A geotechnical engineering slope deformation monitoring method and system

Abstract

The application provides a geotechnical engineering slope deformation monitoring method and system, belonging to the field of geotechnical engineering monitoring, and is used for solving the problems of low monitoring precision and poor early warning reliability caused by time and space dislocation of multi-source data, static geological parameters, poor adaptation of swivel scene and no closed-loop optimization in the related art. The method collects multi-source deformation data, geological parameters and swivel characteristics, and identifies the swivel type and progress. After time and space alignment, multi-resolution layered fusion (dynamic weight adaptation data reliability and environmental interference) is performed, dynamic inversion of geological parameters is performed to construct a prediction model, model parameters and early warning thresholds are dynamically scheduled in combination with swivel characteristics, and the precision is improved through closed-loop feedback optimization iteration. The system provides hardware support for the method, and realizes accurate slope deformation monitoring and efficient early warning.

Description

Technical Field

[0001] This application relates to the field of geotechnical engineering monitoring, and more particularly to a method and system for monitoring slope deformation in geotechnical engineering. Background Technology

[0002] In the field of geotechnical engineering, slopes are a common engineering structure, and their deformation state is directly related to the safety of the project and the stability of the surrounding environment. Therefore, slope deformation monitoring is one of the core links in the operation and maintenance of the project. Accurate monitoring is needed to identify instability risks in a timely manner and ensure the safe operation of the project.

[0003] Currently, slope deformation monitoring often employs multi-source data acquisition methods, with common data including InSAR data, GNSS data, inclinometer data, crack gauge data, etc. After simple fusion processing of these data, early warning information is generated by combining them with preset fixed early warning thresholds to achieve monitoring and risk management of slope deformation.

[0004] However, existing monitoring technologies have significant shortcomings: First, multi-source data are prone to spatiotemporal misalignment due to differences in acquisition frequency and spatial resolution, and simple fusion cannot guarantee data consistency, affecting monitoring accuracy. Second, the geological parameters of the slope are not fully correlated, or the geological parameters are statically assigned, which cannot adapt to changes in the mechanical properties of the slope's soil and rock, resulting in a disconnect between the deformation prediction model and the actual working conditions. Third, there is a lack of targeted monitoring strategy adjustments for various transformation scenarios such as slope excavation and backfilling, and temporary support to permanent structure conversion, and the early warning thresholds and model parameters cannot dynamically adapt to the transformation requirements. Fourth, there is no closed-loop feedback optimization mechanism, which cannot iteratively improve monitoring accuracy based on prediction deviations and early warning effects, easily leading to false alarms and missed alarms, making it difficult to meet the precise monitoring needs of complex slopes. Summary of the Invention

[0005] This application provides a method and system for monitoring slope deformation in geotechnical engineering. It can solve the problems of low accuracy and poor adaptability of existing monitoring by accurately fusing multi-source data, dynamically associating geological parameters, and adapting to rotating scenarios, thereby achieving accurate monitoring and efficient early warning of slope deformation.

[0006] Firstly, this application provides a method for monitoring slope deformation in geotechnical engineering. This includes collecting multi-source data related to slope deformation, wherein the multi-source data includes at least large-scale overall deformation data, local deep deformation data, fine surface deformation data, and slope geological parameter data. The collected multi-source data is then fused in a multi-resolution layer to form unified data characterizing the slope deformation state. Based on the fused data and slope geological parameters, a deformation prediction model is constructed, outputting the future deformation trend of the slope, and generating early warning information in conjunction with preset early warning standards.

[0007] By adopting the above technical solution, firstly, multi-source data covering slope deformation and geological characteristics at different scales are comprehensively collected to ensure the integrity of the monitoring data; then, through multi-resolution layered fusion processing, the consistency problem caused by the differences in multi-source data is eliminated, forming unified data that can accurately reflect the slope deformation state; finally, a prediction model is constructed and an early warning is generated by combining geological parameters, so that the prediction results are more in line with the actual mechanical properties of the slope, improving the accuracy of deformation prediction and the reliability of early warning, and solving the problems of poor multi-source data fusion effect and disconnect between prediction and reality in existing technologies.

[0008] Furthermore, the multi-resolution layered fusion includes at least three layers: macroscopic layer, mesoscopic layer, and microscopic layer. The macroscopic layer corresponds to the large-scale overall deformation data and includes at least InSAR data and GNSS data. The mesoscopic layer corresponds to the local deep deformation data and includes at least inclinometer data. The microscopic layer corresponds to the fine surface deformation data and includes at least crack gauge data and piezometer data.

[0009] By adopting the above technical solution, the specific hierarchical division of multi-resolution layering and the data source corresponding to each level are clearly defined, making the layered fusion more targeted, avoiding information loss or interference caused by the mixed fusion of data of different scales, further improving the accuracy of the fused data, and providing high-quality data support for subsequent prediction models.

[0010] Furthermore, the slope geological parameters include at least the internal friction angle, cohesion, pore water pressure, and slope angle of the soil and rock. When constructing the deformation prediction model, the slope geological parameters are correlated with the fused data to make the prediction model adapt to the mechanical properties of the slope soil and rock.

[0011] By adopting the above technical solutions, the types of key geological parameters are identified and associated with the fused data to construct a prediction model. This enables the model to fully reflect the mechanical nature of the slope's rock and soil, avoids one-sided predictions based solely on data trends, significantly improves the model's fit to the slope's deformation patterns, and reduces prediction bias caused by ignoring geological characteristics.

