A method for dynamically monitoring and analyzing the service state of a highway road body

Abstract

The application discloses a highway road body service state dynamic monitoring and analyzing method, relates to the technical field of intelligent highway monitoring, and realizes the associated monitoring of the service states of a pavement, a base layer and a roadbed by constructing a road body digital twin model and establishing the mapping relationship between multi-source sensing data and each structural unit. A dynamic service state benchmark is generated according to real-time conditions such as vehicle load, temperature, rainfall and water content, false positives caused by environmental disturbance and load fluctuation are reduced, and slow cumulative damage is timely identified. The associated unit enhancement sampling is driven by abnormal state evidence, the main abnormal reasons are determined by combining forward and parameter inversion, and the accuracy of abnormal positioning and reason identification is improved. Further, the structural evolution chain is constructed according to the inversion parameters, the abnormal expansion trend is predicted and graded early warning is implemented, the sensing strategy is reversely adjusted, the dynamic benchmark is updated, and a monitoring, analyzing, predicting and early warning closed loop is formed.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring technology for highways, specifically to a method for dynamic monitoring and analysis of the service status of highway road bodies. Background Technology

[0002] During long-term service, highways are continuously subjected to the combined effects of cyclical vehicle loads, impacts from overweight vehicles, temperature changes, rainfall infiltration, freeze-thaw cycles, and differential settlement of the subgrade. This causes gradual changes in the material properties and interlayer bonding of the pavement surface layer, base course, subbase course, and subgrade, potentially leading to anomalies such as surface cracks, interlayer voids, base course loosening, abnormal subgrade moisture content, and uneven settlement. These anomalies typically exhibit characteristics of spreading from localized areas to adjacent regions, propagating from the surface to deeper layers, or reflecting from the subgrade to the pavement.

[0003] Existing road inspection methods mainly include manual inspection, vehicle-mounted image inspection, ground-penetrating radar inspection, deflection inspection, and fixed-point monitoring using strain, vibration, displacement, or settlement sensors. However, the data obtained by different inspection devices vary greatly in terms of sampling time, spatial coverage, detection depth, and physical quantity type, making it difficult to directly reflect the correlation status and abnormal evolution process between the various structural layers of the road body.

[0004] Existing technologies have proposed using vehicle loads and sensor responses to monitor the structural condition of highways. For example, Chinese patent CN110725297B discloses a self-powered highway traffic and subgrade modulus monitoring system and method. This system places piezoelectric sensors and weighing devices in the highway subgrade, calculates the subgrade modulus using vehicle weight, vehicle speed, and the peak output voltage of the piezoelectric sensors, and issues a maintenance warning when the subgrade modulus decays beyond a preset ratio. This solution can achieve real-time monitoring of the subgrade modulus, but it mainly relies on the piezoelectric response under vehicle loads to calculate the subgrade modulus at a single location.

[0005] However, existing road condition monitoring methods typically compare measured data with fixed thresholds or initial values. When vehicle loads increase, pavement temperatures rise, or rainfall causes changes in structural moisture content, monitoring data may deviate significantly from the fixed thresholds, even without permanent damage to the road structure, leading to false alarms. Conversely, structural damage in a slow accumulation phase may not be identified in time because it has not yet reached the fixed threshold. Therefore, accurately monitoring and analyzing the true service condition of highway pavements under complex loads and environmental disturbances is a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0006] In order to solve the above-mentioned technical problems, this application proposes the following technical solution: In a first aspect, embodiments of this application provide a method for dynamic monitoring and analysis of the service status of a highway road body, including: Construct a hierarchical digital twin model of the road body, establish the mapping relationship between multi-source sensing data and the structural units of the twin model, and generate a dynamic service status benchmark; The real-time sensing data is compared with the dynamic service status benchmark to form evidence of abnormal status and increase the sampling frequency of abnormal units and their associated units. The abnormal state evidence is input into the road ontology digital twin model for forward modeling and parameter inversion to determine the main causes of the anomalies. Based on the main causes of anomalies and inversion parameters, a structural evolution chain is constructed to predict anomaly development trends and generate graded early warnings. The perception strategy is adjusted and the dynamic service status benchmark is updated according to the early warning results.

[0007] In one possible implementation, the construction of a hierarchical digital twin model of the road entity, the establishment of a mapping relationship between multi-source sensing data and the structural units of the twin model, and the generation of a dynamic service status benchmark include: By combining multi-dimensional conditions, road body monitoring units are divided and the relationships between units are sorted out to construct a spatial topology network; Collect data from various end-side sensing devices on the road, match the road body monitoring units according to the device deployment and detection parameters, and assign weights to cross-unit data according to spatial and signal conditions to achieve accurate spatial mapping between sensing data and digital twin units; Based on the monitoring data of the road during its normal service life and the historical data of the load environment, a multi-index service status benchmark field was established for each road body monitoring unit according to different working conditions.

[0008] In one possible implementation, the step of establishing a multi-index service status reference field for each road body monitoring unit based on historical data of road normal service life monitoring and load environment, categorized by working condition, includes: Acquire historical monitoring data of the road body during its normal service period, at any given time. Construct load environment excitation coordinates and road structure hysteresis memory ; Based on historical data from normal service status, from the... Three reference states adjacent to the current load environment excitation coordinates are selected from each monitoring unit. , and And calculate the first The monitoring unit State response curvature of class state features ,in: ; according to , and Generate the first under the current conditions Dynamic service state baseline value of class state characteristics ; The dynamic service state reference field of the road body is composed of dynamic reference values ​​corresponding to different state characteristics of each road body monitoring unit: in: This represents the set of road body monitoring units. This represents the number of types of state characteristics.

