated monitoring is also real-time monitoring, and the historical monthly high-cut slope risk assessment table describes the target high-cut slope safety level score corresponding to the monitored area in each month; obtaining land use type based on remote sensing images, obtaining slope, slope direction, and cumulative catchment area based on contour data, obtaining support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data based on historical maintenance records in high-cut slope statistical documents, and obtaining rainfall based on meteorological data. Based on the topographic map, surface water areas are obtained to obtain the instability disaster factors of high cut slopes; the high cut slope risk level assessment model is trained in combination with the historical monthly high cut slope risk assessment table and the instability disaster factors of high cut slopes. The high cut slope risk level assessment model is also the aforementioned machine learning model, which adopts a decision tree (XGBoost) model; after the model training, the current meteorological conditions or meteorological forecast data can be obtained for the monitored area to estimate the precipitation in the monitored area, and the estimated precipitation is combined with other high cut slope instability disaster factors and input into the trained high cut slope risk level assessment model, which outputs the current status of the monitored area or a list of high cut slopes that may pose risks within the next 48 hours, so as to achieve the safety level assessment of the high cut slopes in the monitored area. In addition, after model training, for the new monitored area, in addition to obtaining the current meteorological conditions or meteorological forecast data, other high-cut slope instability disaster factors need to be obtained and input into the trained high-cut slope risk level assessment model to output the current status of the new monitored area or a list of high-cut slopes that may pose risks within the next 48 hours, thereby realizing the safety level assessment of the high-cut slopes in the new monitored area.

[0034] Based on this, an embodiment of the present invention provides a specific implementation of a high-cut slope safety level assessment method based on multi-source data.

[0035] For the aforementioned step S102, the embodiment of the present invention provides a specific implementation method for evaluating the safety level score of the target high-cut slope corresponding to the monitored area according to the multi-source high-cut slope monitoring data corresponding to the monitored area, including the following steps 1a to 1b: Step 1a: for any monitoring data in the multi-source high-cut slope monitoring data corresponding to the area to be monitored, an initial high-cut slope safety level score corresponding to the area to be monitored is evaluated based on the monitoring data.

[0036] Among them, multi-source high-cut slope monitoring data includes group monitoring and prevention data, professional monitoring data, real-time monitoring data and InSAR deformation monitoring data. The sources of group monitoring and prevention, professional monitoring, real-time monitoring and InSAR satellite deformation data are as follows: Mass monitoring and prevention data: This type of data usually comes from daily observations and reports from local residents or community organizations, including descriptions and photos of cracks, sliding, and subsidence on high-cut slopes. Although these data are highly subjective, they can provide first-hand on-site information and help to detect potential risks in a timely manner.

[0037] Professional monitoring data: Data obtained by geological experts or engineering teams using professional instruments such as total stations, GPS (Global Positioning System), and tilt sensors for regular or continuous monitoring. These data are highly accurate and can accurately reflect key indicators such as displacement and deformation of high-cut slopes, and are an important basis for evaluating the stability of high-cut slopes.

[0038] Real-time monitoring data: data transmitted in real time by automated monitoring equipment installed on high-cut slopes, such as remote video monitoring and automated data acquisition systems. This type of data can reflect the dynamic changes of high-cut slopes in real time and is of great value for early warning and emergency response.

[0039] InSAR satellite deformation data: high-cut slope surface deformation data obtained using synthetic aperture radar interferometry (InSAR). InSAR technology can penetrate clouds and rain and fog to achieve large-scale, high-precision surface deformation monitoring, especially suitable for areas with complex terrain and inconvenient transportation.

[0040] Specifically, the collection of historical monitoring data of high-cut slopes requires comprehensive consideration of multiple sources and technical means to ensure the comprehensiveness and accuracy of the data. At the same time, it is also necessary to establish a sound data management and analysis system to effectively process and utilize the collected data to provide strong support for stability assessment and disaster warning of high-cut slopes.

[0041] Based on the above monitoring data, an embodiment of the present invention provides an implementation method for evaluating the initial high-cut slope safety level score corresponding to the monitored area based on the monitoring data, see the weight coefficients and safety level score table of various monitoring data shown in Table 1: Table 1 Weight coefficients and safety level scores of various monitoring data

[0042] In the specific implementation: (1) Based on the group monitoring and group prevention data, it is judged that there are abnormal events in the monitored area, and the abnormal events are judged to be cracks or abnormal damage of the retaining wall, and the initial high-cut slope safety level score is determined according to the judgment results. Please continue to refer to Table 1. If there is a situation described in Table 1 that "the overall situation is abnormal, and cracks or abnormal damage of the retaining wall occur", the initial high-cut slope safety level score obtained based on the evaluation of this type of monitoring data is determined to be "1".

