Methods and systems for full life cycle health monitoring and risk management of road subgrade
By deploying dense sensing nodes on the roadbed and utilizing moving load excitation and dispersion analysis, a global mechanical model is constructed, which solves the problems of insufficient coverage and dynamic response of traditional monitoring methods. This enables intelligent health assessment and risk management of the roadbed, improving identification accuracy and lifespan extension.
Patent Information
- Application Number
- CN202611241625.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional roadbed monitoring methods suffer from sparse monitoring points, limited data dimensions, lack of dynamic response feature mining and full life cycle assessment capabilities, making it difficult to achieve effective coverage and intelligent management of the entire roadbed area.
By employing distributed sensor networks and mobile load excitation technology, a global mechanical model of the roadbed is constructed by deploying dense intelligent sensor nodes on the roadbed and using dispersion analysis and sparse constraint inversion algorithms. This model is then combined with long short-term memory neural networks for health assessment and risk management.
It enables dynamic mechanical parameter inversion across the entire roadbed, improving the identification rate of potential weak areas and the accuracy of risk identification, extending the service life of roads and reducing maintenance costs.
Smart Images

Figure CN123117772A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of civil engineering, specifically relating to a method and system for health monitoring and risk management of road subgrade throughout its entire life cycle. Background Technology
[0002] With the continuous expansion of national infrastructure construction, road engineering, as a core component of the transportation network, has become a major issue concerning national welfare and people's livelihood, particularly regarding its operational safety and durability. Roadbeds, as the foundation system bearing the weight of the superstructure and traffic loads, face complex mechanical evolution processes throughout their entire life cycle, including but not limited to soil consolidation, humidity changes, freeze-thaw cycles, and cumulative damage caused by repeated loading. These factors collectively determine the load-bearing capacity and performance of the roadbed, thus affecting the overall service level and service life of the road. Therefore, implementing scientific and efficient health monitoring and risk management of roadbeds has significant engineering value and social importance.
[0003] Road subgrade life-cycle health monitoring technology mainly involves the real-time sensing and assessment of the internal mechanical parameters and their evolution of the subgrade structure. Traditional subgrade monitoring methods primarily rely on displacement sensors, pressure sensors, or settlement observation points embedded within the subgrade to periodically collect static or quasi-static data to assess the subgrade's operational status. However, this monitoring method based on discrete point deployment has significant limitations in terms of spatial coverage, data acquisition efficiency, and dynamic response capabilities, making it difficult to fully reflect the mechanical behavior characteristics of the entire subgrade.
[0004] The shortcomings of existing technologies are mainly reflected in the following aspects: First, traditional monitoring methods rely excessively on the installation of dedicated sensors. The limited number of sensors and high deployment costs result in sparse monitoring points, making it difficult to achieve effective coverage of the entire roadbed and easily overlooking key weak areas. Second, existing methods primarily focus on the collection and analysis of static parameters, lacking in-depth analysis of the dynamic response characteristics of the roadbed under vehicle moving loads, and are unable to obtain the dynamic elastic modulus, stiffness distribution, and damage evolution trends of the roadbed in real time. Third, the data obtained from existing point-based monitoring has a single dimension, lacking the ability to systematically characterize the overall topology and mechanical transmission path of the roadbed, leading to insufficient identification and early warning capabilities for potential risks. Finally, traditional methods struggle to achieve continuous dynamic assessment of the roadbed's health status throughout its entire life cycle, lacking the technological leap from "static point-based monitoring" to "dynamic full-domain mechanical inversion." These problems collectively restrict the intelligent development level of roadbed health monitoring technology, urgently requiring a new technical solution capable of utilizing moving loads to achieve dynamic mechanical parameter inversion across the entire roadbed area. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for monitoring the health and managing the risks of roadbeds throughout their entire life cycle, which can effectively solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for full life-cycle health monitoring and risk management of road subgrade includes the following specific steps: Step 1: Deploy a distributed sensor network: In a representative section of the road subgrade to be monitored, densely distributed intelligent sensor nodes are uniformly deployed according to a preset grid layout strategy. Each sensor node integrates an acceleration sensor, a strain sensor, a humidity sensor, and a temperature sensor to achieve comprehensive perception and data acquisition of the mechanical response parameters of the subgrade structure; Step 2: Excite a moving load source: A test vehicle equipped with a dynamic load sensor travels along the monitored road section. The contact pressure between the test vehicle's tires and the road surface is calculated in real time using a preset calibration equation. The test vehicle passes through the monitoring area at a constant speed, with the speed range set to 20 to 80 kilometers per hour. During the test vehicle's travel, the entire road section is triggered. Synchronous data acquisition; Step 3: Acquire multi-dimensional dynamic response data: Each sensor node synchronously acquires acceleration time history signals, strain time history