Bridge health monitoring method based on digital twinning

Abstract

The invention discloses a bridge health monitoring method based on digital twinning, and belongs to the field of engineering safety. Comprising the following steps: monitoring real-time vibration data of one or more preset positions on a bridge; real-time dynamic load data and environment parameters of the bridge are obtained, the dynamic load data comprise the weight, the speed and the position of the vehicle, and the environment parameters comprise the temperature, the wind speed and the humidity; inputting the dynamic load data and the environmental parameters into a bridge digital twin model, and outputting predicted vibration data of a preset position; and generating a damage index based on the residual error of the real-time vibration data and the predicted vibration data, and judging whether the bridge has local damage based on the damage index. The method can achieve the precise recognition and positioning of the local damage of the bridge structure.

Description

Technical Field

[0001] This invention relates to the field of engineering safety technology, specifically to a bridge health monitoring method based on digital twins. Background Technology

[0002] Traditional bridge structural health monitoring generally relies on deploying sensors at key locations and setting vibration response thresholds for safety warnings. However, in complex actual operating environments, bridge vibration responses are simultaneously influenced by multiple time-varying factors such as vehicle loads, ambient temperature, and wind loads. Existing monitoring methods often struggle to effectively isolate the impact of environmental and operational load changes on the overall structural response when faced with these complex coupled excitations, resulting in the system's insensitivity to weak abnormal signals caused by localized damage to the actual structure. This limitation renders traditional threshold alarm mechanisms ineffective in identifying and accurately locating early, localized damage, frequently leading to false alarms or missed alarms, failing to meet the high-level requirements of modern bridge preventative maintenance for accurate and reliable damage identification. Therefore, the industry urgently needs an innovative monitoring method that can more intelligently distinguish between load effects and structural damage and achieve precise damage location. Summary of the Invention

[0003] To address the technical problems existing in the background art, this invention proposes a bridge health monitoring method based on digital twins, comprising: S1. Monitor real-time vibration data at one or more preset locations on the bridge; S2. Obtain real-time dynamic load data and environmental parameters of the bridge. The dynamic load data includes the weight, speed and position of the vehicles, and the environmental parameters include temperature, wind speed and humidity. S3. Input the dynamic load data and environmental parameters into the bridge digital twin model and output the predicted vibration data at the preset location. S4. Based on the residual between real-time vibration data and predicted vibration data, a damage index is generated, and the bridge is judged to have suffered local damage based on the damage index.

[0004] This invention, by inputting real-time dynamic loads and environmental parameters into a digital twin model, can isolate the influence of normal operation and environmental changes on vibration response, enabling residual-based damage indicators to more sensitively reflect the true local damage of the structure, thereby significantly improving the accuracy and reliability of monitoring.

[0005] Preferably, the real-time vibration data is monitored by a distributed optical fiber sensor, which is based on the Brillouin scattering principle and transforms a single optical cable into a continuous spatial array of measuring points.

[0006] This invention uses distributed fiber optic sensors to monitor real-time vibration data, overcoming the shortcomings of traditional point sensors, such as limited deployment quantity and potential missed detection of local damage, and greatly improving the spatial resolution and coverage of the monitoring.

[0007] Preferably, S3 specifically includes: S31. Perform graded and quantitative processing on environmental parameters to divide the temperature range into three sub-ranges: material brittleness range, elastic deformation range, and thermal expansion range; divide the humidity range into three sub-ranges: accelerated carbonization range, stable range, and critical corrosion range; and divide the wind speed range into four sub-ranges based on the structural wind vibration effect: imperceptible wind range, aerodynamic micro-vibration range, vortex-induced vibration range, and flutter critical range. S32. Based on historical monitoring data, a dynamic weight rule base is constructed. Using a digital twin model, for the combined working conditions of different temperature sub-intervals, humidity sub-intervals, and wind speed sub-intervals, the modal influence weight of each environmental parameter on the predicted vibration data is calculated by inversion, and a weight mapping table corresponding to the combined working conditions is established. Among them, the wind speed weight in the vortex-induced vibration interval or flutter critical interval is higher than the wind speed weight in the imperceptible wind interval or aerodynamic micro-vibration interval. S33. Input the real-time dynamic load data and the graded and quantized environmental parameters into the bridge digital twin model; call the weight mapping table according to the combined working conditions corresponding to the current environmental parameters; use a weighted superposition algorithm to fuse the environmental parameter inputs and generate predicted vibration data at the preset position.

