Earthquake landslide risk assessment method and system for traffic network in high intensity area

Abstract

This application provides a method and system for assessing earthquake-induced landslide risk in transportation networks in high-intensity seismic zones, relating to the field of landslide risk assessment technology. The method includes: screening key influencing factors of earthquake-induced landslides; optimizing the construction of positive and negative sample libraries; using a spatial weighted machine learning model to evaluate landslide susceptibility and generate a susceptibility zoning map; extracting parameters of potential landslide source areas based on the high-susceptibility zone identification results; using a parallel particle discrete element method to simulate the entire landslide movement process to obtain hazard parameters; inputting the hazard parameters into a transportation engineering structure vulnerability model to quantitatively calculate the failure probability; and fusing the susceptibility probability and structural vulnerability results to generate a risk zoning map based on road network units. The final results of this application are directly applicable to the operation and maintenance management of transportation infrastructure. The provided risk zoning map can be directly used to guide the optimal allocation of risk control resources, the formulation of emergency plans, and the seismic resilience design of engineering projects, demonstrating clear application value.

Description

Technical Field

[0001] This application relates to the field of landslide risk assessment technology, and in particular to a method and system for assessing earthquake-induced landslide risks in transportation networks in high-intensity seismic zones. Background Technology

[0002] High-intensity earthquake zones, characterized by basic intensity VII and above, typically have complex topographical and geological conditions, well-developed fault structures, and fractured rock masses. When a strong earthquake occurs, it can easily trigger large-scale landslides, collapses, and other secondary geological disasters. As linear engineering projects, transportation networks often traverse different geomorphic units and geological zones. Their long routes and wide distribution mean they face severe threats after an earthquake, including landslides burying roadbeds, impacting bridge piers, and destroying tunnel entrances, directly leading to traffic disruptions and hindered rescue efforts.

[0003] Currently, risk assessment for earthquake-induced landslides mainly falls into two categories. One is regional susceptibility assessment based on statistics or machine learning. This method can quickly predict large-scale disasters, but often ignores the physical processes of landslides, such as their path, velocity, accumulation range, and impact energy, making it difficult to directly assess their destructive intensity on specific engineering structures. The other is individual landslide simulation based on physical mechanics models. This method is highly accurate and realistically reproduces the process, but due to high computational costs, it is difficult to simulate a large number of potential landslides on a regional scale. Furthermore, existing assessment systems mostly stop at predicting the landslide disaster itself, rarely quantitatively coupling landslide dynamic parameters, such as impact kinetic energy, with the vulnerability of transportation engineering structures such as roadbeds, bridges, and tunnels. Assessment results often remain at the level of landslide susceptibility zoning, failing to directly output risk levels for road sections or bridge / tunnel units. This leads to a disconnect between assessment results and the actual needs of transportation operation and maintenance management, lacking targeted decision-making guidance value.

[0004] Therefore, how to construct a full-chain assessment system that integrates disaster identification, spatial prediction, physical simulation, and structural response to achieve accurate quantification from landslide susceptibility to road network risk is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This application provides a method and system for assessing earthquake landslide risk in transportation networks in high-intensity seismic zones, aiming to address several problems existing in the prior art. First, current earthquake landslide factor selection methods lack specificity. Traditional methods do not optimize factors based on the triggering characteristics of strong earthquakes in high-intensity zones and the linear engineering characteristics of transportation networks, resulting in insufficient applicability and accuracy of susceptibility assessment models in this specific scenario. Second, the sample library construction suffers from bias. Conventional non-disaster sample selection methods are too random, easily including potential landslide areas, leading to a decline in model training quality. Third, the assessment dimensions are singular and lack physical process support. Traditional machine learning susceptibility assessment can only answer where landslides may occur, but cannot answer where the landslide will go after it occurs or the magnitude of the impact force, making it difficult to support quantitative risk assessment of engineering structures. Finally, disaster assessment is disconnected from engineering assessment. Landslide hazard parameters, such as impact kinetic energy, are not effectively converted into the probability of damage or functional loss of transportation engineering structures such as roadbeds, bridges, and tunnels, making it difficult for assessment results to directly serve disaster prevention and mitigation decisions for transportation infrastructure. To address the aforementioned issues, this invention provides a method and system for assessing earthquake-induced landslide risks in high-intensity traffic networks. The aim is to construct a closed-loop technical system covering the entire chain of disaster-causing factor screening, disaster spatial prediction, physical process simulation, vulnerability assessment of disaster-bearing bodies, and road network risk zoning, thereby achieving precise quantification from disaster identification to engineering risk.

