A method and system for processing data of automatic monitoring of roadbed slope

Abstract

The application discloses a kind of roadbed slope automation monitoring data processing method and system, belong to roadbed slope monitoring technical field, comprising: obtaining the physical state monitoring data of roadbed slope, based on the initial geology parameter and design parameter of roadbed slope, corresponding slope digital twin is established, and the risk correlation network in slope is constructed;Real-time surface displacement data is obtained as input, the internal mechanical parameters of digital twin are dynamically inverted by reverse analysis algorithm, to the error of the calculated displacement of digital twin and surface displacement data is less than the preset error threshold value as target, when error reaches preset error threshold value, output real-time evolution slope internal mechanical parameters;Based on the real-time evolution slope internal mechanical parameters obtained by inversion, the real-time stability coefficient of roadbed slope is calculated;Real-time stability coefficient is compared with preset stability threshold value.The application accurately evaluates slope stability, reduces the loss caused by slope disaster.

Description

Technical Field

[0001] This application relates to an automated monitoring data processing method and system for roadbed slopes, belonging to the field of roadbed slope monitoring technology. Background Technology

[0002] Monitoring the stability of roadbed slopes is crucial for ensuring road traffic safety. However, current roadbed slope monitoring technologies have many shortcomings. Traditional roadbed slope monitoring relies heavily on periodic manual inspections and limited data acquisition from single-type sensors, making it difficult to comprehensively and in real-time grasp the physical state of the slope. Furthermore, the acquisition of internal mechanical parameters of the slope largely depends on theoretical assumptions or a limited number of field tests, failing to dynamically and accurately reflect the actual mechanical properties of the slope. Existing technologies lack methods to effectively combine real-time monitoring data with slope mechanical models, making it difficult to accurately assess slope stability in real time. This leads to untimely and inaccurate early warnings of slope hazards, posing significant risks to road traffic safety. Summary of the Invention

[0003] According to one aspect of this application, an automated monitoring data processing method for roadbed slopes is provided, which can accurately assess slope stability in real time and reduce economic losses caused by slope disasters.

[0004] A method for automated monitoring data processing of roadbed slopes includes: Step 1: Acquire physical state monitoring data of the roadbed slope by deploying a sensor array on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. Step 2: Based on the initial geological parameters and design parameters of the roadbed slope, establish a corresponding digital twin of the slope and simultaneously construct a risk correlation network within the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include soil and rock type, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. Step 3: Using the real-time acquired surface displacement data as input, the internal mechanical parameters of the slope digital twin are dynamically inverted through the reverse analysis algorithm. The goal is to make the error between the calculated displacement of the slope digital twin and the surface displacement data less than a preset error threshold. When the error reaches the preset error threshold, the real-time evolved internal mechanical parameters of the slope are output. Step 4: Based on the real-time evolved internal mechanical parameters of the slope obtained by inversion, calculate the real-time stability coefficient of the roadbed slope using the limit equilibrium method, strength reduction method, or numerical simulation method. Step 5: Compare the real-time stability coefficient with the preset stability threshold, and generate corresponding early warning information based on the comparison result.

[0005] Furthermore, in step five, the stability threshold includes a first stability threshold, a second stability threshold, and a critical stability threshold, and the first stability threshold > the second stability threshold > the critical stability threshold; When the real-time stability coefficient is lower than the first stability threshold but higher than the second stability threshold, a first-level early warning message is generated, which indicates that the slope safety reserve has decreased. When the real-time stability coefficient is lower than the second stability threshold but higher than the critical stability threshold, a secondary early warning message is generated, which indicates that the slope is close to instability. When the real-time stability coefficient approaches or reaches the critical stability threshold, a level-three early warning message is generated, which indicates an extremely high risk of instability.

[0006] Furthermore, in step three, the internal mechanical parameters of the slope include at least one of the following: cohesion c of the soil and rock mass, internal friction angle φ, elastic modulus E, and Poisson's ratio ν.

[0007] Furthermore, step five also includes: Abrupt change point detection analysis is performed on the time series data composed of the surface displacement data, internal displacement data, and internal mechanical parameter data of the slope to identify whether there is an accelerated change trend in the data that meets the characteristics of instability precursors. When the trend of change based on internal displacement data exceeds the change threshold determined according to soil environmental parameters, it is determined that the monitoring point has reached the bottom of the potential slip surface, and the displacement value recorded at this moment is used as the second displacement data. Meanwhile, the data representing the top displacement of the potential slip surface is defined as the first displacement data, and the characteristic parameters of the potential slip surface are calculated based on the first displacement data and the second displacement data. When the accelerating change trend is identified, and the risk is determined to reach a critical level based on the slip surface characteristic parameters, a critical instability warning message is generated and issued. The rules for constructing the risk association network include: If the geotechnical materials where the two monitoring nodes are located have similar engineering properties or belong to the same homogeneous unit, and historical monitoring data confirms that the geotechnical materials have mutual influence on pore water pressure and soil pressure parameters, then an undirected connection edge is established between the two monitoring nodes. If historical data clearly indicates that an accident in the area where a certain monitoring node is located will propagate unidirectionally to the area where another monitoring node is located, or if the two monitoring nodes are located on opposite sides of a weak interlayer or unfavorable structural surface, and exploration and engineering data show that the structural surface exhibits stress transmission and seepage penetration, then a directed connection edge is established between the two monitoring nodes.

[0008] Furthermore, the mutation point detection analysis includes any of the following rules: Rule 1: Acquire the time series data, calculate the displacement rate and acceleration of the data series, and when the acceleration value continues to increase and exceeds the preset acceleration threshold, it is determined that an acceleration change trend that meets the characteristics of instability precursors has been identified. Rule 2: Divide historical time series data into training and validation sets to train the model to learn the data evolution patterns; input real-time time series data into the trained model to predict the data trend within a preset time period; calculate the deviation between the predicted value and the real-time observed value; when the deviation continuously exceeds the preset deviation threshold and the data shows non-linear accelerated growth characteristics, it is determined that an accelerated change trend that meets the characteristics of instability precursors has been identified.

[0009] Furthermore, the physical condition monitoring data also includes rainfall data; The method further includes step six: The rainfall data is input as a boundary condition into the slope digital twin updated by the data processing and inversion steps to calculate the predicted displacement data under the rainfall conditions. The predicted displacement data is compared with the real-time acquired surface displacement data, and the deviation between the two is calculated. When the deviation value exceeds the preset tolerance range, a model anomaly warning message is generated, wherein the tolerance range is set based on the accuracy of the monitoring data and the allowable error of the engineering.

