Seismic data preprocessing method, device and equipment and storage medium

Abstract

The application discloses a seismic data preprocessing method, device and equipment and a storage medium. The method comprises the following steps: preprocessing seismic three-component data according to seismic observation data; generating a receiving function according to the seismic three-component data; determining a preset number of sample receiving functions in the receiving function, and generating artificial label classification of the sample receiving functions; taking the receiving function as a node, generating a topological graph structure by establishing edge connection between nodes with a Pearson correlation coefficient greater than a preset value; and taking the topological graph structure as input, performing label classification on each receiving function by using a classification prediction model based on a graph convolutional neural network. The application can automatically filter out a large number of poor-quality receiving functions, thereby effectively reducing the manual workload required for selecting receiving functions.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of crustal structure analysis, and in particular to a seismic data preprocessing method, device, equipment and storage medium. BACKGROUND

[0002] Seismic observation is currently one of the effective means for studying the geodynamic genesis of oil and gas fields. Seismic data processing requires a large amount of manual and time costs. In the work of phase recognition, seismic wave travel time first arrival picking, dispersion curve and receiving function picking, manual repetitive work is required, which requires a large amount of manual cost.

[0003] For receiving function data processing, the prior art has a method of using a convolutional neural network to train an intelligent model through a large amount of labeled data, and then apply it to the selection of receiving functions of new observations of a single station.

[0004] The inventors have found that the prior art has at least the following defects in the selection of receiving functions in the process of seismic data processing:

[0005] A large amount of receiving functions need to be manually labeled and classified to generate a training set, resulting in a large amount of manual work.

[0006] The information disclosed in this BACKGROUND section is only intended to increase an understanding of the general background of the application and is not intended to be a recognition or any form of suggestion that this information constitutes prior art. SUMMARY

[0007] The purpose of the present application is to reduce the manual workload of receiving function selection.

[0008] The present application provides a seismic data preprocessing method, comprising the steps of:

[0009] S11, preprocessing seismic three-component data according to seismic observation data;

[0010] S12, generating a receiving function according to the seismic three-component data;

[0011] S13, determining a preset number of sample receiving functions in the receiving function, and generating manual label classification of the sample receiving functions; the categories of the sample receiving functions include "keep" and "discard";

[0012] S14, taking the receiving function as a node, and generating a topological graph structure by establishing an edge connection between nodes with a Pearson correlation coefficient greater than a preset value;

[0013] S15, labeling and classifying each of the receiver functions by a classification prediction model based on a graph convolutional neural network, with the topological graph structure as input.

[0014] Preferably, in the present application, it further comprises:

[0015] S16, manually labeling and classifying the receiver functions classified as "retained" by the classification prediction model.

[0016] Preferably, in the present application, the preprocessing of the seismic three-component data according to the seismic observation data comprises:

[0017] Making a seismic catalog within the time range of the seismic observation data; for P-wave receiver functions, selecting seismic events with epicentral distance between 30-90° and magnitude greater than 5.5;

[0018] Using a global standard one-dimensional velocity model to calculate the theoretical arrival time of P-wave; according to the theoretical arrival time, cutting off the seismic event data; and after removing the instrument response, performing spike removal, mean removal and tilt removal processing on the seismic event data to generate the seismic three-component data.

[0019] Preferably, in the present application, the generating of the receiver functions according to the seismic three-component data comprises:

[0020] Rotating the horizontal component of the preprocessed seismic three-component data to radial and tangential directions, wherein the horizontal component refers to the amplitude of particle vibration in the north and east directions, the radial component refers to the direction of the great circle path from the source to the observation station, and the tangential direction refers to the direction orthogonal to the radial direction in the horizontal plane, and the formula for rotating the horizontal component of the seismic observation data to the radial-tangential coordinate system comprises:

[0021]

[0022] Wherein, R, T, E and N represent radial, tangential, east and north components, respectively, and Baz represents the back azimuth;

[0023] Generating receiver functions using a frequency domain deconvolution method, wherein the receiver function refers to the response of the underground medium below the observation station, and the receiver function is based on the assumption of an equivalent source, which equates the vertical component of the earthquake to a pulse function; the receiver function includes a radial receiver function and a tangential receiver function; wherein the radial receiver function is generated by deconvolving the vertical component with the rotated radial component; the tangential receiver function refers to the deconvolution of the tangential component with the vertical component; and the calculation formula of the receiver function in the frequency domain comprises:

[0024]

[0025]

[0026] In the formula, E R (ω) and E T (ω) are the frequency spectra of the radial and tangential receiver functions; D R (ω), D T (ω) and D V (ω) are the radial, tangential and vertical event spectra, respectively; I(ω) and S(ω) are the instrument response spectrum and the source function spectrum, respectively;

[0027] After the radial and tangential receiver function spectra are obtained, inverse Fourier transform is performed to generate the corresponding receiver functions in the time domain.

[0028] Preferably, in the present application, the topological graph structure is generated by establishing edge connections between nodes with a Pearson correlation coefficient greater than a preset value, based on the receiver functions, including:

[0029] Pearson correlation coefficients of each receiver function and other receiver functions are calculated respectively;

[0030] A preliminary topological graph structure is established respectively based on each receiver function as a node;

[0031] In the preliminary topological graph structure, edge connections are established between nodes with a Pearson correlation coefficient greater than a preset value;

[0032] The topological graph structure is generated by filtering out nodes without edge connections in the preliminary topological graph structure.

[0033] Preferably, in the present application, the classification prediction model based on the graph convolutional neural network is used to label and classify each receiver function based on the topological graph structure, including:

[0034] An adjacency matrix A and a degree matrix D are obtained based on the topological graph structure; and a feature matrix H is generated based on the data of each node; the feature matrix H is an n×T matrix, where n is the number of nodes, and T is the feature dimension, i.e., the number of sampling points of the receiver function.

[0035] Preferably, in the present application, the classification prediction model includes a three-layer network, in which:

[0036] The input layer is the original adjacency matrix A, the degree matrix D and the feature matrix H;

[0037] The leaky ReLU is used as the activation function in the middle layer;

[0038] The Softmax function is used for output in the output layer;

[0039] Preferably, in the present application, the preset value ranges from 0.6 to 0.8.

[0040] In another aspect of the present application, a seismic data preprocessing device is also provided, comprising:

[0041] A three-component data generation unit is configured to preprocess seismic three-component data according to seismic observation data;

[0042] A receiver function calculation unit is configured to generate a receiver function according to the seismic three-component data;

[0043] An artificial label generation unit is configured to determine a preset number of sample receiver functions in the receiver function and generate an artificial label classification of the sample receiver functions; the categories of the sample receiver functions include "retained" and "rejected";

[0044] A topological graph generation unit is configured to generate a topological graph structure by establishing an edge connection between nodes with a Pearson correlation coefficient greater than a preset value;

[0045] An automatic label classification unit is configured to take the topological graph structure as input and perform label classification on each receiver function by a classification prediction model based on a graph convolutional neural network.

[0046] In another aspect of the present application, a seismic data preprocessing device is also provided, comprising:

[0047] A memory is configured to store a computer program;

[0048] A processor is configured to call and execute the computer program to realize each step of the seismic data preprocessing method according to any one of the above aspects.

[0049] In another aspect of the present application, a storage medium is also provided, which stores a computer program, and the computer program is executed by a processor to realize each step of the seismic data preprocessing method according to any one of the above aspects.

[0050] The seismic data preprocessing device comprises a computer program stored in a medium, and the computer program comprises program instructions, which, when executed by a computer, cause the computer to execute the method according to each of the above aspects and achieve the same technical effects.

