Method for remotely inverting significant wave height from seafloor acoustic sensors

Abstract

The application discloses a method for inversing effective wave height of sea wave by seabed acoustic sensing, comprising the following steps: obtaining original data of a plurality of DAS sensors and pre-processing to obtain first data; the first data comprises a plurality of DAS0 data containing wave information; performing effective wave height calculation processing on the first data to obtain second data; screening the second data to obtain third data; and inversing modeling of the third data and AWAC effective wave height data by deep learning to directly inverse the effective wave height of the remote sea wave. Based on the high-efficiency nonlinear modeling capability of the deep learning model, the deep learning is used for nonlinear modeling of the DAS and the effective wave height of the remote sea wave, the construction of the traditional physical model is skipped, and the effective wave of the remote sea wave is directly inverted.

Description

Technical Field

[0001] This invention relates to marine engineering technology, specifically to a method for remotely inverting the effective wave height of ocean waves using underwater acoustic sensing. Background Technology

[0002] Significant wave height (SWH) is one of the most commonly used wave parameters in marine engineering and hydrometeorology. It represents the average height of the highest third of waves above the sea surface and is used to describe the overall intensity of ocean waves. Since the human eye can more easily perceive larger waves, SWH is a more accurate representation of wave height as perceived by the human eye. SWH has wide applications and is of great significance for shipping safety assessments, coastal engineering protection, and weather forecasting.

[0003] Current field observations of significant wave height primarily rely on buoys and satellite remote sensing. However, buoys are inherently limited by their high maintenance costs and inability to provide effective observations in extreme weather conditions. Furthermore, the limited number of buoys and their restricted observation range prevent the study of distant wave parameters. Therefore, a new method for significant wave height observation is needed to supplement traditional methods.

[0004] While some progress has been made in retrieving significant wave height from marine environmental noise recorded using Distributed Acoustic Sensing (DAS) technology, certain limitations remain. Existing research relies on empirical formulas to derive significant wave height from noise recorded by submarine optical cables and observation data from nearby buoys. For retrieving significant wave height from distant ocean waves, these empirical methods are also used without further advancements. Based on previous research, DAS technology can observe signals related to distant storms and submarine earthquakes, demonstrating its potential for retrieving distant signals. However, the nonlinear relationship between distant wave signals and signals recorded by DAS technology remains unexplored. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for remote inversion of the significant wave height of ocean waves using underwater acoustic sensing, so as to utilize DAS to invert the significant wave height of ocean waves at a distance.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention provides a method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing, comprising:

[0008] Raw data from multiple DAS sensors is acquired and preprocessed to obtain first data; the first data includes multiple DAS0 data containing wave information.

[0009] The first data is processed by performing a type of significant wave height calculation to obtain the second data;

[0010] The second data is filtered to obtain the third data;

[0011] Deep learning was used to invert and model the third data and AWAC significant wave height data to directly invert the significant wave height of distant ocean waves.

[0012] Optionally, the raw data of each DAS sensor is DATA. raw All of them underwent the following preprocessing:

[0013] Original DATA raw Data converted to strain rate data DATA strain rate :

[0014]

[0015] Where λ is the laser wavelength emitted by the DAS system; G is the set calculation reference length; n is the reflectivity of the optical fiber; and r is the axial elastic deformation factor of the optical fiber.

[0016] Optionally, the preprocessing further includes:

[0017] First, all strain rate data are processed through a low-pass filter with a cutoff frequency of 2Hz. Then, according to the Nyquist sampling theorem, one point is collected every 40 time points, and the sampling frequency is reduced from 200Hz to 5Hz. Finally, band-pass filtering of 0.04-0.4Hz is performed to obtain multiple DAS0 data containing wave information.

[0018] Optionally, the step of performing a similar significant wave height calculation on the first data to obtain the second data includes:

[0019] Extract the peaks and troughs from the DAS0 data and calculate the wave height H. j Wave height H arranged from largest to smallest j Take the first third of the wave height and calculate the average value DAS. Hs As shown in equation (2);

[0020]

[0021] DAS Hs The valid DAS data, i.e., the second data, has the same time and frequency as the AWAC valid wave height data.

[0022] Optionally, filtering the second data includes:

[0023] Calculate the DAS of each DAS sensor HsThe Pearson correlation coefficient between the data and the AWAC significant wave height data was compared with a set threshold to filter out multiple DAS sensors whose DAS values ​​exceeded the threshold. Hs Data is used as third-party data.

