Mountain wind power maintenance road side slope stability analysis method

Abstract

The application discloses a mountainous wind power overhauling road side slope stability analysis method and relates to the technical field of road monitoring. The steps of the method comprise the following steps: obtaining overhauling road data and side slope types, measuring rock and soil quality section parameters and building a data set; calculating anti-skid stability safety coefficients according to road load levels, converting three types of data sets into standardized scores according to specifications, forming a parameter standardized score data set, and obtaining conditional scores by using a principal component analysis method; taking monitoring point slope heights, combining side slope types to calculate slope height amplification coefficients, and calculating a first risk value; obtaining image data sets by using a UAV to take pictures, detecting corner points by using a Harris algorithm, matching by using a normalized cross-correlation algorithm, obtaining corner point matching structured parameters, and calculating surface similarity by using a similarity representation algorithm to obtain a second risk value; calculating a comprehensive risk value by using a Bayesian weighted fusion algorithm to calculate two risk values, and evaluating side slope stability. The comprehensive risk value is output by using a hierarchical risk coupling algorithm, and a high-risk side slope section is accurately locked.

Description

Technical Field

[0001] This invention relates to the field of monitoring engineering, specifically to a method for analyzing the stability of slopes along maintenance roads for mountain wind power projects. Background Technology

[0002] Mountain wind power is a crucial area for clean energy development, and the stability of the slopes along maintenance roads directly impacts the transportation safety, operation and maintenance efficiency, and lifespan of wind turbine equipment. In my country, mountain wind power projects are mostly located in areas with complex terrain and variable climates. Slopes are frequently susceptible to instability due to multiple factors, including geological structure, heavy-load transportation, and freezing-thaw cycles caused by heavy rain. Industry statistics indicate that approximately 35% of mountain wind power operation and maintenance accidents are related to slope collapses, resulting in significant economic losses. Therefore, accurately assessing the stability of maintenance road slopes has become a core requirement for ensuring the safe operation of mountain wind power projects.

[0003] Current slope stability analysis methods primarily rely on single geological exploration approaches. These involve drilling core samples in-situ, conducting mechanical tests such as uniaxial compressive strength and shear strength, and combining these with parameters from geological borehole data, including core quality indicators, joint characteristics, and groundwater conditions. A safety factor is calculated using the limit equilibrium method or a standard scoring table to determine the slope's stability. However, the main problem with these methods lies in the limitations of geological exploration. Specifically, the correlation between geological parameters and slope surface morphology is severed. Indicators such as rock mass integrity and soil compaction cannot be coupled with dynamic characteristics such as surface crack development and displacement changes. This makes the assessment results insufficient to fully reflect the true stability under complex conditions, posing a potential safety hazard to the safe operation of maintenance roads for mountain wind power projects. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for analyzing the stability of slopes along maintenance roads in mountainous wind power areas, solving the problem that indicators such as rock mass integrity and soil compaction cannot be coupled with dynamic characteristics such as surface crack development and displacement changes for analysis.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for analyzing the slope stability of a mountain wind power maintenance road, comprising the following steps: Step S1: Obtain the maintenance road dataset and road slope type of the mountain wind power maintenance road, collect slope rock cores on site, and conduct uniaxial compressive strength test on the rock section of the slope rock core based on the road slope type to obtain the rock section parameters, and conduct shear strength test on the soil section of the slope rock core to obtain the soil section parameters. Step S2: Based on the rock segment parameters and soil segment parameters, construct corresponding rock segment datasets and soil segment datasets; Step S3: Based on road design specifications, the rock section dataset, soil section dataset, and maintenance road dataset are converted into standardized scores to obtain a parameter standardized score dataset; the parameter standardized score dataset is analyzed using principal component analysis to calculate the conditional score; the anti-skid stability safety factor is calculated based on the road load level in the maintenance road dataset. Step S4: Obtain the slope height of the monitoring points on the mountain wind power maintenance road; calculate the slope height amplification factor based on the slope height of the monitoring points and the road slope type; analyze and calculate the slope height amplification factor, condition score and anti-slip stability safety factor using a multi-parameter coupled weighted algorithm to obtain the first risk value; Step S5: Take aerial photos of the slopes of the mountain wind power maintenance road using a drone to obtain an aerial image dataset; perform corner detection on the aerial image dataset using the Harris algorithm to obtain a high-confidence corner set; perform corner matching on the high-confidence corner set using the normalized cross-correlation corner matching algorithm to obtain corner matching structured parameters. Step S6: Based on the corner matching structured parameters, the surface similarity of the slope of the mountain wind power maintenance road is calculated using a similarity characterization algorithm to obtain the surface similarity; based on the surface similarity, the second risk value is calculated. Step S7: Calculate the first risk value and the second risk value using a Bayesian weighted fusion algorithm to obtain a comprehensive risk value. Based on the comprehensive risk value, calculate the slope stability of the mountain wind power maintenance road.

[0006] Preferably, the dataset includes the road slope type and maintenance road data for mountain wind power maintenance roads, including: A surveying section was set up every 50 meters along the longitudinal direction of the maintenance road. Differential GPS was used to accurately locate the coordinates of each section. Slope types were classified through geological hammer testing and overburden thickness detection. The classification rules are as follows: When the exposed bedrock thickness accounts for ≥70%, it is defined as a rocky section; when the fill area or the overburden thickness is >1.5 meters, it is defined as a soil section. Road maintenance dataset The construction process is as follows: Road maintenance dataset Including wind turbine transport load Slope roughness index ; The calculation process for the transport load of the wind turbine is as follows:

[0007] in, It is the wind turbine transport load, used to verify whether the road needs to be reinforced; This is the weight of the wind turbine blades, taken as 80 tons; It refers to the number of axles on the transport vehicle; This refers to the single-axle load limit for road design. Slope roughness index The calculation process is as follows: Use a drone equipped with LiDAR to scan the slope and obtain the elevation value of point p in the point cloud. ;

[0008] in, It is the slope roughness index, used to quantify surface deformation and associated with the second risk value; It is the elevation value of point p in the point cloud, collected by a drone; It is the average elevation of a local window, calculated using a GIS sliding window. It represents the number of point clouds within the window.

[0009] Preferably, based on the road slope type, uniaxial compressive strength tests are conducted on the rock section of the slope core to obtain rock section parameters, and shear strength tests are conducted on the soil section of the slope core to obtain soil section parameters, including: The test procedure for geotechnical mechanical parameters is as follows: Processing rock cores into A standard cylindrical specimen measuring 50mm × 100mm was subjected to a 2000kN compression testing machine and loaded at a rate of 0.5MPa / s until failure. The peak load P was recorded, and the saturated uniaxial compressive strength was calculated using the formula. :

[0010] in, It is the destructive load; A is the cross-sectional area of ​​the specimen, calculated after testing perpendicular to the bedding direction in layered rock mass. Its function is to evaluate rock strength and is used in BQ rock mass classification; The shear strength test of the soil section was conducted using a direct shear apparatus for consolidated rapid shear tests. Four levels of normal stress were applied at a shear rate of 0.8 mm / min. Shear stress-displacement curves were plotted, and equations were established based on the Mohr-Coulomb criterion.

[0011] in, To break shear stress; is the normal stress, applied via a lever-weight system; c is the cohesive force obtained from a direct shear test. The internal friction angle is obtained from the direct shear test. After the test, 85% of the peak strength is taken as the design parameter to obtain the soil strength parameters. The safety factor is calculated by inputting the limit equilibrium method.

[0012] Preferably, step S2 includes: Rock segment dataset Including uniaxial compressive strength Rock mass integrity coefficient The proportion of cumulative core length Joint spacing Joint quality score ; Soil Section Dataset Including cohesion internal friction angle On-site measured dry density pore water pressure , hydraulic gradient i.

