CN115201902A - Fault intelligent identification method and system based on deep learning - Google Patents
Fault intelligent identification method and system based on deep learning Download PDFInfo
- Publication number
- CN115201902A CN115201902A CN202210779693.3A CN202210779693A CN115201902A CN 115201902 A CN115201902 A CN 115201902A CN 202210779693 A CN202210779693 A CN 202210779693A CN 115201902 A CN115201902 A CN 115201902A
- Authority
- CN
- China
- Prior art keywords
- fault
- displacement
- network model
- plane
- seismic data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 71
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 31
- 230000006870 function Effects 0.000 claims abstract description 20
- 238000003709 image segmentation Methods 0.000 claims abstract description 16
- 238000013434 data augmentation Methods 0.000 claims abstract description 13
- 238000006073 displacement reaction Methods 0.000 claims description 76
- 238000012545 processing Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 claims description 11
- 230000001131 transforming effect Effects 0.000 claims description 8
- 230000037303 wrinkles Effects 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 5
- 238000003384 imaging method Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 description 17
- 238000004590 computer program Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000010276 construction Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 4
- 230000036039 immunity Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/301—Analysis for determining seismic cross-sections or geostructures
- G01V1/302—Analysis for determining seismic cross-sections or geostructures in 3D data cubes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/64—Geostructures, e.g. in 3D data cubes
- G01V2210/642—Faults
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Remote Sensing (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geophysics (AREA)
- Databases & Information Systems (AREA)
- Geology (AREA)
- Medical Informatics (AREA)
- Environmental & Geological Engineering (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention relates to a fault intelligent identification method and a fault intelligent identification system based on deep learning, wherein the fault intelligent identification method comprises the following steps: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using the training samples and the labels to obtain a network model for intelligently detecting faults in the three-dimensional seismic data; carrying out data standardization and data augmentation on parameters of the network model, selecting an optimized loss function, and improving the fault recognition capability of the network model; inputting real seismic data into the network model for prediction, evaluating a prediction result, and adjusting and perfecting a training sample. The invention has higher efficiency, more accurate result and stronger noise resistance; can be applied in the field of seismic exploration.
Description
Technical Field
The invention relates to the technical field of seismic exploration, in particular to a fault intelligent identification method and system based on deep learning.
Background
The interpretation of fault layers in seismic imaging data is not only a key step in oil and gas exploration and development, but also an important basis for construction and earth dynamics analysis. In the aspect of oil and gas exploration and development, faults are widely concerned at home and abroad due to the migration and aggregation of oil and gas. Meanwhile, faults and cracks and karst cave development zones controlled by the faults are the most important research objects in China. The fine interpretation of fault layers in seismic data plays a vital role in the prediction of oil and gas reservoirs and the domestic oil and gas exploration and development.
Conventional fault automatic interpretation methods, such as coherence, variance, curvature, fault likelihood, etc., to detect the location of faults in seismic data have been widely used in fault identification analysis, where all of these attributes are to detect faults by calculating the continuity or discontinuity of the seismic event. Due to the fact that non-geological factors such as noise and seismic acquisition errors can also generate discontinuity of seismic event axes, the seismic attribute calculation method is very sensitive to the non-geological factors, and some information irrelevant to faults can be highlighted while faults are detected. Therefore, based on these seismic attributes, further fault interpretation still often requires extensive post-processing and human intervention.
It can be seen that the conventional fault automatic interpretation method is limited by the sensitivity of coherence and variance to the discontinuities of seismic reflection caused by noise or non-fault structures, resulting in many noise or interference features in the fault detection results. And the deep learning approach may avoid this problem. The fault detected by the traditional method is particularly discontinuous, and the later fault plane picking or tracking is challenged. Therefore, in practical applications, selecting or designing an accurate, efficient and lightweight network model is an important issue.
Disclosure of Invention
In view of the above problems, an object of the present invention is to provide a fault intelligent recognition method and system based on deep learning, which has higher efficiency, more accurate result and stronger noise immunity.
In order to achieve the purpose, the invention adopts the following technical scheme: a fault intelligent identification method based on deep learning comprises the following steps: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using the training sample and the label to obtain a network model for intelligently detecting faults in the three-dimensional seismic data; carrying out data standardization and data augmentation on the parameters of the network model, selecting an optimized loss function, and improving the fault identification capability of the network model; inputting real seismic data into the network model for prediction, evaluating a prediction result, and adjusting and perfecting a training sample.
Further, the method for generating the fault and the label corresponding to the fault by forward modeling comprises the following steps:
using two types of vertical shear displacement fields S 1 (x, y, z) and S 2 (x, y, z) transforming imaging of seismic data from planar space to folded space with different curved structures by the two types of vertical shear displacement fields;
and simulating a fault structure in a geological fold structure model by adopting a volume vector field, and simulating various generalized faults by randomly selecting fault parameters.
Further, the simulating various generalized faults through randomly selected fault parameters comprises:
simulating a fracture in the seismic data;
giving the longest diameter along the fault trend and the minimum diameter along the fault inclination direction at the section, determining a displacement field on a fault plane, and constructing fault displacement;
constructing fault plane disturbance, and disturbing a fault plane to obtain a real fault plane;
and (3) extrapolating the displacement field from the fault plane to construct a fault volume displacement field so as to estimate inclined displacement components of an upper plate and a lower plate of the fault nearby the fault plane, thereby generating the fault.
Further, the simulating fracture comprises:
randomly selecting a reference point (X) 0 ,Y 0 ,Z 0 ) Trend towards
And the dip angle theta parameter is used for converting the reference point from the global coordinate to the local coordinate and realizing the simulation of fracture in the seismic data.
Further, the constructing a fault plane disturbance comprises:
m disturbance points are randomly generated near the fault, and a smooth and curved surface is interpolated from the disturbance points by adopting a bi-harmonic spline interpolation method.
Further, the extrapolating the displacement field from the fault plane to construct a fault volume displacement field to estimate the inclined displacement components of the upper wall and the lower wall of the fault around the vicinity of the fault plane to generate the fault comprises:
respectively calculating corresponding tendency displacement according to any point in the upper fault plate and any point in the lower fault plate of the displacement field;
inverting the displacement signs of the upper wall and the lower wall of the fault to generate a reverse fault;
calculating the normal displacement in the upper disc and the lower disc of the fault;
and transforming the inverse fault into fold coordinates through the inclination displacement and the normal displacement to generate a fault.
Further, the transforming the inverse fault into wrinkle coordinates by the dip displacement and the normal displacement, generating a fault, comprising:
converting the fold coordinates into plane local coordinates;
loading the dip displacement and the normal displacement of the fault in the plane local coordinates, and converting the plane local coordinates back to wrinkle coordinates.
A fault intelligent recognition system based on deep learning comprises: the system comprises a first processing module, a second processing module and a third processing module, wherein the first processing module generates a fault and a label corresponding to the fault according to input seismic data, constructs a fault sample library, and provides a training sample and the label thereof from the fault sample library; the second processing module is used for constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by the training sample and the label to obtain a network model for intelligently detecting faults in the three-dimensional seismic data; the optimization module is used for carrying out data standardization and data augmentation on parameters of the network model, selecting an optimized loss function and improving the fault recognition capability of the network model; and the intelligent identification module inputs real seismic data into the network model for prediction, evaluates the prediction result and adjusts and perfects the training sample.
A computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the above methods.
A computing device, comprising: one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the above-described methods.
Due to the adoption of the technical scheme, the invention has the following advantages:
1. the invention generates a large amount of diversified training data which accord with the actual situation by a forward modeling method, and simultaneously constructs a complete training sample library by combining the explained fault result. On the basis, an optimized and simple three-dimensional convolution neural network model is designed to efficiently process a large three-dimensional seismic data volume and obtain an accurate fault detection result, and further matched filtering scanning processing is carried out on the fault detection result to obtain an enhanced fault probability volume, fault tendency and trend estimation.
2. The invention can greatly improve the anti-interference capability of the network by introducing actual data noise into the training sample.
In conclusion, the fault detection result based on deep learning is superior to the conventional fault detection method based on seismic attributes in continuity, noise immunity and resolution.
Drawings
FIG. 1 is a schematic flow chart of a seismic fault detection method based on deep learning according to an embodiment of the present invention;
FIG. 2a is a schematic diagram of a composition of training samples for constructing a construction pattern and features in forward modeling considering actual data according to an embodiment of the present invention;
FIG. 2b illustrates the addition of noise in consideration of the actual work area environment in one embodiment of the present invention;
FIG. 3 is a simplified structural model of a convolutional neural network for three-dimensional seismic fault detection in accordance with an embodiment of the present invention;
FIG. 4a is a three-dimensional seismic data volume of fault detection results in an embodiment of the invention;
FIG. 4b is a cross-sectional view of the computed coherence for fault detection in an embodiment of the present invention;
FIG. 4c illustrates the planarity of the fault detection results in accordance with an embodiment of the present invention;
FIG. 4d is a table showing the variance of the fault detection result according to an embodiment of the present invention;
FIG. 4e is a diagram of an embodiment of the ant body for fault detection;
FIG. 4f shows the fault detection result of the deep learning method according to an embodiment of the present invention, wherein the arrows indicate that the conventional method cannot detect a significant fault with surface wave reflection characteristics.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings of the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the described embodiments of the invention, are within the scope of the invention.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Training a convolutional neural network for fault interpretation requires a large amount of seismic data and corresponding fault interpretation results. The workload of acquiring a large amount of training data about faults by a manual interpretation method is huge, the subjectivity is strong, and different interpreters can obtain different interpretation results. And if the fault interpretation result used for training is unreliable, the training of the convolutional neural network and the final fault prediction accuracy are influenced. Therefore, in the embodiment, a method for automatically generating training samples is adopted to obtain a large number of synthetic seismic images and corresponding fully accurate fault calibration is adopted to train the convolutional neural network for fault interpretation. On this basis, it will be modified appropriately to obtain a convolutional neural network for training the fault direction estimation and fault plane construction. Meanwhile, a forward modeling method is further corrected by considering a specific fault mode in actual seismic data to obtain a fault training sample which is more consistent with an actual work area. In addition, the existing manually calibrated fault results are added into a training sample library.
The invention provides a fault intelligent identification method and system based on deep learning, which comprises the following steps: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using a training sample and a label to obtain a network model for intelligently detecting faults in three-dimensional seismic data; carrying out data standardization and data augmentation on parameters of the network model, and selecting an optimized loss function to improve fault recognition capability of the network model; inputting the real seismic data into a network model for prediction, evaluating the prediction result, and adjusting and perfecting the training sample. The fault detection result based on deep learning is superior to the conventional fault detection method based on seismic attributes in continuity, noise immunity and resolution.
In one embodiment of the invention, a fault intelligent identification method based on deep learning is provided. In this embodiment, as shown in fig. 1, the method includes the following steps:
1) Generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library;
2) Constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using a training sample and a label to obtain a network model for intelligently detecting faults in three-dimensional seismic data;
3) Carrying out data standardization and data augmentation on parameters of the network model, and selecting an optimized loss function to improve fault recognition capability of the network model;
4) Inputting the real seismic data into a network model for prediction, evaluating the prediction result, and adjusting and perfecting the training sample.
In the step 1), a forward modeling method is adopted for generating the fault and the label corresponding to the fault, and the method comprises the following steps:
1.1 Using two types of vertical shear displacement fields S 1 (x, y, z) and S 2 (x, y, z) transforming the imaging of the seismic data from planar space to folded space with different curved structures by two types of vertical shear displacement fields;
the method specifically comprises the following steps:
wherein,
representing the point after the origin local coordinate system (x, y, z) is converted into the wrinkle space.
