Bridge multi-parameter fusion health state monitoring method
By using distributed sensor networks and multi-dimensional data processing technology, the problems of multi-dimensional parameter integration and environmental interference removal in bridge monitoring were solved, enabling accurate identification and risk assessment of bridge structural damage and generating a dynamic risk concern list.
Patent Information
- Application Number
- CN202610456947.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-08
- Publication Date
- 2026-11-06
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This application relates to the field of bridge structure monitoring technology, and in particular to a method for monitoring the health status of bridges using multi-parameter fusion. Background Technology
[0002] In modern infrastructure construction, bridges, as a core component of transportation networks, directly impact socio-economic operations and public safety through their safety and stability. With increasing bridge service life and the influence of environmental factors, monitoring the health status of bridge structures has become a crucial research area. Ensuring the long-term safe operation of bridges requires not only real-time monitoring of their operational status but also accurate prediction and timely intervention of potential risks, which is of irreplaceable value in maintaining public safety.
[0003] However, current bridge monitoring methods often struggle to comprehensively capture changes in structural condition when faced with complex environments and multiple influencing factors. Existing methods primarily focus on the acquisition and analysis of single-type monitoring parameters, lacking the synergistic integration of multi-dimensional parameters. Furthermore, they fail to effectively handle cross-timescale interference in time-series data, leading to overlapping interference from monitoring data across different time periods and distorted extraction of parameter change features. Simultaneously, environmental factors, especially temperature, can easily confuse structural responses with actual damage signals, and existing interference removal techniques lack specificity, reducing the reliability of monitoring data.
[0004] Furthermore, existing technologies rely heavily on threshold comparisons of single parameters to determine abnormal states, lacking in-depth data mining and cross-parameter correlation verification. This makes it difficult to distinguish between occasional fluctuations of a single parameter and collaborative anomalies of multiple parameters, easily leading to misjudgments and missed judgments. Summary of the Invention
[0005] To improve the accuracy of bridge health monitoring data, this application provides a method for monitoring the health status of bridges using multi-parameter fusion.
[0006] This application provides a method for monitoring the health status of bridges using multi-parameter fusion, which employs the following technical solution: including the following steps: S1. Data Acquisition and Preprocessing: Multi-dimensional monitoring data during bridge operation is collected through a distributed sensor network. After preprocessing the collected data, an initial monitoring dataset is constructed. S2. Hierarchical Time Series Processing: The time series data of each parameter in the initial monitoring dataset are processed hierarchically to separate independent data subsets from different time periods, thereby obtaining hierarchical time series data and eliminating cross-time scale data interference; S3. Parameter Feature Analysis and Interference Removal: Calculate the variation characteristics of each parameter based on the hierarchical time series data, extract fluctuation patterns, remove environmental interference factors, and determine the temporal evolution trend of each parameter; S4. Anomaly detection and correlation analysis: If the parameter fluctuation pattern exceeds the preset threshold range, the corresponding time series data is deeply mined to locate the abnormal segment, and combined with cross-parameter correlation analysis of other parameters, a comprehensive anomaly judgment result is obtained; S5. Anomaly Priority Ranking and Feature Extraction: Based on the comprehensive anomaly judgment results, a priority evaluation model is constructed to extract features from high-priority anomaly segments and obtain detailed anomaly distribution information; S6. Damage Signal Identification and Location: Analyze and identify structural damage signals within abnormal fluctuation ranges, eliminate interference errors, and determine the specific location of potential damage; S7. Risk Classification and Dynamic Assessment: Classify potential damage risks and generate a list of concerns, identify abnormal trend characteristics by combining historical monitoring data, and complete the dynamic assessment of the bridge's operational status.
[0007] Optionally, in step S1, the multi-dimensional monitoring data at least covers strain parameters, vibration parameters, displacement parameters, and temperature parameters; the preprocessing operation includes noise reduction and invalid value removal; the sensor placement covers key stress-bearing parts such as the mid-span of the main beam, supports, and the top of the piers.
[0008] Optionally, in step S2, the hierarchical processing uses wavelet decomposition or empirical mode decomposition to separate independent data subsets according to a preset time length.