[0009] In one possible implementation, the acquisition of historical monitoring data of the road body during its normal service period, at time... Construct load environment excitation coordinates and road structure hysteresis memory ,include: Regarding the first Each road body monitoring unit, at any time Construct the load environment excitation coordinates: in: For the first Each monitoring unit at time The load environment excitation coordinates, For current vehicle axle load, cumulative equivalent axle load, or dynamic wheel load, This is the reference load under normal service conditions. The current road structure temperature, As the reference temperature, For temperature change scale parameters, This represents the current moisture content of the structural layer or subgrade. As the baseline moisture content, Given the current rainfall intensity, This refers to the current duration of rainfall. As the baseline rainfall effect, The default non-zero positive number; Constructing road structure hysteresis memory: in: For the first Each monitoring unit at time The amount of delayed memory, The length of the historical time window. This refers to the historical sampling period number. For the first The state memory decay coefficient of each monitoring unit.

[0010] In one possible implementation, comparing real-time sensing data with the dynamic service status benchmark to form evidence of abnormal status and increasing the sampling frequency of the abnormal unit and its associated units includes: When real-time monitoring data fluctuations exceed the normal range, sampling is increased and adjacent devices in the topology are linked to collect data collaboratively. After the anomaly subsides, the frequency is gradually reduced to achieve dynamic allocation of sensing resources. The side-side equipment uniformly calibrates and registers the multi-source monitoring data, generates two types of data slices: load and environment, and marks the reliability of each slice according to multiple data quality indicators; By comparing measured data with dynamic benchmarks to extract abnormal features, and combining slice reliability and spatial continuity to determine the strength of evidence, dynamic state evidence of the unit is generated.

[0011] In one possible implementation, the step of comparing measured data with a dynamic benchmark to extract abnormal features, combining slice reliability and spatial continuity to determine the strength of evidence, and generating evidence of the unit's dynamic state includes: Real-time monitoring values Compared with dynamic service status reference value Compare and calculate the first The first road body monitoring unit Anomaly degree of class state features: in: For the first The monitoring unit Anomaly degree of class state features The first under normal conditions Discrete scale of class state features It is an inverse hyperbolic sine function. For the first Each monitoring unit at time The amount of delayed memory; Based on anomaly Data credibility identification and spatial topological continuity form dynamic status evidence for the corresponding monitoring units.

[0012] In one possible implementation, the step of inputting the abnormal state evidence into a road ontology digital twin model for forward modeling and parameter inversion to determine the main causes of the anomaly includes: By inputting load and environmental parameters into the digital twin model for forward modeling, the theoretical response is obtained. The deviation of the dynamic state evidence calculation is compared to reflect the gap between the model's structural parameters and the actual road conditions. Based on the forward modeling deviation inversion of various road structural parameters, and combined with the range of anomaly spatial distribution, local or multi-element joint solutions are carried out to obtain the spatial distribution of structural parameters; Iterative forward and inverse modeling is used to verify structural parameters, and four types of causes are compared to determine the root cause of anomalies. The reliability of road service status is output by combining multiple indicators.

[0013] In one possible implementation, the iterative forward and inverse verification of structural parameters classifies four types of causes for comparison and matching degree to identify the root causes of anomalies, and combines multiple indicators to output the reliability of road service status, including: Establish a set of candidate anomaly causes: ,in: This indicates structural performance degradation; Indicates abnormal traffic load; This indicates a sudden change in environmental conditions; This indicates a malfunction in the sensing device; For each candidate cause of anomaly, the corresponding parameters in the digital twin model of the road body are adjusted, and multiple forward and inverse iterations are performed. For candidate anomaly reasons Calculate the first Multi-feature residual volume after sub-forward-inverse iteration ; Based on the multi-feature residual volume of multiple iterations Calculate the causes of candidate anomalies residual continuous shrinkage ; In the After each iteration, the final forward residual of each monitoring unit is calculated. ; Based on the spatial topology network of the road body, combined with the above Calculate candidate anomaly causes Corresponding spatial residual fracture ; Candidate anomaly reasons Structural state parameters obtained by inversion Re-input the digital twin model of the road ontology to obtain the re-forward modeled simulation response; Using the re-forward modeled response as the inversion data, structural state parameter inversion is performed again to obtain the loop-loop structural state parameters. According to the above Calculate candidate anomaly causes Forward and inverse loop bias ; Based on the residual continuous shrinkage Spatial residual fracture degree and forward and inverse loop deviation Calculate the causes of candidate anomalies Closed-loop consistency index ; Closed-loop consistency index The largest candidate cause of the anomaly was determined to be the main reason for the current road body anomaly. .

[0014] In one possible implementation, the candidate anomaly cause... Calculate the first Multi-feature residual volume after sub-forward-inverse iteration ,include: in: Candidate anomaly cause In the Multi-feature residual volume after the next iteration The number of types of state features. Candidate anomaly cause The corresponding set of monitoring units, Candidate anomaly cause Next State features obtained from forward modeling. This indicates median operations.

[0015] In one possible implementation, the step of constructing a structural evolution chain based on the main causes of the anomaly and inversion parameters, predicting the anomaly development trend and generating a graded early warning, adjusting the perception strategy according to the early warning results and updating the dynamic service status benchmark includes: Store road time-series service data, combine spatial topology analysis to identify the anomaly diffusion pattern, and construct a structural evolution chain based on the anomaly time-series correlation to distinguish different types of road body service status anomalies. Match the abnormal evolution pattern of the road body service status to determine the abnormal type of the road body service status, adjust the twin model parameters to perform multi-condition simulation, and predict the development trend, impact range and maintenance critical time of the abnormal road body service status. By integrating multiple indicators to classify three levels of early warning, coordinating and regulating sensing and data collection strategies, and updating the twin model after maintenance, a complete monitoring and analysis closed loop is formed.

