Slope rainfall infiltration stability dynamic evaluation and landslide prediction method

Abstract

The invention discloses a slope rainfall infiltration stability dynamic evaluation and landslide prediction method. The method comprises the following steps: constructing a numerical model coupling rainfall infiltration and slope stress field response; integrating multi-source data, and updating model boundary conditions in real time; performing dynamic stability evaluation based on a local safety coefficient method; and training a landslide prediction model and realizing intelligent early warning identification. The seepage-stress-strength three-field coupling numerical model is provided, the physical response process of the unsaturated soil body is dynamically simulated, the precursor mechanism of rainfall-induced landslide can be truly restored, and prediction scientificity and accuracy are improved; remote sensing topographic data, soil property survey parameters and real-time meteorological monitoring information are integrated, model boundary conditions and initial states are dynamically updated through a system interface, real-time response to environmental changes is achieved, and simulation reliability is improved; a local safety coefficient evaluation mechanism is introduced, and a continuous weak region discrimination algorithm is combined, so that space identification and time sequence tracking of a potential slip region and a damage zone are realized, and the slope stability analysis refinement level is improved.

Description

Technical Field

[0001] This invention relates to a prediction method, and more particularly to a method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation. Background Technology

[0002] Slope engineering is widely used in infrastructure construction such as highways, railways, water conservancy, hydropower, mines, and urban slopes. Its stability is directly related to the safety of the project and the safety of people's lives and property. Especially in areas with large topographic relief, complex geological conditions, or variable climate, rainfall-induced landslides are one of the most common and destructive geological hazards. Under the influence of heavy or continuous rainfall, the seepage field and stress field inside the slope soil continuously evolve, pore water pressure increases, and effective stress decreases, which can easily trigger slope instability.

[0003] Currently, with the development of remote sensing monitoring, geophysical exploration technology, and numerical analysis methods, slope stability research is gradually evolving from empirical to physical mechanism-based, and from static calculation to dynamic simulation. Common stability evaluation methods include the limit equilibrium method, the finite element strength reduction method, and empirical rainfall threshold models, which have already served to some extent in slope engineering design and landslide risk prediction.

[0004] However, in practical applications, traditional methods generally have the following problems: First, the assessment methods are static and cannot dynamically reflect the changes in slope condition during rainfall; second, the model parameters and boundary conditions are updated in a lagging manner and cannot adapt to real-time responses during rainfall; and third, there is a lack of intelligent early warning mechanisms, and the methods still rely on manual analysis and judgment, which are characterized by lag and uncertainty.

[0005] On the other hand, global climate change is increasing the frequency of extreme weather events, significantly raising the risk of landslides triggered by heavy rainfall. Faced with increasingly complex slope stability issues, engineering practice urgently needs a technical solution that possesses multi-source data fusion capabilities, supports dynamic seepage-stress coupling analysis, and enables real-time identification of slope stability levels and intelligent landslide early warning.

[0006] Therefore, developing a dynamic assessment method for slope rainfall stability and landslide prediction based on the combination of hydro-mechanical coupling numerical simulation and intelligent recognition algorithm not only has important engineering application value, but also provides scientific and efficient technical support for natural disaster prevention and infrastructure safety.

[0007] Currently, the finite element strength reduction method (FEM-SRM) is widely used in academia and engineering to analyze slope stability in response to rainfall-induced landslides. This method typically combines a two-dimensional or three-dimensional slope geometric model, inputs geological parameters, and sets rainfall boundary conditions to calculate the overall safety factor of the slope, thereby assessing whether the slope is in an unstable state.

[0008] In addition, some studies have introduced the theory of unsaturated seepage to simulate the changes in pore water pressure in soil during rainfall and weakly couple it with the stress field to reflect the trend of slope strength attenuation under the influence of rainfall.

[0009] Another type of method uses empirical index methods or critical rainfall models to establish empirical risk assessment models by analyzing historical landslide data. These models are widely used for preliminary landslide sensitivity zoning and early warning threshold setting.

[0010] Currently, slope stability analysis and landslide early warning methods mainly rely on traditional limit equilibrium methods, strength reduction methods, or landslide identification methods based on empirical indicators. However, these methods still have the following objective technical shortcomings in addressing the complex coupling mechanisms of landslides induced by heavy rainfall and in real-time risk identification:

[0011] 1. The assessment method is static and cannot reflect the dynamic evolution of the rainfall process: Existing methods generally adopt static analysis methods, such as single calculation of the overall safety factor or critical slip surface search, which only reflect the stability state at a certain moment. They cannot capture the whole process response of soil pore pressure increase, strength degradation and stress change during rainfall, which can easily lead to delayed early warning or misjudgment.

