A high-cut slope stability assessment method based on multi-source data fusion

Abstract

The present invention discloses a high-cut slope stability assessment method based on multi-source data fusion, which specifically relates to the field of data analysis technology. By integrating terrain data, geological exploration data, remote sensing images, monitoring sensor data and geophysical detection data, an inversion algorithm is used to automatically extract abnormal structural features such as shear wave velocity mutations and resistivity troughs, and cross-validate with geological profiles to generate underground structure risk layers. The fusion model combines machine learning with mechanical algorithms to comprehensively output slope stability indicators, including safety factors and sliding probabilities, to achieve intelligent identification and classification of potential sliding hazard areas. Finally, the assessment results are visualized in a three-dimensional GIS platform in the form of layer overlays, which significantly improves the identification accuracy and scientific nature of the assessment of hidden risks in the slope, and provides a more reliable decision-making basis for the early warning and disposal of high-risk slopes.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a high road cutting slope stability assessment method based on multi-source data fusion. Background Art

[0002] High-cut slope stability assessment involves analyzing and evaluating the stability of slopes formed during highway excavation (i.e., cutting slopes). By comprehensively analyzing factors such as the slope's geological structure, soil and rock properties, slope gradient, rainfall, and earthquakes, the risk of geological hazards such as landslides and collapses, whether under natural conditions or external disturbances, is determined. This provides a scientific basis for slope design, construction, and subsequent maintenance, ensuring safe highway operation.

[0003] The existing technology has the following shortcomings:

[0004] During slope stability assessments based on multi-source data fusion, abnormal underground structures may not be identified in a timely manner, leading to severe distortion in the assessment results. Specifically, in some karst-developed areas, conventional remote sensing and surface monitoring methods have difficulty acquiring information on abnormal structures such as underground cavities, fissures, or weak interlayers. If the data fusion process does not incorporate geological profiles or geophysical exploration data of sufficient depth, these hidden structures will be misjudged as homogeneous bodies, causing the model to severely overestimate the overall stability of the slope. For example, during the construction of a certain highway, the model assessment results indicated that the slope was "stable." However, during actual construction, the slope became unstable due to the penetration of unidentified karst caves, resulting in a major landslide. Summary of the Invention

[0005] The purpose of the present invention is to provide a high road cutting slope stability assessment method based on multi-source data fusion to address the shortcomings of the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a high-cut slope stability assessment method based on multi-source data fusion, comprising:

[0007] Obtain topographic data, geological survey data, remote sensing image data, monitoring sensor data, and geophysical detection data of high road cutting slopes;

[0008] Based on geophysical exploration data, an inversion algorithm is used to automatically extract underground abnormal structural features, which are cross-validated with geological profile data to generate an underground structure risk map. The underground abnormal structural features include shear wave velocity mutation features and resistivity valley abnormal features.

[0009] Among them, after analyzing the resistivity valley anomaly characteristics, the resistivity valley anomaly value is generated. The acquisition method is: construct a feature set of resistivity profile data, with N sampling points or pixel points, each point contains a resistivity value , local mean , local standard deviation and adjacent gradients ; The eigenvector is expressed as: ; Select the number of clusters K and use the K-means algorithm to classify the feature vectors of N points Clustering: The initial centroid is randomly selected or initialized using K-means optimization until the centroid converges or the maximum number of rounds is reached: From the final K clusters, the cluster with the lowest mean value is selected. The cluster with the lowest resistivity average value is regarded as the resistivity valley anomaly class and marked as , for each sampling point i, calculate the resistivity valley anomaly value: ; Where VC is the resistivity valley abnormal value, Indicates that at all sampling points , calculate its relationship with the low resistivity cluster center The maximum Euclidean distance between Represents the feature vector of the jth sample point;

[0010] Standardized topographic, geological, monitoring, and underground structural anomaly risk maps are fed into a fusion model built by coupling machine learning algorithms with mechanical models to output slope stability indicators, including safety factors and slip probabilities.

[0011] Based on the output results of the fusion model, the slope stability is graded and evaluated. Combined with the location of abnormal underground structures, potential sliding risk areas are identified, and the identification results are visualized in a 3D GIS platform in the form of layer overlays.

