Bridge structure health monitoring system based on intelligent sensor network and deep learning

Abstract

The present invention relates to the technical field of bridge structure monitoring, and specifically to a bridge structure health monitoring system based on intelligent sensor networks and deep learning, comprising: a data acquisition module, used to set multiple acquisition points on a bridge, and obtain bridge structure monitoring data related to the bridge, wherein the bridge structure monitoring data includes temperature and humidity, strain, deflection, cable frequency data and vibration acceleration on the bridge; a data monitoring module, used to match and connect the bridge structure monitoring data, set abnormal indicators corresponding to the bridge structure monitoring data, and obtain abnormal related data corresponding to the abnormal indicators; a modal abnormality analysis module, used to identify abnormal related data, determine abnormal labels of the abnormal related data, and cluster according to the abnormal labels to obtain abnormal classification results; the accuracy and reliability of bridge structure health status assessment are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge structure monitoring, and in particular to a bridge structure health monitoring system based on intelligent sensor networks and deep learning. Background Art

[0002] As an important part of road traffic construction, the safety of bridges has attracted increasing attention. Traditional monitoring methods often rely on manual inspections and regular testing, which have problems such as low monitoring efficiency, high cost, and difficulty in real-time monitoring. With the development of intelligent sensing technology and deep learning, bridge structure health monitoring systems are gradually moving towards intelligence and automation. However, the existing bridge structure health monitoring systems still have some problems, such as inaccurate abnormality identification and single evaluation results.

[0003] For example, Chinese patent publication number CN111486902A discloses a real-time monitoring system for the safety performance of a large-span bridge based on big data, including a detection area division module, a bearing pressure detection module, a wind parameter acquisition module, a deformation displacement monitoring module, a water flow impact parameter detection module, a modeling and analysis module, a parameter database, a central server, an early warning module and a display terminal. The real-time monitoring system for the safety performance of a large-span bridge based on big data provided by the present invention can comprehensively reflect the health status of the bridge by integrating the bearing pressure, longitudinal wind force, deformation displacement and water flow impact force influencing factors of the bridge, and combining the modeling and analysis module to quantify the influence of the above influencing factors on the safety performance of the bridge, and then statistically calculate the comprehensive safety factor of the bridge, so as to comprehensively reflect the health status of the bridge.

[0004] For example, Chinese patent publication No. CN109029589A discloses a bridge structure safety status monitoring system, which is characterized in that it includes: a sensing subsystem, a signal acquisition and storage subsystem, a signal processing subsystem, a signal communication subsystem and a safety evaluation subsystem; the sensing subsystem monitors the response of the bridge structure under external stimulation through sensors or measuring instruments; the signal acquisition and storage subsystem is used to collect various types of information monitored by the sensing subsystem and store the various types of information in a storage space; the signal processing subsystem includes multiple types of signal processing units for detecting various types of abnormal signals; the safety evaluation subsystem is used to evaluate the damage status of the bridge structure, and then generate a bridge structure health detection report.

[0005] The existing technology tends to identify various signals generated by the bridge structure and complete the assessment of the current bridge health based on these contents; however, when verifying the health of the bridge structure, it is necessary to identify the correlation, difference, and distribution of multiple related data on the bridge structure to verify the specific situation of the current bridge structure, thereby identifying the location that will be mainly affected when an abnormality occurs, so as to perform maintenance on the bridge in a timely manner. Summary of the invention

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a bridge structure health monitoring system based on intelligent sensor networks and deep learning, including: a data acquisition module, used to set multiple acquisition points on the bridge to obtain bridge structure monitoring data related to the bridge, wherein the bridge structure monitoring data includes temperature and humidity, strain, deflection, cable frequency data and vibration acceleration on the bridge.

[0007] The data monitoring module is used to match and connect the bridge structure monitoring data, set abnormal indicators corresponding to the bridge structure monitoring data, and obtain abnormal related data corresponding to the abnormal indicators.

[0008] The modal anomaly analysis module is used to identify anomaly-related data, determine anomaly labels of anomaly-related data, and cluster them according to the anomaly labels to obtain anomaly classification results.

[0009] The distribution identification module is used to verify the abnormal labels of the abnormal classification results after clustering, obtain the location points corresponding to the abnormal labels, determine the abnormal distribution coefficient related to the distribution of abnormal labels and location points, and the relative safety factor between each location point.

[0010] The difference assessment module is used to analyze the temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the bridge structure monitoring data one by one, determine the difference in each measurement, and obtain the difference assessment coefficient.

[0011] The health monitoring module is used to comprehensively identify the difference assessment coefficient, relative safety factor and abnormal distribution coefficient to obtain the health monitoring results corresponding to the current bridge structure.

[0012] The beneficial effects of the present invention are as follows: 1. The present invention realizes real-time monitoring of bridge structures through an intelligent sensor network and multiple collection points, thereby improving monitoring efficiency and accuracy.

