A method and system for three-dimensional localization of geological fault zones based on high-frequency electromagnetic attenuation

Abstract

This invention discloses a method and system for three-dimensional location of geological fault zones based on high-frequency electromagnetic attenuation, belonging to the field of underground structural detection technology. The method acquires the original electromagnetic wave signal of the target exploration area, extracts initial physical property characteristics, and determines the initial boundary contour of the region; it extracts the attenuation pattern vectors of spatial nodes in the boundary neighborhood, obtains abnormal attenuation scores through a random forest discriminant model, and divides the area into normal and abnormal attenuation zones; it writes the normal zone nodes into a basic three-dimensional point set, extracts local interference feature vectors from the abnormal attenuation zone nodes, and clusters them into homogeneous interference sub-regions; it establishes a local compensation model for each homogeneous interference sub-region to obtain compensated abnormal node data; it fuses the basic three-dimensional point set and the compensated abnormal node data to construct a three-dimensional mesh, and performs spatial interpolation repair based on topological residuals to output a three-dimensional boundary model of the fault zone. This improves the reliability and spatial continuity of three-dimensional fault zone boundary location.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration and underground structure detection technology, specifically to a three-dimensional positioning method and system for geological fault zones based on high-frequency electromagnetic attenuation. Background Technology

[0002] In geophysical exploration and subsurface structural detection, the three-dimensional spatial distribution of fault zones directly affects the interpretation of subsurface media and the assessment of engineering risks. Existing technologies have attempted to use electromagnetic wave signals to detect subsurface structures, identifying changes in rock strata structure and anomalous areas through the propagation response of electromagnetic waves in the subsurface media. However, in complex geological environments, the subsurface media are usually not uniformly distributed, especially at rock strata boundaries or fault zone edges. Factors such as rock fracturing, fracture development, and tectonic stress concentration can cause local attenuation anomalies and phase changes in electromagnetic wave signals.

[0003] Existing processing methods typically employ uniform inversion, uniform filtering, or uniform models for the target exploration area. These methods struggle to distinguish between normal signal variations in stable regions and genuine anomalous attenuation at the edges of fault zones. They easily treat anomalous attenuation information at the edges as ordinary noise, weakening or smoothing it out. This leads to shifts in fault zone boundary information or excessive smoothing during subsequent 3D reconstruction.

[0004] Furthermore, since the abnormal attenuation in the edge region of the fault zone is usually local and discontinuous, if there is a lack of identification, diversion and targeted processing of abnormal attenuation nodes, the final three-dimensional boundary result is prone to problems such as local offset, excessive smoothing of the boundary or insufficient spatial continuity.

[0005] Currently, there is no effective solution to address the above problems.

[0006] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a three-dimensional positioning method and system for geological fault zones based on high-frequency electromagnetic attenuation, in order to solve the technical problem in the prior art that when a unified inversion, unified filtering or unified model is used to process the target exploration area, it is difficult to distinguish between normal signal changes in stable areas and abnormal attenuation at the edge of the fault zone, which leads to local offset, excessive smoothing of the boundary or insufficient spatial continuity of the three-dimensional boundary of the fault zone.

[0008] The technical solution of this invention is as follows: On one hand, this invention provides a three-dimensional positioning method for geological fault zones based on high-frequency electromagnetic attenuation. The method includes: acquiring the original electromagnetic wave signal of the target exploration area; extracting initial physical property features from the original electromagnetic wave signal; and determining the initial regional boundary contour based on the initial physical property features; extracting the attenuation pattern vectors of spatial nodes within the neighborhood of the initial regional boundary contour; inputting the attenuation pattern vectors into a pre-trained random forest discriminant model to obtain an abnormal attenuation score for the spatial nodes; and dividing the spatial nodes into a normal zone node set and an abnormal attenuation zone node set based on the comparison result of the abnormal attenuation score and a first preset threshold; writing the normal zone node set into a basic three-dimensional point set; and extracting the attenuation patterns of each spatial node within the abnormal attenuation zone node set. A local interference feature vector is used to perform cluster analysis, dividing the set of nodes in the abnormal attenuation zone into several homogeneous interference sub-regions. A local compensation model is established for each homogeneous interference sub-region, and the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal in the corresponding homogeneous interference sub-region are compensated using the local compensation model. Based on the comparison between the evaluation value of the compensated residual and the second preset threshold, the compensated abnormal node data is obtained. The basic three-dimensional point set and the compensated abnormal node data are fused to construct a three-dimensional mesh, and the topological residual of the edge nodes in the three-dimensional mesh is calculated. Spatial interpolation is performed on the edge nodes whose topological residual is greater than the third preset threshold for repair, and a three-dimensional boundary model of the fracture zone is output based on the repaired three-dimensional mesh.

[0009] Optionally, acquiring the original electromagnetic wave signal of the target exploration area includes: collecting transient electromagnetic signal sequences or ground-penetrating radar echo signals of the target exploration area through a surface electromagnetic transmission and receiving array and / or a downhole electromagnetic transmission and receiving array, as the original electromagnetic wave signal.

[0010] Optionally, the step of extracting initial physical property features from the original electromagnetic wave signal includes: performing synchronization and time correction, filtering, abnormal sampling removal, gain compensation, and time-to-depth conversion on the original electromagnetic wave signal; and extracting at least one of amplitude attenuation correlation, phase delay correlation, apparent resistivity proxy, and frequency band energy ratio from the processed original electromagnetic wave signal as the initial physical property features.

[0011] Optionally, extracting the attenuation mode vector of spatial nodes in the neighborhood of the initial region boundary contour includes: extracting the attenuation trend of the spatial nodes in multiple frequency bands, the energy difference with adjacent measurement points, the phase change and spatial gradient information within a preset spatial window, and constructing the attenuation mode vector.

[0012] Optionally, the first preset threshold is determined by: determining a benchmark threshold based on the statistical distribution of stable area samples and abnormal area samples in the historical exploration data of the target exploration area; correcting the benchmark threshold according to the noise basis calibration result of the current batch of original electromagnetic wave signals to obtain the first preset threshold; the step of writing the normal area node set into the basic three-dimensional point set includes: retaining the initial physical property characteristics corresponding to the spatial nodes in the normal area node set, or writing the spatial nodes in the normal area node set into the basic three-dimensional point set after baseline smoothing.

[0013] Optionally, the local interference feature vector includes at least one of high-frequency energy dissipation rate, phase fluctuation amount, and continuity deviation of adjacent nodes; when the acquisition system acquires electromagnetic responses in at least two different polarization directions, the local interference feature vector also includes polarization difference amount; the clustering analysis based on the local interference feature vector includes: standardizing the local interference feature vector, and using a density-based clustering algorithm to cluster the standardized local interference feature vector, dividing spatial nodes with similar physical interference characteristics and continuous spatial distribution into the same homogeneous interference sub-region.

[0014] Optionally, the step of establishing a local compensation model for each homogeneous interference sub-region includes: extracting signal propagation interference samples from each homogeneous interference sub-region; and constructing a support vector machine regression compensation model corresponding to each homogeneous interference sub-region based on the signal propagation interference samples and the reference sample.

[0015] The reference sample is derived from at least one of borehole control points, trench control points, stable zone templates, and small-scale construction samples.

[0016] Optionally, obtaining the compensated abnormal node data based on the comparison result between the compensated residual evaluation value and the second preset threshold includes: calculating the root mean square error or mean absolute error of the support vector machine regression compensation model on the validation sample set as the compensated residual evaluation value; when the compensated residual evaluation value is less than or equal to the second preset threshold, outputting the compensated abnormal node data using the current support vector machine regression compensation model; when the compensated residual evaluation value is greater than the second preset threshold, iteratively adjusting at least one of the penalty parameter, kernel scale parameter, and sample weight of the support vector machine regression compensation model; when the compensated residual evaluation value recalculated based on the adjusted support vector machine regression compensation model is less than or equal to the second preset threshold, outputting the compensated abnormal node data using the adjusted support vector machine regression compensation model; when the number of iterative adjustments reaches the preset maximum number of iterations, and the compensated residual evaluation value is still greater than the second preset threshold, using low-order local smoothing compensation as the downgrade compensation result for the corresponding homogeneous interference sub-region, and setting a low-confidence label for the homogeneous interference sub-region.

[0017] Optionally, calculating the topological residual of the edge nodes in the 3D mesh includes: determining the deviation of the edge node relative to the neighborhood continuity constraint surface and / or control points, and using the deviation as the topological residual; wherein, the third preset threshold is determined based on the control point deviation, neighborhood continuity deviation, and engineering tolerance error; the spatial interpolation repair of edge nodes whose topological residual is greater than the third preset threshold includes: obtaining the main strike and dip information of the fault zone, and using anisotropic kriging interpolation or spline interpolation with directional constraints to spatially interpolate and repair the edge node based on the low residual nodes around the edge node, so that the main interpolation direction is consistent with the main strike and dip of the fault zone; if it is detected that the proportion of edge nodes whose topological residual is greater than the third preset threshold in the 3D mesh after spatial interpolation repair exceeds the preset allowable proportion, an anomaly feedback mechanism is triggered, and the feedback is rolled back to the clustering analysis step for regrouping, or the feedback is rolled back to the local compensation model establishment step for retraining and compensation.

[0018] On the other hand, the present invention also provides a three-dimensional positioning system for geological fault zones based on high-frequency electromagnetic attenuation. The system includes: an electromagnetic signal acquisition module for acquiring the original electromagnetic wave signal of the target exploration area; an initial boundary determination module for extracting initial physical property features from the original electromagnetic wave signal and determining the initial regional boundary contour based on the initial physical property features; an abnormal attenuation identification module for extracting the attenuation pattern vectors of spatial nodes in the neighborhood of the initial regional boundary contour, inputting the attenuation pattern vectors into a pre-trained random forest discriminant model to obtain the abnormal attenuation score of the spatial nodes, and dividing the spatial nodes into a normal zone node set and an abnormal attenuation zone node set according to the comparison result of the abnormal attenuation score and a first preset threshold; and a node splitting and homogeneous grouping module for writing the normal zone node set into a basic three-dimensional point set and extracting each spatial node in the abnormal attenuation zone node set. The system employs a clustering analysis method based on the local interference feature vectors of the nodes to divide the set of nodes in the abnormal attenuation zone into several homogeneous interference sub-regions. A local compensation module is used to establish a local compensation model for each homogeneous interference sub-region, and to compensate the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal within the corresponding homogeneous interference sub-region using the local compensation model. Based on the comparison between the compensated residual evaluation value and a second preset threshold, the compensated abnormal node data is obtained. A three-dimensional mesh construction module is used to fuse the basic three-dimensional point set with the compensated abnormal node data to construct a three-dimensional mesh, and to calculate the topological residuals of the edge nodes in the three-dimensional mesh. A spatial repair output module is used to perform spatial interpolation repair on edge nodes whose topological residuals are greater than a third preset threshold, and to output a three-dimensional boundary model of the fracture zone based on the repaired three-dimensional mesh.

[0019] The beneficial effects of this invention are as follows: By extracting initial physical property features from the original electromagnetic wave signal and determining the initial region boundary contour, and then identifying abnormal attenuation regions based on attenuation mode vectors, this invention allows nodes in the normal region and nodes in the abnormal attenuation region to enter different processing paths, avoiding indiscriminate and complex compensation processing of stable regions, thereby reducing the risk of over-correction of normal regions. Simultaneously, this invention divides the nodes in the abnormal attenuation region into several homogeneous interference sub-regions according to local interference feature vectors, and establishes a local compensation model for each homogeneous interference sub-region, enabling different types of edge abnormal attenuation to receive corresponding compensation, reducing compensation deviations caused by a single model covering all abnormal regions. Furthermore, after compensation, this invention fuses the basic 3D point set with the compensated abnormal node data to construct a 3D mesh, and performs spatial interpolation repair on the edge nodes based on topological residuals, ensuring that the final output 3D boundary model of the fault zone takes into account both electromagnetic signal compensation results and 3D spatial continuity, improving the reliability of fault zone boundary positioning and model continuity. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation provided by this invention.

[0021] Figure 2 A detailed flowchart illustrating a three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation provided by this invention.

[0022] Figure 3 This is a schematic diagram of the structure of a three-dimensional positioning system for geological fault zones based on high-frequency electromagnetic attenuation provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0024] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0025] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0026] As mentioned earlier, existing processing methods typically employ a unified inversion, unified filtering, or unified model for the target exploration area. This approach struggles to distinguish between normal signal variations in stable regions and genuine anomalous attenuation at the fault edges. It easily treats anomalous attenuation information as ordinary noise, weakening or smoothing it out. This leads to shifts or over-smoothing of the fault boundary information during subsequent 3D reconstruction. Furthermore, since anomalous attenuation at the fault edge is usually localized and discontinuous, without the identification, separation, and targeted processing of anomalous attenuation nodes, the final 3D boundary results are prone to local shifts, over-smoothing of boundaries, or insufficient spatial continuity.

[0027] To address these issues, the present invention provides a method and system for three-dimensional localization of geological fault zones based on high-frequency electromagnetic attenuation, which solves the aforementioned problems in the following manner. This will be further explained below with reference to the accompanying drawings.

[0028] Example 1:

[0029] like Figure 1 As shown in Embodiment 1 of the present invention, a three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation can be implemented by a processor, controller, geophysical data processing workstation, or a data processing system consisting of electromagnetic acquisition equipment, memory, and processor. The method mainly includes the following steps:

[0030] S100. Obtain the original electromagnetic wave signal of the target exploration area, extract the initial physical property features from the original electromagnetic wave signal, and determine the initial area boundary contour based on the initial physical property features.

[0031] S200. Extract the decay pattern vector of the spatial nodes in the neighborhood of the initial region boundary contour, input the decay pattern vector into the pre-trained random forest discriminant model to obtain the abnormal decay score of the spatial nodes, and divide the spatial nodes into a normal region node set and an abnormal decay region node set according to the comparison result of the abnormal decay score and the first preset threshold.

[0032] S300. Write the set of normal region nodes into the basic three-dimensional point set, and extract the local interference feature vectors of each spatial node in the set of abnormal attenuation region nodes. Perform cluster analysis based on the local interference feature vectors to divide the set of abnormal attenuation region nodes into several homogeneous interference sub-regions.

[0033] S400. Establish a local compensation model for each homogeneous interference sub-region, and use the local compensation model to compensate the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal in the corresponding homogeneous interference sub-region. Based on the comparison result of the compensation residual evaluation value and the second preset threshold, obtain the compensated abnormal node data.

[0034] S500. The basic three-dimensional point set is fused with the compensated abnormal node data to construct a three-dimensional mesh, and the topological residual of the edge nodes in the three-dimensional mesh is calculated.

[0035] S600. Spatial interpolation repair is performed on the edge nodes whose topological residuals are greater than the third preset threshold, and a three-dimensional boundary model of the fracture zone is output based on the repaired three-dimensional mesh.

[0036] Based on the above steps, Embodiment 1 of the present invention first determines the initial regional boundary contour in the target exploration area through initial physical property characteristics, and then identifies abnormal attenuation of spatial nodes in the neighborhood of the initial regional boundary contour, so that nodes in the normal zone and nodes in the abnormal attenuation zone enter different processing paths. Simultaneously, the present invention does not directly output the fault zone boundary after obtaining the compensated abnormal node data, but continues to fuse the basic 3D point set with the compensated abnormal node data to construct a 3D mesh, and uses topological residuals to determine whether there is spatial discontinuity at the edge nodes. For edge nodes with excessive topological residuals, the present invention further performs spatial interpolation repair, so that the 3D boundary model of the fault zone is simultaneously constrained by the electromagnetic signal compensation result and the 3D spatial continuity condition.

