CN103487832A - Method for classifying supervised waveforms in three-dimensional seismic signal - Google Patents

Abstract

The invention provides a method for classifying supervised waveforms in a three-dimensional seismic signal. The method mainly includes the steps of data preprocessing, feature selection and classified identification. The method has the advantages that the method is based on three-dimensional seismic signal resources and well logging data information, extracted attributive characters are optimized through a genetic algorithm, analyzed three-dimensional seismic target interval data undergoes waveform classification by means of an SVM classification algorithm, different seismic facies are recognized, and therefore reliable supports are provided for follow-up seismic resource explanation, and reliability of lithology prediction, sand body prediction, fractured reservoir prediction, elusive reservoir prediction and the like is improved. Compared with an SVM design classifier, the method carries out feature selection through addition of the genetic algorithm, design complexity of the SVM classifier is reduced, and therefore waveform classification processing efficiency is improved.

Description

In a kind of 3-D seismics signal the supervision waveform classification arranged

Technical field

The invention belongs to the sorting technique field of waveform in seismic signal, relate to a kind of waveform classification of 3-D seismics signal, particularly wherein have the supervision waveform sorting technique.

Background technology

Waveform separation technology based on seismic signal is the important means that the seismic interpretation personnel carry out subsurface reservoir and stratal configuration analysis.Rationally and accurately the seismic signal waveform classification results can reflect subsurface reservoir and stratal configuration structure truly, thereby be conducive to the seismic interpretation personnel underground structure is carried out to structure elucidation accurately, and then raising is to the reliability of lithology prediction, Sand-body Prediction, fractured reservoirs prediction and disguised prediction of oil-gas reserve, thereby minimizing exploration risk, save prospecting prime cost, bring huge economic and social benefit.So, the waveform separation research of seismic signal is had to important practical significance.

In the seismic prospecting of oil gas, the purpose of seismic data interpretation is to carry out the description of underground structure explanation and stratum and rock signature in order to extract more information from geological data.Obtaining one of effective method of these information from seismic data is exactly extraction and analysis and the waveform separation technology of seismic properties feature.

Along with the development of scientific and technological level and improving constantly the seismic data acquisition technology, make the earthquake information comprised in seismic signal abundanter, and wherein many useful earthquake informations only depend on the observation of naked eyes can not detect out, must be extracted, be analyzed it by seismic data processing technology and computer technology, and, by certain mathematical method, the geologic feature of these earthquake informations is explained.For the existing geological data gathered, current waveform separation technology mainly is based on without the supervised classification algorithm, particularly based on artificial neural network theories, as business software OpendTect, Petrel and Stratimgic etc.The method, to containing noisy seismic signal data, having higher holding capacity, also has the classification capacity without the training sample data.But, also there is some shortcoming without supervision waveform separation algorithm: at first, ignored without supervision waveform separation algorithm the well logging information that there is the important references meaning in the geological data, taxonomic structure is internal distribution and the statistical nature based on earthquake work area data just, with the undertighten that contacts of actual conditions, classification results is not accurately with rationally; In addition, artificial neural network has that the computation complexity of network training is higher, arranging of some parameter needs us to have neural network in some prior imformations and some situation may to be absorbed in the problems such as local optimum.

Therefore, the difficulty of the waveform separation problem based on the 3-D seismics signal is mainly manifested in following several aspect:

(1) along with the raising of seismic data acquisition technology, the earthquake information comprised in the 3-D seismics signal is more and more abundanter, and how extracting exactly effective information in seismic signal, to carry out that waveform separation processes be the large difficult point in geological data waveform separation problem.Reasonable and effective seismic properties feature, the accuracy that can improve well waveform separation, improve the reliability to lithology prediction, Sand-body Prediction, fractured reservoirs prediction and disguised prediction of oil-gas reserve.

(2) sorter model that how to build superior performance is based on another large difficult point that the waveform separation of 3-D seismics signal is analyzed.Selecting suitable sorting algorithm, not only can set up the superior sorter model of classification performance, also can improve the waveform separation analysis efficiency, is the emphasis in waveform separation so select suitable sorting algorithm.

(3) for the characteristics of geological data, data can be sneaked into a certain amount of noise in gatherer process.Due to adding of random noise, can cause serious impact to the waveform separation result of seismic signal, reduce the accuracy of classification results, and then can affect the reliability of lithology prediction, Sand-body Prediction, fractured reservoirs prediction and disguised prediction of oil-gas reserve.