signals, humidity time history signals, and temperature time history signals induced by the moving load. The sampling frequency is set to 500 to 2000 Hz, and the data acquisition duration covers the time window during which the test vehicle completely passes through the monitoring area; Step 4: Construct a global mechanical model of the roadbed: Based on the acquired multi-dimensional dynamic response data, the propagation characteristic parameters of Rayleigh waves and Love waves are extracted using a dispersion analysis algorithm. Based on the propagation characteristic parameters, a sparse constraint inversion algorithm is used to invert the shear wave velocity distribution of the roadbed soil, and a three-dimensional shear wave velocity imaging model of the roadbed is constructed; Step 5: Calculate the dynamic elastic modulus distribution: Based on the three-dimensional shear wave velocity imaging model, according to the theoretical relationship between shear wave velocity and elastic modulus... Calculate the shear modulus of the subgrade soil ,in Given the soil density, the dynamic elastic modulus of the subgrade is further calculated based on the shear modulus. , Using Poisson's ratio, a dynamic elastic modulus distribution map of the entire subgrade is obtained; Step 6: Assess the subgrade bearing capacity and damage status: Based on the dynamic elastic modulus distribution map, a subgrade bearing capacity assessment model is established using finite element forward modeling. Combined with historical health baseline data, the bearing capacity degradation index of each zone is calculated. Based on the degradation index, potential weak areas of the subgrade are identified and risk level classifications are generated; Step 7: Generate risk management strategies: For weak areas with different risk levels, corresponding risk management strategies are generated based on a preset maintenance decision matrix, including measures such as increasing monitoring frequency, limiting load passage, grouting reinforcement or excavation and reconstruction, and the management instructions are issued to the subgrade operation and maintenance management platform.
[0007] In step 1, the spacing between sensor nodes is set to 1 to 3 meters along the road and 5 to 10 meters along the road. The number of sensor nodes in a single monitoring section is no less than 200. The sensor nodes adopt a low-power design and have an average fault-free operating time of greater than or equal to 50,000 hours.
[0008] In step 2, the single axle load of the test vehicle is set to 8 to 12 tons. The number of test vehicles is determined according to the length of the monitoring section, with at least one test vehicle configured for every 100 meters. The tire pressure of the test vehicles is maintained at 0.7 to 0.9 MPa, and the contact area between the tire and the road surface is determined by calibration using a pressure-sensitive film.
[0009] In step 3, each sensor node is configured with an independent synchronization clock module with a clock synchronization accuracy of less than or equal to 1 microsecond. The data acquisition trigger signal is uniformly allocated by the global positioning system timing signal to ensure that the time synchronization error of the sensor nodes across the entire road segment is less than or equal to 10 microseconds.
[0010] In step 4, the dispersion analysis algorithm adopts a high-resolution dispersion curve extraction method based on time-frequency analysis, with the frequency analysis range set to 10 to 500 Hz. The sparse constraint inversion algorithm adopts L1 norm regularization constraint, and the inversion grid resolution is set to 0.5 meters by 0.5 meters.
[0011] In step 5, the soil density is determined by on-site nuclear density meter calibration. The Poisson's ratio is obtained by referring to a table based on the subgrade soil type. The Poisson's ratio for sandy soil is 0.25 to 0.30, and the Poisson's ratio for cohesive soil is 0.35 to 0.45. The spatial interpolation of the dynamic elastic modulus calculation results is performed using the Kriging interpolation algorithm.
[0012] In step 6, the finite element forward modeling uses quadrilateral element meshes with a mesh size no larger than 0.5 meters. The historical health baseline data is derived from the initial inspection data during road completion and acceptance. The formula for calculating the bearing capacity degradation index is: ,in The initial elastic modulus, This is the current elastic modulus.
[0013] Areas with a degradation index greater than 0.3 are classified as high-risk areas, areas with a degradation index between 0.15 and 0.3 are classified as medium-risk areas, and areas with a degradation index less than 0.15 are classified as low-risk areas. High-risk areas will automatically trigger an early warning signal and initiate an emergency response process.
[0014] This invention also includes establishing a roadbed full life cycle health monitoring database. The database records the spatial coordinates of sensor nodes, the raw data of dynamic response for each test, the elastic modulus distribution data obtained by inversion, the bearing capacity assessment results, and maintenance records. The database capacity is not less than 10 gigabytes, and the data retention period is not less than 10 years.
[0015] The present invention also includes using a long short-term memory neural network model based on a historical database to predict the evolution trend of roadbed bearing capacity, with a prediction time span of not less than 12 months. The prediction results are used to guide the formulation of preventive maintenance plans. The input features of the prediction model include historical elastic modulus sequences, ambient temperature sequences, humidity sequences, and traffic flow sequences.
[0016] Compared with the prior art, the present invention has the following beneficial effects: Through a densely deployed intelligent sensor network, continuous spatial coverage of the entire roadbed is achieved. The spacing between sensor nodes can be as close as 1 meter, and the number of sensor nodes in a single monitoring section is no less than 200. This effectively solves the problem of insufficient coverage of traditional point-based monitoring and increases the identification rate of potential weak areas of the roadbed from 60% to over 95% compared to traditional methods.