[0010] Furthermore, the reverse analysis algorithm is executed using any of the following control methods: Method 1: Initialize the particle swarm using the internal mechanical parameters of the slope as optimization variables; The error between the displacement calculated by the digital twin of the slope based on the internal mechanical parameters of the slope and the real-time surface displacement data is used as the fitness function. Iterative updates of particle position and velocity; When the number of iterations reaches a preset value or the fitness function value is less than a preset error threshold, the parameters corresponding to the current global optimal particle are output as the real-time evolution of the slope's internal mechanical parameters. Method 2: Set the prior probability distribution of the internal mechanical parameters of the slope based on the initial geological parameters; A likelihood function is constructed based on the error between the displacement calculated from the digital twin of the slope and the real-time surface displacement data. The posterior probability distribution is solved using the Markov chain Monte Carlo method. The mean or mode of the posterior distribution is taken as the internal mechanical parameter for real-time evolution. When the standard deviation of the posterior distribution is less than the preset accuracy threshold, the inversion stops.

[0011] Furthermore, it also includes step seven: Obtain weather forecast information for a future preset time period, the weather forecast information including at least rainfall and temperature, and input the weather forecast information as a driving condition into the slope digital twin updated in step three; The stability evolution of the roadbed slope in a future preset period is predicted using the slope digital twin, and a predictive stability report including the predicted stability coefficient and its changing trend is output.

[0012] According to another aspect of this application, an automated monitoring data processing system for roadbed slopes is also provided, comprising: The data acquisition module is used to acquire physical state monitoring data of the roadbed slope through a sensor array deployed on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. The model building module is used to establish a corresponding digital twin of the roadbed slope based on the initial geological parameters and design parameters, and simultaneously build a risk association network inside the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include the type of rock and soil, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. The data processing and inversion module is communicatively connected to both the data acquisition module and the model building module. It receives real-time surface displacement data output by the data acquisition module and uses it as input to dynamically invert the internal mechanical parameters of the slope's digital twin generated by the model building module using a reverse analysis algorithm. The dynamic inversion aims to ensure that the error between the calculated displacement of the slope's digital twin and the surface displacement data is less than a preset error threshold. When the error reaches the preset error threshold, it outputs the real-time evolving internal mechanical parameters of the slope. A state assessment module, which is communicatively connected to the data processing and inversion module, is used to calculate the real-time stability coefficient of the roadbed slope based on the real-time evolving internal mechanical parameters of the slope output by the data processing and inversion module. The early warning decision module is communicatively connected to the state assessment module. It is used to compare the real-time stability coefficient output by the state assessment module with a preset stability threshold and generate corresponding early warning information based on the comparison result.

[0013] Furthermore, it also includes: A meteorological data acquisition module is used to acquire weather forecast information for a future preset period, the weather forecast information including at least rainfall and temperature; The stability prediction module is communicatively connected to the meteorological data acquisition module, the data processing and inversion module, and the model building module. The stability prediction module receives weather forecast information output by the meteorological data acquisition module and uses it as a driving condition, inputting it into the updated slope digital twin by the data processing and inversion module. The stability prediction module also uses the updated slope digital twin to predict the stability evolution of the roadbed slope within a preset future time period and outputs a predictive stability report containing the predicted stability coefficient and its changing trend.

[0014] The beneficial effects that this application can produce include: The method and system for automated monitoring data processing of roadbed slopes provided in this application comprehensively acquire slope physical state data through a sensor array and establish a digital twin of the slope, realizing digital simulation of the slope state. By using a reverse analysis algorithm to dynamically invert internal mechanical parameters, it can accurately acquire changes in the internal mechanical properties of the slope in real time. At the same time, based on real-time parameters, it calculates the stability coefficient and generates early warning information, which can timely and accurately assess slope stability, detect potential dangers in advance, provide a scientific basis for road maintenance and disaster prevention, effectively ensure road traffic safety, and reduce economic losses and social impacts caused by slope disasters. Attached Figure Description

[0015] Figure 1 This is a flowchart of an automated monitoring data processing method for roadbed slopes according to one embodiment of this application; Figure 2 This is a system diagram of an automated monitoring data processing system for roadbed slopes according to one embodiment of this application. Detailed Implementation

[0016] The present application is described in detail below with reference to the embodiments, but the present application is not limited to these embodiments.

[0017] See Figure 1 A method for automated monitoring data processing of roadbed slopes, characterized by comprising: Step 1: Acquire physical state monitoring data of the roadbed slope by deploying a sensor array on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. Step 2: Based on the initial geological parameters and design parameters of the roadbed slope, establish a corresponding digital twin of the slope and simultaneously construct a risk correlation network within the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include soil and rock type, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. Step 3: Using the real-time acquired surface displacement data as input, the internal mechanical parameters of the slope digital twin are dynamically inverted through the reverse analysis algorithm. The goal is to make the error between the calculated displacement of the slope digital twin and the surface displacement data less than a preset error threshold. When the error reaches the preset error threshold, the real-time evolved internal mechanical parameters of the slope are output. Step 4: Based on the real-time evolved internal mechanical parameters of the slope obtained by inversion, calculate the real-time stability coefficient of the roadbed slope using the limit equilibrium method, strength reduction method, or numerical simulation method. Step 5: Compare the real-time stability coefficient with the preset stability threshold, and generate corresponding early warning information based on the comparison result.

[0018] Specifically, step one involves the layout of surface and internal sensor arrays. Surface displacement sensors, such as GNSS and laser displacement meters, are used to capture macroscopic deformations of the slope's exterior, while internal displacement sensors, such as inclinometers and fiber optic sensors, are used to sense microscopic changes such as slippage and settlement within the soil and rock mass. These two types complement each other. Clearly defining the sensor types and data types being monitored is crucial. Surface displacement, directly reflecting slope stability, is a direct indicator of slope instability, while internal displacement serves as an early warning signal. Combining both enables comprehensive state perception. This addresses the limitations of traditional single-dimensional monitoring; monitoring only surface displacement cannot predict internal slippage risks, and monitoring only internal displacement is insufficient to reflect overall stability. It provides realistic feedback data for the digital twin, ensuring the relevance of subsequent modeling and calculations. In step two, the slope digital twin is a parametric mechanical calculation model based on the finite element method, finite difference method, or limit equilibrium method. Mechanical properties can be dynamically updated by adjusting input parameters, such as soil and rock mechanical and geometric parameters. The mechanical calculation model can simulate the stress, strain, and displacement mechanical responses of the slope under different working conditions, achieving accurate replication of the physical slope. The initial geological parameters encompass the types, distribution characteristics, geological structure information, and survey point data of the soil and rock masses, forming the basis for reflecting the inherent geological conditions of the slope. Design parameters include slope height, slope ratio, and support structure parameters. The risk correlation network implicitly models the mechanical interactions between various components of the slope, such as different soil and rock layers, support structures, and geological structures, including the force transmission from the slippage of a certain soil and rock layer to adjacent layers and the support structure. This provides a structured analytical framework for subsequent parameter inversion and stability calculations, transforming abstract slope mechanical behavior into calculable and simulable digital objects, and providing a core platform for subsequent real-time data calibration of model parameters and stability assessment. Furthermore, using the surface displacement data collected in step one as input, which is readily available and highly accurate, it is suitable as an inversion benchmark. The internal mechanical parameters in the digital twin are inferred through a reverse analysis algorithm. The objective is to ensure that the error between the calculated displacement of the slope digital twin and the surface displacement data is less than a preset error threshold. When the error meets the requirement, the real-time evolving internal mechanical parameters are output. Slope construction and environmental changes lead to dynamic variations in the mechanical parameters of the soil and rock mass. Static parameters obtained solely from the initial survey cannot reflect the real-time state of the slope, resulting in significant discrepancies between model calculations and actual conditions. This step utilizes dynamic inversion to achieve real-time updates of the digital twin parameters, ensuring the model remains consistent with the physical slope and providing accurate parameter support for subsequent stability calculations. Based on the real-time evolving internal mechanical parameters of the slope obtained from the inversion in step three, the real-time nature and accuracy of the calculation results are ensured.It supports limit equilibrium method, strength reduction method, and numerical simulation method, and can select the appropriate method according to slope type, geological complexity, and engineering requirements. It outputs a real-time stability coefficient, which is a core indicator for measuring the safety status of a slope. Generally, a stability coefficient ≥1.0 indicates a safe state, while <1.0 indicates a risk of instability. Specific thresholds can be adjusted according to the engineering level; the higher the coefficient value, the higher the slope safety reserve. The real-time stability coefficient calculated in step four is compared with a preset stability threshold. The threshold can be set according to the importance level of the slope, such as 1.2 for high-grade highway slopes and 1.0 for ordinary roadbeds. Based on the comparison results, corresponding early warning information is generated. For example, a safety prompt is output when the stability coefficient is ≥ the threshold, an early warning prompt is output when it is between the threshold and the risk threshold, and an emergency early warning prompt is output when it is below the risk threshold. This provides a basis for slope safety management decisions, enabling early detection, early warning, and early handling of risks.

