Bridge safety monitoring system based on sensor data

Abstract

The invention relates to the technical field of safety monitoring, in particular to a bridge safety monitoring system based on sensor data. The method comprises the steps that a sensor measuring unit monitors key physical parameters of a bridge in real time; the data acquisition and transmission unit reads sensor data and transmits the read sensor data; the data processing and analysis unit extracts key features reflecting the health state of the bridge in the sensor data based on a multi-modal physical perception feature fusion method, identifies an abnormal mode in the sensor data through a support vector machine model, and evaluates the remaining life of the bridge based on a carbonization-corrosion-crack closed-loop feedback control rule; and the information management unit stores the analysis result of the data processing and analysis unit and displays the monitored bridge physical parameters. According to the design of the invention, a multi-modal physical perception feature fusion method is introduced, and adversarial learning, graph wavelet transformation and a three-dimensional convolutional neural network are combined, so that high-precision feature extraction of bridge structure response is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of safety monitoring, and more specifically, to a bridge safety monitoring system based on sensor data. Background Art

[0002] With the increase in traffic volume and the growth of service life, infrastructure such as bridges is gradually facing the risks of aging and damage. However, traditional manual inspection methods are difficult to achieve real-time and comprehensive monitoring of the bridge's health status. Especially under complex natural environmental conditions, such as the influence of adverse weather and vibration, accurately separating the structural response signal from environmental noise has become a major challenge. Traditional monitoring methods often struggle to effectively distinguish between the two, thus affecting the accuracy of monitoring data. At the same time, with the continuous progress of sensor technology, the amount of data obtained has increased explosively. How to efficiently process this massive amount of data and extract useful information from it to evaluate the bridge's condition has become an urgent problem to be solved. In addition, most traditional life prediction models are based on static or linear assumptions and cannot fully reflect the true degradation process of the bridge structure over time, especially failing to consider non-linear degradation effects caused by factors such as carbonation and corrosion. Therefore, a bridge safety monitoring system based on sensor data is designed. Summary of the Invention

[0003] The purpose of the present invention is to provide a bridge safety monitoring system based on sensor data to solve the problems raised in the above background art, namely, that traditional manual inspection cannot achieve real-time and comprehensive monitoring, environmental noise interference affects data accuracy, it is difficult to efficiently process and extract features from massive sensor data, and traditional life prediction models cannot accurately reflect the non-linear degradation process of the structure.

[0004] To achieve the above object, the present invention provides a bridge safety monitoring system based on sensor data, including: A sensor measurement unit that continuously monitors the key physical parameters of the bridge; A data acquisition and transmission unit that reads sensor data and transmits the read sensor data; A data processing and analysis unit that extracts key features reflecting the bridge's health status from sensor data based on a multi-modal physical perception feature fusion method, identifies abnormal patterns in sensor data through a support vector machine (SVM) model, and evaluates the remaining life of the bridge based on a carbonation-corrosion-crack closed-loop feedback control rule; An information management unit that stores the analysis results of the data processing and analysis unit and displays the monitored bridge physical parameters.

[0005] As a further improvement of this technical solution, the data acquisition and transmission unit includes a data reading module and a data transmission module; Among them, the data reading module reads sensor data through a connected sensor, and preliminarily sorts and packages the read sensor data; The data transmission module sends the packaged sensor data to the server where the data processing and analysis unit is located by means of wireless transmission.

[0006] As a further improvement of this technical solution, the data processing and analysis unit includes a data feature extraction module and a data anomaly analysis module; Among them, the data feature extraction module receives the preprocessed sensor data, and extracts key features reflecting the bridge health status based on the multi-modal physical perception feature fusion method; The data anomaly analysis module, based on the extracted key features, identifies abnormal data patterns through a support vector machine model, and evaluates the remaining life of the bridge using the carbonization-corrosion-crack closed-loop feedback control rule based on the extracted key features and abnormal data patterns.

[0007] As a further improvement of this technical solution, the extraction of key features reflecting the bridge health status based on the multi-modal physical perception feature fusion method includes the following steps: S1.1. Receive the preprocessed multi-dimensional time series sensor data transmitted from the data acquisition and transmission unit, use the preprocessed multi-dimensional time series sensor data as the input of the bridge structural dynamics model, generate corresponding structural response prediction values under given loads and boundary conditions, and use the structural response prediction values as the physical benchmark, calculate the physical residual between the actual structural response measurement value and the structural response prediction value, apply the adversarial learning method to separate environmental noise from the structural response, and generate a structural response feature vector and environmental features; S1.2. Perform a fast Fourier transform on the physical residual signal to obtain the frequency domain features of the physical residual signal, and align the frequency components of the physical residual signal with the main frequency characteristics of the bridge structure through weighted adjustment, construct an optimized frequency band division of the sensor graph structure, and extract the main frequency amplitude and frequency band energy anomaly index of the aligned frequency; S1.3. Based on the sensor deployment topology structure, construct a sensor physical relationship graph, perform spatial decomposition on the structural response features separated in step S1.1 and the frequency domain features in step S1.2 through graph wavelet transform, and calculate the sub-graph energy distribution entropy and cross-sensor frequency response consistency; S1.4. Integrate the time domain, frequency domain, and spatial features extracted in step S1.2 and step S1.3, construct a three-dimensional feature tensor, and use three-dimensional convolution operations to fuse the time domain, frequency domain, and spatial features, and output the fused high-dimensional feature tensor; S1.5. Screen the initial feature vectors using the mutual information method based on the fused high-dimensional feature tensor, and perform standardization processing on each screened feature vector; S1.6. Combine the standardized feature vectors into the final feature vector.