[0008] This invention can accurately quantify the nonlinear effects of different environmental conditions on bridge structural vibration, making the predictions of the digital twin model closer to reality, thereby more effectively suppressing the interference caused by environmental factors in subsequent residual analysis.

[0009] Preferably, S4 specifically includes: S41. Calculate the residual between the real-time vibration data and the predicted vibration data at the preset position; S42. If there is only one preset location, a damage index for the preset location is generated based on the residual. If the damage index exceeds the first dynamic threshold, it is determined that the bridge has suffered local damage. The first dynamic threshold is determined based on the time series statistical analysis of historical residual data. S43. If there are multiple preset positions, perform the following steps: S431. Using each preset position as a node, generate a residual correlation network based on the spatial correlation between the residuals of the real-time vibration data and the predicted vibration data of each node; S432. Identify abnormal transmission paths through residual correlation networks and eliminate conductive residual anomalies caused by vehicle load transmission or environmental interference. S433. Based on the spatial correlation, local abrupt residuals are screened, and damage indicators are generated by combining the modal shape shift degree. If the damage indicators exceed the second dynamic threshold and form an isolated high-value region in the continuous spatial measurement point array, it is determined that local damage has occurred in the region. The second dynamic threshold is determined based on the statistical distribution of historical residual data in the spatial dimension.

[0010] This invention provides adaptive solutions for different monitoring scales. In particular, for multi-point monitoring, the introduction of spatial correlation analysis is helpful in locating the true local damage.

[0011] Preferably, S432 specifically involves: based on a network clustering algorithm, identifying high-cohesion node clusters in the residual association network whose internal connection density is higher than the average connection density of the network, and determining residual anomalies within the high-cohesion node clusters whose residual amplitude exceeds a preset residual amplitude as conductive anomalies caused by vehicle load transmission or environmental interference, and removing them.

[0012] This invention can intelligently distinguish between spatially correlated conductive anomalies caused by load transfer or environmental disturbances and true local mutations using network clustering algorithms, effectively reducing the risk of misjudgment.

[0013] Preferably, the step of filtering local abrupt change residuals based on the spatial correlation and generating damage indices by combining the modal shape shift degree specifically involves: After eliminating conductive residual anomalies, each node is taken as the target node in turn. The ratio of the residual amplitude of the target node to the average residual amplitude of the K nearest nodes in the residual correlation network is calculated. Nodes with a ratio greater than a set value are selected as local mutation nodes. Modal parameters are identified and the real-time vibration mode shapes are extracted from the real-time vibration data of the local mutation nodes. The modal guarantee criterion values ​​between the real-time vibration mode shapes and the pre-stored reference mode shapes are calculated. Damage indices are generated by weighted fusion based on the residual ratio of the local mutation nodes and the modal guarantee criterion value.

[0014] This invention combines residual amplitude and mode shift to generate a comprehensive damage index with dual evidence, which significantly improves the confidence and anti-interference ability of damage identification.

[0015] Preferably, the vehicle weight is obtained through a dynamic weighing system deployed at the bridgehead or on the bridge deck.

[0016] Preferably, the vehicle speed is obtained through multi-source sensor data fusion, specifically: the instantaneous speed of the vehicle is measured by a radar speed measuring unit deployed on the bridge surface, and the vehicle sequence images are captured by a video monitoring unit deployed on the bridge to analyze the average speed of the vehicle; the instantaneous speed and the average speed are fused and confidence verified to generate the vehicle speed.

[0017] This invention obtains vehicle speed through multi-source data fusion and verification, which can obtain more stable and accurate dynamic speed parameters than a single sensor, thereby accurately characterizing the time-varying characteristics of moving loads.

[0018] Preferably, the vehicle position is obtained by: real-time tracking and calculation of the vehicle's longitudinal and lateral coordinates on the bridge surface using a video monitoring unit or radar unit deployed on the bridge, and determining the vehicle position based on the longitudinal and lateral coordinates.

