Mountainous high-speed road cutting high slope stability evaluation method based on multi-source data fusion

Abstract

The present application provides a mountainous high-speed highway cutting high slope stability evaluation method based on multi-source data fusion, relates to the technical field of slope stability evaluation, and determines an external disturbance class feature vector according to a vehicle-by-vehicle passing record, calculates a characteristic value, obtains a macro evolution class feature vector, forms an abnormal structure space matrix according to spatial distribution data of apparent resistivity and spatial distribution data of shear wave velocity, calculates a disturbance sensitivity index, an evolution offset index and a structure damage index of the mountainous high-speed highway cutting high slope, forms a fusion feature matrix of each grid, inputs the fusion feature matrix of each grid into a trained slope stability evaluation model, obtains a stability score spatial distribution map, screens low stability grids from the stability score spatial distribution map and obtains a potential sliding risk grid set, determines the instability leading factor of a contiguous risk block, and improves the evaluation accuracy and instability leading factor discrimination ability.

Description

Technical Field

[0001] This invention relates to the field of slope stability assessment technology, specifically a method for assessing the stability of high slopes in mountainous highway cuttings based on multi-source data fusion. Background Technology

[0002] As the construction of mountain expressways extends to areas with complex terrain, the stability assessment of high slopes in road cuts has become a core technical aspect of ensuring the safe operation of vital transportation routes. Traditional slope monitoring methods have limitations such as limited data dimensions and delayed early warnings. In contrast, fusion assessment methods that integrate multi-source heterogeneous data, including surface deformation, deep displacement, geophysical exploration, and meteorological observations, can characterize the evolution of slopes from different physical dimensions. This has become an important development trend in the field and has demonstrated significant engineering application value in the dynamic risk prevention and control of mountain expressways.

[0003] In the prior art, patent application CN120196909A discloses a method for assessing the stability of high road cut slopes based on multi-source data fusion. This method acquires five types of data: topographic, geological, remote sensing, sensor data, and geophysical data. Based on geophysical data inversion, it extracts abrupt changes in shear wave velocity and resistivity trough anomalies, and combines these with geological profiles to generate a subsurface structure risk layer. Standardized data is input into a fusion model constructed by coupling a multinomial regression model and a mechanical model, outputting a safety factor and sliding probability, which are then weighted and averaged to obtain a comprehensive instability score. Finally, potential sliding hazard areas are identified through dual-threshold grading and spatial overlay, and visualized on a 3D GIS platform. The basic principle of this scheme is to enhance the identification capability of hidden structures through geophysical inversion and to achieve quantitative stability assessment through the coupled output of mechanics and machine learning.

[0004] While the aforementioned scheme improves the ability to identify hidden structures to some extent, it still has limitations in adaptability to mountainous highway cutting scenarios. The scheme fails to incorporate the unique traffic load of mountainous highways as an independent first-order data mode into the fusion framework, making it impossible for the model to characterize the coupled physical mechanism of "external traffic dynamic load—macroscopic deformation evolution—internal damage accumulation." This can easily lead to systematic evaluation biases in heavily loaded, densely trafficked road sections. Furthermore, the scheme uses a weighted average of the mechanical safety factor and machine learning sliding probability as the comprehensive score output, which is a shallow data fusion architecture. It does not construct an indexed expression system for the three physical dimensions of external disturbance, macroscopic evolution, and internal damage. Consequently, the evaluation results can only provide a stability level, but cannot identify the dominant factors leading to instability while outputting conclusions, making it difficult to provide a basis for subsequent differentiated treatment decisions.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for evaluating the stability of high slopes in mountainous highway cuttings based on multi-source data fusion, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion includes the following steps: S1. Collect slope attribute data, vehicle-by-vehicle records, three-dimensional surface displacement data, deep dip angle data, groundwater level data, rainfall data, apparent resistivity spatial distribution data, and shear wave velocity spatial distribution data within the high slope area of ​​the highway cutting in the target mountainous area. S2. Determine the feature vector of external disturbance based on the vehicle passing record, calculate the feature value of the three-dimensional displacement data of the ground surface, the deep dip angle data, the groundwater level data and the rainfall data to obtain the feature vector of macro-evolution, and calculate the deviation value of each sampling point relative to the local neighborhood based on the spatial distribution data of apparent resistivity and the spatial distribution data of shear wave velocity and interpolate it to a unified spatial grid to form an abnormal structure spatial matrix. S3. Based on the external disturbance type feature vector, macro-evolution type feature vector and abnormal structure space matrix, calculate the disturbance sensitivity index, evolutionary offset index and structural damage index of the high slope of the highway cutting in the mountainous area, and align and stitch the three indices with the corresponding slope attribute data according to a unified spatial grid to form the fusion feature matrix of each grid. S4. Input the fusion feature matrix of each grid into the trained slope stability assessment model, output the stability score, and arrange each stability score in space according to the spatial location of the high slope of the mountain highway cutting to obtain the spatial distribution map of the stability score. S5. Screen low-stability grids from the stability score spatial distribution map and obtain a set of grids with potential sliding hazards. Cluster them into contiguous hazard blocks, determine the dominant instability factors of each grid, and count the proportion of grids with various dominant instability factors in the contiguous hazard blocks to determine the dominant instability factors of the contiguous hazard blocks.

