Fracture identification method based on aviation gravity and magnetic data

Abstract

The invention discloses a fracture identification method based on aviation gravity and magnetic data, and relates to the technical field of fracture identification, historical aviation gravity and magnetic data are processed through Euler deconvolution, historical field source position and depth data are obtained, and a mapping model is established; based on the field source position and the depth data, adopting a preset fracture position prediction model to perform fracture position prediction, and obtaining fracture position data; meanwhile, fracture type data is obtained based on fuzzy clustering; on the basis of the to-be-measured area, to-be-measured aviation gravity and magnetic data are acquired; determining an initial estimated fracture parameter of the to-be-detected area, calculating estimated aviation gravity and magnetic data based on the mapping model, obtaining a loss function, judging whether the loss function value exceeds a preset precision threshold, if so, ending iteration, and outputting an optimal fracture characteristic parameter; and outputting the fracture position, the fracture type and the fracture parameters of the final iteration. According to the invention, accurate identification of fractures under complex geological conditions is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of fracture identification, and more specifically, to a fracture identification method based on airborne gravity and magnetic data. Background Art

[0002] In the fields of geoscience and resource exploration, fractures, as a key component of geological structures, profoundly influence various geological processes and the distribution pattern of resources. From vast onshore oil and gas fields to deep underground mineral layers and then to oceanic crustal plates, fractures not only maintain the stability of geological bodies, but their formation and distribution directly determine the formation and migration paths of oil, gas, and mineral resources. Therefore, accurately identifying the location, shape, and scale of fractures is of inestimable significance for in-depth research on geological evolution, efficient exploration of potential resources, and effective assessment of geological disaster risks.

[0003] Early traditional fracture identification methods such as geological sampling and shallow seismic profiling can obtain relatively detailed geological information in local areas, but generally suffer from defects such as low efficiency, high cost, and limited coverage. With the continuous expansion of exploration scale, traditional methods can no longer meet the urgent need for rapid and accurate identification of large areas. With the continuous progress of geophysical exploration technology, airborne geophysical surveys, with their significant advantages of high efficiency and wide coverage, have gradually become an important technical means in the field of geological exploration. Airborne gravity and magnetic surveys can quickly collect gravity and magnetic field data over large areas. Different geological bodies will produce unique abnormal responses in the gravity and magnetic data due to differences in density and magnetism. The presence of fractures will significantly change the density and magnetic distribution of surrounding rocks, and thus corresponding abnormal characteristics will appear in the airborne gravity and magnetic data. However, relying solely on these gravity and magnetic anomaly data, it is difficult to accurately determine key parameters such as the location, depth, and scale of fractures, and the accuracy and reliability of fracture identification need to be further improved.

[0004] As a mature potential field inversion technology, Euler deconvolution has been widely used in the field of geological exploration. Based on the Euler homogeneous equation, this method can quickly invert and calculate the location and depth of the field source by using potential field anomalies and the "structural index" of geological bodies. In practical fracture identification work, by reasonably setting the structural index related to fractures and combining it with airborne gravity and magnetic anomaly data, Euler deconvolution provides a new technical idea for determining the location and depth of fractures, and to a certain extent, improves the efficiency and accuracy of fracture identification.

[0005] Although a large number of studies have currently applied various gravity and magnetic anomaly inversion technologies to geological structure detection, in the field of fracture identification, the existing inversion methods are significantly insufficient in the accuracy and stability of gravity and magnetic data processing when facing complex and variable geological environments, and are difficult to handle fracture identification tasks under various complex geological conditions.

[0006] Therefore, how to accurately identify faults under complex geological conditions is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a fault identification method based on airborne gravity and magnetic data to solve the problems existing in the above-mentioned background technology.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A fault identification method based on airborne gravity and magnetic data, comprising:

[0010] Obtain historical airborne gravity and magnetic data, process the historical airborne gravity and magnetic data through Euler deconvolution to obtain historical field source position and depth data; based on the neural network algorithm, establish a mapping model of the historical airborne gravity and magnetic data, fault characteristic parameters, and field source position and depth data;

[0011] Based on the historical field source position and depth data in the Euler deconvolution process, use a preset fault position prediction model to predict the fault position to obtain fault position data; at the same time, obtain fault type data based on fuzzy clustering;

[0012] Based on the area to be measured, obtain the airborne gravity and magnetic data to be measured; confirm the initial estimated fault parameters of the area to be measured, calculate the estimated airborne gravity and magnetic data based on the mapping model, calculate the loss function of the airborne gravity and magnetic data to be measured and the estimated airborne gravity and magnetic data, and determine whether the loss function value exceeds a preset accuracy threshold. If it does not exceed the preset accuracy threshold, optimize the fault characteristic parameters through particle swarm intelligence; if it is satisfied, end the iteration and output the optimal fault characteristic parameters;

[0013] Output the fault position, fault type, and fault parameters output in the final iteration step.

[0014] Preferably, the historical airborne gravity and magnetic data is obtained through gravity and magnetic measurements at two different measurement heights, namely the first historical airborne gravity and magnetic data Q1 and the second historical airborne gravity and magnetic data Q2.

[0015] Preferably, the fault characteristic parameters include fault length, fault width, fault dip angle, and fault type.

[0016] Preferably, the mapping model specifically includes:

[0017]

[0018] where l is the fault length, w is the fault width, d is the fault depth, θ is the fault dip angle, F cis the fracture type, t1 and t2 are the measured heights, G1 and G2 are the historical gravity and magnetic data, and (x, y) is the field source position.

[0019] Preferably, the fracture position prediction specifically includes:

[0020] Each particle i represents a set of SVM parameter combinations p i =(γ i , C i ), where γ i is the kernel parameter and C i is the penalty factor;

[0021]

[0022] where k is the number of iterations, is the velocity vector of the i-th particle at the (k + 1)-th iteration, corresponding to the velocity components of the kernel parameter γ and the penalty factor C respectively, ω is the inertia weight, c1 and c2 are the learning factors, r1 and r2 are random numbers between [0, 1], pBest i,γ and pBest i,C are the kernel parameter and penalty factor corresponding to the i-th particle when it reaches the optimal fitness value in its own history, gBest γ and gBest C are the global optimal kernel parameter and penalty factor corresponding to all particles when they reach the optimal fitness value in the historical iterations, and are the kernel parameter and penalty factor of the i-th particle at the k-th iteration, m is the number of validation samples, is the result of predicting the j-th validation sample using the SVM parameter combination (γ i , C i ) represented by the i-th particle, y j is the actual fracture position label of the j-th validation sample;

[0023] SVM model:

[0024]

[0025] 0 ≤ α i ≤ C, i = 1, …, n;

[0026] K(x i , x j ) = exp(-γ‖x i - x j ‖ 2 );

[0027]

[0028] where xi , x j , where x is the input sample data, which are respectively gravity gradient, magnetic gradient and depth data; γ is the kernel parameter, and α i and α j are both Lagrange multipliers, y i is the class label of the sample, and b is the bias term;

[0029] Initialize the particle swarm and randomly generate the initial positions of a group of particles and velocities

[0030] For each particle i, use its SVM parameter combination to train the SVM model, solve for the Lagrange multipliers α i , α j and the bias term b, and calculate the fitness value Fitness i ;

[0031] Update the pBest i,γ , pBest i,C and gBest γ , gBest C ;

[0032] Update the velocities and positions of the particles according to the particle velocity and position update formulas;

[0033] Repeat the iterative update until the maximum number of iterations is reached, obtain the final fracture position prediction model, and perform fracture position prediction.

[0034] Preferably, obtaining the fracture type data based on fuzzy clustering specifically includes:

[0035] Determine the number of clusters c of the fracture type and select the fuzzy factor m, m ∈ (1, +∞), and set the iteration stop threshold ε; construct the membership matrix U of n × c = [u ij , where u ij is the membership degree of sample q i belonging to the j-th class,

[0036] Calculate the cluster center v j of each class according to the membership matrix U, and the calculation formula is:

[0037]

[0038] where v j is the cluster center of the j-th class, obtained by weighted averaging the sample data, and the weights are determined by the m-th power of the membership degree;

[0039] Update the membership matrix U using the current cluster center v j , and the update formula is:

[0040]

[0041] Among them, ||q i -v j || represents the distance between the sample x i and the clustering center v j .