[0019] Compared to existing technologies, the advantages of this invention are as follows: By using distributed fiber optic sensors to continuously monitor the vibration of the entire bridge, it effectively avoids the problem of traditional single-point sensors potentially missing localized damage. It can also acquire data on the weight, speed, and position of vehicles on the bridge, as well as ambient temperature, humidity, and wind speed. This real-world information is input into the bridge's digital twin simulation model, thereby calculating the expected normal vibration state of the bridge under the current actual load and environment. By comparing the difference between the actual measured vibration and the model's predicted vibration, vibration interference caused by vehicle traffic and environmental factors can be eliminated, and abnormal signals caused solely by localized structural damage can be accurately detected. Especially when multiple monitoring points exist, the spatial correlation of abnormal signals can be analyzed, ultimately pinpointing the true location of localized damage. This invention provides a more timely and reliable basis for bridge safety management decisions. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, this embodiment provides a bridge health monitoring method based on digital twins, including: S1. Monitor real-time vibration data at one or more preset locations on the bridge. Specifically: On both sides of the main bridge deck of the bridge to be monitored, distributed sensing optical cables based on the Brillouin scattering principle are laid longitudinally along the direction of the maintenance road. Special attention is paid to the area near the anchor base of the cable stay, and the optical cables are laid straight during the installation to form a monitoring array with continuous spatial measurement point function. The spacing between the measurement points is set to 0.5m.

[0023] This invention employs a multi-layered, multi-material fixing method to ensure the stability of optical cables under the complex environment of strong vibration, strong winds, and humidity on bridges, thereby improving the accuracy and reliability of monitoring data. First, the inspection walkway surface is thoroughly cleaned to remove debris and dust, creating favorable conditions for the subsequent bonding of fixing materials. Next, the optical cable is secured every 1 meter using snap-fit ​​steel nails, ensuring the cable line remains straight and stable. The high strength of the snap-fit ​​steel nails allows them to withstand significant tensile force, ensuring initial fixation of the cable to the bridge deck. Building upon this, for uneven areas, silicone sealant is used for reinforcement, ensuring the monitoring optical cable adheres tightly to the ground. This effectively reduces cable loosening caused by uneven ground, enhancing cable stability. The high adhesion and elasticity of the silicone sealant tightly binds the cable to the bridge structure, forming a unified whole and improving the cable's vibration and impact resistance. Finally, cloth tape is used to level the cable, ensuring a smooth contact with the bridge deck. The waterproof properties of the cloth tape prevent moisture penetration, protecting the cable from the effects of the humid environment and extending its service life.

[0024] In the approach bridge section, the "dark optical fiber" in the existing communication pipeline of the bridge is used to connect with the distributed sensing optical cable of the main bridge section through optical fiber fusion splicing technology, thereby forming a continuous monitoring loop from the approach bridge to the main bridge.

[0025] In this embodiment, one or more measuring points on the distributed sensing optical cable deployed on the main bridge section of the bridge to be monitored are selected as preset locations for monitoring to obtain real-time vibration data of one or more preset locations on the bridge to be monitored.

[0026] S2. Obtain real-time dynamic load data and environmental parameters of the bridge to be monitored. The dynamic load data includes vehicle weight, speed, and location, while the environmental parameters include temperature, wind speed, and humidity. Specifically: By setting a precision of at the bridge entrance or on the bridge deck The dynamic weighing system obtains the vehicle weight.

[0027] By deploying it on the bridge deck with an error less than The radar speed measurement unit measures the instantaneous speed of the vehicle at a sampling frequency of 100Hz, and the average speed of the vehicle is analyzed by capturing a sequence of vehicle images at 30fps through a video monitoring unit deployed on the bridge.

[0028] The instantaneous speed and average speed are fused and confidence levels are verified to generate the vehicle speed, specifically as follows: Establish a unified time-stamping system to time-align the data from the radar velocity measurement unit and the video monitoring unit, with synchronization accuracy controlled within 1 millisecond.

[0029] The raw instantaneous velocity data acquired by radar is processed using a moving average filter with a window length of 0.5 seconds to eliminate random signal fluctuations. For video sequence images, vehicle outlines are first identified using a deep learning-based target detection algorithm, and then the vehicle displacement between consecutive frames is calculated using an improved optical flow method. The average velocity sequence is then calculated based on the frame rate.

