A method and device for early warning of dynamic stability of landslide based on slope surface distortion

Abstract

The application provides a landslide dynamic stability early warning method and device based on slope surface distortion, relates to the technical field of landslide monitoring and early warning, and comprises the following steps: obtaining rock-soil mechanics parameters and slope body geological structure; calculating a slope heterogeneity index of the slope body geological structure and topographic mapping data, dividing a distortion sensitive area through the heterogeneity index, and constructing a targeted layout scheme; constructing a slope model based on the targeted layout scheme and the rock-soil mechanics parameters, searching for a potential sliding surface through the slope model, and critically screening to construct a slope stability sample pair; inputting a feature vector of the slope stability sample pair into a convolutional neural network for training, combining a triple physical constraint mechanism to construct, and obtaining an early warning model; inputting the slope stability sample pair and real-time targeted displacement data into the early warning model to inverse the sliding surface position, and triggering a hierarchical early warning through the sliding surface position. The application solves the problem that landslide risks cannot be timely warned.

Description

Technical Field

[0001] This invention relates to the field of landslide monitoring and early warning technology, and more specifically, to a method and device for early warning of landslide dynamic stability based on slope distortion. Background Technology

[0002] In existing landslide monitoring and early warning technologies, the deployment of monitoring points typically relies on experience or uniform distribution principles, lacking analysis of the slope's internal structure, geological formations, and historical deformation characteristics. Monitoring networks struggle to cover key deformation areas, resulting in insufficient data representativeness. Traditional numerical simulations for dynamic stability evaluation are computationally complex and time-consuming, heavily reliant on manual labor and professional experience. Furthermore, model parameters are largely based on static assumptions, making real-time updates difficult and failing to reflect the spatial and temporal characteristics of landslide instability mechanisms. The spatiotemporal sparsity of monitoring data limits the model's accurate representation of slip surface formation and evolution, impacting the timeliness and accuracy of early warnings and leading to the inability to promptly warn of landslide risks.

[0003] Therefore, there is an urgent need for a landslide dynamic stability early warning method and device based on slope distortion, which solves the problem of the inability to provide timely early warning of landslide risks. Summary of the Invention

[0004] The purpose of this invention is to provide a method and device for early warning of landslide dynamic stability based on slope distortion, so as to improve the above-mentioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0005] Firstly, this application provides a landslide dynamic stability early warning method based on slope distortion, including:

[0006] Obtain geotechnical parameters and slope geological structure;

[0007] The slope heterogeneity index is calculated based on the geological structure and topographic mapping data of the slope. The distortion-sensitive area is divided by the heterogeneity index, and a targeted deployment scheme is constructed.

[0008] Based on the targeted deployment scheme and the geotechnical parameters, a slope model is constructed. By searching for potential slip surfaces and critically screening the slope model, slope stability sample pairs are constructed.

[0009] The feature vectors of the slope stability sample pairs are input into a convolutional neural network for training, and a warning model is obtained by combining a triple physical constraint mechanism.

[0010] The slope stability sample pairs and real-time target displacement data are input into the early warning model to invert the slip surface location, and the slip surface location triggers a graded early warning.

[0011] Secondly, this application also provides a landslide dynamic stability early warning device based on slope distortion, comprising:

[0012] The acquisition module is used to acquire geotechnical mechanical parameters and slope geological structure;

[0013] The calculation module is used to calculate the slope heterogeneity index based on the slope geological structure and topographic mapping data, and to divide the distortion-sensitive area through the heterogeneity index to construct a targeted deployment scheme.

[0014] The construction module is used to construct a slope model based on the targeted layout scheme and the geotechnical parameters, and to construct slope stability sample pairs by searching for potential sliding surfaces and critically screening the slope model.

[0015] The training module is used to input the feature vectors of the slope stability sample pairs into the convolutional neural network for training, and to construct an early warning model by combining a triple physical constraint mechanism.

[0016] The inversion module is used to input the slope stability sample pairs and real-time target displacement data into the early warning model to invert the slip surface location, and trigger graded early warnings based on the slip surface location.