[0019] In step five, the stability threshold includes a first stability threshold, a second stability threshold, and a critical stability threshold, and the first stability threshold > the second stability threshold > the critical stability threshold; When the real-time stability coefficient is lower than the first stability threshold but higher than the second stability threshold, a first-level early warning message is generated, which indicates that the slope safety reserve has decreased. When the real-time stability coefficient is lower than the second stability threshold but higher than the critical stability threshold, a secondary early warning message is generated, which indicates that the slope is close to instability. When the real-time stability coefficient approaches or reaches the critical stability threshold, a level-three early warning message is generated, which indicates an extremely high risk of instability.

[0020] Specifically, based on the safety risk level of slope engineering, a stability threshold system is defined: First stability threshold > Second stability threshold > Critical stability threshold, where: First stability threshold K1 = 1.25, corresponding to the critical boundary where the slope safety reserve is sufficient; Second stability threshold K2 = 1.15, corresponding to the critical boundary where the slope safety reserve is significantly reduced; Critical stability threshold K... 临 =1.05, corresponding to a slope approaching the critical boundary of instability; values ​​below this indicate a risk of landslide. Warning Level Classification and Response Measures: Level 1 Warning Information: Triggered when the real-time stability coefficient meets K2 < K < K1, indicating that the slope is affected by environmental or load factors, resulting in reduced safety redundancy, but not yet reaching a dangerous state. Response measures include: adjusting the monitoring frequency from once / 30min to once / 15min; arranging technical personnel to check the support structure daily for deformation and cracking; generating a daily slope safety reserve monitoring report and synchronizing it with the project management team; prohibiting the addition of temporary loads at the slope crest. Level 2 Warning Information: When the real-time stability coefficient meets K2 < K < K1... 临When K ≤ K2, the slope slip surface has initially formed, the deformation rate may accelerate, and instability is imminent. Response measures include: increasing monitoring frequency to once every 5 minutes, focusing on tracking deep displacement data from inclinometers to determine the slip surface development trend; closing off the construction area around the slope, retaining only monitoring and emergency access; initiating a support and reinforcement preparatory plan; and assigning dedicated personnel for 24-hour monitoring, submitting a slope deformation monitoring report every 2 hours. Level 3 warning information: When the real-time stability coefficient K ≤ K2... 临 The warning is triggered when a sudden increase in displacement rate or the penetration of the plastic zone is detected, indicating that the slope is in an extremely unstable state with a very high risk of instability. Response measures include: immediately activating the emergency evacuation plan and evacuating all personnel and equipment within 50 meters of the slope; cutting off power and water sources near the slope to prevent secondary disasters; initiating emergency reinforcement measures and requesting support from professional emergency rescue teams if necessary; and reporting the slope status to local transportation and emergency management departments in real time, with advance notice required if road closures are necessary. Warning information output and closed-loop management output methods: Level 1 warning: cloud platform push notification, project manager SMS; Level 2 warning: cloud platform pop-up warning, project team SMS, on-site audible and visual alarm; Level 3 warning: cloud platform forced pop-up, team SMS, on-site high-frequency audible and visual alarm, emergency command center hotline notification. Information content includes the warning level, current stability coefficient K value, and threshold range (e.g., current K=1.12, within K...). 临 The data includes: K2 (between -K2), inversion parameters (c / φ value), displacement rate, and a list of specific response measures. Warning cancellation conditions: Level 1 warning requires 24 consecutive hours, K≥1.25 and stable deformation; Level 2 warning requires 48 consecutive hours, K≥1.15 and no increase in deep displacement; Level 3 warning requires 72 consecutive hours after reinforcement completion, K≥1.10 and no abnormalities in any monitoring data. Cancellation process: The technical team submits a warning cancellation assessment report, attached with a monitoring data trend chart. After approval by the project leader, the warning is stopped and regular monitoring resumes. Simultaneously, the digital twin parameters are updated, and this warning case is included in the historical database for subsequent threshold optimization.

[0021] In step three, the internal mechanical parameters of the slope include at least one of the following: cohesion c of the soil and rock mass, internal friction angle φ, elastic modulus E, and Poisson's ratio ν.