[0042] Define the objective function J of fuzzy clustering, and the formula is:

[0043]

[0044] Calculate the objective function value J of this iteration (t) and the objective function value J of the previous iteration (t-1) . If it satisfies J (t) -J (t -1) ≤ε, it is considered that the iteration converges and the iteration stops; otherwise, continue the iteration, continuously update the clustering center and the membership matrix until the termination condition is satisfied; when the iteration ends, determine the fracture type to which each sample belongs according to the final membership matrix U. For each sample q i , the fracture type is the category with the largest membership degree.

[0045] Preferably, the loss function specifically includes:

[0046]

[0047] Among them, Loss is the loss function value, n represents the number of data samples, and are the measured gravity data and measured magnetic data of the i-th sample respectively, and are the historical gravity data and historical magnetic data of the i-th sample respectively, the data calculated based on the established mapping model, and the historical gravity and magnetic data include historical gravity data and historical magnetic data.

[0048] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a fracture recognition method based on airborne gravity and magnetic data. It combines the Euler deconvolution method to process historical airborne gravity and magnetic data to obtain the field source position and depth data, uses the neural network algorithm to establish a mapping model, and considers the relationship between fracture characteristic parameters and gravity and magnetic data. It also uses a preset fracture position prediction model and fuzzy clustering to determine the fracture position and type respectively. By integrating a variety of technical means, it analyzes and mines fracture information from different angles. Compared with a single method, it greatly improves the accuracy and reliability of fracture recognition and can more accurately determine key information such as the position, depth, and scale of fractures. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained according to the provided accompanying drawings.

[0050] Figure 1 It is a method step diagram provided by the present invention. Specific embodiments

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0052] An embodiment of the present invention discloses a fracture identification method based on airborne gravity and magnetic data, as Figure 1 shown, including:

[0053] Obtain historical airborne gravity and magnetic data, process the historical airborne gravity and magnetic data through Euler deconvolution to obtain historical field source position and depth data; based on the neural network algorithm, establish a mapping model of historical airborne gravity and magnetic data, fracture characteristic parameters, and field source position and depth data;

[0054] Based on the historical field source position and depth data in the Euler deconvolution process, use a preset fracture position prediction model to predict the fracture position and obtain fracture position data; at the same time, obtain fracture type data based on fuzzy clustering;

[0055] Based on the area to be measured, obtain the airborne gravity and magnetic data to be measured; confirm the initial estimated fracture parameters of the area to be measured, calculate the estimated airborne gravity and magnetic data based on the mapping model, calculate the loss function of the airborne gravity and magnetic data to be measured and the estimated airborne gravity and magnetic data, and determine whether the loss function value exceeds a preset accuracy threshold. If it does not exceed the preset accuracy threshold, optimize the fracture characteristic parameters through particle swarm intelligence; if it is satisfied, end the iteration and output the optimal fracture characteristic parameters;

[0056] Output the fracture position, fracture type, and fracture parameters output in the final iteration step.

[0057] The field source body with the center point at (x0, y0, z0) satisfies the following Euler equation:

[0058]

[0059] The parameter N in the homogeneous equation is defined as the structural index, which needs to be determined according to the shape of the field source or the known information about the nature of the anomaly. It has a certain corresponding relationship with the regular shape of the field source (Table 1). For example, for a gravitational field source, the structural index of a long and narrow two-dimensional rock wall is 0, and that of a uniformly massive sphere is 2, etc. For a field source with a given geometric shape, a structural index is the rate at which an exponential factor corresponding to the field drops off with distance.

