A data analysis method for variable-scale identification of natural fractures

By calculating the evaluation of reference indicators and data point distances of the feature vectors, the selection of principal components is optimized, and the limitations of PCA algorithm in the analysis of natural crack variable scales is solved, the analysis efficiency and accuracy are improved, and the main cracks are revealed Change mode.

CN118708956BActive Publication Date: 2024-11-12BEIJING KAIBORUI PETROLEUM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411194603.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-12
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

When using PCA algorithm to perform natural crack variable-scale analysis, there are limitations in the selection of principal components, resulting in increased computational complexity, loss of information or risk of overfitting, affecting analysis efficiency and accuracy.

Method used

By obtaining the distance between the diagonal matrix of the feature data matrix and the data points, calculate the optimal principal component of the feature vector to evaluate the reference index, and accumulate the feature vectors in sequence until the preset judgment index threshold is reached to determine the appropriate main Ingredient quantity.

Benefits of technology

The traditional method is avoided to rely solely on the principal component sorting that only depends on the size of the eigenvalue, retain important change information, improve the efficiency and accuracy of variable-scale analysis of natural cracks, and reveal the main changes modes and influencing factors of the cracks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118708956B_ABST
    Figure CN118708956B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of data processing, and in particular to a data analysis method for natural fracture variable scale identification. The method obtains natural fracture data at different scales, extracts N feature corresponding data from all natural fracture data to form a feature data matrix, performs principal component analysis on the feature data matrix, and obtains a feature vector sequence; for any feature vector in the feature vector sequence, obtains an evaluation reference index of whether the feature vector belongs to an optimal principal component; selects M feature vectors in the feature vector sequence by sequential accumulation to obtain the number of selected principal components; screens all feature vectors to obtain effective principal components according to the evaluation reference index of whether each feature vector belongs to an optimal principal component and the number of selected principal components, identifies natural fractures according to each effective principal component, can retain important change information of the fracture, and achieves the effect of effectively analyzing and identifying the variable scale characteristics of the fracture.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for analyzing data of natural fracture variable scale identification. Background Art

[0002] The formation and evolution of natural fractures are closely related to geological structures, groundwater, stress fields and other factors. Through variable-scale identification, we can more effectively understand the distribution characteristics and evolution laws of fractures at different scales, which is helpful to assess the risk of geological disasters, such as ground collapse and karst collapse, so as to provide early warning and take preventive measures. In addition, in the field of resource exploration and development, fractures have a significant impact on reservoir properties and fluid migration. By identifying the spatial distribution and connectivity of fractures at different scales, we can optimize resource exploration strategies and improve resource development and utilization. Therefore, variable-scale identification and analysis of natural fractures has a wide range of application significance in geological engineering, petroleum exploration, mining and other fields.

[0003] The PCA principal component analysis in machine learning can be well used for variable scale analysis and identification of natural fractures. Specifically, by collecting relevant characteristic data of fractures, such as obtaining size, shape, position and other geometric parameters through laser scanning or photogrammetry, and standardizing the collected relevant characteristic data before inputting it into the PCA algorithm, the principal components of data at multiple scales are extracted, thereby helping to identify the main change patterns of fractures. The PCA algorithm maximizes the data variance, extracts the main features, and reduces the dimensionality to simplify complex data, thereby identifying and analyzing the main change patterns of fractures. Therefore, the PCA algorithm is a powerful tool in fracture scale analysis, which can reveal important changes in fractures in different characteristics and provide effective scale identification.

[0004] However, the selection of principal components in the variable-scale analysis of cracks by the PCA algorithm is a key and important step. In the traditional way, the principal component in the PCA algorithm is the eigenvector with the largest eigenvalue, that is, the principal component corresponding to the maximum variance is retained; but for the current scenario, the selection of principal components only relies on the traditional evaluation standard of the largest eigenvalue, which has high limitations, and the selection of principal components is difficult to set. The selection of more principal components will lead to a significant increase in computational complexity, affecting the efficiency of scale analysis, and there is a risk of overfitting or making the PCA algorithm too affected by noise; and the selection of fewer principal components will lead to information loss, reduced explanatory power, and affect the performance and accuracy of the PCA algorithm.