[0008] As a further improvement of this technical solution, in S1.1, for the physical residual between the actual structural response measurement value and the structural response prediction value, the adversarial learning method is applied to separate the environmental noise and the structural response, and generate the structural response feature vector and the environmental features, including the following steps: S1.11. Build a physically-driven structural response model based on the bridge structure parameters, input the current load and boundary conditions, and output the theoretical predicted response value at the same sensor position; S1.12. Collect the actual structural response measurement values of the sensors and calculate the physical residual signal; S1.13. Design a generator and a discriminator to build an adversarial network; S1.14. Randomly initialize the parameters of the generator and the discriminator, and train them by alternately optimizing the generator and the discriminator; S1.15. Use the cross-entropy loss function to measure the performance of the generator and the discriminator; S1.16. After training, use the generator to extract the structural response feature vector from the physical residual signal, and perform standardization processing on the generated structural response feature vector and the environmental feature vector.

[0009] As a further improvement of this technical solution, in S1.2, construct a sensor graph structure to optimize the frequency band division, and extract the main frequency amplitude and the frequency band energy anomaly index of the aligned frequencies, including the following steps: S1.21. Collect the bridge sensor deployment information and the physical connection relationships between the sensors; S1.22. Build a graph structure; Among them, each node in the graph structure corresponds to a sensor, and the weight of the edge represents the correlation coefficient of the structural response between two sensors; S1.23. Use the graph structure to analyze the response consistency of each sensor at different frequencies; S1.24. Perform clustering analysis on the graph structure to identify subsets of sensors with similar frequency response patterns; S1.25. Retain the frequency bands within ±a% of the structural main frequency characteristics based on the clustering analysis results; S1.26. Find the maximum amplitude and its corresponding frequency in the current frequency band, record this amplitude as the main frequency amplitude feature, and calculate the energy of the current frequency band, and output the frequency band energy anomaly index.

[0010] As a further improvement of this technical solution, in S1.4, a three-dimensional feature tensor is constructed, and time domain, frequency domain, and spatial features are fused using three-dimensional convolution operations to output a fused high-dimensional feature tensor, including the following steps: S1.41. Determine the tensor dimensions, which include the number of time windows, the number of physically aligned frequency bands, and the number of sensor nodes; S1.42. For each time window, each frequency band, and each sensor node, fill the corresponding eigenvalue into the three-dimensional tensor; S1.43. Define a three-dimensional convolutional layer using a deep learning framework and set the convolutional kernel size; S1.44. Pass the constructed three-dimensional feature tensor as input to the three-dimensional convolutional layer; S1.45. After passing through the three-dimensional convolutional layer, output a high-dimensional feature tensor that fuses time domain, frequency domain, and spatial characteristics.

[0011] As a further improvement of this technical solution, identifying abnormal data patterns through a support vector machine model and evaluating the remaining life of the bridge using the carbonation-rust-crack closed-loop feedback control rule based on the extracted key features and abnormal data patterns includes the following steps: S2.1. Select the data obtained during the fault-free operation of the bridge as the training set, predict the crack propagation rate using a spatio-temporal graph convolutional network, and invert the carbonation depth using a hyperspectral image fusion CNN model; S2.2. Use the extracted key features, crack propagation rate, and carbonation depth as inputs to train a support vector machine model. During the training process, adjust the parameters of the support vector machine model through cross-validation; S2.3. Use the trained support vector machine model to score the training data. If the score exceeds the preset threshold b, mark the bridge data for this time period as abnormal; otherwise, consider it normal; S2.4. Based on the long-term monitoring data of the bridge, construct a bridge performance degradation sequence, and use the ratio of the ion diffusion rate in the carbonation zone to the rust expansion stress intensity as an index to drive the carbonation depth evolution model, and then trigger the correction term in the degradation sequence through the dynamic results output by this carbonation depth evolution model; S2.5. Model the bridge performance degradation sequence as a non-negative and non-decreasing random process, and construct three types of degradation modes according to the value; S2.6. Set the degradation threshold d for bridge structure failure, and dynamically adjust the degradation threshold d based on the value; S2.7. Calculate the time required from now until the predicted degradation threshold d is reached based on the current cumulative degradation level and shape and scale parameters. This time is the remaining life of the bridge.

[0012] As a further improvement of this technical solution, in S2.5, according to values, three types of degradation modes are constructed, including the following steps: S2.51. If the value is less than the benchmark point of the value, multiple degradation increment samples are generated using a Gaussian distribution to simulate the degradation situation in a future period of time, and the generated degradation increments are added to the current degradation level to update the bridge performance degradation sequence; S2.52. If the value is greater than or equal to the benchmark point of the value and less than or equal to the upper limit of the value, multiple degradation increment samples are generated using a Gamma distribution to simulate the degradation situation in a future period of time; S2.53. If the value is greater than the upper limit of the

[0013] value, multiple degradation increment samples are generated using a Weibull distribution to simulate the degradation situation in a future period of time, and the generated degradation increments are added to the current degradation level to update the bridge performance degradation sequence. As a further improvement of this technical solution, the information management unit includes a storage management module, an alarm management module, a report generation module, and a video monitoring module; Among them, the storage management module stores bridge physical parameters and abnormal analysis results by establishing a database; The alarm management module receives abnormal data from the data processing and analysis unit and pushes alarm information according to preset levels; The report generation module generates reports, historical data reports, and diagnostic reports for managing bridge data;

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In this bridge safety monitoring system based on sensor data, by introducing a multi-modal physical perception feature fusion method, combining adversarial learning, graph wavelet transform, and three-dimensional convolutional neural network, high-precision feature extraction of bridge structure responses and effective separation of environmental noise are achieved, thereby significantly improving the accuracy and robustness of bridge health state assessment. It can not only effectively identify early structural damage but also has good anti-interference ability, meeting the long-term bridge monitoring requirements in complex environments.