1.2 ) simulating fault structure in a geological fold structure model by using a volume vector field, and simulating various generalized faults by randomly selecting fault parameters.
In the step 1.2), various generalized faults are simulated through randomly selected fault parameters, and the method comprises the following steps:
1.2.1 Simulating fractures in the seismic data;
1.2.2 Given the longest diameter lx along the fault trend and the smallest diameter ly along the fault inclination direction at the section, determining the displacement field on the fault plane and constructing fault displacement;
in the present embodiment, the displacement field d (X, Y, Z = 0) on the tomographic plane is defined as the origin (X, Y, Z) of the center point (X, Y, Z) 0 ,Y 0 ,Z 0 ) The ideal elliptic function is as follows:
wherein lx, ly and the maximum displacement d of the center point max Are all randomly selected, r (x, y) =1.
1.2.3 Constructing fault plane disturbance, and disturbing a fault plane to obtain a real fault plane;
1.2.4 To construct a fault volume displacement field from the fault plane extrapolated displacement field to estimate the dip displacement components around the upper and lower walls of the fault near the fault plane, thereby generating the fault.
In the step 1.2.1), the method for simulating fracture comprises the following steps:
randomly selecting a reference point (X) 0 ,Y 0 ,Z 0 ) Trend of the Chinese character
And the dip angle theta parameter, converting the reference point from the global coordinate (X, Y, Z) to the local coordinate (X, Y, Z), and realizing the simulation of fracture in the seismic data; the method comprises the following specific steps:
wherein theta is the strike angle,
is an inclination angle. Wherein each fault is associated with a datum plane defined by the strike angle box dip.
1.2.1.1 A translation reference point (X) 0 ,Y 0 ,Z 0 ) A coordinate point of (2);
1.2.1.2 Rotating transformation is carried out by adopting a rotation matrix rotation coordinate, and the simulated fracture is realized.
In the step 1.2.3), fault plane disturbance is constructed to obtain a more real fault plane, and the specific method comprises the following steps: m disturbance points are randomly generated near the fault, and a smooth and curved surface is interpolated from the disturbance points by adopting a bi-harmonic spline interpolation method.
In the step 1.2.4), a fault volume displacement field is constructed by extrapolating a displacement field from the fault plane to estimate inclined displacement components of an upper plate and a lower plate of the fault around the vicinity of the fault plane, thereby generating the fault. Wherein the displacement decreases in the z-axis direction away from the fault surface by the amount of displacement and reaches zero at the reverse drag radius γ.
The method specifically comprises the following steps:
1.2.4.1 Respectively calculating corresponding tendency displacement according to the displacement field for any point on the upper wall of the fault and any point on the lower wall of the fault;
at any point on the disk in the fault, the trend displacement D is calculated by the following formula y (x,y,z)：
D y (x,y,z)＝λ·d(x,y；z＝0)·α(x,y,z)
Wherein α is a non-linear scalar function that gradually decreases away from the cross-section along the z-axis direction: the fault plane is defined as the function z = f (x, y) of all (x, y) on the reference plane. The value of f is the deviation from the reference plane along the z-axis.
And for any point in the fault footwall, the displacement D is inclined y The formula for the calculation of (x, y, z) is:
D y (x,y,z)＝(λ-1)·d(x,y；z＝0)·α(x,y,z)
where λ represents the displacement ratio of the upper and lower disks in the fault.
1.2.4.2 Inverting the displacement signs of the upper wall and the lower wall of the fault to generate an inverse fault;
1.2.4.3 Normal displacements in the top and bottom wall of the fault are calculated since the fault is usually curved and therefore must also be non-zero in the z direction to ensure seamless movement of fault blocks along the fault surface;
normal displacement D z (x, y, z) is:
D z (x,y,z)＝f(x,y+D y (x,y,z))-f(x,y)
the above formula describes the change in fault plane with respect to the dip displacement. The z-direction displacement always follows the extension trend of the fault plane because a higher displacement amount is observed near the region where the fault surface has a higher curvature.