[0013] 2. The present invention can timely discover and identify abnormal conditions in the bridge structure through the data monitoring module and the modal anomaly analysis module, providing an important basis for early warning and maintenance.

[0014] 3. The present invention combines the difference assessment module, the distribution identification module and the health monitoring module to comprehensively assess the health status of the bridge structure and improve the accuracy and reliability of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0016] Figure 1 This is the system framework diagram of the bridge structure health monitoring system based on intelligent sensor network and deep learning.

[0017] Figure 2 It is a system flow chart of the bridge structure health monitoring system based on intelligent sensor network and deep learning.

[0018] Figure 3 It is a flow chart of the modal anomaly analysis module of the bridge structure health monitoring system based on intelligent sensor network and deep learning.

[0019] Figure 4 It is a flow chart of the difference assessment module of the bridge structure health monitoring system based on intelligent sensor network and deep learning. DETAILED DESCRIPTION

[0020] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0021] See also Figure 1 The bridge structure health monitoring system based on intelligent sensor network and deep learning includes: data acquisition module, data monitoring module, modal anomaly analysis module, distribution identification module, difference assessment module and health monitoring module.

[0022] The data acquisition module transmits the collected data to the data monitoring module. The data monitoring module is used to match and connect the data and identify the abnormal indicators in the data, and transmit the abnormal indicators and related data to the modal abnormality analysis module; the modal abnormality analysis module clusters the abnormal indicators and corresponding data to obtain the abnormal classification results, and transmits the classified data to the distribution identification module; the distribution identification module is used to verify the distribution of the clustered data, and output the corresponding identified content to the difference assessment module. The difference assessment module analyzes the differences in the data and outputs the data to the health monitoring module. The health monitoring module comprehensively analyzes all the obtained data to obtain the health monitoring results.

[0023] like Figure 2 As shown, in one embodiment of the present invention, a data acquisition module is used to set multiple acquisition points on a bridge to obtain bridge structure monitoring data related to the bridge, wherein the bridge structure monitoring data includes temperature and humidity, strain, deflection, cable frequency data and vibration acceleration on the bridge.

[0024] The data monitoring module is used to match and connect the bridge structure monitoring data, set abnormal indicators corresponding to the bridge structure monitoring data, and obtain abnormal related data corresponding to the abnormal indicators.

[0025] The modal anomaly analysis module is used to identify anomaly-related data, determine anomaly labels of anomaly-related data, and cluster them according to the anomaly labels to obtain anomaly classification results.

[0026] The distribution identification module is used to verify the abnormal labels of the abnormal classification results after clustering, obtain the location points corresponding to the abnormal labels, determine the abnormal distribution coefficient related to the distribution of abnormal labels and location points, and the relative safety factor between each location point.

[0027] The difference assessment module is used to analyze the temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the bridge structure monitoring data one by one, determine the difference in each measurement, and obtain the difference assessment coefficient.

[0028] The health monitoring module is used to comprehensively identify the difference assessment coefficient, relative safety factor and abnormal distribution coefficient to obtain the health monitoring results corresponding to the current bridge structure.

[0029] Temperature and humidity are measured by thermometers and humidity meters installed at the collection points on the bridge. Temperature and humidity reflect the thermal conditions of the bridge and its surrounding environment, which helps to evaluate the performance changes of bridge materials at different temperatures. Deflection is identified by image deflectometers installed on the main beam or other positions of the bridge. Strain is obtained by vibrating wire strain gauges installed on the bridge structure. Strain reflects the deformation of the bridge structure under force, which helps to evaluate the strength and stiffness of the bridge. Cable frequency data is identified by cable force accelerometers on the bridge's cable-stayed cables. It mainly measures the natural frequency of the cable. This natural frequency is represented here as cable frequency data and is used to calculate the cable force of the cable. It is an important parameter for evaluating the health status of the cable-stayed bridge structure. Vibration acceleration is identified by accelerometers installed on the bridge structure. It mainly identifies the acceleration generated by the vibration of vehicles passing on the bridge. It reflects the acceleration change of the bridge structure under vibration or dynamic load, which helps to evaluate the dynamic response and stability of the bridge.

[0030] In one embodiment of the present invention, the data monitoring module needs to process the acquired bridge result monitoring data, find abnormal indicators existing in the data at this time, and obtain data related to the abnormal indicators.

[0031] The purpose of matching and connecting the bridge structure monitoring data at this time is to align the data at different time points and combine these data into multiple data curves currently required, and select abnormal indicators based on the identification of these data curves.

[0032] For example, verify the strain, deflection, cable frequency data, and vibration acceleration under the corresponding temperature and humidity. At this time, the scenarios in which these data are collected are all scenes where the bridge is in normal use and there is no large-scale congestion. Obtain the data at the collection point at this time, and combine these data together in the form of a time series. At the same time, verify the changing trend of these data to ensure that the current data has a consistent changing trend. Obtain the indicator value of the bridge structure monitoring data, and set the places that are different from the predicted values ​​as abnormal indicators.

[0033] Therefore, the method for obtaining abnormal indicators is to obtain the corresponding time series of the bridge structure monitoring data, verify the fluctuation of the bridge structure monitoring data in each cycle under the corresponding time series, and observe the periodic influence value between the variables in the bridge structure monitoring data in each cycle when the time series changes. The fluctuation amount and the periodic influence value are combined to identify the abnormal indicators in the bridge structure monitoring data.

[0034] The fluctuation of bridge structure monitoring data is expressed as the ratio of the mean value to the standard deviation of the bridge structure monitoring data, which is used to indicate the variation of the current bridge structure monitoring data under corresponding circumstances.

[0035] The cycle impact value verifies the linear correlation degree between the variables in the bridge structure monitoring data in each cycle, that is, obtains the covariance between the variables in the bridge structure monitoring data, and sets the covariance matrix to represent the linear relationship between the variables in the current bridge structure monitoring data. When setting the covariance matrix, the bridge structure monitoring data will be standardized to eliminate the dimensions between the data. Each element in the covariance matrix represents the covariance between the variables in the bridge structure monitoring data, which is expressed as the covariance between temperature and humidity, strain, deflection, cable frequency data and vibration acceleration. The covariance matrix is ​​characteristically decomposed to obtain the eigenvalues ​​and eigenvectors of the covariance matrix, and the eigenvalues ​​are sorted from large to small, and the eigenvectors whose cumulative contribution rate reaches the preset threshold are selected, and the covariance average value of the corresponding eigenvector is used as the cycle impact value. The cycle set at this time can be set according to the time interval of the collection, for example, select data with an interval of 15 minutes as a cycle, or select a shorter or longer time to represent the selected cycle length at this time, and analyze the data in these cycles in turn to obtain the cycle impact value and fluctuation amount to identify whether the data changes in these cycles are normal.

[0036] The cumulative contribution rate represents the ratio of the sum of a part of the eigenvalues ​​sorted from large to small to the sum of the total eigenvalues. When the preset threshold is reached, the eigenvector corresponding to the corresponding eigenvalue is selected, and the average value of the covariance values ​​corresponding to these eigenvectors is used as the periodic influence value selected at this time; the preset threshold will select 80% as the value used at this time; the purpose of selecting covariance and related eigenvalues ​​according to eigenvectors and other methods at this time is to obtain the main influencing components in the bridge structure monitoring data during the normal collection process, and analyze these main components to verify the relevant conditions of the bridge structure at this time.

[0037] The way of combining the fluctuation amount and the periodic impact value can be expressed as calculating the fluctuation amount and the periodic impact value to obtain the abnormal index.

[0038] ;in, Indicates abnormal indicators. represents the fluctuation amount, represents the periodic impact value, represents the standard value of the fluctuation amount, It represents the standard value of the periodic impact value. The fluctuation amount and the standard value of the periodic impact value used at this time can be obtained through the average value of the bridge-related data in the historical data.

[0039] According to the obtained abnormal indicators, the abnormal indicator values ​​and related bridge structure monitoring data in each cycle are output as abnormal related data. At this time, it is necessary to output the current bridge situation according to the different values ​​of these fluctuations and periodic impact values ​​in each cycle, so as to verify and analyze these situations later and obtain the comprehensive analysis results of the current bridge.

[0040] In one embodiment of the present invention, the modal anomaly analysis module needs to identify anomaly labels in anomaly-related data and identify these related data to facilitate subsequent processing of the anomaly-related data.

[0041] Before identifying abnormal related data, it is necessary to verify the data type of the abnormal related data at this time, and whether there are corresponding data drifts and data jumps in the abnormal related data, and identify and process these data situations to discover related problems in the abnormal related data.

[0042] Because the data collected by the current collection point will show a fluctuating form, other data except temperature and humidity will not show simple straight line data, so it is necessary to verify whether other data have data drift type situations. At the same time, it is also necessary to verify whether the value of a certain point becomes larger or smaller. These situations are clustered with the actual use of the bridge to obtain the abnormal labels and abnormal classification results at this time.

[0043] The classification method at this time can be based on the distribution ratio of these data to the collected data, and then identify and classify the relevant data under these distribution ratios during calculation. The data corresponding to temperature and humidity, strain, deflection, frequency data and vibration acceleration can be clustered according to the characteristics of the data collected at the collection point or at the time, and abnormal labels can be set to complete all identification.

[0044] like Figure 3 As shown, the implementation method of the modal anomaly analysis module includes: obtaining the distribution ratio corresponding to the abnormal related data and the collection point, setting multiple clustering clusters for the abnormal related data corresponding to the maximum distribution ratio, and selecting the distribution ratio at this time to indicate that after the data at each collection point is identified, the ratio value of the corresponding abnormal related data to the original data is finally obtained. When this ratio value is large, it means that the data collected at this location point has a large contribution to the current bridge structure analysis and can significantly represent the characteristics of the current abnormal related data; clustering the abnormal related data according to the clustering cluster, calculating the distance between each abnormal related data and the clustering cluster, and recalculating the center of each clustering cluster until the center of the clustering cluster no longer changes significantly, and outputting the clustering centers corresponding to the multiple clustering clusters at this time as the clustering result, setting abnormal labels for the abnormal related data in the clustering result according to the clustering result and the actual use of the bridge, and outputting the obtained clustering result as the abnormal classification result.

[0045] The distance between the above-mentioned abnormal related data and the clustering cluster is expressed as calculating the distance between the two data using a vector method. At the same time, the center of the clustering cluster is set to the average value of the current clustering data. At this time, recalculating the center of each clustering cluster is to continuously calculate the average value of the current clustered data until the average value no longer changes. From this, multiple stable clustering centers can be obtained, thereby outputting the corresponding clustering results, and finally obtaining the abnormal classification results.

[0046] The abnormal labels set at this time will display the content related to the current temperature and humidity, strain, deflection, cable frequency data and vibration acceleration, such as temperature and humidity abnormal labels, strain abnormal labels, deflection abnormal labels, cable frequency data abnormal labels, vibration acceleration abnormal labels, etc.

[0047] When the monitored temperature and humidity data exceeds the preset normal range, the temperature and humidity anomaly label is triggered; this can help determine whether the bridge structure is in unfavorable environmental conditions. For example, excessive humidity may lead to accelerated structural corrosion. If the bridge is located in a high humidity area and the humidity inside the bridge is continuously higher than 80%, the "high humidity anomaly" label may be triggered, indicating that dehumidification measures need to be taken to protect the bridge structure.

[0048] When the strain value of the bridge structure exceeds the preset threshold, the strain anomaly label is triggered; this usually indicates that the bridge structure is subjected to excessive load or other abnormal conditions exist; if the strain value of a certain part of the bridge suddenly increases to above the first-level warning value under normal use, the "strain anomaly" label may be triggered, indicating that the bridge needs to be further inspected to determine whether there are cracks, damage and other problems.

[0049] When the deflection of a bridge due to stress or non-uniform temperature changes exceeds the preset allowable range, the deflection anomaly label is triggered; this may reflect changes in the stiffness of the bridge structure or other structural problems; if the deflection value of a certain span exceeds the maximum deflection value allowed by the design when a heavy vehicle passes through the bridge, the "deflection anomaly" label may be triggered, indicating the need to evaluate the bearing capacity of the bridge and carry out necessary reinforcement.

[0050] When the frequency data of a bridge cable or sling changes and exceeds the normal range, the cable frequency data abnormality label is triggered; this usually indicates that the cable or sling may be at risk of loosening or breaking; if the bridge cable frequency monitoring data suddenly drops and is lower than the preset safety range, the "cable frequency abnormality" label may be triggered, indicating that the cable needs to be inspected and maintained.

[0051] When the vibration acceleration data of the bridge exceeds the preset normal range, the vibration acceleration abnormality label is triggered; this may reflect changes in the dynamic performance of the bridge structure or the existence of other abnormal conditions; if the vibration acceleration value generated by the bridge under wind load or vehicle load continues to increase and exceeds the preset safety range, it may trigger the "vibration acceleration abnormality" label, indicating that the dynamic performance of the bridge needs to be evaluated and corresponding vibration reduction measures need to be taken.

[0052] In one embodiment of the present invention, the distribution recognition module needs to obtain the abnormality label of the abnormal classification, and obtain the abnormality distribution coefficient and the relative safety factor according to the relevant position points of the abnormality label, and use these two coefficients as the output content of the distribution recognition module.

[0053] In this module, we first identify the location points corresponding to these abnormal labels. These location points will indicate the abnormal situation identified at a certain point on the bridge, and the location points will indicate what form of data in the current abnormal classification results will use the same label, as well as the distribution of this label at the corresponding time and situation. At this time, the abnormal distribution coefficient is used to indicate the distribution of the probability of occurrence of abnormal labels at the corresponding location points and the data ratio. For the relative safety factor, the data between the corresponding location points is solved to obtain the optimal matching path of the relevant abnormal labels at this time to obtain the distribution form of the data in these abnormal classification results.

[0054] Therefore, the anomaly distribution coefficient can be expressed as follows: obtain the collection point corresponding to the anomaly classification result, determine the location information of each collection point, the location information includes the location of the collection point, and the location corresponding to the data of the collection point in the bridge structure monitoring data, associate the location information of the collection point with the anomaly classification result, and obtain the location point corresponding to the anomaly label.

[0055] The way to associate at this time is to verify whether the information described by the abnormal label is consistent with the location of the collection point, and whether the corresponding data at the collection point is consistent with the content described by the abnormal label; these consistent collection points and abnormal labels are output to obtain relevant location points. At the same time, the collection point corresponding to the abnormal classification result indicates the location of the corresponding data in the current abnormal classification result. When the abnormal classification result is generated above, the data will be screened once, which may cause the data collected by some collection points to be filtered after no related abnormalities occur. Therefore, it is necessary to verify the collection points at this time to obtain the location points that need to be processed and identified later, and perform distribution analysis on the data at these location points to determine the relative distribution probability of each abnormal label and the probability of occurrence of abnormal labels.

[0056] Analyze the distribution of abnormal labels at the location points, calculate the abnormal distribution density of the abnormal labels at the corresponding location points, the probability of occurrence of abnormal labels and the relative distribution probability corresponding to the location points, consider the spatial or temporal characteristics of the actual distribution of the current abnormal labels, and obtain the abnormal distribution coefficient according to the abnormal distribution density, occurrence probability and relative distribution probability of the abnormal labels.

[0057] The anomaly distribution density is expressed as the inverse of the average distance between the location point containing the anomaly label and the nearest neighboring location point, and the ratio of this inverse to the sum of the inverses of all location points corresponding to the anomaly label is used as the anomaly distribution density at this time. At this time, the distance between the location point currently required to be identified and the adjacent location point will be selected, and the inverse will be taken after the average value is calculated. At the same time, all location points corresponding to the anomaly label currently required to be identified must be verified, and the inverses corresponding to these location points are calculated in turn and summed up. The ratio of the inverse of the location point corresponding to the anomaly label currently required to be identified to the total sum of the inverses is used as the anomaly distribution density that needs to be identified at this time.

[0058] The probability of occurrence of an abnormal label is expressed as the ratio of the number of currently identified abnormal labels to the total number of abnormal labels. The relative distribution probability corresponding to the position point is expressed as the ratio of the number of abnormal labels existing at the identified position point to the number of all abnormal labels.

[0059] Therefore, the abnormal distribution coefficient can be expressed as: ;in, represents the abnormal distribution coefficient, Represents the abnormal distribution density corresponding to the j-th abnormal label at the i-th position; Represents the probability of occurrence of the jth abnormal label; Represents the relative distribution probability of the i-th position point; the value range of i is 1 to I, and the value range of j is 1 to J.

[0060] This anomaly distribution coefficient represents the distribution of related anomaly labels and location points to describe the corresponding situation of the anomaly labels in the bridge monitoring process at this time.

[0061] The relative safety factor can be expressed as: obtaining the first time series and the second time series corresponding to the abnormal label and the location point. Since the bridge structure monitoring data has been divided into the form of time series when setting the abnormal index, the abnormal related data, abnormal classification results and other contents can also be considered as data related to the time series. The first time series and the second time series are different from the form of dividing the abnormal index in the previous step. The first time series and the second time series are to obtain the relevant data of the current abnormal index, and express this part of the data in the form of time series. At the same time, the data is divided into two groups for comparison to determine the changes in the data at each location point at this time. For example, the first time series and the second time series will represent the data at adjacent location points, or the data at multiple non-adjacent location points. Finally, the calculated data will be mapped to the corresponding location points to reflect the related problems that may exist in this calculation process.

[0062] Construct a cumulative distance matrix related to the first time series and the second time series. The length of the matrix is ​​the length of the first time series and the length of the second time series. Each value in the matrix represents the distance from the corresponding value in the first time series to the corresponding value in the second time series. This distance can be obtained by calculating the distance between vectors. At the same time, this value will be added to the minimum value of the adjacent points of the currently calculated point to realize the calculation of the optimal matching path. The reason for choosing the vector calculation formula is that when the abnormal index is obtained, the characteristic vector of the relevant data has been obtained. In this scheme, the distance, data and can all be obtained by vector calculation.

[0063] Starting from one end of the cumulative distance matrix, the connection is made along the direction of the minimum cumulative distance to find the current optimal matching path. The final optimal matching path represents the shortest distance between the first time series and the second time series to verify the relative distance between these location points at this time.

[0064] According to the selected optimal matching path, the abnormal distribution density of each location point on the optimal matching path is calculated to obtain the path abnormal distribution density of the path segment; the risk weight of the location point on the path abnormal distribution density is obtained, and the relative safety factor is calculated.

[0065] The path anomaly distribution density is expressed as the ratio of the sum of the anomaly distribution densities of all location points on the current optimal matching path to the sum of the anomaly distribution densities of all location points; the risk weight will be set according to the anomaly labels existing at the location points, which can be expressed as the ratio of the indicator value corresponding to the anomaly label to the sum of the indicator values ​​of all anomaly labels, and the sum of the ratios of all anomaly labels at the location point is used as the risk weight set at this time; finally, the relative safety factor is calculated based on the path anomaly distribution density and risk weight.

[0066] The relative safety factor is expressed as: ;in, Expressed as a relative safety factor, Expressed as an exponential constant, Expressed as risk weight, It is expressed as the path anomaly distribution density, It represents the standard value of the path anomaly distribution density. This standard value represents the average value of the path anomaly distribution density calculated in the historical data. At this time, the ratio of the anomaly label of the location point on the optimal matching path to all anomaly labels is used as the risk weight used at this time.

[0067] Finally, the anomaly distribution coefficient and relative safety factor obtained at this time are output to complete the analysis of the current anomaly labels and location points. By analyzing the distribution of location points corresponding to the anomaly labels, it is possible to find out which areas have a higher frequency of abnormal activities, helping to locate potential sources of risk; using the anomaly distribution coefficient, it is possible to measure the density of anomalies at specific location points relative to the overall level, revealing whether the anomaly distribution is uniform or whether there is a clustering phenomenon; calculating the relative safety factor helps to compare the safety between different location points and provide decision support for planning the safest path; through detailed verification and analysis of anomaly labels, the system can better capture the real anomaly pattern and improve the ability to predict unknown anomalies. After understanding the anomaly distribution, monitoring resources can be allocated more specifically, such as increasing the sensor density in certain areas or strengthening inspection efforts; ultimately making the bridge monitoring process more dynamic and specific, and improving the overall monitoring effect.

[0068] In one embodiment of the present invention, the difference assessment module needs to analyze the temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the bridge structure monitoring data to determine whether there are corresponding differences in the current processing, and use the difference assessment coefficient to describe the difference.

[0069] The differences identified in this module represent the trend difference, deviation, and probability of the deviation and trend difference corresponding to the bridge structure monitoring data at the current time point. These three values ​​are used to quantify the differences generated at this time.

[0070] When verifying these data, the identified bridge structure monitoring data are uniformly normalized, and the time points corresponding to the temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the bridge structure monitoring data are obtained. There are N time points, and the relevant data at these time points are calculated to obtain the difference evaluation coefficient.

[0071] like Figure 4 As shown, the method for obtaining the difference evaluation coefficient is as follows: normalizing the bridge structure monitoring data, sequentially obtaining the trend difference and deviation of the bridge structure monitoring data at corresponding time points, and calculating the probability value of the co-occurrence of the corresponding deviation and trend difference to obtain the difference evaluation coefficient.

[0072] The trend difference is represented by the difference between the slope and intercept after fitting the bridge structure monitoring data at the current time point, and the deviation is represented by the difference between the value at the corresponding time point in the bridge structure monitoring data and the predicted value; since there are multiple data types in the bridge structure monitoring data judged at this time, the bridge structure monitoring data is divided into three groups to calculate the corresponding trend difference and deviation; these three groups can be represented as environmental factor group, structural response group and dynamic characteristic group; these three groups are calculated separately to get the comprehensive result at this time; the environmental factor group includes temperature and humidity data; the structural response group includes strain, deflection and cable frequency data; the dynamic characteristic group covers vibration acceleration data.

[0073] Therefore, the calculation method of the trend difference includes performing linear regression on the bridge structure monitoring data, obtaining a first linear curve related to temperature and humidity after fitting the bridge structure monitoring data, obtaining a second linear curve related to strain, deflection and cable frequency data, and obtaining a third linear curve related to vibration acceleration; calculating the slope value and intercept value of the first linear curve, the second linear curve and the third linear curve to obtain the trend difference.

[0074] The first linear curve is a linear regression of temperature and humidity. After fitting the relevant data, a linear curve that changes with time points is obtained, and the slope value and intercept value at the corresponding time point are calculated. At this time, temperature and humidity represent two values. The obtained slope value and intercept value will change with the change of temperature and humidity. At the same time, the corresponding changes of temperature and humidity at the time point selected at this time will not be too obvious. At this time, the maximum value of temperature and humidity in the calculated slope value and intercept value will be selected as the slope value and intercept value of the first linear curve at this time; the second linear curve is a multiple regression analysis of strain, deflection and cable frequency data to obtain the slope and intercept of the multiple regression at this time. At this time, deflection can be selected as the dependent variable, strain and cable frequency data can be selected as independent variables for linear regression analysis, and the intercept and slope calculated at this time are used as the intercept and slope of the second linear curve; the third linear curve is a linear regression of vibration acceleration to obtain the slope and intercept corresponding to the vibration acceleration, so as to obtain the intercept and slope expressed by the third linear curve.

[0075] The trend difference can be expressed as: ;in, Indicates the trend difference at time point t, represents the slope of the first linear curve at time point t, represents the slope of the second linear curve at time point t, represents the slope of the third linear curve at time point t, represents the intercept of the first linear curve at time point t, represents the intercept of the second linear curve at time point t, represents the intercept of the third linear curve at time point t, represents the weight coefficient of the slope, Represents the weight coefficient of the intercept. The two weight coefficients can be set to 0.5 and 0.5 respectively in the order of slope and weight. At this time, the slope and intercept both represent corresponding dimensionless values ​​to realize the comprehensive calculation of the trend difference at this time.

[0076] When calculating the deviation, the comprehensive deviation of the data in the bridge structure monitoring data is used as the deviation at this time. The deviation at each time point can be expressed as the weighted average of the deviation values ​​of temperature and humidity, strain, deflection, cable frequency data and vibration acceleration. These data are subtracted from the predicted value to obtain the deviation value, and then weighted averaging is performed. The weights used can be set according to the proportion of the corresponding values ​​of temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the corresponding data volume in the historical data.

[0077] After the deviation amount and the trend difference amount are obtained, the difference evaluation coefficient combines the deviation amount and the trend difference amount with the corresponding co-occurrence probability to obtain the difference evaluation coefficient at this time.

[0078] ;in, represents the difference evaluation coefficient, Indicates the trend difference at time point t, represents the deviation at time point t, Indicates the number of time points, the value of t ranges from 1 to T, It represents the probability value of the deviation and trend difference at the time point t. This probability value can represent the probability value of the deviation and trend difference in the historical data. The difference evaluation coefficient helps to determine which parts or parameters have changed most significantly, thereby guiding the maintenance team to develop targeted maintenance plans, optimize resource allocation, and reduce maintenance costs. Through continuous monitoring and difference evaluation, problems that may affect the safety of the bridge can be discovered and resolved in a timely manner to ensure the safety of traffic on the bridge.

[0079] In one embodiment of the present invention, the health monitoring module ultimately comprehensively identifies the difference assessment coefficient, the relative safety coefficient and the abnormal distribution coefficient to obtain a comprehensive health monitoring result at three levels. This health monitoring result is represented as a health assessment of the current bridge result.

[0080] In this module, these three coefficients are solved comprehensively to identify the specific manifestations of the bridge structure under relevant circumstances, so as to obtain the health monitoring results of the current bridge results.

[0081] Therefore, the health monitoring result is expressed as: the difference assessment coefficient, relative safety factor and abnormal distribution coefficient are comprehensively calculated to obtain the health monitoring coefficient, and the health monitoring coefficient is output as the health monitoring result.

[0082] ;in, represents the health monitoring coefficient, represents the difference evaluation coefficient, Expressed as a relative safety factor, represents the abnormal distribution coefficient, , , Represents the weight coefficient, and the values ​​of the weight coefficient are set to 0.3, 0.4, and 0.3 respectively.

[0083] The difference assessment coefficient can measure the degree of discreteness of the data, the relative safety factor reflects the safety margin of the structure, and the abnormal distribution coefficient can reveal the abnormal state in the data. Combining the three, a more comprehensive and detailed assessment of the bridge structure can be made; by comprehensively considering multiple coefficients, the limitations of a single coefficient can be compensated, thereby improving the accuracy of the assessment results; when one or more coefficients exceed the preset threshold, an early warning mechanism can be triggered to remind management personnel to take timely measures to prevent accidents and improve monitoring efficiency and accuracy.

[0084] Suppose a bridge is undergoing health monitoring, and a large amount of data is collected through sensors. First, the data is preprocessed to ensure the accuracy and consistency of the data; the data is monitored and analyzed in real time using the abnormal distribution coefficient, and the vibration data of the bridge in a certain period of time is found to be abnormal; then the difference evaluation coefficient is calculated, and it is found that the discrete degree of some data is large, and there may be abnormalities; then the relative safety factor is calculated, and it is found that the safety margin of the bridge under certain loads is small; finally, the three coefficients are comprehensively identified to conclude that the current bridge structure has potential hidden dangers; based on this conclusion, managers can take timely measures to maintain and reinforce it to ensure the safe operation of the bridge.

[0085] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. A bridge structure health monitoring system based on intelligent sensor network and deep learning, characterized by: include: A data acquisition module is used to set multiple acquisition points on the bridge to obtain bridge structure monitoring data related to the bridge, wherein the bridge structure monitoring data includes temperature and humidity, strain, deflection, cable frequency data and vibration acceleration on the bridge; The data monitoring module is used to match and connect the bridge structure monitoring data, set the abnormal indicators corresponding to the bridge structure monitoring data, and obtain the abnormal related data corresponding to the abnormal indicators; The modal anomaly analysis module is used to identify anomaly-related data, determine anomaly labels of anomaly-related data, and cluster them according to anomaly labels to obtain anomaly classification results; The distribution identification module is used to verify the abnormal labels of the abnormal classification results after clustering, obtain the location points corresponding to the abnormal labels, determine the abnormal distribution coefficient related to the distribution of abnormal labels and location points, and the relative safety factor between each location point; The difference assessment module is used to analyze the temperature and humidity, strain, deflection, cable frequency data and vibration acceleration in the bridge structure monitoring data one by one, determine the difference in each measurement, and obtain the difference assessment coefficient; The health monitoring module is used to comprehensively identify the difference assessment coefficient, relative safety factor and abnormal distribution coefficient to obtain the health monitoring results corresponding to the current bridge structure; The method for obtaining the difference evaluation coefficient is as follows: normalizing the bridge structure monitoring data, sequentially obtaining the trend difference and deviation of the bridge structure monitoring data at corresponding time points, and calculating the probability value of the corresponding deviation and trend difference appearing together to obtain the difference evaluation coefficient; The calculation method of the trend difference includes: performing linear regression on the bridge structure monitoring data, obtaining a first linear curve related to temperature and humidity after fitting the bridge structure monitoring data, obtaining a second linear curve related to strain, deflection and cable frequency data, and obtaining a third linear curve related to vibration acceleration; calculating the slope value and intercept value of the first linear curve, the second linear curve and the third linear curve to obtain the trend difference; The deviation amount is expressed as the comprehensive deviation of the data in the bridge structure monitoring data as the deviation amount used at this time; The difference evaluation coefficient is expressed as: ； in, represents the difference evaluation coefficient, Indicates the trend difference at time point t, represents the deviation at time point t, Indicates the number of time points, the value of t ranges from 1 to T, It indicates the probability value of the deviation and trend difference occurring together at time point t.

2. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 1 is characterized in that: The method for obtaining abnormal indicators is to obtain the corresponding time series of bridge structure monitoring data, verify the fluctuation amount of the bridge structure monitoring data in each cycle under the corresponding time series, and observe the periodic influence value between each variable in the bridge structure monitoring data in each cycle when the time series changes, combine the fluctuation amount and the periodic influence value, and identify the abnormal indicators in the bridge structure monitoring data.

3. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 2 is characterized in that: The way of combining the fluctuation amount and the periodic impact value can be expressed as: calculating the fluctuation amount and the periodic impact value to obtain the abnormal index; ； in, Indicates abnormal indicators. represents the fluctuation amount, represents the periodic impact value, represents the standard value of the fluctuation amount, Indicates the standard value of the period impact value.

4. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 1 is characterized in that: The implementation methods of the modal anomaly analysis module include: Obtain the distribution ratio of the abnormal related data and the collection points, set multiple clusters for the abnormal related data corresponding to the maximum distribution ratio, cluster the abnormal related data according to the clustering clusters, calculate the distance between each abnormal related data and the clustering cluster, and recalculate the center of each clustering cluster until the center of the clustering cluster no longer changes significantly, output the clustering centers corresponding to the multiple clustering clusters at this time as the clustering result, and output the obtained clustering result as the abnormal classification result.

5. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 1 is characterized in that: The anomaly distribution coefficient can be expressed as follows: obtain the collection point corresponding to the anomaly classification result, determine the location information of each collection point, associate the location information of the collection point with the anomaly classification result, and obtain the location point corresponding to the anomaly label; Analyze the distribution of abnormal labels at the location points, calculate the abnormal distribution density of abnormal labels at the corresponding location points, the occurrence probability of abnormal labels and the relative distribution probability corresponding to the location points, and obtain the abnormal distribution coefficient.

6. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 5 is characterized in that: The abnormal distribution coefficient can be expressed as: ； in, represents the abnormal distribution coefficient, Represents the abnormal distribution density corresponding to the j-th abnormal label at the i-th position; Represents the probability of occurrence of the jth abnormal label; Represents the relative distribution probability of the i-th position point; the value range of i is 1 to I, and the value range of j is 1 to J.

7. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 1 is characterized in that: The relative safety factor is expressed as: Obtain the first time series and the second time series corresponding to the abnormal label and the location point, construct a cumulative distance matrix related to the first time series and the second time series, start from one end of the cumulative distance matrix, connect along the direction of the minimum cumulative distance, and find the current optimal matching path; according to the selected optimal matching path, calculate the abnormal distribution density of each location point on the optimal matching path, and obtain the path abnormal distribution density of the path segment; obtain the risk weight of the location point on the path abnormal distribution density, and calculate the relative safety factor; The relative safety factor is expressed as: ;in, Expressed as a relative safety factor, Expressed as an exponential constant, Expressed as risk weight, It is expressed as the path anomaly distribution density, The standard value representing the path anomaly distribution density.

8. The bridge structure health monitoring system based on intelligent sensor network and deep learning according to claim 1 is characterized in that: The health monitoring result is expressed as: the difference assessment coefficient, the relative safety coefficient and the abnormal distribution coefficient are comprehensively calculated to obtain the health monitoring coefficient, and the health monitoring coefficient is output as the health monitoring result; ； in, represents the health monitoring coefficient, represents the difference evaluation coefficient, Expressed as a relative safety factor, represents the abnormal distribution coefficient, , , Represents the weight coefficient.

Images