[0013] Preferably, the conditional score is calculated by analyzing the standardized score dataset of the parameters using principal component analysis, including: The principal component extraction process is as follows: Select the smallest k that satisfies the cumulative contribution rate to calculate the principal component scores of k samples; The process of selecting the minimum k is as follows:

[0014] in, These are the eigenvalues ​​arranged in descending order, and their function is to determine the number of principal components k to retain. The principal component score calculation process for the i-th sample is as follows:

[0015] Among them, principal component scores These are standardized values, used to project the samples onto the principal component space; It is the standardized vector of the i-th sample; It is the eigenvector of the l-th principal component; Conditional scoring The synthesis process is as follows:

[0016]

[0017] in, These are the principal component weights, and their function is to allocate weights according to variance contribution. It is the eigenvalue of the l-th principal component; It is a conditional score, and its function is to generate a comprehensive stability score.

[0018] Preferably, the slope height of the monitoring points on the mountain wind power maintenance road is obtained, and based on the slope height of the monitoring points and the road slope type, the slope height amplification factor is calculated, including: An RTK real-time dynamic positioning system is used to collect the three-dimensional coordinates of monitoring points on site and extract elevation values. Select monitoring points based on their elevation. Elevation of the foot of the slope Calculate the slope height of the monitoring point The process is as follows:

[0019] in, It is the elevation of the monitoring point; It is the elevation of the foot of the slope; Slope height magnification factor The definition is as follows:

[0020] in, It is the slope height amplification factor, which quantifies the slope height effect; It is the slope height of the monitoring point; It is the rock slope height index, with values ​​assigned according to rock mass type: 0.35 for granite and 0.45 for shale. It is a type of road slope; It is a conditional scoring; It is the conditional score attenuation index, which is determined according to the rock mass integrity. For BQ>550, it is 0.8. This is the soil sensitivity coefficient, with a default value of 0.05. The slope height is determined by the power of the soil slope, calibrated through direct shear tests: 1.3 for clay and 1.2 for sand.

[0021] Preferably, a first risk value is obtained by analyzing and calculating the slope height amplification factor, condition score, and anti-slide stability safety factor using a multi-parameter coupled weighted algorithm, including: First risk value The calculation process is as follows:

[0022]

[0023]

[0024] in, It is the first risk value, and its function is to comprehensively quantify risk, with a limit of 10 on the upper limit of the risk value. It is the slope height amplification factor; It is a conditional scoring; It is the anti-slip stability safety factor. ; It is the rainfall-landslide sensitivity coefficient. = This data is derived from historical data. This is the cumulative rainfall over the past 7 days; This is a rainy season detection switch function; it is used when the rainy season is in effect. =1, dry season =0, The higher the value, the higher the risk, especially during periods of heavy rain. When the thickness exceeds 100mm, the risk increases exponentially, which is consistent with the characteristics of landslide disasters.

[0025] Preferably, the corner matching structured parameters are obtained, including: The displacement vector calculation process is as follows:

[0026]

[0027] This formula is used to calculate the displacement of each corner point from the reference image to the target image; The formula for calculating quality weights is as follows:

[0028] in, It is the quality weight, which is used to assign weights to each displacement vector for subsequent weighted fusion. The weights are determined by the significance of the corner points and the matching quality. It is the normalized Harris response value of the corner point; This is the NCC correlation coefficient for that point pair; The definition is as follows:

[0029] in, These are the corner matching structured parameters, which are a set of affine transformation matrices, displacement vector sets, and mass weights.

[0030] Preferably, step S6 includes: Structural similarity The calculation process is as follows (deformation stability assessment):

[0031] in, It is structural similarity, and its function is to quantify the stability of structural deformation. It is the number of points within RANSAC; It is the sum of the weights of the effective matching points. Quality weights from S5; It is the determinant of the affine matrix; H comes from the affine matrix of S5; It is the wind power rock mass deformation sensitivity coefficient. =0.3; It is the scaling sensitivity factor. =0.5; It is the Frobenius norm, used to capture overall deformation; It is the Frobenius norm of the affine matrix and the identity matrix; Displacement similarity The calculation process is as follows:

[0032]

[0033] in, It is displacement similarity, and its function is to quantify positional stability; This represents the number of valid matches. It is the cumulative rainfall over seven days, and the infiltration amplification factor of the rainstorm; It is the impact coefficient of rainstorm. ; It is the magnitude of the single-point displacement vector; The displacement vector from S5; The formula for the dual-metric weighted fusion algorithm is as follows:

[0034] in, It is surface similarity; under normal circumstances, = =0.5, equilibrium deformation and displacement; during heavy rain, i.e. >100, =0.3, =0.7, focusing on the risk of seepage displacement; during freeze-thaw cycles, i.e., the average temperature over seven consecutive days. <0 , = , =1- , The focus is on the structural deterioration caused by frost heave; The calculation process for the second risk value is as follows:

[0035] in, It is the dimension conversion coefficient, k=100; the second risk value. It is negatively correlated with surface similarity.

[0036] Preferably, a comprehensive risk value is obtained by calculating the first risk value and the second risk value using a Bayesian weighted fusion algorithm. Based on the comprehensive risk value, the slope stability of the mountain wind power maintenance road is calculated, including: The calculation process for the comprehensive risk value is as follows:

[0037]

[0038] in, It is a comprehensive risk value, which linearly integrates geological and deformation risks as well as extreme rainfall risk assessments. When the thickness is >150mm and the fusion value is >5, the maximum risk is forced to be 10. It is a dynamic factor, under normal conditions Balancing geological and deformation risks; It is the first risk value; It is the second risk level; it is in the midst of a rainstorm. When >100, 3. Focus on the risk of deformation over time; during the freeze-thaw cycle, i.e., the average temperature over seven consecutive days. <0 hour, 7. Focus on structural degradation risk; Traditional risk assessment methods, such as fault tree analysis, have single node states, making it difficult to handle polymorphic risks and complex dependencies, and they cannot be dynamically updated. The analytic hierarchy process (AHP) is highly subjective with fixed weights. In contrast, this algorithm can handle uncertainty by weighted fusion of multi-source information. It supports node polymorphism correction, which is more in line with engineering realities. It can dynamically update probabilities to adapt to dynamic changes in the construction process. It accurately identifies key risk factors through conditional probability tables and posterior inference, and verifies causal relationships by combining sensitivity analysis. It integrates fault tree logic and fuzzy evaluation, taking into account both qualitative and quantitative aspects, and significantly improves the scientificity and accuracy of risk assessment. Stability coefficient The formula is as follows:

[0039] in, It is a stability coefficient that maps risk values ​​to engineering stability indicators; when When, it is in a stable state; when At that time, it is in a basically stable state; when At that time, it was in an unstable state.

[0040] This invention provides a method for analyzing the slope stability of maintenance roads in mountainous wind power areas, involving machine learning and deep learning technologies, which has the following beneficial effects: (1) The method for analyzing the stability of the mountain wind power maintenance road slope breaks through the limitations of the single geological survey method by using a dual coupling analysis mechanism of geological survey and surface monitoring. That is, it integrates the mechanical parameters of rock core, geological survey data and surface deformation information obtained by UAV aerial photography, and models the characteristics of rock and soil slopes differently. It quantitatively correlates static geological indicators such as rock integrity and soil shear strength with dynamic deformation characteristics such as surface displacement and crack development, so as to realize the collaborative analysis of multi-dimensional data and comprehensively reflect the stability of the slope.

[0041] (2) The method for analyzing the stability of the mountain wind power maintenance road slope, with the improved Harris algorithm and dual-metric weighted fusion algorithm, improves the accuracy of corner matching, drives the calculation of slope surface similarity, and outputs the coupling results of the first and second risk values ​​by combining dynamic weight adjustment, which significantly improves the synchronization of deformation monitoring and risk quantification under complex terrain.