[0037] Example 2:

[0038] Based on the above embodiments, in order to provide a clearer and more complete explanation of the technical solutions therein, the present invention also provides an embodiment two. For example... Figure 2 As shown, in Embodiment 2 of the present invention, the method can also be executed by a processor or a controller.

[0039] In the second embodiment of the present invention, step S100 may include the following steps:

[0040] S110. Acquire transient electromagnetic signal sequences or ground-penetrating radar echo signals of the target exploration area through a surface electromagnetic transmission and receiving array and / or a downhole electromagnetic transmission and receiving array, as the original electromagnetic wave signal.

[0041] S120. Perform synchronization and time-to-depth conversion on the original electromagnetic wave signal, filter processing, abnormal sampling rejection, gain compensation, and time-to-depth conversion.

[0042] S130. Extract at least one of amplitude attenuation correlation, phase delay correlation, apparent resistivity proxy and bandwidth energy ratio from the processed original electromagnetic wave signal as the initial physical property feature.

[0043] S140. Based on the initial physical properties, the target exploration area is initially spatially divided to determine the initial area boundary contour.

[0044] For example, steps S110 to S140 may further include the following steps. In this second embodiment, firstly, the high-frequency electromagnetic attenuation refers to the frequency band attenuation characteristics extracted from transient electromagnetic signal sequences or ground-penetrating radar echo signals, which are more susceptible to the influence of fractured media at the edge of the fracture zone compared to the low-frequency background. This frequency band attenuation characteristic is used for subsequent initial physical property feature extraction, attenuation mode vector construction, and local interference feature extraction.

[0045] Secondly, the data acquisition unit receives the raw electromagnetic wave signals from each channel and adds the acquisition time, spatial coordinates of the measurement point, and channel identifier to each set of signals. The acquisition time comes from the unified clock of the acquisition equipment; the spatial coordinates of the measurement point come from the measurement results of the surface measurement point, the positioning results of the downhole measurement point, or the coordinate registration results; the channel identifier comes from the hardware number of the transmitting and receiving array.

[0046] Then, the processor synchronizes the original electromagnetic wave signal to align the signals from different channels to the same time reference; it filters the synchronized signal to reduce interference components unrelated to the target detection frequency band; it removes abnormal samples that are obviously missing, jump, or exceed the effective range of the acquisition equipment; it performs gain compensation on signals that weaken with propagation time or propagation distance; and it converts the time domain signal into data corresponding to the spatial node according to the time-to-depth conversion rules determined on site.

[0047] The time-to-depth conversion rule can be determined based on the velocity model of the target exploration area, known medium parameters, borehole control points, trench control points, or field calibration results. In this second embodiment, it will be used as a routine coordinate conversion step in the preprocessing of electromagnetic detection data. The converted node depths and spatial coordinates are used for the subsequent formation of the initial area boundary profile.

[0048] Subsequently, the processor extracts initial physical property features from the processed original electromagnetic wave signal. To ensure that these initial physical property features correspond to the medium differences at the fracture zone edge, this embodiment two organizes the amplitude attenuation correlation, phase delay correlation, apparent resistivity surrogate, and bandgap energy ratio of the same spatial node into a node-level initial physical property feature vector. This node-level initial physical property feature vector serves as the common data basis for determining the initial region boundary profile and constructing subsequent attenuation mode vectors. Its expression can be:

[0049]

[0050] in, Indicates the first The initial physical property feature vectors of each spatial node; Indicates the first The amplitude attenuation correlation quantity corresponding to each spatial node; Indicates the first Phase delay correlation quantity corresponding to each spatial node; Indicates the first Apparent resistivity proxy quantity for each spatial node; Indicates the first The frequency band energy ratio corresponding to each spatial node. All of the above characteristic components originate from the original electromagnetic wave signal or its preprocessing result corresponding to the same spatial node, without introducing new data sources that deviate from the original electromagnetic wave signal.

[0051] Specifically, the amplitude attenuation correlation is obtained by extracting the envelope amplitude through Hilbert transform of the preprocessed time-domain signal and calculating the logarithmic attenuation rate of the envelope amplitude with depth; the phase delay correlation is obtained by cross-correlation calculation of the signals of the current spatial node and adjacent stable region nodes; the apparent resistivity surrogate is obtained by power-law fitting of the late attenuation curve of the signal; and the frequency band energy ratio is obtained by calculating the ratio of the preset high-frequency band energy integral to the full-frequency band energy integral after performing a fast Fourier transform on the signal.

[0052] Finally, the processor performs preliminary spatial division of the target exploration area based on the node-level initial physical property feature matrix. Specifically, the processor first performs dimensionless processing on each feature component in the node-level initial physical property feature vector according to the mean and standard deviation of the stable zone samples, so that the amplitude attenuation correlation, phase delay correlation, apparent resistivity surrogate, and band energy ratio are included in the same comparison caliber; then, based on the dimensionless initial physical property feature vector, the processor calculates the feature difference value and spatial gradient between adjacent spatial nodes. The feature difference value is used to characterize the comprehensive difference of adjacent spatial nodes in initial physical property features, and the spatial gradient is used to characterize the magnitude of change of initial physical property features along the spatial direction of the measuring point. The boundary judgment threshold (e.g., 1.5) is determined by the statistical distribution of feature differences of the stable zone samples as a baseline value, and is corrected by combining the measuring point spacing, target exploration depth, and field calibration results; the gradient threshold (e.g., 0.8) corresponding to the boundary judgment threshold is determined by the spatial gradient statistical results of the stable zone samples and the spacing between adjacent measuring points.

[0053] When the feature difference value of adjacent spatial nodes is greater than the boundary determination threshold, the processor determines the corresponding spatial node as a candidate boundary node; when neither of them exceeds the corresponding threshold, the processor does not determine the corresponding spatial node as a candidate boundary node. Subsequently, the processor connects the spatially continuous candidate boundary nodes according to the spatial adjacency relationship and survey line connection relationship between the candidate boundary nodes, and uses the connected boundary node sequence as the initial region boundary contour.

[0054] In the second embodiment of the present invention, step S200 may include the following steps:

[0055] S210. Based on the initial region boundary contour, determine the boundary neighborhood spatial nodes that need to be identified for abnormal attenuation.

[0056] S220. Extract the attenuation trend of the spatial node in multiple frequency bands, the energy difference between it and adjacent measurement points, the phase change and spatial gradient information within the preset spatial window, and construct the attenuation mode vector.

[0057] S230. Input the decay pattern vector into a pre-trained random forest discriminant model to obtain the abnormal decay score of the spatial node.

[0058] S240. Based on the statistical distribution of stable area samples and abnormal area samples in the historical exploration data of the target exploration area, a benchmark threshold is determined, and the benchmark threshold is corrected according to the noise floor calibration results of the current batch of original electromagnetic wave signals to obtain the first preset threshold.

[0059] S250. Based on the comparison result between the abnormal attenuation score and the first preset threshold, the spatial nodes are divided into a set of normal zone nodes and a set of abnormal attenuation zone nodes.

[0060] For example, steps S210 to S250 may further include the following steps. First, the processor selects boundary neighborhood spatial nodes centered on the initial region boundary contour obtained in step S100, according to the topological relationship of the measurement points or the size of the coarse grid cells. A preset spatial window is used to limit the neighborhood range of each spatial node to be determined. This window can be jointly determined by the size of the coarse grid cells, the spacing between measurement points, the estimated width of the target fault zone, and the field calibration results.

[0061] Then, the processor extracts the attenuation mode vector within a preset spatial window. The attenuation mode vector reflects at least four types of information: the first type is the attenuation trend across multiple frequency bands, reflecting the energy change of the electromagnetic wave along the propagation path at different frequency bands; the second type is the energy difference between adjacent measurement points, specifically represented by calculating the relative difference in the high-frequency envelope area of ​​the signal between the node and adjacent measurement points within the preset spatial window; the third type is the phase change, specifically represented by calculating the phase angle difference corresponding to the peak time offset of the cross-correlation between the node and the reference node; the fourth type is spatial gradient information, specifically represented by taking the partial derivatives of the initial physical property characteristics in three orthogonal directions in space and calculating the magnitude of its gradient vector.