[0012] Preferably, an inversion algorithm is used to automatically extract underground abnormal structural features, specifically: multiple seismic receivers are deployed on the target slope, and active sources are used to excite seismic waves to collect surface wave signals; the collected original seismic signals are denoised, band filtered, gain adjusted and normalized, and a frequency-wavenumber analysis technique or a phase velocity-frequency spectrum analysis technique is used to extract multiple surface wave dispersion curves; the extracted dispersion curves are input into the MASW inversion algorithm to reconstruct the shear wave velocity distribution of each depth point in the longitudinal profile of the slope to form a two-dimensional or three-dimensional shear wave velocity model; the velocity difference between adjacent points in the shear wave velocity model is calculated to obtain a velocity gradient field: a shear wave velocity mutation threshold is set, and the area that meets the conditions is marked as a shear wave velocity mutation zone, that is, a candidate area for a potential weak structural surface, and the coordinates and depth of the mutation zone are projected into a vector or raster layer as a shear wave velocity mutation structure layer.

[0013] Preferably, an electrode array is arranged in the target area, and a multi-pole configuration method is used to collect apparent resistivity data. The apparent resistivity data is inverted and calculated using a finite element inversion algorithm to generate a two-dimensional or three-dimensional underground resistivity distribution model. The resistivity model is standardized, the resistivity value is scaled to the [0,1] interval, and low-resistance blocks are identified through image segmentation. The identified low-resistance anomaly is projected into a spatial layer, and its geometric boundary, depth range and center point coordinates are extracted to generate a resistivity valley anomaly layer.

[0014] Preferably, the machine learning model module is used to convert the shear wave velocity mutation characteristics and resistivity valley anomaly characteristics into a comprehensive feature vector, and use the comprehensive feature vector as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the sliding probability label of each spatial unit as the prediction target, and uses minimizing the sum of the prediction errors of the sliding probability labels of all spatial units as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The sliding probability of each spatial unit is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0015] Preferably, the shear wave velocity mutation value is generated after analyzing the shear wave velocity mutation characteristics, and the acquisition method is:

[0016] Obtain the two-dimensional or three-dimensional shear wave velocity data volume Vs(x,z) of the slope, divide the model into several spatial units, and calculate the shear wave velocity difference in each vertical direction section : ; Where: i represents the horizontal grid index; j represents the depth index; represents the shear wave velocity value of the i,jth unit; Indicates the vertical shear wave velocity mutation value at the depth position; calculate the shear wave velocity gradient on each horizontal section : ; Define the shear wave velocity mutation value as the weighted superposition of the mutation values ​​in two directions.

[0017] Preferably, the obtained sliding probability of each spatial unit and the mechanical safety factor of each unit cell are de-dimensionalized and normalized, and then a weighted average summation calculation is performed to obtain a comprehensive instability score for each unit.

[0018] Preferably, the obtained comprehensive instability score of each unit is compared with a gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and the comprehensive instability score of each unit is compared with the first standard threshold and the second standard threshold respectively;

[0019] If the comprehensive instability score of each unit is greater than the second standard threshold, the corresponding slope will be marked as a high-risk area, and a first-level warning signal will be generated;

[0020] If the comprehensive instability score of each unit is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding slope will be marked as a medium risk area, and a second-level warning signal will be generated;

[0021] If the comprehensive instability score of each unit is less than the first standard threshold, the corresponding slope will be marked as a low-risk area, and a level 3 warning signal will be generated.

[0022] Preferably, the identified high-risk and medium-risk areas are spatially overlaid with the underground abnormal structure layer to identify the units within the spatial overlap or adjacent range and mark them as potential sliding hazard areas to provide decision-making reference for managers.

[0023] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0024] 1. This invention overcomes the difficulty in identifying abnormal underground structures in traditional slope stability assessments by integrating geophysical exploration data and an automated inversion algorithm. It accurately extracts shear wave velocity mutations and resistivity dips, and, combined with cross-validation using geological profiles, generates a highly reliable underground structure risk map. Furthermore, this invention standardizes topographic, geological, monitoring, and underground structure data before inputting them into a fusion model. This model, combined with machine learning algorithms and mechanical analysis methods, achieves high-precision predictions of slope stability indicators and outputs quantitative metrics, including slip probability and safety factor, enhancing the scientific nature and robustness of the assessment.

[0025] 2. This invention uses a graded threshold judgment mechanism to classify slopes into high, medium, and low risk categories. It then identifies potential sliding risk areas based on the spatial location of abnormal underground structures, triggering a graded early warning mechanism. Identification results are visualized as layers overlaid on a 3D GIS platform, making risk information more intuitive and providing project managers with an efficient and actionable decision-making tool. This significantly improves the proactiveness and reliability of high-cut slope hazard warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0027] Figure 1 This is a schematic diagram of the present invention.

[0028] Figure 2 This is a diagram of the machine learning model principle. DETAILED DESCRIPTION

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0030] For examples, see Figure 1 As shown, the high cutting slope stability assessment method based on multi-source data fusion described in this embodiment includes:

[0031] Obtain topographic data, geological survey data, remote sensing image data, monitoring sensor data, and geophysical detection data of high road cutting slopes;

[0032] Based on geophysical exploration data, an inversion algorithm is used to automatically extract underground abnormal structural features and cross-validate with geological profile data to generate an underground structure risk layer;

[0033] Standardized topographic, geological, monitoring, and underground structural anomaly risk maps are fed into a fusion model built by coupling machine learning algorithms with mechanical models to output slope stability indicators, including safety factors and slip probabilities.

[0034] Based on the output results of the fusion model, the slope stability is graded and evaluated. Combined with the location of abnormal underground structures, potential sliding risk areas are identified, and the identification results are visualized in a 3D GIS platform in the form of layer overlays.

[0035] Topographic data is mainly used to describe the elevation, slope, aspect, and spatial morphology of slopes, providing basic geometric structure information for stability analysis. Common ways to obtain topographic data include:

[0036] Unmanned aerial vehicle (UAV) surveying: Equipped with high-precision laser radar (LiDAR) or multispectral cameras, it can quickly obtain three-dimensional point clouds and orthophotos of slopes and generate high-resolution digital elevation models (DEMs) or digital surface models (DSMs).

[0037] GNSS measurement and total station measurement: Set ground measurement points at key control points to obtain high-precision terrain coordinate information for correcting remote sensing data.

[0038] Topographic map vectorization: Combined with 1:1000 or 1:2000 scale topographic maps, digital extraction is performed through the GIS system.

[0039] Geological survey data is used to reveal the rock and soil structure, physical and mechanical properties, and hydrogeological conditions of the slope. The acquisition methods include:

[0040] Drilling sampling: Drill holes are laid out to obtain rock and soil samples at different depths, and laboratory tests (such as triaxial shear tests and direct shear tests) are conducted to analyze soil shear strength, porosity, moisture content and other parameters.

[0041] In-situ testing: such as standard penetration test (SPT), cone penetration test (CPT), and cross-plate shear test (VST), used to obtain stratum bearing capacity and shear strength indicators.

[0042] Geological cataloging: Describe the drill core, identify structural surface features such as faults, cracks, and joints, and draw cross-sections based on regional geological maps.

[0043] Remote sensing image data is used to analyze macro-slope deformation, vegetation cover changes, drainage paths, and other conditions. Common acquisition methods include:

[0044] Satellite remote sensing images: Multispectral or high-resolution satellites (such as Sentinel-2, WorldView, and Gaofen series) can be used for periodic monitoring.

[0045] InSAR technology (Synthetic Aperture Radar Interferometry): uses radar image comparison to generate deformation interference maps to monitor small surface displacement changes, which is suitable for early identification of landslides.

[0046] Infrared remote sensing and thermal imaging: Identify aquifer changes or potential slip surfaces in areas with large surface temperature differences.

[0047] Monitoring sensor data can achieve real-time perception and dynamic feedback of slope stability. Common sensor types include:

[0048] Inclinometer: Installed on the slope surface or in a borehole to monitor changes in sliding direction and angle.

[0049] Displacement meters and inclinometers: used to measure the displacement trend of the sliding zone layer and support shallow and deep sliding monitoring.

[0050] Piezometers and groundwater level gauges: monitor groundwater pressure and water level changes, reflecting changes in potential sliding forces.

[0051] Rain gauges and weather stations: collect meteorological elements such as rainfall intensity, rainfall duration, temperature and humidity, and evaluate inducing factors.

[0052] Wireless Internet of Things acquisition system: Sensor data is transmitted to the slope safety management platform via wireless networks, supporting remote access and automatic early warning.

