CN110441820A - Architectonic intelligent interpretation method - Google Patents
Architectonic intelligent interpretation method Download PDFInfo
- Publication number
- CN110441820A CN110441820A CN201910777505.1A CN201910777505A CN110441820A CN 110441820 A CN110441820 A CN 110441820A CN 201910777505 A CN201910777505 A CN 201910777505A CN 110441820 A CN110441820 A CN 110441820A
- Authority
- CN
- China
- Prior art keywords
- architectonic
- neural networks
- data
- convolutional
- layer
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000010276 construction Methods 0.000 claims abstract description 24
- 238000003062 neural network model Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims 1
- 238000007405 data analysis Methods 0.000 abstract description 2
- 230000004913 activation Effects 0.000 description 5
- 239000003245 coal Substances 0.000 description 4
- 238000003325 tomography Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000001427 coherent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 102000008169 Co-Repressor Proteins Human genes 0.000 description 1
- 108010060434 Co-Repressor Proteins Proteins 0.000 description 1
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 description 1
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000299 nuclear matrix Anatomy 0.000 description 1
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
- G01V1/30—Analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Remote Sensing (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Geophysics (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geology (AREA)
- Environmental & Geological Engineering (AREA)
- Acoustics & Sound (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The present invention provides a kind of architectonic intelligent interpretation methods, it is related to seismic data analysis technical field, it include: the preparatory explanation results using original seismic amplitude data, label data body is generated by the method to different types of structure assignment, as the training set of convolutional neural networks, convolutional neural networks model prediction geological structure is constructed.The correlation between layer position and variety classes construction is considered, the layer position at construction is explained and is more clear, it is more accurate to the geologic interpretation of different configuration.
Description
Technical field
The present invention relates to seismic data analysis technical field, especially a kind of architectonic intelligent interpretation method.
Background technique
With the continuous development of coal resources exploration, the required precision of seismic data interpretation is higher and higher.Initially, layer position solves
It releases and structure interpretation is completed using manual manual interpretation, this method is geologic knowledge of the explanation personnel according to oneself
It is explained with experience, not only time-consuming but also lacks certain objectivity.
In view of the above-mentioned problems, current a solution is to be identified using full convolutional neural networks to fault plane,
Seismic amplitude sectional slice data are obtained from 3-D seismics amplitude data body, and slice of data is calibrated by manual interpretation method
On position of fault, using this slice of data as label data.This method does not have using slice of data as label data
Representativeness, it is not accurate to architectonic explanation results.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of architectonic intelligent interpretation methods, to solve existing skill
The problem of art.
In a first aspect, a kind of architectonic intelligent interpretation method is provided, method includes the following steps:
The original seismic amplitude data explained in advance according to construction is in advance based on, determines the explanation knot of the structure interpretation
Fruit generates the 3D data volume of the structure interpretation result;
Based on the explanation results of the construction, the type of the structure interpretation result is determined;
Based on the type of the structure interpretation result, assignment is carried out to the 3D data volume of the structure interpretation result, it is raw
At the label data body of the structure interpretation result;
According to the label data body and original seismic amplitude data of the structure interpretation result, training set is established;
Predetermined initial convolution neural network model is trained using the training set, the volume after being trained
Product neural network model, it is pre- in order to be carried out using the convolutional neural networks model after the training to geological structure to be predicted
It surveys.
Above-mentioned architectonic intelligent interpretation method proposed by the present invention, utilizes the preparatory explanation of original seismic amplitude data
As a result, label data body is generated by the method to different types of structure assignment, as the training set of convolutional neural networks, structure
Build convolutional neural networks model prediction geological structure.The correlation between layer position and variety classes construction is considered, makes to construct
The layer position at place is explained and is more clear, more accurate to the geologic interpretation of different configuration.
Further, the original seismic amplitude data is the coalfield actual seismic with a variety of typical geology structural types
Data, and the data of the accurate result of manual interpretation in advance;The manual interpretation result includes: layer position, construction and background.
Further, the different types of structure assignment of 3D data volume progress to the explanation results refers to: will
The label value of layer position is assigned to 1, and the label value of construction is assigned to 2, and by background, other without the label value at construction are assigned to 0, generate institute
State label data body;The background refers at no construction.
Further, the training set includes: the label data body and the original seismic amplitude data.
Further, the convolutional neural networks model is Three dimensional convolution neural network model.
Further, the building convolutional neural networks model includes: selection initiation parameter;Utilize the original earthquake
Amplitude data and the label data body training convolutional neural networks model;It minimizes loss function and obtains final training
Model.
Further, the convolutional neural networks include three convolutional layers, are used for feature extraction;
The convolution kernel size of first convolutional layer is 3 × 3 × 3, step-length 2;The convolution of second convolutional layer and third convolutional layer
Core size is 3 × 3 × 3, step-length 1;
Three convolutional layers apply compensation line to activate.
Further, the convolutional neural networks further include two pond layers, and the feature that the convolutional layer is extracted carries out
Selection and filtering;
The size of first pond layer is 2 × 2 × 2, and step-length 2 is set to after first convolutional layer;
The size of second pond layer is 2 × 2 × 2, and step-length 2 is set to after second convolutional layer.
Further, the convolutional neural networks further include a full articulamentum and a softmax classifier;
The full articulamentum, is activated using linearity correction, is set to after the third convolutional layer;
The softmax classifier is set to after the full articulamentum, for generating trained and prediction result.
Second aspect provides a kind of machine readable storage medium, and the machine readable storage medium is stored with machine can
It executes instruction, when being called and being executed by processor, the machine-executable instruction promotes described the machine-executable instruction
Processor realizes method described in first aspect.
Above-mentioned architectonic intelligent interpretation method proposed by the present invention utilizes the artificial preparatory of coalfield actual seismic data
Explanation results generate label data body by the method to different types of structure assignment, the training as convolutional neural networks
Collection constructs Three dimensional convolution Neural Network model predictive geological structure.Trained convolutional neural networks model is set to be more suitable coalfield
Real data, it is contemplated that the correlation between layer position and variety classes construction explains the layer position at construction more accurate;Instruction
The convolutional neural networks model practised can be used directly to the prediction of other any coalfield data, can be quickly and accurately right
Geological structure carries out automatic interpretation.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of geologic structure interpretation method flow diagram provided in an embodiment of the present invention;
Fig. 2 is a kind of training set of geologic structure interpretation method provided in an embodiment of the present invention;
Fig. 3 is a kind of convolutional neural networks structure of geologic structure interpretation method provided in an embodiment of the present invention;
Fig. 4 provides a kind of practical coal field geology structure forecast result of geologic structure interpretation method for the embodiment of the present invention;
Fig. 5 is another geologic structure interpretation method flow diagram provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
With the continuous development of coal resources exploration, the required precision of seismic data interpretation is higher and higher.Initially, layer position solves
It releases and structure interpretation is completed using manual manual interpretation, this method is geologic knowledge of the explanation personnel according to oneself
It is explained with experience, it is not only time-consuming but also lack certain objectivity.
For defect existing for manual interpretation, there is following two mode to realize that earthquake fault is explained at present:
The first: identifying fault plane using full convolutional neural networks.It is obtained from 3-D seismics amplitude data body
Seismic amplitude sectional slice data calibrate the position of fault on slice of data by manual interpretation method, with this slice of data
As label data.This method does not account for relationship mutual between contiguous slices, causes using slice of data as label
Prediction result is not accurate, does not have representativeness.
Second: doing label combination seismic data cube using attribute of coherent data volume to carry out model training, the net of application training
Network model predicts the tomography of new seismic data cube.The quality of the prediction result of this method depends on training pattern
Precision, training pattern are that label is made using coherent body, and classification results are inaccurate.
In conclusion at present aiming at the problem that earthquake fault is explained, the time-consuming subjectivity of manual interpretation method, automatic prediction method
Precision is lower, and to solve the above-mentioned problems, the embodiment of the invention provides a kind of architectonic intelligent interpretation methods.For convenient for
The present embodiment is understood, is described in detail below to the embodiment of the present invention.
A kind of architectonic intelligent interpretation method is present embodiments provided, the flow chart of this method is as shown in Figure 1, include
Following steps:
S110 generates the three-dimensional of the explanation results according to the explanation results of predetermined original seismic amplitude data
Data volume;
S120 is based on the explanation results, determines the type of the explanation results;
S130 is carried out assignment to the 3D data volume of the explanation results, is generated institute based on the type of the explanation results
State the label data body of explanation results;
S140 establishes training set according to the label data body and original seismic amplitude data of the explanation results;
S150 is trained predetermined initial convolution neural network model using the training set, is trained
Convolutional neural networks model afterwards, in order to utilize the convolutional neural networks model after the training to geological structure to be predicted
It is predicted.
For step S110, it may be predetermined that original seismic amplitude data, the original seismic amplitude data can refer to tool
There are many coalfield actual seismic data of typical geology structural type, as shown in part (a) of Fig. 2.Next, can be to original
Seismic amplitude data carries out manual interpretation, obtains explanation results, which may include layer position and construction etc.;To explanation
As a result three-dimensional geologic structure modeling is carried out, 3D data volume is generated.
For step S120, the type of explanation results may include layer position type, structural type and without structural type etc..
For step S130, assignment can be carried out to 3D data volume according to the different type of explanation results, which can
Using the label as the 3D data volume.For example, the label that can be 1 by the 3D data volume add value of layer position type, by structure
The label that the 3D data volume add value of type is 2 is made, the label for being 0 by the 3D data volume add value of background type generates
Label data body;As shown in part (b) of Fig. 2, wherein gray layer represents layer position, the dotted representative construction of black, three face of Dark grey
Represent background, background i.e. without construction at.
It, can be shown in part (b) of original earthquake data and Fig. 2 shown in part (a) according to fig. 2 for step S140
Label data body, establish the training set of neural network model.
For step S150, initial convolution neural network model is initially set up, initializes relevant parameter.Such as in some realities
Apply in mode, initial Three dimensional convolution neural network structure as shown in figure 3, include three convolutional layers, two pond layers, one entirely
Articulamentum and a softmax classifier.
Convolutional layer is made of many features figure, and convolution layer parameter includes convolution kernel size, step-length and filling, and three determines jointly
The size for having determined convolutional layer output characteristic pattern, is the hyper parameter of convolutional neural networks.Wherein, it is small that convolution kernel size, which can specify,
In the arbitrary value of input image size, convolution kernel is bigger, and extractible input feature vector is more complicated.The input of convolution kernel and preceding layer
Image carries out convolution algorithm and obtains the lesser figure of size.As shown in formula 1, each element in convolution kernel is weight ginseng
Number, convolution kernel are multiplied and are added with the pixel value in the respective range in the input picture of this layer, and pass through Softmax function call
To the pixel value of output figure.The corresponding matrix of j-th of characteristic pattern of kth layer is weighted simultaneously by several characteristic pattern convolution of preceding layer
It is obtained by activation primitive operation.
Wherein: f is activation primitive, NjIt is the combination of input feature vector figure,It is the eigenmatrix of previous tomographic image,It is
Weight in convolution nuclear matrix,It is j-th of bias matrix of kth layer.
Activation primitive is exactly the Nonlinear Mapping layer in convolutional neural networks, and ReLU function is used in convolutional neural networks
A kind of most activation primitives, when input is less than 0, exporting is 0, when input is greater than 0, is exported as initial value, have calculate it is simple,
Derivation quick advantage, while effective solution convolutional neural networks the problem of gradient disappears in training.ReLU function meter
Calculate formula:
F (x)=max (0, x) (2)
In convolutional neural networks structure shown in Fig. 3, the convolution kernel size that the first convolutional layer uses is 3 × 3 × 3, step-length
It is 2;Second and the convolution kernel size that uses of third convolutional layer be 3 × 3 × 3, step-length 1;Each convolutional layer can be learnt by one group
Filter composition, these filters have small subscriber loops domain, but can expand the depth of input, and convolutional layer is mainly used
In feature extraction.
A pond layer would generally be added between each convolutional layer.After convolutional layer carries out feature extraction, the spy of output
Sign figure can be passed to pond layer and carry out feature selecting and information filtering.Pond layer includes presetting pond function, function
It is the characteristic pattern statistic that the result of a single point in characteristic pattern is replaced with to its adjacent area.Pond layer choosing takes pond region and volume
Product Nuclear receptor co repressor characteristic pattern step is identical, You Chihua size, step-length and filling control.Pond is divided into maximum pond, average Chi Huahe
Random pool.Present embodiment is extracted special using maximum pond using the maximum value of element in block as the output of function
Plane local maximum is levied, maximum pond calculation formula:
Wherein:It is the characteristic value at the position (i, j) of w-th of convolution kernel acquisition, uwIt is the maximum that block is calculated
Value.
In convolutional neural networks structure shown in Fig. 3, it is 2 that size is all employed after the first convolutional layer and the second convolutional layer
× 2 × 2, the maximum pond that step-length is 2.The effect of pond layer is the ruler for reducing picture in the case where reserved graph piece essential information
It is very little, and operated using pondization to reduce the total amount of parameter and calculating.
In convolutional neural networks model finally, the general full articulamentum that can connect one or more, full articulamentum are being rolled up
Act as the role of classifier in product neural network, before one layer of all neurons all mutually interconnected with the neuron of later layer
It connects.In convolutional neural networks structure shown in Fig. 3, full articulamentum is set to after third convolutional layer, each convolutional layer and complete
The linear activation (ReLU) of correction is all applied after articulamentum.
Obtained all distributed nature figures are mapped to sample space by full articulamentum, then using softmax function
Classified calculating exports the class label of maximum probability corresponding to the input picture.The calculation formula of Softmax function:
Wherein: n is categorical measure.
In the training stage, convolutional neural networks need to calculate current predicted value and true value when each iteration
Between gap, these gaps are exactly current penalty values.Softmax Loss function can calculate the classification in current iteration
Loss:
Wherein: N is every batch number of samples, is the label of current training sample, for the correspondence class of current iteration output
Other score.
In convolutional neural networks structure shown in Fig. 3, the softmax classifier after being set to full articulamentum generates one
As a result, indicating a possibility that each classification occur in the central point of input, Fig. 4 is the convolutional neural networks model for using Fig. 3
To the prediction result of practical coal field geology.
Step S150 is obtained after being trained using training set to predetermined initial convolution neural network model
Convolutional neural networks model after to training.It is architectonic to target area finally using obtained convolutional neural networks model
Classification is predicted, realizes that automatic seism explains work.
Description below is carried out in conjunction with specific embodiment of the Fig. 5 to one of geologic structure interpretation method:
Step S101: 3D data volume is generated;According to based on the original seismic amplitude data explained in advance, determines and explain knot
Fruit generates the 3D data volume of explanation results by constructing three-dimensional geological modeling.
Step S102: classify to explanation results;Explanation results are divided into layer position, construction and background three classes.
Step S103: label data body is established;Type based on explanation results carries out the 3D data volume for explaining explanation
Assignment, is assigned to 1 for the label value of layer position, and the label value of construction is assigned to 2, and by background, other without the label value at construction are assigned to 0,
Generate the label data body of explanation results.
Step S104: training set is established;According to the label data body and original seismic amplitude data of explanation results, instruction is established
Practice collection.
Step S105: building neural network model;Three convolutional neural networks models are constructed, relevant parameter is initialized.
Step S106: training neural network model;Using training set data to predetermined initial convolutional neural networks
Model is trained, and obtains final Three dimensional convolution neural network model by minimizing loss function.
Step S107: geological structure to be predicted is predicted;Using obtained convolutional neural networks model, to target
The classification of regional tectonics is predicted, realizes automatic geologic structure interpretation work.
The present invention utilizes the manual interpretation achievement, including layer position, tomography, karst collapse col umn etc. of coalfield actual seismic data, passes through
The method of model construction of tectonic geology generates label data body, and trained convolutional neural networks model is made to be more suitable the practical money in coalfield
Material;The constructions such as layer position, tomography and karst collapse col umn are predicted simultaneously and automatic interpretation, it is contemplated that the mutual pass between layer position and construction
System explains the layer position at construction more accurate;It is any that the convolutional neural networks model trained can be used directly to other
The prediction of coalfield data quickly and accurately can carry out automatic interpretation to geological structure.
In the description of the present invention, it should be noted that term " center ", "upper", "lower", "left", "right", "vertical",
The orientation or positional relationship of the instructions such as "horizontal", "inner", "outside" be based on the orientation or positional relationship shown in the drawings, merely to
Convenient for description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation,
It is constructed and operated in a specific orientation, therefore is not considered as limiting the invention.In addition, term " first ", " second ",
" third " is used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features;And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (10)
1. a kind of architectonic intelligent interpretation method, which comprises the following steps:
According to the explanation results of predetermined original seismic amplitude data, the 3D data volume of the explanation results is generated;
Based on the explanation results, the type of the explanation results is determined;
Based on the type of the explanation results, assignment is carried out to the 3D data volume of the explanation results, generates the explanation knot
The label data body of fruit;
According to the label data body and original seismic amplitude data of the explanation results, training set is established;
Predetermined initial convolution neural network model is trained using the training set, the convolution mind after being trained
Through network model, in order to be predicted using the convolutional neural networks model after the training geological structure to be predicted.
2. architectonic intelligent interpretation method according to claim 1, which is characterized in that the original seismic amplitude number
According to for the coalfield actual seismic data with a variety of typical geology structural types, and the data of accurate explanation results in advance;Institute
Stating explanation results includes: layer position, construction and background.
3. architectonic intelligent interpretation method according to claim 1, which is characterized in that described to the explanation results
3D data volume carry out assignment refer to: the label value of layer position is assigned to 1, the label value of construction is assigned to 2, by the label of background
Value is assigned to 0, generates the label data body;The background refers at no construction.
4. architectonic intelligent interpretation method according to claim 1, which is characterized in that the training set includes: institute
State label data body and the original seismic amplitude data.
5. architectonic intelligent interpretation method according to claim 1, which is characterized in that the convolutional neural networks mould
Type is Three dimensional convolution neural network model.
6. architectonic intelligent interpretation method according to claim 1, which is characterized in that described predetermined initial
Convolutional neural networks model includes: selection initiation parameter;Utilize the original seismic amplitude data and the label data body
The training convolutional neural networks model;It minimizes loss function and obtains final training pattern.
7. architectonic intelligent interpretation method according to claim 1, which is characterized in that the convolutional neural networks packet
Three convolutional layers are included, feature extraction is used for;
The convolution kernel size of first convolutional layer is 3 × 3 × 3, step-length 2;The convolution kernel of second convolutional layer and third convolutional layer is big
Small is 3 × 3 × 3, step-length 1;
Three convolutional layers apply compensation line to activate.
8. architectonic intelligent interpretation method according to claim 7, which is characterized in that the convolutional neural networks are also
Including two pond layers, the feature that the convolutional layer extracts is selected and filtered;
The size of first pond layer is 2 × 2 × 2, and step-length 2 is set to after first convolutional layer;
The size of second pond layer is 2 × 2 × 2, and step-length 2 is set to after second convolutional layer.
9. architectonic intelligent interpretation method according to claim 7, which is characterized in that the convolutional neural networks are also
Including a full articulamentum and a softmax classifier;
The full articulamentum, is activated using linearity correction, is set to after the third convolutional layer;
The softmax classifier is set to after the full articulamentum, for generating result.
10. a kind of machine readable storage medium, which is characterized in that the machine readable storage medium is stored with the executable finger of machine
It enables, for the machine-executable instruction when being called and being executed by processor, the machine-executable instruction promotes the processor
Realize the described in any item methods of claim 1-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910777505.1A CN110441820B (en) | 2019-08-21 | 2019-08-21 | Intelligent interpretation method of geological structure |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910777505.1A CN110441820B (en) | 2019-08-21 | 2019-08-21 | Intelligent interpretation method of geological structure |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110441820A true CN110441820A (en) | 2019-11-12 |
CN110441820B CN110441820B (en) | 2020-06-16 |
Family
ID=68437061
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910777505.1A Active CN110441820B (en) | 2019-08-21 | 2019-08-21 | Intelligent interpretation method of geological structure |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110441820B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111273352A (en) * | 2020-01-15 | 2020-06-12 | 中国煤炭地质总局勘查研究总院 | Intelligent detection method and device for geological structure and electronic equipment |
CN112130200A (en) * | 2020-09-23 | 2020-12-25 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
CN113296152A (en) * | 2020-02-21 | 2021-08-24 | 中国石油天然气集团有限公司 | Fault detection method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104280771A (en) * | 2014-10-27 | 2015-01-14 | 中国石油集团川庆钻探工程有限公司地球物理勘探公司 | Three-dimensional seismic data waveform semi-supervised clustering method based on EM algorithm |
WO2016073483A1 (en) * | 2014-11-05 | 2016-05-12 | Shell Oil Company | Systems and methods for multi-dimensional geophysical data visualization |
CN109086773A (en) * | 2018-08-29 | 2018-12-25 | 电子科技大学 | Fault plane recognition methods based on full convolutional neural networks |
CN109492775A (en) * | 2018-11-19 | 2019-03-19 | 中国矿业大学(北京) | A kind of detection method of geologic structure interpretation, detection device and readable storage medium storing program for executing |
CN109597129A (en) * | 2019-01-07 | 2019-04-09 | 中国地质大学(武汉) | Fracture-pore reservoir beading reflectance signature recognition methods based on target detection |
WO2019118658A1 (en) * | 2017-12-14 | 2019-06-20 | Schlumberger Technology Corporation | System and method for simulating reservoir models |
-
2019
- 2019-08-21 CN CN201910777505.1A patent/CN110441820B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104280771A (en) * | 2014-10-27 | 2015-01-14 | 中国石油集团川庆钻探工程有限公司地球物理勘探公司 | Three-dimensional seismic data waveform semi-supervised clustering method based on EM algorithm |
WO2016073483A1 (en) * | 2014-11-05 | 2016-05-12 | Shell Oil Company | Systems and methods for multi-dimensional geophysical data visualization |
WO2019118658A1 (en) * | 2017-12-14 | 2019-06-20 | Schlumberger Technology Corporation | System and method for simulating reservoir models |
CN109086773A (en) * | 2018-08-29 | 2018-12-25 | 电子科技大学 | Fault plane recognition methods based on full convolutional neural networks |
CN109492775A (en) * | 2018-11-19 | 2019-03-19 | 中国矿业大学(北京) | A kind of detection method of geologic structure interpretation, detection device and readable storage medium storing program for executing |
CN109597129A (en) * | 2019-01-07 | 2019-04-09 | 中国地质大学(武汉) | Fracture-pore reservoir beading reflectance signature recognition methods based on target detection |
Non-Patent Citations (1)
Title |
---|
韩万林 等: "断层推断的改进BP神经网络方法", 《合肥工业大学学报(自然科学版)》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111273352A (en) * | 2020-01-15 | 2020-06-12 | 中国煤炭地质总局勘查研究总院 | Intelligent detection method and device for geological structure and electronic equipment |
CN113296152A (en) * | 2020-02-21 | 2021-08-24 | 中国石油天然气集团有限公司 | Fault detection method and device |
CN112130200A (en) * | 2020-09-23 | 2020-12-25 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
CN112130200B (en) * | 2020-09-23 | 2021-07-20 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
Also Published As
Publication number | Publication date |
---|---|
CN110441820B (en) | 2020-06-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113392775B (en) | Sugarcane seedling automatic identification and counting method based on deep neural network | |
CN107480341B (en) | A kind of dam safety comprehensive method based on deep learning | |
CN109359519A (en) | A kind of video anomaly detection method based on deep learning | |
CN110441820A (en) | Architectonic intelligent interpretation method | |
CN109522967A (en) | A kind of commodity attribute recognition methods, device, equipment and storage medium | |
CN109446992A (en) | Remote sensing image building extracting method and system, storage medium, electronic equipment based on deep learning | |
CN110084294A (en) | A kind of Remote Image Classification based on multiple dimensioned depth characteristic | |
CN110516539A (en) | Remote sensing image building extracting method, system, storage medium and equipment based on confrontation network | |
CN107316066A (en) | Image classification method and system based on multi-path convolutional neural networks | |
CN106920243A (en) | The ceramic material part method for sequence image segmentation of improved full convolutional neural networks | |
CN109886217A (en) | A method of it is high that wave being detected from Nearshore Wave video based on convolutional neural networks | |
CN108830285A (en) | A kind of object detection method of the reinforcement study based on Faster-RCNN | |
CN109508360A (en) | A kind of polynary flow data space-time autocorrelation analysis method of geography based on cellular automata | |
CN110189255A (en) | Method for detecting human face based on hierarchical detection | |
CN109902715A (en) | A kind of method for detecting infrared puniness target based on context converging network | |
CN110264484A (en) | A kind of improvement island water front segmenting system and dividing method towards remotely-sensed data | |
CN106991666B (en) | A kind of disease geo-radar image recognition methods suitable for more size pictorial informations | |
CN104021396A (en) | Hyperspectral remote sensing data classification method based on ensemble learning | |
CN109872311A (en) | A kind of Rock Mass Integrality sentences knowledge method | |
CN110222773A (en) | Based on the asymmetric high spectrum image small sample classification method for decomposing convolutional network | |
CN109409441A (en) | Based on the coastal waters chlorophyll-a concentration remote sensing inversion method for improving random forest | |
CN109544537A (en) | The fast automatic analysis method of hip joint x-ray image | |
CN109886103A (en) | Urban poverty measure of spread method | |
CN109961434A (en) | Non-reference picture quality appraisement method towards the decaying of level semanteme | |
CN109919045A (en) | Small scale pedestrian detection recognition methods based on concatenated convolutional network |
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 |