An AI-based method and system for evaluating the health status of a highway bridge and tunnel structure

Abstract

The application provides an AI-based highway bridge and tunnel structure health state evaluation method and system, which comprises the following steps: according to a continuous damage point position sequence, a clustering algorithm is used to identify a turning point and divide a life cycle section, and the boundaries of a stable section, a degradation section and an acceleration section are determined; the extension rate data in each life cycle section is obtained, the data between sections is connected through a linear interpolation method for the curve jump position, and a smooth evolution curve is obtained; if there is a symmetric filling residual error in the evolution curve, the error source is judged by comparing the difference of the extension rate of adjacent sections, and the corrected curve parameters are obtained; crack extension characteristics are extracted from the corrected curve parameters, a time series analysis model is used to predict the evolution trend, and the whole process mapping of the crack from initiation to extension is determined; according to the whole process mapping of the crack, it is judged whether the extension rate is close to a preset critical threshold, and a threshold point of the remaining life of the structure is obtained.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to an AI-based method and system for assessing the health status of highway bridge and tunnel structures. Background Technology

[0002] Bridge and tunnel structural health monitoring is an important research area for ensuring highway traffic safety and the long-term service of infrastructure. Its core lies in predicting the degradation trend of structures through real-time data analysis, timely identifying potential risks, and thus avoiding major accidents.

[0003] The importance of this field is self-evident, especially given that bridges and tunnels are vital transportation arteries; any minor damage could trigger a chain reaction, affecting public safety and economic operations.

[0004] Currently, although many monitoring methods can collect a large amount of structural status data, these methods often struggle to integrate scattered data points into a coherent trend, especially at different stages of structural damage from the initial stage to accelerated deterioration, where the continuity and consistency of the data are often disrupted.

[0005] Existing technologies focus more on the static state at a certain moment when processing data, ignoring the dynamic changes in time and space during the damage evolution process. This leads to biases in the judgment of the structural health status and makes it difficult to accurately reflect the true development path of the damage.

[0006] Against this backdrop, a core technical challenge in bridge and tunnel structural health monitoring is how to handle the data correlation issues caused by the slight spatial drift of damage points.

[0007] Spatial position drift refers to the slight shift in the actual location of the damage point during monitoring due to structural deformation or external environmental influences. Although this shift is subtle, it can disrupt the continuity of data recording and thus affect the accurate calculation of the damage development speed.

[0008] Especially when the structure transitions from a stable state to a stage of accelerated degradation, this drift accumulates, making it difficult to seamlessly connect data from different time periods, resulting in analytical gaps.

[0009] Specifically, in actual business scenarios, the monitoring system assigns a fixed identifier to each crack or damage point to track its changes. However, when the location drifts beyond a certain range, the system may not be able to correctly identify that it is the same damage point.

[0010] For example, in the monitoring of cracks in a tunnel sidewall, if the location of the crack moves slightly due to foundation settlement, the data recorded by the system may be misjudged as a new damage point, causing the originally continuous crack propagation record to be interrupted, and the calculated propagation rate to have a significant error.

[0011] Therefore, ensuring the continuity of damage point identification and the integrity of data recording under conditions of spatial displacement has become a critical issue that urgently needs to be addressed in the health assessment of bridge and tunnel structures. Summary of the Invention

[0012] This invention provides an AI-based method for assessing the health status of highway bridge and tunnel structures, mainly including: When the coordinate drift threshold exceeds the preset value by scanning the timestamp sequence, the Kalman filter algorithm is used to calibrate the spatial coordinates in real time to obtain a continuous sequence of damage point locations. Based on the continuous sequence of damage point locations, a clustering algorithm is used to identify turning points and divide the life cycle into segments, determining the boundaries of the stable segment, the degradation segment, and the acceleration segment. Acquire expansion rate data within each life cycle segment, and connect the data between segments using linear interpolation at curve transition points to obtain a smooth evolution curve; If there is a symmetrical filling residual error in the evolution curve, the source of the error can be determined by comparing the difference in the expansion rate of adjacent segments, and the corrected curve parameters can be obtained. Crack propagation characteristics are extracted from the corrected curve parameters, and the evolution trend is predicted using a time series analysis model to determine the entire process mapping from crack initiation to propagation. Based on the mapping of the entire crack process, it is determined whether the propagation rate is close to the preset critical threshold, and the threshold point of the remaining life of the structure is obtained.