[0006] In a first aspect, this application provides a method for assessing the risk of earthquake-induced landslides in transportation networks in high-intensity seismic zones, the method comprising: A list of landslides in high-intensity transportation networks induced by historical earthquakes was obtained, and a disaster sample database was initially constructed by selecting multiple influencing factors. The information content of each influencing factor was calculated using an information content model, and a multi-index correlation analysis was conducted on the initially selected influencing factors in combination with the information content value, the proportion of disaster area, and the number of disaster points. Key factors suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity transportation networks were then selected. Based on the aforementioned key factors, the landslide disaster surface is converted into a grid cell, with the grid center point as the positive sample; a dynamic buffer zone is set outside the historical landslide boundary, and an equal number of negative samples are randomly selected in the extremely low information value region outside the buffer zone to construct an optimized sample library; wherein, the extremely low information value region is the region where the information value is lower than a set threshold. Based on the optimized sample library, a spatial weighted machine learning model considering the neighborhood characteristics of grid cells is constructed. The sample library is divided into training and testing sets according to a set ratio for model training and accuracy verification. Earthquake landslide hazard susceptibility assessment is carried out, and a landslide susceptibility zoning map is generated. The landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas. Based on the identification results of high-risk or extremely high-risk areas, the topographic and lithological parameters of potential landslide source areas are automatically extracted. A high-fidelity simulation model is constructed using the parallel particle discrete element method to simulate the entire process of landslide instability initiation, movement and deposition, and to obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy and deposition range. The aforementioned hazard parameters are input as loads into predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances in traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures are then quantitatively calculated. These quantitative vulnerability results include the probability of failure or the degree of functional loss of each structure under landslides of different intensities. By integrating the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of transportation engineering structures, a risk zoning map is generated with transportation network units as the assessment objects using a risk matrix or risk calculation formula. The map displays the characteristics of the disaster-causing body and the information of the disaster-bearing body.

[0007] As a preferred technical solution, the initially selected multiple influencing factors include elevation, slope, aspect, profile curvature, stratigraphic lithology, distance from road, distance from fault, distance from epicenter, and distance from water system. Through multi-index correlation analysis, key factor combinations suitable for traffic network scenarios in high-intensity seismic zones are selected.

[0008] As a preferred technical solution, the distance of the dynamic buffer is 80~120 meters, and the selection area of ​​negative samples is limited to the area outside the dynamic buffer and the information value is lower than a preset threshold.

[0009] As a preferred technical solution, the spatial weight machine learning model includes at least one of random forest, extreme gradient boosting, or support vector machine. The construction process of the spatial weight machine learning model includes: constructing a neighborhood weight matrix based on factor importance and Euclidean distance between sample points, introducing spatial weights into the machine learning framework for training and optimization, and calculating the probability of landslide disasters occurring at each point in the region.

[0010] As a preferred technical solution, the method for generating landslide susceptibility zoning maps includes: mapping the calculated probability of disaster occurrence to the original geographic spatial domain, and dividing it into five categories: extremely low susceptibility zone, low susceptibility zone, medium susceptibility zone, high susceptibility zone, and extremely high susceptibility zone using the natural breakpoint method. The identification results of the high susceptibility zone and the extremely high susceptibility zone serve as the basis for extracting potential landslide source areas.

[0011] As a preferred technical solution, the simulation of the entire physical process of landslide movement is based on a supercomputing platform to achieve parallel computing.

[0012] As a preferred technical solution, the vulnerability curve or model of the traffic engineering structure is pre-established based on the seismic, shock-resistant, and burial-resistant properties of the engineering structure itself.

[0013] As a preferred technical solution, the risk zoning map uses road sections, bridges or tunnels as basic assessment units. The risk level of each assessment unit in the map is calculated by coupling the susceptibility probability and structural vulnerability results through a risk matrix, and the specific disaster-causing body characteristics and disaster-bearing body information are displayed in association.