[0043] (2) Based on professional monitoring data, real-time monitoring data and InSAR deformation monitoring data, the monthly displacement change rate of the monitored area is determined respectively, and the initial high-cut slope safety level score is determined according to the mapping relationship between the monthly displacement change rate range and the safety level score. Please refer to Table 1. Taking professional monitoring data as an example, the monthly displacement change rate of the monitored area is determined according to the professional monitoring data. If the monthly displacement change rate is within the range of 0-2.5mm, the initial high-cut slope safety level score obtained based on the evaluation of this type of monitoring data is determined to be "0"; if the monthly displacement change rate is within the range of 2.5-5mm, the initial high-cut slope safety level score obtained based on the evaluation of this type of monitoring data is determined to be "0.25"; if the monthly displacement change rate is within the range of 5-10mm, the initial high-cut slope safety level score obtained based on the evaluation of this type of monitoring data is determined to be "0.5"; if the monthly displacement change rate is within the range of 10-25mm, the initial high-cut slope safety level score obtained based on the evaluation of this type of monitoring data is determined to be "1".

[0044] Step 1b, normalizing the weight coefficient corresponding to each monitoring data, and using the normalized weight coefficient to fuse the initial high-cut slope safety level score to obtain the target high-cut slope safety level score corresponding to the monitored area.

[0045] The target high-cut slope safety grade score can be determined according to the following formula: ; in, Score the safety level of the target high-cut slope. The initial high-cut slope safety level score obtained from various monitoring data, is the weight coefficient of each type of monitoring data (as shown in Table 1). For example, referring to the Technical Specifications for Highway Slope Monitoring (DB42 / T 1496-2019), the safety level of high cut slopes is divided into 5 levels, and the safety level evaluation model is shown in Table 2 below: Table 2 Safety level evaluation model

[0046] For the aforementioned step S104, the embodiment of the present invention provides an implementation method for performing correlation screening and contribution screening on a plurality of pre-configured candidate high-cut slope instability disaster factors to obtain a plurality of target high-cut slope instability disaster factors, see the following steps 2a to 2b: Step 2a: determining the correlation between a plurality of pre-configured candidate high-cut slope instability disaster factors, and selecting a plurality of intermediate high-cut slope instability disaster factors from the candidate high-cut slope instability disaster factors according to the correlation.

[0047] In one example, the Pearson correlation coefficient, Spearman correlation coefficient and other methods can be used to determine the correlation between the candidate high-cut slope instability disaster factors, and the initial high-cut slope instability disaster factors with high correlation (generally referring to the absolute similarity value greater than or equal to 0.9) can be found through pairing. One of the factors is directly selected and retained, and the other factor is discarded. According to the results of all pairwise correlation analysis, the intermediate high-cut slope instability disaster factors are obtained to reduce the complexity and calculation amount of the model.

[0048] Step 2b, evaluate the contribution of each intermediate high shear slope instability hazard factor to the high shear slope risk prediction, and select multiple target high shear slope instability hazard factors from the intermediate high shear slope instability hazard factors in descending order of contribution.

[0049] In one example, the SHAP value analysis method of the artificial intelligence interpretable model was used to evaluate the contribution of each intermediate high cut slope instability hazard factor to the high cut slope risk prediction; the intermediate high cut slope instability hazard factor with higher importance was selected as the target high cut slope instability hazard factor, which was also the input of the model, to improve the performance and generalization ability of the model.