[0013] This invention provides an AI-based health status assessment system for highway bridge and tunnel structures, mainly comprising: The coordinate drift calibration module is used to perform real-time calibration of spatial coordinates by using a Kalman filter algorithm when the coordinate drift threshold exceeds a preset value by detecting the scan timestamp sequence, so as to obtain a continuous sequence of damage point positions. The lifecycle segmentation module is used to identify turning points and divide the lifecycle segments based on a continuous sequence of damage point locations using a clustering algorithm, and to determine the boundaries of the stable segment, the degradation segment, and the acceleration segment. The evolution curve smoothing module is used to acquire expansion rate data within each life cycle segment, and connects the data between segments using a linear interpolation method at the curve jump points to obtain a smooth evolution curve. The residual error correction module is used to determine the source of error by comparing the difference in the expansion rate of adjacent segments if there is a symmetrical filling residual error in the evolution curve, and to obtain the corrected curve parameters. The crack feature extraction module is used to extract crack propagation features from the corrected curve parameters, use a time series analysis model to predict the evolution trend, and determine the entire process mapping from crack initiation to propagation. The remaining life assessment module is used to determine whether the propagation rate is close to a preset critical threshold based on the crack's entire process mapping, and to obtain the threshold point for the structure's remaining life.

[0014] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses an AI-based method for assessing the health status of highway bridge and tunnel structures. Addressing the health degradation caused by crack propagation during long-term use, it proposes a complete monitoring and prediction solution. The invention detects coordinate drift by scanning timestamp sequences, uses a Kalman filter algorithm to calibrate spatial coordinates in real time, obtains a continuous sequence of damage point locations, and combines this with a clustering algorithm to identify inflection points, divide the life cycle, and determine the boundaries of stable, deteriorating, and accelerating segments. Subsequently, linear interpolation smooths curve transitions, corrects residual errors from symmetrical filling, extracts crack propagation characteristics, and uses a time-series analysis model to predict evolution trends, constructing a full-process mapping of cracks from initiation to propagation. Finally, it determines whether the propagation rate approaches a critical threshold, accurately determining the remaining lifespan threshold of the structure. The core innovation of this invention lies in the seamless integration of real-time calibration, life cycle division, and trend prediction, enabling dynamic assessment and early warning of structural health status, providing a scientific basis for the safe maintenance of highway bridges and tunnels. Attached Figure Description

[0015] Figure 1 This is a flowchart of an AI-based method for assessing the health status of highway bridge and tunnel structures according to the present invention.

[0016] Figure 2 This is a schematic diagram of an AI-based method for assessing the structural health of highway bridges and tunnels according to the present invention.

[0017] Figure 3 This is another schematic diagram of an AI-based method for assessing the structural health of highway bridges and tunnels according to the present invention.

[0018] Figure 4 This is a schematic diagram of an AI-based highway bridge and tunnel structural health status assessment system according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figures 1-4 This embodiment of an AI-based method and system for assessing the health status of highway bridge and tunnel structures may specifically include: S101. When the coordinate drift threshold detected by scanning the timestamp sequence exceeds the preset value, the Kalman filter algorithm is used to calibrate the spatial coordinates in real time to obtain a continuous sequence of damage point positions.

[0021] The process involves obtaining the original spatial coordinates and their corresponding timestamp sequences, calculating the Euclidean distance between the original spatial coordinates of adjacent timestamp sequences to obtain the coordinate drift, extracting the corresponding original spatial coordinates to determine the coordinates to be calibrated, constructing a state vector based on the coordinates to be calibrated, establishing an observation matrix based on the state vector, iteratively updating the observation matrix using a Kalman filter algorithm to obtain the target spatial coordinates, and concatenating the target spatial coordinates according to the chronological order of the timestamp sequences to obtain a continuous sequence of damage point locations.