[0008] Furthermore, based on the vehicle-by-vehicle passing records, the feature vectors of external disturbance types are determined. The specific steps are as follows: Extract the axle load and speed of each vehicle from the vehicle-by-vehicle records according to a unified assessment time window, filter out heavy-load vehicles whose axle load exceeds the preset load threshold, count the number of filtered vehicles, and obtain the number of times heavy-load vehicles pass. A unified assessment time window refers to a standard time interval that divides a continuous monitoring period into fixed durations. The arithmetic mean of the axle loads of heavy-duty vehicles within a unified assessment time window is calculated to obtain the average axle load. At the same time, the arithmetic mean of the speeds of heavy-duty vehicles is calculated to obtain the average speed. The number of times heavy-load vehicles pass through, the average axle load, and the average vehicle speed are combined into a one-dimensional vector in a fixed order to obtain the feature vector of external disturbance.

[0009] Furthermore, feature values ​​are calculated from the three-dimensional surface displacement data, deep dip angle data, groundwater level data, and rainfall data to obtain macro-evolutionary feature vectors. The specific steps are as follows: The corresponding time series information was extracted from the three-dimensional surface displacement data, deep dip angle data, groundwater level data and rainfall data according to a unified evaluation time window, and the time series of three-dimensional surface displacement, deep dip angle, groundwater level and rainfall were obtained. Calculate the characteristic values, including mean, maximum and standard deviation, in the three-dimensional displacement time series of the Earth's surface, the deep dip angle time series, the groundwater level time series and the rainfall time series respectively; The mean, maximum value, and standard deviation of the four types of data are concatenated and spliced ​​into a one-dimensional array in the order of three-dimensional surface displacement, deep dip angle, groundwater level, and rainfall. This one-dimensional array is the macro-evolutionary feature vector.

[0010] Furthermore, based on the spatial distribution data of apparent resistivity and shear wave velocity, the deviation values ​​of each sampling point relative to its local neighborhood are calculated and interpolated to a unified spatial grid to form an anomaly structure spatial matrix. The specific steps are as follows: Based on the spatial distribution data of apparent resistivity, the arithmetic mean of the resistivity values ​​of all sampling points within the neighborhood set defined by a preset radius with each sampling point as the center is calculated. The resistivity value of each sampling point is subtracted from the arithmetic mean to obtain the resistivity deviation value of each sampling point. Based on the spatial distribution data of shear wave velocity, the arithmetic mean of the shear wave velocity values ​​of all sampling points within the neighborhood set defined by a preset radius with each sampling point as the center is calculated. The shear wave velocity value of each sampling point is subtracted from the arithmetic mean to obtain the shear wave velocity deviation value of each sampling point. A unified spatial grid covering the high slopes of highway cuts in the target mountainous area was established. The resistivity deviation and shear wave velocity deviation of each sampling point were interpolated to each grid of the unified spatial grid using the inverse distance weighted interpolation method to form an anomaly structure spatial matrix.

[0011] Furthermore, based on the external disturbance feature vectors, macro-evolution feature vectors, and the anomalous structure space matrix, the disturbance sensitivity index, evolutionary offset index, and structural damage index of the high slope of the highway cutting in mountainous areas are calculated respectively. The specific steps are as follows: Obtain the historical maximum values ​​of the feature vectors of external disturbances, macro-evolution, and the spatial matrix of abnormal structures in the high slope of the highway cutting in the target mountainous area; Based on the maximum value of historical data, the feature vectors of external disturbance, macro-evolution, and abnormal structure space matrix are normalized to obtain the external disturbance component group, macro-evolution component group, and internal damage component group. The characteristic data within the external disturbance component group, the macro-evolution component group, and the internal damage component group are weighted and summed to obtain the disturbance sensitivity index, the evolutionary deviation index, and the structural damage index.

[0012] Furthermore, the disturbance sensitivity index, evolutionary offset index, and structural damage index are aligned and stitched together with the corresponding slope attribute data according to a unified spatial grid to form a fused feature matrix for each grid. The specific steps are as follows: Extract all grid positions of the unified spatial grid from the abnormal structure spatial matrix to obtain the grid set; The disturbance sensitivity index, evolution offset index, and structural damage index are respectively assigned to each grid in the grid set; For each grid, extract the slope, slope height, and uniquely thermally encoded soil and rock type from the slope attribute data corresponding to that grid; The disturbance sensitivity index, evolution offset index, structural damage index, slope, slope height, and uniquely thermally encoded soil and rock type of each grid are sequentially spliced ​​together to form a fused feature matrix.