[0050] In a specific application example, the target high-cut slope instability disaster factors screened by the embodiment of the present invention include disaster factor data such as land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall, surface waters, etc. in the monitoring area at each period, as shown in Table 3 below: Table 3 Factors causing instability of high-cut slopes

[0051] Specifically, the land use type in the monitoring area affects the stability of the high cut slope. Different land use methods may produce different pressures and impacts on the slope. The slope aspect diagram determines the direction and intensity of water flow and gravity on the high cut slope, and is an important factor in assessing the risk of the high cut slope. The support structure / whether secondary reinforcement is required / engineering quality evaluation / slope geological condition data determine the physical properties and mechanical characteristics of the high cut slope. Different structural conditions have different effects on the stability of the high cut slope. Longitude / latitude / district and county determine the land use, infrastructure construction, regulatory requirements, governance measures and other factors in the area. Therefore, the geographical location distribution of the high cut slope is closely related to its stability, and it is necessary to comprehensively consider the geographical location and administrative divisions for evaluation and analysis. Precipitation and surface water are important external factors. An increase in rainfall may lead to slope saturation and reduce its stability. Different distances from the surface water will affect the stress inside the slope, thereby affecting the stability of the high cut slope.

[0052] For the aforementioned step S106, an embodiment of the present invention provides an implementation method for constructing a sample data set based on the target high-cut slope safety level score and the target high-cut slope instability disaster factor, that is, the target high-cut slope safety level score is used as a training label, and the target high-cut slope instability disaster factor is used as a model input.

[0053] Preferably, before using the sample data set to train the decision tree (XGBoost) model, the sample data set may be preprocessed, including data cleaning, monitoring data standardization and normalization, and disaster factor normalization: Data cleaning: Clean, standardize and normalize all input data to ensure data quality and consistency. Specifically, if missing values ​​in monitoring data are checked and processed, interpolation methods (such as linear interpolation, mean interpolation, etc.) can be used to fill in missing data points to ensure data continuity. Identify and remove outliers. For monitoring data that obviously deviates from the normal range, statistical methods (such as the 3σ criterion) can be used to identify them and correct or delete them according to the specific situation.

[0054] Standardization and normalization of monitoring data: Standardize different types of monitoring data to make them have the same dimension and scale.

[0055] Normalization of hazard factors: Normalize the hazard factors of high-cut slope instability and map their values ​​to the interval [0, 1] to facilitate model training and comparison. In one example, categorical variable encoding (LabelEncoder) can be used to convert categorical data (such as high-cut slope safety level) into a format that is easy for machine learning algorithms to process. Taking categorical variable encoding as an example, categorical variable encoding can convert the target high-cut slope instability hazard factor into a binary vector, so that it can be used by most machine learning algorithms, preventing the machine learning model from incorrectly interpreting the possible numerical relationship between categorical values.

[0056] For the aforementioned step S108, the decision tree (XGBoost) model in the embodiment of the present invention is used to establish the correlation between the target high-cut slope instability disaster factor (input variable X) and the target high-cut slope safety level score (output variable Y). Among them, the X decision tree (XGBoost) model is an additive combination of multiple decision trees, and the core mechanism is the optimization process of the gradient boosting tree (GBDT), which improves the model prediction results by gradually reducing the residual or fitting the gradient. The expression of the decision tree (XGBoost) model is as follows: ; in: is the final prediction function, which is accumulated from multiple trees; is the initial prediction value (usually the sample mean); is the mth tree, indicating the fitting of the residual by the current round learner; M is the number of trees. It should be noted that the predicted value here is the safety grade score of the high-cut slope.

[0057] The decision tree (XGBoost) model adjusts the model prediction value through the loss function, and its loss function is: ; ; The first part is the gradient boosting algorithm loss, and the second part is the regularization term. represents the loss function, It can be understood as the set of all trainable parameters in the decision tree (XGBoost) model, n is the number of training function samples; l is the loss of a single sample; is the predicted value of safety level of high cutting slope; is the true value of the high slope cutting safety level; Ω( ) is the regularization term of the mth tree; and λ are manually set parameters; w is the vector formed by the values ​​of all leaf nodes of the decision tree; T is the number of leaf nodes.

[0058] For the classification task, the final prediction value (i.e., the target high-cut slope safety level score) is obtained by the additive combination of all trees: ; In the decision tree (XGBoost) model, the construction process of a single decision tree adopts the structure of the CART (Classification and Regression Trees) tree and is optimized based on gradients and second-order derivatives.

