Rainfall-induced landslide deformation early warning method

Abstract

The present application belongs to the technical field of landslide monitoring, and particularly relates to a rainfall-induced landslide deformation early warning method, comprising the following steps: constructing a landslide dynamic digital twin based on coupled physical mechanism; laying monitoring sensors to collect rainfall, displacement and underground water level data in real time; adopting ensemble Kalman filter data assimilation technology to fuse real-time monitoring data and digital twin prediction results, dynamically adjusting rock-soil mass parameters to realize adaptive calibration of the digital twin; based on the calibrated digital twin, inputting future rainfall forecast data to simulate instability scenarios, and triggering graded early warning according to the predicted stability coefficient change curve. The present application realizes the transition of landslide early warning from static threshold to dynamic prediction, and significantly improves the early warning accuracy and prediction period.

Description

Technical Field

[0001] This invention belongs to the field of landslide monitoring technology, specifically relating to a method for early warning of landslide deformation induced by rainfall. Background Technology

[0002] Rainfall-induced landslides are one of the most common geological hazards. Existing landslide early warning methods have the following limitations: Traditional statistical experience-based models, while simple and easy to use, lack physical mechanism support and cannot accurately reflect the uniqueness of individual landslides, resulting in limited early warning accuracy. Existing technologies lack early warning methods that can integrate physical mechanisms with real-time monitoring data and possess self-learning and adaptive capabilities, leading to low early warning accuracy, insufficient lead time, and inability to perform scenario extrapolation and mechanism analysis. Therefore, this invention proposes an improved early warning method for rainfall-induced landslide deformation. This invention achieves a transformation from static to dynamic prediction in landslide early warning by constructing a dynamic digital twin and employing ensemble Kalman filtering data assimilation technology. Summary of the Invention

[0003] To address the aforementioned shortcomings in the existing technology, this invention provides a method for early warning of rainfall-induced landslide deformation to solve the problems mentioned in the background technology.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A method for early warning of rainfall-induced landslide deformation includes a landslide digital twin module, a monitoring device network module, a data assimilation and dynamic update module, and a graded early warning module. The landslide digital twin module includes constructing a three-dimensional geometric model encompassing the terrain surface, geological interfaces, and potential sliding surfaces. The monitoring device network module deploys monitoring devices at different locations on the landslide site to monitor rainfall and displacement. The data assimilation and dynamic update module updates data variables in the digital twin in real time by combining monitoring network data with Kalman filter data assimilation processing. The graded early warning module simulates future rainfall data to predict the stability coefficient change curve and provide early warning of instability.

[0005] Geological exploration data is used for modeling. The three-dimensional geometric model is discretized by the finite element method. The discretization process requires densification of the sliding zone and toe area in the landslide body.

[0006] The seepage field control equation and the stress field control equation are constructed. The seepage field control equation and the stress field control equation are bidirectionally coupled through the effective stress principle. The coupling process is set with extended state vectors of displacement field, pore water pressure field and material parameter field.

[0007] The Kalman filter data assimilation process generates multiple set members, and the prediction equation is: Where M[·] is the physical model operator, and the update equation in the Kalman filter data assimilation process is... K is the Kalman gain.

[0008] The Kalman gain is calculated using the following formula: ,in For the prediction error covariance matrix, The observation error covariance matrix, For observation operators.

[0009] The existing actual value is corrected according to the Kalman gain K, and the corrected value is used to update the predicted curve.

[0010] The parameter constraint processing includes: constraining the permeability coefficient to a positive value within a preset range; constraining the internal friction angle within a preset angle range; constraining the cohesion to be no less than a preset lower limit of strength; applying smoothness constraints to the soil and rock parameters of adjacent units within the same stratum; applying engineering experience constraints to the soil and rock parameters of the slip zone region; and adopting a graded processing strategy according to the degree of parameter constraint violation.

[0011] The tiered early warning system calculates the stability coefficient Fs based on the updated digital twin.

[0012] A three-level early warning mechanism is established based on the stability coefficient, and tiered response measures are established based on the three-level early warning mechanism.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves real-time tracking and evolution prediction of physical landslide status through the construction of a dynamic digital twin. By using material parameters as state variables in data assimilation, it enables online updating of early warning model parameters, thus completing the transformation from static threshold early warning to dynamic prediction early warning.

[0014] 2. This invention improves early warning accuracy and forecast period by integrating physical models with data assimilation technology and continuously correcting model biases through the data assimilation process. At the same time, it ensures the interpretability of the early warning decision-making process based on clear physical principles, avoiding the problems of black box models. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a method for early warning of landslide deformation induced by rainfall. Figure 2 This is a schematic diagram of the core process of data assimilation and updating. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0017] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this application. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0018] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present application. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0019] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0020] Example 1: like Figure 1-2As shown, this invention provides a method for early warning of rainfall-induced landslide deformation. Specifically, it first includes establishing a geometric model. Based on actual geological survey data and topographic measurement data, a three-dimensional geometric model is constructed, including the topographic surface, stratigraphic interfaces, and potential sliding surfaces. The topographic surface refers to the undulating outline of the landslide body exposed to the natural environment, encompassing key functional areas such as the catchment area at the top of the slope, the runoff zone on the slope, and the deposition area at the toe of the slope. It is the interface where rainfall and the landslide body first interact, and it is also the starting point for "rainfall infiltration simulation" in deformation early warning. The stratigraphic interface refers to the boundary between soil and rock masses with different mechanical properties within the landslide body, such as the boundary between cohesive soil layers and gravel layers, or weathered rock layers and fresh rock layers. The key parameters such as "permeability coefficient, cohesion, and internal friction angle" of the soil and rock masses on both sides of the interface are significantly different, which is the fundamental reason for the "non-uniform deformation" of the landslide body. The potential sliding surface refers to the weak interface of the soil and rock mass within the landslide body with the lowest strength and most susceptible to shear failure. It is mostly a weak interlayer, weathered mudstone layer, or a "saturated weak zone" formed by long-term seepage. It is the key occurrence surface of the landslide through continuous deformation accumulation until it finally becomes unstable and forms a landslide. It is also the core carrier of "stability coefficient Fs calculation" in this early warning method.

[0021] S1: Finite element simulation is used to discretize or refine the mesh of the constructed model components; that is, using finite element software (such as FLAC3D, ABAQUS), the previously constructed complete model of "terrain surface, strata, and potential sliding surface" is divided into countless "regular small units". Each unit has independent spatial coordinates and number, and each unit can be assigned values ​​individually, making it more consistent with the differences in real strata, and providing a basis for subsequent calculation of pore water pressure changes and rock and soil stress. Computers can also calculate the "displacement" of each unit through the mechanical relationship between units. For example, the displacement of the slope toe unit is 5mm and the displacement of the sliding zone unit is 10mm. After these discrete displacement data are summarized, the overall deformation distribution of the landslide body can be reconstructed in the computer. This deformation data is the key basis for "judging the risk trend" in the early warning. For example, if the deformation data accelerates, the risk of landslide increases.

[0022] During the discretization process, not all discrete units are the same size. Instead, for the "core areas that affect the early warning results", the units are broken down into smaller units, i.e., densified. For example, the predicted sliding zone area is densified to 0.5m, and the critical area at the toe of the slope is densified to 1.0m. For non-critical areas, larger units are needed to balance computational efficiency, such as setting the grid size to 2.0m.

[0023] S2: Establish a physical coupling model, including establishing the governing equations for the seepage field: in, , , The permeability coefficient is the tensor component, i.e., the permeability coefficient in the three directions of x, y, and z. Through the three-dimensional geometric model constructed above, and because the permeability of different strata or directions varies greatly, only differentiated assignment can simulate how water flows in complex strata. That is, borehole seepage tests are carried out on different strata to obtain the "layered and directional" permeability coefficients.

[0024] Unlike traditional early warning systems that often use fixed empirical values ​​for rainfall infiltration, this invention can dynamically update rainfall intensity by combining real-time weather forecast data, making the infiltration simulation more closely resemble the actual rainfall process and improving the foresight of dynamic early warning.