[0012] 2. Fixed model boundary conditions, lacking the ability to respond to real-time rainfall and changes in operating conditions: In traditional analysis, model parameters are mostly derived from previous surveys. Factors such as rainfall boundaries, topographic changes, and groundwater levels are often set as constant or approximate conditions, failing to integrate measured meteorological or remote sensing data, resulting in deviations between simulation results and actual slope responses.

[0013] 3. The overall safety factor is roughly determined, making it difficult to accurately identify potential landslide development areas: The practice of using the overall safety factor as the sole criterion cannot identify local damage or the development of small-scale shear zones. Especially in loess areas, it is easy to overlook landslide areas controlled by local weak layers or structural surfaces, and the stability assessment results lack specificity.

[0014] 4. Lack of landslide probability prediction mechanism, making it impossible to achieve quantitative risk warning: Most current methods only make manual judgments when simulation results are abnormal, lacking a data-driven landslide probability modeling and risk classification mechanism, which makes it impossible to achieve quantitative, automated and dynamic updates of warnings, affecting the timeliness and scientific nature of landslide prevention and control.

[0015] 5. Fragmented technology chain, low system integration, and difficulty in engineering deployment: Simulation calculation, data acquisition and early warning output are often completed by multiple independent systems. The lack of a unified integration platform and process management mechanism leads to low information transmission efficiency and delayed system response, making it difficult to meet the engineering application needs of rapid assessment and real-time early warning for regional landslide prevention and control.

[0016] In summary, existing technologies have significant shortcomings in terms of dynamism, real-time performance, precision, intelligence, and systemicity. There is an urgent need for a new technical method that can achieve full-process coupled modeling, real-time data fusion, local stability identification, and intelligent landslide early warning to meet the practical needs of slope and landslide monitoring and safety management in complex geological environments. Summary of the Invention

[0017] To address the shortcomings of existing technologies, this invention discloses a method for dynamic assessment of slope stability under rainfall infiltration and landslide prediction based on hydraulic coupling numerical simulation. The technical solution is as follows:

[0018] A method for dynamic assessment of slope stability under rainfall infiltration and landslide prediction based on hydraulic coupling numerical simulation, characterized by:

[0019] S1: Construct a numerical model that couples rainfall infiltration with slope stress field response;

[0020] S2: Integrates multi-source data and updates model boundary conditions in real time;

[0021] S3: Dynamic stability assessment based on the local safety factor method;

[0022] S4: Train the landslide prediction model and achieve intelligent early warning and identification.

[0023] Beneficial effects

[0024] 1. The shift from static evaluation to dynamic simulation

[0025] Existing technical problems: Traditional slope stability analysis often uses the limit equilibrium method or the strength reduction finite element method, and the analysis results are a single safety factor, which cannot characterize the evolution process of pore pressure increase, strength decrease and slope instability during rainfall infiltration.

[0026] Advantages of this invention: It proposes a three-field coupled numerical model of seepage-stress-strength to dynamically simulate the physical response process of unsaturated soil, which can realistically reproduce the precursor mechanism of rainfall-induced landslides and improve the scientificity and accuracy of prediction.

[0027] 2. The shift from fixed parameters to real-time updates

[0028] Existing technical issues: Most model parameters are input only once, ignoring the influence of time-varying data such as topography, soil properties, and weather, resulting in analysis results that are out of sync with on-site changes.

[0029] Advantages of this invention: It integrates remote sensing topographic data, soil survey parameters and real-time meteorological monitoring information, and dynamically updates the model boundary conditions and initial state through the system interface, thereby achieving real-time response to environmental changes and significantly improving simulation reliability.

[0030] 3. Shift from overall discrimination to local recognition

[0031] Existing technical problems: Conventional methods only provide the overall safety factor of the slope, and it is difficult to identify the early development of potential slip surfaces or local failure areas.

[0032] Advantages of this invention: It introduces a local safety factor (LFS) assessment mechanism and combines it with a continuous weak zone discrimination algorithm to achieve spatial identification and temporal tracking of potential slip zones and failure zones, thereby improving the precision of slope stability analysis.

[0033] 4. Shift from post-event judgment to early warning

[0034] Existing technical problems: Most systems are only used for result analysis and lack probability prediction and risk classification of landslide occurrence, thus failing to provide effective early warning information.

[0035] Advantages of this invention: It integrates machine learning methods with historical landslide case data to construct a landslide probability prediction model, and, in conjunction with numerical simulation output results, generates landslide risk levels and trigger window judgments, thereby achieving early warning and automatic alarm.

[0036] 5. The shift from isolated analysis to system integration

[0037] Existing technical problems: The traditional technology chain is fragmented, with numerical simulation, data updates, and early warning identification operating independently, resulting in low information transmission efficiency and difficulty in engineering deployment.

[0038] Advantages of this invention: It proposes a fully integrated and modular landslide dynamic assessment and early warning system that can be embedded in slope monitoring platforms or geological disaster prevention and control systems to achieve multi-module collaborative linkage and improve practicality and scalability. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the process of the present invention;

[0040] Figure 2 This is a schematic diagram illustrating the changes in slope saturation under the influence of rainfall infiltration.

[0041] Figure 3 This is a schematic diagram of the equivalent plastic strain and potential yield surface;

[0042] Figure 4 This is a schematic diagram for calculating the local safety factor.

[0043] Figure 5 This is a schematic diagram of the local safety factor. Detailed Implementation

[0044] This invention discloses a method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation. The overall process is as follows: Figure 1 As shown, it mainly includes the following four steps:

[0045] S1: Constructing a Coupled Numerical Model. This step is executed jointly by the computation module and the finite element simulator. When the system initiates a new rainfall infiltration-stability assessment task, an unsaturated soil-slope coupled model is established in the COMSOL Multiphysics simulation platform. The model incorporates a permeability coefficient function related to saturation, a strength reduction mechanism induced by pore water pressure, and sets rainfall boundary conditions and initial conditions to complete time-varying coupled simulation calculations. This step obtains dynamic evolution information of the slope seepage field and stress field under different rainfall histories, providing a data foundation for subsequent stability analysis.

[0046] This invention establishes a numerical analysis model for slope stability under rainfall infiltration conditions. It mainly consists of three parts: the slope's soil and rock mass, the saturated zone below the groundwater level, and the rainfall infiltration boundary conditions. This includes the slope's geometry, material properties, and boundary conditions. The entire process can be implemented and coupled as a whole using the finite element software COMSOL Multiphysics, and applied to the analysis and research of slope stability. The model includes two main modules: solid mechanics and Darcy's law. The solid mechanics module simulates the slope's mechanical response, simulating soil stress and deformation, while the Darcy's law module simulates rainfall infiltration, describing the movement of water in unsaturated soil. The model uses the Drucker-Prager constitutive relation to describe the mechanical behavior of the slope's soil and rock mass, and the Richards equations in the fluid equations are used to describe the nonlinear flow process in a variablely saturated porous medium.

[0047] Numerical simulations of rainfall infiltration require consideration of the complex interaction between water flow and soil, i.e., the coupling of seepage analysis and mechanical analysis. In the mechanical analysis, the total stress distribution in the soil... By combining pore water pressure with the effective stress of the soil, we can analyze its deformation, stress, and shear strength. The mechanical response of the soil is described by the stress equilibrium equation, and the deformation and stress need to be calculated based on the constitutive relationship of the soil.

[0048] (1)

[0049] In the formula, f represents the body force per unit volume.

[0050] To analyze seepage in unsaturated soil slopes, the Richards equation or the unsteady Darcy equation is usually used to describe the movement of water in the soil, and to obtain the hydraulic head, permeability coefficient, and pore water pressure.

[0051] (2)

[0052] In the formula, ρ is the soil volumetric water content; k is the unsaturated soil permeability coefficient (m / s); h is the pressure head (m). It is the gradient operator. The total head gradient determines the flow direction and velocity distribution. This represents unsteady seepage driven by the total water potential gradient; S represents the source term. The Van Genuchten model (VG model) is a widely used mathematical model to describe soil moisture characteristics curves (SWCC) and soil water transport processes. In seepage analysis, it is mainly used to describe the changes in soil water saturation under different water heads. It is represented as:

[0053] (3)

[0054] In the formula, , These represent the soil saturation and residual volumetric water content, respectively; Se represents the effective saturation, expressed as:

[0055] (4)

[0056] in, For VG model parameters, usually , It is a constant related to soil particle size and pore structure, which controls the steepness of the soil moisture curve.