[0059] Based on this, an embodiment of the present invention provides an implementation method for optimizing the hyperparameters of a machine learning model, wherein the hyperparameters include a learning rate, the number of decision trees, the maximum depth of the decision tree, the depth required for internal node redivision, the minimum leaf node sample weight and, and the feature sampling ratio of each tree. Specifically, refer to the following steps: determine a parameter combination consisting of different learning rates, the number of decision trees, the maximum depth of the decision tree, the depth required for internal node redivision, the minimum leaf node sample weight and, and the feature sampling ratio of each tree; for each parameter combination, use the parameter combination to initialize a decision tree (XGBoost) model, use a sample data set to train the initialized decision tree (XGBoost) model, determine the performance indicators of the decision tree (XGBoost) model trained based on the parameter combination, and determine the target parameter combination based on the performance indicators. The specific process is as follows: (1) Learning rate: It is a hyperparameter that controls the influence of each newly added decision tree on the final prediction. It determines the update amplitude of the model at each step in the training process and balances the learning speed and the generalization ability of the model.

[0060] Specifically, if the learning rate is too high, the model may converge quickly within a smaller number of iterations, but the final performance may not be ideal (prone to fall into local optimality); lowering the learning rate will greatly increase the training time, but will make the training process more stable and have stronger generalization ability, but more decision trees (n_estimators) are required to achieve the same effect.

[0061] Use the grid search method to select a suitable learning rate between 0.01-1 to ensure high prediction accuracy while ensuring the training efficiency of the model.

[0062] (2) Number of decision trees (n_estimators): This is the number of decision trees iteratively generated during the XGBoost training process.

[0063] Specifically, generally speaking, increasing the number of decision trees can improve the stability and accuracy of the model, but it will also increase the calculation time. For high-cut slope monitoring data, more decision trees may help to better capture the complex relationship between different monitoring indicators, thereby improving the accuracy of high-cut slope risk prediction. For example, when dealing with a large amount of professional monitoring (monthly displacement change rate), real-time monitoring (monthly displacement change rate) and InSAR satellite deformation data, more decision trees can more comprehensively analyze the characteristics of these data and reduce the impact of randomness.

[0064] Use the grid search method to select an appropriate number of decision trees between 100 and 1000 to prevent model overfitting while ensuring high prediction accuracy.

[0065] (3) Maximum depth of decision tree (max_depth): Start from a smaller depth and gradually increase the maximum depth of the decision tree to observe the performance changes of the model.

[0066] Specifically, too large a depth will lead to overfitting, while too small a depth may cause the model to underfit. When considering high-cut slope monitoring data, the choice of depth needs to balance the degree of fit to the data and the generalization ability of the model. If the depth is too large, it may overfit the high-cut slope monitoring data of a specific time period or a specific area, and fail to adapt well to future changes. For example, there may be abnormal fluctuations in the high-cut slope monitoring data for certain periods of time. If the decision tree depth is too large, these outliers may be over-learned, resulting in inaccurate predictions of future risks.

[0067] Use methods such as grid search to search for the optimal maximum depth within a certain range (3-10) to improve the performance and generalization ability of the model.

[0068] (4) Maximum depth required for internal node repartitioning (max_depth): Parameter adjustment can control the growth of the decision tree and prevent overfitting.

[0069] Specifically, a larger value can prevent the decision tree from overgrowing, but may cause the model to underfit. A smaller value allows the decision tree to grow deeper, but it is prone to overfitting. For high-cut slope monitoring data, the adjustment of these two parameters needs to consider the data volume and data stability of different monitoring types. For example, if the sample size of professional monitoring data and real-time monitoring data is large and relatively stable, the minimum number of samples required for internal node redivision can be appropriately reduced to allow the decision tree to more deeply explore the potential relationships in the data. For InSAR satellite deformation data with a smaller sample size, it may be necessary to increase the minimum number of samples required for leaf nodes to prevent overfitting.

[0070] Through methods such as grid search, we try different parameter combinations to find a suitable range of values ​​so that the model can achieve good performance on both the training set and the test set.

[0071] (5) The minimum leaf node sample weight sum (min_child_weight) during each tree training: the sum of the first-order gradients (residuals) of all samples in each leaf node. It is an important parameter that affects the complexity and generalization ability of the model. Its function is to limit the splitting of leaf nodes and avoid excessive growth of trees, which will affect the performance of the model.

[0072] Specifically, when the minimum leaf node sample weight and value are small, the leaf node can be easily split, the resulting tree will be deeper and more complex, and the model can fit the complex patterns of the data. When the minimum leaf node sample weight and value are large, the growth of the tree will be restricted, the resulting tree will be shallower, and the model complexity will be reduced.