[0022] Specifically, the core inversion parameters include soil cohesion (c), internal friction angle (φ), elastic modulus (E), and Poisson's ratio (ν). c and φ have the greatest impact on slope anti-sliding stability, accounting for ≥70%, and directly determine the shear strength of the sliding surface. E affects the accuracy of slope deformation calculation, especially the matching degree between surface displacement and deep displacement. ν affects the lateral deformation calculation results; for slopes with support structures, deviations in ν can lead to errors in support stress calculation, and are inverted synchronously with E. The reverse analysis algorithm is adapted to multi-parameter inversion requirements: the genetic algorithm has strong global optimization capabilities and is noise-resistant, suitable for multi-parameter synchronous inversion scenarios (such as c+φ+E+ν, c+φ+E), especially suitable for scenarios requiring simultaneous correction of strength and deformation parameters. The convergence condition is that the error of three consecutive iterations is ≤ a preset threshold. Particle swarm optimization (PSO) is fast and its parameters are easy to adjust. It is suitable for two-parameter / three-parameter inversions (such as c+φ, E+ν, c+φ+E). When only deformation-related parameters (E, ν) need to be corrected, its convergence efficiency is higher. The convergence condition is that the error change rate is ≤0.5% / iteration. Least squares is simple in principle and suitable for linear problems. It is suitable for single-parameter fine-tuning or two-parameter correlation inversions. It has high correction accuracy in small deformation scenarios. The convergence condition is that the sum of squared residuals is minimized. Inversion implementation process and error control input data: In addition to surface displacement data that has been continuous and anomaly-free for the past 24 hours, if ν needs to be inverted, deep lateral displacement data must be supplemented to ensure the accuracy of deformation parameter inversion. In the iteration, the first step is to set the initial range of parameters (based on geological test values ​​fluctuating by 30%, reference range: c∈[10,50]kPa, φ∈[15°,35°], E∈[10,50]MPa, ν∈[0.2,0.4], ν for cohesive soil is 0.3~0.4, and ν for sandy soil is 0.2~0.3); the second step is to select the inversion combination according to the scenario: if both surface displacement and deep displacement have large deviations: select c+φ+E+ν synchronous inversion; if only surface displacement has large deviations and deep displacement matches: select c+φ inversion; if deformation tends to Potential matching but large deviation in absolute displacement value: Select E+ν synchronous inversion; the third step is to adjust the parameters in each iteration, calculate the surface displacement and deep lateral displacement of the slope digital twin, and compare it with the measured data; the fourth step is to preset the error threshold, and when the error reaches the standard, output the real-time mechanical parameters. The parameter combination and inversion algorithm need to be marked, such as c=28kPa±3%, φ=25°±2%, E=25MPa±5%, ν=0.32±0.02, and use the genetic algorithm for synchronous inversion. If only some parameters are inverted, it needs to be stated that the other parameters use the latest calibration value.

[0023] Step five also includes: Abrupt change point detection analysis is performed on the time series data composed of the surface displacement data, internal displacement data, and internal mechanical parameter data of the slope to identify whether there is an accelerated change trend in the data that meets the characteristics of instability precursors. When the trend of change based on internal displacement data exceeds the change threshold determined according to soil environmental parameters, it is determined that the monitoring point has reached the bottom of the potential slip surface, and the displacement value recorded at this moment is used as the second displacement data. Meanwhile, the data representing the top displacement of the potential slip surface is defined as the first displacement data, and the characteristic parameters of the potential slip surface are calculated based on the first displacement data and the second displacement data. When the accelerating change trend is identified, and the risk is determined to reach a critical level based on the slip surface characteristic parameters, a critical instability warning message is generated and issued. The rules for constructing the risk association network include: If the geotechnical materials where the two monitoring nodes are located have similar engineering properties or belong to the same homogeneous unit, and historical monitoring data confirms that the geotechnical materials have mutual influence on pore water pressure and soil pressure parameters, then an undirected connection edge is established between the two monitoring nodes. If historical data clearly indicates that an accident in the area where a certain monitoring node is located will propagate unidirectionally to the area where another monitoring node is located, or if the two monitoring nodes are located on opposite sides of a weak interlayer or unfavorable structural surface, and exploration and engineering data show that the structural surface exhibits stress transmission and seepage penetration, then a directed connection edge is established between the two monitoring nodes.

[0024] The mutation point detection and analysis includes any of the following rules: Rule 1: Acquire the time series data, calculate the displacement rate and acceleration of the data series, and when the acceleration value continues to increase and exceeds the preset acceleration threshold, it is determined that an acceleration change trend that meets the characteristics of instability precursors has been identified. Rule 2: Divide historical time series data into training and validation sets to train the model to learn the data evolution patterns; input real-time time series data into the trained model to predict the data trend within a preset time period; calculate the deviation between the predicted value and the real-time observed value; when the deviation continuously exceeds the preset deviation threshold and the data shows non-linear accelerated growth characteristics, it is determined that an accelerated change trend that meets the characteristics of instability precursors has been identified.

[0025] Specifically, addressing the gradual characteristic of slope instability arising from slow changes in displacement / mechanical parameters to accelerated abrupt changes, this method analyzes time series data of surface displacement, internal displacement, and internal mechanical parameters to identify early signs of instability, overcoming the problem of delayed warnings caused by relying solely on stability coefficients. Rule 1 first performs differential calculations on the time series data to obtain displacement rate and acceleration; then, a preset acceleration threshold is set; finally, when the acceleration value continuously increases and exceeds the preset threshold, an accelerating trend is identified. This method features simple calculation logic, strong real-time performance, no need for complex model training, low engineering implementation cost, and suitability for slope engineering projects with high real-time requirements and small monitoring data volumes, such as ordinary highway subgrade slopes. Rule 2 first divides historical data into training and validation sets. An LSTM (Long Short-Term Memory) model is trained on the training set to learn the normal evolutionary patterns of the data; then, real-time data is input into the trained model to predict data trends for a preset future period; finally, the deviation between the predicted value and the real-time observed value is calculated. When the deviation continuously exceeds a preset deviation threshold and the data exhibits a non-linear accelerating growth characteristic, an accelerating trend is identified. It captures the nonlinear evolution characteristics of data, has high prediction accuracy, and can predict trend changes in advance. It is suitable for slopes with complex geological conditions, large data fluctuations, and high requirements for early warning and foresight, such as steep slopes in mountainous areas and soft soil roadbed slopes.

[0026] Furthermore, by analyzing the changing characteristics of internal displacement data, the spatial location of potential slip surfaces is determined, and core characteristic parameters are calculated, addressing the problem that traditional methods cannot clearly identify the location of risk sources. The determination is based on soil environmental parameters, such as soil and rock type, water content, and pore water pressure, to establish an internal displacement change threshold. The determination logic is that when the internal displacement change trend of a monitoring point exceeds this threshold, it indicates that the point has entered an active soil and rock slip zone and is determined to be the bottom of a potential slip surface. The final output is the recorded displacement value at this moment as the second displacement data, i.e., the displacement at the bottom of the slip surface. The determination logic is that the monitoring point data located above the potential slip surface and whose displacement changes conform to the characteristics of the top of the slip surface are defined as the first displacement data, i.e., the displacement at the top of the slip surface. The characteristic parameters of the slip surface include: slip surface thickness, calculated based on the depth difference between monitoring points corresponding to the first and second displacement data (e.g., if the top monitoring point is 5m deep and the bottom monitoring point is 10m deep, the slip surface thickness is approximately 5m); slip surface inclination angle, calculated using trigonometric functions by combining the horizontal and vertical displacement components of the monitoring points (e.g., if the horizontal displacement is 3mm and the vertical displacement is 4mm, the inclination angle is approximately 53°); slip rate, the average rate of change of the first and second displacement data over time (e.g., if the top displacement increases by 2mm and the bottom displacement increases by 3mm within 1 hour, the average slip rate is 2.5mm / h); and slip surface integrity, determined by comparing the characteristic parameters of multiple monitoring sections to determine if the slip surface is continuous (e.g., if multiple sections detect continuous top and bottom slip surfaces with consistent parameters, the slip surface is considered continuous). In terms of spatial positioning, clearly defining the depth, inclination angle, and other location information of potential slip surfaces provides precise targeting for emergency response; in terms of risk quantification, parameters such as slip surface thickness and continuity determine the scope of risk spread and the degree of damage, providing a basis for the intensity of response measures.

