Bridge vortex vibration real-time identification and early warning system and method based on multi-imu fusion

Abstract

The application belongs to the technical field of bridge structure health monitoring and wind engineering safety early warning, and discloses a bridge vortex vibration real-time identification and early warning system and method based on multi-IMU fusion, which comprises the following steps: each IMU terminal synchronously collects acceleration and angular velocity original data at a dynamically adjusted sampling rate and pre-processes the data at an edge end; the pre-processed signal is subjected to sliding window fast Fourier transform, the main frequency component and the corresponding amplitude spectral density are continuously tracked, and a "vortex vibration suspected" flag is triggered; the edge node uploads the "vortex vibration suspected" flag, the main frequency, the amplitude, and the timestamp characteristic data to a cloud platform for convergence and spatio-temporal fusion analysis; a hierarchical early warning decision model is executed based on the fusion analysis result, and intelligent early warning decision and release are realized. The application converts the originally high-cost vortex professional monitoring into an economic and intelligent solution that can be deployed on a large scale, and realizes the leap from "post-event analysis" to "real-time early warning".

Description

Technical Field

[0001] This invention belongs to the field of bridge structural health monitoring and wind engineering safety early warning technology, specifically involving a bridge vortex-induced vibration real-time identification and early warning system and method based on multi-IMU fusion. Background Technology

[0002] Vortex-induced vibration (VEM) is a limited, periodic wind-induced vibration that occurs in long-span bridges under specific wind speeds. While it usually does not directly cause structural damage, sustained VEM can significantly exacerbate structural fatigue, affect driving comfort and safety, and may raise public concerns. Therefore, real-time and accurate monitoring and early warning of VEM are crucial.

[0003] Existing vortex-induced vibration monitoring methods have the following limitations: Traditional sensing methods have limited capabilities: commonly used accelerometers can only capture linear acceleration and are insensitive to the crucial torsional component in vortex-induced vibrations, making it difficult to comprehensively characterize vibration modes. Anemometers can only provide excitation conditions and cannot directly obtain the structural response.

[0004] The system is complex to deploy and costly: most existing high-precision monitoring systems use wired transmission, which requires large-scale cabling on bridges, making installation and maintenance difficult and the total cost high, making it difficult to popularize on existing bridges.

[0005] The identification algorithms are lagging and lack intelligence: most methods rely on sending data back to cloud servers for post-processing, resulting in poor real-time performance. The early warning logic is mostly based on simple amplitude threshold judgment, lacking intelligent analysis of the spatiotemporal evolution characteristics and multimodal coupling of vibration, leading to a high false alarm rate and the inability to provide early warnings in the early stages of vibration.

[0006] Lack of integrated, low-power dedicated solutions: Existing solutions typically treat vibration monitoring and environmental monitoring as independent subsystems, failing to integrate them into a unified low-power node. This results in system redundancy, high power consumption, and makes them unsuitable for long-term unattended online monitoring.

[0007] Therefore, there is an urgent need for a technical solution that can achieve low cost, easy deployment, high real-time performance, accurate identification of vortex-induced vibration characteristics, and intelligent early warning. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a real-time identification and early warning system and method for bridge vortex-induced vibration based on multi-IMU fusion. It utilizes the ability of IMUs to simultaneously measure linear acceleration and angular velocity to capture the vertical, lateral, and key torsional components of vortex-induced vibration from all angles. Through a sensor network composed of multiple IMU terminals deployed at key sections of the bridge, it senses the spatial distribution, propagation characteristics, and modal shape of the vortex-induced vibration. A hierarchical processing architecture of "edge perception - feature extraction - cloud fusion - intelligent early warning" is constructed to achieve real-time and automated processing of the entire process from initial vibration identification to amplitude exceeding the limit early warning. An integrated, wireless, and low-power hardware terminal and networking solution are provided, significantly reducing system deployment and maintenance costs.

[0009] To achieve the above objectives, the present invention provides the following solution: A bridge vortex-induced vibration real-time identification and early warning system based on multi-IMU fusion, the system comprising: IMU sensing terminal, sensor network, and auxiliary weather station; The IMU sensing terminal is used to capture the vertical, lateral and key torsional components of bridge vortex-induced vibration. The sensor network is used to sense the spatial distribution, propagation characteristics, and modal shape of bridge vortex-induced vibration. The auxiliary meteorological station is deployed at the bridge site to collect wind speed and direction data in real time, which serves as the basis for judging the excitation conditions of bridge vortex-induced vibration.

[0010] Preferably, the IMU sensing terminal integrates a three-axis MEMS accelerometer, a three-axis gyroscope, a microprocessor, a LoRa / Wi-Fi / 4G wireless communication module and a battery, and is encapsulated in a waterproof shell to form an independent intelligent sensing node; An integrated triaxial MEMS accelerometer is used to collect triaxial acceleration data at the location of measurement points on a bridge. A three-axis gyroscope is used to collect the three-axis angular velocities of the bridge measuring points. A microprocessor for processing triaxial acceleration and triaxial angular velocity; The LoRa / Wi-Fi / 4G wireless communication module is used to upload processed data to a gateway or cloud and receive external commands. The battery is used to power the IMU sensing terminal.

[0011] Preferably, the sensor network consists of multiple IMU terminals deployed at key sections of the bridge at 1 / 4 span, 1 / 2 span, and 3 / 4 span. A distributed network dynamic optimization method based on real-time modal shape matching enables sensor networks to possess "self-awareness" and "adaptive encryption" capabilities. Using the initially identified mode vectors Perform MAC matching with the theoretical mode shape library to determine the modal order corresponding to the current vibration. ; Based on the modal order corresponding to the current vibration Extracting theoretical vibration modes Continuous distribution; Based on theoretical vibration modes Given a continuous distribution, calculate the position of each node. absolute value of theoretical mode shape This absolute value reflects the relationship between the node and the first node. Sensitivity of first-order modes.

[0012] Preferably, the process of deploying equipment at the bridge site to collect real-time wind speed and direction data as a basis for determining the vortex-induced vibration excitation conditions of the bridge includes: The "predictive wake-up" mechanism is adopted, which uses external meteorological commands as the active wake-up source for low-power terminals and constructs a real-time linkage closed loop of "meteorological perception - edge decision-terminal response". The terminal no longer relies on its own vibration detection to wake up, but continuously listens for external meteorological commands. When the meteorological station at the bridge site detects that the wind speed has entered the theoretical vortex-induced vibration locking range of the bridge and the wind direction is valid, the edge gateway broadcasts a "pre-wake-up" command. After receiving the command, the terminal actively enters the standby state before the vortex-induced vibration actually occurs, realizing "the system is awake before the wind arrives".

[0013] Preferably, a "predictive wake-up" mechanism is adopted, using external meteorological commands as the active wake-up source for low-power terminals, and constructing a real-time linkage closed loop of "meteorological perception - edge decision-making - terminal response" including: Let the bridge be the first The locked wind speed range for the step vortex vibration mode is The effective wind direction deviation threshold is Define the vortex excitation probability As a quantitative basis for awakening decision-making: ; in, For the current moment The 10-minute average wind speed For the current moment The minimum angle between the wind direction and the normal to the bridge axis. Let wind speed be the membership function. This is the wind direction membership function.

