A highway infrastructure health monitoring method and system based on intelligent sensing network

Abstract

This invention belongs to the field of facility monitoring technology, specifically relating to a method and system for health monitoring of highway infrastructure based on an intelligent sensing network. The method includes: real-time acquisition and preprocessing of raw strain field and temperature field data of highway infrastructure using pre-embedded sensors to obtain a stable strain field sequence and a stable temperature field sequence; separation of low-frequency periodic components caused by environmental temperature fluctuations from the stable temperature field sequence, combined with the stable strain field sequence to obtain a temperature-compensated strain field sequence; separation of the reference drift component from the temperature-compensated strain field sequence, and elimination of vehicle load influence factors to obtain a residual strain field sequence; obtaining a set of abnormal deviation points based on the residual strain field sequence; and obtaining the location and degree of damage based on the set of abnormal deviation points. This invention can automatically complete damage identification and safety level assessment of highway infrastructure.

Description

Technical Field

[0001] This invention belongs to the field of facility monitoring technology, specifically relating to a method and system for monitoring the health of highway infrastructure based on intelligent sensing networks. Background Technology

[0002] Health monitoring of highway infrastructure is a key research area for ensuring traffic safety and extending structural lifespan. Its importance lies in the fact that hidden damage to bridges, pavements, or tunnels can trigger sudden disasters, causing enormous loss of life and property. Current monitoring methods mostly rely on periodic manual inspections combined with the deployment of external sensors. While these methods can acquire some surface data, they are insufficient to comprehensively capture real-time changes within the infrastructure. Consequently, damage detection often lags behind actual developments, hindering preventative maintenance.

[0003] The core changes in infrastructure health occur within the concrete and in concealed areas such as reinforcing steel. Information such as strain fields, temperature fields, and crack propagation directly reflect the degradation of structural load-bearing capacity and durability. However, the continuous collection of this critical information faces significant challenges because embedded sensors need to operate stably under complex loads and environmental conditions for extended periods. Any signal drift or attenuation will affect data reliability, making it difficult for the monitoring system to accurately distinguish between normal fluctuations and actual damage.

[0004] This reliability issue further complicates health diagnosis: the system needs to accurately compare the currently collected internal strain, temperature, and crack information with the baseline state at the time of design to determine the degree of deviation. However, changes in ambient temperature, vehicle dynamic loads, and material aging effects can cause the baseline state itself to drift slowly, leading to misjudgments in the comparison results. For example, in long-term operation, the natural differences in the internal strain field of the same bridge in winter and summer may be misidentified as damage expansion, or the minute strain changes caused by steel corrosion may be masked by normal temperature effects, making it impossible for the system to quantify and present key indicators such as the degree of steel corrosion, the depth of concrete carbonation, and the prestress damage rate in a timely manner.

[0005] Therefore, how to accurately compare the current state with the design benchmark based on the long-term stable acquisition of internal multi-field information by pre-embedded sensors, and automatically complete damage identification and safety level assessment, has become a key issue that needs to be addressed in the health monitoring of highway infrastructure. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method and system for monitoring the health of highway infrastructure based on an intelligent sensing network, which aims to automatically identify damage and assess the safety level of highway infrastructure.

[0007] To achieve the above objectives, the present invention provides the following solution: A method for health monitoring of highway infrastructure based on intelligent sensing networks, the method comprising: Based on pre-embedded sensors, the raw strain field data and temperature field data of highway infrastructure are collected in real time and pre-processed to obtain a stable strain field sequence and a stable temperature field sequence. The low-frequency periodic components caused by ambient temperature fluctuations are separated from the stationary temperature field sequence, and combined with the stationary strain field sequence to obtain the temperature-compensated strain field sequence. The reference drift component was separated from the temperature-compensated strain field sequence, and the influence of vehicle load was eliminated to obtain the residual strain field sequence. Based on the residual strain field sequence, the set of abnormal deviation points is obtained; Based on the set of abnormal deviation points, the location and degree of damage of the damaged area are obtained.

[0008] Preferably, the method for separating the low-frequency periodic components caused by ambient temperature fluctuations from a stationary temperature field sequence and combining them with a stationary strain field sequence to obtain a temperature-compensated strain field sequence includes: The frequency spectrum of the stationary temperature field sequence is obtained by performing a frequency domain transformation on the Fourier transform. In the frequency spectrum of the temperature field, the peak frequency in the low-frequency range is determined. The component corresponding to this peak frequency is the low-frequency periodic component generated by the fluctuation of ambient temperature. Based on the peak frequency of the low-frequency periodic components, a linear combination of sine and cosine waves is constructed to reconstruct the low-frequency periodic component sequence. Subtracting the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence yields the residual temperature field sequence after removing ambient temperature fluctuations. The linear relationship between the residual temperature field sequence and the stationary strain field sequence was established by least squares regression, and the regression coefficients were determined. The temperature compensation value is calculated by using the regression coefficients and the residual temperature field sequence. The temperature compensation value is then subtracted from the stationary strain field sequence to obtain the temperature-compensated strain field sequence.

