Bridge wind-induced vibration adaptive early warning method and system thereof

Abstract

The application discloses a bridge wind-induced vibration adaptive early warning method and system, the method acquires bridge acceleration time history data and synchronous wind speed data, and divides and analyzes samples according to a preset time interval; the acceleration time history is pretreated, and the response is decomposed into multiple intrinsic mode functions by using variational mode decomposition; the instantaneous frequency, instantaneous energy and instantaneous frequency entropy of each mode are calculated, the dominant mode of wind-induced vibration is screened according to the instantaneous frequency entropy and central frequency, and the wind-induced vibration time history is reconstructed; characteristics such as root mean square, peak factor, spectral centroid and the like are extracted from the wind-induced vibration time history, and an average wind speed is fused to construct a comprehensive early warning index; a peak threshold extreme value model is used to dynamically update the early warning threshold, and a Bayesian online change point detection is combined to identify risk mutations, and finally, graded early warning information is output. The application can inhibit the influence of vehicle vibration and other operation interference on wind-induced vibration risk identification, and improve the early warning capability of wind-induced risks such as vortex-induced vibration and buffeting.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology for structural health monitoring in civil engineering, and particularly relates to an adaptive early warning method and system for wind-induced vibration of bridges. Background Technology

[0002] Long-span bridges are typically subjected to both wind and vehicle loads during actual operation. The acceleration responses recorded by structural health monitoring systems over long periods often include multiple components such as wind-induced vibration, vehicle-induced vibration, and environmental noise. Wind-induced buffeting and vortex-induced vibration can lead to decreased bridge comfort, fatigue accumulation in components, and even safety risks. Therefore, it is necessary to identify wind-induced vibration risks and issue early warnings in a timely manner based on monitoring data.

[0003] Existing methods for early warning of wind-induced vibrations in bridges typically rely on fixed thresholds, human experience, or single amplitude indicators. Since the dynamic characteristics of bridge structures change slowly with operational status, and the wind environment at bridge sites varies seasonally, fixed thresholds can easily lead to false alarms or missed alarms. Furthermore, precursory characteristics of vortex-induced vibrations, such as frequency locking and energy accumulation, appear before a significant increase in amplitude, making it difficult to identify risks in a timely manner simply based on amplitude exceeding limits.

[0004] Existing wind-induced vibration early warning methods mostly employ fixed amplitude thresholds, fixed frequency band energy thresholds, or manual empirical rules. Since bridge structural stiffness, damping, traffic flow conditions, and the wind environment at the bridge site change over time, fixed thresholds are prone to false alarms or missed alarms. In particular, vortex-induced vibration typically undergoes a process of frequency locking, energy accumulation, and amplitude growth; traditional amplitude over-limit alarms often only trigger after a significant increase in response, making it difficult to capture early warning signs in a timely manner. Therefore, it is necessary to propose an adaptive, online detection method for bridge wind-induced vibration early warning. Summary of the Invention

[0005] Based on this, in order to solve the above-mentioned technical problems, the present invention aims to provide an adaptive early warning method and system for wind-induced vibration of bridges.

[0006] To achieve the above objectives, this invention proposes a technical solution for adaptive early warning of wind-induced vibration in bridges, comprising the following steps: S1: Acquire acceleration time history data collected by the bridge structural health monitoring system. a ( t The acceleration time history data is divided into multiple analysis samples according to a preset time interval; the average wind speed corresponding to the time period of the analysis sample is acquired simultaneously. U .

[0007] S2: Acceleration time history data a ( t Preprocessing is performed, and variational mode decomposition is used to decompose each analysis sample into... Keigenmode functions u k ( t ) and its center frequency f k .

[0008] S3: For each eigenmode function u k ( t Perform a Hilbert transform to calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k .

[0009] S4: Based on instantaneous frequency entropy H k and center frequency f k Screening of dominant modes of wind-induced vibration and reconstruction of wind-induced vibration time history. a w ( t ).

[0010] S5: Time history of wind-induced vibration a w ( t Extracting the root mean square of wind-induced vibrations RMS w Peak factor C p Spectral centroid SC w Features such as average wind speed U Construct a comprehensive early warning index I pre .

[0011] S6: The early warning threshold is dynamically updated using a peak-to-threshold extreme value model, and Bayesian online change point detection is used to identify abrupt changes in the distribution of the comprehensive early warning index.

[0012] S7: Generate and release graded early warning information based on the degree of exceeding the limit of the comprehensive early warning index and the posterior probability of the change point.

[0013] Optionally, the preprocessing in step S2 includes mean removal, trend removal, outlier removal, missing segment verification, and bandpass filtering in the range of 0.02Hz to 5Hz; the preset time interval is 5 minutes to 30 minutes.

[0014] Optionally, the number of modes in the variational mode decomposition in step S2 KThe number of candidate modes is determined by adaptive spectral envelope peak detection and optimized based on the mode aliasing index.

[0015] Preferably, the preferred method for determining the number of candidate modes includes: calculating the power spectral density of the acceleration time history data. P ( f ),right P ( f Smoothing is performed to obtain a smoothed spectrum; local peaks that satisfy the amplitude constraints are extracted and adjacent peaks are merged; the number of merged peaks is used as the initial number of modes. K 0, respectively for K 0-1 , K 0、 K 0+1 Perform variational mode decomposition and select the candidate modes with the smallest average correlation coefficients between modes as the final number of modes. K .

[0016] Optionally, the instantaneous frequency entropy mentioned in step S3 H k From instantaneous frequency IF k ( t probability density distribution of ) p k ( f ) Calculate and normalize.

[0017] Optionally, the wind-sensitive frequency band mentioned in step S4 [ f L , f H The entropy threshold is defined as [0.05Hz, 1.0Hz]. H th It is obtained through adaptive learning from historical benchmark samples with low wind speeds and no abnormal events.

[0018] Preferably, the entropy threshold H th The adaptive learning includes: calculating the minimum instantaneous frequency entropy of each benchmark sample within the wind-sensitive frequency band, forming a benchmark entropy sequence, and taking the preset low quantile of the benchmark entropy sequence as... H th .

[0019] Optionally, the wind-induced vibration characteristic parameters in step S5 include the root mean square of the wind-induced vibration. RMS w Peak factor C p Spectral centroid SC w and average wind speed U .

[0020] Ideally, the comprehensive early warning index in step S5 I pre Root mean square of wind-induced vibration RMS w Peak factor C p Spectral centroid SC w and average wind speed U Relative to the corresponding reference value RMS 0、 Cp 0、 SC 0、 U The weighted average of the normalized ratios of 0 is used to construct the structure; where the weight coefficients are... w i It is not less than 0, and the sum of the weighting coefficients is 1.

[0021] Furthermore, the reference value RMS 0、 Cp 0、 SC 0、 U 0 represents the median or mean of historical normal samples.

[0022] Optionally, the weighting coefficient w i The coefficients are determined based on feature stability in historical normal samples, expert preset values, or the entropy weight method. When using the entropy weight method, the difference coefficients are first determined based on the information entropy of each feature within the rolling window, and then the difference coefficients are normalized to obtain the final coefficients. w i .

[0023] Optionally, the peak-to-threshold extreme value model described in step S6 uses a generalized Pareto distribution to fit the peak exceeding the threshold. u The comprehensive early warning index exceeds the limit value, the threshold u The average excess function and parameter stability are automatically determined and updated on a rolling basis based on recent data without anomalies.

[0024] Optionally, the Bayesian online change point detection in step S6 uses the Student-t prediction distribution to recursively calculate the running length probability, and triggers a risk escalation judgment when the posterior probability of the change point exceeds a set threshold; the risk level is jointly determined by the comprehensive early warning index exceeding the limit level, the posterior probability of the change point, and the number of consecutively triggered samples.

[0025] This invention also provides an adaptive early warning system for bridge wind-induced vibration, which is embedded in a bridge structural health monitoring platform to achieve online early warning of bridge wind-induced vibration risks, including: The data acquisition module is used to acquire bridge acceleration time history data collected by the bridge structural health monitoring system. a ( tThe data is divided into multiple analysis samples according to a preset time interval; the average wind speed corresponding to the time period of each analysis sample is acquired simultaneously. U ; An adaptive decomposition module is used to process the acceleration time history data. a ( t Preprocessing is performed, and variational mode decomposition is used to decompose each analysis sample into... K eigenmode functions u k ( t ) and its center frequency f k ; The wind-induced modal recognition and reconstruction module is used to identify and reconstruct each intrinsic mode function. u k ( t Perform a Hilbert transform to calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k According to instantaneous frequency entropy H k and center frequency f k Screening of dominant modes of wind-induced vibration and reconstruction of wind-induced vibration time history. a w ( t ); The feature extraction and index construction module is used to extract features from the time history of wind-induced vibrations. a w ( t Extract wind-induced vibration characteristic parameters and combine them with average wind speed. U Construct a comprehensive early warning index I pre ; The dynamic threshold update and risk assessment module is used to model the tail distribution of the comprehensive early warning index sequence under historical normal conditions using a peak-over-threshold extreme value model, dynamically update the early warning threshold, and use Bayesian online change point detection for real-time monitoring. I pre Distribution changes; The early warning release module is used when... I pre When the warning threshold is exceeded or the posterior probability of the change point exceeds the set threshold, the risk level is determined based on the degree of exceeding the limit, the posterior probability of the change point, and the duration, and bridge wind-induced vibration classification warning information is generated.