[0024] Optionally, the step of using deep learning to invert and model the third data and the distant AWAC significant wave height data includes:

[0025] Using the LSTM model in deep learning as a modeling tool, the DAS of multiple DAS sensors was selected. Hs The data is used as input data, and the AWAC significant wave height data is used as the output target to fit the inversion model.

[0026] Optionally, the LSTM model consists of three LSTM layers. The last time step of the last LSTM output passes through a fully connected layer to obtain an output value. The inversion target is the 12th step value of the AWAC significant wave height in time.

[0027] Optionally, the fitting of the inversion model uses the mean squared error (MSE) as the loss function, Adam as the optimization algorithm, a learning rate of 0.0005, 100 epochs, and a batch size of 32.

[0028] In a second aspect, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of any of the methods described above.

[0029] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described above.

[0030] Compared with the prior art, the advantages of this invention are as follows:

[0031] This invention leverages the efficient nonlinear modeling capabilities of deep learning models to perform nonlinear modeling of DAS and distant effective wave heights, bypassing the construction of traditional physical models and directly inverting distant effective waves. Attached Figure Description

[0032] Figure 1 A flowchart illustrating the method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing, as provided in this application embodiment;

[0033] Figure 2 Here is a diagram of the LSTM model structure;

[0034] Figure 3Divide the AWAC significant wave height time series and dataset;

[0035] Figure 4 For DAS Hs Pearson correlation coefficient with AWAC significant wave height;

[0036] Figure 5 The effective wave height inversion results for the training set;

[0037] Figure 6 For the test set, the effective wave height inversion results are provided.

[0038] Figure 7 This is a schematic diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0039] Example:

[0040] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] See Figure 1 As shown in this embodiment, the method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing mainly includes the following steps:

[0042] 110. Acquire raw data from multiple DAS sensors and preprocess it to obtain first data; the first data includes multiple DAS0 data containing wave information.

[0043] In practice, theoretically, DAS technology allows each point on the optical fiber to become an independent sensor. However, due to hardware limitations, currently one sensor station is placed every 4 meters along the fiber. With a fiber length of approximately 30 km, this translates to a total of 7335 sensors.

[0044] Each DAS sensor directly outputs a 200Hz optical phase signal DATA. raw To accurately obtain DAS0 data containing wave information, it needs to be converted into strain rate DATA. strainrate Based on the empirical relationship between optical phase and fiber strain rate, i.e.

[0045]

[0046] Among them, DATA strainrate Strain rate (nm / m / s); λ = 1.550 × 10⁻⁶ -6 m is the laser wavelength emitted by the DAS system; G = 10m is the manually set calculation reference length; n = 1.467 is the reflectivity of the optical fiber; r = 0.78 is the axial elastic deformation factor of the optical fiber.

[0047] To improve data processing efficiency and filter out other interference noise, the strain data was downsampled, with the DAS data frequency reduced from 200Hz to 5Hz. This frequency is sufficient to observe a complete wave cycle. The specific steps are as follows: 1. The strain rate data is processed by a low-pass filter with a cutoff frequency of 2Hz; 2. According to the Nyquist sampling theorem, one point is collected every 40 time points, and the sampling frequency is reduced from 200Hz to 5Hz.

[0048] The wave signals that DAS can sense are mainly dynamic water pressure signals of seawater, whose magnitude attenuates with depth, and whose frequency band is generally between 0.04-0.4Hz. Therefore, by performing bandpass filtering on the above-processed data at 0.04-0.4Hz, the DAS0 signal containing wave information, i.e., the first data, is obtained.

[0049] In this way, the DAS0 signal containing wave information can be obtained accurately and efficiently through the above operations.

[0050] 120. Perform a significant wave height calculation on the first data to obtain the second data;

[0051] In the specific implementation, to ensure a one-to-one correspondence between the DAS data and the significant wave height data from the long-range Wave Dragon Profiling Wave Meter (AWAC) in terms of time frequency, DAS0 is processed using a method similar to significant wave height calculation. Wave crests and troughs are extracted from DAS0, and the wave height H is calculated. j Wave height H arranged from largest to smallest j Take the first third of the wave height and calculate the average value DAS. Hs As shown in equation (1);

[0052]

[0053] DAS Hs The valid DAS data, i.e., the second data, has the same time and frequency as the AWAC valid wave height data.

[0054] 130. Filter the second data to obtain the third data;

[0055] Due to DAS Hs The data is two-dimensional, composed of time-series data from multiple sensors (channels). Due to the coupling between the DAS fiber optic cable and the seabed sediment, as well as the presence of knots in the cable, the sensing capability of the DAS sensor becomes extremely weak. Therefore, it is necessary to further improve the DAS... Hs Data quality control requires screening.

[0056] Calculate the DAS of each DAS sensor HsPearson correlation coefficient between the signal and long-range AWAC (Acoustic Wave and Current) significant wave height data. A threshold is set to select sensors (DAS) with high correlation to the observed data. Hs Data is used to ensure the accuracy of subsequent inversion results.

[0057] 140. Deep learning was used to invert and model the third data and the distant AWAC significant wave height data to directly invert the distant significant wave height.

[0058] In the specific implementation, the LSTM model from deep learning is used as the modeling tool to process the selected DAS. Hs Data was used for inversion modeling with distant AWAC significant wave height data. Multiple DAS Hs As input data, DAS from multiple sensors Hs The input data is considered as multiple dimensions; simultaneously, the input data needs to be processed into a 24-time-step format to conform to the model's input format; the distant AWAC significant wave height data is used as the output target. The LSTM model consists of three LSTM layers. The output of the last LSTM layer at the last time step passes through a fully connected layer to obtain a final output value; the inversion target is the temporally corresponding AWAC significant wave height at the 12th time step. The model structure is as follows: Figure 2 As shown

[0059] Therefore, this method is based on the efficient nonlinear modeling capability of deep learning models. It uses deep learning to perform nonlinear modeling of DAS and distant significant wave height, skips the construction of traditional physical models, and directly inverts the distant significant wave height, thereby solving the problem.

[0060] The following application scenario example will be used to further verify and illustrate this method:

[0061] The distributed acoustic sensing submarine fiber optic cable is approximately 30 km long. Dividing the fiber optic cable along the seabed into evenly spaced intervals of 4, it creates 7335 sensor stations (also called channels). This existing submarine fiber optic cable can be connected to a DAS interferometer at the fiber optic interface of the shore-based control center, constructing a DAS system and transforming the submarine fiber optic cable into an observation sensor. The rise and fall of ocean waves creates dynamic water pressure, acting on the seabed and generating micro-vibrations. The DAS is sensitive to these weak vibrations and can record these micro-vibration signals. These micro-vibration signals contain a large amount of wave-related signals, which can be used to infer wave parameters.

[0062] First, the raw data from DAS observations is an optical phase signal with a frequency of 200 Hz, which needs to be converted into a strain rate signal using empirical formula (1), with units of nm / m / s. The signals recorded by DAS on the seabed are not only wave-related signals, but also various other noises. In order to remove other noise signals from the data, the strain rate data is processed by a low-pass Butterworth filter with a cutoff frequency of 2 Hz (order 4); then, according to the Nyquist sampling theorem, one point is collected every 40 time points to reduce the data frequency to 5 Hz; finally, the above data is filtered by a Butterworth bandpass filter with a cutoff frequency of 0.04-0.4 Hz (order 4) to obtain the signal DAS0, which theoretically contains only wave information.

[0063] Before calculating the effective DAS, it is necessary to clarify the attributes of the significant wave height data of the target AWAC. In this example, the AWAC is more than 10km away from the submarine optical cable, and the significant wave height data spans one month. Figure 3 As shown, the dataset is divided into two parts: the first part is the training set, and the second part is the test set. Meanwhile, any missing parts are supplemented using cubic spline interpolation. AWAC significant wave height is calculated by taking the wave heights observed within 15 minutes after a specified time, and then calculating the significant wave height, which is recorded every half hour. Therefore, the calculation of significant DAS also needs to be consistent with AWAC in terms of time span, recording frequency, and dataset allocation. Using the preprocessed DAS0 data, the significant wave height is calculated according to formula (2) to obtain the DAS. Hs Then calculate the DAS of these channels. Hs The Pearson correlation coefficient between the DAS and AWAC significant wave height data is used, with a correlation coefficient greater than 0.5 being considered. Hs Data is used as model input. Correspondingly, AWAC significant wave height data is used as model output data. This example randomly selects DAS data from 200 sensors (channels). Hs A reverse simulation demonstration was performed. The DAS (Data Acoustics System) of 200 sensors was also shown. Hs The Pearson coefficient of the AWAC significant wave height, such as Figure 4 As shown.