[0009] Preferably, the method for separating the reference drift component from the temperature-compensated strain field sequence and eliminating the influence of vehicle load to obtain the residual strain field sequence includes: A low-pass filter was used to process the strain field sequence after temperature compensation to obtain the slowly changing reference drift component. Subtract the slowly changing reference drift component from the temperature-compensated strain field sequence to obtain the initial residual strain field sequence after deducting the reference drift. Vehicle load record data is acquired, and a synchronous temporal relationship between the initial residual strain field sequence and the vehicle load record data is established through a timestamp alignment algorithm to obtain time-matched data pairs. Based on the time-matched data pairs, the initial residual strain field sequence and vehicle load record in the time-matched data pairs are converted to the frequency domain using fast Fourier transform to obtain their respective spectral amplitude response information. Based on the spectral amplitude response information, the correlation coefficient between the initial residual strain field and the vehicle load on each frequency component is calculated. If the correlation coefficient exceeds the preset threshold, it is determined that the frequency component is affected by the vehicle load, and the load-related frequency component set is obtained. A frequency domain filtering method is used to remove the spectral components corresponding to the vehicle load-related frequency components from the initial residual strain field spectrum, and an inverse fast Fourier transform is used to obtain the residual strain field sequence after removing the influence of vehicle load factors.

[0010] Preferably, the method for obtaining the set of abnormal deviation points based on the residual strain field sequence includes: The residual strain field sequence was divided into regions using the sliding window technique to obtain the region division results. The amplitude of strain field values ​​in each sub-region of the regional division results is statistically analyzed by amplitude calculation, and the maximum amplitude value of strain field values ​​in each sub-region is calculated to obtain regional amplitude distribution data. If the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, the region is marked as a candidate abnormal region, forming a list of candidate abnormal regions. A clustering algorithm is used to perform spatial clustering analysis on the candidate abnormal region list. The spatial distribution pattern of abnormal deviation points is determined based on the spatial adjacency relationship and amplitude similarity between regions, and the clustering analysis results are obtained. The clustering analysis results are screened a second time by a preset anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a real anomaly deviation point. Based on the spatial coordinates and time series information of the actual abnormal deviation points, the location coordinates, amplitude values ​​and detection time of each actual abnormal deviation point are recorded to form the final set of abnormal deviation points.

[0011] Preferably, the method for obtaining the location and extent of damage based on a set of abnormal deviation points includes: Based on the spatial coordinate information of the highway infrastructure corresponding to the set of abnormal deviation points, calculate the Euclidean distance between each abnormal deviation point and construct a distance matrix. The K-means clustering algorithm is used to perform cluster analysis on outliers, and the cluster centers are determined based on the distance matrix to obtain multiple clusters of outliers. For each cluster of abnormal deviation points, the spatial boundary range of abnormal deviation points within the cluster is calculated. The spatial range of each cluster of abnormal deviation points is determined by the minimum bounding rectangle method, thus obtaining the damage area. The distribution density of abnormal deviation points in each damaged area is analyzed using density statistics methods. Based on the distribution density and the preset density threshold, the degree of damage in the damaged area is determined. A damage assessment report is generated based on the spatial location and extent of the damage.

[0012] The present invention also provides a highway infrastructure health monitoring system based on an intelligent sensing network. The system is used to implement the aforementioned method and includes: a data acquisition and processing module, a temperature compensation module, a separation and rejection module, an anomaly acquisition module, and a damage location module. The data acquisition and processing module is used to collect raw strain field data and temperature field data of highway infrastructure in real time based on pre-embedded sensors and perform preprocessing to obtain stable strain field sequence and stable temperature field sequence. The temperature compensation module is used to separate the low-frequency periodic components caused by ambient temperature fluctuations from the steady temperature field sequence, and combine them with the steady strain field sequence to obtain the temperature-compensated strain field sequence. The separation and elimination module is used to separate the reference drift component from the temperature-compensated strain field sequence and eliminate the influence of vehicle load to obtain the residual strain field sequence. The anomaly acquisition module is used to obtain a set of abnormal deviation points based on the residual strain field sequence; The damage localization module is used to obtain the location and degree of damage of the damaged area based on the set of abnormal deviation points.

[0013] Preferably, the temperature compensation module includes: a frequency domain conversion unit, a low-frequency filtering unit, a reconstruction unit, a removal unit, a linear regression unit, and a compensation unit; The frequency domain transformation unit is used to perform frequency domain transformation on the stationary temperature field sequence through Fourier transform to obtain the temperature field frequency spectrum; The low-frequency screening unit is used to determine the peak frequency in the low-frequency range in the temperature field frequency spectrum. The component corresponding to the peak frequency is the low-frequency periodic component generated by the ambient temperature fluctuation. The reconstruction unit is used to construct a linear combination of sine and cosine waves based on the peak frequency of the low-frequency periodic components, and reconstruct the low-frequency periodic component sequence. The removal unit is used to subtract the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence to obtain the residual temperature field sequence after removing ambient temperature fluctuations. The linear regression unit is used to establish the linear relationship between the residual temperature field sequence and the stationary strain field sequence using least squares regression, and to determine the regression coefficients; The compensation unit is used to calculate the temperature compensation value through the regression coefficient and the residual temperature field sequence, and to subtract the temperature compensation value from the stationary strain field sequence to obtain the temperature-compensated strain field sequence.

[0014] Preferably, the separation and elimination module includes: a filtering unit, a separation unit, an alignment unit, a spectrum amplitude unit, a correlation analysis unit, and an elimination unit; The filtering unit is used to process the temperature-compensated strain field sequence with a low-pass filter to obtain the slowly changing reference drift component. The separation unit is used to subtract the slowly changing reference drift component from the temperature-compensated strain field sequence to obtain the initial residual strain field sequence after deducting the reference drift. The alignment unit is used to acquire vehicle load record data and establish a synchronous temporal relationship between the initial residual strain field sequence and the vehicle load record data through a timestamp alignment algorithm to obtain time-matched data pairs. The spectral amplitude unit is used to convert the initial residual strain field sequence and vehicle load record in the time-matched data pair to the frequency domain using fast Fourier transform, thereby obtaining their respective spectral amplitude response information. The correlation analysis unit is used to calculate the correlation coefficient between the initial residual strain field and the vehicle load on each frequency component based on the spectral amplitude response information. If the correlation coefficient exceeds the preset threshold, it is determined that the frequency component is affected by the vehicle load, and the load-related frequency component set is obtained. The elimination unit is used to remove the spectral components corresponding to the set of frequency components related to vehicle load from the initial residual strain field spectrum using a frequency domain filtering method, and then use inverse fast Fourier transform to obtain the residual strain field sequence after removing the influence factors of vehicle load.

