Slope stability monitoring and early warning method and system

Abstract

The invention relates to the technical field of slope monitoring and early warning, and discloses a slope stability monitoring and early warning method and system. The method comprises the following steps: acquiring multi-source monitoring data of a slope area, and performing noise suppression and abnormal value elimination operation on the acquired original monitoring data; based on the division result of the geographic space grid, a space-time evolution model of a slope displacement field is constructed, and space and time dimension analysis of slope displacement changes is achieved; fusing the stress response data in the rock mass with the external environment driving factor to generate an evaluation index capable of dynamically reflecting the stable state of the slope; triggering a multi-stage early warning strategy according to the change gradient of the stability evaluation index, and generating an early warning instruction containing the risk position coordinate and the change rate; resource reconfiguration of the slope monitoring network is executed according to the early warning instruction, and self-adaptive adjustment of monitoring equipment parameters is completed. According to the method, comprehensive and dynamic monitoring and accurate early warning of the slope stability are achieved through the synergistic effect of multiple links, and slope safety management and control work is assisted.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring and early warning technology, specifically to a slope stability monitoring and early warning method and system. Background Technology

[0002] In fields such as civil engineering and geological engineering, slope instability often triggers geological disasters such as landslides and collapses, posing a serious threat to the lives and property of surrounding residents and the normal operation of infrastructure. With the increase in construction projects in mountainous areas and the frequent occurrence of extreme weather events, the importance of slope stability monitoring has become increasingly prominent.

[0003] Currently, slope stability monitoring largely relies on single-type monitoring data, such as displacement monitoring or stress monitoring, which fails to comprehensively reflect the actual state of the slope. While some monitoring methods attempt to integrate multi-source data, significant shortcomings exist in data processing. Noise and outliers in the raw data are not effectively processed, leading to substantial deviations in subsequent analysis results. Furthermore, existing technologies for describing slope displacement fields are mostly limited to changes at local points, lacking overall spatiotemporal evolution analysis based on geospatial grids, and thus failing to accurately capture the spatial distribution patterns and temporal trends of slope deformation.

[0004] In stability assessment, traditional methods often neglect the synergistic effect of internal rock mass stress and external environmental factors, resulting in static assessment indicators that fail to reflect the dynamic changes in slope stability in real time. Early warning strategies also suffer from relatively simple triggering mechanisms, often based on a single threshold, failing to provide multi-level warnings based on the gradient of stability changes, leading to insufficient accuracy and timeliness of warning information. Furthermore, the resource allocation of monitoring networks lacks adaptive adjustment capabilities, failing to optimize the parameters and layout of monitoring equipment according to early warning instructions, thus affecting the effectiveness and efficiency of monitoring data. Summary of the Invention

[0005] The purpose of this invention is to provide a slope stability monitoring and early warning method and system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a slope stability monitoring and early warning method, the method comprising:

[0007] Collect multi-source monitoring data of the slope area, and perform noise suppression and outlier removal operations on the collected raw monitoring data;

[0008] Based on the geospatial grid division results, a spatiotemporal evolution model of the slope displacement field is constructed;

[0009] By integrating internal stress response data of the rock mass with external environmental driving factors, dynamic assessment indicators for slope stability are generated.

[0010] Based on the change gradient of stability assessment indicators, a multi-level early warning strategy is triggered to generate early warning instructions that include the coordinates of the risk location and the rate of change.

[0011] Based on the early warning instructions, the resources of the slope monitoring network are reconfigured to complete the adaptive adjustment of the monitoring equipment parameters.

[0012] Preferably, the step of collecting multi-source monitoring data of the slope area specifically includes:

[0013] Simultaneously acquire displacement sequence data output by the surface displacement monitoring unit, inclination change output by the deep underground inclinometer unit, and seepage pressure value output by the pore water pressure sensor unit.

[0014] The displacement sequence data is subjected to wavelet threshold denoising to generate denoised displacement temporal features;

[0015] The box plot analysis method is used to detect outliers in the tilt angle change and to remove abnormal tilt angle data that exceed the preset confidence interval;

[0016] The permeation pressure value is estimated using a Kalman filter algorithm to generate a smoothed pore water pressure time series curve.

[0017] Preferably, the step of constructing the spatiotemporal evolution model of the slope displacement field specifically includes:

[0018] The slope area is divided into equally spaced spatial grids, and each grid is associated with its corresponding denoised displacement time series features.

[0019] Extract the spatial correlation coefficient of displacement vectors between adjacent grids to generate a displacement propagation path topology map;

[0020] Establish cross-time-period dependencies of grid displacement based on temporal convolutional networks to predict the displacement field distribution in future time windows;

[0021] By integrating the displacement propagation path topology map with the displacement field distribution prediction results, a spatiotemporal evolution map of slope deformation is generated.

[0022] Preferably, the step of generating dynamic evaluation indicators for slope stability specifically includes:

[0023] Obtain the rainfall intensity data collected by the current weather station and the water level change output by the groundwater level monitoring unit;

[0024] The smoothed pore water pressure time series curve is input into the geotechnical mechanics parameter inversion model, and the rock mass saturated permeability coefficient matrix is ​​output.

[0025] By correlating the spatiotemporal evolution map of slope deformation with the rock mass saturated permeability coefficient matrix, the temporal variation of the overall safety factor of the slope is calculated;

[0026] By combining the rainfall intensity data and water level changes, the time-series changes in the safety factor are weighted and corrected to generate a dynamically updated set of stability assessment indicators.