[0005] Therefore, how to select appropriate principal components when using the PCA algorithm to perform variable-scale analysis of fractures in order to improve the efficiency of variable-scale analysis and identification of natural fractures has become an urgent problem to be solved. Summary of the invention

[0006] In view of this, an embodiment of the present invention provides a natural fracture variable scale identification data analysis method to solve the problem of how to select appropriate principal components when using the PCA algorithm to perform variable scale analysis of fractures, so as to improve the efficiency of variable scale analysis and identification of natural fractures.

[0007] An embodiment of the present invention provides a natural fracture variable scale identification data analysis method, the method comprising the following steps:

[0008] Acquire natural fracture data at different scales, extract N feature corresponding data from all natural fracture data to form a feature data matrix, wherein each column of the feature data matrix represents a feature, and each row represents a natural fracture data, perform principal component analysis on the feature data matrix using a PCA algorithm, and obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalues;

[0009] For any eigenvector in the eigenvector sequence, obtain the diagonal matrix of the eigendata matrix, and obtain an evaluation reference index of whether the eigenvector belongs to the optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix;

[0010] Selecting M feature vectors from the feature vector sequence in order and accumulating them, and obtaining the preferred judgment index of the M feature vectors according to the feature values ​​and the evaluation reference index of the M feature vectors, until the preferred judgment index is greater than or equal to the preset judgment index threshold, and then taking the corresponding M value as the number of principal component selection;

[0011] According to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected, effective principal components are screened from all eigenvectors, and natural fractures are identified based on each effective principal component.

[0012] Preferably, obtaining an evaluation reference index of whether the eigenvector belongs to an optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix includes:

[0013] According to the distance and number between the data points on the eigenvector, the data distribution smoothness index of the eigenvector is obtained, the eigenvalue of the diagonal matrix is ​​obtained, and according to the eigenvalue of the diagonal matrix and the data distribution smoothness index of the eigenvector, the evaluation reference index of whether the eigenvector belongs to the optimal principal component is obtained.

[0014] Preferably, obtaining the data distribution stability index of the feature vector according to the distance and number between the data points on the feature vector includes:

[0015] Obtaining the distance between the first data point and the last data point on the feature vector, and normalizing the distance to obtain a corresponding normalized value;

[0016] Respectively obtaining the distance between every two adjacent data points on the feature vector, obtaining the sum of the distances, the maximum distance, and the minimum distance, obtaining a first ratio of the maximum distance to the sum of the distances, obtaining a second ratio of the minimum distance to the sum of the distances, and obtaining an absolute value of a ratio difference between the first ratio and the second ratio;

[0017] Using an exponential function with a natural constant as the base, the absolute value of the ratio difference is negatively mapped to obtain a corresponding mapping value, and a weighted sum is performed on the mapping value and the normalized value, and the corresponding result is used as the data distribution stability indicator of the eigenvector.

[0018] Preferably, obtaining the evaluation reference index of whether the eigenvector belongs to the optimal principal component according to the eigenvalue of the diagonal matrix and the data distribution stability index of the eigenvector includes:

[0019] The mean between the eigenvalues ​​of the diagonal matrix and the data distribution stability index of the eigenvector is calculated as a reference index for evaluating whether the eigenvector belongs to the optimal principal component.

[0020] Preferably, obtaining the optimal determination index of the M feature vectors according to the feature values ​​of the M feature vectors and the evaluation reference index includes:

[0021] According to the eigenvalues ​​of the M eigenvectors, counting the number of eigenvectors whose eigenvalues ​​are greater than a preset value in the M eigenvectors, and calculating the quantity ratio between the number of eigenvectors and the M value;

[0022] According to the evaluation reference index of each of the feature vectors, a sum of the evaluation reference indexes is obtained, and the index ratio of the evaluation reference index of each feature vector in the M feature vectors to the sum of the evaluation reference indexes is respectively calculated to obtain an index ratio sequence;

[0023] The difference between every two adjacent indicator ratios in the indicator ratio sequence is calculated to obtain a difference cumulative value, and the mean between the reciprocal of the difference cumulative value and the quantity ratio is calculated as a preferred determination indicator for the M feature vectors.

[0024] Preferably, the step of screening out effective principal components from all eigenvectors based on the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected comprises:

[0025] According to the order of the evaluation reference indexes from large to small, k eigenvectors equal to the number of the principal components are selected from all the eigenvectors as effective principal components.