[0057] Geometric model: Construct a two-dimensional or three-dimensional geometric model based on slope topographic data (measured profile), including structural units such as surface overburden, bedrock layer, and water table.

[0058] Boundary conditions: In the solid mechanics module, it is assumed that the soil layer is supported by a hard, rough foundation, and fixed constraints are applied to the lower horizontal boundary. Roller support boundary conditions are applied to the left and right vertical boundaries, specifying that the horizontal displacement on both sides of the model is zero. The upper boundary of the model is a free boundary.

[0059] In the Darcy's Law module, the area from the bottom of the model to the groundwater level is designated as pressure boundaries on both sides, while the rest are stagnant boundaries. A normal inflow velocity is then set on the slope surface as a rainfall infiltration boundary condition. See [link to relevant documentation]. Figure 2 As shown.

[0060] Mesh generation: An irregular mesh generation method (triangular / tetrahedral elements) is adopted, and the surface and potential slip surface regions are refined to enhance the accuracy of local solutions.

[0061] The finite element method (FEM) strength reduction method is a commonly used method for slope stability analysis and calculation. Essentially, it uses numerical analysis techniques to solve for the limit state of a slope within an elastoplastic finite element model. Specifically, the principle is to continuously reduce the strength parameters of the slope's soil and rock mass using numerical analysis software, lowering its shear strength until failure occurs. The strength reduction factor at this point is considered the slope's strength reserve safety factor. The software can also plot the potential slip surface of the slope based on equivalent plastic strain.

[0062] COMSOL Multiphysics software can be used to study slope stability using the finite element strength reduction method. The specific setup method is as follows: based on the aforementioned calculation results, a new solid mechanics module is added. Given the load conditions, the cohesion and internal friction angle are parameterized as functions of the safety factor (FOS), expressed as:

[0063] (14)

[0064] In the formula, c, c represents the cohesion and internal friction angle of the actual soil. F , F Here, FOS represents the reduced soil cohesion and internal friction angle; FOS is the reduction factor. Subsequently, the auxiliary scanning function in the software gradually increases FOS, thereby continuously reducing the soil's material parameters and shear strength until the slope becomes unstable and fails. At this point, the slope's stability safety factor is obtained, and the corresponding reduced values ​​of c and φ are also obtained. The value. The equivalent plastic strain distribution map of the slope soil and rock mass under seepage can be obtained by calculating using the strength reduction method. Its potential yield surface can be determined based on the equivalent plastic strain map in the calculation results. An equivalent plastic strain greater than 0 indicates that the soil has yielded. Areas with large equivalent plastic strains generally exhibit greater plastic deformation of the soil and rock mass; therefore, this can be used to characterize the slip surface of slope instability and failure. See [reference needed]. Figure 3 As shown.

[0065] S2: Real-time update of boundary conditions. Working in collaboration with the data acquisition module and modeling system, real-time meteorological information, topographic data, groundwater level, and soil parameters are acquired through remote sensing mapping and on-site monitoring equipment before or during rainfall. Integrating with a GIS platform or IoT monitoring platform, it automatically calls API interfaces to periodically synchronize and dynamically update model boundary conditions and initial field information.

[0066] This system enables dynamic synchronization and real-time updating of boundary conditions in a slope rainfall-infiltration coupling model. This ensures the model calculations reflect the latest field conditions and achieves more accurate dynamic stability assessments. The process is primarily accomplished through collaboration between the data acquisition module and the modeling system. The detailed implementation principles and procedures are as follows:

[0067] 1. Boundary and Initial Conditions

[0068] In the coupled rainfall infiltration slope stability model, the key boundary conditions include: meteorological factors such as rainfall intensity and duration applied to the upper boundary of the model, which are the main external conditions driving the infiltration process; considering the seepage path and saturation zone development of the slope soil, pressure head boundaries are set on the side of the model and no-flow boundaries are set at the bottom.

[0069] Initial field information: including initial soil moisture content, pore water pressure distribution, and spatial distribution of soil parameters (such as permeability coefficient, shear strength index, etc.), used to describe the initial state of the model.

[0070] 2. Implementation principle and process of real-time updates

[0071] (1) Data acquisition

[0072] Real-time updates rely on the continuous acquisition of field and remote sensing data, including the following information sources:

[0073] Meteorological data: Obtain hourly or minute-level information such as rainfall intensity and duration through weather stations or API interfaces;

[0074] In-situ monitoring equipment: Soil moisture sensors, groundwater level gauges, rain gauges and pore pressure gauges deployed on the slope to obtain the physical condition of the slope in real time;

[0075] By integrating the aforementioned monitoring equipment through a web service platform, a real-time data interface that can be accessed and operated is formed.

[0076] (2) API interface call and data synchronization mechanism

[0077] To achieve automatic data synchronization to the model, this invention designs the following mechanism:

[0078] The system is equipped with a scheduled task module that automatically triggers at set intervals (every 10 minutes or hour); it sends requests to the data platform through a standard API interface to retrieve the latest data; the returned data is parsed in a structured format; and the parsed data is automatically assigned to the boundary condition function expressions and variables of the numerical model, such as mapping rainfall intensity to the flux term of the upper boundary of the model.

[0079] (3) Modeling and execution method for boundary update

[0080] In the COMSOL platform, calling the MATLAB interface enables automatic updating of modeling parameters: real-time meteorological data is input as the model's "global definition variable"; the rainfall-stop change point in the model is set using function expressions, and the model's rainfall intensity expression is updated in real time; the initial field information is reassigned through the "initial value setting module" and the calculation is restarted.

[0081] The core of the real-time update mechanism lies in automatically sensing changes on-site and adjusting model conditions in real time. It achieves a complete technology chain through data acquisition, API parsing, model synchronization, and boundary reconstruction, which significantly improves the response speed, accuracy, and engineering adaptability of slope instability simulation based on rainfall.

[0082] S3: Dynamic Stability Assessment. The system calls upon the seepage-stress data obtained from S1, calculates the stability level of each unit based on the Local Factor of Safety (LFS) method, identifies potential instability zones using the Mohr-Coulomb failure criterion and strength reduction concept, and identifies the development process and spatial evolution of slip zones, providing higher spatial resolution and early warning foresight.

[0083] The Local Safety Factor (LSF) reflects the ability of soil in a specific area of ​​a slope to resist failure under the current stress state. The calculations based on the Local Safety Field theory follow the Mohr-Coulomb failure criterion and are defined as the ratio of the Coulomb force at failure of the calculated element to the current Coulomb force, such as... Figure 4 As shown. It can be represented as:

[0084] (9)

[0085] In the formula, This represents the shear failure resistance of the current element under its current stress state. This represents the actual shear stress borne by the current calculation unit, reflecting the driving force of failure.

[0086] The local safety factor defined by this evaluation method at each subdivision point enables the coupling and analysis of the evolution process of the slope moisture field and effective stress field under uniform rainfall infiltration.

[0087] Based on the stress expression within a two-dimensional slope, the following can be derived:

[0088] (10)

[0089] In the formula, , It refers to the effective cohesion and effective internal friction angle of the slope soil. , Let the principal stress difference and the mean effective principal stress be represented as follows:

[0090] (11) (12)

[0091] In the formula, , The first and third effective principal stresses are respectively used to introduce absorption stress. ,

[0092] (6)

[0093] In the formula, , These represent the gas phase pressure (usually taken as 0) and the pore water pressure, respectively. For effective saturation, The parameters are those for the van Genuchten model. The effective stress is obtained by replacing pore pressure with absorbed stress. formula:

[0094] (7) (8)

[0095] In the formula, , These are the first and third principal stresses, respectively.

[0096] The LSF distribution accurately characterizes the evolution path of slopes from stable to unstable during rainfall infiltration, both temporally and spatially. In the early and middle stages of rainfall, shallow unstable zones with localized LSF < 1 initially appear, gradually developing downwards to form continuous slip surfaces. The LSF and equivalent plastic strain field show a high degree of spatial agreement; the evolution path of the plastic zone can be considered as the deformation response result of the LSF-discriminated region. This coupling relationship provides a mutual verification mechanism for landslide identification and early warning. (See also...) Figure 5 As shown.

[0097] S4: Landslide Prediction and Early Warning Identification. This is accomplished jointly by a machine learning module and an early warning system. When the system detects abnormal fluctuations in the LFS value or obtains new simulated data, it invokes algorithms such as Random Forest, Support Vector Machine, and Long Short-Term Memory Neural Network (LSTM) to predict the probability of landslide occurrence and outputs the early warning level and time window. Input variables include rainfall, cumulative rainfall, slope, geological type, pore pressure growth rate, and LFS evolution curve; output variables include key indicators such as landslide probability, risk level, and expected occurrence period. This enables intelligent identification and early warning of high-risk landslide events on slopes, assisting management units in disaster prevention and mitigation deployment.

[0098] By integrating multiple machine learning algorithms with historical landslide case data, a landslide probability prediction model is constructed. Combined with numerical simulation output, this model generates landslide risk levels and trigger windows, enabling early warning and automatic alarm for landslide events. The landslide probability prediction model is built using the Stacking ensemble learning method. It employs multiple highly differentiated and accurate base learners, including random forests, support vector machines, and long short-term memory neural networks, to form the first layer of the model. A stable prediction output is obtained using 5-fold cross-validation, and this output is used as the input feature of the Stacking meta-learner to train the final landslide probability prediction model.

[0099] The model training process includes: ① cleaning and feature selection of historical landslide case samples; ② constructing multiple basic learners and optimizing hyperparameters; ③ obtaining the prediction output of the first layer based on the k-fold cross-validation method, which is used as the input of the second layer Stacking meta-learner, and finally completing the training of the Stacking landslide prediction model.

[0100] The model inputs include rainfall, cumulative rainfall, slope, geological type, pore water pressure growth rate, and local safety factor evolution curves as the probability of landslide occurrence. The outputs are the landslide probability and trigger window.

[0101] In practical applications, the local safety factor (LFS) evolution data output from numerical simulations (rainfall infiltration-stress coupling analysis based on the COMSOL platform) and input features (such as rainfall amount and rainfall time) are input into a trained Stacking model to output the landslide probability in real time. To improve the stability of the judgment, a sliding window mechanism is introduced, using the average absolute error within the sliding window as the state evaluation index, and setting a dynamic early warning threshold based on interval estimation theory. When the landslide probability exceeds the set threshold and the state index remains higher than the statistical threshold, the system determines the landslide risk level (e.g., high / medium / low) and outputs a trigger window (e.g., within the next 6 hours or 12 hours) based on the rate of change of the predicted probability or the LSTM time series trend, thereby achieving dual intelligent judgment of landslide risk level and expected trigger time. Specifically, high risk: landslide probability greater than 0.8, trigger window less than or equal to 6 hours; medium risk: landslide probability greater than 0.5 but less than 0.6, trigger window less than or equal to 24 hours; low risk: landslide probability less than 0.5, continuous monitoring.

[0102] This invention proposes a systematic method for dynamic stability assessment and landslide prediction of slopes under rainfall infiltration, with the following key innovations:

[0103] 1. Integrated construction of a multi-physics field coupled simulation framework: Unlike traditional static stability analysis methods, this invention is the first to construct a coupled simulation system for the entire process of unsaturated seepage-stress field-strength evolution, which can realistically reproduce the dynamic changes in soil saturation, pore water pressure, effective stress and strength under rainfall, thus improving the scientificity and timeliness of landslide precursor identification.

[0104] 2. Multi-source data-driven real-time boundary update mechanism: To address the problem of model input lag, this invention introduces an integrated interface for high-precision terrain data, soil parameters, and real-time meteorological data to achieve dynamic updates of rainfall boundary conditions and initial field states, ensuring that simulation results closely reflect the evolution of actual working conditions.

[0105] 3. Combined application of local safety factor method and instability criterion: In the slope safety evaluation, this invention adopts local element LFS calculation and combines it with the slip zone continuity identification algorithm, which is more refined and sensitive than the traditional global safety factor. It can detect potential damage areas in advance and achieve earlier and more accurate risk zoning.

[0106] 4. Integration of Landslide Prediction Model and Intelligent Early Warning System: This invention utilizes historical landslide case data and trains a landslide prediction model using machine learning methods to achieve data-driven modeling of landslide occurrence probabilities and risk level identification. The model supports real-time input, rolling prediction, and threshold-based early warning, realizing intelligent and automated landslide early warning.

[0107] 5. Full-process integrated and modular deployment capability: The technical solution of this invention achieves modular encapsulation from modeling, data updating, calculation and evaluation to risk output, which has good engineering application and promotion capabilities and is easy to deploy and operate in different regions and different types of unsaturated slopes.

[0108] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for dynamic assessment of slope stability under rainfall infiltration and landslide prediction based on hydraulic coupling numerical simulation, characterized by: S1: Construct a numerical model that couples rainfall infiltration with slope stress field response; S2: Integrates multi-source data and updates model boundary conditions in real time; S3: Dynamic stability assessment based on the local safety factor method; S4: Train the landslide prediction model and achieve intelligent early warning and identification.

2. The method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation according to claim 1, characterized in that: S1 includes the following: The calculation module and the finite element simulator work together to perform the simulation. When the system starts a new rainfall infiltration-stability assessment task, an unsaturated soil-slope coupled model is established in the COMSOL Multiphysics simulation platform. The model introduces a permeability coefficient function related to saturation and a strength reduction mechanism induced by pore water pressure. Rainfall boundary conditions and initial conditions are set to complete the time-varying coupled simulation calculation. Through this step, the dynamic evolution information of the slope seepage field and stress field under different rainfall histories can be obtained, providing a data basis for subsequent stability analysis.

3. The method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation according to claim 1, characterized in that: The S2 includes the following: the data acquisition module and the modeling system work together to acquire real-time meteorological information, topographic data, groundwater level and soil parameters through remote sensing mapping and on-site monitoring equipment before or during rainfall; and integrate a GIS platform or IoT monitoring platform to automatically call API interfaces to synchronize and dynamically update model boundary conditions and initial field information on a regular basis.

4. The method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation according to claim 1, characterized in that: The S3 includes the following: the system calls the seepage-stress data obtained from S1, calculates the stability level of each unit based on the local safety factor method (LFS), identifies potential instability regions using the Mohr-Coulomb failure criterion and strength reduction concept, and identifies the development process and spatial evolution of slip zones.

5. The method for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation according to claim 1, characterized in that: The S4 includes the following: it is jointly completed by the machine learning module and the early warning system; when the system detects abnormal fluctuations in the LFS value or obtains new simulated data, it calls the random forest, support vector machine, and long short-term memory neural network (LSTM) algorithm models to predict the probability of landslide occurrence and outputs the early warning level and time window; the input variables include rainfall, cumulative rainfall, slope, geological type, pore pressure growth rate, and LFS evolution curve. The output variables are key indicators such as landslide probability, risk level, and expected occurrence time, enabling intelligent identification and early warning of high-risk landslide events on slopes, and assisting management units in disaster prevention and mitigation deployment.

6. A device for dynamic assessment of slope rainfall infiltration stability and landslide prediction based on hydraulic coupling numerical simulation, characterized by: Numerical model building module: Constructs a numerical model that couples rainfall infiltration with slope stress field response; executed jointly by the calculation module and the finite element simulator. When the system starts a new rainfall infiltration-stability assessment task, an unsaturated soil-slope coupled model is established in the COMSOL Multiphysics simulation platform. The module for updating model boundary conditions integrates multi-source data and updates model boundary conditions in real time. It works in collaboration with the data acquisition module and the modeling system to acquire real-time meteorological information, topographic data, groundwater level and soil parameters through remote sensing mapping and on-site monitoring equipment before or during rainfall. It integrates a GIS platform or IoT monitoring platform and automatically calls API interfaces to synchronize and dynamically update model boundary conditions and initial field information on a regular basis. Dynamic stability assessment module: The system calls the seepage-stress data obtained from S1, calculates the stability level of each unit based on the local safety factor method (LFS), and identifies potential instability areas using the Mohr-Coulomb failure criterion and strength reduction concept, as well as the development process and spatial evolution of slip zones. Training the prediction model and its early warning identification module: This is jointly completed by the machine learning module and the early warning system. When the system detects abnormal fluctuations in the LFS value or obtains new simulated data, it calls random forest, support vector machine, and long short-term memory neural network (LSTM) algorithm models to predict the probability of landslide occurrence and outputs the early warning level and time window.

7. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored program, wherein the program, when executed, controls the device where the non-volatile storage medium is located to perform the method described in any one of claims 1 to 5.

8. A terminal device, characterized in that, The terminal device includes: a processor, a memory, a communication interface, and a bus; the processor, the memory, and the communication interface are connected through the bus and communicate with each other; the memory stores executable program code; the processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to perform the method as described in any one of claims 1-5 above.

Images