CN115576014B - Crack type reservoir intelligent identification method based on acoustic wave far detection imaging - Google Patents
Crack type reservoir intelligent identification method based on acoustic wave far detection imaging Download PDFInfo
- Publication number
- CN115576014B CN115576014B CN202211317841.6A CN202211317841A CN115576014B CN 115576014 B CN115576014 B CN 115576014B CN 202211317841 A CN202211317841 A CN 202211317841A CN 115576014 B CN115576014 B CN 115576014B
- Authority
- CN
- China
- Prior art keywords
- network
- imaging
- field
- wave
- segmentation
- 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.)
- Active
Links
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
- G01V1/40—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging
- G01V1/44—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging using generators and receivers in the same well
- G01V1/48—Processing data
- G01V1/50—Analysing data
-
- 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
- G01V2210/61—Analysis by combining or comparing a seismic data set with other data
- G01V2210/616—Data from specific type of measurement
- G01V2210/6169—Data from specific type of measurement using well-logging
-
- 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
- G01V2210/62—Physical property of subsurface
- G01V2210/624—Reservoir parameters
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Landscapes
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- Remote Sensing (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Image Processing (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a crack type reservoir intelligent identification method based on acoustic wave far detection imaging, belonging to the technical field of geophysical acoustic logging application; the method comprises the following steps: step 1: preparing a theoretical simulation data set; step 2: manually labeling a field data set; step 3: constructing a segmentation network model; step 4: training a neural network; step 5: network knowledge migration; step 6: in-situ imaging image processing. According to the invention, the crack reflector in the imaging diagram is intelligently identified through the image segmentation angle, a segmentation network model is constructed, an end-to-end segmentation network is trained by using analog data, and the segmentation network is migrated and learned to an actual data domain, so that the crack type reservoir in the imaging diagram can be automatically identified under the condition of strong noise interference.
Description
Technical Field
The invention belongs to the technical field of geophysical acoustic logging, and particularly relates to an intelligent identification method for a fracture type reservoir based on acoustic wave remote detection imaging.
Background
With the exploration and development of unconventional oil and gas in recent years, more and more attention is paid to the identification of fractured reservoirs. Acoustic long-range detection techniques can use the reflection information to image the geologic structure outside of the well with high resolution. However, both monopole and dipole acoustic sources excite and generate direct mode waves, such as sliding longitudinal waves, sliding transverse waves, stoneley waves, bending waves, and the like, propagating along the borehole wall near the borehole, which interfere with the measurement of the reflected information. Therefore, in the process of processing the acoustic far detection signal, the direct signal needs to be suppressed by a reflected wave extraction method. However, under the actual measurement condition, the amplitude of the reflected wave is weak, and aliasing is generated between the time domain and the direct wave in the well, so that the effect of the reflected wave extraction method is reduced, and a large amount of residual signals still exist in the wave field after processing. This portion of the residual signal, after being processed by a subsequent offset imaging algorithm, will appear as irregular coherent noise in the imaging map.
The Chinese patent discloses a noise reduction method for an acoustic wave far detection imaging chart (application number: 202210845762.6, application bulletin number: CN 115170428A), which comprises the following steps: step 1: preparing a data set; step 2: constructing a noise reduction network model; step 3: preprocessing data; step 4: training a network model; step 5: in-situ imaging image processing. And removing interference noise in the imaging graph by constructing a noise reduction neural network model.
The noise reduction method for the acoustic wave far detection imaging chart has the following defects: there are also vertically distributed residual direct wave artefacts near the borehole that interfere with accurate identification of the effective reflectors. How to weaken or eliminate the influence of direct wave false images on reflector identification and the explanation of effective reservoirs in an imaging diagram still needs to be solved, so that the popularization and application of the acoustic wave far detection technology can be improved.
Disclosure of Invention
Aiming at the defects, the invention provides an intelligent identification method for a fracture type reservoir based on acoustic wave far detection imaging.
The invention aims at realizing the following steps: a crack type reservoir intelligent identification method based on acoustic wave far detection imaging is characterized in that: the method comprises the following steps:
step 1: preparing a theoretical simulation data set;
step 2: manually labeling a field data set;
step 3: constructing a segmentation network model;
step 4: training a neural network;
step 5: network knowledge migration;
step 6: in-situ imaging image processing.
Preferably, the specific operation of preparing the theoretical simulation dataset in step 1 is as follows:
step 1-1: setting the size, the longitudinal and transverse wave speed and the density value of a uniform stratum serving as a background medium, setting the longitudinal and transverse wave speed and the density value of a non-uniform medium, determining the size, the shape and the position of the non-uniform medium by using a random function, and superposing a non-uniform region on the uniform stratum to construct a geological model;
step 1-2: carrying out rapid calculation on a scattering sound field of the well-side fracture model by using a numerical method based on the Born approximation;
step 1-3: returning the time domain waveform obtained in the step 1-2 to the declining direction of the inclined phase shaft by using an offset imaging algorithm to obtain an offset imaging diagram of the well periphery geological structure;
step 1-4: setting a background medium in the geological model as 0, and setting a non-uniform area as 1 to obtain a corresponding label image;
step 1-5: repeating the steps 1-1, 1-2, 1-3 and 1-4 to obtain a large number of paired labels and offset imaging graphs, and establishing a large-scale theoretical simulation data set.
Preferably, the specific operation of preparing the manually labeled field data set in the step 2 is as follows: and marking the collected actual imaging images, and marking only continuous and inclined reflectors in the marking process to generate a label image only comprising 0 and 1, wherein a pixel 1 represents the reflector and a pixel 0 represents the background medium.
Preferably, the step 3 of constructing a segmentation network model includes the following steps:
step 3-1: building an encoder network: constructing an encoder network by adopting Effecientenet-b 4 as a backbone structure, wherein the encoder network is divided into 6 stages, and the output characteristics of each stage are { E } 1 ，E 2 ，E 3 ，E 4 ，E 5 ，E 6 }；
Step 3-2: building a decoder network: output E to encoder network by nearest neighbor interpolation 6 Up-sampling and then combining with E 4 Residual connection is carried out to obtain an output characteristic D 4 Further up-sampling the output stage by stage, and then respectively mixing with { E } 2 ，E 3 Residual connection is carried out to obtain output characteristics { D }, respectively 2 ，D 3 }；
Step 3-3: building a segmentation module: for the output characteristics { D } of the encoder and decoder, respectively 3 ，D 4 ，E 6 2,4,8 times up-sampling operation, unifying the feature sizes to the same size, and then combining it with feature D 2 4 times of up-sampling is carried out after the joint superposition, and finally, a prediction segmentation diagram is obtained after convolution kernel filtering with the size of 1 multiplied by 1 and the step length of 1 multiplied by 1 is utilized;
step 3-4: and combining the encoder network, the decoder network module and the segmentation module in the steps 3-1,3-2 and 3-3 to build a feature pyramid network for the subsequent image segmentation task.
Preferably, the specific operation of training the neural network in the step 4 is:
step 4-1: dividing data by adopting a 10-fold cross validation method, namely firstly dividing 10% of the data set into 10 mutually exclusive subsets with the same size as a test set, dividing the rest data into 10 mutually exclusive subsets with the same size, taking each subset as a validation set and the other subsets as training sets once, and taking an average value as an evaluation result after repeating ten times;
step 4-2: randomly sampling from a training set according to the set batch sampling number, and inputting a label graph and an offset imaging graph into the segmentation neural network model established in the step 3;
step 4-3: normalizing the image:
wherein μ represents the mean value of the data, σ is the standard deviation of the data;
step 4-4: in the training process, carrying out a series of random data enhancement processing on input data, wherein the random data enhancement processing comprises transverse and vertical overturning, and adding 0-30 dB white Gaussian noise;
step 4-5: the joint network outputs the label data corresponding to the label data, and a loss function is calculated by using a Dice coefficient:
wherein Y is a label, f (theta) is a predictive segmentation map of the model, and n in the molecule represents an overlapping part in the former two;
step 4-6: the calculated Dice loss function is back propagated by utilizing an Adam algorithm, and the neural network parameters are updated;
step 4-7: repeating steps 4-2 to 4-6 to continuously optimize the network until all data in the training set are sampled, and calculating lou coefficients on the verification set:
step 4-8: repeating the steps 4-7 until the iteration times exceed a set threshold, or the average lou coefficient increment value on the verification set in the set iteration times does not exceed the set threshold;
step 4-9: re-executing the step 4-8 by using the Lovsz-high function as a loss function, and performing reinforcement training on the network trained by the Dice coefficient to overcome the gradient vanishing phenomenon faced by the network trained by the Dice coefficient;
step 4-10: and applying the trained network to a test data set, and evaluating the generalization performance of the network.
Preferably, the specific operation of the network knowledge migration in step 5 is:
step 5-1: initializing a new encoder network and a new decoder network in the FPN network by utilizing the parameters of the trained network obtained in the step 4, and randomly initializing parameters of the segmentation module;
step 5-2: and (3) setting a smaller learning rate by using an Adam optimizer, and performing network optimization on the field data set established in the step (2) by using the Dice coefficient in the step (4-2) as a loss function.
Preferably, the specific operation of the in-situ imaging map processing in step 6 is as follows:
step 6-1: solidifying and storing the FPN network model parameters trained in the step 5;
step 6-2: slicing the field imaging diagram to be processed, inputting the field imaging diagram to the FPN network stored in the step 6-1, outputting a corresponding prediction segmentation diagram, and finally merging all the output reflectors in the field imaging diagram and intelligently identifying the reflectors.
Preferably, the specific operation of performing rapid calculation on the scattering sound field of the fracture model beside the well by using a numerical method based on the Born approximation in the step 1-2 is as follows:
step 1-2-1: under weak scattering conditions, the local inhomogeneity ζ (x) < 1 caused by the inhomogeneous medium velocity variations, which together with the incident wave field controls the propagation of the scattered wave, the equation is as follows:
where u is the scalar displacement potential, V 0 Background velocity for non-uniform media;
step 1-2-2: based on the Bern approximation principle, the total sound field can be equivalently the incident field u 0 And a scattered field u 1 Is superimposed on the (d):
u＝u 0 +u 1 (2)
wherein the incident field satisfies the homogeneous medium equation:
step 1-2-3: neglecting infinitely small amounts of higher order
The relation equation of the scattered field and the incident field can be obtained by combining the formulas (1), (2) and (3):
step 1-2-4: further deducing scattering sound field in the non-uniform area by using green function and superposition principle as follows:
where x' is a point in the inhomogeneous medium region V and G is the incident wave field u 0 The corresponding green's function, R= |x-x' | represents the distance between two points; when the scattered sound field calculation is carried out, firstly, the non-uniform region is discretized into smaller volume units, and then the sound fields of all the subunits are overlapped through a time domain integral equation method to obtain the scattered sound field of the non-uniform medium.
Preferably, the specific operation of obtaining the imaging chart including only the reflector by using the offset imaging method in the step 1-3 is as follows:
step 1-3-1: back-propagating the waveform data obtained in step 1-2 to the offset domain using Kirchhoff's equation, at a point ζ in the offset domain 0 At t 0 The back-propagation sound field expression for time is:
in the method, in the process of the invention,
represents the boundary of the offset domain Ω, where +.>
Is the derivative of the omega outer normal direction;
step 1-3-2 Using Gaussian beam G GB (ζ,t；ζ 0 ,t 0 ) The green's function in (6) is replaced by the superimposed asymptotic form to obtain a point ζ in the offset domain 0 At t 0 Time-of-day back-propagation wavefield expression:
zeta in well s The wave equation of the direct wave field of the excitation of the sound source at the position is:
where f (t) is the acoustic excitation function of the logging instrument, typically a Rake wavelet; the expression of the direct wavefield obtained by the same derivation step as the counter-propagating wavefield described above is:
wherein Re represents the complex number taking part, f F (omega) is a frequency domain representation of the excitation function, G GB (ζ,ζ s The method comprises the steps of carrying out a first treatment on the surface of the ω) represents the secondary sound source point ζ s A Gaussian cluster to a point ζ in the offset domain is an asymptotic solution of the green's function in the wave equation;
asymptotic solution G of Green's function GB (ζ,ζ 0 The method comprises the steps of carrying out a first treatment on the surface of the ω) is represented by a series of principal points ζ 0 The single Gaussian beam rays are emitted and pass through a Gaussian beam composition of an imaging point zeta, wherein the expression of the single Gaussian beam rays under a ray center coordinate system is as follows:
s is the arc length of the ray, n is the distance from one point outside the ray to the ray, v(s) is the speed on the ray, p, q is the dynamic parameter, and τ(s) is the arrival time on the central ray; by superposing Gaussian beams in different emergent directions, an asymptotic solution of the green function is obtained:
where Φ (θ; ω) is the initial amplitude of the gaussian beam with an exit angle θ, and the formula is as follows:
transforming the green's function of the frequency domain representation to the time domain:
step 1-3-3, solving the Gaussian beam by using a formula (10), then calculating a direct wave field by using a formula (11) and a formula (9), further calculating a counter-propagating wave field by using a formula (13) and a formula (7), and then calculating the cross-correlation of the two wave fields for imaging, wherein the formula is as follows:
I(ζ；ζ s )＝∫dt 0 u D (ζ,t 0 ；ζ s )u(ζ,t 0 ) (14)
by calculating the direct wave field u D (ζ,t 0 ；ζ s ) With counter-propagating wave fields u (ζ, t 0 ) Correlation values between the two to image the geological structure outside the well.
Preferably, the specific operation of building the encoder network in the step 3-1 is as follows:
step 3-1-1: adopting an Effecientenet-b 4 network as a backbone, wherein the structure is a continuous convolution structure, and the input of each stage is the output of the last stage;
step 3-1-2: the core component of the above-mentioned Effecientenet-b 4 network is a mobile convolution module (MBConv), which firstly uses a convolution check input feature of 1×1 to perform channel expansion, the expansion coefficient is r, and then uses a batch normalization layer (BatchNorm) to perform feature normalization, and then uses a Swish activation function to perform nonlinear mapping:
Swish(x)＝x·sigmoid(βx) (15)
where β is a constant or trainable parameter, typically set to a constant of 1.
Step 3-1-3: in order to reduce the parameter quantity of the network, the characteristic of the expanded channels is filtered by adopting depth separable convolution, each channel of the input characteristic is respectively processed by utilizing two-dimensional convolution check, the convolution diagram of each channel is obtained, the output values of a plurality of channels are mapped into a single channel by utilizing 1X 1 convolution, and the richer multi-level characteristic is extracted on the premise that the calculated quantity is slightly lower than that of a standard convolution kernel with the same size.
Step 3-1-4: the method comprises the steps of further refining features by using a channel attention mechanism, firstly compressing the space dimension of the features by using global average pooling operation, and changing the space dimension into one-dimensional features; and then the one-dimensional feature is subjected to channel compression by using a convolution kernel, then the one-dimensional feature is expanded to an original channel by using an extraction operation, and finally the one-dimensional feature is multiplied by an input feature.
Step 3-1-5: and scaling the characteristics processed by the channel attention mechanism to an original input channel by using a 1 multiplied by 1 projection convolution kernel, and then performing residual connection with the original input channel to inhibit the gradient vanishing phenomenon of the deep network.
The invention has the beneficial effects that: 1. the simulation data set is utilized to train the network and is migrated to the actual data set, so that the problem that the actual data set is rare and the effective training of the network is difficult to support is effectively solved; meanwhile, a split network model is built, and the split network model is transferred and learned to an actual data domain, so that a crack type reservoir in an imaging diagram can be automatically identified under the condition of strong noise interference.
2. The invention trains the end-to-end reflector identification network without complex parameter setting and expert knowledge support, can overcome noise interference in an imaging diagram, intelligently identifies the reflectors therein, and is favorable for field popularization and application of the acoustic wave remote detection technology.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is an offset imaging map of a stratigraphic model.
Fig. 3 is a label diagram corresponding to the stratum model.
FIG. 4 is a collected field offset imaging map.
Fig. 5 is a manually noted result of an offset-in-place imaging map.
Fig. 6 is a schematic diagram of the architecture of the effeeientnet-b 4 backbone network.
Fig. 7 is a schematic diagram of a constructed split network model.
Fig. 8 is a schematic diagram of a mobile convolution module.
Fig. 9 is a training set learning graph.
Fig. 10 is a verification set learning graph.
FIG. 11 is a graph of the results of a process of training a network for in situ imaging.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described with reference to the accompanying drawings and specific embodiments. The specific embodiments described herein are to be considered in an illustrative sense only and are not intended to limit the invention. The invention is further summarized in connection with the accompanying drawings.
As shown in fig. 1, the crack type reservoir intelligent identification method based on acoustic wave far detection imaging comprises the following steps:
step 1: preparing a theoretical simulation data set;
step 1-1: setting uniform stratum as background medium, with size of 20x20m, longitudinal wave velocity of 6220m/s, transverse wave velocity of 3208m/s, and density of 2700kg/m 3 The method comprises the steps of carrying out a first treatment on the surface of the The longitudinal wave speed of the crack type heterogeneous medium is 4010m/s, the transverse wave speed is 2102m/s, and the density is 2650kg/m 3 The method comprises the steps of carrying out a first treatment on the surface of the Overlapping the non-uniform region on the uniform stratum to construct a geological model;
step 1-2: carrying out rapid calculation on a scattering sound field of the well-side fracture model by using a numerical method based on the Born approximation;
step 1-2-1: under weak scattering conditions, the local inhomogeneity ζ (x) < 1 caused by the inhomogeneous medium velocity variations, which together with the incident wave field controls the propagation of the scattered wave, the equation is as follows:
where u is the scalar displacement potential, V 0 Background velocity for non-uniform media;
step 1-2-2: based on the Bern approximation principle, the total sound field can be equivalently the incident field u 0 And a scattered field u 1 Is superimposed on the (d):
u＝u 0 +u 1 (2)
wherein the incident field satisfies the homogeneous medium equation:
step 1-2-3: neglecting infinitely small amounts of higher order
The relation equation of the scattered field and the incident field can be obtained by combining the formulas (1), (2) and (3):
step 1-2-4: further deducing scattering sound field in the non-uniform area by using green function and superposition principle as follows:
where x' is a point in the inhomogeneous medium region V and G is the incident wave field u 0 The corresponding green's function, R= |x-x' | represents the distance between two points; when the scattered sound field calculation is carried out, firstly, the non-uniform region is discretized into smaller volume units, and then the sound fields of all the subunits are overlapped through a time domain integral equation method to obtain the scattered sound field of the non-uniform medium.
Step 1-3: returning the time domain waveform obtained in the step to the declining direction of the inclined phase shaft by utilizing an offset imaging algorithm to obtain a tag imaging diagram of the well periphery geological structure;
step 1-3-1: back-propagating the waveform data obtained in step 1-2 to the offset domain using Kirchhoff's equation, at a point ζ in the offset domain 0 At t 0 The back-propagation sound field expression for time is:
in the method, in the process of the invention,
represents the boundary of the offset domain Ω, where +.>
Is the derivative of the omega outer normal direction;
step 1-3-2 Using Gaussian beam G GB (ζ,t；ζ 0 ,t 0 ) The green's function in (6) is replaced by the superimposed asymptotic form to obtain a point ζ in the offset domain 0 At t 0 Time-of-day back-propagation wavefield expression:
zeta in well s The wave equation of the direct wave field of the excitation of the sound source at the position is:
where f (t) is the acoustic excitation function of the logging instrument, typically a Rake wavelet; the expression of the direct wavefield obtained by the same derivation step as the counter-propagating wavefield described above is:
wherein Re represents the complex number taking part, f F (omega) is a frequency domain representation of the excitation function, G GB (ζ,ζ s The method comprises the steps of carrying out a first treatment on the surface of the ω) represents the secondary sound source point ζ s A Gaussian cluster to a point ζ in the offset domain is an asymptotic solution of the green's function in the wave equation;
asymptotic solution G of Green's function GB (ζ,ζ 0 The method comprises the steps of carrying out a first treatment on the surface of the ω) is represented by a series of principal points ζ 0 The single Gaussian beam rays are emitted and pass through a Gaussian beam composition of an imaging point zeta, wherein the expression of the single Gaussian beam rays under a ray center coordinate system is as follows:
s is the arc length of the ray, n is the distance from one point outside the ray to the ray, v(s) is the speed on the ray, p, q is the dynamic parameter, and τ(s) is the arrival time on the central ray; by superposing Gaussian beams in different emergent directions, an asymptotic solution of the green function is obtained:
where Φ (θ; ω) is the initial amplitude of the gaussian beam with an exit angle θ, and the formula is as follows:
transforming the green's function of the frequency domain representation to the time domain:
step 1-3-3, solving the Gaussian beam by using a formula (10), then calculating a direct wave field by using a formula (11) and a formula (9), further calculating a counter-propagating wave field by using a formula (13) and a formula (7), and then calculating the cross-correlation of the two wave fields for imaging, wherein the formula is as follows:
l(ζ；ζ s )＝∫dt 0 u D (ζ,t 0 ；ζ s )u(ζ,t 0 ) (14)
by calculating the direct wave field u D (ζ,t 0 ；ζ s ) With counter-propagating wave fields u (ζ, t 0 ) The correlation values between them are used to image the geological structure outside the well, and the results are shown in fig. 2.
Step 1-4: setting a background medium in the geological model to 0, and setting a non-uniform area to 1 to obtain a corresponding label image, as shown in fig. 3;
step 1-5: repeating the steps 1-1, 1-2, 1-3 and 1-4 to obtain 15000 paired tag images and corresponding offset imaging images, and constructing a large-scale theoretical simulation data set.
Step 2: manually labeling a field data set;
step 2-1: gathering a field imaging map as shown in fig. 4;
step 2-2: and (3) manually marking the field imaging diagram, wherein in the marking process, only continuous and inclined reflectors are marked, and a label image only comprising 0 and 1 is generated, wherein a pixel 1 represents the reflector, a pixel 0 is a background medium, and the result is shown in fig. 5.
Step 2-3: slicing the field imaging image and the corresponding label image to obtain a slice with the size of 256 multiplied by 256;
step 2-4: repeating the steps 2-1, 2-2 and 2-3 to obtain 1315 Zhang Chengxiang diagram and manual label slice, and obtaining the manual labeling field data set.
Step 3: constructing a segmentation network model;
step 3-1: constructing an encoder network;
step 3-1-1: the encoder network is built by using Effecientenet-b 4 as a backbone structure, and the structure comprises 6 stages, corresponding output characteristics are { E }, respectively, as shown in figure 6 1 ，E 2 ，E 3 ，E 4 ，E 5 ，E 6 ]. The method comprises the steps of performing downsampling operation by adopting a standard convolution kernel in the first stage, wherein the first stage comprises 2 groups of mobile convolution modules with the size of 3 multiplied by 3 and the size of 24 channels, and 4 groups of mobile convolution modules with the size of 3 multiplied by 3 and the size of 32 channels, the third stage comprises 4 groups of mobile convolution modules with the size of 5 multiplied by 5 and the size of 56 channels, the fourth stage comprises 6 groups of mobile convolution modules with the size of 3 multiplied by 3 and the size of 112 channels, the fifth stage comprises 6 groups of mobile convolution modules with the size of 5 multiplied by 5 and the size of 160 channels, the sixth stage comprises 8 groups of mobile convolution modules with the size of 5 multiplied by 5, the size of 272 channels and 2 groups of mobile convolution modules with the size of 3 multiplied by 3 and the size of 448 channels;
step 3-1-2: the mobile convolution module (MBConv) is a core module of the above-mentioned Effecientnet-b4 backbone network, as shown in fig. 8, and it firstly uses a convolution kernel of 1×1 to perform channel expansion, the expansion coefficient r is set to 6, then uses a batch normalization layer (batch norm) to perform feature normalization, and then uses a Swish activation function to perform nonlinear mapping:
Swish(x)＝x·sigmoid(βx) (15)
where β is a constant or a trainable parameter, set to a constant of 1.
Step 3-1-3: in order to reduce the parameter quantity of the network, the characteristic of the expanded channels is filtered by adopting depth separable convolution, each channel of the input characteristic is respectively processed by utilizing two-dimensional convolution check, the convolution diagram of each channel is obtained, the output values of a plurality of channels are mapped into a single channel by utilizing 1X 1 convolution, and the richer multi-level characteristic is extracted on the premise that the calculated quantity is slightly lower than that of a standard convolution kernel with the same size.
Step 3-1-4: the method comprises the steps of further extracting features by using a channel attention mechanism, firstly compressing the space dimension of the features by using global average pooling, and changing the space dimension into one-dimensional features; then the one-dimensional feature is compressed by a convolution kernel with a compression ratio of 8, then the one-dimensional feature is expanded to an original channel by an extraction operation, and finally multiplied by an input feature.
Step 3-1-5: and scaling the characteristics processed by the channel attention mechanism to an original input channel by using a 1 multiplied by 1 projection convolution kernel, and then performing residual connection with the original input channel to inhibit the gradient vanishing phenomenon of the deep network.
Step 3-2: building a decoder network: output E to encoder network by nearest neighbor interpolation 6 Up-sampling and then combining with E 4 Residual connection is carried out to obtain an output characteristic D 4 Further up-sampling the output stage by stage, and then respectively mixing with { E } 2 ，E 3 Residual connection is carried out to obtain output characteristics { D }, respectively 2 ，D 3 }；
Step 3-3: building a segmentation module: for the output characteristics { D } of the encoder and decoder, respectively 3 ，D 4 ，E 6 2,4,8 times up sampling operation is carried out to unify the characteristic dimensionsTo the same size and then match it with feature D 2 4 times of up-sampling is carried out after the joint superposition, and finally, a prediction segmentation diagram is obtained after convolution kernel filtering with the size of 1 multiplied by 1 and the step length of 1 multiplied by 1 is utilized;
step 3-4: and combining all the modules in the steps 3-1,3-2 and 3-3, and constructing a feature pyramid network shown in fig. 7 for the subsequent image segmentation task.
Step 4: training a network model;
step 4-1: the data is divided by adopting a 10-fold cross validation method, namely, firstly 10% of the data is divided into a test set, the rest data is divided into 10 mutually exclusive subsets with the same size, each subset is used as a validation set at one time, the other subsets are used as training sets, and the average value is used as an evaluation result after ten times of repetition.
Step 4-2: randomly sampling from a training set according to the set batch sampling number, and inputting a label graph and an offset imaging graph into the segmentation neural network model established in the step 3;
step 4-3: normalizing the image:
wherein μ represents the mean value of the data, σ is the standard deviation of the data;
step 4-4: in the training process, a series of random data enhancement processing is carried out on input data, wherein the random data enhancement processing comprises transverse and vertical overturning, and 0-30 dB of white Gaussian noise is added.
Step 4-5: the joint network outputs the label data corresponding to the label data, and a loss function is calculated by using a Dice coefficient:
where Y is a label, f (θ) is a predictive segmentation map of the model, and ∈in the molecule indicates the overlap in the former two.
Step 4-6: the calculated Dice loss function is back propagated by utilizing an Adam algorithm, and the neural network parameters are updated;
step 4-7: repeating steps 4-2 to 4-6 to continuously optimize the network until all data in the training set are sampled, and calculating the Dice loss function and lou coefficient on the verification set:
step 4-8: repeating the steps 4-7 until the iteration times exceed a set threshold, or the average Iou coefficient increment value on the verification set in the set iteration times does not exceed the set threshold; as shown by the dashed line in fig. 9, the Dice coefficient on the validation set drops rapidly from 1 until it drops to 0.02 after the 55 th round, and the corresponding lou coefficient is given by the dashed line in fig. 10, and finally the Iou coefficient reaches 0.936.
Step 4-9: re-executing the step 4-8 by using the Lovsz-high function as a loss function, and performing reinforcement training on the network trained by the Dice coefficient to overcome the gradient vanishing phenomenon faced by the network trained by the Dice coefficient; the results are shown in the implementations of fig. 9 and 10, with a Iou coefficient on the validation set eventually reaching 0.947.
Step 4-10: and applying the trained network to a test data set, and evaluating the generalization performance of the network.
Step 5: network knowledge migration:
step 5-1: and (3) initializing the encoder and decoder networks in the new FPN network by utilizing the parameters of the trained network obtained in the step (4), and randomly initializing the parameters of the segmentation module.
Step 5-2: and (3) setting a smaller learning rate by using an Adam optimizer, and performing network optimization on the field data set established in the step (2) by taking the Dice coefficient adopted in the step (4-5) as a loss function.
Step 6: in-situ imaging image processing:
step 6-1: performing parameter curing and storage on the FPN network model trained in the step 5;
step 6-2: slicing the field imaging image to be processed, inputting the field imaging image to the network to generate a corresponding prediction segmentation image, and finally merging all the output reflectors in the field imaging image to be processed and intelligently identifying the reflectors in the field imaging image. As shown in fig. 11, the third path is the evaluation result of the network trained in step 4, which can learn only the low-level semantics of the amplitude abnormality of the reflector, but cannot effectively identify other coherent noise. The fourth is the evaluation result of the network trained in the step 5, so that higher-level semantics can be learned, the difference between random noise and the reflector can be accurately identified, the prediction result is more continuous, and the position, the inclination angle and other information of the reflector can be accurately evaluated on the basis of the continuous prediction result of the reflector.
Working principle: the acoustic wave remote detection logging technology utilizes an acoustic source in a well to excite elastic waves to an underground stratum, the elastic waves are collected by a receiver in the well after being reflected by an underground non-uniform body, and further, a high-resolution imaging diagram of an underground geological structure is obtained through an offset imaging algorithm.
The foregoing description is only illustrative of the invention and is not to be construed as limiting the invention. Various modifications and variations of the present invention will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, or the like, which is within the spirit and principles of the present invention, should be included in the scope of the claims of the present invention.
Claims (8)
1. A crack type reservoir intelligent identification method based on acoustic wave far detection imaging is characterized in that: the method comprises the following steps:
step 1: preparing a theoretical simulation data set;
step 2: manually labeling a field data set;
step 3: constructing a segmentation network model;
the construction of the segmentation network model comprises the following steps:
step 3-1: building an encoder network: constructing an encoder network by adopting Effecientenet-b 4 as a backbone structure, wherein the encoder network is divided into 6 stages, and the output characteristics of each stage are { E } 1 ，E 2 ，E 3 ，E 4 ，E 5 ，E 6 }；
Step 3-2: building a decoder network: output E to encoder network by nearest neighbor interpolation 6 Up-sampling and then combining with E 4 Residual connection is carried out to obtain an output characteristic D 4 Further up-sampling the output stage by stage, and then respectively mixing with { E } 2 ，E 3 Residual connection is carried out to obtain output characteristics { D }, respectively 2 ，D 3 }；
Step 3-3: building a segmentation module: for the output characteristics { D } of the encoder and decoder, respectively 3 ，D 4 ，E 6 2,4,8 times up-sampling operation, unifying the feature sizes to the same size, and then combining it with feature D 2 4 times of up-sampling is carried out after the joint superposition, and finally, a prediction segmentation diagram is obtained after convolution kernel filtering with the size of 1 multiplied by 1 and the step length of 1 multiplied by 1 is utilized;
step 3-4: combining the encoder network, the decoder network module and the segmentation module in the steps 3-1,3-2 and 3-3 to build a feature pyramid network for subsequent image segmentation tasks;
step 4: training a neural network;
the specific operation of training the neural network is as follows:
step 4-1: dividing data by adopting a 10-fold cross validation method, namely firstly dividing 10% of the data set into 10 mutually exclusive subsets with the same size as a test set, dividing the rest data into 10 mutually exclusive subsets with the same size, taking each subset as a validation set and the other subsets as training sets once, and taking an average value as an evaluation result after repeating ten times;
step 4-2: randomly sampling from a training set according to the set batch sampling number, and inputting a label graph and an offset imaging graph into the segmentation neural network model established in the step 3;
step 4-3: normalizing the image:
wherein μ represents the mean value of the data, σ is the standard deviation of the data;
step 4-4: in the training process, carrying out a series of random data enhancement processing on input data, wherein the random data enhancement processing comprises transverse and vertical overturning, and adding 0-30 dB white Gaussian noise;
step 4-5: the joint network outputs the label data corresponding to the label data, and a loss function is calculated by using a Dice coefficient:
wherein Y is a label, f (theta) is a predictive segmentation map of the model, and n in the molecule represents an overlapping part in the former two;
step 4-6: the calculated Dice loss function is back propagated by utilizing an Adam algorithm, and the neural network parameters are updated;
step 4-7: repeating steps 4-2 to 4-6 to continuously optimize the network until all data in the training set are sampled, and calculating Iou coefficients on the verification set:
step 4-8: repeating the steps 4-7 until the iteration times exceed a set threshold, or the average Iou coefficient increment value on the verification set in the set iteration times does not exceed the set threshold;
step 4-9: re-executing the step 4-8 by using the Lovsz-high function as a loss function, and performing reinforcement training on the network trained by the Dice coefficient to overcome the gradient vanishing phenomenon faced by the network trained by the Dice coefficient;
step 4-10: applying the trained network to a test data set, and evaluating the generalization performance of the network;
step 5: network knowledge migration;
step 6: in-situ imaging image processing.
2. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 1, is characterized in that: the specific operation of preparing the theoretical simulation dataset in said step 1 is:
step 1-1: setting the size, the longitudinal and transverse wave speed and the density value of a uniform stratum serving as a background medium, setting the longitudinal and transverse wave speed and the density value of a non-uniform medium, determining the size, the shape and the position of the non-uniform medium by using a random function, and superposing a non-uniform region on the uniform stratum to construct a geological model;
step 1-2: carrying out rapid calculation on a scattering sound field of the well-side fracture model by using a numerical method based on the Born approximation;
step 1-3: returning the time domain waveform obtained in the step 1-2 to the declining direction of the inclined phase shaft by using an offset imaging algorithm to obtain an offset imaging diagram of the well periphery geological structure;
step 1-4: setting a background medium in the geological model as 0, and setting a non-uniform area as 1 to obtain a corresponding label image;
step 1-5: repeating the steps 1-1, 1-2, 1-3 and 1-4 to obtain a large number of paired labels and offset imaging graphs, and establishing a large-scale theoretical simulation data set.
3. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 1, is characterized in that: the specific operation of preparing the manual labeling field data set in the step 2 is as follows:
and marking the collected actual imaging images, and marking only continuous and inclined reflectors in the marking process to generate a label image only comprising 0 and 1, wherein a pixel 1 represents the reflector and a pixel 0 represents the background medium.
4. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 1, is characterized in that: the specific operation of the network knowledge migration in the step 5 is as follows:
step 5-1: initializing a new encoder network and a new decoder network in the FPN network by utilizing the parameters of the trained network obtained in the step 4, and randomly initializing parameters of the segmentation module;
step 5-2: and (3) setting a smaller learning rate by using an Adam optimizer, and performing network optimization on the field data set established in the step (2) by using the Dice coefficient in the step (4-2) as a loss function.
5. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 1, is characterized in that: the specific operation of the field imaging diagram processing in the step 6 is as follows:
step 6-1: solidifying and storing the FPN network model parameters trained in the step 5;
step 6-2: slicing the field imaging diagram to be processed, inputting the field imaging diagram to the FPN network stored in the step 6-1, outputting a corresponding prediction segmentation diagram, and finally merging all the output reflectors in the field imaging diagram and intelligently identifying the reflectors.
6. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 2, is characterized in that: the specific operation of rapidly calculating the scattering sound field of the well-side fracture model by using a numerical method based on the Born approximation in the step 1-2 is as follows:
step 1-2-1: under weak scattering conditions, local inhomogeneities |ζ (x) | < 1 caused by inhomogeneous medium velocity variations, which together with the incident wave field control the propagation of the scattered wave, the equation is as follows:
where u is the scalar displacement potential, V 0 Background velocity for non-uniform media;
step 1-2-2: based on the Bern approximation principle, the diffuse sound field is equivalent to the incident field u 0 And a scattered field u 1 Is superimposed on the (d):
u＝u 0 +u 1 . (2)
wherein the incident field satisfies the homogeneous medium equation:
step 1-2-3: neglecting infinitely small amounts of higher order
The relation equation of the scattered field and the incident field is obtained by combining the formulas (1), (2) and (3):
step 1-2-4: further deducing scattering sound field in the non-uniform area by using green function and superposition principle as follows:
where x' is a point in the inhomogeneous medium region V and G is the incident wave field u 0 The corresponding green's function, R= |x-x' | represents the distance between two points; when the scattered sound field calculation is carried out, firstly, the non-uniform area is discretized into smaller volume units, and then the sound fields of all the subunits are overlapped through a time domain integral equation method to obtain the scattered sound field of the non-uniform medium.
7. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 2, is characterized in that: the specific operation of obtaining the imaging diagram only comprising the reflector by using the offset imaging method in the step 1-3 is as follows:
step 1-3-1: back-propagating the waveform data obtained in step 1-2 to the offset domain using Kirchhoff's equation, at a point ζ in the offset domain 0 At t 0 The back-propagation sound field expression for time is:
in the method, in the process of the invention,
represents the boundary of the offset domain Ω, where +.>
Is the derivative of the omega outer normal direction;
step 1-3-2 Using Gaussian beam G GB (ζ,t；ζ 0 ,t 0 ) The green's function in (6) is replaced by the superimposed asymptotic form to obtain a point ζ in the offset domain 0 At t 0 Time-of-day back-propagation wavefield expression:
zeta in well s The wave equation of the direct wave field of the excitation of the sound source at the position is:
wherein f (t) is an acoustic excitation function of the logging instrument and is a Rake wavelet; the expression of the direct wavefield obtained by the same derivation step as the counter-propagating wavefield described above is:
wherein Re represents a complexNumber of the real part f F (omega) is a frequency domain representation of the excitation function, G GB (ζ,ζ s The method comprises the steps of carrying out a first treatment on the surface of the ω) represents the secondary sound source point ζ s A Gaussian cluster to a point ζ in the offset domain is an asymptotic solution of the green's function in the wave equation;
asymptotic solution G of Green's function GB (ζ,ζ 0 The method comprises the steps of carrying out a first treatment on the surface of the ω) is represented by a series of principal points ζ 0 The single Gaussian beam rays are emitted and pass through a Gaussian beam composition of an imaging point zeta, wherein the expression of the single Gaussian beam rays under a ray center coordinate system is as follows:
s is the arc length of the ray, n is the distance from one point outside the ray to the ray, v(s) is the speed on the ray, p, q is the dynamic parameter, and τ(s) is the arrival time on the central ray; by superposing Gaussian beams in different emergent directions, an asymptotic solution of the green function is obtained:
where Φ (θ; ω) is the initial amplitude of the gaussian beam with an exit angle θ, and the formula is as follows:
transforming the green's function of the frequency domain representation to the time domain:
step 1-3-3, solving the Gaussian beam by using a formula (10), then calculating a direct wave field by using a formula (11) and a formula (9), further calculating a counter-propagating wave field by using a formula (13) and a formula (7), and then calculating the cross-correlation of the two wave fields for imaging, wherein the formula is as follows:
I(ζ；ζ s )＝∫dt 0 u D (ζ,t 0 ；ζ s )u(ζ,t 0 ) (14)
by calculating the direct wave field u D (ζ,t 0 ；ζ s ) With counter-propagating wave fields u (ζ, t 0 ) Correlation values between the two to image the geological structure outside the well.
8. The intelligent identification method for the fractured reservoir based on acoustic wave far detection imaging, which is disclosed in claim 1, is characterized in that: the specific operation of building the encoder network in the step 3-1 is as follows:
step 3-1-1: adopting an Effecientenet-b 4 network as a backbone, wherein the structure is a continuous convolution structure, and the input of each stage is the output of the last stage;
step 3-1-2: the core component of the Effecientenet-b 4 network is a mobile convolution module (MBConv), firstly, a 1X 1 convolution is utilized to check input characteristics to expand channels, the expansion coefficient is r, then, a batch normalization layer (BatchNorm) is utilized to normalize the characteristics, and then, a Swish activation function is utilized to carry out nonlinear mapping:
Swish(x)＝x·sigmoid(βx) (15)
wherein, beta is a constant or a trainable parameter, and beta is set to be a constant 1;
step 3-1-3: filtering the characteristics of the channels after expansion by adopting depth separable convolution, respectively processing each channel of input characteristics by utilizing a two-dimensional convolution check, mapping output values of a plurality of channels into a single channel by utilizing 1X 1 convolution after obtaining a convolution map of each channel, and extracting more abundant multi-level characteristics on the premise that the calculated amount is slightly lower than a standard convolution kernel with the same size;
step 3-1-4: the method comprises the steps of further refining features by using a channel attention mechanism, firstly compressing the space dimension of the features by using global average pooling operation, and changing the space dimension into one-dimensional features; then, the one-dimensional feature is subjected to channel compression by utilizing a convolution kernel, then the one-dimensional feature is expanded to an original channel by utilizing extraction operation, and finally, the one-dimensional feature is multiplied with the input feature;
step 3-1-5: and scaling the characteristics processed by the channel attention mechanism to an original input channel by using a 1 multiplied by 1 projection convolution kernel, and then performing residual connection with the original input channel to inhibit the gradient vanishing phenomenon of the deep network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211317841.6A CN115576014B (en) | 2022-10-26 | 2022-10-26 | Crack type reservoir intelligent identification method based on acoustic wave far detection imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211317841.6A CN115576014B (en) | 2022-10-26 | 2022-10-26 | Crack type reservoir intelligent identification method based on acoustic wave far detection imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115576014A CN115576014A (en) | 2023-01-06 |
| CN115576014B true CN115576014B (en) | 2023-07-11 |
Family
ID=84586287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211317841.6A Active CN115576014B (en) | 2022-10-26 | 2022-10-26 | Crack type reservoir intelligent identification method based on acoustic wave far detection imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115576014B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116069035B (en) * | 2023-02-10 | 2023-09-05 | 济宁鲁科检测器材有限公司 | Laser radar-based flaw detection robot weld path recognition device and method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111913217A (en) * | 2020-08-10 | 2020-11-10 | 中海石油(中国)有限公司 | Fracture type stratum seismic scattering wave field characteristic simulation method based on imaginary number domain Rytov approximation |
| WO2021258096A1 (en) * | 2020-06-19 | 2021-12-23 | Exxonmobil Upstream Research Company | Classifying geologic features in seismic data through image analysis by neural network |
| WO2022052367A1 (en) * | 2020-09-10 | 2022-03-17 | 中国科学院深圳先进技术研究院 | Neural network optimization method for remote sensing image classification, and terminal and storage medium |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8594962B2 (en) * | 2009-05-28 | 2013-11-26 | Westerngeco L.L.C. | Distributing a clock in a subterranean survey data acquisition system |
| EP3338115A4 (en) * | 2015-08-18 | 2019-04-24 | Services Petroliers Schlumberger | Reservoir simulations with fracture networks |
| CN113534251A (en) * | 2020-04-22 | 2021-10-22 | 中国石油天然气股份有限公司 | Seismic data fracture identification method and device based on convolutional neural network |
| CN111580166B (en) * | 2020-05-21 | 2022-03-18 | 中国石油大学(华东) | Fracture effectiveness evaluation method based on sound wave remote detection and rock mechanics |
| CN114609682A (en) * | 2020-12-09 | 2022-06-10 | 中国石油天然气股份有限公司 | Method and system for extracting parameters of electrical imaging seam hole and computer readable storage medium |
| CN112711072A (en) * | 2020-12-23 | 2021-04-27 | 西安交通大学 | Res U-net-based three-dimensional seismic data fault identification method |
| CN113156515A (en) * | 2021-04-16 | 2021-07-23 | 中国石油大学(华东) | Acoustic wave remote detection imaging noise reduction processing method and device |
| CN114140415A (en) * | 2021-11-25 | 2022-03-04 | 长江大学 | Ultrasonic logging image crack extraction method and device based on deep learning |
| CN115170428A (en) * | 2022-07-18 | 2022-10-11 | 江苏科技大学 | Noise reduction method for acoustic wave remote detection imaging graph |
| CN115166831A (en) * | 2022-07-28 | 2022-10-11 | 中国石油大学(华东) | Acoustic logging method for guiding fracturing and sidetracking operations |
-
2022
- 2022-10-26 CN CN202211317841.6A patent/CN115576014B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021258096A1 (en) * | 2020-06-19 | 2021-12-23 | Exxonmobil Upstream Research Company | Classifying geologic features in seismic data through image analysis by neural network |
| CN111913217A (en) * | 2020-08-10 | 2020-11-10 | 中海石油(中国)有限公司 | Fracture type stratum seismic scattering wave field characteristic simulation method based on imaginary number domain Rytov approximation |
| WO2022052367A1 (en) * | 2020-09-10 | 2022-03-17 | 中国科学院深圳先进技术研究院 | Neural network optimization method for remote sensing image classification, and terminal and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115576014A (en) | 2023-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stork et al. | Application of machine learning to microseismic event detection in distributed acoustic sensing data | |
| CN111562611B (en) | Semi-supervised depth learning seismic data inversion method based on wave equation drive | |
| CN112232392B (en) | Data interpretation and identification method for three-dimensional ground penetrating radar | |
| CN111164462B (en) | Artificial source surface wave exploration method, surface wave exploration device and terminal equipment | |
| Van den Ende et al. | A self-supervised deep learning approach for blind denoising and waveform coherence enhancement in distributed acoustic sensing data | |
| CN107688201B (en) | RBM-based seismic prestack signal clustering method | |
| EA004551B1 (en) | Method for seismic facies interpretation using textural analysis and neutral networks | |
| Sun et al. | Deep learning for low-frequency extrapolation of multicomponent data in elastic FWI | |
| CN115576014B (en) | Crack type reservoir intelligent identification method based on acoustic wave far detection imaging | |
| CN112731522B (en) | Intelligent recognition method, device and equipment for seismic stratum and storage medium | |
| US20220206176A1 (en) | System and method for identifying subsurface structures | |
| CN116400426B (en) | Electromagnetic method-based data survey system | |
| CN110515122B (en) | Forward grid search positioning and micro-seismic signal identification method and device | |
| Huang et al. | Machine learning using U-net convolutional neural networks for the imaging of sparse seismic data | |
| CN111461988A (en) | Seismic velocity model super-resolution technology based on multi-task learning | |
| CN104730572A (en) | Diffracted wave imaging method and device based on L0 semi-norm | |
| CN108398719A (en) | The processing method and processing device of seismic data | |
| CN116256722B (en) | Method and device for suppressing multiple interference of B-scan image of ground penetrating radar | |
| US20160034612A1 (en) | Re-ordered Interpolation and Convolution for Faster Staggered-Grid Processing | |
| CN116660996A (en) | Drifting type shallow sea local earth sound parameter prediction method based on deep learning | |
| CN115170428A (en) | Noise reduction method for acoustic wave remote detection imaging graph | |
| CN116934603A (en) | Logging curve missing segment completion method and device, storage medium and electronic equipment | |
| Chen et al. | Seismic ahead-prospecting based on deep learning of retrieving seismic wavefield | |
| US20220397691A1 (en) | System and method for reducing statics in seismic imaging | |
| Almansouri et al. | Development of acoustic model-based iterative reconstruction technique for thick-concrete imaging |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |