High fill deep excavation roadbed slope state real-time monitoring and early warning method

Abstract

The application provides a high filling and deep excavation roadbed slope state real-time monitoring and early warning method, a sensing terminal is arranged according to the geological conditions and potential risks of the high filling and deep excavation roadbed slope, and multi-dimensional monitoring data of the slope are collected; the collected multi-source monitoring data are optimized to improve the data quality and consistency; the data after the optimization are subjected to multi-dimensional fusion analysis to form comprehensive data capable of reflecting the overall state of the slope; the safety risks of the slope are evaluated according to the comprehensive data, and corresponding early warning information and disposal suggestions are output. The problems of single slope monitoring information, blind area, early warning lag and lack of self-adaptive capability are solved.

Description

Technical Field

[0001] This invention relates to the field of road construction technology, and in particular to a method for real-time monitoring and early warning of the condition of slopes in high-fill and deep-cut roadbeds. Background Technology

[0002] With the continuous and rapid development of my country's transportation infrastructure construction and the ever-expanding scale of the expressway network, a large number of high-fill and deep-cut roadbed slope projects have emerged. In these projects, high-fill slopes are highly susceptible to uneven settlement of the fill material and slope slippage and instability; deep-cut slopes are often subject to complex geological structures, groundwater activity, and strong unloading effects, facing serious threats from geological disasters such as collapses, landslides, and even debris flows. The stability of high-fill and deep-cut roadbed slopes is directly related to the long-term operational safety and service performance of lifeline projects such as highways and railways.

[0003] Existing monitoring methods mainly rely on manual inspections, single-point displacement gauges, and rain gauges, which provide limited information and are insufficient to comprehensively reflect the overall condition and internal changes of the slope. Manual inspections are time-consuming, and traditional sensors have low data acquisition frequencies, making it difficult to capture sudden or accelerated deformations in a timely manner. Point sensors cannot cover the entire slope, especially large and complex slopes, resulting in monitoring blind spots. Traditional methods are sensitive to surface deformation but are insufficient in monitoring key internal factors such as deep stress and strain and seepage. Data analysis relies on human experience, and the warning thresholds are fixed, making it difficult to adapt to complex working conditions and early risk identification. Summary of the Invention

[0004] The main objective of this invention is to provide a method for real-time monitoring and early warning of the condition of roadbed slopes with high fill and deep cut, which solves the problems of limited slope monitoring information, blind spots, delayed early warning, and lack of adaptive capability.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for real-time monitoring and early warning of the condition of roadbed slopes with high fill and deep cut, the method comprising: S1. Based on the geological conditions and potential risks of high-fill and deep-cut roadbed slopes, deploy suitable sensing terminals to collect multi-dimensional monitoring data of the slopes. S2. Optimize the collected multi-source monitoring data to improve data quality and consistency; S3. Perform multi-dimensional fusion analysis on the optimized data to form comprehensive data that reflects the overall condition of the slope; S4. Assess slope safety risks based on the comprehensive data and output corresponding early warning information and handling suggestions.

[0006] In the preferred embodiment, in step S1: Deploy ground-based synthetic aperture radar to acquire large-scale, high-precision, planar surface deformation fields, and deploy high-precision GNSS receivers as key point displacement benchmarks and verifications; Array-type microelectromechanical system inclinometers are installed at key profiles and potential slip zone locations to monitor tilt changes, and fiber optic grating sensors are embedded to form a distributed network for real-time monitoring of soil strain, temperature, and vibration. Multi-parameter inclinometers are installed in deep boreholes to monitor deep displacement, pore water pressure sensors are deployed to monitor groundwater changes, and earth pressure cells are buried to monitor the internal stress state of the soil. Set up meteorological stations to collect data on rainfall, temperature, humidity, wind speed and direction, and set up video monitoring to assist in visualization and anomaly confirmation; In deep excavation areas prone to rock mass fracturing, microseismic or acoustic emission sensor arrays are deployed to capture signals of the initiation and propagation of fractures within the rock mass.

[0007] In the preferred embodiment, in step S2: Based on the progress of slope construction and the law of deformation evolution, the data acquisition frequency of the deployed sensing terminals is dynamically adjusted. The acquisition frequency is increased in areas with high deformation rate or large construction disturbance, and the acquisition frequency is decreased in areas where deformation tends to be stable. The reliability coefficient of each sensing terminal is evaluated in real time. The reliability coefficient is determined by comprehensively analyzing data integrity, data consistency, data trend rationality and hardware working status. When the reliability coefficient is lower than the failure threshold, the sensor data is automatically removed or its weight in subsequent fusion calculations is reduced. Based on the comprehensive evaluation results of deformation evolution law and sensor reliability coefficient, similar sensors are added in areas with insufficient monitoring capabilities, and sensors are removed in areas that have been in long-term failure or are redundant.

[0008] In the preferred embodiment, step S2 involves preprocessing the collected multi-source monitoring data, including: The data collected by each sensing terminal is timestamped to ensure that the data from different sensors at the same time are comparable; By unifying monitoring data from different coordinate systems to the same spatial reference datum, precise matching of spatial locations can be achieved. An adaptive filtering algorithm based on operating conditions is used to filter out environmental noise and abnormal jump values ​​in the original signal, while retaining the effective signal components. Missing values ​​that occur during data collection are imputed to form a continuous and complete monitoring time series. Normalize or standardize monitoring data with different dimensions to eliminate the impact of dimensional differences on subsequent analysis.

