Marine area cave group transient electromagnetic cooperative survey data processing method and system

Abstract

The application discloses a seawater area cave group transient electromagnetic cooperative survey data processing method and system, and the method comprises the following steps: transmitting a wide frequency excitation signal to the seabed through a ship-borne transient electromagnetic transmitting system, synchronously collecting secondary field attenuation signals of seabed medium responses by using a towed receiving array, and simultaneously collecting background noise base data by arranging a distributed sensor network at a preset seabed node; performing motion attitude compensation on the secondary field attenuation signals to generate transient electromagnetic response data; constructing an adaptive filter based on the background noise base data, performing noise suppression processing on the transient electromagnetic response data, and outputting noise suppression data; and performing space-time alignment on the noise suppression data of the towed array and the seabed node data to form a three-dimensional space sampling data set. The application realizes high-precision fusion of deep and shallow detection data and quantitative evaluation of cave stability, and significantly improves the detection reliability and engineering risk early warning capability of a complex marine karst area.

Description

Technical Field

[0001] This application relates to the field of data monitoring technology, and in particular to a method and system for processing transient electromagnetic collaborative exploration data of karst cave groups in marine areas. Background Technology

[0002] With the intensive construction of mega-scale cross-sea projects, the accurate detection of submarine cave systems in marine karst areas has become a core technical challenge concerning the safety of major infrastructure. According to statistics from the International Tunneling Association (ITA), approximately 34% of global submarine tunnel construction accidents in the past decade were induced by water inrushes from hidden caves, with an average direct economic loss of US$280 million per accident. As a potential carrier for deep-sea carbon sequestration, the stability assessment of cave systems is even more directly related to the geological sequestration safety of hundreds of billions of tons of CO2.

[0003] However, traditional transient electromagnetic methods face two major bottlenecks in their application in marine environments:

[0004] First, the multi-source heterogeneous data from dynamic towed platforms and static seabed nodes make it difficult to achieve spatial consistency in electrical structure inversion. The difference in spatiotemporal reference between the shallow high-resolution data of the towed array and the deep low-noise data of the seabed nodes leads to structural misalignment during fusion inversion. Second, existing karst cave stability assessments lack cross-scale coupling mechanisms for electrical and mechanical parameters. Conventional resistivity models cannot directly correlate with rock mass mechanical behavior, making it difficult to quantitatively predict the risk of karst cave vault instability.

[0005] Although signal processing technologies that combine motion compensation and noise suppression have been developed in recent years, they still cannot meet the accuracy requirements for multi-scale topological reconstruction of karst systems under strong marine environmental interference. Furthermore, stability prediction methods based on geostatistics have not yet established a physical correlation model between electrical conductivity distribution and rock mass compressive strength. Therefore, it is urgent to develop new methods for electromagnetic exploration data processing that are collaborative across multiple platforms to overcome the dual technical barriers of reconstructing the consistent electrical structure of shallow and deep parts and quantitatively assessing the mechanical stability of karst systems. Summary of the Invention

[0006] To address the aforementioned problems, embodiments of the present invention provide a method for processing transient electromagnetic collaborative exploration data of karst cave systems in marine areas, the method comprising:

[0007] Broadband excitation signals are transmitted to the seabed through a shipborne transient electromagnetic transmission system. Simultaneously, a towed receiving array is used to collect the secondary field attenuation signal of the seabed medium response. At the same time, a distributed sensor network is deployed at preset seabed nodes to collect background noise baseline data.

[0008] Motion attitude compensation is performed on the secondary field attenuation signal to generate transient electromagnetic response data; an adaptive filter is constructed based on the background noise baseline data to perform noise suppression processing on the transient electromagnetic response data and output noise-suppressed data.

[0009] The noise suppression data of the towed array is spatiotemporally aligned with the seabed node data to form a three-dimensional spatial sampling dataset; the three-dimensional spatial sampling dataset is then subjected to multi-resolution joint inversion based on the equivalent eddy current diffusion model to reconstruct the three-dimensional electrical structure model of the seabed medium.

[0010] High-resistivity anomaly regions that satisfy the electrical characteristics of karst caves are extracted from the three-dimensional electrical structure model; connected regions are identified using morphological constraint clustering algorithm, and a spatial topology map of karst caves representing their spatial distribution and connectivity is constructed by combining prior geological structural information.

[0011] Furthermore, the motion attitude compensation includes: acquiring spatial pose parameters through the inertial measurement unit built into the receiving array, establishing a direction cosine matrix of the receiving coil, and performing vector rotation correction on the secondary field attenuation signal to eliminate signal distortion caused by tilt angle.

[0012] Furthermore, the method for constructing the adaptive filter includes: establishing an equivalent conductivity and frequency response model with seawater depth and salinity gradient as variables based on the electric field and magnetic field noise spectrum characteristics of the background noise substrate data, and designing a Wiener filter bank with frequency and orientation selectivity accordingly.

[0013] Furthermore, the multi-resolution joint inversion method includes: using fast quasi-two-dimensional inversion to obtain shallow electrical structure from towed array noise suppression data, using full three-dimensional finite element inversion to obtain deep electrical structure from seabed node data, and forcing structural consistency between the two inversion results in overlapping areas through cross-gradient constraint terms.

[0014] Furthermore, the cross-gradient constraint term achieves structural consistency control through a multi-fidelity deep kernel learning framework, specifically including:

[0015] The high-precision conductivity values ​​obtained by inverting seabed node data are used as high-fidelity sampling points.

[0016] Constructing a hyperbolic tangent kernel function Deep Gaussian process model: In the formula, It is a three-dimensional spatial coordinate vector. Scaling the covariance matrix for features Kernel function length scale parameter, This is the signal variance hyperparameter. This is the noise variance hyperparameter. The Kronecker function;

[0017] The pseudo-two-dimensional conductivity distribution inverted by the towed array is mapped to the same kernel-induced feature space as the seabed node data through variational inference.

[0018] Calculate the cosine similarity of the gradient field angle in the feature space. As a measure of structural consistency:

[0019] ;

[0020] In the formula, To simulate a two-dimensional inverted conductivity gradient field, For the full three-dimensional inversion of the conductivity gradient field, For the regenerating kernel Hilbert space product operator, For the regenerating kernel Hilbert space norm operator.

[0021] Furthermore, the morphologically constrained clustering algorithm includes:

[0022] Calculation of Helmholtz decomposition tensor field based on three-dimensional electrical structure model;

[0023] Anisotropic diffusion level set evolution is used to extract closed high-resistivity regions;

[0024] A multi-scale topological framework for the cave system was constructed using persistent homology analysis.

[0025] Furthermore, methods for identifying cave filling materials include:

[0026] Extract the multi-relaxation time spectrum of each cave unit;

[0027] Wasserstein distance between the measured spectrum and the characteristic spectrum of a typical filling material was calculated using optimal transport theory. :

[0028] ;

[0029] In the formula, and Defined in the real number field Probability measure on for and The set of all joint probability distributions for and Transmission planning between them, for defining in Probability measure on For joint probability measure The differential form, and Representing the spectrum and The relaxation time variable, | for and of Order cost function, This is the open bound operator.

[0030] Furthermore, methods for identifying cave filling materials also include cave system stability assessment:

[0031] Calculation of key arch structure parameters based on the topological relationship diagram of the karst cave space ,in, Let be the radius of curvature. The arch span ratio;

[0032] Through resistivity-intensity correlation model Estimate the compressive strength of the rock mass, among which and For regional calibration coefficients;

[0033] Discrete element numerical simulation is used to predict the probability of structural instability under critical water pressure. :

[0034] ;

[0035] In the formula, For the number of simulations, For the first The tensile stress at the crown of the arch in this simulation. To correspond to compressive strength, For indicator functions, For indexing.

[0036] Furthermore, the method also includes:

[0037] The noise suppression data of the towed array and the background noise baseline data collected by the seabed nodes are spatiotemporally correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream.

[0038] A transient electromagnetic collaborative exploration data processing system for karst cave systems in marine areas, comprising:

[0039] A broadband excitation acquisition module transmits a broadband excitation signal to the seabed through a shipborne transient electromagnetic transmission system, and simultaneously uses a towed receiving array to acquire the secondary field attenuation signal of the seabed medium response. At the same time, a distributed sensor network is deployed at preset seabed nodes to acquire background noise baseline data.

[0040] The motion noise suppression module performs motion attitude compensation on the secondary field attenuation signal to generate transient electromagnetic response data; it constructs an adaptive filter based on the background noise baseline data to perform noise suppression processing on the transient electromagnetic response data and outputs noise-suppressed data.

[0041] The multi-source 3D inversion module aligns the noise suppression data of the towed array with the seabed node data in time and space to form a 3D spatial sampling dataset; and performs multi-resolution joint inversion on the 3D spatial sampling dataset based on the equivalent eddy current diffusion model to reconstruct the 3D electrical structure model of the seabed medium.

[0042] The cave topology reconstruction module extracts high-resistivity anomaly regions that satisfy the electrical characteristics of the cave from the three-dimensional electrical structure model; identifies connected regions using a morphological constraint clustering algorithm; and constructs a cave spatial topology map characterizing the spatial distribution and connectivity of the cave by combining prior geological structural information.

[0043] The technical effects and advantages of the transient electromagnetic collaborative exploration data processing method for karst cave groups in marine areas provided by this invention are as follows:

[0044] This invention breaks through the limitations of traditional transient electromagnetic methods, achieving high-precision fusion of shallow and deep exploration data and quantitative assessment of cave stability, significantly improving the reliability of exploration and engineering risk early warning capabilities in complex marine karst areas. Through motion attitude synchronization compensation and filtering technology, this invention eliminates the spatiotemporal reference differences between the towed platform and seabed nodes, solving the structural misalignment problem in dynamic and static data fusion, and achieving seamless coupling between shallow fine structures and deep macroscopic features. A physical correlation model between resistivity and rock mass strength is established, and combined with discrete element simulation technology, electromagnetic detection results are directly converted into the probability of cave vault instability for the first time, breaking through the limitation of traditional methods that can only provide qualitative assessments. The anti-interference algorithm effectively suppresses the effects of ocean current disturbances and salinity abrupt changes, maintaining sub-meter-level accuracy in identifying cave boundaries even in strong noise environments. Helmholtz field decomposition technology accurately identifies the type of infill material, significantly reducing the misjudgment rate of cave system topology reconstruction. Attached Figure Description

[0045] Figure 1 This is a flowchart of the transient electromagnetic collaborative exploration data processing method for karst cave groups in marine areas in Example 1;

[0046] Figure 2 This is a flowchart of the transient electromagnetic collaborative exploration data processing method for karst cave groups in marine areas in Example 2;

[0047] Figure 3 This is a schematic diagram of the connection of the transient electromagnetic collaborative exploration data processing system for the karst cave system in the marine area in Example 3. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Example 1:

[0050] Please see Figure 1 As shown, embodiments of the present invention provide a method for processing transient electromagnetic collaborative exploration data of karst cave systems in marine areas, the method comprising:

[0051] S1. A wide-spectrum excitation signal is transmitted to the seabed through a shipborne transient electromagnetic transmission system. Simultaneously, a towed receiving array is used to collect the secondary field attenuation signal of the seabed medium response. At the same time, a distributed sensor network is deployed at preset seabed nodes to collect background noise baseline data.

[0052] S2. Perform motion attitude compensation on the secondary field attenuation signal to generate transient electromagnetic response data; construct an adaptive filter based on the background noise baseline data, perform noise suppression processing on the transient electromagnetic response data, and output noise-suppressed data.

[0053] S3. The noise suppression data of the towed array is spatiotemporally aligned with the seabed node data to form a three-dimensional spatial sampling dataset; based on the equivalent eddy current diffusion model, the three-dimensional spatial sampling dataset is subjected to multi-resolution joint inversion to reconstruct the three-dimensional electrical structure model of the seabed medium.

[0054] S4. Extract high-resistivity anomaly regions that satisfy the electrical characteristics of karst caves from the three-dimensional electrical structure model; identify connected regions using morphological constraint clustering algorithm, and construct a karst cave spatial topology map that characterizes the spatial distribution and connectivity of karst caves by combining prior geological structural information.

[0055] Methods for motion posture compensation include:

[0056] Spatial pose parameters are obtained by the inertial measurement unit built into the receiving array, and the direction cosine matrix of the receiving coil is established. Vector rotation correction is performed on the secondary field attenuation signal to eliminate signal distortion caused by tilt angle.

[0057] In the exploration of karst cave systems in marine areas, the towed receiving array undergoes spatial attitude changes due to the movement of the ship, resulting in vector distortion of the acquired secondary field attenuation signal. Therefore, motion attitude compensation is required, and specific methods for motion attitude compensation include:

[0058] The inertial measurement unit built into the receiver array acquires three-axis angular velocity, three-axis acceleration, and three-axis geomagnetic data in real time. These parameters together constitute nine-degree-of-freedom spatial pose information, including:

[0059] Angular velocity reflects the rotational speed of the receiving coil around the carrier coordinate axis;

[0060] Acceleration characterizes the linear motion tendency of an array under wave action;

[0061] The geomagnetic vector provides an absolute direction reference (with the geomagnetic north pole pointing as a reference);

[0062] Based on sensor fusion algorithms (such as quaternion filtering), the pose parameters are converted into a direction cosine matrix. This matrix is ​​essentially a three-dimensional rotation operator, and its elements represent the direction cosine relationship between the coordinate axes of the local coordinate system of the receiving coil and the geodetic coordinate system. Specifically:

[0063] The first row of the matrix describes the cosine of the angle between the coil's X-axis and the geographic north, east, and vertical directions; the second row corresponds to the directional projection of the coil's Y-axis; and the third row characterizes the spatial orientation of the coil's Z-axis (usually the effective receiving axis).

[0064] The original secondary field signal is transformed from the coil coordinate system to the geodetic coordinate system. This transformation eliminates signal component projection errors caused by array tilt, ensuring that the received signal always reflects the electromagnetic field vector in the true geodetic coordinate system. For example:

[0065] When the towed array encounters a lateral current and rolls 30°:

[0066] Original signal B raw =(15.2, -4.3, 6.8)nT (Y component is affected by tilt angle).

[0067] After rotation correction using the direction cosine matrix: B corrected =(12.1, 0.5, 8.9) nT Compared with the seabed node reference value (12.0, 0, 9.0) nT, the signal error is reduced after correction.

[0068] Using static field data synchronously collected by the seabed node network as a benchmark, the effect was verified by calculating the Euclidean distance ratio between the signal and the benchmark value before and after motion compensation.

[0069] The aforementioned motion posture compensation method provides distortion-free electromagnetic response data for subsequent cave identification, and in particular ensures the accurate extraction of the boundaries of high-resistivity anomaly regions.

[0070] The method for constructing adaptive filters includes: establishing an equivalent conductivity and frequency response model with seawater depth and salinity gradient as variables based on the electric field and magnetic field noise spectrum characteristics of the background noise substrate data, and designing Wiener filter banks with frequency and orientation selectivity based on this model.

[0071] In deep-sea transient electromagnetic detection, the background noise generated by the dynamic environment of seawater has significant spatiotemporal heterogeneity. This method achieves dynamic tracking and precise suppression of noise patterns by intelligently fusing multi-source environmental data.

[0072] Dynamic modeling methods for noise characteristics include:

[0073] Based on the electromagnetic background field acquired in real time by the seabed array, water salinity and temperature profile data are simultaneously fused; a noise frequency fingerprint database is constructed to identify characteristic spectral patterns such as low-frequency swells and mid-frequency eddies; a spatial mapping relationship between noise intensity and hydrological environment is established to capture noise jump phenomena in salinity abrupt change layers; and areas with strong interference (such as ocean current confluence zones and bioaccumulation zones) are dynamically labeled.

[0074] Methods for constructing equivalent conductivity and frequency response models include:

[0075] Construct a noise propagation model with depth h and salinity s as variables:

[0076] ;

[0077] In the formula, To characterize the nonlinear variation of seawater conductivity with depth and salinity, m is the frequency attenuation exponent. This indicates the baseline shift caused by salinity. It represents the equivalent conductivity.

[0078] This technology, through the deep coupling of environmental perception and signal processing, successfully solves the noise suppression bottleneck of traditional methods in salinity jump zones and active ocean current zones, enabling a breakthrough in the ability to restore geological details in deep-sea resistivity imaging, and providing a reliable data base for the topological analysis of complex karst systems.

[0079] The multi-resolution joint inversion (S3) method includes: using fast quasi-2D inversion to obtain shallow electrical structure from noise-suppressed data of towed arrays, using full 3D finite element inversion to obtain deep electrical structure from seabed node data, and forcing structural consistency between the two inversion results in overlapping areas through cross gradient constraint terms.

[0080] Based on the electromagnetic data with noise suppression, a hierarchical collaborative inversion is implemented to address the heterogeneity of the towed array and the seabed node acquisition system. The continuous scanning data of the towed array is processed by a fast quasi-two-dimensional inversion technique, which involves constructing a dynamic sliding window along the survey line and extracting shallow electrical structures based on the wave equation characteristics. This method utilizes the high-density sampling advantage of the track direction to accurately capture thin electrical interfaces within a scale of hundreds of meters below the seabed, such as high-resistivity cemented zones or low-resistivity fluid channels in sedimentary layers. Its computational efficiency meets the requirements for real-time processing of survey lines spanning thousands of kilometers.

[0081] The broadband electromagnetic field data deployed at the seabed nodes are analyzed using full three-dimensional finite element inversion. By constructing an anisotropic mesh model, the spatial distribution of deep structures is finely characterized. The spatial distribution characteristics of the node array effectively enhance the ability to identify the location of deep anomalies, and are particularly suitable for the electrical reconstruction of complex structures such as ocean current channels and basement fractures at depths of up to 1,000 meters. The mesh density is adaptively adjusted during the inversion process, and local densification is achieved in areas of structural abrupt changes to ensure the geometric fidelity of key features such as cave boundaries.

[0082] To eliminate the conflict in the solution sets of the two types of inversions in the transition region, a structural consistency fusion mechanism is introduced. In the overlapping area of ​​the towed array and the seabed node (typical depth of 150-300 meters), a cross gradient constraint field is established. This constraint forces the electrical gradient vectors of the two inversion results to maintain directional consistency at the same point in space. When the dip angle of the shallow strata shown by the quasi-2D inversion deviates significantly from the strike of the deep structure constructed by the 3D inversion, the constraint term automatically increases its weight, driving the inversion system to rebalance the contribution ratio of the data fitting term and the structural coupling term.

[0083] By hierarchical analysis and structural field fusion of heterogeneous data, the inherent contradiction between vertical resolution and spatial coverage in traditional inversion methods is overcome. For the thin-layer-cavity system unique to salinity stratification areas, cross-gradient constraints enable accurate reconstruction of the spatial topological relationship between shallow cemented layers and deep cavities. For example, the inversion results of a target area show that vertically extending cavities are developed below a shallow high-resistivity thin layer (about 8 meters thick). This structural model was verified by subsequent core sampling as a paleofluid transport path, confirming the analytical capability of joint inversion for complex geological systems.

[0084] The cross-gradient constraint term achieves structural consistency control through a multi-fidelity deep kernel learning framework, specifically including:

[0085] S101. Use the high-precision conductivity values ​​obtained by inverting the seabed node data as high-fidelity sampling points.

[0086] S102. Construct a kernel based on the hyperbolic tangent kernel function. Deep Gaussian process model: In the formula, It is a three-dimensional spatial coordinate vector. Scaling the covariance matrix for features Kernel function length scale parameter, This is the signal variance hyperparameter. This is the noise variance hyperparameter. The Kronecker function;

[0087] S103. The pseudo-two-dimensional conductivity distribution inverted by the towed array is mapped to the same nuclear induced feature space as the seabed node data through variational inference.

[0088] S104. Calculate the cosine similarity of the gradient field angle in the feature space. As a measure of structural consistency:

[0089] ;

[0090] In the formula, To simulate a two-dimensional inverted conductivity gradient field, For the full three-dimensional inversion of the conductivity gradient field, For the regenerating kernel Hilbert space product operator, For the regenerating kernel Hilbert space norm operator.

[0091] The above method utilizes deep kernel learning to intelligently fuse multi-precision data in the feature space, and adapts to the complex morphology of the cave through nonlinear kernel functions, forcing structural consistency and significantly improving inversion accuracy and efficiency.

[0092] Morphologically constrained clustering algorithms include:

[0093] S201. Calculation of Helmholtz decomposition tensor field based on three-dimensional electrical structure model. :

[0094] ;

[0095] In the formula, For three-dimensional spatial position coordinates, For three-dimensional vector differential operators, Let be a scalar potential function. It is a vector potential function;

[0096] S202. Anisotropic diffusion level set evolution is used to extract closed high-resistivity regions;

[0097] S203. Construct a multi-scale topological framework for the cave system through persistent homology analysis.

[0098] This completes the segmentation of the karst cave units. After segmentation, material composition analysis is performed on each independent karst cave unit, i.e., karst cave filling material identification.

[0099] The algorithm described above distinguishes real cave boundaries through Helmholtz decomposition, adapts anisotropic level sets to complex shapes, and automatically constructs multi-scale topological skeletons with persistent homology, significantly improving the accuracy and robustness of identifying the spatial distribution and connectivity of caves.

[0100] The steps for identifying cave filling materials include:

[0101] S301. Extract the multi-relaxation time spectrum of each cave unit;

[0102] S302. Calculate the Wasserstein distance between the measured spectrum and the characteristic spectrum of a typical filling material using optimal transport theory. :

[0103] ;

[0104] In the formula, and Defined in the real number field Probability measure on for and The set of all joint probability distributions for and Transmission planning between them, for defining in Probability measure on For joint probability measure The differential form, and Representing the spectrum and The relaxation time variable, | for and of Order cost function, To remove the bounds operator, we need to find the minimum transmission cost.

[0105] The above discrimination method utilizes Wasserstein distance to quantify the overall morphological differences of the multi-relaxation spectrum, which is highly robust to noise and can distinguish infills with similar composition but different structures (such as muddy gravel), significantly improving discrimination accuracy.

[0106] The process for identifying cave filling materials also includes a stability assessment of the cave system:

[0107] S401. Calculation of key arch structure parameters based on the topological relationship diagram of karst cave space. ,in, Let be the radius of curvature. The arch span ratio;

[0108] S402, Using the resistivity-intensity correlation model Estimate the compressive strength of the rock mass, among which and For regional calibration coefficients;

[0109] S403. Predicting the probability of structural instability under critical water pressure using discrete element numerical simulation. :

[0110] ;

[0111] In the formula, For the number of simulations, For the first The tensile stress at the crown of the arch in this simulation. To correspond to compressive strength, For indicator functions, For indexing.

[0112] The above assessment method integrates spatial topological vault parameters, resistivity intensity inversion, and discrete element hydraulic simulation to achieve dynamic assessment of instability probability through multi-factor coupling, significantly improving the accuracy of engineering risk early warning.

[0113] Example 2:

[0114] like Figure 2 As shown, this embodiment further improves the design based on Embodiment 1. The difference is that in the actual operation of Embodiment 1, it was found that there is a spatiotemporal reference drift between the towed array and the seabed nodes. The spatiotemporal reference drift is directly caused by the technical gap between S2 (motion attitude compensation) and S3 (spatiotemporal alignment) of the original method. That is, it only compensates for the motion error of a single device and does not solve the cross-device coordination error. This leads to the possibility of ghosting of the cave boundary in the three-dimensional electrical model after spatiotemporal alignment in step S3, resulting in misjudgment of the cave boundary. Based on this, the transient electromagnetic collaborative exploration data processing method for seawater cave groups also includes:

[0115] The noise suppression data of the towed array output by S2 and the background noise baseline data collected by the seabed nodes are spatiotemporally correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream as input to S3.

[0116] After generating the noise suppression data for the towed array in step S2, it needs to be spatiotemporally fused with the background noise baseline data collected from the seabed nodes. The specific implementation process is as follows:

[0117] The towed array generates electromagnetic response data during its movement, while the fixed seabed nodes continuously record environmental noise. Because the two types of equipment have independent clock sources and dynamic spatial relationships, direct fusion would lead to blurred cave boundaries in the 3D inversion. Therefore, two-way acoustic timing and Doppler positioning technology is employed, including:

[0118] The towed array periodically transmits acoustic pulse signals, and the seabed nodes immediately return response signals after receiving them. By calculating the propagation delay of the two-way signals, the clock deviation between platforms can be accurately measured. For example, in a survey of the Pearl River Estuary, the initial clock deviation caused a 3.2-millisecond misalignment in the data of adjacent survey lines. After real-time calibration, the timing alignment accuracy reached the microsecond level.

[0119] By utilizing the acoustic Doppler frequency shift effect generated by the towed array's navigation, and combining the known fixed coordinates of the seabed nodes (calibrated by the acoustic positioning network in step S1), the spatial offset vector of the array relative to the nodes is dynamically calculated.

[0120] The calibrated spatiotemporal reference is applied to the original dataset, including:

[0121] Add spatiotemporal stamps (including time synchronization calibration time and Doppler positioning coordinates) to the noise suppression data of the towed array; add node location stamps to the background noise baseline data of the seabed nodes to form a joint observation data stream organized in a three-dimensional grid. Each grid cell in this data stream contains the array-corrected electromagnetic response (from S2 output), the node background noise spectrum (from S1 output), and spatiotemporal registration parameters.

[0122] Example 3:

[0123] like Figure 3 As shown, based on the same inventive concept as the transient electromagnetic collaborative exploration data processing method for karst cave systems in the aforementioned embodiments, this application provides a transient electromagnetic collaborative exploration data processing system for karst cave systems in the sea. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0124] The broadband excitation acquisition module transmits broadband excitation signals to the seabed through a shipborne transient electromagnetic transmission system. Simultaneously, it uses a towed receiving array to acquire the secondary field attenuation signal of the seabed medium response. At the same time, it deploys a distributed sensor network at preset seabed nodes to acquire background noise baseline data.

[0125] The motion noise suppression module performs motion attitude compensation on the secondary field attenuation signal to generate transient electromagnetic response data; it constructs an adaptive filter based on the background noise baseline data to perform noise suppression processing on the transient electromagnetic response data and outputs noise-suppressed data.

[0126] The multi-source 3D inversion module aligns the noise suppression data of the towed array with the seabed node data in time and space to form a 3D spatial sampling dataset. Based on the equivalent eddy current diffusion model, the 3D spatial sampling dataset is subjected to multi-resolution joint inversion to reconstruct the 3D electrical structure model of the seabed medium.

[0127] The cave topology reconstruction module extracts high-resistivity anomaly regions that satisfy the electrical characteristics of the cave from the three-dimensional electrical structure model; it uses a morphological constraint clustering algorithm to identify connected regions and combines prior geological structural information to construct a cave spatial topology map that characterizes the spatial distribution and connectivity of the cave.

[0128] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0129] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present application, based on the technical solution and concept of the present application, should be covered within the scope of protection of the present application.

Claims

1. A method for processing transient electromagnetic co-existing survey data of karst cave systems in marine areas, characterized in that, The methods include: Broadband excitation signals are transmitted to the seabed through a shipborne transient electromagnetic transmission system. Simultaneously, a towed receiving array is used to collect the secondary field attenuation signal of the seabed medium response. At the same time, a distributed sensor network is deployed at preset seabed nodes to collect background noise baseline data. Motion attitude compensation is performed on the secondary field attenuation signal to generate transient electromagnetic response data; An adaptive filter is constructed based on the background noise baseline data to perform noise suppression processing on the transient electromagnetic response data and output noise-suppressed data. The noise suppression data of the towed array is spatiotemporally aligned with the seabed node data to form a three-dimensional spatial sampling dataset; A multi-resolution joint inversion is performed on a three-dimensional spatial sampling dataset based on an equivalent eddy current diffusion model to reconstruct the three-dimensional electrical structure model of the seabed medium. The multi-resolution joint inversion method includes: using fast quasi-two-dimensional inversion to obtain the shallow electrical structure from towed array noise-suppressed data, and using full three-dimensional finite element inversion to obtain the deep electrical structure from seabed node data; a cross-gradient constraint term is used to force structural consistency between the two inversion results in overlapping regions. The cross-gradient constraint term achieves structural consistency control through a multi-fidelity deep kernel learning framework, specifically including: The high-precision conductivity values ​​obtained by inverting seabed node data are used as high-fidelity sampling points. Constructing a hyperbolic tangent kernel function Deep Gaussian process model: In the formula, It is a three-dimensional spatial coordinate vector. Scaling the covariance matrix for features Kernel function length scale parameter, This is the signal variance hyperparameter. This is the noise variance hyperparameter. The Kronecker function; The pseudo-two-dimensional conductivity distribution inverted by the towed array is mapped to the same kernel-induced feature space as the seabed node data through variational inference. Calculate the cosine similarity of the gradient field angle in the feature space. As a measure of structural consistency: ； In the formula, To simulate a two-dimensional inverted conductivity gradient field, For the full three-dimensional inversion of the conductivity gradient field, For the regenerating kernel Hilbert space product operator, For the regenerating kernel Hilbert space norm operator; High-resistivity anomaly regions that satisfy the electrical characteristics of karst caves are extracted from the three-dimensional electrical structure model; connected regions are identified using morphological constraint clustering algorithm, and a spatial topology map of karst caves representing their spatial distribution and connectivity is constructed by combining prior geological structural information.

2. The method for processing transient electromagnetic collaborative exploration data of karst cave systems in marine areas according to claim 1, characterized in that, The motion attitude compensation includes: acquiring spatial pose parameters through the inertial measurement unit built into the receiving array, establishing the direction cosine matrix of the receiving coil, and performing vector rotation correction on the secondary field attenuation signal to eliminate signal distortion caused by tilt angle.

3. The method for processing transient electromagnetic co-survey data of karst cave systems in marine areas according to claim 1, characterized in that, The method for constructing the adaptive filter includes: establishing an equivalent conductivity and frequency response model with seawater depth and salinity gradient as variables based on the electric field and magnetic field noise spectrum characteristics of the background noise substrate data, and designing a Wiener filter bank with frequency and orientation selectivity accordingly.

4. The method for processing transient electromagnetic collaborative exploration data of karst cave systems in marine areas according to claim 1, characterized in that, The morphologically constrained clustering algorithm includes: Calculation of Helmholtz decomposition tensor field based on three-dimensional electrical structure model; Anisotropic diffusion level set evolution is used to extract closed high-resistivity regions; A multi-scale topological framework for the cave system was constructed using persistent homology analysis.

5. The method for processing transient electromagnetic co-survey data of karst cave systems in marine areas according to claim 4, characterized in that, Methods for identifying cave filling materials include: Extract the multi-relaxation time spectrum of each cave unit; Wasserstein distance between the measured spectrum and the characteristic spectrum of a typical filling material was calculated using optimal transport theory. : ； In the formula, and Defined in the real number field Probability measure on for and The set of all joint probability distributions for and Transmission planning between them, for defining in Probability measure on For joint probability measure The differential form, and Representing the spectrum and The relaxation time variable, | for and of Order cost function, This is the open bound operator.

6. The method for processing transient electromagnetic collaborative exploration data of karst cave groups in marine areas according to claim 5, characterized in that, Methods for identifying cave filling materials also include cave system stability assessment: Calculation of key arch structure parameters based on the topological relationship diagram of the karst cave space ,in, Let be the radius of curvature. The arch span ratio; Through resistivity-intensity correlation model Estimate the compressive strength of the rock mass, among which and For regional calibration coefficients; Discrete element numerical simulation is used to predict the probability of structural instability under critical water pressure. : ； In the formula, For the number of simulations, For the first The tensile stress at the crown of the arch in this simulation. To correspond to compressive strength, For indicator functions, For indexing.

7. The method for processing transient electromagnetic collaborative exploration data of karst cave systems in marine areas according to claim 1, characterized in that, The method also includes: The noise suppression data of the towed array and the background noise baseline data collected by the seabed nodes are spatiotemporally correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream.

8. A transient electromagnetic collaborative exploration data processing system for karst cave systems in marine areas, characterized in that the system... include: A broadband excitation acquisition module transmits a broadband excitation signal to the seabed through a shipborne transient electromagnetic transmission system, and simultaneously uses a towed receiving array to acquire the secondary field attenuation signal of the seabed medium response. At the same time, a distributed sensor network is deployed at preset seabed nodes to acquire background noise baseline data. A motion noise suppression module performs motion attitude compensation on the secondary field attenuation signal to generate transient electromagnetic response data. An adaptive filter is constructed based on the background noise baseline data to perform noise suppression processing on the transient electromagnetic response data and output noise-suppressed data. The multi-source three-dimensional inversion module aligns the noise suppression data of the towed array with the seabed node data in time and space to form a three-dimensional spatial sampling dataset. A multi-resolution joint inversion is performed on a three-dimensional spatial sampling dataset based on an equivalent eddy current diffusion model to reconstruct the three-dimensional electrical structure model of the seabed medium. The multi-resolution joint inversion method includes: using fast quasi-two-dimensional inversion to obtain the shallow electrical structure from towed array noise-suppressed data, and using full three-dimensional finite element inversion to obtain the deep electrical structure from seabed node data; a cross-gradient constraint term is used to force structural consistency between the two inversion results in overlapping regions. The cross-gradient constraint term achieves structural consistency control through a multi-fidelity deep kernel learning framework, specifically including: The high-precision conductivity values ​​obtained by inverting seabed node data are used as high-fidelity sampling points. Constructing a hyperbolic tangent kernel function Deep Gaussian process model: In the formula, It is a three-dimensional spatial coordinate vector. Scaling the covariance matrix for features Kernel function length scale parameter, This is the signal variance hyperparameter. This is the noise variance hyperparameter. The Kronecker function; The pseudo-two-dimensional conductivity distribution inverted by the towed array is mapped to the same kernel-induced feature space as the seabed node data through variational inference. Calculate the cosine similarity of the gradient field angle in the feature space. As a measure of structural consistency: ； In the formula, To simulate a two-dimensional inverted conductivity gradient field, For the full three-dimensional inversion of the conductivity gradient field, For the regenerating kernel Hilbert space product operator, For the regenerating kernel Hilbert space norm operator; The cave topology reconstruction module extracts high-resistivity anomaly regions that satisfy the electrical characteristics of the cave from the three-dimensional electrical structure model; identifies connected regions using a morphological constraint clustering algorithm; and constructs a cave spatial topology map characterizing the spatial distribution and connectivity of the cave by combining prior geological structural information.

Images