A multiple longitudinal and transverse wave combined detection method based on a seabed crawler

Abstract

The application provides a multiple longitudinal and transverse wave joint detection method based on a seabed crawler, and belongs to the technical field of geophysical detection.The original longitudinal and transverse wave signal matrix is constructed by emitting multi-frequency longitudinal and transverse wave signals and receiving stratum reflection signals.Frequency characteristics are extracted by using Fourier transform, physical analysis is carried out in combination with wave equations, a longitudinal and transverse wave interference matrix is constructed and weighted fusion is carried out.An error compensation matrix is constructed according to the attitude data of the crawler, and interference compensation is carried out on the joint interference matrix.Time-frequency analysis is carried out through wavelet transform, specific time-frequency characteristics are extracted, a pre-trained stratum classification prediction model is input, anti-interference prediction of the seabed stratum structure type and thickness is realized, and finally, a seabed three-dimensional geological profile is drawn according to the prediction result, key geological structures and resource distribution areas are marked, and the technical problems that there are many external interferences in the seabed geological exploration process and the accurate identification of seabed stratum structure and the positioning of resource distribution are influenced are solved.

Description

Technical Field

[0001] This invention belongs to the field of geophysical exploration technology, and specifically relates to a multi-wave and transverse wave joint detection method based on a seabed crawler. Background Technology

[0002] Seafloor geological exploration is a crucial prerequisite for marine resource development and submarine engineering construction. Traditional seafloor geological exploration primarily relies on shipborne sonar, seismic waves, or single-type waves for detection. These methods typically employ a single wave source at a fixed frequency to transmit detection signals downwards from the water surface, analyzing the seafloor strata structure through the echo signals. In practical applications, these technologies have been widely used in offshore oil exploration, submarine cable laying route planning, and deep-sea mineral resource surveys.

[0003] However, traditional seabed exploration methods face serious interference problems. First, signals emitted from the surface must travel through the entire body of water to reach the seabed, and are subject to interference from factors such as waves, currents, and thermoclines, resulting in severe signal attenuation and distortion. Second, single-type wave sources (such as pure P-waves or pure S-waves) have limited ability to distinguish different geological structures, and the detection results for complex geological structures are easily affected by ambient noise. Third, fixed-site or shipborne detection methods are affected by factors such as ship movement and positioning errors, making it difficult to obtain stable and continuous detection data.

[0004] These various interfering factors severely affect the accurate identification of seabed stratigraphic structures and the precise location of resource distribution. This is particularly true in areas with complex geological structures and variable sea conditions, where traditional detection methods struggle to eliminate the impact of various interferences on detection accuracy, thus limiting the effective exploration and development of seabed resources. In other words, existing technologies face the technical challenge of numerous external interferences during seabed geological exploration, affecting the accurate identification of seabed stratigraphic structures and the location of resource distribution. Summary of the Invention

[0005] In view of this, the present invention provides a multi-wave and transverse wave joint detection method based on a seabed crawler, which can solve the technical problem in the prior art that there are many external interferences during the seabed geological exploration process, which affect the accurate identification of seabed strata structure and the location of resource distribution.

[0006] This invention is implemented as follows: It provides a method for joint detection of multiple P-waves and S-waves based on a seabed crawler, comprising: deploying a joint detection device for multiple P-waves and S-waves on a seabed crawler, enabling the crawler to crawl along a preset detection path and emit multi-frequency P-wave and S-wave signals; receiving P-wave and S-wave signals reflected from the seabed strata and constructing an original P-wave and S-wave signal matrix; extracting signal frequency features and constructing a P-wave and S-wave interference matrix; weighted fusing the P-wave interference matrix and the S-wave interference matrix to generate a joint interference matrix; constructing an interference error compensation matrix based on real-time attitude data recorded by the seabed crawler's attitude sensor and compensating the joint interference matrix; performing time-frequency analysis through wavelet transform to extract specific time-frequency features; inputting the specific time-frequency features into a pre-trained stratigraphic classification and prediction model to achieve classification of seabed stratigraphic structure types and thickness prediction; and drawing a three-dimensional geological profile of the seabed, marking key geological structures and resource distribution areas.

[0007] The multi-frequency P-wave and S-wave joint detection device consists of a multi-frequency wave source generator, a P-wave transmitting unit, a S-wave transmitting unit, an omnidirectional receiving sensor array, a signal processing module, and a data storage unit. It is installed on the bottom of the seabed crawler and can simultaneously transmit multiple frequencies of P-waves and S-waves and receive the reflected signals.

[0008] Among them, the P-wave interference matrix refers to the matrix constructed by extracting the amplitude and phase of each frequency component after spectral analysis of the original P-wave signal matrix, which represents the interference characteristics generated by the P-wave propagating in different geological structures; the S-wave interference matrix refers to the matrix constructed by extracting the amplitude and phase of each frequency component after spectral analysis of the original S-wave signal matrix, which represents the interference characteristics generated by the S-wave propagating in different geological structures; the joint interference matrix refers to the matrix formed by combining the P-wave interference matrix and the S-wave interference matrix through a weighted fusion algorithm, which comprehensively reflects the comprehensive interference characteristics of the P-wave and S-wave propagating in the seabed strata.

[0009] Among them, the interference error compensation matrix refers to the compensation matrix calculated based on the pitch angle, roll angle and depth data recorded by the attitude sensor of the seabed crawler, which is used to eliminate the interference caused by the attitude change of the seabed crawler to signal acquisition.

[0010] Among them, the unique time-frequency features refer to the time-domain and frequency-domain joint features extracted after performing time-frequency analysis on the joint interference matrix through wavelet transform, which can reflect the unique response characteristics of different geological structures.

[0011] The submarine strata wave equation is a partial differential equation describing the propagation characteristics of P-waves and S-waves in the submarine strata. It is used to calculate the propagation speed and attenuation characteristics of P-waves and S-waves in different geological structures. The inputs to the submarine strata wave equation include the strata density obtained from the seabed crawler, the first Lamé parameter obtained from rock sampling analysis of the seabed crawler, the second Lamé parameter obtained from rock sampling analysis of the seabed crawler, and the displacement vector calculated from the original P-wave signal matrix and the original S-wave signal matrix. The outputs are the wave displacement field distribution and the wave energy attenuation law, which are used to construct the P-wave interference matrix and the S-wave interference matrix.

[0012] Among them, the layer classification prediction model refers to the neural network model used for stratigraphic structure identification and thickness prediction; the specific structure of the stratigraphic classification prediction model is a U-shaped convolutional neural network based on the self-attention mechanism, which includes an encoder and a decoder; the encoder consists of five convolutional layers and four pooling layers, with the number of convolutional kernels in each layer being 32, 64, 128, 256, and 512, respectively, and the kernel size being 3×3; the decoder consists of five deconvolutional layers and four concatenation layers, with the number of feature maps in each layer being the same as the corresponding layer in the encoder.

[0013] Among them, a self-attention mechanism is introduced between the encoder and the decoder to enhance the feature extraction capability. The sparse attention coefficient in the self-attention mechanism is dynamically adjusted according to three key parameters: the ratio of the propagation velocities of the P-wave and S-wave calculated from the P-wave interference matrix and the S-wave interference matrix, the formation density obtained from the seabed crawler, and the seabed depth obtained from the seabed crawler.

[0014] The steps involved in establishing the training dataset during the training process of the stratigraphic classification prediction model include collecting P-wave and S-wave reflection data of seafloor strata with various geological structures from different sea areas, preprocessing the collected data to remove noise and outliers, determining the stratigraphic type and thickness using manual annotation, and expanding the number of training samples through data augmentation techniques.

[0015] The training steps for the stratigraphic classification prediction model include: first, initializing the model using a preprocessed training dataset; then, optimizing the parameters using the Adam optimizer with an initial learning rate of 0.001, which decays to 0.1 times every 50 training epochs; and finally, using a weighted combination of stratigraphic type classification cross-entropy loss and thickness prediction mean square error loss, with weights of 0.6 and 0.4, respectively.

[0016] Compared with existing technologies, this invention provides a multi-frequency P-wave and S-wave joint detection method based on a seabed crawler. This method achieves direct detection close to the seabed by deploying a detection device on the seabed crawler capable of simultaneously transmitting and receiving multiple frequencies of P-waves and S-waves. The method utilizes the seabed crawler to crawl along a predetermined path, continuously transmitting multiple frequencies of P-wave and S-wave signals, and receiving signals reflected from the strata to construct an original P-wave and S-wave signal matrix.

[0017] Frequency features were extracted using Fourier transform, and combined with the physical mechanism analysis of the seafloor strata wave equation, P-wave and S-wave interference matrices were constructed and weighted and fused. Simultaneously, an interference error compensation matrix was constructed based on real-time attitude data recorded by the crawler's attitude sensor, effectively eliminating signal interference introduced by crawler attitude changes. Wavelet transform was used to perform time-frequency analysis on the compensated joint interference matrix, extracting specific time-frequency features. Combined with a pre-trained strata classification prediction model, accurate prediction of seafloor strata structure type and thickness was achieved.