[0025] To determine the water storage rate, core samples were taken from the landslide body and measured using indoor geotechnical tests. This reflects the differences in water storage capacity of soil and rock at different test locations, avoiding distortions caused by water storage or release during the simulation process.

[0026] To simulate the time of the process, and based on the aforementioned topographic surface and stratum interface models, a more accurate definition was made of the "surface rainfall infiltration boundary" (such as the infiltration characteristics of the slope top catchment area and slope fissures) and the "seepage boundary of the stratum interface" (such as the water infiltration law of the interface between sand and gravel and cohesive soil).

[0027] This equation is used to calculate the water head at different times and locations within a landslide mass. The model is then assigned values ​​by computer to simulate the trend of water head increase or decrease in the landslide soil mass after rainfall.

[0028] Through the equation: The pore water pressure p at any point within the landslide can be calculated.

[0029] in It is pore water pressure. It is the density of water. It is gravitational acceleration. It is the location elevation, that is, the vertical position value of each point inside the landslide body where the pore water pressure needs to be calculated in the constructed three-dimensional geometric model.

[0030] According to the effective stress principle, the effective stress of the soil and rock mass at the landslide detection location is the one that truly plays a role in resisting landslides. (Effective stress) = (Total stress) - (Pore water pressure). It can be deduced that the higher the pore water pressure, the lower the effective stress, and the weaker the shear strength of the soil and rock mass. This shear strength is determined by the Mohr-Coulomb criterion. "It can be seen that cohesion" internal friction angle Relatedly, the weaker the shear strength, the more prone the landslide is to instability.

[0031] Therefore, pore water pressure is the core mechanical indicator for determining whether a landslide will deform and become unstable. In the effective stress principle, total stress... This includes the initial stress in its natural stable state and the total stress distribution under the combined action of external loads generated by external factors such as rainfall. The initial stress includes the vertical pressure of the soil and rock mass at the landslide location on that point. The downward tension along the slope caused by gravity in the soil and rock mass above the landslide location: The additional vertical pressure exerted on the slope by the weight of rainwater at the landslide location: The additional sliding pull generated by rainwater flowing along the slope at the landslide location: ,in The density of the soil and rock at that location. It is the acceleration due to gravity. Let be the cosine of the slope angle. Let be the sine of the slope angle. To calculate the depth of the point within the landslide body, The intensity of external loads, such as rainfall loads.

[0032] Finally, the initial stress and the components of the same direction and type from the external load stress are added together to obtain the total stress at the calculation point. .

[0033] When a landslide does not satisfy the "uniform, simple load" condition and is subjected to complex loads (seismic force, tectonic stress), the mechanical equilibrium is described by the stress field governing equations: in, Let τ be the components of the normal stress and shear stress in that direction. The density of the soil and rock at this location can be obtained by drilling core samples at the landslide site and conducting laboratory density tests on different strata of the landslide body to obtain the layered density values. This represents the component of gravitational acceleration in that direction.

[0034] In establishing a Cartesian coordinate system conforming to this invention, the x-axis direction is along the potential sliding direction of the landslide, such as downwards on the slope, and the angle between it and the horizontal direction is equal to the average slope angle of the landslide. The landslide slope angle Based on the digital surface model (DSM) obtained by UAV aerial survey, the average slope of the landslide can be extracted using ArcGIS software. The z-axis is vertically downward, consistent with the direction of gravity, and corresponds to the main direction of self-weight stress. The y-axis is perpendicular to the xz plane, that is, perpendicular to the direction of the landslide sliding surface. Since there is no significant sliding tendency and load in this direction, it is used as an auxiliary coordinate system.

[0035] Meanwhile, regarding shear stress, in the natural state, the slope has no relative sliding tendency, and there is no load in the y direction. Therefore, the shear stress related to y is: ; The landslide body is homogeneous in its strata, which can be verified by borehole exploration. Therefore, the stress distribution at the same depth is uniform, i.e., the partial derivatives of stress with respect to the x and y directions are: ; Considering only the effect of gravity, gravity is decomposed into x- and z-components (y-component) ),Right now (The force that drives the landslide downwards) (The component force that compresses the soil and rock mass).

[0036] Set boundary conditions: The landslide surface z=0 with no soil or rock cover and stress of 0 are the boundary conditions. Perform definite integral on the processed equation to obtain the stress value at any depth z=h. The calculation depth h can be combined with borehole exploration and inclinometer monitoring data to determine the burial depth of the potential sliding zone (e.g., 8m, 10m), which is the core calculation depth, i.e. the key area of ​​landslide instability.

[0037] The effective stress within the landslide body can be calculated by applying the principle that the landslide body must maintain a force balance under its own weight, external load, and internal stress. Considering the core triggers of rainfall-induced landslides (Pore water pressure), according to the effective stress principle: The stress values ​​calculated in different directions are corrected to obtain the effective normal stress actually acting on the soil and rock mass. For example, the normal stress in the z-direction is corrected as follows: S3: Establish a monitoring device network module, including 3 automatic rain gauges, which are set up at the top, middle and bottom of the slope to detect rainfall. The function of the automatic rain gauges is to accurately and continuously record the amount and intensity of rainfall in the monitoring area. One rain gauge is set up at the top, middle and bottom of the slope to capture the differences in rainfall at different slope positions. For example, the top of the slope may be more affected by wind and rain, while the bottom of the slope may have water accumulation, resulting in localized more rainfall.

[0038] Eight GNSS monitoring stations detect surface displacement. The role of the GNSS monitoring stations is to use satellite positioning technology to monitor the horizontal and vertical displacement of the monitoring points with high precision and record the time variation of the displacement. The setting of eight stations can cover key areas of the slope, such as weak sections that may slide and the edge of the slope top, forming a displacement monitoring network.

[0039] Four inclinometer boreholes are used to detect deep displacement. Inclinometer boreholes are exploration channels that go deep into the interior of the slope. Their function is to install inclinometers inside the boreholes and calculate the displacement of soil or rock layers at different depths by measuring the inclination changes of the borehole axis. The four inclinometer boreholes are usually placed in key sections of the slope that may be unstable, such as areas with thick slopes or historical signs of landslides.

[0040] Six piezometers are used to measure pore water pressure. The piezometers are buried in the soil or rock inside the slope to directly measure the pressure of pore water. The six piezometers are placed at different depths and in different permeable layers, such as near the groundwater level or in soil layers with high permeability, depending on the hydrological conditions of the slope.

[0041] S4: Establish a data assimilation dynamic update module In the data assimilation and prediction steps, the system is first initialized. For the initial displacement field, based on the initial monitoring data of the monitoring instruments deployed during the observation phase and the aforementioned three-dimensional geometric model, the initial displacement of each point inside the landslide body is determined. If the landslide has no obvious deformation under natural stable conditions, the initial displacement field is set to 0. For local micro-deformation areas discovered during the investigation, such as the shallow loosening area at the toe of the slope, combined with the initial readings of the monitoring instruments, the computer assigns small initial displacements to the corresponding spatial positions of the model, making the initial state more consistent with the non-uniform state of the landslide.

[0042] For the initial pore water pressure field: Combining the initial monitoring data of the borehole pore water pressure gauge with the basic parameters of the above seepage control equation, the spatial distribution of the initial pore water pressure is calculated. The initial water head obtained through exploration, i.e. the initial water level below the surface measured by the piezometer, is used to obtain the value through the above method, and the value is assigned in layers in the twin to ensure that the initial pore water pressure field is consistent with the actual formation seepage characteristics.

[0043] The initial state is transformed into a dynamic estimation variable by extending the state vector, which is defined as follows: Where u, v, and w are displacement components, p is pore water pressure, K is permeability coefficient, c is cohesion, and φ is internal friction angle. The permeability coefficient K, cohesion c, and internal friction angle φ are used as components of the state vector. At the same time, the state covariance matrix P is set.

[0044] During the prediction step, the system state needs to be advanced using a physical model: Each member of the set independently runs the physical model operator M[·], based on the current state and parameters. The set members, i.e., the set for predicting landslides, create dozens to hundreds of slightly different digital twin copies. Real landslides have a lot of uncertainties: for example, during exploration and sampling, the permeability coefficient K of the same stratum may fluctuate, and the internal friction angle may deviate. These small deviations, when calculated with a single twin using fixed parameters, may lead to misjudgment of the early warning due to parameter bias, such as calculating low risk when the actual risk is high. The role of digital twin copies is to cover these uncertainties with multiple copies, which is equivalent to simulating all possible real landslide states. Thus, combined with real-time rainfall data, the landslide state at the next moment can be predicted more accurately. The execution process of running the physical model operator M[·] includes the following steps: First, input the pore water pressure field data. That is, the pore water pressure at each point in the landslide body at time t, as well as the permeability coefficient k and rainfall infiltration. The above seepage control equation was solved to obtain... Water head distribution at different times Then, using the pore water pressure formula mentioned earlier: , transformed Pore ​​water pressure distribution at time t That is, the output of this step is Pore ​​water pressure distribution at time t .

[0045] Then input the soil and rock parameters using the method described above: density ,gravity slope angle Then through total stress and the principle of pore water pressure p combined with effective stress p” can obtain the effective stress As can be seen from the above, pore water pressure affects shear strength through the effective stress principle. By solving the stress field control equations in S2 using the finite element method in the computer, the shear stress and shear stress in different directions at different spatial points in the landslide body can be obtained. The stress values ​​of all nodes can then be aggregated into a stress distribution using software.

[0046] Strain is calculated using effective stress, first by combining the yield condition of the Mohr-Coulomb model: like < When the soil and rock mass is in the elastic stage, the strain consists only of elastic deformation; if ≥ When soil and rock masses enter the elastoplastic stage, the strain includes elastic strain + plastic strain.

[0047] Then, based on the "elastic / elastoplastic stage" and soil parameters ( Determine the elastic-plastic matrix. Then, the strain increment is obtained through the constitutive equation. : The strain is the deformation per unit length. To obtain the actual displacement of the landslide at that location, it is necessary to integrate the geometric equation, i.e., the differential relationship between strain and displacement. The core form of the geometric equation, taking a one-dimensional slope along the x-axis as an example, shows that the landslide mainly slides along the x-direction. Let be the strain along the x-direction. For the displacement along the x-direction, the boundary condition is: if the initial displacement at the landslide surface (x=0) is... The integral result is "the cumulative strain from the surface to depth x".

[0048] Finally, the displacement results of the landslide body at "different depths and different horizontal positions" are integrated to form the deformation field.

[0049] Next, a convergence check of the coupled iterations is performed, with the following specific steps: The system determines convergence by comparing the relative changes in the results of two adjacent iterations. The convergence criterion for the pore water pressure field is as follows: The convergence criterion for the displacement field is: in, and Let these represent the pore water pressure vector and displacement vector in the k-th iteration, respectively. and The preset convergence tolerance is typically set to a value of [value to be filled in]. .

[0050] The model performs convergence checks according to the following process: Calculate the relative change norm of the pore water pressure field between the current iteration step and the previous iteration step; calculate the relative change norm of the displacement field between the current iteration step and the previous iteration step; determine whether the two relative change norms are simultaneously less than the corresponding convergence tolerance; if the convergence condition is met, the iteration is terminated and the next calculation step is initiated; if the convergence condition is not met and the number of iterations has not reached the upper limit, the model returns to the seepage field solution step to continue iterating; if the number of iterations reaches the preset upper limit and convergence is still not achieved, an error is reported and the exception handling mechanism is initiated.

[0051] The convergence criterion, or convergence tolerance setting, is based on a balance between engineering accuracy requirements and computational efficiency, ensuring that a physically compatible solution meeting engineering accuracy requirements is obtained within a reasonable computation time. This convergence checking mechanism guarantees the numerical stability of the seepage-stress coupling solution process and the physical rationality of the calculation results.

[0052] For each set member i, the mathematical expression of the prediction step is: in, This is the model error term, representing the imperfections of the physical model. This represents the predicted state vector of the i-th set member at time t+Δt, which includes the predicted displacement field, pore water pressure field, and material parameter field. The superscript f means "prediction" and the subscript i means the i-th set member. This represents the analytical state vector of the i-th set member at time t. The superscript a indicates "analysis", which is the optimal state after data assimilation correction, including the latest estimates of the displacement field, pore water pressure field and material parameter field.

[0053] Correction Step: First, based on the predicted displacement, pore water pressure, cohesion c, internal friction angle, and other states and parameters from the prediction step, calculate the theoretically measurable values ​​for GNSS, inclinometers, and piezometers using the model; then compare these theoretical values ​​with the actual rainfall, displacement, and pore water pressure data monitored by the instruments to obtain the deviation; finally, use the Kalman gain formula... in The prediction error covariance matrix is ​​calculated using ensemble statistical methods to characterize the uncertainty of the model prediction; R is the observation error covariance matrix, determined based on sensor accuracy, to characterize the uncertainty of the observation data; H is the observation operator, which realizes the mapping from the state space to the observation space.

[0054] In the Kalman filter update process, a unified mathematical framework is used to synchronously correct the state variables and soil parameters. The update formula is as follows: The final corrected state Predicted state, K: Kalman gain, : Actual observed data, Theoretical observations calculated using predicted states; : Observation error.

[0055] When a discrepancy arises between model predictions and field monitoring data, the system adjusts the estimated values ​​of state variables and optimizes soil and rock parameters in reverse based on physical mechanisms. For example, when the predicted displacement is consistently less than the observed displacement, the estimated values ​​of cohesion and internal friction angle are automatically reduced through the following calculation process: in and These are the components in the Kalman gain matrix corresponding to parameters c and φ. and These are predicted values.

[0056] Considering the uncertainty of the prediction results and the errors in the monitoring data itself, we calculate the extent to which the prediction should be corrected based on the monitoring data, i.e., the weight K value. Finally, we use this weight and the deviation to adjust the displacement, pore water pressure, cohesion, internal friction angle, etc., so that they are closer to the actual situation on site.

[0057] Next, parameter constraints are applied, imposing physical constraints on the updated parameters: permeability coefficient K > 0 to ensure the seepage direction conforms to physical reality; internal friction angle 10° ≤ φ ≤ 40°, covering a typical range from soft clay to dense sand; cohesion c ≥ 5 kPa to avoid unrealistically low strength values. The constraint range is set based on extensive geotechnical engineering practice experience.

[0058] S4. Implementation of early warning decisions: For the updated digital twin, perform stability analysis and calculate the stability coefficient Fs. Where c' is the effective cohesion, and in the static model, c' is a fixed value determined experimentally. However, in the innovation of this invention, c' is a dynamic state variable. It is contained in the extended state vector X = [u,v,w,p,K,c,φ]^T, and is continuously corrected and updated based on real-time monitored data such as displacement and pore water pressure through a Kalman filter data assimilation process, so that the digital twin is closer to the real state of the landslide. The c' obtained at this time may exceed the physically reasonable range. After parameter constraint processing, the final c' used is obtained. σ' is the effective normal stress, n is the normal vector, which is a direction vector used to extract the normal stress component perpendicular to the sliding surface in the stress tensor. A is the area, that is, the area of ​​the sliding surface at the bottom of a single soil strip. In the three-dimensional model, this is a micro-element area on the sliding surface. The unit is square meters (m²). W is the weight of the sliding surface, referring to the weight of a single soil strip. α is the inclination angle of the sliding surface, referring to the inclination angle of the potential sliding surface at the position of the soil strip (the angle with the horizontal plane).

[0059] A graded early warning mechanism was established based on the calculated stability coefficient Fs. A yellow warning was issued when Fs < 1.05 for 12 consecutive hours, indicating that the landslide had entered a state of extreme equilibrium, with stability approaching the critical value and a significant risk of instability. An orange warning was issued when Fs < 1.00 for 6 consecutive hours, indicating that the landslide was in a critical state of instability, with safety reserves exhausted and a clear trend of accelerated deformation. A red warning was issued when Fs < 0.95 and the displacement rate > 10 mm / d, indicating that the landslide had entered a stage of accelerated deformation, with obvious signs of imminent collapse and the possibility of failure in the short term.

[0060] In response to the aforementioned warning grading system, a yellow warning system will be implemented with enhanced patrols, increasing the frequency to twice a day; an orange warning will restrict activities in dangerous areas and initiate professional monitoring; and a red warning will prompt immediate evacuation and activation of emergency plans.

[0061] The above are merely embodiments of the present invention. The circuits, electronic components, and modules involved are all prior art, fully achievable by those skilled in the art, and require no further explanation. The content protected by this application does not involve improvements to the software and methods. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all prior art in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the application.

Claims

1. A method for early warning of rainfall-induced landslide deformation, characterized in that, Includes the following steps: S1. Establish a three-dimensional geometric model, including constructing a three-dimensional geometric model based on geological data that includes the topographic surface, stratigraphic interface and potential slip surface. At the same time, the finite element method is used to discretize the mesh, and the mesh is refined in the slip zone and slope toe area. S2. Establish a physical coupling model, including establishing the seepage field control equation and the stress field control equation, and then realize bidirectional coupling through the effective stress principle; S3. Establish a monitoring device network module, including deploying monitoring devices at different points on the landslide site, including rain gauges, GNSS displacement monitoring stations, inclinometers and pore water pressure gauges, to monitor rainfall, displacement and groundwater level data; S4. Establish a data assimilation and dynamic update module, including using ensemble Kalman filter data assimilation technology to generate multiple ensemble members, and realize the real-time update of the state variables of the digital twin through prediction and update steps; S4. Establish a tiered early warning module, including stability simulation based on the updated digital twin, inputting future rainfall forecast data, calculating the stability coefficient Fs, and establishing a three-level early warning mechanism.

2. The method for early warning of rainfall-induced landslide deformation as described in claim 1, characterized in that: The digital twin is modeled using geological exploration data. The three-dimensional geometric model is discretized using the finite element method. The discretization process requires fine-tuning of the locations of the sliding zone and toe area in the landslide body. The digital twin includes the construction of seepage field control equations and stress field control equations.

3. The method for early warning of rainfall-induced landslide deformation as described in claim 2, characterized in that: The seepage field control equation and the stress field control equation are bidirectionally coupled through the effective stress principle. This coupling process incorporates extended state vectors for the displacement field, pore water pressure field, and material parameter field. These extended state vectors are defined as follows: In this vector, u, v, and w are displacement components, p is pore water pressure, K is the permeability coefficient, c is cohesion, and φ is the internal friction angle. The permeability coefficient K, cohesion c, and internal friction angle φ are used as components of the state vector.

4. The method for early warning of rainfall-induced landslide deformation as described in claim 1, characterized in that: The Kalman filter data assimilation process includes generating multiple set members, and the prediction equation in the Kalman filter data assimilation process is: M[·] is the physical model operator, and the For the model error term, the Let K represent the predicted state vector of the i-th set member at time t+Δt, where K is the Kalman gain.

5. The method for early warning of rainfall-induced landslide deformation as described in claim 4, characterized in that: The formula for calculating the Kalman gain is as follows: ,in For the prediction error covariance matrix, The observation error covariance matrix, For observation operators.

6. The method for early warning of rainfall-induced landslide deformation as described in claim 5, characterized in that: The Kalman gain K performs state correction on the existing actual value, and the corrected value is used to update the predicted curve.

7. The method for early warning of rainfall-induced landslide deformation as described in claim 1, characterized in that: The parameter constraint processing specifically includes: constraining the permeability coefficient to a positive value within a preset range; constraining the internal friction angle within a preset angle range; constraining the cohesion to be no less than a preset lower limit of strength; applying smoothness constraints to the soil and rock parameters of adjacent units within the same stratum; applying engineering experience constraints to the soil and rock parameters of the sliding zone region; and adopting a graded processing strategy according to the degree of parameter violation of constraints.

8. The method for early warning of rainfall-induced landslide deformation as described in claim 1, characterized in that: The tiered early warning system calculates the stability coefficient Fs based on the updated digital twin and establishes a three-tiered early warning mechanism based on the stability coefficient.

Images