[0030] A time-velocity correspondence matrix is ​​then constructed, and the instantaneous radar velocity sequence and the average velocity sequence from video analysis are registered on the same time baseline. The dynamic correlation coefficient between the two sets of data within a 2-second time window is calculated. When the correlation coefficient consistently reaches 0.85 or higher, an adaptive weighted fusion algorithm is adopted: radar data is given a higher weight on straight road segments, for example, a weight of 0.7 for radar data and 0.3 for video data; while the weight of video data is appropriately increased in curves and lane-changing areas, with a weight of 0.6 for radar data and 0.4 for video data, achieving complementary advantages.

[0031] Then, a confidence verification mechanism is initiated, establishing a three-level verification system. Specifically: Level 1 verification: Real-time comparison of the relative differences between radar and video speed values. When the difference exceeds a set threshold, such as 3 m / s, an early warning is triggered. Level 2 verification: Combined with the vehicle dynamics model, it is determined whether the acceleration change is within a reasonable range, and physically infeasible data is filtered out. Level 3 verification: Historical speed data of the same vehicle type on the same road segment is introduced as a reference benchmark to cross-validate abnormal fluctuation data.

[0032] For data segments that fail verification, a repair procedure is initiated: a Kalman filter algorithm based on time series prediction is used to smoothly interpolate missing data, while consecutive outliers are marked. When data from a single sensor is determined to be abnormal for 5 consecutive cycles, the system automatically switches to the backup sensor-dominated mode to ensure the continuity of data output.

[0033] The final output is the vehicle speed data after fusion and verification. This data is used as a reliable parameter input to the digital twin model after quality verification, providing an accurate basis for predicting the dynamic response of the bridge.

[0034] By deploying video monitoring units or radar units on the bridge, the longitudinal and lateral coordinates of vehicles on the bridge surface are tracked and calculated in real time. The longitudinal and lateral coordinates are combined to generate vehicle coordinates as the vehicle position.

[0035] This embodiment collects real-time data by deploying meteorological monitoring stations at key locations on the bridge: temperature data (measurement range: -20℃ to +50℃, accuracy: ±0.5℃); humidity data (measurement range: 0%RH to 100%RH, accuracy: ±3%RH); and wind speed data (measurement range: 0m / s to 30m / s, accuracy: ±0.5m / s).

[0036] S3. Input the dynamic load data and environmental parameters into the bridge's digital twin model, and output the predicted vibration data at the preset location. Specifically: S31. Perform graded and quantified processing on environmental parameters, dividing the temperature range into three sub-ranges: a material brittleness range of -20℃ to 0℃, an elastic deformation range of 0℃ to 35℃, and a thermal expansion range of 35℃ to 50℃; dividing the humidity range into three sub-ranges: an accelerated carbonization range of 0% to 60%RH, a stable range of 60%RH to 80%RH, and a critical corrosion range of 80% to 100%RH; and dividing the wind speed range into four sub-ranges based on the structural wind vibration effect: an imperceptible wind range of 0m / s to 5m / s, an aerodynamic micro-vibration range of 5m / s to 10m / s, a vortex-induced vibration range of 10m / s to 15m / s, and a flutter critical range of 15m / s to 25m / s. S32. Based on historical monitoring data, a dynamic weighted rule base is constructed. Using a digital twin model, for the combined working conditions of different temperature sub-intervals, humidity sub-intervals, and wind speed sub-intervals, the modal influence weights of each environmental parameter on the predicted vibration data are calculated by inversion. A weight mapping table corresponding to the combined working conditions is established. The modal influence weights characterize the contribution of each environmental parameter to the bridge vibration response under specific combined working conditions, and the sum of all weights is 1. Specifically: S321. Extract from the bridge historical monitoring database the continuous monitoring values ​​of temperature, humidity, and wind speed for at least one complete year, as well as the acceleration, displacement, or strain response of key bridge locations during the corresponding time period. S322. Classify the historical environmental parameter data according to the temperature sub-interval, humidity sub-interval and wind speed sub-interval divided in S31 to form multiple environmental parameter combination conditions. Each combination condition corresponds to a specific temperature range, humidity range and wind speed range. S323. For each combination of environmental parameters, use a digital twin model to perform parameter inversion analysis: with the vehicle load input fixed, change the values ​​of each environmental parameter individually in turn, fit the model using the least squares method to the difference between the predicted vibration data and the actual monitored vibration data, and invert and calculate the modal influence weight of each environmental parameter on the vibration response under this combination of operating conditions. S324. Establish a combined operating condition-weight mapping table to associate and store the combined operating condition of each environmental parameter with the modal influence weights of its corresponding environmental parameters. Among them, the wind speed weight in the vortex-induced vibration range or flutter critical range is higher than the wind speed weight in the imperceptible wind range or aerodynamic micro-vibration range.