[0017] The beneficial effects of this invention are as follows:

[0018] This invention delineates distortion-sensitive areas using heterogeneity indicators, enabling targeted deployment of monitoring equipment and avoiding the blind spots and resource waste associated with traditional experience-based deployment. Building upon this, a convolutional neural network training method under a triple physical constraint mechanism is employed, endowing the model with data-driven learning capabilities and integrating slope mechanics mechanisms, significantly improving the stability and generalization ability of the early warning model. Subsequently, by leveraging real-time displacement data and sample pairs to invert the location of potential slip surfaces, it overcomes the limitations of traditional methods that rely solely on displacement thresholds, achieving a shift from superficial monitoring to mechanism identification. Based on the evolution of the slip surface, multi-level early warning mechanisms can be triggered, effectively addressing landslide risks. In summary, this invention solves the problem of the inability to provide timely early warnings of landslide risks.

[0019] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the landslide dynamic stability early warning method based on slope distortion as described in an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of the landslide dynamic stability early warning device based on slope distortion as described in an embodiment of the present invention.

[0023] The markings in the diagram are: 800, landslide dynamic stability early warning device based on slope distortion; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0026] Example 1:

[0027] This embodiment provides a method for early warning of landslide dynamic stability based on slope distortion.

[0028] See Figure 1 The figure shows that the method includes steps S1 to S5, including:

[0029] S1: Obtain geotechnical parameters and slope geological structure;

[0030] S2: Calculate the slope heterogeneity index based on the geological structure and topographic mapping data of the slope, divide the distortion-sensitive area through the heterogeneity index, and construct a targeted deployment scheme;

[0031] This step uses heterogeneity indicators to delineate distortion-sensitive areas, enabling targeted deployment of monitoring equipment and avoiding the blind spots and resource waste associated with traditional experience-based deployment.

[0032] To clarify the specific method for obtaining the targeted deployment plan, step S2 includes S21 to S24, specifically:

[0033] S21: By comprehensively judging the sliding mechanism based on the geological structure of the slope and the topographic mapping data, the landslide deformation zone is obtained;

[0034] In this step, the sliding mechanism is comprehensively judged based on the geological structure of the slope and the topographic mapping data. Combined with the landslide type results, the landslide deformation influence range is delineated, and the potential main sliding direction, sliding surface development area and key deformation parts are identified to obtain the landslide deformation zone.

[0035] S22: Extract the topographic curvature and lithological differences of adjacent rock layers based on the landslide deformation zone to obtain heterogeneity indices;

[0036] To clarify the specific method for obtaining the heterogeneity index, step S22 includes S221 to S223, which are as follows:

[0037] S221: Based on the landslide deformation zone, the curvature of adjacent rock layers is calculated using a preset surface elevation model to obtain the topographic curvature;

[0038] This step calculates the second derivative of curvature of adjacent rock layers based on the landslide deformation zone using a preset surface elevation model to obtain the topographic curvature. The larger the absolute value of the curvature, the more significant the topographic abrupt change, and the more likely the slope is to experience stress concentration and local damage. The topographic curvature reflects the sensitivity of the slope to concave and convex deformation.

[0039] S222: Extract the lithological combination and mechanical parameters of adjacent rock layers based on the borehole data in the landslide deformation zone to obtain the lithological difference coefficient;

[0040] In this step, the lithological combination and mechanical parameters of adjacent rock layers are extracted based on the borehole data in the landslide deformation zone, the strength abrupt change of adjacent rock layers is quantified, and the lithological difference coefficient is obtained. The larger the value of the lithological difference coefficient, the more significant the lithological difference, and the easier it is for the structural surface to form a potential slip zone.

[0041] S223: Based on the topographic curvature and the lithological difference coefficient, a heterogeneity index is obtained.

[0042] S23: Based on the heterogeneity index, the calculation is performed in the preset distortion-sensitive weight calculation model to obtain the distortion-sensitive weight parameters;

[0043] In this step, the heterogeneity index is input into the preset distortion sensitivity weight calculation model to calculate the distortion sensitivity weight parameter for each spatial point, which is used to measure the sensitivity of that point to the overall deformation of the slope.

[0044] The expression for the distortion-sensitive weight parameter is:

[0045] (1);

[0046] In the above formula (1), These are distortion-sensitive weighting parameters; For terrain curvature; This represents the lithological difference coefficient.

[0047] S24: Divide the distortion-sensitive region according to the distortion-sensitive weight parameter, and construct the targeted monitoring point layout through the distortion-sensitive region to obtain the targeted layout scheme.

