A fault dip prediction method based on tensor voting

Abstract

This invention discloses a fault dip angle prediction method based on tensor voting, comprising the following steps: S1, acquiring post-stack seismic data and constructing an initial gradient structure tensor; S2, using the gradient structure tensor to extract preliminary fault detection probability attributes; S3, based on the fault detection results, performing spherical voting on the gradient structure tensor to reconstruct the spatial topological connectivity of the fault surface under complex noise background; S4, using the fault detection results after spherical voting to reconstruct a guiding tensor with structural consistency; S5, using the reconstructed gradient structure tensor to perform bar voting, finally obtaining the fault dip angle prediction result. This invention achieves deep coupling of physical texture information and geometric structural features through multi-level tensor voting evolution, effectively avoiding interference from stratigraphic texture on fault attitude identification, solving the technical problem of inconsistent local orientation with the strike of large faults, and significantly improving the accuracy and robustness of fault dip angle prediction.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a method for predicting fault dip angle based on tensor voting. Background Technology

[0002] Fault dip angle, as a core parameter describing fault geometry, is fundamental for accurate calculation in geological modeling, geostress analysis, and engineering hazard prediction. In current geological exploration and geotechnical engineering practices, accurate calculation of fault dip angle has the following irreplaceable necessities, but existing technologies struggle to meet high-precision requirements:

[0003] (1) Mathematical basis for reconstructing real underground geological structure

[0004] The attitude of a fault in three-dimensional space is defined by its strike, dip, and dip angle. Among these, the dip angle determines the steepness of the fault and is a key indicator for distinguishing between high-angle faults (>45°) and low-angle faults (<45°). If the dip angle calculation deviation exceeds ±5°, it will directly cause the spatial distribution of the fault in the geological model to shift by tens or even hundreds of meters, leading to incorrect judgments about trap structures and the direction of ore body extension.

[0005] (2) Key input parameters that determine the distribution of geostress and the determination of rock mass instability

[0006] In tunnel, slope, dam foundation, and underground chamber engineering, faults often become weak points controlling the stability of the surrounding rock. Theoretical mechanics and the Mohr-Coulomb failure criterion indicate that:

[0007] The fault dip angle directly affects the relationship between the sliding force and the normal stress. When the fault dip angle is close to the friction angle within the rock mass, the fault is most prone to sliding instability.

[0008] Inaccurate dip angle values ​​can lead to serious deviations in support design. For example, underestimating the dip angle may underestimate the risk of collapse after excavation; overestimating the dip angle may result in over-support and unnecessary engineering costs.

[0009] (3) Optimize oil and gas migration paths and predict ore body location

[0010] In oil and gas exploration, faults can be both channels for oil and gas migration and barriers that can form traps.

[0011] The dip angle determines the sealing capacity of a fault. Faults with different dip angles exhibit significant differences in their fault gouge ratio (SGR) under compaction and diagenesis. Accurate calculation of the dip angle is a necessary step for quantitatively evaluating lateral sealing capacity.

[0012] For hydrothermal deposits, areas where the fault dip changes (such as from steep to gentle) are often favorable spaces for ore body placement. The lack of accurate dip calculations will lead to inaccurate drilling target selection, increasing exploration risks.

[0013] (4) Distinguish between evidence of structural superposition and evidence of multiple phases of activity

[0014] The same fault often undergoes multiple phases of tectonic movement, and the kinematic characteristics (such as dip angle changes) of the early and late stages record the rotation of the regional stress field. By calculating the differences in fault dip angles at different depths and strata with high precision, the transition history of the fault from extensional to compressive can be inverted. Existing methods (such as relying solely on downhole core dip logging or surface outcrop compass measurements) cannot effectively isolate the effects of later alteration, therefore, new calculation methods must be developed. Summary of the Invention

[0015] To address the above problems, this invention proposes a fault dip angle prediction method based on tensor voting.

[0016] The technical solution of this invention is: a fault dip angle prediction method based on tensor voting, comprising the following steps:

[0017] S1. Acquire post-stack seismic data and construct the initial gradient structure tensor;

[0018] S2. Use the gradient structure tensor to extract the preliminary fault detection probability attributes;

[0019] S3. Based on the fault detection results, perform ball voting on the gradient structure tensor to obtain the gradient structure tensor, eigenvalues ​​and eigenvectors after ball voting, so as to reconstruct the spatial topological connectivity of the fracture surface under complex noise background.

[0020] S4. Utilize the fault detection results after ball voting and fuse them with the original gradient structure tensor in the feature space to reconstruct a guiding tensor with construction consistency.

[0021] S5. Using the reconstructed gradient structure tensor, bar voting is performed to achieve nonlinear enhancement of cross-sectional energy, ultimately obtaining the fault dip angle prediction result.

[0022] Furthermore, in S1, the formula for calculating the gradient structure tensor is:

[0023] ;

[0024] ;

[0025] ;

[0026] in, Represents post-stack seismic data. Represents the gradient in the x-direction. This represents the gradient in the y-direction. Represents the gradient structure tensor. This represents the first component of the gradient structure tensor. This represents the second component of the gradient structure tensor. This represents the third component of the gradient structure tensor. This indicates the smoothing of the outer product of the gradients of x and y. This indicates that the outer product of x and x's gradient is smooth. This indicates that the outer product of y and y gradient is smooth. Represents the first eigenvector. Represents the second eigenvector. Represents the first eigenvalue. This represents the second eigenvalue.

[0027] Furthermore, in S2, the formula for calculating the fault detection results is:

[0028] ;

[0029] in, This indicates the results of the fault detection. Represents the directional derivative. This represents the sensitivity coefficient.

[0030] Furthermore, S3 includes the following sub-steps:

[0031] S31. Using the gradient structure tensor, perform isotropic spherical voting, and bridge the fault fracture response through spatial energy diffusion to reconstruct the topological connectivity of the fracture surface, thereby obtaining the gradient structure tensor enhanced by spherical voting.

[0032] S32. Perform eigenvalue decomposition on the enhanced gradient structure tensor to extract eigenvalues ​​representing local energy intensity and eigenvectors representing the initial orientation direction.

[0033] Furthermore, in S31, the formula for calculating the gradient structure tensor after ball voting is:

[0034] ;

[0035] ;

[0036] in, Represents the ball voting operator. This represents the arc length between the voting point and the receiving point during the ball voting process. A constant parameter representing the degree of curvature decay. This represents the curvature of the curve connecting the voting points and the receiving points during the ball voting process. Indicates the voting neighborhood range. The gradient structure tensor represents the result of ball voting;

[0037] In S32, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after ball voting are as follows:

[0038] ;

[0039] in, Let represent the first eigenvector of the gradient structure tensor after ball voting. Let represent the second eigenvector of the gradient structure tensor after ball voting. This represents the first eigenvalue of the gradient structure tensor after ball voting. This represents the second eigenvalue of the gradient structure tensor after ball voting.

[0040] Furthermore, in S4, the formula for calculating the reconstructed gradient structure tensor is:

[0041] ;

[0042] in, Let s represent the reconstructed gradient structure tensor, and s represent the initial fault detection result. The tensor representing the result of the ball's vote. Represents the fusion coefficient. Represents the trace operator, This represents the gradient tensor of the more continuous tomographic detection results after the ball voting.

[0043] Furthermore, S5 includes the following sub-steps:

[0044] S51. Perform anisotropic rod voting using the reconstructed gradient structure tensor, and perform orientation calibration and connectivity enhancement on the initial tensor field using spatial curvature consistency constraints to obtain the enhanced structure tensor field.

[0045] S52. Perform eigenvalue decomposition on the enhanced structural tensor field, extract its principal eigenvector as the normal vector of the fault, and then calculate the spatial orientation information of the fault.

[0046] Furthermore, in S51, the formula for calculating the gradient structure tensor after bar voting is:

[0047] ;

[0048] ;

[0049] in, This represents the stick voting operator. The gradient structure tensor after bar voting;

[0050] In S52, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after bar voting are as follows:

[0051] ;

[0052] in, The first eigenvector of the gradient structure tensor after bar voting is represented. This represents the second eigenvector of the gradient structure tensor after bar voting. The first eigenvalue of the gradient structure tensor after bar voting is represented. The second eigenvalue represents the gradient structure tensor after the bar voting.

[0053] The beneficial effects of this invention are:

[0054] Based on post-stack seismic data, this invention introduces the tensor voting method to calculate fault angles, effectively avoiding interference from strata and noise on fault direction identification. It solves the problem of inconsistency between local fault directions and the strike of major faults, identifies the strike of major faults, and facilitates subsequent fault interpretation work. Attached Figure Description

[0055] Figure 1 This is a flowchart of a fault dip angle prediction method based on tensor voting;

[0056] Figure 2 This is a schematic diagram of the fault detection results;

[0057] Figure 3 A schematic diagram of the tomographic results after voting for the ball;

[0058] Figure 4 A schematic diagram of the initial fault direction after voting for the ball;

[0059] Figure 5 A schematic diagram showing the fault direction after the bar voting;

[0060] Figure 6 This is a schematic diagram of the fault identification results based on the fault direction after bar voting. Detailed Implementation

[0061] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0062] like Figure 1 As shown, this invention provides a fault dip angle prediction method based on tensor voting, comprising the following steps:

[0063] S1. Acquire post-stack seismic data and construct the initial gradient structure tensor;

[0064] S2. Use the gradient structure tensor to extract the preliminary fault detection probability attributes;

[0065] S3. Based on the fault detection results, perform ball voting on the gradient structure tensor to obtain the gradient structure tensor, eigenvalues ​​and eigenvectors after ball voting, so as to reconstruct the spatial topological connectivity of the fracture surface under complex noise background.

[0066] S4. Utilize the fault detection results after ball voting and fuse them with the original gradient structure tensor in the feature space to reconstruct a guiding tensor with construction consistency.

[0067] S5. Using the reconstructed gradient structure tensor, bar voting is performed to achieve nonlinear enhancement of cross-sectional energy, and finally the fault dip angle prediction result is obtained.

[0068] In this embodiment of the invention, in S1, the formula for calculating the gradient structure tensor is:

[0069] ;

[0070] ;

[0071] ;

[0072] in, Represents post-stack seismic data. Represents the gradient in the x-direction. This represents the gradient in the y-direction. Represents the gradient structure tensor. This represents the first component of the gradient structure tensor. This represents the second component of the gradient structure tensor. This represents the third component of the gradient structure tensor. This indicates the smoothing of the outer product of the gradients of x and y. This indicates that the outer product of x and x's gradient is smooth. This indicates that the outer product of y and y gradient is smooth. Represents the first eigenvector. Represents the second eigenvector. Represents the first eigenvalue. This represents the second eigenvalue.

[0073] Using post-stack seismic data Construct gradient structure tensor And obtain its eigenvalues ​​and eigenvectors. Let T be the gradient in the x and y directions, obtained by smoothing the gradient through the outer product. '—' indicates the outer product smoothing sign. Perform eigenvalue decomposition to obtain eigenvalues. , and eigenvectors , .

[0074] In this embodiment of the invention, in S2, the calculation formula for the fault detection result is as follows:

[0075] ;

[0076] in, This indicates the results of the fault detection. Represents the directional derivative. This represents the sensitivity coefficient.

[0077] like Figure 2 As shown, in an embodiment of the present invention, The smaller the value, the more fault points are detected. The fault detection results in this step. Alternatively, properties such as coherence and curvature can be used as substitutes.

[0078] In this embodiment of the invention, S3 includes the following sub-steps:

[0079] S31. Using the gradient structure tensor, perform isotropic spherical voting, and bridge the fault fracture response through spatial energy diffusion to reconstruct the topological connectivity of the fracture surface, thereby obtaining the gradient structure tensor enhanced by spherical voting.

[0080] S32. Perform eigenvalue decomposition on the enhanced gradient structure tensor to extract eigenvalues ​​representing local energy intensity and eigenvectors representing the initial orientation direction.

[0081] In this embodiment of the invention, in S31, the formula for calculating the gradient structure tensor after ball voting is:

[0082] ;

[0083] ;

[0084] in, Represents the ball voting operator. This represents the arc length between the voting point and the receiving point during the ball voting process. A constant parameter representing the degree of curvature decay. This represents the curvature of the curve connecting the voting points and the receiving points during the ball voting process. Indicates the voting neighborhood range. The gradient structure tensor represents the result of ball voting;

[0085] In S32, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after ball voting are as follows:

[0086] ;

[0087] in, Let represent the first eigenvector of the gradient structure tensor after ball voting. Let represent the second eigenvector of the gradient structure tensor after ball voting. This represents the first eigenvalue of the gradient structure tensor after ball voting. This represents the second eigenvalue of the gradient structure tensor after ball voting.

[0088] In this embodiment of the invention, a ball vote is performed on the tensor T to obtain the preliminary fault direction at the location with the strongest fault energy, that is, a more continuous fault detection result. The voting neighborhood (scale parameter) determines the search radius for the sphere voting. The tensor after sphere voting... Perform eigenvalue decomposition to obtain eigenvalues. , and eigenvectors , .

[0089] like Figure 3 As shown, after the ball voting, the fault detection results were relatively... Figure 2 More continuous. For example... Figure 4 As shown, at fault locations with higher energy, the eigenvectors after ball voting... The direction is consistent with the fault strike.

[0090] In this embodiment of the invention, in S4, the formula for calculating the reconstructed gradient structure tensor is:

[0091] ;

[0092] in, Let s represent the reconstructed gradient structure tensor, and s represent the initial fault detection result. The tensor representing the result of the ball's vote. Represents the fusion coefficient. Represents the trace operator, This represents the gradient tensor of the more continuous tomographic detection results after the ball voting.

[0093] In this embodiment of the invention, based on the fault detection results after ball voting, the tensor obtained from ball voting and the tensor constructed from the fault detection results after ball voting are fused. The specific implementation steps are as follows: Based on the fault detection attribute field enhanced by ball voting, its spatial gradient vector is extracted using a gradient operator, and a geometric structure tensor representing the fault geometry is constructed based on this vector; simultaneously, the energy trace information of the tensor obtained from ball voting is extracted, and this is used as an amplitude scaling factor to map onto the geometric structure tensor to retain the energy intensity characteristics of the original seismic data; the geometric structure tensor and the tensor obtained from ball voting are weighted and fused to construct a reconstructed gradient structure tensor; the geometric normal constraint provided by the geometric structure tensor is used to correct the directional singularity of the fault core area, thereby eliminating the directional convergence effect introduced by the ball voting process.

[0094] In this embodiment of the invention, the newly constructed tensor The process involves bar voting, which exhibits "long-range radiation" characteristics. It allows fault energy to extend outwards along the direction obtained from S33 to obtain more accurate fault strike information, "welding" previously broken and isolated fault points into continuous curves or surfaces, significantly improving the linear characteristics of fault properties. Furthermore, through mutual voting support within the neighborhood, it can offset residual random directions, making the final direction field more consistent with the general trend of regional geological structure.

[0095] In this embodiment of the invention, S5 includes the following sub-steps:

[0096] S51. Perform anisotropic rod voting using the reconstructed gradient structure tensor, and perform orientation calibration and connectivity enhancement on the initial tensor field using spatial curvature consistency constraints to obtain the enhanced structure tensor field.

[0097] S52. Perform eigenvalue decomposition on the enhanced structural tensor field, extract its principal eigenvector as the normal vector of the fault, and then calculate the spatial orientation information of the fault.

[0098] In this embodiment of the invention, in S51, the formula for calculating the gradient structure tensor after bar voting is:

[0099] ;

[0100] ;

[0101] in, This represents the stick voting operator. The gradient structure tensor after bar voting;

[0102] In S52, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after bar voting are as follows:

[0103] ;

[0104] in, The first eigenvector of the gradient structure tensor after bar voting is represented. This represents the second eigenvector of the gradient structure tensor after bar voting. The first eigenvalue of the gradient structure tensor after bar voting is represented. The second eigenvalue represents the gradient structure tensor after the bar voting.

[0105] In this embodiment of the invention, the tensor after bar voting... Perform eigenvalue decomposition to obtain eigenvalues. , and eigenvectors , .like Figure 5 As shown, That is, the fault direction.

[0106] In embodiments of the present invention, such as Figure 6 As shown, based on the fault calculation results of S52, subsequent work such as fault direction determination or fault smoothing can be performed more accurately.

[0107] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A fault dip angle prediction method based on tensor voting, characterized in that, Includes the following steps: S1. Acquire post-stack seismic data and construct the initial gradient structure tensor; S2. Use the gradient structure tensor to extract the preliminary fault detection probability attributes; S3. Based on the fault detection results, perform ball voting on the gradient structure tensor to obtain the gradient structure tensor, eigenvalues ​​and eigenvectors after ball voting, so as to reconstruct the spatial topological connectivity of the fracture surface under complex noise background. S4. Utilize the fault detection results after ball voting and fuse them with the original gradient structure tensor in the feature space to reconstruct a guiding tensor with construction consistency. S5. Using the reconstructed gradient structure tensor, bar voting is performed to achieve nonlinear enhancement of cross-sectional energy, and finally the fault dip angle prediction result is obtained.

2. The fault dip angle prediction method based on tensor voting according to claim 1, characterized in that, In S1, the formula for calculating the gradient structure tensor is: ； ； ； in, Represents post-stack seismic data. Represents the gradient in the x-direction. This represents the gradient in the y-direction. Represents the gradient structure tensor. This represents the first component of the gradient structure tensor. This represents the second component of the gradient structure tensor. This represents the third component of the gradient structure tensor. This indicates the smoothing of the outer product of the gradients of x and y. This indicates that the outer product of x and x's gradient is smooth. This indicates that the outer product of y and y gradient is smooth. Represents the first eigenvector. Represents the second eigenvector. Represents the first eigenvalue. This represents the second eigenvalue.

3. The fault dip angle prediction method based on tensor voting according to claim 1, characterized in that, In S2, the formula for calculating the fault detection result is as follows: ； in, This indicates the results of the fault detection. Represents the directional derivative. This represents the sensitivity coefficient.

4. The fault dip angle prediction method based on tensor voting according to claim 1, characterized in that, S3 includes the following sub-steps: S31. Using the gradient structure tensor, perform isotropic spherical voting, and bridge the fault fracture response through spatial energy diffusion to reconstruct the topological connectivity of the fracture surface, thereby obtaining the gradient structure tensor enhanced by spherical voting. S32. Perform eigenvalue decomposition on the enhanced gradient structure tensor to extract eigenvalues ​​representing local energy intensity and eigenvectors representing the initial orientation direction.

5. The fault dip angle prediction method based on tensor voting according to claim 4, characterized in that, In S31, the formula for calculating the gradient structure tensor after ball voting is: ； ； in, Represents the ball voting operator. This represents the arc length between the voting point and the receiving point during the ball voting process. A constant parameter representing the degree of curvature decay. This represents the curvature of the curve connecting the voting points and the receiving points during the ball voting process. Indicates the voting neighborhood range. The gradient structure tensor represents the result of ball voting; In step S32, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after ball voting are as follows: ； in, Let represent the first eigenvector of the gradient structure tensor after ball voting. Let represent the second eigenvector of the gradient structure tensor after ball voting. This represents the first eigenvalue of the gradient structure tensor after ball voting. This represents the second eigenvalue of the gradient structure tensor after ball voting.

6. The fault dip angle prediction method based on tensor voting according to claim 1, characterized in that, In S4, the formula for calculating the reconstructed gradient structure tensor is: ； in, Let s represent the reconstructed gradient structure tensor, and s represent the initial fault detection result. The tensor representing the result of the ball's vote. Represents the fusion coefficient. Represents the trace operator, This represents the gradient tensor of the more continuous tomographic detection results after the ball voting.

7. The fault dip angle prediction method based on tensor voting according to claim 1, characterized in that, S5 includes the following sub-steps: S51. Perform anisotropic rod voting using the reconstructed gradient structure tensor, and perform orientation calibration and connectivity enhancement on the initial tensor field using spatial curvature consistency constraints to obtain the enhanced structure tensor field. S52. Perform eigenvalue decomposition on the enhanced structural tensor field, extract its principal eigenvector as the normal vector of the fault, and then calculate the spatial orientation information of the fault.

8. The fault dip angle prediction method based on tensor voting according to claim 7, characterized in that, In S51, the formula for calculating the gradient structure tensor after bar voting is: ； ； in, This represents the stick voting operator. The gradient structure tensor after bar voting; In step S52, the formulas for calculating the eigenvalues ​​and eigenvectors of the gradient structure tensor after bar voting are as follows: ； in, The first eigenvector of the gradient structure tensor after bar voting is represented. This represents the second eigenvector of the gradient structure tensor after bar voting. The first eigenvalue of the gradient structure tensor after bar voting is represented. The second eigenvalue represents the gradient structure tensor after the bar voting.

Images