[0018] This invention effectively overcomes various interference factors in traditional exploration, such as water propagation interference, environmental noise interference, and equipment attitude interference, through multi-wave and transverse wave joint detection and multi-level interference compensation technology. This significantly improves the anti-interference capability and detection accuracy of seabed geological exploration. In particular, the stratigraphic classification and prediction model combined with deep learning methods can achieve stable and reliable stratigraphic structure identification and resource distribution location in complex and variable seabed environments. This solves the technical problem in existing technologies where numerous external interferences affect the accurate identification of seabed stratigraphic structures and resource distribution location during seabed geological exploration, providing more accurate technical support for the efficient development of seabed resources. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a schematic diagram of the multiple longitudinal and transverse wave joint detection device in Embodiment 1 of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0022] like Figure 1 The diagram shown is a flowchart of a multi-wave (P-wave and S-wave) joint detection method based on a seabed crawler provided by the present invention. This method includes the following steps:

[0023] S01. Deploy a multi-wave and transverse wave joint detection device on the seabed crawler, so that the seabed crawler crawls along a preset detection path, and at the same time activate the multi-wave and transverse wave joint detection device to transmit multi-frequency wave and transverse wave signals to the seabed strata.

[0024] S02. The multi-wave P-wave and S-wave combined detection device is used to receive the P-wave and S-wave signals reflected from the seabed strata, convert them into digital signals, and construct the original P-wave signal matrix and the original S-wave signal matrix.

[0025] S03. Extract the frequency characteristics of the original P-wave signal matrix and the original S-wave signal matrix through Fourier transform, construct the P-wave interference matrix and the S-wave interference matrix, and use the seafloor strata wave equation to analyze the physical mechanism of P-wave and S-wave propagation.

[0026] S04. The P-wave interference matrix and the S-wave interference matrix are weighted and fused to generate a joint interference matrix. The weighting coefficients are dynamically adjusted based on the seabed geological conditions.

[0027] S05. Based on the real-time attitude data recorded by the attitude sensor of the seabed crawler, construct an interference error compensation matrix and perform compensation processing on the joint interference matrix.

[0028] S06. Perform time-frequency analysis on the compensated joint interference matrix using wavelet transform, extract specific time-frequency features, and establish a submarine stratigraphic structure feature recognition model.

[0029] S07. Input the specific time-frequency features into the pre-trained stratigraphic classification and prediction model to achieve classification of seafloor stratigraphic structure types and thickness prediction.

[0030] S08. Based on the geological structure type and the thickness prediction results, draw a three-dimensional geological profile of the seabed and mark the key geological structures and resource distribution areas.

[0031] The multi-frequency P-wave and S-wave joint detection device consists of a multi-frequency wave source generator, a P-wave transmitting unit, a S-wave transmitting unit, an omnidirectional receiving sensor array, a signal processing module, and a data storage unit. It is installed on the bottom of the seabed crawler and can simultaneously transmit multiple frequencies of P-waves and S-waves and receive the reflected signals.

[0032] Specifically, the longitudinal wave interference matrix refers to the matrix constructed by extracting the amplitude and phase of each frequency component after performing spectral analysis on the original longitudinal wave signal matrix, thus characterizing the interference generated when the longitudinal wave propagates in different geological structures.

[0033] Specifically, the transverse wave interference matrix refers to the matrix constructed by extracting the amplitude and phase of each frequency component after performing spectral analysis on the original transverse wave signal matrix, thus characterizing the interference generated when the transverse wave propagates in different geological structures.

[0034] Specifically, the joint interference matrix refers to the matrix formed by combining the P-wave interference matrix and the S-wave interference matrix through a weighted fusion algorithm, which can comprehensively reflect the combined interference characteristics of P-waves and S-waves propagating in the seabed strata.

[0035] Specifically, the interference error compensation matrix refers to the compensation matrix calculated based on the pitch angle, roll angle, and depth data recorded by the attitude sensor of the seabed crawler, used to eliminate the interference caused by the attitude change of the seabed crawler to signal acquisition.

[0036] Specifically, the unique time-frequency features refer to the combined time-domain and frequency-domain features extracted after performing time-frequency analysis on the joint interference matrix through wavelet transform, which can reflect the unique response characteristics of different geological structures.

[0037] Specifically, the seafloor strata wave equation refers to the partial differential equation describing the propagation characteristics of P-waves and S-waves in the seafloor strata. The seafloor strata wave equation is used to calculate the propagation speed and attenuation characteristics of P-waves and S-waves in different geological structures. The inputs include the strata density obtained from the seafloor crawler, the Lamé first parameter obtained from the rock sampling analysis of the seafloor crawler, the Lamé second parameter obtained from the rock sampling analysis of the seafloor crawler, and the displacement vector calculated from the original P-wave signal matrix and the original S-wave signal matrix. The outputs are the wave displacement field distribution and the wave energy attenuation law. The wave displacement field distribution and the wave energy attenuation law are used to construct the P-wave interference matrix and the S-wave interference matrix.

[0038] Specifically, the stratigraphic classification prediction model refers to a neural network model used for stratigraphic structure identification and thickness prediction. The specific structure of the stratigraphic classification prediction model is a U-shaped convolutional neural network based on a self-attention mechanism, comprising an encoder and a decoder. The encoder consists of five convolutional layers and four pooling layers, with the number of convolutional kernels in each layer being 32, 64, 128, 256, and 512, respectively, and the kernel size being 3×3. The decoder consists of five deconvolutional layers and four concatenation layers, with the number of feature maps in each layer being the same as the corresponding layer in the encoder. A self-attention mechanism is introduced between the encoder and the decoder to enhance feature extraction capabilities. The sparse attention coefficient in the self-attention mechanism is dynamically adjusted based on three key parameters: the ratio of the propagation velocities of the P-wave and S-wave calculated from the P-wave interference matrix and the S-wave interference matrix, the stratigraphic density obtained from the seabed crawler, and the seabed depth obtained from the seabed crawler.

[0039] The steps for establishing the training dataset in the training process of the stratigraphic classification prediction model specifically include collecting P-wave and S-wave reflection data of seafloor strata with various geological structures from different sea areas, preprocessing the collected data to remove noise and outliers, determining the stratigraphic type and thickness using manual annotation, expanding the number of training samples through data augmentation techniques, including adding different levels of Gaussian noise, random pruning, and random rotation operations, and finally constructing a training set containing 100,000 sets of labeled data and a validation set containing 10,000 sets of labeled data.

[0040] The specific steps for training the stratigraphic classification prediction model include: first, initializing the model using a preprocessed training dataset; optimizing parameters using the Adam optimizer with an initial learning rate of 0.001, which decays to 0.1 times every 50 training epochs; using a weighted combination of stratigraphic type classification cross-entropy loss and thickness prediction mean square error loss with weights of 0.6 and 0.4, respectively; introducing an early stopping mechanism to prevent overfitting during training, stopping training when the validation set loss does not decrease for 10 consecutive training epochs; and finally selecting the model parameters with the best performance on the validation set as the final model.

[0041] The specific implementation methods of the above steps are described in detail below.

[0042] The specific implementation of step S01 involves installing a multi-frequency P-wave and S-wave combined detection device on the bottom of a seabed crawler, using a hydraulic drive mechanism to control the crawler along a preset detection path. The preset detection path is pre-planned based on the geological characteristics of the target area, typically using a grid-like or spiral path to cover the target area, with a grid spacing of 5–10 meters. The multi-frequency wave source generator in the multi-frequency P-wave and S-wave combined detection device generates excitation signals in the range of 1–500 Hz. These excitation signals are amplified by a power amplifier and then transmitted to the P-wave transmitting unit and the S-wave transmitting unit respectively. The P-wave transmitting unit is composed of a piezoelectric ceramic transducer array, converting electrical signals into mechanical vibration energy; the S-wave transmitting unit is made of shear-type piezoelectric material, capable of generating transverse vibrations perpendicular to the propagation direction. Both wave source units operate simultaneously, emitting multi-frequency P-wave and S-wave signals into the seabed strata, penetrating to a depth of up to 100 meters below the seabed. The purpose of this step is to achieve a comprehensive scanning detection of the seabed strata, providing raw signal data for subsequent analysis.

[0043] The specific implementation of step S02 involves using an omnidirectional receiving sensor array installed on the bottom of the seabed crawler to receive P-wave and S-wave signals reflected from the seabed strata. This sensor array consists of 32 triaxial accelerometers arranged in a 4×8 matrix, capable of simultaneously capturing P-wave and S-wave reflection signals in three-dimensional space. The received analog signals are amplified by a preamplifier and then converted into digital signals at a sampling rate of 10 kHz by a 24-bit high-precision analog-to-digital converter. The digital signals undergo preliminary filtering by a data processing module to remove 50 Hz power frequency interference and high-frequency noise above 500 Hz. The processed data is rearranged according to the time sequence and sensor position information to construct the original P-wave signal matrix and the original S-wave signal matrix. The original P-wave signal matrix has a dimension of m×n, where m is the number of sensors and n is the number of time sampling points; the original S-wave signal matrix consists of two m×n matrixes, corresponding to the vibration components of the S-wave in two orthogonal directions. The purpose of this step is to obtain high-quality original reflection signal data, laying the foundation for subsequent feature extraction.