[0048] In this step, when the distortion-sensitive weight parameter is When the value is greater than 0.75, the area is determined to be a first-level distortion-sensitive area, and dual-frequency GNSS monitoring points are deployed in a 50m × 50m grid; when the distortion sensitivity weight parameter is less than 0.5... When the value is ≤0.75, the area is determined to be a secondary sensitive area, and points are placed according to a 100m × 100m grid; when the distortion sensitivity weight parameter is... When the value is ≤0.5, the area is determined to be a non-sensitive area and no monitoring points are deployed.

[0049] The distortion-sensitive area is used to optimize the allocation of monitoring resources, ensuring that displacement data can accurately capture key parts of slope deformation, and solving the problems of insufficient representativeness and blind spots in monitoring data caused by traditional uniform point distribution. The primary distortion-sensitive area includes the hanging wall traction section and the outcrop of weak interlayers, while the secondary sensitive area is the rock strata dip angle variation zone.

[0050] S3: Based on the targeted layout scheme and the geotechnical parameters, construct a slope model, and construct slope stability sample pairs by searching for potential sliding surfaces and critical screening of the slope model;

[0051] To clarify the specific method for obtaining slope stability sample pairs, step S3 includes S31 to S36, specifically:

[0052] S31: Based on the target deployment scheme, the monitoring points are spatially deployed, and the slope model is constructed by combining the peak-residual interval of the geotechnical parameters.

[0053] In this step, based on the primary and secondary distortion-sensitive areas in the targeted deployment scheme, targeted monitoring points are deployed according to a grid precision of 50m×50m or 100m×100m to form a spatial coordinate monitoring network. Simultaneously, combining the engineering geological survey report and indoor geotechnical test data, the peak strength and residual strength of the slip zone soil are determined, and a reasonable variation range for the geotechnical mechanical parameters (i.e., the peak-residual range) is constructed. Finally, by integrating topographic data, geological structure, and monitoring point coordinates, an initial slope model is established.

[0054] S32: Import the slope profile of the slope model into the two-dimensional slope stability analysis model to search for potential slip surfaces and obtain multiple sets of potential slip surfaces;

[0055] In this step, the slope profile of the slope model is imported into a two-dimensional slope stability analysis model (Slide). Based on limit equilibrium theory, potential slip surfaces at different locations are automatically searched. The default search count is no less than 500 groups to ensure coverage of potential failure modes in all parts of the slope. The search results include the geometry of each slip surface, its safety factor Fs, and its corresponding slip surface path coordinates.

[0056] S33: Based on a preset safety factor threshold, the multiple groups of potential slip surfaces are screened for critical slip surfaces to obtain a monitoring point displacement dataset;

[0057] To clarify the specific method for obtaining the monitoring point displacement dataset, step S3 includes S331 to S334, specifically:

[0058] S331: Based on a preset safety factor threshold, the multiple sets of potential slip surfaces are screened for critical slip surfaces to obtain a set of spatial location coordinates;

[0059] In this step, based on a preset safety factor threshold of Fs≤1.2, the multiple groups of potential slip surfaces are screened to extract those in critical or unstable states. For each critical slip surface, the two-dimensional slope stability analysis model (Slide) automatically outputs the coordinates of key nodes on its geometric path, forming a spatial coordinate set.

[0060] S332: Based on the discrete element model, the spatial coordinate set is transformed and assigned different shear strength parameters to form a shear strength parameter set;

[0061] In this step, the spatial coordinate set is input into the discrete element model to define joints or contact surfaces, thereby characterizing the potential weak slip surface. Subsequently, based on the peak-residual interval, multiple sets of shear strength parameters are generated using Latin hypercube to form a shear strength parameter set. Each set of parameters represents the mechanical state of the slip zone soil at different weakening stages (stable, creep, critical instability).

[0062] S333: Based on the shear strength parameter set, perform discrete element numerical simulation, calculate the deformation and failure process of the slope under different strength conditions through simulation results, and obtain the displacement field of the entire slope;

[0063] In this step, based on the shear strength parameter set, a non-continuous medium numerical simulation is performed in the discrete element model UDEC to simulate the deformation evolution of the slope during the gradual weakening of its strength until the model converges. The output is the displacement field data of the entire slope as the whole slope displacement field. The displacement field data of the entire slope includes the displacement vector and resultant displacement scalar of each grid node at different time steps, and fully records the entire process response of the slope from stability to instability.

[0064] S334: Based on the targeted deployment scheme, a monitoring point displacement dataset is constructed by extracting the combined displacement time series data of the monitoring points from the entire slope displacement field.

