Bridge modal identification method and system based on distributed optical fiber and power spectral density, medium and product

Abstract

The invention discloses a bridge modal identification method and system based on distributed optical fibers and power spectral density, a medium and a product, and relates to the technical field of bridge modal identification, and the method comprises the steps: obtaining a multi-channel vibration signal collected by a distributed optical fiber sensor system; performing frequency domain analysis on the multi-channel vibration signals to obtain a power spectral density energy diagram; and according to the power spectral density energy diagram, identifying a bridge modal. According to the method, the more accurate bridge modality can be obtained.

Description

Technical Field

[0001] This invention relates to the field of bridge modal identification technology, and in particular to a bridge modal identification method, system, medium, and product based on distributed optical fiber and power spectral density. Background Technology

[0002] Due to the complexity of long-span bridge structures, their variable stress states, and the significant influence of environmental factors, traditional design phases struggle to accurately identify weak points under strong earthquakes. Therefore, acquiring measured response data of bridges under natural or man-made disturbances not only helps verify the rationality of structural analysis models and design parameters but also reveals nonlinear behavior, boundary condition variations, and other potential uncertainties within the bridge system. While traditional methods such as EEMD and SVD possess strong feature extraction capabilities, their complex algorithm implementation and high computational resource requirements make them unsuitable for the real-time and efficiency demands of large-scale bridge monitoring scenarios. Summary of the Invention

[0003] To address the technical problems existing in the background art, this invention proposes a bridge mode identification method, system, medium, and product based on distributed optical fiber and power spectral density.

[0004] This invention proposes a bridge mode identification method based on distributed optical fiber and power spectral density, comprising:

[0005] Acquire multi-channel vibration signals collected by a distributed fiber optic sensor system;

[0006] Frequency domain analysis was performed on the multi-channel vibration signal to obtain the power spectral density energy map;

[0007] The bridge modes were identified based on the power spectral density energy map.

[0008] Preferably, frequency domain analysis is performed on the multi-channel vibration signals to extract the frequency response characteristics of the bridge structure, specifically including:

[0009] The vibration signal of each channel is divided into K segments according to the time sequence;

[0010] Windowing is applied to each segment of the vibration signal;

[0011] Perform a fast Fourier transform on each windowed segment to calculate the power spectral density of the vibration signal for each segment;

[0012] The average power spectral density of each channel is calculated based on the power spectral density of each segment of each channel.

[0013] Based on the power spectral density of each channel, output a power spectral density energy map.

[0014] Preferably, the vibration signals of any adjacent segments overlap by a point D, where D is 30% to 70%.

[0015] Preferably, the window function used in the windowing process is a Hamming window or a Hanning window.

[0016] Preferably, the power spectral density of each vibration signal segment is

[0017]

[0018] In the formula, P k (f) represents the power spectral density of the k-th vibration signal, U is the energy normalization factor of the window function, and x k [n] is the discrete representation of the k-th signal segment, n = 0, 1, 2, ..., L-1, where L is the length of each signal segment; w[n] is the window function; k = 1, 2, 3, ..., K, where K is the total number of vibration signal segments in each channel;

[0019] Wherein, the average power spectral density of each channel is

[0020] In the formula, This represents the average power spectral density per channel.

[0021] Preferably, the bridge modes are identified based on the power spectral density energy map, including:

[0022] The power spectral density energy map is identified using a pre-trained bridge mode recognition model to obtain the bridge modes.

[0023] Preferably, before identifying the power spectral density energy map using a pre-trained bridge modal recognition model to obtain the bridge modes, the method further includes:

[0024] Construct a bridge modality recognition model based on deep learning;

[0025] The bridge mode recognition model was pre-trained using a semi-supervised learning method with power spectral density energy maps that included some bridge mode labels and some that did not.

[0026] The pre-trained bridge modal recognition model was fine-tuned using a small number of power spectral density energy maps with specific bridge modal labels.

[0027] Secondly, this invention also proposes a bridge mode identification system based on distributed optical fiber and power spectral density, comprising:

[0028] The acquisition module is used to acquire multi-channel vibration signals collected by the distributed fiber optic sensor system.

[0029] The analysis module is used to perform frequency domain analysis on multi-channel vibration signals to obtain the power spectral density energy map;

[0030] The identification module is used to identify the bridge modes based on the power spectral density energy map.

[0031] Thirdly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of the first aspects.

[0032] Fourthly, the present invention also proposes a computer program product, including a computer program that, when executed by a processor, implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of the first aspects.