[0013] Furthermore, the dominant factors for instability in contiguous hazardous areas are identified through the following steps: Extract grids with scores higher than a preset threshold from the stability score spatial distribution map and form a set of low-stability grids. By performing spatial intersection analysis between the low-stability mesh set and the resistivity and shear wave velocity anomalies within the spatial matrix of the anomalous structure that meet the preset deviation threshold, a potential sliding hazard mesh set is obtained. Clustering spatially adjacent potential sliding hazard grids in the potential sliding hazard grid set to form contiguous hazard blocks; For a set of grids with potential sliding hazards, the normalized values ​​of the external disturbance component group, macro-evolution component group, and internal damage component group of each grid in the set are compared, and the component dimension with the largest value is taken as the dominant factor of instability corresponding to the grid. The grid proportions of various dominant instability factors within each contiguous danger zone are statistically analyzed, and the dimension with the highest proportion is selected as the dominant instability factor for the contiguous danger zone.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention collects vehicle-by-vehicle records within the high slope area of ​​a mountainous highway cutting and combines them to obtain external disturbance feature vectors. This enables the inclusion of the unique traffic load of mountainous highways as an independent external disturbance feature into a multi-source data fusion framework. This allows the impact of traffic dynamic load on slope stability to be quantitatively characterized and participate in the overall evaluation process. This overcomes the fundamental deficiency of existing evaluation methods that do not consider traffic load, resulting in insufficient adaptability to mountainous highway cutting scenarios, and significantly improves the consistency between the evaluation results and real working conditions.

[0015] Meanwhile, this invention calculates the disturbance sensitivity index, evolutionary offset index, and structural damage index based on external disturbance feature vectors, macro-evolution feature vectors, and anomalous structure spatial matrices, respectively. These three indices are then aligned and stitched with slope attribute data according to a unified spatial grid to form a fused feature matrix. This constructs an index system covering three physical dimensions: external disturbance, macro-evolution, and internal damage. This enables unified alignment and structured integration of multi-source heterogeneous data in space, providing input features with clear physical meaning and dimensional boundaries for subsequent slope stability assessment models. This effectively enhances the model's ability to analyze complex coupling relationships and further improves the accuracy and interpretability of stability scoring.

[0016] Furthermore, this invention integrates the feature matrix input trained into the slope stability assessment model, outputs stability scores, and spatially arranges them to obtain a spatial distribution map of the stability scores. By filtering low-stability grids from the spatial distribution map of stability scores, a set of grids with potential sliding hazards is obtained, and clustering is used to obtain contiguous hazardous blocks. Then, the dominant instability factors of each grid are determined, and the dominant instability factors at the block level are determined based on the proportion of various dominant factor grids within the contiguous hazardous blocks. This not only achieves accurate spatial positioning of potential sliding hazard areas, but also distinguishes the dominant physical dimensions of instability from two scales: grid and contiguous block. This makes the assessment results accurate, spatially positioned, and explanatory of causes, providing a direct decision-making basis for differentiated treatment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the stability assessment method for high slopes in mountainous highway cuttings based on multi-source data fusion. Figure 2 This is a graph showing the relationship between structural damage and physical property deviation. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1-2 The present invention provides a technical solution: A method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion includes the following steps: S1. Collect slope attribute data, vehicle-by-vehicle records, three-dimensional surface displacement data, deep dip angle data, groundwater level data, rainfall data, apparent resistivity spatial distribution data, and shear wave velocity spatial distribution data within the high slope area of ​​the highway cutting in the target mountainous region.

[0021] In this embodiment, data acquisition equipment is deployed along the slope direction and depth in the high slope area of ​​the highway cutting in the target mountainous area, and multi-source data is acquired synchronously according to a unified time base. The slope attribute data includes slope, slope height, and soil and rock type coded by unique heat. The slope is obtained by surveying the slope topography using a total station or UAV aerial survey, and the slope inclination angle is calculated based on the ratio of the elevation difference between the two ends of the slope to the horizontal distance. The slope height is obtained by measuring the ground elevation at the top and bottom of the slope using a level or total station, and the difference between the elevation at the top and bottom of the slope is taken as the slope height.

[0022] Vehicle-to-vehicle records are obtained through quartz dynamic weighing devices deployed near the roadbed. Each record includes the vehicle's passing time, axle load, speed, vehicle type, and lane number. Three-dimensional displacement data of the ground surface is obtained through global navigation satellite system displacement monitoring stations deployed on the slope surface. Each monitoring station outputs three-dimensional coordinate time series in the east-west, north-south, and vertical directions according to a preset sampling frequency, reflecting the deformation process of the slope surface.

[0023] Deep dip angle data were obtained using fixed inclinometers installed in boreholes on the slope; groundwater level data were obtained using vibrating wire piezometers installed at different elevations on the slope; rainfall data were collected by tipping bucket rain gauges installed in open areas around the slope at a preset sampling frequency, representing the cumulative rainfall over a unit time period; apparent resistivity spatial distribution data were collected by deploying a high-density resistivity array on the slope surface and using the Wenner or Schlenbel electrode configuration method to identify groundwater-rich fractured zones.

[0024] Shear wave velocity spatial distribution data is obtained by deploying multiple seismic receivers on the slope surface, using active source to excite seismic waves, and generating a two-dimensional subsurface shear wave velocity distribution model through surface wave dispersion curve extraction and MASW inversion algorithm. Each sampling point in the model contains the shear wave velocity value corresponding to the spatial coordinates, which is used to identify underground weak interlayers or cavities.

[0025] S2. Based on the vehicle-by-vehicle records, determine the feature vectors of external disturbances. Calculate the feature values ​​of the three-dimensional displacement data, deep dip angle data, groundwater level data, and rainfall data to obtain the feature vectors of macro-evolution. Based on the spatial distribution data of apparent resistivity and shear wave velocity, calculate the deviation values ​​of each sampling point relative to the local neighborhood and interpolate them to a unified spatial grid to form an abnormal structure spatial matrix.