On the waveform separation problem of seismic signal, at present more implementation is arranged both at home and abroad, in these methods, major part all is based on without supervised classification thought carries out the waveform separation analysis to seismic signal.Be below three kinds of implementations wherein: (1) waveform classification based on self-organizing map neural network: at first the method sets up the model trace data to the seismic signal data of the objective interval that extracts, by the seismic trace sample data, train the output layer node of self organizing neural network to set up disaggregated model, and then utilize model trace to carry out classifying and dividing to seismic signal.(2) waveform classification based on hierarchical cluster: at first the method builds the cluster spanning tree to the seismic signal data of the objective interval that extracts, and then the cluster spanning tree is carried out to classifying and dividing to seismic signal.(3) waveform classification based on the mixed Gaussian probability model: the method is at first according to the seismic signature of the objective interval extracted, utilize the probability statistical analysis theory to set up the mixed Gaussian probability model, and then utilize the mixed Gaussian probability model to carry out classifying and dividing to seismic signal.

Above prior art, all obtained application significantly in actual seismic data analysis, but these methods all are based on without supervised classification thought and are processed, ignore so to a great extent the log data information that there is the important references meaning in the earthquake data acquisition, caused very large difficulty to follow-up seismic interpretation; And these algorithms have the shortcomings such as computation complexity is high, and operational efficiency is lower.

Summary of the invention

The present invention is directed to difficult point and the existing shortcoming without supervision waveform separation technology of seismic signal waveform sorting technique, having proposed has a supervision waveform classification in a kind of 3-D seismics signal, for solving the existing shortcoming without the supervision waveform classification and the difficult point of waveform separation.

Technical scheme of the present invention is: in a kind of 3-D seismics signal the supervision waveform classification arranged, it is characterized in that, comprise the following steps:

A, data pre-service: comprise that Noise reducing of data is processed, extraction objective interval data reach well logging data analysis is set up to training sample and label;

B, feature selecting: analyze training sample and the tag along sort data thereof extracted, select in sample feature with the maximally related a certain amount of feature of tag along sort as character subset, the character subset of usining after preferred in objective interval data and training sample is as the input sample set of class indication in this programme;

C, class indication: set up disaggregated model seismic signal is carried out to class indication;

Further, in above-mentioned steps a, described denoise processing method is as follows:

A11, input geological data;

A12, judge whether to reach maximum iteration time, if finish the noise reduction process process, otherwise execution step a13;

A13, the Grad of calculating earthquake number strong point on three directions the further structure tensor that calculates this data point;

A14, utilize three-dimensional gauss low frequency filter to carry out the filtering processing to described structure tensor;

A15, structure diffusion matrix and discontinuous factor;

A16, according to diffusion equation to data filtering.

Further, in above-mentioned steps a, described extraction objective interval data method is as follows:

A21, the data of step a16 gained are done to the single layer bit analyzing and processing;

A22, the data of step a16 gained are done to two-layer interdigit analyzing and processing.

Further, described step a21 is further comprising the steps of:

A211, true timing window and analysis layer position;

Window data when a212, extraction edge layer;

A213, acquisition objective interval data.

Described step a22 is further comprising the steps of:

A221, the two-layer position of definite analysis;

A222, extraction are along the interlayer data;

A223, acquisition objective interval data.

Further, in above-mentioned steps a, the described concrete grammar that well logging data analysis is set up to training sample and label is: according to log analysis data in the geological data work area, well logging in work area is carried out to the manual sort to be meaned, set up the training sample label, and in the extraction objective interval well logging and the data of well lie position as training sample, for feature selecting and class indication.

Further, in above-mentioned steps b, feature selecting adopts genetic algorithm, and the training sample set that wherein extracts well logging and well lie is combined into T, and corresponding tag along sort is y, and the feature selecting number is that k and maximum iteration time are maxI, specifically comprises the following steps:

B1, according to the gene code rule, obtain the initialization population of genetic algorithm;

B2, basis

fitness function calculates the fitness of population at individual, wherein x _i(i=1,2 ..., k) be x in the selected training sample set of population at individual T _ithe dimensional feature component, the tag along sort that y is corresponding well logging; MIx _iy means the xi dimensional feature component of log data and the mutual information of tag along sort y; MIx _ix _jthe x that means log data _icharacteristic component and characteristic component x _jbetween mutual information;

B3, judge whether to reach maximum iteration time maxI or the continuous optimum solutions that keep for five times are constant, if, export current optimal feature subset, otherwise execution step b4;

B4, according to the population at individual fitness, carry out to select operation;

B5, execution interlace operation;

B6, execution mutation operation;

B7, return to b2 operation;

B8, output characteristic subset.