[0060] Table 1 Corresponding relationship between field source and structural index

[0061] Gravity anomaly source Structure index Magnetic anomaly source Structure index Dyke or fault structure 0 Point pole 1 Uniform mass horizontal cylinder 1 <![CDATA[Inclined thin plate (Z a )]]> 1 Finite inclined step 1 Uniformly magnetized horizontal cylinder (bipolar line) 2 Uniform mass sphere 2 Uniformly magnetized sphere (dipole) 3

[0062] When performing Euler deconvolution inversion calculation, generally the influence of the regional field or background field B needs to be considered. Regarding the potential field as the sum of a point source field and a regional field, the Euler equation is as follows:

[0063]

[0064] By solving the Euler equation, the position and depth of the field source can be inversely obtained.

[0065] In a specific embodiment, the historical airborne gravity and magnetic data are obtained through gravity and magnetic measurements at two different measurement heights, namely the first historical airborne gravity and magnetic data Q1 and the second historical airborne gravity and magnetic data Q2.

[0066] In a specific embodiment, the fracture characteristic parameters include fracture length, fracture width, fracture dip angle, and fracture type.

[0067] In a specific embodiment, the mapping model specifically includes:

[0068]

[0069] where l is the fracture length, w is the fracture width, d is the fracture depth, θ is the fracture dip angle, F c is the fracture type, t1 and t2 are the measurement heights, G1 and G2 are the historical gravity and magnetic data, and (x, y) is the position of the field source.

[0070] In a specific embodiment, the prediction of the fracture position specifically includes:

[0071] Each particle i represents a set of SVM parameter combinations p i =(γ i , C i ), where γ i is the kernel parameter and C i is the penalty factor;

[0072]

[0073] where k is the number of iterations, is the velocity vector of the $i$-th particle at the $(k + 1)$-th iteration, corresponding to the velocity components of the kernel parameter $\gamma$ and the penalty factor $C$ respectively. $\omega$ is the inertia weight, $c_1$ and $c_2$ are learning factors, $r_1$ and $r_2$ are random numbers between $[0, 1]$, $pBest$ i,γ and $pBest$ i,C are the kernel parameter and penalty factor corresponding to the optimal fitness value in the history of the $i$-th particle itself, $gBest$ γ and $gBest$ C are the global optimal kernel parameter and penalty factor corresponding to the optimal fitness value in the historical iterations of all particles. and are the kernel parameter and penalty factor of the $i$-th particle at the $k$-th iteration, $m$ is the number of validation samples. is the result of predicting the $j$-th validation sample using the SVM parameter combination ($\gamma$ i , $C$ i ) represented by the $i$-th particle, $y$ j is the actual fracture position label of the $j$-th validation sample;

[0074]

[0075] where $r$ is the weight vector, $b$ is the bias term, $\xi$ i is the slack variable used to handle the linearly inseparable case, and $C$ is the penalty factor that controls the penalty degree for misclassified samples.

[0076] By solving the above optimization problem using the Lagrange multiplier method, the dual problem can be obtained:

[0077] SVM model:

[0078]

[0079] $0\leq\alpha$ i $\leq C, i = 1, \ldots, n$;

[0080] $K(x$ i , $x$ j ) = $\exp(-\gamma\lVert x$ i - $x$ j $\rVert$ 2 );

[0081]

[0082] where $x$ i , $x$ j , $x$ are the input sample data, which are gravity gradient, magnetic gradient, and depth data respectively; $\gamma$ is the kernel parameter (radial basis function RBF) that controls the width of the kernel function and affects the fitting and generalization ability of the model; $\alpha$ i and $\alpha$j are all Lagrange multipliers, obtained by solving the dual problem; y i is the class label of the sample, and the fracture position can be encoded in the fracture position prediction; b is the bias term;

[0083] Initialize the particle swarm and randomly generate the initial positions of a group of particles and velocities

[0084] For each particle i, use its SVM parameter combination to train the SVM model, solve for the Lagrange multipliers α i 、α j and the bias term b, and calculate the fitness value Fitness i ;

[0085] Update the pBest i,γ 、pBest i,C and gBest γ 、gBest C ;

[0086] Update the velocities and positions of the particles according to the particle velocity and position update formulas;

[0087] Repeat the iterative update until the maximum number of iterations is reached, obtain the final fracture position prediction model, and perform fracture position prediction.

[0088] In a specific embodiment, the gravity gradient and magnetic gradient are ▽G, ▽M;

[0089]

[0090] wherein, is the gradient of the gravity field G in the x, y, and z directions; is the gradient of the magnetic field M in the x, y, and z directions; the airborne gravity and magnetic data include the data of the gravity field and the magnetic field.

[0091] In a specific embodiment, obtaining the fracture type data based on fuzzy clustering specifically includes:

[0092] Determine the number of clusters c of the fracture type and select the fuzzy factor m, m ∈ (1, +∞), and a common value is 2; set the iteration stop threshold ε, such as ε = 10 -6 , used to judge whether the iteration converges; construct the membership matrix U of n × c = [u ij , u ij is the membership degree of the sample q i belonging to the j-th class, the sample q i includes the gravity gradient, magnetic field gradient, and depth data;

[0093] Calculate the clustering center v of each class according to the membership matrix U j , and the calculation formula is:

[0094]

[0095] where v j is the clustering center of the j-th class, which is obtained by weighted averaging the sample data, and the weight is determined by the m-th power of the membership degree;

[0096] Use the current clustering center v j to update the membership matrix U, and the update formula is:

[0097]

[0098] where ||q i -v j || represents the distance between the sample x i and the clustering center v j ;

[0099] Define the objective function J of fuzzy clustering, and the formula is:

[0100]

[0101] Calculate the objective function value J of this iteration (t) and the objective function value J of the previous iteration (t-1) . If J (t) -J (t -1) ≤ε, it is considered that the iteration converges and the iteration stops; otherwise, continue the iteration, continuously update the clustering center and the membership matrix until the termination condition is satisfied; when the iteration ends, determine the fracture type to which each sample belongs according to the final membership matrix U. For each sample q i , the fracture type is the category with the largest membership degree, that is In this way, all samples are classified into the corresponding fracture types, and the acquisition of fracture type data based on fuzzy clustering is completed.

[0102] In a specific embodiment, the loss function specifically includes:

[0103]

[0104] where Loss is the loss function value, n represents the number of data samples, and are the measured gravity data and measured magnetic data of the i-th sample respectively, and They are the historical gravity data and historical magnetic data of the i-th sample respectively, the data calculated based on the established mapping model, and the historical gravity and magnetic data include historical gravity data and historical magnetic data.

[0105] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0106] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fracture identification method based on airborne gravity and magnetic data, characterized in that, Including: Obtain historical airborne gravity and magnetic data, process the historical airborne gravity and magnetic data through Euler deconvolution, and obtain historical field source position and depth data; based on a neural network algorithm, establish a mapping model between the historical airborne gravity and magnetic data, fracture characteristic parameters, and field source position and depth data; Based on the historical field source position and depth data in Euler deconvolution processing, use a preset fracture position prediction model to predict the fracture position and obtain fracture position data; at the same time, obtain fracture type data based on fuzzy clustering; Based on the area to be measured, obtain airborne gravity and magnetic data to be measured; confirm the initial estimated fracture parameters of the area to be measured, calculate the estimated airborne gravity and magnetic data based on the mapping model, calculate the loss function between the airborne gravity and magnetic data to be measured and the estimated airborne gravity and magnetic data, and determine whether the loss function value exceeds a preset accuracy threshold. If it does not exceed the preset accuracy threshold, optimize the fracture characteristic parameters through particle swarm intelligence; if it is satisfied, end the iteration and output the optimal fracture characteristic parameters; Output the fracture position, fracture type, and fracture parameters output in the final iteration step.

2. The fracture identification method based on airborne gravity and magnetic data according to claim 1, wherein The obtaining of the historical airborne gravity and magnetic data is obtained through gravity and magnetic measurements at two different measurement heights, namely the first historical airborne gravity and magnetic data Q1 and the second historical airborne gravity and magnetic data Q2.

3. The fracture recognition method based on airborne gravity and magnetic data according to claim 1, characterized in that The fracture characteristic parameters include fracture length, fracture width, fracture dip angle, and fracture type.

4. A fracture identification method based on airborne gravity and magnetic data according to claim 3, characterized in that The mapping model specifically includes: Among them, l is the fracture length, w is the fracture width, d is the fracture depth, θ is the fracture dip angle, F c is the fracture type, t1 and t2 are the measurement heights, G1 and G2 are the historical gravity and magnetic data, and (x, y) is the field source position.

5. The fracture identification method based on airborne gravity and magnetic data according to claim 3, wherein, The fracture position prediction specifically includes: Each particle $i$ represents a set of parameter combinations $p$ of SVM i $=$ ($\gamma$ i , $C$ i ), where $\gamma$ i is the kernel parameter and $C$ i is the penalty factor; where k is the number of iterations, is the velocity vector of the i-th particle at the (k + 1)-th iteration, corresponding to the velocity components of the kernel parameter γ and the penalty factor C respectively, ω is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers between [0, 1], pBest i,γ and pBest i,C are the kernel parameter and penalty factor corresponding to the i-th particle when it reaches the optimal fitness value in its own history, gBest γ and gBest C are the global optimal kernel parameter and penalty factor corresponding to all particles when they reach the optimal fitness value in historical iterations, and are the kernel parameter and penalty factor of the i-th particle at the k-th iteration, m is the number of validation samples, is the result of predicting the j-th validation sample using the SVM parameter combination (γ i , C i ) represented by the i-th particle, y j is the actual fracture position label of the j-th validation sample; SVM model: 0 ≤ α i ≤ C, i = 1, …, n; K(x i , x j ) = exp(-γ‖x i - x j ‖ 2 ); where x i 、x j 、x are the input sample data, which are gravity gradient, magnetic gradient and depth data respectively; γ is the kernel parameter, and α i and α j are both Lagrange multipliers, y i is the class label of the sample, and b is the bias term; Initialize the particle swarm and randomly generate the initial positions of a group of particles and velocities For each particle i, use its SVM parameter combination to train the SVM model, solve the Lagrange multipliers α i , α j and the bias term b, and calculate the fitness value Fitness i ; Update the pBest of each particle i,γ pBest i,C and gBest γ gBest C ; Update the velocity and position of the particle according to the particle velocity and position update formula; Repeat the iterative update until the maximum number of iterations is reached to obtain the final fracture position prediction model and perform fracture position prediction.

6. A fracture identification method based on airborne gravity and magnetic data according to claim 1, characterized in that The obtaining of the fracture type data based on fuzzy clustering specifically includes: Determine the number of clusters \(c\) for the fracture type and select the fuzzy factor \(m\), where \(m\in(1, +\infty)\), and set the iteration stop threshold \(\varepsilon\); construct the membership matrix \(U = [u_{ij}]\) of size \(n\times c\), where \(u_{ij}\) is the membership degree of sample \(q_i\) belonging to the \(j\)-th class, and \(0\leq u_{ij}\leq1\). ij _{ij} ij _{i} i belonging to the \(j\)-th class, _{ij} ij \leq1; Calculate the clustering center v of each class according to the membership matrix U j , and the calculation formula is: where, v j is the clustering center of the j-th class, which is obtained by weighted averaging the sample data, and the weights are determined by the m-th power of the membership degree; Using the current cluster center v j Update the membership matrix U, and the update formula is: Among them, ||q i -v j || represents the distance between the sample x i and the clustering center v j ; Define the objective function J of fuzzy clustering, and the formula is: Calculate the objective function value J of this iteration (t) and the objective function value J of the previous iteration (t-1) . If J (t) - J (t-1) ≤ ε, it is considered that the iteration converges and the iteration stops; otherwise, continue the iteration, continuously update the cluster centers and membership matrix until the termination condition is satisfied; after the iteration ends, determine the fracture type to which each sample belongs according to the final membership matrix U. For each sample q i , the fracture type is the category with the largest membership degree.

7. A fracture identification method based on airborne gravity and magnetic data according to claim 1, characterized in that, The loss function specifically includes: where Loss is the loss function value, and n represents the number of data samples. and are the measured gravity data and measured magnetic data of the i-th sample respectively. and are the historical gravity data and historical magnetic data of the i-th sample respectively, which are data calculated based on the established mapping model. The historical gravity and magnetic data include historical gravity data and historical magnetic data.

Images