[0073] Using the grid search method, we search for the optimal minimum leaf node sample weight and between 1 and 50, which improves the model's ability to capture subtle patterns in the data while ensuring the model's generalization ability.

[0074] (6) The feature sampling ratio of each tree used when training each tree (colsample_bytree): controls the ratio of the feature subset used when training each tree.

[0075] Specifically, when processing high-cut slope monitoring data and disaster factor data, it is necessary to select an appropriate maximum number of features based on the characteristics of the data. For example, if there are multiple strongly correlated features, you can consider reducing the value of colsample_bytree to avoid overfitting of the model. For example, if there are many highly correlated features (correlation exceeds 0.9), you can set colsample_bytree to 0.7 or 0.8. If the correlation between features is low, you can consider setting colsample_bytree to a higher value (such as 0.9 or 1.0) to ensure that the model can use more information for learning. Smaller values ​​can reduce the risk of overfitting, but may cause the model to underfit. Larger values ​​allow the decision tree to make more full use of feature information, but it is easy to overfit.

[0076] Through the grid search method, this parameter is adjusted between 0.5-1 to find a suitable balance between model complexity and generalization ability.

[0077] Furthermore, the embodiment of the present invention also includes real-time updating, including: if new multi-source high-cut slope monitoring data corresponding to the monitored area is monitored, then the new target high-cut slope safety level score corresponding to the monitored area is evaluated based on the new multi-source high-cut slope monitoring data; the machine learning model is optimized based on the new target high-cut slope safety level score and the target high-cut slope instability disaster factor. The embodiment of the present invention optimizes the decision tree (XGBoost) model and re-evaluates the safety level of the high-cut slope by updating the monitoring data and the corresponding initial high-cut slope safety level score in real time, as well as updating the high-cut slope disaster factor. In specific implementation, new data is collected regularly to retrain and optimize the model. Specifically, the appropriate update cycle can be determined based on the data update frequency and the performance changes of the model. When updating the model, a combination of new data and historical data can be used to make full use of historical information and improve the stability and generalization ability of the model.

[0078] Furthermore, the embodiment of the present invention also provides an implementation method of using a trained decision tree (XGBoost) model to output a new target high-cut slope safety level score for the monitored area in a target time period based on a new target high-cut slope instability disaster factor, including the following steps 4a to 4c: Step 4a, obtaining the rainfall in the monitored area in the target time period. The target time period may be the current time or the forecast time (such as the next 48 hours). In one example, the current meteorological conditions or meteorological forecast data may be obtained to obtain the rainfall at the current time or the rainfall in the next 48 hours.

[0079] Step 4b, using the rainfall in the target time period in the monitored area to replace the rainfall in the target high-cut slope instability factor, so as to obtain the new target high-cut slope instability disaster factor in the monitored area in the target time period. Considering that the rainfall in the target high-cut slope instability disaster factor changes significantly over time, while the disaster factor data such as land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data and surface water area do not change significantly over time, the rainfall in the target time period can be combined with the original land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data and surface water area to form a new target high-cut slope instability disaster factor.

[0080] Step 4c, inputting the new target high-cut slope instability disaster factor into the trained machine learning model, so that the trained machine learning model outputs a new target high-cut slope safety level score for the monitored area in the target time period.

[0081] Further, the embodiment of the present invention also provides an implementation method of using a trained decision tree (XGBoost) model to output a new target high-cut slope safety level score for a new area to be monitored in a target time period based on a new target high-cut slope instability disaster factor, including: obtaining a combination of new target high-cut slope instability disaster factors for a new area to be monitored in a target time period; inputting the combination of new target high-cut slope instability disaster factors into a trained machine learning model, so that the trained machine learning model outputs a new target high-cut slope safety level score. In specific implementation, for a new area to be monitored, in addition to rainfall, other high-cut slope instability disaster factors will also change, so the land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall and surface waters of the new area to be monitored can be obtained, and the above-mentioned high-cut slope instability disaster factors are input into the trained machine learning model, so that it outputs the target high-cut slope safety level score corresponding to the new area to be monitored, so as to evaluate the safety level of the high-cut slope in the absence of monitoring data.