[0026] In this embodiment, the axle load and speed of each vehicle are extracted from the vehicle-by-vehicle records according to a unified evaluation time window. Heavy-load vehicles with axle loads exceeding a preset load threshold are screened, and the number of screened vehicles is counted to obtain the number of times heavy-load vehicles pass. The preset load threshold is determined according to the design load standard of the target mountain highway, for example, it is set to 10 tons. Vehicles with axle loads exceeding 10 tons are marked as heavy-load vehicles. The preset load threshold is used to distinguish between ordinary passenger cars and heavy trucks that have a significant dynamic load impact on the slope.

[0027] A unified assessment time window refers to a standard time interval that divides a continuous monitoring period into fixed durations. The fixed duration of the unified assessment time window can be set to 1 hour, 24 hours, or 7 days, depending on the response frequency of slope monitoring and the fluctuation cycle of traffic flow, to ensure that all time series data are aligned under the same time reference.

[0028] The arithmetic mean of the axle loads of heavy-duty vehicles within a unified assessment time window is calculated to obtain the average axle load. Simultaneously, the arithmetic mean of the vehicle speeds of heavy-duty vehicles is calculated to obtain the average vehicle speed. The formula for calculating the average axle load is: In the formula, Indicates average axle load. This indicates the total number of heavy-duty vehicles within the unified assessment time window. This represents the index of heavily loaded vehicles within the unified evaluation time window. ; Indicates the first The axle load value of a heavy-duty vehicle.

[0029] The formula for calculating average vehicle speed is: In the formula, Indicates average vehicle speed. This indicates the total number of heavy-duty vehicles within the unified assessment time window. Indicates the first The speed of a heavy-duty vehicle.

[0030] The number of times heavy-load vehicles pass through, the average axle load, and the average vehicle speed are combined in sequence into a one-dimensional vector to obtain the external disturbance class feature vector. The order is as follows: the three components in the external disturbance class feature vector are arranged in the order of "number of times heavy-load vehicles pass through - average axle load - average vehicle speed".

[0031] Time series information was extracted from the three-dimensional surface displacement data, deep dip angle data, groundwater level data, and rainfall data according to a unified evaluation time window. During extraction, for each type of data, all data points whose acquisition time falls between the start and end times (including the endpoints) of the unified evaluation time window were selected and arranged in chronological order from earliest to latest to obtain the time series of three-dimensional surface displacement, deep dip angle, groundwater level, and rainfall.

[0032] Calculate the characteristic values, including mean, maximum and standard deviation, in the three-dimensional displacement time series of the Earth's surface, the deep dip angle time series, the groundwater level time series and the rainfall time series.

[0033] The formula for calculating the mean is: The formula for calculating the maximum value is: The formula for calculating standard deviation is: In the formula, This represents the total number of data points in the three-dimensional displacement time series, deep dip angle time series, groundwater level time series, and rainfall time series. Indicates the index of the data point. ; Indicates the first The values ​​of each data point; This represents the mean; Indicates the maximum value; It represents the standard deviation.

[0034] The mean, maximum value, and standard deviation of the four types of data are concatenated and spliced ​​into a one-dimensional array in the order of three-dimensional surface displacement, deep dip angle, groundwater level, and rainfall. This one-dimensional array is the macro-evolutionary feature vector.

[0035] Based on the spatial distribution data of apparent resistivity, the arithmetic mean of the resistivity values ​​of all sampling points within a neighborhood defined by a preset radius centered on each sampling point is calculated. The resistivity deviation of each sampling point is obtained by subtracting this arithmetic mean from the resistivity value of each sampling point. The formula for calculating this resistivity deviation is as follows: In the formula, Indicates the first The resistivity deviation value of each sampling point represents the degree of resistivity anomaly of that sampling point relative to the surrounding local area. Indicates the first Measured apparent resistivity values ​​at each sampling point; Indicates the first The set consisting of all sampling points within a neighborhood defined by a preset radius, centered on a sampling point; Represents the set of sampling points The number of sampling points in the middle; Represents the neighborhood set Inner Measured apparent resistivity values ​​at each sampling point For neighborhood set Index of internal sampling points.

[0036] Based on the spatial distribution data of shear wave velocity, the arithmetic mean of the shear wave velocity values ​​of all sampling points within a neighborhood set defined by a preset radius centered on each sampling point is calculated. The shear wave velocity deviation value of each sampling point is obtained by subtracting this arithmetic mean from the shear wave velocity value of each sampling point. The formula for calculating this shear wave velocity deviation value is as follows: In the formula, Indicates the first Shear wave velocity deviation value at each sampling point; Indicates the first Measured shear wave velocity values ​​at each sampling point; Represents the neighborhood set Inner Shear wave velocity values ​​at each sampling point.

[0037] Based on the above, a unified spatial grid is established to cover the high slopes of the highway cuttings in the target mountainous area. The unified spatial grid is based on the digital elevation model of the target slope and is divided at fixed intervals of 1 to 5 meters along the horizontal and vertical directions. Each grid has unique row and column coordinates. The resistivity deviation and shear wave velocity deviation values ​​of each sampling point are interpolated onto the unified spatial grid using the inverse distance weighted interpolation method to form an anomaly structure spatial matrix.