[0012] Furthermore, before performing multi-resolution hierarchical fusion on multi-source data, it also includes spatiotemporal alignment processing of multi-source data. The spatiotemporal alignment processing includes at least timestamp unification and spatial grid adaptation. Timestamp unification maps multi-source data with different acquisition frequencies to the same time reference. Spatial grid adaptation divides spatial grids of different densities according to the slope risk level, so that multi-source data are matched in spatial dimension.

[0013] By adopting the above technical solution, spatiotemporal alignment of multi-source data is performed before hierarchical fusion. First, a unified time base is used to eliminate time misalignment caused by differences in collection frequency. Then, spatial grid density is adapted according to risk level to ensure that data in high-risk areas is more refined. Data consistency is guaranteed from both spatiotemporal dimensions, laying a good foundation for hierarchical fusion and further improving fusion accuracy.

[0014] Furthermore, during the multi-resolution layered fusion, dynamic weights are configured for each layer of data. The adjustment of the dynamic weights is based on at least the reliability of each layer of data and the degree of environmental interference of the slope. The reliability of the data is determined based on the accuracy of the data acquisition equipment and the consistency of historical data. The degree of environmental interference includes at least the impact levels corresponding to rainfall, temperature, and vibration interference.

[0015] By adopting the above technical solution, dynamic weights are configured for each layer of data in the hierarchical fusion. The weight adjustment combines the data's own credibility with environmental interference, so that data with high credibility and less interference account for a higher proportion in the fusion, avoiding low-quality data interference caused by fixed weights, and improving the reliability of the fusion results. It is especially suitable for slope monitoring in complex environments.

[0016] Furthermore, the slope geological parameters are updated through dynamic inversion. The dynamic inversion combines at least the fused data, slope rock and soil mechanics theory, and prior information on geological parameters, and adopts an inversion algorithm with global search capabilities to ensure that the inverted geological parameters conform to the actual mechanical state of the slope.

[0017] By adopting the above technical solution, dynamic inversion and updating of geological parameters can be achieved. The inversion process combines data, mechanical theory and prior information, and adopts a global search algorithm to avoid the problem of outdated static parameters or local optima. This ensures that the geological parameters always fit the actual mechanical state of the slope, providing real-time and accurate parameter support for the prediction model and further improving the prediction accuracy.

[0018] Furthermore, it also includes collecting multi-morphological rotation characteristics of slope structures. The rotation characteristics include at least changes in slope construction volume, stress of temporary support structures, and changes in topographic elevation. Based on the rotation characteristics, the slope rotation type and rotation progress are identified. The rotation type includes at least excavation and backfilling rotation, temporary support and permanent structure rotation, and dangerous rock removal and slope greening rotation.

[0019] By adopting the above technical solutions, the collection and identification of slope rotation features are added, the types of rotation features and the classification of rotation types are clarified, and the monitoring method can accurately identify different morphological rotation scenarios of slopes. This lays the foundation for subsequent targeted adjustments to the monitoring strategy and solves the problem that existing technologies cannot be adapted to rotation scenarios.

[0020] Furthermore, when generating early warning information, dynamic scheduling based on rotation characteristics is also included. The dynamic scheduling includes at least adjusting the parameter configuration of the deformation prediction model and the threshold range of the early warning standard according to the rotation type and rotation progress. The parameter configuration adjustment includes at least the model calculation complexity and data input dimension. The threshold range adjustment is at least combined with the updated geological parameters after inversion and the slope risk level of the rotation stage.

[0021] By adopting the above technical solution, dynamic scheduling is achieved based on the rotation characteristics. The model parameters and early warning thresholds are adjusted according to the rotation situation, so that the prediction model and early warning standards can be adapted to the risk characteristics of each stage of rotation, avoiding false alarms and missed alarms in rotation scenarios, and improving the adaptability of the monitoring method to complex working conditions.

[0022] Furthermore, it also includes closed-loop feedback optimization, which is based at least on the deviation between deformation prediction results and actual deformation data, the matching degree between early warning information and actual slope risk status, adjusting the weight configuration of multi-source data fusion, the prior information of geological parameter inversion, and the rules of dynamic scheduling, so as to achieve iterative improvement of monitoring accuracy.

[0023] By adopting the above technical solutions, a closed-loop feedback optimization mechanism is constructed. Based on the matching degree between prediction deviation and early warning, the fusion weight, inversion information and scheduling rules are iteratively adjusted, so that the monitoring method can continuously optimize its performance, improve the monitoring accuracy and early warning reliability, and solve the defects of existing technologies that lack iterative optimization.

[0024] Secondly, this application provides a geotechnical engineering slope deformation monitoring system. It includes a multi-source data acquisition module, a data processing module, a prediction and early warning module, and a hardware support environment. The multi-source data acquisition module includes at least a large-scale deformation acquisition device, a deep deformation acquisition device, a fine deformation acquisition device, a geological parameter acquisition device, and a rotation feature acquisition device. The large-scale deformation acquisition device includes at least an InSAR receiver terminal and a GNSS receiver; the deep deformation acquisition device includes at least an inclinometer; the fine deformation acquisition device includes at least a crack gauge and a piezometer; the geological parameter acquisition device includes at least a geotechnical physical parameter tester and a pore water pressure sensor; and the rotation feature acquisition device includes at least an industrial drone and a stress sensor. The data processing module includes at least edge computing nodes and a central server. The data processing module is used to execute the data processing-related steps in the geotechnical engineering slope deformation monitoring method as described in any of the first aspects above. The edge computing nodes are used for preprocessing and spatiotemporal alignment of multi-source data. The central server is used for multi-source data fusion, geological parameter inversion, and deformation prediction model calculation. The prediction and early warning module includes at least an early warning threshold configuration unit, an early warning information generation unit, and an interactive terminal. The interactive terminal includes at least a monitoring center large screen and a mobile terminal. The hardware support environment includes at least a data transmission network and a power supply unit. The data transmission network includes at least a wireless network and a fiber optic network to realize data interaction between modules. The power supply unit includes at least solar power equipment and a backup power supply to ensure the stable operation of each module.

[0025] By adopting the above technical solutions, a hardware system adapted to the monitoring method is constructed. The multi-source data acquisition module ensures the comprehensive acquisition of various monitoring data. The data processing module collaborates with the central server through edge computing to realize the data processing steps in the method. The prediction and early warning module completes the generation and interaction of early warnings. The hardware support environment ensures the stable operation of the system, providing reliable hardware support for the implementation of the monitoring method and ensuring that the method can efficiently and stably realize slope deformation monitoring.

[0026] In summary, this application has at least the following beneficial effects:

[0027] 1. A method and system that accurately adapt to the needs of slope monitoring are provided, improving the accuracy of deformation monitoring and the reliability of early warning;

[0028] 2. Solve the consistency problem of multi-source data by multi-resolution layered fusion and spatiotemporal alignment;

[0029] 3. Improve the adaptability of monitoring methods to complex working conditions by dynamically inverting geological parameters and adapting to rotating scenarios;

[0030] 4. Through closed-loop feedback optimization, continuous iterative improvement of monitoring accuracy is achieved.

[0031] It should be understood that the description in the Summary Section is not intended to limit the key or essential features of the embodiments of this application, nor is it intended to restrict the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0032] The above and other features, advantages, and aspects of the embodiments of this application will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0033] Figure 1 A schematic diagram of a geotechnical engineering slope deformation monitoring system according to an embodiment of this application is shown.

[0034] Figure 2 A flowchart of a geotechnical engineering slope deformation monitoring method according to an embodiment of this application is shown. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0037] This application provides a method and system for monitoring slope deformation in geotechnical engineering. It can solve the problems of low monitoring accuracy and poor adaptability by accurately fusing multi-source data, dynamically inverting geological parameters, adapting to rotating scenarios, and optimizing closed loops. This enables accurate monitoring and efficient early warning of slope deformation, ensuring engineering safety.

[0038] In one aspect, embodiments of this application disclose a geotechnical engineering slope deformation monitoring system.

[0039] Figure 1 A schematic diagram of a geotechnical engineering slope deformation monitoring system according to an embodiment of this application is shown.

[0040] Reference Figure 1The system includes a multi-source data acquisition module, a data processing module, a prediction and early warning module, and a hardware support environment. The modules work together to achieve accurate monitoring and early warning of slope deformation in geotechnical engineering, providing hardware support for the implementation of the aforementioned geotechnical engineering slope deformation monitoring method.

[0041] The multi-source data acquisition module is used to comprehensively acquire various types of data related to slope deformation. It includes at least large-scale deformation acquisition equipment, deep deformation acquisition equipment, fine deformation acquisition equipment, geological parameter acquisition equipment, and rotation feature acquisition equipment. Specifically, the large-scale deformation acquisition equipment includes at least an InSAR receiver and a GNSS receiver to acquire deformation data over the entire slope area; the deep deformation acquisition equipment includes at least an inclinometer to acquire deformation data of the deep soil and rock mass of the slope; the fine deformation acquisition equipment includes at least a crack gauge and a piezometer to acquire fine surface deformation and pore water pressure data of the slope; the geological parameter acquisition equipment includes at least a soil and rock physical parameter tester and a pore water pressure sensor to acquire the core physical and mechanical parameters of the slope's soil and rock mass; and the rotation feature acquisition equipment includes at least an industrial drone and a stress sensor to acquire data on changes in engineering quantities, temporary support stress, and topographic elevation changes during the multi-form rotation process of the slope structure.

[0042] The data processing module is used to execute the data processing steps in the aforementioned monitoring method, and it includes at least edge computing nodes and a central server. Edge computing nodes are deployed near the slope monitoring area and are mainly used to preprocess the raw data acquired by the multi-source data acquisition module, including outlier removal, data standardization, and spatiotemporal alignment, reducing data transmission pressure and improving subsequent processing efficiency. The central server is deployed at the monitoring center and is mainly used to receive the preprocessed data from the edge computing nodes, perform multi-source data multi-resolution layered fusion, dynamic inversion of slope geological parameters, and deformation prediction model calculations, providing data and model support for the generation of early warning information.

[0043] The prediction and early warning module is used to generate and display slope deformation early warning information. It includes at least an early warning threshold configuration unit, an early warning information generation unit, and an interactive terminal. The early warning threshold configuration unit is used to configure dynamic early warning thresholds based on slope geological parameters, rotation stage, and other information. The early warning information generation unit is used to generate corresponding early warning information by combining the future deformation trend output by the deformation prediction model with the early warning threshold. The interactive terminal includes at least a monitoring center large screen and a mobile terminal. The monitoring center large screen is used to intuitively display monitoring data and early warning information at the monitoring center, while the mobile terminal is used to push early warning information to relevant personnel for convenient remote viewing and response.

[0044] The hardware support environment ensures the stable operation and data interaction of each module, and includes at least a data transmission network and a power supply unit. The data transmission network includes at least a wireless network and a fiber optic network. The wireless network enables data transmission between the multi-source data acquisition modules and edge computing nodes, while the fiber optic network enables high-speed data interaction between edge computing nodes and the central server, and between the central server and the prediction and early warning module. The power supply unit includes at least solar power equipment and a backup power supply. The solar power equipment provides power to the multi-source data acquisition modules and edge computing nodes deployed in the field, while the backup power supply ensures the operation of critical equipment and prevents system monitoring interruptions during severe weather or power outages.

[0045] Each module interacts with data and instructions through a data transmission network. The multi-source data acquisition module transmits the acquired data to the data processing module, which processes the data and then transmits the results to the prediction and early warning module to generate an early warning. The hardware support environment provides a stable guarantee for the entire process, ensuring that the system can continuously and efficiently support the implementation of the aforementioned geotechnical engineering slope deformation monitoring method.

[0046] Secondly, this application discloses a method for monitoring slope deformation in geotechnical engineering.

[0047] Figure 2 A flowchart of a geotechnical engineering slope deformation monitoring method according to an embodiment of this application is shown.

[0048] Reference Figure 2 The method specifically includes the following steps:

[0049] S1: Collect multi-source data related to slope deformation. The multi-source data includes at least large-scale overall deformation data, local deep deformation data, surface fine deformation data, and slope geological parameter data.

[0050] In this step of the method, it is necessary to first clarify the acquisition standards and initial parameter definitions for various types of multi-source data to ensure data consistency and usability: For large-scale overall deformation data, InSAR data should use a spatial resolution of not less than 3m and a temporal resolution of [missing information]. The monitoring data for the day, its raw phase data needs to be converted into deformation data after atmospheric delay correction. ,in These are planar coordinates in the WGS84 coordinate system. For monitoring time;

[0051] GNSS data was acquired using a dual-mode (BeiDou / GPS) receiver with a sampling rate of 1Hz, and the output is three-dimensional coordinates. Calculate the horizontal deformation by the coordinate difference between adjacent time points. With vertical deformation For localized deep deformation data, the inclinometer is set up with a monitoring point every 1m along the borehole depth direction to collect data at different depths. Horizontal deflection angle at the location Through formula ( (For depth interval indexing) is converted into deep horizontal deformation. For fine surface deformation data, the crack gauge collects crack opening and closing information. (Unit: mm, resolution) ), piezometers collect pore water pressure (Data collected by a piezometer, denoted as) To distinguish it from other pore water pressure data units in kPa, the range For slope geological parameter data, initial geological parameters were obtained through borehole sampling combined with indoor direct shear tests, including the internal friction angle. Cohesion The initial value is calculated using a weighted average of regional geological database data and field test data, using the following formula: In the formula (Database weight) (Weight of field test) These are parameter values ​​for similar rock and soil bodies in the regional geological database. The results are from indoor tests conducted using samples taken on-site; slope angle. The formula is calculated by fitting UAV point cloud data. ,in The elevation difference between the top and bottom of the slope. This is the horizontal distance from the top of the slope to the bottom of the slope.

[0052] It also includes collecting multi-form rotation features of slope structures. The rotation features include at least changes in slope construction volume, stress of temporary support structures, and changes in topographic elevation. Based on the rotation features, the slope rotation type and rotation progress are identified. The rotation type includes at least excavation and backfilling rotation, temporary support and permanent structure rotation, and dangerous rock removal and slope greening rotation.

[0053] During the feature acquisition and recognition process of the slope rotation, the features need to be quantified and modeled: changes in the amount of work involved in slope construction. The formula is calculated based on the volume difference between two consecutive drone point cloud data sets. ,in The planar range of the rotation area. The current point cloud elevation. The moment the rotation begins The point cloud elevation, if Indicates the amount of backfill work. This indicates the amount of excavation work. Stress in the temporary support structure. Data was collected using a vibrating wire stress sensor (sampling rate 10 min / time, measurement range). Output stress time history data ,in Establish coordinates for the sensors. Terrain elevation changes. The average elevation difference of the slope at the current moment and at the initial moment of rotation is given by the formula: This represents the area of ​​the rotation region.

[0054] Rotation type recognition uses a random forest classification model, with the input feature vector being... ,in The number of days the rotation lasts; the model outputs the probability distribution of the rotation type. (These correspond to excavation and backfilling rotation, temporary support and permanent structure rotation, and dangerous rock removal and slope greening rotation, respectively. The type with the highest probability is taken as the final identification result.) Rotation progress The calculation is performed using the quantity percentage method, and the formula is as follows: ,in The total design workload of the rotation project (determined by the construction drawings). ,when This indicates that the rotation has not yet started. This indicates that the rotation is complete; if it is a rotation of a permanent structure with temporary support (no significant change in the amount of work), then the stress change percentage is used for calculation. ,in The initial stress during rotation, The final stress is designed for the support structure.

[0055] S2: Multi-resolution layered fusion of the collected multi-source data to form unified data characterizing the slope deformation state.

[0056] In this step of the method, before performing multi-resolution hierarchical fusion on multi-source data, it also includes spatiotemporal alignment processing of multi-source data. The spatiotemporal alignment processing includes at least timestamp unification and spatial grid adaptation. Timestamp unification maps multi-source data with different acquisition frequencies to the same time reference. Spatial grid adaptation divides spatial grids of different densities according to the slope risk level, so that multi-source data are matched in spatial dimension.

[0057] In the process of unifying timestamps, the first step is to determine a global time base. (Selecting a specific GNSS data sampling time as the benchmark, since GNSS has the highest sampling frequency, reaching 1Hz), for data sources with different acquisition frequencies, a hybrid method of "trend extension-linear interpolation" is used to achieve time alignment. InSAR data (time resolution...) (day) and GNSS data (time resolution) Taking (minutes) alignment as an example, let InSAR at time... The shape variable is GNSS in The sequence of shape variables within the interval is First, a deformation trend function is fitted using GNSS data. (in For deformation rate, For the intercept, use the least squares method. (Solve the problem), and then perform time interpolation on the InSAR data based on this trend to obtain an arbitrary reference time. InSAR alignment values: For high-frequency data such as crack gauges and piezometers (sampling rate) If the time interval is less than 1 second, then linear interpolation is directly used to map it to the reference time axis, and the formula is: ,in These are the original values ​​of high-frequency data at adjacent sampling times. The deformation rate is uniformly represented using... This indicates that the suffix distinguishes the data source (e.g., To improve the speed of data fusion computing, (The fitted trend rate).

[0058] Spatial grid adaptation requires prior quantification of slope risk levels. The calculation was performed using a two-factor weighted approach: "deformation rate - geological parameters". In the formula (Deformation rate weight) (Geological parameter weights); This represents the real-time deformation rate at that location (calculated from GNSS data). (Industry-recognized high-risk deformation rate threshold); The cohesion at this location (initial value from geological parameter collection in S1, dynamically updated subsequently). Pore ​​water pressure, For the slope angle, (Refer to cohesion value); ,in It is a high-risk area. It is a medium-risk area. This is a low-risk area. Grid side lengths are determined based on risk level. : Then, all aligned data are interpolated to the corresponding grid nodes to achieve spatial dimension matching. The interpolation uses the inverse distance weighting method. ,in For grid node coordinates, For the aligned data point values, The Euclidean distance from the data point to the grid node. Number of data points participating in interpolation .

[0059] The multi-resolution layered fusion includes at least three layers: macroscopic layer, mesoscopic layer and microscopic layer. The macroscopic layer corresponds to the large-scale overall deformation data and includes at least InSAR data and GNSS data. The mesoscopic layer corresponds to the local deep deformation data and includes at least inclinometer data. The microscopic layer corresponds to the fine surface deformation data and includes at least crack gauge data and piezometer data.

[0060] Each layer of fusion employs a "dynamic weighted fusion" method, first calculating the fusion results from different data sources within each layer, and then performing cross-layer fusion. In the macro-level fusion, the InSAR and GNSS aligned grid data are as follows: Its fusion weight , Based on data credibility Decision, satisfaction Macro-level fusion results: The meso-level fusion is based on deformation data from inclinometers at different depths, employing depth-weighted fusion, with the formula as follows: ,in This is the maximum monitoring depth of the inclinometer (usually taken as 20m). For depth The weight of the mesh deformation value decreases with increasing depth (deep data is more critical for the overall stability of the slope); in the micro-layer fusion, the crack opening degree With pore water pressure Through coupling via the "deformation-permeability correlation coefficient", the formula is: ,in (Crack deformation weight) (Upper limit of the pressure measurement range) normalizes the pressure value and integrates it with the crack data to highlight the micro deformation characteristics.

[0061] Furthermore, during the multi-resolution layered fusion, dynamic weights are configured for each layer of data. The adjustment of the dynamic weights is based on at least the reliability of each layer of data and the degree of environmental interference of the slope. The reliability of the data is determined based on the accuracy of the data acquisition equipment and the consistency of historical data. The degree of environmental interference includes at least the impact levels corresponding to rainfall, temperature, and vibration interference.

[0062] Data credibility The calculation uses a two-factor model of "equipment accuracy - historical consistency": ,in (Equipment accuracy weight); This refers to the equipment's accuracy factor, which is converted from the accuracy specifications in the equipment manual, such as the GNSS horizontal accuracy. ,but (10mm is the maximum allowable error), InSAR accuracy is ,but The historical data consistency factor is calculated by determining the standard deviation of the distortion difference between the data source and a reference data source (such as GNSS as a macroscopic reference) over the past 30 days. The formula is The closer to 1, the better the data consistency.

[0063] The degree of environmental interference is quantified using the interference coefficient. This indicates that the factors taken into account rainfall, temperature, and vibration disturbance are as follows: In the formula (Rainfall weight) (Temperature weighting) (Vibration weight) , , The weighting coefficients are artificially set based on experience in slope monitoring projects, giving priority to the impact of rainfall on data quality. maximum); Real-time rainfall ( (Data collected by weather stations) (Threshold for extreme rainstorms); Real-time ambient temperature ( ), (Suitable operating temperature for the equipment) (Upper limit of equipment temperature); Real-time vibration acceleration (Data collected by vibration sensors) (Upper limit of equipment vibration resistance); The larger the value, the more severe the environmental interference.

[0064] Final dynamic weights of each layer The calculation formula is: ,in This represents the average confidence level of the corresponding layer. This dynamic weighting enables adaptive fusion of multi-resolution data, resulting in unified data characterizing the slope deformation state. .

[0065] S3: Based on the fused data and slope geological parameters, construct a deformation prediction model, output the future deformation trend of the slope, and generate early warning information in combination with preset early warning standards.

[0066] In this step of the method, the slope geological parameters include at least the internal friction angle of the soil and rock. Cohesion Pore ​​water pressure and slope angle When constructing the deformation prediction model, the slope geological parameters are associated with the fused data to make the prediction model adapt to the mechanical properties of the slope rock and soil.

[0067] The deformation prediction model adopts a "spatiotemporal attention-LSTM" hybrid architecture, first constructing the input feature vector. It includes the fused deformation data and geological parameters: ,in The unified fusion deformation amount output by S2. (where the time interval is for calculating the deformation rate) , These are the geological parameters at the current moment (initial values ​​are collected from S1 and subsequently updated through dynamic inversion). A fixed slope angle is assigned to the slope (S1 is obtained through point cloud fitting). The model's spatiotemporal attention layer is used to reinforce the weights of key spatiotemporal features, including temporal attention weights. Through formula calculate( (Attention window size, defaults to 10 time steps) Spatial attention weights pass calculate( To integrate the spatial gradient of deformation (reflecting areas of concentrated deformation), the attention-weighted features are input into the LSTM layer. The hidden layer state update formula of the LSTM is as follows: ,in It is the sigmoid activation function. This is the weight matrix. As a bias term, the final output is the future. Deformation trend . , These are the learnable parameters obtained by training the model using historical monitoring data, with the training objective being to minimize the deformation prediction bias.

[0068] The slope geological parameters are updated through dynamic inversion. The dynamic inversion combines at least the fused data, the slope rock and soil mechanics theory, and prior information on geological parameters. It employs an inversion algorithm with global search capabilities to ensure that the inverted geological parameters conform to the actual mechanical state of the slope.

[0069] Dynamic inversion is based on a two-way iterative process of "mechanical forward modeling - data inversion". First, a geotechnical mechanics forward model is established: based on the Mohr-Coulomb strength criterion, the theoretical deformation rate of the slope grid nodes is calculated. The formula is ,in The unit weight of the soil and rock mass (S1) is obtained through indoor tests and is the default value. ), The slope height (S1 is obtained through point cloud measurement). This represents the change in pore water pressure. The viscosity coefficient (empirical value, taken as...) The objective function of the inversion is to minimize the deviation between the theoretical deformation rate and the observed deformation rate, i.e. ,in The observed deformation rate is calculated for the S2 fused data.

[0070] To avoid local optima, a Bayesian prior distribution is introduced: ,correspond (physical range) ,correspond , (Corresponding to the osmotic pressure measurement range); the global search uses the Expected Improvement (EI) function to select initial values. The EI function formula is ,in For Gaussian processes Predicting the mean and standard deviation, The current optimal observation rate, The CDF and PDF are derived from the standard normal distribution, respectively. The Gaussian process uses the radial basis function (RBF), and the hyperparameters of the kernel function are trained from historical geological parameter inversion data through cross-validation. , This represents the prediction result of the Gaussian process for the inversion parameter G. Local optimization uses the Adam algorithm, with the iterative formula as follows: ,in Learning rate (critical period of rotation) hour Stable period hour First-order and second-order momentum, respectively, iterated to or number of iterations Output the updated geological parameters after inversion. .

[0071] When generating early warning information, dynamic scheduling based on rotation characteristics is also included. The dynamic scheduling includes at least adjusting the parameter configuration of the deformation prediction model and the threshold range of the early warning standard according to the rotation type and rotation progress. The parameter configuration adjustment includes at least the model calculation complexity and data input dimension. The threshold range adjustment is at least combined with the updated geological parameters after inversion and the slope risk level of the rotation stage.

[0072] Model parameter configuration adjustments are based on rotation type With progress For excavation, backfilling, and rotation ( ) or temporary support permanent structure rotation ( critical period of rotation "Lightweight" parameter configuration (number of neurons in the LSTM hidden layer) Attention window Input feature dimensions Only retain To improve computational speed (time taken per prediction) Rotation stabilization period Or dangerous rock removal, slope greening and rotation "High-precision" parameter configuration (number of neurons in the LSTM hidden layer) Attention window Input feature dimensions Complete preservation All features) to improve prediction accuracy (error) ).

[0073] The threshold range for the early warning standard is calculated using a coupled formula of "geological parameters - rotation risk": ,in The baseline warning threshold is set based on industry safety standards. The initial cohesion collected by S1, Rotation sensitivity coefficient hour hour hour ;In the formula, It reflects the contribution of current geological parameters to slope stability (the larger the value, the stronger the stability; the threshold can be appropriately increased). Reflecting the risks during the rotation phase (initial stage of rotation) When this value increases, the threshold decreases to avoid false negatives. (When predicting deformation...) When a "yellow alert" is generated; when When a "red alert" is generated, the alert information includes the alert level, the coordinates of the grid area involved, the predicted deformation, and the geological parameters.

[0074] S4: Perform closed-loop feedback optimization.

[0075] In this step, the closed-loop feedback optimization is based at least on the deviation between the deformation prediction results and the actual deformation data, the matching degree between the early warning information and the actual risk status of the slope, and adjusts the weight configuration of multi-source data fusion, the prior information of geological parameter inversion, and the rules of dynamic scheduling to achieve iterative improvement of monitoring accuracy.

[0076] First, the core basis for feedback optimization needs to be quantitatively calculated: one is the deformation prediction deviation, defined as the root mean square error (RMSE) between the predicted deformation and the actual fused deformation in the next monitoring cycle, as shown in the formula. ,in The total number of grid nodes in the slope monitoring area (determined by S2 spatial grid adaptation). The future output for S3 Predict the amount of deformation. The actual deformation amount collected and fused in the next monitoring cycle (i.e., S2 in) (Output result at time step). A smaller RMSE indicates higher prediction accuracy, and its threshold is set to... (Set based on the industry's high-precision monitoring needs). Secondly, the early warning matching degree is quantified using F1 scores, with the formula as follows: TP (True Positive) represents the number of times a warning was issued and on-site investigation confirmed a risk; FP (False Positive) represents the number of times a warning was issued but no risk was detected on-site; and FN (False Negative) represents the number of times a warning was not issued but a risk existed on-site. All data are derived from warning records and daily on-site risk investigation reports. The larger the value, the higher the degree of alignment between the warning and the actual risk; its target value is set to... .