[0026] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0027] The present invention obtains natural fracture data at different scales, extracts N feature corresponding data from all natural fracture data to form a feature data matrix, wherein each column of the feature data matrix represents a feature, and each row represents a natural fracture data, and uses a PCA algorithm to perform principal component analysis on the feature data matrix to obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalues; for any eigenvector in the feature vector sequence, obtains a diagonal matrix of the feature data matrix, and obtains an evaluation reference index of the eigenvector belonging to an optimal principal component according to the distance between data points on the eigenvector and the diagonal matrix; selects M eigenvectors from the eigenvector sequence by sequential accumulation, and obtains preferred judgment indexes of the M eigenvectors according to the eigenvalues ​​and evaluation reference indexes of the M eigenvectors, until the preferred judgment index is greater than or equal to a preset judgment index threshold, and then uses the corresponding M value as the number of principal component selections; according to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal component selections, screens out effective principal components from all eigenvectors, and identifies natural fractures according to each effective principal component. Among them, by analyzing the evaluation reference indicators of the optimal principal component and the required number of principal components for each eigenvector after principal component analysis, it is possible to avoid sorting the principal components only according to the size of the eigenvalues ​​in the traditional way, and retain the important change information of the cracks without affecting the efficiency of the computational complexity, so as to achieve the effect of effectively analyzing and identifying the variable-scale characteristics of the cracks and revealing the main change patterns and influencing factors of the cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0029] Figure 1 This is a method flow chart of a method for analyzing data of natural fracture variable scale identification provided by the first embodiment of the present invention;

[0030] Figure 2 is a schematic diagram of data distribution with relatively uniform intervals provided by an embodiment of the present invention;

[0031] Figure 3 It is a schematic diagram of a data distribution including discrete points provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0033] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0034] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0035] See also Figure 1 , is a method flow chart of a method for analyzing data of natural fracture variable scale identification provided by the first embodiment of the present invention, such as Figure 1 As shown, the method may include:

[0036] Step S101, obtain natural fracture data at different scales, extract N feature corresponding data from all natural fracture data to form a feature data matrix, wherein each column of the feature data matrix represents a feature, and each row represents a natural fracture data, and use the PCA algorithm to perform principal component analysis on the feature data matrix to obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalue.

[0037] Natural fracture data at different scales refer to data of various data types about natural fractures, including but not limited to the geometric shape, distribution, length, width, depth and other characteristics of fractures, and these natural fracture data can come from geological surveys, remote sensing images, drilling data, etc. After obtaining the natural fracture data, these data are preprocessed, specifically: processing missing values, outliers, and normalizing the data, for example, standardizing the length data and width data of natural fractures to make their mean 0 and variance 1 to ensure the effectiveness of the subsequent PCA algorithm.

[0038] After obtaining the natural fracture data at different scales, N feature corresponding data are extracted from all the natural fracture data to form a feature data matrix, where N is greater than 2, each column of the feature data matrix represents a feature, and each row represents a natural fracture data. Afterwards, the PCA algorithm is used to perform principal component analysis on the feature data matrix to obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalue.

[0039] The specific steps of the PCA algorithm are: decentralized the feature data matrix to obtain a new matrix, that is, zero-mean each column, calculate the covariance matrix of the new matrix, perform eigendecomposition on the covariance matrix, find the eigenvalues ​​of the covariance matrix and the corresponding eigenvectors, and arrange the eigenvectors in descending order from left to right according to the corresponding eigenvalues ​​into a matrix (eigenvector sequence). It is worth noting that the PCA algorithm belongs to the prior art and will not be described in detail here.

[0040] At this point, the eigenvector sequence is obtained after principal component analysis of all natural fracture data.

[0041] Step S102, for any eigenvector in the eigenvector sequence, obtain the diagonal matrix of the characteristic data matrix, and obtain the evaluation reference index of whether the eigenvector belongs to the optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix.

[0042] The core idea of ​​the PCA algorithm is to find the direction (principal component) that maximizes the data variance. These directions are the most important directions of change in the data. For crack data, the PCA algorithm can find the main change characteristics of cracks at different scales and shapes, and use the principal components to reflect the most significant changes in the data, thereby helping to identify the main crack patterns. In crack analysis and identification, principal component analysis is more helpful in extracting the main shape and structural characteristics of cracks. The principal components describe the changes in cracks in these characteristic directions, and thus can reveal the main change patterns of cracks.

[0043] However, the traditional principal component analysis method only sorts the principal components according to the size of the eigenvalue, which has certain limitations and the number of principal components is usually difficult to select. The size of the eigenvalue is the variance of the data mapping distribution in the corresponding direction. The traditional principal component analysis believes that the larger the variance, the more discrete the data mapping distribution is, and the more valid data points there are. Figure 2 , which is a schematic diagram of a relatively evenly spaced data distribution. Figure 2 It represents the distribution of valid data points corresponding to the optimal principal component. The smaller the variance, the more concentrated the distribution of the data mapping, and the more invalid data points. However, under the principal component analysis, there are inevitably individual discrete data points that affect the variance calculation. Figure 3 , which is a schematic diagram of a data distribution formed by discrete points, Figure 3This reflects that although it is the optimal principal component through variance calculation, it is not actually Figure 3 In this case, the eigenvectors whose distribution is not discrete in theory will be considered as the principal components with better effects. Therefore, in the embodiment of the present invention, the distribution analysis of the data points on each eigenvector in the eigenvector sequence is performed to obtain the evaluation reference index for judging whether the eigenvector is the optimal principal component.

[0044] Specifically, for any eigenvector in the eigenvector sequence, the diagonal matrix of the eigendata matrix is ​​obtained, and based on the distance between the data points on the eigenvector and the diagonal matrix, the evaluation reference index of whether the eigenvector belongs to the optimal principal component is obtained, wherein the process of obtaining the evaluation reference index is:

[0045] (1) Obtaining a data distribution stability index of the feature vector according to the distance and number between the data points on the feature vector.

[0046] Specifically, the distance between the first data point and the last data point on the feature vector is obtained, and the distance is normalized to obtain a corresponding normalized value;

[0047] Respectively obtaining the distance between every two adjacent data points on the feature vector, obtaining the sum of the distances, the maximum distance, and the minimum distance, obtaining a first ratio of the maximum distance to the sum of the distances, obtaining a second ratio of the minimum distance to the sum of the distances, and obtaining an absolute value of a ratio difference between the first ratio and the second ratio;

[0048] Using an exponential function with a natural constant as the base, the absolute value of the ratio difference is negatively mapped to obtain a corresponding mapping value, and a weighted sum is performed on the mapping value and the normalized value, and the corresponding result is used as the data distribution stability indicator of the eigenvector.

[0049] In one embodiment, since there are multiple data points in the direction corresponding to each eigenvector during the principal component analysis process, the distance between the first data point and the last data point is calculated based on the data points on the eigenvector and recorded as the discrete distance , and calculate the distance between every two adjacent data points on the feature vector to get the mean distance , Maximum distance and minimum distance , and then according to the discrete distance corresponding to the eigenvector , distance mean , Maximum distance and minimum distance , obtain the data distribution stability index of the eigenvector, then the calculation expression of the data distribution stability index is:

[0050]

[0051] in, It represents the data distribution stability index of the eigenvector, Norm() represents the normalization function, and exp() represents the exponential function with a natural constant as the base. Indicates the difference calculation symbol, Indicates the maximum distance, represents the minimum distance, represents the mean distance, represents the first weight, Represents the second weight.

[0052] It should be noted that The larger the value is, the more discrete the overall distribution of the data points on the eigenvector is, the more valid data points on the corresponding eigenvector are, and the larger the data distribution stability index of the corresponding eigenvector is; The larger the value, the greater the difference between the maximum and minimum interval ratios between data points, and the more unstable the interval distribution between the corresponding data points, the smaller the data distribution stability index of the eigenvector, the greater the overall distribution dispersion, and the more stable the interval distribution between data points, the larger the data distribution stability index of the eigenvector, and the greater the possibility that it belongs to the optimal principal component.

[0053] Preferably, in the embodiment of the present invention, the first weight and the second weight are empirical values, and a reference weight value is given: There is no restriction here and it can be set according to the implementation scenario.

[0054] (2) Obtaining the eigenvalues ​​of the diagonal matrix, and obtaining an evaluation reference index for whether the eigenvector belongs to the optimal principal component according to the eigenvalues ​​of the diagonal matrix and the data distribution stability index of the eigenvector.

[0055] Specifically, the mean between the eigenvalues ​​of the diagonal matrix and the data distribution stability index of the eigenvector is calculated as a reference index for evaluating whether the eigenvector belongs to the optimal principal component.

[0056] In one embodiment, the diagonal elements of the diagonal matrix are its eigenvalues. Therefore, after obtaining the data distribution stability index of the eigenvector, the evaluation reference index of whether the eigenvector belongs to the optimal principal component is obtained according to the eigenvalue of the diagonal matrix and the data distribution stability index of the eigenvector. The calculation expression of the evaluation reference index is:

[0057]

[0058] in, Indicates that the eigenvector belongs to the optimal principal component evaluation reference index, represents a diagonal matrix, is the eigenvalue of a diagonal matrix, Indicator of the stationary distribution of the data representing the eigenvector.

[0059] It should be noted that the evaluation reference index is used to evaluate whether the eigenvector is an optimal principal component. The larger the evaluation reference index, the greater the possibility that the corresponding eigenvector becomes the optimal principal component.

[0060] At this point, it is possible to obtain an evaluation reference index for whether each eigenvector in the eigenvector sequence belongs to the optimal principal component.

[0061] Step S103, sequentially and cumulatively select M feature vectors from the feature vector sequence, and obtain the preferred judgment index of the M feature vectors according to the feature values ​​and evaluation reference index of the M feature vectors, until the preferred judgment index is greater than or equal to the preset judgment index threshold, and then use the corresponding M value as the number of principal component selections.

[0062] After obtaining the evaluation reference index of the optimal principal component for each eigenvector in the eigenvector sequence, the number of principal components required for identifying and analyzing natural fractures can be analyzed to ensure that multiple effective principal components can be selected to better carry out subsequent natural fracture identification and analysis. Therefore, in an embodiment of the present invention, M eigenvectors are selected in sequence in the eigenvector sequence, and the preferred judgment index of the M eigenvectors is obtained according to the eigenvalues ​​and evaluation reference index of the M eigenvectors, until the preferred judgment index is greater than or equal to the preset judgment index threshold, and the corresponding M value is used as the number of principal component selections. It should be noted that the value of M starts from 1 until all eigenvectors in the eigenvector sequence are selected, and after each selection (that is, M takes a different value), the preferred judgment index of the selected M eigenvectors is analyzed to adaptively find the optimal M value in the scenario.

[0063] Among them, taking any value of M as an example, according to the eigenvalues ​​of the M eigenvectors and the evaluation reference index, the optimal judgment index of the M eigenvectors is obtained, including:

[0064] According to the eigenvalues ​​of the M eigenvectors, counting the number of eigenvectors whose eigenvalues ​​are greater than a preset value in the M eigenvectors, and calculating the quantity ratio between the number of eigenvectors and the M value;

[0065] According to the evaluation reference index of each of the feature vectors, a sum of the evaluation reference indexes is obtained, and the index ratio of the evaluation reference index of each feature vector in the M feature vectors to the sum of the evaluation reference indexes is respectively calculated to obtain an index ratio sequence;

[0066] The absolute value of the difference between every two adjacent indicator ratios in the indicator ratio sequence is calculated to obtain a difference cumulative value, and the mean between the reciprocal of the difference cumulative value and the quantity ratio is calculated as the preferred determination indicator of the M feature vectors.

[0067] In one embodiment, the calculation expression of the optimal determination index of the M feature vectors is:

[0068]

[0069] in, represents the optimal judgment index of M feature vectors, represents the number of eigenvectors with eigenvalues ​​greater than 1 among the M eigenvectors, and M represents the number of selected eigenvectors. represents the evaluation reference index of the j-th eigenvector, Represents the sum of the evaluation reference indicators of all feature vectors in the feature vector sequence, Represents the evaluation reference indicator of the j-1th eigenvector.

[0070] It should be noted that The larger the value of is, the more meaningful feature vectors there are in the M feature vectors, and the larger the optimal judgment index of the corresponding M feature vectors is; Indicates the proportion of the evaluation reference index of each feature vector in the selected M feature vectors, It represents the difference in proportions between two adjacent eigenvectors. The smaller the difference in proportions, the smaller the difference in variances of the corresponding eigenvectors, the more consistent with the effective principal components, and the larger the optimal judgment index of the corresponding selected M eigenvectors.