[0062]

[0063] in, Indicates the first The decay mode vector of each spatial node; This indicates the multi-frequency attenuation trend within the preset spatial window; Indicates the first The energy difference between each spatial node and its adjacent measuring points; Indicates the first The amount of phase change of a spatial node relative to its neighboring nodes; It represents the spatial gradient information of the initial physical property characteristics in the neighborhood.

[0064] The processor then inputs the decay pattern vector into a pre-trained random forest discriminant model. The training samples for this random forest discriminant model can come from historical exploration data of the target exploration area, stable zone samples, anomaly zone samples, samples corresponding to borehole control points, or boundary anomaly samples identified on-site. During training, the decay pattern vector serves as the model input, and the stable zone label or anomaly zone label serves as the model output target. After training, the model is used to output anomaly decay scores for the spatial nodes of the current batch.

[0065] The abnormal decay score is used to represent the degree to which a spatial node exhibits abnormal decay characteristics. In this second embodiment, the random forest discriminant model outputs the voting ratio of each decision tree for the abnormal decay category, and normalizes the voting ratio into an abnormal decay score, so that the abnormal decay score can be compared with a first preset threshold (e.g., 0.75). In the same batch of data processing, the abnormal decay score, the first preset threshold, and the corresponding judgment rule adopt the same scoring caliber; the abnormal decay score is uniformly defined as the voting ratio of the number of decision trees in the random forest model that are judged as the abnormal decay category to the total number of decision trees.

[0066] Then, the processor determines a first preset threshold. This first preset threshold is used to determine whether a spatial node has entered an abnormal attenuation zone. The processor first extracts the upper limit of the abnormal attenuation score for stable zone samples and the lower limit of the abnormal attenuation score for abnormal zone samples based on historical detection data of the target exploration area, and determines the arithmetic mean of the two as the benchmark value of the first preset threshold. Subsequently, it calculates the ratio of the noise floor amplitude of the current batch's original electromagnetic wave signal to the historical noise floor amplitude as the deviation rate. The benchmark value is multiplied by this deviation rate for dynamic amplification or reduction correction, thereby obtaining the first preset threshold used for the current batch. The noise floor comes from at least one of the following: pre-acquisition calibration section, statistical results of the stable background section, or the operation record of the acquisition equipment.

[0067] Finally, the processor compares the abnormal attenuation score of each spatial node with a first preset threshold. When the abnormal attenuation score is less than or equal to the first preset threshold, the processor divides the corresponding spatial node into a normal zone node set; when the abnormal attenuation score is greater than the first preset threshold, the processor divides the corresponding spatial node into an abnormal attenuation zone node set. The normal zone node set proceeds to the bypass write processing in step S300, while the abnormal attenuation zone node set proceeds to the local interference feature extraction and homogenization grouping processing in step S300.

[0068] In Embodiment 2 of the present invention, step S300 may include the following steps:

[0069] S310. Retain the initial physical property characteristics corresponding to the spatial nodes in the normal region node set, or perform baseline smoothing on the spatial nodes in the normal region node set and then write them into the basic three-dimensional point set.

[0070] S320. Extract the local interference feature vector of each spatial node in the abnormal attenuation region node set;

[0071] S330. Standardize the local interference feature vector;

[0072] S340. A density-based clustering algorithm is used to cluster the standardized local interference feature vectors, and spatial nodes with similar physical interference characteristics and continuous spatial distribution are divided into the same homogeneous interference sub-region.

[0073] S350, Output an abnormal node sample pool with sub-region labels.

[0074] For example, steps S310 to S350 may further include the following steps. First, for the set of normal region nodes determined in step S200, the processor treats it as part of the basic three-dimensional point set. For spatial nodes in the set of normal region nodes, the processor can directly retain their initial physical property characteristics and write the spatial coordinates, initial physical property characteristics, and mass markers into the basic three-dimensional point set.

[0075] When nodes in the normal zone exhibit small random fluctuations not exceeding a first preset threshold, the processor can perform baseline smoothing based on the trend of changes in the physical properties of adjacent normal zone nodes. Then, for the set of nodes in the abnormal attenuation zone, the processor further extracts local interference feature vectors. These local interference feature vectors include at least one of high-frequency energy dissipation rate, phase fluctuation, and continuity deviation between adjacent nodes. When the acquisition system obtains electromagnetic responses with at least two different polarization directions at the same measurement point or a registerable measurement point, the local interference feature vectors also include polarization difference. When the acquisition system does not obtain electromagnetic responses with different polarization directions, the processor does not include polarization difference as a component of the local interference feature vectors for this batch.

[0076] The components in the aforementioned local interference feature vector are obtained as follows: the high-frequency energy dissipation rate is obtained by calculating the exponential decay constant of the signal energy in the target high-frequency band with the propagation depth; the phase fluctuation is obtained by calculating the local statistical variance of the phase delay correlation between adjacent nodes within a preset spatial window; the continuity deviation between adjacent nodes is obtained by calculating the mean of the Euclidean distance between the node and the initial physical property feature vector of the first-order adjacent node; and the polarization difference is obtained by extracting the normalized difference of the response amplitudes of two orthogonal polarization directions at the same measurement point.

[0077] Then, the processor standardizes the local disturbance feature vectors, enabling features with different dimensions and value ranges to enter the same clustering process. For any local disturbance feature component, the standardization result can be expressed as:

[0078]

[0079] in, Indicates the first The abnormal node is at the _ ... The original values ​​on each local disturbance feature; This indicates the node in the current batch's abnormal decay region node set. The mean of a local disturbance feature; This indicates the node in the current batch's abnormal decay region node set. The standard deviation of each local disturbance feature; This represents the standardized feature value. When the standard deviation of a local disturbance feature in the current batch is zero or lower than the device's quantization resolution, the processor will not use this feature as the clustering input for this batch, or it will set the standardized result corresponding to the feature to zero and retain the quality label.

[0080] Subsequently, the processor employs a density-based clustering algorithm to cluster the standardized local disturbance feature vectors. This clustering process considers both the similarity of local disturbance features and the continuity of spatial distribution.

[0081] If the number of samples in a homogeneous interference sub-region is less than the preset minimum number of trainable samples, or if the intra-class dispersion of the local interference feature vector in the homogeneous interference sub-region is greater than the preset dispersion threshold (e.g., 2.0), the processor expands the clustering neighborhood or merges the homogeneous interference sub-region with adjacent homogeneous interference sub-regions that have similar features and are spatially continuous. If the merging still cannot meet the modeling sample requirements, the processor inputs the homogeneous interference sub-region into the degradation compensation channel and sets a low confidence label.

[0082] Finally, the processor generates sub-region labels for each homogeneous interference sub-region and outputs an anomalous node sample pool with sub-region labels. The anomalous node sample pool includes spatial node coordinates, local interference feature vectors, corresponding original electromagnetic wave signal segments, sub-region labels, and quality markers.

[0083] In Embodiment 2 of the present invention, step S400 may include the following steps:

[0084] S410. Extract signal propagation interference samples from each of the homogeneous interference sub-regions;

[0085] S420. Based on the signal propagation interference samples and reference samples, construct support vector machine regression compensation models corresponding to each homogeneous interference sub-region.

[0086] S430. Calculate the root mean square error or mean absolute error of the support vector machine regression compensation model on the validation sample set, and use it as the evaluation value of the compensation residual.

[0087] S440. When the compensated residual evaluation value is less than or equal to the second preset threshold, the compensated abnormal node data is output using the current support vector machine regression compensation model.

[0088] S450. When the post-compensation residual evaluation value is greater than the second preset threshold, at least one of the penalty parameter, kernel scale parameter and sample weight of the support vector machine regression compensation model is iteratively adjusted, and the post-compensation residual evaluation value is recalculated.

[0089] S460. When the compensated residual evaluation value recalculated based on the adjusted support vector machine regression compensation model is less than or equal to the second preset threshold, the compensated abnormal node data is output using the adjusted support vector machine regression compensation model.

[0090] S470. When the number of iterations reaches the preset maximum number of iterations, and the residual evaluation value after compensation is still greater than the second preset threshold, low-order local smoothing compensation is used as the downgrade compensation result for the corresponding homogeneous interference sub-region, and a low confidence flag is set for the homogeneous interference sub-region.

[0091] For example, steps S410 to S470 may further include the following steps. First, the processor organizes the modeling data into units of homogeneous interference sub-regions. For any homogeneous interference sub-region, the processor extracts signal propagation interference samples from that sub-region. The signal propagation interference samples at least include the local interference feature vectors of spatial nodes within the homogeneous interference sub-region, the corresponding original electromagnetic wave signal segments, the preprocessed signal parameters, spatial coordinates, and sub-region labels.

[0092] Then, the processor acquires reference samples. These reference samples come from at least one of borehole control points, trench control points, stable zone templates, and pilot-scale structural samples. Borehole control points and trench control points provide known geological boundary locations or known medium states; stable zone templates provide signal references that do not exhibit abnormal attenuation; and pilot-scale structural samples provide pre-calibrated local signal compensation references. For the same homogeneous interference sub-region, the processor determines a source of reference sample target values ​​before model training and writes this source of target values, along with the compensation object, into the model training record for that sub-region. Subsequently, the processor establishes a support vector machine regression compensation model for each homogeneous interference sub-region. The input to this model is the local interference feature vector of the corresponding homogeneous interference sub-region and its associated signal parameters; the output of the model is the compensated node-level signal parameters or node-level physical property characteristics. These node-level signal parameters or node-level physical property characteristics all originate from the original electromagnetic wave signal within the corresponding homogeneous interference sub-region. Therefore, the compensation processing of the original electromagnetic wave signal by the local compensation model specifically manifests as compensating and correcting the node-level signal parameters or node-level physical property characteristics obtained from the original electromagnetic wave signal.

[0093] The penalty parameters, kernel scale parameters, and sample weights of the support vector machine regression compensation model can be determined based on validation set search, cross-validation, field calibration results, sample quality labeling, control point reliability, or reference sample source level.

[0094] Then, the processor uses a validation sample set to validate the support vector machine regression compensation model. The post-compensation residual evaluation value is used to evaluate the deviation between the model output and the reference sample target value. Before model training, the processor uniformly sets the compensation object to the correction amount of the node-level initial physical property characteristics. The reference sample target value is preferably extracted from the non-anomaly calibrated physical property characteristics at the depth of the borehole control point of the known geological boundary in the target exploration area; if there is no borehole control point, adjacent stable zone nodes at the same depth are extracted, and baseline smoothed and used as the reference sample target value. The output of the model is the compensated physical property characteristic correction value, which is used to superimpose with the original physical property characteristics to obtain the compensated anomaly node data. In Example 2, when the root mean square error is used as the post-compensation residual evaluation value, the post-compensation residual evaluation value is expressed as:

[0095]

[0096] in, This represents the evaluation value of the compensated residual obtained using the root mean square error caliber; Indicates the first The node-level signal parameters or node-level physical properties of a verification sample after compensation by the local compensation model; Indicates the first The target value of the reference sample corresponding to each validation sample; This indicates the number of validation samples. and Use the same physical quantity specifications and the same dimensions.

[0097] In other embodiments, when the mean absolute error is used as the evaluation value of the compensated residual, the evaluation value of the compensated residual is expressed as:

[0098]

[0099] in, This represents the compensated residual evaluation value obtained using the mean absolute error caliber. The second preset threshold (e.g., 0.05) is determined based on the historical statistical distribution of the compensated residual evaluation value on the validation sample set and the engineering tolerance error.

[0100] In a single model validation process, the same homogeneous interference sub-region uses either root mean square error or mean absolute error as the evaluation caliber, and the second preset threshold is kept consistent with the selected evaluation caliber.

[0101] The second preset threshold is used to determine whether the local compensation model passes validation. The second preset threshold is determined based on the statistical distribution of the compensated residual evaluation value on the validation sample set, pilot test results, calibration results during the maintenance phase, and engineering tolerance error. When the compensated residual evaluation value is less than or equal to the second preset threshold, the processor determines that the local compensation model corresponding to the homogeneous interference sub-region meets the output conditions and uses this model to output the compensated abnormal node data.

[0102] When the compensated residual evaluation value exceeds the second preset threshold, the processor iteratively adjusts at least one of the penalty parameter, kernel scale parameter, and sample weights of the support vector machine regression compensation model. After adjustment, the processor retrains or updates the support vector machine regression compensation model and recalculates the compensated residual evaluation value. This process is repeated until the compensated residual evaluation value meets the second preset threshold, or the number of iterations reaches the maximum number of iterations.

[0103] When the number of iterations reaches the maximum number of iterations, and the residual evaluation value after compensation is still greater than the second preset threshold, the processor uses low-order local smoothing compensation as the downgrade compensation result for the corresponding homogeneous interference sub-region. Low-order local smoothing compensation can be determined based on the compensation change trend of adjacent compensated nodes or adjacent low-residual nodes within the corresponding homogeneous interference sub-region. The processor uses the compensation amount or post-compensation physical property characteristics of adjacent nodes as a reference to perform local weighted smoothing on the abnormal nodes in the current homogeneous interference sub-region that have not passed model validation, thus obtaining the downgrade compensation result. When performing low-order local smoothing compensation, the support vector machine regression compensation model training is not re-executed. At this time, the processor sets a low-confidence flag for the homogeneous interference sub-region and outputs the low-confidence flag along with the compensated abnormal node data to step S500.

[0104] In the second embodiment of the present invention, step S500 may include the following steps:

[0105] S510. Convert the basic three-dimensional point set and the compensated abnormal node data to a unified coordinate system;

[0106] S520. The converted basic 3D point set is fused with the compensated abnormal node data to form a 3D boundary node set.

[0107] S530. Construct a three-dimensional mesh based on the set of three-dimensional boundary nodes;

[0108] S540. Determine the deviation of the edge nodes in the three-dimensional mesh relative to the neighborhood continuity constraint surface and / or control points, and use the deviation as the topological residual;

[0109] S550. Based on the comparison result between the topological residual and the third preset threshold, determine the set of nodes with excessive residuals.

[0110] For example, steps S510 to S550 may further include the following steps. First, the processor receives the basic three-dimensional point set output in step S300 and the compensated abnormal node data output in step S400. The basic three-dimensional point set mainly corresponds to the normal region node set, and its node data includes spatial coordinates, initial physical property characteristics, and quality labels; the compensated abnormal node data mainly corresponds to the abnormal attenuation region node set, and its node data includes spatial coordinates, compensated signal parameters or physical property characteristics, sub-region labels, and compensated quality labels.

[0111] Then, the processor transforms the basic 3D point set and the compensated anomaly node data to a unified coordinate system based on the measurement point coordinates, time-to-depth conversion results, control point data, and coordinate registration parameters. The specific process of coordinate transformation is as follows: using the surface or a known benchmark control point as the origin, the relative coordinates of the measurement points are multiplied by a rotation and translation matrix calibrated with the equipment; then, the depth data is mapped to Z-axis coordinates, thereby generating a 3D boundary point cloud. Coordinate registration parameters can be derived from field measurement control points, combined well-ground measurement data, existing geological interpretation results, or system initialization calibration results.

[0112] Subsequently, the processor fuses the base 3D point set in a unified coordinate system with the compensated anomalous node data to form a 3D boundary node set. This 3D boundary node set includes both base nodes in stable regions and anomalous nodes after local compensation. For nodes in anomalous attenuation regions, the processor redetermines the corresponding candidate boundary positions based on the spatial gradient changes of the compensated node-level signal parameters or node-level physical properties, and uses these candidate boundary positions as the anomalous region boundary node positions in the 3D boundary node set. This process of redetermining the candidate boundary positions does not change the original acquired signal itself, but only changes the boundary representation position of the anomalous node in the 3D boundary node set. For anomalous nodes with low-confidence markers, the processor retains their quality markers during fusion to identify the reliability of the region during subsequent topology residual detection and result display.

[0113] Then, the processor constructs a 3D mesh based on the set of 3D boundary nodes. The 3D mesh can consist of 3D boundary nodes, mesh edges, and mesh patches, used to represent the initial morphology of the fault zone boundary in 3D space. When constructing the 3D mesh, the processor can determine the mesh connectivity based on node spatial adjacency, local surface continuity constraints, and control point constraints. Among them, node spatial adjacency is used to determine whether adjacent nodes can form mesh edges, local surface continuity constraints are used to limit the degree of abrupt changes between adjacent patches, and control point constraints are used to ensure that the 3D mesh corresponds to borehole control points, trench control points, or other known geological control data.

[0114] Next, the processor calculates the topological residuals of the edge nodes in the 3D mesh. These topological residuals are used to evaluate the deviation of the edge nodes relative to the neighborhood continuity constraint surface and / or control points. The neighborhood continuity constraint surface can be obtained by fitting low-residue nodes around the edge node, and the control points can be derived from borehole control points, trench control points, or other known geological control data. If effective control points are lacking in Example 2, the topological residuals can be determined primarily based on the neighborhood continuity constraint surface. To clarify the objective calculation object of the topological residuals, Example 2 can be expressed as follows:

[0115]

[0116] in, Indicates the first Topological residuals of each edge node; Indicates the first The spatial location of each edge node in the 3D mesh; Indicates the first The projection position of each edge node on the neighborhood continuity constraint surface; Indicates the first The control point constraint position corresponding to each edge node, or the control constraint position obtained by fitting adjacent control points; Represents the neighborhood continuity constraint weight; Indicates the control point constraint weights; This represents the spatial Euclidean distance. Specifically, for edge nodes exceeding the effective interpolation radius of known geological control points, the constraint weights of the control points are forcibly assigned to 0, and the topological residual is determined solely based on the neighborhood continuity constraint surface. .

[0117] When both neighborhood continuity constraint surfaces and control point constraint locations exist simultaneously. + =1, and the topological residual is determined based on the neighborhood continuity deviation and the control point deviation; when effective control points are lacking... =1, =0, and the topological residual is mainly determined based on the neighborhood continuity constraint surface; when there is a lack of effective neighborhood continuity constraint surface but there are control point constraint positions. =0, =1, and determine the topological residual based on the control point constraint position. The topological residual is a distance scalar corresponding to the spatial coordinates, and its unit is consistent with the spatial coordinate unit of the edge node.

[0118] The third preset threshold (e.g., 0.5 meters) is used to determine whether edge nodes require spatial repair, and is determined based on control point deviation, neighborhood continuity deviation, and engineering tolerance error. Control point deviation is derived from the statistical deviation between control points and corresponding grid nodes; neighborhood continuity deviation is derived from the local continuity constraints formed by surrounding low residual nodes; and engineering tolerance error is derived from the error requirements of subsequent engineering applications in the target exploration area or the results of on-site calibration.

[0119] Finally, the processor compares the topological residual of each edge node with a third preset threshold. When the topological residual is less than or equal to the third preset threshold, the processor determines that the edge node meets the spatial continuity requirement in the current 3D mesh; when the topological residual is greater than the third preset threshold, the processor adds the edge node to the residual excess node set and outputs it to step S600.

[0120] In Embodiment 2 of the present invention, step S600 may include the following steps:

[0121] S610. Obtain the main strike and dip information of the fault zone;

[0122] S620. Based on the low residual nodes around the edge node, anisotropic kriging interpolation or spline interpolation with directional constraints is used to spatially interpolate and repair the edge node, so that the main interpolation direction is consistent with the main strike and dip angle of the fault zone.

[0123] S630. Recalculate the topological residuals of edge nodes based on the repaired 3D mesh;

[0124] S640. If it is detected that the proportion of edge nodes with topological residuals greater than the third preset threshold in the three-dimensional mesh after spatial interpolation repair exceeds the preset allowable proportion (e.g., 5%), then an abnormal feedback mechanism is triggered, which feeds back to the clustering analysis step to regroup, or feeds back to the local compensation model establishment step to retrain and compensate the model.

[0125] S650. If the three-dimensional mesh after spatial interpolation repair meets the output conditions, then output the three-dimensional boundary model of the fracture zone based on the repaired three-dimensional mesh.

[0126] For example, steps S610 to S650 may further include the following steps. First, the processor acquires the main strike and dip angle information of the fault zone. The main strike is used to characterize the extension direction of the fault zone in the horizontal projection direction, and the dip angle is used to characterize the inclination state of the fault zone relative to the horizontal plane.

[0127] The preset allowable ratio is determined based on the engineering allowable error, the statistical results of control point deviation, the result display requirements of the target exploration area, or the calibration results of the maintenance stage. It is used to determine whether the proportion of nodes with excessive residuals in the repaired 3D mesh still exceeds the acceptable range.

[0128] Then, for each edge node in the set of nodes with excessive residuals, the processor selects its surrounding low-residual nodes as interpolation reference nodes. Low-residual nodes can be neighboring nodes whose topological residuals are less than or equal to a third preset threshold, or control nodes whose quality markers meet the requirements. The interpolation neighborhood radius is used to limit the range of low-residual nodes participating in the repair; it can be determined based on mesh density, fracture zone geometry, boundary continuity requirements, and field calibration results.

[0129] Subsequently, the processor uses anisotropic kriging interpolation or spline interpolation with directional constraints to spatially repair nodes with excessive residuals. Anisotropic kriging interpolation is used to set different spatial correlations in different directions, while spline interpolation with directional constraints is used to ensure that the repaired surface remains continuous along a preset direction. In this second embodiment, the main interpolation direction is consistent with the main strike and dip angle of the fault zone, the interpolation reference node is derived from surrounding low residual nodes, and the interpolation object is the edge node with excessive topological residual.

[0130] After spatial interpolation repair is completed, the processor updates the boundary node positions of nodes with excessive residuals in the 3D mesh based on the interpolation results, and retains the compensated signal parameters, compensated physical property characteristics, and quality markers corresponding to the nodes, thus reforming the repaired 3D mesh. Subsequently, the processor recalculates the topological residuals of the edge nodes in the repaired 3D mesh and counts the proportion of edge nodes whose topological residuals are still greater than the third preset threshold out of the total number of edge nodes.

[0131] If the proportion of nodes exceeding the topological residual limit exceeds the preset allowable proportion, the processor triggers an anomaly feedback mechanism. The anomaly feedback mechanism includes the following paths: First, when the nodes exceeding the residual limit are mainly concentrated in the same homogeneous interference sub-region or a few spatially adjacent homogeneous interference sub-regions, the feedback reverts to the local compensation model establishment step in step S400, and the model is retrained, parameters are adjusted, and compensation is processed again; Second, when the nodes exceeding the residual limit span multiple homogeneous interference sub-regions, and the location of the exceeding residual limit highly coincides with the boundary of the homogeneous interference sub-region, the feedback reverts to the clustering analysis step in step S300, and the local interference feature vector is readjusted or the homogeneous interference sub-regions are re-divided; Third, when the processor determines that the exceeding residual limit is mainly due to insufficient original acquisition quality, it outputs a re-acquisition prompt, causing the acquisition end to re-acquire the original electromagnetic wave signal in the corresponding area.

[0132] The inadequate quality of the original acquisition includes at least one of the following: missing channel measurements, signal-to-noise ratio lower than the preset quality control requirements, channel delay deviation still existing after synchronization and time calibration, or missing effective spatial coordinates for the corresponding measurement point.

[0133] When the number of feedbacks reaches the preset maximum number of feedbacks, and the proportion of nodes with excessive topological residuals in the repaired 3D mesh still exceeds the preset allowable proportion, the processor outputs a 3D boundary model of the fault zone with low-confidence region markers, and simultaneously outputs a prompt for re-acquisition or manual review of the corresponding region.

[0134] Finally, based on the repaired 3D mesh that meets the output conditions, the processor outputs a 3D boundary model of the fault zone. The 3D boundary model of the fault zone includes at least boundary node coordinates, mesh patches, node quality markers, low-confidence region markers, topological residual distribution results, and 3D visualization data. The node quality markers are passed along with the spatial nodes from the original signal preprocessing, abnormal attenuation identification, local compensation, 3D mesh construction, and spatial interpolation repair processes. When multiple adjacent edge nodes have low-confidence markers, or when the corresponding region still has excessive topological residuals after reaching the maximum number of feedback iterations, the processor identifies that region as a low-confidence region.

[0135] Example 3:

[0136] Based on the same general inventive concept, the present invention also provides a three-dimensional positioning system for geological fault zones based on high-frequency electromagnetic attenuation, which is used to perform the methods described in any of the above embodiments.

[0137] like Figure 3 As shown, the system may include an electromagnetic signal acquisition module, an initial boundary determination module, an abnormal attenuation identification module, a node splitting and homogeneous grouping module, a local compensation module, a three-dimensional mesh construction module, a spatial repair output module, a memory, and a processor. The electromagnetic signal acquisition module is connected to a surface electromagnetic transmission and reception array and / or a downhole electromagnetic transmission and reception array to form physical measurement input. The data acquisition unit writes the raw electromagnetic wave signal, containing the acquisition time, spatial coordinates of the measurement point, and channel identifier, into the memory. The memory is used to cache the raw electromagnetic wave signal, node coordinates, quality markers, intermediate features, model parameters, thresholds, and residual results. The processor is used to call the programs and data in the memory and control each functional module to execute in the order of steps S100 to S600.

[0138] During system operation, the output data of the previous functional module is written into the memory and used as the input data of the next functional module. When any module outputs a low confidence flag, residual over-limit result, or re-acquisition prompt, the processor transmits the quality status along with the node data to the subsequent module, and the space repair output module outputs it to the display module or external exploration interpretation terminal.

[0139] The electromagnetic signal acquisition module is used to acquire the raw electromagnetic wave signal of the target exploration area. The electromagnetic signal acquisition module may include a surface electromagnetic transmitter / receiver array and / or a downhole electromagnetic transmitter / receiver array. After acquiring the raw electromagnetic wave signal, the surface electromagnetic transmitter / receiver array and / or the downhole electromagnetic transmitter / receiver array transmits the raw electromagnetic wave signal to a processor or data acquisition unit.

[0140] The initial boundary determination module is used to extract initial physical property features from the original electromagnetic wave signal and determine the initial region boundary contour based on the initial physical property features. During system operation, the initial boundary determination module receives the original electromagnetic wave signal output by the electromagnetic signal acquisition module, performs synchronization and time correction, filtering, abnormal sampling rejection, gain compensation, and time-to-depth conversion on it, and extracts the initial physical property features. The initial region boundary contour output by the initial boundary determination module is written to the memory and used as the input of the abnormal attenuation identification module.

[0141] The abnormal attenuation identification module extracts attenuation pattern vectors of spatial nodes within the neighborhood of the initial region boundary contour. These attenuation pattern vectors are then input into a pre-trained random forest discriminant model to obtain an abnormal attenuation score for each spatial node. Based on a comparison of the abnormal attenuation score with a first preset threshold, the spatial nodes are divided into a normal region node set and an abnormal attenuation region node set. The abnormal attenuation identification module sends the normal region node set to the bypass processing channel in the node splitting and homogeneous grouping module, and sends the abnormal attenuation region node set to the homogeneous grouping processing channel in the same module.

[0142] The node splitting and homogeneous grouping module is used to write the normal region node set into the basic 3D point set, and extract the local interference feature vectors of each spatial node in the abnormal attenuation region node set. Based on the local interference feature vectors, cluster analysis is performed to divide the abnormal attenuation region node set into several homogeneous interference sub-regions. This module outputs the basic 3D point set, homogeneous interference sub-region labels, and an abnormal node sample pool.

[0143] The local compensation module is used to establish a local compensation model for each of the homogeneous interference sub-regions. This model is used to compensate for the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal within the corresponding homogeneous interference sub-region. Based on the comparison between the compensated residual evaluation value and a second preset threshold, the compensated abnormal node data is obtained. The local compensation module is also used to iteratively adjust the parameters of the local compensation model or output a downgraded compensation result with a low-confidence marker when the compensated residual evaluation value does not meet the requirements.

[0144] Before establishing the local compensation model, the processor determines a compensation caliber within the same homogeneous interference sub-region. This caliber is either a node-level signal parameter or a node-level physical property. When the compensation caliber is a node-level signal parameter, the target value of the reference sample is the calibration value or stable region reference value corresponding to that node-level signal parameter in the reference sample. When the compensation caliber is a node-level physical property, the target value of the reference sample is the calibration value or stable region reference value corresponding to that node-level physical property in the reference sample. The training input, target value of the reference sample, verification evaluation value, and compensation output of the local compensation model maintain the same compensation caliber and the same dimensions within the same homogeneous interference sub-region.

[0145] The 3D mesh construction module is used to fuse the basic 3D point set with the compensated abnormal node data to construct a 3D mesh, and calculate the topological residuals of the edge nodes in the 3D mesh. The 3D mesh construction module sends edge nodes with topological residuals greater than a third preset threshold to the spatial repair output module.

[0146] The spatial repair output module is used to perform spatial interpolation repair on edge nodes whose topological residuals exceed a third preset threshold, and outputs a three-dimensional boundary model of the fracture zone based on the repaired three-dimensional mesh. The spatial repair output module is also used to send feedback instructions to the node splitting and homogeneous grouping module or the local compensation module when there are still nodes with residuals exceeding a preset allowable proportion in the repaired three-dimensional mesh.

[0147] The memory is used to store the original electromagnetic wave signal, initial physical property characteristics, attenuation mode vector, random forest discrimination model, first preset threshold, local interference feature vector, homogeneous interference sub-region label, local compensation model, second preset threshold, compensated abnormal node data, three-dimensional mesh, topological residual, third preset threshold, and three-dimensional boundary model of the fault zone.

[0148] The processor is used to call programs and data in the memory, so that the above modules are executed sequentially according to a preset process.

[0149] In this system, the electromagnetic signal acquisition module provides physical measurement input, the initial boundary determination module converts the physical measurement input into the initial region boundary profile, the abnormal attenuation identification module divides the spatial nodes in the neighborhood of the initial region boundary profile into normal zone nodes and abnormal attenuation zone nodes, the node splitting and homogeneous grouping module writes the normal zone nodes into the basic three-dimensional point set and divides the abnormal attenuation zone nodes into several homogeneous interference sub-regions, the local compensation module outputs the compensated abnormal node data for each homogeneous interference sub-region, the three-dimensional mesh construction module merges the basic three-dimensional point set and the compensated abnormal node data to form a three-dimensional mesh, and the spatial repair output module performs spatial interpolation repair on the edge nodes with excessive topological residuals.

[0150] It should be understood that, in the embodiments of the present invention, "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean that B is determined solely based on A; B can also be determined based on A and / or other information.

[0151] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0152] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0153] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, or may be electrical, mechanical, or other forms of connection.

[0154] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0155] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0156] From the above description of the embodiments, those skilled in the art will clearly understand that the present invention can be implemented in hardware, firmware, or a combination thereof. When implemented in software, the above-described functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium accessible to a computer. For example, but not limited to, computer-readable media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible to a computer. Furthermore, any connection can suitably be considered a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. As used in this invention, disk and disc include compressed optical discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, wherein disks typically store data magnetically, while discs typically store data optically using lasers. The combinations described above should also be included within the scope of protection for computer-readable media.

[0157] In summary, the above description is merely a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation, characterized in that, The method includes: Acquire the original electromagnetic wave signal of the target exploration area, extract initial physical property features from the original electromagnetic wave signal, and determine the initial area boundary contour based on the initial physical property features; Extract the decay pattern vector of the spatial nodes in the neighborhood of the initial region boundary contour, input the decay pattern vector into a pre-trained random forest discriminant model to obtain the abnormal decay score of the spatial nodes, and divide the spatial nodes into a normal region node set and an abnormal decay region node set according to the comparison result of the abnormal decay score and a first preset threshold. Write the set of nodes in the normal region into the basic three-dimensional point set, and extract the local interference feature vectors of each spatial node in the set of nodes in the abnormal attenuation region. Based on the local interference feature vectors, perform cluster analysis to divide the set of nodes in the abnormal attenuation region into several homogeneous interference sub-regions. For each homogeneous interference sub-region, a local compensation model is established. The local compensation model is used to compensate the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal in the corresponding homogeneous interference sub-region. Based on the comparison between the compensation residual evaluation value and the second preset threshold, the compensation abnormal node data is obtained. The basic 3D point set is fused with the compensated abnormal node data to construct a 3D mesh, and the topological residual of the edge nodes in the 3D mesh is calculated. Spatial interpolation is performed on edge nodes whose topological residuals are greater than a third preset threshold for repair, and a three-dimensional boundary model of the fracture zone is output based on the repaired three-dimensional mesh.

2. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The acquisition of the raw electromagnetic wave signal of the target exploration area includes: The transient electromagnetic signal sequence or ground-penetrating radar echo signal of the target exploration area is collected by a surface electromagnetic transmission and receiving array and / or a downhole electromagnetic transmission and receiving array, and used as the original electromagnetic wave signal.

3. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The extraction of initial physical property features from the original electromagnetic wave signal includes: The original electromagnetic wave signal is subjected to synchronization and time correction, filtering, abnormal sampling rejection, gain compensation, and time-to-depth conversion. At least one of the following is extracted from the processed original electromagnetic wave signal: amplitude attenuation correlation, phase delay correlation, apparent resistivity proxy, and band energy ratio, as the initial physical property feature.

4. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The step of extracting the decay mode vector of spatial nodes in the neighborhood of the initial region boundary contour includes: Extract the attenuation trend of the spatial node in multiple frequency bands, the energy difference between it and adjacent measurement points, the phase change and spatial gradient information within the preset spatial window, and construct the attenuation mode vector.

5. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The first preset threshold is determined in the following way: The baseline threshold is determined based on the statistical distribution of stable area samples and abnormal area samples in the historical exploration data of the target exploration area; Based on the noise floor calibration results of the current batch of original electromagnetic wave signals, the reference threshold is corrected to obtain the first preset threshold. The step of writing the set of normal zone nodes into the basic three-dimensional point set includes: retaining the initial physical property characteristics corresponding to the spatial nodes in the set of normal zone nodes, or writing the spatial nodes in the set of normal zone nodes into the basic three-dimensional point set after baseline smoothing.

6. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The local interference feature vector includes at least one of high-frequency energy dissipation rate, phase fluctuation amount, and continuity deviation of adjacent nodes; when the acquisition system acquires electromagnetic responses in at least two different polarization directions, the local interference feature vector also includes polarization difference amount. The clustering analysis based on the local interference feature vector includes: standardizing the local interference feature vector and using a density-based clustering algorithm to cluster the standardized local interference feature vector, dividing spatial nodes with similar physical interference characteristics and continuous spatial distribution into the same homogeneous interference sub-region.

7. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The step of establishing a local compensation model for each of the homogeneous interference sub-regions includes: Extract signal propagation interference samples from each of the homogeneous interference sub-regions; Based on the signal propagation interference samples and reference samples, support vector machine regression compensation models corresponding to each homogeneous interference sub-region are constructed respectively. The reference sample is derived from at least one of borehole control points, trench control points, stable zone templates, and small-scale construction samples.

8. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 7, characterized in that, The comparison between the compensated residual evaluation value and the second preset threshold yields the compensated abnormal node data, including: Calculate the root mean square error or mean absolute error of the support vector machine regression compensation model on the validation sample set, and use it as the evaluation value of the compensation residual. When the compensated residual evaluation value is less than or equal to the second preset threshold, the compensated abnormal node data is output using the current support vector machine regression compensation model. When the post-compensation residual evaluation value is greater than the second preset threshold, at least one of the penalty parameter, kernel scale parameter and sample weight of the support vector machine regression compensation model is iteratively adjusted. When the compensated residual evaluation value recalculated based on the adjusted support vector machine regression compensation model is less than or equal to the second preset threshold, the compensated abnormal node data is output using the adjusted support vector machine regression compensation model. When the number of iterations reaches the preset maximum number of iterations, and the residual evaluation value after compensation is still greater than the second preset threshold, low-order local smoothing compensation is used as the downgrade compensation result for the corresponding homogeneous interference sub-region, and a low confidence flag is set for the homogeneous interference sub-region.

9. The three-dimensional location method for geological fault zones based on high-frequency electromagnetic attenuation according to claim 1, characterized in that, The calculation of the topological residuals of the edge nodes in the 3D mesh includes: Determine the deviation of the edge node relative to the neighborhood continuity constraint surface and / or control point, and use the deviation as the topological residual; The third preset threshold is determined based on control point deviation, neighborhood continuity deviation, and engineering tolerance error. The step of performing spatial interpolation repair on edge nodes with topological residuals greater than a third preset threshold includes: obtaining the main strike and dip information of the fault zone; and performing spatial interpolation repair on the edge nodes based on the low residual nodes around the edge nodes using anisotropic kriging interpolation or spline interpolation with directional constraints, so that the main interpolation direction is consistent with the main strike and dip of the fault zone. If it is detected that the proportion of edge nodes with topological residuals greater than the third preset threshold in the 3D mesh after spatial interpolation repair exceeds the preset allowable proportion, an anomaly feedback mechanism is triggered, which either reverts to the clustering analysis step for regrouping or reverts to the local compensation model establishment step for retraining and compensation.

10. A three-dimensional positioning system for geological fault zones based on high-frequency electromagnetic attenuation, characterized in that, The system includes: The electromagnetic signal acquisition module is used to acquire the original electromagnetic wave signals of the target exploration area; An initial boundary determination module is used to extract initial physical property features from the original electromagnetic wave signal and determine the initial region boundary contour based on the initial physical property features. An abnormal decay identification module is used to extract the decay pattern vector of spatial nodes in the neighborhood of the initial region boundary contour, input the decay pattern vector into a pre-trained random forest discriminant model to obtain the abnormal decay score of the spatial node, and divide the spatial node into a normal region node set and an abnormal decay region node set according to the comparison result of the abnormal decay score and a first preset threshold. The node splitting and homogeneous grouping module is used to write the normal region node set into the basic three-dimensional point set, extract the local interference feature vector of each spatial node in the abnormal attenuation region node set, perform cluster analysis based on the local interference feature vector, and divide the abnormal attenuation region node set into several homogeneous interference sub-regions. The local compensation module is used to establish a local compensation model for each of the homogeneous interference sub-regions, and to use the local compensation model to compensate the node-level signal parameters or node-level physical properties obtained from the original electromagnetic wave signal in the corresponding homogeneous interference sub-region. Based on the comparison result of the compensation residual evaluation value and the second preset threshold, the compensated abnormal node data is obtained. A three-dimensional mesh construction module is used to fuse the basic three-dimensional point set with the compensated abnormal node data to construct a three-dimensional mesh, and to calculate the topological residual of the edge nodes in the three-dimensional mesh; The spatial repair output module is used to perform spatial interpolation repair on edge nodes whose topological residuals are greater than a third preset threshold, and output a three-dimensional boundary model of the fracture zone based on the repaired three-dimensional mesh.

Images