[0022] When dealing with spatial coordinate drift, the system first checks whether the coordinate drift exceeds a preset threshold by scanning the timestamp sequence. For example, if the threshold is set to 0.5 meters, and the coordinate data corresponding to the collected timestamp sequence are (1.2,2.3,3.1), (1.3,2.4,3.2), and (1.8,2.9,3.7), the Euclidean distance between adjacent coordinates is calculated. It is found that the distance between the third point and the second point is 0.707 meters, which exceeds the threshold of 0.5 meters. The system automatically triggers the drift detection mechanism, records the timestamp as an anomaly, and analyzes that the deviation direction is positive offset in the x-axis and y-axis. The offset is calculated by the algorithm as (0.5,0.5,0.5). Subsequently, in response to the detected drift, the system automatically invoked the Kalman filter algorithm for real-time calibration. The filter's state vector was initialized to the current coordinates (1.8, 2.9, 3.7), the process noise covariance matrix Q was set to an identity matrix of 0.01, and the measurement noise covariance matrix R was set to 0.1. In the prediction phase, the next state was estimated based on a uniform motion model. In the update phase, the state was adjusted in conjunction with the sensor measurements (1.7, 2.8, 3.6), resulting in calibrated coordinates of (1.75, 2.85, 3.65). Analysis showed that the coordinate deviation was reduced to within 0.05 meters after calibration. Finally, the system stores the calibrated coordinate sequence as a continuous sequence of damage point locations, for example, the output sequence is (1.2,2.3,3.1), (1.3,2.4,3.2), (1.75,2.85,3.65). By associating the damage points with the equipment operating status through timestamps, the system analyzes that the drift may be related to the equipment vibration frequency of 10Hz. The system automatically generates a report, suggesting that the equipment stability can be optimized to reduce the drift. Logically, this forms a complete chain from detection to calibration to correlation analysis, ensuring that the coordinate data accuracy is improved to 0.05 meters, meeting the requirements of real-time monitoring.

[0023] S102. Based on the continuous sequence of damage point locations, a clustering algorithm is used to identify turning points and divide the life cycle into segments, determining the boundaries of the stable segment, the degradation segment, and the acceleration segment.

[0024] A continuous sequence of damage point locations is obtained, and a multidimensional feature vector matrix containing tangential slope and curvature is constructed based on the damage point location sequence. The multidimensional feature vector matrix is ​​processed using a clustering algorithm to obtain a set of clusters, and the intersection time of adjacent sets of clusters is identified to determine the turning point. The damage point location sequence is segmented according to the turning point to obtain evolutionary sub-segments, and the average damage propagation rate of the evolutionary sub-segments is calculated. The average damage propagation rate is compared with a preset threshold to determine the boundaries of stable segments, degenerative segments, and acceleration segments.

[0025] Based on a continuous sequence of damage point locations, the system employs a clustering algorithm to identify inflection points and divide the lifecycle into segments, determining the boundaries of stable, deteriorating, and accelerated segments, forming a complete logical chain from data processing to stage division. First, the system reads the calibrated damage point location sequence, such as (1.2,2.3,3.1), (1.25,2.35,3.15), (1.4,2.5,3.3), (1.6,2.7,3.5), and (1.9,3.0,3.8), calculates the rate of change between adjacent points, obtaining a rate of change sequence of 0.05, 0.21, 0.28, and 0.46. Setting a rate of change threshold of 0.3, the system automatically detects that the rate of change at the fourth point (0.28) is close to the threshold, and the rate of change at the fifth point (0.46) significantly exceeds the threshold, marking the fifth point as a potential inflection point. Next, the system invoked the K-means clustering algorithm to divide the location sequence into three clusters, setting the initial cluster centers as (1.2, 2.3, 3.1), (1.4, 2.5, 3.3), and (1.9, 3.0, 3.8). After iterative calculation, cluster 1 was found to contain the first two points, cluster 2 to contain the third and fourth points, and cluster 3 to contain the fifth point. Analyzing the distances between clusters, the system found that the distance from cluster 1 to cluster 2 was 0.22, and the distance from cluster 2 to cluster 3 was 0.42. The system determined that there was a significant jump between cluster 2 and cluster 3, confirming that the turning point was located at the fifth point. Subsequently, the system divided the lifecycle segments based on the rate of change and the clustering results. Cluster 1 was defined as the stable segment, with a rate of change below 0.1, indicating that the damage point location remained almost unchanged; cluster 2 was defined as the degenerative segment, with a rate of change between 0.2 and 0.3, indicating that the damage point location shifted slowly; and cluster 3 was defined as the accelerating segment, with a rate of change exceeding 0.4, indicating that the damage point location changed rapidly. The boundaries were set between the second and third points and between the fourth and fifth points, respectively. Finally, the system correlates the segmentation results with the equipment operating parameters. The analysis reveals that the stable segment corresponds to low equipment load operation, the degradation segment corresponds to a load increase to 50%, and the acceleration segment corresponds to a load exceeding 80%. A segmentation report is generated to alert users to the high-risk status of the acceleration segment. Logically, from location sequence to cluster analysis to life cycle segmentation, a closed loop is formed to ensure the accuracy of damage point trend judgment.

[0026] S103. Obtain the expansion rate data within each life cycle segment, and connect the data between segments using a linear interpolation method at the curve transition points to obtain a smooth evolution curve.

[0027] The expansion rate sequence within each lifecycle segment is obtained, and the rate difference at the boundary between adjacent lifecycle segments is calculated to determine the jump position. A set of boundary breakpoints is extracted based on the jump position, and the time span of adjacent breakpoints in the set of boundary breakpoints is calculated to determine the interpolation interval. A linear interpolation algorithm is used to process the inter-segment data for the interpolation interval to obtain a smoothed reconstructed rate sequence. The discrete nodes of each lifecycle segment are connected through the reconstructed rate sequence to obtain a smooth evolution curve.

[0028] The system first extracts the expansion rate data for each life cycle segment, and calculates the expansion rate of the damage points in each segment based on the divided stable segment, degradation segment and acceleration segment.

[0029] For example, within the stable segment, the positional change of two points is 0.05 with a time interval of 2 hours, and the system calculates the expansion rate to be approximately 0.025 units / hour; within the degenerate segment, the positional change of two points is 0.2 with a time interval of 1.5 hours, and the expansion rate is approximately 0.133 units / hour; within the accelerated segment, the positional change is 0.3 with a time interval of 0.5 hours, and the expansion rate is approximately 0.6 units / hour. Next, the system stores these rate data as a time series, forming a rate distribution map for each segment. Analysis reveals that the rate fluctuation in the stable segment is less than 0.01, the rate in the degenerate segment gradually increases to 0.15, and the rate in the accelerated segment increases sharply to over 0.6, logically reflecting the characteristics of different stages of damage evolution. Subsequently, for the curve abrupt changes, i.e., the data discontinuities at the boundary between the degenerate and accelerated segments, the system uses linear interpolation for smoothing.

[0030] For example, between the fourth point (positions 1.6, 2.7, 3.5) and the fifth point (positions 1.9, 3.0, 3.8), the position jumps to 0.3. The system calculates the interpolation point position as (1.75, 2.85, 3.65) with a time interval of 0.5 hours, generating a smooth transition curve. Analyzing the slope change of the curve before and after interpolation, the process from 0.28 to 0.46 is smoother, reducing the visual impact of data abrupt changes. Finally, the system correlates the smoothed evolution curve with the equipment temperature parameters, finding that the temperature at the jump position rises sharply from 60 degrees Celsius to 85 degrees Celsius, inferring that thermal stress may be the cause of accelerated expansion. Logically, from rate extraction to curve smoothing to parameter correlation, a complete data processing chain is formed, ensuring the visualization and reliability of the evolution trend.

[0031] S104. If there is a symmetrical filling residual error in the evolution curve, the source of the error is determined by comparing the difference in the expansion rate of adjacent segments, and the corrected curve parameters are obtained.

[0032] An abnormal interval corresponding to the symmetric filling residual error in the evolutionary trajectory is obtained, the abnormal interval containing the expansion rate of adjacent segments; the absolute value of the difference between the expansion rates of adjacent segments is calculated, and the absolute value of the difference is compared with a preset fluctuation threshold to determine the distribution range of the symmetric filling residual error; if the numerical offset within the distribution range exceeds a preset range, it is determined that the symmetric filling residual error originates from data alignment deviation; the evolutionary trajectory is reconstructed based on the data alignment deviation to obtain a reconstructed sequence, and numerical correction is performed through the reconstructed sequence to obtain the corrected curve parameters.

[0033] When processing evolution curves, if the system detects residual errors from symmetrical filling, it first automatically identifies the error location and analyzes its source using an algorithm, then corrects it to optimize curve parameters. In specific implementation, the system scans the smoothed curve data and detects a symmetrical filling error at the boundary between the end of the degradation segment and the beginning of the acceleration segment, with an error value of 0.12 units, representing a symmetrical deviation in the rates of the two segments. By comparing the expansion rate differences between adjacent segments, the system calculates that the average rate at the end of the degradation segment is 0.18 units / hour, while the average rate at the beginning of the acceleration segment is 0.42 units / hour, resulting in a rate difference of 0.24 units / hour, exceeding the preset threshold of 0.1. The system determines that the error originates from insufficient consideration of the rate gradient changes between segments during data filling. Next, the system uses a weighted average algorithm for correction. Based on the time series, the rates of three data points before and after the error location are weighted, with weighting coefficients decreasing over time: 0.3, 0.25, 0.2, 0.15, 0.05, and 0.05. The corrected rate value is calculated to be 0.29 units / hour, reducing the error to 0.03 units, meeting the accuracy requirements. Analysis of the curves before and after correction reveals a more natural rate transition, with the slope changing smoothly from 0.35 to 0.38, avoiding interference from abrupt changes in trend judgment. To further verify the correction effect, the system correlates the corrected curve parameters with the equipment vibration frequency data. It finds that the rate change is consistent with the trend of the vibration frequency increasing from 5.2 Hz to 7.8 Hz, suggesting that vibration may be one of the causes of the error. Logically, from error identification to rate correction to parameter correlation, a closed-loop verification is formed, ensuring the accuracy and reliability of the curve parameters.

[0034] S105. Extract crack propagation characteristics from the corrected curve parameters, use a time series analysis model to predict the evolution trend, and determine the entire process mapping from crack initiation to propagation.

[0035] Extract crack propagation feature set from the corrected curve parameters; input the crack propagation feature set into the time series analysis model to output a predicted evolution trend sequence; perform backtracking analysis on the predicted evolution trend sequence to locate abrupt change moments, and mark the abrupt change moments as crack initiation state points; generate a continuous state sequence based on the crack initiation state points and the predicted evolution trend sequence; map the continuous state sequence to the physical space coordinate system to determine the entire process mapping from crack initiation to propagation.

[0036] The system extracts crack propagation features from the corrected curve parameters and uses a time-series analysis model to predict the evolution trend, determining the entire process mapping from crack initiation to propagation. The specific implementation method is as follows: First, the system analyzes the corrected curve parameters using a feature extraction algorithm to identify key indicators of crack propagation, such as the crack length growth rate and stress concentration factor. The calculated average growth rate in the initial stage of crack propagation is 0.05 mm / hour, and the stress concentration factor is 2.3, indicating a relatively stable crack initiation period. Subsequently, the system inputs the extracted feature data into a time-series analysis model based on a Long Short-Term Memory (LSTM) network. The model uses crack propagation data from the past 24 hours as a training set to predict the propagation trend for the next 12 hours. The prediction results show that the crack length will reach 1.2 mm in the 8th hour, and the growth rate will increase to 0.08 mm / hour, indicating the entry into an accelerated propagation stage. Next, the system constructs a mapping of the entire crack evolution process, dynamically linking the slow growth of cracks from the initiation stage (0.05 mm / h) to the accelerated growth stage (0.08 mm / h) on the time axis, generating a visual curve. The inflection point of the curve occurs when the stress concentration factor increases from 2.3 to 3.1, which coincides with the predicted accelerated propagation time point, logically forming a closed-loop analysis from feature extraction to trend prediction to process mapping. To enhance the reliability of the analysis, the system compares the crack propagation trend with a material fatigue life database, finding that the error between the predicted crack propagation rate and the fatigue crack growth rate under similar stress conditions in the database is only 0.01 mm / h, verifying the accuracy of the model prediction and ensuring the scientific validity and credibility of the crack evolution trend.

[0037] S106. Based on the crack full-process mapping, determine whether the propagation rate is close to the preset critical threshold, and obtain the threshold point of the remaining life of the structure.

[0038] Based on the entire process mapping from crack initiation to propagation, a spatial coordinate sequence and a corresponding time span sequence are extracted. A first-order derivative algorithm is used to process the spatial coordinate sequence and the time span sequence to obtain a propagation rate sequence. A rate gradient set is obtained by performing a difference operation on the propagation rate sequence. The propagation rate sequence is then weighted using the rate gradient set to obtain the approximation value of the current evolution stage. If the approximation value is greater than a preset critical threshold, the structural decay node corresponding to the approximation value is extracted from the entire process mapping. The structural decay node is then matched with a pre-established failure boundary model to obtain the threshold point for the remaining lifespan of the structure.

[0039] Based on the crack propagation process mapping, the system automatically determines whether the propagation rate is close to a preset critical threshold and calculates the threshold point for the remaining life of the structure. The specific implementation method is as follows: First, the system extracts the current propagation rate data from the crack evolution mapping and compares it with the preset critical threshold of 0.1 mm / h. Analysis shows that the currently predicted accelerated propagation rate has reached 0.08 mm / h, close to the critical value, triggering an early warning mechanism. Next, the system calls a damage accumulation algorithm based on finite element analysis, inputting parameters such as a crack length of 1.2 mm and a stress concentration factor of 3.1, to simulate the stress distribution of the structure at the current propagation rate. The calculation shows that when the propagation rate reaches 0.1 mm / h, the crack length will increase to 1.5 mm, corresponding to a remaining life of approximately 36 hours. Furthermore, through correlation analysis with a historical damage database, the system extracts the failure time distribution of similar materials at the critical rate, verifying that the deviation of the calculation results is only 2 hours, confirming the reliability of the remaining life threshold point. Finally, the system compares the remaining lifespan threshold with the structural design lifespan and finds that the current predicted lifespan is 15% lower than the design value of 48 hours. It then automatically generates a risk assessment report, logically forming a closed-loop process from rate judgment to lifespan calculation and risk assessment, ensuring the scientific rigor and practicality of the analysis.

[0040] This invention provides an AI-based health status assessment system for highway bridge and tunnel structures, mainly comprising: The coordinate drift calibration module is used to perform real-time calibration of spatial coordinates by using a Kalman filter algorithm when the coordinate drift threshold exceeds a preset value by detecting the scan timestamp sequence, so as to obtain a continuous sequence of damage point positions. The lifecycle segmentation module is used to identify turning points and divide the lifecycle segments based on a continuous sequence of damage point locations using a clustering algorithm, and to determine the boundaries of the stable segment, the degradation segment, and the acceleration segment. The evolution curve smoothing module is used to acquire expansion rate data within each life cycle segment, and connects the data between segments using a linear interpolation method at the curve jump points to obtain a smooth evolution curve. The residual error correction module is used to determine the source of error by comparing the difference in the expansion rate of adjacent segments if there is a symmetrical filling residual error in the evolution curve, and to obtain the corrected curve parameters. The crack feature extraction module is used to extract crack propagation features from the corrected curve parameters, use a time series analysis model to predict the evolution trend, and determine the entire process mapping from crack initiation to propagation. The remaining life assessment module is used to determine whether the propagation rate is close to a preset critical threshold based on the crack's entire process mapping, and to obtain the threshold point for the structure's remaining life.

[0041] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, the personal information processing rules are clearly informed through signs / information, and authorization is obtained through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0042] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the principle of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. An AI-based method for assessing the structural health of highway bridges and tunnels, characterized in that, The method includes: When the coordinate drift threshold exceeds the preset value by scanning the timestamp sequence, the Kalman filter algorithm is used to calibrate the spatial coordinates in real time to obtain a continuous sequence of damage point locations. Based on the continuous sequence of damage point locations, a clustering algorithm is used to identify turning points and divide the life cycle into segments, determining the boundaries of the stable segment, the degradation segment, and the acceleration segment. Acquire expansion rate data within each life cycle segment, and connect the data between segments using linear interpolation at curve transition points to obtain a smooth evolution curve; If there is a symmetrical filling residual error in the evolution curve, the source of the error can be determined by comparing the difference in the expansion rate of adjacent segments, and the corrected curve parameters can be obtained. Crack propagation characteristics are extracted from the corrected curve parameters, and the evolution trend is predicted using a time series analysis model to determine the entire process mapping from crack initiation to propagation. Based on the mapping of the entire crack process, it is determined whether the propagation rate is close to the preset critical threshold, and the threshold point of the remaining life of the structure is obtained.

2. The AI-based method for assessing the structural health of highway bridges and tunnels according to claim 1, characterized in that, When the coordinate drift threshold detected by scanning the timestamp sequence exceeds a preset value, a Kalman filter algorithm is used to calibrate the spatial coordinates in real time to obtain a continuous sequence of damage point locations, including: Obtain the original spatial coordinates and the corresponding timestamp sequence, and calculate the Euclidean distance between the original spatial coordinates of adjacent timestamp sequences to obtain the coordinate drift. If the coordinate drift is greater than the preset drift threshold, the corresponding original spatial coordinates are extracted to determine the coordinates to be calibrated. A state vector is constructed based on the coordinates to be calibrated, and an observation matrix is ​​established based on the state vector. The target spatial coordinates are obtained by iteratively updating the observation matrix using the Kalman filter algorithm. The target spatial coordinates are spliced ​​together according to the chronological order of the timestamp sequence to obtain a continuous sequence of damage point locations.

3. The AI-based method for assessing the health status of highway bridge and tunnel structures according to claim 1, characterized in that, The process of identifying turning points and dividing the lifecycle into segments based on a continuous sequence of damage point locations using a clustering algorithm, and determining the boundaries of stable, deteriorating, and accelerated segments, includes: Obtain a continuous sequence of damage point locations, and construct a multidimensional feature vector matrix containing tangential slope and curvature based on the damage point location sequence; Clustering algorithms are used to process the multidimensional feature vector matrix to obtain cluster sets, and the intersection time of adjacent cluster sets is identified to determine the turning point; The sequence of damage points is segmented based on the inflection point to obtain evolutionary sub-segments, and the average damage propagation rate of the evolutionary sub-segments is calculated. The average damage propagation rate is compared with a preset threshold to determine the boundaries of the stable segment, the degradation segment, and the acceleration segment.

4. The AI-based method for assessing the structural health of highway bridges and tunnels according to claim 1, characterized in that, The process of acquiring expansion rate data within each lifecycle segment, and connecting the inter-segment data using a linear interpolation method at curve transition points to obtain a smooth evolution curve, includes: Obtain the expansion rate sequence within each lifecycle segment, and calculate the rate difference at the boundary between adjacent lifecycle segments to determine the jump position; Extract the set of boundary breakpoints based on the jump position, and calculate the time span of adjacent breakpoints in the set of boundary breakpoints to determine the interpolation interval; A linear interpolation algorithm is used to process the inter-segment data within the interpolation interval to obtain a smoothed reconstruction rate sequence. By connecting the discrete nodes of each of the lifecycle segments through the reconstruction rate sequence, a smooth evolution curve is obtained.

5. The AI-based method for assessing the structural health of highway bridges and tunnels according to claim 1, characterized in that, If a symmetrical filling residual error exists in the evolution curve, the source of the error is determined by comparing the differences in the expansion rates of adjacent segments, and the corrected curve parameters are obtained, including: Obtain the abnormal interval corresponding to the symmetric filling residual error in the evolution trajectory, wherein the abnormal interval includes the expansion rate of adjacent segments; Calculate the absolute value of the difference between the expansion rates of the adjacent segments, and compare the absolute value of the difference with a preset fluctuation threshold to determine the distribution range of the symmetrical filling residual error; If the numerical offset within the distribution range exceeds the preset range, it is determined that the symmetrical filling residual error originates from data alignment deviation. The evolutionary trajectory is reconstructed based on the data alignment deviation to obtain a reconstructed sequence, and numerical correction is performed using the reconstructed sequence to obtain the corrected curve parameters.

6. The AI-based method for assessing the structural health of highway bridges and tunnels according to claim 1, characterized in that, The process of extracting crack propagation characteristics from the corrected curve parameters, using a time-series analysis model to predict evolution trends, and determining the entire process mapping from crack initiation to propagation includes: Extract the crack propagation feature set from the corrected curve parameters; The crack propagation feature set is input into the time series analysis model to output a predicted evolution trend sequence; The predicted evolution trend sequence is back-analyzed to locate the moment of abrupt change, and the moment of abrupt change is marked as the crack initiation state point; A continuous state sequence is generated based on the crack initiation state points and the predicted evolution trend sequence; The continuous state sequence is mapped to a physical space coordinate system to determine the entire process mapping of cracks from initiation to propagation.

7. The AI-based method for assessing the health status of highway bridge and tunnel structures according to claim 1, characterized in that, The step of determining whether the propagation rate is close to a preset critical threshold based on the crack's entire process mapping, and obtaining the threshold point for the structure's remaining lifespan, includes: Based on the mapping of the entire process from crack initiation to expansion, a spatial coordinate sequence and a corresponding time span sequence are extracted. The first derivative algorithm is used to process the spatial coordinate sequence and the time span sequence to obtain the expansion rate sequence. The rate gradient set is obtained by performing a difference operation on the extended rate sequence, and the approximation value of the current evolution stage is obtained by weighting the extended rate sequence with the rate gradient set. If the approximation value is greater than a preset critical threshold, then the structural decay node corresponding to the approximation value is extracted from the whole process mapping; The threshold point for the remaining life of the structure is obtained by matching the structural decay node with the pre-established failure boundary model.

8. An AI-based highway bridge and tunnel structural health status assessment system, used to perform the method described in any one of claims 1-7, characterized in that, The system includes: The coordinate drift calibration module is used to perform real-time calibration of spatial coordinates by using a Kalman filter algorithm when the coordinate drift threshold exceeds a preset value by detecting the scan timestamp sequence, so as to obtain a continuous sequence of damage point positions. The lifecycle segmentation module is used to identify turning points and divide the lifecycle segments based on a continuous sequence of damage point locations using a clustering algorithm, and to determine the boundaries of the stable segment, the degradation segment, and the acceleration segment. The evolution curve smoothing module is used to acquire expansion rate data within each life cycle segment, and connects the data between segments using a linear interpolation method at the curve jump points to obtain a smooth evolution curve. The residual error correction module is used to determine the source of error by comparing the difference in the expansion rate of adjacent segments if there is a symmetrical filling residual error in the evolution curve, and to obtain the corrected curve parameters. The crack feature extraction module is used to extract crack propagation features from the corrected curve parameters, use a time series analysis model to predict the evolution trend, and determine the entire process mapping from crack initiation to propagation. The remaining life assessment module is used to determine whether the propagation rate is close to a preset critical threshold based on the crack's entire process mapping, and to obtain the threshold point for the structure's remaining life.

Images