[0065] In this step, based on the coordinates of the targeted monitoring points deployed in the targeted deployment scheme, the time series data of the corresponding locations (i.e., the displacement scalar of each monitoring point in each calculation step) are extracted from the full slope displacement field to form a monitoring point displacement dataset. The monitoring point displacement dataset reflects the deformation response of key parts of the slope under specific slip surface and strength conditions.

[0066] S34: Normalize the displacement dataset of the monitoring points and construct a three-dimensional displacement curve matrix;

[0067] In this step, the monitoring point displacement dataset is normalized to eliminate dimensional differences and unify the numerical range. Subsequently, the normalized displacement curves of all the monitoring point displacement datasets are aligned by time steps to construct a three-dimensional displacement curve matrix.

[0068] S35: Extract the overall shape and trend of the curve from the displacement dataset of the monitoring points, and calculate the slope distortion index.

[0069] In this step, based on the normalized displacement curves, the overall morphological features of the monitoring point displacement dataset are extracted. Principal component analysis (PCA), curve fitting, or feature engineering methods are used to quantify the overall deformation coordination, local abrupt changes, and spatiotemporal evolution of the slope, constructing a slope distortion degree index. This slope distortion degree index is used as channel two of the input features to characterize the macroscopic deformation state of the slope, improving the physical interpretability of the model for the slip surface position-intensity inversion.

[0070] S36: Based on the three-dimensional displacement curve matrix and the slope distortion index, pairing is performed to obtain slope stability sample pairs.

[0071] In this step, the three-dimensional displacement curve matrix and the slope distortion index are used as input features, and the corresponding spatial coordinate set and shear strength parameter set are used as output labels. A complete slope stability sample pair that can be used for supervised learning is constructed.

[0072] S4: Input the feature vectors of the slope stability sample pairs into a convolutional neural network for training, and combine them with a triple physical constraint mechanism to construct an early warning model;

[0073] In this step, a convolutional neural network training method under a triple physical constraint mechanism is adopted to endow the model with data-driven learning capabilities. Combined with the slope mechanics mechanism, the stability of the early warning model is improved.

[0074] To clarify the specific method for obtaining the early warning model, step S4 includes S41 to S45, specifically:

[0075] S41: Input the displacement curve matrix and slope distortion index of the slope stability sample pair into the convolutional neural network in parallel for multidimensional prediction to obtain landslide surface prediction parameters.

[0076] In this step, the displacement curve matrix and slope distortion index are used as dual-channel data for feature construction. The displacement curve matrix is ​​used to characterize the temporal evolution of the displacement scalar at each point, and the distortion index is used to describe the overall distortion state of the slope. The convolutional neural network adopts a CNN-BiLSTM hybrid architecture.

[0077] First layer: 1D-CNN, with 32 one-dimensional convolutional kernels of width 7, which slide along the time dimension to extract local abrupt features in the displacement curve;

[0078] The second layer: BiLSTM, with 64 hidden units, jointly models forward and backward to capture long-term temporal dependencies in the displacement field evolution;

[0079] Output layer: a four-branch fully connected structure that predicts the sliding surface position encoding, cohesion, and internal friction angle respectively.

[0080] S42: Apply a boundary penalty term to the landslide surface prediction parameters according to the triple physical constraint mechanism to obtain the physical enhancement parameters;

[0081] In this step, a boundary penalty term is applied to the geotechnical parameters output by the landslide surface prediction parameters according to the first layer of the triple physical constraint mechanism—parameter rationality constraint. When the predicted value exceeds the reasonable range given by the geological survey, a penalty term is added to the loss function to construct the physical enhancement parameters.

[0082] S43: Add total variational regularization to the displacement curve matrix to obtain a smooth displacement field;

[0083] In this step, the displacement curve matrix is ​​subjected to total variational regularization through the second layer of the triple physical constraint mechanism—the displacement field smoothness constraint—to obtain a smooth displacement field, which effectively suppresses non-physical oscillations in the displacement curve.

[0084] S44: Based on the focus loss, the sliding frequency of different geometric shapes in the convolutional neural network is enhanced to obtain the physical perception focus loss;

[0085] In this step, the third layer of the triple physical constraint mechanism—class imbalance correction—is adopted. Since different slip surface geometries appear with different frequencies in the training set, the model is prone to insufficient ability to identify a few classes. The focus loss reduces the loss contribution of easily classified samples by adjusting the weight factor and balance factor, strengthens the learning of slip surface classes in small samples, and effectively improves the model's sensitivity to rare but dangerous slip surface types.