[0015] 2. In the bridge safety monitoring system based on sensor data, a carbonization-corrosion-crack closed-loop feedback control rule is adopted. By combining real-time sensor data with a support vector machine (SVM) anomaly recognition model, a dynamically evolving bridge performance degradation model is constructed. Based on this model, the intelligent prediction of the remaining life of the bridge is realized. By introducing the coupling coefficient k value of carbonization and corrosion to dynamically adjust the degradation process and failure threshold, the limitation of the traditional linear degradation model is broken through, making the bridge life assessment closer to the actual degradation mechanism, providing a scientific basis for bridge maintenance decision-making, and improving the safety of bridge operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall flow block diagram of the present invention; The meanings of each label in the figure are as follows: 1. Sensor measurement unit; 2. Data acquisition and transmission unit; 21. Data reading module; 22. Data transmission module; 3. Data processing and analysis unit; 31. Data feature extraction module; 32. Data anomaly analysis module; 4. Information management unit; 41. Storage management module; 42. Alarm management module; 43. Report generation module; 44. Video monitoring module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment: Please refer to Figure 1 As shown, a bridge safety monitoring system based on sensor data is provided, including: The sensor measurement unit 1 monitors the key physical parameters of the bridge in real time (including stress, strain, displacement, etc.). In this embodiment, these sensors are usually installed at the key parts of the bridge to ensure that the health data of the bridge can be obtained comprehensively and accurately. The sensors include stress sensors, strain sensors, displacement sensors, and hyperspectral imaging sensors deployed in the key areas of the bridge deck. The hyperspectral imaging sensors are installed on the bottom surface of the bridge main girder and the surface of the pier and abutment, fixed by brackets and equipped with an anti-shake module. Each sensor covers an area of 10m×10m. When collecting data, the natural light compensation mode is adopted for the light source, and the output data is a three-dimensional hyperspectral cube; The data acquisition and transmission unit 2 reads the sensor data and transmits the read sensor data. In this embodiment, the data acquisition and transmission unit 2 includes a data reading module 21 and a data transmission module 22; The data acquisition and transmission unit 2 is responsible for securely and efficiently transmitting the data collected by the sensors to the data processing and analysis unit 3, which includes a data collector, a transmission device, and the corresponding communication network; Among them, the data reading module 21 reads the sensor data by connecting to the sensors, and preliminarily arranges and packages the read sensor data; The data transmission module 22 sends the packaged sensor data to the server or data center where the data processing and analysis unit 3 is located by means of wired or wireless transmission.

[0019] The data processing and analysis unit 3 extracts the key features reflecting the bridge health status from the sensor data based on the multi-modal physical perception feature fusion method, identifies the abnormal patterns in the sensor data through the support vector machine model, and evaluates the remaining life of the bridge based on the carbonation-rust-crack closed-loop feedback control rule; In this embodiment, the data processing and analysis unit 3 includes a data feature extraction module 31 and a data anomaly analysis module 32; Among them, the data feature extraction module 31 receives the preprocessed sensor data (such as stress, strain, displacement, etc.), and extracts the key features reflecting the bridge health status based on the multi-modal physical perception feature fusion method; The data anomaly analysis module 32 identifies the abnormal data patterns through the support vector machine model based on the extracted key features, and evaluates the remaining life of the bridge by using the carbonation-rust-crack closed-loop feedback control rule based on the extracted key features and the abnormal data patterns.

[0020] Furthermore, the multi-modal physical perception feature fusion method is to accurately extract the key features reflecting the essence of the bridge structure health from the complex sensor data; separate the environmental noise and the structural response through adversarial learning (S1.1), combine the physical model benchmark (S1.11), eliminate the interference of temperature, humidity, etc., ensure the reliability of the features, integrate the time-domain residual, the frequency-domain main frequency amplitude (S1.2) and the spatial sensor topology relationship (S1.3), and use three-dimensional convolution to generate a high-dimensional feature tensor (S1.4) to comprehensively capture the bridge dynamics behavior; the band energy anomaly index (S1.26) and the sub-graph energy distribution entropy (S1.3) can identify micro-damage (such as early cracks), and support the support vector machine model to accurately locate the abnormal patterns (S2.3); through the fusion of physical model guidance and deep learning, triple breakthroughs in noise robustness, early damage sensitivity, and life prediction accuracy are achieved, which is the key technical support for the bridge monitoring to move from "post-alarm" to "pre-prediction"; The steps of extracting the key features reflecting the bridge health status based on the multi-modal physical perception feature fusion method include: S1.1. Receive the pre - processed multi - dimensional time - series sensor data (including stress, strain, displacement, etc.) transmitted from the data acquisition and transmission unit 2, and use the pre - processed multi - dimensional time - series sensor data as the input of the bridge structure dynamics model (the bridge structure dynamics model is constructed based on the structural mechanics and dynamics control equations, and is used to simulate the dynamic response behavior of the bridge under various loads and boundary conditions. First, the geometric topology of the bridge is established, the material properties of each component (including elastic modulus, density, damping, etc.) are defined, and time - varying external excitations such as vehicle loads, wind loads, and temperature changes are input. At the same time, boundary conditions such as the type of bearing and connection form are set. On this basis, the structural dynamics equation is solved to obtain the time - varying responses at key nodes, including displacement, acceleration, stress, or strain values, as the structural response prediction results. Its output is aligned with the measured sensor data to construct the physical residual, and then support subsequent functions such as feature extraction, anomaly identification, and remaining life prediction). Generate corresponding structural response prediction values under given loads and boundary conditions, and use the structural response prediction values as the physical benchmark to calculate the physical residual between the actual structural response measurement value and the structural response prediction value. Apply the adversarial learning method to separate environmental noise from the structural response, and generate the structural response feature vector and environmental features (improve the ability to identify the true state of the structure and enhance the accuracy and robustness of feature extraction); Among them, calculating the physical residual between the actual structural response measurement value and the structural response prediction value, applying the adversarial learning method to separate environmental noise from the structural response, and generating the structural response feature vector and environmental features include the following steps: S1.11. Construct a physics - driven structural response model based on bridge structure parameters (including geometric parameters, material parameters, etc.) (the physics - driven structural response model is an analytical or numerical dynamics model constructed based on bridge structure parameters, boundary conditions, and load inputs. Its core function is to predict the theoretical response values (stress, strain, displacement) at the positions of each sensor. By inputting the external load information and boundary conditions at the current moment, combining the geometric information, material properties, and structural topology of the bridge body, it outputs the response variables corresponding to the sensor measurement points as the physical reference values of the structural response). Input the current load and boundary conditions, and output the theoretical predicted response values (stress, strain, displacement) at the same sensor position; S1.12. Collect the actual structural response measurement values of the sensors and calculate the physical residual signal (the difference between the actual structural response measurement value and the predicted value. The physical residual is used to reflect structural anomalies. By comparing the actual structural response measurement values of the sensors with the predicted values obtained from the bridge dynamics model, the residual signal obtained can effectively reflect the deviation and potential anomalies of the structure, which helps to locate the source of the anomaly); S1.13. The generator and discriminator are used to construct an adversarial network. The generator attempts to extract structural response features from the input (physical residual signal) while mimicking the distribution of environmental noise as much as possible; the discriminator tries to distinguish the generated structural response features from the real ones. S1.14. Randomly initialize the parameters of the generator and discriminator, and train by alternately optimizing the generator and discriminator. In each iteration, first fix the discriminator parameters and update the generator to minimize the discriminator's recognition ability of its output (that is, make the generated structural response features more difficult to be recognized as forged); then fix the generator parameters and update the discriminator to better distinguish real and generated structural response features. S1.15. Use the cross - entropy loss function to measure the performance of the generator and discriminator (the measurement criterion is to minimize the discriminator's accuracy in distinguishing real and generated samples (the discriminator hopes to minimize the loss, that is, correct classification), while maximizing the deception ability of the generator to generate samples (the generator hopes to maximize the discriminator's loss, that is, misclassification). The two are in mutual confrontation and jointly optimize until a balance is reached). S1.16. After training, use the generator to extract the structural response feature vector from the physical residual signal, and standardize the generated structural response feature vector and the environmental feature vector.

[0021] S1.2. Perform a fast Fourier transform on the physical residual signal to obtain the frequency - domain features of the physical residual signal, and align the frequency components of the physical residual signal with the structural main - frequency features of the bridge through weighted adjustment (for the amplitude spectrum of each sensor, define a weighting function so that higher weights are given at positions close to the structural main - frequency features. To eliminate frequency - offset problems caused by changes in boundary conditions or external disturbances, adopt a weighting strategy to adjust the spectrum), construct a sensor graph structure (nodes = sensors, edges = physical association strength) to optimize the frequency - band division, focus on the structure - sensitive frequency band, and extract the main - frequency amplitude and frequency - band energy anomaly index of the aligned frequencies (establish a graph structure based on the physical connection relationship between sensors, and use the graph structure for clustering analysis, which can help identify subsets of sensors with similar frequency - response patterns. This not only helps to focus on the key features that truly reflect the structural state but also effectively filters out noise information irrelevant to structural health and improves the quality of feature extraction in subsequent analysis steps). Among them, constructing a sensor graph structure (nodes = sensors, edges = physical association strength) to optimize the frequency - band division, focus on the structure - sensitive frequency band, and extract the main - frequency amplitude and frequency - band energy anomaly index of the aligned frequencies includes the following steps: S1.21. Collect the bridge sensor deployment information (location, type) and the physical connection relationship between each sensor. S1.22. Construct a graph structure. Among them, each node in the graph structure corresponds to a sensor, and the weight of the edge represents the correlation coefficient of the structural response between two sensors (the correlation coefficient between two sensors is calculated based on the modal vibration mode vector); S1.23. Analyze the response consistency of each sensor at different frequencies using the graph structure; S1.24. Perform clustering analysis (spectral clustering) on the graph structure to identify subsets of sensors with similar frequency response patterns; S1.25. Based on the clustering analysis results, retain the frequency bands within ±a% of the main frequency characteristics of the structure (ensuring that the frequency bands of interest are closely centered around the frequency components that best represent the dynamic characteristics of the structure, further strengthening the pertinence and effectiveness of the structural health state assessment. The main frequency amplitude and the frequency band energy anomaly index in this process are used as important characteristic indicators, which can intuitively reflect the response characteristics and their change trends of the structure at specific frequencies); S1.26. Find the maximum amplitude and its corresponding frequency within the current frequency band, record this amplitude as the main frequency amplitude characteristic, calculate the energy of the current frequency band, and output the frequency band energy anomaly index; S1.3. Based on the sensor deployment topology structure, construct a sensor physical relationship graph (the node attributes include the sensor type (including strain gauges), installation location, and the edge weight is determined according to the structural physical connection strength between sensors). Through graph wavelet transform, perform spatial decomposition on the structural response characteristics separated in step S1.1 and the frequency domain characteristics in step S1.2, extract the sub-graph energy distribution at different scales, and calculate the sub-graph energy distribution entropy (calculate the sub-graph energy distribution entropy to quantify the spatial heterogeneity of the structural response, that is, to measure the degree of uniformity of the energy distribution within each sub-graph) and the cross-sensor frequency response consistency (analyze and compare the similarity or difference in the responses of each sensor within the selected frequency band, so as to reflect the change of the response synergy between sensors); S1.4. Integrate the time domain, frequency domain, and spatial characteristics extracted in steps S1.2 and S1.3 to construct a three-dimensional feature tensor (including the number of time windows, the number of physically aligned frequency bands, and the number of sensor nodes), and use three-dimensional convolution operations to fuse the time domain, frequency domain, and spatial characteristics, and output the fused high-dimensional feature tensor (by integrating the time series segments (time domain), physically aligned frequency components (frequency domain), and the spatial distribution of sensors (spatial domain) of sensor data into a three-dimensional tensor, this method can comprehensively capture the multi-level characteristics of the bridge structural health state. This multi-dimensional information integration helps to more accurately reflect the actual condition of the bridge; three-dimensional convolution operations can automatically learn and extract the complex relationships between different dimensions, and dig out the deep-level characteristics hidden in the original data. Compared with traditional single-dimensional analysis methods, this method can provide richer and more representative feature representations, which helps to improve the accuracy of subsequent tasks (such as anomaly detection, remaining life prediction, etc.)); Among them, constructing a three-dimensional feature tensor and using three-dimensional convolution operations to fuse time-domain, frequency-domain, and spatial features, and outputting a fused high-dimensional feature tensor includes the following steps: S1.41. Determine the tensor dimensions, which include the number of time windows (the number of time series segments determined according to actual monitoring requirements), the number of physically aligned frequency bands (the number of key frequency bands determined in S1.2), and the number of sensor nodes (the total number of sensors deployed on the bridge); S1.42. For each time window, each frequency band, and each sensor node, fill the corresponding eigenvalue into the three-dimensional tensor; S1.43. Define a three-dimensional convolutional layer using a deep learning framework (PyTorch) and set the convolutional kernel size; S1.44. Pass the constructed three-dimensional feature tensor as input to the three-dimensional convolutional layer. The convolution operation will automatically slide the window to traverse the entire tensor, perform dot product and summation operations at each position, thereby generating a new feature map; S1.45. After a series of three-dimensional convolutional layers, output a high-dimensional feature tensor that fuses time-domain, frequency-domain, and spatial characteristics; S1.5. Based on the fused high-dimensional feature tensor, use the mutual information method to screen the initial feature vectors and perform standardization processing on each screened feature vector; S1.6. Combine the standardized feature vectors into the final feature vector.

[0022] Furthermore, identify abnormal data patterns through a support vector machine model, and evaluate the remaining life of the bridge based on the extracted key features and abnormal data patterns using the carbonation-corrosion-crack closed-loop feedback control rule (the carbonation-corrosion-crack closed-loop feedback control rule integrates three fields: chemistry (carbonation), electrochemistry (corrosion), and mechanics (cracks), captures the full-link changes in degradation evolution, and improves the dynamic accuracy of prediction; introduces a feedback term, rust expansion stress causes cracks → cracks in turn accelerate carbonation, avoiding traditional "static" assumptions and improving the dynamic accuracy of prediction; regulates the change of degradation rate through the k value, and supports the simulation of degradation jump and self-amplification effect), including the following steps: S2.1. Select the data obtained during the fault-free operation of the bridge as the training set, and use the spatio-temporal graph convolutional network to predict the crack propagation rate (the input of the spatio-temporal graph convolutional network is: the spatio-temporal graph structure (nodes represent sensor positions or specific structural key points, and edges represent the physical association strength (distance) between these points) and time series data (for each node, there is a series of data records that change over time, and these records reflect the historical changes of the structural parameters at this position); the intermediate processing includes graph convolutional operations, time convolutional operations, and fusion mechanisms; the final output is an estimate of the crack propagation rate in the future for a period of time, which is a numerical value or a set of numerical values representing the prediction results for different time periods), and use the hyperspectral image fusion CNN model to invert the carbonation depth (combine the hyperspectral image data with the convolutional neural network (CNN) to identify and quantify the carbonation degree in the bridge concrete, wherein, in the input layer of the hyperspectral image fusion CNN model, the hyperspectral data is taken, and at the same time, the time series data of the strain sensor at the corresponding position (that is, the strain parameters of the bridge monitored in real time by the sensor measurement unit 1) is used as the collaborative input, and is mapped to an H×W×32 feature map through the fully connected layer (H represents the height of the image feature map, and W represents the width of the image feature map). In the fusion layer, the spectral and strain features are concatenated in the channel dimension to obtain an H×W×182 tensor, and the spatio-temporal-spectral features are extracted through 3 layers of 3D convolution (kernel size 3×3×5, stride 1×1×2). The key bands are dynamically weighted through the channel attention module, and finally, by the output layer, the carbonation depth is predicted through the regression head. The spatio-temporal graph convolutional network is used to predict the crack propagation rate, and at the same time, data is collected through the hyperspectral imaging sensor on the bridge deck and input into the hyperspectral CNN model to invert the carbonation depth); S2.2. Use the extracted key features, crack propagation rate, and carbonation depth as inputs to train the support vector machine model. During the training process, adjust the parameters of the support vector machine model through cross-validation; S2.3. Use the trained support vector machine model to score the training data. If the score exceeds the preset threshold b, mark the bridge data for this time period as abnormal; otherwise, consider it normal; S2.4. Based on the long-term monitoring data of the bridge, construct a bridge performance degradation sequence to reflect the change trend of the structural performance over time, and use the ratio of the ion diffusion rate in the carbonated area to the rust expansion stress intensity ( value) as an index to drive the carbonation depth evolution model , and then, through the dynamic results output by the carbonation depth evolution model, trigger the correction term in the degradation sequence (the correction term is , when When the value changes, the evolution method and degradation threshold of the entire degradation model will be dynamically adjusted, so as to more accurately simulate the actual degradation process. By using the trigger correction term, the degradation process is no longer a simple linear accumulation, but a non-linear evolution with a feedback amplification effect, reflecting the closed-loop feedback of carbonation depth → steel corrosion → microcrack propagation → carbonation acceleration; Among them, the bridge performance degradation sequence is: In the formula, represents the cumulative degradation level of the bridge at time , that is, the bridge performance degradation sequence, represents the basic degradation increment (determined by environmental conditions), , represents the amplification function of the carbonation-corrosion coupling factor, , represents the ion diffusion rate in the carbonation zone, represents the stress intensity of rust expansion, represents the carbonation depth at the current moment, represents the influence function of carbonation depth on degradation, which is a linear function, represents the weight coefficient, represents the small perturbation term or random noise, represents time; Taking the ratio of the ion diffusion rate in the carbonation zone to the stress intensity of rust expansion ( value) as the index-driven carbonation depth evolution model is specifically: ; In the formula, represents the carbonation depth at the current moment, represents the material damage coefficient; S2.5. Model the bridge performance degradation sequence as a non-negative and non-decreasing random process, and construct three types of degradation modes according to the value; Among them, constructing three types of degradation modes according to the value includes the following steps: These three types of degradation modes use the real-time ratio of chemical diffusion (carbonation) to mechanical response (rust expansion stress) as the degradation state criterion, breaking through the traditional time / single physical field-dependent model; S2.51. If the value is less than the reference point of the (Carbonation and corrosion do not form a significant coupling, and the structure is in a relatively safe state with a slow degradation rate). Generate multiple degradation increment samples using the Gaussian distribution to simulate the degradation situation in the future for a period of time, and add the generated degradation increments to the current degradation level to update the bridge performance degradation sequence; In this stage, the bridge performance degradation increment ; Among them, is the mean of the basic degradation rate, reflecting the slow and uniform degradation trend, is the standard deviation of fluctuations, reflecting the environmental random perturbations; Simulation method: Randomly generate multiple degradation increment samples from the Gaussian distribution, and for each sample : ; In the formula, represents the time step index; S2.52. If the value is greater than or equal to the reference point of the value and less than or equal to the upper limit of the value (The coupling effect begins to appear, and some problems begin to occur in the structure, but it has not reached the dangerous level), use the Gamma distribution to generate multiple degradation increment samples to simulate the degradation situation in the future for a period of time; Among them, use the Gamma distribution for modeling: ; represents the shape parameter, determining the skewness of the distribution, represents the scale parameter, controlling the degradation rate. The Gamma distribution is suitable for describing the occasional acceleration during the degradation process (including microcrack connection and cover layer failure); Simulation method: Sample multiple Gamma distribution samples as future degradation increments, but do not immediately update the degradation sequence, only as alternative paths for deduction or input for Monte Carlo prediction; S2.53. If the value is greater than the upper limit of the value (The corrosion rate surges non-linearly (exponential amplification effect), and the structure enters the stage of rapid deterioration), use the Weibull distribution to generate multiple degradation increment samples to simulate the degradation situation in the future for a period of time, and add the generated degradation increments to the current degradation level to update the bridge performance degradation sequence; ; represents the scale parameter, controlling the degradation speed, Denote the shape parameter reflecting the "acceleration" characteristic; At each moment , sample multiple samples , and update the degradation level: ; S2.6. Set the degradation threshold d for the failure of the bridge structure. This value reflects that when the bridge structure reaches this level of degradation, it will no longer be able to operate safely, and dynamically adjust the degradation threshold d based on value; Based on value, the dynamic adjustment of the degradation threshold is specifically: ; In the formula, denotes the reference point of the dynamic coupling of carbonation and corrosion; denotes the width of the transition interval, indicating that the change within this range has a greater impact on the structural degradation; denotes the maximum proportion by which the degradation threshold can be reduced; denotes the initial degradation threshold; denotes the degradation threshold dynamically adjusted based on the value; S2.7. Calculate the time required from now until the predicted degradation threshold d is reached based on the current cumulative degradation level, shape parameter, and scale parameter. This time is the remaining life of the bridge.

[0023] The information management unit 4 stores the analysis results of the data processing and analysis unit 3 and displays the monitored physical parameters of the bridge; In this embodiment, the information management unit 4 includes a storage management module 41, an alarm management module 42, a report generation module 43, and a video monitoring module 44; Among them, the storage management module 41 stores the physical parameters of the bridge and the abnormal analysis results by establishing a database. Users can access the system at any time and place to view the real-time data and historical records of the bridge. Through the traceability and analysis of the monitoring data, the system can timely discover potential safety problems of the bridge; The alarm management module 42 receives the abnormal data from the data processing and analysis unit 3 and pushes alarm information according to the preset level. Through various types of sensors deployed on the surface of the bridge, the system can monitor the structural parameters of the bridge and set a multi-level alarm function. When the data exceeds the alarm upper limit, the system will automatically issue alarm information at different levels and convey it to relevant leaders and responsible persons in various ways; The report generation module 43 generates reports, historical data reports, and diagnostic reports for managing bridge data; The video monitoring module 44 accesses and controls the cameras on the bridge site for remote video monitoring. Users can view the live video images on the site in real time from the monitoring terminal. At the same time, they can also control the camera lens and pan-tilt head on the site through software, meeting the users' needs for multi-directional video image monitoring of the site and ensuring the safe operation of the bridge.

[0024] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A bridge safety monitoring system based on sensor data, characterized in that, Including: A sensor measurement unit (1) that monitors the key physical parameters of the bridge in real time; A data acquisition and transmission unit (2) that reads the sensor data and transmits the read sensor data; A data processing and analysis unit (3) that extracts the key features reflecting the bridge health state from the sensor data based on the multi-modal physical perception feature fusion method, identifies the abnormal patterns in the sensor data through the support vector machine SVM model, and evaluates the remaining life of the bridge based on the carbonation-corrosion-crack closed-loop feedback control rule; An information management unit (4) that stores the analysis results of the data processing and analysis unit (3) and displays the monitored bridge physical parameters.

2. The bridge safety monitoring system based on sensor data according to claim 1, wherein: The data acquisition and transmission unit (2) includes a data reading module (21) and a data transmission module (22); Wherein, the data reading module (21) reads the sensor data by connecting to the sensor, and preliminarily arranges and packages the read sensor data; The data transmission module (22) sends the packaged sensor data to the server where the data processing and analysis unit (3) is located by wireless transmission.

3. The bridge safety monitoring system based on sensor data according to claim 1, characterized in that: The data processing and analysis unit (3) includes a data feature extraction module (31) and a data anomaly analysis module (32); Wherein, the data feature extraction module (31) receives the preprocessed sensor data and extracts the key features reflecting the bridge health state based on the multi-modal physical perception feature fusion method; The data anomaly analysis module (32) identifies the abnormal data patterns through the support vector machine model based on the extracted key features, and evaluates the remaining life of the bridge by using the carbonation-corrosion-crack closed-loop feedback control rule based on the extracted key features and the abnormal data patterns.

4. The bridge safety monitoring system based on sensor data according to claim 3, characterized in that: The extraction of the key features reflecting the bridge health state based on the multi-modal physical perception feature fusion method includes the following steps: S1.

1. Receive the preprocessed multi-dimensional time series sensor data transmitted from the data acquisition and transmission unit (2), use the preprocessed multi-dimensional time series sensor data as the input of the bridge structural dynamics model, generate the corresponding structural response prediction value under the given load and boundary conditions, and use this structural response prediction value as the physical benchmark, calculate the physical residual between the actual structural response measurement value and the structural response prediction value, and apply the adversarial learning method to separate the environmental noise and the structural response, generating the structural response feature vector and the environmental feature; S1.

2. Perform a fast Fourier transform on the physical residual signal, obtain the frequency domain features of the physical residual signal, and make the frequency components of the physical residual signal align with the main frequency features of the bridge structure through weighted adjustment, construct the sensor graph structure optimization frequency band division, and extract the main frequency amplitude and frequency band energy anomaly index of the aligned frequency. S1.

3. Based on the sensor deployment topology, construct a physical relationship graph of sensors, perform spatial decomposition on the structural response features separated in step S1.1 and the frequency domain features in step S1.2 through graph wavelet transform, and calculate the sub-graph energy distribution entropy and cross-sensor frequency response consistency; S1.

4. Integrate the time domain, frequency domain, and spatial features extracted in steps S1.2 and S1.3, construct a three-dimensional feature tensor, and use three-dimensional convolution operations to fuse the time domain, frequency domain, and spatial features, and output the fused high-dimensional feature tensor; S1.

5. Based on the fused high-dimensional feature tensor, use the mutual information method to screen the initial feature vectors, and perform standardization processing on each screened feature vector; S1.

6. Combine the standardized feature vectors into the final feature vector.

5. The bridge safety monitoring system based on sensor data according to claim 4, wherein: In S1.1, calculate the physical residual between the actual structural response measurement value and the structural response prediction value, and apply the adversarial learning method to separate the environmental noise and the structural response, generating a structural response feature vector and environmental features, including the following steps: S1.

11. Construct a physically-driven structural response model based on the bridge structure parameters, input the current load and boundary conditions, and output the theoretical predicted response value at the same sensor position; S1.

12. Collect the actual structural response measurement values of the sensors and calculate the physical residual signal; S1.

13. Design a generator and a discriminator to construct an adversarial network; S1.

14. Randomly initialize the parameters of the generator and the discriminator, and perform training by alternately optimizing the generator and the discriminator; S1.

15. Use the cross-entropy loss function to measure the performance of the generator and the discriminator; S1.

16. After training, use the generator to extract the structural response feature vector from the physical residual signal, and perform standardization processing on the generated structural response feature vector and environmental feature vector.

6. The bridge safety monitoring system based on sensor data according to claim 4, characterized in that: In S1.2, construct an optimized frequency band division for the sensor graph structure, and extract the main frequency amplitude and frequency band energy anomaly index of the aligned frequencies, including the following steps: S1.

21. Collect the bridge sensor deployment information and the physical connection relationship between each sensor; S1.

22. Construct a graph structure; Among them, each node in the graph structure corresponds to a sensor, and the weight of the edge represents the correlation coefficient of the structural response between two sensors; S1.

23. Use the graph structure to analyze the response consistency of each sensor at different frequencies; S1.

24. Perform clustering analysis on the graph structure to identify subsets of sensors with similar frequency response patterns; S1.

25. Based on the clustering analysis results, retain the frequency bands within ±a% of the main frequency characteristics of the structure; S1.

26. Find the maximum amplitude and its corresponding frequency within the current frequency band, record this amplitude as the main frequency amplitude feature, and calculate the energy of the current frequency band, and output the frequency band energy anomaly index.

7. The bridge safety monitoring system based on sensor data according to claim 4, characterized in that: In S1.4, construct a three-dimensional feature tensor, and use three-dimensional convolution operations to fuse the time domain, frequency domain, and spatial features, and output the fused high-dimensional feature tensor, including the following steps: S1.

41. Determine the tensor dimensions, and the tensor dimensions include the number of time windows, the number of physically aligned frequency bands, and the number of sensor nodes; S1.

42. For each time window, each frequency band, and each sensor node, fill the corresponding eigenvalue into a three-dimensional tensor. S1.

43. Define a three-dimensional convolutional layer using a deep learning framework and set the convolutional kernel size. S1.

44. Pass the constructed three-dimensional feature tensor as input to the three-dimensional convolutional layer. S1.

45. After passing through the three-dimensional convolutional layer, output a high-dimensional feature tensor that integrates time-domain, frequency-domain, and spatial characteristics.

8. The bridge safety monitoring system based on sensor data according to claim 3, characterized in that: The method of identifying abnormal data patterns through a support vector machine model and evaluating the remaining life of the bridge based on the extracted key features and abnormal data patterns using the carbonation-corrosion-crack closed-loop feedback control rule includes the following steps: S2.

1. Select the data obtained during the fault-free operation of the bridge as the training set, use a spatio-temporal graph convolutional network to predict the crack propagation rate, and use a hyperspectral image fusion CNN model to invert the carbonation depth. S2.

2. Use the extracted key features, crack propagation rate, and carbonation depth as input to train a support vector machine model. During the training process, adjust the parameters of the support vector machine model through cross-validation. S2.

3. Use the trained support vector machine model to score the training data. If the score exceeds the preset threshold b, mark the bridge data for this time period as abnormal; otherwise, consider it normal. S2.

4. Based on the long-term bridge monitoring data, construct the bridge performance degradation sequence, and use the ratio of the ion diffusion rate in the carbonation zone to the rust expansion stress intensity as an index to drive the carbonation depth evolution model, and then trigger the correction term in the degradation sequence through the dynamic results output by the carbonation depth evolution model; S2.

5. Model the bridge performance degradation sequence as a non - negative and non - decreasing stochastic process, and construct three types of degradation patterns according to the value. S2.

6. Set the degradation threshold d for the failure of the bridge structure, and dynamically adjust the degradation threshold d based on the value; S2.

7. According to the current cumulative degradation level, shape parameter, and scale parameter, calculate the time required from now until the degradation threshold d is expected to be reached. This time is the remaining life of the bridge.

9. The bridge safety monitoring system based on sensor data according to claim 8, characterized in that: In S2.5, according to value, three types of degradation modes are constructed, including the following steps: S2.

51. If the value is less than the reference point of the value , generate multiple degenerate increment samples using the Gaussian distribution to simulate the degradation situation in a future period of time, add the generated degenerate increments to the current degradation level, and update the bridge performance degradation sequence; S2.

52. If the value is greater than or equal to the reference point of the value and less than or equal to the upper limit of the value , generate multiple degenerate increment samples using the Gamma distribution to simulate the degradation situation in a future period of time; S2.

53. If The value is greater than The upper limit of the value , generate multiple degradation increment samples using the Weibull distribution to simulate the degradation situation in a future period of time, add the generated degradation increments to the current degradation level, and update the bridge performance degradation sequence.

10. The bridge safety monitoring system based on sensor data according to claim 1, characterized in that: The information management unit (4) includes a storage management module (41), an alarm management module (42), a report generation module (43), and a video monitoring module (44). Among them, the storage management module (41) stores the bridge physical parameters and abnormal analysis results by establishing a database. The alarm management module (42) receives the abnormal data from the data processing and analysis unit (3) and pushes alarm information according to the preset level. The report generation module (43) generates reports, historical data reports, and diagnostic reports for managing bridge data. The video monitoring module (44) accesses and controls the bridge site cameras for remote video monitoring.

Images