1.2.4.4 The inverse fault is converted into fold coordinates by the dip displacement and the normal displacement, and a fault is generated.
In the step 1.2.4.4), the inverse fault is converted into fold coordinates through the dip displacement and the normal displacement, and a fault is generated, specifically: the volume displacement field D is defined in a local coordinate system y (x, y, z) and D x (x,y,z)(D x (x, y, z) = 0), after which the fault configuration is generated on the basis of the wrinkle structure model by:
(1) Coordinate the fold from
Conversion to planar local coordinates (x, y, z);
(2) The dip displacement and the normal displacement of the fault are loaded in the plane local coordinates (x, y, z), and the plane local coordinates are converted back to wrinkle coordinates.
Specifically, the dip displacement and the normal displacement loading the fault in the plane local coordinates (x, y, z) are:
transforming the planar local coordinates back to the wrinkle coordinates as:
in use, a series of seismic data (as shown in figure 2 a) is synthesised using the method described above. When a fault is generated in the synthetic model, a fault displacement distribution is defined as a gaussian function or a linear function. The maximum fault displacement of each fault is randomly selected between 0 and 40 sampling points. Images with more faults can train the convolutional neural network more effectively for fault segmentation than images with fewer faults. Therefore, 5 or more slices are added to a 128 × 128 × 128 training image. The reflection model is convolved with wavelets after (rather than before) the folds and faults are created in the model, because the convolution blurs sharp discontinuities near the faults, thus making the faults appear more realistic. To further improve the realism of the composite seismic image, some random noise is also added to the image as shown in figure 2 b.
In the step 2), a convolutional neural network based on image segmentation is constructed, and the convolutional neural network is trained by training samples and labels, wherein the training samples are automatically generated.
Specifically, the construction method of the convolutional neural network comprises the following steps: the detection of the position of the seismic data (or image) break layer is regarded as a problem of binary image segmentation, namely, a sampling point of the position of the seismic data break layer is marked as 1, and all other sampling points are marked as 0. In order to realize the binary image segmentation about the fault, a U-net convolution neural network (as shown in FIG. 3) which has been greatly successful in medical image segmentation is adopted and simplified and optimized to efficiently and accurately detect the fault. Since image segmentation is an end-to-end or graph-to-graph process, computational efficiency is very high.
In the step 3), data standardization and data augmentation are carried out on the parameters of the network model, and an optimized loss function is selected to improve the fault recognition capability of the network model. In this embodiment, the optimization of the network model parameters is as follows:
when the three-dimensional synthetic seismic image is used as training data, the data is standardized. Training data can be linearly mapped to the 0-1 interval by data normalization. The data standardization can make the model training process more stable, and can reduce the calculated amount under the condition of not changing the numerical value sequencing in the training data. In the model training process, a smaller fixed learning rate is adopted, and the stability of the training process is ensured. And a learning rate scheduling scheme is added in due time according to the training condition at the later stage, and the optimization process of the model is more accurate by dynamically adjusting the learning rate.
In order to improve the generalization of the model and obtain more effective training data at lower cost, the embodiment performs data augmentation on the training data. The main form of the method is to rotate the training data and the corresponding sample labels by 0 degrees, 90 degrees, 180 degrees and 270 degrees around the Z axis so as to obtain additional effective training data and further improve the generalization of the model.
In the step 4), the real seismic data (shown in fig. 4 a) is input into the network model for prediction, the prediction result (shown in fig. 4 f) is evaluated, and the training sample is adjusted and perfected. The method specifically comprises the following steps: compared with the most common conventional methods in the industry, including coherence properties (shown in fig. 4 b) and fault planarity, variance and ant body (shown in fig. 4c, 4d and 4 e), the method provided by the invention is proved to be superior to the conventional fault detection method based on seismic properties in continuity, noise resistance and resolution.