[0086] S45: Based on the physical enhancement parameters, the smooth displacement field, and the physical sensing focus loss, an early warning model is constructed.

[0087] In this step, the physical enhancement parameters, the smooth displacement field, and the physical sensing focus loss are integrated into a unified loss function to complete model training and optimization, and output an early warning model.

[0088] S5: Input the slope stability sample pairs and real-time target displacement data into the early warning model to invert the slip surface location, and trigger graded early warning based on the slip surface location.

[0089] This invention constructs a complete system extending from monitoring apparent slope displacement to identifying internal failure mechanisms. Based on the evolution of the slip surface, a multi-level early warning mechanism can be triggered, thereby effectively addressing landslide risks. Compared to the traditional delayed strategy of alarming when a single point displacement exceeds a threshold, this invention effectively alleviates the problems of delayed early warning and high false alarm rates in traditional methods, significantly improving the timeliness and accuracy of landslide risk early warning.

[0090] To clarify the specific procedures for tiered early warning, step S5 includes S51 to S55, specifically:

[0091] S51: Calculate the time-series combined displacement scalar from the three-dimensional displacement data and real-time target displacement data in the slope stability sample pair, and construct the combined displacement matrix;

[0092] In this step, real-time three-dimensional displacement data of the targeted monitoring points are collected at a frequency of minutes for the slope stability sample. The slope stability sample is then processed synchronously in combination with historical data. The resultant displacement scalar is calculated for each monitoring point, and a resultant displacement matrix is ​​constructed.

[0093] S52: Extract morphological features from the combined displacement matrix to obtain the slope distortion index vector;

[0094] In this step, the combined displacement matrix is ​​extracted based on temporal morphological analysis and spatial synergy analysis methods to obtain the slope distortion index vector;

[0095] S53: Input the combined displacement matrix and the slope distortion index vector into the early warning model for learning time-series evolution to obtain the slip surface position parameters;

[0096] In this step, the combined displacement matrix is ​​used as channel one, and the slope distortion index vector is used as channel two. The combined displacement matrix and the slope distortion index vector are input into the early warning model for processing. Through time-series evolution learning, the similarity between the current displacement pattern and historical landslide cases is identified, and the sliding surface location parameters are output to realize the sliding surface location inversion and intensity parameter estimation.

[0097] S54: Based on the slip surface position parameters, perform a two-dimensional coordinate set transformation, and calculate the confidence score by combining the slip surface occurrence probability distribution of the early warning model;

[0098] In this step, the slip surface position encoding of the convolutional neural network is decoded based on the slip surface position parameters, and the 128-bit binary vector is restored to a two-dimensional coordinate sequence to form the slip surface geometric path. Combined with the slip surface occurrence probability distribution of the early warning model, the confidence score of the slip surface position is calculated. When the confidence score is <0.6, it is judged as low confidence and a manual review mechanism is triggered. When the confidence score is ≥0.6, the next level of early warning judgment process is entered.

[0099] S55: Based on the real-time overall safety factor in the slip surface position parameters and the credibility score, the slope distortion type is classified, and a graded early warning is triggered by the classification level.

[0100] Example 2:

[0101] This embodiment provides a landslide dynamic stability early warning device based on slope distortion, the device comprising:

[0102] The acquisition module is used to acquire geotechnical mechanical parameters and slope geological structure;

[0103] The calculation module is used to calculate the slope heterogeneity index based on the slope geological structure and topographic mapping data, and to divide the distortion-sensitive area through the heterogeneity index to construct a targeted deployment scheme.

[0104] To clarify the specific methods for obtaining the calculation module, the following are included:

[0105] The judgment unit is used to comprehensively judge the sliding mechanism based on the geological structure of the slope and the topographic mapping data to obtain the landslide deformation zone.

[0106] The first extraction unit is used to extract the topographic curvature and lithological difference characteristics of adjacent rock layers based on the landslide deformation zone, and obtain heterogeneity indicators.

[0107] The calculation unit is used to calculate the distortion-sensitive weight parameters by inputting the heterogeneity index into a preset distortion-sensitive weight calculation model.

[0108] The partitioning unit is used to divide the distortion-sensitive region according to the distortion-sensitive weight parameter, and to construct the targeted monitoring point layout through the distortion-sensitive region to obtain the targeted layout scheme.