[0053] Geophysical exploration data is used to identify hidden abnormal underground structures (such as caves, fault zones, and weak interlayers) and is a key data source for improving the accuracy of depth assessment. It includes:

[0054] Seismic wave method (reflection wave / refracted wave): By artificially exciting the seismic source, the propagation behavior of the wave in the underground medium is recorded, and the underground layer interface and abnormal body are inverted.

[0055] Geoelectrical method (resistivity imaging): Using electrodes to measure resistivity variations in different areas of the subsurface, weak structures such as aquifers, cavities, and faults can be identified.

[0056] Gravity and magnetic methods: used to identify areas with significant lithologic differences and assist in determining underground dissolution or fault characteristics.

[0057] Geological Radar (GPR): Suitable for scanning shallow structures and obtaining high-resolution images for detailed supplementation.

[0058] Based on geophysical exploration data, an inversion algorithm is used to automatically extract underground abnormal structural features and cross-validate them with geological profile data to generate underground structural risk layers, including:

[0059] Acquire raw seismic wave, georesistivity, or radar data from the slope area. Perform preprocessing operations such as filtering, denoising, and normalization on the data. Calibrate data acquisition points and spatial coordinates to facilitate spatial registration with geological profile data.

[0060] The steps for identifying the shear wave velocity mutation region include:

[0061] Multi-channel seismic receivers are deployed on the target slope, and active sources are used to excite seismic waves and collect surface wave signals.

[0062] The collected original seismic signals are subjected to denoising, frequency band filtering, gain adjustment and normalization to remove background noise and enhance signal quality.

[0063] Multiple surface wave dispersion curves are extracted using frequency-wavenumber (fk) analysis or phase velocity-frequency spectrum analysis techniques.

[0064] The extracted dispersion curves are input into the MASW inversion algorithm to reconstruct the shear wave velocity (Vs) distribution at each depth point in the slope longitudinal profile and form a two-dimensional or three-dimensional Vs model.

[0065] The velocity gradient field is obtained by calculating the velocity difference between adjacent points (in the vertical or horizontal direction) in the Vs model:

[0066] Set a shear wave velocity mutation threshold (for example, ΔVs > 200 m / s) and mark the area that meets this condition as a shear wave velocity mutation zone, that is, a candidate area for potential weak structural surfaces.

[0067] The coordinates and depth of the mutation area are projected into a vector or raster layer as a shear wave velocity mutation structure layer for subsequent fusion modeling.

[0068] The steps for extracting resistivity valley anomalies include:

[0069] An electrode array is placed in the target area, using a multi-pole configuration (such as Wenner or Schlumberger) to collect apparent resistivity data.

[0070] The apparent resistivity data is inverted and calculated using the finite element inversion algorithm to generate a two-dimensional or three-dimensional underground resistivity distribution model. The model grid accuracy can be set to 1~2 meters.

[0071] The resistivity model is normalized and the resistivity value is scaled to the interval [0,1] to enhance the image segmentation effect.

[0072] Identify low-resistance areas through image segmentation: Option 1: Otsu algorithm (automatic threshold method) converts the resistivity model into a grayscale image; uses the Otsu method to calculate the optimal threshold T and segment the image into low-resistance and high-resistance areas; extracts areas with resistivity values ​​less than T as low-resistance anomaly areas.

[0073] Option 2: K-means clustering: Take the resistivity value of each grid point and its neighborhood statistical characteristics as input; set the number of clusters K to 3-5 to cluster the resistivity values; select the cluster category with the lowest resistivity mean as the low-resistance anomaly area.

[0074] The identified low-resistance anomaly is projected into a spatial layer, and its geometric boundary, depth range and center point coordinates are extracted to generate a resistivity valley anomaly layer.

[0075] Project the inverted underground structural anomalies, including shear wave velocity mutations and resistivity dips, into a unified spatial coordinate system (e.g., the UTM coordinate system). Construct a 3D point cloud or raster model and spatially annotate the anomalies.

[0076] Key landmarks such as structural surfaces, weak zones, and aquifers are extracted from the borehole log. Section intersection points corresponding to the inversion feature points are established, and the depth, thickness, and properties of the two are compared for consistency. If the positional deviation is less than 2 meters, the structure is marked as abnormal; if the difference is significant, a low confidence rating is assigned, and the subsequent correction process is initiated.