[0014] Secondly, this application provides a risk assessment system for earthquake-induced landslides in high-intensity seismic transport networks, the system comprising: The factor screening module is configured to obtain a list of landslides in high-intensity transportation networks induced by historical earthquakes, initially select multiple influencing factors to construct a disaster sample database, use an information content model to calculate the information content of each influencing factor, and combine the information content value, the proportion of disaster area and the number of disaster points to conduct multi-index correlation analysis on the initially selected multiple influencing factors, and screen out key factors suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity transportation networks. The sample library construction module is configured to convert landslide disaster surfaces into grid cells based on the key factors, with the grid center point as the positive sample; set a dynamic buffer outside the historical landslide boundary, and randomly select an equal number of negative samples in the extremely low information value region outside the buffer to construct an optimized sample library; wherein, the extremely low information value region is the region where the information value is lower than a set threshold. The susceptibility assessment module is configured to construct a spatial weighted machine learning model that considers the neighborhood features of grid cells based on the optimized sample library, divide the sample library into training and testing sets according to a set ratio for model training and accuracy verification, conduct earthquake and landslide susceptibility assessment, and generate a landslide susceptibility zoning map; wherein the landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas; The physical simulation module is configured to automatically extract topographic and lithological parameters of potential landslide source areas based on the identification results of high-risk or extremely high-risk areas, construct a high-fidelity simulation model using the parallel particle discrete unit method, simulate the entire process of landslide instability initiation and deposition, and obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy, and deposition range. The vulnerability assessment module is configured to input the hazard parameters as loads into predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances of traffic engineering structures, and quantitatively calculate the quantitative vulnerability results of the traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures include the probability of failure or the degree of functional loss of each structure under landslide action of different intensities. The risk zoning module is configured to integrate the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of transportation engineering structures. It uses a risk matrix or risk calculation formula to generate a risk zoning map with transportation network units as the assessment objects. The map displays the characteristics of the disaster-causing body and the information of the disaster-bearing body.

[0015] The earthquake-induced landslide risk assessment method and system for transportation networks in high-intensity seismic zones provided in this application have at least the following beneficial effects: 1) This application tightly integrates disaster-causing factor screening, disaster spatial prediction, physical process simulation, vulnerability assessment of disaster-bearing bodies, and road network risk zoning, thus establishing a seamless data and information flow from basic data to key factors, vulnerability zoning, physical simulation parameters, structural vulnerability, and road network risk zoning. This end-to-end design of disaster-causing factor identification, disaster spatiotemporal prediction, physical process simulation, and disaster-bearing body response assessment overcomes the shortcomings of existing technologies that separate vulnerability assessment from engineering risk assessment, forming a logically rigorous technical closed loop.

[0016] 2) This application improves the spatial distribution accuracy of regional landslide susceptibility assessment by employing a machine learning model that considers spatial neighborhood characteristics. Based on this, high-fidelity physical simulations are performed only in high-susceptibility areas using a supercomputing-based parallel particle discrete element method. This avoids the enormous computational burden of performing physical simulations across the entire region while simultaneously reconstructing key dynamic parameters such as landslide movement paths, velocities, and impact kinetic energy through the physical model. This coarse-screening plus fine-calculation approach significantly improves the physical realism and numerical accuracy of landslide hazard intensity assessment while ensuring efficiency across a wide range of assessments.

[0017] 3) This application achieves deep coupling between disaster load and structural response by quantitatively inputting parameters such as landslide impact kinetic energy obtained from physical simulation into the vulnerability model of transportation engineering structures. The final result is a risk zoning map with road sections, bridges, and tunnels as basic units. This map is no longer merely a geologically defined disaster zone, but directly correlates the risk level results of disaster-causing bodies such as high-impact-energy landslides and disaster-bearing bodies such as a roadbed section or a bridge pier. This allows transportation infrastructure operation and maintenance managers to intuitively identify high-risk road sections, thereby guiding the optimal allocation of risk control resources, the formulation of emergency plans, and the seismic resilience design of projects. It has clear application value and promising prospects for wider application.

[0018] 4) This application employs a dynamic buffer optimization negative sample sampling strategy, effectively eliminating ambiguous regions near landslide boundaries and selecting negative samples from areas with extremely low information content. This significantly improves the purity and discriminative power of the sample library, providing a reliable data foundation for high-precision training of machine learning models. Simultaneously, by combining an information content model with a factor selection method based on multi-indicator correlation analysis, the representativeness and independence of the factor set input to the model in specific scenarios are ensured, guaranteeing the reliability of the evaluation results from the outset. Attached Figure Description The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] Figure 1 A flowchart illustrating a method for assessing the risk of earthquake-induced landslides in transportation networks in high-intensity seismic zones, provided as an embodiment of this application; Figure 2 This is a structural diagram of the earthquake landslide risk assessment system for high-intensity traffic network provided in this application embodiment.