[0027] Preferably, the step of triggering a multi-level early warning strategy based on the change gradient of the stability assessment index specifically includes:

[0028] Monitor the rate of decrease of the safety coefficient in the set of stability assessment indicators, and activate the early warning mode when the rate of decrease exceeds a preset threshold;

[0029] Based on the numerical range of the descent rate, the corresponding level of early warning response rules are matched to generate an early warning level identifier;

[0030] Locate the set of grid coordinates with the largest deformation gradient in the spatiotemporal evolution map of the slope deformation, and associate it with the corresponding displacement change rate;

[0031] The warning level identifier, grid coordinate set, and displacement change rate are integrated to generate a structured warning instruction data packet.

[0032] Preferably, the step of reconfiguring the resources of the slope monitoring network according to the early warning instruction specifically includes:

[0033] Parse the warning level identifier in the structured warning instruction data packet to determine the list of monitoring device types that need to be adjusted;

[0034] Based on the grid coordinate set, the spatial coverage of high-risk areas is extracted, and the optimal deployment density of monitoring equipment is calculated.

[0035] The sampling frequency parameters of the displacement monitoring unit are adjusted according to the displacement change rate, and the communication bandwidth of the pore water pressure sensing unit is reallocated.

[0036] Send a configuration update command containing sampling frequency parameters and communication bandwidth parameters to the target monitoring equipment.

[0037] Preferably, the method further includes:

[0038] Obtain rock mass structural surface distribution data from the historical slope instability case database and construct a rock mass stress propagation path map;

[0039] The spatiotemporal evolution map of slope deformation and the rock mass stress propagation path map are overlaid and analyzed to identify the location of potential sliding surfaces;

[0040] The safety factor weight parameters in the stability assessment index set are adjusted according to the potential sliding surface location;

[0041] The matching rules for the warning level identifier are updated based on the revised safety factor weight parameters.

[0042] Preferably, the method further includes:

[0043] Establish a coupled response model of groundwater seepage pressure field and surface displacement field;

[0044] The smoothed pore water pressure time-series curve and the denoised displacement time-series characteristics are input in real time.

[0045] The dynamic correction value of the deformation modulus of the soil and rock mass is output through the coupled response model;

[0046] The dynamic correction value of the deformation modulus of the soil and rock mass is fed back to the parameter update interface of the soil and rock mechanics parameter inversion model.

[0047] Preferably, the method further includes:

[0048] Construct a diagnostic module for the operating status of monitoring equipment, and periodically collect battery voltage and signal strength indicators of the monitoring equipment;

[0049] When the battery voltage is below the critical threshold or the signal strength index continues to decay, a device maintenance request command is generated.

[0050] By associating the equipment maintenance request command with the set of grid coordinates of high-risk areas, a maintenance task queue with priority sorting is generated;

[0051] The maintenance task queue is injected into the device type list filtering logic of the resource reconfiguration step.

[0052] Preferably, the present invention further includes a slope stability monitoring and early warning system for performing the slope stability monitoring and early warning method described above, the system comprising:

[0053] The multi-source data acquisition module is used to collect multi-source monitoring data of the slope area and perform data preprocessing.

[0054] The spatiotemporal evolution modeling module is used to construct a spatiotemporal evolution model of the slope displacement field;

[0055] The dynamic evaluation module is used to generate dynamic evaluation indicators for slope stability.

[0056] The early warning decision module is used to trigger multi-level early warning strategies and generate early warning instructions;

[0057] The resource optimization module is used to perform resource reconfiguration and device parameter adjustment for the monitoring network.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] By collecting multi-source monitoring data and performing noise suppression and outlier removal, data interference can be reduced, making the original data closer to the true state and providing a reliable data foundation for subsequent analysis. Constructing a spatiotemporal evolution model of the slope displacement field based on geospatial grid division results allows for the spatial refinement of slope displacement changes to grid cells and the tracking of the evolution process over time. This comprehensively presents the distribution characteristics and development trends of slope displacement, contributing to a deeper understanding of the patterns of slope deformation.

[0060] Dynamic assessment indicators are generated by integrating internal rock mass stress response data with external environmental driving factors. This approach considers the combined impact of internal stress and external environmental factors on slope stability, enabling the assessment indicators to be updated in real time as various factors change, thus better reflecting the actual dynamic situation of slope stability. A multi-level early warning strategy is triggered based on the gradient of changes in the stability assessment indicators. This allows for the issuance of corresponding warning levels based on the degree of change, making the warnings more targeted and facilitating appropriate countermeasures by relevant personnel. Furthermore, the warning instructions include the coordinates of the risk location and the rate of change, enabling precise location of risk areas and understanding of their changes, thus improving the accuracy and timeliness of responses.

[0061] By reconfiguring the monitoring network resources and adaptively adjusting the parameters of monitoring equipment in accordance with the early warning instructions, monitoring resources can be tilted towards risk areas, the working parameters of monitoring equipment can be optimized, and more intensive and accurate monitoring data can be obtained in key areas. At the same time, resource waste can be avoided, the overall efficiency of the monitoring network can be improved, and the monitoring system can efficiently and accurately capture relevant information when slope stability changes, thus providing strong support for slope safety management. Attached Figure Description

[0062] Figure 1 This is a time-series diagram of the slope stability monitoring and early warning method described in this invention;

[0063] Figure 2 This is a flowchart of multi-source monitoring data acquisition and preprocessing.

[0064] Figure 3 A flowchart for constructing a spatiotemporal evolution model of slope displacement field;

[0065] Figure 4 A flowchart for monitoring network resource reconfiguration;

[0066] Figure 5 A flowchart for modifying the historical case library and sliding surface. Detailed Implementation

[0067] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Please see Figure 1 This invention provides a method for monitoring and early warning of slope stability, the method comprising:

[0069] Real-time monitoring and early warning of slope conditions are achieved through multi-source data acquisition, spatiotemporal evolution modeling, dynamic assessment, early warning decision-making, and resource optimization. First, multi-source monitoring data of the slope area are collected, including surface displacement, subsurface dip angle changes, and pore water pressure data, with noise suppression and outlier removal performed on the raw data. Based on the geospatial grid division results, a spatiotemporal evolution model of the slope displacement field is constructed, analyzing displacement propagation paths and cross-time period dependencies. Internal rock mass stress response data and external environmental driving factors are integrated to generate dynamically updated slope stability assessment indicators. Multi-level early warning strategies are triggered based on the gradient changes in the assessment indicators, generating early warning commands that include the coordinates of risk locations and the rate of change. Finally, the resource configuration of the monitoring network is adjusted according to the early warning commands, optimizing equipment sampling frequency and communication bandwidth to achieve adaptive monitoring.