In one embodiment of the present invention, a fault intelligent recognition system based on deep learning is provided, which includes:
the first processing module is used for generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library and providing a training sample and the label thereof by the fault sample library;
the second processing module is used for constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using the training samples and the labels to obtain a network model for intelligently detecting faults in the three-dimensional seismic data;
the optimization module is used for carrying out data standardization and data augmentation on parameters of the network model, selecting an optimized loss function and improving the fault recognition capability of the network model;
and the intelligent identification module inputs real seismic data into the network model for prediction, evaluates the prediction result and adjusts and perfects the training sample.
The system provided in this embodiment is used for executing the above method embodiments, and for details of the process and the details, reference is made to the above embodiments, which are not described herein again.
In the computing device structure provided in an embodiment of the present invention, the computing device may be a terminal, and the computing device may include: a processor (processor), a communication Interface (communication Interface), a memory (memory), a display screen and an input device. The processor, the communication interface and the memory are communicated with each other through a communication bus. The processor is used to provide computing and control capabilities. The memory comprises a nonvolatile storage medium and an internal memory, wherein the nonvolatile storage medium stores an operating system and a computer program, and the computer program is executed by the processor to realize the fault intelligent identification method based on deep learning; the internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a manager network, NFC (near field communication) or other technologies. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on a shell of the computing equipment, an external keyboard, a touch pad or a mouse and the like. The processor may call logic instructions in the memory to perform the following method: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using a training sample and a label to obtain a network model for intelligently detecting faults in three-dimensional seismic data; data standardization and data augmentation are carried out on parameters of the network model, an optimized loss function is selected, and fault recognition capability of the network model is improved; inputting the real seismic data into a network model for prediction, evaluating the prediction result, and adjusting and perfecting the training sample.
In addition, the logic instructions in the memory may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention or a part thereof which substantially contributes to the prior art may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Those skilled in the art will appreciate that the above-described configurations of computing devices are merely some of the configurations associated with the present application and do not constitute limitations on the computing devices to which the present application may be applied, as a particular computing device may include more or fewer components, or some components in combination, or have a different arrangement of components.
In one embodiment of the invention, a computer program product is provided, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, enable the computer to perform the methods provided by the above-described method embodiments, for example, comprising: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using a training sample and a label to obtain a network model for intelligently detecting faults in three-dimensional seismic data; data standardization and data augmentation are carried out on parameters of the network model, an optimized loss function is selected, and fault recognition capability of the network model is improved; inputting the real seismic data into a network model for prediction, evaluating the prediction result, and adjusting and perfecting the training sample.
In one embodiment of the invention, a non-transitory computer-readable storage medium is provided, which stores server instructions that cause a computer to perform the methods provided by the above embodiments, for example, including: generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library; constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using a training sample and a label to obtain a network model for intelligently detecting faults in three-dimensional seismic data; carrying out data standardization and data augmentation on parameters of the network model, and selecting an optimized loss function to improve fault recognition capability of the network model; inputting the real seismic data into a network model for prediction, evaluating the prediction result, and adjusting and perfecting the training sample.
The implementation principle and technical effect of the computer-readable storage medium provided by the above embodiments are similar to those of the above method embodiments, and are not described herein again.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A fault intelligent identification method based on deep learning is characterized by comprising the following steps:
generating a fault and a label corresponding to the fault according to input seismic data, constructing a fault sample library, and providing a training sample and the label thereof by the fault sample library;
constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by using the training samples and the labels to obtain a network model for intelligently detecting faults in the three-dimensional seismic data;
carrying out data standardization and data augmentation on the parameters of the network model, selecting an optimized loss function, and improving the fault identification capability of the network model;
inputting real seismic data into the network model for prediction, evaluating a prediction result, and adjusting and perfecting a training sample.
