A method and device for intelligent quality control of seismic data based on machine learning

Abstract

The present invention discloses a method and device for intelligent quality control of seismic data based on machine learning, which belongs to the technical field of seismic data processing. The method comprises the following steps: S1. Calculating single-channel attributes with correlation and sensitivity based on input seismic data and the purpose of processing, and organizing them according to different data domains; S2. Extracting features of the attributes of different data domains using a deep learning network to generate a feature matrix; S3. Classifying the feature matrix using a clustering analysis algorithm to achieve seismic data classification; S4. Applying the classification results of the seismic data to perform targeted data processing to achieve seismic data quality control. The present invention realizes automated intelligent quality control by introducing deep learning networks and unsupervised learning technologies in machine learning, significantly improving the efficiency of seismic data processing, and effectively reducing the error rate. The present invention is suitable for quality control in seismic data processing.

Description

Technical Field

[0001] The present invention belongs to the technical field of seismic data processing, and relates to a quality control method in the process of seismic data processing, and specifically to an intelligent quality control method and device for seismic data based on machine learning. Background Art

[0002] Seismic exploration is a key technology in modern geological science and resource exploration. By collecting and processing the propagation of seismic waves through different geological structures, it can effectively interpret the underground geological structure. This technology plays a vital role in the exploration of oil, natural gas, mineral resources, and groundwater resources. Furthermore, seismic exploration provides crucial data support for the assessment of geological hazards such as earthquakes and landslides.

[0003] With technological advancements and increasing exploration needs, the scale of seismic data acquisition continues to expand, as evidenced by the widespread adoption of the "two-width, one-high" seismic acquisition method. This method increases trace density and the size of the work area, rapidly increasing the amount of seismic data in a single work area from terabytes to petabytes. This surge in data volume poses unprecedented challenges to the computational efficiency of seismic data processing, with quality control (QC) being particularly critical. Effective data quality control not only ensures data accuracy but also effectively improves the efficiency and refinement of subsequent data processing and interpretation.

[0004] Traditional methods for seismic data quality control rely primarily on manual inspection, but this approach proves insufficient when faced with petabyte-scale data. Manual inspection is not only time-consuming but also susceptible to the skills and experience of data processors, often leading to inconsistent results. While automated quality control techniques, such as rule-based anomaly detection, have made some progress, these methods generally lack the flexibility to handle complex seismic data. This is particularly true when transitioning from land-based exploration to the more complex marine exploration process. As the exploration area expands and the data complexity increases, the task of quality control becomes increasingly complex and time-consuming.

[0005] To address the challenges faced by existing quality control methods, research has shown that machine learning-based techniques can be used to automate data quality testing. However, most applications still rely on supervised learning methods, which require large amounts of labeled data and are not suitable for rapid analysis of large-scale seismic data. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for intelligent quality control of seismic data based on machine learning, so as to solve the problem that there is currently no efficient quality control method applicable to large-scale seismic data.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A method for intelligent quality control of seismic data based on machine learning, comprising the following steps:

[0009] S1. Seismic Data Attribute Calculation: Based on the input seismic data and processing objectives, single-channel attributes with correlation and sensitivity are calculated and organized according to different data domains.

[0010] S2. Seismic data attribute feature extraction: Attributes from different data domains are extracted using a deep learning network to generate a feature matrix.

[0011] S3. Seismic data classification: Use cluster analysis algorithms to classify the feature matrix and achieve seismic data classification;

[0012] S4. Seismic data quality control: Apply the classification results of seismic data to perform targeted data processing to achieve seismic data quality control.

[0013] As a limitation, when the single-channel attribute with correlation and sensitivity calculated in step S1 is multiple attributes, principal component analysis is used to achieve data dimensionality reduction.

[0014] As another limitation, the method further includes pre-processing the seismic data before step S1.

[0015] As a third limitation, the seismic data includes pre-stack data or post-stack data.

[0016] As a fourth limitation, the attributes include statistical domain attributes, spectral domain attributes and time domain attributes.

[0017] As a further limitation, the statistical domain attributes include maximum value, minimum value, mean value, median value, skewness, kurtosis and time difference; the spectral domain attributes include Fourier transform coefficients, wavelet transform parameters and spectrum distance; the time domain attributes include autocorrelation function, differential mean value and entropy.

[0018] As a fifth limitation, the data domain includes a shot domain or a detection point domain.

[0019] As a sixth limitation, the deep learning network includes a convolutional neural network (CNN).

[0020] As a seventh limitation, the clustering analysis algorithm includes a K-means algorithm.

[0021] The present invention also provides an intelligent quality control device for seismic data based on machine learning, comprising the following modules connected in sequence:

[0022] Seismic data attribute calculation module: used to calculate single-channel attributes with correlation and sensitivity based on the input seismic data and processing purpose, and organize them according to different data domains;

[0023] Seismic data attribute feature extraction module: used to extract features from attributes of different data domains using a deep learning network to generate a feature matrix;

[0024] Seismic data classification module: uses cluster analysis algorithm to classify the feature matrix and realize seismic data classification;

[0025] Seismic data quality control module: applies the classification results of seismic data to perform targeted data processing to achieve quality control of seismic data.

[0026] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned intelligent quality control method for seismic data based on machine learning is implemented.

[0027] The present invention also provides a computer-readable storage medium, which stores a computer program for executing the above-mentioned intelligent quality control method for seismic data based on machine learning.

[0028] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared with the prior art:

[0029] The present invention provides a machine learning-based intelligent seismic data quality control method and device. By introducing deep learning networks and unsupervised learning technologies in machine learning, it can realize automated intelligent quality control of the seismic data processing process, significantly improving the efficiency of seismic data processing. It can accurately identify abnormal nodes and other quality problems in the data (such as noise, abnormal instrument response and positioning errors, etc.), effectively reduce the error rate, and reduce the misoperation of normal data. It also has good adaptability and scalability, and can adapt to more complex and larger-scale seismic data processing needs in the future, thereby improving the quality and efficiency of the entire seismic exploration project. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of the method in Example 1;

[0031] Figure 2 The display result of the pre-stack seismic data in Example 1 in GeoEast software;

[0032] Figure 3 This is the node classification result diagram in Example 1, where Figure 3 (a) is the first type of node, Figure 3(b) is the second type of node, Figure 3 (c) is the third type of node;

[0033] Figure 4 This is a schematic structural diagram of the device in Example 2.

[0034] 1. Seismic data attribute calculation module; 2. Seismic data attribute feature extraction module; 3. Seismic data classification module; 4. Seismic data quality control module. DETAILED DESCRIPTION

[0035] The present invention will be further described in detail below by way of specific examples. It should be understood that the described examples are only used to illustrate the present invention and are not intended to limit the present invention.

[0036] Example 1

[0037] This embodiment discloses a method for intelligent quality control of seismic data based on machine learning, the flow chart of which is as follows: Figure 1 As shown in Figure 3, this method is applied to the quality control of the secondary positioning process of a certain ocean bottom node (OBN) seismic receiver point.

[0038] During offshore exploration operations, GPS and other positioning systems are used to deploy nodes on the sea surface according to designed coordinates (primary positioning). However, due to factors such as ocean currents, tides, ship speed, and wind and waves, the actual placement of the nodes often deviates from the intended location. Even if the nodes are initially precisely placed, fishing vessel activity and changes in ocean climate during the exploration period can cause the nodes to shift. This positional shift not only affects the quality of seismic data acquisition but also increases the difficulty of subsequent seismic data processing.

[0039] The method provided in this embodiment can accurately identify nodes of different categories based on pre-stack seismic data by classifying nodes in the detection point domain, and then implement targeted secondary positioning processing and parameter selection, thereby improving the efficiency and accuracy of secondary positioning. Specifically, the following steps are included:

[0040] S0. Seismic Data Preprocessing

[0041] The collected seismic data (including pre-stack data or post-stack data, in this embodiment, pre-stack data, specifically OBN pre-stack seismic data) are input into a specific seismic processing and interpretation software (in this embodiment, GeoEast software is used). The results are as follows: Figure 2 shown.

[0042] During this process, depending on the equipment and environmental conditions used during data acquisition, the seismic data can be preprocessed (e.g., inspecting the observation system, removing damaged data channels, suppressing noise, etc.). Alternatively, preprocessing can be omitted and the next step can be performed directly. In this embodiment, no preprocessing was performed and the next step was performed directly.

[0043] S1. Calculation of earthquake data attributes

[0044] According to the input seismic data and the purpose of processing, various statistical domain attributes (including maximum, minimum, mean, median, skewness, kurtosis and time difference, etc.), spectral domain attributes (Fourier transform coefficients, wavelet transform parameters and spectral distance, etc.) and time domain attributes (autocorrelation function, differential mean and entropy, etc.) of a single channel can be calculated. The selection of which attributes to calculate depends on the processing purpose and data sensitivity to ensure that the calculated attributes are relevant and sensitive to the data processing process.

[0045] The attribute calculated in this embodiment is the time difference. Based on the first arrival data of the pre-stack direct wave, the difference between the theoretical arrival time and the actual arrival time of each direct wave is calculated, that is, the single-channel time difference. The time difference is organized according to the detection point domain to form a time difference matrix for each node in the detection point domain.

[0046] If the calculated single-channel attributes with correlation and sensitivity are multiple attributes, principal component analysis can be used to achieve data dimensionality reduction and reduce computational complexity. Alternatively, no processing can be performed and the next step can be performed directly. If the single-channel attribute is a single attribute, principal component analysis is not required and the next step can be performed directly. In this example, only the time difference attribute with correlation and sensitivity is calculated, so principal component analysis is not required and the next step can be performed directly.

[0047] S2. Extraction of seismic data attribute features

[0048] The time difference matrix in the detection point domain is subjected to feature extraction using a deep learning network (a convolutional neural network is used in this embodiment) to generate a feature matrix.

[0049] S3. Earthquake data classification

[0050] Using cluster analysis algorithm (K-means algorithm is used in this embodiment), the feature matrix is ​​classified to achieve node classification. The results are as follows: Figure 3 As shown;

[0051] Depend on Figure 3As can be seen, after classification, based on visual intuition and parameter calculation, the seafloor nodes can be divided into three categories. Combined with practical work experience, we can see that the first category of nodes is characterized by inaccurate X, Y, and Z coordinates of the geophone points; the second category is characterized by inaccurate X, Y, and Z coordinates of the geophone points; and the third category is characterized by inaccurate Z coordinates of the geophone points or inaccurate water speeds. The classification results show significant differences between different categories, while the results within each category are similar or even consistent, demonstrating the effectiveness and accuracy of the classification results.

[0052] S4. Seismic Data Quality Control

[0053] Applying the classification results of the identified nodes, targeted secondary positioning processing is performed according to their different characteristics, and the most appropriate positioning parameters are selected to ensure the accuracy and efficiency of positioning (for example, the first type of node feature is that the XYZ coordinates of the detection point are inaccurate, so the data in the three directions of XYZ need to be corrected; the second type of node feature is that the XY coordinates are inaccurate, so only the data in the XY direction needs to be corrected, and the data in the Z direction does not need to be corrected. This not only greatly reduces the amount of data processing and improves computing efficiency, but also prevents the erroneous processing of data that does not need to be modified, and improves the accuracy of data processing), thereby achieving quality control.

[0054] If the traditional direct wave first arrival location method is used, the direct wave first arrival information is directly used to analyze the overall seismic data through spatial and temporal consistency parameters to infer the most likely location of the node. This requires solving an overdetermined linear equation system, which is computationally difficult and has low accuracy.

[0055] By adopting the method of this embodiment, the direct wave first arrival data is first deeply processed and classified, and then targeted secondary positioning processing is performed based on the classification results, which greatly reduces the difficulty of calculation and improves the accuracy and efficiency of positioning.