[0051] Compared with the prior art, the present application has the following beneficial effects:

[0052] The application selects a small amount of sample receiving functions in the receiving function for artificial marking classification as the labeled data in semi-supervised learning after calculating and generating the receiving function by using the frequency domain deconvolution method; in the application, when generating the topology graph structure according to the receiving function, the Pearson correlation coefficient preset value is set as the condition for connecting the edges between the receiving functions in the generated topology graph structure, so that a part of the receiving functions with poor quality can be preliminarily filtered out; then, the topology graph structure is taken as the input, and each receiving function is marked and classified by the classification prediction model based on the graph convolution neural network, so that more receiving functions with poor quality can be further filtered out.

[0053] Due to the application, a large amount of receiving functions with poor quality can be effectively and automatically filtered out, so the data amount of the receiving functions that need to be manually marked can be effectively reduced, and thus the manual workload required for selecting the receiving functions is effectively reduced.

[0054] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application and can be implemented according to the content of the specification, at the same time, in order to make the above and other purposes, technical features and advantages of the application more easily understood, one or more preferred embodiments are listed below, and are described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical scheme of the application, the drawings required by the embodiments will be briefly introduced below, and obviously, the drawings in the following description can also be used to obtain other drawings without creative labor for those skilled in the art.

[0056] Figure 1 is a step diagram of the seismic data preprocessing method described in the application;

[0057] Figure 2 is a structural schematic diagram of the seismic data preprocessing device described in the application

[0058] Figure 3 is a structural schematic diagram of the seismic data preprocessing device described in the application DETAILED DESCRIPTION

[0059] The specific embodiments of the application will be described in detail below with reference to the accompanying drawings, but it should be understood that the protection scope of the application is not limited by the specific embodiments.

[0060] Unless otherwise explicitly stated, throughout the specification and claims, the term "comprise" or its variants such as "comprises" or "comprising" will be understood to include the stated element or component, but not to exclude other elements or components.

[0061] In this document, the terms "first", "second", etc. are used to distinguish two different elements or parts, and are not used to define a particular position or relative relationship. In other words, in some embodiments, the terms "first", "second", etc. can also be interchanged with each other.

[0062] Example One

[0063] In order to reduce the artificial workload of the receiver function selection, as shown in the embodiments of the application, a seismic data preprocessing method is provided, comprising the steps of: Figure 1

[0064] S11, preprocessing the seismic three-component data according to the seismic observation data;

[0065] CCP is the abbreviation of Cmmom Conversion Point, which refers to the wave field back propagation according to the kinematic properties of the ray path between the underground seismic phase conversion point and the surface receiving station, the underground conversion point is arranged in two dimensions or three dimensions, the amplitude of the receiving function corresponding to the time of the surface observation station position is back propagated to the underground conversion point, and imaging is performed in the depth domain.

[0066] In the embodiments of the application, the preprocessed seismic three-component data refers to: first, a seismic catalog within the time range of the seismic observation data is made, for the P-wave receiving function, usually a seismic event with a hypocentral distance of 30-90° and a magnitude greater than 5.5 is selected; then the theoretical arrival time of P-wave is calculated using the global standard one-dimensional velocity model (IASP91, AK135 model, etc.), the seismic event data is intercepted according to the theoretical arrival time, the instrument response is removed, and then the seismic event data is processed for spike removal, mean removal and tilt removal.

[0067] S12, generating a receiving function according to the seismic three-component data;

[0068] In this step, the horizontal component of the preprocessed seismic three-component data is rotated to the radial and tangential directions, which can specifically include:

[0069] The horizontal component refers to the amplitude of the northward and eastward particle vibration on the horizontal plane, the radial component refers to the direction of the great circle path between the source and the observation station, and the tangential direction refers to the direction orthogonal to the radial direction on the horizontal plane. The formula for rotating the horizontal component of the seismic observation data to the radial and tangential coordinate system includes:

[0070]

[0071] Wherein, R, T, E and N represent radial, tangential, eastward and northward components respectively, and Baz represents the back azimuth.

[0072] ​Then, a receiving function is generated by using a frequency domain deconvolution method, and specifically can include:

[0073] The receiving function refers to a response of a subsurface medium below the observation station, and is based on an equivalent source assumption, in which a vertical component of a seismic wave is equivalent to an impulse function; the receiving function includes a radial receiving function and a tangential receiving function; the radial receiving function is deconvolution of the vertical component by using a rotated radial component; the tangential receiving function refers to deconvolution of a tangential component on the vertical component; and a calculation formula of the receiving function in the frequency domain includes:

[0074]

[0075]

[0076] wherein, E R (ω) and E T (ω) are frequency spectrums of the radial receiving function and the tangential receiving function; D R (ω), D T (ω) and D V (ω) are frequency spectrums of radial, tangential and vertical seismic events respectively; I(ω) and S(ω) are an instrument response spectrum and a source function frequency spectrum respectively.

[0077] After the radial and tangential receiving function spectrums are obtained, the corresponding receiving functions can be generated by inverse Fourier transform to the time domain.

[0078] S13, determining a preset number of sample receiving functions in the receiving function, and generating an artificial label classification of the sample receiving functions; the sample receiving function categories include "retained" and "eliminated";

[0079] In the present application, a classification prediction model of a graph convolutional neural network based on semi-supervised learning is included, and for this purpose, a small amount of sample receiving functions in the receiving function need to be artificially labeled to generate labeled data. Since the purpose of the embodiment of the present application is to filter out poor quality receiving functions, when the artificial label classification is performed, the categories of the receiving functions can be determined as the two categories of retained and eliminated.

[0080] S14, taking the receiving function as a node, and generating a topological graph structure by establishing an edge connection between nodes with a Pearson correlation coefficient greater than a preset value;

[0081] In the embodiment of the present application, the poor quality receiving functions are preliminarily filtered out when generating the topological graph structure for the classification prediction model by establishing the edge connection between the nodes with the preset Pearson correlation coefficient. In actual application, the following sub-steps can be specifically included:

[0082] S41, respectively calculate the Pearson correlation coefficient of each receiving function and other receiving functions;

[0083] S42, respectively establish a preliminary topology structure with each receiving function as a node;

[0084] S43, in the preliminary topology structure, an edge connection is established between nodes with a Pearson correlation coefficient greater than a preset value; in actual application, the preset value in the embodiment of the application can be set to 0.6 to 0.8. Further, the preferred value can be set to 0.63.

[0085] S44, by filtering out the nodes without edge connections in the preliminary topology structure, the topology structure is generated.

[0086] S15, taking the topology structure as input, each receiving function is labeled and classified by a classification prediction model based on a graph convolutional neural network.

[0087] The nodes in the topology structure in the embodiment of the application are receiving functions that have been preliminarily filtered out of poor quality receiving functions; in order to further filter out poor quality receiving functions, the topology structure is taken as input, each receiving function is labeled and classified by a classification prediction model based on a graph convolutional neural network, and then only the receiving functions labeled "keep" by the classification prediction model are retained.

[0088] The specific process of this step can include: obtaining an adjacency matrix A and a degree matrix D according to the topology structure; and generating a feature matrix H according to the data of each node; the feature matrix H is an n x T matrix, where n is the number of nodes; T is the feature dimension, i.e., the number of sampling points of the receiving function.

[0089] In actual application, the classification prediction model in the embodiment of the application can include a three-layer network, wherein:

[0090] The input layer is the original adjacency matrix A, the degree matrix D and the feature matrix H;

[0091] The intermediate layer uses leaky ReLU as the activation function;

[0092] The output layer uses the Softmax function for output.

[0093] The inventors have found that the method proposed in the embodiment of the application cannot completely automatically select receiving functions, and some poor quality receiving functions are also retained in the receiving functions labeled "keep" by the classification prediction model, therefore, further, the embodiment of the application can further include the following steps:

[0094] S16. Manually marking the received functions classified as “retained” by the classification prediction model.

[0095] The working principle and technical effect of the embodiment of the present application are described below with a specific example:

[0096] The observation data of a seismic station (MDJ) was selected to extract received functions from all data in 2017, including a total of 620 received functions, each of which is a 600-dimensional vector. To verify the accuracy of the seismic data preprocessing method in the embodiment of the present application, all received functions were manually marked, and the One-hot encoding method was used, in which the received functions to be retained were marked as [1, 0], and the received functions to be discarded were marked as [0, 1]. According to the statistics, 103 received functions were marked as “retained”, and 517 received functions were marked as “discarded”.

[0097] Each received function was taken as a node in the topological graph structure, and the Pearson correlation coefficient between each pair of received functions was calculated. If the Pearson correlation coefficient is greater than 0.6, an edge is established, otherwise no edge is established. All received functions are traversed to form the final topological graph structure. For the nodes without edge connection, they are directly preliminarily filtered out in the topological graph structure, and the corresponding received functions are marked as “discarded”.

[0098] In this example, a total of 182 received functions were discarded in the preliminary filtering. When checking the manual marks of these discarded received functions, it was found that the mark results were all [0, 1], that is, the manual marks of these preliminarily filtered received functions were all “discarded”. That is, the received functions filtered out are all poor quality received functions.

[0099] Next, the final topological graph structure was generated using the 438 received functions still retained, which was used for classification prediction based on the classification prediction model of the graph convolutional neural network.

[0100] In this example, the classification prediction model based on the graph convolutional neural network was written in python language, mainly calling the Pytorch library to realize the establishment of the graph convolutional neural network, and the received function reading and output operation was realized through the Obspy library. The graph neural network in this example includes three layers, the dimension of the first layer is set to 600, which is the same as the feature number of the received function, the feature number of the second layer is set to 64, and the feature number of the third layer is consistent with the dimension of the label of the received function, which is set to 2. The Epoch of training is set to 50, the learning efficiency is set to 0.01, and the loss function adopts the cross-entropy function. In the topological graph structure, 30 nodes are randomly selected as the training set, and the remaining all received functions are taken as the test set.

[0101] The test result shows that after 50 cycles of training, the accuracy of the final training set is close to 100%, in the test set, the recall rate is 92.47%, the precision rate is 47.25%, and the accuracy is 74.75%. Among them, the recall rate is a coverage measure. It measures the identification ability of the prediction classification model to positive examples (reserved acceptance functions); the precision rate is for the prediction result, which means the probability of actually being a positive sample in all samples predicted as positive. The precision rate represents the prediction accuracy in the positive sample result; and the accuracy rate is the ratio of the predicted label and the true label.

[0102] From the three indexes of recall rate, precision rate and accuracy rate, the acceptance function automatic selection method based on the classification prediction model of the graph convolutional neural network in the embodiment of the application can preferably select the acceptance functions to be reserved, but about half of the acceptance functions marked as "discard" will be included in the prediction of the reserved acceptance functions. Therefore, further screening by manual is still needed.

[0103] Although the method proposed in the embodiment of the application cannot realize complete automatic selection of acceptance functions, manual participation is still needed in the annotation of the training set and the final determination of the reserved acceptance functions. However, through the preprocessing of the seismic data in the embodiment of the application, the artificial workload of selecting acceptance functions can be greatly reduced. For example, in the training set, after classification and prediction by the classification prediction model, 182 acceptance functions are marked as "reserved", which need to be further selected. Since the number of acceptance functions that need to be manually marked can be reduced to about 40% of the total acceptance functions by the embodiment of the application, about 60% of the workload is saved. Therefore, the acceptance function selection method proposed in the application can effectively reduce the artificial workload of selecting acceptance functions.

[0104] The present example includes a total of 620 acceptance functions. After the classification prediction model in the embodiment of the application classifies and marks each of the acceptance functions, only 7 acceptance functions are manually marked as "reserved" acceptance functions among a total of 408 acceptance functions marked as "discard" acceptance functions, but are automatically filtered out, with an error rate of only 1.7%. Therefore, the embodiment of the application can accurately automatically filter out poor quality acceptance functions and reduce the workload of manual selection.

[0105] To sum up, the embodiment of the present application selects a small amount of sample receiving functions in the receiving function for artificial marking classification after calculating and generating the receiving function by using the frequency domain deconvolution method, to serve as the marked data in semi-supervised learning; in the present application, when generating the topology graph structure according to the receiving function, the Pearson correlation coefficient preset value is set to serve as the condition for connecting the edges between the receiving functions in the generated topology graph structure, so that a part of the receiving functions with poor quality can be preliminarily filtered out; then, the topology graph structure is taken as the input, and the classification prediction model based on the graph convolutional neural network is used to mark and classify each receiving function, to further filter out more receiving functions with poor quality.

[0106] Due to the fact that a large amount of receiving functions with poor quality can be effectively and automatically filtered out by the present application, the data amount of the receiving functions that need to be artificially marked can be effectively reduced, so that the artificial workload required for selecting the receiving functions is effectively reduced.

[0107] Embodiment two

[0108] Corresponding to the method embodiment, the present application also provides a seismic data preprocessing device, Figure 2 The structural schematic diagram of the seismic data preprocessing device provided by the embodiment of the present application is shown, and the seismic data preprocessing device is Figure 1 The device corresponding to the seismic data preprocessing method in the corresponding embodiment, that is, the virtual device is used to realize Figure 1 The seismic data preprocessing method in the corresponding embodiment, each virtual module of the seismic data preprocessing device can be executed by an electronic device, such as a network device, a terminal device or a server. Specifically, the seismic data preprocessing device in the embodiment of the present application comprises:

[0109] The three-component data generation unit 01 is used for preprocessing the seismic three-component data according to the seismic observation data;

[0110] The receiving function calculation unit 02 is used for generating the receiving function according to the seismic three-component data, and filtering the receiving function by using the time domain prediction deconvolution or the frequency domain resonance filter, to eliminate the multiple reverberation of the sediment layer;

[0111] The artificial marking generation unit 03 is used for determining a preset amount of sample receiving functions in the receiving function, and generating the artificial marking classification of the sample receiving functions; the categories of the sample receiving functions include “retention” and “elimination”;

[0112] The topology graph generation unit 04 is used for taking the receiving function as the node, and generating the topology graph structure by establishing the edge connection between the nodes with the Pearson correlation coefficient greater than the preset value;

[0113] The automatic labeling classification unit 05 is configured to label and classify each of the received functions by using a graph convolutional neural network-based classification prediction model, with the topological graph structure as input.

[0114] It should be noted that the specific implementation manner and technical effects of the seismic data preprocessing device in the embodiments of the present application can refer to the corresponding seismic data preprocessing method in the embodiments of the present application, which will not be repeated here. Figure 1 The corresponding seismic data preprocessing method in the embodiments of the present application, which will not be repeated here.

[0115] Embodiment three

[0116] Corresponding to the method embodiments, the embodiments of the present application also provide a seismic data preprocessing device, such as a terminal, a server, etc. The server can be a stand-alone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc., but is not limited thereto.

[0117] An example diagram of a hardware structure block diagram of the seismic data preprocessing device provided by the embodiments of the present application is shown in FIG. 1, which can include: Figure 3

[0118] a processor 1, a communication interface 2, a memory 3, and a communication bus 4;

[0119] The processor 1, the communication interface 2, and the memory 3 can communicate with each other through the communication bus 4;

[0120] Optionally, the communication interface 2 can be an interface of a communication module, such as an interface of a GSM module;

[0121] The processor 1 can be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0122] The memory 3 can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0123] The processor 1 is specifically configured to execute a computer program stored in the memory 3 to perform the following steps:

[0124] S11, preprocessing seismic three-component data according to seismic observation data;​

[0125] S12, generating a receiving function according to the seismic three-component data;

[0126] S13, determining a preset number of sample receiving functions in the receiving function, and generating artificial marking classification of the sample receiving functions; the categories of the sample receiving functions include "retention" and "rejection";

[0127] S14, taking the receiving function as a node, and generating a topological graph structure by establishing edge connection between nodes with Pearson correlation coefficient greater than a preset value;

[0128] S15, taking the topological graph structure as input, and marking classification of each receiving function by a classification prediction model based on a graph convolutional neural network.

[0129] Preferably, in the embodiments of the present application, further comprising:

[0130] S16, marking classification of the receiving function classified as "retention" by the classification prediction model.

[0131] The above product can execute the method provided by the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method. Technical details not described in detail in the embodiments can be referred to the seismic data preprocessing method provided by the embodiments of the present application.

[0132] Embodiment four

[0133] In the embodiments of the present application, a storage medium is also provided, which can store a program suitable for execution by a processor, and the program is used for:

[0134] S11, preprocessing seismic three-component data according to seismic observation data;

[0135] S12, generating a receiving function according to the seismic three-component data;

[0136] S13, determining a preset number of sample receiving functions in the receiving function, and generating artificial marking classification of the sample receiving functions; the categories of the sample receiving functions include "retention" and "rejection";

[0137] S14, taking the receiving function as a node, and generating a topological graph structure by establishing edge connection between nodes with Pearson correlation coefficient greater than a preset value;

[0138] S15, taking the topological graph structure as input, and marking classification of each receiving function by a classification prediction model based on a graph convolutional neural network.

[0139] Preferably, in the embodiments of the present application, further comprising:

[0140] S16, manually marking the received function classified as "retained" by the classification prediction model.

[0141] Optionally, the refinement function and the extension function of the program can refer to the above description.

[0142] The product described above can execute the method provided by the embodiments of the application, and has the corresponding function modules and beneficial effects of the execution method. Technical details not described in detail in the embodiments can refer to the method provided by other embodiments of the application.

[0143] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0144] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. In addition, the coupling or direct coupling or communication connection between the displayed or discussed elements can be indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0145] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0146] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0147] It should be understood that the features of the embodiments of the present application, the various embodiments, and the features can be combined with each other, and all can achieve the purpose of solving the above technical problems.

[0148] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0149] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A seismic data preprocessing method, characterized in that, Including the following steps: S11. Based on earthquake observation data, preprocess the three-component earthquake data; S12. Generate a receiving function based on the three-component earthquake data; S13. Determine a preset number of sample receiving functions from the receiving functions, and generate artificial label classifications for the sample receiving functions; the categories of the sample receiving functions include "retain" and "discard"; S14. Using the receiving function as a node, a topology graph structure is generated by establishing edge connections between nodes whose Pearson correlation coefficient is greater than a preset value. S15. Using the topological graph structure as input, each of the receiving functions is labeled and classified using a classification prediction model based on a graph convolutional neural network.

2. The seismic data preprocessing method according to claim 1, characterized in that, Also includes: S16. Manually label and classify the receiver functions that are classified as "reserved" by the classification prediction model.

3. The seismic data preprocessing method according to claim 1, characterized in that, The preprocessing of the three-component seismic data based on seismic observation data includes: Create an earthquake catalog within the time range of earthquake observation data; for P-wave receiver functions, select earthquake events with epicentral distances between 30° and 90° and magnitudes greater than 5.

5. The theoretical arrival time of P-waves is calculated using a global standard one-dimensional velocity model; seismic event data is extracted based on the theoretical arrival time; and after removing instrument response, the seismic event data is processed to remove spikes, mean, and tilt to generate the three-component seismic data.

4. The seismic data preprocessing method according to claim 3, characterized in that, The step of generating a receiving function based on the three-component seismic data includes: Rotating the horizontal component of the preprocessed three-component seismic data to the radial and tangential directions includes: the horizontal component refers to the amplitude of particle vibration in the north and east directions horizontally; the radial component refers to the direction of the great circle path of the epicenter and the observation station; and the tangential direction refers to the direction orthogonal to the radial direction horizontally. The formula for rotating the horizontal component of the seismic observation data to the radial-tangential coordinate system includes: Where R, T, E, and N represent the radial, tangential, eastward, and northward components, respectively, and Baz represents the reverse azimuth angle; The receiver function is generated using a frequency-domain deconvolution method, comprising: the receiver function being the response of the subsurface medium beneath the observation station; the receiver function being based on the equivalent source assumption, equating the vertical component of the earthquake to an impulse function; the receiver function including a radial receiver function and a tangential receiver function; wherein, the radial receiver function is the deconvolution of the vertical component using the rotated radial component; the tangential receiver function is the deconvolution of the tangential component with respect to the vertical component; the frequency-domain calculation formula for the receiver function includes: In the formula, E R (ω) and E T (ω) represents the spectrum of the radial and tangential receiver functions; D R (ω), D T (ω) and D V (ω) represents the radial, tangential, and vertical seismic event spectra, respectively; I(ω) and S(ω) represent the instrument response spectrum and the source function spectrum, respectively. After obtaining the radial and tangential receiver function spectra, the spectrum is transformed to the time domain by inverse Fourier transform to generate the corresponding receiver function.

5. The seismic data preprocessing method according to claim 1, characterized in that, The step of generating a topology graph structure by establishing edge connections between nodes with a Pearson correlation coefficient greater than a preset value, using the receiving function as the node, includes: Calculate the Pearson correlation coefficients for each of the aforementioned receiving functions and other receiving functions respectively; A preliminary topology graph structure is established using each of the aforementioned receiving functions as nodes; In the initial topology graph structure, edges are established between nodes whose Pearson correlation coefficient is greater than a preset value; The topology graph structure is generated by filtering out nodes without edge connections in the initial topology graph structure.

6. The seismic data preprocessing method according to claim 1, characterized in that, The step of labeling and classifying each of the receiving functions using the topological graph structure as input and a classification prediction model based on a graph convolutional neural network includes: The adjacency matrix A and degree matrix D are obtained based on the topology graph structure; and a feature matrix H is generated based on the data of each node; the feature matrix H is an n×T dimensional matrix, where n is the number of nodes; T is the feature dimension, that is, the number of sampling points of the receiving function.

7. The seismic data preprocessing method according to claim 6, characterized in that, The classification prediction model comprises a three-layer network, wherein: The input layer consists of the original adjacency matrix A, the degree matrix D, and the feature matrix H; The intermediate layer uses leaky ReLU as the activation function; The output layer uses the Softmax function for output.

8. The seismic data preprocessing method according to claim 1, characterized in that, The preset value ranges from 0.6 to 0.

8.

9. A seismic data preprocessing device, characterized in that, include: The three-component data generation unit is used to preprocess the three-component seismic data based on seismic observation data. A receiver function calculation unit is used to generate a receiver function based on the three-component seismic data. The manual label generation unit is used to determine a preset number of sample receiving functions in the receiving functions and generate manual label classifications for the sample receiving functions; the categories of the sample receiving functions include "retain" and "discard"; The topology graph generation unit is used to generate a topology graph structure by establishing edge connections between nodes with a Pearson correlation coefficient greater than a preset value, using the receiving function as the node. An automatic labeling and classification unit is used to label and classify each of the receiving functions by taking the topological graph structure as input and using a classification prediction model based on a graph convolutional neural network.

10. A seismic data preprocessing device, characterized in that, include: Memory, used to store computer programs; A processor for invoking and executing the computer program to implement the steps of the seismic data preprocessing method as described in any one of claims 1-8.

11. A storage medium, characterized in that, Includes a software program adapted for execution by a processor of the steps of the seismic data preprocessing method as described in any one of claims 1-8.

Images