Bridge structure monitoring method, system, and medium based on sensor data

Abstract

This application provides a method, system, and medium for bridge structure monitoring based on sensor data. The method includes: marking measurement point locations to establish a data acquisition system to collect and validate structured raw datasets; decomposing the obtained valid raw datasets; processing the noise coefficients; fusing the obtained normalized datasets with temporal and spatial correlations; extracting and combining features from the obtained global representation datasets; comparing and calculating thresholds to obtain multi-dimensional temporal feature sets; obtaining a comprehensive safety index and classifying safety levels; and finally obtaining a structured safety assessment report. Through synchronous data acquisition from multiple sensors, robust wavelet and sliding window anomaly cleaning, hierarchical weighted spatiotemporal fusion of component importance, extraction of structural response sensitive features, and comprehensive safety assessment using multi-level thresholds and standard indicators, a closed-loop data flow is achieved throughout the process, accurately identifying potential safety hazards in the flower box structure and improving the completeness of monitoring data processing and the reliability of assessment results.

Description

Technical Field

[0001] This application relates to the field of bridge structural health monitoring technology, and more specifically, to bridge structural monitoring methods, systems, and media based on sensor data. Background Technology

[0002] Urban elevated flower boxes, fixed to the top of bridge crash barriers, are subjected to wind loads, vehicle vibrations, temperature fluctuations, and rainwater erosion over long periods. This makes them prone to problems such as loosening of the U-shaped grooves, aging and embrittlement of the PP flower pots, degradation of the soil blocks, corrosion of steel components, and detachment of irrigation pipes, directly threatening the safety of vehicles and pedestrians beneath the bridge. Existing bridge monitoring technologies primarily focus on the main structure, such as the main beams, towers, and supports. Dedicated monitoring methods for lightweight ancillary structures like flower boxes are lacking. Conventional data processing often employs single filtering or simple threshold judgments, failing to establish a complete data flow of "collection—cleaning—fusion—feature analysis—evaluation." Furthermore, the results of previous processing steps are often unusable as input for subsequent steps, leading to data link breaks and resulting in missed anomalies, distorted evaluations, and delayed early warnings. Summary of the Invention

[0003] The purpose of this application is to provide a bridge structure monitoring method, system, and medium based on sensor data, which can realize a closed-loop data flow throughout the process, accurately identify safety hazards such as loosening, aging, corrosion, and deformation of the flower box structure, and improve the integrity of monitoring data processing and the reliability of evaluation results.

[0004] This application also provides a bridge structure monitoring method based on sensor data, including the following steps: Mark the locations of the measurement points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets; The effective original dataset is decomposed to obtain noise coefficients, which are then processed to obtain a threshold. The noise coefficients are then subjected to inverse wavelet transform to obtain the denoised time series signal, which is then completed and integrated to obtain a normalized dataset. Based on the normalized dataset, temporal and spatial correlation fusion is performed to obtain a unified feature dataset of the overall structural state, and then weighted fusion is performed to obtain a global representation dataset. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; Threshold comparisons and calculations are performed based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report.

[0005] Optionally, in the bridge structure monitoring method based on sensor data described in this application embodiment, the step of marking the measurement point locations according to the structural drawings, building a synchronous data acquisition system to collect a structured raw dataset for verification, and obtaining a valid raw dataset includes: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

[0006] Optionally, in the bridge structure monitoring method based on sensor data described in this application embodiment, the step of decomposing the effective original dataset to obtain noise coefficients for processing, obtaining a threshold-processed noise coefficient, performing inverse wavelet transform, and then obtaining a denoised time-series signal for completion and integration to obtain a normalized dataset includes: Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

[0007] Optionally, in the bridge structure monitoring method based on sensor data described in this application embodiment, the step of performing temporal and spatial correlation fusion based on the normalized dataset to obtain a unified feature dataset of the overall structural state and performing weighted fusion to obtain a global representation dataset includes: The normalized dataset is fused in the time domain by a preset algorithm model to obtain integrated characterization data for a single measurement point. Spatial correlation fusion is performed based on the integrated characterization data of the single measurement point to obtain a unified special dataset of the overall structural state.

[0008] The global representation dataset is obtained by weighted fusion of the unified feature dataset of the overall structure state.

[0009] Optionally, in the bridge structure monitoring method based on sensor data described in this application embodiment, the step of extracting and combining features based on the global representation dataset to obtain a multi-dimensional time-series feature set includes: Based on the global representation dataset, feature extraction is performed to obtain stress amplitude features; Feature extraction is performed based on the global representation dataset to obtain the peak feature of horizontal displacement. Feature extraction is performed based on the global representation dataset to obtain the vibration dominant frequency features; Feature extraction is performed based on the global representation dataset to obtain corrosion rate features; Feature extraction is performed based on the global representation dataset to obtain the tilt angle change rate feature; A multidimensional time series feature set is obtained by combining the stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics.

[0010] Optionally, in the bridge structure monitoring method based on sensor data described in the embodiments of this application, the step of comparing and calculating thresholds according to the multi-dimensional time-series feature set to obtain a comprehensive safety index and classify safety levels, and finally obtaining a structured safety assessment report, includes: Obtain technical specifications and structural design parameters, and construct a multi-level threshold system; Based on the multidimensional time-series feature set, threshold comparison and calculation are performed to obtain the safe membership degree; The comprehensive security index is calculated based on the security membership degree. Security levels are classified based on the comprehensive security index, and a structured security assessment report is obtained.

[0011] Secondly, embodiments of this application provide a bridge structure monitoring system based on sensor data. The system includes a memory and a processor. The memory includes a program for a bridge structure monitoring method based on sensor data. When the program for the bridge structure monitoring method based on sensor data is executed by the processor, it implements the following steps: Mark the locations of the measurement points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets; The effective original dataset is decomposed to obtain noise coefficients, which are then processed to obtain a threshold. The noise coefficients are then subjected to inverse wavelet transform to obtain the denoised time series signal, which is then completed and integrated to obtain a normalized dataset. Based on the normalized dataset, temporal and spatial correlation fusion is performed to obtain a unified feature dataset of the overall structural state, and then weighted fusion is performed to obtain a global representation dataset. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; Threshold comparisons and calculations are performed based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report.

[0012] Optionally, in the bridge structure monitoring system based on sensor data described in this application embodiment, the step of marking the measurement point locations according to the structural drawings, building a synchronous data acquisition system to collect structured raw datasets for verification, and obtaining valid raw datasets includes: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

[0013] Optionally, in the bridge structure monitoring system based on sensor data described in this application embodiment, the step of decomposing the effective original dataset to obtain noise coefficients, processing them to obtain a threshold, performing inverse wavelet transform on the noise coefficients, and then completing and integrating the denoised time-series signal to obtain a normalized dataset includes: Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

[0014] Thirdly, embodiments of this application also provide a computer-readable storage medium, which includes a bridge structure monitoring method program based on sensor data. When the bridge structure monitoring method program based on sensor data is executed by a processor, it implements the steps of the bridge structure monitoring method based on sensor data as described in any of the above claims.

[0015] As can be seen from the above, the bridge structure monitoring method, system, and medium based on sensor data provided in this application, through marking the location of measuring points according to structural drawings, building a synchronous data acquisition system to collect structured raw datasets for verification, obtaining valid raw datasets, decomposing the valid raw datasets to obtain noise coefficients for processing, obtaining threshold processing noise coefficients for inverse wavelet transform, and then obtaining denoised time-series signals for completion and integration to obtain normalized datasets, performing temporal and spatial correlation fusion based on the normalized datasets to obtain a unified feature dataset of the overall structural state, and performing weighted fusion to obtain a global representation. The dataset is used to extract and combine features from the global representation dataset to obtain a multi-dimensional temporal feature set. Threshold comparisons and calculations are performed based on the multi-dimensional temporal feature set to obtain a comprehensive safety index and classify safety levels. Finally, a structured safety assessment report is obtained. Through synchronous data acquisition from multiple sources of sensors, robust wavelet and sliding window anomaly cleaning, hierarchical weighted spatiotemporal fusion of component importance, extraction of structural response sensitive features, and comprehensive safety assessment based on multi-level thresholds and standard indicators, a closed-loop data flow is achieved throughout the process. This accurately identifies safety hazards such as loosening, aging, corrosion, and deformation of the flower box structure, improving the completeness of monitoring data processing and the reliability of assessment results.

[0016] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the bridge structure monitoring method based on sensor data provided in this application embodiment; Figure 2 A flowchart illustrating the multi-source sensor synchronous data acquisition process for a bridge structure monitoring method based on sensor data provided in this application embodiment; Figure 3 A high-level flowchart of a bridge structure monitoring method based on sensor data provided in this application embodiment. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating bridge structure monitoring based on sensor data in some embodiments of this application. This bridge structure monitoring method based on sensor data is used in terminal devices, such as mobile phones and computers. The bridge structure monitoring method based on sensor data includes the following steps: S11. Mark the locations of the measuring points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets. S12. Decompose the effective original dataset to obtain noise coefficients and process them to obtain a threshold. Perform inverse wavelet transform on the noise coefficients to obtain the denoised time series signal. Complete and integrate the signal to obtain a normalized dataset. S13. Perform temporal and spatial correlation fusion based on the normalized dataset to obtain a unified feature dataset of the overall structural state and perform weighted fusion to obtain a global representation dataset. S14. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; S15. Based on the multidimensional time-series feature set, threshold comparison and calculation are performed to obtain a comprehensive security index and classify security levels, and finally a structured security assessment report is obtained.

[0022] The process involves marking measurement points based on structural drawings, establishing a synchronous data acquisition system to collect and validate structured raw datasets, obtaining valid raw datasets, decomposing these datasets to obtain noise coefficients, processing them to obtain thresholds, performing inverse wavelet transforms on the noise coefficients, obtaining denoised time-series signals, completing and integrating them to obtain regularized datasets, performing temporal and spatial correlation fusion on the regularized datasets to obtain unified feature datasets of the overall structural state, performing weighted fusion to obtain global representation datasets, extracting and combining features from the global representation datasets to obtain multi-dimensional temporal feature sets, comparing and calculating thresholds based on the multi-dimensional temporal feature sets to obtain a comprehensive safety index and classify safety levels, and finally obtaining a structured safety assessment report. Through synchronous data acquisition from multi-source sensors, robust wavelet and sliding window anomaly cleaning, hierarchical weighted spatiotemporal fusion of component importance, extraction of structural response sensitive features, and comprehensive safety assessment using multi-level thresholds and standard indicators, a closed-loop data flow is achieved throughout the process. This accurately identifies safety hazards such as loosening, aging, corrosion, and deformation of the flower box structure, improving the completeness of monitoring data processing and the reliability of assessment results.

[0023] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the multi-source sensor synchronous data acquisition process of a bridge structure monitoring method based on sensor data in some embodiments of this application. According to an embodiment of the present invention, the step of marking the measurement point locations according to the structural drawings, building a synchronous data acquisition system to collect a structured raw dataset for verification, and obtaining a valid raw dataset specifically involves: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

[0024] In this process, based on the structural drawings of the elevated flower box, measuring points were marked one by one at the following locations: the stress-bearing section of the metal U-shaped channel, the welding and load-bearing nodes of the metal support, the contact points between the anchor bolts and nuts, the load-bearing area of ​​the flower pot container sidewall, and the pipe body and joints of the irrigation pipeline. These points were uniformly numbered and a measuring point ledger was established. A multi-channel synchronous data acquisition instrument was connected, with all sensor signals connected to the same acquisition host to avoid asynchronous errors from multiple devices. A hardware synchronous clock was enabled to achieve millisecond-level time alignment for all sensor channels. The sampling frequency was uniformly set to ≥10Hz, remaining constant throughout the entire process for continuous and uninterrupted acquisition, ensuring the integrity of the raw data. All sensors were triggered in parallel according to a unified clock to acquire structured raw datasets, simultaneously completing strain, vibration, and other data acquisition. The acquisition of tilt angle, corrosion, and temperature / humidity signals involves amplification, filtering, and AD conversion of analog signals, followed by direct parsing of digital signals. Obvious spikes and interference are removed without modifying the original physical quantities. Each raw data point is encapsulated in a fixed format, including timestamp, measurement point number, component type, sensor type, physical quantity value, and acquisition device number. Data is stored continuously in a time series. Synchronization checks are performed on the structured raw dataset, verifying the consistency of timestamps from each sensor, removing abnormal sampling points with excessive time differences, and performing integrity checks by verifying the number of data points by measurement point number to ensure no packet loss or missing channels. The principle of preserving original values ​​is upheld; no deep processing such as filtering, noise reduction, or conversion is performed, only obvious anomalies are marked, and the true acquisition form is retained. Valid raw datasets are obtained and stored in separate files by day / by structural component, named with the project name, location, and date, and indexed for easy traceability.

[0025] According to an embodiment of the present invention, the step of decomposing the effective original dataset to obtain noise coefficients, processing them to obtain a threshold-processed noise coefficient, performing inverse wavelet transform, and then obtaining a denoised time-series signal for completion and integration to obtain a normalized dataset is specifically as follows: Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

[0026] The effective raw datasets were sorted by measurement point number and timestamp to ensure temporal continuity and channel independence. Single-channel time series sequences were established for each sensor type, including strain, vibration, tilt, corrosion, and temperature / humidity sequences. The consistency of sampling intervals was verified, and the sampling frequency was confirmed to be stable at ≥10Hz, providing an equally spaced time series basis for wavelet decomposition and sliding window calculations. A standardized raw dataset was obtained. A three-level discrete wavelet decomposition was uniformly performed on the standardized raw dataset using a pre-set wavelet decomposition model, such as the db4 wavelet basis. Each single-channel time series signal was independently decomposed to obtain noise figures, including high-frequency detail coefficients at levels 1 to 3, corresponding to environmental noise such as wind-induced vibration, vehicle vibration, and electromagnetic interference, and low-frequency approximation coefficients at level 3, corresponding to the true structural response trend. These were obtained using a pre-set function model, such as soft thresholding. The function performs threshold shrinking on the noise figure, retaining effective weak signals and suppressing random and environmental vibration noise. The threshold calculation uses an unbiased risk estimation threshold, adaptively adapting to different sensor signal characteristics to obtain the threshold-processed noise figure. Based on the threshold-processed noise figure, an inverse wavelet transform is performed to reconstruct the denoised time series signal, retaining the original timestamps, measurement point numbers, component types, and physical quantity units. Only the physical quantities are replaced with denoised values, without changing the data structure. Calculations are performed based on the denoised time series signal using a fixed-length sliding window. The window length is adapted to the sampling frequency; 15-30 sampling points are recommended to ensure sufficient statistical samples are included within the window. The window slides point by point with a step size of 1, covering the entire time series to avoid boundary omissions. The median Me of the denoised sample sequence within the current window is calculated, and the absolute deviation of each point from the median is calculated. -Me|, takes the median of the deviation sequence, and obtains the median absolute deviation within the window: MAD = median(| -Me|), employing robust statistics to avoid the interference of outliers in the traditional Z-score, the calculation formula is: Where the constant c=1.4826 is the equivalent standard deviation conversion factor under normal distribution. A robust Z-score > 3 is preset as an outlier. Sampling points meeting the condition are marked and removed, with the physical quantity at that point set to null, while the timestamp and measurement point information are retained. The denoised time-series signal sequence is traversed to identify single-point and continuous short-segment missing data caused by outlier removal, transmission packet loss, or instantaneous sensor failure. Segments with continuous missing lengths exceeding 1 / 2 window length are marked as segments to be interpolated. Neighborhood weighted interpolation is used for completion, with 5-10 effective sampling points before and after the missing point as neighborhood samples. Weights are assigned according to time distance, with closer times having higher weights. The weight is inversely proportional to the time interval. The weighting calculation formula is as follows: , These are time points 0 and i, respectively, and the difference between them represents the time interval between the i-th sampling point and the 0th sampling point. For missing parts at the beginning and end of the sequence, one-sided neighborhood weighted interpolation is used. For slowly varying signals such as temperature, humidity, and tilt angle, the neighborhood range can be appropriately expanded. For rapidly varying signals such as vibration and strain, the neighborhood length is strictly limited to ensure local features. After completion, the temporal continuity is rechecked to ensure no null values ​​or jumps, and the sampling interval remains consistent. The denoised, anomaly-removed, and completed data from all measurement points and all sensor channels are integrated, maintaining the same structure as the effective original dataset: timestamp + measurement point number + component type + physical quantity value. This yields a normalized dataset with equal temporal intervals, no missing values, no outliers, and no significant environmental noise, which serves as the standard input for subsequent feature extraction and state assessment.

[0027] According to an embodiment of the present invention, the step of performing temporal and spatial correlation fusion based on the normalized dataset to obtain a unified feature dataset of the overall structural state and then performing weighted fusion to obtain a global representation dataset specifically involves: The normalized dataset is fused in the time domain by a preset algorithm model to obtain integrated characterization data for a single measurement point. Spatial correlation fusion is performed based on the integrated characterization data of the single measurement point to obtain a unified special dataset of the overall structural state.

[0028] The global representation dataset is obtained by weighted fusion of the unified feature dataset of the overall structure state.

[0029] This process involves extracting time-series monitoring data from multiple heterogeneous sensors deployed at the same measurement point from a normalized dataset. Using a unified timestamp as a benchmark, strict time-domain alignment is achieved to ensure synchronous matching of multiple physical quantities at the same time. A pre-defined algorithm, such as the Kalman filter, is used to process the multi-type sensor data at this measurement point. The structural response at this measurement point is taken as the system state variable, and the observations of each sensor are used as the measurement quantity. A linear discrete Kalman filter model is established, including a state equation describing the dynamic change of the structural response over time and an observation equation characterizing the observation relationship of each sensor to the same state. The observation noise covariance is set according to the sensor type, prioritizing high stability and accuracy. Higher sensor quality results in higher confidence levels. Kalman filtering recursion is performed, and through prediction and update cycles, the optimal fusion estimate at the measurement point level is obtained for fusion. The fusion result eliminates single-sensor random errors, jumps, and local anomalies, obtaining integrated characterization data for single measurement points. The original timestamp and component affiliation are retained. Monitoring data of similar measurement points between different bridge piers are extracted from the integrated characterization data of single measurement points. Measurement points of similar components are divided into spatially adjacent measurement point groups, such as adjacent supports, U-shaped channels in the same span, and anchor bolts in the same row, according to pier, layout section, and installation location. The Pearson correlation coefficient is calculated for the integrated characterization data of each group of adjacent measurement points. The calculation formula is: in For the target measurement point time series, The time series of adjacent measurement points The spatial correlation coefficient between two points is compared with a preset spatial correlation threshold, such as 0.7 to 0.9. If it is lower than the threshold, it is determined to be a spatially abnormal point. The deviation is corrected by using the temporal trend of adjacent highly correlated measurement points. Single-point mutations and local distortions are suppressed by spatial weighted averaging. Under the constraint of spatial consistency, regional spatial fusion is completed for similar components to obtain a unified feature dataset of the overall structural state. According to the structural safety importance of the flower box, the unified feature dataset of the overall structural state is divided into metal U-shaped channels, metal supports, anchor bolts, flower pot containers, and irrigation. The pipeline consists of five substructures, each assigned a weight: the metal U-shaped channel (0.30), the metal support (0.25), the anchor bolts (0.20), the flowerpot container (0.15), and the irrigation pipe (0.10). These weights are normalized and summed to 1. An independent data subset is created for each substructure. The datasets for each substructure are aligned by timestamp to generate a full-point time-series data matrix for each substructure, ensuring a one-to-one correspondence between all measurement points at the same time. All measurement points within a substructure have equal weights. A weighted average of the weighted features of the overall substructure's state is calculated time-by-time using the following formula: in Let be the unified characteristic value of the overall structural state of the S-type substructure at time t. Let k be the internal weight of the measurement point. For the state characteristic value of measurement point k at time t, the unified state characteristic values ​​of the five substructures are linearly weighted and fused according to preset importance weights. The calculation formula is as follows: The weighted results are regularized and trend smoothed to form a temporal feature that can uniformly represent the overall safety status of the ancillary structures. A unified timestamp sequence is retained, and each record contains a timestamp, an overall status index, the status values ​​of each substructure, and a weight configuration identifier. The data format is kept continuous, equally spaced, and structured to obtain a global representation dataset after spatiotemporal fusion, which is used for subsequent status assessment, trend prediction, anomaly warning, and safety level determination.

[0030] According to an embodiment of the present invention, the step of extracting and combining features based on the global representation dataset to obtain a multidimensional temporal feature set specifically includes: Based on the global representation dataset, feature extraction is performed to obtain stress amplitude features; Feature extraction is performed based on the global representation dataset to obtain the peak feature of horizontal displacement. Feature extraction is performed based on the global representation dataset to obtain the vibration dominant frequency features; Feature extraction is performed based on the global representation dataset to obtain corrosion rate features; Feature extraction is performed based on the global representation dataset to obtain the tilt angle change rate feature; A multidimensional time series feature set is obtained by combining the stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics.

[0031] The data is split and recombined according to timestamp, component type, and sensor type to form dedicated time series sequences for each sensor. Data continuity and equal intervals are verified to ensure no missing values ​​or outliers, meeting the requirements for feature extraction such as amplitude calculation, FFT transformation, and differential operations. A unified feature index is established, with each feature record corresponding one-to-one with the original timestamp to ensure time series traceability. Feature extraction is performed based on the global representation dataset. Using strain time series data as a foundation, combined with the component material's elastic modulus E and cross-sectional parameters, strain is converted into stress time series. A fixed-duration sliding window is used to traverse the stress sequence, calculating the maximum and minimum stress values ​​σ_max and σ_min within the window. The stress amplitude is defined as: A_σ = σ_max - σ_min. Stress amplitude features are obtained by outputting the timestamp, serving as the first dimension feature. Using tilt angle and vibration fusion data as input, a system is established... A simplified dynamic model of the structure is used. The structural rotation angle is converted from the tilt angle value, and the rigid body displacement component is calculated by combining the component height. The dynamic vibration displacement component is obtained by the second integral of the vibration acceleration. A sliding window statistical analysis is performed on the total horizontal displacement time series, and the maximum absolute displacement value within the window is extracted, which is the horizontal displacement peak value. The peak value is smoothed to eliminate local abrupt changes, and the horizontal displacement peak value feature is obtained as the second dimension feature. Equal-length vibration data segments are extracted, and a Hanning window is added to reduce spectral leakage. A Fast Fourier Transform (FFT) is performed on the signal to obtain the power spectral density (PSD). The main peak frequency with the highest energy in the spectrum is identified and defined as the first-order natural frequency, i.e., the vibration dominant frequency. The dominant frequency value is updated window by window to obtain a stationary and continuous dominant frequency feature sequence as the third dimension feature. Based on the corrosion depth time series output by the corrosion sensor, the central difference method is used to calculate the depth change per unit time, i.e., the corrosion rate. For corrosion rate, for The degree of corrosion over time, Let t be the degree of corrosion. For the time difference, Short-term fluctuations are low-pass filtered to preserve the true corrosion development trend and obtain corrosion rate characteristics as the fourth dimension. The time-domain first derivative of the tilt angle time series is directly performed to calculate the time-varying slope. The tilt angle change rate is calculated using a least-squares linear fitting method with adjacent multi-points to suppress high-frequency noise and obtain the tilt angle change rate characteristics as the fifth dimension. The stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics at the same moment are combined into a 5-dimensional feature vector. All feature vectors are arranged in time stamp order to obtain a multi-dimensional time series feature set. The timestamp, corresponding component information, and feature meaning labels are retained. The structure is regular, without redundancy, and can be directly used for pattern recognition and state assessment. The magnitude of each feature quantity is uniformly regularized to avoid excessive numerical differences affecting the subsequent model training and discrimination effect.

[0032] According to an embodiment of the present invention, the step of performing threshold comparison and calculation based on the multi-dimensional time-series feature set to obtain a comprehensive security index and classify security levels, and finally obtaining a structured security assessment report, specifically includes: Obtain technical specifications and structural design parameters, and construct a multi-level threshold system; Based on the multidimensional time-series feature set, threshold comparison and calculation are performed to obtain the safe membership degree; The comprehensive security index is calculated based on the security membership degree. Security levels are classified based on the comprehensive security index, and a structured security assessment report is obtained.

[0033] Based on the "Technical Standard for Urban Bridge Maintenance" (CJJ99-2017), the "Code for Design of Building Structures" (GB50009-2012), and structural design parameters, a multi-level threshold system was constructed, with the upper limit threshold for stress amplitude being [σ]. a ] max The upper limit threshold for the peak horizontal displacement is [d]. max The upper limit threshold for the rate of change of tilt angle is [θ']. max The lower limit threshold of the dominant vibration frequency is f min =0.85 × structural design fundamental frequency, the upper limit threshold for corrosion rate is [v ] max Thresholds are adjusted separately according to the component importance coefficient. Thresholds for key load-bearing components such as metal U-shaped channels, metal supports, and anchor bolts are strictly controlled. Five types of characteristic data, namely stress amplitude, peak horizontal displacement, vibration dominant frequency, corrosion rate, and tilt angle change rate, are read one by one according to the timestamp. For each time point, the thresholds of the five types of characteristics are compared and the safe membership interval [0,1] is calculated. The upper limit index includes stress amplitude, peak displacement, corrosion rate, and tilt angle change rate. The calculation formula is: The closer the indicator is to the threshold, the closer the safety membership degree is to 0. The lower limit indicator is the dominant vibration frequency, and the calculation formula is: When the main frequency is lower than 85% of the design fundamental frequency, stiffness is directly determined to be deteriorated, and the safety membership is set to 0. Get the number of years the structure has been used (T) By comparing the long-term performance degradation curves of similar elevated flower boxes, the degradation correction coefficient λ∈(0,1) was obtained. The longer the service life and the higher the degree of degradation, the smaller λ becomes, and the safety index is appropriately lowered to reflect the aging effect. The initial comprehensive safety index was calculated using a weighting method that matches the importance of the structure. The weight w is consistent with the importance hierarchy of components, with stress, displacement, and tilt angle indicators having higher weights. The final comprehensive safety index is then calculated by combining the decay correction coefficient λ. S∈[0,1], the higher the value, the better the structural safety. Safety levels are classified according to the comprehensive safety index S. The acceptable range is S≥0.90, all structural indicators are normal, there is no safety risk, routine monitoring is sufficient. The caution range is 0.75≤S<0.90, some indicators are close to the threshold, performance is slightly degraded, and closer monitoring is needed. The abnormal range is 0.60≤S<0.75, indicators are significantly out of limit or continuously deteriorating, there is a safety hazard, inspection and maintenance are recommended. The dangerous range is S<0.60, multiple indicators are seriously out of standard, structural stiffness, strength, or durability are significantly compromised. A decline indicates a risk of failure. Immediate warning and action are required to obtain a structured safety assessment report, which includes the assessment time, structure number, service life, comparison of measured values ​​of five characteristic indicators with standard thresholds, single indicator exceedances and their locations, calculation process and results of the comprehensive safety index, and safety level determination conclusion. The report outputs graded warning instructions, including a qualified green normal instruction, a caution yellow attention instruction, an abnormal orange alarm instruction, and a dangerous red emergency shutdown and handling instruction. The report includes trend curves and distribution maps of exceedance points to support maintenance decisions and traceability archiving.

[0034] Please refer to Figure 3 , Figure 3 This is a high-level flowchart of a bridge structure monitoring method based on sensor data in some embodiments of this application.

[0035] This invention also discloses a bridge structure monitoring system based on sensor data, including a memory and a processor. The memory includes a bridge structure monitoring method program based on sensor data. When the processor executes the bridge structure monitoring method program based on sensor data, it performs the following steps: Mark the locations of the measurement points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets; The effective original dataset is decomposed to obtain noise coefficients, which are then processed to obtain a threshold. The noise coefficients are then subjected to inverse wavelet transform to obtain the denoised time series signal, which is then completed and integrated to obtain a normalized dataset. Based on the normalized dataset, temporal and spatial correlation fusion is performed to obtain a unified feature dataset of the overall structural state, and then weighted fusion is performed to obtain a global representation dataset. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; Threshold comparisons and calculations are performed based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report.

[0036] The process involves marking measurement points based on structural drawings, establishing a synchronous data acquisition system to collect and validate structured raw datasets, obtaining valid raw datasets, decomposing these datasets to obtain noise coefficients, processing them to obtain thresholds, performing inverse wavelet transforms on the noise coefficients, obtaining denoised time-series signals, completing and integrating them to obtain regularized datasets, performing temporal and spatial correlation fusion on the regularized datasets to obtain unified feature datasets of the overall structural state, performing weighted fusion to obtain global representation datasets, extracting and combining features from the global representation datasets to obtain multi-dimensional temporal feature sets, comparing and calculating thresholds based on the multi-dimensional temporal feature sets to obtain a comprehensive safety index and classify safety levels, and finally obtaining a structured safety assessment report. Through synchronous data acquisition from multi-source sensors, robust wavelet and sliding window anomaly cleaning, hierarchical weighted spatiotemporal fusion of component importance, extraction of structural response sensitive features, and comprehensive safety assessment using multi-level thresholds and standard indicators, a closed-loop data flow is achieved throughout the process. This accurately identifies safety hazards such as loosening, aging, corrosion, and deformation of the flower box structure, improving the completeness of monitoring data processing and the reliability of assessment results.

[0037] According to an embodiment of the present invention, the step of marking the measurement point locations according to the structural drawings, building a synchronous data acquisition system to collect a structured raw dataset for verification, and obtaining a valid raw dataset specifically includes: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

[0038] In this process, based on the structural drawings of the elevated flower box, measuring points were marked one by one at the following locations: the stress-bearing section of the metal U-shaped channel, the welding and load-bearing nodes of the metal support, the contact points between the anchor bolts and nuts, the load-bearing area of ​​the flower pot container sidewall, and the pipe body and joints of the irrigation pipeline. These points were uniformly numbered and a measuring point ledger was established. A multi-channel synchronous data acquisition instrument was connected, with all sensor signals connected to the same acquisition host to avoid asynchronous errors from multiple devices. A hardware synchronous clock was enabled to achieve millisecond-level time alignment for all sensor channels. The sampling frequency was uniformly set to ≥10Hz, remaining constant throughout the entire process for continuous and uninterrupted acquisition, ensuring the integrity of the raw data. All sensors were triggered in parallel according to a unified clock to acquire structured raw datasets, simultaneously completing strain, vibration, and other data acquisition. The acquisition of tilt angle, corrosion, and temperature / humidity signals involves amplification, filtering, and AD conversion of analog signals, followed by direct parsing of digital signals. Obvious spikes and interference are removed without modifying the original physical quantities. Each raw data point is encapsulated in a fixed format, including timestamp, measurement point number, component type, sensor type, physical quantity value, and acquisition device number. Data is stored continuously in a time series. Synchronization checks are performed on the structured raw dataset, verifying the consistency of timestamps from each sensor, removing abnormal sampling points with excessive time differences, and performing integrity checks by verifying the number of data points by measurement point number to ensure no packet loss or missing channels. The principle of preserving original values ​​is upheld; no deep processing such as filtering, noise reduction, or conversion is performed, only obvious anomalies are marked, and the true acquisition form is retained. Valid raw datasets are obtained and stored in separate files by day / by structural component, named with the project name, location, and date, and indexed for easy traceability.

[0039] According to an embodiment of the present invention, the step of decomposing the effective original dataset to obtain noise coefficients, processing them to obtain a threshold-processed noise coefficient, performing inverse wavelet transform, and then obtaining a denoised time-series signal for completion and integration to obtain a normalized dataset is specifically as follows: Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

[0040] The effective raw datasets were sorted by measurement point number and timestamp to ensure temporal continuity and channel independence. Single-channel time series sequences were established for each sensor type, including strain, vibration, tilt, corrosion, and temperature / humidity sequences. The consistency of sampling intervals was verified, and the sampling frequency was confirmed to be stable at ≥10Hz, providing an equally spaced time series basis for wavelet decomposition and sliding window calculations. A standardized raw dataset was obtained. A three-level discrete wavelet decomposition was uniformly performed on the standardized raw dataset using a pre-set wavelet decomposition model, such as the db4 wavelet basis. Each single-channel time series signal was independently decomposed to obtain noise figures, including high-frequency detail coefficients at levels 1 to 3, corresponding to environmental noise such as wind-induced vibration, vehicle vibration, and electromagnetic interference, and low-frequency approximation coefficients at level 3, corresponding to the true structural response trend. These were obtained using a pre-set function model, such as soft thresholding. The function performs threshold shrinking on the noise figure, retaining effective weak signals and suppressing random and environmental vibration noise. The threshold calculation uses an unbiased risk estimation threshold, adaptively adapting to different sensor signal characteristics to obtain the threshold-processed noise figure. Based on the threshold-processed noise figure, an inverse wavelet transform is performed to reconstruct the denoised time series signal, retaining the original timestamps, measurement point numbers, component types, and physical quantity units. Only the physical quantities are replaced with denoised values, without changing the data structure. Calculations are performed based on the denoised time series signal using a fixed-length sliding window. The window length is adapted to the sampling frequency; 15-30 sampling points are recommended to ensure sufficient statistical samples are included within the window. The window slides point by point with a step size of 1, covering the entire time series to avoid boundary omissions. The median Me of the denoised sample sequence within the current window is calculated, and the absolute deviation of each point from the median is calculated. -Me|, takes the median of the deviation sequence, and obtains the median absolute deviation within the window: MAD = median(| -Me|), employing robust statistics to avoid the interference of outliers in the traditional Z-score, the calculation formula is: Where the constant c=1.4826 is the equivalent standard deviation conversion factor under normal distribution. A robust Z-score > 3 is preset as an outlier. Sampling points meeting the condition are marked and removed, with the physical quantity at that point set to null, while the timestamp and measurement point information are retained. The denoised time-series signal sequence is traversed to identify single-point and continuous short-segment missing data caused by outlier removal, transmission packet loss, or instantaneous sensor failure. Segments with continuous missing lengths exceeding 1 / 2 window length are marked as segments to be interpolated. Neighborhood weighted interpolation is used for completion, with 5-10 effective sampling points before and after the missing point as neighborhood samples. Weights are assigned according to time distance, with closer times having higher weights. The weight is inversely proportional to the time interval. The weighting calculation formula is as follows: , These are time points 0 and i, respectively, and the difference between them represents the time interval between the i-th sampling point and the 0th sampling point. For missing parts at the beginning and end of the sequence, one-sided neighborhood weighted interpolation is used. For slowly varying signals such as temperature, humidity, and tilt angle, the neighborhood range can be appropriately expanded. For rapidly varying signals such as vibration and strain, the neighborhood length is strictly limited to ensure local features. After completion, the temporal continuity is rechecked to ensure no null values ​​or jumps, and the sampling interval remains consistent. The denoised, anomaly-removed, and completed data from all measurement points and all sensor channels are integrated, maintaining the same structure as the effective original dataset: timestamp + measurement point number + component type + physical quantity value. This yields a normalized dataset with equal temporal intervals, no missing values, no outliers, and no significant environmental noise, which serves as the standard input for subsequent feature extraction and state assessment.

[0041] According to an embodiment of the present invention, the step of performing temporal and spatial correlation fusion based on the normalized dataset to obtain a unified feature dataset of the overall structural state and then performing weighted fusion to obtain a global representation dataset specifically involves: The normalized dataset is fused in the time domain by a preset algorithm model to obtain integrated characterization data for a single measurement point. Spatial correlation fusion is performed based on the integrated characterization data of the single measurement point to obtain a unified special dataset of the overall structural state.

[0042] The global representation dataset is obtained by weighted fusion of the unified feature dataset of the overall structure state.

[0043] This process involves extracting time-series monitoring data from multiple heterogeneous sensors deployed at the same measurement point from a normalized dataset. Using a unified timestamp as a benchmark, strict time-domain alignment is achieved to ensure synchronous matching of multiple physical quantities at the same time. A pre-defined algorithm, such as the Kalman filter, is used to process the multi-type sensor data at this measurement point. The structural response at this measurement point is taken as the system state variable, and the observations of each sensor are used as the measurement quantity. A linear discrete Kalman filter model is established, including a state equation describing the dynamic change of the structural response over time and an observation equation characterizing the observation relationship of each sensor to the same state. The observation noise covariance is set according to the sensor type, prioritizing high stability and accuracy. Higher sensor quality results in higher confidence levels. Kalman filtering recursion is performed, and through prediction and update cycles, the optimal fusion estimate at the measurement point level is obtained for fusion. The fusion result eliminates single-sensor random errors, jumps, and local anomalies, obtaining integrated characterization data for single measurement points. The original timestamp and component affiliation are retained. Monitoring data of similar measurement points between different bridge piers are extracted from the integrated characterization data of single measurement points. Measurement points of similar components are divided into spatially adjacent measurement point groups, such as adjacent supports, U-shaped channels in the same span, and anchor bolts in the same row, according to pier, layout section, and installation location. The Pearson correlation coefficient is calculated for the integrated characterization data of each group of adjacent measurement points. The calculation formula is: in For the target measurement point time series, The time series of adjacent measurement points The spatial correlation coefficient between two points is compared with a preset spatial correlation threshold, such as 0.7 to 0.9. If it is lower than the threshold, it is determined to be a spatially abnormal point. The deviation is corrected by using the temporal trend of adjacent highly correlated measurement points. Single-point mutations and local distortions are suppressed by spatial weighted averaging. Under the constraint of spatial consistency, regional spatial fusion is completed for similar components to obtain a unified feature dataset of the overall structural state. According to the structural safety importance of the flower box, the unified feature dataset of the overall structural state is divided into metal U-shaped channels, metal supports, anchor bolts, flower pot containers, and irrigation. The pipeline consists of five substructures, each assigned a weight: the metal U-shaped channel (0.30), the metal support (0.25), the anchor bolts (0.20), the flowerpot container (0.15), and the irrigation pipe (0.10). These weights are normalized and summed to 1. An independent data subset is created for each substructure. The datasets for each substructure are aligned by timestamp to generate a full-point time-series data matrix for each substructure, ensuring a one-to-one correspondence between all measurement points at the same time. All measurement points within a substructure have equal weights. A weighted average of the weighted features of the overall substructure's state is calculated time-by-time using the following formula: in Let be the unified characteristic value of the overall structural state of the S-type substructure at time t. Let k be the internal weight of the measurement point. For the state characteristic value of measurement point k at time t, the unified state characteristic values ​​of the five substructures are linearly weighted and fused according to preset importance weights. The calculation formula is as follows: The weighted results are regularized and trend smoothed to form a temporal feature that can uniformly represent the overall safety status of the ancillary structures. A unified timestamp sequence is retained, and each record contains a timestamp, an overall status index, the status values ​​of each substructure, and a weight configuration identifier. The data format is kept continuous, equally spaced, and structured to obtain a global representation dataset after spatiotemporal fusion, which is used for subsequent status assessment, trend prediction, anomaly warning, and safety level determination.

[0044] According to an embodiment of the present invention, the step of extracting and combining features based on the global representation dataset to obtain a multidimensional temporal feature set specifically includes: Based on the global representation dataset, feature extraction is performed to obtain stress amplitude features; Feature extraction is performed based on the global representation dataset to obtain the peak feature of horizontal displacement. Feature extraction is performed based on the global representation dataset to obtain the vibration dominant frequency features; Feature extraction is performed based on the global representation dataset to obtain corrosion rate features; Feature extraction is performed based on the global representation dataset to obtain the tilt angle change rate feature; A multidimensional time series feature set is obtained by combining the stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics.

[0045] The data is split and recombined according to timestamp, component type, and sensor type to form dedicated time series sequences for each sensor. Data continuity and equal intervals are verified to ensure no missing values ​​or outliers, meeting the requirements for feature extraction such as amplitude calculation, FFT transformation, and differential operations. A unified feature index is established, with each feature record corresponding one-to-one with the original timestamp to ensure time series traceability. Feature extraction is performed based on the global representation dataset. Using strain time series data as a foundation, combined with the component material's elastic modulus E and cross-sectional parameters, strain is converted into stress time series. A fixed-duration sliding window is used to traverse the stress sequence, calculating the maximum and minimum stress values ​​σ_max and σ_min within the window. The stress amplitude is defined as: A_σ = σ_max - σ_min. Stress amplitude features are obtained by outputting the timestamp, serving as the first dimension feature. Using tilt angle and vibration fusion data as input, a system is established... A simplified dynamic model of the structure is used. The structural rotation angle is converted from the tilt angle value, and the rigid body displacement component is calculated by combining the component height. The dynamic vibration displacement component is obtained by the second integral of the vibration acceleration. A sliding window statistical analysis is performed on the total horizontal displacement time series, and the maximum absolute displacement value within the window is extracted, which is the horizontal displacement peak value. The peak value is smoothed to eliminate local abrupt changes, and the horizontal displacement peak value feature is obtained as the second dimension feature. Equal-length vibration data segments are extracted, and a Hanning window is added to reduce spectral leakage. A Fast Fourier Transform (FFT) is performed on the signal to obtain the power spectral density (PSD). The main peak frequency with the highest energy in the spectrum is identified and defined as the first-order natural frequency, i.e., the vibration dominant frequency. The dominant frequency value is updated window by window to obtain a stationary and continuous dominant frequency feature sequence as the third dimension feature. Based on the corrosion depth time series output by the corrosion sensor, the central difference method is used to calculate the depth change per unit time, i.e., the corrosion rate. For corrosion rate, for The degree of corrosion over time, Let t be the degree of corrosion. For the time difference, Short-term fluctuations are low-pass filtered to preserve the true corrosion development trend and obtain corrosion rate characteristics as the fourth dimension. The time-domain first derivative of the tilt angle time series is directly performed to calculate the time-varying slope. The tilt angle change rate is calculated using a least-squares linear fitting method with adjacent multi-points to suppress high-frequency noise and obtain the tilt angle change rate characteristics as the fifth dimension. The stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics at the same moment are combined into a 5-dimensional feature vector. All feature vectors are arranged in time stamp order to obtain a multi-dimensional time series feature set. The timestamp, corresponding component information, and feature meaning labels are retained. The structure is regular, without redundancy, and can be directly used for pattern recognition and state assessment. The magnitude of each feature quantity is uniformly regularized to avoid excessive numerical differences affecting the subsequent model training and discrimination effect.

[0046] According to an embodiment of the present invention, the step of performing threshold comparison and calculation based on the multi-dimensional time-series feature set to obtain a comprehensive security index and classify security levels, and finally obtaining a structured security assessment report, specifically includes: Obtain technical specifications and structural design parameters, and construct a multi-level threshold system; Based on the multidimensional time-series feature set, threshold comparison and calculation are performed to obtain the safe membership degree; The comprehensive security index is calculated based on the security membership degree. Security levels are classified based on the comprehensive security index, and a structured security assessment report is obtained.

[0047] Based on the "Technical Standard for Urban Bridge Maintenance" (CJJ99-2017), the "Code for Design of Building Structures" (GB50009-2012), and structural design parameters, a multi-level threshold system was constructed, with the upper limit threshold for stress amplitude being [σ]. a ] max The upper limit threshold for the peak horizontal displacement is [d]. max The upper limit threshold for the rate of change of tilt angle is [θ']. max The lower limit threshold of the dominant vibration frequency is f min =0.85 × structural design fundamental frequency, the upper limit threshold for corrosion rate is [v ] max Thresholds are adjusted separately according to the component importance coefficient. Thresholds for key load-bearing components such as metal U-shaped channels, metal supports, and anchor bolts are strictly controlled. Five types of characteristic data, namely stress amplitude, peak horizontal displacement, vibration dominant frequency, corrosion rate, and tilt angle change rate, are read one by one according to the timestamp. For each time point, the thresholds of the five types of characteristics are compared and the safe membership interval [0,1] is calculated. The upper limit index includes stress amplitude, peak displacement, corrosion rate, and tilt angle change rate. The calculation formula is: The closer the indicator is to the threshold, the closer the safety membership degree is to 0. The lower limit indicator is the dominant vibration frequency, and the calculation formula is: When the main frequency is lower than 85% of the design fundamental frequency, stiffness is directly determined to be deteriorated, and the safety membership is set to 0. Get the number of years the structure has been used (T) By comparing the long-term performance degradation curves of similar elevated flower boxes, the degradation correction coefficient λ∈(0,1) was obtained. The longer the service life and the higher the degree of degradation, the smaller λ becomes, and the safety index is appropriately lowered to reflect the aging effect. The initial comprehensive safety index was calculated using a weighting method that matches the importance of the structure. The weight w is consistent with the importance hierarchy of components, with stress, displacement, and tilt angle indicators having higher weights. The final comprehensive safety index is then calculated by combining the decay correction coefficient λ. S∈[0,1], the higher the value, the better the structural safety. Safety levels are classified according to the comprehensive safety index S. The acceptable range is S≥0.90, all structural indicators are normal, there is no safety risk, routine monitoring is sufficient. The caution range is 0.75≤S<0.90, some indicators are close to the threshold, performance is slightly degraded, and closer monitoring is needed. The abnormal range is 0.60≤S<0.75, indicators are significantly out of limit or continuously deteriorating, there is a safety hazard, inspection and maintenance are recommended. The dangerous range is S<0.60, multiple indicators are seriously out of standard, structural stiffness, strength, or durability are significantly compromised. A decline indicates a risk of failure. Immediate warning and action are required to obtain a structured safety assessment report, which includes the assessment time, structure number, service life, comparison of measured values ​​of five characteristic indicators with standard thresholds, single indicator exceedances and their locations, calculation process and results of the comprehensive safety index, and safety level determination conclusion. The report outputs graded warning instructions, including a qualified green normal instruction, a caution yellow attention instruction, an abnormal orange alarm instruction, and a dangerous red emergency shutdown and handling instruction. The report includes trend curves and distribution maps of exceedance points to support maintenance decisions and traceability archiving.

[0048] A third aspect of the present invention provides a readable storage medium including a bridge structure monitoring method program based on sensor data, wherein when the bridge structure monitoring method program based on sensor data is executed by a processor, it implements the steps of the bridge structure monitoring method based on sensor data as described in any of the preceding claims.

[0049] This invention discloses a bridge structure monitoring method, system, and medium based on sensor data. The method involves marking measurement point locations according to structural drawings, establishing a synchronous data acquisition system to collect and validate structured raw datasets, obtaining valid raw datasets, decomposing these datasets to obtain noise coefficients, processing them to obtain a threshold, performing inverse wavelet transform on the noise coefficients, and then completing and integrating the denoised time-series signal to obtain a normalized dataset. Finally, the normalized dataset undergoes temporal and spatial correlation fusion to obtain a unified feature dataset of the overall structural state, which is then weighted and fused to obtain a global representation dataset. Feature extraction and combination are performed on the global representation dataset to obtain a multi-dimensional temporal feature set. Threshold comparison and calculation are then performed on the multi-dimensional temporal feature set to obtain a comprehensive safety index and classify safety levels. Finally, a structured safety assessment report is obtained. Through synchronous data acquisition from multiple sources of sensors, robust wavelet and sliding window anomaly cleaning, hierarchical weighted spatiotemporal fusion of component importance, extraction of structural response sensitive features, and comprehensive safety assessment based on multi-level thresholds and standard indicators, a closed-loop data flow is achieved throughout the process. This accurately identifies safety hazards such as loosening, aging, corrosion, and deformation of the flower box structure, improving the completeness of monitoring data processing and the reliability of assessment results.

[0050] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0051] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0052] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0053] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0054] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A method for bridge structure monitoring based on sensor data, characterized in that, Includes the following steps: Mark the locations of the measurement points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets; The effective original dataset is decomposed to obtain noise coefficients, which are then processed to obtain a threshold. The noise coefficients are then subjected to inverse wavelet transform to obtain the denoised time series signal, which is then completed and integrated to obtain a normalized dataset. Based on the normalized dataset, temporal and spatial correlation fusion is performed to obtain a unified feature dataset of the overall structural state, and then weighted fusion is performed to obtain a global representation dataset. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; Threshold comparisons and calculations are performed based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report.

2. The method of claim 1, wherein, The process involves marking the locations of measurement points according to the structural drawings, building a synchronous data acquisition system to collect structured raw datasets for verification, and obtaining valid raw datasets, including: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

3. The method of claim 2, wherein, The process involves decomposing the effective original dataset to obtain noise coefficients, processing them to obtain a threshold, performing inverse wavelet transform on the noise coefficients, and then completing and integrating the denoised time-series signal to obtain a normalized dataset. Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

4. The sensor data based bridge structure monitoring method of claim 1, wherein, The step involves performing temporal and spatial correlation fusion based on the normalized dataset to obtain a unified feature dataset of the overall structural state, followed by weighted fusion to obtain a global representation dataset, including: The normalized dataset is fused in the time domain by a preset algorithm model to obtain integrated characterization data for a single measurement point. Based on the integrated characterization data of the single measurement point, spatial correlation and fusion are performed to obtain a unified special dataset of the overall structural state. The global representation dataset is obtained by weighted fusion of the unified feature dataset of the overall structure state.

5. The bridge structure monitoring method based on sensor data according to claim 1, characterized in that, The step of extracting and combining features based on the global representation dataset to obtain a multidimensional temporal feature set includes: Based on the global representation dataset, feature extraction is performed to obtain stress amplitude features; Feature extraction is performed based on the global representation dataset to obtain the peak feature of horizontal displacement. Feature extraction is performed based on the global representation dataset to obtain the vibration dominant frequency features; Feature extraction is performed based on the global representation dataset to obtain corrosion rate features; Feature extraction is performed based on the global representation dataset to obtain the tilt angle change rate feature; A multidimensional time series feature set is obtained by combining the stress amplitude characteristics, horizontal displacement peak characteristics, vibration dominant frequency characteristics, corrosion rate characteristics, and tilt angle change rate characteristics.

6. The bridge structure monitoring method based on sensor data according to claim 1, characterized in that, The process involves threshold comparison and calculation based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report, including: Obtain technical specifications and structural design parameters, and construct a multi-level threshold system; Based on the multidimensional time-series feature set, threshold comparison and calculation are performed to obtain the safe membership degree; The comprehensive security index is calculated based on the security membership degree. Security levels are classified based on the comprehensive security index, and a structured security assessment report is obtained.

7. A bridge structure monitoring system based on sensor data, characterized in that, The system includes a memory and a processor. The memory contains a program for a bridge structure monitoring method based on sensor data. When the program for the bridge structure monitoring method based on sensor data is executed by the processor, it performs the following steps: Mark the locations of the measurement points according to the structural drawings, build a synchronous data acquisition system to collect structured raw datasets for verification, and obtain valid raw datasets; The effective original dataset is decomposed to obtain noise coefficients, which are then processed to obtain a threshold. The noise coefficients are then subjected to inverse wavelet transform to obtain the denoised time series signal, which is then completed and integrated to obtain a normalized dataset. Based on the normalized dataset, temporal and spatial correlation fusion is performed to obtain a unified feature dataset of the overall structural state, and then weighted fusion is performed to obtain a global representation dataset. Based on the global representation dataset, feature extraction and combination are performed to obtain a multidimensional temporal feature set; Threshold comparisons and calculations are performed based on the multidimensional time-series feature set to obtain a comprehensive security index and classify security levels, ultimately resulting in a structured security assessment report.

8. The bridge structure monitoring system based on sensor data according to claim 7, characterized in that, The process involves marking the locations of measurement points according to the structural drawings, building a synchronous data acquisition system to collect structured raw datasets for verification, and obtaining valid raw datasets, including: Mark the locations of the measuring points according to the structural drawings and build a synchronous data acquisition system; Collect structured raw datasets, including timestamps, measurement point numbers, component types, sensor types, physical quantity values, and acquisition device numbers; The structured original dataset is validated to obtain a valid original dataset.

9. The bridge structure monitoring system based on sensor data according to claim 8, characterized in that, The process involves decomposing the effective original dataset to obtain noise coefficients, processing them to obtain a threshold, performing inverse wavelet transform on the processed noise coefficients, and then completing and integrating the denoised time-series signal to obtain a normalized dataset. Preprocess the valid original dataset to obtain a standardized original dataset; The standardized original dataset is decomposed using a preset wavelet decomposition model to obtain noise coefficients, including high-frequency detail coefficients and low-frequency approximation coefficients. The noise coefficients are processed using a preset function model to obtain threshold-processed noise coefficients; Based on the threshold, the noise coefficient is processed by inverse wavelet transform to obtain the denoised time sequence signal; The absolute deviation of the median within the window is calculated based on the denoised time-series signal. Calculate the robust Z-score based on the absolute deviation of the median within the window and perform anomaly detection to obtain marked and discarded sampling points; The denoised time-series signal is completed and integrated to obtain a normalized dataset.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a bridge structure monitoring method program based on sensor data. When the bridge structure monitoring method program based on sensor data is executed by a processor, it implements the steps of the bridge structure monitoring method based on sensor data as described in any one of claims 1 to 6.

Images