2. The fault intelligent identification method based on deep learning as claimed in claim 1, characterized in that the fault generation and fault corresponding label adopt forward modeling method, which includes:
using two types of vertical shear displacement fields S 1 (x, y, z) and S 2 (x, y, z) transforming imaging of seismic data from planar space to folded space with different curved structures by the two types of vertical shear displacement fields;
and simulating a fault structure in a geological fold structure model by adopting a volume vector field, and simulating various generalized faults by randomly selecting fault parameters.
3. The intelligent fault identification method based on deep learning as claimed in claim 2, wherein the simulation of various generalized faults through randomly selected fault parameters comprises:
simulating a fracture in the seismic data;
giving the longest diameter along the fault trend and the smallest diameter along the fault inclination direction at the section, determining a displacement field on a fault plane, and constructing fault displacement;
constructing fault plane disturbance, and disturbing a fault plane to obtain a real fault plane;
and (3) extrapolating the displacement field from the fault plane to construct a fault volume displacement field so as to estimate inclined displacement components of an upper plate and a lower plate of the fault nearby the fault plane, thereby generating the fault.
4. The intelligent fault identification method based on deep learning of claim 3, wherein the simulated fracture comprises:
randomly selecting a reference point (X) 0 ,Y 0 ,Z 0 ) Trend of the Chinese character
And the dip angle theta parameter is used for converting the reference point from the global coordinate to the local coordinate and realizing the simulation of fracture in the seismic data.
5. The intelligent fault identification method based on deep learning of claim 3, wherein constructing fault plane disturbance comprises:
m disturbance points are randomly generated near the fault, and a smooth and curved surface is interpolated from the disturbance points by adopting a bi-harmonic spline interpolation method.
6. The intelligent fault identification method based on deep learning as claimed in claim 3, wherein the step of constructing a fault volume displacement field by extrapolating a displacement field from a fault plane to estimate inclined displacement components of an upper plate and a lower plate of faults around the fault plane near the fault plane so as to generate the fault comprises the following steps:
respectively calculating corresponding tendency displacement according to any point in the upper fault plate and any point in the lower fault plate of the displacement field;
inverting the displacement signs of the upper wall and the lower wall of the fault to generate a reverse fault;
calculating the normal displacement in the upper disc and the lower disc of the fault;
and transforming the inverse fault into fold coordinates by the inclination displacement and the normal displacement to generate a fault.
7. The intelligent fault identification method based on deep learning of claim 6, wherein the step of transforming the inverse fault into fold coordinates through the tendency displacement and the normal displacement to generate the fault comprises the following steps:
converting the fold coordinate into a plane local coordinate;
loading the dip displacement and the normal displacement of the fault in the plane local coordinates, and converting the plane local coordinates back to wrinkle coordinates.
8. The utility model provides a fault intelligent recognition system based on deep learning which characterized in that includes:
the system comprises a first processing module, a second processing module and a third processing module, wherein the first processing module generates a fault and a label corresponding to the fault according to input seismic data, constructs a fault sample library, and provides a training sample and the label thereof from the fault sample library;
the second processing module is used for constructing a convolutional neural network based on image segmentation, and training the convolutional neural network by the training samples and the labels to obtain a network model for intelligently detecting faults in the three-dimensional seismic data;
the optimization module is used for carrying out data standardization and data augmentation on the parameters of the network model, selecting an optimized loss function and improving the fault identification capability of the network model;
and the intelligent identification module inputs real seismic data into the network model for prediction, evaluates the prediction result, and adjusts and perfects the training sample.
9. A computer readable storage medium storing one or more programs, wherein the one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-7.
10. A computing device, comprising: one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210779693.3A CN115201902A (en) | 2022-07-04 | 2022-07-04 | Fault intelligent identification method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210779693.3A CN115201902A (en) | 2022-07-04 | 2022-07-04 | Fault intelligent identification method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115201902A true CN115201902A (en) | 2022-10-18 |
Family
ID=83577257
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210779693.3A Pending CN115201902A (en) | 2022-07-04 | 2022-07-04 | Fault intelligent identification method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115201902A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115639605A (en) * | 2022-10-28 | 2023-01-24 | 中国地质大学（武汉） | Automatic high-resolution fault identification method and device based on deep learning |
| CN115657132A (en) * | 2022-12-29 | 2023-01-31 | 成都捷科思石油天然气技术发展有限公司 | Reservoir prediction method and system based on machine learning |
| CN116821642A (en) * | 2023-06-13 | 2023-09-29 | 北京建筑大学 | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning |
-
2022
- 2022-07-04 CN CN202210779693.3A patent/CN115201902A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115639605A (en) * | 2022-10-28 | 2023-01-24 | 中国地质大学（武汉） | Automatic high-resolution fault identification method and device based on deep learning |
| CN115639605B (en) * | 2022-10-28 | 2024-05-28 | 中国地质大学（武汉） | Automatic identification method and device for high-resolution fault based on deep learning |
| CN115657132A (en) * | 2022-12-29 | 2023-01-31 | 成都捷科思石油天然气技术发展有限公司 | Reservoir prediction method and system based on machine learning |
| CN115657132B (en) * | 2022-12-29 | 2023-03-10 | 成都捷科思石油天然气技术发展有限公司 | Reservoir prediction method and system based on machine learning |
| CN116821642A (en) * | 2023-06-13 | 2023-09-29 | 北京建筑大学 | Building earthquake damage rapid assessment method and system based on data augmentation and deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115201902A (en) | Fault intelligent identification method and system based on deep learning | |
| Liu et al. | Neural subdivision | |
| US20110115787A1 (en) | Visulation of geologic features using data representations thereof | |
| CN103091395B (en) | Method and device for three-dimensionally dynamically detecting fluid phase state in porous medium under reservoir conditions | |
| CN107221028A (en) | A kind of geologic body closed surface three-dimensional rebuilding method based on seismic interpretation data | |
| CN107705359A (en) | The method and apparatus for obtaining wireless channel model | |
| Galley et al. | Geophysical inversion for 3D contact surface geometry | |
| CN107204040A (en) | Multiple-Point Geostatistics modeling method and device, computer-readable storage medium | |
| US10943037B2 (en) | Generating a CAD model from a finite element mesh | |
| EP2803043B1 (en) | 3-d surface-based waveform inversion | |
| Peng et al. | Higher Dimensional Vector Field Visualization: A Survey. | |
| Wang et al. | EQSM: An efficient high quality surface grid generation method based on remeshing | |
| CN104637075A (en) | Automatic rapid mapping method for sand body | |
| Wang | Computing length-preserved free boundary for quasi-developable mesh segmentation | |
| WO2017152123A1 (en) | Context based bounded hydrocarbon formation identification | |
| Bi et al. | Improving fault surface construction with inversion-based methods | |
| CN115639605B (en) | Automatic identification method and device for high-resolution fault based on deep learning | |
| US20160245951A1 (en) | Method for Parameterizing a 3D Domain With Discontinuities | |
| Jacobson | Part II: Automatic Skinning via Constrained Energy Optimization | |
| CN115984503A (en) | Geological profile generation method, geological profile generation system, electronic equipment and medium | |
| Doloi et al. | Super-resolution reconstruction of reservoir saturation map with physical constraints using generative adversarial network | |
| Varela et al. | Software architecture of an interface for three-dimensional collision handling in maritime Virtual Environments | |
| CN203299170U (en) | Three-dimensional dynamic detection device for phase states of fluid in porous medium under oil deposit conditions | |
| Pan et al. | HeterSkinNet: A heterogeneous network for skin weights prediction | |
| Wu et al. | Real‐Time Symmetry‐Preserving Deformation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |