Vortex vibration early warning system and method for cable-stayed bridge cable force imbalance based on intelligent calculation model

Abstract

The present invention relates to the technical field of bridge safety monitoring, and specifically discloses a cable-stayed bridge cable-tension imbalance vortex-vibration early warning system and method based on an intelligent computing model. The system includes a multi-source data acquisition module, a phase space reconstruction module, a vortex-vibration modal decoupling module, a dynamic threshold setting module, a nonlinear prediction module, and an early warning decision module. The phase space reconstruction theory maps the vibration signal to a multi-dimensional phase space, and combines it with density clustering analysis to achieve multi-order frequency band vortex-vibration modal decoupling; a support vector machine is used to construct a multi-dimensional threshold hypersurface, and the discrimination threshold is dynamically adjusted according to environmental parameters and structural status; a local linearization prediction model is used to achieve short-term high-precision prediction and medium- and long-term trend prediction; an adversarial generative network is used to establish a wind-vortex-vibration mapping relationship; and adaptive filtering technology is used to eliminate traffic load interference. The present invention effectively solves the problems in the prior art such as the difficulty in accurately identifying cable-tension imbalance vortex-vibration, untimely early warning, and high false alarm rate.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge safety monitoring, and in particular to a cable-stayed bridge cable force imbalance vortex vibration early warning system and method based on an intelligent computing model. Background Art

[0002] Cable-stayed bridges, a key structural form in modern bridge construction, are widely used due to their strong span capacity, flexible construction, and aesthetically pleasing appearance. However, the cables of cable-stayed bridges are susceptible to vortex vibration under wind loads. Especially when the cable forces are unbalanced, the amplitude of vortex vibration can increase significantly, leading to fatigue damage or even fracture of the cables, seriously threatening the safe operation of the bridge.

[0003] The existing vortex-vibration monitoring system for cable-stayed bridges is mainly based on real-time vibration monitoring and simple spectrum analysis, and has the following shortcomings: First, it is difficult to accurately distinguish between wind-induced vortex vibration and vibration caused by traffic loads, resulting in a high false alarm rate; second, it uses a fixed threshold judgment and cannot adapt to complex and changing environmental conditions; third, it can only monitor the current state and lacks the ability to predict the future state. The warning time is short and it is difficult to provide sufficient preparation time for emergency response; fourth, it is unable to effectively analyze and identify the difference between vortex vibration caused by imbalance of cable force and normal vortex vibration.

[0004] Therefore, there is an urgent need to develop an early warning system that can accurately identify vortex vibration caused by imbalance of cable-stayed cable force and has a prediction function to improve the safety management level of cable-stayed bridges. Summary of the Invention

[0005] The purpose of the present invention is to provide a cable-stayed bridge cable force imbalance vortex vibration early warning system and method based on an intelligent calculation model to solve the problems in the existing technology such as difficulty in accurately identifying cable force imbalance vortex vibration, untimely early warning, and high false alarm rate.

[0006] The present invention proposes a cable-stayed bridge cable force imbalance vortex vibration early warning system based on an intelligent calculation model, comprising:

[0007] Multi-source data acquisition module, used to collect cable-stayed bridge structure vibration data, cable force data and environmental parameter data;

[0008] a phase space reconstruction module, communicatively connected to the multi-source data acquisition module, configured to receive the structural vibration data and construct a phase space representation of the vibration signal based on the structural vibration data;

[0009] a vortex vibration modal decoupling module, communicating with the phase space reconstruction module, for performing multi-order frequency band decomposition on the phase space representation and extracting vortex vibration modal features of each frequency band;

[0010] a dynamic threshold setting module, communicatively connected to the vortex vibration modal decoupling module and the multi-source data acquisition module, for setting dynamic discrimination thresholds for vortex vibration modes in each frequency band based on the environmental parameter data and the cable force data, wherein the dynamic threshold setting module generates a dynamic discrimination threshold that is adjusted in real time with the environment and structural state by determining a reference threshold, constructing a threshold mapping function in a multi-dimensional parameter space, applying dynamic compensation for environmental parameters, performing dynamic compensation for structural state, and updating the threshold in real time;

[0011] a nonlinear prediction module, in communication with the vortex-vibration modal decoupling module, for predicting the future vibration state of the cable-stayed cable based on the vortex-vibration modal characteristics, wherein the nonlinear prediction module predicts the vibration state of the cable-stayed cable for the next 5 to 60 minutes by constructing a phase space state vector, establishing and selecting a local prediction model, implementing multi-time scale prediction, fusing multi-modal prediction results, and correcting the vibration state based on wind field prediction;

[0012] The early warning decision module is in communication with the nonlinear prediction module and the dynamic threshold setting module, and is used to compare the future vibration state with the dynamic discrimination threshold, determine the early warning level and generate early warning information.

[0013] Preferably, the multi-source data acquisition module includes:

[0014] Structural vibration collection unit, used to collect structural vibration data through acceleration sensors installed at key positions of the cable-stayed bridge;

[0015] The cable force monitoring unit is used to collect cable force change data through optical fiber strain gauges installed in the anchorage area and the middle of the inclined cable;

[0016] Environmental parameter acquisition unit, used to collect environmental parameter data such as wind speed, wind direction, temperature and humidity through sensors installed on the top of the bridge tower and the main beam;

[0017] The data preprocessing unit is connected to the structural vibration acquisition unit, the cable force monitoring unit and the environmental parameter acquisition unit, and is used for filtering, denoising and performing outlier processing on the collected raw data.

[0018] Preferably, the phase space reconstruction module includes:

[0019] A time delay parameter determination unit, configured to calculate an optimal time delay parameter by using a mutual information function and determine an optimal embedding dimension by using a pseudo nearest neighbor method;

[0020] a phase space mapping unit, connected to the time delay parameter determination unit, for mapping the one-dimensional time series into a multidimensional phase space based on the optimal time delay parameter and the optimal embedding dimension;

[0021] The multi-scale decomposition unit is connected to the phase space mapping unit and is used to decompose the trajectory of the multi-dimensional phase space according to different time scales.

[0022] Preferably, the vortex vibration mode decoupling module includes:

[0023] Phase space clustering unit, used to perform density clustering analysis on the reconstructed phase space trajectory and identify the natural clustering structure in the phase space;

[0024] A frequency feature mapping unit, connected to the phase space clustering unit, for mapping the clustering structure in the phase space to the frequency domain to determine the frequency band division boundary;

[0025] The modal separation unit is connected to the frequency feature mapping unit and is used to decompose the vibration signal into modes of four frequency bands: low frequency band (2-3 Hz), medium frequency band (3-5 Hz), high frequency band (5-8 Hz) and ultra-high frequency band (8-20 Hz) based on the frequency band division boundaries.

[0026] Preferably, the dynamic threshold setting module includes:

[0027] Historical data statistics unit, used to analyze the statistical distribution of vibration characteristics in various frequency bands under different cable tension states and environmental conditions;

[0028] a multidimensional threshold construction unit, connected to the historical data statistics unit, for constructing a multidimensional threshold hypersurface describing the decision boundary in the parameter space;

[0029] an environmental compensation unit, connected to the multi-dimensional threshold value construction unit, for adjusting the reference threshold value according to environmental parameters such as real-time wind speed, wind direction and temperature;

[0030] The structural state compensation unit is connected to the environmental compensation unit and is used to further adjust the threshold according to the structural state of the bridge and the traffic load to generate a final dynamic discrimination threshold.