[0070] Example 1: See Figure 2 The acquisition and preprocessing of multi-source monitoring data involves the collaborative work of surface displacement monitoring units, deep underground inclinometer units, and pore water pressure sensing units. The surface displacement monitoring unit continuously records the movement of the slope surface using high-precision displacement sensors installed at specific locations on the slope, outputting displacement sequence data, which includes real-time captured displacement changes. The deep underground inclinometer unit is deployed in boreholes, using inclinometer sensors to monitor changes in the tilt angle of deep underground rock masses, outputting tilt angle change data to reflect minute deformations in the internal structure of the rock mass. The pore water pressure sensing units are distributed within hydrological monitoring boreholes, measuring water pressure fluctuations in the pores of the rock mass, outputting a sequence of seepage pressure values ​​to indicate groundwater flow dynamics. Data from these units is synchronously transmitted to the central processing unit via wired or wireless communication protocols, achieving timestamp-aligned data acquisition.

[0071] Displacement sequence data output by surface displacement monitoring units are susceptible to environmental noise interference, including mechanical vibration, electromagnetic interference, and sensor drift. Wavelet threshold denoising is applied to displacement sequence data to eliminate high-frequency noise components. This method is based on discrete wavelet transform, decomposing the displacement sequence into multiple sub-bands of different frequencies. The decomposition process uses specific wavelet basis functions, such as Daubechies or Haar wavelets, to achieve a hierarchical representation of the signal through multi-scale analysis. In the thresholding stage, soft thresholding is applied to high-frequency detail coefficients, setting coefficients below a set threshold to zero, effectively filtering out random noise components. Subsequently, the reconstruction operation integrates the processed sub-band coefficients to generate denoised displacement time-series features. This process preserves the main trend of the displacement signal, removes short-term disturbances, and improves the analyzability of the displacement data. The threshold setting depends on the statistical characteristics of the data itself; for example, the threshold level is calculated through data adaptation, and the processing parameters are dynamically adjusted based on the standard deviation of the displacement sequence. During implementation, an iterative optimization mechanism is used to verify the denoising effect and avoid excessive smoothing that could lead to loss of detail.

[0072] The dip angle variation data output by deep underground inclinometer units, including the sequence of rock mass dip angle changes over time, may contain outliers, such as abrupt changes caused by sensor malfunctions or external impacts. Box plot analysis is used for outlier detection, performing operations based on the statistical characteristics of the data distribution. The first step is to calculate the quartile statistics of the dip angle variation data. After sorting the complete data sequence, the lower and upper quartile positions are determined, and the interquartile range is calculated. Data points exceeding a specified multiple of this range are identified as outliers; a common threshold is 1.5 times the interquartile range. Specific implementation includes constructing a box plot of the data sequence to visualize the median, upper and lower quartiles, and the coordinates of possible outliers. Data exceeding the upper limit or falling below the lower limit are marked as outliers and removed, retaining the valid dataset. This process, combined with automated scripts, enables batch processing, continuously applying the data to the real-time data stream of the dip angle variation sequence to ensure statistical integrity. In subsequent processing, interpolation is used to fill in the positions of the removed points, employing a linear interpolation strategy between adjacent data points to maintain temporal continuity. The advantage of this method is that it processes large-scale data efficiently and transparently, avoiding human intervention.

[0073] The osmotic pressure value output by the pore water pressure sensing unit is affected by measurement errors and system noise. A Kalman filter algorithm performs state estimation to smooth the osmotic pressure data sequence. This algorithm is based on a state-space model framework, establishing a relationship model between the system state and the observed values. The process includes two iterative phases: a prediction step and an update step. In the prediction step, the system state and covariance matrix at the current time are estimated using the state estimate from the previous time step and the control input. In the update step, the current observed values ​​and predicted values ​​are fused, and the state estimate and error covariance are corrected using Kalman gain calculation. After repeated iterations, a smoothed pore water pressure time series curve is output, eliminating random fluctuations in the noise. Regarding parameter settings, the state transition matrix is ​​initialized based on physical laws such as Darcy's law, and the observation noise covariance is calibrated using sensor accuracy data. During implementation, a recursive calculation mechanism is used to process the real-time data stream, maintaining low-latency output. The smoothed curve shows the steady-state change trend of the osmotic pressure, reducing the impact of high-frequency disturbances. This method is integrated into the data processing pipeline, synchronizing with the outputs of other units for time alignment and format normalization. Ultimately, all preprocessed data is stored in a structured database to support subsequent analysis and model building. The entire process emphasizes automated processing logic, achieving seamless integration through software tools, with processing frequency dynamically matched to the data sampling rate. System logs record the execution status of each operation, including noise reduction processing time, outlier ratio, and filtering convergence status, for process traceability. The synchronization of data streams between units is maintained through a timestamp verification protocol, ensuring precise temporal correlation between multi-source data.