Further, in the process of processing in above-mentioned genetic algorithm, the gene code of step b1 adopts that { 0,1} binary string means, which position in binary string just means the seismic channel data u that extracts in the data pre-service _i={ x _i1, x _i2..., x _iMand log data w _i={ x _i1, x _i2..., x _iMin which dimensional feature of feature space, 0 means that this characteristic component is selected, 1 means to select this characteristic component.

Further, in above-mentioned steps c, use support vector machine to set up disaggregated model, and the objective interval data are carried out to class indication, finally make waveform separation division figure.

The detailed process of above-mentioned steps c is:

C1, the character subset data normalization is processed;

C2, use 10 folding cross-validation methods to receive the best SVM parameters of rope;

C3, set up the svm classifier model;

C4, to selecting the objective interval data after feature, carry out Classification and Identification;

C5, drawing waveforms classifying and dividing figure.

Beneficial effect of the present invention: of the present invention to have the supervision waveform classification to take 3-D seismics signal data and log data information be basis, by genetic algorithm, the attributive character of extracting is optimized, utilize the svm classifier algorithm, analyzed 3-D seismics objective interval data are carried out to the waveform separation division, identify different seismic facies, and then support for the latter earthquake data interpretation provides reliably, improve the reliability to lithology prediction, Sand-body Prediction, fractured reservoirs prediction and disguised prediction of oil-gas reserve etc.Simultaneously, utilize well logging and well lie information to set up training sample, be applied to like this well logging information with important references meaning in waveform separation, improve nicety of grading, reduce the difficulty of later stage explanation; In feature selecting is processed, introduce genetic algorithm, reduce the redundance of feature, reduce the complexity of disaggregated model design, improve classification effectiveness, and reduce the impact of redundance feature on classification results simultaneously, improve to a certain extent classify accuracy; In class indication is processed, the mode identification method algorithm of support vector machine of introducing based on Statistical Learning Theory carries out Classification and Identification.With nothing supervision waveform separation algorithm, compare: owing to having used, there is the well logging information that important references is worth, improved the accuracy of waveform separation, the reliability of raising to lithology prediction, Sand-body Prediction, fractured reservoirs prediction and disguised prediction of oil-gas reserve, for follow-up seismic interpretation provides very big facility; With respect to only comparing with SVM design category device, added genetic algorithm to carry out feature selecting, reduce the design complexities of svm classifier device, thereby improved the waveform separation treatment effeciency.

The accompanying drawing explanation

The evolution cycle schematic diagram of Fig. 1 genetic algorithm;

The method flow diagram that Fig. 2 is the embodiment of the present invention;

Structure directing filtering process flow diagram in the scheme that Fig. 3 is the embodiment of the present invention;

Extract the objective interval data flowchart in the scheme that Fig. 4 is the embodiment of the present invention;

In the scheme that Fig. 5 is the embodiment of the present invention, character subset extracts process flow diagram;

Set up the disaggregated model process flow diagram in the scheme that Fig. 6 is the embodiment of the present invention;

Fig. 7 is SOM waveform separation division figure;

Fig. 8 is SVM waveform separation division figure;

The waveform separation division figure that Fig. 9 is the embodiment of the present invention.

Embodiment

Do specific descriptions below in conjunction with accompanying drawing pair a plurality of specific embodiments identical with the principle of the invention, to promote the understanding to the principle of the invention.