[0014] This invention also provides a method for real-time identification and early warning of bridge vortex-induced vibration based on multi-IMU fusion. The method is implemented through the aforementioned system and includes: Step S1: Each IMU terminal synchronously collects the raw acceleration and angular velocity data of each measuring point on the bridge by dynamically adjusting the sampling rate and performs preprocessing at the edge end; Step S2: Perform a sliding window fast Fourier transform on the preprocessed signal to continuously track the main frequency component and its corresponding amplitude spectral density, triggering the "vortex oscillation suspected" flag; Step S3: The edge nodes upload the "suspected vortex-induced vibration" flag, main frequency, amplitude, and timestamp feature data to the cloud platform for aggregation and spatiotemporal fusion analysis; Step S4: Based on the results of the fusion analysis, execute the hierarchical early warning decision model to realize intelligent early warning decision-making and issuance.

[0015] Preferably, the judgment conditions that trigger the "suspected vortex-induced vibration" flag include: The main frequency amplitude of one or more IMU channels is higher than the background noise level; This main frequency appears synchronously in the same type of signal in multiple adjacent IMU nodes; The clock frequency value remains stable within a certain narrow band range.

[0016] Preferably, the method for edge nodes to upload "suspected vortex-induced vibration" flags, main frequency, amplitude, and timestamp characteristic data to the cloud platform for aggregation and spatiotemporal fusion analysis includes: Data alignment and synchronization: Align the data streams of all IMU nodes based on timestamps; Modal shape recognition: By comparing the amplitude ratio and phase difference of the vertical and torsional vibration signals from IMUs at different locations, the approximate modal shape of the current vibration is derived and matched with the theoretical modes of the bridge; Propagation characteristics analysis: Analyze the distribution of vibration amplitude along the bridge span direction to determine the initiation location and development of vortex-induced vibration.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: Comprehensive monitoring dimensions: By utilizing IMU to simultaneously sense multi-directional linear acceleration and angular velocity, it is the first time that effective monitoring of vortex-induced torsional components has been achieved in a low-cost solution, resulting in more accurate identification.

[0018] Strong spatial awareness: Through data fusion of multiple IMU nodes, it can identify the modal shape and spatial evolution process of vortex-induced vibration, which surpasses the limitations of single-point monitoring.

[0019] High real-time performance and intelligence: The architecture of "preliminary edge judgment + deep cloud decision-making" not only ensures the timeliness of early warning, but also improves the accuracy of identification and the scientific nature of decision-making through complex cloud algorithms, effectively reducing false alarms.

[0020] The system is economical and easy to use: the use of wireless, low-power integrated terminals significantly reduces installation, wiring and maintenance costs, making the technology widely applicable to long-term health monitoring of various large and medium-span bridges. Attached Figure Description

[0021] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the overall architecture of the system of the present invention; Figure 2 This is a schematic diagram showing the typical deployment location of IMU sensor terminals on the main beam of a bridge. Figure 3 This is a general flowchart of the method of the present invention; Figure 4 This is a schematic diagram of the vibration amplitude time history curve at a certain point on the bridge during vortex-induced vibration. Figure 5 This is a schematic diagram of the vibration spectrum curve at a certain point on the bridge during vortex-induced vibration. Figure 6 This is a schematic diagram of the data spectrum distribution of multiple IMU nodes during vortex-induced vibration. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] Example 1 like Figures 1-2 As shown, the present invention provides a bridge vortex-induced vibration real-time identification and early warning system based on multi-IMU fusion. The system includes: IMU sensing terminal, sensor network, and auxiliary weather station. IMU sensing terminal is used to capture the vertical, lateral and key torsional components of bridge vortex-induced vibration. Sensor networks are used to sense the spatial distribution, propagation characteristics, and modal shape of bridge vortex-induced vibrations. An auxiliary meteorological station is deployed at the bridge site to collect wind speed and direction data in real time, which serves as the basis for judging the excitation conditions of bridge vortex-induced vibration.

[0026] In this embodiment, the IMU sensing terminal integrates a three-axis MEMS accelerometer (for acquiring three-axis acceleration at the bridge measuring point location), a three-axis gyroscope (for acquiring three-axis angular velocity at the bridge measuring point location), a microprocessor, a LoRa / Wi-Fi / 4G wireless communication module, and a battery, all encapsulated in a waterproof housing, forming an independent intelligent sensing node.

[0027] The IMU sensing terminal is an intelligent sensing node deployed at the forefront of the bridge structure. Its core function can be summarized as follows: converting the physical vibrations (acceleration and angular velocity) of the bridge structure into digitized spatiotemporal characteristic information, and performing preliminary processing and intelligent triggering at the edge, providing a high-value data foundation for multi-node fusion analysis in the cloud. Unlike traditional "dumb sensors" (which only collect and upload raw data), this terminal has edge computing capabilities and an intelligent working mode, enabling it to identify suspected vortex-induced vibration events locally, thereby significantly reducing power consumption and communication load while ensuring monitoring effectiveness. Six-axis fusion (acceleration + angular velocity) simultaneously captures translational and torsional motion, comprehensively characterizing vortex-induced vibration modes. This invention is the first to introduce gyroscopes into bridge vortex-induced vibration monitoring, enabling direct measurement of torsional vibration. It also combines time-domain energy, frequency-domain stability, spatial consistency (multi-node collaboration), and wind speed conditions for multi-dimensional feature fusion triggering. This significantly improves the accuracy and reliability of vortex-induced vibration identification and reduces invalid data.

[0028] Sensing layer: A triaxial MEMS accelerometer is used to measure the triaxial linear acceleration at the location of the bridge measuring point. This reflects the translational vibration of the structure in three spatial directions. Among them, the vertical acceleration... It is the core data source for identifying vertical bending vortex-induced vibrations. A three-axis MEMS gyroscope is used to measure the three-axis angular velocity at the location of measuring points on a bridge. This reflects the torsional vibration of the structure about three axes. The angular velocity about the longitudinal axis is... It is the core data for identifying torsional vortex-induced vibrations.

[0029] Processing layer: The microprocessor (MCU) is the brain of the terminal, responsible for: ① controlling the sensor sampling timing; ② running embedded algorithms (filtering, FFT, feature extraction); ③ executing the intelligent state machine (switching between sleep / inspection / working / reporting modes); ④ managing power and communication.

[0030] Communication Layer: LoRa / Wi-Fi / 4G module, data transmission channel, responsible for uploading edge-processed feature data and alarm information to the gateway or cloud, and receiving external commands (such as wind speed trigger commands). LoRa is preferred for low-power scenarios, while 4G is suitable for occasions with high real-time requirements.

[0031] Power supply layer: High-capacity batteries (lithium thionyl chloride) are the core of the energy supply, powering all modules. Combined with intelligent power management, it achieves several years of maintenance-free operation.

[0032] Structural layer: Waterproof outer shell provides physical protection, protecting internal circuitry from moisture, salt spray, and dust corrosion, adapting to the harsh environment of bridges.

[0033] Sensor network: Multiple IMU terminals are deployed along the longitudinal direction of the main girder of the bridge at key sections such as 1 / 4 span, 1 / 2 span, and 3 / 4 span. Each section can be symmetrically arranged upstream and downstream to monitor torsion. All terminals communicate with an edge computing gateway or directly with the cloud.

[0034] A sensor network is a distributed intelligent sensing system composed of multiple IMU terminals deployed according to spatial topological patterns. Its core function is to transform discrete, single-point vibration measurements into a holographic understanding of vortex-induced vibration events—including their spatial distribution, energy propagation paths, vibration modal orders, and dynamic evolution processes—through spatiotemporal fusion and collaborative analysis. Compared to single-point monitoring, the value of sensor networks lies in achieving a leap from "point-like information" to "area-like understanding," enabling the system to answer four core questions in vortex-induced vibration monitoring: Where is the vibration occurring? (Spatial distribution and location of vibration initiation) How does it vibrate? (Modal shape and propagation characteristics) How long will this tremor last? (Sustainability and development trends) How serious is it? (Energy concentration and global amplitude) Auxiliary meteorological station: Deployed at the bridge site, it collects wind speed and direction data in real time, serving as the basis for judging vortex-induced vibration excitation conditions. By linking wind speed and direction data with structural response data in real time, the system can filter out "vibrations truly caused by wind-induced vortex-induced vibrations" from "potential vibrations," thereby significantly improving the accuracy of identification and the reliability of early warning. Compared with traditional methods that rely solely on structural response, the innovation of this invention lies in using the wind environment as the "first filter" for vortex-induced vibration identification. Before vibration energy is triggered, wind conditions are used to eliminate a large number of non-wind-induced interferences (such as traffic loads and construction vibrations), achieving a leap from "post-event response" to "pre-event screening." In this invention, wind condition data directly participates in the terminal's wake-up decision and cloud-based fusion identification, linking with the front-end trigger in real time. Meteorological data is upgraded from "background information" to "active judgment basis," realizing wind-vibration collaborative perception.

[0035] Auxiliary weather stations include: Anemometer: An ultrasonic anemometer or a mechanical three-cup anemometer is used, with a measurement range of 0~60 m / s, an accuracy of ±0.2m / s, and a response time of ≤0.25 s.

[0036] Wind vane: Integrated with or independently configured with anemometer, measuring range 0~360°, accuracy ±3°.

[0037] Data acquisition unit: Low-power microprocessor, continuously acquires raw wind speed and direction data at a sampling rate of 1 Hz to 10 Hz.

[0038] Communication module: Transmits data to the edge gateway or cloud in real time via RS485 / 4G / LoRa.

[0039] Installation location: Installed on the upstream side of the bridge (in the direction of the prevailing wind), at a height of 10 meters above the bridge deck (avoiding interference from the bridge structure), to ensure that the measured value represents the free-flowing wind conditions.

[0040] The original collected data is the instantaneous wind speed V. inst (t) and instantaneous wind direction θ inst (t). To eliminate the effects of turbulent fluctuations, statistical characteristic values ​​need to be calculated.

[0041] Specifically, this invention employs a "predictive wake-up" mechanism, using external meteorological commands as the active wake-up source for low-power terminals and constructing a real-time linkage closed loop of "meteorological perception - edge decision-making - terminal response." The terminal no longer relies on its own vibration detection for wake-up but continuously listens for external meteorological commands (via an ultra-low-power wireless receiving unit). When the meteorological station at the bridge site detects that the wind speed has entered the theoretical vortex-induced vibration locking range of the bridge (e.g., 8-12 m / s) and the wind direction is valid, the edge gateway broadcasts a "pre-wake-up" command. Upon receiving the command, the terminal proactively enters a standby state before the actual vortex-induced vibration occurs, achieving "system wake-up before the wind arrives."

[0042] Specifically, the technical implementation of the predictive wake-up mechanism: The core is to establish a quantifiable, progressively layered wake-up judgment model. Let the bridge be... The locked wind speed range for the step vortex vibration mode is The effective wind direction deviation threshold is (usually taken) Define the vortex excitation probability. As a quantitative basis for awakening decision-making: in: Current moment The 10-minute average wind speed.

[0043] Current moment The minimum angle between the wind direction and the bridge axis normal.

[0044] Wind speed membership function, using a trapezoidal distribution: in The transition zone width (e.g., 2 m / s) is used to smooth the threshold boundary and avoid frequent wake-ups caused by small fluctuations in wind speed.

[0045] Wind direction membership function, using a triangular distribution: This indicates the impact of the current wind conditions on the first... The excitation probability of the first mode. When Exceeding the preset wake-up threshold When the value is 0.7, the system determines that "vortex oscillation may be about to occur" and triggers a wake-up command.

[0046] The edge gateway continuously receives weather station data and calculates it at fixed intervals (e.g., every 1 minute). A pre-wake-up instruction frame is generated when the following conditions are met: Pre-wake-up conditions: The instruction frame structure is as follows: Frame header (1B) Command type (1B) Current wind speed (2B) Current wind direction (1B) Target mode order (1B) Estimated duration (2B) Checksum (1B) 0xAA 0x01 0x0BB8 0x02 0x01 0x012C 0xXX Instruction type: 0x01 indicates "pre-wake-up".

[0047] Target mode order: Indicates which mode the terminal should focus on (e.g., 0x01 indicates first-order vertical bend).

[0048] Expected duration: The length of the current risk window (in seconds) estimated based on weather forecasts or wind speed trends, for use by the terminal to plan and switch working modes.

[0049] The command is sent to all IMU terminals in the area via LoRa broadcast or 4G multicast.

[0050] The terminal has a built-in ultra-low power wireless receiver unit (power consumption <50 μA) that continuously monitors the broadcast channel.

[0051] Upon receiving the command, the terminal does not wake up immediately. Instead, it uses the wind speed and direction information contained in the command, combined with its own stored local correction parameters (such as the microclimate correction coefficient for the bridge location), to recalculate. Only when The system will only be officially activated when the threshold is exceeded. This "dual verification" mechanism can avoid false activation caused by the deviation between the data from a single weather station and the actual wind conditions at the bridge site.

[0052] Tiered wake-up strategy: like : Directly enter full-speed working mode (IMU 100 Hz sampling, real-time spectrum analysis).

[0053] like Enter standby inspection mode (IMU 20 Hz sampling, only calculates RMS and main frequency, does not upload raw data).

[0054] like Ignore this wake-up and return to sleep.

[0055] The "real-time linkage closed loop" consists of a meteorological perception sub-layer, an edge decision-making sub-layer, and a terminal response sub-layer. The three layers form a closed loop through bidirectional command flow and confirmation feedback mechanism.

[0056] Wake-up confirmation and state synchronization: After each terminal is woken up, it must return an acknowledgment frame to the gateway within 1 second, including: node ID, wake-up timestamp, current working mode (standby / full speed), and battery level (for energy management). After the gateway receives acknowledgments from all expected terminals, it enters the monitoring and coordination state. If no acknowledgment is received from a terminal within the timeout period, the command is resent or the node is marked as abnormal.

[0057] Dynamic threshold update: The closed loop is not a one-way "meteorology -> terminal" process; it also includes reverse calibration from structural response to meteorological thresholds. After a vortex-induced vibration event ends, the cloud platform compares the event record with the wake-up command at the time and calculates: like (i.e., waking up too early), then appropriately increase... If vortex-induced vibration has occurred but a wake-up call has not been triggered (missed report), then the warning level will be reduced. Alternatively, the membership function parameters can be adjusted. This online learning mechanism allows the system's prediction accuracy to continuously improve over time.

[0058] Dynamic extension / termination of the risk window: The gateway reassesses every 5 minutes based on the latest weather data. : like Extend the risk window and broadcast a "continue monitoring" instruction to the terminal.

[0059] like (Lag Threshold): If the risk is deemed to have been eliminated, broadcast a "hibernate" command.

[0060] Specifically, this invention proposes a distributed network dynamic optimization method based on real-time modal shape matching, enabling sensor networks to possess "self-awareness" and "adaptive encryption" capabilities.

[0061] 1. Modal matching and node importance assessment Using the initially identified mode vectors MAC matching is performed with the theoretical mode shape library to determine the most likely modal order corresponding to the current vibration. Extracting theoretical mode shapes The continuous distribution (obtainable through finite element model or function fitting). Calculate the position of each node. absolute value of theoretical mode shape This value reflects the node's relationship to the first... Sensitivity (i.e. importance) of the first mode.

[0062] Extraction of preliminary mode shape vectors: Arrangement along the bridge span direction One IMU node, node The vertical position is In the initial stage of a vortex-induced vibration event (e.g., the first 30 seconds), all nodes operate at the base sampling rate. (e.g., 20 Hz) Synchronously acquire vertical acceleration signals.

[0063] For each node Extract the current dominant frequency Normalized amplitude at the location and phase (This can be obtained through narrowband filtering followed by Hilbert transform). Using the node with the largest amplitude (usually the mid-span) as the phase reference, construct the measured mode shape vector: Where sgn is the sign function, with +1 indicating in-phase and -1 indicating out-of-phase.

[0064] MAC matching with the theoretical mode shape library: The theoretical mode shape library for bridges is pre-established through finite element analysis or environmental vibration testing, containing mode shape vectors for each order of modes. and at the node position Discrete values ​​are obtained by sampling at point .

[0065] Calculate the measured vibration mode and the first Modal Assurance Criterion (MAC) value for the first theoretical mode shape: The MAC value ranges from 0 to 1, with a higher correlation indicating a stronger correlation. The modal order with the highest MAC value exceeding a preset threshold (e.g., 0.8) is taken as the most probable mode of the current vibration. Obtaining the continuous distribution of theoretical mode shapes: For the matched first To obtain the first mode, we need to obtain its continuous mode shape function along the bridge span. This can be achieved in two ways: Finite element interpolation: Using the nodal mode values ​​calculated by the finite element model, a continuous curve is obtained by cubic spline interpolation.

[0066] Analytical function fitting: For simple bridge types (such as simply supported beams and continuous beams), the mode shape can be approximated by a sine function: (Vertical vibration mode of a simply supported beam) in This represents the bridge span length. For complex structures such as suspension bridges, a multi-segment function fitting method can be used.

[0067] Quantization of node sensitivity: node For the first Sensitivity of first mode Defined as the absolute value of the theoretical mode shape at that location: 2. Dynamic Sampling Rate Allocation Strategy The sampling rate is dynamically adjusted based on node importance. Highly important nodes ( Near the antinode, the sampling rate is increased to (100 Hz) is used to accurately capture amplitude and phase, providing core data for high-precision modal analysis.

[0068] Medium importance nodes ( Maintaining the base sampling rate (20 Hz) was used to verify mode continuity and propagation characteristics.

[0069] Low importance nodes ( Near the modal node, the sampling rate drops to (1 Hz) or enter deep sleep (only communication listening is retained), greatly reducing power consumption.

[0070] 3. Dynamic Reconstruction of Network Topology If the spacing between adjacent high-importance nodes is too large (exceeding 1 / 4 of the theoretical mode shape half-wavelength), the system can automatically insert "virtual nodes"—that is, require the sampling rate of medium-importance nodes in the region to temporarily increase in order to fill the gap in spatial resolution.

[0071] When the vortex resonant frequency changes (switches to another mode), the system re-executes the above process to dynamically adjust the network configuration.

[0072] Case scenario: The initial stage of first-order vertical bending vortex-induced vibration (frequency 0.185 Hz) of a suspension bridge.

[0073] Step 1: All 10 nodes are sampled at a base of 20 Hz. The initial identification frequency in the cloud is 0.185 Hz. The mode shape is approximately symmetrical. The MAC value matches the theoretical mode shape of the first-order vertical bending with a degree of 0.92.

[0074] Step 2: Extract the theoretical first-order vertical bending mode distribution from the system and calculate the theoretical mode values ​​at each node location: Mid-span node (x=150m): |Φ|=1.0 → High importance 1 / 4 span across nodes (x=75m): |Φ|=0.71 → Medium importance 3 / 4 spanning nodes (x=225m): |Φ|=0.71 → Medium importance Nodes near the abutment (x=0, 300m): |Φ|≈0.1 → Low importance Step 3: Send reconfiguration command from the cloud: Cross-node: The sampling rate is increased to 100 Hz, and continuous uploading of raw data is enabled.

[0075] 1 / 4 span and 3 / 4 span nodes: Maintain 20 Hz, but require uploading characteristic values ​​(RMS, frequency).

[0076] Bridge node: frequency downgraded to 1 Hz, only uploading energy values ​​for background monitoring.

[0077] Step 4: After 1 minute, the vortex-induced vibration develops to a steady state. The cloud uses high-precision data to accurately identify the vibration mode and finds that the amplitude on both sides of the mid-span is slightly asymmetrical, indicating that there may be slight structural nonlinearity, which is recorded in the health record.

[0078] Step 5: After the vortex vibration ends, all nodes return to the basic low-power inspection mode.

[0079] Example 2 like Figures 3-6As shown, this invention discloses a method for real-time identification and early warning of bridge vortex-induced vibration based on multi-IMU fusion. Specifically, it relates to an intelligent method for real-time sensing, identification, evaluation, and early warning of bridge vortex-induced vibration using a distributed inertial measurement unit (IMU) sensor array. The method includes: Step S1: Each IMU terminal synchronously acquires raw acceleration and angular velocity data by dynamically adjusting the sampling rate and performs preprocessing at the edge. Step S2: Perform a sliding window fast Fourier transform on the preprocessed signal to continuously track the main frequency component and its corresponding amplitude spectral density, triggering the "vortex oscillation suspected" flag; Step S3: The edge nodes upload the "suspected vortex-induced vibration" flag, main frequency, amplitude, and timestamp feature data to the cloud platform for aggregation and spatiotemporal fusion analysis; Step S4: Based on the results of the fusion analysis, execute the hierarchical early warning decision model to realize intelligent early warning decision-making and issuance.

[0080] The specific implementation process is as follows: Step S1: Distributed data synchronous acquisition and edge preprocessing Each IMU terminal synchronously collects raw data of acceleration and angular velocity at each measuring point on the bridge by dynamically adjusting the sampling rate (50Hz).

[0081] Filtering: A bandpass filter is used to retain the main vibration frequency band of the bridge while suppressing high-frequency noise and low-frequency drift. The passband frequency range of the bandpass filter is set according to the bridge structure type and monitoring target. For the main girder of a suspension bridge, the passband range is preferably 0.1 Hz to 1 Hz; for suspension cables and suspenders, it is preferably 1 Hz to 10 Hz; for continuous beam bridges, it is preferably 0.5 Hz to 5 Hz. The filter can be implemented using a fourth-order or higher Butterworth filter or an FIR filter to obtain a flat response within the passband and effectively suppress out-of-band noise.

[0082] Initial Feature Extraction: The real-time root mean square (RMS) value and short-term peak amplitude of six channels (three acceleration signals and three angular velocity signals) are calculated. Specifically, ACC_X represents the longitudinal (longitudinal) linear acceleration of the bridge, monitoring longitudinal sway; ACC_Y represents the transverse (lateral) linear acceleration of the bridge, monitoring transverse bending vibration, sensitive to certain asymmetric vortex-induced vibrations; ACC_Z represents the vertical (vertical) linear acceleration of the bridge, the core monitoring channel for vortex-induced vibrations, and the main response direction of the vertical bending mode; GYRO_X represents the torsional angular velocity about the longitudinal axis of the bridge, monitoring torsional vibration about the longitudinal axis, crucial for the torsional component of vortex-induced vibrations; GYRO_Y represents the transverse torsional angular velocity about the transverse axis of the bridge, monitoring pitch motion about the transverse axis; and GYRO_Z represents the torsional angular velocity about the vertical axis, monitoring torsion in the horizontal plane. Key Monitoring Channels: ACC_Z (vertical acceleration) and GYRO_X (torsion about the longitudinal axis) are the core channels for vortex-induced vibration monitoring. This is because vortex-induced vibrations mainly manifest as vertical bending and torsional vibrations, and these two channels directly correspond to the main response directions of the bridge. ACC_Y (lateral acceleration) and GYRO_Z (lateral torsion) can be used as aids to identify complex coupled vibration modes or eliminate specific disturbances.

[0083] For each activated channel signal The edge computing unit executes: in This represents the number of sampling points within the sliding window (e.g., corresponding to a 1-second data window). The real-time root mean square (RMS) reflects the vibration energy level of the channel within the current time window and is a fundamental indicator for judging vibration intensity.

[0084] in For a shorter sliding window (e.g., 0.5 seconds). Short-term amplitude peak. Sensitive to instantaneous impacts, it is used to capture the maximum transient response of vibrations and helps to identify the moment of vibration onset.

[0085] Fusion analysis: In this invention, the features of these channels are not used in isolation, but are fused together in the following way: The relationship between vertical and torsional forces: If the RMS of ACC_Z increases and the peak value of GYRO_X rises synchronously, it can be preliminarily determined to be a vertical bending-torsional coupled vortex-induced vibration.

[0086] Multi-channel spatial consistency: By comparing the timing and amplitude relationship of peak occurrences in the same channel (such as ACC_Z) of different IMU terminals, the propagation direction and modal shape of vibration can be inferred.

[0087] Energy ratio analysis: Calculating the energy ratio of GYRO_X to ACC_Z can quantify the proportion of the torsional component in the total vibration, providing a basis for early warning classification.

[0088] Step S2: Real-time extraction and preliminary identification of vortex vibration features (edge ​​side) Spectrum monitoring: A sliding-window Fast Fourier Transform is performed on the preprocessed signal (the signal after filtering and feature extraction at the edges) to continuously track the dominant frequency components and their corresponding amplitude spectral density. Specifically: I. Input Data and Preprocessing Input signal: Discrete time series of each channel after bandpass filtering Sampling rate (50 Hz).

[0089] Data segmentation: An infinitely long signal is truncated into fixed-length frames using a sliding window, with each frame containing... Each sampling point. Frame length. Balancing frequency resolution and real-time performance is necessary; typically, a suitable approach is chosen. Seconds, corresponding frequency resolution This is sufficient to distinguish the dominant frequency of bridge vortex-induced vibration (usually in between).

[0090] II. Sliding Window Design and Spectrum Calculation 1. Window Function Selection: To suppress spectral leakage, each frame of data needs to be windowed. The Hanning window or Hamming window is selected, as it has a moderate main lobe width and fast side lobe attenuation, making it suitable for vibration signal analysis. The windowed signal is as follows: in, Let be the original discrete-time signal sequence, representing the _th _ The values ​​of each sampling point (such as the sampled values ​​of physical quantities like acceleration and angular velocity), among which . For a sequence of window functions, and These are window functions of equal length, used to weight each sample point to suppress spectral leakage in the FFT. Common window functions include the Hanning window and the Hamming window, whose expressions are determined by their specific types. The number of sampling points per frame of data, i.e., the window length, corresponds to the time length. ( (Sampling rate). This is the windowed signal sequence, which is the pointwise product of the original signal and the window function, used for subsequent FFT analysis.

[0091] 2. Fast Fourier Transform (FFT): For windowed... Performing an FFT on a point sequence yields a complex spectrum. , Corresponding frequency .

[0092] 3. Amplitude Spectral Density (ASD) Calculation: To obtain a spectrum consistent with the physical meaning of the signal amplitude, calculate the one-sided amplitude spectral density (considering energy conservation): in The energy normalization coefficient of the window function ensures that the total energy remains constant after windowing. DC component. Usually ignored.

[0093] For vortex-induced vibration analysis, focusing on the power spectral density (PSD) may be more common, but the amplitude spectral density directly reflects the vibration amplitude. In practical applications, the choice can be made based on specific needs. As an energy indicator.

[0094] 4. Sliding and Overlapping: To improve temporal resolution, an overlap ratio (such as 50% or 75%) is set between adjacent frames. This means the starting point of the next frame moves relative to the current frame. Each sampling point. Overlap processing can smooth the spectrum changes over time and avoid missing short-term events.

[0095] III. Real-time tracking of the main frequency component In the spectrum of each frame In this process, it is necessary to identify the dominant frequency components that may represent vortex-induced vibrations. The specific algorithm is as follows: 1. Frequency band limitation: Preset the frequency band of interest according to the bridge type. (For example, 0.5 Hz ~ 5 Hz), searching for peaks only within this range can effectively reduce computational load and avoid high-frequency noise interference.

[0096] 2. Peak Detection: Within the frequency band of interest, find all local maxima (i.e., those that satisfy...) and point).

[0097] Set amplitude threshold (For example, 3 standard deviations of background noise) Remove spurious peaks caused by noise. Retain the peak with the largest amplitude. Peak (usually) ), as a candidate main frequency.

[0098] 3. Continuous Frequency Tracking: To eliminate random fluctuations, frequency trajectory smoothing is introduced. The maximum peak frequency detected in each frame is... Tracking can be performed using a first-order low-pass filter or a Kalman filter. in This is a smoothing factor (e.g., 0.7). This is the smoothed estimate of the main frequency.

[0099] If the peak frequency detected in multiple consecutive frames remains stable within a certain narrow band (fluctuation less than...), If the frequency is 0.1 Hz, then the frequency is considered to be the current dominant vibration frequency.

[0100] 4. Amplitude Recording: Simultaneously record the amplitude corresponding to this main frequency. This forms an amplitude-time series, which is used for subsequent amplitude growth trend analysis.

[0101] IV. Implementation Optimization of Edge Computing To meet the requirements of low power consumption and real-time performance, the following optimizations need to be considered when implementing this on the terminal MCU: Variable frame rate strategy: Normally, FFT is performed at a lower frequency (e.g., once every 10 seconds) for background monitoring; when vibration energy is detected to exceed the threshold (triggered by time domain RMS) or when an external wind speed warning is received, the system switches to a high frame rate mode (e.g., once per second) for fine tracking.

[0102] Fixed-point FFT library: Employs fixed-point FFT algorithms optimized for MCUs (such as the ARM CMSIS-DSP library) to avoid floating-point operation overhead.

[0103] Peak search acceleration: Quickly locate local maxima using DSP instructions or lookup table methods.

[0104] V. Output and Application The real-time output of the spectrum monitoring module includes: Current dominant frequency and its smoothing value.

[0105] Corresponding amplitude And its changing trend (such as the growth rate within 10 seconds).

[0106] Spectral entropy (optional): measures the degree of energy concentration. During vortex-induced vibration, the spectral entropy should be close to 0 (energy is concentrated at a single frequency).

[0107] These features will be input into the subsequent vortex-induced vibration identification logic, and fused with data such as wind speed and spatial distribution to make graded early warning decisions.

[0108] Preliminary judgment: The "suspected vortex-induced vibration" flag is triggered when the following conditions are met simultaneously: The main frequency amplitude of one or more IMU channels is significantly higher than the background noise level (e.g., more than 3 standard deviations above the background value).

[0109] This main frequency appears synchronously in the same signal of multiple adjacent IMU nodes.

[0110] The clock frequency value remains stable within a certain narrow band range.

[0111] Take a suspension bridge with a main span of 1200 meters as an example. According to wind tunnel tests and the bridge's design data, the frequency of its first-order symmetrical vertical bending mode is approximately 0.185 Hz, and the frequency of its first-order torsional mode is approximately 0.362 Hz.

[0112] To capture these two main vortex-induced vibration modes, multiple IMU nodes were deployed on the main girder of the bridge. It is assumed that one terminal was deployed at each of the three key sections: 1 / 4 span, mid-span, and 3 / 4 span.

[0113] When a typical vortex-induced vibration event occurs, the monitoring data will exhibit the following characteristics: First, observe the vertical acceleration channel (ACC_Z) at the mid-span node. At t=4 min, the amplitude increases sharply, and then remains high between t=6 min and t=10 min, with the frequency very stably locked around 0.185 Hz (fluctuation less than ±0.01 Hz). This dominant frequency value stabilizes within a narrow band. If, after t=4 min, the frequency jumps wildly between 0.1 Hz and 0.3 Hz, it indicates broadband random oscillation rather than vortex-induced vibration.

[0114] To verify that this is not a localized disturbance but rather a systemic vibration of the bridge, data from different nodes need to be compared. Below are the vibration amplitudes at 0.185 Hz for three different nodes at the same time (e.g., t=8min): 1 / 4 span node: amplitude approximately 12 mg; Mid-span node: amplitude approximately 22 mg; 3 / 4 span node: amplitude approximately 11 mg. This amplitude distribution shows a pattern of maximum amplitude at the mid-span and symmetrical decrease at both ends, perfectly matching the characteristics of a first-order symmetrical vertical bending vibration mode of a suspension bridge. This satisfies the first condition: the dominant frequency appears synchronously in similar signals from multiple adjacent IMU nodes. If only the ACC_Z signal at the mid-span node shows significant vibration, while the 1 / 4 and 3 / 4 span nodes show no response, or the response amplitudes are completely irregular, it may simply be a localized deformation or sensor malfunction at the mid-span, rather than vortex-induced vibration.

[0115] Step S3: Multi-node data cloud aggregation and spatiotemporal fusion analysis Edge nodes upload characteristic data such as "suspected vortex-induced vibration" flag, main frequency, amplitude, and timestamp to the cloud platform. The process is then performed in the cloud. Data Alignment and Synchronization: Based on high-precision timestamps, the data streams of all IMU nodes are aligned. To achieve spatiotemporal fusion of multi-node data, this invention adopts a layered synchronization strategy: the terminal layer stamps data based on a local high-precision real-time clock; the gateway layer periodically broadcasts time information through LoRaWAN Class B mode or a custom synchronization frame to calibrate the drift of the terminal's local clock, controlling the time error to the sub-millisecond level; the cloud layer further utilizes the correlation of the vortex-induced vibration event itself, and performs secondary fine-tuning of the key node data through a cross-correlation function, ultimately achieving microsecond-level data alignment, laying the foundation for high-precision modal recognition.

[0116] Modal shape recognition: By comparing the amplitude ratio and phase difference of vertical and torsional vibration signals from IMUs at different locations, the approximate modal shape of the current vibration (such as symmetrical vertical bending or anti-symmetrical torsion) is deduced and matched with the theoretical modes of the bridge. The specific process includes: Data preprocessing: After synchronization, the vertical acceleration signal and torsional angular velocity signal of each node are bandpass filtered to focus on possible vortex-induced vibration frequency bands.

[0117] Reference node selection: Select a reference node (such as the middle node) to calculate the phase difference.

[0118] Amplitude extraction: For each node, the amplitude (vertical and torsional) at the main frequency is extracted through spectrum analysis or narrowband filtering.

[0119] Phase difference calculation: The phase difference of each node relative to the reference node is calculated by cross-correlation or transfer function.

[0120] Normalized amplitude: Divide the amplitude of each node by the amplitude of the reference node to obtain the normalized mode amplitude.

[0121] Construct experimental mode vectors: obtain the vertical mode amplitude sequence (along the bridge span) and torsional mode amplitude sequence for each node.

[0122] Matching with theoretical mode shapes: The experimental mode shape is compared with the pre-stored theoretical mode shape (obtained through finite element calculation) to determine which mode the current vibration belongs to.

[0123] Modal Assurance Criterion: The degree of matching can be quantified using the MAC value (Modal Assurance Criterion).

[0124] Propagation Characteristic Analysis: This analysis examines the distribution of vibration amplitude along the bridge span to determine the initiation location and development of vortex-induced vibration. The aim of this analysis is to utilize the spatial resolution of a distributed IMU network to reveal the dynamic evolution of vortex-induced vibration energy along the bridge span—including the initiation location (where the vibration initially begins), propagation direction and velocity (whether the vibration spreads towards both ends or propagates unidirectionally), and development trend (amplitude growth pattern, standing wave / traveling wave characteristics). This analysis is crucial for understanding the spatial non-uniformity of vortex-induced vibration, identifying potential structural weaknesses, and developing precise early warning strategies. Specifically: I. Data Fundamentals and Preprocessing Input data: Vertical acceleration amplitude of each node after time synchronization and modal filtering. or angular velocity amplitude (The dominant channel, such as vertical acceleration, is usually chosen.) Each node corresponds to a fixed bridge span coordinate. (Taking one end as the origin, increasing along the bridge direction).

[0125] Time window: Select a time series during the duration of vortex-induced vibration. This constitutes a space-time amplitude matrix. : in For the number of nodes, This represents the number of sampling points within the time window (downsampling by time can reduce computational load).

[0126] II. Identification of the Oscillation Start Location The initiation point refers to the spatial point where the vibration energy initially begins to increase significantly. It can be determined through the following steps: 1. Energy growth initiation detection: For each node Calculate the rate of change of its amplitude over time (approximate derivative): in, The node index represents the node number laid out along the bridge. One IMU terminal, . For nodes At any moment The vibration amplitude usually refers to the amplitude of the vertical acceleration or torsional angular velocity after preprocessing. It can be the short-time root mean square (RMS), peak value, or amplitude of a certain frequency band (such as the dominant frequency of vortex vibration). For discrete time points, the first Each sampling moment corresponds to the center or starting point of the time window for amplitude calculation. The time step is the time interval between adjacent amplitude sampling points, which is usually equal to the sliding step of the amplitude calculation window (e.g., 1 second) or the data sampling interval. For nodes At any moment The approximate rate of change of amplitude reflects the instantaneous rate of increase in vibration energy; a positive value indicates an increase in amplitude, and a negative value indicates a decrease.

[0127] Find the first time each node exceeds the background noise threshold. Time point That is, satisfying: in These represent the mean and standard deviation of the amplitude during the quiet period.

[0128] 2. Spatial-temporal analysis: Compare the start-up time of each node. The earliest occurrence of the node position is taken as the oscillation start position. Plot a scatter plot of "start-up time - location" and fit a linear trend: in, The node index represents the node number laid out along the longitudinal direction of the bridge. One IMU terminal, . For nodes The start-up time, the moment when the vibration amplitude of the node first exceeds the background noise threshold, is obtained by energy growth initiation detection (forward difference method). For nodes The longitudinal position coordinates, along the bridge span direction, are usually taken as the origin at one end of the bridge (such as the abutment) and increase towards the other end. The slope of the fitted straight line represents the rate at which the oscillation start time changes with position. Its physical meaning is related to the direction of vibration propagation: if Vibration from Spreading from small beginnings to larger ones; if Conversely, if the opposite is true. The intercept of the fitted line represents the position. The theoretical start-up time at a certain point is mainly used for fitting calculations, and its absolute value is affected by the choice of the coordinate origin. The initial propagation velocity of energy along the bridge direction is the reciprocal of the slope. When the vibration exhibits obvious unidirectional propagation characteristics, the propagation velocity of energy along the bridge span direction can be roughly estimated.

[0129] 3. Energy Centroid Method: Define the centroid (energy center) of the amplitude distribution at each moment: If the center of gravity shifts in a certain direction over time, it indicates that energy is spreading outward from the initial position.

[0130] III. Characterization of Vibration Development 1. Spatial distribution evolution of amplitude envelope: plotting at different times The amplitude distribution curve along the bridge span (A continuous curve can be obtained by spline interpolation).

[0131] Observe the changes in the peak position, peak magnitude, and curve shape of the distribution curve over time. The development stage of vortex-induced vibration is usually characterized by a continuous increase in peak amplitude, and the distribution curve maintains a shape similar to the mode shape (standing wave). If the peak position drifts or the curve asymmetry intensifies, it may indicate a change in the structural state or the influence of a complex wind field.

[0132] 2. Energy Concentration Index: Calculate the standard deviation (or half-power width) of the amplitude distribution at each time step: like The gradual increase over time indicates that the vibrational energy is spreading from the starting point to a wider area; if it remains basically unchanged, the energy is concentrated in a fixed area.

[0133] 3. Standing wave / traveling wave discrimination: Calculate the phase difference between adjacent nodes The time series. For a pure standing wave, the phase difference should be 0° or 180° (depending on whether it is within the same half-wave) and should not change with time; for a traveling wave, the phase difference will increase linearly.

[0134] The instantaneous wavenumber can be obtained through the space-time phase gradient. This allows us to determine whether a traveling wave component exists. Vortex-induced vibrations are usually dominated by standing waves, but if the wind field is non-uniform or the structure has asymmetrical damping, a traveling wave component may appear.

[0135] Step S4: Intelligent Early Warning Decision-Making and Issuance The cloud platform integrates the following multi-dimensional information to execute tiered early warnings: Vibration intensity: based on the maximum amplitude / torsion angle after fusion of data from multiple IMUs.

[0136] Vibration duration: The length of time that vortex-induced vibration lasts.

[0137] Wind speed correlation: Whether the current wind speed is within the theoretical vortex-induced vibration locked wind speed range of the bridge.

[0138] Spatial consistency: Whether the vibration exhibits good spatial correlation (non-local random vibration).

[0139] The warning system is divided into three levels: Level 1 (Level of Concern): The vibration characteristics initially conform to vortex-induced vibration, but the amplitude is relatively small. Notify management to pay attention.

[0140] Level 2 (Alarm Level): The amplitude continues to increase, exceeding the preset daily operational safety threshold. An alert is issued to the maintenance department, recommending preparations for an inspection.

[0141] Level 3 (Action Level): The amplitude exceeds the preset traffic impact or fatigue concern threshold and continues without decay. It is recommended to initiate traffic control measures and conduct a detailed structural condition assessment.

[0142] In summary, this invention proposes a dedicated low-power sensing terminal integrating a multi-axis IMU, wireless communication, and battery management, and a method for deploying it on bridges to form a spatial sensing network. It creatively proposes a two-stage vortex-induced vibration identification process of "edge-level primary spectrum identification + cloud-based spatiotemporal modal fusion," balancing real-time performance and analytical depth. A hierarchical intelligent early warning decision-making model combining multi-dimensional information such as vibration amplitude, duration, spatial modes, and wind speed correlation is established, changing the traditional coarse mode of single-threshold alarms. This transforms the originally high-cost specialized monitoring of vortex-induced vibration into an economical and intelligent solution that can be deployed on a large scale, achieving a leap from "post-event analysis" to "real-time early warning."

[0143] Example 3 Taking a steel box girder cable-stayed bridge with a main span of 300 meters as an example, the specific implementation of the present invention will be explained.

[0144] Hardware Deployment: Nine IMU terminals are deployed at the 1 / 4, 1 / 2, and 3 / 4 span sections of the main girder, respectively, near the centerline of the top slab inside the box girder and near the vents on both sides. The terminals use a LoRa self-organizing network to aggregate data to the edge gateway at the bridge tower, and then upload it to the cloud via fiber optic cable.

[0145] Algorithm parameter settings: The edge sampling rate is set to 100Hz, and the bandpass filter range is 0.1-10Hz. The cloud-based early warning thresholds are set based on the finite element analysis and wind tunnel test results of the bridge. The first, second, and third level amplitude thresholds correspond to accelerations of 0.05g, 0.1g, and 0.2g, respectively (g is the acceleration due to gravity).

[0146] Workflow Example: One day, the anemometer measured an average wind speed that entered the vortex-induced vibration lock-in range of the bridge. The edge gateway detected a synchronous spike in the vertical acceleration spectrum of multiple mid-span IMU nodes at 1.2Hz, with amplitude exceeding the background value, triggering the upload of a "suspected" signal. Platform analysis confirmed that this frequency component exhibited a symmetrical vertical bending mode distribution across 9 nodes, and the amplitude continuously increased to 0.12g within 10 minutes. Considering the wind speed conditions, the platform immediately issued a Level 2 alarm to the maintenance unit and displayed a "Beware of Crosswinds" warning on the variable message signs on the bridge. Based on the warning, the maintenance unit activated video monitoring for confirmation and prepared for a focused inspection after the vibration ceased.

[0147] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A real-time identification and early warning system for bridge vortex-induced vibration based on multi-IMU fusion, characterized in that, The system includes: an IMU sensor terminal, a sensor network, and an auxiliary weather station; The IMU sensing terminal is used to capture the vertical, lateral and key torsional components of bridge vortex-induced vibration. The sensor network is used to sense the spatial distribution, propagation characteristics, and modal shape of bridge vortex-induced vibration. The auxiliary meteorological station is deployed at the bridge site to collect wind speed and direction data in real time, which serves as the basis for judging the excitation conditions of bridge vortex-induced vibration.

2. The system according to claim 1, characterized in that, The IMU sensing terminal integrates a three-axis MEMS accelerometer, a three-axis gyroscope, a microprocessor, a LoRa / Wi-Fi / 4G wireless communication module and a battery, all encapsulated in a waterproof housing, forming an independent intelligent sensing node; An integrated triaxial MEMS accelerometer is used to collect triaxial acceleration data at the location of measurement points on a bridge. A three-axis gyroscope is used to collect the three-axis angular velocities of the bridge measuring points. A microprocessor for processing triaxial acceleration and triaxial angular velocity; The LoRa / Wi-Fi / 4G wireless communication module is used to upload processed data to a gateway or cloud and receive external commands. The battery is used to power the IMU sensing terminal.

3. The system according to claim 1, characterized in that, The sensor network consists of multiple IMU terminals deployed at key sections of the bridge at 1 / 4 span, 1 / 2 span, and 3 / 4 span. A distributed network dynamic optimization method based on real-time modal shape matching enables sensor networks to possess "self-awareness" and "adaptive encryption" capabilities. Using the initially identified mode shape vectors Perform MAC matching with the theoretical mode shape library to determine the modal order corresponding to the current vibration. ; Based on the modal order corresponding to the current vibration Extracting theoretical vibration modes Continuous distribution; Based on theoretical vibration modes Given a continuous distribution, calculate the position of each node. absolute value of theoretical mode shape This absolute value reflects the relationship between the node and the first node. Sensitivity of first-order modes.

4. The system according to claim 3, characterized in that, The process of deploying equipment at the bridge site to collect real-time wind speed and direction data as a basis for determining the excitation conditions of bridge vortex-induced vibration includes: The "predictive wake-up" mechanism is adopted, which uses external meteorological commands as the active wake-up source for low-power terminals and constructs a real-time linkage closed loop of "meteorological perception - edge decision-terminal response". The terminal no longer relies on its own vibration detection to wake up, but continuously listens for external meteorological commands. When the meteorological station at the bridge site detects that the wind speed has entered the theoretical vortex-induced vibration locking range of the bridge and the wind direction is valid, the edge gateway broadcasts a "pre-wake-up" command. After receiving the command, the terminal actively enters the standby state before the vortex-induced vibration actually occurs, realizing "the system is awake before the wind arrives".

5. The system according to claim 4, characterized in that, The "predictive wake-up" mechanism uses external meteorological commands as the active wake-up source for low-power terminals, and constructs a real-time linkage closed loop of "meteorological perception - edge decision-making - terminal response," which includes: Set the bridge number The locked wind speed range for the step vortex vibration mode is The effective wind direction deviation threshold is Define the vortex excitation probability As a quantitative basis for awakening decision-making: ； in, For the current moment The 10-minute average wind speed For the current moment The minimum angle between the wind direction and the normal to the bridge axis. Let wind speed be the membership function. This is the wind direction membership function.

6. A method for real-time identification and early warning of bridge vortex-induced vibration based on multi-IMU fusion, wherein the method is implemented by the system described in any one of claims 1-5, characterized in that, The method includes: Step S1: Each IMU terminal synchronously collects the raw acceleration and angular velocity data of each measuring point on the bridge by dynamically adjusting the sampling rate and performs preprocessing at the edge end; Step S2: Perform a sliding window fast Fourier transform on the preprocessed signal to continuously track the main frequency component and its corresponding amplitude spectral density, triggering the "vortex oscillation suspected" flag; Step S3: The edge nodes upload the "suspected vortex-induced vibration" flag, main frequency, amplitude, and timestamp feature data to the cloud platform for aggregation and spatiotemporal fusion analysis; Step S4: Based on the results of the fusion analysis, execute the hierarchical early warning decision model to realize intelligent early warning decision-making and issuance.

7. The method according to claim 6, characterized in that, The conditions that trigger the "suspected vortex-induced vibration" flag include: The main frequency amplitude of one or more IMU channels is higher than the background noise level; This main frequency appears synchronously in the same type of signal in multiple adjacent IMU nodes; The clock frequency value remains stable within a certain narrow band range.

8. The method according to claim 6, characterized in that, The methods for edge nodes to upload "suspected vortex-induced vibration" flags, main frequency, amplitude, and timestamp characteristic data to the cloud platform for aggregation and spatiotemporal fusion analysis include: Data alignment and synchronization: Align the data streams of all IMU nodes based on timestamps; Modal shape recognition: By comparing the amplitude ratio and phase difference of the vertical and torsional vibration signals from IMUs at different locations, the approximate modal shape of the current vibration is derived and matched with the theoretical modes of the bridge; Propagation characteristics analysis: Analyze the distribution of vibration amplitude along the bridge span direction to determine the initiation location and development of vortex-induced vibration.

Images