[0027] The critical instability early warning trigger is set under two conditions: Condition 1 is the identification of an accelerated change trend conforming to the characteristics of instability precursors through abrupt change point detection and analysis; Condition 2 is the determination that the risk has reached a critical level based on the characteristic parameters of the slip surface, such as a continuous slip surface with a slip rate exceeding 10 mm / h, a slip surface thickness exceeding 3 m, and a dip angle greater than 30°. Early warning elements include the critical instability warning level, the precise location of the slip surface, details of characteristic parameters, and emergency response suggestions. The transmission priority is the highest level, simultaneously triggering on-site audible and visual alarms, emergency SMS / APP push notifications, and alarms from the superior management platform to ensure timely delivery of response instructions. By defining the connection relationships between monitoring nodes, a global slope risk correlation network is constructed, enabling the analysis of the transmission of risk from a single node to the global risk, solving the problem of traditional methods that isolate monitoring point data and cannot predict risk diffusion paths. The conditions for constructing undirected connecting edges include two aspects: first, the geotechnical materials of the two monitoring nodes have similar engineering properties or belong to the same homogeneous unit; second, historical monitoring data proves that the pore water pressure and soil pressure parameters of the two nodes influence each other. For example, if the pore water pressure at node A increases, the pore water pressure at node B will increase synchronously within one hour. This serves to characterize the mutual influence of risks between nodes, and its engineering significance lies in predicting the bidirectional diffusion of local risks. For instance, if the area where node A is located becomes unstable, it may lead to synchronous instability in the area where node B is located. The conditions for constructing directed connecting edges include three types of situations: first, historical data... The risk transmission mechanism is characterized by three key aspects: First, it clearly demonstrates the unidirectional transmission of risk between nodes. First, it identifies instances where an accident in one node propagates unidirectionally to another, such as the instability of a node at the top of a slope inevitably leading to instability at a node in the middle of the slope. Second, it identifies instances where the two nodes are located on opposite sides of weak interlayers or unfavorable structural surfaces (such as faults or joints). Third, it identifies instances where exploration and engineering data show stress transmission within the structural surface, such as stress increasing on one side and then simultaneously transmitting or seeping through to the other side, or groundwater flowing through the structural surface. The engineering significance lies in predicting the path and sequence of risk diffusion, such as reinforcing the stress-transmission source node first to block risk transmission. Regarding risk transmission simulation, when a monitoring node triggers an early warning, a risk association network is used to simulate the transmission path and impact range of the risk to surrounding nodes. For example, nodes on undirected edges experience simultaneous risk, while nodes on directed edges experience escalating risk sequentially according to the transmission direction. Regarding monitoring resource optimization, based on network connectivity, monitoring equipment is densified at key risk transmission nodes, such as the starting point of directed edges and core nodes on undirected edges, to improve the sensitivity of risk detection.

[0028] It is worth noting that by detecting and analyzing mutation points, early signs of instability can be identified several hours to days earlier than traditional warnings that rely solely on stability coefficients, allowing ample time for emergency response. By calculating slip surface characteristic parameters, the spatial location, scale, and evolution trend of risk sources are clearly defined, avoiding blind intervention. In terms of overall control, the risk diffusion path and impact range are predicted through a risk correlation network. Warning triggering requires the fusion of multiple conditions: mutation point detection, stability coefficient, slip surface parameters, and risk correlation, significantly reducing the probability of false and missed warnings. For example, a single data mutation will not trigger a warning; a comprehensive judgment must be made based on slip surface parameters and risk transmission characteristics. Furthermore, all judgment thresholds are determined based on conventional engineering parameters, eliminating the need for additional survey costs. Targeted intervention is provided; the warning information includes the risk location, characteristic parameters, and targeted intervention suggestions, which can be directly implemented by maintenance personnel. High adaptability is ensured; the two rules for mutation point detection and the two connection edges of the risk correlation network can be flexibly selected according to slope geological conditions and engineering grade, making it applicable to various roadbed slopes.

[0029] The physical condition monitoring data also includes rainfall data; The method further includes step six: The rainfall data is input as a boundary condition into the slope digital twin updated by the data processing and inversion steps to calculate the predicted displacement data under the rainfall conditions. The predicted displacement data is compared with the real-time acquired surface displacement data, and the deviation between the two is calculated. When the deviation value exceeds the preset tolerance range, a model anomaly warning message is generated, wherein the tolerance range is set based on the accuracy of the monitoring data and the allowable error of the engineering.

[0030] Specifically, real-time rainfall data and the average rainfall over the past hour are used as boundary conditions and input into the updated digital twin after parameter inversion to ensure that the model parameters are consistent with the current slope condition. The calculation time is set to the next hour, and the seepage-stress coupling calculation module is used to output predicted surface displacement data under this rainfall condition, including displacement and velocity in the X, Y, and Z directions, with a calculation step size of 5 minutes. Real-time surface displacement data within the same 1-hour period, synchronized with the prediction calculation, is obtained, and the absolute deviation between the predicted and measured values ​​is calculated. The deviation value Δs = |predicted displacement value s| pred -Measured displacement value s meas |, Relative deviation rate η=Δs / s meas ×100%, when s measWhen the deviation is less than 0.1 mm, the absolute deviation Δs is used as the judgment index. Based on the accuracy of monitoring data and the allowable error in engineering, the tolerance range is set according to different scenarios: no rainfall: absolute deviation Δs ≤ 0.3 mm or relative deviation rate η ≤ 8%; rainfall: absolute deviation Δs ≤ 0.5 mm or relative deviation rate η ≤ 12%. It is worth noting that the tolerance range can be adjusted according to the importance of the slope. For example, the tolerance range for railway slopes can be reduced by 20%, while that for highway slopes can be increased by 10%. When the deviation value exceeds the tolerance range for two consecutive calculation steps, it is judged as a digital twin anomaly, and a model anomaly warning message is generated, with the priority equivalent to a level two warning. The warning includes the abnormal time period, such as "2025-10-29 14:00-14:10", rainfall conditions, such as "rainfall 4.2mm / h", deviation details, such as "predicted displacement in the X direction 0.8mm, measured displacement 1.5mm, absolute deviation 0.7mm > tolerance 0.5mm; relative deviation rate 46.7% > tolerance 12%", and preliminary judgment of the cause of the anomaly, such as "possibly due to a deviation in the permeability coefficient k setting, or a discrepancy between the rainfall infiltration coefficient α and the actual value". A corresponding response is initiated: within 15 minutes, a model check is launched, prioritizing the checking of parameters related to rainfall, such as permeability coefficient k and infiltration coefficient α, comparing the deviation of current parameters with historical optimal values; the monitoring frequency is temporarily increased to once every 2 minutes, supplementing the collection of deep lateral displacement data and soil pressure data to verify whether there are any local deformations not captured by the model; if the check finds parameter deviations, such as the current value of k being 3×10... -6 cm / s, the optimal value for historical inversion is 5×10 -6 If the error rate is less than 1 cm / s, immediately recalibrate the parameters based on real-time monitoring data. After calibration, re-input rainfall data to calculate the predicted displacement. If the deviation falls back to within the tolerance range, issue a notification to resolve the model anomaly. If the deviation still exceeds the tolerance, check the sensor status and replace faulty equipment if necessary. After resolving the model anomaly, record the cause of the anomaly, the verification process, calibration parameters, and the effect. For example, after correcting the k value, the X-direction deviation decreased from 0.7 mm to 0.3 mm. Update the calibrated parameters to the digital twin and include them in the historical parameter database for subsequent optimization of the initial parameter range of the inversion algorithm.

[0031] The reverse analysis algorithm is executed using any of the following control methods: Method 1: Initialize the particle swarm using the internal mechanical parameters of the slope as optimization variables; The error between the displacement calculated by the digital twin of the slope based on the internal mechanical parameters of the slope and the real-time surface displacement data is used as the fitness function. Iterative updates of particle position and velocity; When the number of iterations reaches a preset value or the fitness function value is less than a preset error threshold, the parameters corresponding to the current global optimal particle are output as the real-time evolution of the slope's internal mechanical parameters. Method 2: Set the prior probability distribution of the internal mechanical parameters of the slope based on the initial geological parameters; A likelihood function is constructed based on the error between the displacement calculated from the digital twin of the slope and the real-time surface displacement data. The posterior probability distribution is solved using the Markov chain Monte Carlo method. The mean or mode of the posterior distribution is taken as the internal mechanical parameter for real-time evolution. When the standard deviation of the posterior distribution is less than the preset accuracy threshold, the inversion stops.

[0032] Specifically, Method 1 is a particle swarm optimization control method. In the initialization phase, the internal mechanical parameters of the slope are used as optimization variables. The particle swarm size and parameter search range are set, and the position and velocity of each particle are randomly initialized. The surface displacement data calculated by the digital twin based on the current particle parameters is compared with the real-time collected surface displacement data, and the average absolute error between the two is used as the fitness function. In the iterative update phase, the position and velocity of each particle are updated based on the individual optimal particle and the globally optimal particle. When the number of iterations reaches a preset value, or the fitness function value is less than a preset error threshold (e.g., ≤0.3mm for a first-level slope and ≤0.5mm for a second-level slope), the iteration stops, and the parameter combination corresponding to the current globally optimal particle is output as the real-time evolving internal mechanical parameters of the slope.

[0033] Method two is the Markov Chain Monte Carlo (MCMC) control method. Based on the mechanical parameters obtained from the initial geological exploration, a prior probability distribution of each parameter is set, usually using a normal or uniform distribution, for example, c ~ N(25kPa, 5 2 ), φ~N(28°, 3 2 The mean is the initial experimental value, and the variance is set according to the dispersion of the experimental data. When the dispersion is high, the variance increases by 20%~30%. Assuming that the error of the real-time surface displacement data follows a normal distribution, a likelihood function is constructed by calculating the error between the displacement and the real-time displacement using a digital twin. A Markov chain is constructed using the Metropolis-Hastings algorithm, with the prior distribution as the initial sampling distribution. Each time, a new parameter candidate value is generated, and the current value is randomly perturbed. The perturbation amplitude is 5%~10% of the parameter range. The acceptance probability is calculated through the likelihood function. If the acceptance probability ≥ [0,1] random number, the candidate value is accepted; otherwise, the current value is retained. This sampling is repeated 10000~20000 times to obtain the posterior probability distribution of the parameters. The standard deviation of the posterior distribution is calculated (e.g., the posterior standard deviation of c ≤ 2 kPa, the posterior standard deviation of k ≤ 1 × 10⁻⁶). -7 (cm / s) When the standard deviation is less than the preset accuracy threshold, the inversion stops; the mean or mode of the posterior distribution is taken as the internal mechanical parameter of the real-time evolution, and the 95% confidence interval of the posterior distribution is output to evaluate the parameter uncertainty.

[0034] It also includes step seven: Obtain weather forecast information for a future preset time period, the weather forecast information including at least rainfall and temperature, and input the weather forecast information as a driving condition into the slope digital twin updated in step three; The stability evolution of the roadbed slope in a future preset period is predicted using the slope digital twin, and a predictive stability report including the predicted stability coefficient and its changing trend is output.

[0035] Specifically, weather forecast information for a preset time period is obtained through the API interface of the national meteorological data platform, such as the China Meteorological Administration's meteorological data network or local meteorological department dedicated lines. The preset time period can be set according to project needs, typically 72 hours, but can be extended to 7 days for major projects. The core content must include: rainfall data, temperature data, and auxiliary information. If the predicted rainfall deviates from the historical measured rainfall for the same period by more than 30%, a historical meteorological-measured rainfall correlation model is used for correction; obvious outliers are removed and replaced with the average of adjacent time periods; the correlation between temperature and soil parameters is established, such as for every 5°C decrease in temperature, the cohesion c of cohesive soil increases by 3%~5%; for every 5°C increase in temperature, the permeability coefficient k of sandy soil increases by 2%~3%, and hourly temperature data is converted into a parameter correction coefficient sequence; the preprocessed weather forecast data is converted into a format recognizable by the digital twin. Digital twin prediction settings and execution model initialization: The digital twin updated in step three is called, and the latest inversion parameters (such as c=26kPa, φ=24°, k=5×10) are loaded. -6 The parameters are set to cm / s, E=23MPa, and ν=0.31 to ensure the initial state of the model matches the actual state of the slope. Simultaneously, the seepage-stress coupling module verified in step six is ​​loaded. Preprocessed weather forecast information is used as the driving condition, and the data is input into the digital twin in modules: hourly rainfall data is input into the seepage module to set the slope crest infiltration boundary (the infiltration coefficient α is dynamically adjusted according to the soil type; for example, when rainfall intensity > 5mm / h, α for cohesive soil is reduced to 0.2~0.3 to avoid ignoring surface runoff); the temperature correction coefficient is input into the parameter module to adjust the c, φ, and k values ​​of the soil in real time (e.g., when the temperature drops to 5℃, c is corrected to 26kPa×1.06=27.56kPa, and k is corrected to 5×10). -6 cm / s × 0.96 = 4.8 × 10 -6(cm / s); Simultaneously input the current groundwater depth and surface load to ensure the synergistic effect of multiple driving conditions. Predictive calculation execution: Set the prediction calculation duration and adopt a dynamic step-size calculation strategy; during the calculation process, record the slope surface displacement, deep displacement, pore water pressure distribution per hour, stability coefficient K every 6 hours, and the evolution curve of key parameters in real time. The core content of the predictive stability report generation and output report: The predictive stability report should include the following modules to ensure complete information and facilitate engineering decision-making: a summary table of preprocessed rainfall, temperature, and parameter correction coefficients for the next 72 hours; a time-series variation curve of the stability coefficient K for the next 72 hours, clearly identifying key nodes (such as periods when K drops below 1.15, and safe periods when K ≥ 1.25); predicted evolution curves of surface displacement and deep displacement, marking periods with displacement rates exceeding 2 mm / h (potential risk periods); based on the predicted K value and deformation rate, classifying the risk level for future periods (safe: K ≥ 1.25 and displacement rate < 1 mm / h; watch: 1.15 ≤ K < 1.25 or 1 mm / h ≤ displacement rate < 2 mm / h; warning: K < 1.15 or displacement rate ≥ 2 mm / h), and displaying the distribution of risk periods in the form of a heat map; and engineering response suggestions for risk periods (such as when rainfall intensity > 5 mm / h and K < 1.15, it is recommended to start the slope toe sandbag loading and temporary drainage pumps in advance; during periods of extreme low temperatures, it is recommended to strengthen the stress monitoring of the support structure). Report output and updates: Generate formal reports and visualization reports; re-perform forecast calculations every 12 hours based on the latest weather forecast information and update the report content to avoid forecast deviations caused by changes in weather forecasts; push the reports to the project management team through the cloud platform, and for major projects, report to the local transportation and emergency management departments simultaneously to ensure multi-party collaboration in responding to potential risks.

[0036] See Figure 2 An automated monitoring data processing system for roadbed slopes, characterized in that it includes: The data acquisition module is used to acquire physical state monitoring data of the roadbed slope through a sensor array deployed on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. The model building module is used to establish a corresponding digital twin of the roadbed slope based on the initial geological parameters and design parameters, and simultaneously build a risk association network inside the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include the type of rock and soil, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. The data processing and inversion module is communicatively connected to both the data acquisition module and the model building module. It receives real-time surface displacement data output by the data acquisition module and uses it as input to dynamically invert the internal mechanical parameters of the slope's digital twin generated by the model building module using a reverse analysis algorithm. The dynamic inversion aims to ensure that the error between the calculated displacement of the slope's digital twin and the surface displacement data is less than a preset error threshold. When the error reaches the preset error threshold, it outputs the real-time evolving internal mechanical parameters of the slope. A state assessment module, which is communicatively connected to the data processing and inversion module, is used to calculate the real-time stability coefficient of the roadbed slope based on the real-time evolving internal mechanical parameters of the slope output by the data processing and inversion module. The early warning decision module is communicatively connected to the state assessment module. It is used to compare the real-time stability coefficient output by the state assessment module with a preset stability threshold and generate corresponding early warning information based on the comparison result.

[0037] Also includes: A meteorological data acquisition module is used to acquire weather forecast information for a future preset period, the weather forecast information including at least rainfall and temperature; The stability prediction module is communicatively connected to the meteorological data acquisition module, the data processing and inversion module, and the model building module. The stability prediction module receives weather forecast information output by the meteorological data acquisition module and uses it as a driving condition, inputting it into the updated slope digital twin by the data processing and inversion module. The stability prediction module also uses the updated slope digital twin to predict the stability evolution of the roadbed slope within a preset future time period and outputs a predictive stability report containing the predicted stability coefficient and its changing trend.

[0038] Specifically, the data acquisition module acquires physical state monitoring data of the roadbed slope by deploying sensor arrays, providing basic data for subsequent analysis. The model building module establishes a digital twin of the slope based on the initial geological and design parameters, and simultaneously constructs a risk correlation network within the slope, providing a model foundation for subsequent analysis. The data processing and inversion module receives real-time surface displacement data from the data acquisition module and dynamically inverts the internal mechanical parameters of the digital twin using a reverse analysis algorithm. The goal is to ensure that the error between the calculated displacement and the surface displacement data is less than a preset threshold. When the error reaches the target, the module outputs the real-time evolving internal mechanical parameters of the slope, achieving real-time updates to the internal mechanical state of the slope. The state assessment module calculates the real-time stability coefficient of the roadbed slope based on the real-time internal mechanical parameters output by the data processing and inversion module, assessing the current safety status of the slope. The early warning decision module compares the real-time stability coefficient output by the state assessment module with a preset threshold and generates corresponding early warning information based on the comparison results, promptly reminding relevant personnel to take measures. The meteorological data acquisition module acquires weather forecast information for a preset future period, providing external condition data for slope stability prediction. The stability prediction module receives weather forecast information from the meteorological data acquisition module and uses it as a driving condition to input into the updated digital twin of the slope after data processing and inversion. This model predicts the stability evolution of the roadbed slope over a preset period and outputs a predictive stability report containing predicted stability coefficients and their trends, allowing for advance understanding of the slope's future safety status. This application achieves full automation and intelligence throughout the entire process from data acquisition to state assessment, early warning decision-making, and future stability prediction, helping to promptly identify potential slope safety hazards, take preventative measures in advance, and ensure the safety and stability of the roadbed slope.

[0039] The above description is merely a few embodiments of this application and is not intended to limit this application in any way. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any changes or modifications made by those skilled in the art without departing from the scope of the technical solution of this application using the disclosed technical content are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A method for automated monitoring data processing of roadbed slopes, characterized in that, include: Step 1: Acquire physical state monitoring data of the roadbed slope by deploying a sensor array on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. Step 2: Based on the initial geological parameters and design parameters of the roadbed slope, establish a corresponding digital twin of the slope and simultaneously construct a risk correlation network within the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include soil and rock type, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. Step 3: Using the real-time acquired surface displacement data as input, the internal mechanical parameters of the slope digital twin are dynamically inverted through the reverse analysis algorithm. The goal is to make the error between the calculated displacement of the slope digital twin and the surface displacement data less than a preset error threshold. When the error reaches the preset error threshold, the real-time evolved internal mechanical parameters of the slope are output. Step 4: Based on the real-time evolved internal mechanical parameters of the slope obtained by inversion, calculate the real-time stability coefficient of the roadbed slope using the limit equilibrium method, strength reduction method, or numerical simulation method. Step 5: Compare the real-time stability coefficient with the preset stability threshold, and generate corresponding early warning information based on the comparison result; Step five also includes: Abrupt change point detection analysis is performed on the time series data composed of the surface displacement data, internal displacement data, and internal mechanical parameter data of the slope to identify whether there is an accelerated change trend in the data that meets the characteristics of instability precursors. When the trend of change based on internal displacement data exceeds the change threshold determined according to soil environmental parameters, it is determined that the monitoring point has reached the bottom of the potential slip surface, and the displacement value recorded at this moment is used as the second displacement data. Meanwhile, the data representing the top displacement of the potential slip surface is defined as the first displacement data, and the characteristic parameters of the potential slip surface are calculated based on the first displacement data and the second displacement data. When the accelerating change trend is identified, and the risk is determined to reach a critical level based on the slip surface characteristic parameters, a critical instability warning message is generated and issued. The rules for constructing the risk association network include: If the geotechnical materials where the two monitoring nodes are located have similar engineering properties or belong to the same homogeneous unit, and historical monitoring data confirms that the geotechnical materials have mutual influence on pore water pressure and soil pressure parameters, then an undirected connection edge is established between the two monitoring nodes. If historical data clearly indicates that an accident in the area where a certain monitoring node is located will propagate unidirectionally to the area where another monitoring node is located, or if the two monitoring nodes are located on opposite sides of a weak interlayer or unfavorable structural surface, and exploration and engineering data show that the structural surface exhibits stress transmission and seepage penetration, then a directed connection edge is established between the two monitoring nodes.

2. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, In step five, the stability threshold includes a first stability threshold, a second stability threshold, and a critical stability threshold, and the first stability threshold > the second stability threshold > the critical stability threshold; When the real-time stability coefficient is lower than the first stability threshold but higher than the second stability threshold, a first-level early warning message is generated, which indicates that the slope safety reserve has decreased. When the real-time stability coefficient is lower than the second stability threshold but higher than the critical stability threshold, a secondary early warning message is generated, which indicates that the slope is close to instability. When the real-time stability coefficient approaches or reaches the critical stability threshold, a level-three early warning message is generated, which indicates an extremely high risk of instability.

3. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, In step three, the internal mechanical parameters of the slope include at least one of the following: cohesion c of the soil and rock mass, internal friction angle φ, elastic modulus E, and Poisson's ratio ν.

4. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, The mutation point detection and analysis includes any of the following rules: Rule 1: Acquire the time series data, calculate the displacement rate and acceleration of the data series, and when the acceleration value continues to increase and exceeds the preset acceleration threshold, it is determined that an acceleration change trend that meets the characteristics of instability precursors has been identified. Rule 2: Divide historical time series data into training and validation sets to train the model to learn the data evolution patterns; input real-time time series data into the trained model to predict data trends within a preset future time period; The deviation between the predicted value and the real-time observed value is calculated. When the deviation continuously exceeds the preset deviation threshold and the data shows a non-linear accelerating growth characteristic, it is determined that an accelerating change trend that meets the characteristics of instability precursors has been identified.

5. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, The physical condition monitoring data also includes rainfall data; The method further includes step six: The rainfall data is input as a boundary condition into the slope digital twin updated by the data processing and inversion steps to calculate the predicted displacement data under the rainfall conditions. The predicted displacement data is compared with the real-time acquired surface displacement data, and the deviation between the two is calculated. When the deviation value exceeds the preset tolerance range, a model anomaly warning message is generated, wherein the tolerance range is set based on the accuracy of the monitoring data and the allowable error of the engineering.

6. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, The reverse analysis algorithm is executed using any of the following control methods: Method 1: Initialize the particle swarm using the internal mechanical parameters of the slope as optimization variables; The error between the displacement calculated by the digital twin of the slope based on the internal mechanical parameters of the slope and the real-time surface displacement data is used as the fitness function. Iterative updates of particle position and velocity; When the number of iterations reaches a preset value or the fitness function value is less than a preset error threshold, the parameters corresponding to the current global optimal particle are output as the real-time evolution of the slope's internal mechanical parameters. Method 2: Set the prior probability distribution of the internal mechanical parameters of the slope based on the initial geological parameters; A likelihood function is constructed based on the error between the displacement calculated from the digital twin of the slope and the real-time surface displacement data. The posterior probability distribution is solved using the Markov chain Monte Carlo method. The mean or mode of the posterior distribution is taken as the internal mechanical parameter for real-time evolution. When the standard deviation of the posterior distribution is less than the preset accuracy threshold, the inversion stops.

7. The method for processing automated monitoring data of roadbed slopes according to claim 1, characterized in that, It also includes step seven: Obtain weather forecast information for a future preset time period, the weather forecast information including at least rainfall and temperature, and input the weather forecast information as a driving condition into the slope digital twin updated in step three; The stability evolution of the roadbed slope in a future preset period is predicted using the slope digital twin, and a predictive stability report including the predicted stability coefficient and its changing trend is output.

8. An automated monitoring data processing system for roadbed slopes, employing the automated monitoring data processing method for roadbed slopes as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire physical state monitoring data of the roadbed slope through a sensor array deployed on the roadbed slope, wherein the sensor array includes at least surface displacement sensors and internal displacement sensors, and the physical state monitoring data includes at least surface displacement data and internal displacement data. The model building module is used to establish a corresponding digital twin of the roadbed slope based on the initial geological parameters and design parameters, and simultaneously build a risk association network inside the slope. The digital twin of the slope is a parametric mechanical calculation model based on the finite element, finite difference, or limit equilibrium method. The initial geological parameters include the type of rock and soil, distribution characteristics, geological structure information, and survey point data. The design parameters include slope height, slope ratio, and support structure parameters. The data processing and inversion module is communicatively connected to both the data acquisition module and the model building module. It receives real-time surface displacement data output by the data acquisition module and uses it as input to dynamically invert the internal mechanical parameters of the slope's digital twin generated by the model building module using a reverse analysis algorithm. The dynamic inversion aims to ensure that the error between the calculated displacement of the slope's digital twin and the surface displacement data is less than a preset error threshold. When the error reaches the preset error threshold, it outputs the real-time evolving internal mechanical parameters of the slope. A state assessment module, which is communicatively connected to the data processing and inversion module, is used to calculate the real-time stability coefficient of the roadbed slope based on the real-time evolving internal mechanical parameters of the slope output by the data processing and inversion module. The early warning decision module is communicatively connected to the state assessment module. It is used to compare the real-time stability coefficient output by the state assessment module with a preset stability threshold and generate corresponding early warning information based on the comparison result.

9. The automated monitoring data processing system for roadbed slopes according to claim 8, characterized in that, Also includes: A meteorological data acquisition module is used to acquire weather forecast information for a future preset period, the weather forecast information including at least rainfall and temperature; The stability prediction module is communicatively connected to the meteorological data acquisition module, the data processing and inversion module, and the model building module. The stability prediction module receives weather forecast information output by the meteorological data acquisition module and uses it as a driving condition, inputting it into the updated slope digital twin by the data processing and inversion module. The stability prediction module also uses the updated slope digital twin to predict the stability evolution of the roadbed slope within a preset future time period and outputs a predictive stability report containing the predicted stability coefficient and its changing trend.

Images