[0038] S3. Based on the external disturbance type feature vector, macro-evolution type feature vector and abnormal structure space matrix, calculate the disturbance sensitivity index, evolutionary offset index and structural damage index of the high slope of the highway cutting in the mountainous area. Align and stitch the disturbance sensitivity index, evolutionary offset index and structural damage index with the corresponding slope attribute data according to a unified spatial grid to form the fusion feature matrix of each grid.

[0039] In this embodiment, the historical maximum values ​​of the feature vectors of external disturbance, macro-evolution, and abnormal structure space matrix in the high slope of the highway cutting in the target mountainous area are obtained. Specifically, the historical maximum values ​​are: within the historical monitoring period of the high slope of the highway cutting in the target mountainous area, the maximum values ​​of each dimension element in the feature vector of external disturbance are counted to form the historical maximum value vector of external disturbance; the maximum values ​​of each dimension element in the feature vector of macro-evolution are counted to form the historical maximum value vector of macro-evolution; and the historical maximum values ​​of resistivity deviation and shear wave velocity deviation in the abnormal structure space matrix are counted.

[0040] Based on the historical maximum value, the external disturbance type feature vector, the macroscopic evolution type feature vector, and the abnormal structure space matrix are normalized to obtain the external disturbance component group, the macroscopic evolution component group, and the internal damage component group. That is, the historical maximum value is used as the denominator benchmark for subsequent normalization operations to uniformly map data of different physical quantities to the interval [0,1].

[0041] The characteristic data within the external disturbance component group, the macro-evolution component group, and the internal damage component group are weighted and summed to obtain the disturbance sensitivity index, the evolutionary deviation index, and the structural damage index.

[0042] The formula for calculating the disturbance sensitivity index is: In the formula, Indicates the disturbance sensitivity index; Indicates the number of characteristic data within a group of external disturbance components; Indicates the first group of external disturbance components The preset perturbation weight coefficients corresponding to each feature data satisfy the condition that the sum of the weights of each component equals 1; In the normalized external disturbance component group, the first... Each feature data value, This serves as an index for the intra-group characteristic data of the external disturbance component group. .

[0043] The formula for calculating the evolutionary shift index is: In the formula, Indicates the evolutionary shift index; This indicates the number of characteristic data points within a macro-evolutionary component group. Represents the first macroscopic evolution component group The preset evolution weight coefficients corresponding to each feature data satisfy the condition that the sum of the weights of each component equals 1; Represents the first macroscopic evolution component group after normalization. Each feature data value; Index of intragroup characteristic data representing macro-evolutionary components. .

[0044] The formula for calculating the structural damage index is: In the formula, Indicates the structural damage index; This indicates the number of characteristic data within the internal damage component group. The characteristic data within the internal damage component group includes two components: the normalized resistivity deviation value and the shear wave velocity deviation value, and the sum of the weights of each component is equal to 1. Indicates the internal damage component group number Preset damage weight coefficients corresponding to each feature data; Indicates the first element in the normalized internal damage component group. Each feature data value, .

[0045] It should be noted that the preset disturbance weight coefficient, preset evolution weight coefficient, and preset damage weight coefficient of each feature data in the group are pre-calibrated based on the historical instability case data of the target slope through the analytic hierarchy process or expert scoring method.

[0046] Table 1: Statistical Table of Structural Damage Index Based on the 15 sets of data shown in Table 1, combined with Figure 2It can be seen that the normalized resistivity deviation and the normalized shear wave velocity deviation generally show a gradually increasing trend, and exhibit different variation amplitudes at different sampling points. This is consistent with the uneven spatial distribution characteristics of underground water-rich fractured zones and weak interlayers in actual slopes. For example, when the sampling point index is 4, the normalized resistivity deviation value suddenly increases from 0.17 to 0.35, while the normalized shear wave velocity deviation value decreases from 0.14 to 0.11, indicating that there is a significant local water-rich fracturing anomaly at this location, but the weak interlayer is not obvious. When the sampling point index is 7, the normalized resistivity deviation value jumps from 0.28 to 0.41, while the normalized shear wave velocity deviation value is also at a low level of 0.15, which again reflects the spatial decoupling characteristics of resistivity anomaly and wave velocity anomaly. The structural damage index generally shows an upward trend with the increase of the normalized deviation value, gradually increasing from 0.068 at sampling point index 1 to 0.728 at sampling point index 15. However, when the two deviation values ​​show different changes, the growth rate of the structural damage index also fluctuates, reflecting the difference in the contribution of damage signals from different physical dimensions to the final index.

[0047] The above data variation patterns verify that the structural damage index constructed in this invention can effectively quantify the severity of damage to underground anomalous structures and has good adaptability to asynchronous changes in resistivity deviation and shear wave velocity deviation. It provides reliable damage dimension features for the subsequent construction of the fusion feature matrix. When the structural damage index is high, the structural damage index is used as an important unstable signal to participate in the fusion, thereby realizing differentiated responses to underground structures under different damage modes and effectively improving the accuracy of the assessment results.

[0048] The grid set is obtained by extracting all grid positions of the unified spatial grid from the abnormal structure space matrix. Specifically, the abnormal structure space matrix is ​​a two-dimensional numerical array built according to the row and column structure of the unified spatial grid. The row index and column index corresponding to each storage position in the two-dimensional numerical array are directly used as the grid coordinates. All combinations of row index and column index are summarized to obtain the grid set.

[0049] The disturbance sensitivity index, evolutionary offset index, and structural damage index are respectively assigned to each grid in the grid set.

[0050] For each grid, extract the slope, slope height, and uniquely coded soil and rock type from the slope attribute data corresponding to that grid.

[0051] It should be noted that the specific implementation of the unique hot coding of soil and rock types is as follows: a set of all soil and rock types appearing in the high slope area of ​​the highway cutting in the target mountainous area is predefined, a unique position index is assigned to each soil and rock type, the position of the index is set to 1 and the rest are set to 0, forming a binary vector of the same length as the total number of soil and rock types; for example, limestone, sandstone and mudstone are encoded as [1,0,0], [0,1,0] and [0,0,1] respectively.

[0052] The disturbance sensitivity index, evolution offset index, structural damage index, slope, slope height, and uniquely coded soil and rock type of each grid are sequentially spliced ​​together, and the splicing results are summarized according to grid position to form a fused feature matrix.

[0053] S4. Input the fusion feature matrix of each grid into the trained slope stability assessment model, output the stability score, and arrange each stability score in space according to the spatial location of the high slope of the mountain highway cutting to obtain the spatial distribution map of the stability score.

[0054] In this embodiment, the slope stability assessment model is constructed using a neural network based on a multilayer perceptron. The multilayer perceptron network includes an input layer, a hidden layer, and an output layer. The hidden layer has at least two neurons and uses ReLU as the activation function.

[0055] The fusion feature matrix and corresponding historical stability score of multiple historical mountain highway cutting high slope samples were collected. Multiple sets of samples were constructed and divided into training set, validation set and test set in a ratio of 7:2:1. The training set is used to learn the model parameters; the validation set is used to adjust the hyperparameters during training to prevent overfitting; and the test set is used to evaluate the generalization ability of the model after training.

[0056] The structure of the slope stability assessment model is as follows: Input layer: used to receive the fused feature matrix; Hidden layer: Set to 1 layer with 8 neurons, using ReLU as the activation function; Output layer: has 1 neuron, used to output stability score; The process of training the slope stability assessment model is as follows: The model input is a fusion feature matrix of multiple historical mountain highway cutting slope samples. The historical stability score of the corresponding sample is used as the real supervision label. The historical stability score is a continuous value between 0 and 1, which is based on the overall stability state of each sample during the corresponding monitoring period. The mean squared error loss function is selected, and the average of the sum of squared differences between the stability score predicted by the model and the historical stability score is calculated to obtain the model loss value. The gradient descent optimization algorithm is used to update each parameter according to the direction of the learning rate multiplied by the gradient. The model parameters are iteratively updated through the backpropagation algorithm. When the model loss value is less than the preset loss threshold or the preset maximum number of iterations is reached, the slope stability assessment model is determined to have converged and training is stopped, resulting in a trained slope stability assessment model. In this embodiment, the loss threshold is set to 0.01 and the maximum number of iterations is set to 1000.

[0057] It should be noted that the historical stability score is the true supervision label during the model training phase and is not used as an input feature in the fusion feature matrix. The historical stability score is obtained as follows: each segment of monitoring data divided into a unified evaluation time window within the historical monitoring period is used as a training sample. Combined with the actual deformation and failure monitoring data of the slope in the corresponding time period, the sample is assigned a continuous score value between 0 and 1 based on the actual deformation and failure severity of the historical mountain highway cutting high slope in the actual deformation and failure data. Here, 0 represents that the slope is completely stable in the time period, and 1 represents that the slope has become unstable or is in a critical state of instability in the time period. The closer the score value is to 1, the worse the slope stability in that time period.

[0058] The fused feature matrix of each grid is input into the trained slope stability assessment model to output a stability score; each stability score is mapped one-to-one with the grid of the unified spatial grid to obtain a set of spatial score points.

[0059] Spatial scoring points are spatially interpolated and rasterized within the actual geographic space of the high slope of the highway cutting in the target mountainous area according to the coordinate system of a unified spatial grid, to obtain a spatial distribution map of stability scores.

[0060] The specific method of rasterization rendering is as follows: the range of stability scores is uniformly mapped to a continuous color band from cool to warm tones. The higher the stability score, the more it corresponds to warm colors, and the lower the stability score, the more it corresponds to cool colors. The score value of each grid is assigned a corresponding color according to the continuous color band from cool to warm tones, generating a raster heat map that can be overlaid and displayed in a 3D GIS platform, which is the spatial distribution map of stability scores.

[0061] S5. Screen low-stability grids from the stability score spatial distribution map and obtain a set of grids with potential sliding hazards. Cluster them into contiguous hazard blocks, determine the dominant instability factors of each grid, and count the proportion of grids with various dominant instability factors in the contiguous hazard blocks to determine the dominant instability factors of the contiguous hazard blocks.

[0062] In this embodiment, grids with scores higher than a preset threshold are extracted from the stability score spatial distribution map to form a set of low-stability grids. The preset threshold is obtained by collecting samples of slopes that have historically become unstable and slopes that have remained stable within the area of ​​the high slope of the highway cutting in the target mountainous region. The fusion feature matrix of each sample is extracted and input into the trained slope stability assessment model to obtain the stability score of each sample. Based on the stability scores of all unstable and stable slope samples, the receiver operating characteristic (ROC) curve analysis method is used to calculate the true positive rate and false positive rate under different score thresholds. The stability score corresponding to the maximum Youden index is selected as the preset threshold.

[0063] By performing spatial intersection analysis between the low-stability mesh set and the resistivity and shear wave velocity anomaly meshes within the anomalous structure spatial matrix that meet the preset deviation threshold, a potential sliding hazard mesh set is obtained.

[0064] The resistivity within the abnormal structure space matrix that meets the preset deviation threshold refers to the grid positions where the interpolated resistivity deviation value selected from the abnormal structure space matrix is ​​greater than the resistivity trough threshold. Shear velocity anomaly grids refer to the grid locations where the interpolated shear velocity deviation value is greater than the velocity mutation threshold, selected from the anomaly structure space matrix. The methods for obtaining the resistivity trough threshold and the wave velocity mutation threshold are as follows: statistical analysis is performed on the normalized resistivity deviation value sequence and shear wave velocity deviation value sequence in the historical instability samples, and the threshold is determined by the mean plus twice the standard deviation method. The specific threshold is calibrated according to the actual geological conditions during implementation.

[0065] Potentially adjacent sliding hazard grids in the set of potential sliding hazard grids are clustered to form contiguous hazard blocks. The criterion for determining adjacency is as follows: for any two potential sliding hazard grids, if the absolute value of the difference between their row indices and the absolute value of the difference between their column indices in the same spatial grid do not exceed 1, then the two potential sliding hazard grids are determined to be adjacent to each other. All potential sliding hazard grids are traversed, and adjacent grids are grouped into the same contiguous hazard block until all grids are grouped into a certain block, and each block contains at least one potential sliding hazard grid.

[0066] For a potential sliding hazard grid set, the normalized values ​​of the external disturbance component group, macro-evolution component group, and internal damage component group of each grid within the set are compared. The dimension of the component with the largest value is taken as the dominant instability factor corresponding to the grid. During the comparison, the normalized values ​​of the external disturbance component group, macro-evolution component group, and internal damage component group are traversed to find the component with the largest value: if the largest component belongs to the external disturbance component group, the dominant instability factor is determined to be traffic load disturbance; if the largest component belongs to the macro-evolution component group, the dominant instability factor is determined to be surface deformation or hydrological evolution; if the largest component belongs to the internal damage component group, the dominant instability factor is determined to be underground structural damage.

[0067] The proportion of each dominant instability factor in each contiguous danger zone is statistically analyzed. Specifically, for any contiguous danger zone, the number of grids for which the dominant instability factors are traffic load disturbance, surface deformation or hydrological evolution, and underground structural damage are counted. The total number of grids in the contiguous danger zone is used as the denominator, and the number of grids corresponding to each of the three dominant instability factors is divided by the total number of grids to obtain the proportion of the three dominant instability factors. The dimension with the highest proportion is selected as the dominant instability factor of the contiguous danger zone.

[0068] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0069] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0070] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0071] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion, characterized in that, The specific steps include: S1. Collect slope attribute data, vehicle-by-vehicle records, three-dimensional surface displacement data, deep dip angle data, groundwater level data, rainfall data, apparent resistivity spatial distribution data, and shear wave velocity spatial distribution data within the high slope area of ​​the highway cutting in the target mountainous area. S2. Determine the feature vector of external disturbance based on the vehicle passing record, calculate the feature value of the three-dimensional displacement data of the ground surface, the deep dip angle data, the groundwater level data and the rainfall data to obtain the feature vector of macro-evolution, and calculate the deviation value of each sampling point relative to the local neighborhood based on the spatial distribution data of apparent resistivity and the spatial distribution data of shear wave velocity and interpolate it to a unified spatial grid to form an abnormal structure spatial matrix. S3. Based on the external disturbance type feature vector, macro-evolution type feature vector and abnormal structure space matrix, calculate the disturbance sensitivity index, evolutionary offset index and structural damage index of the high slope of the highway cutting in the mountainous area, and align and stitch the three indices with the corresponding slope attribute data according to a unified spatial grid to form the fusion feature matrix of each grid. S4. Input the fusion feature matrix of each grid into the trained slope stability assessment model, output the stability score, and arrange each stability score in space according to the spatial location of the high slope of the mountain highway cutting to obtain the spatial distribution map of the stability score. S5. Screen low-stability grids from the stability score spatial distribution map and obtain a set of grids with potential sliding hazards. Cluster them into contiguous hazard blocks, determine the dominant instability factors of each grid, and count the proportion of grids with various dominant instability factors in the contiguous hazard blocks to determine the dominant instability factors of the contiguous hazard blocks.

2. The method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion as described in claim 1, characterized in that: The external disturbance feature vector is determined based on the vehicle-by-vehicle passing records. The specific steps are as follows: Extract the axle load and speed of each vehicle from the vehicle-by-vehicle records according to a unified assessment time window, filter out heavy-load vehicles whose axle load exceeds the preset load threshold, count the number of filtered vehicles, and obtain the number of times heavy-load vehicles pass. A unified assessment time window refers to a standard time interval that divides a continuous monitoring period into fixed durations. The arithmetic mean of the axle loads of heavy-duty vehicles within a unified assessment time window is calculated to obtain the average axle load. At the same time, the arithmetic mean of the speeds of heavy-duty vehicles is calculated to obtain the average speed. The number of times heavy-load vehicles pass through, the average axle load, and the average vehicle speed are combined into a one-dimensional vector in a fixed order to obtain the feature vector of external disturbance.

3. The method for assessing the stability of high slopes in mountainous highway cuttings based on multi-source data fusion according to claim 2, characterized in that: Feature values ​​are calculated from surface three-dimensional displacement data, deep dip angle data, groundwater level data, and rainfall data to obtain macro-evolutionary feature vectors. The specific steps are as follows: The corresponding time series information was extracted from the three-dimensional surface displacement data, deep dip angle data, groundwater level data and rainfall data according to a unified evaluation time window, and the time series of three-dimensional surface displacement, deep dip angle, groundwater level and rainfall were obtained. Calculate the characteristic values, including mean, maximum and standard deviation, in the three-dimensional displacement time series of the Earth's surface, the deep dip angle time series, the groundwater level time series and the rainfall time series respectively; The mean, maximum value, and standard deviation of the four types of data are concatenated and spliced ​​into a one-dimensional array in the order of three-dimensional surface displacement, deep dip angle, groundwater level, and rainfall. This one-dimensional array is the macro-evolutionary feature vector.

4. The method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion according to claim 1, characterized in that: Based on the spatial distribution data of apparent resistivity and shear wave velocity, the deviation values ​​of each sampling point relative to its local neighborhood are calculated and interpolated to a unified spatial grid to form an anomaly structure spatial matrix. The specific steps are as follows: Based on the spatial distribution data of apparent resistivity, the arithmetic mean of the resistivity values ​​of all sampling points within the neighborhood set defined by a preset radius with each sampling point as the center is calculated. The resistivity value of each sampling point is subtracted from the arithmetic mean to obtain the resistivity deviation value of each sampling point. Based on the spatial distribution data of shear wave velocity, the arithmetic mean of the shear wave velocity values ​​of all sampling points within the neighborhood set defined by a preset radius with each sampling point as the center is calculated. The shear wave velocity value of each sampling point is subtracted from the arithmetic mean to obtain the shear wave velocity deviation value of each sampling point. A unified spatial grid covering the high slopes of highway cuts in the target mountainous area was established. The resistivity deviation and shear wave velocity deviation of each sampling point were interpolated to each grid of the unified spatial grid using the inverse distance weighted interpolation method to form an anomaly structure spatial matrix.

5. The method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion according to claim 1, characterized in that: Based on external disturbance feature vectors, macro-evolution feature vectors, and anomalous structure space matrices, the disturbance sensitivity index, evolutionary offset index, and structural damage index of high slopes in mountainous highway cuttings are calculated. The specific steps are as follows: Obtain the historical maximum values ​​of the feature vectors of external disturbances, macro-evolution, and the spatial matrix of abnormal structures in the high slope of the highway cutting in the target mountainous area; Based on the maximum value of historical data, the feature vectors of external disturbance, macro-evolution, and abnormal structure space matrix are normalized to obtain the external disturbance component group, macro-evolution component group, and internal damage component group. The characteristic data within the external disturbance component group, the macro-evolution component group, and the internal damage component group are weighted and summed to obtain the disturbance sensitivity index, the evolutionary deviation index, and the structural damage index.

6. The method for assessing the stability of high slopes in mountainous highway cuttings based on multi-source data fusion as described in claim 1, characterized in that: The disturbance sensitivity index, evolutionary offset index, and structural damage index are aligned and stitched together with the corresponding slope attribute data according to a unified spatial grid to form a fused feature matrix for each grid. The specific steps are as follows: Extract all grid positions of the unified spatial grid from the abnormal structure spatial matrix to obtain the grid set; The disturbance sensitivity index, evolution offset index, and structural damage index are respectively assigned to each grid in the grid set; For each grid, extract the slope, slope height, and uniquely thermally encoded soil and rock type from the slope attribute data corresponding to that grid; The disturbance sensitivity index, evolution offset index, structural damage index, slope, slope height, and uniquely thermally encoded soil and rock type of each grid are sequentially spliced ​​together to form a fused feature matrix.

7. The method for stability assessment of high slopes in mountainous highway cuttings based on multi-source data fusion according to claim 1, characterized in that: The specific steps to determine the dominant factors causing instability in contiguous hazardous areas are as follows: Extract grids with scores higher than a preset threshold from the stability score spatial distribution map and form a set of low-stability grids. By performing spatial intersection analysis between the low-stability mesh set and the resistivity and shear wave velocity anomalies within the spatial matrix of the anomalous structure that meet the preset deviation threshold, a potential sliding hazard mesh set is obtained. Clustering spatially adjacent potential sliding hazard grids in the potential sliding hazard grid set to form contiguous hazard blocks; For a set of grids with potential sliding hazards, the normalized values ​​of the external disturbance component group, macro-evolution component group, and internal damage component group of each grid in the set are compared, and the component dimension with the largest value is taken as the dominant factor of instability corresponding to the grid. The grid proportions of various dominant instability factors within each contiguous danger zone are statistically analyzed, and the dimension with the highest proportion is selected as the dominant instability factor for the contiguous danger zone.

Images