[0044] The specific implementation of step S03 involves performing Fourier transform processing on the original P-wave signal matrix and the original S-wave signal matrix. First, a Fast Fourier Transform is performed on the time-series signals recorded by each sensor to extract frequency components within the range of 1–500 Hz, obtaining spectral amplitude and phase information. For the P-wave signal, the propagation characteristics of the P-wave at different frequencies are extracted to construct a P-wave interference matrix; for the S-wave signal, spectral analysis is performed on the S-wave components in two orthogonal directions to construct a S-wave interference matrix. Simultaneously, the physical mechanism of P-wave and S-wave propagation is analyzed based on the seafloor strata wave equation. This wave equation considers physical characteristics such as stratum density, Lamé first parameter, and Lamé second parameter to calculate the propagation speed and attenuation characteristics of P-wave and S-wave in different geological structures. P-wave velocity v p With transverse wave velocity v s pass and The calculations yielded λ, representing the first Lamé parameter, μ, representing the second Lamé parameter, and ρ, representing the formation density. The wave equation was numerically solved using the finite-difference time-domain method, with a grid spacing of one-tenth of the propagation wavelength and a time step determined based on the Coulant stability condition. The wave displacement field distribution and wave energy attenuation law were obtained by solving the wave equation, which were used to construct the P-wave and S-wave interference matrices. The purpose of this step was to extract the characteristic information of P-wave and S-wave propagation in the seafloor strata through frequency domain analysis, providing a basis for subsequent signal fusion.

[0045] The specific implementation of step S04 involves weighted fusion of the P-wave interference matrix and the S-wave interference matrix to generate a joint interference matrix. The weighted fusion employs an adaptive weighting algorithm, with the weighting coefficients dynamically adjusted based on seabed geological conditions. First, the signal-to-noise ratio (SNR) of the P-wave and S-wave under the current seabed geological conditions is calculated, denoted as SNR P-wave and SNR S-wave, respectively. Then, the initial weighting coefficients α and β are determined based on the SNR, satisfying α + β = 1. When SNR P-wave > SNR S-wave, the value of α is increased; conversely, the value of β is increased. In practical applications, the value of α is generally adjusted within the range of 0.3 to 0.7. Furthermore, the influence of stratigraphic type on the weighting is considered: for hard strata such as basalt, where P-wave penetration is strong, the value of α can be set to 0.6–0.7; for soft strata such as argillaceous sediments, where S-waves are more sensitive to interface responses, the value of β can be set to 0.6–0.7. The formula for calculating the joint interference matrix is: Joint interference matrix = α × P-wave interference matrix + β × S-wave interference matrix. The purpose of this step is to comprehensively utilize the complementary advantages of P-waves and S-waves to improve the accuracy of stratigraphic identification.

[0046] The specific implementation of step S05 involves constructing an interference error compensation matrix based on the real-time attitude data recorded by the seabed crawler's attitude sensors, and then compensating for the joint interference matrix. The seabed crawler's attitude sensors include a high-precision gyroscope, accelerometer, and depth sensor, which record the crawler's pitch, roll, and depth data in real time, with a sampling frequency of 100 Hz. First, a coordinate transformation model is established to transform the sensor's measurement coordinate system to a global reference coordinate system, eliminating signal distortion caused by changes in the crawler's attitude. The coordinate transformation is implemented using a rotation matrix, calculated based on Euler angles. Additional compensation is required when the crawler's pitch or roll angle exceeds 5 degrees. The constructed interference error compensation matrix has the same dimension as the joint interference matrix, and its element values ​​are calculated non-linearly based on the attitude deviation angle. The compensation process is implemented using matrix multiplication: the compensated joint interference matrix = interference error compensation matrix × joint interference matrix. The purpose of this step is to eliminate the impact of seabed crawler attitude changes on signal acquisition and improve data quality.

[0047] The specific implementation of step S06 involves performing time-frequency analysis on the compensated joint interference matrix using wavelet transform to extract specific time-frequency features. First, a suitable wavelet basis function for stratigraphic analysis is selected, typically the db4 or db6 wavelet from the Daubechies wavelet family, due to its good time-frequency localization characteristics. Multi-scale wavelet decomposition is performed on each row of data in the joint interference matrix, with five decomposition levels, yielding wavelet coefficients for different frequency bands. The energy distribution of the wavelet coefficients at each level is calculated, constructing a time-frequency energy map to reflect the energy distribution characteristics of the signal at different times and frequencies. Based on the time-frequency energy map, energy peak points and their distribution characteristics are extracted as specific time-frequency features. Furthermore, statistical characteristics of the wavelet coefficients in each frequency band are calculated, including mean, variance, skewness, and kurtosis, as supplementary features to the feature vector. Based on prior geological knowledge, the feature vectors are screened and combined, retaining features that contribute significantly to stratigraphic structure identification and removing redundant features. The final feature vector has a dimension of 128, including time-domain features, frequency-domain features, and joint time-frequency features. The purpose of this step is to extract key features that reflect the characteristics of the stratigraphic structure through time-frequency analysis, so as to provide a basis for subsequent stratigraphic classification and thickness prediction.

[0048] The specific implementation of step S07 involves inputting specific time-frequency features into a pre-trained stratigraphic classification and prediction model to achieve classification of seafloor stratigraphic structure types and thickness prediction. The stratigraphic classification and prediction model is based on a U-shaped convolutional neural network with a self-attention mechanism, comprising an encoder and a decoder. The encoder reshapes the 128-dimensional feature vector into a 16×8 feature map, and then extracts features through five convolutional layers and four pooling layers. The convolutional layers use 3×3 convolutional kernels with the following kernel numbers: 32, 64, 128, 256, and 512. Each convolutional layer is followed by batch normalization and a ReLU activation function. The pooling layers use 2×2 max pooling. The decoder reconstructs features through five deconvolutional layers and four concatenation layers. The deconvolutional layers use 3×3 convolutional kernels, and the number of feature maps is the same as the corresponding layers in the encoder. A self-attention mechanism is introduced between the encoder and decoder to enhance feature extraction capabilities. The sparse attention coefficients in the self-attention mechanism are dynamically adjusted based on three key parameters: the ratio of P-wave to S-wave propagation velocities, stratigraphic density, and seafloor depth. The model output includes the probability distribution of stratigraphic types and predicted thickness values. Stratigraphic types are divided into eight categories: sandy sedimentary layers, muddy sedimentary layers, carbonate rock layers, volcanic rock layers, metamorphic rock layers, gas-permeable zones, liquid-permeable zones, and solid mineral enrichment zones. The thickness prediction accuracy can reach over 90% of the actual thickness. The purpose of this step is to utilize deep learning technology to achieve automatic identification and parameter prediction of seafloor stratigraphic structures.

[0049] Step S08 is implemented by drawing a three-dimensional geological profile of the seabed based on the predicted stratigraphic structure type and thickness, and marking key geological structures and resource distribution areas. First, the predicted results are reconstructed into three-dimensional spatial distribution data according to the crawler path information. Kriging interpolation is used to estimate unexplored areas, generating a continuous three-dimensional stratigraphic distribution model. Then, different colors and textures are set according to different stratigraphic types, and three-dimensional visualization rendering is achieved using the OpenGL graphics library. Key geological structures, including faults, folds, and intrusions, are marked on the three-dimensional profile. Simultaneously, resource distribution areas are predicted based on stratigraphic characteristics, mainly including oil and gas resources, natural gas hydrates, and seafloor hydrothermal deposits. The marking process combines automatic identification with manual verification. Areas with a predicted probability exceeding 85% are automatically marked, while areas with a probability between 60% and 85% require manual confirmation before marking. The completed three-dimensional geological profile is displayed interactively, supporting rotation, scaling, and slicing operations for intuitive analysis of geological structures. The purpose of this step is to present the exploration results in an intuitive form, providing a basis for decision-making in resource exploration and development.