[0074] In the example of processing displacement sequence data, the discrete wavelet transform level for wavelet threshold denoising is set to 5 levels, with the decomposition depth adaptively selected based on signal characteristics. The threshold function adopts a general soft thresholding rule, and the threshold level is dynamically adjusted based on the root mean square error of the displacement sequence. For box plot analysis, the confidence interval for outlier detection is trained based on historical data distribution, and the threshold factor is configurable in the range of 1.3 to 2.0 to balance sensitivity and specificity. The initial estimation of Kalman filter model parameters depends on sensor specifications, such as the resolution and sampling rate of the pore water pressure sensing unit, and the system noise and observation noise parameters are refined through a calibration process. The overall preprocessing workflow is deployed on an embedded system or cloud platform, running in a real-time operating system environment, with processing latency controlled at the millisecond level, consistent with the data acquisition frequency. The output of the preprocessing unit includes a denoised displacement time-series feature data file, a dataset of tilt angle changes after outlier removal, and a smoothed pore water pressure curve file. These output formats adopt a standardized time-series structure, facilitating compatibility with the input interfaces of subsequent modules. In the system design, the data preprocessing unit has an error recovery mechanism. For example, when specific sensor data is missing, default values ​​or historical data are used as replacements to ensure the robustness of the processing flow. The modular architecture supports plug-in algorithm replacement; for example, different wavelet bases or filter variants can be selected to adapt to specific slope environmental conditions. The entire implementation emphasizes the scalability and repeatability of technical details, avoiding reliance on any single hardware platform. The data processing code is implemented using general-purpose programming languages, such as Python or C++, and deployed on distributed computing nodes to improve throughput. After preprocessing, the data is directly streamed to the spatiotemporal modeling module, maintaining the continuous operation of the entire system pipeline.

[0075] Example 2: See Figure 3The spatiotemporal evolution model of the slope displacement field begins with geospatial grid partitioning, dividing the target slope area into equally spaced rectangular grid cells. The grid size is set according to the slope scale and monitoring point density, typically 5m × 5m or 10m × 10m. Each grid cell is bound to a unique spatial coordinate identifier, associated with denoised displacement time-series feature data from the preprocessing module. The displacement data contains a sequence of displacement changes in three dimensions (X, Y, Z axes), stored aligned to a unified time reference. Displacement propagation path analysis calculates spatial correlation based on displacement vectors between adjacent grids, using a sliding time window mechanism to process the time-series data. For any two adjacent grids, the Pearson correlation coefficient of their displacement changes is calculated, generating a spatial correlation matrix. The correlation coefficient calculation covers a preset time span, such as displacement sequences within 24 or 72 hours. After standardization, the correlation matrix is ​​used to construct a displacement propagation topology network using graph theory algorithms, where grids act as nodes and spatial correlation strength is used as directed edge weights. The topology network is stored using an adjacency list data structure, supporting fast path lookup and analysis. Paths with edge weights exceeding a set threshold are marked as strongly correlated propagation channels and visualized as displacement propagation path topology maps, revealing the deformation transmission law of rock mass.

[0076] A temporal convolutional network architecture is used to model the cross-temporal dependencies of displacement fields. The network input layer receives denoised displacement temporal features from a single grid cell, with the sampling frequency uniformly adjusted to one data point per hour. The network structure contains multiple layers of dilated causal convolutional modules, each with a different dilation factor. The convolutional kernel size is selected as 3×1 or 5×1, dynamically adjusted according to the length of the temporal features. The dilation factor is configured as an exponentially growing sequence, allowing the receptive field to cover a long-term time span. After each convolutional layer, a normalization operation and activation function are applied, using a specific function to avoid the gradient vanishing problem. A residual connection structure is added during feature extraction to retain shallow feature information. The fully connected layer outputs the predicted 3D displacement values ​​for a specific future time window (e.g., 6 / 12 / 24 hours). The prediction results are inversely normalized to restore the actual dimensions, generating displacement distribution data for each grid cell at future times. The grid-level prediction data are aggregated to form a global displacement field distribution map, which is continuously updated and output at preset time intervals. This map includes grid coordinates, predicted displacement, and confidence indices, stored in raster data format.

[0077] In the fusion processing stage of the displacement propagation path topology map and displacement field distribution map, a mapping mechanism between spatial topological relationships and displacement prediction results is established. A grid coordinate indexing system is used to associate key nodes in the displacement propagation path with their predicted values ​​in the displacement field map. A spatial interpolation algorithm is applied to areas with low monitoring density, calculating displacement estimates for unmonitored grids based on topological path weights. The fusion output is organized using a spatiotemporal cube data model, adding a time dimension to the three-dimensional X, Y, and Z spatial coordinates. Cube cells store displacement amount, direction of change, and spatial correlation strength attributes. The final generated spatiotemporal evolution map of slope deformation supports dynamic rendering, displaying displacement gradient changes in the form of a heatmap, supplemented by vector arrows indicating the main deformation directions. The map data version is iteratively updated over time, retaining historical states for trend analysis.

[0078] The dynamic assessment index generation process for slope stability is integrated with external environmental data sources. Meteorological monitoring stations provide minute-level rainfall intensity data streams, which are accessed in real-time via standard communication protocols. Hydrological monitoring units output groundwater level change sequences, establishing a spatial correspondence with monitoring well locations. Preprocessed pore water pressure time-series curves are transmitted to the geotechnical parameter inversion module. The inversion model constructs an input-output mapping relationship based on unsaturated soil mechanics theory. The model receives input variables such as pore water pressure time series, soil porosity parameters, and water level depth. The inversion process employs an iterative optimization algorithm to solve for the saturated permeability coefficient of the rock mass, outputting a spatial distribution matrix. Each cell in this matrix corresponds to a permeability coefficient value at a specific grid location, and the data format is compatible with spatial grid systems.