[0077] Based on the above quantitative criteria, the weight configuration for multi-source data fusion is first adjusted: in the S2 dynamic weight formula... Based on this, a deviation correction factor is introduced. The corrected weight formula is as follows: ,in RMSE The bias in predictions made using only this layer of data (e.g., the RMSE calculated using only macroscopic layer data), RMSE This represents the total RMSE after full-layer fusion; if the RMSE of any layer is greater than RMSEtotal, then... The weight of the layer is reduced when the prediction accuracy is high, and vice versa, so as to achieve adaptive adjustment that "the layer with high prediction accuracy gets higher fusion weight".

[0078] Secondly, adjust the prior information for geological parameter inversion: for the prior distribution of geological parameters in S3 ( , ), combining the inversion results with laboratory validation data from on-site borehole sampling (denoted as ), Prior parameters are updated (data collected every 30 days), using the following formula: , (because Physical range is (Take 60 as the upper limit reference value). The prior parameter update is similar: (Will scope (Normalized to a factor of 10); if If the early warning matching degree is high and the inversion credibility is high, the original prior parameter weights are retained; otherwise, the laboratory data weights are increased to ensure that the prior information fits the actual geological conditions of the slope.

[0079] Finally, the dynamic scheduling rules were adjusted: For the "critical rotation period" in S3... The rule for "switching model parameter configurations" is based on RMSE and F1 to adjust the threshold for the rotation phase, and the formula is as follows: , , , where is the preset monitoring accuracy threshold. These are the new critical phase start and end progress thresholds for rotation; if or ,but reduce, Increase the critical period duration to ensure the model uses high-precision parameter configurations (such as LSTM hidden layer neurons) over a longer period. At the same time, the sensitivity coefficient in the early warning threshold adjustment formula was corrected. If the warning matching degree is low, then increase This makes the adjustment of the early warning threshold during the rotation phase more sensitive and reduces the risk of missed reports.

[0080] Closed-loop feedback optimization is performed at a fixed period (consistent with the deformation prediction period, i.e.) Iterative execution, recalculating RMSE and F1 after each optimization, until... and This forms a complete closed loop of "prediction-verification-optimization-re-prediction," continuously improving monitoring accuracy and early warning reliability.

[0081] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0082] By collecting multi-source data related to slope deformation (including large-scale overall deformation, local deep deformation, fine surface deformation data, and slope geological parameters), and simultaneously collecting multi-morphological rotation characteristics of the slope structure (changes in construction volume, stress of temporary support structures, and changes in topographic elevation) to identify the rotation type and progress, the limitations of a single data source can be overcome—avoiding the problems of atmospheric interference when relying solely on InSAR data and obstruction when relying solely on GNSS data. At the same time, the current morphological stage of the slope can be clearly identified through rotation characteristic identification, laying the foundation for subsequent monitoring strategy adaptation and ensuring that the monitoring data covers all dimensions of information about the slope from the whole to the part, from static geological characteristics to dynamic morphological changes, rather than fragmented data accumulation.

[0083] First, spatiotemporal alignment processing is performed on multi-source data (timestamps are unified to map data from different acquisition frequencies to the same time base, and spatial grid adaptation divides the grid into different densities according to the slope risk level). Then, multi-resolution hierarchical fusion is carried out and dynamic weights are configured (weights are adjusted according to data reliability and environmental interference). This can solve the problems of spatiotemporal asynchrony and quality differences of multi-source data: time alignment eliminates the time deviation between InSAR weekly data and GNSS minute-level data, spatial grid adaptation allows for more refined data acquisition in high-risk areas to capture local deformation, and dynamic weights give a higher proportion of data with high reliability and less interference from rainfall / temperature / vibration in the fusion. The final output unified fused data can accurately characterize the slope deformation state and avoid deformation information distortion caused by data misalignment or low-quality data interference.

[0084] A deformation prediction model is constructed based on the fused data and dynamically updated slope geological parameters (internal friction angle, cohesion, pore water pressure, etc.). The geological parameter inversion combines fused data, slope geotechnical mechanics theory (Mohr-Coulomb strength criterion), and prior information, employing a hybrid algorithm of global search and local optimization. This ensures that the geological parameters closely match the actual mechanical state of the slope, avoiding the problem that static parameters cannot reflect the stress redistribution of the slope. The prediction model incorporates a spatiotemporal attention mechanism to focus on key deformation areas and time steps. At the same time, it combines geological parameters to enhance the model's adaptability to the slope's mechanical properties, enabling it to output accurate future deformation trends, rather than relying solely on data trends for one-sided predictions, thus reducing prediction bias caused by ignoring the fundamental mechanical nature of the geotechnical system.

[0085] When generating early warning information, dynamic scheduling is performed based on the rotation characteristics. The parameter configuration of the deformation prediction model is adjusted according to the rotation type and progress (e.g., lightweight parameters are used to improve calculation efficiency during the critical rotation period, and high-precision parameters are used to ensure prediction accuracy during the stable period) and the threshold range of the early warning standard (adjusted in combination with the updated geological parameters and the risk level of the rotation stage). This allows the monitoring strategy to adapt to the dynamic characteristics of slope morphology rotation: avoiding missed early warnings due to sudden changes in slope characteristics in the early stage of rotation (by appropriately lowering the threshold), or false early warnings due to fixed parameters during the stable period of rotation, ensuring a balance between monitoring accuracy and early warning reliability at different rotation stages, rather than using a single rule to deal with all working conditions.

