Bridge monitoring alarm data processing method based on dynamic threshold

Abstract

The application discloses a kind of bridge monitoring alarm data processing method and system based on dynamic threshold, which comprises: real-time acquisition bridge structure monitoring data and environmental condition data, constructs standardization monitoring baseline database and sets initial alarm threshold;Dynamic alarm threshold interval is generated using environmental condition and service state double-factor model;Real-time monitoring data is compared with dynamic alarm threshold interval, and out-of-limit abnormal data is screened;Out-of-limit abnormal data is subjected to condition tracing, location tracing and cause tracing analysis, and invalid alarm is filtered through triple nested logic;Effective anomaly is subjected to four-level quantification grading and archiving, and output standardization abnormal alarm data set.The problems of high false alarm rate of fixed threshold in the prior art, single-factor dynamic threshold unable to quantify service aging effect, lack of abnormal tracing capability and lack of quantitative basis for grading standard are solved, adaptive accurate determination and closed-loop processing of bridge monitoring alarm data are realized, and alarm accuracy is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of bridge structural health monitoring data processing technology, and in particular to a bridge monitoring alarm data processing method and system based on dynamic thresholds. Background Technology

[0002] Bridges are a core component of transportation infrastructure, and their structural health directly affects traffic safety and the service life of the infrastructure. Currently, bridge structural health monitoring systems are widely used in various bridge maintenance operations. These systems collect structural response data in real time by deploying sensors such as strain gauges, displacement gauges, cable force sensors, and accelerometers, and use threshold judgment mechanisms to trigger alarms for structural anomalies.

[0003] Currently, most domestic bridge health monitoring platforms use fixed alarm thresholds for data processing. This means that fixed values ​​are pre-set based on bridge design drawings and national industry standards as the criteria for triggering alarms exceeding limits. The core flaw of this technology lies in the strong environmental sensitivity of bridge structural monitoring parameters such as strain, deflection, cable tension, and displacement. Changes in external conditions, such as diurnal temperature variations, seasonal temperature changes, ambient humidity, wind loads, and random vehicle loads, can cause continuous and non-linear shifts in the monitoring baseline. Fixed thresholds cannot adapt to the dynamic changes in monitoring data and cannot adjust the judgment criteria according to changes in operating conditions, resulting in a large number of invalid false alarms caused by environmental interference.

[0004] In recent years, some studies have begun to try dynamic threshold methods, but the following significant defects still exist: (1) The threshold correction factor is singular, only considering the single-factor correction of external factors such as ambient temperature, without incorporating internal state changes such as bridge service life, structural damage evolution rate, and historical disease cumulative effects into the threshold update system. (2) There is a lack of systematic anomaly identification and tracing mechanism, and a three-in-one tracing analysis system of traffic conditions, disease location, and causes has not been formed, making it impossible to accurately distinguish between false anomalies caused by instantaneous environmental fluctuations and permanent real anomalies caused by structural performance degradation. (3) The alarm classification standards are not uniform and lack quantitative basis. Existing graded warnings are mostly based on empirical qualitative classification (such as "minor", "general", "severe"), lacking dimensionless unified graded quantitative indicators, resulting in incomparable and non-universal standards between different monitoring indicators and different bridge types. (4) Bridge type adaptation depends on customized development. Existing threshold algorithms are mostly customized development for a single bridge type, with fixed parameters. No configurable parameter weight architecture has been established for the differences in monitoring items of each bridge type specified in the standard, which cannot meet the needs of large-scale bridge cluster monitoring promotion and application.

[0005] In summary, existing bridge monitoring and alarm technologies generally suffer from the following industry pain points: (1) high false alarm rate of fixed thresholds; (2) single-factor dynamic thresholds cannot quantify the service aging effect; (3) lack of anomaly tracing capability; (4) grading standards are based on empirical qualitative analysis and lack universal quantitative indicators; (5) bridge type adaptation depends on customized development.

[0006] Therefore, the industry urgently needs a general bridge monitoring alarm data processing solution that can adapt to dynamic thresholds, integrate intelligent rejection of invalid alarms, and perform hierarchical tracing of anomalies. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a bridge monitoring alarm data processing method and system based on dynamic thresholds, so as to realize adaptive and accurate judgment of bridge monitoring anomalies, intelligent identification of false anomalies, quantitative and hierarchical management of risk levels, and full-chain tracing of anomalies, thereby comprehensively improving the intelligence and refinement of bridge health monitoring.

[0008] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A bridge monitoring alarm data processing method based on dynamic thresholds, comprising the following steps:

[0009] Real-time acquisition of bridge structure monitoring data and environmental condition data, wherein the bridge structure monitoring data includes at least one of strain, displacement, settlement, cable force, vibration frequency, and deflection, and the environmental condition data includes at least one of temperature, humidity, wind speed, and vehicle load;

[0010] A standardized monitoring baseline database is constructed, and corresponding initial alarm thresholds are determined according to monitoring indicators. The standardized monitoring baseline database is established based on historical normal monitoring data of the target bridge with no structural damage records for a preset time period.

