CN110674868A - Stratum lithology identification system and method based on high-dimensional drilling parameter information - Google Patents

Abstract

The invention provides a stratum lithology identification system and method based on high-dimensional drilling parameter information, including a drilling test bench for constructing data holes to obtain high-dimensional drilling parameters, each group of high-dimensional drilling parameters are respectively The corresponding training samples and prediction samples are formed; the data dimensionality reduction system is used to calculate the high-dimensional drilling parameters in the training samples, determine the number of principal components, and obtain the classification number of the preliminary training samples and the principal components of the training samples. data set, and then obtain the principal component data set of the prediction sample; the data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training sample, and obtain the best clustering of the principal component data set of the training sample. The optimal number K is used to obtain the principal component data set of the training samples of the definite classification; the prediction recognition system is used to establish the discriminant criterion for the principal component data set of the training samples of the definite classification, classify the predicted samples, and obtain the predicted samples The lithology category to which it belongs.

Description

A formation lithology identification system and method based on high-dimensional drilling parameter information

technical field

The invention belongs to the field of geotechnical engineering survey, relates to a formation lithology identification system, in particular to a formation lithology identification system and method based on high-dimensional drilling parameter information.

Background technique

Due to the vast territory of our country and the extremely complex geological conditions, the rock properties and rock formations in the stratum have obvious spatial differences, and the rationality of engineering design and the safety guarantee of the mining process have been greatly affected. At present, geological drilling and geophysical exploration methods are usually used to identify the stratum. The lithology identification of the stratum has shortcomings such as low identification accuracy and rough identification methods. , The real-time drilling parameter set obtained by drilling equipment has a large amount of data and a high dimension. The lithological characteristics of the formation are usually related to a variety of drilling parameters, and the information contained in each drilling parameter is crossed, and some are even redundant. Therefore, it is impossible to use the multi-parameter linear method for lithology identification, and there is no suitable data processing and analysis method. At present, domestic and foreign scholars have carried out various numerical simulation and experimental methods to study drilling parameter data sets. Since a single or multiple index parameters and their thresholds are used to judge, the amount of information is relatively small, and the thresholds of various parameters in different mines are also different. When multiple indexes are close to the thresholds to different degrees, it is still not very clear how to comprehensively judge. It is a good solution. At the same time, the amount of calculation of the measured data is huge, and each parameter data has the problem of mutual interference. It is impossible to perform calculation and analysis in real time, so that it is impossible to make real-time and reliable prediction of the formation.

SUMMARY OF THE INVENTION

In view of the deficiencies in the prior art, the purpose of the present invention is to provide a formation lithology identification system and method based on high-dimensional drilling parameter information, so as to solve the technical problem of insufficient identification accuracy in the prior art.

In order to solve the above-mentioned technical problems, the present invention adopts the following technical solutions to realize:

A formation lithology identification system based on high-dimensional drilling parameter information, including a drilling test bench, a data dimensionality reduction system, a data clustering system, and a prediction and identification system connected in sequence;

The drilling test bench is used for constructing data holes to obtain high-dimensional drilling parameters, and each group of high-dimensional drilling parameters constitutes corresponding training samples and prediction samples respectively;

The data dimensionality reduction system is used to calculate the high-dimensional drilling parameters in the training samples to determine the number of principal components; obtain the classification number of the preliminary training samples and the data sets of the principal components of the training samples; and then obtain the prediction the principal component data set of the sample;

The data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training sample, obtain the optimal number K of the principal component data set clustering of the training sample, and obtain the training sample of the determined classification Principal Component Component Dataset;

The prediction and identification system is used to establish a discriminant criterion for the main component data set of the training samples for which classification is determined, to classify the predicted samples, and to obtain the lithology category to which the predicted samples belong.

The present invention also has the following technical features:

The drilling test bench includes a hydraulic pump station, an operation platform, a flushing fluid circulation unit and a data acquisition unit respectively connected with the main engine.

The data dimensionality reduction system includes an input terminal 1, a data processor 1 and an output terminal 1 which are connected in sequence.

The data clustering system includes two input terminals, two data processors and two output terminals connected in sequence.

The predictive identification system includes three input terminals, three data processors and three output terminals which are connected in sequence.

The present invention also protects a formation lithology identification method based on high-dimensional drilling parameter information, which adopts the above-mentioned formation lithology identification system based on high-dimensional drilling parameter information; the high-dimensional drilling parameters include: ROP, swing torque, weight on bit, rotational speed, swing pressure and mud pump pressure.

The method specifically includes the following steps:

Step 1: According to the production data of the target stratum area, determine the rock formation characteristics and prediction range of the typical stratum, use the main engine construction data hole of the drilling test bench, and obtain high-dimensional drilling parameters by using the data acquisition unit of the drilling test bench. The group of high-dimensional drilling parameters constitutes the corresponding training samples and prediction samples respectively;

In step 2, the data dimensionality reduction system is used to calculate the high-dimensional drilling parameters in the training samples, and the correlation coefficients between the high-dimensional drilling parameters are obtained, and then the contribution rate of each preset principal component is obtained. Sort from small to small, and obtain the cumulative contribution rate of each preset pivot component; when the cumulative contribution rate of a preset pivot component is greater than 90%, then all preset pivot components before the preset pivot The main component is the main component, and the number of main components and the corresponding eigenvectors are finally determined; and the contribution rate of each main component is used as the weight for weighted calculation, and the weighted score of each training sample is obtained. The weighted scores are sorted from high to low. According to the weighted score of each training sample, the training samples are initially classified and the number of classifications of the preliminary training samples is obtained. According to the feature vector of the main component, the main component data set and Principal component data set of predicted samples;

In step 3, the data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training samples obtained in step 2, and the number of classifications of the preliminary training samples obtained in step 2 is used as the number of original clusters, and the ambiguity is set. And construct a kernel function, establish a membership matrix, and optimize the parameters through continuous iteration, and finally complete the clustering of the principal component data set of the training sample, and obtain the optimal number K of the principal component data set clustering of the training sample, so The optimal number K is the number of lithological classifications, and at the same time, the cluster center of each lithological classification and the corresponding data set of each lithological classification are calculated, and the main component data set of the training sample for the classification is obtained;

In step 4, the prediction and recognition system is used to establish a discrimination criterion for the principal component data set of the training sample of the definite classification obtained in step 3, and each principal component data set of the predicted sample obtained in step 2 is calculated by the Mahalanobis distance judgment method. The Mahalanobis distance of the main component data set of the training sample is selected, the smallest Mahalanobis distance is selected, and the predicted samples are classified to obtain the lithology category to which the predicted samples belong.

Compared with the prior art, the present invention has the following technical effects:

The invention has high identification accuracy, short clustering time, fast data processing speed, effectively utilizes high-dimensional drilling parameter data sets, filters redundant and incorrect parameter information, and can identify formation lithology and structure information in real time. The prediction of formation lithology is of great practical significance, and it is of great significance to realize dynamic intelligent detection of hidden disaster-causing factors in rock formations. It can not only provide accurate information for the intelligent identification of strata, but also provide reference and guidance for the prediction of rock characteristics in other geotechnical engineering constructions such as wells and roadways, slope engineering, etc. At the same time, it provides ideas and methods for the processing of high-dimensional data in engineering practice.

Description of drawings

FIG. 1 is a schematic structural diagram of a formation lithology identification system based on high-dimensional drilling parameter information of the present invention.

FIG. 2 is a schematic structural diagram of the drilling test bench of the present invention.

FIG. 3 is a schematic structural diagram of the data dimensionality reduction system of the present invention.

FIG. 4 is a schematic structural diagram of the data clustering system of the present invention.

FIG. 5 is a schematic structural diagram of the predictive identification system of the present invention.

FIG. 6 is a schematic diagram of the clustering effect of the data clustering system of the practical application example of the present invention.

The meanings of the symbols in the figure are: 1 is the drilling test bench, 2 is the data dimensionality reduction system, 3 is the data clustering system, 4 is the prediction and identification system, 5 is the hydraulic pump station, 6 is the operating table, and 7 is the flushing fluid Circulation system, 8 is the host computer, 9 is the data acquisition system, 10 is the input terminal one, 11 is the data processing system one, 12 is the output terminal one, 13 is the input terminal two, 14 is the data processing system two, and 15 is the output terminal two , 16 is the input terminal three, 17 is the data processing system three, and 18 is the output terminal three.

The specific content of the present invention will be further explained in detail below in conjunction with the embodiments.

Detailed ways

The purpose of the present invention is to provide a formation lithology identification system and method based on high-dimensional drilling parameter information, which can process high-dimensional drilling parameter information, reduce the mutual influence of drilling parameters, and obtain accurate formation lithology characteristics in real time. , which can provide effective means for scientific design and optimization of construction parameters of underground drilling in coal mines. It can not only provide theoretical support for stratum lithology identification, but also provide reference and guidance for the processing of other high-dimensional parameter data sets such as geotechnical engineering construction, and provide ideas and methods for fault diagnosis of drilling equipment.

The invention provides a formation lithology identification system based on high-dimensional drilling parameter information, including a drilling test bench, a data dimensionality reduction system, a data clustering system, and a prediction and identification system connected in sequence.

The drilling test bench is used for constructing data holes to obtain high-dimensional drilling parameters, and each group of high-dimensional drilling parameters constitutes corresponding training samples and prediction samples respectively;

The data dimensionality reduction system is used to reduce the dimensionality of the high-dimensional drilling parameter information set to a new comprehensive principal parameter that is independent of each other, while maintaining the information of the original data set to the greatest extent, and is used to eliminate the amount of each parameter. Calculate the high-dimensional drilling parameters in the training samples to determine the number of principal components; obtain the classification number of the preliminary training samples and the principal component data set of the training samples; and then obtain the principal component data of the predicted samples set;

The data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training sample, obtain the optimal number K of the principal component data set clustering of the training sample, and obtain the training sample of the determined classification Principal Component Component Dataset;

The described prediction and identification system uses the Mahalanobis distance judgment method to calculate the Mahalanobis distances between the predicted samples and the training sample data set of the determined classification, and compare them to identify the lithology of the predicted samples.

The drilling test bench includes a hydraulic pump station, an operation platform, a flushing fluid circulation unit and a data acquisition unit respectively connected with the main engine;

The host is used to drive drilling tool construction data to drill;

The hydraulic pump station is used to provide the motive power to the main engine;

The operating console is used to operate the control host;

The data acquisition unit is used for collecting and transmitting high-dimensional drilling parameters;

The flushing fluid circulation unit is used for cooling the drill bit and discharging the slag;

The data dimensionality reduction system comprises an input terminal 1, a data processor 1 and an output terminal 1 which are connected in sequence;

Described input terminal one is used for receiving training samples and prediction samples;

The data processor 1 is used to standardize the high-dimensional drilling parameter information data set in the training sample and the prediction sample, calculate the correlation coefficient, the contribution rate and the cumulative contribution rate, and finally determine the number of principal components and the corresponding feature vector; obtain a preliminary The number of classifications of the training samples and the principal component data set of the training samples; and then the principal component data set of the predicted samples is obtained;

The output terminal 1 is mainly used for the output of the classification number of the preliminary training samples, the principal component data set of the training samples and the principal component data set of the prediction samples.

The data clustering system includes two input terminals, two data processors and two output terminals connected in sequence,

Described input end 2 is used for receiving the classification number of preliminary training samples and the principal component data set of training samples obtained by output end 1;

The second data processor is used to use the classification number of the preliminary training samples as the original number of clusters, set the ambiguity, construct a kernel function, establish a membership matrix, optimize parameters through continuous iteration, calculate the cluster center, and obtain Obtain the optimal number K of the training sample principal component data set clustering and determine the training sample principal component data set of classification;

The second output terminal is mainly used for outputting the training sample principal component data set for determining the classification.

The predictive identification system includes three input terminals, three data processors and three output terminals connected in sequence;

The third input terminal is used to receive the main component data set of the prediction sample obtained by the output device 1 and the training sample main component data set of the determined classification obtained by the output device 2;

The data processor 3 establishes a discriminant function, establishes a discriminant criterion, and calculates the Mahalanobis distance between each main component data set of the predicted sample and the main component data set of the training sample of the determined classification through the Mahalanobis distance judgment method, and selects the smallest among them. The Mahalanobis distance of the predicted samples is classified to obtain the lithology category to which the predicted samples belong;

The output terminal 3 can be used for outputting lithology identification results.

In the present invention, the method for determining the accuracy of the system and the correction is as follows:

Step 1: Measure the physical and mechanical parameters of hardness and strength of each predicted sample through on-site tests, obtain the material properties of the predicted sample on site, determine the actual lithology category of the predicted sample, and compare it with the predicted result to obtain the predicted accuracy of the predicted sample. .

Step 2, when the prediction accuracy rate is less than 99%, drill holes into the test bench construction data to increase the training sample data set, and repeat steps 1 to 3 to obtain a new training sample principal component data set that determines the classification. Use step 4 to identify the prediction samples, and obtain a new prediction accuracy rate. By increasing the number of training samples, the prediction results are continuously optimized until the prediction accuracy rate is greater than or equal to 99%. Otherwise, steps 1 to 4 will be repeated.

Step 3: Using the final optimized and classified training sample principal component data set, through real-time drilling construction, a high-dimensional drilling parameter set is obtained in real time as a prediction sample, and a real-time prediction sample is obtained by a data dimensionality reduction system. Through the above steps, the lithology of the formation can be identified in real time and the operating conditions of the drilling equipment can be obtained.

Specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent transformations made on the basis of the technical solutions of the present application fall into the protection scope of the present invention.

Example 1:

Following the above technical solutions, as shown in Figures 1 to 5, this embodiment provides a formation lithology identification system based on high-dimensional drilling parameter information, including a drilling test bench, a data dimensionality reduction system, a data Clustering system, prediction and identification system;

Drilling into the test bench is used to construct data holes and obtain high-dimensional drilling parameters. Each group of high-dimensional drilling parameters is composed of corresponding training samples and prediction samples;

The data dimensionality reduction system is used to calculate the high-dimensional drilling parameters in the training samples to determine the number of principal components; obtain the classification number of the preliminary training samples and the principal component data set of the training samples; and then obtain the main components of the prediction samples. meta-component dataset;

The data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training sample, and obtain the optimal number K of the training sample principal component data set clustering, and obtain the training sample principal component of the classification. data set;

The prediction and identification system is used to establish a discriminant criterion for the main component data set of the training samples for which classification is determined, to classify the predicted samples, and to obtain the lithology category to which the predicted samples belong.

Specifically, the drilling test bench includes a hydraulic pump station, an operation platform, a flushing fluid circulation unit and a data acquisition unit respectively connected with the main engine.

Specifically, the data dimensionality reduction system includes an input terminal 1, a data processor 1 and an output terminal 1 which are connected in sequence.

Specifically, the data clustering system includes two input terminals, two data processors, and two output terminals that are connected in sequence.

Specifically, the predictive identification system includes three input terminals, three data processors and three output terminals, which are connected in sequence.

Example 2:

Following the above technical solution, the present embodiment provides a formation lithology identification method based on high-dimensional drilling parameter information, and the method adopts the formation lithology identification system based on high-dimensional drilling parameter information as in Embodiment 1; the The high-dimensional drilling parameters include ROP, swing torque, WOB, rotational speed, swing pressure and mud pump pressure. The method specifically includes the following steps:

Step 1: According to the production data of the target stratum area, determine the rock formation characteristics and prediction range of the typical stratum, use the main engine construction data hole of the drilling test bench, and obtain high-dimensional drilling parameters by using the data acquisition unit of the drilling test bench. The group of high-dimensional drilling parameters constitutes the corresponding training samples and prediction samples respectively;

In step 2, the data dimensionality reduction system is used to calculate the high-dimensional drilling parameters in the training samples, and the correlation coefficients between the high-dimensional drilling parameters are obtained, and then the contribution rate of each preset principal component is obtained. Sort from small to small, and obtain the cumulative contribution rate of each preset pivot component; when the cumulative contribution rate of a preset pivot component is greater than 90%, then all preset pivot components before the preset pivot The main component is the main component, and the number of main components and the corresponding feature vector are finally determined; and the contribution rate of each main component is used as the weight for weighted calculation to obtain the weighted score of each training sample. Sort the weighted scores from high to low. According to the weighted score of each training sample, initially classify the training samples and obtain the classification number of the preliminary training samples, and obtain the main component data set of the training samples according to the feature vector of the main component. and the Principal Component Component dataset of predicted samples.

In step 2, the following steps are specifically included:

Step 2.1, data standardization of high-dimensional drilling parameters:

In order to make the results not affected by the dimension, the high-dimensional drilling parameters are first standardized, so that the mean of each index data is 0 and the standard deviation is 1. The standardization method is:

in,

Among them, α _ij is the index value of the j-th high-dimensional drilling parameter of the i-th evaluation object in the data set, i is the serial number of the evaluation object corresponding to a high-dimensional drilling parameter in the data set, and j is the high-dimensional drilling parameter in the data set. The serial number of the input parameter, μ _j is the average value of α _ij , and s _j is the variance of α _ij .

Step 2.2, calculation of correlation coefficient of high-dimensional drilling parameters:

Calculate the correlation coefficient between the high-dimensional drilling parameters, and construct the correlation coefficient matrix. Let the correlation coefficient matrix be R(r _wq ), where

Among them, r _wq is the correlation coefficient between the w-th high-dimensional drilling parameter and the q-th high-dimensional drilling parameter in the correlation coefficient matrix, and w and q are the serial numbers of the high-dimensional drilling parameters in the data set.

Step 2.3, construct the preset pivot component:

Calculate the eigenvalues λ ₁ ≥λ ₂ ≥...≥λ _N ≥0 of the correlation coefficient matrix R(r _wq ) and its corresponding eigenvectors μ ₁ , μ ₂ ,..., μ _N , denoted μ _j = (μ _1j , μ _2j ,…,μ _Nj ) ^T , do a linear combination:

Wherein, y _N represents the index value of the Nth preset principal component.

Step 2.4 Select the principal component:

The contribution rate of the preset pivot component is calculated, and its size reflects the influence of the preset pivot component. And calculate the cumulative contribution rates of the preset principal components as follows:

Among them, b _j represents the contribution rate of the preset principal component, and α _j represents the cumulative contribution rate of the preset principal component.

When the cumulative contribution rate α _j is close to 1, extract the preset principal component corresponding to the eigenvalues whose cumulative contribution rate is greater than 90%, and select the first p (p≤N) index variables y ₁ , y ₂ ,...,y _p As the p principal component, where p represents the sequence number of the preset principal component corresponding to the eigenvalue whose cumulative cumulative contribution rate is greater than 90%, that is, the number of the principal component.

Step 2.5 Get the weighted score of the training samples:

Let the weighted score be Z, then

According to the weighted score of each training sample, sort from high to low, according to the weighted score of each training sample, preliminarily classify the training samples and obtain the classification number of the initial training sample as q, and obtain the training according to the feature vector of the principal component. A dataset of principal component components of the sample and a dataset of principal component components of the predicted sample.

In step 3, the data clustering system is used to perform fuzzy kernel clustering on the principal component data set of the training samples obtained in step 2, and the number of classifications of the preliminary training samples obtained in step 2 is used as the number of original clusters, and the ambiguity is set. And construct a kernel function, establish a membership matrix, and optimize the parameters through continuous iteration, and finally complete the clustering of the principal component data set of the training sample, and obtain the optimal number K of the principal component data set clustering of the training sample, so The optimal number K is the number of lithological classifications. At the same time, the cluster center of each lithological classification and the corresponding data set of each lithological classification are calculated, and the main component data set of the training sample for the classification is obtained.

Step 3 specifically includes the following steps:

Step 3.1, set the number of original classifications as q, set the ambiguity m, the accuracy of the objective function ε, set the parameters of the appropriate kernel function and construct the kernel function;

Step 3.2, establish a membership degree matrix, and initialize the membership degree matrix;

Step 3.3, calculate the cluster center;

为高维特征空间中第i个样本x_i的距离。where x _i is the i-th sample in the original feature space, i=1, 2,...,n; μ _iu is the membership degree of the i-th sample x _i to the u-th class, u=1,..., q;, μ _iu ∈ [0,1]; m is the ambiguity; v _a is the cluster center of the a-th class in the high-dimensional feature space, a=1,2,…,q;

is the distance of the _ith sample xi in the high-dimensional feature space.

The objective function can be minimized by making its partial derivative with respect to the membership matrix zero, then

r=1,2,…,n; s=1,2,…,q

where μ _rs is the degree of membership of the r-th sample x _r to the s-th class, is the distance between the rth sample x _r in the high-dimensional feature space and the cluster center of the jth class in the high-dimensional feature space.

Step 3.4, compare the iterative membership degree matrix according to the matrix norm, if it converges, stop the iteration, otherwise return to step 3.3.

The optimal number K of the principal component data set clustering of the training sample is obtained, and the optimal number K is the number of lithological classifications. At the same time, the cluster center and each lithological classification are calculated. The corresponding data set of lithology classification is obtained, and the main component data set of the training sample for the classification is obtained.

In step 4, the prediction and recognition system is used to establish a discrimination criterion for the principal component data set of the training sample of the definite classification obtained in step 3, and each principal component data set of the predicted sample obtained in step 2 is calculated by the Mahalanobis distance judgment method. The Mahalanobis distance of the main component data set of the training sample is selected, the smallest Mahalanobis distance is selected, and the predicted samples are classified to obtain the lithology category to which the predicted samples belong.

The Mahalanobis distance judgment method described in step 4 includes the following steps:

Each principal component data set of the prediction sample is calculated in pairs with the training sample principal component data set of the determined classification, and two of the training sample principal component data sets of the determined classification of A and B are used as examples to introduce.

Step 4.1, calculate the mean vector and covariance matrix of A and B;

ma=mean(A), mb=mean(B), S ₁ =cov(A), S ₂ =cov(B)

Step 4.2, calculate the covariance matrix of the population;

where n ₁ and n ₂ are the capacities of A and B, respectively, ma is the mean vector of A, mb is the mean vector of B, S ₁ is the covariance matrix of A, S ₂ is the covariance matrix of B, and S is the The covariance matrix of the population.

Step 4.3, calculate the difference d between the Mahalanobis squared distances between the main component data set x of the predicted sample and the two classes of A and B;

d=(x-ma)S ^-1 (x-ma) ^T- (x-mb)S ^-1 (x-mb) ^T

Step 4.4, if d<0, then x belongs to class A; if d>0, then x belongs to class B.

Application example:

Take a coal mine roadway field test as an example for analysis. A total of 2900 sets of high-dimensional drilling parameters were measured using the drilling test bench, including 2900 training samples and 50 test samples.

First, standardize the high-dimensional drilling parameter data of the training samples. In order to eliminate the influence of the dimensions of each parameter, the measured high-dimensional drilling parameters are standardized, and the ROP, rotary torque, drilling pressure, rotational speed, rotary pressure, The standardized variables of the mud pump pressure are respectively x ₁ , x ₂ , x ₃ , x ₄ , x ₅ , x ₆ , and each preset pivot component is defined as y ₁ , y ₂ , y ₃ , y ₄ , y ₅ , y ₆ , the correlation coefficient matrix can be calculated, and the results are shown in Table 1 below.

Table 1 Training sample correlation coefficient table

Then, the contribution rate and cumulative contribution rate of each preset principal component can be calculated through the data dimensionality reduction system, as shown in Table 2.

Table 2 Preset principal component result table

According to relevant empirical principles, the principle that the cumulative contribution rate is greater than 90% determines that the number of selected principal component is 3, and the feature vector of the principal component of the training sample can be obtained as shown in Table 3 below.

Table 3 Eigenvectors of principal components of training samples

The first principal component mainly reflects x ₂ (swing torque); the second principal component mainly reflects x ₁ (ROP); the third principal component mainly reflects x ₅ (swing pressure). Calculate the dimension reduction of the training samples to obtain the principal component data set of the training samples. Through the weighted score, the number of classifications of the initial training sample is 3, and then the data set of the main component of the predicted sample is obtained.

By selecting an appropriate kernel function and using the data clustering system to cluster the principal component data sets of the training samples, three data sets of the principal components of the training samples with definite classification can be obtained, and the cluster centers can be obtained as follows: v ₁ = (0.1262, -1.0885, -0.7335) ^T , v ₂ =(1.6849, -0.7538, -1.2834) T, v3 = (0.4876, -0.8157, 0.7405) T, corresponding to sandstone, coal and mudstone layers respectively.

The clustering effect is shown in Figure 6. Then, the Mahalanobis distance between each principal component data set of the predicted sample and the three main component data sets of the training samples determined to be classified is calculated by the Mahalanobis distance judgment method, and the predicted samples are classified. , obtain the lithology category to which the predicted sample belongs, and compare it with the actual lithology category of the predicted sample, and continuously optimize the prediction results until the prediction accuracy rate is greater than or equal to 99%. The prediction results are shown in Table 4 below.

Table 4 Prediction effect table

In this embodiment, although the above-described methods are illustrated and described as a series of acts for simplicity of explanation, it is to be understood and appreciated that the methods are not limited by the order of the acts, as some acts in accordance with one or more embodiments It may occur in a different order and/or concurrently with other actions from or not illustrated and described herein but which can be understood by those skilled in the art.

Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

It is noted that references in the specification to "one embodiment," "an embodiment," "example embodiment," "some embodiments," etc. indicate that the described embodiment may include the particular feature, structure, or characteristic, but each implementation Examples may not necessarily include the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Furthermore, when a particular feature, structure or characteristic is described in conjunction with one embodiment, whether explicitly described or not, it will be within the knowledge of those skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments.

The previous description of the present disclosure is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to the present disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the examples and designs described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (7)

1. A stratum lithology recognition system based on high-dimensional drilling parameter information is characterized by comprising a drilling test bed, a data dimension reduction system, a data clustering system and a prediction recognition system which are sequentially connected;

the drilling test bed is used for constructing data holes and obtaining high-dimensional drilling parameters, and each group of high-dimensional drilling parameters respectively form a corresponding training sample and a corresponding prediction sample;

the data dimension reduction system is used for calculating high-dimensional drilling parameters in the training sample and determining the number of principal component components; obtaining the classification number of the preliminary training sample and the principal component data set of the training sample; further obtaining a principal component data set of the prediction sample;

the data clustering system is used for carrying out fuzzy kernel clustering on the principal component data sets of the training samples to obtain the optimal number K of the principal component data sets of the training samples to be clustered, and obtaining the principal component data sets of the training samples to be classified;

the prediction identification system is used for establishing a discrimination criterion for the principal component data set of the training samples determined to be classified, classifying the prediction samples and obtaining the lithology categories to which the prediction samples belong.