[0015] Preferably, the anomaly detection module includes: a partitioning unit, an amplitude detection unit, a candidate unit, a first clustering unit, a verification unit, and an detection unit; The partitioning unit is used to divide the residual strain field sequence into regions using the sliding window technique to obtain the region partitioning results. The amplitude acquisition unit is used to perform amplitude statistics on the strain field values ​​of each sub-region in the regional division result through amplitude calculation, calculate the maximum amplitude value of the strain field values ​​in each sub-region, and obtain regional amplitude distribution data. Candidate units are used to mark a region as a candidate anomalous region if the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, thus forming a list of candidate anomalous regions. The first clustering unit is used to perform spatial clustering analysis on the candidate abnormal region list using clustering algorithms. Based on the spatial adjacency relationship and amplitude similarity between regions, the spatial distribution pattern of abnormal deviation points is determined to obtain the clustering analysis results. The verification unit is used to perform secondary screening of the clustering analysis results through a preset anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a real anomaly deviation point. The acquisition unit is used to record the location coordinates, amplitude values, and detection time of each real abnormal deviation point based on the spatial coordinates and time series information of the real abnormal deviation points, forming the final set of abnormal deviation points.

[0016] Preferably, the damage localization module includes: a distance calculation unit, a second clustering unit, a loss localization unit, a severity judgment unit, and a report generation unit; The distance calculation unit is used to calculate the Euclidean distance between each abnormal deviation point based on the spatial coordinate information of the highway infrastructure corresponding to the abnormal deviation point set, and to construct a distance matrix. The second clustering unit is used to perform cluster analysis on outlier points using the K-means clustering algorithm, determine the cluster centers based on the distance matrix, and obtain multiple clusters of outlier points. The loss localization unit is used to calculate the spatial boundary range of abnormal deviation points within each abnormal deviation point cluster, and to determine the spatial range of each abnormal deviation point cluster by using the minimum bounding rectangle method to obtain the damage area. The degree judgment unit is used to analyze the distribution density of abnormal deviation points in each damaged area using density statistics methods, and determine the degree of damage in the damaged area based on the distribution density and the preset density threshold. The report generation unit is used to generate a damage area assessment report based on the spatial location and degree of damage of the damaged area.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention discloses a method and system for health monitoring of highway infrastructure based on intelligent sensing networks. Addressing the problem in operational scenarios where strain field data in highway infrastructure is easily affected by high-frequency noise, ambient temperature fluctuations, reference drift, and vehicle load interference, making it difficult to accurately extract true structural damage signals, the invention achieves precise damage localization through multi-step signal processing and analysis. First, raw strain and temperature field data collected by embedded sensors are acquired. High-frequency noise is removed through adaptive filtering to obtain a stationary sequence. Then, low-frequency periodic components are separated from the stationary temperature field sequence, and temperature compensation is performed to obtain a pure strain field. Next, the slow reference drift component is subtracted, and the influence of vehicle loads is eliminated to obtain the residual strain field. When the residual amplitude exceeds a threshold in a local area, an abnormal deviation point set is identified. This set is then clustered to form damage regions, and density statistics are used to determine the degree of damage in these regions. This invention progressively eliminates interference factors, achieving precise separation from noisy raw data to true damage signals. This improves the sensitivity and accuracy of early damage identification and localization, demonstrating significant engineering application value. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the process of a highway infrastructure health monitoring method based on an intelligent sensing network according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a highway infrastructure health monitoring system module based on an intelligent sensing network, according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 like Figure 1 As shown, this invention provides a method for monitoring the health of highway infrastructure based on an intelligent sensing network, comprising: Based on pre-embedded sensors, the raw strain field data and temperature field data of highway infrastructure are collected in real time and pre-processed to obtain a stable strain field sequence and a stable temperature field sequence. The low-frequency periodic components caused by ambient temperature fluctuations are separated from the stationary temperature field sequence, and combined with the stationary strain field sequence to obtain the temperature-compensated strain field sequence. The reference drift component was separated from the temperature-compensated strain field sequence, and the influence of vehicle load was eliminated to obtain the residual strain field sequence. Based on the residual strain field sequence, the set of abnormal deviation points is obtained; Based on the set of abnormal deviation points, the location and degree of damage of the damaged area are obtained.

[0023] The specific implementation process of this invention is as follows: Based on embedded sensors, raw strain field data and temperature field data of highway infrastructure are acquired in real time and preprocessed to obtain stationary strain field sequences and stationary temperature field sequences. The preprocessing methods specifically include: First, the original strain field data and temperature field data are synchronized to obtain a synchronized field dataset. This synchronization is achieved through timestamp calibration and interpolation algorithms. When data is missing from a pre-embedded sensor, linear interpolation compensation is performed using data from adjacent pre-embedded sensors. For example, if the strain data of pre-embedded sensor A is lost at 10:30, interpolation is performed using its data from 10:29 and 10:31. Simultaneously, spatial interpolation correction is performed by referencing the synchronous data from adjacent pre-embedded sensors B and C to ensure data integrity and accuracy.