For the shortcoming that has 3-D seismics signal waveform sorting technique commonly used now and the characteristics of waveform separation itself, the supervision waveform separation scheme that has that the present invention proposes mainly comprises data pre-service, the characteristic optimization based on the GA algorithm and the Classification and Identification three parts based on the svm classifier algorithm.General flow chart is as shown in Figure 2: in a kind of 3-D seismics signal of the present embodiment the supervision waveform classification arranged, comprise step: a, data pre-service: comprise that Noise reducing of data processes, extract the objective interval data and well logging data analysis is set up to training sample and label; B, feature selecting: analyze training sample and the tag along sort data thereof extracted, select in sample feature with the maximally related a certain amount of feature of tag along sort as character subset, and the input sample set of the character subset class indication in this programme after preferred in objective interval data and training sample; C, class indication: set up disaggregated model seismic signal is carried out to class indication.In step a, described denoise processing method is as follows: a11, input geological data; A12, judge whether to reach maximum iteration time, if finish the noise reduction process process, otherwise execution step a13; A13, the Grad of calculating earthquake number strong point on three directions the further structure tensor that calculates this data point; A14, utilize three-dimensional gauss low frequency filter to carry out the filtering processing to described structure tensor; A15, structure diffusion matrix and discontinuous factor; A16, according to diffusion equation to data filtering.Extract the objective interval data method as follows: a21, the data of step a16 gained are done to the single layer bit analyzing and processing; A22, the data of step a16 gained are done to two-layer interdigit analyzing and processing.Step a21 comprises the following steps: a211, true timing window and analysis layer position; Window data when a212, extraction edge layer; A213, acquisition objective interval data.Step a22 comprises the following steps: a221, the two-layer position of definite analysis; A222, extraction are along the interlayer data; A223, acquisition objective interval data.

In step a, the concrete grammar of well logging data analysis being set up to training sample and label is: according to log analysis data in the geological data work area, well logging in work area is carried out to the manual sort to be meaned, set up the training sample label, and in the extraction objective interval well logging and the data of well lie position as training sample, for feature selecting and class indication.

In step b, feature selecting adopts genetic algorithm, the training sample set that wherein extracts well logging and well lie is combined into T, corresponding tag along sort is y, the feature selecting number is that k and maximum iteration time are maxI, specifically comprise the following steps: b1, according to the gene code rule, obtain the initialization population of genetic algorithm; B2, basis

fitness function calculates the fitness of population at individual, wherein x _i(i=1,2 ..., k) be x in the selected training sample set of population at individual T _ithe dimensional feature component, the tag along sort that y is corresponding well logging; MIx _iy means the x of log data _ithe mutual information of dimensional feature component and tag along sort y; MIx _ix _jthe x that means log data _icharacteristic component and characteristic component x _jbetween mutual information; B3, judge whether to reach maximum iteration time maxI or the continuous optimum solutions that keep for five times are constant, if, export current optimal feature subset, otherwise execution step b4; B4, according to the population at individual fitness, carry out to select operation; B5, execution interlace operation; B6, execution mutation operation; B7, return to b2 operation; B8, output characteristic subset.Further, in the process of processing in above-mentioned genetic algorithm, the gene code of step b1 adopts that { 0,1} binary string means, which position in binary string just means the seismic channel data u that extracts in the data pre-service _i={ x _i1, x _i2..., x _iMand log data w _i={ x _i1, x _i2..., x _iMin which dimensional feature of feature space, 0 means that this characteristic component is selected, 1 means to select this characteristic component.

Above-mentioned step c is used support vector machine to set up disaggregated model, and the objective interval data are carried out to class indication, finally makes waveform separation division figure.Detailed process is: c1, the character subset data normalization is processed; C2, use 10 folding cross-validation methods to receive the best SVM parameters of rope; C3, set up the svm classifier model; C4, to selecting the objective interval data after feature, carry out Classification and Identification; C5, drawing waveforms classifying and dividing figure.

Scheme below in conjunction with principle of work and effect to above-described embodiment is further described.

A, data pre-service: it is to prepare for feature selecting and class indication that the 3D seismic data data is carried out to pretreated fundamental purpose, and its groundwork flow process mainly comprises: Noise reducing of data is processed, the extraction of objective interval data and the analysis of log data is set up to training sample and label three parts.Due to the physical limitation of system for acquiring seismic data equipment, the limitation of transmission medium and the uncertainty of environmental baseline, the geological data that collection is obtained can contain a certain amount of noise information, and owing in geologic structure, can containing the detailed information such as tomography and geological objects boundary.So should be noted that in the filtering to geological data oriention analysis, rim detection and edge keep three main points of guiding smothing filtering, therefore choice structure Steerable filter method is carried out noise reduction to data, improved signal to noise ratio (S/N ratio), and kept or strengthened original geologic structure.Its work for the treatment of flow process as shown in Figure 3.Extract the objective interval data: the layer position according to analyzing extracts the objective interval data that we will carry out the waveform separation analysis.Extract the overall procedure of objective interval data as shown in Figure 4.Analyze log data, set up training sample and label: according to log analysis data in the geological data work area, we carry out the manual sort by well logging in work area and mean, set up the training sample label, and in the extraction objective interval well logging and the data of well lie position as training sample, for feature selecting and the class indication of back.