[0082] In summary, the high slope safety level assessment method based on multi-source data provided by the embodiment of the present invention has at least the following characteristics: 1. Data diversity and comprehensiveness: The embodiment of the present invention combines multiple data sources such as group monitoring and prevention, professional monitoring, real-time monitoring and InSAR satellite deformation data, ensuring data diversity and comprehensiveness; this multi-source data fusion method can provide more comprehensive high-cut slope status information, which helps to improve the accuracy of assessment; under sparse sample conditions, the prediction accuracy is relatively high, and it has significant advantages in improving assessment accuracy, reliability, dynamic adaptability and decision support.

[0083] 2. High precision and reliability: By using the XGBoost model and combining a variety of monitoring data, the embodiments of the present invention can effectively handle complex nonlinear relationships and improve the accuracy and reliability of predictions; the integrated learning characteristics of the XGBoost model also help reduce the risk of overfitting and improve the generalization ability of the model.

[0084] 3. Dynamic update and adaptability: The embodiments of the present invention support real-time updating of monitoring data and corresponding high-cut slope safety levels, and can adapt to environmental changes and new monitoring data; this dynamic update mechanism enables the model to be continuously optimized to maintain an accurate assessment of high-cut slope risks.

[0085] 4. Flexible feature selection and parameter optimization: The embodiments of the present invention include feature selection and hyperparameter optimization steps, which can be adjusted according to specific application scenarios and data characteristics; this helps to find the best model configuration and improve the performance and efficiency of the model.

[0086] 5. Easy to implement and expand: Since the embodiment of the present invention is based on a machine learning algorithm, it is relatively easy to implement and deploy; at the same time, it also has good scalability, and can easily add new data sources or adjust the model structure to meet different needs.

[0087] 6. Comprehensive assessment and decision support: The embodiment of the present invention not only provides a safety level assessment for high-cut slopes, but also can output a list of areas where risks may occur in the future; this provides important reference information for decision makers and helps to formulate more effective disaster prevention and mitigation measures.

[0088] Based on the above embodiments, the present invention provides a high slope safety level assessment device based on multi-source data. Figure 3 The schematic diagram of the structure of a high slope safety level assessment device based on multi-source data is shown, comprising the following parts: The level evaluation module 302 is used to evaluate the safety level score of the target high-cut slope corresponding to the monitored area according to the multi-source high-cut slope monitoring data corresponding to the monitored area; The factor screening module 304 is used to perform correlation screening and contribution screening on a plurality of pre-configured candidate high-cut slope instability disaster factors to obtain a plurality of target high-cut slope instability disaster factors; A data set construction module 306 is used to construct a sample data set based on the target high-cut slope safety grade score and the target high-cut slope instability disaster factor; The model training module 308 is used to train the model parameters of the machine learning model using the sample data set. The trained machine learning model is used to evaluate the safety level score of the new target high-cut slope corresponding to the monitored area according to the instability disaster factors of the new target high-cut slope corresponding to the monitored area.

[0089] The high-cut slope safety level assessment device based on multi-source data provided by an embodiment of the present invention combines multi-source high-cut slope monitoring data to assess the target high-cut slope safety level score corresponding to the monitored area, thereby ensuring the diversity and comprehensiveness of the data, while screening out the target high-cut slope instability disaster factors, and then constructing a sample data set for training a machine learning model based on the target high-cut slope safety level score and the target high-cut slope instability disaster factors, thereby realizing the training of the model parameters of the machine learning model, which can significantly improve the performance and efficiency of the machine learning model in high-cut slope safety level assessment, and comprehensively improve the accuracy of high-cut slope safety level assessment.

[0090] In one implementation, the level assessment module 302 is specifically configured to: For any monitoring data in the multi-source high-cut slope monitoring data corresponding to the area to be monitored, an initial high-cut slope safety level score corresponding to the area to be monitored is evaluated based on the monitoring data; The weight coefficient corresponding to each monitoring data is normalized, and the initial high-cut slope safety level score is fused using the normalized weight coefficient to obtain the target high-cut slope safety level score corresponding to the monitored area.

[0091] In one embodiment, the multi-source high-cut slope monitoring data includes group monitoring and prevention data, professional monitoring data, real-time monitoring data and InSAR deformation monitoring data; the level assessment module 302 is specifically used for: Based on the group monitoring and group prevention data, it is judged that there are abnormal events in the monitored area, and the abnormal events are judged to be cracks or abnormal damage of the retaining wall, and the initial high-cut slope safety level score is determined according to the judgment results; Based on professional monitoring data, real-time monitoring data and InSAR deformation monitoring data, the monthly displacement change rate of the monitored area is determined respectively. According to the mapping relationship between the interval of the monthly displacement change rate and the safety level score, the initial high-cut slope safety level score is determined.

[0092] In one implementation, the factor screening module 304 is specifically used to: Determine the correlation between a plurality of pre-configured candidate high-cut slope instability disaster factors, and select a plurality of intermediate high-cut slope instability disaster factors from the candidate high-cut slope instability disaster factors according to the correlation; The contribution of each intermediate high shear slope instability hazard factor to the high shear slope risk prediction is evaluated, and multiple target high shear slope instability hazard factors are screened out from the intermediate high shear slope instability hazard factors in descending order of contribution.

[0093] In one embodiment, the machine learning model adopts a decision tree model, and the model parameters of the decision tree model include a learning rate, a number of decision trees, a maximum depth of a decision tree, a depth required for internal node re-division, a minimum leaf node sample weight and a feature sampling ratio of each tree; the model training module 308 is specifically used for: Determine different parameter combinations consisting of learning rates, number of decision trees, maximum depth of decision trees, depth required for internal node re-division, minimum leaf node sample weight and feature sampling ratio of each tree. For each parameter combination, use the parameter combination to initialize the decision tree model, use the sample data set to train the initialized decision tree model, determine the performance indicators of the decision tree model after training based on the parameter combination, and determine the target parameter combination based on the performance indicators.

[0094] In one embodiment, a real-time update and optimization module 310 is further included, which is used to: If new multi-source high-cut slope monitoring data corresponding to the monitored area is detected, the safety level score of the new target high-cut slope corresponding to the monitored area is evaluated based on the new multi-source high-cut slope monitoring data; The machine learning model is optimized based on the new target high-cut slope safety level score and target high-cut slope instability disaster factor.

[0095] In one embodiment, the target high-cut slope instability disaster factors include land use type, slope aspect map, support structure / whether secondary reinforcement / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall and surface waters; and also include a safety level classification and evaluation module 312 for: Obtaining the rainfall in the target time period in the area to be monitored, and using the rainfall in the target time period in the area to be monitored to replace the rainfall in the target high-cut slope instability factor, so as to obtain a new combination of target high-cut slope instability disaster factors in the area to be monitored in the target time period; and / or, obtaining a new combination of target high-cut slope instability disaster factors in a new area to be monitored in the target time period; The new combination of target high-cut slope instability disaster factors is input into the trained machine learning model so that the trained machine learning model outputs a new target high-cut slope safety level score.

[0096] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0097] An embodiment of the present invention provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned implementation methods.

[0098] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43, wherein the processor 40, the communication interface 43 and the memory 41 are connected via the bus 42; the processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0099] The memory 41 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0100] The bus 42 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0101] Among them, the memory 41 is used to store programs, and the processor 40 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0102] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 40. The above processor 40 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and completes the steps of the above method in combination with its hardware.

[0103] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0104] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0105] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A high-cut slope safety level assessment method based on multi-source data, characterized in that: include: According to the multi-source high-cut slope monitoring data corresponding to the monitored area, the safety level score of the target high-cut slope corresponding to the monitored area is evaluated; Perform correlation screening and contribution screening on multiple pre-configured candidate high-cut slope instability disaster factors to obtain multiple target high-cut slope instability disaster factors; Constructing a sample data set based on the safety grade score of the target high-cut slope and the instability disaster factor of the target high-cut slope; The model parameters of the machine learning model are trained using the sample data set, and the trained machine learning model is used to evaluate the safety level score of the new target high-cut slope corresponding to the monitored area based on the new instability disaster factor of the target high-cut slope corresponding to the monitored area.

2. The high-cut slope safety level assessment method based on multi-source data according to claim 1 is characterized in that: According to the multi-source high-cut slope monitoring data corresponding to the monitored area, the safety level score of the target high-cut slope corresponding to the monitored area is evaluated, including: For any monitoring data in the multi-source high-cut slope monitoring data corresponding to the area to be monitored, evaluating the initial high-cut slope safety level score corresponding to the area to be monitored based on the monitoring data; The weight coefficient corresponding to each monitoring data is normalized, and the initial high-cut slope safety level score is integrated with the normalized weight coefficient to obtain the target high-cut slope safety level score corresponding to the monitored area.