[0077] The confirmed anomalous structural features and their spatial distribution are exported as a raster layer (GeoTIFF or NetCDF). The layer attributes include: anomaly type (shear wave velocity mutation / low-resistance anomaly); depth range (e.g., 8–15 m); risk level (high / medium / low, based on the structure size and impact on stability); and confidence score (based on cross-validation consistency). The layer can be overlaid on a GIS platform or 3D slope model for subsequent assessment.

[0078] When new geophysical data or drilling data is input, the automatic comparison mechanism is triggered; the spatial boundaries and risk levels of abnormal structures in the layer are updated to achieve data-driven dynamic correction.

[0079] Standardized topographic, geological, monitoring, and underground structural anomaly risk maps are fed into a fusion model, which is built by coupling machine learning algorithms with mechanical models to output slope stability indicators, including safety factors and slip probabilities. Specifically, these include:

[0080] To ensure the comparability and integration of multi-source heterogeneous data, the following data are first standardized and preprocessed to unify the spatial resolution, coordinate system, and data format:

[0081] Input data categories include: Topographic data (DEM / DSM): slope, aspect, curvature, height difference, etc. Geological data: lithology code, shear strength parameters (c, φ), porosity, interlaminar shear strength difference. Monitoring data: displacement rate, seepage pressure change, groundwater level change, and rainfall intensity time series for each monitoring point.

[0082] Data normalization methods include: Numerical data: Z-score normalization or Min-Max normalization. Categorical data (such as lithology): One-Hot Encoding. Time series data (such as displacement or rainfall): Sliding window statistical feature extraction (such as mean, volatility, slope).

[0083] The traditional safety factor (FS) is calculated based on input geotechnical parameters, slope angle, groundwater level, and other information, using methods such as the simplified Bishop method and the Janbu method. The model incorporates weak structural areas from the underground structure risk layer as sliding surface control factors and assigns reduced shear strength parameters to these areas. The mechanical FS value for each cell is output.

[0084] The machine learning model module is used to convert the shear wave velocity mutation characteristics and resistivity valley anomaly characteristics into comprehensive feature vectors, and use the comprehensive feature vectors as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the sliding probability label of each spatial unit as the prediction target, and minimizes the sum of the prediction errors of the sliding probability labels of all spatial units as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The sliding probability of each spatial unit is determined according to the model output results. Among them, the machine learning model is a polynomial regression model.

[0085] Among them, the shear wave velocity mutation value is generated after analyzing the shear wave velocity mutation characteristics, and the acquisition method is:

[0086] Obtain the 2D or 3D shear wave velocity data volume Vs(x,z) of the slope from seismic surface wave inversion (such as MASW). Divide the model into several spatial units (such as 1m×1m grid).

[0087] Calculate the shear wave velocity difference at each vertical section (along the depth direction) : ; Where: i represents the horizontal grid index; j represents the depth index; represents the shear wave velocity value of the i,jth unit (unit: m / s); Indicates the sudden change in vertical shear wave velocity at this depth.

[0088] Calculate the shear wave velocity gradient at each horizontal section (along the slope longitudinal direction) : ; Indicates the horizontal variation of wave velocity at the same depth.

[0089] The shear wave velocity mutation value is defined as the weighted superposition of mutation values ​​in two directions. A velocity mutation recognition threshold is set, and the area that meets the conditions is extracted as a mutation zone: if the shear wave velocity mutation value is greater than the velocity mutation recognition threshold, it is marked as a shear wave velocity mutation area. The velocity mutation recognition threshold can be set based on regional experience, such as 200–300 m / s. Grids that continuously meet the conditions can be aggregated into a potential structural surface zone to generate a spatial vector layer.

[0090] After analyzing the resistivity valley anomaly characteristics, the resistivity valley anomaly value is generated. The acquisition method is as follows: construct a feature set of resistivity profile data, with N sampling points or pixel points, each point contains: resistivity value , local mean , local standard deviation and adjacent gradients ; The eigenvector is expressed as: .

[0091] The number of clusters K is usually determined by one of the following methods:

[0092] Empirical method: Set K = 3~5, corresponding to low, medium and high resistivity regions respectively;

[0093] Elbow Method: Calculate the total squared error (SSE) for different K values ​​and select the point where the SSE does not decrease significantly.