[0009] In the preferred scheme, step S3 involves feature extraction and index construction of the preprocessed data, including: Based on surface deformation monitoring data, the comprehensive deformation rate and deformation acceleration are calculated to characterize the dynamic features of slope surface deformation. Based on distributed optical fiber strain monitoring data, a strain concentration index is calculated to identify local strain anomaly areas. Based on pore water pressure monitoring data, the rate of change of pore water pressure is calculated to reflect the dynamic changes in groundwater activity. Based on meteorological monitoring data, rainfall intensity and cumulative rainfall are extracted, and the rainfall impact coefficient is calculated by combining historical data; Based on microseismic monitoring data, the frequency of microseismic events, energy release rate, and microseismic activity index are calculated to assess the degree of rock mass fracturing activity.

[0010] In the preferred scheme, multi-source monitoring data are collaboratively fused to generate a continuous state field distribution, including: Multi-scale collaborative fusion of point sensor monitoring data, area radar monitoring data, and distributed optical fiber monitoring data is carried out. The three-dimensional strain field constructed by the distributed optical fiber monitoring data is used as the basic framework. The overall trend of the basic framework is corrected by the area deformation field obtained by the ground-based synthetic aperture radar. The accuracy of key points in the corresponding positions in the basic framework is calibrated by the point monitoring data obtained by the global navigation satellite system receiver and array inclinometer. High-precision, blind-zone-free three-dimensional displacement field and three-dimensional strain field are generated. Spatiotemporal coupling analysis was performed on monitoring data of deep displacement, deep stress and pore water pressure to construct a deep multi-parameter coupling matrix. The correlation coefficients of each parameter at the same spatial location and time node were calculated. When the correlation coefficients exceeded the set threshold and showed a continuous upward trend, they were identified as deep multi-parameter coupling anomaly areas. By correlating meteorological monitoring data, pore water pressure monitoring data, and soil deformation monitoring data, a temporal mapping relationship between rainfall infiltration and soil response is established, generating a dynamic state field that takes into account hydrological influences.

[0011] In the preferred scheme, the quantitative index of slope stability is calculated based on the results of the fusion analysis, including: The extracted key feature indicators are input into a pre-built risk assessment model, and a dynamic weight allocation mechanism is used to calculate the comprehensive risk index. The weight of each indicator is adjusted in real time based on the sensor reliability coefficient, current environmental conditions, and historical performance. Based on the generated three-dimensional displacement field and three-dimensional strain field, combined with the slope geological model, the current safety factor of the slope is estimated in real time using the limit equilibrium method or the finite element simplified model. The comprehensive risk index and safety factor are combined to calculate a quantitative assessment value of the slope instability probability, which is used for subsequent risk level determination.

[0012] In the preferred scheme, the slope safety risk is dynamically assessed and the risk level is determined based on the fusion analysis results, including: The calculated comprehensive risk index, safety factor, and instability probability are used as inputs. Combined with the slope geological model and historical monitoring data, a dynamic Bayesian network is used to update the slope instability probability in real time, and the failure mode of the current slope is identified based on the changing trend of the instability probability. Based on real-time operating condition characteristic parameters, the early warning thresholds of each evaluation index are dynamically calibrated. The operating condition characteristic parameters include rainfall intensity, construction disturbance degree, and microseismic activity index. The risk level is divided into three levels: concern level, warning level, and alarm level by comprehensively judging the instability probability output by the dynamic Bayesian network, the damage mode identification result, and the dynamically calibrated threshold.

[0013] In the preferred scheme, the risk level is divided into three levels: Attention Level, Warning Level, and Alert Level, as follows: When the probability of instability is in the first preset range, or when a single key indicator exceeds the dynamic calibration threshold but does not show a significant deterioration trend, it is judged as a warning of concern level, and a warning of concern is issued and it is recommended to strengthen daily inspections. When the probability of instability is in the second preset range, or the key indicators deteriorate significantly, the deep multi-parameter coupling is abnormally significant, or the safety factor is predicted to be lower than the critical value within a set time in the future, it is judged as a warning level warning, and a warning message is issued and a reminder is given to strengthen manual verification and encrypted monitoring. When the probability of instability is in the third preset range, or when accelerated deformation, large-scale micro-seismic events, or the safety factor is below the critical value and continues to decline, it is determined to be an alarm-level warning, an emergency alarm is issued, and it is recommended to take emergency measures immediately.

[0014] In the preferred solution, early warning information and corresponding handling suggestions are output, including: Based on the determined warning level and risk location, a basic response plan is obtained by matching the preset rule base. At the same time, the most similar successful response case is retrieved from the historical case database to modify the basic response plan and generate an emergency response instruction that includes response measures, responsible parties, resource allocation paths and time requirements. The platform uses a 3D visualization method to display slope topography and geological models, multi-source sensor locations, real-time monitoring data, 3D state field cloud maps, risk level distribution, and early warning information. Warning information and emergency response instructions are pushed to relevant management personnel through at least one of the following methods: platform pop-up, SMS, or email, and execution feedback is received to complete closed-loop management.

