Road Settlement Monitoring Method and System Based on BeiDou Satellite Positioning System
By constructing a multi-level monitoring network through the BeiDou satellite positioning system, and combining multi-source data fusion with mechanism-data dual-driven prediction, the problems of scattered equipment, isolated data, and single prediction in existing technologies have been solved, realizing high-precision and intelligent road settlement monitoring and early warning across the entire area.
Patent Information
- Application Number
- CN202610969404.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-01
- Publication Date
- 2026-11-17
AI Technical Summary
Existing road settlement monitoring technologies suffer from problems such as fragmented equipment, isolated data, limited forecasts, resource waste, and delayed early warnings, making it difficult to achieve high-precision and intelligent monitoring and early warning across the entire area.
A multi-level monitoring network is constructed using the BeiDou satellite positioning system. Combined with InSAR, MEMS inertial sensors, and municipal IoT data, spatiotemporal registration and error correction are performed. A mechanism-data dual-driven prediction model is established, and an event-triggered control unit is integrated to achieve adaptive resource scheduling and hierarchical early warning.
It has achieved high-precision, seamless monitoring across the entire area, improved the reliability of monitoring results and the robustness of predictions, promoted the transformation of operation and maintenance from post-event handling to pre-event early warning, reduced energy consumption and improved the pertinence and timeliness of early warning.
Smart Images

Figure CN123111186A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent monitoring technology, specifically to a road settlement monitoring method and system based on the BeiDou satellite positioning system; Background Technology
[0002] With the acceleration of new urbanization construction in China and the large-scale implementation of underground rail transit and integrated utility tunnel projects, coupled with multiple external disturbances such as the year-by-year increase in road traffic load and the continuous extraction of regional groundwater, road subsidence has become a key hidden danger restricting the structural safety of urban road networks and threatening the safety of road traffic and the integrity of underground pipelines. Road subsidence is characterized by its insidious development, gradual accumulation over the years, and sudden instability. The continuous expansion of subsidence can easily induce road surface cracking and collapse, underground pipeline rupture and leakage, and in turn induce traffic safety accidents such as vehicle rollovers and road collapses that cause injuries. The industry's demand for refined monitoring and proactive risk management of road subsidence throughout its entire life cycle continues to rise.
[0003] Currently, mainstream settlement monitoring relies on a variety of monitoring equipment, including InSAR, GNSS reference stations, MEMS sensing units, and hydrostatic levels. However, these monitoring devices belong to different technical systems and are mostly deployed and used separately, lacking an integrated network and collaborative logic. This makes it difficult to balance large-scale surveys with refined management of key locations. Conventional data processing uses a crude approach of calculating each type of monitoring data separately and simply comparing the results, lacking unified data benchmarks and error correction loops across technical systems. Settlement assessment and trend prediction either rely on pure data fitting algorithms or geotechnical finite element mechanism simulations. These two technologies are disconnected, lacking an evaluation framework driven by both mechanism and data.
[0004] Existing monitoring platforms generally use fixed sampling frequencies to collect data continuously around the clock. They cannot dynamically change the collection strategy based on disturbances such as weather changes, construction of surrounding foundation pits, and extreme rainfall. They also have high energy consumption and storage redundancy in terms of hardware and software resources. At the same time, the early warning rules use a one-size-fits-all logic based on single-point settlement thresholds, without considering the regional geological environment and the spatiotemporal evolution of settlement to achieve graded early warnings. The monitoring data is disconnected from the road maintenance and disposal process, making it difficult for the data to be directly applied to guide operation and maintenance decisions.
[0005] In practical use, it has some drawbacks, such as:
[0006] 1. Existing settlement monitoring technologies are scattered and have fragmented network logic, making it impossible to build a comprehensive, three-dimensional monitoring architecture that combines full-area surveys, detailed investigations of key points, and filling gaps in monitoring blind spots. InSAR large-scale remote sensing survey technology is prone to coherence failure due to dense vegetation, continuous high-rise buildings, and elevated bridges in urban areas, resulting in insufficient precision in detailed point monitoring. High-risk nodes such as bridgeheads and pipeline intersections cannot achieve high-precision continuous monitoring. GNSS monitoring stations have excellent single-point measurement accuracy, but their deployment cost is high and deployment points are sparse. In urban areas with high-rise buildings, satellite signal blockage occurs frequently, creating large areas of monitoring blanks. Static levels and MEMS miniature sensors can only be deployed in small, localized areas, and data from each measuring point is stored independently, generally forming data silos. Various monitoring equipment lack a layered, complementary, and coordinated network solution, making it impossible to achieve seamless integration from macroscopic assessment of the deformation of the entire road network to precise tracking of local hidden danger points. Overall monitoring suffers from the shortcomings of having macroscopic trends over a large area but lacking local details, having monitoring data for key points but lacking control over surrounding blind spots, and having isolated and uncoordinated multi-source monitoring information.
[0007] 2. Multi-source heterogeneous monitoring data are processed independently across different channels, lacking a closed-loop optimization mechanism for spatiotemporal registration, deep fusion, and error correction. This results in inconsistent accuracy and insufficient reliability of the final settlement results. InSAR, GNSS, and roadbed-buried sensors differ significantly in monitoring scale, sampling period, and measurement accuracy. Existing processing methods merely list and compare data results and superimpose surface values, failing to unify spatiotemporal benchmarks for deep information fusion. InSAR monitoring results are inherently subject to atmospheric delays and satellite orbital deviations, introducing inherent systematic errors. High-precision GNSS measured data is only used for result verification and cannot be used in reverse-engineering InSAR calculation link optimization algorithms for atmospheric correction and phase unwrapping. GNSS observation data also suffers from random errors due to multipath interference in urban areas, lacking adaptive optimization and removal models. These various errors cannot be mutually canceled or corrected, limiting the reliability of settlement monitoring results and hindering the support for refined operation and maintenance decisions for urban road safety.
[0008] 3. The technical approaches to settlement status assessment and trend prediction are limited, and geotechnical mechanistic mechanism models and big data statistical models are incompatible, resulting in poor environmental adaptability and ambiguous physical interpretation of prediction results. Currently, settlement prediction is divided into two independent systems: data-driven and mechanism modeling. Pure data algorithms such as time series and machine learning rely on historical monitoring data to fit change patterns. However, they have weak extrapolation and prediction capabilities when faced with unconventional disturbances such as sudden construction of foundation pits nearby or extreme heavy rainfall, and the predicted values lack theoretical support from geotechnical mechanics. Mechanism simulation models such as finite element methods have clear physical logic, but the modeling is highly dependent on stratum parameters and boundary environmental conditions, resulting in high parameter survey costs and the inability to dynamically correct model parameters in real time by accessing massive online monitoring data. Existing solutions lack a mechanism- and data-driven coupled prediction architecture. During the normal settlement phase, they cannot quickly extrapolate trends based on massive monitoring data. Under sudden external disturbances, they cannot quickly switch to mechanism simulation for emergency assessment. The probability of prediction failure is high in complex geological and variable external environments, and the predictive ability of early risk warning is insufficient.
[0009] 4. The monitoring equipment operation and management mode is rigid, lacking an intelligent triggering mechanism based on changes in on-site risks and the ability to dynamically allocate monitoring resources. The existing monitoring system uses a fixed sampling period for continuous data collection. GNSS stations and buried sensors along the entire line operate at full capacity year-round, resulting in enormous consumption of equipment power, data transmission, and storage space. However, road settlement is mostly in a slow creep state, accelerating only under special conditions such as rainstorm infiltration, surrounding earthwork excavation, and short-term traffic surges. The fixed collection mode leads to unnecessary waste of resources during stable periods and missed collection of critical data during periods of accelerated risk development. The entire monitoring system lacks an intelligent scheduling hub for multi-source information linkage. It cannot automatically identify high-risk areas by combining InSAR large-scale hazard survey results, weather warnings, and construction project reporting information, nor can it adjust the collection frequency of corresponding points and optimize sensor network operating parameters as needed. The mismatch between resource allocation and the state of disease development is prominent.
[0010] 5. The monitoring platform suffers from insufficient data mining depth, resulting in a disconnect between settlement monitoring results and road operation and maintenance. It lacks a comprehensive, visualized management and control system integrating data aggregation, intelligent analysis, tiered early warning, and operation and maintenance dispatch. The existing monitoring system only stores raw data and exports basic reports, failing to integrate multi-source monitoring data, geological archives, road design parameters, and information from surrounding engineering activities for correlation analysis. This prevents the generation of a unified management and control map that intuitively displays the safety level and settlement evolution trend of the entire road network. Early warning rules are limited to triggering alarms when settlement values exceed limits at single points, without considering the spatial contiguous characteristics of settlement, temporal development patterns, and geological disaster mechanisms for tiered and differentiated early warning. Monitoring data cannot be efficiently transformed into a basis for road maintenance and remediation, hindering the transformation of road network operation and maintenance from post-incident emergency repairs to pre-incident risk warnings and precise, targeted interventions.
[0011] In summary, existing road settlement monitoring technologies suffer from systemic shortcomings in five key dimensions: collaborative networking of multi-source devices from air, ground, and underground; closed-loop correction through heterogeneous data fusion; mechanistic data coupling for prediction; intelligent scheduling of monitoring resources; and integrated implementation of monitoring, early warning, and operation and maintenance. The overall technical architecture remains at a developmental stage characterized by simple equipment stacking, fragmented data processing, one-sided prediction methods, and low levels of system intelligence. Therefore, developing an intelligent road settlement monitoring solution that constructs a three-dimensional collaborative sensing network integrating air, ground, and underground systems, establishes a complete closed loop for multi-source data acquisition, fusion, and error optimization, relies on both mechanistic and data-driven approaches to achieve intelligent settlement assessment and prediction, and incorporates event-triggered adaptive resource scheduling logic is an urgent technological need to address the shortcomings in urban infrastructure safety monitoring and promote the digital, intelligent, and precise transformation and upgrading of road traffic operation and maintenance. Summary of the Invention
[0012] In order to overcome the above-mentioned defects of the prior art, the present invention provides a road settlement monitoring method and system based on the Beidou satellite positioning system, which solves the problems mentioned in the background art through the following scheme.