[0020] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0021] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0022] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0023] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0024] This application provides a method for assessing the risk of earthquake-induced landslides in transportation networks in high-intensity seismic zones. The method specifically includes the following steps S10-S60.

[0025] S10: Obtain a list of landslides in high-intensity transportation networks induced by historical earthquakes, and initially select multiple influencing factors to construct a disaster sample database; use an information content model to calculate the information content of each influencing factor, and combine the information content value, the proportion of disaster area and the number of disaster points to conduct multi-index correlation analysis on the initially selected multiple influencing factors, and screen out key factors suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity transportation networks.

[0026] Step S10 involves data preparation and factor selection. Specifically, a list of landslides induced by historical earthquakes in high-intensity areas (VII and above) along transportation networks is collected. Nine influencing factors are initially selected: elevation, slope, aspect, profile curvature, stratigraphic lithology, distance from roads, distance from faults, distance from epicenter, and distance from water systems to construct a disaster sample database. An information content model is used to calculate the information content of each influencing factor. Using the information content value, the proportion of disaster area (the proportion of disaster area relative to the total landslide area within the factor classification), and the number of disaster points, a multi-index correlation analysis is conducted on the nine initially selected factors to screen out the key factors most suitable for earthquake landslide susceptibility assessment in the "high-intensity area" and "transportation network" scenarios.

[0027] The multi-index correlation analysis process is as follows: First, statistical analysis was performed on the data of each evaluation factor layer using Origin software. A Pearson correlation coefficient matrix was constructed by calculating the standard deviation, variance, and covariance to quantitatively measure the correlation between factors. Highly redundant factors with an absolute correlation coefficient |R| ≥ 0.7 were removed. Second, to further ensure the independence of the factor system, multicollinearity was tested using tolerance (TOL) and variance inflation factor (VIF). Severely collinear factors with TOL < 0.1 and VIF > 10 were removed, finally completing the screening of key evaluation factors. Taking the landslide susceptibility assessment of the Hailuogou scenic area in the high-intensity earthquake zone of Luding as an example, the above analysis showed that the absolute values ​​of the correlation coefficients between factors were all less than 0.5, and VIF and TOL met the threshold requirements, indicating that there was no significant correlation among the initially selected 9 evaluation factors, which can be used for subsequent landslide susceptibility prediction in the model.

[0028] S20: Based on the key factors, the landslide disaster surface is converted into a grid cell, and the grid center point is used as a positive sample; a dynamic buffer is set outside the historical landslide boundary, and an equal number of negative samples are randomly selected in the extremely low information value area outside the buffer to construct an optimized sample library; wherein, the extremely low information value area is the area where the information value is lower than a set threshold.

[0029] Step S20 involves constructing and optimizing the sample library. Specifically, the landslide hazard surface is converted into a grid cell, and each hazard (positive) sample is represented by the center point of the grid cell. Drawing on the research idea of ​​optimizing negative sample sampling strategies, a dynamic buffer zone (e.g., 100 meters) is set outside the historical landslide boundary. An equal number of non-hazard (negative) samples are randomly selected from the area outside this buffer zone with extremely low information values ​​to improve the quality of the sample library. The constructed and optimized sample library is used as the input dataset for susceptibility assessment, and the training and test sets are divided in a 7:3 ratio.

[0030] S30: Based on the optimized sample library, construct a spatial weighted machine learning model that considers the neighborhood features of grid cells, divide the sample library into training and testing sets according to a set ratio for model training and accuracy verification, conduct earthquake landslide hazard susceptibility assessment, and generate a landslide susceptibility zoning map; wherein the landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas.

[0031] The purpose of step S30 is to construct a spatial weighted machine learning model and evaluate its susceptibility. Specifically, this embodiment constructs a spatial weighted machine learning (SWML) model that considers the neighborhood characteristics of grid cells. Improved machine learning models (including Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM)) are used to evaluate the susceptibility of earthquake-induced landslides in high-intensity seismic zones. The predictive accuracy of the machine learning model is evaluated by combining the receiver operating characteristic (ROC) curve, resulting in a model suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity seismic zones.