[0015] This invention provides a method for real-time monitoring and early warning of the condition of slopes in high-fill and deep-cut roadbeds. Addressing the practical needs of slope monitoring in high-fill and deep-cut roadbeds, it achieves multi-dimensional and comprehensive coverage of slope monitoring. By deploying various sensing terminals in different regions and combining point, area, and distributed monitoring methods, it not only achieves large-scale monitoring of surface deformation but also completes real-time monitoring of key parameters such as deep stress, displacement, and groundwater. Simultaneously, it also considers the synchronous monitoring of rock mass fracture and meteorological environment, effectively compensating for the blind spots of traditional monitoring and comprehensively reflecting the overall condition and internal changes of the slope.

[0016] The optimized design of the data processing stage has significantly improved the quality and effectiveness of monitoring data. By dynamically adjusting the acquisition frequency, the monitoring resources are rationally allocated. Combined with sensor reliability assessment and adaptive optimization of the monitoring network, invalid data and resource waste are avoided. At the same time, the standardized preprocessing process eliminates the differences in time, space and units of different monitoring data. Multi-scale and multi-dimensional data fusion analysis enables in-depth mining of monitoring data. The generated three-dimensional state field can accurately present the real-time state of the slope, providing scientific and comprehensive data support for risk assessment.

[0017] The slope risk assessment system achieves dynamic and precise early warning judgment. By constructing a multi-dimensional feature index system and combining dynamic weight allocation with physical and mechanical models, it completes the quantitative assessment of slope stability. The application of dynamic Bayesian networks can update the instability probability in real time. With the dynamic calibration of the early warning threshold, it effectively adapts to complex and ever-changing engineering conditions and avoids the false alarm and missed alarm problems caused by traditional fixed thresholds. The hierarchical early warning mechanism realizes differentiated responses to different risk levels and improves the pertinence of risk management.

[0018] The early warning, response, and management processes form a complete closed-loop workflow. By combining the rule base and historical case database, the generated emergency response instructions are both practical and scientific. The 3D visualization platform enables an intuitive display of monitoring and early warning information, multi-channel information push ensures the timely transmission of early warning information, and the execution feedback mechanism ensures the effective implementation of response measures. This significantly improves the prevention and control capabilities of geological disasters on high-fill and deep-cut roadbed slopes, and ensures the operational safety and service performance of transportation infrastructure. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the overall scheme of the present invention; Figure 2 This is a flowchart of the multi-source monitoring data optimization processing of the present invention. Detailed Implementation

[0020] Example 1 like Figure 1-2 As shown, a method for real-time monitoring and early warning of the condition of roadbed slopes with high fills and deep cuts is presented. The method includes: S1. Based on the geological conditions and potential risks of high-fill and deep-cut roadbed slopes, deploy suitable sensing terminals to collect multi-dimensional monitoring data of the slopes. S2. Optimize the collected multi-source monitoring data to improve data quality and consistency; S3. Perform multi-dimensional fusion analysis on the optimized data to form comprehensive data that reflects the overall condition of the slope; S4. Assess slope safety risks based on the comprehensive data and output corresponding early warning information and handling suggestions.

[0021] In the preferred embodiment, in step S1: Deploy ground-based synthetic aperture radar to acquire large-scale, high-precision, planar surface deformation fields, and deploy high-precision GNSS receivers as key point displacement benchmarks and verifications; Array-type microelectromechanical system inclinometers are installed at key profiles and potential slip zone locations to monitor tilt changes, and fiber optic grating sensors are embedded to form a distributed network for real-time monitoring of soil strain, temperature, and vibration. Multi-parameter inclinometers are installed in deep boreholes to monitor deep displacement, pore water pressure sensors are deployed to monitor groundwater changes, and earth pressure cells are buried to monitor the internal stress state of the soil. Set up meteorological stations to collect data on rainfall, temperature, humidity, wind speed and direction, and set up video monitoring to assist in visualization and anomaly confirmation; In deep excavation areas prone to rock mass fracturing, microseismic or acoustic emission sensor arrays are deployed to capture signals of the initiation and propagation of fractures within the rock mass.

[0022] In the preferred embodiment, in step S2: Based on the progress of slope construction and the law of deformation evolution, the data acquisition frequency of the deployed sensing terminals is dynamically adjusted. The acquisition frequency is increased in areas with high deformation rate or large construction disturbance, and the acquisition frequency is decreased in areas where deformation tends to be stable. The reliability coefficient of each sensing terminal is evaluated in real time. The reliability coefficient is determined by comprehensively analyzing data integrity, data consistency, data trend rationality and hardware working status. When the reliability coefficient is lower than the failure threshold, the sensor data is automatically removed or its weight in subsequent fusion calculations is reduced. Based on the comprehensive evaluation results of deformation evolution law and sensor reliability coefficient, similar sensors are added in areas with insufficient monitoring capabilities, and sensors are removed in areas that have been in long-term failure or are redundant.

[0023] To achieve efficient allocation of monitoring resources, this embodiment dynamically adjusts the acquisition frequency of each sensing terminal based on the slope deformation evolution law. Let the deformation rate at a certain sensor location at time t be... Construction disturbance intensity is Then the target acquisition frequency of the sensor Determined dynamically according to the following formula:

[0024] in, and These are the minimum and maximum sampling frequencies of the sensor, respectively. The reference deformation rate is 50%, which is set to 50% of the design warning rate in this embodiment; , These are weighting coefficients, reflecting the degree of influence of deformation rate and construction disturbance on frequency adjustment, respectively, and are selected according to the importance of the project; Cut off to interval, Also return to oneness .

[0025] This formula enables automatic intensification of monitoring during periods of accelerated deformation or intense construction disturbances, and automatic frequency reduction during periods of stability, thereby ensuring monitoring effectiveness while reducing data redundancy and energy consumption.

[0026] To ensure the reliability of the monitoring data, a reliability coefficient is calculated in real time for each sensor. This coefficient is calculated by weighting four dimensions:

[0027] in, For data integrity, it is the ratio of the actual number of data collections to the theoretical number of data collections; Data consistency refers to the degree of matching with neighboring sensors or historical averages; For the rationality of the trend, it is the ratio of the rate of change to the historical average rate of change; The hardware status is determined by self-test information; weights , , , satisfy It can be preset according to the sensor type and engineering experience.

[0028] When the reliability coefficient of a certain sensor Below the preset failure threshold If the sensor's current data is deemed unavailable, its weight will be determined in subsequent data fusion calculations. Correct it as follows:

[0029] in, This represents the basic weight of the sensor in the fusion model. The normal threshold is when When the data is below the failure threshold, the sensor data is completely discarded; when it is in an intermediate state, the weight is reduced proportionally to avoid unreliable data interfering with the analysis results.

[0030] Based on a comprehensive evaluation of deformation evolution patterns and sensor reliability coefficients, this application dynamically decides whether to add or remove sensors. A monitoring capability index is defined. Characterizes the coverage capability of the current detection network at point x:

[0031] in, Let be the reliability coefficient of sensor i. Let x be the distance between position x and sensor i, and L be the radius of influence. Essentially, it is a spatially weighted sum of sensor reliability, reflecting the effective coverage of monitoring at the current location.

[0032] In areas with high deformation rates or significant construction disturbances. Below the preset coverage threshold If the reliability coefficient of a sensor remains below the failure threshold for an extended period, it is determined that the monitoring capability is insufficient, triggering an instruction to install additional sensors of the same type in that area; If other sensors in the vicinity can provide sufficient coverage, then the sensor is determined to be redundant or malfunctioning, triggering a cancellation command.

[0033] The above quantitative model enables adaptive optimization of the monitoring network, ensuring high-reliability monitoring capabilities in key areas while avoiding resource waste.

[0034] In the preferred embodiment, step S2 involves preprocessing the collected multi-source monitoring data, including: The data collected by each sensing terminal is timestamped to ensure that the data from different sensors at the same time are comparable; By unifying monitoring data from different coordinate systems to the same spatial reference datum, precise matching of spatial locations can be achieved. An adaptive filtering algorithm based on operating conditions is used to filter out environmental noise and abnormal jump values ​​in the original signal, while retaining the effective signal components. Missing values ​​that occur during data collection are imputed to form a continuous and complete monitoring time series. Normalize or standardize monitoring data with different dimensions to eliminate the impact of dimensional differences on subsequent analysis.

[0035] Preprocessing multi-source monitoring data is a fundamental step in ensuring the quality of subsequent fusion analysis. Since the sampling frequencies, time bases, spatial coordinate systems, and data formats of various sensing terminals differ, this embodiment first performs time synchronization processing. Using a unified time server as the reference, the timestamps of data collected by each sensor are aligned. For sensors with inconsistent sampling frequencies, linear interpolation is used to resample the data to a unified time axis, ensuring the comparability of data from different sensors at the same time. Based on this, spatial registration is performed. Monitoring data from ground-based synthetic aperture radar, global navigation satellite system receivers, array inclinometers, and other different coordinate systems are unified into an engineering-independent coordinate system using a seven-parameter coordinate transformation model, achieving precise spatial matching and laying the foundation for subsequent spatial analysis.

[0036] To address the environmental noise and anomalous fluctuations in the original monitoring signals, this invention employs an adaptive filtering algorithm for noise reduction. For microseismic signals, wavelet thresholding is used, with the threshold dynamically adjusted based on the background noise level. For fiber optic strain signals, sliding window mean filtering is employed. For pore water pressure signals, median filtering is used, and dynamic calibration is performed to address baseline drift caused by rainfall. For data gaps due to communication interruptions or sensor malfunctions, short-term gaps are filled using linear interpolation, while long-term gaps are filled using multivariate regression interpolation combined with data from nearby sensors. Finally, monitoring data with different dimensions are normalized to eliminate the impact of dimensional differences on subsequent analysis.

[0037] The formula for time synchronization resampling is:

[0038] in, and For the time of the known data points, , t represents the data value at the corresponding time, and t represents the target time.

[0039] The wavelet thresholding denoising method is as follows:

[0040] in, These are wavelet coefficients. The adaptive threshold for the j-th layer decomposition can be determined based on the noise level estimation of that layer.

[0041] The formula for normalizing the maximum and minimum values ​​is:

[0042] in, This is the original data. , These are the minimum and maximum values ​​of the indicator, respectively. This is the normalized data.

[0043] In the preferred scheme, feature extraction and index construction are performed on the preprocessed data, including: Based on surface deformation monitoring data, the comprehensive deformation rate and deformation acceleration are calculated to characterize the dynamic features of slope surface deformation. Based on distributed optical fiber strain monitoring data, a strain concentration index is calculated to identify local strain anomaly areas. Based on pore water pressure monitoring data, the rate of change of pore water pressure is calculated to reflect the dynamic changes in groundwater activity. Based on meteorological monitoring data, rainfall intensity and cumulative rainfall are extracted, and the rainfall impact coefficient is calculated by combining historical data; Based on microseismic monitoring data, the frequency of microseismic events, energy release rate, and microseismic activity index are calculated to assess the degree of rock mass fracturing activity.

[0044] To extract key information reflecting slope stability from massive monitoring data, this invention constructs a multi-dimensional characteristic index system, transforming raw monitoring data into stability indicators with clear physical meaning. Based on surface deformation monitoring data, the comprehensive deformation rate is first calculated, which is the average displacement change rate of all effective measuring points within the monitoring area, used to characterize the overall activity level of slope surface deformation. Further, the deformation acceleration, i.e., the rate of change of the deformation rate, is calculated. This indicator can effectively identify the trend of accelerating deformation and is an important criterion for determining whether a slope has entered the accelerated creep stage.

[0045] Based on distributed fiber optic strain monitoring data, this invention calculates a strain concentration index, defined as the ratio of local strain to the regional average strain. When this ratio exceeds a set threshold, it indicates the presence of strain concentration in the region, serving as an important basis for identifying potential slip zones. Based on pore water pressure monitoring data, the pore water pressure change rate is calculated, i.e., the change in pore water pressure per unit time. A rapidly increasing pore pressure change rate is often closely related to rainfall infiltration or groundwater recharge, effectively reflecting the dynamic changes in groundwater activity.

[0046] Based on meteorological monitoring data, a rainfall impact coefficient is constructed, which comprehensively considers the current rainfall intensity and the cumulative rainfall considering the attenuation effect. By setting an attenuation coefficient, the coefficient reflects the continuous impact of previous rainfall on slope stability and can quantify the potential impact of rainfall on slope stability.

[0047] Based on microseismic monitoring data, the microseismic activity index is used to assess the degree of rock mass fracturing activity. When the frequency of microseismic events increases and the energy release rate rises, the b-value decreases, and the microseismic activity index increases accordingly, indicating an increase in microfractures within the rock mass and a higher risk of instability. Through the above multi-dimensional characteristic index system, the slope condition is comprehensively reflected from multiple dimensions such as deformation, stress, seepage, environment, and fracturing, providing a quantitative basis for subsequent risk assessment.

[0048] In the preferred scheme, multi-source monitoring data are collaboratively fused to generate a continuous state field distribution, including: Multi-scale collaborative fusion of point sensor monitoring data, area radar monitoring data, and distributed optical fiber monitoring data is carried out. The three-dimensional strain field constructed by the distributed optical fiber monitoring data is used as the basic framework. The overall trend of the basic framework is corrected by the area deformation field obtained by the ground-based synthetic aperture radar. The accuracy of key points in the corresponding positions in the basic framework is calibrated by the point monitoring data obtained by the global navigation satellite system receiver and array inclinometer. High-precision, blind-zone-free three-dimensional displacement field and three-dimensional strain field are generated. Spatiotemporal coupling analysis was performed on monitoring data of deep displacement, deep stress and pore water pressure to construct a deep multi-parameter coupling matrix. The correlation coefficients of each parameter at the same spatial location and time node were calculated. When the correlation coefficients exceeded the set threshold and showed a continuous upward trend, they were identified as deep multi-parameter coupling anomaly areas. By correlating meteorological monitoring data, pore water pressure monitoring data, and soil deformation monitoring data, a temporal mapping relationship between rainfall infiltration and soil response is established, generating a dynamic state field that takes into account hydrological influences.

[0049] Different types of monitoring data are complementary in terms of spatial coverage and physical meaning. This application adopts a point-surface-volume collaborative fusion strategy. A three-dimensional strain field constructed from distributed fiber optic monitoring data serves as the basic framework. A planar deformation field acquired by ground-based synthetic aperture radar is used to correct the overall trend. Point monitoring data acquired by a global navigation satellite system receiver and an array inclinometer are used to calibrate the accuracy of key points, generating a high-precision, blind-spot-free three-dimensional displacement field and a three-dimensional strain field.

[0050] For deep monitoring data, this invention performs spatiotemporal coupling analysis on deep displacement, deep stress, and pore water pressure monitoring data, calculating the correlation coefficients of each parameter at the same spatial location and time node. When the correlation coefficient exceeds a threshold and shows a continuous upward trend, it is identified as a deep multi-parameter coupling anomaly area. Regarding hydrological factors, the optimal physical time delay of rainfall's impact on pore water pressure is determined by calculating the transfer entropy, establishing a time-series mapping model between rainfall infiltration and soil response, and generating a dynamic state field considering hydrological influences.

[0051] In the preferred scheme, the quantitative index of slope stability is calculated based on the results of the fusion analysis, including: The extracted key feature indicators are input into a pre-built risk assessment model, and a dynamic weight allocation mechanism is used to calculate the comprehensive risk index. The weight of each indicator is adjusted in real time based on the sensor reliability coefficient, current environmental conditions, and historical performance. Based on the generated three-dimensional displacement field and three-dimensional strain field, combined with the slope geological model, the current safety factor of the slope is estimated in real time using the limit equilibrium method or the finite element simplified model. The comprehensive risk index and safety factor are combined to calculate a quantitative assessment value of the slope instability probability, which is used for subsequent risk level determination.

[0052] Based on the three-dimensional state field and characteristic indicators obtained from fusion analysis, this invention calculates the quantitative assessment value of slope stability. In the calculation of the comprehensive risk index, a dynamic weight allocation mechanism is adopted. First, the basic weight of each indicator is calculated using the entropy weight method, and the weight is determined according to the dispersion of the indicator data. Then, the sensor reliability coefficient is introduced to dynamically correct the basic weight, so that the risk assessment model can adapt to different working conditions and sensor states.

[0053] In terms of safety factor estimation, based on the generated three-dimensional displacement and strain fields and combined with the slope geological model, the limit equilibrium method is used to estimate the current safety factor of the slope in real time. This factor reflects the stability of the slope under current load and hydrological conditions. Finally, the logistic regression method is used to integrate the comprehensive risk index and the safety factor to calculate the slope instability probability. This combines the real-time state reflected by the monitoring data with the constraints of the physical and mechanical model, providing a scientific basis for subsequent early warning level determination.

[0054] In the preferred scheme, the slope safety risk is dynamically assessed and the risk level is determined based on the fusion analysis results, including: The calculated comprehensive risk index, safety factor, and instability probability are used as inputs. Combined with the slope geological model and historical monitoring data, a dynamic Bayesian network is used to update the slope instability probability in real time, and the failure mode of the current slope is identified based on the changing trend of the instability probability. Based on real-time operating condition characteristic parameters, the early warning thresholds of each evaluation index are dynamically calibrated. The operating condition characteristic parameters include rainfall intensity, construction disturbance degree, and microseismic activity index. The risk level is divided into three levels: concern level, warning level, and alarm level by comprehensively judging the instability probability output by the dynamic Bayesian network, the damage mode identification result, and the dynamically calibrated threshold.

[0055] In the preferred scheme, the risk level is divided into three levels: Attention Level, Warning Level, and Alert Level, as follows: When the probability of instability is in the first preset range, or when a single key indicator exceeds the dynamic calibration threshold but does not show a significant deterioration trend, it is judged as a warning of concern level, and a warning of concern is issued and it is recommended to strengthen daily inspections. When the probability of instability is in the second preset range, or the key indicators deteriorate significantly, the deep multi-parameter coupling is abnormally significant, or the safety factor is predicted to be lower than the critical value within a set time in the future, it is judged as a warning level warning, and a warning message is issued and a reminder is given to strengthen manual verification and encrypted monitoring. When the probability of instability is in the third preset range, or when accelerated deformation, large-scale micro-seismic events, or the safety factor is below the critical value and continues to decline, it is determined to be an alarm-level warning, an emergency alarm is issued, and it is recommended to take emergency measures immediately.

[0056] Building upon the completion of multi-source data fusion and stability quantification index calculation, this invention further enables dynamic assessment and graded early warning of slope risk. To accurately reflect the temporal evolution characteristics of slope condition, this invention employs a dynamic Bayesian network to construct a risk assessment model. The comprehensive risk index, safety factor, and instability probability calculated in step S6 are used as observation nodes. Combined with the slope geological model and historical monitoring data, the slope instability probability is updated in real time through Bayesian inference. This dynamic Bayesian network effectively handles uncertainties in monitoring data and continuously corrects the judgment of slope condition based on newly acquired monitoring evidence, achieving dynamic tracking of slope instability risk.

[0057] To address the problem that traditional early warning methods use fixed thresholds that are difficult to adapt to complex operating conditions, this invention introduces a dynamic self-calibration mechanism for early warning thresholds. Based on real-time operating condition characteristics, including rainfall intensity, construction disturbance levels, and microseismic activity indices, the early warning thresholds for each evaluation indicator are adjusted in real time. For example, under conditions of continuous heavy rainfall, the threshold for the pore water pressure change rate is appropriately increased to eliminate short-term fluctuations, while the threshold for the deformation rate is appropriately decreased to enhance sensitivity to rainfall-induced deformation. Through this dynamic threshold calibration, the early warning system can adapt to different operating conditions, effectively reducing false alarm and missed alarm rates.

[0058] In terms of risk level determination, this invention uses the instability probability output by the dynamic Bayesian network as the core criterion, and combines it with the damage mode identification results and dynamically calibrated thresholds for comprehensive judgment, classifying the risk level into three levels: Attention Level, Warning Level, and Alert Level. Attention Level warnings correspond to situations where the instability probability is low but shows an abnormal trend; in this case, an attention alert is issued and daily inspections are recommended to be strengthened. Warning Level warnings correspond to situations where the instability probability reaches the moderate risk threshold, or key indicators deteriorate significantly, deep multi-parameter coupling is abnormally significant, or the safety factor is predicted to fall below the critical value within the next 24 hours; in this case, a warning message is issued and manual verification and increased monitoring are recommended. Alert Level warnings correspond to situations where the instability probability reaches the high risk threshold, or accelerated deformation, large-scale micro-seismic events are detected, or the safety factor is below the critical value and continues to decline; in this case, an emergency alarm is issued and immediate emergency measures are recommended. This tiered warning mechanism enables differentiated responses to different risk levels, avoiding resource waste caused by overreaction and ensuring timely handling of high-risk conditions.

[0059] In the preferred solution, early warning information and corresponding handling suggestions are output, including: Based on the determined warning level and risk location, a basic response plan is obtained by matching the preset rule base. At the same time, the most similar successful response case is retrieved from the historical case database to modify the basic response plan and generate an emergency response instruction that includes response measures, responsible parties, resource allocation paths and time requirements. The platform uses a 3D visualization method to display slope topography and geological models, multi-source sensor locations, real-time monitoring data, 3D state field cloud maps, risk level distribution, and early warning information. Warning information and emergency response instructions are pushed to relevant management personnel through at least one of the following methods: platform pop-up, SMS, or email, and execution feedback is received to complete closed-loop management.

[0060] To ensure the effective implementation of early warning information, this invention constructs an intelligent linkage decision-making model for emergency response, directly linking early warning levels with specific response measures. Once an early warning level is determined, the system first matches the warning level and risk location against a pre-defined rule base to obtain a basic response plan. This rule base is pre-built based on engineering experience and safety standards, covering general response measures for different early warning levels. To further enhance the relevance and effectiveness of the response plan, this invention retrieves the most similar successful response cases from a historical case database. A case similarity matching algorithm calculates the similarity between the current early warning information and historical cases in terms of geological conditions, damage modes, and early warning levels. The response plan from the case with the highest similarity is selected to refine the basic response plan. Finally, an emergency response instruction is generated, including specific response measures, responsible parties, resource allocation paths, and time requirements, enabling on-site personnel to quickly understand the appropriate countermeasures.

[0061] In terms of information output, this invention establishes a three-dimensional visualization platform that integrates and displays slope topographic and geological models, multi-source sensor locations, real-time monitoring data, three-dimensional state field cloud maps, risk level distribution, and early warning information. The platform utilizes Geographic Information System (GIS) technology to achieve spatial presentation of monitoring data, allowing managers to quickly grasp the overall safety status of the slope through an intuitive graphical interface. Simultaneously, the system pushes early warning information and emergency response instructions to relevant managers in real time through various channels such as platform pop-ups, SMS, and emails. It also establishes an execution feedback mechanism to receive feedback data on on-site handling, forming a closed-loop management process that ensures timely and effective responses to early warning information.

[0062] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring and early warning of the condition of slopes in high-fill and deep-cut roadbeds, characterized by: The method includes: S1. Based on the geological conditions and potential risks of high-fill and deep-cut roadbed slopes, deploy suitable sensing terminals to collect multi-dimensional monitoring data of the slopes. S2. Optimize the collected multi-source monitoring data to improve data quality and consistency; S3. Perform multi-dimensional fusion analysis on the optimized data to form comprehensive data that reflects the overall condition of the slope; S4. Assess slope safety risks based on the comprehensive data and output corresponding early warning information and handling suggestions.

2. The method for real-time monitoring and early warning of the condition of roadbed slopes with high fill and deep cut as described in claim 1. Its characteristic is: In step S1: Deploy ground-based synthetic aperture radar to acquire large-scale, high-precision, planar surface deformation fields, and deploy high-precision GNSS receivers as key point displacement benchmarks and verifications; Array-type microelectromechanical system inclinometers are installed at key profiles and potential slip zone locations to monitor tilt changes, and fiber optic grating sensors are embedded to form a distributed network for real-time monitoring of soil strain, temperature, and vibration. Multi-parameter inclinometers are installed in deep boreholes to monitor deep displacement, pore water pressure sensors are deployed to monitor groundwater changes, and earth pressure cells are buried to monitor the internal stress state of the soil. Set up meteorological stations to collect data on rainfall, temperature, humidity, wind speed and direction, and set up video monitoring to assist in visualization and anomaly confirmation; In deep excavation areas prone to rock mass fracturing, microseismic or acoustic emission sensor arrays are deployed to capture signals of the initiation and propagation of fractures within the rock mass.

3. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 1, characterized in that: in step S2: Based on the progress of slope construction and the law of deformation evolution, the data acquisition frequency of the deployed sensing terminals is dynamically adjusted. The acquisition frequency is increased in areas with high deformation rate or large construction disturbance, and the acquisition frequency is decreased in areas where deformation tends to be stable. The reliability coefficient of each sensing terminal is evaluated in real time. The reliability coefficient is determined by comprehensively analyzing data integrity, data consistency, data trend rationality and hardware working status. When the reliability coefficient is lower than the failure threshold, the sensor data is automatically removed or its weight in subsequent fusion calculations is reduced. Based on the comprehensive evaluation results of deformation evolution law and sensor reliability coefficient, similar sensors are added in areas with insufficient monitoring capabilities, and sensors are removed in areas that have been in long-term failure or are redundant.

4. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 3, characterized in that: In step S2, the collected multi-source monitoring data is preprocessed, including: The data collected by each sensing terminal is timestamped to ensure that the data from different sensors at the same time are comparable; By unifying monitoring data from different coordinate systems to the same spatial reference datum, precise matching of spatial locations can be achieved. An adaptive filtering algorithm based on operating conditions is used to filter out environmental noise and abnormal jump values ​​in the original signal, while retaining the effective signal components. Missing values ​​that occur during data collection are imputed to form a continuous and complete monitoring time series. Normalize or standardize monitoring data with different dimensions to eliminate the impact of dimensional differences on subsequent analysis.

5. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 1, characterized in that: In step S3, feature extraction and index construction are performed on the preprocessed data, including: Based on surface deformation monitoring data, the comprehensive deformation rate and deformation acceleration are calculated to characterize the dynamic features of slope surface deformation. Based on distributed optical fiber strain monitoring data, a strain concentration index is calculated to identify local strain anomaly areas. Based on pore water pressure monitoring data, the rate of change of pore water pressure is calculated to reflect the dynamic changes in groundwater activity. Based on meteorological monitoring data, rainfall intensity and cumulative rainfall are extracted, and the rainfall impact coefficient is calculated by combining historical data; Based on microseismic monitoring data, the frequency of microseismic events, energy release rate, and microseismic activity index are calculated to assess the degree of rock mass fracturing activity.

6. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 5, characterized in that: Multi-source monitoring data are collaboratively fused to generate a continuous state field distribution, including: Multi-scale collaborative fusion of point sensor monitoring data, area radar monitoring data, and distributed optical fiber monitoring data is carried out. The three-dimensional strain field constructed by the distributed optical fiber monitoring data is used as the basic framework. The overall trend of the basic framework is corrected by the area deformation field obtained by the ground-based synthetic aperture radar. The accuracy of key points in the corresponding positions in the basic framework is calibrated by the point monitoring data obtained by the global navigation satellite system receiver and array inclinometer. High-precision, blind-zone-free three-dimensional displacement field and three-dimensional strain field are generated. Spatiotemporal coupling analysis was performed on monitoring data of deep displacement, deep stress and pore water pressure to construct a deep multi-parameter coupling matrix. The correlation coefficients of each parameter at the same spatial location and time node were calculated. When the correlation coefficients exceeded the set threshold and showed a continuous upward trend, they were identified as deep multi-parameter coupling anomaly areas. By correlating meteorological monitoring data, pore water pressure monitoring data, and soil deformation monitoring data, a temporal mapping relationship between rainfall infiltration and soil response is established, generating a dynamic state field that takes into account hydrological influences.

7. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 6, characterized in that: Based on the results of the fusion analysis, quantitative indicators of slope stability are calculated, including: The extracted key feature indicators are input into a pre-built risk assessment model, and a dynamic weight allocation mechanism is used to calculate the comprehensive risk index. The weight of each indicator is adjusted in real time based on the sensor reliability coefficient, current environmental conditions, and historical performance. Based on the generated three-dimensional displacement field and three-dimensional strain field, combined with the slope geological model, the current safety factor of the slope is estimated in real time using the limit equilibrium method or the finite element simplified model. The comprehensive risk index and safety factor are combined to calculate a quantitative assessment value of the slope instability probability, which is used for subsequent risk level determination.

8. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 7, characterized in that: Based on the results of the fusion analysis, the slope safety risk is dynamically assessed and the risk level is determined, including: The calculated comprehensive risk index, safety factor, and instability probability are used as inputs. Combined with the slope geological model and historical monitoring data, a dynamic Bayesian network is used to update the slope instability probability in real time, and the failure mode of the current slope is identified based on the changing trend of the instability probability. Based on real-time operating condition characteristic parameters, the early warning thresholds of each evaluation index are dynamically calibrated. The operating condition characteristic parameters include rainfall intensity, construction disturbance degree, and microseismic activity index. The risk level is divided into three levels: concern level, warning level, and alarm level by comprehensively judging the instability probability output by the dynamic Bayesian network, the damage mode identification result, and the dynamically calibrated threshold.

9. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 8, characterized in that: The risk levels are divided into three levels: Attention Level, Warning Level, and Alert Level, as follows: When the probability of instability is in the first preset range, or when a single key indicator exceeds the dynamic calibration threshold but does not show a significant deterioration trend, it is judged as a warning of concern level, and a warning of concern is issued and it is recommended to strengthen daily inspections. When the probability of instability is in the second preset range, or the key indicators deteriorate significantly, the deep multi-parameter coupling is abnormally significant, or the safety factor is predicted to be lower than the critical value within a set time in the future, it is judged as a warning level warning, and a warning message is issued and a reminder is given to strengthen manual verification and encrypted monitoring. When the probability of instability is in the third preset range, or when accelerated deformation, large-scale micro-seismic events, or the safety factor is below the critical value and continues to decline, it is determined to be an alarm-level warning, an emergency alarm is issued, and emergency measures are recommended to be taken immediately.

10. The method for real-time monitoring and early warning of the condition of high-fill and deep-cut roadbed slopes according to claim 9, characterized in that: Output early warning information and corresponding handling suggestions, including: Based on the determined warning level and risk location, a basic response plan is obtained by matching the preset rule base. At the same time, the most similar successful response cases are retrieved from the historical case database to modify the basic response plan and generate an emergency response instruction that includes response measures, responsible parties, resource allocation paths and time requirements. The platform uses a 3D visualization method to display slope topography and geological models, multi-source sensor locations, real-time monitoring data, 3D state field cloud maps, risk level distribution, and early warning information. Warning information and emergency response instructions are pushed to relevant management personnel through at least one of the following methods: platform pop-up, SMS, or email, and execution feedback is received to complete closed-loop management.

Images