[0050] Furthermore, the detailed structure of the stratigraphic classification prediction model and the specific implementation steps for constructing the training dataset are as follows: The stratigraphic classification prediction model adopts a U-shaped convolutional neural network structure based on the self-attention mechanism, which can effectively process multi-scale information of stratigraphic features. The encoder consists of five convolutional layers and four pooling layers. The first convolutional layer uses 32 3×3 kernels with a stride of 1, taking a 16×8×1 feature map as input and outputting a 16×8×32 feature map. The second convolutional layer uses 64 3×3 kernels, taking an 8×4×32 feature map as input and outputting an 8×4×64 feature map. The third convolutional layer uses 128 3×3 kernels, taking a 4×2×64 feature map as input and outputting a 4×2×128 feature map. The fourth convolutional layer uses 256 3×3 kernels, taking a 2×1×128 feature map as input and outputting a 2×1×256 feature map. The fifth convolutional layer uses 512 3×3 kernels, taking a 1×1×256 feature map as input and outputting a 1×1×512 feature map. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function to enhance the network's expressive power and prevent gradient vanishing. The pooling layer uses 2×2 max pooling with a stride of 2 to gradually reduce the spatial dimension of the feature maps. The decoder consists of five deconvolutional layers and four concatenation layers. The deconvolutional layers use transposed convolution to upsample the feature maps, with the number of feature maps being the same as the corresponding layers in the encoder. The concatenation layers use skip connections to concatenate the feature maps of the corresponding encoder layers with those of the decoder, preserving high-resolution details. A self-attention mechanism is located between the encoder and decoder, enhancing the model's ability to capture long-distance dependencies by calculating the correlation between different locations in the feature maps. The attention matrix is ​​calculated as: Attention = softmax(Q×K / √d)×V, where Q, K, and V are the query matrix, key matrix, and value matrix, respectively, and d is the feature dimension. The sparse attention coefficients are dynamically adjusted based on three key parameters: when the ratio of P-wave to S-wave propagation velocities is greater than 1.9, the attention coefficients tilt towards P-wave features; when the formation density exceeds 2.5 g / cm³, the attention coefficients tilt towards high-frequency features; and when the seabed depth exceeds 3000 meters, the attention coefficients tilt towards low-frequency features. The model output layer consists of two parts: a classification head outputs the probability distribution of eight categories using a softmax activation function; and a regression head outputs the predicted stratigraphic thickness using a linear activation function. The training dataset setup involves four stages: data acquisition, preprocessing, annotation, and augmentation. During the data acquisition stage, P-wave and S-wave reflection data of various geological structures on the seabed were collected from different sea areas, including the Beibu Gulf, the South China Sea, and the East China Sea. The sampling point spacing was 5 meters, covering an area of ​​over 500 square kilometers and encompassing various typical seabed geological structures. In the preprocessing stage, bandpass filtering was applied to the collected data to remove noise signals above 500 Hz and below 1 Hz. Then, median filtering was used to remove outliers. Finally, data normalization was performed to scale the signal amplitude to the range [-1, 1].During the annotation phase, a team of at least five experts with geological backgrounds manually annotated the stratigraphic types and thicknesses based on drilling verification data. The annotation consistency requirement was an Inter-rater agreement coefficient of no less than 0.85. In the data augmentation phase, Gaussian noise of varying degrees was added to increase the model's noise robustness, with noise levels set to 5%–20% of the signal standard deviation. Random pruning expanded the model's adaptability to incomplete data. Random rotation operations enhanced the model's robustness to directional changes, with rotation angles ranging from -10° to 10°. Finally, a training set containing 100,000 annotated data sets and a validation set containing 10,000 annotated data sets were constructed, with a training-to-validation ratio of 10:1. The number of samples from different stratigraphic types was roughly balanced, and the ratio of the largest to smallest class samples did not exceed 3:1. Model training employed mini-batch gradient descent with a batch size of 64 and an initial learning rate of 0.001. A cosine annealing strategy was used, with the learning rate decreasing to 0.1 times its original value every 50 training epochs. The loss function employs a weighted combination of cross-entropy loss for stratigraphic type classification and mean squared error loss for thickness prediction, with weights of 0.6 and 0.4 respectively. The total loss function is expressed as: Loss = 0.6 × CrossEntropy + 0.4 × MSE. An early stopping mechanism is introduced during training: training stops when the validation set loss does not decrease for 10 consecutive training epochs to avoid overfitting. Furthermore, regularization techniques such as weight decay and Dropout are used to improve the model's generalization ability. Finally, the model parameters that perform optimally on the validation set are selected as the final model, which typically converges after 200–300 training epochs.

[0051] The mathematical model or calculation process involved in this invention will be described in detail below.

[0052] In step S03, when analyzing the physical mechanism of P-wave and S-wave propagation using the wave equation, the specific expression of the seafloor strata wave equation is as follows:

[0053]

[0054] In the formula, ρ is the density of the seafloor strata, with units of kilograms per cubic meter; λ is the displacement vector, in meters; t is the time variable, in seconds; λ is the first parameter of Lamé, in Pascals; μ is the second parameter of Lamé, in Pascals. For gradient operators; For the Laplace operator; This represents the external force, measured in Newtons per cubic meter.

[0055] The parameters were obtained as follows: ρ was directly measured using a density measuring instrument carried by the seabed crawler, with a measurement range of 1500–3500 kg / m³; λ and μ were obtained through static mechanical testing of rock samples collected by the seabed crawler in a laboratory, with typical ranges of 10. 9～10 11 Pascals and 5 × 10 8～5×10 10 Pascals; The acoustic signal is provided by a multi-wave longitudinal and transverse joint detection device, and its magnitude and direction are dynamically adjusted according to the frequency and amplitude of the excitation wave.

[0056] In practical numerical calculations, the finite difference time-domain method is used to solve the wave equation, discretizing the continuous space and time to form a difference equation:

[0057]

[0058] In the formula, Δt represents the displacement vector at position coordinates (i, j, k) and time step n; Δt is the time step size in seconds. and These are the discrete gradient operator and the discrete Laplacian operator, respectively.

[0059] The time step Δt is determined based on the Kurant stability condition:

[0060]

[0061] In the formula, Δh is the spatial grid spacing in meters; d is the spatial dimension, with a value of 3; v max The maximum wave velocity in the medium is expressed in meters per second (m / s), and the calculation formula is as follows: Where ρ min This is the minimum density value.

[0062] Longitudinal wave velocity v p and transverse wave velocity v s The calculation formula is as follows:

[0063]

[0064] The wave energy density E can be calculated from the displacement field obtained by solving the wave equation.

[0065]

[0066] In the formula, u i Represents the displacement vector The i-th component; x j This represents the j-th component of the spatial coordinates.

[0067] In step S02, the original longitudinal wave signal matrix P and the original transverse wave signal matrix S are constructed.x S y The expression is as follows:

[0068]

[0069]

[0070] In the formula, p i，j This represents the amplitude of the longitudinal wave signal received by the i-th sensor at the j-th time sampling point; and These represent the amplitudes of the transverse wave signals received by the i-th sensor at the j-th time sampling point in two orthogonal directions, respectively; m is the total number of sensors, with a value of 32; n is the number of time sampling points, which depends on the sampling rate and sampling duration, with a typical value of 10000 to 50000.

[0071] In step S03, frequency features are extracted using Fourier transform to construct the longitudinal wave interference matrix P. F and the shear wave interference matrix S F The expression is as follows:

[0072]

[0073] In the formula, This represents the amplitude of the longitudinal wave signal received by the i-th sensor at the k-th frequency component; Indicates the corresponding phase; and θ represents the amplitude of the two orthogonal transverse wave signals received by the i-th sensor at the k-th frequency component; i，k The phase of the synthesized transverse wave is represented by q; j is the imaginary unit; q is the number of frequency components, usually taken as 500.

[0074] The above amplitude and phase are calculated using Fast Fourier Transform (FFT):

[0075]

[0076] Among them, P i (f k ), and The complex spectrum can be represented as:

[0077]

[0078] Phase θ of the synthesized transverse wave i，k The calculation formula is:

[0079]

[0080] In step S04, the P-wave interference matrix and the S-wave interference matrix are weighted and fused to generate a joint interference matrix C:

[0081] C=αP F +βS F ;

[0082] In the formula, α and β are weighting coefficients, satisfying α+β=1, 0.3≤α≤0.7.

[0083] The weighting coefficients are dynamically adjusted based on the signal-to-noise ratio of the P-wave and S-wave, and the calculation formula is as follows:

[0084]

[0085] β = 1 - α;

[0086] In the formula, SNR p Signal-to-noise ratio (SNR) of the longitudinal wave signal; s α is the signal-to-noise ratio of the shear wave signal; γ is the formation type adjustment factor, ranging from -0.2 to 0.2. When the formation is hard (such as basalt), γ takes a positive value and increases α; when the formation is soft (such as muddy sedimentary layer), γ takes a negative value and decreases α.

[0087] The formula for calculating the signal-to-noise ratio is:

[0088]

[0089] In the formula, and The noise amplitudes for the longitudinal wave and the two transverse waves are respectively obtained through Fourier analysis of the no-signal interval.

[0090] In step S05, based on the real-time attitude data recorded by the attitude sensor of the seabed crawler, an interference error compensation matrix E is constructed:

[0091]

[0092] In the formula, e i，k The compensation coefficient is calculated using the following formula:

[0093] e i，k =1+a1(θ) p -θ p0 ) 2 +a2(θ r -θ r0 ) 2 +a3(d-d0);

[0094] In the formula, θ p θ is the pitch angle, in degrees. r θ is the roll angle in degrees; d is the seabed depth in meters;p0 θ r0 d0 and d0 are the pitch angle, roll angle and depth under the reference state, respectively; a1, a2 and a3 are compensation coefficients, obtained through experimental calibration, with typical values ​​of 0.01, 0.01 and 0.001, respectively.

[0095] The formula for calculating the compensated joint interference matrix C′ is:

[0096]

[0097] In the formula, This represents the element-wise product of matrices (Hadamard product).

[0098] In step S06, time-frequency analysis is performed on the compensated joint interference matrix using wavelet transform. Specifically, wavelet transform is applied to each row of matrix C′, which can be expressed as:

[0099]

[0100] In the formula, c′ i Let ψ represent the signal from the i-th row of matrix C′, i.e., the signal from the i-th sensor; k，l (t) indicates a scale of 2 k Wavelet basis functions with a translation of l; W i，k，l This represents the wavelet coefficients.

[0101] The db4 wavelet from the Daubechies wavelet family is adopted. Its mother wavelet function ψ(t) has no analytical expression and is obtained numerically through iterative methods.

[0102] Wavelet coefficient energy distribution E i，k，l The calculation formula is:

[0103] E i，k，l =|W i，k，l | 2 ;

[0104] Time-frequency energy diagram TFE i (f, t) is obtained through wavelet coefficient energy reconstruction:

[0105]

[0106] In the formula, f k To be consistent with scale 2 k The corresponding frequency; t l δ represents the time corresponding to the translation amount l; δ is the Dirac function.

[0107] Specific time-frequency feature extraction includes the following aspects:

[0108] 1. Energy peak characteristics:

[0109] EP i =max f，t TFE i (f, t);

[0110] FP i =argmax f ∑ t T FE i (f, t);

[0111] TP i =argmax t ∑ f T FE i (f, t);

[0112] 2. Frequency band energy distribution:

[0113] BE i，k =∑ l E i，k，l ;

[0114] 3. Statistical characteristics:

[0115] Mean:

[0116] variance:

[0117] Skewness:

[0118] Kuroshi:

[0119] In the formula, L k The total number of wavelet coefficients at scale k.

[0120] The final specific time-frequency eigenvector F i It is composed of the above features:

[0121]

[0122] By processing data from 32 sensors collected by the seabed crawler, the feature matrix F, which is then input into the stratigraphic classification prediction model, is obtained:

[0123] F = [F1, F2, ..., F 32 ] T ;

[0124] The principles and significance of selecting these functional relationships in the above equations are as follows: the wave equation is the fundamental physical equation describing the propagation of waves in an elastic medium, and its partial differential equation form can accurately reflect the propagation, reflection, and attenuation characteristics of waves; the Fourier transform can convert time-domain signals into frequency-domain representations, facilitating the extraction of the response characteristics of strata to different frequency components; the adaptive weighted fusion algorithm based on signal-to-noise ratio can dynamically adjust the weights of P-wave and S-wave information according to signal quality, improving the fusion effect; the attitude compensation matrix adopts a quadratic function form, which can better fit the nonlinear influence of attitude changes on the signal; wavelet transform has good time-frequency localization characteristics, and can simultaneously acquire time-domain and frequency-domain information of the signal, making it particularly suitable for analyzing the non-stationary characteristics of seafloor strata; statistical features (mean, variance, skewness, kurtosis, etc.) can comprehensively describe the distribution characteristics of wavelet coefficients, helping to identify differences in different strata structures. These equations comprehensively consider the physical principles, signal processing techniques, and feature extraction methods of seafloor strata detection, forming a complete theoretical system for a multi-P-wave and S-wave joint detection method, which improves the accuracy and resolution of seafloor strata structure identification compared to existing technologies.

[0125] Specifically, the principle of this invention is as follows: The core principle of this invention's technical solution lies in overcoming various interference factors in seabed geological exploration through the synergistic effect of multiple anti-interference technologies, thereby achieving accurate identification of the geological structure. First, a seabed crawler is used to directly probe close to the seabed, avoiding the attenuation and interference problems caused by the signal passing through the entire water body, and reducing the adverse effects of water factors such as currents and thermoclines. Second, both P-waves and S-waves are used simultaneously for detection, utilizing their complementary propagation characteristics in the strata to enhance the anti-interference capability and resolution of the detection signal.

[0126] The design of multi-frequency P-wave and S-wave signals enhances the adaptability of the detection system. Waves of different frequencies can maintain the effectiveness of certain frequency bands under different interference environments. High-frequency waves are suitable for detecting shallow details, while low-frequency waves can penetrate deeper strata and have stronger anti-interference capabilities. By performing frequency domain analysis on the original signal through Fourier transform and combining it with the seafloor strata wave equation for physical mechanism analysis, it is possible to distinguish the characteristic differences between effective signals and interference signals, and construct P-wave and S-wave interference matrices characterizing the properties of different geological structures.

[0127] The key innovation of this invention lies in the introduction of an interference error compensation mechanism based on the attitude data of a seabed crawler. The attitude changes (pitch angle, roll angle, etc.) of the crawler when moving on uneven seabeds can significantly interfere with signal acquisition. By constructing an interference error compensation matrix using real-time attitude data, the signal fluctuations and distortions caused by these attitude changes can be effectively eliminated. The time-frequency analysis capability of wavelet transform enables the system to simultaneously identify and filter interference signals in both the time and frequency domains, extracting joint time-frequency features that reflect the unique response characteristics of different geological structures.

[0128] A U-shaped convolutional neural network based on a self-attention mechanism was used as a stratigraphic classification and prediction model. Through training on a large amount of labeled data, it learned signal characteristic patterns under different interference environments, enabling it to adaptively identify and suppress various interferences, thus improving its ability to identify complex geological structures. The model dynamically adjusts the attention coefficient based on key parameters such as stratigraphic density, depth, and the ratio of P-wave to S-wave propagation velocity, further enhancing the system's adaptability and recognition accuracy under various interference conditions.

[0129] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0130] The specific implementation method of step S01 is the same as described above, and will not be repeated here.

[0131] The specific implementation of step S02 involves using an omnidirectional receiving sensor array installed on the bottom of the seabed crawler to receive P-wave and S-wave signals reflected from the seabed strata. This sensor array consists of 32 triaxial accelerometers arranged in a 4×8 matrix, capable of simultaneously capturing P-wave and S-wave reflection signals in three-dimensional space. The received analog signals are amplified by a preamplifier and then converted into digital signals at a sampling rate of 10 kHz by a 24-bit high-precision analog-to-digital converter. The digital signals undergo preliminary filtering by a data processing module to remove 50 Hz power frequency interference and high-frequency noise above 500 Hz. The processed data is rearranged according to the time sequence and sensor position information to construct the original P-wave signal matrix and the original S-wave signal matrix. The original P-wave signal matrix has a dimension of m×n, where m is the number of sensors and n is the number of time sampling points; the original S-wave signal matrix consists of two m×n matrixes, corresponding to the vibration components of the S-wave in two orthogonal directions. The purpose of this step is to obtain high-quality original reflection signal data, laying the foundation for subsequent feature extraction. Original P-wave signal matrix P and original S-wave signal matrix S x S y The expression is as follows:

[0132]

[0133] In the formula, p i，j This represents the amplitude of the longitudinal wave signal received by the i-th sensor at the j-th time sampling point; and These represent the amplitudes of the transverse wave signals received by the i-th sensor at the j-th time sampling point in two orthogonal directions, respectively; m is the total number of sensors, with a value of 32; n is the number of time sampling points, which depends on the sampling rate and sampling duration, with a typical value of 10000 to 50000.

[0134] The specific implementation of step S03 involves performing Fourier transform processing on the original P-wave signal matrix and the original S-wave signal matrix. First, a Fast Fourier Transform is performed on the time-series signals recorded by each sensor to extract frequency components within the range of 1–500 Hz, obtaining spectral amplitude and phase information. For the P-wave signal, the propagation characteristics of the P-wave at different frequencies are extracted to construct a P-wave interference matrix; for the S-wave signal, spectral analysis is performed on the S-wave components in two orthogonal directions to construct a S-wave interference matrix. Simultaneously, the physical mechanism of P-wave and S-wave propagation is analyzed based on the seafloor strata wave equation. This wave equation considers physical characteristics such as stratum density, Lamé first parameter, and Lamé second parameter to calculate the propagation speed and attenuation characteristics of P-wave and S-wave in different geological structures. P-wave velocity v p With transverse wave velocity v s pass and The calculations yielded λ, representing the first Lamé parameter, μ, representing the second Lamé parameter, and ρ, representing the formation density. The wave equation was numerically solved using the finite-difference time-domain method, with a grid spacing of one-tenth of the propagation wavelength and a time step determined based on the Coulant stability condition. The wave displacement field distribution and wave energy attenuation law were obtained by solving the wave equation, which were used to construct the P-wave and S-wave interference matrices. The purpose of this step was to extract the characteristic information of P-wave and S-wave propagation in the seafloor strata through frequency domain analysis, providing a basis for subsequent signal fusion.

[0135] The equation for seafloor strata fluctuation is expressed as follows:

[0136]

[0137] In the formula, ρ is the density of the seafloor strata, with units of kilograms per cubic meter; λ is the displacement vector, in meters; t is the time variable, in seconds; λ is the first parameter of Lamé, in Pascals; μ is the second parameter of Lamé, in Pascals. For gradient operators; For the Laplace operator; This represents the external force, measured in Newtons per cubic meter.

[0138] The parameters were obtained as follows: ρ was directly measured using a density measuring instrument carried by the seabed crawler, with a measurement range of 1500–3500 kg / m³; λ and μ were obtained through static mechanical testing of rock samples collected by the seabed crawler in a laboratory, with typical ranges of 10. 9 ~10 11 Pascal and 5×10 8 ~5×10 10 Pascal; The acoustic signal is provided by a multi-wave longitudinal and transverse joint detection device, and its magnitude and direction are dynamically adjusted according to the frequency and amplitude of the excitation wave.

[0139] In practical numerical calculations, the finite difference time-domain method is used to solve the wave equation, discretizing the continuous space and time to form a difference equation:

[0140]

[0141] In the formula, Δt represents the displacement vector at position coordinates (i, j, k) and time step n; Δt is the time step size in seconds. and These are the discrete gradient operator and the discrete Laplacian operator, respectively.

[0142] The time step Δt is determined based on the Kurant stability condition:

[0143]

[0144] In the formula, Δh is the spatial grid spacing in meters; d is the spatial dimension, with a value of 3; v max The maximum wave velocity in the medium is expressed in meters per second (m / s), and the calculation formula is as follows: Where ρ min This is the minimum density value.

[0145] Frequency features were extracted using Fourier transform to construct the longitudinal wave interference matrix P. F and the shear wave interference matrix S F The expression is as follows:

[0146]

[0147] In the formula, This represents the amplitude of the longitudinal wave signal received by the i-th sensor at the k-th frequency component; Indicates the corresponding phase; and θ represents the amplitude of the two orthogonal transverse wave signals received by the i-th sensor at the k-th frequency component; i，k The phase of the synthesized transverse wave is represented by q; j is the imaginary unit; q is the number of frequency components, usually taken as 500.

[0148] The above amplitude and phase are calculated using Fast Fourier Transform (FFT):

[0149]

[0150] Among them, P i (f k ), and The complex spectrum can be represented as:

[0151]

[0152] Phase θ of the synthesized transverse wave i，k The calculation formula is:

[0153]

[0154] The specific implementation of step S04 involves weighted fusion of the P-wave interference matrix and the S-wave interference matrix to generate a joint interference matrix. The weighted fusion employs an adaptive weighting algorithm, with the weighting coefficients dynamically adjusted based on seabed geological conditions. First, the signal-to-noise ratio (SNR) of the P-wave and S-wave under the current seabed geological conditions is calculated, denoted as SNR P-wave and SNR S-wave, respectively. Then, the initial weighting coefficients α and β are determined based on the SNR, satisfying α + β = 1. When SNR P-wave > SNR S-wave, the value of α is increased; conversely, the value of β is increased. In practical applications, the value of α is generally adjusted within the range of 0.3 to 0.7. Furthermore, the influence of stratigraphic type on the weighting is considered: for hard strata such as basalt, where P-wave penetration is strong, the value of α can be set to 0.6–0.7; for soft strata such as argillaceous sediments, where S-waves are more sensitive to interface responses, the value of β can be set to 0.6–0.7. The formula for calculating the joint interference matrix is: Joint interference matrix = α × P-wave interference matrix + β × S-wave interference matrix. The purpose of this step is to comprehensively utilize the complementary advantages of P-waves and S-waves to improve the accuracy of stratigraphic identification.

[0155] The P-wave interference matrix and the S-wave interference matrix are weighted and fused to generate a joint interference matrix C:

[0156] C=αP F +βS F ;

[0157] In the formula, α and β are weighting coefficients, satisfying α+β=1, 0.3≤α≤0.7.

[0158] The weighting coefficients are dynamically adjusted based on the signal-to-noise ratio of the P-wave and S-wave, and the calculation formula is as follows:

[0159]

[0160] In the formula, SNR p Signal-to-noise ratio (SNR) of the longitudinal wave signal; s α is the signal-to-noise ratio of the shear wave signal; γ is the formation type adjustment factor, ranging from -0.2 to 0.2. When the formation is hard (such as basalt), γ takes a positive value and increases α; when the formation is soft (such as muddy sedimentary layer), γ takes a negative value and decreases α.

[0161] The formula for calculating the signal-to-noise ratio is:

[0162]

[0163] In the formula, and The noise amplitudes for the longitudinal wave and the two transverse waves are respectively obtained through Fourier analysis of the no-signal interval.

[0164] The specific implementation of step S05 involves constructing an interference error compensation matrix based on the real-time attitude data recorded by the seabed crawler's attitude sensors, and then compensating for the joint interference matrix. The seabed crawler's attitude sensors include a high-precision gyroscope, accelerometer, and depth sensor, which record the crawler's pitch, roll, and depth data in real time, with a sampling frequency of 100 Hz. First, a coordinate transformation model is established to transform the sensor's measurement coordinate system to a global reference coordinate system, eliminating signal distortion caused by changes in the crawler's attitude. The coordinate transformation is implemented using a rotation matrix, calculated based on Euler angles. Additional compensation is required when the crawler's pitch or roll angle exceeds 5 degrees. The constructed interference error compensation matrix has the same dimension as the joint interference matrix, and its element values ​​are calculated non-linearly based on the attitude deviation angle. The compensation process is implemented using matrix multiplication: the compensated joint interference matrix = interference error compensation matrix × joint interference matrix. The purpose of this step is to eliminate the impact of seabed crawler attitude changes on signal acquisition and improve data quality.

[0165] Based on the real-time attitude data recorded by the attitude sensor of the seabed crawler, an interference error compensation matrix E is constructed:

[0166]

[0167] In the formula, e i，k The compensation coefficient is calculated using the following formula:

[0168] e i，k =1+a1(θ) p -θ p0 ) 2 +a2(θ r -θ r0 ) 2 +a3(d-d0);

[0169] In the formula, θ p θ is the pitch angle, in degrees. r θ is the roll angle in degrees; d is the seabed depth in meters; p0 θ r0 d0 and d0 are the pitch angle, roll angle and depth under the reference state, respectively; a1, a2 and a3 are compensation coefficients, obtained through experimental calibration, with typical values ​​of 0.01, 0.01 and 0.001, respectively.

[0170] The formula for calculating the compensated joint interference matrix C′ is:

[0171] C′=E°C;

[0172] In the formula, ° represents the element-wise product of matrices (Hadamard product).

[0173] The specific implementation of step S06 involves performing time-frequency analysis on the compensated joint interference matrix using wavelet transform to extract specific time-frequency features. First, a suitable wavelet basis function for stratigraphic analysis is selected, typically the db4 or db6 wavelet from the Daubechies wavelet family, due to its good time-frequency localization characteristics. Multi-scale wavelet decomposition is performed on each row of data in the joint interference matrix, with five decomposition levels, yielding wavelet coefficients for different frequency bands. The energy distribution of the wavelet coefficients at each level is calculated, constructing a time-frequency energy map to reflect the energy distribution characteristics of the signal at different times and frequencies. Based on the time-frequency energy map, energy peak points and their distribution characteristics are extracted as specific time-frequency features. Furthermore, statistical characteristics of the wavelet coefficients in each frequency band are calculated, including mean, variance, skewness, and kurtosis, as supplementary features to the feature vector. Based on prior geological knowledge, the feature vectors are screened and combined, retaining features that contribute significantly to stratigraphic structure identification and removing redundant features. The final feature vector has a dimension of 128, including time-domain features, frequency-domain features, and joint time-frequency features. The purpose of this step is to extract key features that reflect the characteristics of the stratigraphic structure through time-frequency analysis, so as to provide a basis for subsequent stratigraphic classification and thickness prediction.

[0174] Time-frequency analysis of the compensated joint interference matrix is ​​performed using wavelet transform. Each row of matrix C′ can be represented by a wavelet transform as follows:

[0175]

[0176] In the formula, c′ i Let ψ represent the signal from the i-th row of matrix C′, i.e., the signal from the i-th sensor; k，l (t) indicates a scale of 2 k Wavelet basis functions with a translation of l; W i，k，l This represents the wavelet coefficients.

[0177] Wavelet coefficient energy distribution E i，k，l The calculation formula is:

[0178] E i，k，l =|W i，k，l | 2 ;

[0179] Time-frequency energy diagram TFE i (f, t) is obtained through wavelet coefficient energy reconstruction:

[0180]

[0181] In the formula, f k To be consistent with scale 2 k The corresponding frequency; t l δ represents the time corresponding to the translation amount l; δ is the Dirac function.

[0182] Specific time-frequency feature extraction includes the following aspects:

[0183] 1. Energy peak characteristics:

[0184] EP i =max f， tTFEi (f, t);

[0185] FP i =argmax f ∑ t T FE i (f, t);

[0186] TP i =argmax t ∑ f T FE i (f, t);

[0187] 2. Frequency band energy distribution:

[0188] BE i，k =∑ l E i，k，l ;

[0189] 3. Statistical characteristics:

[0190] Mean:

[0191] variance:

[0192] Skewness:

[0193] Kuroshi:

[0194] In the formula, L k The total number of wavelet coefficients at scale k.

[0195] The final specific time-frequency eigenvector F i It is composed of the above features:

[0196]

[0197] By processing data from 32 sensors collected by the seabed crawler, the feature matrix F, which is then input into the stratigraphic classification prediction model, is obtained:

[0198] F = [F1, F2, ..., F 32 ] T ;

[0199] The specific implementation methods for steps S07-S08 are the same as those described above, and will not be repeated here.

[0200] Optional, such as Figure 2 As shown, the multi-wave P-wave and S-wave combined detection device is the core component of the seabed crawler detection system. It consists of a multi-frequency wave source generator, a P-wave transmitting unit, a S-wave transmitting unit, an omnidirectional receiving sensor array, a signal processing module, and a data storage unit. The whole device is encapsulated in a titanium alloy shell and has a pressure resistance of 20 MPa, making it suitable for deep-sea environments up to 6,000 meters.

[0201] The multi-frequency wave source generator employs digital signal synthesis technology to generate continuously adjustable frequency signals within the range of 1–500 Hz, with a frequency resolution of 0.1 Hz. Internally, the generator integrates a high-precision digital frequency synthesis chip (DDS, Direct Digital Synthesizer), controlled by a 32-bit microprocessor, capable of simultaneously generating up to 16 excitation signals of different frequencies. These signals undergo waveform shaping via digital filters to ensure high signal purity, with a total harmonic distortion (THD) of less than 0.01%. The multi-frequency wave source generator also features adaptive frequency selection, intelligently adjusting the excitation signal frequency combination according to different seabed geological environment characteristics to achieve optimal detection results. In practical applications, logarithmically spaced frequency combinations, such as 1, 2, 5, 10, 20, 50, 100, 200, and 500 Hz, are typically used to cover a wide frequency band and meet the detection needs of different depths and geological strata with varying physical properties.

[0202] The longitudinal wave transmitting unit employs a piezoelectric ceramic transducer array structure, consisting of 64 high-performance piezoelectric ceramic elements arranged in an 8×8 matrix, with a total area of ​​400 square centimeters. Each piezoelectric element is made of modified PZT-5H material, exhibiting high motor conversion efficiency and good temperature stability, with an operating temperature range of -10 to 60 degrees Celsius. The piezoelectric elements are connected to the titanium alloy shell via a special backing material. This backing material uses a polymer and tungsten powder composite formula, possessing high damping characteristics to effectively suppress the transducer's back radiation and improve forward acoustic energy output. The working principle of the longitudinal wave transmitting unit utilizes the mechanical vibration in the thickness direction generated by the piezoelectric material under the action of an electric field to excite longitudinal sound waves parallel to the propagation direction. The overall acoustic power of the unit can reach 2 kilowatts, with a maximum sound intensity of 5 watts per square centimeter, capable of generating sufficient energy to penetrate complex geological structures. To prevent overheating, the longitudinal wave transmitting unit incorporates a built-in water circulation cooling system to ensure stable operation over extended periods.

[0203] The transverse wave emitting unit employs a shear-mode piezoelectric ceramic design, consisting of 32 pairs of orthogonally arranged shear-mode piezoelectric elements, capable of generating transverse waves in two orthogonal directions. Each pair of shear elements measures 30×30×15 mm and operates in d15 mode, generating shear vibrations perpendicular to the polarization direction through an electric field applied to the electrode surface. The transverse wave emitting unit is arranged in a 4×8 array, with a total area of ​​300 square centimeters. The shear elements are fixed to a specially designed flexible coupling layer made of silicone rubber material filled with metal particles, with an acoustic impedance of 5×10⁻⁶. 6 With a speed of Pa·s / meter, close to the acoustic impedance of seabed sediments, efficient energy transfer is achieved. The shear wave transmitting unit operates in the frequency range of 5–200 Hz, lower than the P-wave transmitting unit, because the propagation speed of shear waves in the medium is generally lower than that of P-waves. The acoustic power of the shear wave unit is 1.5 kW, and the maximum acoustic intensity can reach 3 W / cm². Shear wave signals can provide stratigraphic information complementary to P-waves, making them particularly suitable for identifying fluid-filled structures and interface characteristics.

[0204] The omnidirectional receiving sensor array consists of 32 triaxial accelerometers arranged in a 4×8 rectangular array, covering an area of ​​600 square centimeters. Each triaxial sensor contains three orthogonally aligned high-sensitivity MEMS accelerometers with a sensitivity of 100 mV / g, a frequency response of 0.1–1000 Hz, and a dynamic range of ±50 g. The sensors employ a low-noise design, with an equivalent noise level below 1 μg / √Hz and a signal-to-noise ratio greater than 80 dB. The triaxial sensors can simultaneously receive longitudinal and transverse wave signals, distinguishing wave components in different vibration directions. The receiving array uses a flexible, suspended mounting method, maintaining good contact with the bottom of the crawler while a micro-spring isolation system reduces interference from the crawler's vibration on the received signal. An acoustic barrier ring is placed around the array to reduce lateral noise interference. The omnidirectional receiving sensor array has a sampling rate of 10 kHz, satisfying the Nyquist sampling theorem and avoiding spectral aliasing.

[0205] The signal processing module adopts a high-performance digital signal processor architecture, with a dual-core DSP main chip at 3GHz to meet real-time signal processing requirements. This module includes multiple preamplifiers with an adjustable gain range of 20–60 dB and an input impedance of 10 kΩ. The signal conditioning circuit uses a differential input design with a high common-mode rejection ratio (>80 dB), effectively suppressing environmental electromagnetic interference. The analog-to-digital converter uses a 24-bit Sigma-Delta ADC with a sampling rate of 10 kHz and a signal-to-noise ratio greater than 110 dB. Digital filtering employs a programmable filter bank, including high-pass, low-pass, band-pass, and band-stop filters, with the cutoff frequency dynamically adjustable according to actual needs. Signal processing algorithms include adaptive noise cancellation, waveform feature extraction, and spectrum analysis, implemented through FPGA hardware acceleration with a processing latency of less than 10 milliseconds. This module also features automatic gain control, automatically adjusting the gain based on the received signal strength to maintain the optimal signal level and avoid saturation distortion or noise overload.

[0206] The data storage unit employs a solid-state drive array design with a RAID 5 redundancy strategy to improve data reliability. This unit also features a low-power standby mode, reducing power consumption by over 90% when not in operation, extending battery life. The data storage unit connects to the signal processing module via a fiber optic interface, achieving a data transfer rate of 10 gigabits per second while avoiding electromagnetic interference. The storage unit also includes a data backup module that automatically backs up critical data to an independent storage area every hour to prevent accidental data loss.

[0207] The power supply system of the multi-wave and transverse wave joint detection device uses a 100 kWh lithium-ion battery pack, supporting 12 hours of continuous operation. The power management system achieves intelligent power distribution, dynamically adjusting power supply according to the operating status of each module to maximize energy efficiency. The device's outer shell is made of high-strength titanium alloy with a strength grade of HY-100, capable of withstanding pressure depths up to 6000 meters, and coated with anti-corrosion polyurethane material, providing a seawater corrosion resistance lifespan exceeding 5 years. The entire device weighs 120 kg, has a volume of 0.6 cubic meters, and adopts a modular design for easy maintenance and upgrades.

[0208] The operation of the multi-wave (P-wave) and (S-wave) combined detection device is as follows: a multi-frequency wave source generator produces excitation signals with a preset frequency combination, driving the P-wave and S-wave transmitting units to transmit detection signals to the seabed strata; an omnidirectional receiving sensor array receives the P-wave and S-wave signals reflected from different strata interfaces; the received signals are amplified, filtered, and subjected to preliminary feature extraction by a signal processing module; the processed data, along with timestamps and location information, is saved to a data storage unit, providing raw material for subsequent data analysis. This device, through complementary P-wave and S-wave detection, significantly improves the accuracy and resolution of seabed strata structure identification, and is particularly suitable for the detection and analysis of complex geological structures and multiphase media.

[0209] To better understand and implement this invention, the following is a specific application scenario example 2: Researchers conducted a multi-wave (P-wave and S-wave) joint detection experiment based on a seabed crawler in a hydrothermal vent area of ​​a certain sea trough. This area, with a water depth of approximately 1400–1600 meters, is a known hydrothermal vent area rich in sulfide deposits. The seabed crawler used in the experiment was equipped with a multi-wave (P-wave and S-wave) joint detection device to detect the seabed stratigraphic structure and resource distribution.

[0210] The multi-wave (P-wave) and transverse-wave combined detection device consists of a multi-frequency wave source generator, a P-wave transmitting unit, a transverse-wave transmitting unit, an omnidirectional receiving sensor array, a signal processing module, and a data storage unit. The wave source generator operates in the frequency range of 1–500 Hz and has a transmitting power of 2500 W. The P-wave transmitting unit consists of 24 piezoelectric ceramic transducers, the transverse-wave transmitting unit is made of 16 shear-type piezoelectric materials, and the omnidirectional receiving sensor array consists of 32 triaxial accelerometers with a signal sampling rate of 10 kHz. The crawler is equipped with a high-precision attitude sensor system, including a gyroscope, accelerometer, and depth sensor, with a sampling frequency of 100 Hz.

[0211] Researchers planned a detection path approximately 3.6 km long, using a grid layout with 8 m spacing between grids. The seabed crawler moved along the pre-set path at an average speed of 0.3 m / s, simultaneously activating a multi-frequency P-wave and S-wave joint detection device to transmit multi-frequency P-wave and S-wave signals in the range of 1–500 Hz to the seabed strata. The crawler's onboard density measuring instrument measured the density of the seabed strata in real time, and simultaneously collected rock samples for subsequent laboratory analysis to obtain Lamé parameters.

[0212] Researchers constructed the original P-wave signal matrix and the original S-wave signal matrix using reflected signals received by a multi-pole P-wave and S-wave joint detection device. Fourier transforms were performed on these matrices to extract frequency features, and P-wave interference matrices and S-wave interference matrices were constructed. Typical stratigraphic physical parameters obtained in the experiment are shown in Table 1.

[0213] Table 1. Measurement results of physical parameters of different strata in the hydrothermal vent area of ​​the Okinawa Trough.

[0214]

[0215] Based on these physical parameters, researchers analyzed the physical mechanism of P-wave and S-wave propagation using the wave equation of the seafloor strata. The numerical simulation of the wave equation employed the finite-difference time-domain method, with the spatial grid spacing set to one-tenth of the propagation wavelength, specifically 5 cm. The time step was determined to be 5 × 10⁻⁶ based on the Kurant stability condition. -6 By solving the wave equation, the distribution of the wave displacement field and the wave energy attenuation law are obtained, further refining the longitudinal wave interference matrix and the transverse wave interference matrix.

[0216] The signal-to-noise ratio (SNR) evaluation results before fusing the P-wave interference matrix and the S-wave interference matrix are shown in Table 2.

[0217] Table 2. Evaluation results of signal-to-noise ratio of P-wave and S-wave signals under different stratigraphic types.

[0218]

[0219] Based on the signal-to-noise ratio and stratigraphic type, researchers determined weighted fusion coefficients suitable for each stratigraphic type, and weighted the P-wave interference matrix and S-wave interference matrix to generate a joint interference matrix. During the exploration, the attitude data recorded by the seabed crawler showed that the maximum pitch angle variation was ±8.3° and the maximum roll angle variation was ±6.7°, which interfered with signal acquisition. Researchers constructed an interference error compensation matrix based on the attitude sensor data, with compensation coefficients a1, a2, and a3 set to 0.012, 0.011, and 0.0008, respectively.

[0220] Researchers performed time-frequency analysis on the compensated joint interference matrix using wavelet transform. They selected the db4 wavelet from the Daubechies wavelet family as the basis function, performed a 5-level wavelet decomposition, and extracted specific time-frequency features. These features were then input into a pre-trained stratigraphic classification prediction model, which is based on a U-shaped convolutional neural network with a self-attention mechanism and consists of an encoder and a decoder.

[0221] The model was trained using a training set containing 105,000 labeled data points and a validation set containing 10,000 labeled data points. The Adam optimizer was used for training, with an initial learning rate of 0.001, which decreased to 0.1 times the initial learning rate every 50 training epochs. The loss function was a weighted combination of cross-entropy loss for stratigraphic type classification and mean squared error loss for thickness prediction, with weights of 0.6 and 0.4, respectively. After 178 training epochs, the model achieved its best performance on the validation set.

[0222] The accuracy of the final stratigraphic classification and thickness prediction results is shown in Table 3:

[0223] Table 3. Accuracy Assessment of Stratigraphic Classification and Thickness Prediction Results

[0224] Stratigraphic type Classification accuracy (%) Relative error in thickness prediction (%) Sample size sandy sedimentary layers 94.6 6.8 1263 muddy sedimentary layer 95.2 5.4 1587 hydrothermal deposits 93.8 7.2 826 sulfide ore bodies 96.3 4.5 342 basalt basement 97.5 3.8 982

[0225] Based on stratigraphic classification and thickness prediction results, researchers created a three-dimensional geological profile of the Okinawa Trough hydrothermal vent area, marking key geological structures and resource distribution areas. Through this process, three sulfide ore-rich areas and two potential natural gas hydrate distribution areas were successfully identified. Their specific locations and estimated resource quantities are shown in Table 4.

[0226] Table 4. Locations of identified resource-rich areas and resource estimates.

[0227]

[0228] Traditional seafloor stratigraphy relies primarily on single acoustic detection methods, such as single-frequency P-wave sonar or side-scan sonar. These methods suffer from limitations in detection depth and insufficient ability to distinguish different geological structures. Particularly in complex hydrothermal vent areas, traditional methods struggle to accurately identify ore bodies and resource distribution. Traditional methods typically achieve only 70%–80% stratigraphic classification accuracy, with thickness prediction relative errors usually between 15% and 20%. In contrast, this invention employs a multi-P- and S-wave combined detection method. By utilizing the complementary characteristics of P- and S-waves simultaneously, combined with the advantages of a seafloor crawler platform, it significantly improves detection accuracy. Experimental data shows that this method achieves stratigraphic classification accuracy of 93.8%–97.5%, and reduces the relative error of thickness prediction to 3.8%–7.2%. Furthermore, traditional methods often require multiple devices, making operation complex. This invention integrates multiple detection functions into a single system, simplifying the operation process and improving efficiency. Most importantly, by fusing P- and S-wave information and employing advanced machine learning algorithms, this invention can more accurately identify seafloor resource distribution, providing more reliable technical support for seafloor resource exploration.

[0229] It should be noted that the variables involved in this invention are explained in detail in Tables 5 and 6 below.

[0230] Table 5. Variable Explanation Table (Part 1)

[0231]

[0232]

[0233] Table 6. Variable Explanation Table (Part Two)

[0234]

[0235] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for joint detection of multiple P-waves and S-waves based on a seabed crawler, characterized in that, include: Multiple P-wave and S-wave joint detection devices are deployed on the seabed crawler, enabling the seabed crawler to crawl along a preset detection path and emit multi-frequency P-wave and S-wave signals. The system receives P-wave and S-wave signals reflected from the seabed strata. For P-wave signals, it extracts the propagation characteristics of P-waves at different frequencies and constructs a P-wave interference matrix. For S-wave signals, it performs spectral analysis on the S-wave components in two orthogonal directions and constructs a S-wave interference matrix. The P-wave and S-wave interference matrices are then weighted and fused to generate a joint interference matrix. Based on real-time attitude data recorded by the seabed crawler's attitude sensor, an interference error compensation matrix is ​​constructed to compensate the joint interference matrix. Time-frequency analysis is performed using wavelet transform to extract unique time-frequency features. These unique time-frequency features are then input into a pre-trained strata classification and prediction model to classify seabed strata structure types and predict their thickness. Draw a three-dimensional geological profile of the seabed and mark key geological structures and resource distribution areas.

2. The method for joint detection of multiple P-waves and S-waves according to claim 1, characterized in that, The multi-frequency P-wave and S-wave combined detection device consists of a multi-frequency wave source generator, a P-wave transmitting unit, a S-wave transmitting unit, an omnidirectional receiving sensor array, a signal processing module, and a data storage unit. It is installed on the bottom of the seabed crawler and is used to simultaneously transmit multiple frequencies of P-waves and S-waves and receive the reflected signals.

3. The method for joint detection of multiple P-waves and S-waves according to claim 2, characterized in that, The longitudinal wave interference matrix refers to the matrix that characterizes the interference generated when longitudinal waves propagate in different geological structures by extracting the amplitude and phase of each frequency component after performing spectral analysis on the original longitudinal wave signal matrix. The transverse wave interference matrix is ​​a matrix that characterizes the interference generated by transverse waves propagating in different geological structures by extracting the amplitude and phase of each frequency component after spectral analysis of the original transverse wave signal matrix. The joint interference matrix is ​​a matrix formed by combining the longitudinal wave interference matrix and the transverse wave interference matrix through a weighted fusion algorithm, which comprehensively reflects the comprehensive interference characteristics of longitudinal and transverse waves propagating in the seabed strata.

4. The method for joint detection of multiple P-waves and S-waves according to claim 3, characterized in that, The interference error compensation matrix is ​​a compensation matrix calculated based on the pitch angle, roll angle, and depth data recorded by the attitude sensors of the seabed crawler, used to eliminate the interference caused by the attitude changes of the seabed crawler to signal acquisition.

5. The method for joint detection of multiple P-waves and S-waves according to claim 4, characterized in that, Specific time-frequency features refer to the combined time-domain and frequency-domain features extracted after performing time-frequency analysis on the joint interference matrix through wavelet transform, which are used to reflect the unique response characteristics of different geological structures.

6. The method for joint detection of multiple P-waves and S-waves according to claim 5, characterized in that, The seafloor strata wave equation is a partial differential equation describing the propagation characteristics of P-waves and S-waves in seafloor strata. It is used to calculate the propagation speed and attenuation characteristics of P-waves and S-waves in different geological structures. The inputs to the seafloor strata wave equation include the strata density obtained from a seafloor crawler, the first Lamé parameter obtained from rock sampling analysis of the seafloor crawler, the second Lamé parameter obtained from rock sampling analysis of the seafloor crawler, and the displacement vectors calculated from the original P-wave signal matrix and the original S-wave signal matrix. The outputs are the wave displacement field distribution and the wave energy attenuation law, which are used to construct the P-wave interference matrix and the S-wave interference matrix.

7. The method for joint detection of multiple P-waves and S-waves according to claim 6, characterized in that, The stratigraphic classification prediction model refers to a neural network model used for stratigraphic structure identification and thickness prediction. The specific structure of the stratigraphic classification prediction model is a U-shaped convolutional neural network based on the self-attention mechanism, which includes an encoder and a decoder. The encoder consists of five convolutional layers and four pooling layers, with the number of convolutional kernels in each layer being 32, 64, 128, 256, and 512, respectively, and the kernel size being 3×3. The decoder consists of five deconvolutional layers and four concatenation layers, with the number of feature maps in each layer being the same as the corresponding layer in the encoder.

8. The method for joint detection of multiple P-waves and S-waves according to claim 7, characterized in that, A self-attention mechanism is introduced between the encoder and decoder to enhance feature extraction capabilities. The sparse attention coefficients in the self-attention mechanism are dynamically adjusted based on three key parameters: the ratio of P-wave to S-wave propagation velocities calculated from the P-wave interference matrix and the S-wave interference matrix, the formation density obtained from the seabed crawler, and the seabed depth obtained from the seabed crawler.

9. The method for joint detection of multiple P-waves and S-waves according to claim 8, characterized in that, The steps for establishing the training dataset in the training process of the stratigraphic classification prediction model include collecting P-wave and S-wave reflection data of seafloor strata with various geological structures from different sea areas, preprocessing the collected data to remove noise and outliers, determining the stratigraphic type and thickness using manual annotation, and expanding the number of training samples through data augmentation techniques.

10. The method for joint detection of multiple P-waves and S-waves according to claim 9, characterized in that, The steps for training the stratigraphic classification prediction model include first initializing the model using a preprocessed training dataset, then optimizing the parameters using the Adam optimizer with an initial learning rate of 0.001, which is reduced to 0.1 times every 50 training epochs. The loss function is a weighted combination of stratigraphic type classification cross-entropy loss and thickness prediction mean square error loss, with weights of 0.6 and 0.4, respectively.

Images