[0042] (3) The method for analyzing the stability of slopes along maintenance roads in mountainous wind power areas achieves a closed loop from slope stability assessment to decision-making through the engineering mapping of comprehensive risk values ​​and stability coefficients. Based on the comprehensive risk values, the method classifies the slopes into stable, basically stable, and unstable states. Combined with the special judgment rules for extreme working conditions such as rainstorms and freeze-thaw cycles, it strengthens the response capability to the special environment of mountainous wind power areas and provides accurate quantitative basis for the safety control of maintenance road slopes. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for analyzing the stability of mountain wind power maintenance road slopes proposed in this invention.

[0044] Figure 2 This invention provides a hierarchical diagram of the first risk value obtained from a method for analyzing the stability of slopes along mountain wind power maintenance roads.

[0045] Figure 3 This invention provides a hierarchical diagram of the second risk value obtained from a method for analyzing the stability of mountain wind power maintenance road slopes. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Please see Figure 1 This invention provides a technical solution: a method for analyzing the slope stability of mountain wind power maintenance roads. Specifically, the following method for analyzing the slope stability of mountain wind power maintenance roads is provided; please refer to [link / reference]. Figure 1 The method includes the following steps: Step S1: Obtain the road slope type and maintenance road dataset of the mountain wind power maintenance road, collect slope rock cores on site, and conduct uniaxial compressive strength tests on the rock section of the slope rock cores based on the road slope type to obtain the rock section parameters, and conduct shear strength tests on the soil section of the slope rock cores to obtain the soil section parameters.

[0048] A surveying section was established every 50 meters along the longitudinal direction of the maintenance road, and the coordinates of each section were accurately located using differential GPS. Slope types were classified based on geological hammer testing and overburden thickness detection, according to the following rules: When the exposed bedrock thickness accounts for ≥70%, it is defined as a rocky section; when the fill area or the overburden thickness is >1.5 meters, it is defined as a soil section.

[0049] Road maintenance dataset The construction process is as follows: Road maintenance dataset Including wind turbine transport equivalent load Slope roughness index .

[0050] The calculation process for the wind turbine transport load is as follows (modeled as a three-axle heavy vehicle):

[0051] in, It is the equivalent load for wind turbine transportation, used to verify whether the road needs to be reinforced; This is the weight of the wind turbine blades, taken as 80 tons; This is the number of axles on the transport vehicle (default is 3). This refers to the design limit for a single axle load on a road, typically 260 kN. The verification rules are as follows: when... When the load is greater than 260kN, output the "reinforcement required" flag.

[0052] Slope roughness index The calculation process is as follows: Use a drone equipped with LiDAR to scan the slope and obtain the elevation value of point p in the point cloud. .

[0053]

[0054] in, It is the slope roughness index, used to quantify surface deformation and associated with the second risk value; It is the elevation value of point p in the point cloud, collected by a drone; It is the average elevation of a local window (5m×5m), calculated using a GIS sliding window; It represents the number of point clouds within the window.

[0055] The core sample drilling and preparation process is as follows: In the rocky section, a fully hydraulic drilling rig was used to drill to a depth of 3 meters below the stable bedrock surface, with each drilling pass being ≤1 meter. After the complete rock core was extracted, it was immediately sealed and the azimuth was marked. It was then saturated for 24 hours to simulate heavy rain conditions.

[0056] In soil sections, undisturbed soil samples are obtained using a thin-walled soil sampler with static pressure, at a depth of 0.5-1.2 meters below the roadbed. In embankment sections, samples are collected primarily at the interface of compacted layers, with a sample diameter ≥100mm and a height-to-diameter ratio strictly maintained at 2:1.

[0057] The test procedure for geotechnical mechanical parameters is as follows: The uniaxial compressive strength test of the rock section was performed in accordance with the "Standard for Testing Methods of Engineering Rock Mass". The rock core was processed into... A standard cylindrical specimen measuring 50 mm × 100 mm was loaded onto a 2000 kN pressure testing machine at a rate of 0.5 MPa / s until failure. The peak load P was recorded, and the saturated uniaxial compressive strength was calculated using the formula. :

[0058] in, It is the destructive load; A is the cross-sectional area of ​​the specimen (mm²). 2 ), and the calculation is performed after testing the layered rock mass in the direction perpendicular to the bedding. Its function is to evaluate rock strength and is used in BQ rock mass classification.

[0059] The shear strength test of the soil section was conducted using a direct shear apparatus for consolidated rapid shear tests, with four levels of normal stress (50 kPa, 100 kPa, 200 kPa, 400 kPa) applied at a shear rate of 0.8 mm / min. Shear stress-displacement curves were plotted, and equations were established based on the Mohr-Coulomb criterion.

[0060] in, To break shear stress; is the normal stress, applied via a lever-weight system; c is the cohesive force obtained from a direct shear test. The internal friction angle was obtained from a direct shear test. After the test, 85% of the peak strength was taken as the design parameter to obtain the soil strength parameters, which were then input into the limit equilibrium method to calculate the safety factor.

[0061] This step involves conducting on-site surveys of mountain wind power maintenance roads to obtain road slope types and maintenance road datasets. Targeted rock core samples are then collected and mechanical tests are carried out to obtain core mechanical parameters such as uniaxial compressive strength of the rock section and shear strength (cohesion, internal friction angle) of the soil section. This provides basic data for subsequent analysis of the fundamental strength characteristics of the rock and soil masses and is a key step in quantifying the mechanical properties of slope materials, directly supporting the assessment of the essential characteristics of rock and soil slope stability.

[0062] Step S2: Based on the rock segment parameters and soil segment parameters, construct corresponding rock segment datasets and soil segment datasets.

[0063] Rock segment dataset The structure is as follows: Rock segment dataset Including uniaxial compressive strength Rock mass integrity coefficient The proportion of cumulative core length Joint spacing Joint quality score .

[0064] Quality index values ​​of core samples obtained from geological boreholes, calculated as the percentage of total intact core samples with a length ≥100mm:

[0065] in, It represents the cumulative length percentage of the rock core, used for rapid evaluation of rock mass integrity; denoted as 'a', the length of the complete rock core segment, measured manually after drilling; 'L' represents the drilling footage, recorded from the borehole depth gauge reading.

[0066] The calculation process for the basic quality index BQ of the rock mass is as follows:

[0067]

[0068] in, It is a basic quality indicator of rock mass and a core indicator for rock mass classification; It is the saturated uniaxial compressive strength; =1MPa, its function is to perform normalization; It is the rock mass integrity coefficient, which quantifies the degree of rock mass fracture development; It is the longitudinal wave velocity of the rock mass, which is obtained by testing in a borehole on the slope using a sonic detector. It is the longitudinal wave velocity of the rock block. Standard samples are cut from the drill core and measured in the laboratory using a rock acoustic wave tester.

[0069] Geological logging quantifies joint characteristics, joint spacing is measured every 20m along the survey line, survey lines are laid out along the slope dip, avoiding unloading fracture zones, and infill types are described according to ISRM standards, which are divided into no infill, calcite and clay infill.

[0070] Joint spacing The calculation process is as follows:

[0071] in, This is the measured value of the joint spacing of the a-th joint; It represents the total number of joints within the interval.

[0072] Joint quality score Based on the following:

[0073] Soil Section Dataset The structure is as follows: Soil Section Dataset Including cohesion internal friction angle On-site measured dry density pore water pressure , hydraulic gradient i.

[0074] Pore ​​water pressure The calculation process is as follows:

[0075] in, This is the water specific weight, corrected for water temperature. = ; It is the depth of the groundwater level, which is monitored in real time by a piezometer.