[0017] By utilizing the moving load excitation method and combining dispersion analysis and sparse constraint inversion algorithm, three-dimensional imaging of the shear wave velocity across the entire roadbed was achieved, thereby accurately calculating the dynamic elastic modulus distribution. This marks the first technological leap from static point monitoring to dynamic global mechanical inversion, with the dynamic elastic modulus inversion accuracy reaching over 90%.
[0018] 3. A finite element bearing capacity assessment model was established based on the dynamic elastic modulus distribution map. The bearing capacity degradation index of each zone was automatically calculated and the risk level was divided. The intelligent hierarchical assessment of the roadbed health status was realized. The accuracy rate of high-risk area identification was greater than or equal to 92%, providing a scientific basis for roadbed life cycle maintenance decisions.
[0019] Based on long short-term memory neural networks and long-term accumulated historical monitoring data, it is possible to predict the evolution trend of roadbed bearing capacity in advance, with a prediction time span of more than 12 months. This enables maintenance management to shift from passive response to proactive prevention, significantly extending the service life of roads and reducing the total life cycle maintenance cost. Attached Figure Description
[0020] Figure 1 This is a schematic diagram of the overall technical solution architecture proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework for constructing a global mechanical model of the roadbed based on dispersion analysis and sparse constraint inversion algorithm in this invention; Figure 3 This is a logical flowchart of the calculation of the dynamic elastic modulus of the roadbed and the assessment of its bearing capacity in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the distributed sensor network, the mobile payload excitation source, and the data processing platform in this invention; Figure 5This is a schematic diagram of the roadbed bearing capacity evolution trend prediction framework based on long short-term memory neural network in this invention. Detailed Implementation
[0021] Please refer to Figures 1 to 5 To further illustrate the technical means and effects of the present invention in order to achieve the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.
[0022] Example 1 This embodiment focuses on a roadbed health monitoring scenario on a provincial highway section from K120+000 to K120+500, and fully deploys the roadbed full life-cycle health monitoring and risk management system proposed in this invention. The monitored section is 500 meters long, 12 meters wide, and designed as a two-way four-lane road with a daily traffic flow exceeding 8,000 vehicles, classifying it as a heavy-duty, busy traffic section.
[0023] Distributed sensor network subsystem: The data acquisition module consists of a densely distributed array of intelligent sensor nodes. The sensor nodes are spaced 2 meters apart laterally along the road and 8 meters apart longitudinally, with a total of 280 sensor nodes deployed within a single monitoring section, forming a grid-like sensing system. (Refer to...) Figure 1 As shown, each sensing node integrates the following four types of sensors: the accelerometer uses a MEMS triaxial accelerometer chip with a range of ±50g, a resolution of 0.001g, and a sampling frequency that can be configured within the range of 500 to 2000 Hz; the strain sensor uses a foil resistance strain gauge with a sensitivity coefficient of 2.0, a substrate resistance of 120 ohms, and a measurement range of ±3000 micro-strain; the humidity sensor uses a capacitive soil moisture sensor with a measurement range of 0 to 100% volumetric water content and an accuracy of ±2%; and the temperature sensor uses a platinum resistance temperature sensor with a measurement range of -40 to +85 degrees Celsius and an accuracy of ±0.1 degrees Celsius.
[0024] Each sensor node achieves data aggregation via a low-speed wireless communication module, employing LoRa spread spectrum communication technology. Operating in the 470-510 MHz frequency band, it achieves a transmission distance of up to 2000 meters in open areas, with an air speed ranging from 0.3 to 50 kilobits per second. Each sensor node is equipped with an independent synchronization clock module, featuring a built-in TCXO temperature-compensated crystal oscillator, ensuring clock synchronization accuracy of less than or equal to 1 microsecond. The sensor nodes utilize a low-power design, equipped with solar charging panels and lithium thionyl chloride battery packs, achieving a mean time between failures (MTBF) of 55,000 hours, meeting the requirements for continuous monitoring around the clock.
[0025] The data concentrator is deployed in a roadside cabinet in the middle of the monitored section. It uses an industrial-grade embedded computer with an eight-core ARM processor, a 2.0 GHz clock speed, 8 gigabytes of memory, and 256 gigabytes of solid-state drive storage. The data concentrator establishes a star network connection with each sensor node via a high-speed wireless communication module, polling and collecting raw sensor data from each node. The data concentrator internally runs a time synchronization service, receiving the second pulse signal output from the GPS timing receiver and using the IEEE 1588 precise time protocol to control the time synchronization error of all sensor nodes along the entire road section to within 10 microseconds.