[0086] By optimizing through closed-loop feedback (adjusting the fusion weights of multi-source data, prior information on geological parameter inversion, and dynamic scheduling rules based on the deviation between deformation prediction results and actual data, and the matching degree between early warning information and actual risks), an iterative mechanism of "prediction-verification-optimization" can be formed: when the prediction deviation is large, the contribution of high-quality data is increased by adjusting the fusion weights, or the prior information on inversion is updated to optimize the accuracy of geological parameters; when the matching degree of early warning is low, the dynamic scheduling rules are corrected to adapt to the actual risk characteristics of the slope, ultimately achieving a continuous improvement in monitoring accuracy and early warning reliability, and avoiding the system from having fixed errors for a long time and being unable to adapt to changes in slope conditions.

[0087] Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-mentioned technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-mentioned technical features or their equivalent features without departing from the aforementioned disclosed concept. For example, technical solutions formed by substituting the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for monitoring slope deformation in geotechnical engineering, characterized in that, include: S1. Collect multi-source data related to slope deformation, including large-scale overall deformation data, local deep deformation data, surface fine deformation data, slope geological parameter data, and multi-morphological rotation characteristics of the slope structure; the slope geological parameters include the internal friction angle of the soil and rock. Cohesion pore water pressure and slope angle The rotation feature includes changes in the amount of slope construction work. Stress in temporary support structures Topographic elevation changes , Identify slope rotation type based on the aforementioned rotation characteristics. and rotation progress , The types of rotation include excavation and backfilling rotation, temporary support and permanent structure rotation, and dangerous rock removal and slope greening rotation. S2. Perform multi-resolution layered fusion on the collected multi-source data to form unified data characterizing the slope deformation state; multi-resolution layered fusion includes the fusion of three layers: macroscopic layer, mesoscopic layer, and microscopic layer. The macroscopic layer corresponds to large-scale overall deformation data, including InSAR data and GNSS data. Mesoscopic data corresponds to local deep deformation data, including inclinometer data. The micro-layer corresponds to fine surface deformation data, including crack gauge data and piezometer data. S3. Based on the fused data and slope geological parameters, construct a deformation prediction model, output the future deformation trend of the slope, and generate early warning information in combination with preset early warning standards. The slope geological parameters are updated through dynamic inversion to establish a forward model of soil and rock mechanics: based on the Mohr-Coulomb strength criterion, the theoretical deformation rate of the slope grid nodes is calculated. ,in For the density of rock and soil mass, The slope height This represents the change in pore water pressure. Let be the viscosity coefficient. The objective function for the inversion is to minimize the deviation between the theoretical deformation rate and the observed deformation rate. ,in The observed deformation rate was calculated from the S2 fused data; To avoid local optima, a Bayesian prior distribution is introduced: , ,correspond The physical range , ,correspond , Corresponding to the osmotic pressure measurement range; the global search uses the expected improvement EI function to screen initial values. The EI function formula is: ,in For Gaussian processes Predicting the mean and standard deviation, The current optimal observation rate, These represent the CDF and PDF of the standard normal distribution, respectively. The Gaussian process uses the radial basis function (RBF) kernel, whose hyperparameters are trained from historical geological parameter inversion data through cross-validation. , To predict the inversion parameter G using this Gaussian process, the Adam algorithm is used for local optimization, with the iterative formula being: ,in Learning rate, critical period of body rotation hour Stable period hour , First-order and second-order momentum, respectively, iterated to or number of iterations Output the updated geological parameters after inversion. .

2. The method for monitoring slope deformation in geotechnical engineering according to claim 1, characterized in that, Before performing multi-resolution hierarchical fusion of multi-source data, spatiotemporal alignment processing of the multi-source data is also included. The spatiotemporal alignment process includes timestamp unification and spatial grid adaptation. The timestamps uniformly map multi-source data with different acquisition frequencies to the same time base. The spatial grid adaptation divides spatial grids into different densities according to the slope risk level, enabling multi-source data to match in spatial dimensions.

3. The method for monitoring slope deformation in geotechnical engineering according to claim 2, characterized in that, During the multi-resolution layered fusion process, dynamic weights are configured for each layer of data. The adjustment of the dynamic weights is based on the reliability of the data at each layer and the degree of environmental disturbance to the slope. The reliability of the data is determined based on the accuracy of the data acquisition equipment and the consistency of historical data. The degree of environmental disturbance includes at least the impact levels corresponding to rainfall, temperature, and vibration disturbance.

4. A slope deformation monitoring system for geotechnical engineering, characterized in that, It includes a multi-source data acquisition module, a data processing module, a prediction and early warning module, and a hardware support environment. The multi-source data acquisition module includes large-scale deformation acquisition equipment, deep deformation acquisition equipment, fine deformation acquisition equipment, geological parameter acquisition equipment, and rotation feature acquisition equipment. Large-scale deformation acquisition equipment includes InSAR receivers and GNSS receivers. Deep deformation acquisition equipment includes inclinometers. Fine deformation acquisition equipment includes crack gauges and piezometers. Geological parameter acquisition equipment includes a rock and soil physical parameter tester and a pore water pressure sensor. Rotation feature acquisition equipment includes industrial drones and stress sensors. The data processing module includes edge computing nodes and a central server. The data processing module is used to execute the geotechnical engineering slope deformation monitoring method as described in any one of claims 1-3. Edge computing nodes are used for preprocessing and spatiotemporal alignment of multi-source data, while the central server is used for multi-source data fusion, geological parameter inversion, and deformation prediction model calculation. The prediction and early warning module includes an early warning threshold configuration unit, an early warning information generation unit, and an interactive terminal. Interactive terminals include a large screen in the monitoring center and mobile terminals. The hardware support environment includes the data transmission network and power supply unit. Data transmission networks, including wireless and fiber optic networks, enable data exchange between modules. The power supply unit includes solar power equipment and backup power to ensure the stable operation of each module.

Images