[0076] Piezometers were installed at the toe of the slope to record pore water pressure during the rainy and dry seasons. The hydraulic gradient i is derived as follows:

[0077] Where i is the hydraulic gradient, used to calculate slope instability driven by seepage force under heavy rainfall conditions; It is the change in pore water pressure during the dry and rainy seasons; The specific gravity of water is taken as 9.8 kN / m³. 3 Corrected according to local water temperature; It is the seepage path length, taken as the horizontal projection distance between piezometers.

[0078] The formula for permeability coefficient inversion is as follows:

[0079] in, Q is the permeability coefficient, which quantifies the soil's water permeability; the higher the value, the stronger the permeability. Q is the seepage flow rate, which is measured in real time at the seepage outlet by a flow meter. It is the cross-sectional area of ​​the water passage, which is calculated geometrically from the cross-sectional area of ​​the borehole or the cross-sectional area where the piezometer is installed. It is the unsteady flow correction term; T is the time constant; t is the time, recorded by the timer after the start of the experiment.

[0080] Construction data was extracted, compaction test reports of the embankment section were retrieved, and the road load rating was verified to meet the transportation requirements of the wind turbine equipment.

[0081] The formula for determining the warning coefficient is as follows:

[0082] in, It is the actual dry density measured on site; It is the maximum dry density of the compaction test; if it does not meet the standard, the warning coefficient α=0.85 is marked in the dataset; if it meets the standard, the warning coefficient α=1.

[0083] Output rock segment dataset ,include , , , , Soil Section Dataset ,include , , , , i.

[0084] This step involves geological drilling and logging to obtain core quality indicators, joint characteristics, groundwater conditions, and soil compaction. Joint parameters are used for subsequent joint condition scoring calculations, groundwater parameters support groundwater scoring and slope height effect correction under heavy rainfall conditions, and compaction and load data are used to verify the anti-sliding safety factor (which is reduced when compaction is insufficient). This provides scenario-based geological basis and construction quality benchmarks for subsequent stability scoring algorithms, ensuring the accuracy of slope analysis under wind turbine transportation conditions.

[0085] Step S3: Based on road design specifications, convert the rock section dataset, soil section dataset, and maintenance road dataset into standardized scores to obtain a parameter standardized score dataset; analyze the parameter standardized score dataset using principal component analysis to calculate the conditional score; and calculate the anti-skid stability safety factor based on the road load level in the maintenance road dataset.

[0086] The input sample is defined as follows: Rock section parameter set Soil Section Parameter Set Road maintenance dataset This constitutes the observation sample matrix X.

[0087] The process of constructing the observation sample matrix X is as follows:

[0088] Where X is the observation sample matrix, which integrates data and provides structured input for standardized scoring; n is the number of monitoring points, determined by the survey cross section in step S1; and m is the total parameter dimension (rock section + soil section + road data). It is a parameter vector of the rock segment; It is the soil section parameter vector; It is a road dataset, derived from the load level and roughness of S1.

[0089] The following are the correction items for the wind power scenario: When the scenario is high-altitude wind power, the following modifications will be enabled to adapt to the scenario.

[0090] The original pore water pressure parameters Revised to (Heavy rain condition correction), the correction process is as follows:

[0091] in, It is a correction of pore water pressure parameters, which is used to improve the estimated pore water pressure during heavy rain. It is a pore water pressure parameter; This is the cumulative rainfall (mm) over the past 7 days; It's a conversion factor, obtained from the data. .

[0092] Original rock mass strength parameters Revised to (Freeze-thaw cycle correction), the correction process is as follows:

[0093]

[0094] in, It is a correction of rock mass strength parameters, which reduces the rock mass strength in the freeze-thaw zone; It is the rock mass strength parameter, namely the uniaxial compressive strength of the rock mass obtained by S1; It is the lithological influence coefficient; Number of freeze-thaw cycles that year , from meteorological records.

[0095] The mapping rules for the standardized scoring sheet are as follows: Based on road design specifications, a scoring function is defined.

[0096] Define scoring function The formula is as follows:

[0097] Based on road design specifications, the rock type is clearly defined. , , , , ), soil ( , , , , i), road ( , Establish the correspondence between parameter values ​​and the "performance grading threshold" (e.g., corresponding to excellent and good) and the grades of excellent, good, medium and poor.

[0098] Distinguish the positive effect of parameters on slope stability (the larger the value, the more stable, such as...). , ) and negative effects (the smaller the value, the more stable, such as , The scoring logic is unified through reverse mapping (negative parameters are converted according to "the smaller the value → the higher the score", ensuring that "the higher the score, the better the stability" for all scores).

[0099] For parameters with critical abrupt changes (such as a sharp decrease in stability increase after rock strength exceeds 60 MPa, or a sharp increase in instability risk after soil slope i exceeds 25°), a stepped scoring method is adopted (the score changes linearly within the threshold, and the score is capped or jumps outside the threshold), which conforms to the nonlinear law of engineering failure.

[0100] Output parameters: Standardized scoring dataset.

[0101] The principal component analysis process is as follows: Because the parameter-standardized scoring dataset has multicollinearity, PCA is used to eliminate redundancy and extract principal components to improve scoring stability.

[0102] The data standardization formula is as follows:

[0103]

[0104]

[0105] in, It is a standardized data matrix, where rows represent samples and columns represent parameters; It is the standardized value of the j-th parameter in the i-th sample; n is the number of samples; These are the parameter values ​​after standardization and scoring; It is the sample mean of the j-th parameter; It is the sample standard deviation of the j-th parameter. Its function is to eliminate dimensional differences and provide comparable data for PCA.

[0106] The formula for calculating the covariance matrix is ​​as follows:

[0107] in, It is a standardized data matrix; yes The transpose of the matrix is ​​used to quantify the correlation between parameters and support eigenvalue decomposition.

[0108] The eigenvalue decomposition process is as follows:

[0109] in, These are eigenvalues, and the eigenvalues ​​are sorted as follows: ≥ ≥⋯≥ ≥0; This is a unit feature vector. Its purpose is to extract the main direction of change in the data.

[0110] The principal component extraction process is as follows: We select the smallest k that satisfies the cumulative contribution rate to calculate the principal component scores of k samples.

[0111] The process of selecting the minimum k is as follows:

[0112] in, These are the eigenvalues ​​arranged in descending order. Their function is to determine the number of principal components, k, to retain.

[0113] The principal component score of the i-th sample is calculated as follows:

[0114] Among them, principal component scores These are standardized values, used to project the samples onto the principal component space; It is the standardized vector of the i-th sample; It is the eigenvector of the l-th principal component.

[0115] Conditional scoring The synthesis process is as follows:

[0116]

[0117] in, These are the principal component weights, and their function is to allocate weights according to variance contribution. It is the eigenvalue of the l-th principal component; It is a conditional score (dimensionless, ranging from 0 to 100), and its function is to generate a comprehensive stability score.

[0118] Heavy load impact factor The definition is as follows:

[0119] in, It is a heavy load influencing factor, and its function is to quantify the risk of wind turbines being transported under heavy load. This refers to the maximum single-axle load of the wind turbine transport vehicle, extracted from the S1 maintenance road dataset. When the vehicle axle load exceeds the limit... 1.2 The safety level has been significantly improved, reflecting the high risk of transporting wind turbine blades.

[0120] The road load class L is defined as follows:

[0121] Wherein, L is the road load level, which is used to convert the road bearing capacity into a graded parameter; It is the maximum permissible single-axle load in the road design, obtained from the road design documents extracted from the S1 construction data.

[0122] Security Level The calculation process is as follows:

[0123]

[0124] in, It represents the safety level, which is used to quantify the static risk level of the slope by comprehensively considering load and terrain factors; L represents the road load level; and d represents the horizontal distance from the toe of the slope to the centerline of the road. It is the height difference between the foot of the slope and the road; It is a heavy load influencing factor; It is the angle of inclination of the line connecting the road and the toe of the slope.

