Floating support mounting structure health monitoring method based on optical fiber sensor

Abstract

The invention provides a floating support installation structure health monitoring method based on an optical fiber sensor, and the method comprises the steps: obtaining the original strain data and original temperature data of a floating support installation region based on the optical fiber sensor, and obtaining the inspection visual data of an underwater floating support installation region; de-noising the original strain data based on time-space domain combination to obtain strain de-noised data, and constructing a temperature-strain transfer model based on the strain de-noised data and the original temperature data; compensating the original temperature data based on a temperature-strain transfer model to obtain temperature compensation data; performing feature extraction and fusion on the temperature compensation data, the strain denoising data and the inspection visual data to obtain a structural damage probability distribution diagram; and analyzing and evaluating the structural damage probability distribution diagram, and determining a structural health monitoring evaluation result. According to the invention, multi-dimensional and all-directional automatic monitoring of the floating support mounting area is realized, the blind area of manual monitoring is compensated, and the monitoring range is greatly expanded.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology, and in particular to a method for monitoring the health of floating installation structures based on fiber optic sensors. Background Technology

[0002] With the development of technology, fiber optic sensors have been widely used in the field of structural health monitoring due to their advantages such as high precision and resistance to electromagnetic interference.

[0003] Floating-mounted structural health refers to a series of monitoring and evaluation activities conducted on structures constructed or installed using the floating-mounting method (such as platforms, bridges, and docks in marine engineering). Traditional monitoring methods rely on direct observation of the structural surface by professional technicians. However, due to limitations in human vision and accessibility, the effectiveness is limited for large structures or hard-to-reach areas, resulting in the detection of only specific types of defects or specific parts of the structure, leading to poor monitoring results. Furthermore, methods using fiber optic sensors for floating-mounted structural health monitoring mainly focus on collecting data on key parameters (such as strain and temperature) using fiber optic grating sensors. Due to the complex marine environment, such as extreme weather or underwater operations, fiber optic signal transmission may be affected, thus reducing the accuracy of health monitoring. Summary of the Invention

[0004] This invention provides a method for health monitoring of floating installation structures based on fiber optic sensors, which enables multi-dimensional and all-round automatic monitoring of the floating installation area, making up for the blind spots of manual monitoring and greatly expanding the monitoring range.

[0005] In a first aspect, the present invention provides a method for health monitoring of a floating installation structure based on fiber optic sensors, comprising:

[0006] The original strain and temperature data of the floating installation area are obtained based on fiber optic sensors, and the inspection visual data of the underwater floating installation area are also obtained.

[0007] The original strain data is denoised based on spatiotemporal domain joint processing to obtain denoised strain data. Based on the denoised strain data and the original temperature data, a temperature-strain transfer model is constructed.

[0008] The original temperature data is compensated based on the temperature-strain transfer model to obtain temperature compensation data.

[0009] Feature extraction and fusion are performed on the temperature compensation data, strain denoising data and the inspection visual data to obtain a structural damage probability distribution map;

[0010] The structural damage probability distribution map is analyzed and evaluated to determine the structural health monitoring and evaluation results.

[0011] Secondly, the present invention also provides a health monitoring system for a floating installation structure based on fiber optic sensors, applied to the health monitoring method for a floating installation structure based on fiber optic sensors as described in the first aspect; the health monitoring system for a floating installation structure based on fiber optic sensors includes:

[0012] The data acquisition module is used to acquire raw strain data and raw temperature data of the floating installation area based on fiber optic sensors, and to acquire inspection visual data of the underwater floating installation area.

[0013] The denoising and construction module denoises the original strain data based on spatiotemporal domain joint processing to obtain denoised strain data, and constructs a temperature-strain transfer model based on the denoised strain data and the original temperature data.

[0014] The temperature compensation module is used to compensate the original temperature data based on the temperature-strain transfer model to obtain temperature compensation data.

[0015] The fusion diagnostic module is used to extract and fuse features from the temperature compensation data, strain denoising data and the inspection visual data to obtain a structural damage probability distribution map.

[0016] The analysis and evaluation module is used to analyze and evaluate the probability distribution map of structural damage and determine the structural health monitoring and evaluation results.

[0017] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby realizing the floating installation structure health monitoring method based on fiber optic sensors as described above.

[0018] Fourthly, the present invention also provides a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements the above-described method for health monitoring of a floating installation structure based on an optical fiber sensor.

[0019] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for health monitoring of a floating installation structure based on a fiber optic sensor.

[0020] The floating installation structure health monitoring method based on fiber optic sensors provided in this invention combines strain and temperature data detected by fiber optic sensors with inspection visual data to achieve multi-dimensional and all-round automatic monitoring of the floating installation area, making up for the blind spots of manual monitoring and greatly expanding the monitoring range. Furthermore, by using spatiotemporal domain joint denoising and temperature-strain field coupling for temperature compensation, the accuracy of strain measurement is improved. This not only effectively suppresses noise interference from wave impacts and vortex-induced vibrations but also reduces the impact of large diurnal temperature differences on data measurement, ensuring that the monitoring data truly reflects the structural stress state. Finally, by combining multiple types of data for structural damage identification, structural damage can be more accurately located and diagnosed, potential safety hazards can be detected in advance, and the accuracy of health monitoring can be further improved. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the health monitoring method for a floating installation structure based on fiber optic sensors provided in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the health monitoring system for a floating installation structure based on an optical fiber sensor provided in an embodiment of the present invention;

[0023] Figure 3 An embodiment diagram of the electronic device provided in this invention;

[0024] Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with the present invention. Detailed Implementation

[0025] 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.