[0032] The SWML model construction method can be broken down into four parts: spatial neighborhood weight definition, weight matrix construction, modeling and spatial mapping. The specific implementation process is as follows: First, in order to reflect the spatial proximity effect and environmental similarity effect, spatial distance weight and feature similarity weight are constructed separately and then fused to obtain the comprehensive spatial weight. Among them, the spatial distance weight is calculated using the Gaussian kernel function, which reflects that the closer the distance, the stronger the correlation. The feature similarity weight is constructed using the exponential decay function, which reflects that the more similar the geographical environment, the more consistent the disaster development pattern. The comprehensive correlation between sample points is obtained by linear superposition.

[0033]

[0034]

[0035]

[0036] in( x i , y i ), ( x j , y j ) are units i With neighboring units j Geographic coordinates α This is the distance attenuation coefficient. f i , f j Units i and j eigenfactors, βThe sensitivity coefficient is used for the feature. Secondly, based on the factor importance identification results, the three dominant evaluation factors with the greatest impact on model prediction are selected. Then, with each grid cell as the center, a search is performed for its spatially nearest neighbors. k Each cell is treated as a local neighborhood, and a spatial weight matrix is ​​calculated using the aforementioned formula. This weight matrix is ​​then incorporated into a machine learning framework, with grid search used to optimize parameters, completing model training and optimization. The SWML model is then used to predict the landslide probability of the entire study area's grid. Finally, the model's output probability values ​​are mapped to the original geospatial data, and the natural breakpoint method is used for classification (extremely low susceptibility area, low susceptibility area, medium susceptibility area, high susceptibility area, extremely high susceptibility area), generating the final landslide susceptibility zoning map. This spatial representation of landslide susceptibility along transportation routes in high-intensity earthquake zones is achieved.

[0037] S40: Based on the identification results of high-risk or extremely high-risk areas, automatically extract the topographic and lithological parameters of potential landslide source areas, and use the parallel particle discrete element method to construct a high-fidelity simulation model to simulate the entire process of landslide instability initiation, movement and accumulation, and obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy and accumulation range.

[0038] Step S40 is the process of simulating the entire physical process of landslide movement based on supercomputing. Specifically, based on the high / extremely high susceptibility areas obtained in step S30, the topographic and lithological parameters of potential landslide source areas are automatically extracted. A high-fidelity simulation model is constructed using the parallel particle discrete element method (DEM). Based on supercomputing algorithms, the entire process of instability initiation, movement, and deposition of a large number of potential landslides is efficiently simulated, and key hazard parameters such as the movement path, velocity time history, impact kinetic energy, and deposition range of each landslide are obtained in batches.

[0039] The high-fidelity simulation model construction process is as follows: First, based on the high / extremely high susceptibility areas obtained in step S30, the geometric parameters such as slope, aspect, profile curvature, sliding surface dip angle, and sliding mass thickness of the landslide source area are automatically extracted using a high-precision digital elevation model. Physical and mechanical parameters such as density, internal friction angle, cohesion, elastic modulus, and Poisson's ratio of the soil and rock mass are assigned to each landslide source area. The initial water content and pore water pressure distribution of the landslide body are determined based on groundwater depth and rainfall infiltration conditions. Second, the discrete element method (DEM) is used. A simulation model was constructed, taking into account the type and structural characteristics of the soil and rock mass. The landslide body was discretized into spherical particles with multi-level particle size distribution. A contact constitutive model such as linear contact or parallel bonding was selected. A fixed bottom boundary and trailing edge initiation conditions were set, and reasonable damping and adaptive time step were configured. Finally, the model parameters were calibrated and verified using historical earthquake landslide measured data. The particle contact parameters were corrected through trial and error to ensure that the simulation results were in good agreement with the actual accumulation range and movement distance of no less than 90%, providing high-fidelity input for subsequent high-efficiency supercomputing simulations.

[0040] The efficient simulation implementation process based on supercomputing algorithms is as follows: First, a hierarchical parallel strategy is designed, employing a job queue scheduling system to achieve landslide task-level concurrency. For a single landslide, a hybrid parallel mode of MPI+OpenMP is used to accelerate particle domain decomposition and contact calculation, and dynamic load balancing is introduced. Second, algorithm optimization is implemented, using linked list method and spatial hashing technology to reduce the particle neighbor search complexity to linear level. Simultaneously, GPU heterogeneous acceleration is utilized for core computing, while the CPU handles task scheduling and data exchange. Finally, batch simulation automation is achieved, i.e., scripts are developed to automatically read parameters from high / extremely volatile areas and generate input files, submitting them in batches to the supercomputing job system and monitoring the running status in real time. After the simulation is completed, key parameters such as motion path, velocity time history, impact kinetic energy, and accumulation range are automatically extracted and structured and stored in a geospatial database. This process overcomes the limitation of traditional DEMs being applicable only to single landslides, achieving efficient physical simulation of a large number of potential landslides at a large regional scale.