[0071] After obtaining the optimal judgment index of M eigenvectors, obtain the judgment index threshold value of 0.8. If the optimal judgment index of M eigenvectors is greater than or equal to 0.8, it is considered that the M eigenvectors selected at this time just meet the effective principal component, and the value of M at this time is used as the number of principal component selection k; on the contrary, if the optimal judgment index of M eigenvectors is less than 0.8, continue to accumulate and select eigenvectors until the optimal judgment index of the cumulatively selected M eigenvectors is greater than or equal to 0.8. It is worth noting that the cumulative selection means starting from the first eigenvector in the eigenvector sequence, that is, M=1, and calculating the optimal judgment index when M=1. If the optimal judgment index is less than 0.8, select the first eigenvector and the second eigenvector, that is, M=2, and calculate the optimal judgment index when M=2. If the optimal judgment index is less than 0.8, continue to select the first eigenvector, the second eigenvector and the third eigenvector, that is, M=3, and calculate the optimal judgment index when M=3. If the optimal judgment index is greater than or equal to 0.8, the number of principal component selection is 3.

[0072] At this point, the number of principal components required for the eigenvector sequence is obtained.

[0073] Step S104, based on the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected, effective principal components are screened from all eigenvectors, and natural fractures are identified based on each effective principal component.

[0074] The principal components have the following characteristics: the first principal component (PC1) usually explains the largest variance in the data, that is, the most significant direction of change. In crack analysis, this may correspond to the main characteristics of the crack, such as the main direction or length change of the crack; the next M principal components (PC2, ..., PCM) explain the second largest variance and provide another important direction of change orthogonal to PC1. In crack analysis, this may correspond to the secondary characteristics of the crack, such as width or other morphological characteristics. Therefore, in the embodiment of the present invention, according to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected, the effective principal components are screened from all eigenvectors, including:

[0075] According to the order of the evaluation reference indexes from large to small, k eigenvectors equal to the number of the principal components are selected from all the eigenvectors as effective principal components.

[0076] It should be noted that in the eigenvector sequence, the eigenvector corresponding to the largest evaluation reference index is the first principal component, the eigenvector corresponding to the second largest evaluation reference index is the second principal component, and so on, to obtain the kth principal component, and these k principal components are taken as effective principal components.

[0077] After obtaining the effective principal components, the data points corresponding to the effective principal components are all effective data points, and then the natural fractures are identified and analyzed based on the data corresponding to these effective data points, such as analyzing the load coefficient of each effective principal component, where the load coefficient describes the coefficient of the relationship between the original feature and the principal component, which can reveal the importance or weight of each original feature on each principal component, and understand the relationship between each effective principal component and the original feature. A larger load coefficient indicates that a specific feature contributes more to the effective principal component, or by observing the characteristics of the effective principal components, the fracture characteristics of different scales can be identified. For example, some effective principal components may represent the long-scale characteristics of natural fractures, while other effective principal components may represent short-scale characteristics.

[0078] It is worth noting that the focus of the present invention is on how to select effective principal components. The identification of natural fractures based on each effective principal component is not the focus of the present invention and will not be described in detail here.

[0079] In summary, the embodiment of the present invention obtains natural fracture data at different scales, extracts N feature corresponding data from all natural fracture data to form a feature data matrix, wherein each column of the feature data matrix represents a feature, and each row represents a natural fracture data, and uses the PCA algorithm to perform principal component analysis on the feature data matrix to obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalue; for any eigenvector in the feature vector sequence, obtain the diagonal matrix of the feature data matrix, and obtain the evaluation reference index of the eigenvector belonging to the optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix; select M eigenvectors in the eigenvector sequence by sequential accumulation, and obtain the preferred judgment index of the M eigenvectors according to the eigenvalues ​​and the evaluation reference index of the M eigenvectors, until the preferred judgment index is greater than or equal to the preset judgment index threshold, and the corresponding M value is used as the number of principal component selections; according to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal component selections, screen out effective principal components from all eigenvectors, and identify natural fractures according to each effective principal component. Among them, by analyzing the evaluation reference indicators of the optimal principal component and the required number of principal components for each eigenvector after principal component analysis, it is possible to avoid sorting the principal components only according to the size of the eigenvalues ​​in the traditional way, and retain the important change information of the cracks without affecting the efficiency of the computational complexity, so as to achieve the effect of effectively analyzing and identifying the variable-scale characteristics of the cracks and revealing the main change patterns and influencing factors of the cracks.

[0080] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A natural fracture variable scale identification data analysis method, characterized in that: The method comprises: Acquire natural fracture data at different scales, extract N feature corresponding data from all natural fracture data to form a feature data matrix, wherein each column of the feature data matrix represents a feature, and each row represents a natural fracture data, perform principal component analysis on the feature data matrix using a PCA algorithm, and obtain a feature vector sequence after sorting the feature vectors from large to small according to the eigenvalues; For any eigenvector in the eigenvector sequence, obtain the diagonal matrix of the eigendata matrix, and obtain an evaluation reference index of whether the eigenvector belongs to the optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix; Selecting M feature vectors from the feature vector sequence in order and accumulating them, and obtaining the preferred judgment index of the M feature vectors according to the feature values ​​and the evaluation reference index of the M feature vectors, until the preferred judgment index is greater than or equal to the preset judgment index threshold, and then taking the corresponding M value as the number of principal component selection; According to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected, effective principal components are screened from all eigenvectors, and natural fractures are identified based on each effective principal component.

2. A natural fracture variable scale identification data analysis method according to claim 1, characterized in that: The step of obtaining an evaluation reference index of whether the eigenvector belongs to an optimal principal component according to the distance between the data points on the eigenvector and the diagonal matrix includes: According to the distance and number between the data points on the eigenvector, the data distribution smoothness index of the eigenvector is obtained, the eigenvalue of the diagonal matrix is ​​obtained, and according to the eigenvalue of the diagonal matrix and the data distribution smoothness index of the eigenvector, the evaluation reference index of whether the eigenvector belongs to the optimal principal component is obtained.

3. A natural fracture variable scale identification data analysis method according to claim 2, characterized in that: The step of obtaining a data distribution stability index of the feature vector according to the distance and quantity between the data points on the feature vector comprises: Obtaining the distance between the first data point and the last data point on the feature vector, and normalizing the distance to obtain a corresponding normalized value; Respectively obtaining the distance between every two adjacent data points on the feature vector, obtaining the sum of the distances, the maximum distance, and the minimum distance, obtaining a first ratio of the maximum distance to the sum of the distances, obtaining a second ratio of the minimum distance to the sum of the distances, and obtaining an absolute value of a ratio difference between the first ratio and the second ratio; Using an exponential function with a natural constant as the base, the absolute value of the ratio difference is negatively mapped to obtain a corresponding mapping value, and a weighted sum is performed on the mapping value and the normalized value, and the corresponding result is used as the data distribution stability indicator of the eigenvector.

4. A natural fracture variable scale identification data analysis method according to claim 2, characterized in that: The step of obtaining an evaluation reference index of whether the eigenvector belongs to an optimal principal component according to the eigenvalue of the diagonal matrix and the data distribution stability index of the eigenvector comprises: The mean between the eigenvalues ​​of the diagonal matrix and the data distribution stability index of the eigenvector is calculated as a reference index for evaluating whether the eigenvector belongs to the optimal principal component.

5. The method for analyzing data of natural fracture variable scale identification according to claim 1, characterized in that: The obtaining, according to the eigenvalues ​​of the M eigenvectors and the evaluation reference index, the preferred determination index of the M eigenvectors comprises: According to the eigenvalues ​​of the M eigenvectors, counting the number of eigenvectors whose eigenvalues ​​are greater than a preset value in the M eigenvectors, and calculating the quantity ratio between the number of eigenvectors and the M value; According to the evaluation reference index of each of the feature vectors, a sum of the evaluation reference indexes is obtained, and the index ratio of the evaluation reference index of each feature vector in the M feature vectors to the sum of the evaluation reference indexes is respectively calculated to obtain an index ratio sequence; The difference between every two adjacent indicator ratios in the indicator ratio sequence is calculated to obtain a difference cumulative value, and the mean between the reciprocal of the difference cumulative value and the quantity ratio is calculated as a preferred determination indicator for the M feature vectors.

6. A natural fracture variable scale identification data analysis method according to claim 1, characterized in that: The effective principal components are screened from all the eigenvectors according to the evaluation reference index of each eigenvector belonging to the optimal principal component and the number of principal components selected, including: According to the order of the evaluation reference indexes from large to small, k eigenvectors equal to the number of the principal components are selected from all the eigenvectors as effective principal components.