[0026] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0028] In existing technologies, health monitoring of floating structures mainly relies on manual observation and single-sensor data acquisition. Professional technicians inspect the structural surface with the naked eye or handheld devices, which is insufficient to cover hidden areas of large structures, resulting in monitoring blind spots. When using fiber optic grating sensors, factors such as temperature changes and mechanical vibrations in complex underwater environments introduce noise interference, causing deviations in strain measurements. Traditional methods lack effective multi-source data fusion mechanisms; visual information and physical sensor data are processed independently, failing to accurately reflect the spatial distribution characteristics of structural damage.

[0029] See Figure 1 , Figure 1 This is a flowchart illustrating the health monitoring method for a floating support structure based on fiber optic sensors provided by the present invention. In this embodiment, the executing entity of the health monitoring method for a floating support structure based on fiber optic sensors is a health monitoring system. Therefore, the health monitoring method for a floating support structure based on fiber optic sensors includes:

[0030] Step 10: Obtain raw strain data and raw temperature data of the floating installation area based on fiber optic sensors, and obtain inspection visual data of the underwater floating installation area.

[0031] Optionally, the health monitoring system deploys a fiber optic sensor network, including fiber Bragg grating sensors and distributed fiber optic sensors, at key locations on the floating installation structure. Fiber Bragg grating sensors are primarily deployed at connection nodes, critical stress points, and areas with high historical damage rates on the floating installation structure for high-precision local strain monitoring; distributed fiber optic sensors are deployed around the main structure of the floating installation structure to achieve large-scale continuous monitoring. The reflected spectrum or scattered light signals from the fiber optic sensors are acquired by a fiber optic sensor demodulator, and then converted into raw strain and temperature data through signal processing. Simultaneously, underwater robots or fixed underwater camera equipment are deployed to conduct regular inspections of the floating installation area, acquiring high-definition underwater visual images and video data. This visual data includes information such as the surface condition of the floating installation structure, the integrity of connection points, corrosion status, and changes in the surrounding environment.

[0032] Step 20: Denoise the original strain data based on spatiotemporal domain joint processing to obtain denoised strain data, and construct a temperature-strain transfer model based on the denoised strain data and the original temperature data.

[0033] Optionally, the health monitoring system denoises the original strain data by combining time and spatial domains, i.e., time series and spatial location, to obtain more accurate denoised strain data, as described in steps 2011-2014. Then, based on the influence of temperature change on strain measurement in the denoised strain data and the original temperature data, a temperature-strain transfer model is established, as described in steps 2021-2024.

[0034] Step 30: Compensate the original temperature data based on the temperature-strain transfer model to obtain temperature compensation data.

[0035] Optionally, the health monitoring system first determines the probability distribution using Bayesian probability based on the obtained temperature-strain transfer model and the original temperature data, and then obtains the error corresponding to the original temperature data according to the temperature-strain transfer model. Finally, it dynamically determines the temperature correction parameters to achieve dynamic compensation of the temperature data, as described in steps 301-304. This ensures that the strain data accurately reflects the deformation of the structure caused by factors such as stress, improving the accuracy and reliability of the strain data and providing more precise data for subsequent structural health assessments.

[0036] Step 40: Perform feature extraction and fusion on temperature compensation data, strain denoising data and inspection visual data to obtain a structural damage probability distribution map.

[0037] Optionally, the health monitoring system extracts features in the same way based on the obtained temperature compensation data and strain denoising data to obtain corresponding spatio-temporal feature vectors. Then, the collected visual inspection data is subjected to feature extraction and cooling to obtain view feature vectors. Based on the spatio-temporal feature vectors and visual feature vectors, a data fusion algorithm is then used to fuse the features of different types of data. According to the fusion result, combined with the structural mechanics model, the damage probability of each part of the floating structure is calculated, and a structural damage probability distribution map is drawn, as specifically described in steps 401 - 405. By extracting and fusing multi-source data features, the advantages of different types of data are fully utilized. Strain data reflects internal forces, and visual data reflects external damage. More comprehensive structural state information is obtained through fusion. The structural damage probability distribution map presents the damage possibility of each part of the structure in an intuitive and visual way, facilitating the rapid positioning of potential problem areas.

[0038] Step 50: Analyze and evaluate the structural damage probability distribution map to determine the structural health monitoring evaluation result.

