Sea water area karst cave group transient electromagnetic collaborative investigation data processing method and system

Abstract

The invention discloses a sea water area karst cave group transient electromagnetic collaborative investigation data processing method and system, and the method comprises the steps: transmitting a wide frequency spectrum excitation signal to the seabed through a shipborne transient electromagnetic transmitting system, and synchronously collecting a secondary field attenuation signal responded by a seabed medium through a pull-type receiving array, meanwhile, a distributed sensor network is arranged at a preset seabed node to collect background noise base data; performing motion attitude compensation on the secondary field attenuation signal 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. According to the method, high-precision fusion of deep and shallow detection data and quantitative evaluation of the stability of the karst cave are realized, and the detection reliability and the engineering risk early warning capability of the complex ocean karst area are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of data monitoring technology, and in particular to a method and system for processing data of transient electromagnetic collaborative survey of marine cave groups. Background Art

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

[0003] However, traditional transient electromagnetic methods face two major bottlenecks in marine applications: First, it is difficult to achieve spatial consistency in electrical structural inversion using multi-source heterogeneous data from a dynamic towed platform and static seabed nodes. Differences in temporal and spatial benchmarks between shallow high-resolution data from the towed array and deep low-noise data from seabed nodes lead to structural misalignment during fusion inversion. Second, existing cave stability assessments lack a cross-scale coupling mechanism for electrical-mechanical parameters, and conventional resistivity models cannot directly correlate with the mechanical behavior of rock masses, making it difficult to provide quantitative early warning of the risk of cave vault instability.

[0004] Although signal processing technology combining motion compensation and noise suppression has been developed in recent years, it still cannot meet the accuracy requirements for multi-scale topological reconstruction of cave systems under strong marine environmental interference; and the stability prediction method based on geostatistics has not yet established a physical correlation model between conductivity distribution and rock compressive strength. Therefore, it is urgent to develop new methods for electromagnetic survey data processing that collaborate on multiple platforms to break through the dual technical barriers of reconstructing the consistency of electrical structures in deep and shallow areas and quantitatively evaluating the mechanical stability of cave systems. Summary of the Invention

[0005] In order to solve the above problems, an embodiment of the present invention provides a method for processing data of transient electromagnetic collaborative survey of marine cave groups, the method comprising: A ship-borne transient electromagnetic transmission system transmits a wide-spectrum excitation signal to the seabed, while a towed receiving array collects the secondary field attenuation signal of the seabed medium response. A distributed sensor network is deployed at pre-set seabed nodes to collect background noise floor data. Performing motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructing an adaptive filter based on the background noise floor data, performing noise suppression processing on the transient electromagnetic response data, and outputting noise suppressed data; The noise suppression data of the towed array is aligned with the seafloor node data in time and space to form a three-dimensional spatial sampling data set. Based on the equivalent eddy diffusion model, a multi-resolution joint inversion is performed on the three-dimensional spatial sampling data set to reconstruct the three-dimensional electrical structure model of the seafloor medium. High-resistance abnormal areas that meet the electrical characteristics of the cave are extracted from the three-dimensional electrical structure model; connected areas are identified using a morphological constrained clustering algorithm, and a cave spatial topological relationship diagram is constructed based on prior geological structure information to characterize the spatial distribution and connectivity of the cave.

[0006] Furthermore, the motion attitude compensation includes: obtaining spatial attitude parameters through the inertial measurement unit built into the receiving array, establishing a receiving coil direction cosine matrix, and performing vector rotation correction on the secondary field attenuation signal to eliminate signal distortion caused by the tilt angle.

[0007] 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 base data, and designing a Wiener filter group with frequency and azimuth selectivity based on this.

[0008] Furthermore, the multi-resolution joint inversion method includes: using fast quasi-two-dimensional inversion for towed array noise suppression data to obtain shallow electrical structure, using full three-dimensional finite element inversion for seabed node data to obtain deep electrical structure, and forcing the structural consistency of the two inversion results in the overlapping area through cross-gradient constraints.

[0009] Furthermore, the cross-gradient constraint term realizes structural consistency control through a multi-fidelity deep kernel learning framework, specifically including: The high-precision conductivity values ​​obtained by inverting the seabed node data are used as high-fidelity sampling points; Constructing a hyperbolic tangent kernel function Deep Gaussian process model of: ; Where, is the three-dimensional space coordinate vector, is the feature scaled covariance matrix, Kernel function length scale parameter, is the signal variance hyperparameter, is the noise variance hyperparameter, is the Kronecker function; The quasi-2D conductivity distribution inverted by the towed array is mapped to the same kernel-induced feature space as the seafloor node data through variational inference; Calculate the gradient field angle cosine similarity in feature space As a measure of structural consistency: ; Where, is the quasi-two-dimensional inversion conductivity gradient field, is the full three-dimensional inversion conductivity gradient field, is the inner product operator on the reproducing kernel Hilbert space, is the reproducing kernel Hilbert space norm operator.

[0010] Furthermore, the morphologically constrained clustering algorithm includes: Calculate the Helmholtz decomposition tensor field based on the three-dimensional electrical structure model; Anisotropic diffusion level set evolution is used to extract closed high-resistance areas; Constructing a multi-scale topological skeleton of a cave system through persistent homology analysis.

[0011] Furthermore, the cave filling identification method includes: Extract the multi-relaxation time spectrum of each cave unit; Calculation of the Wasserstein distance between the measured spectrum and the characteristic spectrum of typical fillings using optimal transport theory : ; Where, and Defined in the field of real numbers The probability measure on , for and The set of all joint probability distributions, for and The transmission plan between The probability measure on , is the joint probability measure The differential form of and Respectively represent the spectrum and The relaxation time variable, | for and of Order cost function, To remove the subfmimum operator.

[0012] Furthermore, the cave filling identification method also includes cave system stability assessment: Calculation of key vault structural parameters based on cave spatial topological relationship diagram ,in, is the radius of curvature, is the arch-span ratio; Through the resistivity-strength correlation model Estimate the compressive strength of rock mass, where and is the regional calibration coefficient; Predicting the probability of structural instability under critical water pressure using discrete element numerical simulation : ; Where, is the number of simulations, For the The tensile stress of the arch is simulated. is the corresponding compressive strength, is the indicator function, is the index.

[0013] Furthermore, the method further comprises: The noise suppression data of the towed array and the background noise floor data collected by the seabed nodes are temporally and spatially correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream.

[0014] Transient electromagnetic collaborative survey data processing system for marine cave groups, the system includes: A wide-spectrum excitation acquisition module transmits a wide-spectrum excitation signal to the seabed via a shipboard transient electromagnetic transmission system, simultaneously collects the secondary field attenuation signal of the seabed medium response using a towed receiving array, and simultaneously deploys a distributed sensor network at preset seabed nodes to collect background noise floor data; a motion noise suppression module, which performs motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructs an adaptive filter based on the background noise floor data, performs noise suppression processing on the transient electromagnetic response data, and outputs noise suppressed data; A multi-source 3D inversion module that performs spatiotemporal alignment of the noise-suppressed data from the towed array with the seafloor node data to form a 3D spatial sampling dataset; performs multi-resolution joint inversion on the 3D spatial sampling dataset based on an equivalent eddy diffusion model to reconstruct a 3D electrical structure model of the seafloor medium; A cave topology reconstruction module extracts high-resistance abnormal areas that meet the electrical characteristics of the cave from the three-dimensional electrical structure model; uses a morphological constrained clustering algorithm to identify connected areas, and combines geological structure prior information to construct a cave spatial topology relationship diagram that characterizes the spatial distribution and connectivity of the cave.

[0015] The technical effects and advantages of the method for processing data of transient electromagnetic collaborative survey of marine caves provided by the present invention are as follows: This invention breaks through the limitations of traditional transient electromagnetic methods, achieving high-precision fusion of deep and shallow detection data and quantitative assessment of cave stability, significantly improving the detection reliability and engineering risk warning capabilities of complex marine karst areas. Through motion posture synchronization compensation and filtering technology, this invention eliminates the time and space reference differences between the towed platform and the seabed node, solves the structural misalignment problem of dynamic and static data fusion, and achieves seamless coupling of shallow fine structures and deep macroscopic features; establishes a physical correlation model between resistivity and rock mass strength, and combines discrete element simulation technology to directly convert electromagnetic detection results into the probability of cave vault instability for the first time, breaking through the limitation of traditional methods that can only be qualitatively assessed; the anti-interference algorithm effectively suppresses the influence of ocean current disturbances and salinity mutations, maintaining sub-meter cave boundary recognition accuracy under strong noise background; Helmholtz field decomposition technology accurately identifies the filling type, significantly reducing the error rate of cave system topology reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the data processing method for transient electromagnetic collaborative survey of marine cave groups in Example 1; Figure 2 This is a flow chart of the data processing method for transient electromagnetic collaborative survey of marine cave groups in Example 2; Figure 3 This is a connection diagram of the transient electromagnetic collaborative survey data processing system for marine cave groups in Example 3. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] Example 1: See also Figure 1 As shown, an embodiment of the present invention provides a method for processing data of transient electromagnetic collaborative survey of marine cave groups, the method comprising: S1. A shipborne transient electromagnetic transmission system transmits a wide-spectrum excitation signal to the seabed, while a towed receiving array collects the secondary field attenuation signal of the seabed medium response. A distributed sensor network is deployed at pre-set seabed nodes to collect background noise floor data. S2. Performing motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructing an adaptive filter based on the background noise floor data, performing noise suppression processing on the transient electromagnetic response data, and outputting noise suppressed data; S3. Temporally and spatially align the noise suppression data of the towed array with the seafloor node data to form a three-dimensional spatial sampling dataset; perform multi-resolution joint inversion on the three-dimensional spatial sampling dataset based on the equivalent eddy diffusion model to reconstruct a three-dimensional electrical structure model of the seafloor medium; S4. Extract high-resistance abnormal areas that meet the electrical characteristics of the cave from the three-dimensional electrical structure model; use a morphological constrained clustering algorithm to identify connected areas, and combine prior geological structure information to construct a cave spatial topological relationship diagram that characterizes the spatial distribution and connectivity of the cave.

[0019] The methods of motion posture compensation include: The spatial posture parameters are obtained through the built-in inertial measurement unit of the receiving array, the direction cosine matrix of the receiving coil is established, and vector rotation correction is performed on the secondary field attenuation signal to eliminate the signal distortion caused by the inclination angle.

[0020] In marine cave surveys, the towed receiving array undergoes spatial posture changes as the ship moves, causing vector distortion in the collected secondary field attenuation signal. Therefore, motion posture compensation is required. Specific methods for motion posture compensation include: The inertial measurement unit built into the receiving array collects three-axis angular velocity, three-axis acceleration, and three-axis geomagnetic data in real time. These parameters together constitute the nine-degree-of-freedom spatial posture information, including: The angular velocity reflects the rotation rate of the receiving coil around the carrier coordinate axis; Acceleration characterizes the linear motion trend of the array under the action of waves; The geomagnetic vector provides an absolute direction reference (with reference to the direction of the geomagnetic north pole); 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. Its elements represent the direction cosine relationship between the local coordinate system of the receiving coil and the coordinate axis of the earth coordinate system. Specifically: 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 represents the spatial orientation of the coil's Z-axis (usually the effective receiving axis).

[0021] The original secondary field signal is converted from the coil coordinate system to the earth coordinate system. This transformation eliminates the projection error of the signal component caused by the tilt of the array, so that the received signal always reflects the electromagnetic field vector in the real earth coordinate system. For example: When the towed array encounters a lateral current and rolls 30°: Original signal B raw =(15.2, -4.3, 6.8)nT (Y component is disturbed by the inclination); After direction cosine matrix rotation correction: B corrected=(12.1, 0.5, 8.9) nT compared with the seabed node benchmark value (12.0, 0, 9.0) nT. The signal error is reduced after correction.

[0022] The static field data collected synchronously by the seabed node network is used as a benchmark, and the effect is verified by calculating the Euclidean distance ratio between the signal before and after motion compensation and the benchmark value.

[0023] The above-mentioned motion posture compensation method provides distortion-free electromagnetic response data for subsequent cave identification, especially ensuring the accurate extraction of the boundaries of high-resistance anomaly areas.

[0024] The method for constructing an adaptive filter includes: based on the electric and magnetic field noise spectrum characteristics of the background noise base data, establishing an equivalent conductivity and frequency response model with seawater depth and salinity gradient as variables, and designing a Wiener filter group with frequency and azimuth selectivity based on this.

[0025] In deep-sea transient electromagnetic detection, the background noise generated by the dynamic environment of seawater has significant temporal and spatial heterogeneity. This method realizes dynamic tracking and precise suppression of noise patterns by intelligently fusing multi-source environmental data.

[0026] Dynamic modeling methods for noise characteristics include: Based on the electromagnetic background field collected in real time by the seabed array, the water salinity and temperature profile data are synchronously integrated; a noise frequency fingerprint library is constructed to identify characteristic spectral patterns such as low-frequency surges and medium-frequency eddies; a spatial mapping relationship between noise intensity and hydrological environment is established to capture the noise jump phenomenon of the salinity mutation layer; and strong interference areas (such as ocean current confluence areas and biological enrichment zones) are dynamically marked.

[0027] The methods for constructing the equivalent conductivity and frequency response model include: Construct a noise propagation model with depth h and salinity s as variables: ; Where, To characterize the nonlinear change of seawater conductivity with depth and salinity, m is the frequency attenuation index, represents the baseline shift caused by salinity, is the equivalent conductivity.

[0028] Through the deep coupling of environmental perception and signal processing, this technology successfully solves the noise suppression bottleneck of traditional methods in salinity jump areas and active ocean current zones, making breakthrough progress in the ability to restore geological details of deep-sea resistivity imaging, and providing a reliable data base for the topological analysis of complex cave systems.

[0029] The multi-resolution joint inversion (S3) method includes: using fast pseudo-two-dimensional inversion on the noise-suppressed data of the towed array to obtain the shallow electrical structure, and using full three-dimensional finite element inversion on the seabed node data to obtain the deep electrical structure, and using cross-gradient constraints to enforce the structural consistency of the two inversion results in the overlapping area.

[0030] Based on the noise-suppressed electromagnetic data, hierarchical collaborative inversion is implemented to address the heterogeneous characteristics of the towed array and the seabed node acquisition system. The continuous scanning data of the towed array are processed using a fast pseudo-two-dimensional inversion technique, that is, a dynamic sliding window is constructed along the survey line, and the shallow electrical structure is extracted based on the characteristics of the wave equation. This method utilizes the advantages of high-density sampling in the track direction to sensitively capture thin electrical interfaces within a hundred meters below the seabed, such as high-resistance cementation zones or low-resistance fluid channels in sedimentary layers. Its computational efficiency meets the needs of real-time processing of thousand-kilometer survey lines.

[0031] The broadband electromagnetic field data of the seabed nodes are analyzed using full three-dimensional finite element inversion. By constructing an anisotropic grid model, the spatial distribution of deep structures is precisely depicted. The spatial distribution characteristics of the node array effectively enhance the ability to identify the orientation of deep anomalies, and are particularly suitable for the electrical reconstruction of complex structures such as ocean current channels and basement faults at depths of kilometers. The grid density is adaptively adjusted during the inversion process, and local encryption is achieved in structural mutation areas to ensure the geometric fidelity of key features such as cave boundaries.

[0032] In order to eliminate the conflict of solutions between the two types of inversion in the transition area, a structural consistency fusion mechanism is introduced; in the detection overlap area between the towed array and the seabed node (typical depth is 150-300 meters), a cross-gradient constraint field is established. This constraint forces the electrical gradient vectors of the two inversion results at the same point in space to maintain directional consistency. When the shallow stratum dip shown by the pseudo-two-dimensional inversion deviates significantly from the deep structure strike constructed by the three-dimensional 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.

[0033] Through the hierarchical analysis and structural field fusion of heterogeneous data, the inherent contradiction between vertical resolution and spatial coverage of traditional inversion methods has been overcome. For the thin layer-cave system unique to the salinity stratification zone, cross-gradient constraints have enabled the accurate reconstruction of the spatial topological relationship between the shallow cementation layer and the deep cave group. For example, the inversion results of a certain target area showed that a vertically extending cave channel developed below the shallow high-resistance thin layer (about 8 meters thick). This structural pattern was verified by subsequent core sampling as a paleofluid migration path, confirming the joint inversion's ability to analyze complex geological systems.

[0034] The cross-gradient constraint achieves structural consistency control through a multi-fidelity deep kernel learning framework, specifically including: S101, using the high-precision conductivity value obtained by inverting the seabed node data as a high-fidelity sampling point; S102, construct a hyperbolic tangent kernel function Deep Gaussian process model of: ; Where, is the three-dimensional space coordinate vector, is the feature scaled covariance matrix, Kernel function length scale parameter, is the signal variance hyperparameter, is the noise variance hyperparameter, is the Kronecker function; S103, mapping the quasi-two-dimensional conductivity distribution inverted by the towed array into the same kernel-induced feature space as the seafloor node data through variational inference; S104, calculating the gradient field angle cosine similarity in the feature space As a measure of structural consistency: ; Where, is the quasi-two-dimensional inversion conductivity gradient field, is the full three-dimensional inversion conductivity gradient field, is the inner product operator on the reproducing kernel Hilbert space, is the reproducing kernel Hilbert space norm operator.

[0035] The above method uses deep kernel learning to intelligently fuse multi-precision data in the feature space, adapts to the complex morphology of caves through nonlinear kernel functions, enforces structural consistency, and significantly improves inversion accuracy and efficiency.

[0036] Morphologically constrained clustering algorithms include: S201. Calculation of Helmholtz decomposition tensor field based on three-dimensional electrical structure model : ; Where, is the three-dimensional space position coordinate, is a three-dimensional vector differential operator, is the scalar potential function, is the vector potential function; S202, extracting closed high-resistance areas using anisotropic diffusion level set evolution; S203. Construct a multi-scale topological skeleton of the cave system through persistent homology analysis.

[0037] In this way, the segmentation of the cave units is completed. After the segmentation of the cave units, the material composition analysis is carried out on each independent cave unit, that is, the cave filling identification.

[0038] The above algorithm distinguishes the real cave boundaries through Helmholtz decomposition, adapts to complex shapes through anisotropic level sets, and automatically constructs a multi-scale topological skeleton through persistent homology, significantly improving the accuracy and robustness of identifying the spatial distribution and connectivity of caves.

[0039] The steps for identifying cave fillings include: S301, extracting the multi-relaxation time spectrum of each cave unit; S302. Calculate the Wasserstein distance between the measured spectrum and the characteristic spectrum of typical fillings using the optimal transport theory : ; Where, and Defined in the field of real numbers The probability measure on , for and The set of all joint probability distributions, for and The transmission plan between The probability measure on , is the joint probability measure The differential form of and Respectively represent the spectrum and The relaxation time variable, | for and of Order cost function, To remove the subfmimum operator, find the minimum transmission cost.

[0040] The above discrimination method uses Wasserstein distance to quantify the overall morphological differences of multi-relaxation spectra. It is highly robust to noise and can distinguish fillings with similar composition but different structures (such as mud-containing gravel), significantly improving the discrimination accuracy.

[0041] The cave filling identification step also includes the cave system stability assessment: S401. Calculate key vault structural parameters based on the cave space topology diagram ,in, is the radius of curvature, is the arch-span ratio; S402, through the resistivity-intensity correlation model Estimate the compressive strength of rock mass, where and is the regional calibration coefficient; S403, Predicting the probability of structural instability under critical water pressure using discrete element numerical simulation : ; Where, is the number of simulations, For the The tensile stress of the arch is simulated. is the corresponding compressive strength, is the indicator function, is the index.

[0042] The above assessment method integrates spatial topological vault parameters, resistivity intensity inversion and discrete element water pressure simulation to realize dynamic assessment of instability probability with multi-factor coupling, significantly improving the accuracy of engineering risk warning.

[0043] Example 2: like Figure 2 As shown, this embodiment further improves the design based on the first embodiment. The difference is that in the actual operation of the first embodiment, it was found that there was a time-space reference drift between the towed array and the seabed node. The time-space reference drift was directly caused by the technical gap between S2 (motion posture compensation) and S3 (time-space alignment) of the original method. That is, only the motion error of a single device was compensated, and the cross-device coordination error was not resolved. As a result, the three-dimensional electrical model after the time-space alignment in the S3 step may have cave boundary ghosting, resulting in misjudgment of the cave boundary. Based on this, the method for processing data of transient electromagnetic collaborative survey of cave groups in marine waters also includes: The noise suppression data of the towed array output by S2 and the background noise floor data collected by the seabed node are temporally and spatially correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream as the input of S3.

[0044] After the noise suppression data of the towed array is generated in step S2, it needs to be fused with the background noise floor data collected by the seabed node in a spatiotemporal basis. The specific implementation process is as follows: The towed array generates electromagnetic response data during its motion, while the fixed seafloor nodes continuously record ambient noise. Because the two types of equipment have independent clock sources and dynamic spatial positions, direct fusion will result in blurred cave boundaries in the 3D inversion. Therefore, two-way acoustic timing and Doppler positioning technology are used, including: The towed array periodically transmits acoustic pulse signals, which are received by the seabed nodes and immediately returned with a response signal. By calculating the two-way signal propagation delay, the clock deviation between platforms is accurately determined. 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.

[0045] The acoustic Doppler frequency shift effect generated by the towed array navigation is used, combined with the known fixed coordinates of the seabed node (calibrated by the acoustic positioning network in step S1), to dynamically calculate the spatial offset vector of the array relative to the node.

[0046] Apply the calibrated spatiotemporal benchmark to the original dataset, including: A time-space stamp (including timing calibration time and Doppler positioning coordinates) is added to the noise suppression data of the towed array; a node position stamp is added to the background noise floor data of the seabed node to form a joint observation data stream organized in a three-dimensional grid. Each grid cell in the data stream contains the array corrected electromagnetic response (from S2 output), the node background noise spectrum (from S1 output) and the time-space alignment parameters.

[0047] Example 3: like Figure 3 As shown, based on the same inventive concept as the method for processing data of transient electromagnetic collaborative survey of marine caves in the aforementioned embodiment, this application provides a system for processing data of transient electromagnetic collaborative survey of marine caves. The system and method embodiments in the present application are based on the same inventive concept. The system includes: The wide-spectrum excitation acquisition module transmits a wide-spectrum excitation signal to the seabed through a ship-borne transient electromagnetic transmission system, and simultaneously uses a towed receiving array 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 floor data; A motion noise suppression module performs motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructs an adaptive filter based on the background noise floor data, performs noise suppression processing on the transient electromagnetic response data, and outputs noise suppressed data; The multi-source 3D inversion module aligns the noise suppression data of the towed array with the seafloor node data in time and space to form a 3D spatial sampling data set. It then performs a multi-resolution joint inversion on the 3D spatial sampling data set based on the equivalent eddy diffusion model to reconstruct the 3D electrical structure model of the seafloor medium. The cave topology reconstruction module extracts high-resistance abnormal areas that meet the electrical characteristics of the cave from the three-dimensional electrical structure model; uses a morphological constrained clustering algorithm to identify connected areas, and combines geological structure prior information to construct a cave spatial topology relationship diagram that characterizes the spatial distribution and connectivity of the cave.

[0048] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

[0049] The above is only a preferred specific implementation method of the embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and concept of the present application within the technical scope disclosed in the present application, and they should be covered by the scope of protection of the present application.

Claims

1. A method for processing data of transient electromagnetic collaborative survey of marine cave groups, characterized in that: Methods include: A ship-borne transient electromagnetic transmission system transmits a wide-spectrum excitation signal to the seabed, while a towed receiving array collects the secondary field attenuation signal of the seabed medium response. A distributed sensor network is deployed at pre-set seabed nodes to collect background noise floor data. Performing motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructing an adaptive filter based on the background noise floor data, performing noise suppression processing on the transient electromagnetic response data, and outputting noise suppressed data; The noise suppression data of the towed array is aligned with the seafloor node data in time and space to form a three-dimensional spatial sampling data set. Based on the equivalent eddy diffusion model, a multi-resolution joint inversion is performed on the three-dimensional spatial sampling data set to reconstruct the three-dimensional electrical structure model of the seafloor medium. High-resistance abnormal areas that meet the electrical characteristics of the cave are extracted from the three-dimensional electrical structure model; connected areas are identified using a morphological constrained clustering algorithm, and a cave spatial topological relationship diagram is constructed based on prior geological structure information to characterize the spatial distribution and connectivity of the cave.

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

3. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 1 is 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 background noise base data, and designing a Wiener filter bank with frequency and azimuth selectivity based on the equivalent conductivity and frequency response model.

4. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 1 is characterized in that: The multi-resolution joint inversion method includes: using fast quasi-two-dimensional inversion on towed array noise suppression data to obtain shallow electrical structure, using full three-dimensional finite element inversion on seabed node data to obtain deep electrical structure, and forcing the structural consistency of the two inversion results in the overlapping area through cross-gradient constraints.

5. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 4 is characterized in that: 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 the seabed node data are used as high-fidelity sampling points; Construct a hyperbolic tangent kernel function Deep Gaussian process model of: ； Where, is the three-dimensional space coordinate vector, is the feature scaled covariance matrix, Kernel function length scale parameter, is the signal variance hyperparameter, is the noise variance hyperparameter, is the Kronecker function; The quasi-2D conductivity distribution inverted by the towed array is mapped to the same kernel-induced feature space as the seafloor node data through variational inference; Calculate the gradient field angle cosine similarity in feature space As a measure of structural consistency: ； Where, is the quasi-two-dimensional inversion conductivity gradient field, is the full three-dimensional inversion conductivity gradient field, is the inner product operator on the reproducing kernel Hilbert space, is the reproducing kernel Hilbert space norm operator.

6. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 1 is characterized in that: The morphologically constrained clustering algorithm includes: Calculate the Helmholtz decomposition tensor field based on the three-dimensional electrical structure model; Anisotropic diffusion level set evolution is used to extract closed high-resistance areas; Constructing a multi-scale topological skeleton of a cave system through persistent homology analysis.

7. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 6 is characterized in that: Methods for identifying cave fillings include: Extract the multi-relaxation time spectrum of each cave unit; Calculation of the Wasserstein distance between the measured spectrum and the characteristic spectrum of typical fillings using optimal transport theory : ； Where, and Defined in the field of real numbers The probability measure on , for and The set of all joint probability distributions, for and The transmission plan between The probability measure on , is the joint probability measure The differential form of and Respectively represent the spectrum and The relaxation time variable, | for and of Order cost function, To remove the subfmimum operator.

8. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 7 is characterized in that: The identification method of cave fillings also includes the assessment of cave system stability: Calculation of key vault structural parameters based on cave spatial topological relationship diagram ,in, is the radius of curvature, is the arch-span ratio; Through the resistivity-strength correlation model Estimate the compressive strength of rock mass, where and is the regional calibration coefficient; Predicting the probability of structural instability under critical water pressure using discrete element numerical simulation : ； Where, is the number of simulations, For the The tensile stress of the arch is simulated. is the corresponding compressive strength, is the indicator function, is the index.

9. The method for processing data of transient electromagnetic collaborative survey of marine caves according to claim 1, characterized in that: The method also includes: The noise suppression data of the towed array and the background noise floor data collected by the seabed nodes are temporally and spatially correlated through two-way acoustic timing and Doppler positioning to generate a strictly synchronized joint observation data stream.

10. Transient electromagnetic collaborative survey data processing system for seawater cave groups, characterized in that the system include: A wide-spectrum excitation acquisition module transmits a wide-spectrum excitation signal to the seabed via a shipboard transient electromagnetic transmission system, simultaneously collects the secondary field attenuation signal of the seabed medium response using a towed receiving array, and simultaneously deploys a distributed sensor network at preset seabed nodes to collect background noise floor data; a motion noise suppression module, which performs motion posture compensation on the secondary field attenuation signal to generate transient electromagnetic response data; constructing an adaptive filter based on the background noise floor data, performing noise suppression processing on the transient electromagnetic response data, and outputting noise suppressed data; A multi-source 3D inversion module that performs spatiotemporal alignment of the noise-suppressed data from the towed array with the seafloor node data to form a 3D spatial sampling dataset; performs multi-resolution joint inversion on the 3D spatial sampling dataset based on an equivalent eddy diffusion model to reconstruct a 3D electrical structure model of the seafloor medium; A cave topology reconstruction module extracts high-resistance abnormal areas that meet the electrical characteristics of the cave from the three-dimensional electrical structure model; uses a morphological constrained clustering algorithm to identify connected areas, and combines geological structure prior information to construct a cave spatial topology relationship diagram that characterizes the spatial distribution and connectivity of the cave.

Images