[0024] Then, an adaptive filter is used to filter the strain field data in the synchronous field dataset to obtain a stationary strain field sequence after removing high-frequency noise. Simultaneously, the same adaptive filter is used to filter the temperature field data in the synchronous field dataset to obtain a stationary temperature field sequence after removing high-frequency noise. This invention utilizes an adaptive filter to effectively improve data quality. By dynamically adjusting the filtering parameters, the adaptive filter can remove high-frequency noise based on signal characteristics, thereby obtaining stationary strain field and temperature field sequences. This processing contributes to the accuracy of subsequent analysis and avoids misjudgments caused by noise interference.

[0025] Furthermore, the low-frequency periodic components caused by ambient temperature fluctuations are separated from the stationary temperature field sequence, and combined with the stationary strain field sequence, a temperature-compensated strain field sequence is obtained, specifically including: The stationary temperature field sequence is transformed into the frequency domain using Fourier transform to obtain its frequency spectrum. By performing a Fourier transform on the stationary temperature field sequence, its frequency components are revealed. In the temperature field frequency spectrum, the low-frequency range typically corresponds to diurnal or seasonal environmental fluctuations, while the peak frequency often appears in the range of 0.00001 Hz to 0.0001 Hz. This component is mainly caused by slow changes in external temperature.

[0026] In the frequency spectrum of the temperature field, the peak frequency in the low-frequency range is determined. The component corresponding to this peak frequency is the low-frequency periodic component generated by the fluctuation of ambient temperature.

[0027] By constructing a linear combination of sine and cosine waves based on the peak frequencies of the low-frequency periodic components, the sequence of low-frequency periodic components can be reconstructed. For example, using an amplitude-matched sine function to reconstruct this low-frequency periodic component sequence can accurately reproduce the slow rising and falling trends of temperature.

[0028] Subtracting the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence yields a residual temperature field sequence that removes environmental temperature fluctuations. This residual temperature field sequence primarily retains short-term local temperature changes, such as rapid fluctuations caused by direct sunlight or brief periods of shading. These changes are more directly correlated with structural strain. This removal of low-frequency environmental fluctuations helps highlight the immediate impact of temperature on the structure and avoids interference from long-period environmental factors in subsequent compensation accuracy.

[0029] Least square regression was used to establish a linear relationship between the residual temperature field sequence and the stationary strain field sequence, and the regression coefficients were determined. The regression coefficients reflect the strain response caused by a unit temperature change. For example, a regression coefficient of -15 microstrain / ℃ indicates that the strain decreases by 15 microstrain when the temperature increases by 1℃.

[0030] Temperature compensation values ​​are calculated using regression coefficients and the residual temperature field sequence. These compensation values ​​are then subtracted from the stationary strain field sequence to obtain the temperature-compensated strain field sequence. The temperature-compensated strain field sequence significantly reduces temperature-induced spurious strain fluctuations and more accurately reflects structural changes caused by loads or damage.

[0031] Furthermore, the reference drift component is separated from the temperature-compensated strain field sequence, and the influence of vehicle load is eliminated to obtain the residual strain field sequence, specifically including: A low-pass filter is used to process the temperature-compensated strain field sequence to obtain the slowly varying reference drift component. By setting a low cutoff frequency, the low-pass filter allows only extremely low-frequency signals to pass through, thereby separating the trend components that slowly rise or fall over time.

[0032] The initial residual strain field sequence after removing the slowly varying reference drift component is obtained by subtracting the reference drift component from the temperature-compensated strain field sequence. This sequence better reflects the true structural response and avoids long-term trends masking local variations.

[0033] Vehicle load record data is acquired, and a time-series synchronization relationship between the initial residual strain field sequence and the vehicle load record data is established using a timestamp alignment algorithm to obtain time-matched data pairs. The low-frequency data is upsampled to a high-frequency data time base using the timestamp alignment algorithm to ensure that each time point has a corresponding data pair, thereby establishing an accurate time-series matching relationship.

[0034] Based on the time-matched data pairs, the initial residual strain field sequence and vehicle load record in the time-matched data pairs are transformed to the frequency domain using Fast Fourier Transform to obtain their respective spectral amplitude response information.

[0035] Based on the spectral amplitude response information, the correlation coefficients between the initial residual strain field and the vehicle load at each frequency component are calculated. If the correlation coefficient exceeds a preset threshold, it is determined that the frequency component is affected by the vehicle load, thus obtaining the set of load-related frequency components. The correlation coefficient can be used to quantify the degree of influence of the vehicle load by calculating the Pearson correlation coefficient of each frequency component. When the correlation coefficient of a certain frequency component exceeds the threshold, it indicates that the frequency component is mainly caused by the vehicle load.

[0036] A frequency domain filtering method is employed to remove the spectral components corresponding to the vehicle load-related frequency components from the initial residual strain field spectrum. An inverse fast Fourier transform is then used to obtain the residual strain field sequence after removing the influence of vehicle load. The frequency domain filtering process uses a band-stop filter to remove vehicle load-related frequency components. For the identified vehicle load-related frequency components, corresponding filter parameters are designed to attenuate the amplitude of these frequency components to less than 5% of their original amplitude. This processing effectively eliminates the interference of vehicle load on the strain field while retaining structural damage information that may be contained in other frequency components.

[0037] Furthermore, based on the residual strain field sequence, a set of anomalous deviation points is obtained, specifically including: A sliding window technique is used to divide the residual strain field sequence into regions, resulting in region division. Based on the spatial coordinate information of the highway infrastructure (such as bridge structures) corresponding to the residual strain field sequence, a fixed-size window, such as a 5m × 5m grid, is set, with a window sliding step of 2m. The window is then slid across the entire area of ​​the bridge structure row by row to obtain several sub-regions. This division can capture local anomalies without being overly fragmented.

[0038] The amplitude of strain field values ​​in each sub-region of the regional division is statistically analyzed through amplitude calculation. The maximum amplitude value of strain field values ​​in each sub-region is calculated to obtain regional amplitude distribution data. This step of statistically analyzing the amplitude of strain field values ​​in each sub-region, mainly calculating the maximum amplitude value, and traversing all sub-regions, forms regional amplitude distribution data. This step helps to quickly locate potential high-strain concentration areas.

[0039] If the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, that region is marked as a candidate anomaly region, forming a candidate anomaly region list. If the maximum amplitude value of a region exceeds a preset threshold (e.g., a threshold of 2.5), it is also marked as a candidate anomaly region. In actual highway infrastructure monitoring, this threshold is determined based on the normal fluctuation range of the historical residual strain field to avoid misjudgment due to environmental noise. The candidate anomaly region list is thus formed and typically contains dozens of scattered or concentrated regions.

[0040] Spatial clustering analysis of the candidate anomaly region list is performed using clustering algorithms. Based on the spatial adjacency and amplitude similarity between regions, the spatial distribution pattern of anomaly deviation points is determined, yielding the clustering analysis results. For example, when using the DBSCAN clustering algorithm, regions less than 3 meters apart are considered adjacent, and amplitude differences less than 0.5 are considered similar, thus clustering multiple close candidate regions into one class. This clustering method can identify anomaly spatial distribution patterns.

[0041] The clustering analysis results are further filtered through a pre-defined anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a true anomaly deviation point. For example, an anomaly is confirmed as a true anomaly deviation point only if it exceeds the threshold for five consecutive time steps and spatially covers at least three adjacent sub-regions. This anomaly verification can significantly reduce the artifacts caused by instantaneous noise or vehicle residue.

[0042] Based on the spatial coordinates and time series information of the actual abnormal deviation points, the location coordinates, amplitude values ​​and detection time of each actual abnormal deviation point are recorded to form the final set of abnormal deviation points.

[0043] Furthermore, based on the set of abnormal deviation points, the location and extent of damage are obtained, specifically including: Based on the spatial coordinates of the highway infrastructure corresponding to the set of anomalous deviation points, the Euclidean distance between each anomalous deviation point is calculated, and a distance matrix is ​​constructed. The distance matrix reflects the density between anomalous deviation points and helps to reveal potential damage patterns.

[0044] K-means clustering algorithm is used to cluster abnormal deviation points, determining cluster centers based on the distance matrix, resulting in multiple clusters of abnormal deviation points. For example, selecting 3 clusters can divide scattered abnormal deviation points of highway infrastructure bridges into three clusters: the center of the bridge deck, near the supports, and the edge. This clustering can clearly distinguish different damage source areas and improve positioning accuracy.

[0045] For each cluster of anomalous deviation points, the spatial boundary range of the anomalous deviation points within the cluster is calculated. The spatial range of each cluster is determined using the minimum bounding rectangle method, thus obtaining the damaged area. The minimum bounding rectangle method is used to calculate the spatial boundary range of each cluster of anomalous deviation points, which can concisely enclose all anomalous deviation points and determine the coordinates of the damaged area. For example, if an anomalous deviation point within a cluster spans a 2-meter area, the minimum bounding rectangle can accurately label the coordinate region with x-coordinates from 10 meters to 12 meters and y-coordinates from 5 meters to 7 meters. This method is computationally efficient and facilitates subsequent visualization and evaluation.

[0046] Density statistical methods are used to analyze the distribution density of abnormal deviation points within each damaged area. Based on the distribution density and a preset density threshold, the degree of damage in the damaged area is determined. The density statistical method analyzes the distribution density of abnormal deviation points within the damaged area. If the number of abnormal deviation points per square meter in the damaged area exceeds a preset first density threshold of 7, it is identified as a high-risk damaged area. If the number of abnormal deviation points per square meter is below the preset first density threshold but above a preset second density threshold of 4, it is identified as a medium-risk damaged area; otherwise, it is identified as a low-risk damaged area. For example, in a bridge, when the number of abnormal deviation points in the damaged area reaches 8 points / square meter, it is marked as a high-risk damaged area. This density threshold judgment can effectively filter isolated noise and highlight concentrated anomalies caused by continuous damage accumulation.

[0047] Based on the spatial location and extent of damage, a damage assessment report is generated, including the area coordinates, the number of abnormal deviation points, and the risk level. For example, a damage assessment report might record that a certain damaged area has coordinates from (15,8) to (18,10), six abnormal deviation points, and a risk level of medium risk. This type of report provides an intuitive basis for maintenance decisions, and timely intervention can prevent damage from spreading, ensure structural safety, and extend service life.

[0048] In summary, this invention discloses a method for health monitoring of highway infrastructure based on intelligent sensing networks. First, raw strain and temperature field data are acquired from pre-embedded sensors. High-frequency noise is removed through adaptive filtering to obtain a stationary sequence. Then, low-frequency periodic components are separated from the stationary temperature field sequence, and temperature compensation is performed to obtain a pure strain field. Next, the slow reference drift component is subtracted, and the influence of vehicle loads is eliminated to obtain the residual strain field. When the residual amplitude exceeds a threshold in a local area, an abnormal deviation point set is identified. This set is then clustered to form damage regions, and density statistics are used to determine the degree of damage in these regions. This invention progressively eliminates interference factors, achieving precise separation from noisy raw data to the true damage signal. This improves the sensitivity and accuracy of early damage identification and positioning, demonstrating significant engineering application value.

[0049] Example 2 like Figure 2 As shown, based on the same inventive concept, the present invention also provides a highway infrastructure health monitoring system based on an intelligent sensing network, used to implement the methods described in the foregoing embodiments. The system includes: a data acquisition and processing module, a temperature compensation module, a separation and rejection module, an anomaly acquisition module, and a damage location module. The data acquisition and processing module is used to collect raw strain field data and temperature field data of highway infrastructure in real time based on pre-embedded sensors and perform preprocessing to obtain stable strain field sequence and stable temperature field sequence. The temperature compensation module is used to separate the low-frequency periodic components caused by ambient temperature fluctuations from the steady temperature field sequence, and combine them with the steady strain field sequence to obtain the temperature-compensated strain field sequence. The separation and elimination module is used to separate the reference drift component from the temperature-compensated strain field sequence and eliminate the influence of vehicle load to obtain the residual strain field sequence. The anomaly acquisition module is used to obtain a set of abnormal deviation points based on the residual strain field sequence; The damage localization module is used to obtain the location and degree of damage of the damaged area based on the set of abnormal deviation points.

[0050] Furthermore, in this embodiment, the temperature compensation module includes: a frequency domain conversion unit, a low-frequency filtering unit, a reconstruction unit, a removal unit, a linear regression unit, and a compensation unit; The frequency domain transformation unit is used to perform frequency domain transformation on the stationary temperature field sequence through Fourier transform to obtain the temperature field frequency spectrum; The low-frequency screening unit is used to determine the peak frequency in the low-frequency range in the temperature field frequency spectrum. The component corresponding to the peak frequency is the low-frequency periodic component generated by the ambient temperature fluctuation. The reconstruction unit is used to construct a linear combination of sine and cosine waves based on the peak frequency of the low-frequency periodic components, and reconstruct the low-frequency periodic component sequence. The removal unit is used to subtract the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence to obtain the residual temperature field sequence after removing ambient temperature fluctuations. The linear regression unit is used to establish the linear relationship between the residual temperature field sequence and the stationary strain field sequence using least squares regression, and to determine the regression coefficients; The compensation unit is used to calculate the temperature compensation value through the regression coefficient and the residual temperature field sequence, and to subtract the temperature compensation value from the stationary strain field sequence to obtain the temperature-compensated strain field sequence.

[0051] Furthermore, in this embodiment, the separation and elimination module includes: a filtering unit, a separation unit, an alignment unit, a spectrum amplitude unit, a correlation analysis unit, and an elimination unit; The filtering unit is used to process the temperature-compensated strain field sequence with a low-pass filter to obtain the slowly changing reference drift component. The separation unit is used to subtract the slowly changing reference drift component from the temperature-compensated strain field sequence to obtain the initial residual strain field sequence after deducting the reference drift. The alignment unit is used to acquire vehicle load record data and establish a synchronous temporal relationship between the initial residual strain field sequence and the vehicle load record data through a timestamp alignment algorithm to obtain time-matched data pairs. The spectral amplitude unit is used to convert the initial residual strain field sequence and vehicle load record in the time-matched data pair to the frequency domain using fast Fourier transform, thereby obtaining their respective spectral amplitude response information. The correlation analysis unit is used to calculate the correlation coefficient between the initial residual strain field and the vehicle load on each frequency component based on the spectral amplitude response information. If the correlation coefficient exceeds the preset threshold, it is determined that the frequency component is affected by the vehicle load, and the load-related frequency component set is obtained. The elimination unit is used to remove the spectral components corresponding to the set of frequency components related to vehicle load from the initial residual strain field spectrum using a frequency domain filtering method, and then use inverse fast Fourier transform to obtain the residual strain field sequence after removing the influence factors of vehicle load.

[0052] Furthermore, in this embodiment, the anomaly acquisition module includes: a partitioning unit, an amplitude acquisition unit, a candidate unit, a first clustering unit, a verification unit, and an acquisition unit; The partitioning unit is used to divide the residual strain field sequence into regions using the sliding window technique to obtain the region partitioning results. The amplitude acquisition unit is used to perform amplitude statistics on the strain field values ​​of each sub-region in the regional division result through amplitude calculation, calculate the maximum amplitude value of the strain field values ​​in each sub-region, and obtain regional amplitude distribution data. Candidate units are used to mark a region as a candidate anomalous region if the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, thus forming a list of candidate anomalous regions. The first clustering unit is used to perform spatial clustering analysis on the candidate abnormal region list using clustering algorithms. Based on the spatial adjacency relationship and amplitude similarity between regions, the spatial distribution pattern of abnormal deviation points is determined to obtain the clustering analysis results. The verification unit is used to perform secondary screening of the clustering analysis results through a preset anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a real anomaly deviation point. The acquisition unit is used to record the location coordinates, amplitude values, and detection time of each real abnormal deviation point based on the spatial coordinates and time series information of the real abnormal deviation points, forming the final set of abnormal deviation points.

[0053] Furthermore, in this embodiment, the damage localization module includes: a distance calculation unit, a second clustering unit, a loss localization unit, a severity judgment unit, and a report generation unit; The distance calculation unit is used to calculate the Euclidean distance between each abnormal deviation point based on the spatial coordinate information of the highway infrastructure corresponding to the abnormal deviation point set, and to construct a distance matrix. The second clustering unit is used to perform cluster analysis on outlier points using the K-means clustering algorithm, determine the cluster centers based on the distance matrix, and obtain multiple clusters of outlier points. The loss localization unit is used to calculate the spatial boundary range of abnormal deviation points within each abnormal deviation point cluster, and to determine the spatial range of each abnormal deviation point cluster by using the minimum bounding rectangle method to obtain the damage area. The degree judgment unit is used to analyze the distribution density of abnormal deviation points in each damaged area using density statistics methods, and determine the degree of damage in the damaged area based on the distribution density and the preset density threshold. The report generation unit is used to generate a damage area assessment report based on the spatial location and degree of damage of the damaged area.

[0054] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for monitoring the health of highway infrastructure based on intelligent sensing networks, characterized in that, The method includes: Based on pre-embedded sensors, the raw strain field data and temperature field data of highway infrastructure are collected in real time and pre-processed to obtain a stable strain field sequence and a stable temperature field sequence. The low-frequency periodic components caused by ambient temperature fluctuations are separated from the stationary temperature field sequence, and combined with the stationary strain field sequence to obtain the temperature-compensated strain field sequence. The reference drift component was separated from the temperature-compensated strain field sequence, and the influence of vehicle load was eliminated to obtain the residual strain field sequence. Based on the residual strain field sequence, the set of abnormal deviation points is obtained; Based on the set of abnormal deviation points, the location and degree of damage of the damaged area are obtained.

2. The method according to claim 1, characterized in that, Methods for separating low-frequency periodic components caused by ambient temperature fluctuations from a stationary temperature field sequence and combining them with a stationary strain field sequence to obtain a temperature-compensated strain field sequence include: The frequency spectrum of the stationary temperature field sequence is obtained by performing a frequency domain transformation on the Fourier transform. In the frequency spectrum of the temperature field, the peak frequency in the low-frequency range is determined. The component corresponding to this peak frequency is the low-frequency periodic component generated by the fluctuation of ambient temperature. Based on the peak frequency of the low-frequency periodic components, a linear combination of sine and cosine waves is constructed to reconstruct the low-frequency periodic component sequence. Subtracting the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence yields the residual temperature field sequence after removing ambient temperature fluctuations. The linear relationship between the residual temperature field sequence and the stationary strain field sequence was established by least squares regression, and the regression coefficients were determined. The temperature compensation value is calculated by using the regression coefficients and the residual temperature field sequence. The temperature compensation value is then subtracted from the stationary strain field sequence to obtain the temperature-compensated strain field sequence.

3. The method according to claim 1, characterized in that, Methods for separating the reference drift component from the temperature-compensated strain field sequence and eliminating the influence of vehicle load to obtain the residual strain field sequence include: A low-pass filter was used to process the strain field sequence after temperature compensation to obtain the slowly changing reference drift component. Subtract the slowly changing reference drift component from the temperature-compensated strain field sequence to obtain the initial residual strain field sequence after deducting the reference drift. Vehicle load record data is acquired, and a synchronous temporal relationship between the initial residual strain field sequence and the vehicle load record data is established through a timestamp alignment algorithm to obtain time-matched data pairs. Based on the time-matched data pairs, the initial residual strain field sequence and vehicle load record in the time-matched data pairs are converted to the frequency domain using fast Fourier transform to obtain their respective spectral amplitude response information. Based on the spectral amplitude response information, the correlation coefficient between the initial residual strain field and the vehicle load on each frequency component is calculated. If the correlation coefficient exceeds the preset threshold, it is determined that the frequency component is affected by the vehicle load, and the load-related frequency component set is obtained. A frequency domain filtering method is used to remove the spectral components corresponding to the vehicle load-related frequency components from the initial residual strain field spectrum, and an inverse fast Fourier transform is used to obtain the residual strain field sequence after removing the influence of vehicle load factors.

4. The method according to claim 1, characterized in that, Methods for obtaining the set of abnormal deviation points based on the residual strain field sequence include: The residual strain field sequence was divided into regions using the sliding window technique to obtain the region division results. The amplitude of strain field values ​​in each sub-region of the regional division results is statistically analyzed by amplitude calculation, and the maximum amplitude value of strain field values ​​in each sub-region is calculated to obtain regional amplitude distribution data. If the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, the region is marked as a candidate abnormal region, forming a list of candidate abnormal regions. A clustering algorithm is used to perform spatial clustering analysis on the candidate abnormal region list. The spatial distribution pattern of abnormal deviation points is determined based on the spatial adjacency relationship and amplitude similarity between regions, and the clustering analysis results are obtained. The clustering analysis results are screened a second time by a preset anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a real anomaly deviation point. Based on the spatial coordinates and time series information of the actual abnormal deviation points, the location coordinates, amplitude values ​​and detection time of each actual abnormal deviation point are recorded to form the final set of abnormal deviation points.

5. The method according to claim 1, characterized in that, Methods for obtaining the location and extent of damage based on a set of abnormal deviation points include: Based on the spatial coordinate information of the highway infrastructure corresponding to the set of abnormal deviation points, calculate the Euclidean distance between each abnormal deviation point and construct a distance matrix. The K-means clustering algorithm is used to perform cluster analysis on outliers, and the cluster centers are determined based on the distance matrix to obtain multiple clusters of outliers. For each cluster of abnormal deviation points, the spatial boundary range of abnormal deviation points within the cluster is calculated. The spatial range of each cluster of abnormal deviation points is determined by the minimum bounding rectangle method, thus obtaining the damage area. The distribution density of abnormal deviation points in each damaged area is analyzed using density statistics methods. Based on the distribution density and the preset density threshold, the degree of damage in the damaged area is determined. A damage assessment report is generated based on the spatial location and extent of the damage.

6. A highway infrastructure health monitoring system based on an intelligent sensing network, the system being used to implement the method according to any one of claims 1-5, characterized in that, The system includes: a data acquisition and processing module, a temperature compensation module, a separation and rejection module, an anomaly acquisition module, and a damage localization module; The data acquisition and processing module is used to collect raw strain field data and temperature field data of highway infrastructure in real time based on pre-embedded sensors and perform preprocessing to obtain stable strain field sequence and stable temperature field sequence. The temperature compensation module is used to separate the low-frequency periodic components caused by ambient temperature fluctuations from the steady temperature field sequence, and combine them with the steady strain field sequence to obtain the temperature-compensated strain field sequence. The separation and elimination module is used to separate the reference drift component from the temperature-compensated strain field sequence and eliminate the influence of vehicle load to obtain the residual strain field sequence. The anomaly acquisition module is used to obtain a set of abnormal deviation points based on the residual strain field sequence; The damage localization module is used to obtain the location and degree of damage of the damaged area based on the set of abnormal deviation points.

7. The system according to claim 6, characterized in that, The temperature compensation module includes: a frequency domain conversion unit, a low-frequency filtering unit, a reconstruction unit, a removal unit, a linear regression unit, and a compensation unit; The frequency domain transformation unit is used to perform frequency domain transformation on the stationary temperature field sequence through Fourier transform to obtain the temperature field frequency spectrum; The low-frequency screening unit is used to determine the peak frequency in the low-frequency range in the temperature field frequency spectrum. The component corresponding to the peak frequency is the low-frequency periodic component generated by the ambient temperature fluctuation. The reconstruction unit is used to construct a linear combination of sine and cosine waves based on the peak frequency of the low-frequency periodic components, and reconstruct the low-frequency periodic component sequence. The removal unit is used to subtract the reconstructed low-frequency periodic component sequence from the stationary temperature field sequence to obtain the residual temperature field sequence after removing ambient temperature fluctuations. The linear regression unit is used to establish the linear relationship between the residual temperature field sequence and the stationary strain field sequence using least squares regression, and to determine the regression coefficients; The compensation unit is used to calculate the temperature compensation value through the regression coefficient and the residual temperature field sequence, and to subtract the temperature compensation value from the stationary strain field sequence to obtain the temperature-compensated strain field sequence.

8. The system according to claim 6, characterized in that, The separation and elimination module includes: a filtering unit, a separation unit, an alignment unit, a spectrum amplitude unit, a correlation analysis unit, and an elimination unit; The filtering unit is used to process the temperature-compensated strain field sequence with a low-pass filter to obtain the slowly changing reference drift component. The separation unit is used to subtract the slowly changing reference drift component from the temperature-compensated strain field sequence to obtain the initial residual strain field sequence after deducting the reference drift. The alignment unit is used to acquire vehicle load record data and establish a synchronous temporal relationship between the initial residual strain field sequence and the vehicle load record data through a timestamp alignment algorithm to obtain time-matched data pairs. The spectral amplitude unit is used to convert the initial residual strain field sequence and vehicle load record in the time-matched data pair to the frequency domain using fast Fourier transform, thereby obtaining their respective spectral amplitude response information. The correlation analysis unit is used to calculate the correlation coefficient between the initial residual strain field and the vehicle load on each frequency component based on the spectral amplitude response information. If the correlation coefficient exceeds the preset threshold, it is determined that the frequency component is affected by the vehicle load, and the load-related frequency component set is obtained. The elimination unit is used to remove the spectral components corresponding to the set of frequency components related to vehicle load from the initial residual strain field spectrum using a frequency domain filtering method, and then use inverse fast Fourier transform to obtain the residual strain field sequence after removing the influence factors of vehicle load.

9. The system according to claim 6, characterized in that, The anomaly detection module includes: a partitioning unit, an amplitude detection unit, a candidate unit, a first clustering unit, a verification unit, and an acquisition unit; The partitioning unit is used to divide the residual strain field sequence into regions using the sliding window technique to obtain the region partitioning results. The amplitude acquisition unit is used to perform amplitude statistics on the strain field values ​​of each sub-region in the regional division result through amplitude calculation, calculate the maximum amplitude value of the strain field values ​​in each sub-region, and obtain regional amplitude distribution data. Candidate units are used to mark a region as a candidate anomalous region if the maximum amplitude value of a sub-region in the regional amplitude distribution data exceeds a preset threshold, thus forming a list of candidate anomalous regions. The first clustering unit is used to perform spatial clustering analysis on the candidate abnormal region list using clustering algorithms. Based on the spatial adjacency relationship and amplitude similarity between regions, the spatial distribution pattern of abnormal deviation points is determined to obtain the clustering analysis results. The verification unit is used to perform secondary screening of the clustering analysis results through a preset anomaly verification module. If the clustered region combination meets the continuity condition and the amplitude exceeds the limit for a preset duration, then the region combination is confirmed as a real anomaly deviation point. The acquisition unit is used to record the location coordinates, amplitude values, and detection time of each real abnormal deviation point based on the spatial coordinates and time series information of the real abnormal deviation points, forming the final set of abnormal deviation points.

10. The system according to claim 6, characterized in that, The damage localization module includes: a distance calculation unit, a second clustering unit, a loss localization unit, a severity judgment unit, and a report generation unit; The distance calculation unit is used to calculate the Euclidean distance between each abnormal deviation point based on the spatial coordinate information of the highway infrastructure corresponding to the abnormal deviation point set, and to construct a distance matrix. The second clustering unit is used to perform cluster analysis on outlier points using the K-means clustering algorithm, determine the cluster centers based on the distance matrix, and obtain multiple clusters of outlier points. The loss localization unit is used to calculate the spatial boundary range of abnormal deviation points within each abnormal deviation point cluster, and to determine the spatial range of each abnormal deviation point cluster by using the minimum bounding rectangle method to obtain the damage area. The degree judgment unit is used to analyze the distribution density of abnormal deviation points in each damaged area using density statistics methods, and determine the degree of damage in the damaged area based on the distribution density and the preset density threshold. The report generation unit is used to generate a damage area assessment report based on the spatial location and degree of damage of the damaged area.

Images