Typhoon safety evaluation system and method for ocean engineering structure

Abstract

The invention belongs to the technical field of intelligent monitoring, and discloses a typhoon safety evaluation system and method for an ocean engineering structure. The method comprises the steps of collecting original data, preprocessing the original data to obtain standard data, and performing time-frequency analysis on the standard data to obtain time-frequency characteristics; performing analysis based on the time-frequency characteristics to obtain modal parameters; analyzing according to the modal parameters to obtain a non-linear state judgment result; carrying out damage positioning according to a non-linear state judgment result; carrying out residual safety life prediction according to a damage positioning result; the safety guarantee level of the ocean engineering structure in the typhoon period is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and more specifically, to a typhoon safety assessment system and method for marine engineering structures. Background Technology

[0002] Marine engineering structures face the severe tests of extreme marine loads such as typhoons during their service life. The strong non-stationary excitation caused by typhoons can easily lead to sudden jumps in the structural vibration modal parameters. Such parameter jumps are often the core precursors to the initiation and expansion of latent structural damage or the entry into a critical state of instability. However, traditional structural health monitoring technologies mostly use fixed thresholds to determine parameter anomalies, rely on single data types to analyze structural states, and conduct analysis based on the assumption of stationary signals. These technologies are difficult to adapt to the complex characteristics of time-varying wind speeds, multimodal coupling, and strong noise interference in typhoon scenarios. As a result, the identification accuracy of modal parameter jumps is insufficient, the location of latent damage is ambiguous, and it is impossible to effectively correlate damage with remaining safe life. This fails to meet the safety management requirements of early warning, accurate location, and precise assessment of structures during typhoons.

[0003] In view of this, the present invention proposes a typhoon safety assessment system and method for marine engineering structures to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a typhoon safety assessment method for marine engineering structures, comprising: Collect raw data, preprocess the raw data to obtain standard data, perform time-frequency analysis on the standard data to obtain time-frequency characteristics; Modal parameters are obtained by analyzing time-frequency characteristics; nonlinear state determination results are obtained by analyzing the modal parameters. Damage localization is performed based on the nonlinear state determination results; Remaining safe life is predicted based on the damage location results.

[0005] Furthermore, the time-frequency features include n IMF components, an instantaneous frequency sequence, and an instantaneous amplitude sequence; Methods for obtaining time-frequency features include: Calculate the mean and standard deviation of all raw data, remove outliers using the three-standard-deviation criterion, and use linear interpolation to fill in missing data to obtain standard data; Empirical mode decomposition (EMD) is performed on the standard acceleration. By subtracting the local mean envelope, the standard acceleration is decomposed into n intrinsic mode functions (IMFs), namely n IMF components and 1 residual term. Perform a Hilbert transform on each IMF component, and construct an analytic signal based on the Hilbert transform results; extract the instantaneous amplitude and instantaneous phase from the analytic signal, and obtain the instantaneous frequency based on the instantaneous phase; Iterate through all IMF components to obtain the instantaneous frequency sequence and instantaneous amplitude sequence.

[0006] Furthermore, the modal parameters include the modal frequencies, mode shapes, and damping ratio sequences of each order; Methods for obtaining the frequencies of each modal order include: Construct the Hankel matrix using standard acceleration as input; Singular value decomposition is performed on the Hankel matrix to extract the modal frequencies of each order.

[0007] Furthermore, methods for obtaining mode shapes of various orders include: Calculate the mode shapes of each order by combining the instantaneous amplitude values ​​of p measuring points; For each modal order, extract its corresponding instantaneous frequency sequence; The Pearson correlation coefficient is calculated using covariance and standard deviation. Based on the Pearson correlation coefficient, the IMF component that matches the instantaneous frequency sequence is found, and the instantaneous amplitude sequence of all p measurement points under the matched IMF component is extracted. The instantaneous amplitude sequence of each measuring point in the nth mode is smoothed by time window, and outliers are removed based on the three-times standard deviation principle to obtain the denoised amplitude. The time period during the typhoon when the coefficient of variation of the mode frequency is lower than the variation threshold is selected, and the mean of the denoised amplitude in the corresponding time period is calculated as the stable amplitude of the measuring point in the nth mode. Obtain the maximum amplitude value among the amplitude vectors of p measurement points in the nth mode, calculate the ratio of the stable amplitude value to the maximum amplitude value of the measurement points in the nth mode, obtain the mode shape components, obtain the mode shape components of p measurement points, and obtain the mode shape of the nth mode.

[0008] Furthermore, methods for obtaining the damping ratio sequence include: Plot the power spectral density curve of the IMF component, obtain the peak frequency, and then obtain the two frequencies when the power spectral density is half of the peak power; The damping ratio is obtained by calculating the ratio of the difference between the two frequencies to twice the peak frequency; the damping ratio sequence is obtained by iterating through all IMF components, where the difference between the two frequencies is the value of the high-frequency side frequency minus the low-frequency side frequency.

[0009] Furthermore, methods for obtaining nonlinear state determination results include: The modal frequency is compared with the frequency threshold range, and the damping ratio is compared with the damping ratio threshold range. If the modal frequency is not within the frequency threshold range, or the damping ratio is not within the damping ratio threshold range, a parameter abrupt jump is determined, and the jump time and jump amount are recorded. The frequency threshold range is calculated based on the normal modal frequency, wind speed influence coefficient, real-time wind speed, and maximum wind speed threshold obtained from historical data statistics. The damping ratio threshold range is calculated based on the normal modal damping ratio, damping ratio wind speed coefficient, real-time wind speed, and maximum wind speed threshold obtained from historical data statistics. For the standard acceleration K seconds before and after the abrupt change, the Takens embedding theorem is used to reconstruct the phase space and obtain the phase space vector. The neighboring vectors of each phase space vector are found in the phase space, and the spatial distance between the phase space vector and the corresponding neighboring vector at the preset time is calculated. The logarithm of the spatial distance is fitted with a linear relationship with time, and the slope of the linear relationship is calculated to obtain the maximum Lyapunov exponent. When the maximum Lyapunov exponent is greater than 0, the structure is determined to have entered a chaotic state; otherwise, it is determined to be in a stable vibration state.

[0010] Furthermore, methods for damage localization based on nonlinear state determination results include: Extract the modal shape changes at the jump moment to obtain the modal shape change sequence; Calculate the mean standard strain at each monitoring point, mark the monitoring points that exceed the strain threshold, and obtain the strain anomaly sequence at the jump time; For underwater images of ROVs, template matching was used to compare them with baseline images before the typhoon to identify areas with abnormal appearances. The correlation between the mode shape variation sequence and the strain anomaly sequence was calculated using the grey relational analysis method. Regions with a correlation degree higher than the correlation degree threshold are identified as highly correlated regions. By comparing overlapping locations with abnormal appearance regions, the location of the damage can be obtained. The additional stress in the damaged area is obtained by calculating the product of the strain anomaly and the elastic modulus of the structural material. The degree of damage is obtained by comparing the additional stress with the yield strength of the material.

[0011] Furthermore, methods for obtaining areas of appearance abnormality include: Align the ROV underwater image with the reference image; using the reference image as the complete template and the aligned ROV underwater image as the search image, use a sliding window to match according to the sliding step size. For each sliding window position, calculate the normalized cross-correlation coefficient between the search image and the complete template. Traverse all sliding window positions to generate a similarity matrix, with each element corresponding to the matrix element value of a window. The mean and standard deviation of the matrix elements in the similarity matrix are calculated. An anomaly threshold is set based on the mean and standard deviation of the similarity matrix. When the matrix element value is less than the anomaly threshold, the corresponding sliding window is determined to be an anomaly window. The anomaly window is binarized to obtain a binarized anomaly map. A conversion factor is calculated based on the shooting distance, pixel size, focal length, and unit conversion factor. The pixel coordinates in the binarized anomaly map are mapped to the physical coordinates of the structure based on the conversion factor to obtain the physical location of the anomaly region.

[0012] Furthermore, methods for predicting remaining safe life based on damage location results include: Stress data during the typhoon were calculated based on standard strain signals. The number of cycles under different stress levels was counted using the rainflow counting method. The stress data were then converted into an average stress spectrum of stress amplitude, and the number of cycles corresponding to the i-th stress amplitude was counted. Calculate the ratio of the number of cycles corresponding to the g-th stress amplitude to the fatigue life corresponding to the g-th stress amplitude, and calculate the ratio corresponding to all stress levels to obtain fatigue damage; The structural resistance is calculated based on the initial resistance, corrosion rate, and cumulative fatigue damage at time t. The time-varying reliability index is calculated based on the mean resistance, standard deviation of resistance, mean load effect, and standard deviation of load. When the time-varying reliability index is not lower than the first threshold, it is considered safe. A value not lower than the second threshold and less than the first threshold is considered moderately reliable. If the value is less than the second threshold, it is considered unreliable; A reliability index decay model is established based on historical data of the reliability index. The target time and the current time are calculated when the reliability index equals the target reliability index and the time-varying reliability index, respectively. The remaining lifetime is obtained by calculating the difference between the target time and the current time.

[0013] A typhoon safety assessment system for marine engineering structures, implementing the aforementioned typhoon safety assessment method for marine engineering structures, including: Data Analysis Module: Collects raw data, preprocesses the raw data to obtain standard data, performs time-frequency analysis on the standard data, and obtains time-frequency characteristics; Feature analysis module: Analyzes time-frequency features to obtain modal parameters; analyzes the modal parameters to obtain nonlinear state determination results; Damage localization module: performs damage localization based on the nonlinear state determination results; Life Prediction Module: Predicts remaining safe life based on damage location results.

[0014] The technical effects and advantages of the typhoon safety assessment system and method for marine engineering structures of this invention are as follows: This invention effectively solves the core challenge of monitoring marine engineering structures during typhoons. By constructing a time-varying threshold range of modal parameters adapted to dynamic wind speed changes, and combining Takens phase space reconstruction and maximum Lyapunov exponent analysis, it not only achieves accurate identification of modal frequency and damping ratio jumps, but also quantifies the nonlinear state of the structure, accurately determines the critical state of chaotic instability, and provides early warning. By fusing multi-source data such as mode shape changes, strain anomalies, and abnormal appearance of ROV underwater images, it can accurately locate hidden damage and classify the degree of damage as minor, moderate, and severe. By statistically analyzing the stress cycle characteristics during typhoons using the rainflow counting method, and coupling fatigue damage and marine corrosion effects to construct a time-varying resistance model and a reliability index decay model, it can dynamically predict the remaining safe life of the structure and quantify the reliability state level. Ultimately, it forms a closed-loop control capability of early warning of parameter mutations, accurate location of hidden damage, and precise assessment of safe life, significantly improving the safety assurance level of marine engineering structures during typhoons. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the typhoon safety assessment method for marine engineering structures according to the present invention; Figure 2 This is a schematic diagram of the method for obtaining nonlinear state determination results according to the present invention; Figure 3 This is a schematic diagram of the method for damage localization based on nonlinear state determination results according to the present invention. Figure 4 This is a schematic diagram of the typhoon safety assessment system for marine engineering structures according to the present invention. Detailed Implementation

[0016] 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.

[0017] Example 1 Please see Figure 1 As shown, this embodiment provides a typhoon safety assessment method for marine engineering structures, including: Raw data was collected, including structural response data and environmental load data. Structural response data included triaxial acceleration, strain, and tilt data; environmental load data included wind speed, wind direction, wave height, wave period, and wave direction. Underwater optical images of the ROV were also acquired. A sensor network deployed at key structural parts of the aquaculture platform collected structural response and environmental data during and after the typhoon. Structural response monitoring sensors, including fiber optic strain sensors, acceleration sensors, and tilt sensors, monitored the dynamic strain and static deformation, structural vibration characteristics, and overall tilt attitude of key components such as the platform trusses and legs. Environmental load monitoring sensors, including anemometers and wave radar, recorded the actual wind and wave loads during the typhoon, providing input for subsequent load-response correlation analysis. Furthermore, an underwater robot equipped with acoustic detection equipment and an optical camera automatically or remotely scanned the underwater structure after the typhoon and detected corrosion and biofouling, obtaining underwater optical images of the ROV.

[0018] The raw data is preprocessed to obtain standard data, and time-frequency analysis is performed on the standard data to obtain time-frequency characteristics; based on the time-frequency characteristics, modal parameters are obtained. Time-frequency characteristics include IMF components, instantaneous frequency sequences, and instantaneous amplitude sequences; Methods for obtaining time-frequency features include: The mean and standard deviation of all raw data are calculated, and outliers are removed using the three-standard-deviation criterion. That is, raw data that are lower than the difference between the mean and three standard deviations or higher than the sum of the mean and three standard deviations are removed. The missing data are then filled in using linear interpolation to obtain standard data. The three-standard-deviation criterion eliminates interference such as sensor noise and instantaneous impacts, providing a highly reliable data source for subsequent time-frequency feature extraction and avoiding misjudging outlier data as modal parameter jumps. Empirical Mode Decomposition (EMD) is performed on standard acceleration. By subtracting the local mean envelope, the standard acceleration is decomposed into n intrinsic mode functions (IMFs), i.e., n IMFs and 1 residual term. The condition that the number of zero-crossing points and extreme points of each IMF is equal to or differs by 1 ensures that each IMF is a stationary signal. The IMF components obtained by EMD accurately lock the independent vibration signals of each single-order mode, solving the feature confusion problem caused by multi-modal coupling under typhoon excitation, and providing an isolated signal carrier for tracking the changes of each order of modal parameters. EMD is used to adapt to the non-stationary excitation scenario of typhoons, breaking the dependence of traditional Fourier analysis on stationary signals. By filtering the local mean envelope, the non-stationary vibration signal is decomposed into stationary IMFs, laying the foundation for time-varying analysis of modal parameters. Perform a Hilbert transform on each IMF, and construct an analytic signal based on the Hilbert transform results; as shown in the example. Hilbert transform of IMFs ,in, For the first One IMF; Integration time; analyzing the signal ,in, The imaginary unit; For the first One IMF in Amplitude at any given moment; It is a mathematical constant; It is the instantaneous phase; combined with Hilbert transform, it realizes instantaneous tracking of modal parameters, breaking through the bottleneck of traditional steady-state analysis that can only obtain average modal parameters. It can accurately capture the jump changes in modal frequency and realize early warning before macroscopic damage appears.

[0019] The instantaneous amplitude and instantaneous phase are extracted from the analytical signal, and the instantaneous frequency is obtained from the instantaneous phase; as in the example... Instantaneous frequency of each IMF ; By traversing all IMF components, instantaneous frequency and instantaneous amplitude sequences are obtained. The instantaneous frequencies extracted by Hilbert transform directly correspond to the time-varying state of the modal parameters, and their jump fluctuations are the core characteristic of latent damage. The instantaneous amplitude sequence is correlated with vibration intensity and can help determine whether parameter jumps are approaching the critical threshold for instability.

[0020] Modal parameters include the frequency of each modal, the mode shape of each modal, and the damping ratio sequence; Methods for obtaining modal parameters include: Construct the Hankel matrix using standard acceleration as input; for example... ,in, for The amount of acceleration data at any given moment; The delay order is selected based on experience and is generally greater than the highest modal order of the structure. This refers to the data length, which is generally less than half of the total data length. Singular value decomposition is performed on the Hankel matrix to extract the modal frequencies of each order. Based on standard acceleration, the vibration time-domain correlation is preserved by reasonably setting the delay order and data length, providing a structured data carrier for the extraction of modal information under non-stationary excitation of typhoons, and avoiding information fragmentation caused by time-varying excitation. Methods for extracting modal frequencies include: Perform singular value decomposition on the Hankel matrix to obtain the left singular vector matrix, the singular value diagonal matrix, and the right singular vector matrix; for example... ,in, It is a left singular vector matrix. The column vectors are matrices eigenvectors; For a singular value diagonal matrix, the elements on the main diagonal are... It is a singular value and satisfies , For matrix The rank of the rank is 0, and all off-diagonal elements are 0. It is a right singular vector matrix, and the column vectors are matrices. eigenvectors; It is the transpose of the vector; The left singular vector matrix and the singular value diagonal matrix are partitioned according to their modal orders. The left singular vector matrix is ​​then divided into orthogonal bases for the signal subspace and the noise subspace, as follows: , ,in, The first n columns form an orthogonal basis for the signal subspace; The remaining columns form an orthogonal basis for the noise subspace; It is a diagonal matrix formed by the first n largest singular values; The diagonal matrix is ​​formed by the remaining singular values; the signal and noise are separated by singular value decomposition, and the effective signal components corresponding to each mode are accurately located, which solves the problem of mode frequency extraction distortion under typhoon background noise interference and ensures the reliability of parameter jump recognition. The orthogonal basis of the signal subspace is further divided into two submatrices. and The system matrix is ​​obtained by combining the shift invariance of the subspace with two submatrices; for example... , ,in, For the system matrix, ; Solve for the eigenvalues ​​of the system matrix: Decompose the system matrix into an orthogonal matrix. and an upper triangular matrix The product, based on the upper triangular matrix Orthogonal matrix Construct a new matrix by multiplying the product of the components; repeatedly decompose the system matrix and construct a new matrix until the new matrix converges. Extract the diagonal elements of the converged upper triangular matrix to obtain the eigenvalues ​​of the system matrix.

[0021] The imaginary part of the eigenvalues ​​is extracted to obtain the modal angular frequency, and the modal angular frequency is converted into the modal frequency.

[0022] By employing Hankel matrix singular value decomposition combined with subspace shift invariance to extract frequencies, the limitations of traditional steady-state analysis are overcome. Through time-domain correlation preservation and signal-noise separation, accurate decoupling of various modes under non-stationary typhoon scenarios is achieved, and frequency jumps can be effectively identified even in multi-modal coupling. By combining the instantaneous amplitude values ​​of p measurement points, calculate the mode shapes of each order; the measurement points are the locations of all sensors during data acquisition. For each modal order, extract its corresponding instantaneous frequency sequence; The Pearson correlation coefficient is calculated using covariance and standard deviation. Based on the Pearson correlation coefficient, the IMF component that matches the instantaneous frequency sequence is found, and the instantaneous amplitude sequence of all p measurement points under the matched IMF component is extracted. The instantaneous amplitude sequence of each measuring point in the nth mode is smoothed by time window, and outliers are removed based on the three-standard-deviation principle to obtain the denoised amplitude. The time period during the typhoon when the coefficient of variation of the mode frequency is lower than the variation threshold is selected, and the mean of the denoised amplitude in the corresponding time period is calculated as the stable amplitude of measuring point k in the nth mode. The variation threshold can be obtained by statistical analysis of historical data. Obtain the maximum amplitude value among the amplitude vectors of p measurement points in the nth mode, calculate the ratio of the stable amplitude value to the maximum amplitude value at the kth measurement point in the nth mode, obtain the mode shape components, obtain the mode shape components of p measurement points, and obtain the mode shape of the nth mode.

[0023] The vibration mode calculation integrates frequency stability screening and amplitude normalization. By using the coefficient of variation to screen stable periods during typhoons, amplitude distortion during parameter jump periods is avoided. Combined with multi-point ratio quantification of vibration modes, the accurate location of hidden damage is achieved. Plot the power spectral density curve of the IMF component to obtain the peak frequency, and then obtain two frequencies when the power spectral density is half of the peak power. Extract the damping ratio based on the IMF power spectrum half-power bandwidth method, and improve the accuracy of spectral peak identification by taking advantage of the stationarity of the IMF. It can capture the time-varying changes of the damping ratio more accurately than traditional overall signal analysis, and provide a highly reliable indicator for early warning of instability critical state.

[0024] The damping ratio is obtained by calculating the ratio of the difference between two frequencies to twice the peak frequency. A damping ratio sequence is obtained by iterating through all IMF components, where the difference between two frequencies is the difference between the high-frequency side frequency and the low-frequency side frequency. The frequency sequence directly captures abrupt changes; the mode shape is quantified by the amplitude ratio at measurement points, and its distortion characteristics pinpoint the damage location; the damping ratio sequence reflects the structure's energy dissipation capacity, and abrupt signal changes are the core precursor to instability and critical states.

[0025] The modal parameters are analyzed to obtain the nonlinear state determination results, and the damage is located based on the nonlinear state determination results; the remaining safe life is predicted based on the damage location results. Reference Figure 2 Methods for obtaining nonlinear state determination results include: The modal frequency is compared with the frequency threshold range, and the damping ratio is compared with the damping ratio threshold range. If the modal frequency or damping ratio is not within the frequency threshold range, a parameter jump is determined, and the jump time and jump amount are recorded. The frequency threshold range is calculated based on the normal modal frequency, wind speed influence coefficient, real-time wind speed, and maximum wind speed threshold obtained from historical data statistics. For example, the lower limit of the frequency threshold range... upper limit of frequency threshold range , These are the normal modal frequencies obtained based on historical data statistics; The wind speed influence coefficient is determined by the structure type. Real-time wind speed; The maximum wind speed threshold is the maximum wind speed the structure can withstand. The damping ratio threshold range is calculated based on historical data, including the normal mode damping ratio, damping ratio wind speed coefficient, real-time wind speed, and the maximum wind speed threshold. For example, the lower limit of the damping ratio threshold range... Upper limit of damping ratio threshold range ; The normal modal damping ratio is obtained based on historical data statistics; The damping ratio wind speed coefficient is greater than The optimal threshold can be obtained through natural heuristic optimization algorithms; combined with dynamic adjustment of wind speed threshold, it solves the problem of misjudgment of parameter changes caused by typhoon wind speed interference of traditional fixed threshold, and provides a benchmark for accurate identification of modal jumps; by constructing time-varying threshold range through wind speed influence coefficient, it breaks through the limitation of insufficient sensitivity of traditional fixed threshold to typhoon wind speed, realizes accurate identification of modal parameter jumps, and avoids misjudging wind speed fluctuations as damage. For the standard acceleration K seconds before and after the abrupt change, the Takens embedding theorem is used to reconstruct the phase space, obtaining phase space vectors. The nearest neighbor vectors of each phase space vector are found in the phase space, and the spatial distance between the phase space vector and its corresponding nearest neighbor vector at a preset time is calculated. A linear relationship between the logarithm of the spatial distance and time is fitted, and the slope of the linear relationship is calculated to obtain the maximum Lyapunov exponent. When the maximum Lyapunov exponent is greater than 0, the structure is determined to have entered a chaotic state; otherwise, it is determined to be in a stable oscillation state. This quantifies the nonlinear state from the perspective of phase space evolution. The result of a value greater than 0 directly indicates that the structure has entered chaos, i.e., the critical state of instability, and is the core quantitative indicator of the risk of instability after parameter jumps.

[0026] Methods for reconstructing phase space using Takens' embedding theorem include: Calculate the autocorrelation function of the standard acceleration during the analysis period, plot the curve of the autocorrelation function with respect to the delay time, and take the delay time corresponding to the first drop of the autocorrelation function to half of the value of the autocorrelation function when the delay time is 0 as the optimal delay time; Determining the optimal embedding dimension using the spurious nearest neighbor method: Initialize the embedding dimension and the maximum candidate embedding dimension; For the current embedding dimension, construct temporary phase space vectors. For each temporary phase space vector, find the nearest neighbor vector with the smallest Euclidean distance among all temporary phase space vectors, and calculate the nearest neighbor distance between the temporary phase space vector and the nearest neighbor vector. Increase the embedding dimension by 1, construct a new temporary phase space vector, and calculate the new nearest neighbor distance; If the ratio of the neighbor distance of the next embedding dimension to the neighbor distance of the current embedding dimension is greater than the distance threshold, then the temporary phase space vector of the current embedding dimension and its corresponding neighbor vector are determined to be false neighbors; where the distance threshold can be obtained through historical data statistics. The ratio of false neighbor points to the maximum candidate embedding dimension is calculated to obtain the false neighbor point ratio. Increment the embedding dimension and calculate the corresponding proportion of false neighbors. When the change in the proportion of false neighbors is lower than the change threshold, the corresponding embedding dimension is set as the optimal embedding dimension. The change threshold can be obtained through historical data statistics.

[0027] Each phase space vector contains the number of data points with the optimal embedding dimension, and the delay between adjacent data points is the optimal delay time. A phase space vector is constructed, and each effective time index is used to construct a phase space vector. All phase space vectors are arranged in chronological order to form a reconstructed phase space.

[0028] Takens phase space reconstruction combined with the maximum Lyapunov exponent directly correlates modal parameter jumps with chaotic and stable states, transforming abstract nonlinear characteristics into quantifiable instability early warning indicators, thus solving the problem of the difficulty in defining the critical state of instability. Reference Figure 3 Methods for damage localization based on nonlinear state determination results include: Extract the modal shape changes at the jump moment to obtain the modal shape change sequence; Calculate the mean standard strain at each monitoring point, mark the monitoring points that exceed the strain threshold, and obtain the strain anomaly sequence at the jump time; where the strain threshold can be obtained through statistical analysis of historical data; For underwater images of ROVs, template matching was used to compare them with baseline images before the typhoon to identify areas with abnormal appearances. Methods for obtaining areas of appearance abnormality include: Align the ROV underwater image with the baseline image; alignment can be achieved by extracting SIFT feature points by constructing a Gaussian difference pyramid. Using the baseline image as the complete template and the aligned ROV underwater image as the search image, a sliding window is used for matching according to the sliding step size. For each sliding window position, the normalized cross-correlation coefficient between the search image and the complete template is calculated. All sliding window positions are traversed to generate a similarity matrix, with each element corresponding to the matrix element value of a window. The mean and standard deviation of the matrix elements in the similarity matrix are calculated. An anomaly threshold is set based on the mean and standard deviation of the similarity matrix. When the matrix element value is less than the anomaly threshold, the corresponding sliding window is determined to be an anomaly window. The anomaly window is binarized to obtain a binarized anomaly map. A conversion factor is calculated based on the shooting distance, pixel size, focal length, and unit conversion factor. The pixel coordinates in the binarized anomaly map are mapped to the physical coordinates of the structure according to the conversion factor to obtain the physical location of the anomaly region. The shooting distance can be directly measured by a range sensor. The anomaly threshold can be obtained by statistical analysis of historical data.

[0029] The correlation between the mode shape variation sequence and the strain anomaly sequence was calculated using the grey relational analysis method. Areas with a correlation degree higher than the correlation degree threshold are identified as highly correlated areas. By comparing the overlapping positions with areas of appearance abnormality, the location of the damage can be obtained. The correlation degree threshold can be obtained through historical data statistics. The spatial correlation area of ​​the damage is located by the mode shape change, and the damage intensity is quantified by strain anomaly, providing visual evidence for appearance abnormalities. The combination of these three methods solves the pain point that it is difficult to identify hidden damage with single data.

[0030] The additional stress in the damaged region is obtained by calculating the product of the strain anomaly and the elastic modulus of the structural material. The degree of damage is then determined by comparing the additional stress with the material's yield strength. When the ratio of the additional stress to the material's yield strength is lower than the first damage threshold, it is considered minor damage; when it is not lower than the first damage threshold but lower than the second damage threshold, it is considered moderate damage; and when it is not lower than the second damage threshold, it is considered severe damage. The first and second damage thresholds can be obtained through statistical analysis of historical data. By using gray relational analysis to correlate mode shapes and strain data, and superimposing the physical coordinate mapping of abnormal appearances in ROV underwater images, three-dimensional localization of modal characteristics, mechanical response, and visual evidence is achieved, significantly improving the accuracy of locating hidden damage.

[0031] Methods for predicting remaining safe life based on damage location results include: Stress data during the typhoon were calculated based on standard strain signals. The number of cycles under different stress levels was counted using the rainflow counting method. The stress data were then converted into a stress amplitude-average stress spectrum, and the number of cycles corresponding to the i-th stress amplitude was counted. Iterate through each time index. If the current point is greater than the two adjacent points, it is determined to be a peak point. If the current point is less than the two adjacent points, it is determined to be a valley point. Record the information of all peaks and valleys in chronological order, and take a set of adjacent peaks and valleys as an extreme point to form a peak-valley sequence. Starting from each peak and valley point, assume that rainwater flows down the stress time history curve from the peak and valley point to the next adjacent extreme point; Cycle determination: If the stress value of the next extreme point encountered during the rainwater flow exceeds the stress value of the starting point, the peak value of the next extreme point is greater than the peak value of the starting point, and the valley value of the next extreme point is less than the valley value of the starting point, then the flow segment forms a complete cycle; if it does not exceed the stress value, then a half cycle is formed, which needs to be merged with the subsequent cycle.

[0032] Store the peak and valley point sequence into the counting stack in order, and arrange the elements in the stack in chronological order. Loop counting: Take the last three consecutive points from the top of the stack to be counted: starting point A, intermediate point B, and ending point C, ensuring that point A and point C are of the same type, and point B is of the opposite type; calculate the stress difference between point A and point B, and the stress difference between point B and point C. If the stress at point C is greater than the stress at point A, then points A, B, and C form a complete cycle. Calculate half of the stress difference between points A and B as the stress amplitude, and calculate half of the sum of the stresses at points A and B as the average stress. If the stress at point C is not greater than the stress at point A, then points B and C form a half-cycle. Calculate half of the stress difference between points B and C as the stress amplitude, and calculate half of the sum of the stresses at points B and C as the average stress. Keep this half-cycle in the half-cycle buffer, remove B from the stack, add C to the top of the stack, and continue processing the next point.

[0033] Repeat the above loop counting steps until there are fewer than 3 points remaining in the stack. At this point, the remaining points in the stack are merged with the semi-loops in the semi-loop buffer to form a new complete loop. The unmerged semi-loops are recorded as residual loops. Calculate the mean of the maximum and minimum stress amplitudes in all complete cycles to obtain the stress amplitude of a complete cycle; calculate the mean of the maximum and minimum average stresses in all complete cycles to obtain the average stress of a complete cycle; calculate the mean of the stress amplitudes at the end and beginning of all half cycles as the stress amplitude of a half cycle; calculate the mean of the stresses at the end and beginning of all half cycles as the stress of a half cycle.

[0034] The number of stress amplitude intervals is set using an equal interval division method. The width of the stress amplitude interval is set according to the stress amplitude range and the number of stress amplitude intervals. The number of stress amplitude intervals is set based on the SN curve of the structural material, which covers all complete cycles. The range of each stress amplitude interval is obtained based on the width of the stress amplitude interval, the range of the stress amplitude, and the number of stress amplitude intervals.

[0035] Loop count statistics: Iterate through all complete cycles, and for each cycle's stress amplitude, determine its corresponding stress amplitude range. Count the number of cycles for each stress amplitude range; If a half-cycle exists, the cycle number of each half-cycle is recorded as 0.5, and its stress amplitude is assigned to the corresponding interval. The results are accumulated in the corresponding stress amplitude interval to obtain the cycle number corresponding to the g-th stress amplitude.

[0036] Calculate the ratio of the number of cycles corresponding to the g-th stress amplitude to the fatigue life corresponding to the g-th stress amplitude. Calculate the ratio corresponding to all stress levels to obtain fatigue damage. The fatigue life corresponding to the g-th stress amplitude is obtained by querying the stress amplitude and fatigue life curve of the structural material. The structural resistance is calculated based on the initial resistance, corrosion rate, and cumulative fatigue damage at time t; for example... ,in, The initial resistance is determined by the cross-sectional dimensions and material strength; The corrosion rate is determined from marine environmental monitoring data; for It continuously accumulates fatigue damage; it simultaneously integrates marine environmental corrosion and typhoon fatigue damage, breaking through the limitations of traditional single-factor lifetime prediction; and it combines a reliable index decay model to achieve dynamic updates of remaining lifetime, adapting to the time-varying characteristics of damage accumulation after a typhoon.

[0037] Time-varying reliability indices are calculated based on the mean resistance, standard deviation of resistance, mean load effect, and standard deviation of load; such as time-varying reliability indices. ,in, for Mean resistance at any given time; for Standard deviation of resistance over time; for The mean load effect at any given time was obtained from long-term monitoring data. for The standard deviation of load at any time moment is obtained from long-term monitoring data statistics; When the time-varying reliability index is not lower than the first threshold, it is considered safe. A value not lower than the second threshold and less than the first threshold is considered moderately reliable. If the value is less than the second threshold, it is considered unreliable; the first and second thresholds can be obtained through statistical analysis of historical data. A reliability index decay model is established based on historical data of the reliability index. The target time and current time are calculated when the reliability index equals the target reliability index and the time-varying reliability index, respectively. The remaining lifetime is obtained by calculating the difference between the target time and the current time. (Example: Reliability index decay model) ,in, , and The parameters are used for fitting; if the time-varying reliability index is less than the target reliability index, data is re-acquired to calculate the predicted remaining life. Fatigue damage is integrated with stress cycle information from rainflow counting, and the time-varying reliability index is coupled with corrosion and fatigue effects, which is the core quantitative basis for remaining life prediction and directly reflects the structural safety reserve.

[0038] Example 2 Please see Figure 4 As shown, this embodiment provides a typhoon safety assessment system for marine engineering structures, including: Data Analysis Module: Collects raw data, preprocesses the raw data to obtain standard data, performs time-frequency analysis on the standard data, and obtains time-frequency characteristics; Feature analysis module: Analyzes time-frequency features to obtain modal parameters; analyzes the modal parameters to obtain nonlinear state determination results; Damage localization module: performs damage localization based on the nonlinear state determination results; Life Prediction Module: Predicts remaining safe life based on damage location results.

[0039] 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 variations or substitutions that can be easily conceived by those 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 determined by the scope of the claims.

[0040] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A typhoon safety assessment method for marine engineering structures, characterized in that, include: Collect raw data, preprocess the raw data to obtain standard data, perform time-frequency analysis on the standard data to obtain time-frequency characteristics; Modal parameters are obtained by analyzing time-frequency characteristics. Based on the analysis of modal parameters, the nonlinear state determination results are obtained; Damage localization is performed based on the nonlinear state determination results; Remaining safe life is predicted based on the damage location results.

2. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, The time-frequency features include n IMF components, an instantaneous frequency sequence, and an instantaneous amplitude sequence; Methods for obtaining time-frequency features include: Calculate the mean and standard deviation of all raw data, remove outliers using the three-standard-deviation criterion, and use linear interpolation to fill in missing data to obtain standard data; Empirical mode decomposition (EMD) is performed on the standard acceleration. By subtracting the local mean envelope, the standard acceleration is decomposed into n intrinsic mode functions (IMFs), namely n IMF components and 1 residual term. Perform a Hilbert transform on each IMF component, and construct an analytic signal based on the Hilbert transform results; extract the instantaneous amplitude and instantaneous phase from the analytic signal, and obtain the instantaneous frequency based on the instantaneous phase; Iterate through all IMF components to obtain the instantaneous frequency sequence and instantaneous amplitude sequence.

3. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, The modal parameters include the modal frequencies, mode shapes, and damping ratio sequences of each order; Methods for obtaining the frequencies of each modal order include: Construct the Hankel matrix using standard acceleration as input; Singular value decomposition is performed on the Hankel matrix to extract the modal frequencies of each order.

4. The typhoon safety assessment method for marine engineering structures according to claim 3, characterized in that, Methods for obtaining mode shapes of various orders include: Calculate the mode shapes of each order by combining the instantaneous amplitude values ​​of p measuring points; For each modal order, extract its corresponding instantaneous frequency sequence; The Pearson correlation coefficient is calculated using covariance and standard deviation. Based on the Pearson correlation coefficient, the IMF component that matches the instantaneous frequency sequence is found, and the instantaneous amplitude sequence of all p measurement points under the matched IMF component is extracted. The instantaneous amplitude sequence of each measuring point in the nth mode is smoothed by time window, and outliers are removed based on the three-times standard deviation principle to obtain the denoised amplitude. The time period during the typhoon when the coefficient of variation of the mode frequency is lower than the variation threshold is selected, and the mean of the denoised amplitude in the corresponding time period is calculated as the stable amplitude of the measuring point in the nth mode. Obtain the maximum amplitude value among the amplitude vectors of p measurement points in the nth mode, calculate the ratio of the stable amplitude value to the maximum amplitude value of the measurement points in the nth mode, obtain the mode shape components, obtain the mode shape components of p measurement points, and obtain the mode shape of the nth mode.

5. The typhoon safety assessment method for marine engineering structures according to claim 3, characterized in that, Methods for obtaining damping ratio sequences include: Plot the power spectral density curve of the IMF component, obtain the peak frequency, and then obtain the two frequencies when the power spectral density is half of the peak power; The damping ratio is obtained by calculating the ratio of the difference between the two frequencies to twice the peak frequency; the damping ratio sequence is obtained by iterating through all IMF components, where the difference between the two frequencies is the value of the high-frequency side frequency minus the low-frequency side frequency.

6. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, Methods for obtaining nonlinear state determination results include: The modal frequency is compared with the frequency threshold range, and the damping ratio is compared with the damping ratio threshold range. If the modal frequency is not within the frequency threshold range, or the damping ratio is not within the damping ratio threshold range, a parameter abrupt jump is determined, and the jump time and jump amount are recorded. The frequency threshold range is calculated based on the normal modal frequency, wind speed influence coefficient, real-time wind speed, and maximum wind speed threshold obtained from historical data statistics. The damping ratio threshold range is calculated based on the normal modal damping ratio, damping ratio wind speed coefficient, real-time wind speed, and maximum wind speed threshold obtained from historical data statistics. For the standard acceleration K seconds before and after the abrupt change, the Takens embedding theorem is used to reconstruct the phase space and obtain the phase space vector. The neighboring vectors of each phase space vector are found in the phase space, and the spatial distance between the phase space vector and the corresponding neighboring vector at the preset time is calculated. The logarithm of the spatial distance is fitted with a linear relationship with time, and the slope of the linear relationship is calculated to obtain the maximum Lyapunov exponent. When the maximum Lyapunov exponent is greater than 0, the structure is determined to have entered a chaotic state; otherwise, it is determined to be in a stable vibration state.

7. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, Methods for damage localization based on nonlinear state determination results include: Extract the modal shape changes at the jump moment to obtain the modal shape change sequence; Calculate the mean standard strain at each monitoring point, mark the monitoring points that exceed the strain threshold, and obtain the strain anomaly sequence at the jump time; For underwater images of ROVs, template matching was used to compare them with baseline images before the typhoon to identify areas with abnormal appearances. The correlation between the mode shape variation sequence and the strain anomaly sequence was calculated using the grey relational analysis method. Regions with a correlation degree higher than the correlation degree threshold are identified as highly correlated regions. By comparing overlapping locations with abnormal appearance regions, the location of the damage can be obtained. The additional stress in the damaged area is obtained by calculating the product of the strain anomaly and the elastic modulus of the structural material. The degree of damage is obtained by comparing the additional stress with the yield strength of the material.

8. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, Methods for obtaining areas of appearance abnormality include: Align the ROV underwater image with the reference image; using the reference image as the complete template and the aligned ROV underwater image as the search image, use a sliding window to match according to the sliding step size. For each sliding window position, calculate the normalized cross-correlation coefficient between the search image and the complete template. Traverse all sliding window positions to generate a similarity matrix, with each element corresponding to the matrix element value of a window. The mean and standard deviation of the matrix elements in the similarity matrix are calculated. An anomaly threshold is set based on the mean and standard deviation of the similarity matrix. When the matrix element value is less than the anomaly threshold, the corresponding sliding window is determined to be an anomaly window. The anomaly window is binarized to obtain a binarized anomaly map. A conversion factor is calculated based on the shooting distance, pixel size, focal length, and unit conversion factor. The pixel coordinates in the binarized anomaly map are mapped to the physical coordinates of the structure based on the conversion factor to obtain the physical location of the anomaly region.

9. The typhoon safety assessment method for marine engineering structures according to claim 1, characterized in that, Methods for predicting remaining safe life based on damage location results include: Stress data during the typhoon were calculated based on standard strain signals. The number of cycles under different stress levels was counted using the rainflow counting method. The stress data were then converted into an average stress spectrum of stress amplitude, and the number of cycles corresponding to the i-th stress amplitude was counted. Calculate the ratio of the number of cycles corresponding to the g-th stress amplitude to the fatigue life corresponding to the g-th stress amplitude, and calculate the ratio corresponding to all stress levels to obtain fatigue damage; The structural resistance is calculated based on the initial resistance, corrosion rate, and cumulative fatigue damage at time t. The time-varying reliability index is calculated based on the mean resistance, standard deviation of resistance, mean load effect, and standard deviation of load. When the time-varying reliability index is not lower than the first threshold, it is considered safe. A value not lower than the second threshold and less than the first threshold is considered moderately reliable. If the value is less than the second threshold, it is considered unreliable; A reliability index decay model is established based on historical data of the reliability index. The target time and the current time are calculated when the reliability index equals the target reliability index and the time-varying reliability index, respectively. The remaining lifetime is obtained by calculating the difference between the target time and the current time.

10. A typhoon safety assessment system for marine engineering structures, implementing the typhoon safety assessment method for marine engineering structures as described in any one of claims 1-9, characterized in that, include: Data Analysis Module: Collects raw data, preprocesses the raw data to obtain standard data, performs time-frequency analysis on the standard data, and obtains time-frequency characteristics; Feature analysis module: Analyzes time-frequency features to obtain modal parameters; Based on the analysis of modal parameters, the nonlinear state determination results are obtained; Damage localization module: performs damage localization based on the nonlinear state determination results; Life Prediction Module: Predicts remaining safe life based on damage location results.

Images