[0033] The bridge mode identification method, system, medium, and product proposed in this invention based on distributed optical fiber and power spectral density achieve continuous, long-term, and high-temporal-resolution vibration signal acquisition of bridge structures by introducing DAS technology. Combined with PSD analysis, the accuracy of frequency domain feature extraction can be improved, that is, a more accurate power spectral density energy map can be obtained, thereby identifying more accurate bridge modes based on the power spectral density energy map. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a bridge mode identification method based on distributed optical fiber and power spectral density in one embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of a multi-channel vibration signal in one embodiment of the present invention.

[0036] Figure 3 This is a schematic diagram of the power spectral density energy map in one embodiment of the present invention. Detailed Implementation

[0037] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0038] Reference Figures 1-3 This invention proposes a bridge mode identification method based on distributed optical fiber and power spectral density, comprising:

[0039] Acquire multi-channel vibration signals collected by a distributed fiber optic sensor system;

[0040] Frequency domain analysis was performed on the multi-channel vibration signal to obtain the power spectral density energy map;

[0041] The bridge modes were identified based on the power spectral density energy map.

[0042] This invention introduces DAS technology to achieve continuous, long-term, and high-temporal-resolution vibration signal acquisition of bridge structures. It utilizes multi-channel signals from a distributed optical fiber system to enhance the robustness of the model. Combined with PSD analysis, it can improve the accuracy of frequency domain feature extraction, i.e., obtain a more accurate power spectral density energy map. Based on the power spectral density energy map, a more accurate bridge mode can be obtained, which is suitable for complex bridge vibration modes.

[0043] The method of this invention has the advantages of flexible deployment, efficient data acquisition, intuitive spectrum analysis, and simple algorithm implementation, and can effectively identify the health status of bridges.

[0044] In this embodiment, frequency domain analysis is performed on the multi-channel vibration signal to extract the frequency response characteristics of the bridge structure, specifically including:

[0045] The vibration signal of each channel is divided into K segments according to the time sequence;

[0046] Windowing is applied to each segment of the vibration signal;

[0047] Perform a fast Fourier transform on each windowed segment to calculate the power spectral density of the vibration signal for each segment;

[0048] The average power spectral density of each channel is calculated based on the power spectral density of each segment of each channel.

[0049] Based on the power spectral density of each channel, output a power spectral density energy map.

[0050] This implementation utilizes windowed averaging of long-term observation data, effectively suppressing non-stationary noise interference and improving the smoothness and reliability of spectral estimation. This is particularly beneficial in real-world scenarios involving non-stationarity and low signal-to-noise ratios in bridge structure vibration response signals, significantly reducing the variance of power spectrum estimation and enhancing spectral stability and reliability. Furthermore, due to the long-term continuous observations of DAS, segmenting the data only reduces the data scale relatively, without decreasing frequency resolution.

[0051] In a further embodiment, the vibration signals of any adjacent segments overlap by a point D, where D is 30% to 70%, to ensure spectral continuity.

[0052] In a further embodiment, the window function used in the windowing process is a Hamming window or a Hanning window to reduce frequency leakage.

[0053] The power spectral density of each vibration signal segment is:

[0054]

[0055] In the formula, P k (f) represents the power spectral density of the k-th vibration signal, U is the energy normalization factor of the window function, and x k [n] is the discrete representation of the k-th signal segment, n = 0, 1, 2, ..., L-1, where L is the length of each signal segment; w[n] is the window function; k = 1, 2, 3, ..., K, where K is the total number of segments of the vibration signal in each channel.

[0056] Wherein, the average power spectral density of each channel is

[0057] In the formula, This represents the average power spectral density per channel.

[0058] In one specific embodiment, identifying bridge modes based on the power spectral density energy map specifically includes:

[0059] Observe the frequency distribution: Examine the distribution along the frequency axis of the PSD plot; different frequency components correspond to different dynamic modes and responses of the bridge structure.

[0060] Analyze energy peaks: In the PSD plot, look for peaks in energy intensity (usually represented by color scales); these peaks correspond to the bridge's natural frequencies, which are the natural vibration frequencies of the bridge at specific frequencies; for each energy peak, record its corresponding frequency value, which can help identify the bridge's modes.

[0061] Comparing different channel numbers: The PSD plot calculated based on the distributed fiber method contains multiple channel numbers (i.e., multiple measurement points). Comparing the energy intensity of the same frequency under different channel numbers helps to determine which modes are significant at different locations on the bridge.

[0062] Combining structural knowledge: Based on the structural characteristics and design parameters of the bridge, analyze and identify the relationship between frequencies and possible modes of the bridge; for example, low-frequency modes may be related to the overall bending or torsion of the bridge, while high-frequency modes may be related to local vibrations or smaller structural components.

[0063] In order to accurately identify bridge modes, in another embodiment, bridge modes are identified based on the power spectral density energy map, including:

[0064] The power spectral density energy map is identified using a pre-trained bridge mode recognition model to obtain the bridge modes.