[0125] Rock mass grade coefficient The calculation process is as follows:

[0126] in, It is the rock mass grade coefficient, which serves as an S-shaped function to correlate rock mass quality with safety factor; It is the basic quality index of the rock mass, and the BQ value is extracted from the S1 geological exploration data; 500 is the critical value for BQ classification, which comes from industry standards.

[0127] Standardized benchmark function The calculation process is as follows:

[0128]

[0129] in, It is a standard benchmark function, whose function is to calculate the initial safety factor by combining the rock mass quality and safety level; It is the minimum safety factor benchmark.

[0130] The freeze-thaw correction process is as follows (specific to high-altitude wind power, altitude ≥2000m):

[0131]

[0132] in, It is a freeze-thaw condition correction, which uses a two-stage correction mechanism to reflect the physical process of "freeze-thaw deteriorated rock mass → decreased stability"; 0.02 is derived from the regression of wind farm freeze-thaw test data; The number of freeze-thaw cycles in a year, expressed in cycles per year, is obtained from freeze-thaw data obtained from the local meteorological station.

[0133] The process for obtaining the anti-skid stability safety factor is as follows:

[0134]

[0135] in, It is the final calculated anti-slip stability safety factor; α is the early warning coefficient, which is obtained by judging the compaction degree of S2.

[0136] This step involves designing relevant scoring tables to convert rock section parameters, soil section parameters, and maintenance road datasets into standardized scores, forming a standardized parameter score dataset. Principal component analysis is then used to perform dimensionality reduction analysis on the standardized parameter score dataset, generating condition scores that comprehensively reflect the slope's foundation conditions, providing standardized indicators for subsequent risk quantification. Simultaneously, based on the road load level in the maintenance road dataset, combined with the relative position of the slope and road and load characteristics, the anti-sliding stability safety factor is calculated, quantifying the impact of load and topographic factors on slope stability. The condition scores and the anti-sliding stability safety factor serve as core parameters input to subsequent steps, supporting the coupled calculation of the first risk value and providing crucial quantitative evidence for dynamic risk assessment and safety control of mountain wind power maintenance road slopes.

[0137] Step S4: Obtain the slope height of the monitoring point of the mountain wind power maintenance road. Based on the slope height of the monitoring point and the road slope type, calculate the slope height amplification factor. Analyze and calculate the slope height amplification factor, condition score, and anti-slip stability safety factor using a multi-parameter coupled weighted algorithm to obtain the first risk value.

[0138] The process of conducting a survey on the slopes of the maintenance roads for mountainous wind power projects is as follows: Based on the road design drawings and preliminary geological survey reports, benchmark monitoring points are set up every 50m along the longitudinal direction of the road, and the monitoring points are densified in risk-sensitive sections such as the rock-soil boundary, dense joint zone, and historical deformation zone; the spatial locations of the top of the slope (monitoring point) and the bottom of the slope (or road surface) are marked simultaneously to form a "top of slope - bottom of slope" correspondence.

[0139] The RTK real-time dynamic positioning system (centimeter-level accuracy) is used to collect the three-dimensional coordinates of the monitoring points (slope top) on site and extract the elevation values; for the slope toe (or road surface), the design elevation is retrieved first from the as-built drawings or survey report; if the drawings are missing, the typical locations of the slope toe are measured by RTK (≥3 locations are averaged) to ensure the uniformity of the elevation benchmark.

[0140] Select monitoring points based on their elevation. Elevation of the toe of the slope (or road surface) Calculate the slope height of the monitoring point The process is as follows:

[0141] in, It is the elevation of the monitoring point; It is the elevation at the foot of the slope.

[0142] Slope height magnification factor The definition is as follows:

[0143] in, It is the slope height amplification factor, which quantifies the slope height effect; It is the slope height of the monitoring point; It is the rock slope height index, with values ​​assigned according to rock mass type: 0.35 for granite and 0.45 for shale. It is a type of road slope; It is a conditional scoring; It is the conditional score attenuation index, which is determined according to the rock mass integrity. For BQ>550, it is 0.8. This is the soil sensitivity coefficient, with a default value of 0.05. The slope height is determined by the power of the soil slope, calibrated through direct shear tests: 1.3 for clay and 1.2 for sand.

[0144] First risk value The calculation process is as follows:

[0145]

[0146]

[0147] in, It is the first risk value, and its function is to comprehensively quantify risk, with a limit of 10 on the upper limit of the risk value. It is the slope height amplification factor; It is a conditional scoring; It is the anti-slip stability safety factor. ; It is the rainfall-landslide sensitivity coefficient. = This data is derived from historical data. This is the cumulative rainfall over the past 7 days; This is a rainy season detection switch function; it is used when the rainy season is in effect. =1, dry season =0. The higher the value, the higher the risk, especially during periods of heavy rain. When the thickness exceeds 100mm, the risk increases exponentially, which is consistent with the characteristics of landslide disasters.

[0148] This step involves obtaining slope height data at various monitoring points through on-site surveys of the slopes along the wind power maintenance road in mountainous areas. Based on this slope height and slope type, a slope height amplification factor is calculated. This factor, along with the conditional score generated in step S3 and the anti-sliding stability safety factor, is then integrated to comprehensively calculate the first risk value, characterizing the slope instability risk. This first risk value serves as a core quantitative indicator and is directly input into subsequent step S7, providing precise support for dynamic risk early warning and graded response for wind power maintenance roads. Through differentiated slope height effects and parameter coupling calculations, the accuracy of risk assessment for steep slopes under complex conditions such as heavy wind turbine transport and torrential rain infiltration is significantly improved.

[0149] Step S5: Take aerial photos of the slopes of the mountain wind power maintenance road using a drone to obtain an aerial image dataset; perform corner detection on the aerial image dataset using the Harris algorithm to obtain a high-confidence corner set; perform corner matching on the high-confidence corner set using the normalized cross-correlation corner matching algorithm to obtain corner matching structured parameters.

[0150] A multi-rotor drone equipped with a high-resolution optical camera was used to collect aerial images of the slopes along the maintenance road for wind power projects in mountainous areas. Before flight, the flight path was planned based on the slope topography data, with a flight altitude of 50-100 meters, a lateral overlap of ≥80%, and a longitudinal overlap of ≥70%, ensuring sufficient overlap between adjacent images to meet subsequent stitching requirements. During flight, the drone's obstacle avoidance function was activated to avoid obstacles such as mountain protrusions and trees. At the same time, camera parameters were adjusted in real time according to lighting conditions to avoid loss of image details due to backlighting or overexposure. Key areas of the slope were photographed in detail, with 3-5 sets of oblique photography added to obtain three-dimensional geometric information. After aerial photography, the images were initially screened to remove invalid data that was blurry, misaligned, or abnormally exposed. Finally, an aerial image dataset covering the entire slope area and including surface texture and morphological features was formed, providing basic data support for subsequent slope surface corner detection and deformation analysis.

[0151] The aerial image dataset acquired by the drone is input into the Harris corner detection algorithm.

[0152] The original Harris algorithm calculation process is as follows:

[0153] in, It is the original Harris algorithm; and These are the eigenvalues ​​of the image autocorrelation matrix; k is an empirical constant. .

[0154] The improved Harris formula calculation process is as follows:

[0155] in, It is an improved Harris formula, i.e., k=0, which eliminates negative interference, makes the background region response value strictly 0, and improves the response value at the fracture endpoint. Times (when k=0.05) ≈1.11).

[0156] The adaptive thresholds for wind power are as follows: Wind power adaptive thresholds include terrain adaptive thresholds. and gradient strength threshold .

[0157] Terrain Adaptive Threshold The calculation process is as follows:

[0158]

[0159] in, It is a dynamic threshold, and its function is to serve as a terrain-adaptive threshold; It is the mean of the global response value of the image, which dynamically adapts to complex terrain and reflects the overall terrain complexity; It is the standard deviation of the response value, which accurately separates real corner points from noise and identifies local abrupt changes. It is the wind power terrain complexity factor, when it is in a flat area. Increase the suppression of false corner points such as turf texture, especially when in rugged rock formations. Reduce the detection rate of fracture corner points in reinforced rock masses; It is the slope roughness index, i.e., the standard deviation of LiDAR point cloud elevation; It is a reference roughness. =1m. Its purpose is to solve the terrain interference problem unique to wind power scenarios.

[0160] Gradient strength threshold The calculation process is as follows:

[0161] in, It is the gradient intensity threshold, which quantifies the local texture intensity and guides the selection of window size; and These are the gradients in the x and y directions.

[0162] The corner preprocessing process is as follows:

[0163]

[0164] in, This is to input the high-confidence corner set of the improved normalized cross-correlation corner matching algorithm NCC. Its purpose is to filter high-confidence corners and retain only corners with response values ​​exceeding 0.6τ, so as to improve the matching efficiency of NCC and prevent false matches. It is the original corner dataset {( , , )}; It is the normalized corner response value; It is the terrain-adaptive threshold.

[0165] Window size The formula is as follows:

[0166] in, It's the window size, and the gradient is weak. At that time, 5×5 window, strong gradient At that time, the expanded window captures large cracks. .

[0167] The normalized cross-correlation corner matching algorithm (NCC) coefficients are defined as follows:

[0168] in, This is the NCC correlation coefficient, which quantifies the similarity of image patches; window size. Dynamic expansion based on local gradient intensity (automatic enlargement window for rock mass fracture zone); It is the average grayscale value of the template area; It is the average gray value of the target area; It is the standard deviation of the template region; It is the standard deviation of the target region.

[0169] The joint verification process for the two responses is as follows:

[0170]

[0171] in, It is a Susan constraint; low values ​​exclude pseudo-corner points in uniform regions, such as grass textures. It is the maximum Susan response value of the image. =1; The Harris constraint ensures that the target point has a significant corner response. Automatic adjustment after heavy rain; joint constraint filtering for local distortion matching; It is the normalized corner response value; It is the NCC coefficient; It represents the number of similar grayscale pixels; This is the total number of neighboring pixels, N=37. The purpose of this formula is to ensure matching reliability using triple verification.

[0172] Traditional Harris algorithms suffer from low contrast in normalized images, poor differentiation between corners and background, difficulty in threshold selection, and a high risk of false detections. Susan algorithms are sensitive to noise and have complex threshold settings. The improved Harris algorithm enhances the contrast of the normalized image by modifying the empirical constant k, making corners more prominent and significantly narrowing the threshold selection range for more accurate localization. Furthermore, combining it with the Susan algorithm forms a joint operator that retains Harris's strong noise resistance while leveraging Susan's powerful ability to extract prominent corners, reducing false corners. This improves both detection accuracy and efficiency, resulting in superior robustness.

[0173] Output the set of matching point pairs that pass through. .

[0174] The RANSAC fine matching process is as follows: by The point set is used as the input to RANSAC.

[0175] The process of solving for the affine matrix is ​​as follows:

[0176] in, It is a 2×2 affine transformation matrix; It is a 2×1 translation vector; , () are the coordinates of a corner point in the reference image; , ( ) represents the coordinates of the matched corner point in the target image. This formula is used to solve for the affine transformation model between two images, describing the overall deformation of the slope.

[0177] The displacement vector calculation process is as follows:

[0178]

[0179] This formula is used to calculate the displacement of each corner point from the reference image to the target image.

[0180] The formula for calculating quality weights is as follows:

[0181] in, These are quality weights, which are used to assign weights to each displacement vector for subsequent weighted fusion (step S6). The weights are determined by both corner significance and matching quality. It is the normalized Harris response value of the corner point; It is the NCC correlation coefficient of that point pair.

[0182] Output and , The definition is as follows:

[0183] in, These are the corner matching structured parameters, which are a set including the affine transformation matrix, the set of displacement vectors, and the mass weights. It represents the number of points within the RANSAC.

[0184] This step first generates a high-confidence corner point set using an improved Harris algorithm, and then calculates the NCC correlation coefficient using an adaptive template. Next, reliable matching point pairs are screened using dual-response constraints (integrating Susan structural verification, Harris response evaluation, and a joint quality-confidence criterion). Finally, RANSAC fine matching outputs three core elements: affine transformation parameters, displacement vectors, and quality weights. These structured parameters (slope deformation matrix, pixel displacement, and weighted fusion coefficients) are directly input into step S6 to support the dual-metric fusion calculation of slope surface similarity, providing a high-precision displacement field analysis foundation for deformation monitoring and stability early warning of wind power maintenance roads.

[0185] Step S6: Based on the corner matching structured parameters, the surface similarity of the slope of the mountain wind power maintenance road is calculated using a similarity characterization algorithm to obtain the surface similarity; based on the surface similarity, the second risk value is calculated.

[0186] The similarity representation algorithm based on corner images is calculated as follows: Compared to traditional similarity representation algorithms, which often rely on global features (such as color histograms) and are prone to misjudgment due to the loss of local details, this corner-based image similarity representation algorithm significantly reduces the false matching rate. Traditional algorithms rely heavily on global features (such as color histograms), making them susceptible to misjudgment due to the loss of local details and sensitive to changes in rotation and lighting. This new algorithm extracts key local features through improved Harris corner detection, combining coarse NCC matching with unidirectional thresholding and fine RANSAC matching. Furthermore, it integrates the proportion of fine matching with Euclidean distance weighted similarity calculations, balancing the number of matching logs and feature distance. This results in stronger robustness to translation and rotation transformations, improved efficiency by reducing response function computations, and results that better reflect human subjective judgment.

[0187] The set obtained by inputting S5 .

[0188] Structural similarity The calculation process is as follows (deformation stability assessment):

[0189] in, It is structural similarity, and its function is to quantify the stability of structural deformation. It represents the number of points within the RANSAC, derived from the RANSAC output of S5. It is the sum of the weights of the effective matching points. Quality weights from S5; It is the determinant of the affine matrix; H comes from the affine matrix of S5; It is the wind power rock mass deformation sensitivity coefficient. =0.3; It is the scaling sensitivity factor. =0.5; It is the Frobenius norm, used to capture overall deformation; It is the Frobenius norm of the affine matrix and the identity matrix.

[0190] Displacement similarity The calculation process is as follows (position stability assessment):

[0191]

[0192] in, It is displacement similarity, and its function is to quantify positional stability; This represents the number of valid matches. It is the cumulative rainfall over seven days, and the infiltration amplification factor of the rainstorm; It is the impact coefficient of rainstorm. ; It is the magnitude of the single-point displacement vector; The displacement vector from S5.

[0193] The formula for the dual-metric weighted fusion algorithm is as follows:

[0194] in, It is surface similarity; under normal circumstances, = =0.5, equilibrium deformation and displacement; during heavy rain, i.e. >100, =0.3, =0.7, focusing on the risk of seepage displacement; during freeze-thaw cycles, i.e., the average temperature over seven consecutive days. <0 , = , =1- , It focuses on the structural deterioration caused by frost heave.

[0195] The calculation process for the second risk value is as follows:

[0196] in, It is the dimension conversion coefficient, k=100; the second risk value. It is negatively correlated with surface similarity.

[0197] This step, based on the affine transformation parameters, displacement vectors, and mass weight dataset output from step S5, calculates the slope surface similarity using a dual-metric weighted fusion algorithm. Structural similarity measures the overall deformation degree of the slope, while displacement similarity characterizes the positional shift. These two metrics are combined with dynamic weights based on the working conditions to generate a comprehensive surface similarity value. Finally, based on the surface similarity principle, this value is converted into a second risk value R2, establishing a mapping relationship between image deformation features and risk parameters, laying the foundation for time-lapse analysis in subsequent comprehensive risk assessment.

[0198] Step S7: Calculate the first risk value and the second risk value using a Bayesian weighted fusion algorithm to obtain a comprehensive risk value. Based on the comprehensive risk value, calculate the slope stability of the mountain wind power maintenance road.

[0199] The first risk value obtained by inputting S4 The second risk value obtained from S6 .

[0200] The calculation process for the comprehensive risk value is as follows (using the Bayesian weighted fusion algorithm):

[0201]

[0202] in, It is a comprehensive risk value, which linearly integrates geological and deformation risks as well as extreme rainfall risk assessments. When the thickness is >150mm and the fusion value is >5, the maximum risk is forced to be 10. It is a dynamic factor, under normal conditions Balancing geological and deformation risks; It is the first risk value; It is the second risk level; it is in the midst of a rainstorm. When >100, 3. Focus on the risk of deformation over time; during the freeze-thaw cycle, i.e., the average temperature over seven consecutive days. <0 hour, 7. Focus on the risk of structural deterioration.

[0203] Traditional risk assessment methods, such as fault tree analysis, suffer from single-node states, making it difficult to handle polymorphic risks and complex dependencies, and they cannot be dynamically updated. The analytic hierarchy process (AHP) is highly subjective with fixed weights. In contrast, this algorithm, through weighted fusion of multi-source information (expert experience, monitoring data, and historical cases), can handle uncertainty; it supports node polymorphism correction, aligning with engineering realities; it can dynamically update probabilities to adapt to dynamic changes during construction; it accurately identifies key risk factors through conditional probability tables and posterior inference, and verifies causal relationships using sensitivity analysis; and it integrates fault tree logic with fuzzy evaluation, balancing qualitative and quantitative approaches, significantly improving the scientific rigor and accuracy of risk assessment.

[0204] Stability coefficient The formula is as follows:

[0205] in, It is a stability coefficient that maps risk values ​​to engineering stability indicators; when When, it is in a stable state; when At that time, it is in a basically stable state; when At that time, it was in an unstable state.

[0206] This step integrates the first risk value from step S4 and the second risk value from step S6, and generates a comprehensive risk value through Bayesian fusion using adaptive weighting based on operating conditions. Based on a simplified three-level stability assessment, and combined with special judgment rules for heavy rain conditions (… >150 and >5 hours (mandatory high risk) to enhance extreme climate response capabilities. Final output: continuity stability coefficient. Achieve the engineering transformation from risk value to stable state, forming a closed loop for wind farm slope risk assessment.

[0207] This invention addresses the assessment bias issues in the risk assessment of slopes along maintenance roads for mountainous wind power projects, caused by difficulties in integrating heterogeneous parameters, delayed operational condition response, and lack of quantified risk stratification. It constructs an assessment system that begins with parameter standardization, uses dynamic operational condition coupling as a pivot, and integrates layered risks as the ultimate goal. Data sets of rock and soil mechanical parameters and road characteristics are obtained through field tests and geological surveys. Parameter normalization is achieved by combining known road design specifications with a design scoring table. A comprehensive condition score is extracted through principal component analysis. The first risk value is calculated by integrating the slope height amplification effect and load characteristics, while a second risk value is generated by coupling surface deformation characteristics from UAV aerial photography. Freeze-thaw and rainy season correction mechanisms are introduced to achieve dynamic weighting of the two risks. Finally, a comprehensive risk value is output through a layered risk coupling algorithm, accurately identifying high-risk slope sections and providing a quantitative decision-making basis for the safety control of maintenance roads.

[0208] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0209] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

Claims

1. A method for analyzing the slope stability of a maintenance road for mountain wind power projects, characterized in that, Includes the following steps: Step S1: Obtain the maintenance road dataset and road slope type of the mountain wind power maintenance road, collect slope rock cores on site, and conduct uniaxial compressive strength test on the rock section of the slope rock core based on the road slope type to obtain the rock section parameters, and conduct shear strength test on the soil section of the slope rock core to obtain the soil section parameters. Step S2: Based on the rock segment parameters and soil segment parameters, construct corresponding rock segment datasets and soil segment datasets; Step S3: Based on road design specifications, the rock section dataset, soil section dataset, and maintenance road dataset are converted into standardized scores to obtain a parameter standardized score dataset; the parameter standardized score dataset is analyzed using principal component analysis to calculate the conditional score; the anti-skid stability safety factor is calculated based on the road load level in the maintenance road dataset. Step S4: Obtain the slope elevation of the monitoring points on the mountain wind power maintenance road; Based on the slope height and road slope type of the monitoring points, the slope height amplification factor is calculated; the slope height amplification factor, condition score and anti-skid stability safety factor are analyzed and calculated using a multi-parameter coupled weighted algorithm to obtain the first risk value; Step S5: Take aerial photos of the slopes of the mountain wind power maintenance road using a drone to obtain an aerial image dataset; use the Harris algorithm to perform corner detection on the aerial image dataset to obtain a high-confidence corner set; Corner matching structured parameters are obtained by performing corner matching on the high-confidence corner set using a normalized cross-correlation corner matching algorithm. Step S6: Based on the corner matching structured parameters, the surface similarity of the slope of the mountain wind power maintenance road is calculated using a similarity characterization algorithm to obtain the surface similarity. Based on the surface similarity, a second risk value is calculated; Step S7: Calculate the first risk value and the second risk value using a Bayesian weighted fusion algorithm to obtain a comprehensive risk value. Based on the comprehensive risk value, calculate the slope stability of the mountain wind power maintenance road.

2. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 1, characterized in that, Obtain the road slope types and maintenance road dataset for mountain wind power maintenance roads, including: A surveying section was set up every 50 meters along the longitudinal direction of the maintenance road. Differential GPS was used to accurately locate the coordinates of each section. Slope types were classified through geological hammer testing and overburden thickness detection. The classification rules are as follows: When the exposed bedrock thickness accounts for ≥70%, it is defined as a rocky section; when the fill area or the overburden thickness is >1.5 meters, it is defined as a soil section. Road maintenance dataset The construction process is as follows: Road maintenance dataset Including wind turbine transport load Slope roughness index ; The calculation process for the transport load of the wind turbine is as follows: ； in, It is the wind turbine transport load, used to verify whether the road needs to be reinforced; This is the weight of the wind turbine blades, taken as 80 tons; It refers to the number of axles on the transport vehicle; This refers to the single-axle load limit for road design. Slope roughness index The calculation process is as follows: Use a drone equipped with LiDAR to scan the slope and obtain the elevation value of point p in the point cloud. ; ； in, It is the slope roughness index, used to quantify surface deformation and associated with the second risk value; It is the elevation value of point p in the point cloud, collected by a drone; It is the average elevation of a local window, calculated using a GIS sliding window. It represents the number of point clouds within the window.

3. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 2, characterized in that, Based on the road slope type, uniaxial compressive strength tests were conducted on the rock section of the slope core to obtain rock section parameters, and shear strength tests were conducted on the soil section of the slope core to obtain soil section parameters, including: The test procedure for geotechnical mechanical parameters is as follows: Processing rock cores into A standard cylindrical specimen measuring 50mm × 100mm was subjected to a 2000kN compression testing machine and loaded at a rate of 0.5MPa / s until failure. The peak load P was recorded, and the saturated uniaxial compressive strength was calculated using the formula. : ； in, It is the destructive load; A is the cross-sectional area of ​​the specimen, calculated after testing perpendicular to the bedding direction in layered rock mass. Its function is to evaluate rock strength and is used in BQ rock mass classification; The shear strength test of the soil section was conducted using a direct shear apparatus for consolidated rapid shear tests. Four levels of normal stress were applied at a shear rate of 0.8 mm / min. Shear stress-displacement curves were plotted, and equations were established based on the Mohr-Coulomb criterion. ； in, To break shear stress; is the normal stress, applied via a lever-weight system; c is the cohesive force obtained from a direct shear test. The internal friction angle is obtained from the direct shear test. After the test, 85% of the peak strength is taken as the design parameter to obtain the soil strength parameters. The safety factor is calculated by inputting the limit equilibrium method.

4. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 3, characterized in that, Step S2 includes: Rock segment dataset Including uniaxial compressive strength Rock mass integrity coefficient The proportion of cumulative core length Joint spacing Joint quality score ; Soil Section Dataset Including cohesion internal friction angle On-site measured dry density pore water pressure , hydraulic gradient i.

5. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 4, characterized in that, By analyzing the standardized score dataset of the parameters using principal component analysis, conditional scores are calculated, including: The principal component extraction process is as follows: Select the smallest k that satisfies the cumulative contribution rate to calculate the principal component scores of k samples; The process of selecting the minimum k is as follows: ； in, These are the eigenvalues ​​arranged in descending order, and their function is to determine the number of principal components k to retain. The principal component score of the i-th sample is calculated as follows: ； Among them, principal component scores These are standardized values, used to project the samples onto the principal component space; It is the standardized vector of the i-th sample; It is the eigenvector of the l-th principal component; Conditional scoring The synthesis process is as follows: ； ； in, These are the principal component weights, and their function is to allocate weights according to variance contribution. It is the eigenvalue of the l-th principal component; It is a conditional score, and its function is to generate a comprehensive stability score.

6. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 5, characterized in that, The slope height of monitoring points along mountain wind power maintenance roads is obtained. Based on the slope height of these monitoring points and the road slope type, a slope height amplification factor is calculated, including: An RTK real-time dynamic positioning system is used to collect the three-dimensional coordinates of monitoring points on site and extract elevation values. Select monitoring points based on their elevation. Elevation of the foot of the slope Calculate the slope height of the monitoring point The process is as follows: ； in, It is the elevation of the monitoring point; It is the elevation of the foot of the slope; Slope height magnification factor The definition is as follows: ； in, It is the slope height amplification factor, which quantifies the slope height effect; It is the slope height of the monitoring point; It is the rock slope height index, with values ​​assigned according to rock mass type: 0.35 for granite and 0.45 for shale. It is a type of road slope; It is a conditional scoring; It is the conditional score attenuation index, which is determined according to the rock mass integrity. For BQ>550, it is 0.

8. This is the soil sensitivity coefficient, with a default value of 0.

05. The slope height is determined by the power of the soil slope, calibrated through direct shear tests: 1.3 for clay and 1.2 for sand.

7. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 6, characterized in that, By using a multi-parameter coupled weighted algorithm to analyze and calculate the slope height amplification factor, condition score, and anti-slide stability safety factor, a first risk value is obtained, including: First risk value The calculation process is as follows: ； ； ； in, It is the first risk value, and its function is to comprehensively quantify risk, with a limit of 10 on the upper limit of the risk value. It is the slope height amplification factor; It is a conditional scoring; It is the anti-slip stability safety factor. ; It is the rainfall-landslide sensitivity coefficient. = This data is derived from historical data. This is the cumulative rainfall over the past 7 days; This is a rainy season detection switch function; it is used when the rainy season is in effect. =1, dry season =0, The higher the value, the higher the risk, especially during periods of heavy rain. When the thickness exceeds 100mm, the risk increases exponentially, which is consistent with the characteristics of landslide disasters.

8. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 7, characterized in that, Obtain the corner matching structured parameters, including: The displacement vector calculation process is as follows: ； ； This formula is used to calculate the displacement of each corner point from the reference image to the target image; The formula for calculating quality weights is as follows: ； in, It is the quality weight, which is used to assign weights to each displacement vector for subsequent weighted fusion. The weights are determined by the significance of the corner points and the matching quality. It is the normalized Harris response value of the corner point; This is the NCC correlation coefficient for that point pair; The definition is as follows: ； in, These are the corner matching structured parameters, which are a set of affine transformation matrices, displacement vector sets, and mass weights.

9. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 8, characterized in that, Step S6 includes: Structural similarity The calculation process is as follows (deformation stability assessment): ； in, It is structural similarity, and its function is to quantify the stability of structural deformation. It is the number of points within RANSAC; It is the sum of the weights of the effective matching points. Quality weights from S5; It is the determinant of the affine matrix; H comes from the affine matrix of S5; It is the wind power rock mass deformation sensitivity coefficient. =0.3; It is the scaling sensitivity factor. =0.5; It is the Frobenius norm, used to capture overall deformation; It is the Frobenius norm of the affine matrix and the identity matrix; Displacement similarity The calculation process is as follows: ； ； in, It is displacement similarity, and its function is to quantify positional stability; This represents the number of valid matches. It is the cumulative rainfall over seven days, and the infiltration amplification factor of the rainstorm; It is the impact coefficient of rainstorm. ; It is the magnitude of the single-point displacement vector; The displacement vector from S5; The formula for the dual-metric weighted fusion algorithm is as follows: ； in, It is surface similarity; under normal circumstances, = =0.5, equilibrium deformation and displacement; during heavy rain, i.e. >100, =0.3, =0.7, focusing on the risk of seepage displacement; during freeze-thaw cycles, i.e., the average temperature over seven consecutive days. <0 , = , =1- , The focus is on the structural deterioration caused by frost heave; The calculation process for the second risk value is as follows: ； in, It is the dimension conversion coefficient, k=100; the second risk value. It is negatively correlated with surface similarity.

10. The method for analyzing the slope stability of a mountain wind power maintenance road according to claim 9, characterized in that, A comprehensive risk value is obtained by calculating the first and second risk values ​​using a Bayesian weighted fusion algorithm. Based on this comprehensive risk value, the slope stability of the mountain wind power maintenance road is calculated, including: The calculation process for the comprehensive risk value is as follows: ； ； in, It is a comprehensive risk value, which linearly integrates geological and deformation risks as well as extreme rainfall risk assessments. When the thickness is >150mm and the fusion value is >5, the maximum risk is forced to be 10. It is a dynamic factor, under normal conditions Balancing geological and deformation risks; It is the first risk value; It is the second risk level; it is in the midst of a rainstorm. When >100, 3. Focus on the risk of deformation over time; during the freeze-thaw cycle, i.e., the average temperature over seven consecutive days. <0 hour, 7. Focus on structural degradation risk; Traditional risk assessment methods, such as fault tree analysis, have single node states, making it difficult to handle polymorphic risks and complex dependencies, and they cannot be dynamically updated. The analytic hierarchy process (AHP) is highly subjective with fixed weights. In contrast, this algorithm can handle uncertainty by weighted fusion of multi-source information. It supports node polymorphism correction, which is more in line with engineering realities. It can dynamically update probabilities to adapt to dynamic changes in the construction process. It accurately identifies key risk factors through conditional probability tables and posterior inference, and verifies causal relationships by combining sensitivity analysis. It integrates fault tree logic and fuzzy evaluation, taking into account both qualitative and quantitative aspects, and significantly improves the scientificity and accuracy of risk assessment. Stability coefficient The formula is as follows: ； in, It is a stability coefficient that maps risk values ​​to engineering stability indicators; when When, it is in a stable state; when At that time, it is in a basically stable state; when At that time, it was in an unstable state.

Images