[0037] The combined working condition-weight mapping table in this embodiment is partially shown below:

[0038] S325. Set up a weight rule base update mechanism. When the amount of newly added monitoring data reaches the set threshold, the weight value will be automatically recalculated and the mapping table will be updated to ensure that the weight rule base adapts to changes in the bridge status.

[0039] S33. Input the real-time dynamic load data and the graded and quantified environmental parameters into the bridge digital twin model; call the weight mapping table according to the combination of working conditions corresponding to the current environmental parameters. For example, when the combination of environmental parameters is 25°C (temperature in elastic deformation range), 65%RH (humidity in stable range), and 12m / s (wind speed in vortex-induced vibration range), the weight mapping table is called to obtain: temperature weight 0.15, humidity weight 0.05, and wind speed weight 0.8; use a weighted superposition algorithm to fuse the environmental parameter inputs to generate predicted vibration data at the preset position.

[0040] S4. Based on the residuals between real-time vibration data and predicted vibration data, a damage index is generated, and the bridge is judged to have suffered localized damage based on the damage index. Specifically: S41. Calculate the residual between the real-time vibration data and the predicted vibration data at the preset position; S42. If there is only one preset location, a damage index for the preset location is generated based on the residual. If the damage index exceeds a first dynamic threshold, it is determined that the bridge has suffered local damage. The first dynamic threshold is determined based on time-series statistical analysis of historical residual data. In this embodiment, the mean of the residual data for the preset location over the past 30 working days is added to 2.5 to 3.5 times the standard deviation.

[0041] S43. If there are multiple preset positions, perform the following steps: S431. Using each preset position as a node, generate a residual correlation network based on the spatial correlation between the residuals of the real-time vibration data and the predicted vibration data of each node.

[0042] S432. Identify abnormal transmission paths through residual correlation networks and eliminate conductive residual anomalies caused by vehicle load transfer or environmental interference. Specifically: Based on network clustering algorithms, such as Leiden et al., high-cohesion node clusters with internal connection density higher than the network average connection density are identified in the residual association network. Residual anomalies within high-cohesion node clusters whose residual amplitude exceeds a preset residual amplitude are determined to be conduction anomalies caused by vehicle load transmission or environmental interference and are removed. In this embodiment, the preset residual amplitude is the average historical residual amplitude plus 1.5 to 2.5 times the standard deviation.

[0043] S433. Based on the spatial correlation, local mutation residuals are screened, and damage indices are generated by combining the modal shape shift degree. Specifically, after eliminating conductive residual anomalies, each node is sequentially taken as a target node, and the ratio of the residual amplitude of the target node to the average residual amplitude of the K nearest nodes in the residual correlation network is calculated. Nodes with a ratio greater than a set value are screened as local mutation nodes. The value of K ranges from 3 to 8, and is 5 in this embodiment; the value of the set value ranges from 1.8 to 2.2, and is 2 in this embodiment.

[0044] Modal parameter identification is performed on the real-time vibration data of the local mutation node to extract its real-time vibration mode shape. Specifically, the random subspace identification method is used for modal parameter identification. In this embodiment, the following steps are taken: Acceleration response data of a 3×3 measuring point array consisting of the local mutation node and its eight adjacent nodes are collected, with a sampling frequency of 200Hz; a Hankel matrix is ​​constructed, with the number of rows in the matrix block being twice the number of measuring points and the number of columns being no less than 500; the system matrix is ​​calculated through QR decomposition and singular value decomposition, with the singular value truncation threshold set to 1% of the maximum singular value; eigenvalues ​​and eigenvectors are extracted from the system matrix to obtain the first eight modal frequencies, damping, and mode shapes of the structure; false modes are eliminated using the modal confidence criterion, retaining the true modes with a modal confidence value greater than 0.8. Subsequently, for each retained true mode, its corresponding eigenvector is normalized to its maximum value, and the resulting normalized vector is the real-time vibration mode shape of that mode.

[0045] Then, the modal guarantee criterion value between the real-time vibration mode shape and the pre-stored reference mode shape is calculated. The pre-stored reference mode shape is obtained under the known health condition of the bridge through standardized field testing and modal identification process, and after verification by a high-fidelity physical model.

[0046] The formula for calculating the modal guarantee criterion value is: ,in This is the real-time vibration mode shape vector. This is the reference mode shape vector.

[0047] The modal guarantee criterion values ​​were calculated for the first 8 identified modes, with a focus on analyzing the changes in the modal guarantee criterion values ​​for the 3rd to 6th main vibration modes. When the modal guarantee criterion value of a certain mode is lower than 0.85, it is determined that the node has a significant mode shift in that mode.

[0048] Based on the residual ratio of the local mutation nodes and the modal guarantee criterion value, a damage index is generated by weighted fusion. In this embodiment, the residual ratio weight is 0.5 to 0.7, and the modal guarantee criterion weight is 0.3 to 0.5.

[0049] If the damage index exceeds the second dynamic threshold and forms an isolated high-value region in the continuous spatial measurement point array, it is determined that the region has local damage. The second dynamic threshold is determined based on the statistical distribution of historical residual data in the spatial dimension. In this embodiment, the second dynamic threshold is the 90% to 95% quantile of all historical values ​​of damage index of all measurement points.

[0050] An isolated high-value region must simultaneously meet the following three characteristics: 1. The damage index value significantly exceeds the second dynamic threshold, exhibiting a statistically abnormally high value. In this embodiment, the damage index value is more than 1.2 times the second dynamic threshold; 2. The high-value region is surrounded by nodes with relatively low damage index values, forming a significant numerical gradient change with the surrounding area, and is not connected to other high-damage index regions. In this embodiment, the average damage index value of all adjacent nodes centered on the target node, and nodes with a Manhattan distance of less than or equal to 2, does not exceed 0.7 times the damage index value of the target node, and the slope of the spatial gradient of the damage index between the target node and any adjacent node is greater than 1; 3. The high-value region has a limited spatial distribution range. In this embodiment, it is concentrated within a 3×3 measuring point array centered on the target node, and there are no other nodes in this array whose damage index values ​​exceed the second dynamic threshold.

[0051] This invention achieves all-weather, high-resolution acquisition of bridge vibration response and vehicle loads through innovative distributed fiber optic deployment technology and multi-source data fusion. Utilizing a digital twin model based on dynamic weight mapping of environmental parameters, it effectively isolates complex interferences caused by normal operation and environmental changes, highlighting subtle damage signals. Finally, by combining residual space network analysis and modal shape migration as dual criteria, it achieves early detection and precise location of localized damage. This not only significantly improves the sensitivity and accuracy of damage identification but also ensures long-term operational stability through rigorous benchmark establishment and adaptive update mechanisms, providing strong technical support for preventative maintenance and safe operation of bridge structures.

[0052] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A bridge health monitoring method based on digital twinning, characterized in that, The method comprises the following steps: S1, monitoring real-time vibration data of one or more preset positions on the bridge; S2, acquiring real-time dynamic load data and environmental parameters of the bridge, wherein the dynamic load data comprises the weight, speed and position of the vehicle, and the environmental parameters comprise temperature, wind speed and humidity; S3, inputting the dynamic load data and the environmental parameters into a bridge digital twin model to output predicted vibration data of the preset positions; S4, generating a damage index based on the residual error between the real-time vibration data and the predicted vibration data, and determining whether the bridge has local damage based on the damage index.

2. The bridge health monitoring method based on digital twinning according to claim 1, characterized in that, The real-time vibration data is monitored by a distributed optical fiber sensor, which converts a single optical cable into a continuous spatial measurement point array based on the principle of Brillouin scattering.

3. The bridge health monitoring method based on digital twinning according to claim 1 or 2, characterized in that, S3 specifically comprises: S31, performing hierarchical quantization processing on the environmental parameters to divide the temperature interval into three temperature subintervals of material embrittlement interval, elastic deformation interval and thermal expansion interval; divide the humidity interval into three humidity subintervals of accelerated carbonization interval, stable interval and corrosion critical interval; and divide the wind speed interval into four wind speed subintervals of non-sensitive wind interval, aerodynamic micro-vibration interval, vortex-induced vibration interval and flutter critical interval according to the structural wind vibration effect; S32, constructing a dynamic weight rule library based on historical monitoring data, and using the digital twin model to inversely calculate the modal influence weight of each environmental parameter on the predicted vibration data for different combinations of temperature subintervals, humidity subintervals and wind speed subintervals, and establishing a weight mapping table corresponding to the combination working conditions; wherein the wind speed weight in the vortex-induced vibration interval or the flutter critical interval is higher than that in the non-sensitive wind interval or the aerodynamic micro-vibration interval; S33, inputting the real-time dynamic load data and the hierarchical quantization processed environmental parameters into the bridge digital twin model; calling the weight mapping table according to the combination working condition corresponding to the current environmental parameters; and using a weighted superposition algorithm to fuse the environmental parameter input to generate the predicted vibration data of the preset positions.

4. The bridge health monitoring method based on digital twinning according to any one of claims 1 to 3, characterized in that, S4 specifically comprises: S41, calculating the residual error between the real-time vibration data and the predicted vibration data of the preset positions; S42, if there is only one preset position, generating a damage index for the preset position based on the residual error, and if the damage index exceeds a first dynamic threshold, determining that the bridge has local damage, wherein the first dynamic threshold is determined based on time series statistical analysis of historical residual error data; S43, if there are multiple preset positions, the following steps are performed: S431, taking each preset position as a node, generating a residual error correlation network based on the spatial correlation of the residual error between the real-time vibration data and the predicted vibration data of each node; S432, identifying an abnormal conduction path through the residual error correlation network, and eliminating conduction residual error caused by vehicle load transmission or environmental interference; S433, screening local mutation residual error according to the spatial correlation, generating a damage index combined with the modal shape offset degree, and if the damage index exceeds a second dynamic threshold and forms an isolated high value area in the continuous spatial measurement point array, determining that the area has local damage, wherein the second dynamic threshold is determined based on the statistical distribution of historical residual error data in the spatial dimension.

5. The bridge health monitoring method based on digital twinning of claim 4, wherein, The S432 specifically comprises: based on a network clustering algorithm, identifying a high-cohesion node community with a higher internal connection density than an average connection density of the network in the residual correlation network, and determining a residual anomaly with a residual amplitude exceeding a preset residual amplitude in the high-cohesion node community as a conductive anomaly caused by vehicle load transmission or environmental interference and removing the conductive anomaly.

6. The bridge health monitoring method based on digital twinning of claim 4, wherein, The screening of the local mutation residual according to the spatial correlation, in combination with the modal shape offset degree, generates an injury index, which specifically comprises: After removing the conductive residual anomaly, each node is sequentially taken as a target node, the ratio of the residual amplitude of the target node to the average residual amplitude of the K nearest nodes in the residual correlation network is calculated, and a node with a ratio greater than a set value is screened as a local mutation node; The real-time vibration data of the local mutation node is subjected to modal parameter identification to extract a real-time vibration modal shape, and a modal assurance criterion value between the real-time vibration modal shape and a pre-stored reference modal shape is calculated. Based on the residual ratio and the modal assurance criterion value of the local mutation node, an injury index is generated through weighted fusion.

7. The bridge health monitoring method based on digital twinning of claim 1, wherein, The vehicle weight is obtained through a dynamic weighing system arranged on a bridge head or a bridge surface.

8. The bridge health monitoring method based on digital twinning of claim 1, wherein, The vehicle speed is obtained through multi-source sensor data fusion, specifically: the instantaneous speed of the vehicle is measured through a radar speed measurement unit arranged on the bridge surface, and the average speed of the vehicle is analyzed through a video monitoring unit deployed on the bridge to capture vehicle sequence images; the instantaneous speed and the average speed are subjected to data fusion and confidence checking to generate the vehicle speed.

9. The bridge health monitoring method based on digital twinning of claim 1, wherein, The vehicle position is obtained by: tracking and solving the longitudinal and lateral coordinates of the vehicle on the bridge surface in real time through a video monitoring unit or a radar unit deployed on the bridge, and determining the vehicle position based on the longitudinal and lateral coordinates.

Images