[0016] In this embodiment, a layered digital twin model of the road body is constructed, and a mapping relationship between multi-source sensing data and each structural unit is established to achieve correlated monitoring of the service status of the pavement, base course, and subgrade. A dynamic service status benchmark is generated based on real-time conditions such as vehicle load, temperature, rainfall, and moisture content, reducing false alarms caused by fixed thresholds under environmental disturbances and load fluctuations, and promptly identifying slowly accumulating damage. Enhanced sampling of associated units is driven by evidence of abnormal states, and the main causes of anomalies are determined by combining forward modeling and parameter inversion, improving the accuracy of anomaly location and cause identification. Furthermore, a structural evolution chain is constructed based on the inversion parameters to predict anomaly expansion trends and implement tiered early warnings. Simultaneously, the sensing strategy is adjusted in reverse, and the dynamic benchmark is updated, forming a closed loop of monitoring, analysis, prediction, and early warning. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a method for dynamic monitoring and analysis of the service status of a highway road body, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the road body monitoring unit division provided in an embodiment of this application; Figure 3 This is a schematic diagram of a road body monitoring system provided in an embodiment of this application; Figure 4 The closed-loop consistency index histogram of candidate anomaly causes provided in the embodiments of this application; Figure 5 This is a schematic diagram of the road body structure evolution chain provided in the embodiments of this application; Figure 6 A comparison curve of dynamic service status reference value and measured response value provided for embodiments of this application. Detailed Implementation

[0018] The present solution will now be described in conjunction with the accompanying drawings and specific embodiments.

[0019] See Figure 1 The method for dynamic monitoring and analysis of the service status of highway road bodies in this embodiment includes: S101, construct a hierarchical digital twin model of the road body, establish the mapping relationship between multi-source sensing data and the structural units of the twin model, and generate a dynamic service status benchmark.

[0020] The road body is continuously segmented along the mileage direction of the highway route. The continuous segmentation results are corrected by combining the road structure, material composition, construction batch, terrain conditions, traffic load characteristics, drainage conditions and abnormal distribution of historical road body service status, so as to obtain multiple road body monitoring units.

[0021] like Figure 2As shown, the 400m test section was first divided into four longitudinal sections with a base length of 100m along the road mileage direction. For the cut-fill junction, culvert transition section, and areas with concentrated existing cracks, the base monitoring unit was further subdivided into special monitoring units with a length of 20-50m.

[0022] In the cross-sectional direction of the road, lateral monitoring units are divided according to the driving lanes, emergency lanes, shoulders, and central median. In the depth direction of the road structure, the road body is divided into four levels of monitoring units: pavement surface layer, base layer, subbase layer, and subgrade. For example, the monitoring unit corresponding to the pavement surface layer in the first longitudinal section is denoted as U11, the monitoring unit corresponding to the pavement surface layer in the second longitudinal section is denoted as U12, and the monitoring unit corresponding to the subgrade in the first longitudinal section is denoted as U21.

[0023] Based on the actual functional relationships of the road structure, establish a spatial topology network for the road body: in, For monitoring unit set, This is the set of topological connection edges between monitoring units.

[0024] The topological connection edges include: The first type is the vertical load transfer connection edge, used to represent the vertical load transfer relationship between the pavement surface layer, base course, subbase course, and subgrade. The second type is the longitudinally adjacent connection edge, used to represent the continuous influence relationship between adjacent mileage sections in terms of vehicle load propagation, crack propagation, and settlement development. The third type is the laterally related connection edge, used to represent the lateral deformation relationship between different lanes, shoulders, and central medians within the same cross-section. The fourth type is the moisture migration connection edge, used to represent the vertical seepage of rainfall moisture between different structural layers, and the relationship of moisture migration along the road's lateral drainage facilities. For cut-fill junctions, bridge abutment overlaps, and culvert transition sections, structural abrupt change connection edges are added to the original topological connection edges to represent the locations where stiffness, boundary conditions, or support conditions change abruptly.

[0025] like Figure 3 As shown, end-side sensing equipment was installed within the test section. A set of road surface image acquisition equipment and millimeter-wave radar was installed every 100m to collect data on road surface cracks, deformation, and vehicle operating status. Strain sensors, vibration sensors, and displacement sensors were installed in each key monitoring area. Moisture content sensors were installed in the base layer or subgrade. Settlement monitoring equipment was installed at cut-fill junctions and culvert transition sections. Dynamic weighing equipment was installed at the entrance of the test section. A weather station was located near the test section.

[0026] In this embodiment, the sampling frequency of the vibration sensor is 200Hz under normal conditions and 1000Hz under abnormal enhanced sampling conditions; the sampling frequency of the strain sensor is 10Hz under normal conditions and 100Hz under abnormal enhanced sampling conditions; the sampling period for water content, temperature and settlement data is 10min under normal conditions and shortened to 1min under abnormal conditions; the sampling period for road surface images is 5min under normal conditions and shortened to 10s under abnormal conditions.

[0027] Based on the actual installation coordinates, detection direction, detection depth, and detection coverage of the sensing device, the sensing data is mapped to the corresponding monitoring unit. For sensing devices that simultaneously cover multiple monitoring units, data attribution is determined according to the distance between the device's detection center and the monitoring unit, the angle of the detection direction, and the effective signal coverage depth. If a certain detection data corresponds to multiple monitoring units, its spatial association identifier with each monitoring unit is retained separately, and the detection data is not directly merged at this stage.

[0028] Regarding the first Each monitoring unit, at any time Forming state feature vectors: in, Indicates the first One road body monitoring unit, Indicates the monitoring time. The number of types of state features. Indicates the first Each monitoring unit at time The State characteristics. In this embodiment, state characteristics include at least one of the following: dominant vibration frequency, vibration attenuation rate, peak strain, strain recovery time, residual displacement, settlement, settlement rate, structural moisture content, crack width, crack length, ground penetrating radar reflection intensity, and interlayer void characteristics.

[0029] To eliminate the problem of different dimensions for characteristics under different states, the central value and discrete scale of each state characteristic are calculated using historical data from normal service conditions. The monitoring unit The central value of the class state feature is: The discrete scale is: in: This indicates taking the median.

[0030] When calculated Smaller than the preset minimum scale season: In this embodiment, Take twice the corresponding sensor resolution. This process prevents the denominator in subsequent calculations from approaching zero due to excessively small changes in normal historical data.

[0031] Transform the original state features into dimensionless state features: in: To prevent positive numbers with a denominator of zero, this embodiment takes... .

[0032] Continuous 90-day monitoring data of roads confirmed by manual inspection to be free of obvious cracks, voids, loosening, and abnormal settlement were selected as historical data for normal service conditions. This embodiment uniformly uses 10 minutes as the calculation window for load environment conditions. Vehicle load. Take the 95th percentile value of the dynamic weighing axle load within the corresponding time window to avoid a small number of abnormal extreme values ​​having too much impact on the calculation results.

[0033] The cumulative effect of rainfall is expressed as: in: Rainfall intensity, The duration of continuous rainfall. The unit is mm.

[0034] Constructing the load-environment excitation coordinates: in: For the first Each monitoring unit at time The load environment excitation coordinates; This refers to the current vehicle axle load, cumulative equivalent axle loads, or dynamic wheel load. This is the reference load under normal service conditions; The current road structure temperature; Reference temperature; The temperature change scale parameter is the standard deviation of temperature change during normal service. This represents the current moisture content of the structural layer or subgrade. The baseline moisture content; This represents the current rainfall intensity. This refers to the current duration of rainfall; The baseline rainfall effect is taken as the 95th percentile of the cumulative rainfall effect during normal service life. The above formula multiplicatively couples load and temperature, resulting in different excitation coordinates for the same vehicle load under different temperature conditions. Simultaneously, it converts moisture content change into a nonlinear radical term retaining the direction of change, and rainfall intensity and duration into a cumulative infiltration effect term.

[0035] In one implementation, when sufficient historical temperature data is lacking, Take 10℃, We set 0.2 as the initial empirical value, and recalibrated it using historical data from normal service in actual applications. This avoids the unreasonable elimination of traffic load due to the temperature correction term approaching zero under low-temperature conditions. All input quantities in the above formula have been normalized to reference values, therefore all terms are dimensionless and can be directly combined nonlinearly.

[0036] To reflect the state hysteresis of road structures after continuous heavy loads, rain immersion, or freeze-thaw cycles, a hysteresis memory of the road structure is further constructed: in: For the first Each monitoring unit at time The amount of delayed memory; The length of the historical time window; This refers to the historical sampling period number; For the first The state memory decay coefficient of each monitoring unit is obtained through calibration using historical normal data; that is, the parameter that minimizes the prediction error of the baseline state after traversing the candidate range is taken as the final parameter. When the road body experiences continuously increasing heavy loads or moisture, It tends towards a positive value; when external forces gradually decrease and the road structure is in the process of recovery. The value tends to be negative; when the external influence is stable over a long period of time, Approaching zero.

[0037] In this embodiment, the history window length Take 18, which corresponds to 3 consecutive hours of historical data; State memory decay coefficient The initial value is 6. In other implementations, A value of 6-72 is acceptable. The range can be 2 to 24. The specific values ​​are determined by calibrating with normal historical data. That is, dynamic benchmarks are calculated for each candidate parameter range, and the parameter set that minimizes the median absolute deviation between the measured values ​​and the dynamic benchmark values ​​during the normal historical period is taken as the final parameter set.

[0038] Normal historical data are arranged according to the load environment excitation coordinates. Arranged in ascending order, bins are created with an initial interval width of 0.05. Each interval retains at least 30 valid samples, and the median of each state feature is used as the reference state feature for that interval. For the current excitation coordinates... Select the reference point that is closest to and smaller than the current excitation coordinates as the reference state. Select the reference point that is closest to and larger than the current excitation coordinates as the reference state. Then select the state in the reference state. The next reference point is used as the reference state. ,satisfy: If the current excitation coordinates are outside the historical reference range, the three reference states closest to the boundary are used, and the current excitation coordinates are truncated to the boundary of the historical reference range. At the same time, the corresponding data is marked as data outside the reference range, triggering enhanced sampling instead of direct infinite extrapolation.

[0039] Based on the characteristics of dimensionless states , and Calculate the state response curvature: because and All are dimensionless quantities, therefore the state response curvature in the formula is also dimensionless, avoiding inconsistencies in dimensions caused by different physical quantities directly participating in the calculation. In this embodiment, It is not used to describe the curvature of geometric curves, but rather to describe the second-order trend of the rate of change of road state response during adjacent load-environment excitation state changes, and to distinguish between normal elastic response and nonlinear response caused by structural degradation.

[0040] Calculate the dimensionless value of the dynamic baseline: Then, the dimensionless dynamic reference is converted into a dynamic service state reference value with actual physical units: The dynamic service status reference field is composed of the dynamic reference values ​​of each state characteristic of each monitoring unit: Thus, the spatial topological relationship of the road body, the mapping relationship of multi-source sensing data, and the reference field of dynamic service status have been established.

[0041] S102, compare the real-time sensing data with the dynamic service status benchmark to form evidence of abnormal status, and increase the sampling frequency of the abnormal unit and its associated units.

[0042] End-side sensing devices collect road condition data at a standard sampling frequency. Side-side computing devices calculate the anomaly degree of the condition characteristics based on the obtained dynamic baseline values ​​after each 10-minute window. in, It is an inverse hyperbolic sine function. When When the corresponding state characteristics are normal fluctuations, the corresponding state characteristics are determined to be normal fluctuations; when three consecutive calculation windows meet the condition... When the corresponding state feature is marked as a suspected anomaly, and when three consecutive calculation windows meet the condition... At that time, enhanced sampling is triggered.