[0026] Moving load excitation subsystem: The dynamic load test vehicle is a three-axle heavy-duty truck, with a single axle load set at 10 tons, for a total weight of 30 tons. A dynamic load sensor is installed on the rear axle of the test vehicle. This sensor is a strain gauge wheel axle load sensor with a range of 0 to 20 tons and an accuracy of 0.5%FS. Tire pressure is maintained at 0.8 MPa. The contact area between the tire and the road surface is pre-calibrated using a pressure-sensitive diaphragm, and the contact pressure is calculated in real-time using a preset calibration equation. The test vehicle is equipped with a high-precision positioning module, employing dual-antenna real-time dynamic positioning technology with a positioning accuracy better than 2 cm, recording the vehicle's trajectory and speed information in real time.
[0027] Reference Figure 2 As shown, the test vehicle travels at a constant speed of 60 kilometers per hour along the monitored section. A pre-acquisition signal is triggered 200 meters before the test vehicle enters the monitoring area to ensure that the acquisition time window covers the entire process of the test vehicle passing through the monitoring area. The onboard data acquisition unit inside the test vehicle transmits dynamic payload data to the data processing platform in real time via a 4G / 5G mobile network, and simultaneously broadcasts the trigger synchronization acquisition signal to the distributed sensor network.
[0028] Data processing and computing platform: The data processing platform adopts a cloud computing architecture, deployed on a private cloud server cluster within the road management center. The platform is configured with 20 computing nodes, each equipped with a 64-core CPU, 256 gigabytes of memory, and four Tesla V100 GPU accelerator cards, achieving a total computing power of 500 TFLOPS. The platform deploys the following core software modules: The dispersion analysis module extracts high-resolution dispersion curves based on time-frequency analysis. It employs a generalized S-transform algorithm to decompose the acceleration time-history signal into time-frequency components, with the frequency analysis range set to 10 to 500 Hz. The module first preprocesses the acquired multi-channel signals, including DC component removal, zero-phase filtering, and resampling unification. Then, it constructs a dispersion energy spectrum matrix and extracts the main dispersion curves of Rayleigh and Love waves using a peak tracking algorithm.
[0029] The sparse-constrained inversion module uses propagation characteristic parameters extracted from dispersion analysis and employs a sparse-constrained inversion algorithm with L1 norm regularization to invert the shear wave velocity distribution of the subgrade soil. This module divides the monitoring area into 0.5m x 0.5m inversion grids, establishes a nonlinear mapping relationship between wave velocity and dispersion curves, and iteratively optimizes the solution to find the optimal shear wave velocity distribution within the sparse solution space.
[0030] The elastic modulus calculation module is based on a three-dimensional shear wave velocity imaging model, according to the theoretical relationship between shear wave velocity and elastic modulus. Calculate the shear modulus of the subgrade soil ,in The soil density was determined to be 1850 kg / m³ using a nuclear density meter calibrated on-site. Poisson's ratio was obtained from a table based on the subgrade soil type; the subgrade fill in this section is cohesive soil, and the Poisson's ratio was taken as 0.40. The dynamic elastic modulus was determined using the formula... The calculation results were obtained, and the spatial interpolation of the calculation results was performed using the Kriging interpolation algorithm to generate a dynamic elastic modulus distribution map of the entire roadbed.
[0031] The bearing capacity assessment module uses finite element forward modeling to establish a roadbed bearing capacity assessment model. The finite element mesh uses quadrilateral elements with a mesh size no larger than 0.5 meters, and the Mohr-Coulomb elastoplastic model is selected as the material constitutive model. The module loads historical health baseline data, which originates from the initial elastic modulus field measured during road completion and acceptance. The bearing capacity degradation index is calculated using the formula... Calculate, where $E_{initial,i}$ is the initial elastic modulus. This is the current elastic modulus.
[0032] The risk level classification module automatically identifies potentially vulnerable areas of the roadbed based on the degradation index and generates a risk level classification. Areas with a degradation index greater than 0.3 are classified as high-risk areas, areas with a degradation index between 0.15 and 0.3 are classified as medium-risk areas, and areas with a degradation index less than 0.15 are classified as low-risk areas. High-risk areas automatically trigger early warning signals and initiate emergency response procedures.
[0033] The predictive analysis module uses a long short-term memory neural network model to predict the evolution trend of roadbed bearing capacity. This module employs a two-layer LSTM network structure containing 128 hidden units. Input features include historical elastic modulus sequences, ambient temperature sequences, humidity sequences, and traffic flow sequences. The prediction time span is set to 12 months. (Refer to...) Figure 5 As shown, the prediction results are used to guide the development of preventative maintenance plans.
[0034] Risk Management and Operations Subsystem: The maintenance and repair decision matrix is stored in the database, containing the mapping relationship of control strategies corresponding to different risk levels. The strategy for medium-risk areas is to increase the frequency of monitoring to once a week and limit the load to 8 tons for axle weight; the strategy for high-risk areas is to immediately limit the load to 5 tons for axle weight and initiate a grouting reinforcement plan; the strategy for extremely high-risk areas is to close the lane and excavate and rebuild.