[0013] To achieve the above objectives, the present invention provides the following technical solution:
[0014] Road settlement monitoring methods based on the BeiDou satellite positioning system include:
[0015] S1. Acquire multi-level monitoring data for the target road area. The multi-level monitoring data includes at least: a general survey layer deformation field obtained by InSAR technology; detailed survey layer positioning data obtained by BeiDou monitoring stations deployed in the abnormal target areas identified in the general survey layer deformation field and at key road nodes; and blind spot layer monitoring data obtained by auxiliary sensors deployed in signal-blocked areas and connected municipal IoT devices. The blind spot layer monitoring data includes at least one of MEMS inertial sensor data, hydrostatic level data, municipal pipeline pressure / flow monitoring data, and traffic monitoring data.
[0016] S2. Spatiotemporal registration and error correction are performed on the multi-level monitoring data, and an adaptive weighted fusion algorithm is used for fusion processing to generate comprehensive settlement monitoring results for the target road area; wherein, the detailed investigation layer and blind spot filling layer data are used to perform error calibration and feedback on the InSAR deformation data to optimize its processing model.
[0017] S3. Based on the comprehensive settlement monitoring results, and combined with road geological survey data, structural design parameters and external event factors, use the mechanism-data dual-drive model to evaluate the safety status of each structural unit of the road and predict its settlement trend.
[0018] S4. Based on the assessment and prediction results, generate graded early warning information and present it visually.
[0019] This invention provides a road settlement monitoring method and system based on the BeiDou satellite positioning system, which has the following advantages compared with the prior art:
[0020] 1. This invention establishes a multi-level collaborative monitoring network consisting of a general survey layer, a detailed survey layer, and a blind spot filling layer, and integrates an event trigger control unit. This enables seamless connection and adaptive dynamic scheduling of monitoring resources from overall perception and key detailed surveys to blind spot coverage. It solves the technical defects of existing technologies, such as the scattered deployment of various monitoring devices, their independent operation, and the lack of three-dimensional networking logic. When InSAR identifies abnormal target areas or senses disturbances such as vibration, rainfall, or surrounding construction, it can automatically wake up and switch the corresponding BeiDou station to high-frequency mode. This achieves intelligent resource allocation for macro-early warning, precise location, and on-demand monitoring, effectively overcoming monitoring blind spots, improving the monitoring density and timeliness in high-risk areas, and significantly reducing the overall energy consumption of the system during periods of no risk.
[0021] 2. This invention constructs a closed-loop optimization mechanism for deep fusion of multi-source heterogeneous data and mutual calibration and correction of errors by using spatiotemporal registration of multi-source heterogeneous data, specific error correction, and adaptive weighted fusion algorithms, and by introducing a closed-loop optimization mechanism for inverse optimization of the InSAR model using high-precision data. This replaces the extensive fusion mode in the prior art where each data source is solved independently and the results are simply compared. Through BeiDou multipath error correction and InSAR atmospheric error correction, and by using BeiDou high-precision measured data to calibrate and provide feedback on the system error of InSAR, dynamic parameter optimization of the InSAR atmospheric correction model and spatial interpolation algorithm is achieved. This solves the problems of inconsistent benchmarks of heterogeneous data, mutual contamination of errors, and low reliability of fusion results, and significantly improves the accuracy and reliability of comprehensive settlement monitoring results.
[0022] 3. This invention constructs a mechanism-data dual-driven prediction model and pre-sets an adaptive weighted composite strategy of normal and event-driven approaches. This achieves complementary advantages and synergistic prediction between physical laws and historical statistical laws, solving the technical bottlenecks of existing technologies where pure data models have poor extrapolation capabilities under sudden strong disturbances and pure mechanism models have complex parameters that are difficult to correct in real time. Under normal operating conditions, the model combines the advantages of both approaches for weighted fusion. When receiving high-disturbance risk event signals such as deep foundation pit excavation or pipeline rupture, or when the prediction deviation between the two sub-models is too large, the model will automatically increase the weight of the mechanism prediction sub-model and prioritize the analysis results based on physical laws. This significantly improves the model's robustness and physical interpretability in dealing with sudden and unconventional operating conditions while ensuring the accuracy of normal predictions.
[0023] 4. This invention establishes a dynamic early warning system based on multi-indicator quantitative assessment and trend prediction by integrating a safety status assessment and hierarchical early warning generation module. It abandons the rigid mode of traditional single threshold-triggered alarms and can comprehensively judge the current safety status (normal, caution, abnormal, dangerous) and future subsidence trends to generate four levels of early warning: blue, yellow, orange, and red. This solves the problems of existing early warning systems being out of touch with geological environment and spatiotemporal evolution laws, and the problems of single and delayed early warning information. It realizes the refined and dynamic differentiation of risk levels, provides more targeted and predictive decision support for maintenance management, and effectively promotes the transformation of road network operation and maintenance from "post-event handling" to "pre-event early warning and precise intervention". Attached Figure Description
[0024] Figure 1 This is a flowchart of the road settlement monitoring method based on the BeiDou satellite positioning system of the present invention;
[0025] Figure 2 This is a flowchart of the multi-level monitoring data of the present invention;
[0026] Figure 3 This is a modular framework diagram of the road settlement monitoring system based on the BeiDou satellite positioning system of the present invention. Detailed Implementation
[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0028] Please see Figures 1 to 3 The road settlement monitoring method based on the BeiDou satellite positioning system shown includes the following steps:
[0029] S1. Obtain multi-level monitoring data for the target road area.