[0041] S50: Input the aforementioned hazard parameters as loads into the predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances of traffic engineering structures, and quantitatively calculate the quantitative vulnerability results of the traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures include the probability of failure or the degree of functional loss of each structure under landslides of different intensities.

[0042] Step S50 is used to quantitatively assess the vulnerability of traffic engineering structures. In practice, vulnerability curves or models are defined for key traffic engineering structures such as roadbeds, bridges (piers, abutments), and tunnel entrances. The impact kinetic energy and accumulation pressure of landslides at specific locations, simulated in step S40, are input as loads into the structural vulnerability model. Combined with the seismic, erosion, and burial resistance properties of the engineering structure itself, the probability of failure or degree of functional loss under landslides of different intensities is quantitatively calculated.

[0043] In step S50, the structural vulnerability model uses a vulnerability curve in the form of a log-normal distribution:

[0044] in This refers to the impact kinetic energy or accumulation pressure of a landslide. This is the median value of the structural resistance. The standard deviation is logarithmic. Separate models are used for the three types of structures: roadbeds are modeled using accretion pressure, bridges using impact kinetic energy, and tunnel entrances using a two-parameter logistic regression model.

[0045] The structural properties (section dimensions, material strength, reinforcement ratio, etc.) are quantified through finite element simulation or code calculation. and The impact kinetic energy, accumulation pressure, and other hazard parameters of the landslide at a specific location obtained from step S40 simulation are used as load inputs into the structural vulnerability model to calculate the failure probability of each structure. Furthermore, the functional loss coefficient is calculated using the weights of the damaged state. The overall loss of the road section is modeled using a series system model. This method achieves quantitative coupling between landslide dynamic parameters and the mechanical response of engineering structures.

[0046] S60: Integrating the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of transportation engineering structures, a risk zoning map is generated using a risk matrix or risk calculation formula, with the characteristics of the disaster-causing body and the information of the disaster-bearing body displayed in the map.

[0047] The purpose of step S60 is to generate a traffic network risk zoning map. In practice, this involves integrating the spatial susceptibility probability of earthquake landslides from step S30 with the quantitative vulnerability results of traffic engineering structures from step S50, and using a risk matrix or risk calculation formula to generate a risk zoning map with traffic network units (such as road sections, bridges, and tunnels) as the assessment objects. This risk zoning map not only displays the risk level but also associates specific characteristics of the disaster-causing body (such as high susceptibility or high impact energy) and information about the disaster-bearing body.

[0048] In step S60, the traffic network risk zoning map is generated using a risk calculation formula:

[0049] in The grid cell susceptibility probability (0~1) is output in step S30. The segment function loss coefficient (0~1) is output in step S50. The overall risk value is 0~1.

[0050] The transportation network is divided into continuous road segment units ranging from 100m to 500m, or bridges, tunnel entrances, and roadbed sections are treated as independent evaluation units. For each road network unit, the maximum susceptibility probability of its covering grid is extracted, and the comprehensive functional loss of multiple structures within the unit is calculated using a series system model. The results are then substituted into the formula to obtain the risk value.

[0051] Based on risk values, road network units are divided into four levels: low risk ( ), medium risk ( High risk Extremely high risk Using a GIS platform, risk levels are classified and rendered to generate a traffic network risk zoning map. The map marks the main disaster-causing factors and disaster-bearing body types of high-risk road sections, providing an intuitive basis for disaster prevention decisions.

[0052] In summary, the method described in this application, through steps S10-S60, constructs a closed-loop technical system for earthquake-induced landslide risk assessment of transportation networks, covering the entire chain from disaster-causing factor identification to disaster spatiotemporal prediction, physical process simulation, and disaster-bearing body response evaluation. By employing a combination of machine learning for coarse screening and physical simulation for fine calculation, it significantly improves the physical realism and accuracy of landslide dynamic processes and their engineering impact assessments while ensuring efficiency across a wide range of scenarios. The final results directly benefit the operation and maintenance management of transportation infrastructure; the provided risk zoning maps can be directly used to guide the optimal allocation of risk management resources, the development of emergency plans, and the seismic resilience design of engineering projects, demonstrating clear application value.