[0079] The stability safety factor calculation module synchronously receives the spatiotemporal evolution map of slope deformation and the permeability coefficient matrix. The calculation algorithm is implemented based on the rigid body limit equilibrium theory framework, considering the balance relationship between anti-sliding force and sliding force. The anti-sliding force calculation combines grid displacement vectors, soil friction angle parameters, and permeability coefficient data. The sliding force calculation introduces mechanical components such as gravity and buoyancy caused by water level changes. Each grid independently calculates the instantaneous safety factor, and then integrates the safety states of all grids. The integration method includes the average value method, the minimum value method, or the weighted average method, with weights allocated according to the importance of the grid in the displacement propagation topology. The calculation process generates the temporal change of the overall safety factor across the entire domain, forming a safety factor sequence ordered by time.

[0080] Environmental driving factors are introduced during the dynamic index correction phase. Rainfall intensity data is converted into influence weighting coefficients, and the portion above the rainfall intensity threshold is converted into correction factors according to a logarithmic relationship. Groundwater level changes are graded and correction parameters are set according to their change gradient. The correction factors are applied to the original safety factor value through multiplication operations to generate a dynamically updated set of stability assessment indicators. The set contains structured data such as the corrected safety factor value, environmental correction factors, and permeability coefficient distribution. The entire assessment process is triggered at fixed time intervals (e.g., 30 minutes), and the data source is the latest received monitoring data stream. The output of the indicator set includes a timestamp version identifier to ensure synchronization with the spatiotemporal map data. All intermediate calculation data and result parameters are stored in a time-series database, supporting historical backtracking queries and analysis interface calls. The system transmits the assessment indicators to the early warning decision module through a message queue mechanism, completing the closed-loop transmission of the data stream.

[0081] Example 3: See Figure 4 The system focuses on triggering early warning mechanisms and dynamically adjusting resource allocation. This process begins with continuous monitoring of the stability assessment index set output by the dynamic evaluation module. The temporal changes in the safety factor within the index set are analyzed to determine the rate of decline, which is obtained through differential calculation over continuous time windows. A tiered early warning threshold system is established, comprising a baseline threshold, an intermediate threshold, and a high-level threshold. When the rate of decline of the safety factor first exceeds the baseline threshold, the system enters an early warning preparation state, continuously tracking the rate change trend. If the rate of decline continues to climb and exceeds the intermediate threshold, a yellow early warning response mode is activated; if it exceeds the high-level threshold, a red early warning mode is triggered. Early warning level identifiers are dynamically generated according to threshold matching rules, and the identifier encoding includes a color code and a risk level value.

[0082] Simultaneously, the spatial analysis function of the slope deformation spatiotemporal evolution map is invoked. The displacement gradient of each grid cell in the map is obtained through vector modulus calculation, and a spatial neighborhood comparison algorithm is used to locate gradient abrupt change regions. Specifically, with the target grid as the center, the sum of the absolute values ​​of the differences in displacement changes between it and its eight neighboring grids is calculated, and this sum is defined as the local deformation gradient intensity. By traversing the gradient intensity values ​​of all grid cells, the grid coordinate set of the top 5% quantiles is selected. This set is marked as the core coordinate group of the high-risk area. The displacement change rate of each grid in this coordinate group in the latest monitoring period is further extracted, and the rate value is calculated using the time series differencing method. The early warning decision module encapsulates the early warning level identifier, the high-risk grid coordinate set, and the corresponding displacement change rate into a structured data packet. The data packet is organized in JSON-LD format and includes a metadata header (timestamp, data version number) and body fields (early warning level numerical code, geographic coordinate string, rate array).

[0083] When the structured early warning command data packet is transmitted to the resource optimization module, the parsing engine first deconstructs the early warning level identifier. The identifier's numerical code is mapped to a preset equipment adjustment rule table: a yellow early warning corresponds to adjusting surface displacement and underground inclinometer units; a red early warning additionally includes pore water pressure sensing units. The high-risk grid coordinate set is input into the spatial analysis unit to calculate the spatial coverage of the high-risk area. The range calculation uses a convex hull algorithm to generate the minimum bounding polygon, along with grid count statistics with spatial indexes. The formula for calculating the optimal monitoring equipment deployment density is:

[0084] ;

[0085] in: This represents the optimized equipment deployment density (number of devices per unit area). Represents the baseline deployment density parameter; The warning level weighting coefficients are (yellow = 0.7, red = 1.2). It is the average displacement rate of the current grid group; It is the baseline value of the displacement rate; It is the maximum allowable displacement rate setting value of the system.

[0086] Based on this density value, the equipment deployment is replanned: new equipment locations are selected using the spatial Thiessen polygon algorithm, and portable monitoring nodes are deployed in areas with coverage gaps; redundant equipment automatically enters sleep mode. The sampling frequency parameters of the displacement monitoring units are adjusted through a rate feedback mechanism.

[0087] The system establishes the sampling frequency. With the rate of change of displacement Mapping function: when hour, ; hour, ; hour, .

[0088] The communication bandwidth of the pore water pressure sensing unit is allocated according to the data volume requirement, and the bandwidth allocation factor is... The calculation is based on the ratio of the current change in the osmotic pressure curve to the historical peak value.

[0089] bandwidth Actual allocation value: Yellow warning mode: Red Alert Mode: .

[0090] The target monitoring equipment receives configuration update commands through the device management interface. Command transmission uses a two-way authentication protocol, including the device identification code, new sampling frequency value, bandwidth allocation parameters, and effective timestamp. The parameter update process is executed in two phases: after the initial command is sent, the device's response status is verified; after confirming a normal response, the formal parameter loading command is issued. The device firmware has a built-in parameter switching module, seamlessly transitioning to the new configuration within a specified time window. All configuration changes are recorded in the audit log, including fields such as device ID, old parameter value, new parameter value, and switching time. When the grid coordinate set in a high-risk area changes, an incremental update mechanism is triggered: the system only initiates partial reconfiguration for devices associated with the newly added grid coordinates, avoiding resource consumption caused by resetting parameters across the entire area. The resource configuration status is visualized in real time on the monitoring platform, using a hierarchical color scheme to mark updated devices (green), devices awaiting updates (yellow), and abnormal devices (red). Administrators can manually intervene in the parameter synchronization process of abnormal nodes. The entire implementation process is managed through a distributed task queue; high-priority warning commands can interrupt low-priority configuration tasks that are currently being executed, achieving immediate response to critical resources.

[0091] Example 4: See Figure 5 The integration of historical case databases and real-time monitoring data enhances the accuracy of early warning. The system accesses a historical slope instability case database, which stores the structural characteristics of typical landslide events worldwide, categorized by geological structure. Each record contains a dataset of the spatial distribution of rock mass structural surfaces, including the surface attitude (strike, dip, angle), continuity index, and mechanical parameters. Case extraction employs a spatial similarity matching algorithm: relevant cases are filtered based on the rock mass type, slope range, and geological age of the target slope. The filtering results generate a set of structural surface distribution points, which are then spatially interpolated into a three-dimensional stress propagation path map. Map nodes represent structural surface intersections, and edge lines mark the potential stress transmission direction. Weight values ​​are determined by the structural surface friction coefficient and continuity rate.

[0092] The spatial overlay of the slope deformation spatiotemporal evolution map and the stress propagation path map was performed on a geographic information system platform. The overlay operation employed raster weighted fusion technology to project the displacement gradient field and stress transmission direction field onto a unified coordinate system. A spatial clustering algorithm identified highly correlated regions: when the angle between the displacement gradient direction and the stress transmission direction was less than 15° and the continuous coverage area exceeded 50 square meters, these regions were marked as potential sliding surface candidate areas. The boundaries of the candidate areas were refined using an edge detection algorithm to generate vector boundary polygons. Further analysis of the displacement-stress synergy index within the candidate areas was conducted, and the peak point of the synergy was located as the center of the sliding surface. The feature data of the two potential sliding surfaces identified in a specific slope case are shown in Table 1.

[0093] Table 1: Shows the feature data of two potential sliding surfaces identified in a slope case.

[0094]

[0095] The safety factor weight parameters of the safety assessment index set are corrected based on the distribution of the sliding surface location. The original weight allocation adopts a uniform distribution model, with an initial weight value of 1.0 for each grid. The correction process is implemented in two steps: For grids located within the boundary polygon of the sliding surface, the weight enhancement factor γ is calculated based on the distance from the center of the sliding surface. The weight of the grid at the center point is increased to 2.0, and the weight of the grid at the boundary is 1.5. For adjacent grids outside the polygon, a distance decay function is used, and the weight coefficient decreases linearly with increasing distance to 1.0. The second step dynamically adjusts the weight based on the stress transfer weight of the sliding surface: when the stress transfer weight exceeds 0.85, the corresponding grid weight is increased by an additional 0.3. The corrected grid weight matrix is ​​then multiplied by the original safety factor matrix using a Hadamard product to generate a risk-weighted safety factor sequence. This sequence serves as a new input source for the dynamic assessment module.

[0096] The update mechanism for the warning level identifier matching rules is linked to the revised weight parameters. The original rules relied solely on the rate of decrease in the safety factor; the new rules introduce a spatial risk distribution dimension. The updated decision matrix includes two evaluation axes: a safety factor change rate axis (slow / medium / fast) and a spatial weight extreme value axis (low / medium / high risk). The matrix intersection areas define nine warning states; a medium-speed change in a high-risk area triggers a yellow warning, while a rapid change in a high-risk area directly activates a red warning. The rule update package is deployed through a version control mechanism, and historical decision records are automatically migrated to the new evaluation system, maintaining warning continuity.

[0097] The coupled response modeling of the groundwater seepage pressure field and the surface displacement field adopts a multi-physics co-simulation framework. The model inputs smoothed pore water pressure curves and denoised displacement time-series characteristics; the preprocessing step resamples both types of data to the same time base. The physical field coupling is achieved through two-phase medium mechanics equations, establishing a bidirectional interaction between the deformation of the soil-rock skeleton and the pore fluid flow. The simulation process consists of three calculation cycles: the initial cycle establishes the equilibrium state of the steady-state seepage field and displacement field; the transition cycle introduces measured pressure fluctuation data to calculate the effect of the fluid pressure gradient on the skeleton stress; and the dynamic cycle superimposes measured displacement disturbances to solve for the adaptive changes in the soil-rock modulus.

[0098] The model outputs a dynamic correction sequence of soil deformation modulus values, including dual-channel data on the time-varying changes of bulk modulus and shear modulus. The correction values ​​are fed back to the soil mechanics parameter inversion model via a standard data interface. Upon receiving the new parameters, the inversion model initiates an incremental optimization process: retaining the original inversion result framework, iterative optimization is performed only on parameters related to the deformation modulus. The maximum number of iterations is set to 10, and the convergence threshold is set to the parameter change rate of 0.5%. The optimized parameter set generates a new version of the permeability coefficient matrix, which is then injected into the real-time calculation pipeline of the dynamic evaluation module. The parameter update process is accompanied by metadata recording, including audit information such as correction timestamps, parameter change magnitudes, and the number of affected grids. The system establishes a version rollback function, allowing restoration to the previous valid state within 72 hours after a parameter update, preventing system misjudgments caused by abnormal corrections.

[0099] The implementation process establishes a closed-loop feedback mechanism: weekly timed updates to the historical case database based on similarity retrieval, and newly added slope monitoring data are automatically added to the case matching feature database; spatial overlay analysis is performed every 6 hours; the coupled response model automatically adjusts its calculation frequency based on the rate of change of input data, switching to real-time calculation mode when the pore water pressure change rate exceeds 2 kPa / h. Technically, the stress propagation path map generation module is deployed on GPU-accelerated computing nodes, processing 800 structural surface data points in approximately 3 minutes; the coupled response model runs in a multi-core parallel environment, with a typical calculation cycle controlled within 15 minutes. All output data is stored in a timestamped binary format, supporting millisecond-level time synchronization accuracy. The sliding surface identification results are visualized as a 3D rendered model, with potential sliding areas marked by semi-transparent red curved surfaces, overlaid with the displacement heatmap on the core interface of the monitoring and early warning platform.

[0100] Example 5: This example focuses on a collaborative mechanism for maintaining the monitoring equipment's own status and optimizing resources. The equipment operation status diagnosis module periodically collects health indicator data from each monitoring node, with the collection interval set to a fixed cycle. The surface displacement monitoring unit acquires battery voltage data through its internal sensor circuitry, using an ADC analog-to-digital converter interface for real-time reading. The signal strength index of the deep underground inclinometer unit is parsed using the underlying protocol of the wireless communication module to extract the received signal strength indicator parameter value. The collected data is encapsulated in an equipment information frame format, including fields such as equipment identification code, timestamp, original voltage value, and signal strength value, and sent to the central processing platform using a low-power transmission protocol.

[0101] The diagnostic logic determines abnormal device states based on dynamic threshold rules. The battery voltage critical threshold is set as a specified percentage of the rated operating voltage, and this parameter varies depending on the device type. The signal strength attenuation threshold uses an adaptive calculation mode: the system records the median signal strength of each device's most recent 200 transmissions; a status alarm is triggered when five consecutive collected values ​​fall below a specific percentage of the median. Anomaly detection outputs a binary result label: 0 for normal status, 1 for voltage anomaly, 2 for signal anomaly, and 3 for combined anomaly.

[0102] The generation process for equipment maintenance request instructions begins after an anomaly status determination. The maintenance instruction code contains three core fields: the coordinate location information of the faulty equipment referencing the deployment topology index of the slope monitoring network; the anomaly status code mapping to the preset maintenance measure category; and a request generation timestamp with an appended expiration date parameter. Instructions are broadcast via the system message bus and simultaneously written to a persistent storage queue. Upon receiving the maintenance request instruction, the spatial association module retrieves the grid coordinate set data for high-risk areas in real time. This set data comes from the latest structured early warning instruction data packet output by the early warning decision module, and access is accelerated through an in-memory database cache.

[0103] The maintenance task queue is constructed using a two-level priority ranking system. The first priority is based on equipment anomaly codes: composite anomaly code 3 is assigned the highest priority level 9; voltage anomaly code 1 and signal anomaly code 2 are further subdivided into levels 7 and 8 based on real-time power consumption and communication quality scores, respectively. The second priority is calculated based on the risk weight of the faulty equipment's spatial location, incorporating spatial overlay analysis results: when the equipment coordinates are within the boundary of the newly identified potential sliding surface, the risk weight is set to 3.0; the weight of the outer buffer zone of the sliding surface is 1.5; and the weight of the ordinary monitoring area is 1.0. The final priority score is generated by multiplying the anomaly level by the spatial weight coefficient, ranging from 3 to 27 points. The task queue is managed using a max-heap data structure to ensure that high-priority tasks are always at the top of the queue.

[0104] When the resource reconfiguration module injects maintenance tasks into the queue, it executes the equipment type list filtering logic. The filter parses the equipment identification code prefix code in the queue and classifies and identifies the equipment type as displacement monitoring unit, inclinometer unit, or pore pressure unit. The filtering outputs two types of results: the list of equipment requiring immediate maintenance is directly pushed to the work order pool of the field maintenance system; the list of equipment requiring configuration adjustments is transferred to the resource optimization process. After receiving the list, the resource optimization module triggers the bypass mechanism for equipment parameter adjustment: the battery or signal problems of abnormal equipment are converted into false status data and injected into the configuration algorithm, forcibly reducing its sampling frequency or transmission bandwidth requirements. At the same time, the sampling frequency of adjacent backup equipment is temporarily increased to compensate for the data gap and maintain the integrity of the monitoring network.

[0105] The abnormal equipment parameter adjustment strategy follows a dynamic compensation principle. For displacement monitoring units with abnormal battery voltage, a forced degradation scheme for sampling frequency is implemented: if the original frequency is higher than once per hour, it is downgraded to once per hour; if the original frequency is already once per hour, it switches to power-saving mode. For pore pressure sensing units with abnormal signals, a data compression transmission scheme is adopted: raw floating-point data is converted into 8-bit quantized data, sacrificing accuracy to maintain basic data transmission. All adjustment parameters have valid time tags, and the initial configuration is automatically restored after the maintenance expiration period. After receiving the task queue, the on-site maintenance work order system generates the optimal inspection path based on the spatial distribution of equipment. The path planning uses a heuristic search algorithm, integrating geographical constraints and priority scores, and outputs the equipment maintenance order and recommended route navigation map. The status of each maintenance work order is synchronized with the system resource optimization module in real time. The maintenance completion confirmation signal triggers the equipment parameter reset command, restoring the normal monitoring status. Throughout the implementation process, the monitoring network topology map dynamically renders the status of abnormal equipment: equipment with abnormal battery voltage is marked with a red lightning bolt icon, equipment with abnormal signal voltage is marked with a yellow wavy icon, and equipment scheduled for maintenance is marked with a purple clock icon, intuitively supporting operation and maintenance decisions. The system retains complete status diagnostic logs, maintenance operation records, and resource configuration change trajectories, forming a closed-loop evidence chain for equipment management.

[0106] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and early warning of slope stability, characterized in that, Includes the following steps: Collect multi-source monitoring data of the slope area, and perform noise suppression and outlier removal operations on the collected raw monitoring data; Based on the geospatial grid division results, a spatiotemporal evolution model of the slope displacement field is constructed; By integrating internal stress response data of the rock mass with external environmental driving factors, dynamic assessment indicators for slope stability are generated. Based on the change gradient of stability assessment indicators, a multi-level early warning strategy is triggered to generate early warning instructions that include the coordinates of the risk location and the rate of change. Based on the early warning instructions, the resources of the slope monitoring network are reconfigured to complete the adaptive adjustment of the monitoring equipment parameters.

2. The slope stability monitoring and early warning method according to claim 1, characterized in that, The specific steps for collecting multi-source monitoring data of the slope area include: Simultaneously acquire displacement sequence data output by the surface displacement monitoring unit, inclination change output by the deep underground inclinometer unit, and seepage pressure value output by the pore water pressure sensor unit. The displacement sequence data is subjected to wavelet threshold denoising to generate denoised displacement temporal features; The box plot analysis method is used to detect outliers in the tilt angle change and to remove abnormal tilt angle data that exceed the preset confidence interval; The permeation pressure value is estimated using a Kalman filter algorithm to generate a smoothed pore water pressure time series curve.

3. The slope stability monitoring and early warning method according to claim 2, characterized in that, The steps for constructing the spatiotemporal evolution model of the slope displacement field specifically include: The slope area is divided into equally spaced spatial grids, and each grid is associated with its corresponding denoised displacement time series features. Extract the spatial correlation coefficient of displacement vectors between adjacent grids to generate a displacement propagation path topology map; Establish cross-time-period dependencies of grid displacement based on temporal convolutional networks to predict the displacement field distribution in future time windows; By integrating the displacement propagation path topology map with the displacement field distribution prediction results, a spatiotemporal evolution map of slope deformation is generated.

4. The slope stability monitoring and early warning method according to claim 3, characterized in that, The steps for generating dynamic assessment indices for slope stability specifically include: Obtain the rainfall intensity data collected by the current weather station and the water level change output by the groundwater level monitoring unit; The smoothed pore water pressure time series curve is input into the geotechnical mechanics parameter inversion model, and the rock mass saturated permeability coefficient matrix is ​​output. By correlating the spatiotemporal evolution map of slope deformation with the rock mass saturated permeability coefficient matrix, the temporal variation of the overall safety factor of the slope is calculated; By combining the rainfall intensity data and water level changes, the time-series changes in the safety factor are weighted and corrected to generate a dynamically updated set of stability assessment indicators.

5. The slope stability monitoring and early warning method according to claim 4, characterized in that, The steps of triggering a multi-level early warning strategy based on the gradient change of stability assessment indicators specifically include: Monitor the rate of decrease of the safety coefficient in the set of stability assessment indicators, and activate the early warning mode when the rate of decrease exceeds a preset threshold; Based on the numerical range of the descent rate, the corresponding level of early warning response rules are matched to generate an early warning level identifier; Locate the set of grid coordinates with the largest deformation gradient in the spatiotemporal evolution map of the slope deformation, and associate it with the corresponding displacement change rate; The warning level identifier, grid coordinate set, and displacement change rate are integrated to generate a structured warning instruction data packet.

6. The slope stability monitoring and early warning method according to claim 5, characterized in that, The specific steps for reconfiguring the resources of the slope monitoring network based on the early warning instruction include: Parse the warning level identifier in the structured warning instruction data packet to determine the list of monitoring device types that need to be adjusted; Based on the grid coordinate set, the spatial coverage of high-risk areas is extracted, and the optimal deployment density of monitoring equipment is calculated. The sampling frequency parameters of the displacement monitoring unit are adjusted according to the displacement change rate, and the communication bandwidth of the pore water pressure sensing unit is reallocated. Send a configuration update command containing sampling frequency parameters and communication bandwidth parameters to the target monitoring equipment.

7. The slope stability monitoring and early warning method according to claim 6, characterized in that, The method further includes: Obtain rock mass structural surface distribution data from the historical slope instability case database and construct a rock mass stress propagation path map; The spatiotemporal evolution map of slope deformation and the rock mass stress propagation path map are overlaid and analyzed to identify the location of potential sliding surfaces; The safety factor weight parameters in the stability assessment index set are adjusted according to the potential sliding surface location; The matching rules for the warning level identifier are updated based on the revised safety factor weight parameters.

8. The slope stability monitoring and early warning method according to claim 7, characterized in that, The method further includes: Establish a coupled response model of groundwater seepage pressure field and surface displacement field; The smoothed pore water pressure time-series curve and the denoised displacement time-series characteristics are input in real time. The dynamic correction value of the deformation modulus of the soil and rock mass is output through the coupled response model; The dynamic correction value of the deformation modulus of the soil and rock mass is fed back to the parameter update interface of the soil and rock mechanics parameter inversion model.

9. The slope stability monitoring and early warning method according to claim 8, characterized in that, The method further includes: Construct a diagnostic module for the operating status of monitoring equipment, and periodically collect battery voltage and signal strength indicators of the monitoring equipment; When the battery voltage is below the critical threshold or the signal strength index continues to decay, a device maintenance request command is generated. By associating the equipment maintenance request command with the set of grid coordinates of high-risk areas, a maintenance task queue with priority sorting is generated; The maintenance task queue is injected into the device type list filtering logic of the resource reconfiguration step.

10. A slope stability monitoring and early warning system, used to execute the slope stability monitoring and early warning method according to any one of claims 1-9, characterized in that, The system includes: The multi-source data acquisition module is used to collect multi-source monitoring data of the slope area and perform data preprocessing. The spatiotemporal evolution modeling module is used to construct a spatiotemporal evolution model of the slope displacement field; The dynamic evaluation module is used to generate dynamic evaluation indicators for slope stability. The early warning decision module is used to trigger multi-level early warning strategies and generate early warning instructions; The resource optimization module is used to perform resource reconfiguration and device parameter adjustment for the monitoring network.

Images