A bridge monitoring data outlier intelligent judgment and processing system

Abstract

The application provides a bridge monitoring data abnormal value intelligent judgment and processing system, comprising a monitoring module, which is used for non-contact acquisition of dynamic displacement data of multiple monitoring points of a bridge; a preprocessing module, which is used for time synchronization and quality fusion of displacement, load and environmental data to generate multidimensional time sequence data; an abnormality identification module, which is used for extraction of cooperative deformation and dynamic fingerprint features and abnormality identification and initial classification based on feature deviation degree; an abnormality diagnosis module, which is used for abnormality cause analysis in combination with load and environmental conditions to output abnormality diagnosis results; and a strategy generation module, which is used for generation of hierarchical early warning and disposal strategies according to diagnosis results and structure key index usage degree. The application can improve the accuracy and efficiency of bridge structure safety monitoring and guarantee the safe operation of the bridge.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering data processing, and more particularly, to a bridge monitoring data abnormal value intelligent judgment and processing system. BACKGROUND

[0002] In the field of bridge engineering, with the increase of traffic flow and the growth of service time, the health monitoring of bridge structure becomes particularly important. The traditional bridge monitoring method mainly relies on manual inspection and simple sensor data collection, and these methods have problems such as low efficiency, incomplete data, and delayed response to abnormal situations. In recent years, with the development of sensor technology, although a large amount of monitoring data can be collected, these data often have noise interference, difficulty in fusing multi-source data, and lack of intelligent abnormality judgment.

[0003] In the implementation process of the present application, there are at least the following problems or defects in the prior art: it is difficult to effectively fuse multi-source monitoring data for comprehensive analysis, it is difficult to quickly and accurately identify and diagnose the abnormal situation of the bridge structure, and there is a lack of adaptive early warning and disposal strategy, resulting in insufficient intelligent level of the bridge monitoring system, which cannot meet the needs of modern bridge safety management. SUMMARY

[0004] The present application provides a bridge monitoring data abnormal value intelligent judgment and processing system, comprising:

[0005] A monitoring module for non-contact synchronous acquisition of dynamic displacement data of multiple monitoring points in a specified area of a bridge;

[0006] A preprocessing module for time synchronization, spatial registration and quality fusion of the dynamic displacement data, synchronously collected load data and environmental data, to generate multi-dimensional time series data;

[0007] An abnormality identification module for extracting cooperative deformation features representing the overall working performance of the structure and dynamic fingerprint features representing the dynamic characteristics of the structure from the multi-dimensional time series data, and performing abnormality identification and preliminary classification based on feature deviation degree, and outputting abnormal preliminary category information;

[0008] An abnormality diagnosis module for receiving the abnormal preliminary category information, and performing deep diagnosis and cause correlation analysis in combination with load working conditions and environmental conditions, and outputting abnormal diagnosis results including abnormal influence level;

[0009] A strategy generation module for generating graded early warning information and corresponding disposal strategies according to the abnormal diagnosis results and the structure key indicator usage degree calculated based on the dynamic displacement data.

[0010] Further, the monitoring module comprises:

[0011] a radar scanning control unit configured to control the microwave radar to scan at least one cross-section or longitudinal-section region of the bridge, and to take a plurality of girders or a plurality of preset monitoring points on the same girder as synchronous monitoring targets;

[0012] a dynamic displacement calculation unit configured to calculate a displacement time-history curve of each of the monitoring points in a three-dimensional space in real time from the radar echo signal, the displacement time-history curve comprising a static displacement component and a dynamic vibration component;

[0013] a data packaging unit configured to package dynamic displacement data of all the monitoring points obtained in a same scanning period and radar state data into a structured data frame.

[0014] Further, the preprocessing module comprises:

[0015] a space-time reference unification unit configured to unify the displacement data of each of the monitoring points in the structured data frame, vehicle passing time sequence data collected by the video unit, and temperature and humidity data collected by the environmental sensor to a same time reference and a space coordinate system;

[0016] a load response correlation unit configured to mark a corresponding load action period in the displacement time-history curve according to the vehicle passing time sequence data, and extract a peak displacement response under load and a corresponding time point;

[0017] a data quality fusion unit configured to calculate a comprehensive quality confidence coefficient for each frame of the multi-dimensional time sequence data according to a signal-to-noise ratio of the radar signal, a data packet loss rate, and an environmental interference intensity.

[0018] Further, the anomaly identification module comprises:

[0019] a feature extraction unit configured to calculate a correlation coefficient matrix or a transverse distribution influence line of a displacement time-history of a plurality of girder monitoring points synchronously monitored in a same cross-section as the cooperative deformation feature;

[0020] a dynamic fingerprint feature extraction unit configured to perform frequency spectrum analysis or modal analysis on a displacement time-history curve of each of the monitoring points to extract a fundamental frequency, a damping ratio, or a vibration mode of a specified order of the structure as the dynamic fingerprint feature;

[0021] an anomaly preliminary judgment unit configured to compare the cooperative deformation feature and the dynamic fingerprint feature extracted in a current monitoring period with a health reference model or a historical statistical reference, calculate a feature deviation degree, and determine that an anomaly occurs in a corresponding feature dimension when the feature deviation degree exceeds a preset threshold, and classify the anomaly preliminary category information as a cooperative deformation anomaly or a dynamic characteristic anomaly.

[0022] Further, the feature deviation degree calculation in the abnormality preliminary judgment unit comprises:

[0023] calculating the transverse distribution coefficient of the displacement response of each beam body monitoring point of the current section wherein , is the peak displacement of the i-th beam piece, is the sum of the peak displacements of all beam pieces of the current section;

[0024] comparing the current calculated transverse distribution coefficient vector with the reference transverse distribution coefficient vector

[0025] quantifying the feature deviation degree by calculating the Euclidean distance or the cosine similarity between the two vectors, wherein the Euclidean distance , and the cosine similarity ;

[0026] wherein denotes the square root operation, and the sum of the corresponding operation results of all beam piece numbers i.

[0027] Further, the abnormality diagnosis module comprises:

[0028] an abnormality pattern comparison unit, which stores a typical abnormality pattern library containing change patterns caused by structural damage, support failure, hinge joint damage, and single plate stress;

[0029] a cause correlation reasoning unit configured to match the abnormality preliminary category information and the corresponding abnormality features output by the abnormality preliminary judgment unit with the typical abnormality pattern library, correlate the load type, weight, and environmental data at the same time, and give a probabilistic judgment of the possible causes of the abnormality;

[0030] an influence evaluation unit configured to evaluate the influence level of the abnormality on the overall safety and applicability of the structure based on the probabilistic judgment of the possible causes of the abnormality and the abnormality feature deviation degree, and generate an abnormality diagnosis result containing the abnormality influence level.

[0031] Further, the strategy generation module comprises:

[0032] a key indicator dynamic calculation unit configured to dynamically calculate bridge key indicators based on the dynamic displacement data, the key indicators including dynamic deflection usage and impact coefficient usage ;

[0033] ​​Wherein, the dynamic deflection usage degree ;

[0034] The impact coefficient usage degree ;

[0035] The measured dynamic deflection peak value, The no-load reference value, The design allowable deflection, The measured impact coefficient, The design impact coefficient;

[0036] The hierarchical early warning triggering unit is pre-set with hierarchical early warning thresholds associated with the impact level in the abnormal diagnosis result and the key indicator usage degree, and triggers early warning of the corresponding level when the condition is met;

[0037] The strategy mapping unit maps to generate corresponding data handling suggestions and structure inspection or maintenance suggestion strategies according to the triggered early warning level and the abnormal diagnosis result.

[0038] Further, the working logic of the hierarchical early warning triggering unit is based on the usage degree output by the key indicator dynamic calculation unit and the abnormal impact level in the abnormal diagnosis result for comprehensive judgment:

[0039] When the abnormal impact level is slight, and all key indicator usage degrees do not exceed the first usage degree threshold, trigger attention level early warning;

[0040] When the abnormal impact level is moderate, or any key indicator usage degree exceeds the first usage degree threshold but does not exceed the second usage degree threshold, trigger warning level early warning;

[0041] When the abnormal impact level is severe, or any key indicator usage degree exceeds the second usage degree threshold, trigger alarm level early warning;

[0042] Wherein, the second usage degree threshold is greater than the first usage degree threshold.

[0043] Further, it further comprises:

[0044] The dynamic maintenance module is used for dynamically updating the health reference model used by the abnormal preliminary judgment unit by using the multi-dimensional time series data of the monitoring period when the abnormal diagnosis module does not output abnormal diagnosis results and the comprehensive quality confidence coefficient output by the data quality fusion unit is higher than the set threshold in the monitoring period.

[0045] Wherein, the dynamic maintenance module specifically performs the following update operations:

[0046] a) calling the feature extraction unit, calculating the average lateral distribution coefficient vector under multiple load events in the period, for updating the statistical benchmark of the cooperative deformation feature;

[0047] b) calling the dynamic fingerprint feature extraction unit, extracting the average fundamental frequency and damping ratio of the structure vibration in the period, for updating the statistical benchmark of the dynamic fingerprint feature.

[0048] Further, the dynamic maintenance module adopts a recursive statistical algorithm with time decay weighting for dynamic updating;

[0049] For the statistical benchmark update of the cooperative deformation feature, the recursive formula is:

[0050]

[0051] wherein, represents the updated benchmark lateral distribution coefficient vector, represents the average lateral distribution coefficient vector calculated in the current effective period, represents the benchmark lateral distribution coefficient vector before updating, represents a forgetting factor between 0 and 1, for controlling the influence weight of historical data;

[0052] For the statistical benchmark update of the dynamic fingerprint feature, the same recursive logic is adopted to update the average fundamental frequency and average damping ratio.

[0053] The above embodiments according to the present application have at least the following beneficial effects:

[0054] 1. Through the monitoring module, non-contact synchronous acquisition of dynamic displacement data of multiple monitoring points of the bridge is realized, the displacement change of the bridge under different loads can be obtained in real time, the problems of limited monitoring points, incomplete data acquisition and easy interference of contact measurement in the traditional monitoring method are solved, the accuracy and timeliness of the monitoring data are improved, and a more comprehensive and reliable data basis is provided for the bridge structure health monitoring.

[0055] 2. The preprocessing module is introduced to time synchronize, space register and quality fuse the dynamic displacement data with the synchronously collected load data and environmental data, and generate multi-dimensional time sequence data. This technical means effectively solves the fusion difficulty problem caused by the inconsistent time benchmarks, unmatched space coordinates and uneven data quality among multiple source data, enables the monitoring system to comprehensively consider the influence of load and environmental factors on the bridge structure, and thus more accurately reflects the actual working state of the bridge, and provides high-quality data support for subsequent anomaly identification and diagnosis.

[0056] 3. The combination of the anomaly identification module and the anomaly diagnosis module realizes the extraction of the collaborative deformation feature representing the overall performance of the structure in the horizontal direction and the dynamic fingerprint feature representing the dynamic performance of the structure from the multi-dimensional time sequence data, and performs anomaly identification and preliminary classification based on the feature deviation degree, and then performs deep diagnosis and cause correlation analysis in combination with the load working condition and the environmental condition. A series of technical measures effectively solve the problems of inaccurate anomaly identification, non-in-depth diagnosis and difficulty in determining the causes of the anomaly of the traditional monitoring system, enable the monitoring system to quickly and accurately identify the abnormal state of the bridge structure, analyze the possible causes, provide a scientific basis for timely taking targeted maintenance measures, and improve the intelligent level and safety guarantee capability of the bridge structure health monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0057] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description read in conjunction with the accompanying drawings. In the drawings, several embodiments of the present application are illustrated by way of example in which:

[0058] Figure 1 A structural schematic diagram of a bridge monitoring data abnormal value intelligent judgment and processing system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0059] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0060] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present application can be embodied in the form of a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0061] It should be noted that the number of any elements in the drawings is used for example only and not limitation, and any naming is only used for distinction and does not have any limiting meaning.

[0062] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art. Figure 1 Figure 1 A structural schematic diagram of a bridge monitoring data abnormal value intelligent judgment and processing system provided by an embodiment of the present application. As shown in Figure 1 the drawing, a bridge monitoring data abnormal value intelligent judgment and processing system includes:

[0063] ​The monitoring module 101 is configured to non-contactingly and synchronously collect dynamic displacement data of a plurality of monitoring points in a designated area of the bridge.

[0064] In some embodiments, the monitoring module comprises:

[0065] The radar scanning control unit is configured to control the microwave radar to scan at least one cross-section or longitudinal-section area of the bridge, and take a plurality of beam bodies or a plurality of preset monitoring points on the same beam body as synchronous monitoring targets.

[0066] The microwave radar is a technology that uses the principle of microwave signal reflection to measure distance, and can obtain displacement information of an object in real time without contacting the object. The radar scanning control unit controls the scanning range and frequency of the radar to ensure comprehensive coverage of the key areas of the bridge. The radar scanning control unit can be configured to scan the cross-section or longitudinal-section area of the bridge, and take a plurality of beam bodies or a plurality of preset monitoring points on the same beam body as synchronous monitoring targets. For example, in the key parts of the bridge, such as the vicinity of the support, the mid-span position of the beam body, etc., high-density monitoring points are set to obtain more detailed dynamic displacement data.

[0067] The radar scanning control unit can dynamically adjust the scanning frequency and scanning range according to the specific structure of the bridge and the monitoring requirements. For example, in areas with high traffic flow or complex structure of the bridge, the scanning frequency can be appropriately increased to obtain more detailed dynamic displacement data.

[0068] The dynamic displacement calculation unit is configured to calculate the displacement time history curve of each monitoring point in three-dimensional space from the radar echo signal in real time, and the displacement time history curve includes a static displacement component and a dynamic vibration component. The dynamic displacement calculation unit extracts the displacement time history curve of each monitoring point from the radar echo signal, which includes a static displacement component and a dynamic vibration component, and can comprehensively reflect the displacement change of the bridge under different loads.

[0069] The dynamic displacement calculation unit calculates the displacement time history curve of each monitoring point in three-dimensional space from the radar echo signal in real time, which includes a static displacement component reflecting the displacement state of the bridge under long-term load and a dynamic vibration component reflecting the vibration response of the bridge under dynamic load such as vehicles. When calculating the displacement time history curve, a signal processing algorithm such as Kalman filtering algorithm can be used. When packing data, time stamp and spatial coordinate information are added to the displacement data of each monitoring point to ensure traceability and accuracy of the data in subsequent processing. For example, the time stamp can be accurate to the millisecond level, and the spatial coordinate information can be determined by GPS positioning or a preset coordinate system of the bridge structure. In this way, the displacement data of each monitoring point not only contains the displacement value, but also contains the time information and spatial position information.

[0070] a data packaging unit configured to package the dynamic displacement data of all the monitoring points obtained in the same scanning period and the radar state data into a structured data frame. The data packaging unit integrates the dynamic displacement data of all the monitoring points in the same scanning period to form the structured data frame.

[0071] The data packaging unit packages the dynamic displacement data of all the monitoring points obtained in the same scanning period and the radar state data, such as the working frequency of the radar and the scanning angle, into a structured data frame. This data frame format facilitates data fusion and processing by subsequent modules. For example, the scanning frequency of the radar scanning control unit can be adjusted according to the dynamic characteristics of the bridge, and is usually set to several times to tens of times per second to ensure that the dynamic response of the bridge when the vehicle passes through can be captured.

[0072] a preprocessing module 102 configured to perform time synchronization, spatial registration and quality fusion on the dynamic displacement data, the synchronously collected load data and the environmental data to generate multi-dimensional time series data;

[0073] In some embodiments, the preprocessing module includes:

[0074] a space-time reference unification unit configured to unify the displacement data of each monitoring point in the structured data frame, the vehicle passing time series data collected by the video unit and the temperature and humidity data collected by the environmental sensor to the same time reference and spatial coordinate system;

[0075] The space-time reference unification unit unifies the dynamic displacement data collected by the microwave radar, the vehicle passing time series data collected by the video unit and the temperature and humidity data collected by the environmental sensor to the same time reference and spatial coordinate system. The unification of the time reference is achieved by high-precision clock synchronization technology, for example, GPS clock synchronization, to ensure the synchronization of all data in time. The unification of the spatial coordinate system is achieved by geographic information system (GIS) technology, which maps the displacement data of all monitoring points to a unified geographic coordinate system.

[0076] a load response correlation unit configured to mark the corresponding load action period in the displacement time history curve according to the vehicle passing time series data, and extract the peak displacement response under the load and the corresponding time;

[0077] The load response correlation unit marks the time period corresponding to the load action in the displacement time history curve by analyzing the vehicle passing time sequence data, and extracts the peak displacement response and its corresponding time. The peak displacement response refers to the maximum displacement value of the bridge structure under a certain load action, and the corresponding time refers to the specific time point at which the peak displacement occurs. When marking the load action period, the load response correlation unit can obtain the accurate time sequence of vehicle passing through vehicle detection sensors such as ground coils or video analysis algorithms, and perform correlation analysis on these time points and displacement data. For example, when a vehicle passes through a certain monitoring point of the bridge, its passing time is recorded, and the corresponding peak displacement response is found on the displacement time history curve.

[0078] The data quality fusion unit is configured to calculate a comprehensive quality confidence coefficient for each frame of the multi-dimensional time sequence data according to a signal-to-noise ratio of radar signals, a data packet loss rate, and an environmental interference intensity.

[0079] The data quality fusion unit calculates a comprehensive quality confidence coefficient for each frame of data according to a signal-to-noise ratio of radar signals, a data packet loss rate, and an environmental interference intensity. The signal-to-noise ratio reflects the quality of the radar signals, the data packet loss rate indicates the integrity during data transmission, and the environmental interference intensity reflects the degree of influence of the external environment on data acquisition. The comprehensive quality confidence coefficient is a quantitative index for evaluating the overall quality of the data, and a weighted average method can be used to assign a quality score to each frame of data according to the specific values of the signal-to-noise ratio, the data packet loss rate, and the environmental interference intensity. For example, the signal-to-noise ratio weight can be set to 0.4, the data packet loss rate weight to 0.3, and the environmental interference intensity weight to 0.3.

[0080] The abnormality identification module 103 is configured to extract a cooperative deformation feature representing the overall lateral working performance of the structure and a dynamic fingerprint feature representing the dynamic characteristics of the structure from the multi-dimensional time sequence data, and perform abnormality identification and preliminary classification based on feature deviation degree, and output abnormality preliminary classification information; specifically, the abnormality identification module includes:

[0081] The feature extraction unit is configured to calculate a correlation coefficient matrix or a lateral distribution influence line of displacement time history for a plurality of beam body monitoring points synchronously monitored at the same section as the cooperative deformation feature. The feature extraction unit calculates a correlation coefficient matrix or a lateral distribution influence line of displacement time history for a plurality of beam body monitoring points synchronously monitored at the same section. The correlation coefficient matrix refers to a matrix formed by calculating the correlation between the displacement data of a plurality of monitoring points to quantify the cooperative deformation relationship between the monitoring points. The lateral distribution influence line is a curve used to describe the deformation distribution of a plurality of beams under lateral load.

[0082] The correlation coefficient matrix can be calculated by using the Pearson correlation coefficient formula, by calculating the linear correlation between the displacement data of each monitoring point, to form a matrix to quantify the cooperative deformation relationship between these monitoring points. For example, for multiple beam monitoring points on the same section, the correlation coefficient of the displacement data of each monitoring point with the displacement data of other monitoring points is calculated to form a symmetric matrix.

[0083] The transverse distribution influence line can be drawn by normalizing the peak displacement of each monitoring point to intuitively reflect the deformation distribution of each monitoring point. The dynamic fingerprint feature extraction unit can use the Fast Fourier Transform (FFT) algorithm to convert the time-domain displacement time history curve into a frequency-domain signal during frequency spectrum analysis, thereby extracting the fundamental frequency and damping ratio of the structure. Modal analysis can extract the mode shape of the structure by eigenvalue decomposition and other methods.

[0084] The dynamic fingerprint feature extraction unit is configured to perform frequency spectrum analysis or modal analysis on the displacement time history curve of each monitoring point to extract the fundamental frequency, damping ratio, or mode shape of a specified order of the structure as the dynamic fingerprint feature. The dynamic fingerprint feature extraction unit performs frequency spectrum analysis or modal analysis on the displacement time history curve of each monitoring point to extract the fundamental frequency, damping ratio, or mode shape of a specified order of the structure. The fundamental frequency refers to the lowest natural frequency of the structure during vibration, the damping ratio reflects the degree of energy dissipation of the structure during vibration, and the mode shape describes the vibration pattern of the structure at a specific frequency.

[0085] The anomaly preliminary judgment unit is configured to compare the cooperative deformation features and the dynamic fingerprint features extracted in the current monitoring period with the health benchmark model or historical statistical benchmark, calculate the feature deviation, and determine that the corresponding feature dimension is abnormal when the feature deviation exceeds a preset threshold, and classify the preliminary anomaly category information as a cooperative deformation anomaly or a dynamic property anomaly.

[0086] The anomaly preliminary judgment unit compares the cooperative deformation features and the dynamic fingerprint features extracted in the current monitoring period with the health benchmark model or historical statistical benchmark to calculate the feature deviation. The health benchmark model is a model constructed based on the feature data of the bridge in a normal working state, which is used as a reference for determining whether the current state is abnormal. When the feature deviation exceeds a preset threshold, it is determined that the corresponding feature dimension is abnormal, and the anomaly is preliminarily classified as a cooperative deformation anomaly or a dynamic property anomaly.

[0087] Deviation can be calculated using methods such as Euclidean distance or cosine similarity. For example, for coordinated deformation features, the lateral distribution coefficient vector of the current monitoring period is compared with the baseline vector, and the Euclidean distance or cosine similarity between the two is calculated to quantify the degree of deviation. When the deviation exceeds a preset threshold, it is judged as an anomaly. The preset threshold can be adjusted according to historical data and the actual operation of the bridge. For example, the Euclidean distance threshold for coordinated deformation features can be set to 0.1, and the deviation threshold for dynamic fingerprint features can be set to 5%.

[0088] In some embodiments, the calculation of feature deviation includes:

[0089] Calculate the lateral distribution coefficient of the displacement response at each monitoring point of the beam in the current cross section. ,in , For the first Peak displacement of the beam This represents the summation of the peak displacements of all beam segments at the current cross-section;

[0090] The currently calculated lateral distribution coefficient vector vector of lateral distribution coefficients relative to the baseline Compare;

[0091] By calculating the Euclidean distance between the two vectors or cosine similarity To quantify feature deviation, where Euclidean distance is used. cosine similarity ;

[0092] in, This represents the square root operation. This indicates that the corresponding calculation results for all beam segments with index i are summed.

[0093] Lateral distribution coefficient By calculating the first Peak displacement of the beam The sum of the peak displacements of all beam segments at the current cross-section The ratio is used to obtain the peak displacement. Peak displacement refers to the maximum displacement response of the beam under load, which can intuitively reflect the degree of deformation of the beam at that moment. This involves summing the peak displacements of all beam segments within the same cross-section for normalization, ensuring that the lateral distribution coefficient remains within a reasonable range. The benchmark lateral distribution coefficient vector is obtained statistically from a large amount of monitoring data of the bridge under normal operating conditions, and includes the proportional relationship of the displacement response of each beam under typical load conditions.

[0094] In practical applications, the establishment of the benchmark transverse distribution coefficient vector needs to consider various typical load cases, such as different vehicle load distributions, thermal expansion and contraction effects under different environmental temperatures, etc. By statistically analyzing the monitoring data under these working conditions, a comprehensive benchmark vector is obtained. The calculation of the Euclidean distance first normalizes the peak displacement of each beam piece to eliminate the dimensional and order differences between different beam pieces, making the calculation results more comparable.

[0095] The calculation of the cosine similarity can be done by introducing a weight factor, which gives different weights to the peak displacements of different beam pieces to highlight the importance of key beam pieces in cooperative deformation. For example, higher weights can be given to main load-bearing beam pieces, while weights can be appropriately reduced for secondary beam pieces. In addition, when determining the deviation threshold, abnormal cases in historical monitoring data can be analyzed, combined with the structural characteristics and safety requirements of the bridge, to set the threshold range to ensure the accuracy and reliability of the initial judgment of abnormalities.

[0096] The abnormal diagnosis module 104 is configured to receive the preliminary abnormality classification information and perform deep diagnosis and cause correlation analysis in combination with load conditions and environmental conditions, and output abnormal diagnosis results including abnormal impact levels.

[0097] In some embodiments, the abnormal diagnosis module includes:

[0098] The abnormal pattern comparison unit stores a typical abnormal pattern library, which includes change patterns caused by structural damage, support failure, hinge joint damage, and single plate stress causes. The abnormal pattern comparison unit stores a typical abnormal pattern library, which includes change patterns caused by structural damage, support failure, hinge joint damage, and single plate stress causes. These patterns are constructed based on historical data and expert experience, and are used to match the current monitored abnormal characteristics.

[0099] Further, the typical abnormal pattern library can be constructed and optimized by machine learning algorithms. For example, support vector machines (SVM) or neural networks are used to train historical abnormal data to generate feature patterns corresponding to different causes. Input parameters include cooperative deformation features, dynamic fingerprint features, and related load and environmental data.

[0100] The cause correlation reasoning unit is configured to match the abnormal preliminary category information and the corresponding abnormal features output by the abnormal preliminary judgment unit with the typical abnormal mode library, and correlate the load type, weight and environmental data at the same time to give a probabilistic judgment of the possible causes of the abnormality. The cause correlation reasoning unit compares the abnormal features with the typical mode library, and analyzes the possible causes of the abnormality in combination with the load type, weight and environmental data at the same time. For example, if the monitored abnormal features match the support failure mode and the load is large at the time, it can be judged that the abnormality is caused by uneven force on the support. The cause correlation reasoning unit makes the cause judgment by using the Bayesian inference algorithm in combination with the prior probability and the current monitoring data to calculate the probability of each cause. For example, for a certain abnormal feature, the probability distribution of the causes such as structural damage and support failure is calculated.

[0101] The impact evaluation unit is configured to evaluate the impact level of the abnormality on the overall safety and applicability of the structure based on the probabilistic judgment of the possible causes of the abnormality and the deviation degree of the abnormal features, and generate an abnormal diagnosis result containing the impact level of the abnormality.

[0102] The impact evaluation unit evaluates the impact level of the abnormality on the overall safety and applicability of the structure according to the probabilistic judgment of the causes of the abnormality and the deviation degree of the features. The impact level is usually divided into three levels of slight, moderate and severe, which are used to guide the subsequent maintenance strategy. For example, a slight abnormality only needs to be monitored regularly, while a severe abnormality needs to be repaired immediately. When implemented, evaluation rules are set according to the degree of feature deviation from the benchmark and the probability of the cause. For example, when the feature deviation degree exceeds 10% and the cause probability points to structural damage, it is determined as a severe abnormality. In addition, in order to improve the accuracy of diagnosis, the typical abnormal mode library can be updated regularly, and new abnormal modes and causes discovered are included in the library to ensure that the system can adapt to the changes of the bridge structure and new abnormal conditions.

[0103] The strategy generation module 105 is configured to generate graded warning information and corresponding disposal strategies according to the abnormal diagnosis result and the structure key indicator usage degree calculated based on the dynamic displacement data.

[0104] In some embodiments, the strategy generation module includes:

[0105] The key indicator dynamic calculation unit is configured to dynamically calculate bridge key indicators based on the dynamic displacement data, wherein the key indicators include dynamic deflection usage degree and impact coefficient usage degree .

[0106] The dynamic deflection usage degree is calculated as follows:

[0107] The impact coefficient usage degree is calculated as follows:

[0108] This represents the measured peak dynamic deflection. This is the baseline value without load. To allow for deflection, The measured impact coefficient, The design impact coefficient;

[0109] The measured peak dynamic deflection is calculated by collecting real-time dynamic displacement data of the bridge during vehicle traffic, removing noise through a filtering algorithm. No-load reference value The allowable deflection can be determined by measuring when there are no vehicles on the bridge. This is determined according to bridge design specifications. Regarding the impact coefficient, the measured impact coefficient... The design impact factor can be obtained by analyzing the vibration response in the displacement-time history curve. It is determined according to the bridge design documents.

[0110] The tiered early warning triggering unit is pre-set with tiered early warning thresholds associated with the impact level in the anomaly diagnosis results and the usage of key indicators. When the conditions are met, an early warning of the corresponding level is triggered. For example, when the anomaly impact level is minor and the usage of all key indicators does not exceed the first usage threshold, a warning at the attention level is triggered; when the anomaly impact level is moderate or the usage of any key indicator exceeds the first usage threshold but does not exceed the second usage threshold, a warning at the alert level is triggered; when the anomaly impact level is severe or the usage of any key indicator exceeds the second usage threshold, an alarm at the alarm level is triggered.

[0111] The first usage threshold can be set to 0.8, indicating that an alert is triggered when the usage of a key indicator approaches 80% of the design allowable value; the second usage threshold can be set to 1.0, indicating that an alarm is triggered when the usage of a key indicator reaches the design allowable value. In practical applications, these thresholds can be calibrated based on historical data and expert experience to ensure the accuracy and reliability of the early warning system. Furthermore, the tiered early warning triggering unit can generate corresponding handling suggestions based on the warning level. For example, for a warning at the attention level, it is recommended to increase the monitoring frequency and record more data for further analysis; for a warning at the alert level, it is recommended to conduct local inspections, focusing on areas where problems may occur; for a warning at the alarm level, it is recommended to take immediate emergency measures, such as restricting traffic or carrying out emergency repairs, to ensure the safe operation of the bridge. In this way, the system can provide targeted handling suggestions based on different warning levels, helping maintenance personnel to take timely and effective measures.

[0112] comprehensive judgment is made based on the usage degree output by the key indicator dynamic calculation unit and the abnormal influence level in the abnormal diagnosis result:

[0113] When the abnormal influence level is slight and the usage degree of all key indicators does not exceed the first usage degree threshold, a notice-level warning is triggered;

[0114] When the abnormal influence level is moderate, or the usage degree of any key indicator exceeds the first usage degree threshold but does not exceed the second usage degree threshold, a warning-level warning is triggered;

[0115] When the abnormal influence level is severe, or the usage degree of any key indicator exceeds the second usage degree threshold, an alarm-level warning is triggered;

[0116] The abnormal influence level refers to the influence of the abnormality output by the abnormal diagnosis module on the overall safety and applicability of the structure, which is usually divided into three levels: slight, moderate and severe. The hierarchical warning triggering unit presets hierarchical warning thresholds associated with parameters and levels, and triggers warnings of corresponding levels when the conditions are met. For example, when the abnormal influence level is slight and the usage degree of all key indicators does not exceed the first usage degree threshold, a notice-level warning is triggered; when the abnormal influence level is moderate or the usage degree of any key indicator exceeds the first usage degree threshold but does not exceed the second usage degree threshold, a warning-level warning is triggered; when the abnormal influence level is severe or the usage degree of any key indicator exceeds the second usage degree threshold, an alarm-level warning is triggered.

[0117] The strategy mapping unit can formulate specific measures according to the warning level and the abnormal diagnosis result, combined with the bridge maintenance manual and expert experience when generating the treatment strategy. For example, when the notice-level warning is triggered, it is recommended to strengthen the monitoring frequency; when the warning-level warning is triggered, it is recommended to perform local inspection and maintenance; when the alarm-level warning is triggered, it is recommended to immediately take emergency measures such as traffic restriction or emergency repair.

[0118] In some embodiments, further comprising:

[0119] The dynamic maintenance module 106 is configured to, when the abnormal diagnosis module does not output an abnormal diagnosis result and the comprehensive quality confidence coefficient output by the data quality fusion unit is higher than a set threshold in a monitoring period, dynamically update the health benchmark model used by the abnormal preliminary judgment unit by using the multi-dimensional time series data in the period;

[0120] The dynamic maintenance module specifically performs the following update operations:

[0121] a) Call the feature extraction unit to calculate the average transverse distribution coefficient vector under multiple load events in the period, which is used to update the statistical benchmark of cooperative deformation characteristics;

[0122] b) calling the dynamic fingerprint feature extraction unit to extract the average fundamental frequency and damping ratio of the structure vibration in this period, for updating the statistical benchmark of dynamic fingerprint features.

[0123] The dynamic maintenance module initiates the updating operation when no abnormality is detected and the comprehensive quality confidence coefficient output by the data quality fusion unit is higher than the set threshold. The comprehensive quality confidence coefficient refers to the data quality score calculated by evaluating multiple factors such as radar signal signal-to-noise ratio, data packet loss rate, and environmental interference intensity, used to judge the reliability of the current monitoring data. When the coefficient is higher than the set threshold, it indicates that the current data has high credibility and can be used to update the health benchmark model.

[0124] The updating operation includes two aspects: one is to call the feature extraction unit to calculate the average lateral distribution coefficient vector under multiple load events in this period, for updating the statistical benchmark of collaborative deformation features; the other is to call the dynamic fingerprint feature extraction unit to extract the average fundamental frequency and damping ratio of the structure vibration in this period, for updating the statistical benchmark of dynamic fingerprint features. These updating operations are realized by time-decay-weighted recursive statistical algorithm, ensuring that the health benchmark model can dynamically reflect the current health status of the bridge.

[0125] Preferably, the dynamic maintenance module uses time-decay-weighted recursive statistical algorithm for dynamic updating. For the statistical benchmark update of collaborative deformation features, the recursive formula is: . Wherein, represents the updated benchmark lateral distribution coefficient vector, represents the average lateral distribution coefficient vector calculated in the current effective period, represents the benchmark lateral distribution coefficient vector before updating, represents the forgetting factor, which is a value between 0 and 1, used to control the influence weight of historical data. The value of the forgetting factor can be adjusted according to the actual operation of the bridge and the stability of the monitoring data. For example, when the bridge operation state is relatively stable, the value of can be appropriately reduced, making the model rely more on historical data; when the bridge operation state changes rapidly, the value of can be increased, making the model adapt to the current state more quickly. For the statistical benchmark update of dynamic fingerprint features, the same recursive logic is used to update the average fundamental frequency and average damping ratio. In this way, the dynamic maintenance module can reflect the changes in the health status of the bridge in real time, ensuring the accuracy and reliability of the abnormality identification module.

[0126] For the statistical benchmark update of dynamic fingerprint features, the same recursive logic is used to update the average fundamental frequency and average damping ratio.

[0127] In practice, according to the actual operation of the bridge and the stability of the monitoring data, The value can be set between 0.1 and 0.5. For example, for a bridge with a relatively stable operating state, The value can be set to 0.1, so that the model relies more on historical data; and for a bridge with a fast-changing operating state, The value can be set to 0.5, so that the model adapts to the current state more quickly. Calculate the average lateral distribution coefficient vector in the current effective period It can be obtained by averaging the lateral distribution coefficients under multiple load events. For example, the system can record the lateral distribution coefficients of multiple load events in each monitoring period, and then calculate the average of these coefficients as .

[0128] Update the reference lateral distribution coefficient vector according to the recursive formula , and use the updated vector for subsequent anomaly identification. For the update of the dynamic fingerprint feature, the average fundamental frequency and average damping ratio in the current effective period also need to be calculated, and the reference value is updated through the recursive formula. In this way, the dynamic maintenance module can reflect the changes in the health status of the bridge in real time, ensuring the accuracy and reliability of the anomaly identification module.

[0129] The above description is only some of the preferred embodiments of the present application and the explanation of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with technical features with similar functions disclosed in the embodiments of the present application (but not limited to) to form technical solutions.

Claims

1. A bridge monitoring data outlier intelligent judgment and processing system, characterized in that, The method comprises the following steps: a monitoring module is used to synchronously collect dynamic displacement data of multiple monitoring points in a designated area of a bridge; a preprocessing module is used to perform time synchronization, spatial registration and quality fusion on the dynamic displacement data, synchronously collected load data and environmental data, to generate multi-dimensional time series data; an anomaly identification module is used to extract cooperative deformation features representing the overall working performance of the structure in the transverse direction and dynamic fingerprint features representing the dynamic characteristics of the structure from the multi-dimensional time series data, perform anomaly identification and preliminary classification based on feature deviation, and output anomaly preliminary classification information; an anomaly diagnosis module is used to receive the anomaly preliminary classification information, perform deep diagnosis and cause correlation analysis in combination with load working conditions and environmental conditions, and output anomaly diagnosis results including anomaly influence levels; a strategy generation module is used to generate hierarchical early warning information and corresponding treatment strategies according to the anomaly diagnosis results and the use degree of key structure indicators calculated based on the dynamic displacement data; the calculation of the feature deviation comprises: The lateral distribution coefficient of the displacement response of each beam body monitoring point of the current section is calculated wherein , is the peak displacement of the first piece beam, denotes the sum of the peak displacements of all beam pieces of the current section; comparing the current calculated lateral distribution coefficient vector with a reference lateral distribution coefficient vector ; calculating the Euclidean distance or cosine similarity of two vectors to obtain the feature deviation; the monitoring module comprises: a radar scanning control unit is used to control the microwave radar to scan at least one transverse section or longitudinal section of the bridge, and take multiple girders or multiple preset monitoring points on the same girder as synchronous monitoring targets; a dynamic displacement solving unit is used to solve the displacement time history curve of each monitoring point in three-dimensional space in real time from the radar echo signal, wherein the displacement time history curve comprises a static displacement component and a dynamic vibration component; a data packaging unit is used to package the dynamic displacement data of all monitoring points obtained in the same scanning period and radar state data into a structured data frame.

2. The system of claim 1, wherein, The preprocessing module comprises: a space-time reference unification unit is used to unify the displacement data of each monitoring point in the structured data frame, vehicle passing time series data collected by a video unit, and temperature and humidity data collected by an environmental sensor to the same time reference and spatial coordinate system; a load response correlation unit is used to mark the corresponding load action period in the displacement time history curve according to the vehicle passing time series data, and extract the peak displacement response under the load and the corresponding time; a data quality fusion unit is used to calculate a comprehensive quality confidence coefficient for each frame of multi-dimensional time series data according to the signal-to-noise ratio of the radar signal, the data packet loss rate and the environmental interference intensity.

3. The system of claim 2, wherein, The anomaly identification module comprises: a feature extraction unit is used to calculate the correlation coefficient matrix or transverse distribution influence line of the displacement time history of multiple girder monitoring points synchronously monitored in the same section as the cooperative deformation features; a dynamic fingerprint feature extraction unit is used to perform frequency spectrum analysis or modal analysis on the displacement time history curve of each monitoring point to extract the fundamental frequency, damping ratio or specified order mode shape of the structure as the dynamic fingerprint features; The abnormality preliminary judgment unit is configured to compare the cooperative deformation features and the dynamic fingerprint features extracted in the current monitoring period with a health benchmark model or historical statistical benchmark, calculate a feature deviation degree, and determine that an abnormality occurs in a corresponding feature dimension when the feature deviation degree exceeds a preset threshold, and classify the abnormality preliminary category information as a cooperative deformation abnormality or a dynamic characteristic abnormality.

4. The system of claim 3, wherein, The abnormality diagnosis module comprises: The abnormal pattern comparison unit stores a typical abnormal pattern library, and the typical abnormal pattern library includes change patterns caused by structural damage, support failure, hinge joint damage, and single plate stress; The cause correlation reasoning unit is configured to match the abnormality preliminary category information and corresponding abnormal features output by the abnormality preliminary judgment unit with the typical abnormal pattern library, correlate load types, weights, and environmental data at the same time, and give a probabilistic judgment of possible causes of the abnormality. The influence evaluation unit is configured to evaluate an influence level of the abnormality on overall safety and applicability of the structure based on the probabilistic judgment of possible causes of the abnormality and the abnormal feature deviation degree, and generate an abnormality diagnosis result including the influence level of the abnormality.

5. The system of claim 4, wherein, The strategy generation module comprises: A key indicator dynamic calculation unit is configured to dynamically calculate a bridge key indicator based on the dynamic displacement data, the key indicator including a dynamic deflection usage degree and an impact coefficient usage degree ; wherein the dynamic deflection usage is ; Impact coefficient usage ; measured deflection peak, unloaded reference value, design allowable deflection, measured impact factor, design impact factor; The hierarchical early warning triggering unit is preset with hierarchical early warning thresholds associated with the influence level in the abnormality diagnosis result and the usage degree of the key indicators, and triggers early warning of a corresponding level when the conditions are met. The strategy mapping unit generates corresponding data handling suggestions and structure inspection or maintenance suggestion strategies based on the triggered early warning level and the abnormality diagnosis result.

6. The system of claim 5, wherein, The hierarchical early warning triggering unit comprehensively judges the usage degree output by the key indicator dynamic calculation unit and the abnormal influence level in the abnormality diagnosis result: When the abnormal influence level is slight and the usage degrees of all key indicators do not exceed a first usage threshold, a notice-level early warning is triggered; When the abnormal influence level is moderate or the usage degree of any key indicator exceeds the first usage threshold but does not exceed a second usage threshold, a warning-level early warning is triggered; When the abnormal influence level is severe or the usage degree of any key indicator exceeds the second usage threshold, an alarm-level early warning is triggered; The second usage threshold is greater than the first usage threshold.

7. The system of claim 6, wherein, Further comprising: The dynamic maintenance module is configured to, when the abnormality diagnosis module does not output an abnormality diagnosis result and the comprehensive quality confidence coefficient output by the data quality fusion unit is higher than a set threshold in a monitoring period, dynamically update the health benchmark model used by the abnormality preliminary judgment unit using the multi-dimensional time series data in the period. The dynamic maintenance module performs the following operations: a) calling the feature extraction unit to calculate an average transverse distribution coefficient vector under multiple load events in the period to update the statistical benchmark of the cooperative deformation features; b) calling the dynamic fingerprint feature extraction unit to extract an average fundamental frequency and a damping ratio of structure vibration in the period to update the statistical benchmark of the dynamic fingerprint features.

8. The system of claim 7, wherein, The dynamic maintenance module uses a time decay weighted recursive algorithm for dynamic update. For the statistical benchmark update of the cooperative deformation feature, the recursive formula is: ； wherein, denotes the updated reference lateral distribution coefficient vector, denotes the average lateral distribution coefficient vector calculated for the current validity period, denotes the reference lateral distribution coefficient vector before update, denotes the forgetting factor.

Images