[0030] This step aims to establish a hierarchical, collaborative monitoring network, comprehensively utilizing multiple data sources such as space-based, ground-based, and IoT technologies to construct a three-dimensional sensing system that ranges from macro to micro and from surface to point. Specific data acquisition methods are as follows:
[0031] S11. Census Layer Data Acquisition
[0032] This level relies on satellite remote sensing time-series data to conduct a comprehensive survey of macroscopic deformation, enabling preliminary screening and location of potential hazards across the entire road area;
[0033] By periodically acquiring synthetic aperture radar satellite imagery data (such as Sentinel-1, with a revisit period of 12 days), and using time-series InSAR technology, a large-scale surface deformation field with meter-level spatial resolution covering the entire target road area is calculated. ;
[0034] Surface subsidence rate calculated pixel-by-pixel from global deformation field data The system presets a first screening threshold. When the regional settlement rate satisfies At that time, the area is automatically marked as a deformation anomaly target area. The preliminary identification and rough location of potential road subsidence hazards were completed.
[0035] S12, Detailed Layer Data Acquisition
[0036] This step involves deploying high-precision monitoring equipment to conduct targeted and refined monitoring of abnormal areas and key road structural points identified in the census screening.
[0037] Abnormal target areas identified in step S11 Within the scope, key risk nodes such as bridge piers, high embankment sections, and soft soil treatment sections are identified based on road design data, and lightweight Beidou integrated monitoring terminals are deployed. The terminal has a built-in choke coil antenna to suppress multipath effects during the monitoring process, and relies on low-power wide area networks such as NB-IoT and LoRa to achieve stable data interaction with the cloud platform.
[0038] The activation and operating status of the monitoring terminal are intelligently controlled by the system event-triggered control unit, and the operating mode is adaptively switched based on multi-source risk signals. Specific triggering conditions include:
[0039] 1. InSAR deformation field shows that the target area settling rate exceeds a higher second threshold. (like, );
[0040] 2. Sensors deployed in the surrounding blind-spot layer monitor the peak vibration acceleration. Exceeding the preset threshold Its identification signal corresponds to disturbance conditions such as heavy vehicle passage and localized small landslides;
[0041] 3. If the meteorological data shows that the cumulative rainfall in 24 hours exceeds 50mm, or if the municipal data identifies deep foundation pit excavation within a 500m radius, it indicates a strong risk of disturbance.
[0042] When any trigger condition is met, the control platform first checks the current working mode of the target monitoring terminal; if it is already in high-frequency precision acquisition mode, the current mode is maintained and there is no need to trigger it again; if it is in sleep or low-frequency mode, a wake-up / switching operation is performed.
[0043] Specifically, when any trigger condition is met, the platform remotely wakes up the dormant terminal, or adaptively switches the terminal from the conventional low-frequency sampling mode (1 time / hour) to the high-frequency precision acquisition mode (1 time / second) to obtain high-precision three-dimensional positioning time-series data. Further calculations yielded the time-series data of vertical settlement at the monitoring points. This enables high-precision targeted detailed investigation of abnormal areas;
[0044] S13. Data acquisition for the blind spot filling layer:
[0045] This step targets areas where satellite and BeiDou signals are blocked or lost. It involves deploying local monitoring equipment and integrating municipal IoT data to fill monitoring blind spots and improve multi-dimensional perception.
[0046] By deploying MEMS inertial sensor arrays and hydrostatic levels in areas with severe satellite signal obstruction, such as under overpasses, inside tunnels, and in dense forests, a ground-based blind spot monitoring network is constructed. Simultaneously, it connects to the municipal IoT system via a secure data exchange interface to synchronously acquire real-time pressure and flow data for underground water supply and drainage, and gas pipelines. And the traffic flow density and vehicle type classification time series data output by the traffic monitoring system. ;
[0047] Integrate all blind spot sensor data, pipeline operation data, and traffic load data to construct a complete blind spot monitoring dataset. This will compensate for the spatial blind spots and dimensional limitations of satellite remote sensing and BeiDou monitoring, and achieve full-area, multi-angle coverage of road monitoring.
[0048] S2. Spatiotemporal registration, fusion, and model optimization of multi-source data
[0049] This step involves spatiotemporal alignment, error correction, and deep data fusion of the three-layer monitoring data, and iteratively optimizes the InSAR solution model based on high-precision measured data.
[0050] S21. Spatiotemporal Registration and Error Correction
[0051] Time synchronization: When unifying the timestamps of all data to the UTC standard, linear interpolation is performed on non-equal interval sampled data based on the minimum common sampling interval of each data source to generate a time-aligned data sequence;
[0052] Spatial reference unification: Through the seven-parameter Bursa model, the InSAR deformation field in the SAR coordinate system, the BeiDou positioning data in the WGS-84 coordinate system, and the blind spot sensor data in the local engineering coordinate system are uniformly transformed to the engineering coordinate system of the target area.
[0053] Specific error correction:
[0054] For BeiDou data: Multipath error correction is performed using a multipath hemispherical map model. This model continuously collects observation data for at least one satellite repetition period (e.g., BeiDou's sidereal period is approximately 23 hours and 56 minutes) at the actual installation location of the monitoring station, extracts multipath error sequences related to satellite elevation and azimuth angles, and constructs a multipath error lookup table function indexed by satellite elevation and azimuth angles. The real-time correction formula is:
[0055] ,in These are the original observations of the carrier phase;
[0056] For InSAR data: Initial correction of atmospheric delay error is performed using accessed meteorological reanalysis data;
[0057] S22, Adaptive Weighted Fusion
[0058] An adaptive weighted fusion algorithm was used to process the registered InSAR survey settlement data. Detailed investigation of subsidence data using BeiDou Conduct point-to-point fusion calculations;
[0059] Using BeiDou stations as fusion points, the optimal fusion settlement value The calculation method is as follows:
[0060]
[0061]
[0062] :
[0063] : , ;in The InSAR coherence coefficient has a value of [0,1]. Estimated BeiDou multipath error (unit: mm); to prevent division by zero, when
[0064] for: , ;
[0065] For areas without BeiDou monitoring stations, the fusion points can be used as control points, and the Kriging spatial interpolation algorithm can be used to interpolate the fusion results across the entire region to generate high-confidence comprehensive settlement monitoring results. ;
[0066] S23, InSAR processing model optimization
[0067] The core of this method lies in using high-precision monitoring data from the detailed investigation layer and the blind spot filling layer to optimize the InSAR processing flow.
[0068] Error calibration: After spatiotemporal registration is completed, the BeiDou time series will be settled. Using hydrostatic leveling data as the baseline, calculate the systematic error of InSAR deformation at the same location during the same period. , ;
[0069] Feature analysis and feedback:
[0070] analyze The correlation between the system error and three influencing factors—atmospheric water vapor content, local topographic relief, and temporal coherence coefficient—was investigated, and a systematic error characteristic model was constructed using multiple linear regression. ,in This refers to atmospheric precipitable water. The topographic standard deviation is calculated based on digital elevation model (DEM) data within a 100m × 100m window centered on the monitoring point. The change in the coherence coefficient, the regression coefficient Obtained through least squares fitting;
[0071] The number of BeiDou monitoring stations participating in the regression calibration should not be less than 10, and their spatial distribution should cover areas with different terrain (flat, undulating) and coherence (high, medium, low) conditions to ensure the statistical significance of the regression coefficients. The fitted error feature model is then sent back to the InSAR data processing engine.
[0072] Dynamic optimization of model parameters:
[0073] Based on the error characteristics of the feedback, two key parameters in the InSAR processing flow are optimized using the following methods:
[0074] Atmospheric error correction model parameter optimization: In the atmospheric phase filtering stage of InSAR processing, the following adaptive parameter adjustment strategy is adopted: The general atmospheric correction empirical model... The globally fixed coefficient k in the original text is replaced with a dynamically adjusted form. ,in These are the initial coefficients of the model. The water vapor sensitivity coefficient obtained from the regression in step S23 is... The preset relaxation factor (range of values) );
[0075] At the same time, according to the error characteristic model Significance (t-test) (If determined to be significant), higher-order terms (such as elevation-related terms) in the atmospheric linear combination model can be adaptively turned on or off.
[0076] Parameter optimization of spatial interpolation algorithm in incoherent regions: For low-coherence regions such as vegetated areas and water bodies, an adaptive Kriging interpolation method is adopted; the coherence coefficient is used as the parameter. The reciprocal of the factor is used as the main control factor for interpolation weights, and the interpolation search radius R and smoothing factor are dynamically adjusted. : ,in The average coherence coefficient of the pixels within the search window. Based on the radius (e.g.) ), Based on the smoothing factor (e.g.) );when When the time comes, the interpolation method will be automatically switched to the trend surface-based generalized kriging method to enhance the physical constraints of the interpolation.
[0077] The obtained optimized model parameters (including dynamic correction coefficient k(t), adaptive interpolation radius R, and smoothing factor) This will be applied to the InSAR data processing workflow of the entire target road area, thereby improving the global InSAR deformation accuracy;
[0078] S24, Decoherence Region Compensation
[0079] For areas where InSAR is completely decoherent and lacks ground-based sensor deployment, pre-acquired airborne lidar point cloud data or high-precision 3D models generated by oblique photography are called up; by comparing digital surface models at different times, the change in surface elevation is extracted as supplementary data for subsidence information and input into the spatial interpolation process to complete regional compensation.
[0080] S3. Mechanism-Data Dual-Driven Security Assessment and Trend Prediction
[0081] This step integrates geomechanical mechanisms with measured historical data to complete the road structure safety classification and settlement trend prediction in units.
[0082] S31. Input Data Preparation
[0083] Input data sources are divided into three categories:
[0084] 1. Comprehensive settlement monitoring results: ;
[0085] 2. Road body data: geological survey data such as soil layer distribution and mechanical parameters, and structural design parameters such as pavement thickness, subgrade fill height, and retaining wall dimensions;
[0086] 3. External event influencing factors: surrounding construction distance and excavation depth, groundwater level time series data, pipeline pressure and flow rate, and traffic load data, denoted as factor vector. ;
[0087] S32. Construction and Fusion of Dual-Drive Prediction Model
[0088] The model employs two sub-models: parallel deployment mechanism prediction and data-driven prediction.
[0089] (1) Mechanism prediction sub-model: A two-dimensional plane strain model is constructed using the finite element method to simulate the settlement response in the cross-sectional direction of the road; the modeling and solution steps are as follows:
[0090] Geometry and Mesh Generation: The model width is twice the width of the roadbed top surface (e.g., road surface width). At that time, the model width is taken The depth is taken to the bedrock surface or below the thickness of the compressed layer. Quadrilateral plane strain elements are used. ) Perform grid division, focusing on key areas (such as roadbed fill areas) with a grid size not exceeding 100 mm. The mesh size of the boundary influence zone can be relaxed to [a certain value]. ;
[0091] Material constitutive properties and parameters: The subgrade fill and foundation soil layers adopt the Mohr-Coulomb ideal elastoplastic constitutive model; the required mechanical parameters include: elastic modulus E (unit: ). Poisson's ratio Cohesion c (unit) ), internal friction angle (Unit: °C) Natural density (unit The parameters are derived from the geological survey data and indoor geotechnical test report described in step S3; the pavement structure layer adopts a linear elastic constitutive model;
[0092] Boundary and load conditions: The bottom boundary of the model constrains all degrees of freedom. ), the horizontal degrees of freedom are constrained by the two side boundaries ( The initial stress field is automatically balanced using a gravity field.
[0093] External event factor vector Traffic loads are applied to the pavement surface as equivalent uniformly distributed pressures (e.g., for urban arterial roads, this can be taken as...). Groundwater level changes are achieved by altering the pore pressure within specific soil layers; the amount of groundwater level change... (unit The resulting change in pore water pressure is ,in Let be the density of water, and take a value of . g is the acceleration due to gravity, and its value is... In the finite element model, by calculating the results... The nodal pore water pressure boundary condition is applied to the soil layer below the groundwater level to realize the soil mechanical effect loading caused by the rise and fall of the groundwater level; the loading process adopts multi-analysis step simulation to correspond to different time stages;
[0094] Solution and Output: The finite element solver is used to perform a nonlinear static solution. The vertical displacements of each node at the top of the model are the predicted settlement values based on the mechanism. Its output frequency is synchronized with the update frequency of external event factors;
[0095] (2) Data-driven prediction sub-model: Employing an attention-based mechanism The model serves as the core architecture, and its specific construction and training steps are as follows:
[0096] Input construction: based on the past N time steps ( A comprehensive settlement sequence (one monitoring point per day) With the external factor vector at the corresponding time point The concatenated data is used as input to the model.
[0097] Network structure: The model contains 3 encoder layers, each with a built-in multi-head self-attention mechanism (the number of attention heads is set to 4), and the feedforward network dimension is set to 128; the position encoding adopts learnable position embedding;
[0098] Hyperparameter settings: Optimizer selection The initial learning rate is set to The batch size was set to 32, the number of training epochs was set to 200, and an early stopping mechanism was used, stopping training when the validation set loss did not decrease for 10 consecutive epochs. To avoid overfitting, an extra layer was added after each encoder layer. Layer, drop rate set to ;
[0099] Training strategy: Organize historical data according to... The dataset is divided into training, validation, and test sets; the loss function used is the mean squared error (MSE). The model uses a sliding window to predict the future. time step (e.g.) The settlement value of (day) is denoted as . ;
[0100] Output: The model output is the predicted settlement value after inverse standardization, in millimeters (mm). );
[0101] It should be noted that for short-term strong disturbances (such as hourly rainstorms, construction shocks, and vibration exceeding limits), the data-driven model may not be able to respond in time due to the limitations of daily sampling resolution; in this case, the system will automatically increase the weight of the mechanism model through an event-driven fusion strategy (see S33) to ensure the reliability of the prediction.
[0102] S33, Adaptive Weighted Composite Prediction
[0103] The dual-drive model adaptively weights and fuses the outputs of two sub-models to obtain the final settlement prediction result. The following two fusion rules shall be followed:
[0104] (1) Normal fusion strategy (when there are no sudden high-risk external conditions): The weighted average fusion is adopted, and the formula is: ;
[0105] Among them, weight Based on the goodness of fit of the data-driven model over the past 30 days Dynamic adjustment ;
[0106] The calculation method is as follows The range of values is between;
[0107] After dynamic adjustment, Finally cut off to Within the range; that is, the better the data-driven model fits ( (The larger the weight), the greater the weight of the mechanistic model. The lower;
[0108] (2) Event-driven fusion strategy: When the system receives a preset high-risk external event signal (including but not limited to: deep foundation pit excavation in the surrounding area, pipeline burst alarm, 24-hour cumulative rainfall) (or the difference in predictions between the two sub-models exceeds a set threshold) At that time, switch to a fusion strategy that primarily uses the output of the mechanistic model;
[0109] Set threshold ,in and These are the mechanism-based prediction sub-model and the data-driven prediction sub-model for the future. The predicted settlement value at time; the triggering condition is: Once triggered, the fusion strategy will be adjusted as follows: The weight Forced to be set (Truncation interval) This significantly increases the weight of the output results of the mechanistic model;
[0110] Under strong disturbance conditions, the output results of the mechanism model are given priority. This is based on the following logic: Under sudden strong disturbance conditions that exceed the distribution range of historical data, the mechanism model based on the law of physical conservation has higher reliability than the purely data-driven statistical model.
[0111] S34. Safety Status Assessment
[0112] Based on quantitative indicators such as current settlement amount, settlement rate, and uneven settlement difference, and referring to industry standards such as the "Highway Subgrade Design Specification", the road is divided into independent assessment units such as ordinary road sections and bridge piers, and the safety status is divided into four levels: normal, caution, abnormal and dangerous.
[0113] S4, Tiered Early Warning and 3D Visualization
[0114] S41, Generation of Tiered Early Warnings
[0115] Based on the real-time safety level and settlement prediction results output in step S3, four warning levels—blue, yellow, orange, and red—are established, with the following clear judgment rules:
[0116] Blue alert: The structural safety status is "normal," and the slope of the predicted settlement rate is obtained by least-squares linear fitting of the settlement prediction sequence for the next 30 days based on the mechanism-data dual-driven model. This means that the predicted settlement rate over the next 30 days shows no statistically significant upward trend.
[0117] Yellow alert: The structural safety status is "Caution," or the settlement rate shows an upward trend; the "upward trend" is specifically defined as the rate slope calculated based on settlement data from the past 20 days. And the rate of increase was positive for five consecutive days;
[0118] Orange alert: The structural safety status is "abnormal," or it is predicted that the indicators will approach the danger threshold in the later stages; "approaching the danger threshold" is specifically defined as: the maximum predicted settlement value for the next 15 days based on the mechanism-data dual-driven model. The differential settlement or total settlement hazard thresholds specified in industry standards (such as the "Highway Subgrade Design Code") have been reached. of ,Right now ;
[0119] Red Alert: The structural safety status is "dangerous," or there is a risk of rapid settlement in the short term; "risk of rapid settlement in the short term" is specifically defined as: based on high frequency (sampling interval) Monitoring data revealed past Settlement within , or the past Settling rate within ;
[0120] Early warnings at all levels are distributed to maintenance management personnel via platform messages, SMS, and mobile push channels.
[0121] S42, 3D Visualization
[0122] When importing a 3D reality model or BIM model of a road using oblique photography into the system, the comprehensive settlement results, unit safety status, and early warning level are bound to corresponding spatial coordinates and dynamically rendered and displayed.
[0123] By mapping settlement amount and settlement rate using color marks, a global settlement cloud map is generated;
[0124] Four types of identification icons or highlighted borders are used to distinguish road structural units with different warning levels;
[0125] The location-click search function can retrieve the time-series change curves, raw monitoring data, and safety assessment reports for the corresponding location;
[0126] Supports animation playback of the settlement evolution process, and overlays settlement prediction trend curves;
[0127] In a specific implementation, taking a section of a main urban road passing under a railway tunnel as an example, the application process of this system is illustrated as follows:
[0128] 1. InSAR surveys revealed a linear settlement zone (target area) in the roadbed above the tunnel, with a rate of 25 mm / year, triggering an event mechanism;
[0129] 2. The platform remotely wakes up two lightweight Beidou monitoring stations deployed in the target area and instructs them to enter high-frequency acquisition mode; at the same time, the data from the MEMS sensors pre-embedded in the tunnel is uploaded synchronously.
[0130] 3. The data fusion engine merges InSAR and GNSS data to generate an accurate subsidence field, and uses GNSS data feedback to discover residual errors in the InSAR atmospheric correction of the area, automatically optimizing the model parameters;
[0131] 4. At this time, the municipal IoT interface sends a message: the pressure of a main water supply pipe near the settlement zone has dropped abnormally; external event factors have been updated.
[0132] 5. The dual-drive prediction model receives a high-risk event signal of "pipeline anomaly"; the mechanism sub-model simulates the scenario of soil softening caused by water pipe seepage and predicts that settlement will accelerate; the data sub-model predicts little change based on historical trends; the system adopts an event-driven strategy to significantly increase the weight of the mechanism model and finally outputs the prediction that "settlement may accelerate".
[0133] 6. Based on this, the risk assessment module adjusts the risk level to "orange". The platform immediately generates an orange warning and highlights the road section on the 3D map. After receiving the warning, the maintenance unit quickly investigates and confirms that there is a hidden leak in the water pipe. It repairs it in time and avoids possible road collapse accidents.