[0026] Due to the adoption of the above technical solution, the beneficial effects obtained by the present invention include: 1. Improve early identification capability of wind-induced vibration: By identifying narrow-band, stable wind-induced vibration modes through instantaneous frequency entropy and frequency locking characteristics, the vortex-induced vibration incubation process can be captured before the vibration amplitude significantly exceeds the limit.

[0027] 2. Enhance operational interference suppression capability: By using mode decomposition and entropy feature screening to screen wind-induced dominant modes, the impact of operational interference such as vehicle vibration on wind-induced feature extraction is reduced, thereby improving the stability of early warning results under complex traffic conditions.

[0028] 3. Achieve adaptive threshold update: The peak-to-threshold extreme value model is used to model the tail distribution of the normal state, so that the warning threshold can be updated in a rolling manner as the structural state and environmental conditions change.

[0029] 4. Facilitates engineering deployment: The method is based on the real-time calculation of time-frequency characteristics and statistical models by the monitoring server, making it suitable for embedding into bridge health monitoring platforms for online early warning. Attached Figure Description

[0030] Figure 1 This is a flowchart of the method of the present invention.

[0031] Figure 2 This is a diagram showing the reconstruction results of acceleration time history and wind-induced vibration time history according to an embodiment of the present invention.

[0032] Figure 3 This is a graph showing the screening results of modal center frequency and instantaneous frequency entropy in an embodiment of the present invention.

[0033] Figure 4 This is a graph showing the comprehensive early warning index and dynamic threshold results of an embodiment of the present invention.

[0034] Figure 5 This is a diagram showing the Bayesian online change point detection results according to an embodiment of the present invention.

[0035] Figure 6 This is a diagram showing the risk level determination results of an embodiment of the present invention. Detailed Implementation

[0036] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are merely preferred implementations of the present invention and are not intended to limit the scope of protection of the present invention. Any simple modifications, equivalent changes, and alterations made to the following embodiments based on the technical essence of the present invention should still fall within the scope of protection of the technical solution of the present invention.

[0037] Example 1: This example provides an adaptive early warning method and system for wind-induced vibration of bridges. The specific process is as follows: Figure 1 As shown, it includes: S1. Data Preparation: Acceleration response data collected by the bridge structural health monitoring system was acquired, along with the average wind speed for the corresponding time period. The acceleration data was divided into several analysis samples with a basic time interval of 10 minutes. For each sample, the mean was removed, the trend term was removed, and outliers were eliminated. Samples with a missing value exceeding a preset proportion were directly marked as invalid. For valid samples, a bandpass filter from 0.02Hz to 5Hz was used to remove ultra-low frequency drift and high-frequency noise.

[0038] S2. Variational Mode Decomposition: The power spectral density of the sample was calculated using the Welch method. P ( f ),right P ( f After smoothing, local peaks are extracted. Each local peak must simultaneously meet a preset ratio where its amplitude is greater than the adjacent frequency point and not less than the maximum amplitude across the entire frequency band. If the frequency interval between two candidate peaks is less than 0.1 Hz, the peak with the larger amplitude is retained. The number of peaks after merging is used as the initial number of modes. K 0, and for K 0-1 , K 0、 K 0+1 Variational mode decomposition is performed separately. An aliasing index, Overlap, is constructed by calculating the inter-modal correlation coefficient. The candidate mode number with the smallest Overlap is selected as the final number of modes. K Punishment factor α Based on sampling frequency f s and the minimum interval Δ between adjacent center frequencies f min Make adaptive adjustments.

[0039]

[0040] S3. Instantaneous frequency entropy calculation: For each eigenmode function u k ( t Perform a Hilbert transform to obtain the analytic signal. z k ( t ), and calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k Instantaneous frequency entropy H k Depend on IF k( t probability density distribution of ) p k ( f The calculation shows that for the dominant mode of wind-induced vibration, the frequency is usually relatively stable and the energy is relatively concentrated, resulting in a lower instantaneous frequency entropy; for the dominant mode of vehicle vibration, the frequency is more random and the frequency band is wider, resulting in a higher instantaneous frequency entropy.

[0041]

[0042] S4. Wind-induced modal screening and reconstruction: Set wind-sensitive frequency bands [ f L , f H ],For example f L =0.05Hz f H =1.0Hz. The intrinsic mode functions that simultaneously meet the screening criteria are identified as the dominant modes of wind-induced vibration, and these modes are superimposed to reconstruct the wind-induced vibration time history. a w ( t Entropy threshold H th It can be adaptively determined from historical low wind speeds and no abnormal events, that is, by calculating the minimum instantaneous frequency entropy of each reference sample in the wind-induced sensitive frequency band, forming a reference entropy sequence, and taking the 5th percentile of this sequence as the reference entropy. H th . Figure 2 A comparison between the simulated acceleration time history and the reconstructed wind-induced vibration time history is shown. Figure 3 The results of the center frequency and instantaneous frequency entropy screening for each mode are shown.

[0043]

[0044] S5. Construction of Comprehensive Early Warning Index: Time history of wind-induced vibration a w ( t Extracting the root mean square of wind-induced vibrations RMS w Peak factor C p and spectral centroid SC w and incorporate average wind speed U Construct a comprehensive early warning index I pre .in, RMS 0、 C p0 , SC0、 U 0 represents the baseline value corresponding to the historical normal state or design control state. w 1. w 2. w 3. w 4 represents the weighting coefficient. Spectral centroid SC w It is obtained by weighted averaging of frequency and power spectrum and is used to characterize the location of energy accumulation.

[0045]

[0046] Furthermore, for each feature X j (benchmark value) RMS 0、 Cp 0、 SC 0、 U 0) Establish a baseline value for historical normal conditions, which can be the median of historical normal samples. P 50j Or the average. Comprehensive Early Warning Index I pre The system is constructed using the weighted ratios of each feature relative to the benchmark value, and the weight coefficients satisfy the non-negative normalization constraint.

[0047] S6. Dynamic threshold update and change point detection: Using a peak-to-threshold extreme value model to analyze historical normal conditions I pre Tail distribution is modeled. First, candidate thresholds are selected. u Calculations exceeding u Exceeding the limit Y = I pre - u And fit the generalized Pareto distribution. Y Tail portion distribution; threshold u The average excess function is approximately linear within a certain range, and its parameter stability is verified. Blue, yellow, orange, and red warning thresholds are calculated based on the target excess probability and updated on a rolling basis using recent data without anomalies. Furthermore, Bayesian online change point detection is used to recursively calculate the posterior probability of the change point. I pre When the dynamic threshold is exceeded or the posterior probability of the change point exceeds the set threshold, a warning of the corresponding level is triggered. Figure 4 The results of the integrated early warning index and dynamic threshold are shown. Figure 5 The results of running the variable-point posterior probability are shown. Figure 6 The results of different risk levels are shown.

[0048]

[0049] Risk levels can be determined according to the following rules: when I pre When the thresholds for blue, yellow, orange, or red are exceeded, the corresponding basic risk level is entered; when the posterior probability of the change point... P c Exceeding the variable point threshold P c_th And the number of consecutively triggered samples reached N c When, the risk level is raised by one level from the basic risk level; when I pre Not exceeding the threshold but P c When the amplitude continues to rise, an early warning alert is issued. This can handle the risk of gradually increasing amplitude and also identify sudden changes in statistical distribution caused by sudden gusts or vortex-induced vibration lock-in.

[0050] S7. Warning Issued: Early warning information is generated based on the risk level. This information includes the bridge number, monitoring point, trigger time, comprehensive warning index, dynamic threshold, dominant frequency, average wind speed, risk level, and recommended actions. Recommendations may include strengthening monitoring, verifying wind speed data, conducting on-site inspections, restricting traffic, or activating emergency plans.

[0051] Example 2: This example provides a bridge wind-induced vibration adaptive early warning system, including a data acquisition module, an adaptive decomposition module, a wind-induced modal identification and reconstruction module, a feature extraction and index construction module, a dynamic threshold update and risk assessment module, and an early warning release module. The system can be deployed on a bridge monitoring server, edge computing device, or cloud platform, enabling both offline playback analysis of historical monitoring data and real-time early warning of online monitoring data.

[0052] The data acquisition module is used to acquire bridge acceleration time history data collected by the bridge structural health monitoring system. a ( t The data is divided into multiple analysis samples according to a preset time interval; the average wind speed corresponding to the time period of each analysis sample is acquired simultaneously. U ; An adaptive decomposition module is used to process the acceleration time history data. a ( t Preprocessing is performed, and variational mode decomposition is used to decompose each analysis sample into... K eigenmode functions u k ( t ) and its center frequency f k ; The wind-induced modal recognition and reconstruction module is used to identify and reconstruct each intrinsic mode function. uk ( t Perform a Hilbert transform to calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k According to instantaneous frequency entropy H k and center frequency f k Screening of dominant modes of wind-induced vibration and reconstruction of wind-induced vibration time history. a w ( t ); The feature extraction and index construction module is used to extract features from the time history of wind-induced vibrations. a w ( t Extract wind-induced vibration characteristic parameters and combine them with average wind speed. U Construct a comprehensive early warning index I pre ; The dynamic threshold update and risk assessment module is used to model the tail distribution of the comprehensive early warning index sequence under historical normal conditions using a peak-over-threshold extreme value model, dynamically update the early warning threshold, and use Bayesian online change point detection for real-time monitoring. I pre Distribution changes; The early warning release module is used when... I pre When the warning threshold is exceeded or the posterior probability of the change point exceeds the set threshold, the risk level is determined based on the degree of exceeding the limit, the posterior probability of the change point, and the duration, and bridge wind-induced vibration classification warning information is generated.

[0053] The foregoing descriptions and embodiments are provided to enable those skilled in the art to understand and apply this invention. Those skilled in the art will readily make various modifications to these contents and apply the general principles described herein to other embodiments without inventive effort. This invention is not limited to the foregoing descriptions and embodiments. Any improvements and modifications made by those skilled in the art based on the disclosure of this invention, without departing from the scope of this invention, should be within the protection scope of this invention.

Claims

1. An adaptive early warning method for wind-induced vibration of bridges, characterized in that, Includes the following steps: S1: Acquire bridge acceleration time history data collected by the bridge structural health monitoring system. a ( t The data is divided into multiple analysis samples according to a preset time interval; the average wind speed corresponding to the time period of each analysis sample is acquired simultaneously. U ; S2: For the acceleration time history data a ( t Preprocessing is performed, and variational mode decomposition is used to decompose each analysis sample into... K eigenmode functions u k ( t ) and its center frequency f k ; S3: For each eigenmode function u k ( t Perform a Hilbert transform to calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k ; S4: Based on instantaneous frequency entropy H k and center frequency f k Screening of dominant modes of wind-induced vibration and reconstruction of wind-induced vibration time history. a w ( t ); S5: Time history of wind-induced vibration a w ( t Extract wind-induced vibration characteristic parameters and combine them with average wind speed. U Construct a comprehensive early warning index I pre ; S6: A peak-to-threshold extreme value model is used to model the tail distribution of the comprehensive early warning index sequence under historical normal conditions, dynamically updating the early warning threshold, and Bayesian online change point detection is used for real-time monitoring. I pre Distribution changes; S7: When I pre When the warning threshold is exceeded or the posterior probability of the change point exceeds the set threshold, the risk level is determined based on the degree of exceeding the limit, the posterior probability of the change point, and the duration, and bridge wind-induced vibration classification warning information is generated.

2. The method according to claim 1, characterized in that, The preprocessing in step S2 includes mean removal, trend removal, outlier removal, missing segment verification, and bandpass filtering in the range of 0.02Hz to 5Hz; the preset time interval is 5 minutes to 30 minutes.

3. The method according to claim 1, characterized in that, The number of modes in the variational mode decomposition described in step S2 K The number of candidate modes is determined by adaptive spectral envelope peak detection and optimized based on the mode aliasing index.

4. The method according to claim 3, characterized in that, The candidate mode number preferably includes: the power spectral density of the acceleration time history data. P ( f ),right P ( f Smoothing is performed to obtain a smoothed spectrum; local peaks that satisfy the amplitude constraints are extracted and adjacent peaks are merged; the number of merged peaks is used as the initial number of modes. K 0, respectively for K 0-1 , K 0、 K 0+1 Perform variational mode decomposition and select the candidate modes with the smallest average correlation coefficients between modes as the final number of modes. K .

5. The method according to claim 1, characterized in that, The instantaneous frequency entropy mentioned in step S3 H k From instantaneous frequency IF k ( t probability density distribution of ) p k ( f ) Calculate and normalize.

6. The method according to claim 1, characterized in that, The wind-sensitive frequency band mentioned in step S4 [ f L , f H The entropy threshold is defined as [0.05Hz, 1.0Hz]. H th It is obtained through adaptive learning from historical benchmark samples with low wind speeds and no abnormal events.

7. The method according to claim 6, characterized in that, The entropy threshold H th The adaptive learning includes: calculating the minimum instantaneous frequency entropy of each benchmark sample within the wind-sensitive frequency band, forming a benchmark entropy sequence, and taking the preset low quantile of the benchmark entropy sequence as... H th .

8. The method according to claim 1, characterized in that, The wind-induced vibration characteristic parameters mentioned in step S5 include the root mean square of the wind-induced vibration. RMS w Peak factor C p Spectral centroid SC w and average wind speed U Comprehensive Early Warning Index I pre Root mean square of wind-induced vibration RMS w Peak factor C p Spectral centroid SC w and average wind speed U Relative to the corresponding reference value RMS 0、 Cp 0、 SC 0、 U The weighted average of the normalized ratios of 0 is used to construct the structure; where the weight coefficients are... w i It is not less than 0, and the sum of the weighting coefficients is 1.

9. The method according to claim 8, characterized in that, The benchmark value RMS 0、 Cp 0、 SC 0、 U 0 represents either the median or the mean of historical normal samples.

10. The method according to claim 8, characterized in that, The weighting coefficient w i The coefficients are determined based on feature stability in historical normal samples, expert preset values, or the entropy weight method. When using the entropy weight method, the difference coefficients are first determined based on the information entropy of each feature within the rolling window, and then the difference coefficients are normalized to obtain the final coefficients. w i .

11. The method according to claim 1, characterized in that, The peak-to-threshold extreme value model described in step S6 uses a generalized Pareto distribution to fit the peak exceeding the threshold. u The comprehensive early warning index exceeds the limit value, the threshold u The average excess function and parameter stability are automatically determined and updated on a rolling basis based on recent data without anomalies.

12. The method according to claim 1, characterized in that, The Bayesian online change point detection described in step S6 uses the Student-t prediction distribution to recursively calculate the running length probability. When the posterior probability of the change point exceeds a set threshold, a risk escalation judgment is triggered. The risk level is determined by the comprehensive early warning index exceeding the limit level, the posterior probability of the change point, and the number of consecutively triggered samples.

13. A bridge wind-induced vibration adaptive early warning system, wherein the system is embedded in a bridge structural health monitoring platform to realize online early warning of bridge wind-induced vibration risk, characterized in that, include: The data acquisition module is used to acquire bridge acceleration time history data collected by the bridge structural health monitoring system. a ( t The data is then divided into multiple analysis samples according to a preset time interval. Simultaneously acquire the average wind speed corresponding to the time period of the analysis sample. U ; An adaptive decomposition module is used to process the acceleration time history data. a ( t Preprocessing is performed, and variational mode decomposition is used to decompose each analysis sample into... K eigenmode functions u k ( t ) and its center frequency f k ; The wind-induced modal recognition and reconstruction module is used to identify and reconstruct each intrinsic mode function. u k ( t Perform a Hilbert transform to calculate the instantaneous frequency. IF k ( t Instantaneous energy IE k ( t and instantaneous frequency entropy H k According to instantaneous frequency entropy H k and center frequency f k Screening of dominant modes of wind-induced vibration and reconstruction of wind-induced vibration time history. a w ( t ); The feature extraction and index construction module is used to extract features from the time history of wind-induced vibrations. a w ( t Extract wind-induced vibration characteristic parameters and combine them with average wind speed. U Construct a comprehensive early warning index I pre ; The dynamic threshold update and risk assessment module is used to model the tail distribution of the comprehensive early warning index sequence under historical normal conditions using a peak-over-threshold extreme value model, dynamically update the early warning threshold, and use Bayesian online change point detection for real-time monitoring. I pre Distribution changes; The early warning release module is used when... I pre When the warning threshold is exceeded or the posterior probability of the change point exceeds the set threshold, the risk level is determined based on the degree of exceeding the limit, the posterior probability of the change point, and the duration, and bridge wind-induced vibration classification warning information is generated.

Images