B, feature selecting: the fundamental purpose of feature selecting is in order to optimize the feature in the objective interval data, the training sample extracted by analysis and tag along sort data thereof, select feature and the maximally related a certain amount of feature of tag along sort in sample, so not only can reduce the complexity of back disaggregated model design, efficiency is provided, also can improve to a certain extent nicety of grading, reduce the impact of irrelevant feature on classification results.The characteristic optimization algorithm that this programme is selected is genetic algorithm, and it is input as the well logging of extraction and the training sample set T of well lie, and corresponding tag along sort y, feature selecting number k and maximum iteration time maxI.And export with the character subset preferably.The process flow diagram of feature selecting as shown in Figure 5.

C, class indication: after the training sample set that objective interval data and well logging are formed carries out feature selecting, carry out most important work and set up disaggregated model seismic signal is carried out to class indication.Few for log data in the 3-D seismics work area, characteristics with little training sample, and the seismic signal waveform classification is generally the Nonlinear Classification problem, and combining support vector machine has and better must solve the small sample that in the past perplexed a lot of learning methods, non-linear, the ability of crossing the practical problemss such as study, high dimension, local minimum point, this programme has selected support vector machine to set up disaggregated model, and the objective interval data are carried out to class indication, finally make waveform separation division figure.The idiographic flow of this section processes as shown in Figure 6.

In the solution of the present invention, the some algorithm principle related to is as follows:

Structure directing filtering

Structure directing filtering technique for 3D seismic data filtering ^[7]on the anisotropic diffusion equation basis of diffusion tensor matrices, a kind of edge preserving smooth filter algorithm that the structure tensor of structure reflection 3-D view partial structurtes is realized.In 3-D view is processed, we definition structure tensor S is:

$S=\left[\begin{array}{ccc}\frac{\∂u}{\∂x}\frac{\∂u}{\∂x}& \frac{\∂u}{\∂x}\frac{\∂u}{\∂y}& \frac{\∂u}{\∂x}\frac{\∂u}{\∂z}\\ \frac{\∂u}{\∂x}\frac{\∂u}{\∂y}& \frac{\∂u}{\∂y}\frac{\∂u}{\∂y}& \frac{\∂u}{\∂y}\frac{\∂u}{\∂z}\\ \frac{\∂u}{\∂x}\frac{\∂u}{\∂z}& \frac{\∂u}{\∂y}\frac{\∂u}{\∂z}& \frac{\∂u}{\∂z}\frac{\∂u}{\∂z}\end{array}\right]---(1-1)$

In above formula: u (x, y, z) is the input original three-dimensional image; In the structure tensor matrix

the partial derivative that means respectively x direction, y direction and these three dimensions of z direction.Can there be certain instability in structure tensor in some marginal texture present positions of 3-D seismics image, so itself and small scale gaussian kernel can be carried out to convolution, new like this structure tensor can have certain robustness, and this process can be expressed as follows:

S _σ=S(G _σ*u(t)) (1-2)

Structure tensor matrix S in above formula _σa positive semidefinite matrix, thus can be by this matrix is carried out to Eigenvalues Decomposition, thus obtain:

$S_{\mathrm{\σ}}=\left(\begin{array}{ccc}{v}_1& v_2& v_3\end{array}\right)\left[\begin{array}{ccc}{\mathrm{\λ}}_1& 0& 0\\ 0& {\mathrm{\λ}}_2& 0\\ 0& 0& {\mathrm{\λ}}_3\end{array}\right]{\left(\begin{array}{ccc}{v}_1& v_2& v_3\end{array}\right)}^T---(1-3)$

In above formula: λ ₁, λ ₂, λ ₃for the structure tensor matrix S calculated _σeigenwert on three directions, and λ ₁>=λ ₂>=λ ₃>=0; v ₁, v ₂, v ₃, be respectively and three eigenvalue λ ₁, λ ₂, λ ₃corresponding proper vector.In the process of diffusing filter, for protect marginal texture and the compacting noise, we by with two the proper vector vs of gradient towards quadrature ₂, v ₃each component v ₂₁, v ₂₂, v ₃₁, v ₃₂construct diffusion tensor D, be expressed as:

$D=\left[\begin{array}{ccc}{v}_{21}{v}_{21}+v_{31}{v}_{31}& v_{21}{v}_{22}+v_{31}{v}_{32}& v_{21}{v}_{23}+v_{31}{v}_{33}\\ v_{21}{v}_{22}+v_{31}{v}_{32}& v_{22}{v}_{22}+v_{32}{v}_{32}& v_{22}{v}_{23}+v_{32}{v}_{33}\\ v_{21}{v}_{23}+v_{31}{v}_{33}& v_{22}{v}_{23}+v_{32}{v}_{33}& v_{23}{v}_{23}+v_{33}{v}_{33}\end{array}\right]---(1-4)$

We just can show that the structure directing filtering equations is like this:

$\begin{array}{cc}u(t+1)=u\left(t\right)+c\·\mathrm{div}(D(G_{\mathrm{\σ}}*u\left(t\right))\▿u)& t\≥0---(1-5)\end{array}$

In the waveform separation application of seismic signal, the marginal texture of underground medium and discontinuous information are very important, so we have constructed a discontinuous factor f at this, and this factor have been applied in structure directing filtering.

$f=\frac{\mathrm{diag}\left(S^{0}S\right)}{\mathrm{diag}\left(S^0\right)\mathrm{diag}\left(S\right)}---(1-6)$

In formula: diag () means the element on the main diagonal angle of matrix is sued for peace; S ⁰the structure tensor that means the input original image; S means the structure tensor of circulation at present; Wherein the span of discontinuous factor f is [0,1], and in the time of near marginal texture and intermittent configuration, the f value is tending towards 0; But, during away from marginal texture, the f value is tending towards 1.After introducing discontinuous factor, the structure directing filtering algorithm can be expressed as:

$\begin{array}{cc}u(t+1)=u\left(t\right)+c\·\mathrm{div}(f\·D(G_{\mathrm{\σ}}*u\left(t\right))\▿u)& t\≥0---(1-7)\end{array}$

Genetic algorithm

Genetic algorithm is that a kind of to take Darwinian natural selection and theory of heredity be basis, the efficient global optimization approach that survival of the fittest rule in the biological evolution process is combined with the interior random exchanging mechanism of chromosomal information of group.A complete genetic algorithm mainly comprises judgement of gene code, initialization of population, selection operation, interlace operation, mutation operation and termination condition etc.Complete genetic algorithm evolution cycle is as shown in Figure 1:

Selection, crossover and mutation operation are referred to as genetic manipulation, and for the setting of selecting to the most important thing is in operation fitness function, and crossover and mutation is the main operation of Random Effect ideal adaptation degree value; Termination condition is the judgement genetic algorithm criterion that could stop, and general to adopt maximum heredity time number of operations be that maximum iteration time or the optimum solution that calculates after genetic manipulation several times are constant.

Support vector machine

Support vector machine (Support Vector Machine) is a kind of machine learning method based on Statistical Learning Theory grown up middle 1990s.It is to improve the learning machine generalization ability by seeking the structuring least risk, realizes minimizing of empiric risk and fiducial range, thereby reaches in the situation that the statistical sample amount is less, also can obtain the purpose of good statistical law.In fields such as recognition of face, genetic test, text identification and network flow classification, be widely used at present.

For non-linear SVM model, support vector machine be by a mapping phi () by original data-mapping in new space, then do the Linear SVM sort operation.By lagrange's method of multipliers, proper objective function just becomes like this:

$\underset{\mathrm{\&alpha;}}{\max }L\left(\mathrm{\&alpha;}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i-\frac{1}{2}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{\mathrm{\&alpha;}}_j{y}_i{y}_j<\mathrm{\&phi;}\left(x_i\right)\&CenterDot;\mathrm{\&phi;}\left(x_j\right)>$

$\mathrm{subjectto}:0\&le;{\mathrm{\&alpha;}}_i\&le;C,i=1,...,n;\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{y}_i=0---(1-8)$

By in formula<φ (x _i) φ (x _j) a kernel function κ (x for inner product operation _i, x _j) replace, the objective function that just can obtain non-linear SVM is:

$\underset{\mathrm{\&alpha;}}{\max }L\left(\mathrm{\&alpha;}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i-\frac{1}{2}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{\mathrm{\&alpha;}}_j{y}_i{y}_j\mathrm{\&kappa;}(x_i,x_j)$

$\mathrm{subjectto}:0\&le;{\mathrm{\&alpha;}}_i\&le;C,i=1,...,n;\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{y}_i=0---(1-9)$

And the optimization lineoid (classification function) of non-linear SVM is:

$f\left(x\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i{y}_i\mathrm{\&kappa;}(x_i,x)+b=\underset{i\&Element;\mathrm{sv}}{\mathrm{\&Sigma;}}{\mathrm{\&alpha;}}_i{y}_i\mathrm{\&kappa;}(x_i,x)+b---(1-10)$

The kernel function of commonly using in non-linear SVM mainly contains the polynomial(polynomial expression) kernel function, RBF(radial basis) kernel function and sigmoid kernel function, functional form is suc as formula 1-11,1-12,1-13.In order to obtain higher nicety of grading, the parameter of kernel function needs correct setting.

(1) Polynomial kernel function:

$\mathrm{\&kappa;}(x_i,x_j)={(1+x_i\&CenterDot;x_j)}^d---(1-11)$

(2) RBF kernel function:

$\mathrm{\&kappa;}(x_i,x_j)=\exp (-\frac{{\left|\right|x_i-x_j\left|\right|}^2}{{2\mathrm{\&sigma;}}^2})---(1-12)$

(3) Sigmoid kernel function:

κ(x _i，x _j)=tanh(kx _i·x _j-δ) （1-13）

In order to estimate the performance of this programme, we carry out application simulation to actual seismic work area data.In experiment simulation, we have compared respectively self-organizing map neural network algorithm commonly used and have only used the classification performance of SVM algorithm process.

Emulation: chosen the demonstration data F3 work area 3D seismic data of business software Opendtect as the real data that will analyze.F3 is the block that the North Sea is positioned at Holland's part.The F3 work area scope that we select is No. Inline: 104-696, No. Xline: 304-1246, wherein well logging always has 20 mouthfuls.The layer position of carrying out seismic facies analysis is MFS4.The objective interval data set that table 1 extracts for us.

The objective interval data set that table 1 extracts

Utilize SOM algorithm, SVM algorithm and this programme GA+SVM algorithm to set up disaggregated model to log data, wherein in this programme, by genetic algorithm, process 20 characteristic components that extracted the lonely data centralization of purpose Cen, result is as shown in table 2.

Table 2 log data collection in the result of setting up disaggregated model relatively

The disaggregated model of setting up with three kinds of algorithms respectively carries out class indication to the work area earthquake data set, and the waveform separation division figure made is respectively as shown in Fig. 7, Fig. 8, Fig. 9.

In the present invention program, following word is special use or the universal word of this area, and the meaning of its representative is as follows: layer position refers to a certain ad-hoc location in succession of strata, the layer position on stratum can be the boundary line of stratigraphic unit, can be also the reference lamina etc. that belongs to a certain specific epoch; Tomography, breaking because of the stressed some strength that reaches in earth's crust rock stratum, and has the structure obviously relatively moved to be called tomography along the plane of fracture; Interpolation, utilize known point to calculate the process of unknown point; Objective interval, refer in the 3-D seismics signal data, to analyze along single layer position the time window data or two layer interdigits will analyzing the interlayer data.

Those of ordinary skill in the art will appreciate that, embodiment described here is in order to help reader understanding's implementation method of the present invention, should be understood to that protection scope of the present invention is not limited to such special statement and embodiment.Those of ordinary skill in the art can make various other various concrete distortion and combinations that do not break away from essence of the present invention according to these technology enlightenments disclosed by the invention, and these distortion and combination are still in protection scope of the present invention.

Claims (10)

1. in a 3-D seismics signal supervision waveform classification arranged, it is characterized in that, comprise the following steps:

A, data pre-service: comprise that Noise reducing of data is processed, extraction objective interval data reach well logging data analysis is set up to training sample and label;

B, feature selecting: analyze training sample and the tag along sort data thereof extracted, select in sample feature with the maximally related a certain amount of feature of tag along sort as character subset, the character subset of usining after preferred in objective interval data and training sample is as the input sample set of class indication in this programme;

C, class indication: set up disaggregated model seismic signal is carried out to class indication.

2. in a kind of 3-D seismics signal according to claim 1 the supervision waveform classification arranged, it is characterized in that, in described step a, described denoise processing method is as follows:

A11, input geological data;

A12, judge whether to reach maximum iteration time, if finish the noise reduction process process, otherwise execution step a13;

A13, the Grad of calculating earthquake number strong point on three directions the further structure tensor that calculates this data point;

A14, utilize three-dimensional gauss low frequency filter to carry out the filtering processing to described structure tensor;

A15, structure diffusion matrix and discontinuous factor;

A16, according to diffusion equation to data filtering.

3. in a kind of 3-D seismics signal according to claim 1 the supervision waveform classification arranged, it is characterized in that, in described step a, described extraction objective interval data method is as follows:

A21, the data of step a16 gained are done to the single layer bit analyzing and processing;

A22, the data of step a16 gained are done to two-layer interdigit analyzing and processing.

4. in a kind of 3-D seismics signal according to claim 3 the supervision waveform classification arranged, it is characterized in that, described step a21 is further comprising the steps of:

A211, true timing window and analysis layer position;

Window data when a212, extraction edge layer;

A213, acquisition objective interval data.

5. in a kind of 3-D seismics signal according to claim 3 the supervision waveform classification arranged, it is characterized in that, described step a22 is further comprising the steps of:

A221, the two-layer position of definite analysis;

A222, extraction are along the interlayer data;

A223, acquisition objective interval data.

6. in a kind of 3-D seismics signal according to claim 1 the supervision waveform classification arranged, it is characterized in that, in described step a, the described concrete grammar that well logging data analysis is set up to training sample and label is: according to log analysis data in the geological data work area, well logging in work area is carried out to the manual sort to be meaned, set up the training sample label, and the data of extracting well logging and well lie position in objective interval are as training sample, for feature selecting and class indication.

7. in a kind of 3-D seismics signal according to claim 1 the supervision waveform classification arranged, it is characterized in that, in described step b, feature selecting adopts genetic algorithm, the training sample set that wherein extracts well logging and well lie is combined into T, corresponding tag along sort is y, the feature selecting number is that k and maximum iteration time are maxI, specifically comprises the following steps:

B1, according to the gene code rule, obtain the initialization population of genetic algorithm;

B2, basis

fitness function calculates the fitness of population at individual, wherein x _i(i=1,2 ..., k) be x in the selected training sample set of population at individual T _ithe dimensional feature component, the tag along sort that y is corresponding well logging; MIx _iy means the x of log data _ithe mutual information of dimensional feature component and tag along sort y; MIx _ix _jthe x that means log data _icharacteristic component and characteristic component x _jbetween mutual information;

B3, judge whether to reach maximum iteration time maxI or the continuous optimum solutions that keep for five times are constant, if, export current optimal feature subset, otherwise execution step b4;

B4, according to the population at individual fitness, carry out to select operation;

B5, execution interlace operation;

B6, execution mutation operation;

B7, return to b2 operation;

B8, output characteristic subset.

8. in a kind of 3-D seismics signal according to claim 7 the supervision waveform classification arranged, it is characterized in that, in the process of processing in described genetic algorithm, the gene code of step b1 adopts { 0, the 1} binary string means, which position in binary string just means the seismic channel data u that extracts in the data pre-service _i={ x _i1, x _i2..., x _iMand log data w _i={ x _i1, x _i2..., x _iMin which dimensional feature of feature space, 0 means that this characteristic component is selected, 1 means to select this characteristic component.

9. in a kind of 3-D seismics signal according to claim 1 the supervision waveform classification arranged, it is characterized in that, in described step c, use support vector machine to set up disaggregated model, and the objective interval data are carried out to class indication, finally make waveform separation division figure.

10. according in the described a kind of 3-D seismics signal of claim 1 or 9 the supervision waveform classification being arranged, it is characterized in that, the detailed process of described step c is:

C1, the character subset data normalization is processed;

C2, use 10 folding cross-validation methods to receive the best SVM parameters of rope;

C3, set up the svm classifier model;

C4, to selecting the objective interval data after feature, carry out Classification and Identification;

C5, drawing waveforms classifying and dividing figure.

Images