[0053] This application also provides a system for assessing the risk of earthquake-induced landslides in high-intensity seismic transport networks, such as... Figure 2 As shown, the earthquake landslide risk assessment system for the transportation network in this high-intensity seismic zone includes: The factor screening module 201 is configured to obtain a list of landslides in the transportation network of high-intensity areas induced by historical earthquakes, initially select multiple influencing factors to construct a disaster sample database, use an information content model to calculate the information content of each influencing factor, and combine the information content value, the proportion of disaster area and the number of disaster points to conduct multi-index correlation analysis on the initially selected multiple influencing factors, and screen out key factors suitable for evaluating the susceptibility of earthquake landslides in the transportation network of high-intensity areas. The sample library construction module 202 is configured to convert the landslide disaster surface into a grid cell based on the key factors, and use the grid center point as a positive sample; set a dynamic buffer outside the historical landslide boundary, and randomly select an equal number of negative samples in the extremely low information value area outside the buffer to construct an optimized sample library; wherein, the extremely low information value area is the area where the information value is lower than a set threshold. The susceptibility assessment module 203 is configured to construct a spatial weighted machine learning model that considers the neighborhood features of grid cells based on the optimized sample library, divide the sample library into training and test sets according to a set ratio for model training and accuracy verification, conduct earthquake landslide susceptibility assessment, and generate a landslide susceptibility zoning map; wherein the landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas; The physical simulation module 204 is configured to automatically extract the topographic and lithological parameters of potential landslide source areas based on the identification results of high-risk or extremely high-risk areas, construct a high-fidelity simulation model using the parallel particle discrete unit method, simulate the entire process of landslide instability initiation and motion accumulation, and obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy, and accumulation range. The vulnerability assessment module 205 is configured to input the hazard parameters as loads into predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances of traffic engineering structures, and quantitatively calculate the quantitative vulnerability results of the traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures include the probability of failure or the degree of functional loss of each structure under landslide action of different intensities. The risk zoning module 206 is configured to integrate the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of traffic engineering structures, and generate a risk zoning map with traffic network units as the assessment objects using a risk matrix or risk calculation formula. The map displays the characteristics of the disaster-causing body and the information of the disaster-bearing body.

[0054] It should be noted that the aforementioned earthquake landslide risk assessment system for transportation networks in high-intensity zones is based on the same technical concept as the prior method and can achieve the same technical effect, so it will not be elaborated here.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for assessing the risk of earthquake-induced landslides in transportation networks in high-intensity seismic zones, characterized in that... The method includes: A list of landslides in high-intensity transportation networks induced by historical earthquakes was obtained, and a disaster sample database was initially constructed by selecting multiple influencing factors. The information content of each influencing factor was calculated using an information content model, and a multi-index correlation analysis was conducted on the initially selected influencing factors in combination with the information content value, the proportion of disaster area, and the number of disaster points. Key factors suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity transportation networks were then selected. Based on the aforementioned key factors, the landslide disaster surface is converted into a grid cell, with the grid center point as the positive sample; a dynamic buffer zone is set outside the historical landslide boundary, and an equal number of negative samples are randomly selected in the extremely low information value region outside the buffer zone to construct an optimized sample library; wherein, the extremely low information value region is the region where the information value is lower than a set threshold. Based on the optimized sample library, a spatial weighted machine learning model considering the neighborhood characteristics of grid cells is constructed. The sample library is divided into training and testing sets according to a set ratio for model training and accuracy verification. Earthquake landslide hazard susceptibility assessment is carried out, and a landslide susceptibility zoning map is generated. The landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas. Based on the identification results of high-risk or extremely high-risk areas, the topographic and lithological parameters of potential landslide source areas are automatically extracted. A high-fidelity simulation model is constructed using the parallel particle discrete element method to simulate the entire process of landslide instability initiation, movement and deposition, and to obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy and deposition range. The aforementioned hazard parameters are input as loads into predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances in traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures are then quantitatively calculated. These quantitative vulnerability results include the probability of failure or the degree of functional loss of each structure under landslides of different intensities. By integrating the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of transportation engineering structures, a risk zoning map is generated with transportation network units as the assessment objects using a risk matrix or risk calculation formula. The map displays the characteristics of the disaster-causing body and the information of the disaster-bearing body.