[0064] Finally, the above-mentioned DAS Hs Inputting AWAC significant wave height data into Figure 2 In the deep learning model, the training set is used to fit the inversion model. The loss function is mean squared error (MSE), the optimization algorithm is Adam, the learning rate is 0.0005, the epoch is 100, and the batch size is 32. The training results are as follows: Figure 5 The yellow line represents the observed true values, and the blue line represents the inverted values ​​from the model. The mean squared error (MSE) is 0.0019m. 2The mean absolute error (MAE) is 0.0318m, and the coefficient of determination (R²) is... 2 The value is 0.9009. Correspondingly, the test set results are as follows: Figure 6 The mean square error (MSE) is 0.0056m. 2 The mean absolute error (MAE) is 0.0547m, and the coefficient of determination (R²) is... 2 The error value is 0.6835. These error values ​​are acceptable in practical applications, and the inversion is feasible.

[0065] like Figure 7 As shown, this application embodiment provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114. The processor 111, communication interface 112, and memory 113 communicate with each other via the communication bus 114. The memory 113 stores a computer program. When the processor 111 executes the program stored in the memory 113, it implements the steps of the intelligent patrol behavior detection method based on spatiotemporal action detection provided in any of the aforementioned method embodiments. This application embodiment also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as provided in any of the aforementioned method embodiments.

[0066] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0067] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the content of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing, characterized in that, include: The raw data from multiple DAS sensors is acquired and preprocessed to obtain the first data. The first data includes multiple DAS0 data sets containing wave information; The first data is processed by performing a type of significant wave height calculation to obtain the second data; The second data is filtered to obtain the third data; Deep learning was used to invert and model the third data and AWAC significant wave height data to directly invert the significant wave height of distant ocean waves.

2. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 1, characterized in that, The raw data for each DAS sensor is DATA raw All of them underwent the following preprocessing: Original DATA raw Data converted to strain rate data DATA strainrate : Where λ is the laser wavelength emitted by the DAS system; G is the set calculation reference length; n is the reflectivity of the optical fiber; and r is the axial elastic deformation factor of the optical fiber.

3. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 2, characterized in that, The preprocessing also includes: First, all strain rate data are processed through a low-pass filter with a cutoff frequency of 2Hz. Then, according to the Nyquist sampling theorem, one point is collected every 40 time points, and the sampling frequency is reduced from 200Hz to 5Hz. Finally, a band-pass filter of 0.04-0.4Hz is applied to obtain multiple DAS0 data containing wave information as the first data.

4. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 1, characterized in that, The first data is processed by performing a type of significant wave height calculation to obtain the second data; include: Extract the peaks and troughs from the DAS0 data and calculate the wave height H. j Wave height H arranged from largest to smallest j Take the first third of the wave height and calculate the average value DAS. Hs As shown in equation (2); DAS Hs The valid DAS data, i.e., the second data, has the same time and frequency as the AWAC valid wave height data.

5. The method for remote inversion of significant wave height using underwater acoustic sensing as described in claim 1 or 4, characterized in that, The filtering of the second data includes: Calculate the DAS of each DAS sensor Hs The Pearson correlation coefficient between the data and the AWAC significant wave height data was compared with a set threshold to filter out multiple DAS sensors whose DAS values ​​exceeded the threshold. Hs Data is used as third-party data.

6. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 1, characterized in that, The method of using deep learning to invert and model the third data and the distant AWAC significant wave height data includes: Using the LSTM model in deep learning as a modeling tool, the DAS of multiple DAS sensors was selected. Hs The data is used as input data, and the AWAC significant wave height data is used as the output target to fit the inversion model.

7. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 6, characterized in that, The LSTM model consists of three LSTM layers. The last time step of the last LSTM layer is passed through a fully connected layer to obtain an output value. The inversion target is the 12th step value of the AWAC significant wave height in time.

8. The method for remotely inverting the significant wave height of ocean waves using underwater acoustic sensing as described in claim 6, characterized in that, The fitting of the inversion model uses the mean squared error (MSE) as the loss function, Adam as the optimization algorithm, a learning rate of 0.0005, 100 epochs, and a batch size of 32.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 8.

Images