3. The high-cut slope safety level assessment method based on multi-source data according to claim 2 is characterized in that: The multi-source high-cut slope monitoring data includes group monitoring and prevention data, professional monitoring data, real-time monitoring data and InSAR deformation monitoring data; The initial high-cut slope safety level score corresponding to the monitored area is evaluated based on the monitoring data, including: Based on the group detection and group prevention data, it is judged that there is an abnormal event in the monitored area, and the abnormal event is judged to be a crack or abnormal damage of the retaining wall, and an initial high-cut slope safety grade score is determined according to the judgment result; Based on the professional monitoring data, the real-time monitoring data and the InSAR deformation monitoring data, the monthly displacement change rate of the monitored area is determined respectively, and the initial high-cut slope safety level score is determined according to the mapping relationship between the interval of the monthly displacement change rate and the safety level score.

4. The high-cut slope safety level assessment method based on multi-source data according to claim 1 is characterized in that: The pre-configured multiple candidate high-cut slope instability disaster factors are screened for correlation and contribution, and multiple target high-cut slope instability disaster factors are obtained, including: Determine the correlation between a plurality of pre-configured candidate high-cut slope instability disaster factors, and select a plurality of intermediate high-cut slope instability disaster factors from the candidate high-cut slope instability disaster factors according to the correlation; The contribution of each of the intermediate high shear slope instability hazard factors to the high shear slope risk prediction is evaluated, and a plurality of target high shear slope instability hazard factors are screened out from the intermediate high shear slope instability hazard factors in descending order of the contribution.

5. The high-cut slope safety level assessment method based on multi-source data according to claim 1 is characterized in that: The machine learning model adopts a decision tree model, and the model parameters of the decision tree model include a learning rate, the number of decision trees, the maximum depth of the decision tree, the depth required for internal node re-division, the minimum leaf node sample weight and the feature sampling ratio of each tree; Using the sample data set to train a machine learning model includes: Determine different parameter combinations consisting of the learning rate, the number of decision trees, the maximum depth of the decision tree, the depth required for the internal node re-division, the minimum leaf node sample weight and the feature sampling ratio of each tree; for each parameter combination, initialize the decision tree model using the parameter combination, train the initialized decision tree model using the sample data set, determine the performance index of the decision tree model after training based on the parameter combination, and determine the target parameter combination according to the performance index.

6. The high-cut slope safety level assessment method based on multi-source data according to claim 1 is characterized in that: The method further comprises: If it is monitored that the area to be monitored has new multi-source high-cut slope monitoring data, then the new target high-cut slope safety level score corresponding to the area to be monitored is evaluated based on the new multi-source high-cut slope monitoring data; The machine learning model is optimized based on the new safety grade score of the target high cutting slope and the instability disaster factor of the target high cutting slope.

7. The method for assessing the safety level of high slope cutting based on multi-source data according to claim 1 is characterized in that: The target high-cut slope instability disaster factors include land use type, slope aspect map, support structure / whether secondary reinforcement is required / engineering quality evaluation / slope geological condition data, longitude / latitude / county data, rainfall and surface waters; the method also includes: Obtaining the rainfall in the target time period of the area to be monitored, and using the rainfall in the target time period of the area to be monitored to replace the rainfall in the target high-cut slope instability disaster factor, so as to obtain a new combination of the target high-cut slope instability disaster factors in the area to be monitored in the target time period; and / or obtaining a new combination of the target high-cut slope instability disaster factors in the new area to be monitored in the target time period; The new combination of the target high-cut slope instability disaster factors is input into the trained machine learning model so that the trained machine learning model outputs a new safety level score of the target high-cut slope.

8. A high-cut slope safety level assessment device based on multi-source data, characterized in that: include: A grade evaluation module, used to evaluate the safety grade score of the target high-cut slope corresponding to the monitored area according to the multi-source high-cut slope monitoring data corresponding to the monitored area; A factor screening module is used to screen the correlation and contribution of multiple pre-configured candidate high-cut slope instability disaster factors to obtain multiple target high-cut slope instability disaster factors; A data set construction module, used to construct a sample data set based on the target high-cut slope safety grade score and the target high-cut slope instability disaster factor; A model training module is used to train the model parameters of the machine learning model using the sample data set. The trained machine learning model is used to evaluate the safety level score of the new target high-cut slope corresponding to the monitored area based on the new instability disaster factor of the target high-cut slope corresponding to the monitored area.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images