[0056] Example 2

[0057] This embodiment provides an intelligent seismic data quality control device based on machine learning, the structural diagram of which is shown in FIG. Figure 4 As shown, it specifically includes the following modules connected in sequence:

[0058] Seismic data attribute calculation module 1: used to calculate single-channel attributes with correlation and sensitivity based on the input seismic data and the purpose of processing, and organize them according to different data domains;

[0059] Seismic data attribute feature extraction module 2: used to extract features from attributes of different data domains using a deep learning network to generate a feature matrix;

[0060] Seismic data classification module 3: Use cluster analysis algorithm to analyze the feature matrix and classify seismic data;

[0061] Seismic data quality control module 4: applies the classification results of seismic data to perform targeted data processing to achieve seismic data quality control.

[0062] Example 3

[0063] This embodiment provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, so as to implement the above-mentioned intelligent quality control method for seismic data based on machine learning.

[0064] The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc.

[0065] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. The processor is configured to execute the computer-readable instructions stored in the memory.

[0066] Those skilled in the art should understand that in order to solve the technical problem of how to obtain a good user experience, this embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the scope of protection of this disclosure.

[0067] For detailed description of this embodiment, please refer to the corresponding description in the aforementioned embodiments, which will not be repeated here.

[0068] Example 4

[0069] This embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the intelligent quality control method of seismic data based on machine learning.

[0070] The computer-readable storage medium stores non-transitory computer-readable instructions, which, when executed by a processor, execute all or part of the steps of the aforementioned methods.

[0071] The above-mentioned computer-readable storage media include, but are not limited to, optical storage media (e.g., CD-ROMs and DVDs), magneto-optical storage media (e.g., MOs), magnetic storage media (e.g., magnetic tapes or mobile hard disks), media with built-in rewritable non-volatile memory (e.g., memory cards), and media with built-in ROM (e.g., ROM cartridges).

Claims

1. A method for intelligent quality control of seismic data based on machine learning, characterized in that: The following steps are involved: S1. Calculate single-channel attributes with correlation and sensitivity based on the input seismic data and processing purpose, and organize them according to data domains; S2. Extract features from attributes of different data domains using a deep learning network to generate a feature matrix; S3. Use cluster analysis algorithm to classify the feature matrix and classify seismic data in different data domains; S4. Apply the classification results of seismic data in different data domains to perform targeted data processing to achieve quality control of seismic data in different data domains; Among them, the deep learning network includes a convolutional neural network; the clustering analysis algorithm includes a K-means algorithm.

2. The intelligent quality control method of seismic data based on machine learning according to claim 1, characterized in that: When the single-channel attribute with correlation and sensitivity calculated in step S1 is multiple attributes, principal component analysis is used to achieve data dimensionality reduction.

3. The intelligent quality control method of seismic data based on machine learning according to claim 1, characterized in that: The method also includes pre-processing the seismic data before step S1.

4. A method for intelligent quality control of seismic data based on machine learning according to any one of claims 1 to 3, characterized in that: The seismic data includes pre-stack data or post-stack data.

5. A method for intelligent quality control of seismic data based on machine learning according to any one of claims 1 to 3, characterized in that: The attributes include statistical domain attributes, spectral domain attributes and time domain attributes.

6. The intelligent quality control method for seismic data based on machine learning according to claim 5, characterized in that: The statistical domain attributes include maximum value, minimum value, mean value, median value, skewness, kurtosis and time difference; the spectral domain attributes include Fourier transform coefficients, wavelet transform parameters and spectrum distance; the time domain attributes include autocorrelation function, differential mean value and entropy.

7. A method for intelligent quality control of seismic data based on machine learning according to any one of claims 1 to 3, characterized in that: The data domain includes a shot domain or a detection point domain.

8. An intelligent seismic data quality control device based on machine learning, characterized in that: It includes the following modules connected in sequence: Seismic data attribute calculation module: used to calculate single-channel attributes with correlation and sensitivity based on the input seismic data and processing purpose, and organize them according to different data domains; Seismic data attribute feature extraction module: used to extract features from attributes of different data domains using a deep learning network to generate a feature matrix; Seismic data classification module: uses cluster analysis algorithm to classify the feature matrix and realize seismic data classification; Seismic data quality control module: applies the classification results of seismic data to perform targeted data processing to achieve quality control of seismic data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the intelligent quality control method for seismic data based on machine learning according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program for executing the intelligent quality control method for seismic data based on machine learning as described in any one of claims 1-7.

Images