[0065] In this embodiment, the bridge modal recognition model includes a convolutional neural network (CNN) and a recurrent neural network (RNN). The CNN is used to extract frequency domain features, while the RNN is used to capture temporal information in multi-channel vibration signals, so as to improve the recognition accuracy by combining the convolutional neural network and the recurrent neural network.

[0066] Moreover, the high-frequency, high-resolution data provided by the DAS system in this embodiment enables the model to learn more modal features, improving its recognition accuracy and adaptability in complex environments, and enabling the model to better cope with various uncertainties and changes in practical applications.

[0067] In one specific embodiment, supervised learning of the bridge modal recognition model can also be performed using labeled power spectral density energy maps, but obtaining a large number of labeled samples is very costly.

[0068] In another specific embodiment, before identifying the power spectral density energy map using a pre-trained bridge mode recognition model to obtain the bridge modes, the method further includes:

[0069] Construct a bridge modality recognition model based on deep learning;

[0070] The bridge mode recognition model was pre-trained using a semi-supervised learning method with power spectral density energy maps that included some bridge mode labels and some that did not.

[0071] The pre-trained bridge modal recognition model is fine-tuned using a small number of power spectral density energy maps with specific bridge modal labels to obtain a trained bridge modal recognition model.

[0072] This embodiment maximizes the use of all available data, including some unlabeled data, through a joint semi-supervised learning method combining labeled and unlabeled data. In this way, the model can still improve recognition performance even with limited data. Furthermore, the model is pre-trained using a semi-supervised learning method, and then fine-tuned based on the power spectral density energy map corresponding to the vibration data of a specific bridge, thereby improving the accuracy of identifying the target bridge modes. The specific bridge mode labels are part of the bridge mode labels used in the pre-training process.

[0073] Secondly, this invention also proposes a bridge mode identification system based on distributed optical fiber and power spectral density, comprising:

[0074] The acquisition module is used to acquire multi-channel vibration signals collected by the distributed fiber optic sensor system.

[0075] The analysis module is used to perform frequency domain analysis on multi-channel vibration signals to obtain the power spectral density energy map;

[0076] The identification module is used to identify the bridge modes based on the power spectral density energy map.

[0077] In this embodiment, a distributed optical fiber monitoring module is also included. The distributed optical fiber monitoring module is electrically connected to the PSD analysis module and is used to continuously collect multi-point vibration signals of the entire bridge structure.

[0078] The distributed fiber optic monitoring module, based on DAS technology, enables continuous acquisition of multi-point vibration signals across the entire bridge structure. The DAS host acquires real-time strain rate change data along the direction of the sensing fiber optic cable using the backscattering Rayleigh principle. In practical applications, existing communication fiber optic cables within the bridge are preferred. Where the cable installation location has good coupling with the bridge structure, it can be directly connected to the DAS system for signal acquisition. If the existing communication fiber optic cable location does not meet coupling requirements (e.g., suspended installation), manual re-laying is necessary. It is recommended to install the cable on the bridge deck or in areas sensitive to structural stiffness, using methods such as pasting, burying, or binding to ensure signal coupling quality.

[0079] In this embodiment, a data storage module is also included. The data storage module is electrically connected between the distributed optical fiber monitoring module and the PSD analysis module. The data storage module is used to complete the configuration of acquisition parameters and data management functions of the distributed optical fiber monitoring module.

[0080] The acquisition parameters include settings for sampling frequency, channel spacing, and signal-to-noise ratio control to ensure that the vibration signal covers the typical structural response frequency band. All acquired raw signals are organized into a two-dimensional matrix (channel × time) and stored in the local data storage module according to time periods. Considering that DAS has long-term, continuous observation capabilities, the system supports periodic automatic storage and compression mechanisms to ensure the real-time performance and efficiency of subsequent batch PSD analysis.

[0081] The analysis module performs frequency domain analysis on the multi-channel vibration signals acquired by DAS and extracts the frequency response characteristics of the bridge structure, namely the power spectral density energy map.

[0082] Specifically, it includes:

[0083] The vibration signal of each channel is divided into K segments according to the time sequence;

[0084] Windowing is applied to each segment of the vibration signal;

[0085] Perform a fast Fourier transform on each windowed segment to calculate the power spectral density of the vibration signal for each segment;

[0086] The average power spectral density of each channel is calculated based on the power spectral density of each segment of each channel.

[0087] Based on the power spectral density of each channel, output a power spectral density energy map.

[0088] This implementation utilizes windowed averaging of long-term observation data, effectively suppressing non-stationary noise interference and improving the smoothness and reliability of spectral estimation. This is particularly beneficial in real-world scenarios involving non-stationarity and low signal-to-noise ratios in bridge structure vibration response signals, significantly reducing the variance of power spectrum estimation and enhancing spectral stability and reliability. Furthermore, due to the long-term continuous observations of DAS, segmenting the data only reduces the data scale relatively, without decreasing frequency resolution.

[0089] In a further embodiment, the vibration signals of any adjacent segments overlap by a point D, where D is 30% to 70%, to ensure spectral continuity.

[0090] In a further embodiment, the window function used in the windowing process is a Hamming window or a Hanning window to reduce frequency leakage.

[0091] The power spectral density of each vibration signal segment is:

[0092]

[0093] In the formula, P k (f) represents the power spectral density of the k-th vibration signal, U is the energy normalization factor of the window function, and x k [n] is the discrete representation of the k-th signal segment, n = 0, 1, 2, ..., L-1, where L is the length of each signal segment; w[n] is the window function; k = 1, 2, 3, ..., K, where K is the total number of segments of the vibration signal in each channel.

[0094] Wherein, the average power spectral density of each channel is

[0095] In the formula, This represents the average power spectral density per channel.

[0096] In order to accurately identify the bridge modes, the identification module in this embodiment uses a pre-trained bridge mode identification model to identify the power spectral density energy map and obtain the bridge modes.

[0097] Of course, in this embodiment, it also includes:

[0098] The model building module is used to build a bridge modality recognition model based on deep learning.

[0099] The training set acquisition module is used to acquire the training dataset; the training dataset includes a power spectral density energy map with bridge mode labels.

[0100] The training module is used to train the bridge modal recognition model using the training dataset to obtain a trained bridge modal recognition model.

[0101] Thirdly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of the first aspects.

[0102] Fourthly, the present invention also proposes a computer program product, including a computer program that, when executed by a processor, implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of the first aspects.

[0103] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A bridge mode identification method based on distributed optical fiber and power spectral density, characterized in that, include: Acquire multi-channel vibration signals collected by a distributed fiber optic sensor system; Frequency domain analysis was performed on the multi-channel vibration signal to obtain the power spectral density energy map; The bridge modes were identified based on the power spectral density energy map.

2. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 1, characterized in that, Frequency domain analysis was performed on multi-channel vibration signals to extract the frequency response characteristics of the bridge structure, specifically including: The vibration signal of each channel is divided into K segments according to the time sequence; Windowing is applied to each segment of the vibration signal; Perform a fast Fourier transform on each windowed segment to calculate the power spectral density of the vibration signal for each segment; The average power spectral density of each channel is calculated based on the power spectral density of each segment of each channel. Based on the power spectral density of each channel, output a power spectral density energy map.

3. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 2, characterized in that, The vibration signals of any adjacent segments overlap by point D, where D is 30% to 70%.

4. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 2, characterized in that, The window function used in windowing processing is either the Hamming window or the Hanning window.

5. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 2, characterized in that, The power spectral density of each vibration signal segment is In the formula, P k (f) represents the power spectral density of the k-th vibration signal, U is the energy normalization factor of the window function, and x k [n] is the discrete representation of the k-th signal segment, n = 0, 1, 2, ..., L-1, where L is the length of each signal segment; w[n] is the window function; k = 1, 2, 3, ..., K, where K is the total number of vibration signal segments in each channel; The average power spectral density per channel is: In the formula, This represents the average power spectral density per channel.

6. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 1, characterized in that, Based on the power spectral density energy map, the bridge modes were identified, including: The power spectral density energy map is identified using a pre-trained bridge modal recognition model to obtain the bridge modes; The bridge modal recognition model includes convolutional neural networks and recurrent neural networks. Convolutional neural networks are used to extract frequency domain features, while recurrent neural networks are used to capture time-series information in multi-channel vibration signals.

7. The bridge mode identification method based on distributed optical fiber and power spectral density according to claim 6, characterized in that, Before using a pre-trained bridge modal recognition model to identify the power spectral density energy map and obtain the bridge modes, the following steps are also included: Construct a bridge modality recognition model based on deep learning; The bridge mode recognition model was pre-trained using a semi-supervised learning method with power spectral density energy maps that included some bridge mode labels and some that did not. The pre-trained bridge modal recognition model is fine-tuned using a small number of power spectral density energy maps with specific bridge modal labels to obtain a trained bridge modal recognition model.

8. A bridge mode identification system based on distributed optical fiber and power spectral density, characterized in that, include: The acquisition module is used to acquire multi-channel vibration signals collected by the distributed fiber optic sensor system. The analysis module is used to perform frequency domain analysis on multi-channel vibration signals to obtain the power spectral density energy map; The identification module is used to identify the bridge modes based on the power spectral density energy map.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the bridge mode identification method based on distributed optical fiber and power spectral density as described in any one of claims 1-7.

Images