2. The formation lithology recognition system based on high-dimensional drilling parameter information as claimed in claim 1, wherein the drilling test bench comprises a hydraulic pump station, an operation bench, a flushing liquid circulation unit and a data acquisition unit which are respectively connected with a host machine.

3. The formation lithology recognition system based on high-dimensional drilling parameter information as claimed in claim 1, wherein the data dimension reduction system comprises a first input end, a first data processor and a first output end which are connected in sequence.

4. The system for identifying the lithology of the stratum based on the high-dimensional drilling parameter information as claimed in claim 1, wherein the data clustering system comprises a second input end, a second data processor and a second output end which are connected in sequence.

5. The formation lithology recognition system of claim 1, wherein the predictive recognition system comprises a third input terminal, a third data processor and a third output terminal connected in series.

6. A stratum lithology recognition method based on high-dimensional drilling parameter information is characterized in that the method adopts the stratum lithology recognition system based on the high-dimensional drilling parameter information; the high-dimensional drilling parameters comprise the mechanical drilling speed, the rotary torque, the drilling pressure, the rotating speed, the rotary pressure and the pressure of a slurry pump.

7. The method for identifying the lithology of the stratum based on the high-dimensional drilling parameter information as claimed in claim 6, wherein the method comprises the following steps:

determining the rock stratum characteristics and the prediction range of a typical stratum according to the production information of a target stratum region, constructing a data hole by adopting a host of a drilling test bed, and obtaining high-dimensional drilling parameters by adopting a data acquisition unit of the drilling test bed, wherein each group of high-dimensional drilling parameters respectively form a corresponding training sample and a prediction sample;

calculating high-dimensional drilling parameters in the training sample by using a data dimension reduction system to obtain correlation coefficients among the high-dimensional drilling parameters, then obtaining the contribution rate of each preset principal component, sequencing from large to small, and obtaining the accumulated contribution rate of each preset principal component; when the accumulated contribution rate of a certain preset principal component is greater than 90%, all preset principal components before the preset principal component are principal components, and finally the number of the principal components and the corresponding feature vectors are determined; performing weighted calculation by taking the contribution rate of each principal component as weight to obtain the weighted score of each training sample, sequencing the training samples from high to low according to the weighted score of each training sample, preliminarily classifying the training samples according to the weighted score condition of each training sample to obtain the classification number of the preliminary training samples, and obtaining a principal component data set of the training samples and a principal component data set of the prediction samples according to the feature vectors of the principal components;

thirdly, fuzzy kernel clustering is carried out on the principal component data sets of the training samples obtained in the second step by adopting a data clustering system, the classification number of the preliminary training samples obtained in the second step is used as the original clustering number, the fuzzy degree is set, a kernel function is constructed, a membership matrix is established, clustering of the principal component data sets of the training samples is finally completed through continuous iteration optimization parameters, the optimal number K of the principal component data sets of the training samples is obtained, the optimal number K is the lithologic classification number, and meanwhile, the clustering center of each lithologic classification and the corresponding data set of each lithologic classification are calculated to obtain the principal component data sets of the training samples with determined classifications;

and step four, establishing a judgment criterion for the principal component data sets of the training samples of the determined classification obtained in the step three by adopting a prediction recognition system, calculating the Mahalanobis distance between each principal component data set of the prediction samples obtained in the step two and the principal component data sets of the training samples of the determined classification by using a Mahalanobis distance judgment method, selecting the smallest Mahalanobis distance, and classifying the prediction samples to obtain the lithology categories to which the prediction samples belong.

Images