[0109] The construction module is used to construct a slope model based on the targeted layout scheme and the geotechnical parameters, and to construct slope stability sample pairs by searching for potential sliding surfaces and critically screening the slope model.

[0110] To clarify the specific methods for obtaining the building modules, the following are included:

[0111] The deployment unit is used to spatially deploy monitoring points based on the targeted deployment scheme, and to construct a slope model by combining the peak-residual interval of the geotechnical parameters.

[0112] The import unit is used to import the slope profile of the slope model into the two-dimensional slope stability analysis model to search for potential slip surfaces and obtain multiple sets of potential slip surfaces.

[0113] The filtering unit is used to perform critical slip surface screening on the multiple groups of potential slip surfaces according to a preset safety factor threshold, so as to obtain a monitoring point displacement dataset.

[0114] The normalization unit is used to normalize the monitoring point displacement dataset and construct a three-dimensional displacement curve matrix.

[0115] The second extraction unit is used to extract the overall shape and trend of the curve from the monitoring point displacement dataset and calculate the slope distortion index.

[0116] The pairing unit is used to pair slope stability sample pairs based on the three-dimensional displacement curve matrix and the slope distortion index.

[0117] The training module is used to input the feature vectors of the slope stability sample pairs into the convolutional neural network for training, and to construct an early warning model by combining a triple physical constraint mechanism.

[0118] To clarify the specific methods for obtaining the training modules, the following are included:

[0119] The multidimensional prediction unit is used to input the displacement curve matrix and slope distortion index of the slope stability sample pair into the convolutional neural network in parallel to perform multidimensional prediction and obtain landslide surface prediction parameters.

[0120] Boundary units are applied to apply boundary penalty terms to the landslide surface prediction parameters according to the triple physical constraint mechanism, so as to obtain physical enhancement parameters.

[0121] Regularized units are used to add total variational regularization to the displacement curve matrix to obtain a smooth displacement field;

[0122] The enhancement unit is used to enhance the sliding frequencies of different geometric shapes in the convolutional neural network based on the focus loss, so as to obtain the physical perception focus loss.

[0123] The early warning construction unit is used to construct an early warning model based on the physical enhancement parameters, the smooth displacement field, and the physical sensing focus loss.

[0124] The inversion module is used to input the slope stability sample pairs and real-time target displacement data into the early warning model to invert the slip surface location, and trigger graded early warnings based on the slip surface location.

[0125] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0126] Example 3:

[0127] Corresponding to the above method embodiments, this embodiment also provides a landslide dynamic stability early warning device based on slope distortion. The landslide dynamic stability early warning device based on slope distortion described below and the landslide dynamic stability early warning method based on slope distortion described above can be referred to in correspondence.

[0128] Figure 2 This is a block diagram illustrating a landslide dynamic stability early warning device 800 based on slope distortion, according to an exemplary embodiment. Figure 2 As shown, the landslide dynamic stability early warning device 800 based on slope distortion may include: a processor 801 and a memory 802. The landslide dynamic stability early warning device 800 based on slope distortion may also include one or more of the following: a multimedia component 803, an I / O interface 804, and a communication component 805.

[0129] The processor 801 controls the overall operation of the landslide dynamic stability early warning device 800 based on slope distortion to complete all or part of the steps in the aforementioned landslide dynamic stability early warning method based on slope distortion. The memory 802 stores various types of data to support the operation of the landslide dynamic stability early warning device 800. This data may include, for example, instructions for any application or method operating on the landslide dynamic stability early warning device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the landslide dynamic stability early warning device 800 based on slope distortion and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0130] In an exemplary embodiment, the landslide dynamic stability early warning device 800 based on slope distortion can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the aforementioned landslide dynamic stability early warning method based on slope distortion.

[0131] Example 4:

[0132] Corresponding to the above method embodiments, this embodiment also provides a medium. The medium described below can be referred to in conjunction with the landslide dynamic stability early warning method based on slope distortion described above.

[0133] A medium storing a computer program, which, when executed by a processor, implements the steps of the landslide dynamic stability early warning method based on slope distortion as described in the above method embodiments.