[0009] Optionally, in step S3, the environmental interference factors are removed, specifically by constructing a correlation model based on temperature parameters to remove temperature interference. This process includes: S31. Extract the synchronous temperature parameter sequence and the structural parameter sequence from the historical monitoring database, wherein the structural parameter sequence includes the strain parameter sequence, vibration parameter sequence and displacement parameter sequence; S32. Using the multiple linear regression method, with the temperature parameter sequence as the independent variable and each structural parameter sequence as the dependent variable, a correlation model between the initial temperature and the structural parameters is constructed. S33. Calculate the residual sequence between the actual observed values of each structural parameter and the predicted values of the initial temperature and structural parameter correlation model, and calculate the standard deviation of the residual sequence; S34. According to the bridge structure type and design standard calibration, if the standard deviation of the residual sequence exceeds the preset threshold, it is determined that the structural parameter is significantly affected by non-temperature factors, and the independent analysis process is initiated; if it does not exceed the preset threshold, the residual sequence is used as the pure structural response data after temperature stripping, and the model dynamic calibration is completed and the temperature contribution deviation is quantified.
[0010] Optionally, the statistical features include mean, variance, fluctuation range, skewness, and kurtosis. The model dynamic calibration uses the least squares method to update the intercept and slope parameters. The calibration cycle is linked to the sensor acquisition frequency. When the acquisition frequency is 10-100Hz, the calibration cycle is set to 1 hour. Independent analysis procedures initiated for structural parameters significantly affected by non-temperature factors include: S341. Simultaneously acquire the time-series data of temperature change, vibration parameters, and displacement parameters corresponding to the structural parameters to form a multi-dimensional raw dataset. Preprocess the dataset using Z-score normalization or min-max normalization to eliminate dimensional differences. S342. Based on Pearson correlation coefficient or mutual information entropy, extract the correlation characteristics between temperature change and vibration parameters, and between temperature change and displacement parameters, construct the parameter interaction matrix under multi-factor disturbance, and quantify the mutual influence intensity between parameters. S343. Apply the analytic hierarchy process or entropy weight method to decompose the superposition effect of each parameter, calculate the contribution weights of temperature change, vibration parameters, and displacement parameters, identify parameters whose weights exceed the preset threshold as the main sources of deviation, extract their time series data segments, and locate abnormal fluctuation intervals using the sliding window method.
[0011] Optionally, in step S4, the deep mining adopts wavelet packet decomposition technology, the anomaly detection algorithm is the isolated forest algorithm or the local anomaly factor algorithm, and the preset threshold is calibrated by the bridge structure design specifications and historical monitoring data; the cross-parameter correlation analysis is based on the change characteristics of each parameter within the same time window to judge the cooperative correlation in order to eliminate the interference of random fluctuations of a single parameter.
[0012] Optionally, the structural damage signal identification in step S6 specifically includes: S61. Analyze the data within the abnormal fluctuation range, and combine the linkage response characteristics of vibration parameters and displacement parameters to identify whether there are periodic abnormal modes that match the load period or sudden abnormal modes that change instantaneously. S62. If the aforementioned periodic abnormality pattern or sudden abnormality pattern exists, it is determined that the data contains structural damage signals, excluding transient environmental interference and equipment errors; S63. Integrate the multi-factor interference characteristics of temperature change, vibration parameters, and displacement parameters, construct a comprehensive evaluation model based on random forest, calculate the correlation coefficient between the main sources of deviation and structural damage, determine the degree of correlation between the two, and generate a distribution map of the contribution of each factor to the damage signal.
[0013] Optionally, the input features of the comprehensive evaluation model may also include the natural frequency and damping ratio of the bridge structure, which are extracted from vibration parameters through a modal recognition algorithm.
[0014] Optionally, the determination of the specific location of the potential damage in step S6 includes: S64. Extract time series segments of abnormal data containing structural damage signals, process them in segments according to a set time window, and extract local data features within each time window, including peak values, valley values, and frequency centroids. S65. Analyze the correlation between local data characteristics and environmental factors, including wind speed and traffic load, screen out abnormal segments that are not significantly related to environmental factors, and match structural damage failure modes in combination with sensor fault judgment criteria. S66. Wavelet packet decomposition is used to perform deep feature decomposition on the successfully matched abnormal fragment data, extract key change points, and mark the potential damage distribution range; retrieve the historical database of structural damage, compare the correspondence between the distribution range and the known damage location, use K-means clustering algorithm or support vector machine classification algorithm to process the overlapping area data, and combine the sensor placement location to determine the final damage location result.
[0015] Optionally, the risk classification and risk concern list generation in step S7 specifically includes: S71. Extract long-term trend records from multi-dimensional monitoring data, process them in segments according to monthly and quarterly time intervals, and calculate the trend characteristic value for each interval. The trend characteristic value includes the rate of change, fluctuation amplitude, and peak mean. S72. Construct a comparison matrix between damage signals and environmental interference, filter out the intervals where there are significant differences between the signals and interference, analyze the correlation between damage signal fluctuation data and bridge safety status within the intervals, and determine whether it exceeds the design allowable range. S73. Combining the spatial distribution of abnormal fluctuations with the damage location results, potential risk points are marked, and the support vector machine algorithm is used to classify the risk points and determine their category. The category classification includes structural damage, environmental disturbance, and equipment failure. S74. Calculate the degree of correlation between risk points and persistent interference factors, classify risk points with a correlation degree higher than a preset threshold as key sources of deviation to be monitored, and generate a final list of concerns including risk location, category, and degree of correlation.
[0016] Optional.
[0017] In summary, this application includes the following beneficial technical effects: This invention collects multi-dimensional parameters such as strain, vibration, displacement, and temperature through a distributed sensor network. The sensors are precisely positioned at key stress-bearing locations such as the mid-span of the main beam, supports, and the tops of piers, achieving comprehensive coverage of bridge structure monitoring data and solving the problem of incomplete information from single-parameter monitoring. Furthermore, preprocessing techniques such as noise reduction and invalid value removal ensure the validity of the collected data. By relying on wavelet decomposition or empirical mode decomposition to process time series data in layers, and separating independent data subsets according to preset time lengths, the interference of data across time scales can be effectively eliminated, and the problem of data superposition distortion in different time periods can be solved; making the extraction of fluctuation patterns and time evolution trends of various parameters more accurate. Wavelet packet decomposition technology is used to deeply mine abnormal data, and detection algorithms adapted to bridge scenarios, such as isolated forest and local anomaly factor, are combined. The preset threshold is calibrated by both standard and historical data to improve the sensitivity of anomaly detection. By constructing a comprehensive evaluation model based on damage pattern recognition, environmental and equipment error elimination, and random forest, the system achieves accurate identification of structural damage signals. By combining wavelet packet feature decomposition, K-means / support vector machine clustering algorithms, and historical data comparison, along with sensor placement to assist in localization, a standardized damage localization process is formed, solving the problem of ambiguous localization in existing methods. A risk classification, list generation, and trend analysis management system is constructed. By extracting trend features for the current month or quarter, classifying risks using support vector machines, and quantifying the degree of correlation, a targeted list of concerns is generated. At the same time, abnormal trend features are identified by combining historical monitoring data, thereby upgrading from single monitoring to dynamic tracking and prediction. Attached Figure Description
[0018] Figure 1 This is a flowchart illustrating a bridge multi-parameter fusion health status monitoring method according to this application; Figure 2 This is a comparison diagram before and after calibration of the temperature and structural parameter correlation model of the multi-parameter fusion health status monitoring method for bridges proposed in this application; Figure 3 This application presents a method for monitoring the health status of bridges using a multi-parameter fusion approach, which yields damage location results and risk classification maps. Detailed Implementation
[0019] The following is in conjunction with the appendix Figure 1-3 This application will be described in further detail.
[0020] This application discloses a method for monitoring the health status of bridges using multi-parameter fusion. For example... Figure 1 , 2 As shown in Figure 3, the following steps are included: S1. Data Acquisition and Preprocessing: Multi-dimensional monitoring data during bridge operation is collected through a distributed sensor network. After preprocessing the collected data, an initial monitoring dataset is constructed.
[0021] S2. Hierarchical time series processing: The time series data of each parameter in the initial monitoring dataset are processed hierarchically to separate independent data subsets of different time periods, thereby obtaining hierarchical time series data and eliminating cross-time scale data interference.
[0022] S3. Parameter feature analysis and interference removal: Calculate the variation characteristics of each parameter based on the hierarchical time series data, extract fluctuation patterns, remove environmental interference factors, and determine the temporal evolution trend of each parameter.
[0023] S4. Anomaly detection and correlation analysis: If the parameter fluctuation pattern exceeds the preset threshold range, perform in-depth mining of the corresponding time series data to locate the abnormal segment, and combine it with cross-parameter correlation analysis of other parameters to obtain a comprehensive anomaly judgment result.
[0024] S5. Anomaly Priority Ranking and Feature Extraction: Based on the comprehensive anomaly judgment results, a priority evaluation model is constructed to extract features from high-priority anomaly segments and obtain detailed anomaly distribution information.
[0025] S6. Damage signal identification and localization: Analyze and identify structural damage signals within abnormal fluctuation ranges, eliminate interference errors, and determine the specific location of potential damage.
[0026] S7. Risk Classification and Dynamic Assessment: Classify potential damage risks and generate a list of concerns; identify abnormal trend characteristics by combining historical monitoring data; and complete the dynamic assessment of the bridge's operational status.
[0027] The detailed features of step S1 are as follows: the multi-dimensional monitoring data at least covers strain parameters, vibration parameters, displacement parameters and temperature parameters; the preprocessing operation includes noise reduction and invalid value removal; the sensor placement covers key stress-bearing parts of the bridge main beam mid-span, supports and the top of the piers.
[0028] The refinement features of step S2: The hierarchical processing adopts wavelet decomposition or empirical mode decomposition to separate independent data subsets according to a preset time length.
[0029] Step S3 involves a detailed step of constructing a correlation model based on temperature parameters to remove temperature interference. S31. Extract the synchronous temperature parameter sequence and the structural parameter sequence from the historical monitoring database, wherein the structural parameter sequence includes the strain parameter sequence, vibration parameter sequence and displacement parameter sequence; S32. Using the multiple linear regression method, with the temperature parameter sequence as the independent variable and each structural parameter sequence as the dependent variable, a correlation model between the initial temperature and the structural parameters is constructed. S33. Calculate the residual sequence between the actual observed values of each structural parameter and the predicted values of the initial temperature and structural parameter correlation model, and calculate the standard deviation of the residual sequence; S34. According to the bridge structure type and design standard calibration, if the standard deviation of the residual sequence exceeds the preset threshold, it is determined that the structural parameter is significantly affected by non-temperature factors, and the independent analysis process is initiated; if it does not exceed the preset threshold, the residual sequence is used as the pure structural response data after temperature stripping, and the model dynamic calibration is completed and the temperature contribution deviation is quantified.
[0030] The statistical characteristics, model calibration and independent analysis process are detailed: The statistical characteristics include mean, variance, fluctuation range, skewness and kurtosis. The model dynamic calibration uses the least squares method to update the intercept and slope parameters. The calibration cycle is linked to the sensor acquisition frequency. When the acquisition frequency is 10-100Hz, the calibration cycle is set to 1 hour. Independent analysis procedures initiated for structural parameters significantly affected by non-temperature factors include: S341. Simultaneously acquire the time-series data of temperature change, vibration parameters, and displacement parameters corresponding to the structural parameters to form a multi-dimensional raw dataset. Preprocess the dataset using Z-score normalization or min-max normalization to eliminate dimensional differences. S342. Based on Pearson correlation coefficient or mutual information entropy, extract the correlation characteristics between temperature change and vibration parameters, and between temperature change and displacement parameters, construct the parameter interaction matrix under multi-factor disturbance, and quantify the mutual influence intensity between parameters. S343. Apply the analytic hierarchy process or entropy weight method to decompose the superposition effect of each parameter, calculate the contribution weights of temperature change, vibration parameters, and displacement parameters, identify parameters whose weights exceed the preset threshold as the main sources of deviation, extract their time series data segments, and locate abnormal fluctuation intervals using the sliding window method.
[0031] The detailed features of step S4 are as follows: the in-depth mining adopts wavelet packet decomposition technology, the anomaly detection algorithm is the isolated forest algorithm or the local anomaly factor algorithm, and the preset threshold is calibrated by the bridge structure design specifications and historical monitoring data; the cross-parameter correlation analysis is based on the change characteristics of each parameter within the same time window to judge the cooperative correlation in order to eliminate the interference of random fluctuations of a single parameter.
[0032] Detailed steps for structural damage signal identification in step S6: S61. Analyze the data within the abnormal fluctuation range, and combine the linkage response characteristics of vibration parameters and displacement parameters to identify whether there are periodic abnormal modes that match the load period or sudden abnormal modes that change instantaneously. S62. If the aforementioned periodic abnormality pattern or sudden abnormality pattern exists, it is determined that the data contains structural damage signals, excluding transient environmental interference and equipment errors; S63. Integrate the multi-factor interference characteristics of temperature change, vibration parameters, and displacement parameters, construct a comprehensive evaluation model based on random forest, calculate the correlation coefficient between the main sources of deviation and structural damage, determine the degree of correlation between the two, and generate a distribution map of the contribution of each factor to the damage signal.
[0033] Supplementary features of the comprehensive evaluation model: The input features of the comprehensive evaluation model also include the natural frequency and damping ratio of the bridge structure, which are extracted from the vibration parameters through a modal recognition algorithm.
[0034] Detailed steps for determining the specific location of potential damage in step S6: S64. Extract time series segments of abnormal data containing structural damage signals, process them in segments according to a set time window, and extract local data features within each time window, including peak values, valley values, and frequency centroids. S65. Analyze the correlation between local data characteristics and environmental factors, including wind speed and traffic load, screen out abnormal segments that are not significantly related to environmental factors, and match structural damage failure modes in combination with sensor fault judgment criteria. S66. Wavelet packet decomposition is used to perform deep feature decomposition on the successfully matched abnormal fragment data, extract key change points, and mark the potential damage distribution range; retrieve the historical database of structural damage, compare the correspondence between the distribution range and the known damage location, use K-means clustering algorithm or support vector machine classification algorithm to process the overlapping area data, and combine the sensor placement location to determine the final damage location result.
[0035] Detailed steps for "Risk Classification and Risk Concern List Generation" in Step S7: S71. Extract long-term trend records from multi-dimensional monitoring data, process them in segments according to monthly and quarterly time intervals, and calculate the trend characteristic value for each interval. The trend characteristic value includes the rate of change, fluctuation amplitude, and peak mean. S72. Construct a comparison matrix between damage signals and environmental interference, filter out the intervals where there are significant differences between the signals and interference, analyze the correlation between damage signal fluctuation data and bridge safety status within the intervals, and determine whether it exceeds the design allowable range. S73. Combining the spatial distribution of abnormal fluctuations with the damage location results, potential risk points are marked, and the support vector machine algorithm is used to classify the risk points and determine their category. The category classification includes structural damage, environmental disturbance, and equipment failure. S74. Calculate the degree of correlation between risk points and persistent interference factors, classify risk points with a correlation degree higher than a preset threshold as key sources of deviation to be monitored, and generate a final list of concerns including risk location, category, and degree of correlation.
[0036] The implementation principle of the bridge multi-parameter fusion health status monitoring method in this application embodiment is as follows: Data Acquisition and Preprocessing: A distributed sensor network is activated to simultaneously collect multi-dimensional monitoring data on strain, vibration, displacement, and temperature during bridge operation at a frequency of 50Hz. The data acquisition unit performs wavelet denoising on the raw data to remove invalid values and fixed biases caused by poor sensor contact. The processed valid data is then integrated to construct an initial monitoring dataset, which is simultaneously uploaded to the embedded processor and a cloud database backup to ensure data traceability.
[0037] Hierarchical time series processing: The embedded processor extracts time series data of each parameter from the initial monitoring dataset, and performs hierarchical processing according to a preset time length using wavelet decomposition. This separates independent data subsets from different time periods to obtain hierarchical time series data, effectively eliminating the mutual interference between short-term traffic load fluctuations and long-term environmental erosion data, and providing accurate time series data for subsequent feature analysis.
[0038] Parameter Feature Analysis and Interference Removal: The processor calculates the statistical characteristics and rates of change of the layered time series data, such as mean, variance, fluctuation amplitude, skewness, and kurtosis, and extracts the fluctuation patterns of each parameter. Simultaneously, the temperature interference removal process is initiated: the temperature parameter series and strain, vibration, and displacement structural parameter series of the same period in the past year are extracted from the cloud database, and an initial temperature-structural parameter correlation model is constructed using the multiple linear regression method. The residual series and standard deviation of each structural parameter between the actual observed values and the model prediction values are calculated. Among them, the standard deviation of the residual of the strain parameter at the mid-span of the main beam exceeds the preset threshold calibrated according to the bridge design standard, indicating that it is significantly affected by non-temperature factors, and the independent analysis process is initiated. The standard deviation of the residuals of the other parameters is lower than the threshold. The residual series is used as the pure structural response data after temperature removal. The least squares method is used to update the intercept and slope parameters of the model to complete the dynamic calibration of the model and quantify the contribution deviation of temperature to each parameter.
[0039] Independent analysis process execution: For the strain parameters at mid-span of the main beam, corresponding temperature changes, vibration, and displacement time-series data are acquired simultaneously to form a multi-dimensional raw dataset. The Z-score standardization method is used for preprocessing to eliminate dimensional differences. Based on the Pearson correlation coefficient, the correlation characteristics between temperature and vibration, and temperature and displacement are extracted, and a multi-factor interference parameter interaction matrix is constructed to quantify the mutual influence intensity between parameters. The analytic hierarchy process is applied to decompose the superposition effect, and the vibration parameter contribution weight is calculated to be 0.71, which exceeds the preset weight threshold of 0.6. The vibration parameter is determined to be the main source of deviation. The corresponding time-series data segments are extracted, and two abnormal fluctuation intervals are located using the sliding window method.
[0040] Anomaly detection and correlation analysis: The isolated forest algorithm was used to deeply mine the time series data of strain parameters at the mid-span of the main beam. Combined with wavelet packet decomposition technology, the specific time points and continuous data segments of the two abnormal fluctuation intervals were accurately located, and the potential anomalies were initially judged. Further, combined with the characteristics of vibration and displacement parameter changes within the same time window, it was found that the vibration parameters and strain parameters in the two intervals showed significant coordinated fluctuations and cross-parameter coordinated correlation. The interference of random fluctuations of single parameters was eliminated, and the comprehensive anomaly judgment result was obtained: two valid anomaly intervals.
[0041] Anomaly Priority Ranking and Feature Extraction: A priority evaluation model is constructed based on the anomaly's impact range and fluctuation amplitude, classifying both valid anomaly intervals as high priority; temporal features, including peak and trough values, as well as frequency domain features, including the frequency centroid, are extracted within the intervals to obtain detailed information such as the spatial location, fluctuation duration, and amplitude of the anomaly distribution, providing support for subsequent damage identification.
[0042] Damage Signal Identification and Location: Two high-priority anomaly intervals were analyzed. Data analysis revealed a periodic anomaly pattern matching the traffic load cycle in one interval, and a sudden, transient anomaly pattern in the other. Both were determined to contain structural damage signals, excluding transient environmental interference and equipment errors. A random forest comprehensive evaluation model was constructed by integrating the interference features of multiple factors including temperature changes, vibration, and displacement. Modal recognition algorithms were used to extract the bridge's natural frequency and damping ratio from vibration parameters as supplementary input features. The correlation coefficient between vibration parameters and damage was calculated to be 0.83, generating a distribution map of the contribution of each factor to the damage signal. Time-series data segments from the two anomaly intervals were extracted and analyzed according to 5... The data was processed in segments with time windows of a few minutes, and local data features such as peak values, valley values, and frequency centroids were extracted within each window. The correlation between local features and wind speed and traffic load was analyzed, and abnormal segments with no significant correlation to environmental factors were screened out. Combined with sensor fault judgment criteria, structural damage fault modes were matched. Wavelet packet decomposition was used to perform deep feature decomposition on the matched segments, extract key change points, and mark the potential damage distribution range. The historical damage database in the cloud was retrieved, and a record of a slight crack was found on the left side of the main beam mid-span. The overlapping area data was processed by K-means clustering algorithm, and combined with the sensor placement, the final damage location result was determined: the crack on the left side of the main beam mid-span has expanded, and the left support has slight settlement.
[0043] Risk Classification and Dynamic Assessment: Risk management and dynamic assessment are conducted on the located damage. Long-term trend records of multi-dimensional monitoring data from the past 6 months are extracted, processed by monthly and quarterly segments, and trend characteristic values such as change rate, fluctuation amplitude, and peak mean are calculated for each interval. A comparison matrix of damage signals and environmental interference is constructed, and intervals with significant differences between signals and interference are screened out. Analysis shows that the expansion of the crack in the mid-span of the main beam has a gradual trend, and the settlement of the left support is an occasional anomaly, neither of which exceeds the allowable range of the bridge design. Combining the spatial distribution of abnormal fluctuations with the damage location results, two potential risk points are marked and classified into the structural damage category using the support vector machine algorithm. The correlation between the risk points and continuous traffic load is calculated. The correlation degree of the expansion of the crack in the mid-span of the main beam is 0.86, which exceeds the preset threshold of 0.7, and is listed as a key source of deviation for attention. A final list of concerns is generated, clearly marking the risk locations: left side of the mid-span of the main beam and above the left support; the category is structural damage. At the same time, the data is compared and analyzed with historical monitoring data in the cloud to complete the dynamic assessment of the bridge's operating status.
[0044] Subsequent maintenance follow-up: According to the risk concern list, the maintenance personnel grouted and reinforced the crack on the left side of the main beam mid-span, and adjusted and sealed the left support for settlement. After reinforcement and maintenance, the device continued to be monitored routinely according to the above procedures. After dynamic calibration of the model, it was shown that the strain, vibration and displacement parameters had all returned to the normal range, the crack expansion trend had stopped, and the support settlement had stabilized. The two risk points were removed from the key concern list, and the regular monitoring frequency was restored, forming a closed loop of maintenance.
[0045] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. A method for monitoring the operational status of a bridge, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing: Multi-dimensional monitoring data during bridge operation is collected through a distributed sensor network. After preprocessing the collected data, an initial monitoring dataset is constructed. S2. Hierarchical Time Series Processing: The time series data of each parameter in the initial monitoring dataset are processed hierarchically to separate independent data subsets from different time periods, thereby obtaining hierarchical time series data and eliminating cross-time scale data interference; S3. Parameter Feature Analysis and Interference Removal: Calculate the variation characteristics of each parameter based on the hierarchical time series data, extract fluctuation patterns, remove environmental interference factors, and determine the temporal evolution trend of each parameter; S4. Anomaly detection and correlation analysis: If the parameter fluctuation pattern exceeds the preset threshold range, the corresponding time series data is deeply mined to locate the abnormal segment, and combined with cross-parameter correlation analysis of other parameters, a comprehensive anomaly judgment result is obtained; S5. Anomaly Priority Ranking and Feature Extraction: Based on the comprehensive anomaly judgment results, a priority evaluation model is constructed to extract features from high-priority anomaly segments and obtain detailed anomaly distribution information; S6. Damage Signal Identification and Location: Analyze and identify structural damage signals within abnormal fluctuation ranges, eliminate interference errors, and determine the specific location of potential damage; S7. Risk Classification and Dynamic Assessment: Classify potential damage risks and generate a list of concerns, identify abnormal trend characteristics by combining historical monitoring data, and complete the dynamic assessment of the bridge's operational status.
2. The bridge operation status monitoring method according to claim 1, characterized in that, In step S1, the multi-dimensional monitoring data covers at least strain parameters, vibration parameters, displacement parameters, and temperature parameters; the preprocessing operation includes noise reduction and invalid value removal; the sensor placement covers key stress-bearing parts such as the mid-span of the main beam, supports, and the top of the piers.
3. The bridge operation status monitoring method according to claim 1, characterized in that, In step S2, the hierarchical processing uses wavelet decomposition or empirical mode decomposition to separate independent data subsets according to a preset time length.
4. The bridge operation status monitoring method according to claim 1, characterized in that, In step S3, the environmental interference factors are removed, specifically by constructing a correlation model based on temperature parameters to remove temperature interference. This process includes: S31. Extract the synchronous temperature parameter sequence and the structural parameter sequence from the historical monitoring database, wherein the structural parameter sequence includes the strain parameter sequence, vibration parameter sequence and displacement parameter sequence; S32. Using the multiple linear regression method, with the temperature parameter sequence as the independent variable and each structural parameter sequence as the dependent variable, a correlation model between the initial temperature and the structural parameters is constructed. S33. Calculate the residual sequence between the actual observed values of each structural parameter and the predicted values of the initial temperature and structural parameter correlation model, and calculate the standard deviation of the residual sequence; S34. According to the bridge structure type and design standard calibration, if the standard deviation of the residual sequence exceeds the preset threshold, it is determined that the structural parameter is significantly affected by non-temperature factors, and the independent analysis process is initiated; if it does not exceed the preset threshold, the residual sequence is used as the pure structural response data after temperature stripping, and the model dynamic calibration is completed and the temperature contribution deviation is quantified.
5. The bridge operation status monitoring method according to claim 4, characterized in that, The statistical features include mean, variance, fluctuation range, skewness, and kurtosis. The model dynamic calibration uses the least squares method to update the intercept and slope parameters. The calibration cycle is linked to the sensor acquisition frequency. When the acquisition frequency is 10-100Hz, the calibration cycle is set to 1 hour. Independent analysis procedures initiated for structural parameters significantly affected by non-temperature factors include: S341. Simultaneously acquire the time-series data of temperature change, vibration parameters, and displacement parameters corresponding to the structural parameters to form a multi-dimensional raw dataset. Preprocess the dataset using Z-score normalization or min-max normalization to eliminate dimensional differences. S342. Based on Pearson correlation coefficient or mutual information entropy, extract the correlation characteristics between temperature change and vibration parameters, and between temperature change and displacement parameters, construct the parameter interaction matrix under multi-factor disturbance, and quantify the mutual influence intensity between parameters. S343. Apply the analytic hierarchy process or entropy weight method to decompose the superposition effect of each parameter, calculate the contribution weights of temperature change, vibration parameters, and displacement parameters, identify parameters whose weights exceed the preset threshold as the main sources of deviation, extract their time series data segments, and locate abnormal fluctuation intervals using the sliding window method.
6. The bridge operation status monitoring method according to claim 1, characterized in that, In step S4, the deep mining adopts wavelet packet decomposition technology, the anomaly detection algorithm is the isolated forest algorithm or the local anomaly factor algorithm, and the preset threshold is calibrated by the bridge structure design specifications and historical monitoring data. Cross-parameter correlation analysis is based on the change characteristics of each parameter within the same time window to determine the co-correlation and eliminate the interference of random fluctuations of a single parameter.
7. The bridge operation status monitoring method according to claim 1, characterized in that, The structural damage signal identification in step S6 specifically includes: S61. Analyze the data within the abnormal fluctuation range, and combine the linkage response characteristics of vibration parameters and displacement parameters to identify whether there are periodic abnormal modes that match the load period or sudden abnormal modes that change instantaneously. S62. If the aforementioned periodic abnormality pattern or sudden abnormality pattern exists, it is determined that the data contains structural damage signals, excluding transient environmental interference and equipment errors; S63. Integrate the multi-factor interference characteristics of temperature change, vibration parameters, and displacement parameters, construct a comprehensive evaluation model based on random forest, calculate the correlation coefficient between the main sources of deviation and structural damage, determine the degree of correlation between the two, and generate a distribution map of the contribution of each factor to the damage signal.
8. A bridge operation status monitoring method according to claim 7, characterized in that, The input features of the comprehensive evaluation model also include the natural frequency and damping ratio of the bridge structure, which are extracted from vibration parameters through a modal recognition algorithm.
9. A bridge operation status monitoring method according to claim 7, characterized in that, The determination of the specific location of the potential damage in step S6 specifically includes: S64. Extract time series segments of abnormal data containing structural damage signals, process them in segments according to a set time window, and extract local data features within each time window, including peak values, valley values, and frequency centroids. S65. Analyze the correlation between local data characteristics and environmental factors, including wind speed and traffic load, screen out abnormal segments that are not significantly related to environmental factors, and match structural damage failure modes in combination with sensor fault judgment criteria. S66. Wavelet packet decomposition is used to perform deep feature decomposition on the successfully matched abnormal fragment data, extract key change points, and mark the potential damage distribution range; retrieve the historical database of structural damage, compare the correspondence between the distribution range and the known damage location, use K-means clustering algorithm or support vector machine classification algorithm to process the overlapping area data, and combine the sensor placement location to determine the final damage location result.
10. A bridge operation status monitoring method according to claim 9, characterized in that, The risk classification and risk concern list generation in step S7 specifically includes: S71. Extract long-term trend records from multi-dimensional monitoring data, process them in segments according to monthly and quarterly time intervals, and calculate the trend characteristic value for each interval. The trend characteristic value includes the rate of change, fluctuation amplitude, and peak mean. S72. Construct a comparison matrix between damage signals and environmental interference, filter out the intervals where there are significant differences between the signals and interference, analyze the correlation between damage signal fluctuation data and bridge safety status within the intervals, and determine whether it exceeds the design allowable range. S73. Combining the spatial distribution of abnormal fluctuations with the damage location results, potential risk points are marked, and the support vector machine algorithm is used to classify the risk points and determine their category. The category classification includes structural damage, environmental disturbance, and equipment failure. S74. Calculate the degree of correlation between risk points and persistent interference factors, classify risk points with a correlation degree higher than a preset threshold as key sources of deviation to be monitored, and generate a final list of concerns including risk location, category, and degree of correlation.