[0094] Silhouette Coefficient: Measure the separation and closeness of clusters and select K corresponding to the maximum silhouette value.

[0095] Use K-means algorithm to calculate the feature vectors of N points Perform clustering:

[0096] The initial centroid is randomly selected or initialized using K-means and optimization;

[0097] Iterate the following two steps until the centroid converges or the maximum number of rounds is reached:

[0098] ; ;in: is the cluster category to which the k-th point belongs, is the centroid vector of the kth class;

[0099] From the final K clusters, the cluster with the lowest mean value is selected. The cluster with the lowest resistivity average value is regarded as the resistivity valley anomaly class and marked as , for each sampling point i, calculate the resistivity valley anomaly value: ; Where VC is the resistivity valley abnormal value, Indicates that at all sampling points , calculate its relationship with the low resistivity cluster center The maximum Euclidean distance between Represents the eigenvector of the jth sample point; the higher the resistivity valley anomaly value, the closer it is to the anomaly center.

[0100] The obtained sliding probability of each spatial unit and the mechanical safety factor of each unit are de-dimensionalized and normalized, and then the weighted average summation is calculated to obtain the comprehensive instability score of each unit.

[0101] Comparing the obtained comprehensive instability score of each unit with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and comparing the comprehensive instability score of each unit with the first standard threshold and the second standard threshold respectively;

[0102] If the comprehensive instability score of each unit is greater than the second standard threshold, the corresponding slope will be marked as a high-risk area, and a first-level warning signal will be generated;

[0103] If the comprehensive instability score of each unit is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding slope will be marked as a medium risk area, and a second-level warning signal will be generated;

[0104] If the comprehensive instability score of each unit is less than the first standard threshold, the corresponding slope will be marked as a low-risk area, and a level 3 warning signal will be generated.

[0105] It should be noted here that the risk of a level one warning signal is higher than that of a level two warning signal, and the risk of a level two warning signal is higher than that of a level three warning signal. Relevant personnel can take corresponding measures based on the different levels of warning signals.

[0106] The identified high- and medium-risk areas are spatially overlaid with the underground structural anomaly layer to identify overlapping or adjacent units and mark them as potential slip zones. This marking is based not only on the scoring results but also on the actual geological evidence of structural weakness, enhancing the reliability of the identification.

[0107] Finally, the DEM, the abnormal structure layer, the comprehensive instability score layer, and the marked potential sliding areas were imported into a 3D GIS platform for visualization. The platform enables 3D rotation, cross-section browsing, click-to-search, and hierarchical control, visually displaying slope morphology, the potential for sliding paths, the location of abnormal structures, and their spatial relationship to risk areas, providing decision-making support for managers.

[0108] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0109] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0110] It should be understood that the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone, where A and B may be singular or plural. In addition, the character " / " herein generally indicates that the objects associated with each other are in an "or" relationship, but it may also indicate an "and / or" relationship, which can be understood by referring to the context. A person of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0111] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A high-cut slope stability assessment method based on multi-source data fusion, characterized by: include: Obtain topographic data, geological survey data, remote sensing image data, monitoring sensor data, and geophysical detection data of high road cutting slopes; Based on geophysical exploration data, an inversion algorithm is used to automatically extract underground abnormal structural features, which are cross-validated with geological profile data to generate an underground structure risk map. The underground abnormal structural features include shear wave velocity mutation features and resistivity valley abnormal features. Among them, after analyzing the resistivity valley anomaly characteristics, the resistivity valley anomaly value is generated. The acquisition method is: construct a feature set of resistivity profile data, with N sampling points or pixel points, each point contains a resistivity value , local mean , local standard deviation and adjacent gradients ; The eigenvector is expressed as: ; Select the number of clusters K and use the K-means algorithm to classify the feature vectors of N points Clustering: The initial centroid is randomly selected or initialized using K-means optimization until the centroid converges or the maximum number of rounds is reached: From the final K clusters, the cluster with the lowest mean value is selected. The cluster with the lowest resistivity average value is regarded as the resistivity valley anomaly class and marked as , for each sampling point i, calculate the resistivity valley anomaly value: ; Where VC is the resistivity valley abnormal value, Indicates that at all sampling points , calculate its relationship with the low resistivity cluster center The maximum Euclidean distance between Represents the feature vector of the jth sample point; Standardized topographic, geological, monitoring, and underground structural anomaly risk maps are fed into a fusion model built by coupling machine learning algorithms with mechanical models to output slope stability indicators, including safety factors and slip probabilities. Based on the output results of the fusion model, the slope stability is graded and evaluated. Combined with the location of abnormal underground structures, potential sliding risk areas are identified, and the identification results are visualized in a 3D GIS platform in the form of layer overlays.

2. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 1, characterized in that: An inversion algorithm is used to automatically extract underground abnormal structural features, specifically: multi-channel seismic receivers are deployed on the target slope, and active sources are used to excite seismic waves to collect surface wave signals; the collected original seismic signals are denoised, band filtered, gain adjusted and normalized, and frequency-wavenumber analysis technology or phase velocity-frequency spectrum analysis technology is used to extract multiple surface wave dispersion curves; the extracted dispersion curves are input into the MASW inversion algorithm to reconstruct the shear wave velocity distribution of each depth point in the slope longitudinal profile to form a two-dimensional or three-dimensional shear wave velocity model; the velocity difference between adjacent points in the shear wave velocity model is calculated to obtain the velocity gradient field: a shear wave velocity mutation threshold is set, and the area that meets the conditions is marked as a shear wave velocity mutation area, that is, a candidate area for a potential weak structural surface; the coordinates and depth of the mutation area are projected into a vector or raster layer as a shear wave velocity mutation structure layer.

3. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 2 is characterized by: An electrode array is deployed in the target area, and a multi-pole configuration method is used to collect apparent resistivity data. The apparent resistivity data is then inverted and calculated using a finite element inversion algorithm to generate a two-dimensional or three-dimensional underground resistivity distribution model. The resistivity model is standardized, and the resistivity value is scaled to the [0,1] interval. Low-resistance blocks are identified through image segmentation, and the identified low-resistance anomalies are projected into a spatial layer. The geometric boundaries, depth range, and center point coordinates of the anomalies are extracted to generate a resistivity valley anomaly layer.

4. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 1, wherein: The shear wave velocity mutation characteristics and resistivity valley anomaly characteristics are converted into comprehensive feature vectors, which are used as inputs of the machine learning model. The machine learning model predicts the slip probability label of each spatial unit with each set of comprehensive feature vectors as the prediction target, and minimizes the sum of the prediction errors of the slip probability labels of all spatial units as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The slip probability of each spatial unit is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

5. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 4 is characterized in that: After analyzing the shear wave velocity mutation characteristics, the shear wave velocity mutation value is generated. The acquisition method is: Obtain the two-dimensional or three-dimensional shear wave velocity data volume Vs(x,z) of the slope, divide the shear wave velocity model into several spatial units with a grid size of 1m×1m, and calculate the shear wave velocity difference in each vertical direction section : ; Where: i represents the horizontal grid index; j represents the depth index; represents the shear wave velocity value of the i,jth unit; Indicates the vertical shear wave velocity mutation value at the depth position; calculate the shear wave velocity gradient on each horizontal section : ; Define the shear wave velocity mutation value as the weighted superposition of the mutation values ​​in two directions.

6. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 1, characterized in that: The obtained sliding probability of each spatial unit and the mechanical safety factor of each unit are de-dimensionalized and normalized, and then the weighted average summation is calculated to obtain the comprehensive instability score of each unit.

7. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 6 is characterized by: Comparing the obtained comprehensive instability score of each unit with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and comparing the comprehensive instability score of each unit with the first standard threshold and the second standard threshold respectively; If the comprehensive instability score of each unit is greater than the second standard threshold, the corresponding slope will be marked as a high-risk area, and a first-level warning signal will be generated; If the comprehensive instability score of each unit is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding slope will be marked as a medium risk area, and a second-level warning signal will be generated; If the comprehensive instability score of each unit is less than the first standard threshold, the corresponding slope will be marked as a low-risk area, and a level 3 warning signal will be generated.

8. The method for evaluating the stability of high cutting slopes based on multi-source data fusion according to claim 7 is characterized by: The identified high-risk and medium-risk areas are spatially overlaid with the underground abnormal structure layer to identify units within the spatially overlapping or adjacent ranges and mark them as potential sliding hazard areas, providing decision-making reference for managers.

Images