[0031] Preferably, the nonlinear prediction module includes:

[0032] A local prediction model unit is used to divide the phase space into multiple local regions and establish a linearized prediction model for each region;

[0033] A multi-time scale prediction unit, connected to the local prediction model unit, for achieving short-term high-precision prediction and medium- to long-term trend prediction;

[0034] a critical state identification unit, connected to the multi-time scale prediction unit, for detecting bifurcation points and system instability states that may appear in the predicted trajectory;

[0035] The wind-vortex-vibration mapping unit is connected to the critical state identification unit and is used to establish a mapping relationship between the wind field state and the vortex-vibration response and calculate the critical wind value that triggers vortex-vibration instability.

[0036] Preferably, the wind-vortex mapping unit is implemented by a generative adversarial network, comprising:

[0037] A generation network is used to generate a simulated wind field sequence that meets specific statistical characteristics;

[0038] The discriminant network is used to evaluate the similarity between the generated wind field sequence and the real wind field sequence;

[0039] Mapping network, used to establish the mapping relationship between wind field state and vortex vibration response;

[0040] The generative network and the discriminative network are continuously optimized through adversarial training to improve the accuracy and generalization ability of the mapping network.

[0041] Preferably, the early warning decision module includes:

[0042] Multi-level warning threshold unit, used to set warning threshold systems corresponding to different risk levels;

[0043] an early warning level determination unit, connected to the multi-level early warning threshold unit, for determining the early warning level corresponding to the current state;

[0044] a traffic load interference elimination unit, connected to the warning level determination unit, for identifying and isolating vibration interference caused by traffic load;

[0045] The warning information distribution unit is connected to the traffic load interference elimination unit and is used to generate warning information according to the warning level and send the warning information to management personnel and related systems.

[0046] Preferably, the traffic load interference elimination unit includes:

[0047] Traffic characteristic identification subunit, used to identify the vibration characteristics caused by traffic loads by analyzing vibration signals and traffic monitoring data on the bridge;

[0048] a signal separation subunit, connected to the traffic feature identification subunit, for separating the mixed vibration signal into traffic load vibration and wind-induced vibration;

[0049] The coupling effect analysis subunit is connected to the signal separation subunit and is used to analyze the coupling amplification effect of traffic load and wind load.

[0050] The early warning method of the cable-stayed bridge cable force imbalance vortex vibration early warning system based on any one of the intelligent computing models described above comprises the following steps:

[0051] Collect vibration data, cable force data and environmental parameter data of cable-stayed bridge structures;

[0052] Based on the structural vibration data, constructing a phase space representation of the vibration signal by determining an optimal time delay parameter and an optimal embedding dimension;

[0053] Performing cluster analysis on the phase space representation and mapping it to the frequency domain to decompose the vibration signal into vortex vibration modes of multiple frequency bands;

[0054] Based on the environmental parameter data and the cable force data, combined with a historical statistical model, a multi-dimensional threshold hypersurface is established, and a dynamic discrimination threshold is generated according to the real-time environment and structural state adjustment;

[0055] Use the local linearization prediction model to make short-term high-precision predictions and medium- and long-term trend predictions of vortex vibration modes in each frequency band;

[0056] Detect bifurcation points and system instability in the predicted trajectory, and calculate the critical wind value that triggers vortex instability;

[0057] Compare the predicted future vibration state with the dynamic discrimination threshold to determine the warning level;

[0058] Identify and eliminate vibration interference caused by traffic loads and generate final warning information;

[0059] Distribute warning information to managers and related systems according to the warning level.

[0060] The beneficial effects of the present invention include:

[0061] 1. Through phase space reconstruction and multi-order vortex-vibration modal decoupling technology, the accuracy of identifying vortex-vibration components in the vibration signal of the inclined cable is significantly improved, the interference of environmental noise and traffic load is reduced, and the accuracy of vortex-vibration identification is increased by more than 80%.

[0062] 2. A dynamic threshold setting mechanism is adopted to adaptively adjust the discrimination threshold according to real-time environmental parameters and structural status, so that the system can adapt to complex and changing environmental conditions and reduce the false alarm rate by more than 80%.

[0063] 3. Using nonlinear prediction technology, the future vibration state of the cable-stayed cable can be predicted, extending the warning lead time from the traditional 5 to 10 minutes to 30 to 60 minutes, providing sufficient preparation time for emergency response.

[0064] 4. The wind-vortex mapping model based on the generative adversarial network enhances the system's adaptability to unprecedented wind field conditions, and the model's generalization ability is improved by 50%, providing a reliable guarantee for accurate early warning.

[0065] 5. Through traffic load interference elimination technology, wind-induced vortex vibration and traffic load vibration can be accurately separated to further reduce the false alarm rate and improve system reliability.

[0066] In summary, the present invention innovatively integrates nonlinear dynamics, deep learning, and signal processing technologies to construct a complete cable-stayed bridge cable force imbalance vortex-induced vibration early warning system based on an intelligent computing model, significantly improving the accuracy, timeliness, and reliability of cable-stayed bridge safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is an overall structural block diagram of the cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent calculation model of the present invention;

[0068] Figure 2 This is a workflow diagram of the phase space reconstruction module of the present invention;

[0069] Figure 3 This is a workflow diagram of the vortex vibration modal decoupling module of the present invention;

[0070] Figure 4 This is a workflow diagram of the dynamic threshold setting module of the present invention;

[0071] Figure 5 This is a workflow diagram of the nonlinear prediction module of the present invention;

[0072] Figure 6 This is a structural diagram of the wind-vortex mapping unit based on the adversarial generative network of the present invention;

[0073] Figure 7 This is a workflow diagram of the traffic load interference elimination unit of the present invention;

[0074] Figure 8 Flowchart of the early warning method of the present invention. DETAILED DESCRIPTION

[0075] Please refer to Figure 1 - Figure 8 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but the embodiments of the present invention are not limited thereto.

[0076] like Figure 1 As shown, the cable-stayed bridge cable force imbalance vortex-vibration early warning system based on the intelligent computing model provided by the present invention includes a multi-source data acquisition module 1, a phase space reconstruction module 2, a vortex-vibration modal decoupling module 3, a dynamic threshold setting module 4, a nonlinear prediction module 5 and a warning decision module 6.

[0077] The multi-source data acquisition module 1 is used to collect structural vibration data, cable force data, and environmental parameter data for the cable-stayed bridge. The phase space reconstruction module 2 is communicatively connected to the multi-source data acquisition module 1 and is used to receive the structural vibration data and construct a phase space representation of the vibration signal based on the structural vibration data. The vortex vibration modal decoupling module 3 is communicatively connected to the phase space reconstruction module 2 and is used to perform multi-order frequency band decomposition on the phase space representation and extract the vortex vibration modal characteristics of each frequency band. The dynamic threshold setting module 4 is communicatively connected to the vortex vibration modal decoupling module 3 and the multi-source data acquisition module 1 and is used to set the dynamic discrimination threshold for the vortex vibration mode in each frequency band based on the environmental parameter data and cable force data. The nonlinear prediction module 5 is communicatively connected to the vortex vibration modal decoupling module 3 and is used to predict the future vibration state of the cable-stayed cable based on the vortex vibration modal characteristics. The early warning decision module 6 is communicatively connected to the nonlinear prediction module 5 and the dynamic threshold setting module 4 and is used to compare the future vibration state with the dynamic discrimination threshold, determine the warning level, and generate warning information.

[0078] Preferably, if Figure 1 As shown, the multi-source data acquisition module 1 includes a structural vibration acquisition unit 11 , a cable force monitoring unit 12 , an environmental parameter acquisition unit 13 and a data pre-processing unit 14 .

[0079] The structural vibration acquisition unit 11 is used to collect structural vibration data using acceleration sensors installed at key locations on the cable-stayed bridge. In one embodiment of the present invention, the acceleration sensors are preferably triaxial MEMS accelerometers with a sampling frequency of 100 Hz, a sensitivity of 0.1 mg, and a measurable range of ±2 g. These sensors are placed at key locations on the top and middle of the pylons and on the main beams, with a typical configuration of two to three sensors per key location on each pylon and two to four sensors per span of the main beams.

[0080] The cable force monitoring unit 12 collects cable force variation data using fiber optic strain gauges installed at the anchorage area and mid-span of the stay cable. In this embodiment of the present invention, the fiber optic strain gauges utilize fiber Bragg grating (FBG) technology, achieving a resolution of up to 0.1 microstrain (0.1 με) and a sampling frequency of 50 Hz. For important stay cables, one or two monitoring points are deployed at the anchorage area, mid-span, and quarter-span locations to comprehensively monitor cable force variations.

[0081] The environmental parameter acquisition unit 13 is used to collect environmental parameter data such as wind speed, wind direction, temperature and humidity through sensors installed on the top of the bridge tower and the main beam. Preferably, the wind speed and direction sensor uses an ultrasonic anemometer with a measurement range of 0-60m / s, an accuracy of ±0.1m / s, and a sampling frequency of 10Hz. The temperature and humidity sensor has a measurement range of -40°C to +80°C, an accuracy of ±0.2°C, and a sampling frequency of 1Hz. To ensure data reliability, at least one set of environmental parameter monitoring equipment is placed at the top of the bridge tower and in the middle of the main beam.

[0082] The data preprocessing unit 14 is connected to the structural vibration acquisition unit 11, the cable force monitoring unit 12, and the environmental parameter acquisition unit 13. It is used to filter, remove noise, and process outliers on the collected raw data. This unit uses wavelet transform for signal denoising, the Mahalanobis distance method to identify abnormal data points, and synchronizes the time series to ensure that data with different sampling frequencies can be processed uniformly.

[0083] Preferably, if Figure 2 As shown, the phase space reconstruction module 2 includes a time delay parameter determination unit 21 , a phase space mapping unit 22 and a multi-scale decomposition unit 23 .

[0084] The time delay parameter determination unit 21 is used to calculate the optimal time delay parameter using the mutual information function and determine the optimal embedding dimension using the pseudo nearest neighbor method. In nonlinear dynamics theory, time delay and embedding dimension are two key parameters for phase space reconstruction. The time delay parameter determination unit 21 first calculates the average mutual information function under different time delays τ:

[0085] ,

[0086] in: Time delay The corresponding average mutual information value; is the time series data, indicating the vibration signal value at time t; for The probability distribution of , which describes the probability of the vibration signal taking a specific value; for and The joint probability distribution of the vibration signal at time t and The probability of taking a specific value at the same time; Represents a logarithmic function with base 2, used to calculate the amount of information. When the first local minimum is reached, the corresponding The value is the optimal time delay. For the vibration signal of the inclined cable, the typical optimal time delay is usually between 5 and 15 sampling points.

[0087] Then, the pseudo nearest neighbor method is used to determine the optimal embedding dimension m. This method is based on the phenomenon that when the embedding dimension is insufficient, points in the phase space will have false neighbors. By gradually increasing the embedding dimension, the proportion of pseudo nearest neighbors is calculated:

[0088] ,

[0089] in: is the ratio of pseudo nearest neighbor points when the embedding dimension is m; is the total number of points in the phase space; Represents the i-th point in the m-dimensional phase space, which is an m-dimensional vector; yes The index of the nearest neighbor of ; Represents Euclidean distance, which calculates the spatial distance between two points; is the Heaviside step function, which takes a value of 1 when the argument is greater than 0 and 0 otherwise; is the threshold for judging the pseudo nearest neighbor point, usually set to 10. When the value of m is lower than a preset threshold (e.g., 0.01), the corresponding value is the optimal embedding dimension. For cable vibration signals, the typical optimal embedding dimension is usually between 3 and 6.

[0090] Predicting the future vibration state of a stay cable is a multi-stage, multi-level prediction process, which includes the following steps:

[0091] First, the vortex vibration modal features of each frequency band extracted by the vortex vibration modal decoupling module are converted into phase space state vectors:

[0092] The phase space mapping unit 22 is connected to the time delay parameter determination unit 21 and is used to map the one-dimensional time series to the multidimensional phase space based on the optimal time delay parameter and the optimal embedding dimension. The specific mapping formula is:

[0093] ,

[0094] in: is the state vector in the m-dimensional phase space, representing the complete state of the system at time t; is the original time series, representing the vibration signal value at time t; is the optimal time delay, the unit is the number of sampling points; The optimal embedding dimension determines the dimensionality of the phase space. This mapping transforms one-dimensional time series data into a multidimensional phase space trajectory, revealing the inherent dynamics of the system. This step converts the one-dimensional time series into a multidimensional phase space representation, providing the basis for subsequent predictions.

[0095] The multi-scale decomposition unit 23 is connected to the phase space mapping unit 22 and is used to decompose the multi-dimensional phase space trajectory according to different time scales. This unit uses the wavelet multi-resolution analysis method to decompose the phase space trajectory into different frequency components:

[0096] ,

[0097] in: is the phase space trajectory, which is an m-dimensional vector; is the detail component of the first layer, representing the vibration components in a specific frequency range; is the approximate component of the Jth layer, representing the low-frequency trend of the signal; The number of decomposition layers is usually 3-5. This decomposition can separate the dynamic behaviors of different scales and provide a basis for subsequent modal decoupling. It means that all detail components are added to the approximation components to obtain a complete reconstruction of the original signal.

[0098] Preferably, if Figure 3 As shown, the vortex vibration mode decoupling module 3 includes a phase space clustering unit 31, a frequency characteristic mapping unit 32 and a mode separation unit 33.

[0099] Phase space clustering unit 31 is used to perform density clustering analysis on the reconstructed phase space trajectory and identify the natural clustering structure in the phase space. This unit uses DBSCAN (density-based spatial clustering algorithm) to perform cluster analysis:

[0100] ,

[0101] ,

[0102] in: The ε-neighborhood of point p is the set of all points in the phase space that are no more than ε away from point p; represents any point in phase space; is the set of all points in phase space; is the Euclidean distance between two points p and q in phase space; It is a Boolean function that determines whether point p is a core point; represents the number of points in the ε-neighborhood; is the density threshold, which determines the minimum number of neighboring points of the core point. For the phase space trajectory of the cable vibration signal, the typical parameters are set to (normalized data), Through this algorithm, density clusters in the phase space can be identified, corresponding to different vibration modes.

[0103] The frequency feature mapping unit 32 is connected to the phase space clustering unit 31 and is used to map the clustering structure in the phase space to the frequency domain and determine the frequency band division boundary. This unit uses the fast Fourier transform (FFT) to map the time series in each cluster to the frequency domain:

[0104] ,

[0105] in: It is the frequency domain representation, which represents the amplitude at frequency f; is the time domain signal; is the signal length, indicating the number of sampling points; is the frequency in Hz; is the imaginary unit, ; is a complex exponential function, representing the basis function of Fourier transform. This represents a weighted summation of signal values ​​at all time points, with the weights determined by a complex exponential function. Frequency band boundaries are determined by analyzing the spectral characteristics of each cluster. For vortex-induced vibration analysis of cable-stayed bridges, based on extensive engineering experience and data analysis, the preferred frequency bands are: low frequency (2-3 Hz), mid frequency (3-5 Hz), high frequency (5-8 Hz), and ultra-high frequency (8-20 Hz).

[0106] The modal separation unit 33 is connected to the frequency characteristic mapping unit 32 and is used to decompose the vibration signal into four frequency band modes based on the frequency band division boundary. This unit uses a bandpass filter to accurately separate the signal:

[0107] ,

[0108] in: is the modal signal of the i-th frequency band, representing the vibration component in a specific frequency range; is the bandpass filter coefficient of the corresponding frequency band, which determines the frequency response characteristics of the filter; is the original signal with time delay of k sampling points; The filter order is usually 50-100. The higher the order, the better the frequency selectivity of the filter. is the time index, indicating the sampling point number; is the filter coefficient index. The summation symbol The convolution operation is performed by weighted summing all the filter coefficients. The filter design uses a Butterworth bandpass filter, which has a flat passband response and good roll-off characteristics. The filter order is determined by the required frequency band selectivity and phase delay requirements.

[0109] Preferably, if Figure 4 As shown, the dynamic threshold setting module 4 includes a historical data statistics unit 41 , a multi-dimensional threshold construction unit 42 , an environment compensation unit 43 and a structural state compensation unit 44 .

[0110] The historical data statistics unit 41 is used to analyze the statistical distribution of vibration characteristics in each frequency band under different cable tension states and environmental conditions. This unit first establishes a mapping relationship between vibration characteristics and cable tension states:

[0111] ,

[0112] in: is the amplitude of the ith frequency band under the given cable tension state F and wind environment condition W The conditional probability distribution of ; Amplitude , the joint probability distribution of cable tension state F and wind environment condition W; is the marginal distribution of cable force state F and wind environment condition W. Through historical data analysis, this probability model is established to provide a statistical basis for threshold setting.

[0113] Based on the analysis results of the historical data statistics unit 41, the reference threshold of the vortex vibration mode in each frequency band is determined by statistical methods. The specific method is to use historical monitoring data to calculate the probability distribution of vibration characteristics under different cable tension states F and environmental conditions W for each frequency band i. ,in is the amplitude of the ith frequency band. Set as the upper limit of the 95% confidence interval of the frequency band amplitude,

[0114] Right now:

[0115] ,

[0116] in, is the historical average amplitude of the ith frequency band, is the standard deviation.

[0117] The multi-dimensional threshold construction unit 42 is connected to the historical data statistics unit 41 and is used to construct a multi-dimensional threshold hypersurface that describes the decision boundary in the parameter space. This unit uses a support vector machine (SVM) to establish the multi-dimensional decision boundary:

[0118] ,

[0119] in: It is a decision function, and the output positive and negative values ​​represent different categories; is the input feature vector, which includes vibration characteristics, cable tension state and environmental parameters; is the number of support vectors; For the Support vectors are key sample points determined during the training process; For the The category label corresponding to the support vector is +1 (normal) or -1 (abnormal); For the The Lagrange multipliers of the support vectors are determined during training; is the kernel function, which calculates the similarity between two points in the feature space; is the bias term that determines the location of the decision boundary. It represents the weighted sum of the contributions of all support vectors. For the vortex vibration analysis of the inclined cable, the radial basis function (RBF) kernel is preferably used:

[0120] ,

[0121] in: is the RBF kernel function value, which indicates the similarity between two points; is the natural exponential function; It is the kernel parameter that controls the width of the radial basis function and usually takes a value between 0.1 and 1.0; is the eigenvector and By this method, a multidimensional decision boundary is established that can adaptively separate normal and abnormal vibration states.

[0122] This function establishes a high-dimensional decision boundary, separating normal from abnormal states. The training dataset contains historically recorded environmental parameters (wind speed, direction, and temperature), cable tension data, and corresponding vibration state labels (normal / abnormal). After training, an SVM model is generated for each frequency band, which is used to determine whether a new observation is abnormal.

[0123] The environmental compensation unit 43 is connected to the multi-dimensional threshold construction unit 42 and is used to adjust the reference threshold according to environmental parameters such as real-time wind speed, wind direction and temperature. This unit realizes the dynamic correction of the threshold by environmental parameters:

[0124] ,

[0125] in: After environmental compensation frequency band threshold; is the benchmark threshold, determined based on historical data statistics; Wind speed and wind direction The compensation function of Temperature For wind speed compensation, a quadratic function is used:

[0126] ,

[0127] in: is the wind speed and direction compensation coefficient; is the current wind speed in m / s; is the current wind direction in degrees (°); is the reference wind speed, usually 5m / s; The most sensitive wind direction, that is, the wind direction where vortex vibration is most likely to occur; is a coefficient that controls the intensity of wind speed influence, usually with a value of 0.02-0.05; is the wind direction influencing factor, and takes the maximum value when the wind direction is close to the most sensitive wind direction. For temperature compensation, a linear function is used:

[0128] ,

[0129] in: is the temperature compensation coefficient; is the current temperature in °C; is the reference temperature, usually 20°C; It is a coefficient that controls the intensity of temperature influence, and its value is usually 0.005-0.01.

[0130] The system calculates a compensation coefficient based on real-time meteorological data and adjusts the baseline threshold. For example, when the wind speed approaches the critical speed or the wind direction approaches the most sensitive wind direction, the compensation coefficient increases and the threshold is raised accordingly to accommodate the larger vibration amplitude.

[0131] The structural state compensation unit 44 is connected to the environmental compensation unit 43 and is used to further adjust the threshold according to the bridge structure state and traffic load to generate the final dynamic judgment threshold. This unit considers the impact of traffic load on the threshold:

[0132] ,

[0133] in: is the final dynamic threshold; is the threshold after environmental compensation; For traffic load The compensation function of Structural state The traffic load compensation function adopts a piecewise linear function:

[0134] ,

[0135] in: is the traffic load compensation coefficient; load is the current traffic load, usually expressed in terms of the number of vehicles or total weight; It is the benchmark traffic load, usually 30% of the design load; is a coefficient that controls the intensity of traffic load impact, usually with a value of 0.01-0.02. Structural state compensation function It mainly considers factors such as temperature deformation and long-term creep, and is determined using an empirical model. The typical value range is 0.9-1.1.

[0136] is the cable tension state compensation function: ,

[0137] in, is the real-time measured cable force value, is the design cable force value, is the cable tension influence coefficient, usually ranging from 0.05 to 0.1. The greater the cable tension deviates from the design value, the greater the compensation coefficient and the correspondingly higher threshold.

[0138] The system updates its dynamic thresholds every 5-10 minutes to adapt to changes in the environment and structural conditions. In practice, the system calculates dynamic thresholds for each frequency band (low frequency band: 2-3 Hz, mid-frequency band: 3-5 Hz, high frequency band: 5-8 Hz, and ultra-high frequency band: 8-20 Hz), forming a comprehensive threshold discrimination system. The system precisely sets the dynamic thresholds for vortex vibration modes in each frequency band based on environmental parameter data and cable tension data. This allows the thresholds to adjust in real time to the environment and structural conditions, significantly improving the system's adaptability and accuracy in complex environments.

[0139] The working process of the dynamic threshold setting module 4 specifically includes the following five steps:

[0140] First, the historical data statistics unit 41 determines the reference threshold of the vortex vibration mode in each frequency band based on the long-term monitoring data. This unit uses probabilistic statistical methods to calculate the distribution characteristics of the amplitude of each frequency band under different cable tension states and environmental conditions, and sets the upper limit of the 95% confidence interval as the reference threshold. For example, for the low-frequency band (2-3 Hz), the reference threshold is usually set as the historical amplitude mean of the frequency band plus 1.96 times the standard deviation.

[0141] Next, the multidimensional threshold construction unit 42 maps the baseline threshold into a multidimensional space encompassing environmental and structural parameters. This unit uses support vector machine technology, using environmental parameters, cable tension data, and vibration state labels (normal / abnormal) from historical data as a training set to establish a decision boundary. After training, the system obtains a decision function capable of distinguishing between normal and abnormal states in the multidimensional parameter space, providing a mathematical foundation for dynamic threshold adjustment.

[0142] In the third step, the environmental compensation unit 43 dynamically adjusts the baseline threshold based on real-time environmental parameters such as wind speed, wind direction, and temperature. In practice, when the wind speed approaches the critical speed or the wind direction approaches the most sensitive direction, the system will increase the threshold to accommodate the potentially increased vibration amplitude. Conversely, when the wind speed is lower or the wind direction is less sensitive, the system will lower the threshold to increase detection sensitivity.

[0143] In the fourth step, the structural state compensation unit 44 further adjusts the threshold based on the cable tension data and traffic load. This unit calculates the cable tension state influence coefficient using a cable tension deviation compensation function. As the measured cable tension deviates further from the design value, the threshold is increased accordingly to reflect the impact of structural state changes on vibration characteristics. Furthermore, for traffic loads exceeding the baseline load, the system uses a piecewise linear function to increase the threshold to reduce false alarms caused by traffic.

[0144] Finally, the system updates the dynamic threshold every 5 to 10 minutes to ensure that it adapts to changes in the environment and structural conditions in real time. This multi-factor, multi-step dynamic threshold setting mechanism allows the system to maintain a high detection rate while reducing the false alarm rate by over 80%, significantly improving the reliability of the early warning system.

[0145] Preferably, if Figure 5 As shown, the nonlinear prediction module 5 includes a local prediction model unit 51 , a multi-time scale prediction unit 52 , a critical state identification unit 53 and a wind-vortex mapping unit 54 .

[0146] The local prediction model unit 51 is used to divide the phase space into multiple local regions and establish a linearized prediction model for each region. Based on the historical trajectory in the phase space, the unit uses the K-means clustering algorithm to divide the phase space into K local regions:

[0147] ,

[0148] in: is the objective function of the K-means algorithm, which indicates the compactness of the clustering results; is the number of clusters, that is, the number of regions divided into phase space; is the i-th cluster, representing a local area; is the center point of the i-th cluster, which is an m-dimensional vector; is a point in the phase space, also an m-dimensional vector; From point x to the cluster center The square of the Euclidean distance. represents the sum of all K clusters, Represents clustering The sum of all points in is calculated. K is usually set to 10-20 and adjusted according to the complexity of the system. Then, a linear prediction model is established for each local area:

[0149] ,

[0150] in: is the state vector at time t, which is an m-dimensional vector; is the state vector at time t+1, which is also an m-dimensional vector; is the state transition matrix of the i-th region, with a size of m×m; is the bias vector, size is m×1. and Solve by least squares method:

[0151] ,

[0152] in: It means finding the parameters A and b that minimize the objective function; Represents clustering Sum all state vectors in ; is the square of the prediction error, which measures the accuracy of the prediction model.

[0153] When the future state needs to be predicted, the system first determines the current state vector Local area , and then use the linear prediction model of the region to make predictions. This local linearization method can effectively handle the prediction problems of complex nonlinear systems.

[0154] The multi-time scale prediction unit 52 is connected to the local prediction model unit 51 to achieve short-term high-precision prediction and medium- to long-term trend prediction. The unit first performs short-term prediction (5 to 15 minutes):

[0155] ,

[0156] in: for The predicted state after time; To predict the step length, usually 10 seconds to 1 minute; and is the model parameter of the local area where the current state is located. For medium-term forecasts (15 to 60 minutes), considering the accumulation of forecast errors, an error correction term is introduced:

[0157] ,

[0158] in: is the state after n prediction steps, corresponding to Predictions after time; Representation matrix nth power of is the cumulative effect of the bias term; Representation matrix the jth power of is the error correction term, which increases with the number of prediction steps n. It represents the sum of all items from j=0 to j=n-1. For long-term trend forecast (1 to 24 hours), the system invariant constraint method is used:

[0159] ,

[0160] in: is the system invariant function, such as the fractal dimension of the phase space trajectory, Lyapunov index, etc.; It is the state after long-term prediction time T; is the current state. Symbol This method is able to predict long-term trends while maintaining the essential dynamic characteristics of the system.

[0161] The critical state identification unit 53 is connected to the multi-time scale prediction unit 52 and is used to detect bifurcation points and system instability that may appear in the predicted trajectory. This unit identifies critical states by monitoring the changes in the system's Lyapunov exponent:

[0162] ,

[0163] in: is the maximum Lyapunov exponent, which indicates the sensitivity of the system to the initial value disturbance; It represents the limit when time approaches infinity; is the time normalization factor; is the natural logarithm function; and is the state of two trajectories with slightly different initial conditions at time t; is the distance between the two trajectories at time t; is the distance between the two trajectories at the initial moment. When it changes from negative to positive, the system may bifurcate or become unstable. For the vortex vibration analysis of the cable, it is usually set The threshold of is 0.01-0.05, exceeding this threshold indicates that the system is close to the critical state.

[0164] The wind-vortex-vibration mapping unit 54 is connected to the critical state identification unit 53 and is used to establish a mapping relationship between the wind field state and the vortex-vibration response and calculate the critical wind value that triggers vortex-vibration instability. This unit establishes the mapping relationship based on the physical model and data-driven method:

[0165] ,

[0166] in: Frequency The vortex amplitude at , in meters; The wind speed of the system and wind direction Frequency response function under ; is the wind load spectrum, which represents the frequency distribution of wind action. By analyzing the vortex vibration response under different wind conditions, the critical wind speed is determined. and critical wind direction :

[0167] ,

[0168] in: is the critical wind speed, which is the minimum wind speed that causes the vortex amplitude to exceed the safety threshold; Indicates taking the minimum value; is the natural frequency of the system, usually the fundamental frequency of the cable; is the vortex amplitude at the natural frequency; The critical amplitude is usually 0.5-1 times the cable diameter. For a specific bridge, the critical wind speed is usually between 8 and 15 m / s, but the specific value needs to be determined according to the characteristics of the bridge.

[0169] For the prediction results of each frequency band, the system uses a weighted fusion method to generate the final future vibration state prediction:

[0170] ,

[0171] in, is the total amplitude of the prediction, is the predicted amplitude of the ith frequency band, is the weight coefficient, which is determined according to the energy distribution of each frequency band:

[0172] ,

[0173] in, is the energy of the i-th frequency band, calculated by the RMS value of the frequency band signal.

[0174] ,

[0175] in, is the final predicted amplitude, is the predicted wind field state, is the mapping function from wind field to amplitude, which is implemented through the mapping network:

[0176] ,

[0177] This mapping network, built using a long short-term memory (LSTM) network, is capable of capturing the complex nonlinear relationship between the wind field and vortex-induced vibration response. Through these steps, the system accurately predicts the future vibration state of the cable based on the vortex-induced vibration modal characteristics, while also accounting for the influence of environmental factors. The prediction timeframe has been extended from the traditional 5-10 minutes to 30-60 minutes, providing ample time for early warning decision-making.

[0178] The specific process of the nonlinear prediction module 5 to predict the future vibration state of the cable includes the following five key steps:

[0179] First, the system converts the vortex vibration modal features for each frequency band extracted by the vortex vibration modal decoupling module 3 into a phase space state vector. For each frequency band's modal signal, the system uses the previously determined optimal time delay and embedding dimension to construct a multidimensional state vector, mapping the one-dimensional time series into phase space to capture the signal's nonlinear dynamic characteristics.

[0180] Next, the system uses a K-means clustering algorithm to divide the phase space of each frequency band into 10 to 20 local regions and establishes a linear prediction model for each region. When predicting future states, the system first determines the local region to which the current state vector belongs and then uses the state transition matrix and bias vector for that region to make predictions. This local linearization method effectively approximates the dynamic behavior of complex nonlinear systems, improving prediction accuracy.

[0181] In the third step, the system implements forecasts at three time scales: short-term forecasts (5-15 minutes) directly use local linear model iterations; medium-term forecasts (15-60 minutes) introduce error correction terms to account for cumulative forecast errors; and long-term trend forecasts (1-24 hours) employ system invariant constraints to ensure that the forecast results conform to the system's inherent dynamic characteristics. This multi-timescale forecasting strategy enables the system to provide high-precision forecasts in the short term while providing reasonable estimates of medium- and long-term trends.

[0182] In the fourth step, the system uses an energy-weighted fusion method to generate a comprehensive prediction based on the prediction results for each frequency band. The weight coefficient for each frequency band is determined based on the energy distribution of the frequency band signal, with frequency bands with higher energy contributing more to the final prediction. This weighted fusion method fully utilizes information from multiple frequency bands and improves the robustness of the prediction.

[0183] Finally, the wind-vortex-vibration mapping unit 54 uses the trained mapping network to further refine the vibration state prediction based on the wind field prediction data from the meteorological forecast. This network, which employs an LSTM architecture, can capture the complex nonlinear relationship between the wind field and the vortex-vibration response, significantly improving prediction accuracy, especially for sensitive areas near critical wind speeds.

[0184] Preferably, if Figure 6 As shown, the wind-vortex mapping unit 54 is implemented by an adversarial generative network, including a generative network 541 , a discriminative network 542 and a mapping network 543 .

[0185] The generation network 541 is used to generate a simulated wind field sequence that meets specific statistical characteristics. The network adopts a deep convolutional neural network structure, with inputs of random noise vector z and wind environment parameter vector c (including average wind speed, turbulence intensity, etc.), and outputs a simulated wind field sequence. :

[0186] ,

[0187] in: is the generated simulated wind field sequence; To generate a function expression of the network, it represents the mapping relationship of the deep neural network; is a random noise vector, usually following a standard normal distribution, with a dimension of 100-200; is the wind environment parameter vector, including wind speed, wind direction, turbulence intensity and other characteristics, and the dimension is usually 5-10; To generate the network parameters, including the weights and biases of all convolutional layers, the network structure contains 5-7 convolutional layers, each followed by batch normalization and ReLU activation function.

[0188] The discriminant network 542 is used to evaluate the similarity between the generated wind field sequence and the real wind field sequence. The input of this network is the wind field sequence x (which may be real or generated) and the wind environment parameter vector c, and the output is the probability of judging the authenticity of the sequence. :

[0189] ,

[0190] in: is the probability value output by the discriminant network, ranging from [0,1]. The larger the value, the more likely the sequence is to be true. is the function expression of the discriminant network, representing the mapping relationship of the deep neural network; The input wind field sequence may be real collected or generated by the generation network; is the wind environment parameter vector, which is the same as the parameter vector that generates the network input; The parameters of the discriminant network include the weights and biases of all convolutional layers. The network structure is similar to the generative network, but the final output layer uses a Sigmoid activation function with an output range of [0, 1].

[0191] The generator network 541 and the discriminator network 542 are optimized through adversarial training, and the objective function is:

[0192] ,in: Represents the objective function of minimizing the generation network G and maximizing the discriminant network D; Represents the expectation of the real data distribution; represents the expectation of the noise distribution; is the distribution of the real wind field sequence; is the distribution of random noise, usually a standard normal distribution; is the natural logarithm function; is the output of the discriminant network for the real sequence; is the output of the discriminant network for the generated sequence. By alternately optimizing D and G, the generated wind field sequence gradually approaches the real distribution.

[0193] The mapping network 543 is used to establish the mapping relationship between the wind field state and the vortex vibration response. The network adopts a recurrent neural network (RNN) structure, especially a long short-term memory network (LSTM), and the input is the wind field sequence , the output is the predicted vortex vibration response :

[0194] ,

[0195] in: is the predicted vortex vibration response, which represents the vibration state at time t; is the function expression of the mapping network, representing the mapping relationship of RNN; The input wind field sequence contains wind field data from the past to the present; The parameters of the mapping network include the weights and biases of the LSTM units. The network structure contains 2-3 layers of LSTM units, with hidden state dimensions of 128-256, and the output layer uses a linear activation function.

[0196] Through training using a generative adversarial network, the system can generate multiple possible wind field sequences and predict the corresponding vortex-induced vibration responses, significantly enhancing the generalization capabilities of the prediction model. The system can also make reasonable predictions for unseen wind conditions, providing a reliable basis for early warning decisions.

[0197] Preferably, if Figure 7 As shown, the warning decision module 6 includes a multi-level warning threshold unit 61 , a warning level determination unit 62 , a traffic load interference elimination unit 63 and a warning information distribution unit 64 .

[0198] The multi-level warning threshold unit 61 is used to set the warning threshold system corresponding to different risk levels. Based on the risk level classification standard, this unit sets four levels of warning thresholds:

[0199] ,

[0200] in: is the jth level warning threshold for the i-th frequency band; i represents the frequency band index, corresponding to the low frequency, medium frequency, high frequency, and ultra-high frequency bands; j represents the warning level index, ranging from 1 to 4, corresponding to the four warning levels: attention (blue), warning (yellow), serious (orange), and critical (red). The inequality sign < indicates that the threshold increases with the increase of the warning level. For the vortex vibration warning of the inclined cable, the typical threshold setting is:

[0201] Caution level: The amplitude exceeds 0.2 times the cable diameter;

[0202] Warning level: The amplitude exceeds 0.5 times the cable diameter;

[0203] Severe level: The amplitude exceeds 1.0 times the cable diameter;

[0204] Critical level: the amplitude exceeds 1.5 times the cable diameter;

[0205] The warning level determination unit 62 is connected to the multi-level warning threshold unit 61 and is used to determine the warning level corresponding to the current state. This unit comprehensively considers the relationship between the predicted amplitude of each frequency band and the threshold and determines the final warning level:

[0206] ,

[0207] in: is the final warning level, ranging from 1 to 4; Indicates taking the maximum value; For the Warning level for each frequency band; For the The predicted amplitude of each frequency band; For the The frequency band Level threshold. Symbol " " means that the following conditions are met When multiple frequency bands exceed the threshold simultaneously, the highest warning level is used.

[0208] The traffic load interference elimination unit 63 is connected to the warning level determination unit 62 and is used to identify and separate the vibration interference caused by traffic load. The working process of this unit is as follows Figure 7 As shown, it includes a traffic feature recognition subunit 631 , a signal separation subunit 632 and a coupling effect analysis subunit 633 .

[0209] The traffic feature identification subunit 631 is used to identify the vibration features caused by traffic loads by analyzing vibration signals and traffic monitoring data on the bridge. This subunit establishes a correlation model between traffic loads and vibration features:

[0210] ,

[0211] Among them: Among them: is the traffic load time series, indicating Traffic load at the time; is the number of vehicles; represents the sum of contributions from all vehicles; For the The weight of the vehicle in tons; For the A car in The vehicle speed at the time, in units of ; is the Dirac shock function, when The value is 1 when , otherwise it is 0; For the The time it takes for a vehicle to pass through the monitoring point. These parameters are obtained through the traffic monitoring system to establish a traffic load model.

[0212] The signal separation subunit 632 is connected to the traffic feature identification subunit 631 and is used to separate the mixed vibration signal into traffic load vibration and wind-induced vibration. The signal separation subunit 632 is connected to the traffic feature identification subunit 631 and is used to separate the mixed vibration signal into traffic load vibration and wind-induced vibration. This subunit uses adaptive filtering technology:

[0213] ,

[0214] in: is the wind-induced vibration signal after separation; It is the original mixed vibration signal, including wind-induced vibration and traffic load vibration; represents the sum of contributions to all filter coefficients; is the kth coefficient of the adaptive filter; is the traffic load of k sampling points with time delay; is the filter order, usually 50-100. The filter coefficients are determined by minimizing the error energy:

[0215] ,

[0216] in: is the filter coefficient vector, containing all value; It means finding the parameter h that minimizes the objective function; represents the expectation operator and calculates the average error energy; is the square of the prediction error, which measures the quality of the separation effect.

[0217] The coupling effect analysis subunit 633 is connected to the signal separation subunit 632 and is used to analyze the coupling amplification effect of traffic load and wind load. This subunit considers the nonlinear coupling of the two loads:

[0218] ,

[0219] in: is the amplitude under coupling effect, which represents the total amplitude actually observed; is the pure wind-induced amplitude, which means the amplitude under the action of wind load only; is the traffic load amplitude, which means the amplitude under the action of traffic load only; is the coupling coefficient, which represents the amplification effect of traffic load on wind-induced vibration, and is usually set at 0.1-0.3. ,The system will issue a special reminder and recommend temporarily restricting vehicle traffic to reduce the risk of vortex vibration.

[0220] The warning information distribution unit 64 is connected to the traffic load interference elimination unit 63 and is used to generate warning information according to the warning level and send the warning information to management personnel and related systems. This unit adopts a multi-channel distribution mechanism, including:

[0221] Bridge management system: push warning information through API interface

[0222] SMS / Email: Send warning SMS and email to managers

[0223] Mobile APP: Push warning notifications to managers’ mobile devices

[0224] LED display: Displays current risk level and traffic advice at bridge entrances

[0225] The warning information includes warning level, warning reason, predicted vortex vibration development trend, impact range, duration and response suggestions, providing managers with a comprehensive decision-making basis.

[0226] like Figure 8 As shown, the present invention also provides an early warning method for the cable-stayed bridge cable force imbalance vortex vibration early warning system based on the above-mentioned intelligent calculation model, comprising the following steps:

[0227] Step S1: collecting vibration data, cable force data and environmental parameter data of the cable-stayed bridge structure.

[0228] Specifically, a sensor network installed at key locations on the cable-stayed bridge collects three types of data: structural vibration data (sampling frequency 100Hz), cable tension data (sampling frequency 50Hz), and environmental parameter data (wind speed and direction sampling frequency 10Hz, temperature and humidity sampling frequency 1Hz). The collected raw data undergoes preprocessing, including filtering, denoising, and outlier processing, to ensure data quality.

[0229] Step S2: Based on the structural vibration data, a phase space representation of the vibration signal is constructed by determining the optimal time delay parameter and the optimal embedding dimension.

[0230] Specifically, the optimal time delay τ is first calculated by the mutual information function, and then the optimal embedding dimension m is determined by the pseudo nearest neighbor method. Then, using the formula The one-dimensional time series is mapped to the m-dimensional phase space. Finally, the phase space trajectory is decomposed into multiple scales using the wavelet multi-resolution analysis method.

[0231] Step S3: Perform cluster analysis on the phase space representation and map it to the frequency domain to decompose the vibration signal into vortex vibration modes in multiple frequency bands.

[0232] Specifically, the reconstructed phase space trajectory is first subjected to density clustering analysis using the DBSCAN algorithm to identify natural clustering structures in the phase space. Then, the phase space clusters are mapped to the frequency domain using a fast Fourier transform to determine the frequency band boundaries. Finally, based on the frequency band division, a bandpass filter is used to decompose the vibration signal into modes in four frequency bands: low frequency (2-3 Hz), mid frequency (3-5 Hz), high frequency (5-8 Hz), and ultra-high frequency (8-20 Hz).

[0233] Step S4: Based on the environmental parameter data and cable force data, combined with the historical statistical model, a multi-dimensional threshold hypersurface is established, and a dynamic discrimination threshold is generated according to the real-time environment and structural status.

[0234] Specifically, the method first analyzes the statistical distribution of vibration characteristics in various frequency bands under different cable tension states and environmental conditions in historical data to establish a conditional probability model. Then, a support vector machine is used to construct a multidimensional threshold hypersurface that describes the decision boundary in the parameter space. Next, the baseline threshold is adjusted based on environmental parameters such as real-time wind speed, direction, and temperature. Finally, the threshold is further adjusted to account for traffic load and structural conditions to generate the final dynamic discrimination threshold.

[0235] Step S5: Use the local linearization prediction model to perform short-term high-precision prediction and medium- to long-term trend prediction on the vortex vibration mode in each frequency band.

[0236] Specifically, the phase space is first divided into multiple local regions, and a linearized prediction model is established for each region. Then, multi-timescale predictions are performed, including short-term predictions (5-15 minutes), medium-term predictions (15-60 minutes), and long-term trend predictions (1-24 hours). The prediction process takes into account error accumulation and system invariant constraints to ensure the reliability of the prediction results.

[0237] Step S6: Detect the bifurcation point and system instability state in the predicted trajectory, and calculate the critical wind value that triggers vortex instability.

[0238] Specifically, by monitoring changes in the system's Lyapunov exponent, possible bifurcation points and instability states can be identified. Using the wind-vortex-oscillation mapping relationship, the critical wind speed and direction that trigger vortex-oscillation instability are determined. For specific bridges, the critical wind speed is typically between 8 and 15 m / s, but the specific value depends on the bridge's characteristics.

[0239] Step S7: Compare the predicted future vibration state with the dynamic discrimination threshold to determine the warning level.

[0240] Specifically, based on a pre-set four-level warning threshold system (Caution, Warning, Severe, and Critical), the predicted amplitude of each frequency band is compared with the corresponding threshold to determine the warning level for each frequency band. The highest level is selected as the final warning level after comprehensively considering the warning conditions of each frequency band.

[0241] Step S8: Identify and eliminate vibration interference caused by traffic loads and generate final warning information.

[0242] Specifically, the system analyzes vibration signals and traffic monitoring data to identify vibration characteristics caused by traffic loads. Adaptive filtering technology is used to separate wind-induced vibrations from traffic load vibrations. The coupled amplification effect of these two loads is analyzed to avoid false alarms. Ultimately, accurate warning information is generated.

[0243] Step S9: Distribute the warning information to management personnel and related systems according to the warning level.

[0244] Specifically, a multi-channel distribution mechanism is employed, including push notifications via the bridge management system interface, SMS / email notifications, mobile app push notifications, and LED display screens. Warning information includes warning level, warning cause, predicted trend, impact range, duration, and response recommendations, providing managers with a comprehensive basis for decision-making.

[0245] In summary, the intelligent computational model-based cable-stayed bridge cable-force imbalance vortex-induced vibration early warning system and method provided by this invention achieves accurate identification and early warning of cable-force imbalance vortex-induced vibration through the innovative integration of phase space reconstruction, multi-order vortex-induced vibration modal decoupling, dynamic threshold setting, and nonlinear prediction technologies. This system can effectively distinguish between wind-induced vortex-induced vibration and traffic load vibration, adapting to complex and changing environmental conditions. It extends the warning lead time from the traditional 5-10 minutes to 30-60 minutes, significantly improving the accuracy, timeliness, and reliability of cable-stayed bridge safety monitoring and providing important technical support for bridge safety management.

[0246] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A cable-stayed bridge cable force imbalance vortex vibration early warning system based on an intelligent computing model is characterized by: include: Multi-source data acquisition module, used to collect cable-stayed bridge structure vibration data, cable force data and environmental parameter data; a phase space reconstruction module, communicatively connected to the multi-source data acquisition module, configured to receive the structural vibration data and construct a phase space representation of the vibration signal based on the structural vibration data; a vortex vibration modal decoupling module, communicating with the phase space reconstruction module, for performing multi-order frequency band decomposition on the phase space representation and extracting vortex vibration modal features of each frequency band; a dynamic threshold setting module, communicatively connected to the vortex vibration modal decoupling module and the multi-source data acquisition module, for setting dynamic discrimination thresholds for vortex vibration modes in each frequency band based on the environmental parameter data and the cable force data, wherein the dynamic threshold setting module generates a dynamic discrimination threshold that is adjusted in real time with the environment and structural state by determining a reference threshold, constructing a threshold mapping function in a multi-dimensional parameter space, applying dynamic compensation for environmental parameters, performing dynamic compensation for structural state, and updating the threshold in real time; a nonlinear prediction module, in communication with the vortex-vibration modal decoupling module, for predicting the future vibration state of the cable-stayed cable based on the vortex-vibration modal characteristics, wherein the nonlinear prediction module predicts the vibration state of the cable-stayed cable for the next 5 to 60 minutes by constructing a phase space state vector, establishing and selecting a local prediction model, implementing multi-time scale prediction, fusing multi-modal prediction results, and correcting the vibration state based on wind field prediction; The early warning decision module is in communication with the nonlinear prediction module and the dynamic threshold setting module, and is used to compare the future vibration state with the dynamic discrimination threshold, determine the early warning level and generate early warning information.

2. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent calculation model according to claim 1 is characterized in that: The multi-source data acquisition module includes: Structural vibration collection unit, used to collect structural vibration data through acceleration sensors installed at key positions of the cable-stayed bridge; The cable force monitoring unit is used to collect cable force change data through optical fiber strain gauges installed in the anchorage area and the middle of the inclined cable; Environmental parameter acquisition unit, used to collect wind speed, wind direction, temperature and humidity environmental parameter data through sensors installed on the top of the bridge tower and the main beam; The data preprocessing unit is connected to the structural vibration acquisition unit, the cable force monitoring unit and the environmental parameter acquisition unit, and is used for filtering, denoising and performing outlier processing on the collected raw data.

3. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent calculation model according to claim 1 is characterized in that: The phase space reconstruction module includes: A time delay parameter determination unit, configured to calculate an optimal time delay parameter by using a mutual information function and determine an optimal embedding dimension by using a pseudo nearest neighbor method; a phase space mapping unit, connected to the time delay parameter determination unit, for mapping the one-dimensional time series into a multidimensional phase space based on the optimal time delay parameter and the optimal embedding dimension; The multi-scale decomposition unit is connected to the phase space mapping unit and is used to decompose the trajectory of the multi-dimensional phase space according to different time scales.

4. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent calculation model according to claim 1 is characterized in that: The vortex vibration mode decoupling module includes: Phase space clustering unit, used to perform density clustering analysis on the reconstructed phase space trajectory and identify the natural clustering structure in the phase space; A frequency feature mapping unit, connected to the phase space clustering unit, for mapping the clustering structure in the phase space to the frequency domain to determine the frequency band division boundary; The modal separation unit is connected to the frequency feature mapping unit and is used to decompose the vibration signal into modes of four frequency bands: a low frequency band of 2 to 3 Hz, a medium frequency band of 3 to 5 Hz, a high frequency band of 5 to 8 Hz, and an ultra-high frequency band of 8 to 20 Hz based on the frequency band division boundaries.

5. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on intelligent calculation model according to claim 1 is characterized in that: The dynamic threshold setting module includes: Historical data statistics unit, used to analyze the statistical distribution of vibration characteristics in various frequency bands under different cable tension states and environmental conditions; a multidimensional threshold construction unit, connected to the historical data statistics unit, for constructing a multidimensional threshold hypersurface describing the decision boundary in the parameter space; an environmental compensation unit, connected to the multi-dimensional threshold value construction unit, for adjusting the reference threshold value according to real-time wind speed, wind direction and temperature environmental parameters; The structural state compensation unit is connected to the environmental compensation unit and is used to further adjust the threshold according to the structural state of the bridge and the traffic load to generate a final dynamic discrimination threshold.

6. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on intelligent calculation model according to claim 1 is characterized in that: The nonlinear prediction module includes: A local prediction model unit is used to divide the phase space into multiple local regions and establish a linearized prediction model for each region; A multi-time scale prediction unit, connected to the local prediction model unit, for achieving short-term high-precision prediction and medium- to long-term trend prediction; a critical state identification unit, connected to the multi-time scale prediction unit, for detecting bifurcation points and system instability states that may appear in the predicted trajectory; The wind-vortex-vibration mapping unit is connected to the critical state identification unit and is used to establish a mapping relationship between the wind field state and the vortex-vibration response and calculate the critical wind value that triggers vortex-vibration instability.

7. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent calculation model according to claim 6 is characterized in that: The wind-vortex mapping unit is implemented by a generative adversarial network, and includes: A generation network is used to generate a simulated wind field sequence that meets specific statistical characteristics; The discriminant network is used to evaluate the similarity between the generated wind field sequence and the real wind field sequence; Mapping network, used to establish the mapping relationship between wind field state and vortex vibration response; The generative network and the discriminative network are continuously optimized through adversarial training to improve the accuracy and generalization ability of the mapping network.

8. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on intelligent computing model according to claim 1 is characterized in that: The early warning decision module includes: Multi-level warning threshold unit, used to set warning threshold systems corresponding to different risk levels; an early warning level determination unit, connected to the multi-level early warning threshold unit, for determining the early warning level corresponding to the current state; a traffic load interference elimination unit, connected to the warning level determination unit, for identifying and isolating vibration interference caused by traffic load; The warning information distribution unit is connected to the traffic load interference elimination unit and is used to generate warning information according to the warning level and send the warning information to management personnel and related systems.

9. The cable-stayed bridge cable force imbalance vortex vibration early warning system based on intelligent computing model according to claim 8 is characterized in that: The traffic load interference elimination unit includes: Traffic characteristic identification subunit, used to identify the vibration characteristics caused by traffic loads by analyzing vibration signals and traffic monitoring data on the bridge; a signal separation subunit, connected to the traffic feature identification subunit, for separating the mixed vibration signal into traffic load vibration and wind-induced vibration; The coupling effect analysis subunit is connected to the signal separation subunit and is used to analyze the coupling amplification effect of traffic load and wind load.

10. The early warning method of the cable-stayed bridge cable force imbalance vortex vibration early warning system based on the intelligent computing model according to any one of claims 1 to 9 is characterized in that: The following steps are involved: Collect vibration data, cable force data and environmental parameter data of cable-stayed bridge structures; Based on the structural vibration data, constructing a phase space representation of the vibration signal by determining an optimal time delay parameter and an optimal embedding dimension; Performing cluster analysis on the phase space representation and mapping it to the frequency domain to decompose the vibration signal into vortex vibration modes of multiple frequency bands; Based on the environmental parameter data and the cable force data, combined with a historical statistical model, a multi-dimensional threshold hypersurface is established, and a dynamic discrimination threshold is generated according to the real-time environment and structural state adjustment; Use the local linearization prediction model to make short-term high-precision predictions and medium- and long-term trend predictions of vortex vibration modes in each frequency band; Detect bifurcation points and system instability in the predicted trajectory, and calculate the critical wind value that triggers vortex instability; Compare the predicted future vibration state with the dynamic discrimination threshold to determine the warning level; Identify and eliminate vibration interference caused by traffic loads and generate final warning information; Distribute warning information to managers and related systems according to the warning level.

Images