[0134] The medium can specifically be any medium capable of storing program code, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0135] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0136] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of landslide dynamic stability based on slope distortion, characterized in that, include: Obtain geotechnical parameters and slope geological structure; The slope heterogeneity index is calculated based on the geological structure and topographic mapping data of the slope. The distortion-sensitive area is divided by the heterogeneity index, and a targeted deployment scheme is constructed. The specific methods for obtaining the slope heterogeneity index include: Based on the borehole data in the landslide deformation zone, the lithological combination and mechanical parameters of adjacent rock layers are extracted, the strength abrupt change of adjacent rock layers is quantified, and the lithological difference coefficient is obtained. The larger the value of the lithological difference coefficient, the more significant the lithological difference, and the easier it is for the structural surface to form a potential slip zone. Based on the topographic curvature and lithological difference characteristics of adjacent rock layers in the landslide deformation zone, the heterogeneity index is obtained. Based on the targeted deployment scheme and the geotechnical parameters, a slope model is constructed. By searching for potential slip surfaces and critically screening the slope model, slope stability sample pairs are constructed. The specific methods for obtaining the slope stability sample pairs include: Based on the targeted deployment scheme, monitoring points are spatially deployed, and the slope model is constructed by combining the peak-residual interval of the geotechnical parameters. The slope profile of the slope model is imported into the two-dimensional slope stability analysis model to search for potential slip surfaces, resulting in multiple sets of potential slip surfaces. Based on a preset safety factor threshold, the multiple groups of potential slip surfaces are screened for critical slip surfaces to obtain a monitoring point displacement dataset. The displacement dataset of the monitoring points is normalized to construct a three-dimensional displacement curve matrix; The overall shape and trend of the curves were extracted from the displacement dataset of the monitoring points, and the slope distortion index was calculated. Based on the three-dimensional displacement curve matrix and the slope distortion index, a pair of slope stability samples is obtained. The feature vectors of the slope stability sample pairs are input into a convolutional neural network for training, and a warning model is obtained by combining a triple physical constraint mechanism. The triple physical constraint mechanism includes parameter rationality constraints, displacement field smoothness constraints, and category imbalance correction. The specific acquisition method of the early warning model includes: The displacement curve matrix and slope distortion index of the slope stability sample pair are input into a convolutional neural network in parallel for multidimensional prediction to obtain landslide surface prediction parameters. Based on the parameter rationality constraints, a boundary penalty term is applied to the landslide surface prediction parameters to obtain the physical enhancement parameters; Based on the parameter rationality constraints, the displacement curve matrix is ​​subjected to total variation regularization to obtain a smooth displacement field; The sliding frequencies of different geometric shapes in the convolutional neural network are enhanced based on class imbalance correction and focus loss to obtain physical perception focus loss; A warning model is constructed based on the physical enhancement parameters, the smooth displacement field, and the physical sensing focus loss; The slope stability sample pairs and real-time target displacement data are input into the early warning model to invert the slip surface location, and the slip surface location triggers a graded early warning.

2. The landslide dynamic stability early warning method based on slope distortion according to claim 1, characterized in that, The slope heterogeneity index is calculated based on the geological structure and topographic mapping data of the slope. Distortion-sensitive areas are delineated using this heterogeneity index, and a targeted deployment scheme is constructed, including: By comprehensively judging the sliding mechanism based on the geological structure of the slope and the topographic mapping data, the landslide deformation zone is obtained. Based on the topographic curvature and lithological differences of adjacent rock layers in the landslide deformation zone, heterogeneity indices are obtained. The heterogeneity index is input into a preset distortion-sensitive weight calculation model to obtain distortion-sensitive weight parameters. The distortion-sensitive region is divided according to the distortion-sensitive weight parameter, and the targeted monitoring points are deployed and constructed through the distortion-sensitive region to obtain the targeted deployment scheme.

3. The landslide dynamic stability early warning method based on slope distortion according to claim 2, characterized in that, Based on the topographic curvature and lithological differences of adjacent rock layers extracted from the landslide deformation zone, heterogeneity indicators are obtained, including: Based on the landslide deformation zone, the curvature of adjacent rock layers is calculated using a preset surface elevation model to obtain the topographic curvature; Based on the borehole data in the landslide deformation zone, the lithological combination and mechanical parameters of adjacent rock layers are extracted to obtain the lithological difference coefficient. The heterogeneity index is constructed based on the topographic curvature and the lithological difference coefficient.