2. The method according to claim 1, characterized in that, The initially selected influencing factors include elevation, slope, aspect, profile curvature, stratigraphic lithology, distance from roads, distance from faults, distance from epicenter, and distance from water systems. Through multi-index correlation analysis, key factor combinations suitable for traffic network scenarios in high-intensity seismic zones are selected.

3. The method according to claim 1, characterized in that, The dynamic buffer distance is 80-120 meters, and the selection area of ​​negative samples is limited to the area outside the dynamic buffer and where the information content value is lower than a preset threshold.

4. The method according to claim 1, characterized in that, The spatial weight machine learning model includes at least one of random forest, extreme gradient boosting, or support vector machine. The construction process of the spatial weight machine learning model includes: constructing a neighborhood weight matrix based on factor importance and Euclidean distance between sample points, introducing spatial weights into the machine learning framework for training and optimization, and calculating the probability of landslide disasters occurring at each point in the region.

5. The method according to claim 1, characterized in that, The methods for generating landslide susceptibility zoning maps include: mapping the calculated probability of disaster occurrence to the original geographic spatial domain, and using the natural breakpoint method to divide the area into five categories: extremely low susceptibility zone, low susceptibility zone, medium susceptibility zone, high susceptibility zone, and extremely high susceptibility zone. The identification results of the high susceptibility zone and the extremely high susceptibility zone serve as the basis for extracting potential landslide source areas.

6. The method according to claim 1, characterized in that, The simulation of the entire physical process of landslide movement is based on parallel computing using a supercomputing platform.

7. The method according to claim 1, characterized in that, The vulnerability curves or models of the transportation engineering structures are pre-established based on the seismic, shock-resistant, and burial-resistant properties of the engineering structures themselves.

8. The method according to claim 1, characterized in that, The risk zoning map uses road sections, bridges, or tunnels as basic assessment units. The risk level of each assessment unit in the map is calculated by coupling the susceptibility probability and structural vulnerability results through a risk matrix, and the specific characteristics of the disaster-causing body and the information of the disaster-bearing body are displayed in association.

9. A system for assessing the risk of earthquake-induced landslides in transportation networks in high-intensity seismic zones, characterized in that, The system includes: The factor screening module is configured to obtain a list of landslides in high-intensity transportation networks induced by historical earthquakes, initially select multiple influencing factors to construct a disaster sample database, use an information content model to calculate the information content of each influencing factor, and combine the information content value, the proportion of disaster area and the number of disaster points to conduct multi-index correlation analysis on the initially selected multiple influencing factors, and screen out key factors suitable for evaluating the susceptibility of earthquake-induced landslides in high-intensity transportation networks. The sample library construction module is configured to convert landslide disaster surfaces into grid cells based on the key factors, with the grid center point as the positive sample; set a dynamic buffer outside the historical landslide boundary, and randomly select an equal number of negative samples in the extremely low information value region outside the buffer to construct an optimized sample library; wherein, the extremely low information value region is the region where the information value is lower than a set threshold. The susceptibility assessment module is configured to construct a spatial weighted machine learning model that considers the neighborhood features of grid cells based on the optimized sample library, divide the sample library into training and testing sets according to a set ratio for model training and accuracy verification, conduct earthquake and landslide susceptibility assessment, and generate a landslide susceptibility zoning map; wherein the landslide susceptibility zoning map includes the identification results of high susceptibility areas or extremely high susceptibility areas; The physical simulation module is configured to automatically extract topographic and lithological parameters of potential landslide source areas based on the identification results of high-risk or extremely high-risk areas, construct a high-fidelity simulation model using the parallel particle discrete unit method, simulate the entire process of landslide instability initiation and deposition, and obtain hazard parameters, including landslide movement path, velocity time history, impact kinetic energy, and deposition range. The vulnerability assessment module is configured to input the hazard parameters as loads into predefined vulnerability curves or models of roadbeds, bridges, or tunnel entrances of traffic engineering structures, and quantitatively calculate the quantitative vulnerability results of the traffic engineering structures. The quantitative vulnerability results of the traffic engineering structures include the probability of failure or the degree of functional loss of each structure under landslide action of different intensities. The risk zoning module is configured to integrate the spatial susceptibility probability of earthquakes and landslides with the quantitative vulnerability results of transportation engineering structures. It uses a risk matrix or risk calculation formula to generate a risk zoning map with transportation network units as the assessment objects. The map displays the characteristics of the disaster-causing body and the information of the disaster-bearing body.

Images