[0035] The operation and maintenance management platform provides road maintenance managers with a visual interface, integrating functions such as electronic maps, monitoring data queries, early warning information push notifications, and maintenance task assignment. The platform interacts with the data processing platform in real time via API interfaces to obtain the latest roadbed health assessment results and risk control instructions.
[0036] In actual operation, this system performs full life cycle health monitoring and risk management tasks according to the following process.
[0037] Step 1: Deployment of Distributed Sensor Network: Within a representative section of the roadbed to be monitored, intelligent sensor nodes are evenly deployed according to a pre-defined grid layout strategy. Technicians first determine the monitoring area based on road design data, then use a total station to conduct on-site layout and positioning, marking the sensor node installation positions at 2-meter lateral and 8-meter longitudinal intervals. Each sensor node is fixed to the roadbed surface using a stainless steel bracket, with the sensor probe buried 5 centimeters below the roadbed surface. Connecting cables are protected by pre-buried PVC conduits and led to a roadside junction box. After installation, power-on testing and parameter configuration are performed to ensure all sensors are functioning correctly and data communication links are unobstructed.
[0038] Step 2: Moving load excitation and data synchronous acquisition: The data processing platform simultaneously sends synchronous acquisition commands to the distributed sensor network and the mobile payload test vehicle. The test vehicle enters the monitoring section at a constant speed of 60 km / h from the starting point, triggering pre-acquisition 200 meters from the starting point. Throughout the test vehicle's passage through the monitoring area, each sensor node synchronously acquires acceleration time history signals, strain time history signals, humidity time history signals, and temperature time history signals at a sampling frequency of 1000 Hz. The data acquisition duration is 30 seconds, covering the time window during which the test vehicle completely passes through the 500-meter monitoring area. The raw data from each sensor node is transmitted to the data concentrator via the LoRa network. After the data concentrator performs data packaging, format conversion, and timestamp correction, it is uploaded to the data processing platform via a fiber optic network.
[0039] Step 3: Multidimensional dynamic response data processing: After receiving the raw acquired data, the data processing platform first performs data quality control, removing outliers and noise interference. Then, the dispersion analysis module performs generalized S-transform time-frequency analysis on the acceleration time-history signal, extracting the phase velocities of Rayleigh and Love waves at different frequencies to generate a dispersion energy spectrum. The peak tracking algorithm automatically extracts the main dispersion curve, obtaining the correlation data between frequency and phase velocity. This data reflects the propagation characteristics of elastic waves at different frequencies in the roadbed and is directly related to the shear wave velocity distribution of the roadbed soil.
[0040] Step 4: Construction of the global mechanical model of the roadbed: The sparse-constrained inversion module receives dispersion analysis results as input and establishes the objective function for the inversion problem. This module divides the monitoring area into 20 layers along the depth direction, with a horizontal grid resolution of 0.5m x 0.5m, resulting in over 10,000 unknowns. L1 norm regularization constraints are used to promote sparsity of the solution, and the nonlinear optimization problem is solved using iterative weighted least squares. During each iteration, the forward modeling module estimates the theoretical dispersion curve based on the current shear wave velocity, compares it with the measured dispersion curve to calculate the residual, and the inversion module updates the shear wave velocity distribution based on the residual. After 50 iterations and convergence, a three-dimensional shear wave velocity imaging model of the roadbed is obtained, achieving a resolution of 0.5m x 0.5m x 0.25m.
[0041] Step 5: Calculation and visualization of dynamic elastic modulus: The elastic modulus calculation module loads the three-dimensional shear wave velocity imaging model of the roadbed and the soil physical parameter database. For each inversion grid cell, the module first calculates the elastic modulus according to the formula... Calculate the shear modulus, and then use the formula... The dynamic elastic modulus was calculated. During the calculation, the soil density was calibrated at 1850 kg / m³ using a field nuclear density meter, and the Poisson's ratio was set to 0.40 based on the cohesive soil type. For the gaps between grids, spatial interpolation smoothing was performed using the Kriging interpolation algorithm. Finally, a dynamic elastic modulus distribution map of the entire roadbed was generated and visualized on the electronic map of the operation and maintenance management platform in the form of a color heatmap. The elastic modulus values ranged from 20 to 200 MPa.
[0042] Step Six: Carrying Capacity Assessment and Risk Identification The bearing capacity assessment module loads the dynamic elastic modulus distribution map of the entire roadbed and historical health baseline data. The historical baseline data comes from the initial inspection report at the time of road completion and acceptance, where the average elastic modulus of the roadbed was measured to be 150 MPa. The module compares the current elastic modulus with the initial elastic modulus grid by grid and calculates the bearing capacity degradation index at each location. Based on the degradation index calculation results, three high-risk areas were identified in this monitored road section, located at K120+120, K120+280, and K120+420, with degradation indices of 0.35, 0.38, and 0.32, respectively; five medium-risk areas were identified, with degradation indices between 0.18 and 0.28; the remaining areas were all low-risk, with degradation indices less than 0.12.
[0043] The risk level classification module labels each zone with a risk level based on preset thresholds. High-risk areas automatically trigger a red alert signal, pushing an alert notification to maintenance management personnel through the operation and maintenance management platform, and simultaneously generating an emergency response work order. Medium-risk areas trigger a yellow alert, initiating encrypted monitoring procedures. Low-risk areas maintain the regular monitoring frequency.
[0044] Step Seven: Risk Management Strategy Generation and Implementation Based on the maintenance and repair decision matrix, the module generates corresponding risk control strategies for vulnerable areas with different risk levels. For the high-risk area at K120+120, the strategies include immediately limiting load to 5 tons, increasing monitoring frequency to once a day, and initiating grouting reinforcement. The system automatically sends control instructions to the roadbed operation and maintenance management platform, which generates maintenance task orders and assigns them to maintenance work teams, while simultaneously publishing load restriction notices via variable message signs. For medium-risk areas, the strategies include limiting load to 8 tons and increasing monitoring frequency to once a week. For low-risk areas, the routine maintenance plan is maintained.
[0045] Step 8: Predictive Maintenance and Full Lifecycle Data Management The predictive analysis module uses a long short-term memory neural network model to predict the evolution trend of roadbed bearing capacity. The model input features include 24 periods of elastic modulus monitoring data collected over the past 12 months, daily ambient temperature series, monthly average humidity series, and annual traffic flow series. The model outputs predicted elastic modulus values for the next 12 months. The prediction results show that the elastic modulus of the high-risk area at K120+280 will continue to decline at a rate of 2% per month, and the degradation index is expected to exceed 0.5 in 6 months. Based on the prediction results, the maintenance management department has initiated a preventative maintenance plan ahead of schedule, arranging for deep grouting reinforcement work to be carried out 3 months later.
[0046] The full lifecycle health monitoring database records complete data from each test in real time, including sensor node spatial coordinates, raw dynamic response data, inverted elastic modulus distribution data, load-bearing capacity assessment results, and maintenance records. The database employs a distributed storage architecture with a total capacity of 20 gigabytes and a data retention period of 15 years, meeting the needs of full lifecycle management.
[0047] Example 2 This embodiment addresses the special working conditions of soft soil sections of highways, illustrating the deployment scheme and technical adjustments of the system under complex geological conditions.
[0048] To address the unique characteristics of soft soil road sections, the deployment scheme of the sensor network subsystem has been adjusted as follows: sensor nodes are densified laterally to a spacing of 1.5 meters and maintained at an 8-meter spacing longitudinally, increasing the number of sensor nodes in a single monitoring section to 350, in order to meet the need for denser monitoring of areas sensitive to soft soil deformation. Regarding sensor node selection, the range of accelerometers has been expanded to ±100g, and the range of strain sensors has been expanded to ±5000 micro-strain, to accommodate the larger structural response amplitudes in soft soil road sections.
[0049] In the elastic modulus calculation module of the data processing platform, the soil density parameter was adjusted to 1650 kg / m³, and the Poisson's ratio was adjusted to 0.45 to reflect the higher compressibility characteristics of soft clay. The material constitutive model in the finite element forward simulation module was switched from the Mohr-Coulomb model to the modified Cambridge model to more accurately describe the elastoplastic behavior of soft clay.
[0050] The monitoring process for soft soil sections was optimized in the following aspects: the test vehicle speed was reduced to 40 km / h to obtain a longer signal duration and improve the signal-to-noise ratio. The data acquisition time was extended to 45 seconds to ensure that the test vehicle completely passed through the monitoring area. The frequency analysis range of the dispersion analysis algorithm was adjusted to 5 to 300 Hz to capture lower-frequency wave propagation characteristics in soft soil. The inversion grid resolution of the sparse constraint inversion algorithm was refined to 0.4 m x 0.4 m to improve the ability to identify local weak interlayers.
[0051] In the calculation of the bearing capacity degradation index, the high-risk threshold was adjusted to 0.25 and the medium-risk threshold was adjusted to 0.12 to adapt to the stricter control standards for soft soil road sections. The LSTM network of the predictive analysis module added historical settlement data as input features, and the prediction time span was extended to 18 months to match the longer deformation evolution cycle of soft soil road sections.
[0052] Example 3 This embodiment addresses the special working conditions of the transition zone between the bridge section and the roadbed on urban expressways, and elaborates on the deployment scheme and technical implementation of the system in the approach road section of the elevated bridge.
[0053] The section of the urban expressway from K5+800 to K6+100 includes a 300-meter bridge approach road. A significant stiffness difference exists between the roadbed and the bridge deck transition area in this section, requiring specialized monitoring. A distributed sensor network is deployed with a denser network in the transition area, with sensor nodes spaced 1 meter laterally and 5 meters longitudinally, totaling 180 nodes. Displacement sensors are added to these nodes to measure vertical settlement deformation. These displacement sensors are wire-type displacement gauges with a range of 0 to 500 mm and an accuracy of 0.1 mm.
[0054] The mobile load excitation subsystem uses two test vehicles traveling in convoy, maintaining a constant speed and a 30-meter interval as they pass through the monitoring area to generate a richer range of excitation frequency components. The axle load of the test vehicles was adjusted to 8 tons per axle to meet the load limits of urban roads.
[0055] The dispersion analysis module of the data processing platform adds coherence function analysis to the bridge approach road section to identify the converted wave characteristics of Rayleigh waves and Love waves, avoiding misinterpretation of bridge deck vibration signals as roadbed responses. The sparse constraint inversion module adopts adaptive mesh refinement technology in the transition region to automatically identify locations of abrupt stiffness changes and refine the local mesh.
[0056] The elastic modulus calculation module establishes elastic modulus models for both the roadbed and the bridge in the transition zone, and calculates the stiffness gradient distribution in the transition zone using a stiffness matching algorithm. The load-bearing capacity assessment module adds a comfort index evaluation function, calculating the ride comfort index based on the vibration acceleration time history when a vehicle passes, and assessing the impact of the transition zone on driving comfort.
[0057] In terms of risk management strategies, the maintenance and repair decision matrix for bridge approach road sections incorporates a bridge approach slab reinforcement strategy, automatically initiating the bridge approach slab reinforcement plan when the stiffness gradient exceeds a threshold. The LSTM network input features of the predictive analysis module incorporate bridge load stress monitoring data to predict the differential settlement development trend of approach road sections.
[0058] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Any equivalent modifications made by those skilled in the art within the spirit of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for monitoring and risk control of a whole life cycle of a road subgrade, characterized in that, Includes the following steps: Within a representative section of the roadbed to be monitored, densely distributed intelligent sensing nodes are uniformly deployed according to a preset grid layout strategy. Each intelligent sensing node integrates multiple types of sensors to achieve comprehensive perception of the mechanical response parameters of the roadbed structure. A mobile load source equipped with a dynamic load sensor travels along the monitored road segment, triggering synchronous data acquisition across the entire road segment during the travel of the mobile load source. Each of the intelligent sensing nodes synchronously collects multi-dimensional dynamic response signals triggered by the moving load source, and the data collection duration covers the time window during which the moving load source completely passes through the monitoring area. Based on the acquired multidimensional dynamic response signal, the propagation characteristic parameters of elastic waves are extracted using a dispersion analysis algorithm. Based on the propagation characteristic parameters, an inversion algorithm is used to invert the shear wave velocity distribution of the subgrade soil and construct a three-dimensional shear wave velocity imaging model of the subgrade. Based on the three-dimensional shear wave velocity imaging model, the dynamic elastic modulus of the roadbed is calculated according to the theoretical mapping relationship between shear wave velocity and elastic modulus, and the dynamic elastic modulus distribution map of the entire roadbed is obtained. Based on the dynamic elastic modulus distribution map, a roadbed bearing capacity assessment model is established using finite element forward modeling. Combined with historical health baseline data, the bearing capacity degradation index of each zone is calculated. Based on the degradation index, potential weak areas of the roadbed are identified and risk level classification is generated. For vulnerable areas with different risk levels, corresponding risk control strategies are generated based on a preset maintenance and repair decision matrix, and control instructions are sent to the roadbed operation and maintenance management platform.
2. The method of claim 1, wherein, The steps of deploying densely distributed intelligent sensing nodes include: Sensor nodes are deployed along the road in both the transverse and longitudinal directions at preset intervals, the preset intervals being determined based on the road grade and geological conditions; Each of the aforementioned intelligent sensing nodes adopts a low-power design and is equipped with an independent power supply unit to meet the requirements for long-term trouble-free operation; Each of the intelligent sensing nodes establishes a data connection through a preset communication protocol, forming a grid-covered sensing system.
3. The method of claim 1, wherein, The step of using a mobile load source equipped with a dynamic load sensor to travel along the monitored road segment includes: The axle load of the mobile load source is set according to the monitoring requirements. The contact pressure between the tire and the road surface is obtained in real time through the dynamic load sensor. The contact pressure is calculated by a preset calibration equation. The mobile load source passes through the monitoring area at a preset speed and triggers a pre-acquisition signal before entering the monitoring area to ensure that the acquisition time window covers the entire monitoring process. The number of test vehicles is determined based on the length of the monitoring section. At least one test vehicle is configured for each preset length section to ensure complete load excitation coverage within the monitoring area.
4. The method of claim 1, wherein, The steps for each of the aforementioned intelligent sensing nodes to synchronously acquire the multidimensional dynamic response signal triggered by the moving load source include: Each of the aforementioned intelligent sensing nodes is equipped with an independent synchronization clock module, which achieves time synchronization through the global positioning system timing signal to ensure that the time synchronization error of the sensing nodes across the entire road segment meets the preset accuracy requirements. The multidimensional dynamic response signal includes acceleration time history signal, strain time history signal, humidity time history signal and temperature time history signal, and each type of signal is collected by the corresponding sensor unit; During the data acquisition process, a quality control procedure is executed to preprocess the multidimensional dynamic response signal, remove outliers and noise interference, and retain valid signal data for subsequent analysis.
5. The method of claim 1, wherein, The steps for extracting the propagation characteristic parameters of elastic waves using the dispersion analysis algorithm include: The collected acceleration time history signal is processed by time-frequency analysis to extract the phase velocities of Rayleigh waves and Love waves at different frequencies and generate a dispersion energy spectrum. The peak tracking algorithm is used to automatically extract the main dispersion curve and obtain the corresponding data of frequency and phase velocity. The inversion algorithm employs L1 norm regularization constraints to promote sparsity of solutions. It solves the nonlinear optimization problem through iterative weighted least squares method and obtains the optimal shear wave velocity distribution after multiple iterations and convergence.
6. The method of claim 1, wherein, The steps for calculating the carrying capacity degradation index of each partition include: The current elastic modulus is compared with the initial elastic modulus grid by grid to calculate the bearing capacity degradation index at each location; The initial elastic modulus is derived from the historical health baseline data at the time of road completion and acceptance, and the historical health baseline data is stored in a pre-stored knowledge repository; Risk levels are determined based on the numerical range of the degradation index. Areas with a degradation index greater than a first threshold are classified as high-risk areas, areas with a degradation index between the first and second thresholds are classified as medium-risk areas, and areas with a degradation index less than the second threshold are classified as low-risk areas.
7. The method of claim 1, wherein, The steps for generating the corresponding risk management strategy include: For high-risk areas, control strategies are generated, including immediately limiting traffic to a preset axle load, increasing monitoring frequency to once a day, and initiating grouting reinforcement programs. For medium-risk areas, control strategies are generated, including limiting load passage to a preset axle weight and increasing monitoring frequency to once a week; For low-risk areas, the control strategy of maintaining the regular monitoring frequency will be maintained, and the existing maintenance plan will remain unchanged; The generated risk control strategy is sent to the roadbed operation and maintenance management platform through a preset communication interface, automatically generating maintenance task work orders and assigning them to maintenance work teams, while triggering the corresponding early warning information release process.
8. A system for implementing the method of claim 1, characterized by include: The distributed sensor network subsystem consists of a densely deployed array of intelligent sensor nodes. Each intelligent sensor node integrates an accelerometer, a strain sensor, a humidity sensor, and a temperature sensor to collect multidimensional response signals of the roadbed structure. The mobile load excitation subsystem consists of a test vehicle equipped with dynamic load sensors, which is used to travel along the monitored road section and apply dynamic loads to the roadbed. A data concentrator, deployed on the roadside of the monitored road section, is used to aggregate the raw data from each of the aforementioned smart sensor nodes and perform format conversion and timestamp correction. The data processing platform, deployed on a remote server cluster, is used to perform dispersion analysis, inversion calculation, and elastic modulus evaluation on the received raw data.
9. The system of claim 8, wherein, include: The dispersion analysis module is used to perform time-frequency analysis on the acquired multi-channel signals after preprocessing, and to extract the propagation characteristic parameters of elastic waves; The sparse constraint inversion module is used to invert the shear wave velocity distribution of the subgrade soil based on the propagation characteristic parameters and the sparse constraint inversion algorithm with L1 norm regularization constraints, and to construct a three-dimensional shear wave velocity imaging model. The elastic modulus calculation module is used to calculate the dynamic elastic modulus distribution of the roadbed based on the three-dimensional shear wave velocity imaging model, and to perform spatial interpolation processing using the Kriging interpolation algorithm to generate a global elastic modulus distribution map. The bearing capacity assessment module is used to establish a roadbed bearing capacity assessment model using finite element forward modeling, calculate the bearing capacity degradation index of each zone, and classify the risk level according to a preset threshold.
10. The system of claim 8, wherein, include: The predictive analysis module is used to build a long short-term memory neural network predictive analysis model based on sequence data in the historical monitoring database to predict the evolution trend of roadbed bearing capacity. The prediction results are used to guide the formulation of preventive maintenance plans. The risk management subsystem is used to generate corresponding risk management strategies by calling the preset maintenance and repair decision matrix according to different risk levels, and to issue management instructions to the roadbed operation and maintenance management platform. The full life cycle health monitoring database is used to record complete data for each test in real time, including the spatial coordinates of sensor nodes, the original dynamic response signal, the elastic modulus distribution data obtained by inversion, the load-bearing capacity assessment results, and maintenance records; The operation and maintenance management platform provides a visual operation interface for road maintenance management personnel, integrating functions such as electronic map display, monitoring data query, early warning information push and maintenance task assignment.