4. The landslide dynamic stability early warning method based on slope distortion according to claim 1, characterized in that, Based on a preset safety factor threshold, the multiple groups of potential slip surfaces are screened for critical slip surfaces to obtain a monitoring point displacement dataset, including: Based on a preset safety factor threshold, the multiple sets of potential slip surfaces are screened for critical slip surfaces to obtain a set of spatial location coordinates. The spatial coordinate set is transformed based on the discrete element model and assigned different shear strength parameters to form a shear strength parameter set; Discrete element numerical simulation is performed based on the shear strength parameter set. The deformation and failure process of the slope under different strength conditions is calculated through the simulation results, and the displacement field of the whole slope is obtained. According to the targeted deployment scheme, the combined displacement time series data of the monitoring points are extracted from the entire slope displacement field to construct a monitoring point displacement dataset.

5. A landslide dynamic stability early warning device based on slope distortion, characterized in that, include: The acquisition module is used to acquire geotechnical mechanical parameters and slope geological structure; The calculation module is used to calculate the slope heterogeneity index based on the slope geological structure and topographic mapping data, and to divide the distortion-sensitive area through the heterogeneity index to construct a targeted deployment scheme. The slope heterogeneity indicators include: Based on the borehole data in the landslide deformation zone, the lithological combination and mechanical parameters of adjacent rock layers are extracted, the strength abrupt change of adjacent rock layers is quantified, and the lithological difference coefficient is obtained. The larger the value of the lithological difference coefficient, the more significant the lithological difference, and the easier it is for the structural surface to form a potential slip zone. Based on the topographic curvature and lithological difference characteristics of adjacent rock layers in the landslide deformation zone, the heterogeneity index is obtained. The construction module is used to construct a slope model based on the targeted layout scheme and the geotechnical parameters, and to construct slope stability sample pairs by searching for potential sliding surfaces and critically screening the slope model. The building module includes: The deployment unit is used to spatially deploy monitoring points based on the targeted deployment scheme, and to construct a slope model by combining the peak-residual interval of the geotechnical parameters. The import unit is used to import the slope profile of the slope model into the two-dimensional slope stability analysis model to search for potential slip surfaces and obtain multiple sets of potential slip surfaces. The filtering unit is used to perform critical slip surface screening on the multiple groups of potential slip surfaces according to a preset safety factor threshold, so as to obtain a monitoring point displacement dataset. The normalization unit is used to normalize the monitoring point displacement dataset and construct a three-dimensional displacement curve matrix. The second extraction unit is used to extract the overall shape and trend of the curve from the monitoring point displacement dataset and calculate the slope distortion index. A pairing unit is used to pair slope stability sample pairs based on the three-dimensional displacement curve matrix and the slope distortion index. The training module is used to input the feature vectors of the slope stability sample pairs into the convolutional neural network for training, and to construct an early warning model by combining a triple physical constraint mechanism. The training module's triple physical constraint mechanism includes parameter rationality constraints, displacement field smoothness constraints, and class imbalance correction. The training module includes: The multidimensional prediction unit is used to input the displacement curve matrix and slope distortion index of the slope stability sample pair into the convolutional neural network in parallel to perform multidimensional prediction and obtain landslide surface prediction parameters. Apply boundary elements to apply boundary penalty terms to the landslide surface prediction parameters according to parameter rationality constraints, and obtain physical enhancement parameters; Regularization units are used to add total variational regularization to the displacement curve matrix according to parameter rationality constraints to obtain a smooth displacement field; The enhancement unit is used to enhance the sliding frequencies of different geometric shapes in the convolutional neural network based on class imbalance correction and focus loss, so as to obtain the physical perception focus loss. The early warning construction unit is used to construct an early warning model based on the physical enhancement parameters, the smooth displacement field, and the physical sensing focus loss. The inversion module is used to input the slope stability sample pairs and real-time target displacement data into the early warning model to invert the slip surface location, and trigger graded early warnings based on the slip surface location.

6. The landslide dynamic stability early warning device based on slope distortion according to claim 5, characterized in that, The computing module includes: The judgment unit is used to comprehensively judge the sliding mechanism based on the geological structure of the slope and the topographic mapping data to obtain the landslide deformation zone. The first extraction unit is used to extract the topographic curvature and lithological difference characteristics of adjacent rock layers based on the landslide deformation zone, and obtain heterogeneity indicators. The calculation unit is used to calculate the distortion-sensitive weight parameters by inputting the heterogeneity index into a preset distortion-sensitive weight calculation model. The partitioning unit is used to divide the distortion-sensitive region according to the distortion-sensitive weight parameter, and to construct the targeted monitoring point layout through the distortion-sensitive region to obtain the targeted layout scheme.

Images