[0039] Optionally, the health monitoring system first performs threshold segmentation on the damage probability distribution map to identify high-risk areas. Multiple levels of risk thresholds are set (such as below 0.3 for low risk, 0.3 - 0.7 for medium risk, and above 0.7 for high risk), and the probability distribution map is marked in layers to highlight potential damage areas. Then, in combination with historical monitoring data and the expert knowledge base, in-depth analysis of high-risk areas is carried out. The current damage probability distribution is compared with the historical trend to identify abnormal changes; similar cases and treatment plans in the expert knowledge base are queried to assist in risk assessment. Then, based on a multi-factor comprehensive scoring model, the overall health status of the floating installation structure is quantitatively evaluated. Considering multi-dimensional indicators such as damage probability, importance of damage location, damage development trend, and environmental impact factors, a structural health index is calculated, ranging from 0 - 100, where 0 represents complete damage and 100 represents complete health. Finally, a structural health monitoring evaluation report is generated, including content such as damage location, damage type, damage degree, development trend, and maintenance suggestions. According to the evaluation result, hierarchical maintenance suggestions are given: SHI > 80 is the normal state, and regular monitoring is recommended; 60 < SHI ≤ 80 is a slight abnormality, and an increased monitoring frequency is recommended; 40 < SHI ≤ 60 is a moderate abnormality, and maintenance is recommended; SHI ≤ 40 is a serious abnormality, and immediate maintenance or replacement is recommended.

[0040] This invention achieves multi-dimensional and comprehensive automatic monitoring of the floating installation area by combining strain and temperature data detected by fiber optic sensors with inspection visual data, thus compensating for blind spots in manual monitoring and significantly expanding the monitoring range. Furthermore, temperature compensation is performed through spatiotemporal domain joint denoising and temperature-strain field coupling, improving the accuracy of strain measurement. This not only effectively suppresses noise interference from wave impacts and vortex-induced vibrations but also reduces the impact of large diurnal temperature variations on data measurement, ensuring that the monitoring data truly reflects the structural stress state. Finally, combining various types of data for structural damage identification enables more precise location and diagnosis of structural damage, early detection of potential safety hazards, and further improvement in the accuracy of health monitoring.

[0041] In one embodiment, steps 2011-2014 are described as follows:

[0042] Step 2011: The original strain data is segmented based on time series and spatial location to obtain multiple sub-blocks, and features are extracted from each sub-block to obtain the spatiotemporal features of the sub-block.

[0043] Optionally, the health monitoring system first segments the raw strain data based on time series and spatial location to obtain multiple sub-blocks, and then extracts features from each sub-block to obtain its spatiotemporal features. Specifically, the raw strain data is segmented according to time windows (e.g., 10 minutes per time window) and spatial location (e.g., 5 meters per spatial unit) to form multiple spatiotemporal sub-blocks. Statistical features (mean, variance, kurtosis, skewness), frequency domain features (power spectral density, dominant frequency component), and morphological features (waveform factor, impulse factor) are extracted from each sub-block to constitute its spatiotemporal feature vector.

[0044] Step 2012: Based on the spatiotemporal characteristics of the sub-blocks, determine the number of decomposition layers and the decomposition function for each sub-block, and decompose the sub-blocks based on the number of decomposition layers and the decomposition function to obtain multiple frequency band coefficients.

[0045] Optionally, the health monitoring system determines the number of decomposition levels and the decomposition function for each sub-block based on its spatiotemporal characteristics, and decomposes the sub-blocks based on the number of decomposition levels and the decomposition function to obtain multiple frequency band coefficients. Specifically, based on the signal-to-noise ratio and spectral characteristics of the sub-blocks, the wavelet function (such as Daubechies wavelet, Symlet wavelet, or Coiflet wavelet) and the optimal number of decomposition levels (usually 3-5 levels) are adaptively selected. Multi-scale wavelet decomposition is performed on each sub-block to obtain wavelet coefficients for different frequency bands, including high-frequency detail coefficients and low-frequency approximation coefficients.

[0046] Step 2013: Based on the frequency band coefficients and the threshold function, determine the frequency band threshold, and process the frequency band coefficients based on the frequency band threshold to obtain the frequency band denoising coefficients.

[0047] Optionally, for each frequency band coefficient obtained after decomposition, the health monitoring system considers the distribution of coefficients at adjacent positions in the spatiotemporal domain, and combines this with the previously calculated local mean μ0 and local variance σ0 of the sub-block, using the formula: Calculate the threshold, where μ0 represents the local mean of the sub-block; σ0 represents the local variance of the sub-block; k0 represents the adjustment coefficient; c i,j This is represented as the frequency band coefficients of adjacent positions of the sub-block in the spatiotemporal domain; Represented as the mean of adjacent coefficients; ω i,j This is represented as the weighting coefficient of adjacent positions in the spatiotemporal domain. The obtained threshold comprehensively considers the average level, fluctuation degree, and differences between adjacent coefficients of the sub-block. Then, each frequency band coefficient is compared with this threshold. Coefficients smaller than the threshold are considered noise correlation coefficients and are set to zero; coefficients larger than the threshold are retained, thus obtaining the frequency band denoising coefficients.

[0048] Step 2014: Perform inverse transformation on the frequency band denoising coefficients to reconstruct the strain denoised data.

[0049] Optionally, the health monitoring system utilizes inverse wavelet packet transform to reconstruct the frequency band denoising coefficients. During wavelet packet decomposition, the signal is decomposed into multiple frequency bands; the inverse transform recombines the coefficients in these frequency bands according to certain rules to recover an approximation of the original signal. Furthermore, spatiotemporal smoothing constraints are introduced during the reconstruction process. For boundary data of adjacent sub-blocks, weighted fusion is performed based on their correlation in the spatiotemporal domain. For example, for data points at the boundary of two adjacent sub-blocks, a weighted average is calculated according to their temporal and spatial distance ratio, making the reconstructed data smoother and more continuous in the spatiotemporal domain, ultimately yielding strain-denoised data. Introducing spatiotemporal smoothing constraints during reconstruction ensures the smoothness and continuity of the reconstructed data in the spatiotemporal domain, making the data more consistent with the characteristics of actual physical signals and improving data quality.

[0050] This invention, through sub-block division and feature extraction, adaptively selects decomposition parameters based on the characteristics of different sub-blocks, enabling processing of data characteristics at different spatiotemporal locations, thus improving the accuracy and effectiveness of denoising. Furthermore, by calculating thresholds based on spatiotemporal correlation, it more accurately identifies noise coefficients, effectively removing noise while preserving the effective features of the signal to the greatest extent, resulting in better denoising performance compared to traditional methods.

[0051] In one embodiment, steps 2021-2024 are described as follows:

[0052] Step 2021: Perform fractal dimension processing on the strain denoised data and the original temperature data at different spatiotemporal scales to obtain the first fractal features and the second fractal features at different scales.

[0053] Optionally, the health monitoring system uses the box dimension calculation method to calculate the fractal dimension, treating the strain-denoised data and the original temperature data as datasets distributed in the spatiotemporal domain. For spatiotemporal windows of different sizes (e.g., different durations in time and different ranges in space), boxes of size ∈ are used to cover the data points, and the number of boxes N(∈) required to cover the data is counted. As ∈ continuously decreases, the fractal dimension is calculated using the formula... Calculate the fractal dimension. The fractal dimension obtained from strain-denoised data constitutes the first fractal feature, and the fractal dimension obtained from the original temperature data constitutes the second fractal feature. For example, when monitoring a dam structure, the time window varies from 1 hour to 1 day, and the spatial scope is different monitoring sections of the dam. Calculate the fractal dimension of strain and temperature data under different windows. The fractal dimension can characterize the complexity and self-similarity of data at different spatiotemporal scales.

[0054] Step 2022: Based on the first fractal feature and the second fractal feature, the local region is divided to obtain multiple local regions.

[0055] Optionally, the health monitoring system divides the spatiotemporal domain into local regions based on the fractal dimension changes of the strain denoised data and the original temperature data obtained in the previous step. Specifically, regions with similar fractal dimension changes and similar data characteristics are grouped into the same local region. For example, in the monitoring data of a certain offshore platform, if the fractal dimension change trends of the strain and temperature data in the central region of the platform are found to be consistent and relatively stable over a period of time, then this region and the corresponding time period are divided into a local region. In this way, multiple local regions are divided in the entire spatiotemporal domain, so that the data in each local region has relatively consistent characteristics.

[0056] Step 2023: For each local region, construct a local regression model using the data points within that local region.

[0057] Optionally, the health monitoring system employs a local polynomial regression model within each local area. With temperature T as the independent variable and strain S as the dependent variable, a polynomial relationship is fitted using the least squares method. For example, for a quadratic polynomial regression model S... i =a i0 +a i1 T+a i2 T 2 …, where i represents the local region number. Temperature and strain data points are collected within the local region, and the data are substituted into the model. The sum of squared errors is minimized. (where S) j Represented as actual strain value, The polynomial coefficients a are determined by (represented as the strain value predicted by the model). i0 a i1 ai2 For different local regions, corresponding local regression models are constructed. Local multinomial regression models can flexibly fit the relationship between temperature and strain based on the characteristics of the data within each local region. Compared to a globally uniform linear model, it can better adapt to the nonlinear characteristics of data within local regions, improving the model's fitting accuracy in local areas.

[0058] Step 2024: The local regression models of each local region are fused to obtain the temperature-strain transfer model.

[0059] Optionally, the health monitoring system considers the transition and connection between adjacent local regions. Each local region's regression model is assigned a weight, determined based on the data point's location in the spatiotemporal domain and the correlation between adjacent regions. For example, for a data point located at the boundary of two adjacent local regions, the weight is determined based on factors such as its distance from the centers of the two regions. Then, the prediction results of different local models are combined based on the data point's location. For instance, for temperature data at a certain location, the strain value is predicted using the regression models of adjacent local regions, and then a weighted average is calculated to obtain the final predicted strain value. This method constructs a global temperature-strain transfer model.

[0060] This invention utilizes fractal dimension to process data, enabling in-depth analysis of the complex characteristics of strain and temperature data at different spatiotemporal scales. This breaks through the limitations of traditional methods that only focus on numerical data, providing richer and more accurate information for subsequent modeling. Furthermore, based on fractal features, local regions are divided, and local regression models are constructed within each local region. This allows for targeted modeling based on the data characteristics of different regions, improving the model's fitting accuracy within local regions and accurately describing the local temperature-strain relationship.

[0061] In one embodiment, steps 301-304 are described as follows:

[0062] Step 301: Based on the temperature-strain transfer model and the original temperature data, determine the probability distribution of the original data.

[0063] Optionally, the health monitoring system utilizes Bayesian inference to determine the probability distribution of the raw temperature data. This Bayesian inference is based on Bayes' theorem. Here, the temperature-strain transfer model is considered as a known conditional relationship, with the original temperature data used as observational data. Assume the temperature is T. t First, a prior probability distribution P(T) for temperature is defined, which can be approximated by statistical analysis of historical temperature data. Then, the probability P(S|T) of strain data occurring at a given temperature T is calculated using a temperature-strain transfer model. Finally, based on the actually observed strain data S, the posterior probability is calculated using Bayes' theorem. This allows us to obtain the probability of temperature taking different values, i.e., the probability distribution of temperature data.

[0064] Step 302: Perform error analysis on the raw temperature data, and determine the error probability based on the error analysis results and probability distribution.

[0065] Optionally, the health monitoring system first analyzes the error between the raw temperature data and the actual situation. This can be done by comparing it with more accurate temperature measuring devices (such as high-precision thermometers) or by considering the physical characteristics of the structure (such as the theoretical temperature range under certain operating conditions). Then, based on the probability distribution of the previously obtained temperature data, the probability of different error values ​​occurring is calculated. For example, if the known temperature measurement value is T... mean The actual temperature range is T min ,T max Calculate T based on the probability distribution P(T). mean Falling into different error ranges such as [T mean -ΔT,T mean The probability of +ΔT]. By combining error analysis and probability distribution to determine the error probability, the error situation of temperature data and its likelihood of occurrence can be quantified.

[0066] Step 303: Based on the error probability and temperature-strain transfer model, determine the correction parameters, and compensate the original temperature data based on the correction parameters to obtain preliminary compensation data.

[0067] Optionally, the health monitoring system uses formula T based on the error probability and the temperature-strain transfer model. * =T+∑ i p i ·δ i To calculate the correction parameters. Where T is the original temperature, p i Let δ be the error probability. i This corresponds to the error correction amount. Error correction amount δ i The value can be determined based on the temperature-strain transfer model and the actual error conditions. For example, if the model analysis shows that the temperature error is ΔT, the effect on strain is ΔS. To eliminate this effect, the corresponding δ needs to be determined. i Then, the calculated correction parameters are applied to the original temperature data to obtain the preliminary compensation data T. * .

[0068] Step 304: Verify the preliminary compensation data to obtain temperature compensation data.

[0069] Optionally, the health monitoring system calculates the posterior probability of the compensated temperature data based on the obtained preliminary compensation data to verify whether the compensation effect meets expectations. This can be determined by comparing the degree of conformity between the temperature data before and after compensation and the actual situation (such as comparing with high-precision measurements or considering the physical constraints of the structure). If the posterior probability indicates that the compensated data still has a large deviation and does not meet expectations, the correction parameters are readjusted, and the compensation process is repeated. For example, an acceptable error range is set; if the error between the compensated temperature data and the actual value exceeds this range, the error correction amount is adjusted, and compensation is performed again until satisfactory temperature compensation data is obtained.

[0070] This invention, from a Bayesian probabilistic perspective, fully considers the uncertainty of temperature data and combines a temperature-strain model for probabilistic correction and compensation, which better reflects the characteristics of error and randomness in actual temperature measurements. Through multi-step analysis and calculation, including determining the probability distribution, error analysis, calculating correction parameters, and verification and adjustment, it can more accurately compensate for the original temperature data and improve the accuracy of the temperature data.

[0071] In one embodiment, steps 401-405 are described as follows:

[0072] Step 401: Based on temperature compensation data and strain denoising data, construct a spatiotemporal feature vector, and perform feature extraction and dimensionality reduction on the inspection visual data to obtain a visual feature vector.

[0073] Optionally, the health monitoring system uses the acquired temperature compensation data T c and strain-denoised data ε d Due to temperature compensation data T c and strain-denoised data ε d This reflects the structural characteristics in time and space. Therefore, data from historical time τ are selected and combined in a fixed order to form a spatiotemporal feature vector f. ts =[T c (t-τ),…,T c (t),ε d (t-τ),…,ε d (t)]. For example, during the floating installation process, the temperature compensation value and strain denoising value are considered every 10 minutes in the past hour (assuming τ is 1 hour and the sampling interval is 10 minutes). They are arranged in sequence to form a spatiotemporal feature vector, so that it comprehensively reflects the spatiotemporal change information of the structure in that time period.

[0074] Further, for the visual inspection data of the health monitoring system, the Histogram of Oriented Gradients (HOG) feature extraction method is first used. HOG constructs features by calculating and statistically analyzing the gradient direction histogram of local regions of an image. The image is divided into multiple small cell units. The gradient direction and magnitude of the pixels within each cell are calculated, and then the gradient direction histogram is statistically analyzed within each cell unit. Next, these histograms are combined to obtain the local gradient histogram. Since the dimension of the original HOG features may be relatively high, principal component analysis (PCA) is then used to reduce its dimension to d dimensions, obtaining the visual feature vector f v .

[0075] Step 402: Take each physical node in the floating installation structure as a vertex of the graph, and construct an adjacency matrix according to the spatial position relationship between the physical nodes, and determine the degree matrix based on the adjacency matrix.

[0076] Optionally, the health monitoring system regards each physical node in the floating installation structure as a vertex in the graph theory. An adjacency matrix A is constructed according to the spatial position relationship between the nodes, and its element A ij is calculated as If the distance between nodes i and j < R, where x i , x j respectively represent the coordinates of nodes i and j

[0077] Otherwise. For example, in an offshore floating installation platform structure, if the distance between two nodes (such as the connection points of two struts) is less than the set R, the adjacency weight between them is calculated according to the above formula: if the distance is greater than R, the adjacency weight is 0. In addition, the degree matrix D is represented as a diagonal matrix, and its diagonal element D ii = ∑ j A ij , that is, the sum of the weights of all adjacent edges of node i. By constructing the adjacency matrix and the degree matrix, the topological relationship of the floating installation structure is explicitly modeled, which conforms to the mechanical conduction law and can reflect the spatial correlation and interaction between nodes.

[0078] Step 403: Align and splice the spatio-temporal feature vector and the visual feature vector to obtain a joint embedding feature vector.

[0079] Optionally, the health monitoring system uses canonical correlation analysis (CCA) to align the temperature-strain features (i.e., spatio-temporal feature vector) f ts and the visual feature vector f v . Specifically, the goal of CCA is to find the linear combinations of two sets of variables (here, the spatio-temporal feature vector and the visual feature vector) such that the correlation between them is maximized. The transformation matrix is calculated through CCA, and f tsand f v Perform transformations separately, then concatenate the transformed features to generate a joint embedding feature vector z = [CCA(f is [CCA(f0)]. For example, suppose that after CCA transformation, the spatiotemporal feature vector changes from m-dimensional to p-dimensional, and the visual feature vector also changes from d-dimensional to p-dimensional (by adjusting the CCA transformation to make the dimensions consistent). These two p-dimensional feature vectors are then concatenated sequentially to obtain a 2p-dimensional joint embedding feature vector. CCA can effectively solve the heterogeneity problem of multi-source data by finding the maximum correlation between two sets of features, aligning different types of data in the feature space, and enabling them to be better integrated.

[0080] Step 404: Perform graph convolution processing on the joint embedded feature vector based on the graph convolution layer to obtain the node feature matrix.

[0081] Optionally, the health monitoring system pre-sets the calculation formula for the graph convolutional layer as follows: in Represented as a normalized adjacency matrix, H (l) Represented as the feature matrix of the l-th layer (initially H) (0) (This can be a matrix composed of jointly embedded feature vectors) W (i) The weights are represented as learnable weights (fixed weights are pre-computed via singular value decomposition to avoid training), and σ is the activation function (such as the RLU function). In graph convolution operations, the adjacency matrix is ​​normalized. For the characteristic matrix H (l) Weighted aggregation is performed so that the features of each node can be fused with the information of its neighboring nodes, and then the result is obtained through the weight matrix W. (l) A linear transformation is performed, followed by an activation function to introduce nonlinearity. After multiple layers of graph convolution, the final node feature matrix H is obtained. out For example, for a model with three graph convolutional layers, starting from the initial joint embedding feature vector matrix H... (0) Initially, after three sequential operations, H is obtained. (3) As a node feature matrix.

[0082] Graph convolutional layers can effectively propagate feature information across graph structures, enabling each node's features to not only contain its own information but also incorporate information from neighboring nodes, thereby capturing the spatial relationships between structural nodes.

[0083] Step 405: Based on the node feature matrix, determine the predicted probability of each physical node, and map the predicted probability of each physical node to the corresponding position of the floating installation structure to obtain the damage probability distribution map.

[0084] Optionally, the health monitoring system outputs a node feature matrix H. out Gaussian process regression (GPR) is applied to determine the predicted probability for each physical node. For node i, the predicted probability formula is: Among them, P i Let β represent the predicted probability of the i-th physical node; β represents the scaling hyperparameter; N(i) represents the set of neighboring nodes of node i; h i h j h k Let represent the eigenvectors of nodes i, j, and k in the node feature matrix, respectively; <*,*> represent the vector inner product operation; and N represents the total number of physical nodes in the floating installation structure. After calculating the predicted probability of each node, it is mapped to the corresponding position on the floating installation structure and visualized using color coding to obtain a damage probability distribution map. For example, areas with higher probabilities are represented in red, and areas with lower probabilities are represented in blue, intuitively showing the probability of damage to different parts of the structure.

[0085] This invention employs multiple methods to extract and fuse features from temperature-compensated data, strain-denoised data, and inspection visual data, fully leveraging the advantages of different data types to comprehensively describe the structural state. Specifically, spatiotemporal feature vectors capture the spatiotemporal dynamic changes of the structure, while visual feature vectors extract surface damage information. Jointly embedding feature vectors integrates both, providing a rich and comprehensive feature representation for subsequent analysis. Furthermore, a graph-based modeling of the floating installation structure accurately captures the spatial relationships and mechanical transmission patterns between nodes through adjacency matrices and graph convolutional layers. This modeling approach aligns with the physical characteristics of the structure, effectively propagating and fusing node feature information, leading to a better understanding of the propagation and distribution of structural damage.

[0086] Furthermore, the health monitoring system for floating installation structures based on fiber optic sensors provided by the present invention will be described below. The health monitoring system for floating installation structures based on fiber optic sensors described below can be referred to in correspondence with the health monitoring method for floating installation structures based on fiber optic sensors described above.

[0087] Optionally, refer to Figure 2 , Figure 2 This is a schematic diagram of the health monitoring system for a floating installation structure based on fiber optic sensors provided by the present invention. The health monitoring system for a floating installation structure based on fiber optic sensors includes:

[0088] The data acquisition module 210 is used to acquire raw strain data and raw temperature data of the floating installation area based on fiber optic sensors, and to acquire inspection visual data of the underwater floating installation area.

[0089] The denoising and construction module 220 denoises the original strain data based on spatiotemporal domain joint processing to obtain denoised strain data, and constructs a temperature-strain transfer model based on the denoised strain data and the original temperature data.

[0090] Temperature compensation module 230 is used to compensate the original temperature data based on the temperature-strain transfer model to obtain temperature compensation data;

[0091] The fusion diagnostic module 240 is used to extract and fuse features from temperature compensation data, strain denoising data and inspection visual data to obtain a structural damage probability distribution map.

[0092] The analysis and evaluation module 250 is used to analyze and evaluate the probability distribution map of structural damage and determine the results of structural health monitoring and evaluation.

[0093] This invention achieves multi-dimensional and comprehensive automatic monitoring of the floating installation area by combining strain and temperature data detected by fiber optic sensors with inspection visual data, thus compensating for blind spots in manual monitoring and significantly expanding the monitoring range. Furthermore, temperature compensation is performed through spatiotemporal domain joint denoising and temperature-strain field coupling, improving the accuracy of strain measurement. This not only effectively suppresses noise interference from wave impacts and vortex-induced vibrations but also reduces the impact of large diurnal temperature variations on data measurement, ensuring that the monitoring data truly reflects the structural stress state. Finally, combining various types of data for structural damage identification enables more precise location and diagnosis of structural damage, early detection of potential safety hazards, and further improvement in the accuracy of health monitoring.

[0094] Please see Figure 3 , Figure 3 An embodiment diagram of an electronic device provided in accordance with the present invention. For example... Figure 3 As shown, this embodiment of the invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, it performs the following steps:

[0095] The original strain and temperature data of the floating installation area are obtained based on fiber optic sensors, and the inspection visual data of the underwater floating installation area are also obtained.

[0096] The original strain data is denoised based on spatiotemporal domain joint denoising to obtain strain denoised data. Based on the strain denoised data and the original temperature data, a temperature-strain transfer model is constructed.

[0097] The original temperature data is compensated based on the temperature-strain transfer model to obtain temperature compensation data;

[0098] Feature extraction and fusion of temperature compensation data, strain denoising data and inspection visual data are performed to obtain a structural damage probability distribution map.

[0099] The probability distribution map of structural damage is analyzed and evaluated to determine the results of structural health monitoring and evaluation.

[0100] Please see Figure 4 , Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with an embodiment of the present invention is shown. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, it performs the following steps:

[0101] The original strain and temperature data of the floating installation area are obtained based on fiber optic sensors, and the inspection visual data of the underwater floating installation area are also obtained.

[0102] The original strain data is denoised based on spatiotemporal domain joint denoising to obtain strain denoised data. Based on the strain denoised data and the original temperature data, a temperature-strain transfer model is constructed.

[0103] The original temperature data is compensated based on the temperature-strain transfer model to obtain temperature compensation data;

[0104] Feature extraction and fusion of temperature compensation data, strain denoising data and inspection visual data are performed to obtain a structural damage probability distribution map.

[0105] The probability distribution map of structural damage is analyzed and evaluated to determine the results of structural health monitoring and evaluation.

[0106] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the fiber optic sensor-based health monitoring method for floating installation structures provided by the above methods, the method comprising:

[0107] The original strain and temperature data of the floating installation area are obtained based on fiber optic sensors, and the inspection visual data of the underwater floating installation area are also obtained.

[0108] The original strain data is denoised based on spatiotemporal domain joint denoising to obtain strain denoised data. Based on the strain denoised data and the original temperature data, a temperature-strain transfer model is constructed.

[0109] The original temperature data is compensated based on the temperature-strain transfer model to obtain temperature compensation data;

[0110] Feature extraction and fusion of temperature compensation data, strain denoising data and inspection visual data are performed to obtain a structural damage probability distribution map.

[0111] The probability distribution map of structural damage is analyzed and evaluated to determine the results of structural health monitoring and evaluation.

[0112] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0113] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the health of a float-on installation structure based on an optical fiber sensor, characterized by, The method comprises the following steps: obtaining original strain data and original temperature data of a float-over installation area based on an optical fiber sensor, and obtaining inspection visual data of the underwater float-over installation area; based on the space-time domain joint, denoising the original strain data to obtain strain denoised data, and based on the strain denoised data and the original temperature data, constructing a temperature-strain transfer model; compensating the original temperature data based on the temperature-strain transfer model to obtain temperature compensation data; extracting and fusing features of the temperature compensation data, strain denoised data and inspection visual data to obtain a structure damage probability distribution map; analyzing and evaluating the structure damage probability distribution map to determine a structure health monitoring evaluation result.

2. The fiber optic sensor based float-over installation structure health monitoring method of claim 1, wherein, The method comprises the following steps: based on time series and spatial position, segmenting the original strain data to obtain a plurality of sub-blocks, and extracting features of each sub-block to obtain sub-block space-time domain features; based on the sub-block space-time domain features, determining the decomposition layer number and decomposition function of each sub-block, and based on the decomposition layer number and decomposition function, decomposing the sub-block to obtain a plurality of frequency band coefficients; based on the frequency band coefficients and a threshold function, determining a frequency band threshold, and based on the frequency band threshold, processing the frequency band coefficients to obtain frequency band denoised coefficients; inverse transforming and reconstructing the frequency band denoised coefficients to obtain strain denoised data.

3. The optical fiber sensor based float-over installation structural health monitoring method of claim 2, wherein, The method comprises the following steps: fractal dimension processing the strain denoised data and the original temperature data at different space-time scales to obtain first fractal features and second fractal features at different scales; based on the first fractal features and second fractal features, dividing a local area to obtain a plurality of local areas; for each local area, constructing a local regression model based on data points in the local area; fuse the local regression models of each local area to obtain the temperature-strain transfer model.

4. The fiber optic sensor based float-over installation structure health monitoring method of claim 2, wherein, The threshold function is: wherein μ0represents a local mean value for the sub-block; σ0represents a local variance for the sub-block; k0represents an adjustment coefficient; c i,j represents a band coefficient for the sub-block at a spatially and temporally adjacent position; represents a mean value of adjacent coefficients; ω i,j represents a weight coefficient for a spatially and temporally adjacent position.

5. The optical fiber sensor based float-over installation structure health monitoring method of claim 1, wherein, The method comprises the following steps: based on the temperature-strain transfer model and the original temperature data, determining the probability distribution of the original data; error analysis of the original temperature data, and based on the error analysis result and the probability distribution, determining an error probability; based on the error probability and the temperature-strain transfer model, determining a correction parameter, and based on the correction parameter, compensating the original temperature data to obtain preliminary compensation data; verifying the preliminary compensation data to obtain the temperature compensation data.

6. The fiber optic sensor-based float-over installation structure health monitoring method of claim 1, wherein, The method comprises the following steps: based on the temperature compensation data and the strain denoised data, constructing a space-time feature vector, and extracting and dimension-reducing features of the inspection visual data to obtain a visual feature vector; each physical node in the float-over installation structure is taken as a vertex of a graph, an adjacency matrix is constructed according to a spatial position relationship between the physical nodes, and a degree matrix is determined based on the adjacency matrix; alignment and splicing are performed between the spatio-temporal feature vector and the visual feature vector to obtain a joint embedding feature vector; graph convolution processing is performed on the joint embedding feature vector based on a graph convolution layer to obtain a node feature matrix; a prediction probability of each physical node is determined based on the node feature matrix, and the prediction probability of each physical node is mapped to a position corresponding to the float-over installation structure to obtain the damage probability distribution map.

7. The optical fiber sensor based float-over installation structural health monitoring method of claim 6, wherein, The prediction probability formula of the physical node is: where P i denotes the predicted probability of the i-th physical node; β denotes the scaling hyper-parameter; N(i) denotes the set of neighboring nodes of node i; h i , h j , h k denote the feature vectors of node i, node j, and node k in the node feature matrix, respectively; <*, *> denotes the vector inner product operation; N denotes the total number of physical nodes in the floating installation structure.

8. A fiber optic sensor based float-over installation structure health monitoring system, characterized by, The float-over installation structure health monitoring method based on the optical fiber sensor is applied to any one of claims 1 to 7. The float-over installation structure health monitoring system based on the optical fiber sensor comprises: A data acquisition module is configured to acquire original strain data and original temperature data of a float-over installation area based on an optical fiber sensor, and to acquire inspection visual data of an underwater float-over installation area. A denoising and construction module is configured to denoise the original strain data based on a spatio-temporal domain joint to obtain strain denoised data, and to construct a temperature-strain transfer model based on the strain denoised data and the original temperature data. A temperature compensation module is configured to compensate the original temperature data based on the temperature-strain transfer model to obtain temperature compensation data. A fusion diagnosis module is configured to extract and fuse features of the temperature compensation data, the strain denoised data, and the inspection visual data to obtain a structure damage probability distribution map. An analysis and evaluation module is configured to analyze and evaluate the structure damage probability distribution map to determine a structure health monitoring evaluation result.

9. An electronic device comprising: A memory is configured to store a computer software program. A processor is configured to read and execute the computer software program, and when the processor executes the computer software program, the float-over installation structure health monitoring method based on the optical fiber sensor is implemented.

10. A non-transitory computer readable storage medium having stored therein a computer software program, characterized in that, When the computer software program is executed by the processor, the float-over installation structure health monitoring method based on the optical fiber sensor is implemented.

Images