[0043] After triggering enhanced sampling, not only is the sampling frequency of the current anomaly monitoring unit increased, but a spatial topology network is also established to perform collaborative enhanced sampling on first-order adjacent monitoring units directly connected to the current monitoring unit. When all anomaly state features have recovered to [a certain value] within six consecutive calculation windows... At that time, the normal sampling frequency is gradually restored.

[0044] The edge computing device converts data from different end-side devices into Coordinated Universal Time (UTC), and uses the moment a vehicle enters the monitoring unit, recorded by a dynamic weighing device, as the load application center moment. For strain, vibration, and displacement data, data are extracted 2 seconds before vehicle entry, during vehicle application, and 5 seconds after vehicle departure to form single-vehicle load application slices. For rainfall and moisture content changes, the moments of rainfall initiation, peak cumulative rainfall, and moisture content recovery to a stable state are used as environmental action nodes to form environmental action slices.

[0045] Each data slice includes the monitoring unit number, data acquisition device number, acquisition time, spatial coordinates, traffic load, environmental status, status characteristics, and data quality information. Data quality information includes data missing rate, time synchronization error, sensor self-test status, and consistency with adjacent sensors.

[0046] The confidence level of a data slice is marked as high when the data missing rate is no greater than 5%, the time synchronization error is no greater than one sampling period, the sensor self-test is normal, and at least two different types of sensors detect state changes in the same direction. It is marked as medium if only two of these conditions are met, otherwise it is marked as low. Low-confidence data is not directly used as evidence of structural performance degradation, but is primarily used to determine malfunctions in sensing devices.

[0047] Edge computing devices combine anomaly degree, data slice confidence, and spatial continuity to form dynamic state evidence. When a single state feature of a single monitoring unit is anomalous and there is no corresponding change in adjacent monitoring units, evidence to be verified is generated. When at least two types of state features are anomalous simultaneously within a monitoring unit, or when the same state feature is anomalous consecutively within two or more topologically adjacent monitoring units, highly reliable dynamic state evidence is generated.

[0048] For example, when the vibration attenuation rate of monitoring unit U12 is abnormal, and the roadbed support response of its lower monitoring unit U22 is abnormal, and the ground penetrating radar shows abnormal reflection intensity at the same location, the location is marked as a suspected interlayer void or loose base area, and the dynamic status evidence is uploaded to the cloud-based digital twin platform.

[0049] S103, input the abnormal state evidence into the road ontology digital twin model, and perform forward modeling and parameter inversion to determine the main causes of the abnormality.

[0050] The current vehicle axle load, vehicle position, vehicle speed, ambient temperature, rainfall conditions, road structure moisture content, and current structural state parameters of each monitoring unit are input into the road body digital twin model. This digital twin model is established using a layered elastic finite element model. Each monitoring unit includes at least the pavement surface layer, base layer, subbase layer, and subgrade structural layer. Initial parameters include the thickness, elastic modulus, damping parameters, density, interlayer bonding degree, subgrade support stiffness, and boundary conditions of each structural layer.

[0051] Input the current vehicle axle load, vehicle position, vehicle speed, structural temperature, moisture content, and rainfall into the digital twin model to calculate the theoretical strain, theoretical displacement, theoretical vibration response, and theoretical settlement response of each monitoring unit.

[0052] Based on the theoretical response, extract the same feature quantities as the measured state characteristics to form: This embodiment sets four types of candidate anomaly causes: in, This indicates structural performance degradation. Indicates abnormal traffic load. This indicates a sudden change in environmental conditions. This indicates an anomaly in the sensing equipment. Different candidate causes of anomalies correspond to different parameters to be retrieved. Parameters corresponding to candidate causes of structural performance degradation include structural layer modulus, interlayer bonding degree, local porosity, subgrade support stiffness, equivalent crack depth, and local settlement. Parameters corresponding to candidate causes of abnormal traffic loads include vehicle axle load correction factor, vehicle position correction factor, and vehicle speed correction factor. Parameters corresponding to candidate causes of abrupt changes in environmental conditions include temperature response correction factor, water content sensitivity factor, and water infiltration boundary correction factor. Parameters corresponding to candidate causes of sensing equipment anomalies include zero-point offset, scale deviation, time offset, and noise amplitude.

[0053] To ensure the inversion process is repeatable, this embodiment employs a bounded coordinate rotation search method. The initial values ​​of each parameter are taken from the current values ​​of the digital twin model, and the parameter search ranges are set as follows: The structural layer modulus is 30%–120% of the initial modulus, the interlayer bonding degree is 0.05–1, the local porosity is 0–0.20, the subgrade support stiffness is 30%–120% of the initial value, the equivalent crack depth is 0 to the corresponding structural layer thickness, the local settlement is 0–50 mm, the vehicle load correction factor is 0.7–1.5, the water content sensitivity coefficient is 0.5–2.0, and the sensor zero-point offset is the corresponding discrete scale. The sensor time offset is -5 to 5 times, and the sensor time offset is -2 to 2 sampling periods.

[0054] At the start of the inversion, the search step size for each parameter is 10% of its allowable range. Only one parameter is changed at a time, and the forward modeling results are calculated for increasing and decreasing the parameter by one step size, retaining the parameter value that minimizes the multi-feature residual volume. After completing one round of searching for all parameters, if the multi-feature residual volume decreases by less than 1%, the step size for all parameters is halved. Maximum number of iterations. The iteration is repeated 20 times. The inversion is terminated when the residual decrease rate is less than 1% for three consecutive iterations, or when the step size of all parameters is less than 0.5% of the allowable range for the corresponding parameters. This yields the candidate causes of the anomalies. Corresponding inversion parameters: For each candidate cause of anomaly, calculate the... Multi-feature residual volume after forward-inverse iteration: in: Candidate anomaly cause In the Multi-feature residual volume after the next iteration; Candidate anomaly cause The corresponding set of monitoring units; Candidate anomaly cause Next State characteristics obtained from forward modeling; This represents the median calculation. Using a geometric product structure of residuals from multiple state features, the multi-feature residual volume only decreases significantly when multiple state features such as strain, displacement, vibration, water content, and settlement are simultaneously improved. When only a single state feature has a good fit, the multi-feature residual volume does not decrease to the same degree.

[0055] Based on the multi-feature residual volume from multiple iterations, candidate anomaly causes are calculated. The residual continuous shrinkage amount: in: Candidate anomaly cause The residual continuous shrinkage amount; This represents the number of forward-inverse iterations. This refers to the initial multi-feature residual volume before structural parameter correction. When each iteration gradually brings the forward modeling result closer to the real-time monitoring result... It is positive and gradually increases; when the residual repeatedly increases and decreases during the iteration process, the geometric product structure suppresses the unstable contraction process only when it accidentally decreases once.

[0056] In the After each iteration, the final forward residuals of each monitoring unit are calculated: Based on the spatial topology network of the road body, candidate anomaly causes are calculated. Corresponding spatial residual fracturing degree: in: Candidate anomaly cause Spatial residual fracture degree; The set of spatial topological connecting edges in the candidate monitoring unit set; For the first The monitoring unit The normal discrete scale of class-state characteristics. When a candidate anomaly cause can only explain a single location, but not the propagation of the anomaly across the pavement, base course, subgrade, or adjacent mileage sections, the spatial residual distribution breaks down, resulting in spatial residual discontinuity. Increase.

[0057] To avoid directly calculating distances using parameters with different dimensions such as elastic modulus, settlement, and degree of adhesion, the inversion parameters are first normalized. Parameter definition: The parameters obtained from the inversion are re-input into the digital twin model for forward modeling, and the inversion is performed again on the re-forward modeled data to obtain the loopback parameters. .

[0058] Calculate candidate anomaly causes Forward and inverse loop bias: in: Candidate anomaly cause Forward-inverse loop bias; This represents the norm. When the candidate anomaly cause can accurately describe the current structural state of the road body, the structural state parameters obtained from the inversion of measured data remain stable after inversion—forward modeling—and another inversion, with the closure bias... The deviation is relatively small. When the candidate anomaly cause cannot accurately describe the current state of the road body, even if the forward modeling response deviation is temporarily reduced by over-adjusting some parameters, it is difficult to recover the original structural state parameters after the inversion, thus resulting in a closure deviation. Larger.

[0059] Calculate the closed-loop consistency index based on the residual continuous shrinkage, spatial residual fragmentation, and forward-inverse loop lag: Since a single residual decrease cannot prove the correctness of the cause of anomalies, spatial continuity constraints and parametric laparosity stability constraints are introduced. Excessive residual shrinkage is compressed using a logarithmic function, and the multiplicative relationship between spatial residual fragmentation and laparosity bias is used to jointly penalize spatial discontinuities and parametric laparosity instability. Furthermore, the difference between spatial residual fragmentation and laparosity bias is used to suppress candidate causes of anomalies that only have unilateral explanatory power.

[0060] Finally, the candidate anomaly cause with the largest closed-loop consistency index was determined as the main cause of the current road body anomaly: .

[0061] In an exemplary anomaly analysis of this embodiment, the calculation results of each candidate anomaly cause are shown in Table 1: Table 1. Anomaly Analysis Results like Figure 4 As shown, the closed-loop consistency index corresponding to structural performance degradation is the largest. The larger the closed-loop consistency index, the stronger the explanatory power of the corresponding candidate anomaly cause for the measured anomaly state. Therefore, structural performance degradation is identified as the main cause of the current anomaly.

[0062] When the difference between the largest and second-largest closed-loop consistency indices is less than 0.08, a single cause is not directly determined. Instead, two composite candidate causes are retained, and supplementary sampling corresponding to these candidate causes is triggered. For example, when the closed-loop consistency indices corresponding to structural performance degradation and abrupt changes in environmental conditions are close, the sampling frequencies of the moisture content sensor, ground-penetrating radar, and strain sensor are simultaneously increased to further determine whether the anomaly is a temporary change in moisture content or permanent structural damage.

[0063] S104. Construct a structural evolution chain based on the main causes of the anomaly and the inversion parameters, predict the anomaly development trend and generate a graded early warning, adjust the perception strategy according to the early warning results and update the dynamic service status benchmark.

[0064] The abnormal state characteristics, anomaly degree, main causes of anomalies, closed-loop consistency index, and inversion structure parameters of each monitoring unit are stored in chronological order. An abnormal event node is generated when the anomaly degree of a monitoring unit is not lower than 2.0 for three consecutive windows and the data reliability is medium or high. When a subsequent abnormal event occurs within a first-order topologically adjacent monitoring unit of the current abnormal monitoring unit, with an occurrence time interval not exceeding 6 calculation windows, and the main causes of the anomalies are the same or related, a directed evolutionary connection edge is established between the two abnormal event nodes. Multiple abnormal event nodes connected in chronological order constitute the road body structure evolution chain.

[0065] For example, if crack growth is first detected in the pavement surface layer monitoring unit, followed by radar reflection anomalies in the corresponding base layer monitoring unit, and then vibration attenuation anomalies and increased settlement rate are detected in the subgrade monitoring unit, then a structural evolution chain is formed: surface crack growth—interlayer voids—base layer loosening—subgrade uneven settlement, as follows. Figure 5 As shown.

[0066] The current structural evolution chain is matched with pre-stored abnormal evolution patterns of road body service status to determine candidate abnormal road body service status types. For each candidate abnormal road body service status, the digital twin model is updated using the obtained structural state parameters, and traffic load and environmental action scenarios for the next 7 days, 30 days, and 90 days are set respectively. The traffic load scenario includes at least a normal traffic volume scenario and a heavy traffic increase scenario. The environmental scenario includes at least a normal environment scenario and a continuous rainfall scenario. Forward modeling of the future state is performed to predict crack width, interlayer void range, structural layer modulus reduction, and cumulative settlement.

[0067] like Figure 6As shown, when the measured response is basically consistent with the dynamic baseline during the normal phase, but deviates significantly from the dynamic baseline within multiple consecutive monitoring windows, and the forward modeling results indicate that the anomaly range will continue to expand, the anomaly is identified as a continuously evolving structural anomaly. When the field detection results are inconsistent with the predicted results, the field detection results are used as new calibration data to re-invert the structural parameters and update the digital twin model.

[0068] A three-level early warning system is generated based on the anomaly degree, closed-loop consistency index, anomaly duration, spatial expansion range, and future trend prediction results.

[0069] A Level 4 alert is generated when any of the following conditions are met: the absolute value of the anomaly is 2.0–3.0; the anomaly exists only in one monitoring unit and has not spread to adjacent monitoring units; the closed-loop consistency index is 0.50–0.65; and the maintenance threshold is predicted not to be reached within 30 days. The corresponding handling method for a Level 4 alert is enhanced sampling and continuous monitoring.

[0070] An inspection-level warning is generated when any of the following conditions are met: the absolute value of the anomaly is 3.0–5.0; the anomaly covers more than two topologically adjacent monitoring units; the closed-loop consistency index is not less than 0.65; and the maintenance threshold is predicted to be reached within 30 days. The corresponding handling methods for inspection-level warnings are on-site verification, ground-penetrating radar testing, deflection testing, or core sampling.

[0071] A Level 3 alert is generated when any of the following conditions are met: the absolute value of the anomaly degree is greater than 5.0; the anomaly propagation speed continues to increase; the digital twin model predicts that the traffic safety threshold will be reached within 7 days; or there is a risk of rapid settlement, significant interlayer voiding, or roadbed instability. The corresponding response measures for a Level 3 alert are speed limits, lane control, emergency detection, and maintenance.

[0072] After an early warning is generated, the warning level, anomaly location, and anomaly propagation path are fed back to the edge computing devices and end-side sensing devices. For warnings at the level of concern, the sampling frequency of the corresponding monitoring unit and its first-order adjacent monitoring units is increased to 2 to 5 times the normal sampling frequency. For warnings at the level of inspection, the sampling frequency is increased to 5 to 10 times the normal sampling frequency, and relevant sensing types such as ground-penetrating radar, displacement, and water content are added. For warnings at the level of handling, the sensors on the anomaly monitoring unit and potential propagation paths are switched to the highest permissible sampling frequency, and the data upload cycle is shortened to less than 1 minute.

[0073] After on-site inspection or maintenance is completed, the on-site inspection results, maintenance methods, maintenance locations, and post-treatment monitoring data are input into the digital twin model to update the structural layer modulus, interlayer bonding degree, subgrade support stiffness, state memory attenuation coefficient, and normal service state reference data.

[0074] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0075] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. The protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for dynamic monitoring and analysis of the service status of highway road bodies, characterized in that, include: Construct a hierarchical digital twin model of the road body, establish the mapping relationship between multi-source sensing data and the structural units of the twin model, and generate a dynamic service status benchmark; The real-time sensing data is compared with the dynamic service status benchmark to form evidence of abnormal status and increase the sampling frequency of abnormal units and their associated units. The abnormal state evidence is input into the road ontology digital twin model for forward modeling and parameter inversion to determine the main causes of the anomalies. Based on the main causes of anomalies and inversion parameters, a structural evolution chain is constructed to predict anomaly development trends and generate graded early warnings. The perception strategy is adjusted and the dynamic service status benchmark is updated according to the early warning results.

2. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 1, characterized in that, The construction of a hierarchical digital twin model of the road body, the establishment of a mapping relationship between multi-source sensing data and the structural units of the twin model, and the generation of a dynamic service status benchmark include: By combining multi-dimensional conditions, road body monitoring units are divided and the relationships between units are sorted out to construct a spatial topology network; Collect data from various end-side sensing devices on the road, match the road body monitoring units according to the device deployment and detection parameters, and assign weights to cross-unit data according to spatial and signal conditions to achieve accurate spatial mapping between sensing data and digital twin units; Based on the monitoring data of the road during its normal service life and the historical data of the load environment, a multi-index service status benchmark field was established for each road body monitoring unit according to different working conditions.

3. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 2, characterized in that, Based on historical data from road normal service life monitoring and load environment, a multi-index service status reference field is established for each road body monitoring unit under different working conditions, including: Acquire historical monitoring data of the road body during its normal service period, at any given time. Construct load environment excitation coordinates and road structure hysteresis memory ; Based on historical data from normal service status, from the... Three reference states adjacent to the current load environment excitation coordinates are selected from each monitoring unit. , and And calculate the first The monitoring unit State response curvature of class state features ,in: ; according to , and Generate the first under the current conditions Dynamic service state baseline value of class state characteristics ; The dynamic service state reference field of the road body is composed of dynamic reference values ​​corresponding to different state characteristics of each road body monitoring unit: in: This represents the set of road body monitoring units. This represents the number of types of state characteristics.

4. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 3, characterized in that, The acquisition of historical monitoring data during the period when the road body was in normal service condition, at time Construct load environment excitation coordinates and road structure hysteresis memory ,include: Regarding the first Each road body monitoring unit, at any time Construct the load environment excitation coordinates: in: For the first Each monitoring unit at time The load environment excitation coordinates, For current vehicle axle load, cumulative equivalent axle load, or dynamic wheel load, This is the reference load under normal service conditions. The current road structure temperature, As the reference temperature, For temperature change scale parameters, This represents the current moisture content of the structural layer or subgrade. As the baseline moisture content, Given the current rainfall intensity, This is the current duration of rainfall. As the baseline rainfall effect, The default value is a non-zero positive number. Constructing road structure hysteresis memory: in: For the first Each monitoring unit at time The amount of delayed memory, The length of the historical time window. This refers to the historical sampling period number. For the first The state memory decay coefficient of each monitoring unit.

5. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 1, characterized in that, The step of comparing real-time sensing data with the dynamic service status benchmark to form evidence of abnormal status and increasing the sampling frequency of abnormal units and their associated units includes: When real-time monitoring data fluctuations exceed the normal range, sampling is increased and adjacent devices in the topology are linked to collect data collaboratively. After the anomaly subsides, the frequency is gradually reduced to achieve dynamic allocation of sensing resources. The side-side equipment uniformly calibrates and registers the multi-source monitoring data, generates two types of data slices: load and environment, and marks the reliability of each slice according to multiple data quality indicators; By comparing measured data with dynamic benchmarks to extract abnormal features, and combining slice reliability and spatial continuity to determine the strength of evidence, dynamic state evidence of the unit is generated.

6. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 5, characterized in that, The process involves comparing measured data with dynamic benchmarks to extract abnormal features, combining slice reliability and spatial continuity to determine the strength of evidence, and generating evidence of the unit's dynamic state, including: Real-time monitoring values Compared with dynamic service status baseline value Compare and calculate the first The first road body monitoring unit Anomaly degree of class state features: in: For the first The monitoring unit Anomaly degree of class state features The first under normal conditions Discrete scale of class state features It is an inverse hyperbolic sine function. For the first Each monitoring unit at time The amount of delayed memory; Based on anomaly Data credibility identification and spatial topological continuity form dynamic status evidence for the corresponding monitoring units.

7. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 1, characterized in that, The step of inputting the abnormal state evidence into the road ontology digital twin model for forward modeling and parameter inversion to determine the main causes of the anomalies includes: By inputting load and environmental parameters into the digital twin model for forward modeling, the theoretical response is obtained. The deviation of the dynamic state evidence calculation is compared to reflect the gap between the model's structural parameters and the actual road conditions. Based on the forward modeling deviation inversion of various road structural parameters, and combined with the range of anomaly spatial distribution, local or multi-element joint solutions are carried out to obtain the spatial distribution of structural parameters; Iterative forward and inverse modeling is used to verify structural parameters, and four types of causes are compared to determine the root cause of anomalies. The reliability of road service status is output by combining multiple indicators.

8. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 7, characterized in that, The iterative forward and inverse verification structure parameters are used to identify the root causes of anomalies by comparing the matching degree of four types of causes, and the road service status credibility is output by combining multiple indicators, including: Establish a set of candidate anomaly causes: ,in: This indicates structural performance degradation; Indicates abnormal traffic load; This indicates a sudden change in environmental conditions; This indicates a malfunction in the sensing device; For each candidate cause of anomaly, the corresponding parameters in the digital twin model of the road body are adjusted, and multiple forward and inverse iterations are performed. For candidate anomaly reasons Calculate the first Multi-feature residual volume after sub-forward-inverse iteration ; Based on the multi-feature residual volume of multiple iterations Calculate the causes of candidate anomalies residual continuous shrinkage ; In the After each iteration, the final forward residual of each monitoring unit is calculated. ; Based on the spatial topology network of the road body, combined with the above Calculate candidate anomaly causes Corresponding spatial residual fracture ; Candidate anomaly reasons Structural state parameters obtained by inversion Re-input the digital twin model of the road ontology to obtain the re-forward modeled simulation response; Using the re-forward modeled response as the inversion data, structural state parameter inversion is performed again to obtain the loop-loop structural state parameters. According to the above Calculate candidate anomaly causes Forward and inverse loop bias ; Based on the residual continuous shrinkage Spatial residual fracture degree and forward and inverse loop deviation Calculate the causes of candidate anomalies Closed-loop consistency index ; Closed-loop consistency index The largest candidate cause of the anomaly was determined to be the main reason for the current road body anomaly. .

9. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 8, characterized in that, Regarding the candidate anomaly cause Calculate the first Multi-feature residual volume after sub-forward-inverse iteration ,include: in: Candidate anomaly cause In the Multi-feature residual volume after the next iteration The number of types of state features. Candidate anomaly cause The corresponding set of monitoring units, Candidate anomaly cause Next State features obtained from forward modeling. This indicates median operations.

10. The method for dynamic monitoring and analysis of the service status of highway road bodies according to claim 1, characterized in that, The step of constructing a structural evolution chain based on the main causes of anomalies and inversion parameters, predicting anomaly development trends and generating tiered early warnings, and adjusting the perception strategy and updating the dynamic service status benchmark according to the early warning results includes: Store road time-series service data, combine spatial topology analysis to identify the anomaly diffusion pattern, and construct a structural evolution chain based on the anomaly time-series correlation to distinguish different types of road body service status anomalies. Match the abnormal evolution pattern of the road body service status to determine the abnormal type of the road body service status, adjust the twin model parameters to perform multi-condition simulation, and predict the development trend, impact range and critical maintenance time of the abnormal road body service status. By integrating multiple indicators to classify three levels of early warning, coordinating and regulating sensing and data collection strategies, and updating the twin model after maintenance, a complete monitoring and analysis closed loop is formed.

Images