Bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering
By installing sensors on bridges and conducting wind tunnel experiments, and using DBSCAN cluster analysis to analyze bridge vibration data, feature values were extracted to identify flutter and vortex-induced vibration, solving the problem of difficulty in distinguishing bridge vibration types and achieving efficient bridge vibration monitoring and processing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-06
- Publication Date
- 2026-03-31
AI Technical Summary
Existing technologies are insufficient to accurately distinguish and identify flutter and vortex-induced vibration in bridge inspections, resulting in inaccurate vibration detection results and an inability to effectively address different vibration patterns.
A DBSCAN-based clustering method was adopted. Displacement sensors were installed on the bridge model to conduct wind tunnel experiments, record data, extract feature values and perform DBSCAN clustering analysis. The displacement standard deviation σy and vibration single frequency degree R value were used as features to identify the wind-induced vibration type of the bridge.
It enables accurate identification and judgment of bridge vibration types, timely handling of abnormal situations, and is applicable to the identification of clusters of various shapes and the processing of noise points. It adapts to non-uniform data and improves the accuracy and efficiency of bridge vibration monitoring.
Smart Images
Figure CN121765411A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of bridge engineering technology, and in particular to a method for classifying, identifying and monitoring bridge vortex-induced vibration and flutter based on DBSCAN clustering. Background Technology
[0002] With economic and technological development, bridge spans are increasing, making wind-induced vibration a more prominent issue. Wind-induced vibration phenomena such as flutter and vortex-induced vibration seriously threaten bridge safety and durability, posing new requirements for bridge safety design and daily maintenance. Based on this, using machine learning methods to identify wind-induced vibration phenomena in bridges offers higher efficiency and accuracy compared to traditional methods, and is of great significance for reducing and preventing the hazards of wind-induced vibration in practical engineering applications.
[0003] Bridges are susceptible to flutter and vortex-induced vibration. The main differences between bridge flutter and vortex-induced vibration lie in their vibration mechanisms, severity, and countermeasures: Flutter is a divergent, self-excited vibration caused by the interaction between aerodynamics and structural elasticity. When aerodynamic loads and structural deformation increase synchronously, a positive feedback loop forms, leading to a sharp increase in amplitude. Flutter is a self-excited vibration, and if not controlled in time, it can lead to structural dynamic instability or even failure, as seen in the Tacoma Narrows Bridge accident. Vortex-induced vibration is forced vibration caused by the periodic shedding of vortices formed by airflow bypassing a bridge structure. When the frequency of these vortex shedding approaches the structure's natural frequency, a "lock-in" phenomenon occurs. Vortex-induced vibration is a type of finite-amplitude vibration that typically does not directly cause structural damage. However, prolonged and continuous vibration can lead to fatigue damage in structural components. The countermeasures for the two types of vibration are also different: Flutter needs to be suppressed by optimizing the aerodynamic design (such as adjusting the cross-sectional shape), adding dampers, or using computational fluid dynamics simulations. Vortex-induced vibration can be prevented by determining the locked wind speed range through segmental model wind tunnel tests and selecting a cross-sectional shape with strong anti-vortex-induced vibration performance.
[0004] However, it is difficult to directly determine the vibration type from the actual vibration data of a bridge during bridge inspection. Usually, it is just a mixture of vibration data, which is a combination of various vibration conditions. As a result, actual bridge vibration inspection can only detect the vibration amplitude and frequency, and cannot accurately determine the vibration type, which brings trouble to bridge vibration monitoring.
[0005] Based on this, the present invention designs a bridge vortex vibration and flutter classification and monitoring method based on DBSCAN clustering to solve the above problems. Summary of the Invention
[0006] The purpose of this invention is to provide a method for classifying and identifying bridge vortex-induced vibration and flutter based on DBSCAN clustering. This method can identify and judge the two types of bridge vibration with large amplitude without relying on prior judgment. Moreover, it can determine the type of bridge vibration with a single detection data, and can also determine the intensity of bridge vibration based on displacement time history diagram and spectrum diagram.
[0007] This invention is implemented as follows: A method for classifying, identifying, and monitoring bridge vortex-induced vibration and flutter based on DBSCAN clustering, comprising the following steps: Step S1: Build a bridge model and install displacement sensors at each box girder and pier of the bridge model. Then conduct a wind tunnel experiment on the bridge model, record the displacement data generated at each part of the bridge model, and establish a sample database of wind-induced vibration of the bridge. Step S2: Plot the wind-induced vibration sample data of the bridge into a displacement time history diagram and a spectrum diagram; Step S3: Process the data from each measuring point in the wind-induced vibration sample database, and process the bridge's vibration and displacement signal data into two feature values as a dataset; then, calculate the average standard deviation of the displacement in the dataset. σ y The single-frequency intensity R value of the vibration at the corresponding measuring point is used as two characteristic values for identifying the type of wind-induced vibration at that location; Step S4: Perform DBSCAN cluster analysis on the wind-induced vibration sample data to identify the most suitable hyperparameters by comparing the identification results of the displacement time history diagram and the spectrum diagram, and obtain the noise points through the hyperparameters. Step S5: By combining the characteristics of the wind speed and displacement time history diagrams of the noise points, the noise points are distinguished, the type of wind-induced vibration of the bridge can be determined, and the results of vortex-induced vibration and flutter can be output.
[0008] Furthermore, the method for plotting the spectrum of the wind-induced vibration sample data of the bridge is as follows: perform a fast Fourier transform on the displacement time history signal to convert the time history signal into a frequency domain signal, and then plot the spectrum with the power spectral density on the vertical axis and the frequency on the horizontal axis. The displacement time history diagram establishes a time and displacement coordinate system, and marks the relationship between time and displacement changes in the coordinate system.
[0009] Furthermore, the average standard deviation of the displacement σ y The calculation formula is: ; Where n is the number of samples, y(t) is the displacement time history, and t is the time step. This represents the time average of the displacement.
[0010] Furthermore, the formula for calculating the characteristic value R of the vibration single-frequency intensity at the corresponding measuring point is as follows: R = P2 / P1 Where P2 is the power spectral density value of the second dominant frequency, and P1 is the power spectral density of the first dominant frequency.
[0011] Furthermore, in step S4, there are only two hyperparameters. The control variable method is used to optimize and adjust the hyperparameters. After adjusting to obtain suitable hyperparameters, multiple noise points are obtained.
[0012] Furthermore, in step S3, the dataset is established by extracting key features of displacement standard deviation and frequency uniformity index from the original vibration data, namely; Plot a phase space diagram using the two characteristics: displacement standard deviation and frequency uniformity index. DBSCAN clustering was used to identify vortex and flutter signals from the phase space diagram. The datasets for vortex vibration and flutter were ultimately confirmed to be accurate.
[0013] Further, in step S5, If the displacement time history diagram of the noise point has obvious periodicity and the amplitude curve density is high, it can be determined that vortex-induced vibration has occurred at that point. If the peak value of the displacement time history graph of the noise level changes little and the amplitude curve density is high, it can be determined that flutter has occurred at that location.
[0014] The beneficial effects of this invention are: this method can monitor the vibration of bridges, and through this method, identify and calculate the conventional vibration data of bridge monitoring, and effectively distinguish the vibration type, thereby effectively identifying the type of bridge vibration, judging the vibration amplitude, and thus being able to deal with abnormal bridge conditions in a timely manner. 2. This method can discover clusters of arbitrary shapes and identify noise points through density-based spatial clustering algorithms. It does not require specifying the number of clusters, can handle abnormal data and noise points, and is applicable to data with uneven density or noisy conditions. It is of greater significance for distinguishing and identifying vortex oscillation and flutter. Attached Figure Description
[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0016] Figure 1 This is a flowchart of the monitoring method of the present invention; Figure 2 This is a schematic diagram of the displacement sensor arrangement in the bridge structure model of the present invention; Figure 3 This is a schematic diagram of the overall vortex-induced vibration and flutter identification results of the bridge according to the present invention; Figure 4The present invention provides time history diagrams and frequency spectrum diagrams of the displacement of monitoring points when the bridge experiences vortex-induced vibration. Figure 5 The time history diagram and spectrum diagram of the displacement of the monitoring point when the bridge experiences flutter are shown in this invention. Detailed Implementation
[0017] Please see Figures 1 to 5 As shown, this invention provides a method for classifying and identifying bridge vortex-induced vibration and flutter based on DBSCAN clustering. To better understand the above technical solution, the following will describe the above technical solution in detail with reference to the accompanying drawings and specific implementation methods.
[0018] In a specific embodiment of the technical solution of the present invention: This includes the following steps: Step 1: Build a bridge model and install displacement sensors at each box girder and pier of the bridge model. Then conduct wind tunnel experiments on the bridge model, record the displacement data generated at each part of the bridge model, and establish a sample database of wind-induced vibration of the bridge. Step 2: Plot the wind-induced vibration sample data of the bridge into displacement time history diagrams and spectrum diagrams; The method for plotting the spectrum of wind-induced vibration sample data of bridges is as follows: perform a fast Fourier transform on the displacement time history signal to convert the time history signal into a frequency domain signal, and then plot the spectrum with power spectral density on the vertical axis and frequency on the horizontal axis. A displacement-time history diagram is generated, and a time-displacement coordinate system is established. The relationship between the changes in time and displacement is marked in the coordinate system.
[0019] Step 3: Process the data from each measuring point in the wind-induced vibration sample database, and process the vibration and displacement signal data of the bridge into two feature values as the dataset.
[0020] Dataset establishment: Extracting key features of displacement standard deviation and frequency uniformity indices from the raw vibration data; Plot a phase space diagram using the two characteristics: displacement standard deviation and frequency uniformity index. DBSCAN clustering was used to identify vortex and flutter signals from the phase space diagram. The data from the three sources were verified, and the accuracy of the datasets for vortex vibration and flutter was confirmed. The scope of the datasets was then expanded.
[0021] and the average standard deviation of displacement in the data σ y This serves as the first characteristic value used to identify the type of wind-induced vibration at that location; Displacement mean standard deviation σ y The calculation formula is: ; Where n is the number of samples, y(t) is the displacement time history, and t is the time step. This represents the time average of the displacement.
[0022] The single-frequency intensity R value of the vibration at the corresponding measuring point is used as the second characteristic value for identifying the type of wind-induced vibration at that location. The formula for calculating the single-frequency intensity R characteristic value of the vibration at the measuring point is as follows: R = P2 / P1 Where P2 is the power spectral density value of the second dominant frequency, and P1 is the power spectral density of the first dominant frequency.
[0023] Step 4: Perform DBSCAN cluster analysis on the wind-induced vibration sample data. Compare the identification results of the displacement time history diagram and the spectrum diagram to obtain the most suitable hyperparameters. Obtain noise points through the hyperparameters. Since there are only two hyperparameters, the control variable method is used to optimize and adjust the hyperparameters. After adjusting to obtain suitable hyperparameters, multiple noise points are obtained.
[0024] Step 5: By combining the characteristics of the wind speed and displacement time history diagrams of the noise points, the noise points are distinguished, which can determine the type of wind-induced vibration of the bridge and output the results of vortex-induced vibration and flutter. If the displacement time history diagram of the noise point has obvious periodicity and the amplitude curve density is high, it can be determined that vortex-induced vibration has occurred at that point. If the peak value of the displacement time history graph of the noise level changes little and the amplitude curve density is high, it can be determined that flutter has occurred at that location.
[0025] It should be noted that: This method can use computers for intelligent processing to calculate the displacement average standard deviation. σ y The algorithm formula for the single-frequency degree R value of the vibration at the corresponding measuring point is entered into the program. Subsequently, the detection data of each sensor only needs to be input into the algorithm program, and the computer can perform data processing, data mining, and pattern recognition. It is particularly suitable for analyzing a large amount of wind tunnel test, field measurement, and computational fluid dynamics simulation data in structural wind engineering.
[0026] Instructions for the deployment of displacement sensors on bridges: Numbers 1-15 are all laser displacement sensors; their specific spatial arrangement on the bridge can be found in [reference needed]. Figure 2 The following is a description of the locations of the displacement sensors installed in various parts of the bridge: The sensors used to detect the vertical displacement of the main girder of the bridge include: Sensor 1, located upstream of the mid-span of the bridge; Sensor 2, located downstream of the mid-span of the bridge; Sensor 4, located upstream of the left quarter point of the mid-span; Sensor 5, located downstream of the left quarter point of the mid-span; Sensor 7, located upstream of the right quarter point of the mid-span; Sensor 8, located downstream of the right quarter point of the mid-span; Sensor 10, located downstream of the mid-span of the side span; and Sensor 12, located upstream of the mid-span of the side span.
[0027] The sensors used to detect the lateral displacement of the main beam of the bridge include: sensor No. 3, which is located downstream of the mid-span; sensor No. 9, which is located downstream of the right quarter point of the mid-span; and sensor No. 11, which is located downstream of the mid-span of the side span.
[0028] The sensors used to detect the lateral displacement of the top of the bridge tower include: Sensor No. 6, which is located downstream of the top of the right bridge tower.
[0029] The sensors used to detect the longitudinal displacement of the bridge tower tops include: sensor 13, located on the upstream side to the right of the top of the right bridge tower; and sensor 14, located on the downstream side to the right of the top of the right bridge tower.
[0030] The sensors used to detect the lateral displacement of the main cable of the bridge include: sensor No. 15, which is located downstream of the mid-span of the main cable; and sensor No. 16, which is located downstream of the right quarter point of the main cable.
[0031] Figure 2 The arrows labeled on each sensor indicate the direction of the displacement they detect. After the bridge construction is completed, only the designed locations on the bridge need to be equipped with sensors such as... Figure 2 The sensors shown can monitor the bridge's vortex vibration and flutter status in real time.
[0032] In a specific example of vibration caused by wind direction: Displacement signals at all 16 locations were obtained from experiments conducted at wind angles of attack of 0°, +3°, and -3° with wind speeds ranging from 0 to 95 m / s. Taking the results at a wind angle of attack of -3° as an example, based on the displacement sensor's sampling frequency of 128 Hz, a total of 864 displacement time-history diagrams and spectrum diagrams were plotted for 54 operating conditions, and the corresponding characteristic values were calculated. σ yAnd R. DBSCAN clustering analysis was performed on the dataset in two-dimensional phase space. After adjusting the appropriate hyperparameters, a total of 13 noise points were found. Based on the characteristics of their wind speed and displacement time histories, the type of wind-induced vibration of the bridge can be clearly identified. Taking the right quarter point of the main cable with a wind speed of 40 m / s downstream and the mid-span of the main girder with a wind speed of 95 m / s downstream as examples, the data set results of this embodiment are shown in […]. Figure 4 and Figure 5 .
[0033] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering, characterized in that, The method comprises the following steps: Step S1, a bridge model is built, displacement sensors are installed at each box girder and pier of the bridge model, a wind tunnel test is performed on the bridge model, displacement data generated at each part of the bridge model are recorded, and a wind-induced vibration sample database of the bridge is established; Step S2, the wind-induced vibration sample data of the bridge are drawn into a displacement time history diagram and a frequency spectrum diagram; Step S3, processing the data of each measuring point of the wind-induced vibration sample database, and processing the vibration and displacement signal data of the bridge into two characteristic values as a data set; and the average standard deviation value of displacement in the data set Step S4, the wind-induced vibration sample data are subjected to DBSCAN clustering analysis and recognition, the comparison of the displacement time history diagram and the frequency spectrum diagram is adjusted to obtain the most suitable hyperparameters, and the noise points are obtained through the hyperparameters; y and the single frequency degree R value of the corresponding measuring point vibration as the two characteristic values used to identify the wind-induced vibration type at this place Step S5, the noise points are distinguished according to the wind speed at which the noise points occur and the characteristics of the displacement time history diagram, the wind-induced vibration type of the bridge is judged, and vortex vibration and flutter results are output. The frequency spectrum diagram in the wind-induced vibration sample data of the bridge is drawn in the following manner: the displacement time history signal is subjected to fast Fourier transform to convert the time history signal into a frequency domain signal, and then the frequency spectrum diagram is drawn with the power spectral density as the vertical coordinate and the frequency as the horizontal coordinate.
2. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering according to claim 1, characterized in that: The displacement time history diagram establishes a time and displacement coordinate system, and marks the change relationship between time and displacement in the coordinate system. The calculation formula of the characteristic value R of the vibration single frequency degree of the corresponding measuring point is:
3. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering according to claim 1, characterized in that: the displacement average standard deviation value R = P2 / P1 y The calculation formula is: ; where n is the number of samples, y(t) is the displacement time course, t is the time step, is the time average of the displacement.
4. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering of claim 1, characterized in that: Wherein P2 is the power spectral density value of the second dominant frequency, and P1 is the power spectral density of the first dominant frequency. In step S4, there are only two hyperparameters, and the control variable method is used to optimize and adjust the hyperparameters, so that suitable hyperparameters are obtained, and a plurality of noise points are obtained. In step S3, the data set is established in the following manner: key features of displacement standard deviation and frequency uniqueness index are extracted from original vibration data, that is, 5. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering of claim 1, characterized in that: The displacement standard deviation and the frequency uniqueness index are drawn into a phase space diagram; 6. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering of claim 1, characterized in that: DBSCAN clustering is used to identify vortex vibration and flutter signals from the phase space diagram; Finally, the data set of vortex vibration and flutter is confirmed to be accurate. In step S5, If the displacement time history diagram of the noise point has obvious periodicity and high amplitude curve density, it can be judged that vortex vibration occurs at the position; 7. The bridge vortex vibration and flutter classification identification monitoring method based on DBSCAN clustering of claim 1, characterized in that: If the displacement time history diagram of the noise point has small peak value change and high amplitude curve density, it can be judged that flutter occurs at the position.