[0011] Based on the environmental operating condition data, the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database is calculated, and the initial alarm threshold is dynamically corrected in the first stage. The initial alarm threshold is set according to the coefficient of variation of the monitoring indicator. When the coefficient of variation is less than the preset coefficient of variation threshold, the upper and lower 2.5% historical quantiles are used. When the coefficient of variation is greater than or equal to the preset coefficient of variation threshold, the upper and lower 5% historical quantiles are used.

[0012] The thresholds after the first level of dynamic correction are adjusted based on the service status correction coefficient to generate dynamic alarm threshold ranges for each monitoring indicator.

[0013] The real-time collected bridge structure monitoring data is compared frame by frame with the dynamic alarm threshold range to filter out preliminary out-of-limit abnormal data that exceed the dynamic alarm threshold range.

[0014] A comprehensive source tracing analysis is performed on the initially screened out-of-limit abnormal data. The comprehensive source tracing analysis includes working condition tracing, location tracing, and cause tracing.

[0015] Invalid alarms are removed according to preset filtering rules, and the remaining valid structural anomaly alarms are classified, clustered and archived according to the severity of the anomaly, the location of the damage and the risk level.

[0016] Output a standardized and hierarchical dataset of bridge structural anomaly alarms to complete closed-loop processing of bridge monitoring alarm data.

[0017] Based on the above technical solution, the present invention can be further improved as follows:

[0018] Further: The specific calculation method for the service condition correction factor K is as follows:

[0019] Let Y be the service life of the bridge, D be the historical damage severity score, D = (100 - BCI) / 100, BCI is the structural condition score (out of 100) in the periodic inspection report of the bridge, and D ranges from 0 to 1. The larger the value, the more severe the damage. The structural damage evolution rate is R, which is obtained by fitting the rate of change of the most recent N monitoring data, N ≥ 10.

[0020] Service status correction factor , where α, β, and γ are weight coefficients, and α+β+γ=1, α, β, and γ are all ∈ [0.1, 0.7]; It is a monotonically increasing function of service life. or .

[0021] The beneficial effects of the above-mentioned further scheme are: by weighting and integrating the three dimensions of service life, disease score and damage evolution rate into a unified correction coefficient, the threshold judgment standard has for the first time achieved a two-factor adaptive matching between the external environmental conditions and the internal service aging state of the bridge, filling the technical gap that the existing single-factor correction scheme cannot quantify the service aging effect.

[0022] Furthermore: The comprehensive source tracing analysis of the initially screened out-of-limit abnormal data specifically includes:

[0023] Operating condition tracing: Match the real-time operating condition data at the time the out-of-limit abnormal data is generated, compare the real-time operating condition data with the historical operating condition data of the same season and time in the standardized monitoring baseline database, and determine whether the abnormality is a false abnormality caused by instantaneous environmental disturbance.

[0024] Location tracing: Based on the binding relationship between the preset sensor locations and the bridge structure zones, the bridge structure area corresponding to the out-of-limit anomaly data is located according to the sensor number and installation location of the source sensor.

[0025] Cause tracing: Extract monitoring time-series data within a preset time period before and after the occurrence of the out-of-limit anomaly data, calculate its time-domain statistical characteristics and frequency-domain energy distribution characteristics, and compare them with historical data under the same operating conditions in the standardized monitoring baseline database to determine whether it is a temporary anomaly caused by instantaneous operating condition fluctuations or a permanent real anomaly caused by structural damage or performance degradation.

[0026] The beneficial effects of the above-mentioned further solutions are: by constructing a three-in-one source analysis system integrating working conditions, location, and causes, and by jointly determining based on time-domain statistical characteristics and FFT frequency-domain energy characteristics, the system can accurately distinguish between transient disturbances and permanent damage, effectively improving the accuracy of anomaly identification.

[0027] Furthermore: The specific determination logic for tracing the cause is as follows:

[0028] Extract monitoring time-series data within a preset time period before and after the occurrence of the anomaly;

[0029] If the out-of-limit abnormal data returns to the baseline range synchronously after the operating conditions are restored, it is determined to be a temporary abnormality caused by instantaneous operating condition fluctuations;

[0030] If the out-of-limit abnormal data continues to deviate from the baseline range after the operating conditions are restored and the deviation shows a cumulative increasing trend, it is determined to be a permanent real anomaly caused by structural damage or performance degradation.

[0031] The beneficial effect of the above-mentioned further solutions is that, through clear time-domain judgment logic, the automatic distinction between temporary anomalies and permanent real anomalies is realized, providing a reliable decision-making basis for subsequent graded handling.

[0032] Further: The step of removing invalid alarms according to preset filtering rules specifically includes the following steps:

[0033] Temporal filtering noise reduction: A combination of sliding median filtering and adaptive wavelet thresholding is used to filter out high-frequency random noise;

[0034] Dynamic matching and verification of operating conditions: The environmental parameters at the time of the occurrence of the out-of-limit abnormal data are matched with the historical operating condition parameters of the same period in the standardized monitoring baseline database. If the similarity of the operating condition parameters is higher than the similarity threshold and the deviation of the monitoring data is lower than the deviation threshold, it is determined to be a false anomaly caused by the operating condition disturbance.

[0035] Baseline deviation threshold verification: Calculate the cumulative deviation of out-of-limit abnormal data relative to the standardized monitoring baseline database. If the cumulative deviation is within the regression baseline range of the baseline regression window, it is determined to be a transient interference and is removed; otherwise, it is transferred to the valid anomaly archiving process.

[0036] The beneficial effects of the above-mentioned further solution are: by using a triple nested filtering logic of time-series filtering, operating condition matching and baseline verification, false anomalies are filtered out layer by layer, which greatly reduces the invalid alarm rate and improves the quality of alarm data.

[0037] Further: The process of hierarchically clustering and archiving the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level specifically includes the following steps:

[0038] Let the measured value of the monitoring indicator be The dynamic threshold range is Calculate the dimensionless deviation ;

[0039] Based on the dimensionless deviation, the effective structural anomaly alarm is divided into four warning intervals: normal operating condition fluctuation, local anomaly, serious defects, and high-risk damage.

[0040] The beneficial effects of the above-mentioned further scheme are: based on the dimensionless deviation δ, a unified full monitoring index grading standard is achieved, which realizes a standardized grading that is quantifiable, reproducible, and comparable across indicators, and solves the deficiency of existing technical experience in qualitative classification lacking quantitative basis.

[0041] Furthermore, the upper and lower limits of the dynamic alarm threshold range are updated in real time based on environmental operating condition data and service status correction coefficients, respectively, with a preset time window as the unit. The length of the preset time window is adaptively set according to the monitoring indicator type, which includes temperature-sensitive indicators and long-term trend indicators.

[0042] The beneficial effects of the above-mentioned further solutions are: by adaptively setting a preset time window and updating iteratively in real time, the dynamic alarm threshold can accurately match the time-varying characteristics of different monitoring indicators, thereby improving the scientific nature and adaptability of threshold determination.

[0043] The present invention also provides a bridge monitoring alarm data processing system based on dynamic thresholds, including a data acquisition module, a baseline database construction module, a two-factor threshold update module, an out-of-limit anomaly screening module, a source tracing and identification module, a filtering and hierarchical archiving module, and a standardized output module;

[0044] The data acquisition module is used to collect bridge structure monitoring data and environmental condition data in real time. The bridge structure monitoring data includes at least one of strain, displacement, settlement, cable force, vibration frequency, and deflection. The environmental condition data includes at least one of temperature, humidity, wind speed, and vehicle load.

[0045] The baseline database construction module is used to construct a standardized monitoring baseline database and determine the corresponding initial alarm threshold according to the monitoring indicators. The standardized monitoring baseline database is established based on historical normal monitoring data of the target bridge with no structural damage records for a preset time period.

[0046] The dual-factor threshold update module is used to calculate the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database based on the environmental condition data, and to perform a first-level dynamic correction on the initial alarm threshold. The initial alarm threshold is set according to the coefficient of variation of the monitoring indicator. When the coefficient of variation is less than the preset coefficient of variation threshold, the upper and lower 2.5% historical quantiles are used. When the coefficient of variation is greater than or equal to the preset coefficient of variation threshold, the upper and lower 5% historical quantiles are used.

[0047] The dual-factor threshold update module is also used to perform a second-level adjustment on the threshold after the first-level dynamic correction based on the service status correction coefficient, and generate dynamic alarm threshold ranges for each monitoring indicator.

[0048] The out-of-limit anomaly screening module is used to perform frame-by-frame adaptive comparison between the real-time collected bridge structure monitoring data and the dynamic alarm threshold range, and to screen out preliminary out-of-limit anomaly data that exceeds the dynamic alarm threshold range.

[0049] The source tracing and identification module is used to perform a full-dimensional source tracing analysis on the initially screened out-of-limit abnormal data. The full-dimensional source tracing analysis includes working condition source tracing, location source tracing, and cause source tracing.

[0050] The filtering and hierarchical archiving module is used to remove invalid alarms according to preset filtering rules, and to hierarchically cluster and archive the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level.

[0051] The standardized output module is used to output a standardized and hierarchical bridge structure anomaly alarm dataset to complete the closed-loop processing of bridge monitoring alarm data.

[0052] The present invention also provides an electronic device, comprising: a processor and a memory;

[0053] The memory stores a computer-readable program that can be executed by the processor;

[0054] When the processor executes the computer-readable program, it implements the bridge monitoring alarm data processing method based on dynamic thresholds.

[0055] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the bridge monitoring alarm data processing method based on dynamic thresholds.

[0056] The beneficial effects of this invention are:

[0057] 1. Pioneering a dual-factor dynamic threshold iteration system based on environmental conditions and service status: This invention introduces a correction coefficient K, and for the first time quantifies the three dimensions of service life Y, damage score D, and damage evolution rate R into a unified correction coefficient, fundamentally solving the baseline drift problem caused by service aging, and realizing real-time iterative updates of alarm thresholds with external environment and internal service damage.

[0058] 2. Constructing a three-in-one anomaly tracing mechanism and a triple invalid alarm filtering logic: This invention systematically integrates three dimensions: operating condition tracing, location tracing, and cause tracing. Cause tracing is based on joint determination using time-domain statistical features and FFT frequency-domain energy features, combined with time-series filtering, operating condition matching, and baseline deviation triple nested verification, achieving accurate differentiation between instantaneous disturbances and permanent damage. Verified through actual engineering applications, the system using this invention achieves an automatic alarm accuracy rate greater than 85% and an overall early warning accuracy rate greater than 95%, effectively eliminating the vast majority of false alarms.

[0059] 3. Establish a standardized four-level alarm classification, clustering, and archiving system: Based on the dimensionless deviation δ unified full monitoring index classification standard, which is different from the existing technical experience-based qualitative classification, this invention realizes a standardized classification that is quantifiable, reproducible, and comparable across indicators, providing an accurate risk level basis for bridge maintenance decisions;

[0060] 4. Adopting a general algorithm architecture with configurable parameter weights: Based on the stress characteristics of various bridge structures and the differentiated requirements of JT / T1037-2022 "Technical Specification for Monitoring of Highway Bridge Structures", this invention configures differentiated weight ranges for beam bridges, cable-stayed / suspension bridges and arch bridges respectively, filling the gap in the existing technology of lacking a multi-bridge type parameter adaptive configuration scheme. When promoted on a large scale, there is no need to develop separate algorithms for each bridge, which has strong reusability and low implementation cost. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of a bridge monitoring alarm data processing method based on dynamic threshold according to an embodiment of the present invention;

[0062] Figure 2 This is a logic diagram of the dynamic threshold adaptive update model according to an embodiment of the present invention;

[0063] Figure 3This is a flowchart of abnormal data tracing and invalid alarm removal according to an embodiment of the present invention;

[0064] Figure 4 This is a schematic diagram of a bridge monitoring and alarm data processing system based on dynamic thresholds according to an embodiment of the present invention. Detailed Implementation

[0065] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0066] like Figure 1 As shown, this invention provides a bridge monitoring alarm data processing method based on dynamic thresholds, comprising the following steps:

[0067] S1: Real-time acquisition of bridge structure monitoring data and environmental condition data. The bridge structure monitoring data includes at least one of strain, displacement, settlement, cable force, vibration frequency, and deflection. The environmental condition data includes at least one of temperature, humidity, wind speed, and vehicle load.

[0068] Specifically, structural monitoring data includes key indicators such as strain, displacement, settlement, cable tension, vibration frequency, and deflection, which are collected in real time by various sensors (strain gauges, displacement gauges, settlement observation instruments, cable tension sensors, accelerometers, and deflectometers) deployed at key sections of the bridge according to a preset sampling frequency. Environmental condition data includes temperature, humidity, wind speed, and vehicle load, which are acquired synchronously through environmental monitoring stations, anemometers, dynamic weighing systems, and other equipment.

[0069] S2: Construct a standardized monitoring baseline database and determine the corresponding initial alarm thresholds based on the monitoring indicators. The standardized monitoring baseline database is established based on historical normal monitoring data of the target bridge during a preset time period of no structural damage records.

[0070] In practice, historical routine monitoring data of the target bridge with no structural damage records for at least three years are retrieved. A statistical distribution-based data cleaning method is used to remove abnormal data fragments exceeding limits during extreme environmental conditions (such as extreme high / low temperatures, typhoons, and excessive flooding) and during maintenance periods. Monitoring data under normal service conditions and standard operating conditions are retained to construct a standardized monitoring baseline database. This database is categorized and stored according to multiple dimensions, including monitoring indicator type, seasonal period, temperature range, and load level.

[0071] S3: Based on the environmental condition data, calculate the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database, and perform a first-level dynamic correction on the initial alarm threshold, such as... Figure 2 As shown;

[0072] Environmental operating condition data includes temperature, humidity, wind speed, vehicle load, etc. The initial alarm threshold is set based on the coefficient of variation of the monitored indicators. When the coefficient of variation is less than the preset coefficient of variation threshold, the upper and lower 2.5% historical quantiles are used. When the coefficient of variation is greater than or equal to the preset coefficient of variation threshold, the upper and lower 5% historical quantiles are used. For temperature-sensitive indicators (such as strain and deflection), a correlation model between the monitored indicators and ambient temperature (such as linear regression or multivariate nonlinear regression) is established to calculate the theoretical baseline value under the current temperature conditions. The difference between the theoretical baseline value and the standard operating condition baseline value is used as the offset to perform the first level of dynamic correction on the initial alarm threshold.

[0073] S4: Based on the service status correction coefficient, the threshold after the first level of dynamic correction is adjusted in the second level to generate the dynamic alarm threshold range of each monitoring indicator.

[0074] Let Y be the service life of the bridge, D be the historical damage severity score, D = (100 - BCI) / 100, BCI is the structural condition score (out of 100) in the periodic inspection report of the bridge, and D ranges from 0 to 1. The larger the value, the more severe the damage. The structural damage evolution rate is R, which is obtained by fitting the rate of change of the most recent N monitoring data, N ≥ 10.

[0075] Service status correction factor , where α, β, and γ are weight coefficients, and α+β+γ=1, α, β, and γ are all ∈ [0.1, 0.7]; It is a monotonically increasing function of service life. or .

[0076] The upper and lower limits of the dynamic threshold range of each monitoring indicator are multiplied by (1+K) and (1-K) respectively to expand or contract in both directions.

[0077] The two levels of correction mentioned above together constitute the two-factor adaptive threshold update model of the present invention. Unlike the existing single-factor environmental correction scheme, the present invention is the first to weight and integrate the three dimensions of service life, disease score and damage evolution rate into a unified correction coefficient, so that the threshold judgment standard can accurately adapt to the real-time environmental conditions and the aging state of bridge service.

[0078] S5: Perform a frame-by-frame adaptive comparison between the real-time collected bridge structure monitoring data and the dynamic alarm threshold range, and filter out the preliminary out-of-limit abnormal data that exceed the dynamic alarm threshold range;

[0079] The bridge monitoring data collected in real time by the sensors is adaptively compared frame by frame and index by index with the latest dynamic threshold range [L, U] of the corresponding monitoring indicators. If the monitoring data falls within the threshold range, it is judged as normal data and directly archived; if it exceeds the upper limit U or is lower than the lower limit L of the threshold range, it is marked as preliminary abnormal data exceeding the limit and enters the subsequent tracing and identification process.

[0080] S6: Perform a full-dimensional source tracing analysis on the initially screened out-of-limit abnormal data. The full-dimensional source tracing analysis includes working condition source tracing, location source tracing, and cause source tracing.

[0081] For the abnormal data exceeding limits obtained from the initial screening, a three-in-one source tracing analysis system integrating operating conditions, location, and causes is constructed, such as... Figure 3 As shown:

[0082] Operating condition tracing: Match real-time operating condition data such as environmental parameters (temperature, humidity), wind speed, and vehicle load at the time of abnormal over-limit data generation, and compare the current operating condition with historical operating conditions of the same season and time in the standardized monitoring baseline database.

[0083] Location tracing: Based on the binding relationship between the preset sensor locations and the bridge structural zones (main beam, arch ring, cable tower, suspenders, piers, foundations, etc.), the bridge structural area corresponding to the abnormal data is accurately located according to the sensor number and installation location of the source of the out-of-limit data.

[0084] Cause tracing: Monitoring time-series data within 10 minutes before and after the occurrence of the out-of-limit anomaly were extracted. Their time-domain statistical characteristics (mean, variance, peak value, skewness) and frequency-domain energy distribution characteristics (obtained through Fast Fourier Transform) were calculated and compared with historical data under the same operating conditions in the standardized monitoring baseline database. If the out-of-limit anomaly data synchronously returns to the baseline range after environmental operating conditions recover, it is determined to be a temporary anomaly caused by instantaneous operating condition fluctuations. If the out-of-limit anomaly data continuously deviates from the baseline range after operating conditions recover and the deviation shows a cumulative increasing trend, it is determined to be a permanent, real anomaly caused by structural damage or performance degradation.

[0085] S7: Eliminate invalid alarms according to preset filtering rules, and classify and archive the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level.

[0086] Employing a triple-nested filtering logic, it comprehensively filters out invalid alarms caused by non-structural interference such as temperature drift, instantaneous wind load, temporary vehicle overloading, and heavy traffic impact.

[0087] (1) Time-series filtering noise reduction: High-frequency random noise is filtered out by a combination of sliding median filtering and adaptive wavelet threshold denoising. Median filtering maintains signal edge features by suppressing impulse noise. Differential detection identifies jump point anomalies by capturing signal mutations. For data with long-period interference within a certain time domain, interference is eliminated by frequency domain analysis (FFT transform), frequency point suppression, and inverse transform reconstruction.

[0088] (2) Dynamic matching verification of operating conditions: The environmental parameters at the time of the occurrence of the out-of-limit abnormal data are matched with the historical operating condition parameters of the same period in the standardized monitoring baseline database. If the similarity of the operating condition parameters is higher than 85% and the deviation of the monitoring data is less than 5% of the baseline value, it is judged as a false anomaly caused by normal operating condition fluctuations. The similarity threshold of 85% is set according to the operating condition matching evaluation standard of JT / T1037-2022 "Technical Specification for Monitoring of Highway Bridge Structures".

[0089] (3) Baseline Deviation Threshold Verification: Calculate the cumulative deviation of out-of-limit abnormal data relative to the standardized monitoring baseline database. The baseline regression window is set to 2 hours based on the temperature stress relaxation period of concrete and steel bridge structures. If the cumulative deviation returns to the baseline range within 2 hours, it is determined to be a transient interference and is removed; otherwise, it is retained as a valid anomaly and transferred to the archiving process.

[0090] Through the above triple nested filtering, the false alarm rate can be effectively controlled within 4%, significantly improving the quality and reliability of alarm data.

[0091] For identified structural anomalies, they are classified into four alarm levels based on the severity of damage, risk level, and scope of structural impact.

[0092] Let the measured value of the monitoring indicator be The dynamic threshold range is Calculate the dimensionless deviation ;

[0093] Based on the dimensionless deviation, effective structural anomaly alarms are divided into four warning levels: normal operating condition fluctuations, local anomalies, severe defects, and high-risk damage. The four-level division is determined according to the progressive logic of the three-level alarm thresholds in JT / T 1037-2022 "Technical Specification for Monitoring Highway Bridge Structures" (normal range → 0.8 times design value → design value → exceeding design value) and the quantitative conversion of risk grading indicators in CJJ 99 "Technical Standard for Urban Bridge Maintenance".

[0094] Level 4 Warning (Normal Operating Condition Fluctuation): 0 < δ ≤ 5%, the structure is experiencing normal operating condition fluctuations, there is no safety risk, and continuous monitoring is recommended;

[0095] Level 3 warning (local anomaly): 5% < δ ≤ 15%, local slight deformation, no structural damage, it is recommended to increase the frequency of monitoring;

[0096] Level II Warning (Severe Defects): 15% < δ ≤ 30%, routine defects in components, affecting local service performance, special testing is recommended;

[0097] Level 1 Warning (High-Risk Damage): δ > 30%, overall structure exceeds limits, posing a safety hazard. It is recommended to immediately activate the emergency response procedure.

[0098] After the hierarchical calibration is completed, the system automatically completes the clustering, hierarchical, and standardized classification and archiving.

[0099] S8: Outputs a standardized and hierarchical dataset of bridge structural anomaly alarms, completing the closed-loop processing of bridge monitoring alarm data.

[0100] By integrating the effective alarm data after classification, and binding complete fields such as the time of anomaly occurrence, monitoring indicator type, abnormal value, dynamic threshold range, deviation, working condition traceability label, location traceability label, cause traceability label, alarm level, and suggested handling measures, a standardized, structured, and highly accurate bridge structure anomaly alarm dataset is output.

[0101] This invention employs a configurable parameter weight architecture. The weight coefficients (α, β, γ) are determined as follows: in the offline stage, the initial weights are determined using grid search or Bayesian optimization with the accuracy of identifying historical anomalies as the optimization objective; in the online stage, the weights are fine-tuned using gradient descent based on real-time feedback of false positive and false negative rates.

[0102] The weighting coefficients (α, β, γ) are configured differently for different bridge types, based on the structural stress characteristics of each bridge type and the differentiated requirements of JT / T 1037-2022 "Technical Specification for Monitoring of Highway Bridge Structures" regarding the "should be selected" and "preferably selected" monitoring items for each bridge type.

[0103] For beam bridges, the focus is on threshold correction for deflection and strain, with weights configured as α∈[0.50,0.65], β∈[0.20,0.30], and γ∈[0.10,0.20].

[0104] For cable-stayed or suspension bridges, the focus is on threshold correction of cable force and vibration frequency, with weights configured as α∈[0.30,0.40], β∈[0.30,0.40], and γ∈[0.25,0.35].

[0105] For arch bridges, the threshold correction for displacement and settlement is emphasized, with weights configured as α∈[0.45,0.55], β∈[0.25,0.35], and γ∈[0.15,0.25].

[0106] Application Examples

[0107] Taking a prestressed concrete continuous beam bridge in service as an example, the bridge has a main span of 120m, has been in service for 12 years, and is equipped with a complete structural health monitoring system.

[0108] (I) Data Acquisition and Baseline Database Construction

[0109] Twenty-four vibrating wire strain gauges, six static levels, and eight displacement gauges were installed at key sections of the main girder. Environmental monitoring stations, anemometers, and dynamic weighing systems were also deployed simultaneously. Historical monitoring data for the bridge over the past five years (2019-2024) were retrieved, excluding data from the maintenance and reinforcement period (45 days) and extreme typhoon weather (7 days). The data were then categorized and stored according to season and temperature range (5°C intervals) to construct a standardized monitoring baseline database.

[0110] (ii) Two-factor threshold iterative update

[0111] Taking the mid-span section strain as an example, the coefficient of variation of this monitoring index is 0.08 (less than 0.1), and the initial threshold is set using the upper and lower 2.5% quantile method as

[105] . , 128 A strain-temperature regression model was established (correlation coefficient 0.73), with a diurnal temperature range of 8. A baseline shift of 15°C occurs at 12°C. twenty two After the first-level correction, the interval is adjusted to

[104] . , 136 ].

[0112] Calculate the service status correction factor: Y=12 years, f(Y)=ln(1+12)=2.565; BCI=92, D=(100-92) / 100=0.08; N=20 monitoring data fitting results in R=0.03. The beam bridge weight configuration is α=0.60, β=0.25, γ=0.15, K=0.60×2.565×0.02+0.25×0.08+0.15×0.03=0.0553 (f(Y) is multiplied by 0.02 for normalization; the normalization coefficient is determined based on the bridge's design service life; when the design service life is 100 years, the normalization coefficient is 2 / 100=0.02). The threshold interval is expanded bidirectionally, ultimately reaching

[98] . , 144 ].

[0113] (III) Anomaly tracing and invalid alarm removal

[0114] In one incident, the strain surged to 152 με, exceeding the threshold. Origin of operating conditions: temperature 38.5℃, wind speed 12.7 m / s gusts, peak traffic flow; Origin of location: sensor S-CS-012 is located on the bottom plate at the mid-span of the main beam; Origin of cause: data from 10 minutes before and after the event were extracted, with a temporal mean of 148.2 με. Skewness 0.82 (positive skewness), FFT frequency domain display 0.2 The 0.5Hz low-frequency energy concentration (wind load frequency) is highly consistent with the characteristics of historical high-temperature gust wind conditions (similarity >90%). Triple verification: condition similarity 87.3% (above 85%), but deviation 10.8% (above baseline 5%), proceeding to the third verification; cumulative deviation 178. If the baseline is reached within 43 minutes (less than 2 hours), it is considered a transient interference and is automatically removed.

[0115] (iv) Comparison of actual anomaly handling and results

[0116] In another incident, the deflection deviation δ remained at 18%. If the deviation continues and accumulates after the working condition is restored (with a daily increase of 0.3mm for 3 consecutive days), it is judged as a level 2 warning, and a structured alarm record is pushed to the operation and maintenance platform.

[0117] This invention provides a bridge monitoring alarm data processing method based on dynamic thresholds. Through in-depth analysis of bridge health monitoring data, comprehensive weighted quantification of bridge service damage, linkage with environmental dual-factor dynamic thresholds, and verification of disease evolution based on time-domain and frequency-domain data, it can effectively achieve refined intelligent assessment of the entire link between monitoring data and structural diseases by filtering out false alarms in a hierarchical manner and providing quantitative and graded early warning.

[0118] like Figure 4 As shown, the present invention also provides a bridge monitoring alarm data processing system based on dynamic thresholds, including a data acquisition module, a baseline database construction module, a two-factor threshold update module, an out-of-limit anomaly screening module, a source tracing and identification module, a filtering and hierarchical archiving module, and a standardized output module;

[0119] The data acquisition module is used to collect bridge structure monitoring data and environmental condition data in real time. The bridge structure monitoring data includes at least one of strain, displacement, settlement, cable force, vibration frequency, and deflection. The environmental condition data includes at least one of temperature, humidity, wind speed, and vehicle load.

[0120] The baseline database construction module is used to construct a standardized monitoring baseline database and determine the corresponding initial alarm threshold according to the monitoring indicators. The standardized monitoring baseline database is established based on historical normal monitoring data of the target bridge with no structural damage records for a preset time period.

[0121] The dual-factor threshold update module is used to calculate the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database based on the environmental condition data, and to perform a first-level dynamic correction on the initial alarm threshold. The initial alarm threshold is set according to the coefficient of variation of the monitoring indicator. When the coefficient of variation is less than the preset coefficient of variation threshold, the upper and lower 2.5% historical quantiles are used. When the coefficient of variation is greater than or equal to the preset coefficient of variation threshold, the upper and lower 5% historical quantiles are used.

[0122] The dual-factor threshold update module is also used to perform a second-level adjustment on the threshold after the first-level dynamic correction based on the service status correction coefficient, and generate dynamic alarm threshold ranges for each monitoring indicator.

[0123] The out-of-limit anomaly screening module is used to perform frame-by-frame adaptive comparison between the real-time collected bridge structure monitoring data and the dynamic alarm threshold range, and to screen out preliminary out-of-limit anomaly data that exceeds the dynamic alarm threshold range.

[0124] The source tracing and identification module is used to perform a full-dimensional source tracing analysis on the initially screened out-of-limit abnormal data. The full-dimensional source tracing analysis includes working condition source tracing, location source tracing, and cause source tracing.

[0125] The filtering and hierarchical archiving module is used to remove invalid alarms according to preset filtering rules, and to hierarchically cluster and archive the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level.

[0126] The standardized output module is used to output a standardized and hierarchical bridge structure anomaly alarm dataset to complete the closed-loop processing of bridge monitoring alarm data.

[0127] The present invention also provides an electronic device, comprising: a processor and a memory;

[0128] The memory stores a computer-readable program that can be executed by the processor;

[0129] When the processor executes the computer-readable program, it implements the bridge monitoring alarm data processing method based on dynamic thresholds.

[0130] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the bridge monitoring alarm data processing method based on dynamic thresholds.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A bridge monitoring alarm data processing method based on dynamic thresholds, characterized in that, Includes the following steps: Real-time acquisition of bridge structure monitoring data and environmental condition data; Establish a standardized monitoring baseline database and determine the corresponding initial alarm thresholds based on the monitoring indicators; Based on the environmental operating condition data, the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database is calculated, and the initial alarm threshold is dynamically corrected in the first stage. The thresholds after the first level of dynamic correction are adjusted based on the service status correction coefficient to generate dynamic alarm threshold ranges for each monitoring indicator. The real-time collected bridge structure monitoring data is compared frame by frame with the dynamic alarm threshold range to filter out preliminary out-of-limit abnormal data that exceed the dynamic alarm threshold range. A comprehensive source tracing analysis is performed on the initially screened out-of-limit abnormal data. The comprehensive source tracing analysis includes working condition tracing, location tracing, and cause tracing. Invalid alarms are removed according to preset filtering rules, and the remaining valid structural anomaly alarms are classified, clustered and archived according to the severity of the anomaly, the location of the damage and the risk level. Output a standardized and hierarchical dataset of bridge structural anomaly alarms to complete closed-loop processing of bridge monitoring alarm data.

2. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 1, characterized in that, The specific calculation method for the service status correction factor K is as follows: Let Y be the service life of the bridge, and D be the historical damage severity score. D = (100-BCI) / 100, where BCI is the structural condition score (out of 100) in the periodic inspection report of the bridge. The value of D ranges from 0 to 1, and the larger the value, the more severe the damage. The structural damage evolution rate is R, which is obtained by fitting the rate of change of the most recent N monitoring data, where N≥10; Service status correction factor , where α, β, and γ are weight coefficients, and α+β+γ=1, α, β, and γ are all ∈ [0.1, 0.7]; It is a monotonically increasing function of service life. or .

3. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 1, characterized in that, The comprehensive source tracing analysis of the initially screened out-of-limit abnormal data specifically includes: Operating condition tracing: Match the real-time operating condition data at the time the out-of-limit abnormal data is generated, compare the real-time operating condition data with the historical operating condition data of the same season and time in the standardized monitoring baseline database, and determine whether the abnormality is a false abnormality caused by instantaneous environmental disturbance. Location tracing: Based on the binding relationship between the preset sensor locations and the bridge structure zones, the bridge structure area corresponding to the out-of-limit anomaly data is located according to the sensor number and installation location of the source sensor. Cause tracing: Extract monitoring time series data within a preset time period before and after the occurrence of the out-of-limit anomaly data, calculate its time domain statistical characteristics and frequency domain energy distribution characteristics, and compare them with historical data under the same operating conditions in the standardized monitoring baseline database to determine whether it is a temporary anomaly caused by instantaneous operating condition fluctuations, or a permanent real anomaly caused by structural damage or performance degradation.

4. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 3, characterized in that, The specific logic for determining the cause of the incident is as follows: Extract monitoring time-series data within a preset time period before and after the occurrence of the anomaly; If the out-of-limit abnormal data returns to the baseline range synchronously after the operating conditions are restored, it is determined to be a temporary abnormality caused by instantaneous operating condition fluctuations; If the out-of-limit abnormal data continues to deviate from the baseline range after the operating conditions are restored and the deviation shows a cumulative increasing trend, it is determined to be a permanent real anomaly caused by structural damage or performance degradation.

5. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 1, characterized in that, The process of removing invalid alarms according to preset filtering rules specifically includes the following steps: Temporal filtering noise reduction: A combination of sliding median filtering and adaptive wavelet thresholding is used to filter out high-frequency random noise; Dynamic matching and verification of operating conditions: The environmental parameters at the time of the occurrence of the out-of-limit abnormal data are matched with the historical operating condition parameters of the same period in the standardized monitoring baseline database. If the similarity of the operating condition parameters is higher than the similarity threshold and the deviation of the monitoring data is lower than the deviation threshold, it is determined to be a false anomaly caused by the operating condition disturbance. Baseline deviation threshold verification: Calculate the cumulative deviation of out-of-limit abnormal data relative to the standardized monitoring baseline database. If the cumulative deviation is within the regression baseline range of the baseline regression window, it is determined to be a transient interference and is removed; otherwise, it is transferred to the valid anomaly archiving process.

6. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 1, characterized in that, The process of hierarchically clustering and archiving the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level includes the following steps: Let the measured value of the monitoring indicator be The dynamic threshold range is Calculate the dimensionless deviation ; Based on the dimensionless deviation, the effective structural anomaly alarm is divided into four warning intervals: normal operating condition fluctuation, local anomaly, serious defects, and high-risk damage.

7. The bridge monitoring alarm data processing method based on dynamic threshold according to claim 1, characterized in that, The upper and lower limits of the dynamic alarm threshold range are updated in real time based on environmental operating condition data and service status correction coefficients, respectively, with a preset time window as the unit. The length of the preset time window is adaptively set according to the monitoring indicator type, which includes temperature-sensitive indicators and long-term trend indicators.

8. A bridge monitoring and alarm data processing system based on dynamic thresholds, characterized in that: It includes a data acquisition module, a baseline database construction module, a two-factor threshold update module, an out-of-limit anomaly screening module, a source tracing and identification module, a filtering and hierarchical archiving module, and a standardized output module; The data acquisition module is used to collect bridge structure monitoring data and environmental condition data in real time. The baseline database construction module is used to construct a standardized monitoring baseline database and determine the corresponding initial alarm thresholds based on the monitoring indicators. The dual-factor threshold update module is used to calculate the real-time baseline offset of each monitoring indicator relative to the standardized monitoring baseline database based on the environmental condition data, and to perform a first-level dynamic correction on the initial alarm threshold. The dual-factor threshold update module is also used to perform a second-level adjustment on the threshold after the first-level dynamic correction based on the service status correction coefficient, and generate dynamic alarm threshold ranges for each monitoring indicator. The out-of-limit anomaly screening module is used to perform frame-by-frame adaptive comparison between the real-time collected bridge structure monitoring data and the dynamic alarm threshold range, and to screen out preliminary out-of-limit anomaly data that exceeds the dynamic alarm threshold range. The source tracing and identification module is used to perform a full-dimensional source tracing analysis on the initially screened out-of-limit abnormal data. The full-dimensional source tracing analysis includes working condition source tracing, location source tracing, and cause source tracing. The filtering and hierarchical archiving module is used to remove invalid alarms according to preset filtering rules, and to hierarchically cluster and archive the remaining valid structural anomaly alarms according to the severity of the anomaly, the location of the damage, and the risk level. The standardized output module is used to output a standardized and hierarchical bridge structure anomaly alarm dataset to complete the closed-loop processing of bridge monitoring alarm data.

9. An electronic device, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the bridge monitoring alarm data processing method based on dynamic thresholds as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the bridge monitoring alarm data processing method based on dynamic thresholds as described in any one of claims 1-7.

Images