[0134] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A road settlement monitoring method based on the BeiDou satellite positioning system, characterized in that, Includes the following steps: S1. Obtain multi-level monitoring data for the target road area; The multi-level monitoring data includes at least: the general survey layer deformation field obtained by InSAR technology, the detailed survey layer positioning data obtained by Beidou monitoring stations deployed in the abnormal target areas identified in the general survey layer deformation field and at key road nodes, and the blind spot layer monitoring data obtained by auxiliary sensors deployed in signal-blocked areas and connected municipal IoT devices. S2. Perform spatiotemporal registration and error correction on the multi-level monitoring data, and use an adaptive weighted fusion algorithm for fusion processing to generate comprehensive settlement monitoring results for the target road area. S3. Based on the comprehensive settlement monitoring results, and combined with road geological survey data, structural design parameters and external event factors, use the mechanism-data dual-drive model to evaluate the safety status of each structural unit of the road and predict its settlement trend. S4. Based on the assessment and prediction results, generate graded early warning information and present it visually.
2. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 1, characterized in that: The specific methods for acquiring multi-level monitoring data include: S11. Based on periodically acquired InSAR image data, the road surface deformation field is calculated, and areas with settlement rates exceeding the first threshold are identified and marked as abnormal target areas. S12. Deploy BeiDou monitoring stations in abnormal target areas and at pre-set key road nodes, and control their activation and working modes according to the event triggering mechanism to obtain high-precision positioning sequences; S13. In known areas where BeiDou signals are blocked, deploy MEMS inertial sensors or hydrostatic levels and connect them to the municipal IoT and traffic monitoring data interfaces to obtain blind spot monitoring data. The blind spot monitoring data includes at least one of the following: MEMS inertial sensor data, hydrostatic level data, municipal pipeline pressure / flow monitoring data, and traffic monitoring data.
3. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 2, characterized in that: The method of controlling its startup and working mode based on the event triggering mechanism includes: When the InSAR deformation field shows that the target area settlement rate exceeds the second threshold, the blind layer sensor detects that the vibration acceleration exceeds the set value, and the accessed meteorological / engineering event information indicates that there is a strong disturbance risk, the control platform remotely wakes up or adjusts the Beidou monitoring station to enter the high-frequency acquisition mode. Using detailed survey layer location data and blind spot layer monitoring data as high-precision benchmarks, error calibration and feedback are performed on InSAR deformation data of the same period and region to optimize the InSAR data processing model.
4. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 1, characterized in that: The adaptive weighted fusion algorithm is used to perform point fusion calculation on the registered InSAR data and BeiDou data; Specifically, for InSAR data, its weight in the fusion is dynamically adjusted based on its coherence coefficient; for BeiDou data, its weight is dynamically adjusted based on the multipath error estimate of its satellite observation environment.
5. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 4, characterized in that: The InSAR data processing includes the optimization of the InSAR solution model: Using detailed survey layer positioning data and blind spot layer monitoring data as high-precision benchmarks, error calibration and analysis were performed on InSAR deformation data of the same period and region to obtain the systematic error characteristics of InSAR data; The system error characteristics are fed back to the InSAR data processing flow to dynamically optimize the atmospheric error correction model parameters and the spatial interpolation algorithm parameters for the decoherent region in the data processing flow.
6. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 4, characterized in that: The multipath error correction of BeiDou data specifically includes: removing multipath errors with repetitive characteristics from the original observations based on a model of a multipath hemispherical map.
7. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 1, characterized in that: The external event factors include one of the following: the construction status of surrounding projects, groundwater level change data, the operational status of key pipelines, and traffic load change information.
8. The road settlement monitoring method based on the BeiDou satellite positioning system according to claim 1, characterized in that: The mechanism-data dual-driven model includes: The mechanism prediction sub-model takes road geological survey data and structural design parameters as inputs and outputs settlement prediction values based on physical and mechanical responses. The data-driven prediction sub-model takes historical settlement sequences and external event factors as inputs and outputs settlement prediction values based on data statistics. And a fusion module, used to fuse the outputs of the mechanism prediction sub-model and the data-driven prediction sub-model to generate the final settlement trend prediction result; The fusion module is configured with a dynamic weight allocation strategy: under normal conditions, the outputs of the two sub-models are weighted and averaged; when a preset high-risk external event signal is received, or the difference between the prediction results of the two sub-models exceeds a set threshold, the fusion strategy is switched to one that prioritizes the output of the sub-model that predicts the mechanism.
9. A road settlement monitoring system based on the BeiDou satellite positioning system, characterized in that, The system for implementing the method of any one of claims 1-8 comprises: Layered monitoring network module: used to acquire the deformation field of the general survey layer obtained by InSAR technology, the positioning data of the detailed survey layer obtained by Beidou monitoring stations, and the monitoring data of the blind spot layer obtained by auxiliary sensors and municipal IoT devices; Data fusion processing engine: Connects to the hierarchical monitoring network module, used for spatiotemporal registration, error correction and adaptive weighted fusion of multi-source monitoring data to generate comprehensive settlement results, and uses detailed survey layer data feedback to optimize the InSAR processing model; Risk assessment and prediction module: Connects to the data fusion processing engine, used to assess the safety status and predict the settlement trend based on comprehensive settlement results, geological survey data and external event factors, through a mechanism-data dual-driven model; Early warning and visualization platform: Connects to the risk assessment and prediction module, used to generate graded early warning information based on the assessment and prediction results, and to display it in three-dimensional visualization.
