CN116108392A - Geological structure identification technology based on improved random forest algorithm - Google Patents
Geological structure identification technology based on improved random forest algorithm Download PDFInfo
- Publication number
- CN116108392A CN116108392A CN202111320576.2A CN202111320576A CN116108392A CN 116108392 A CN116108392 A CN 116108392A CN 202111320576 A CN202111320576 A CN 202111320576A CN 116108392 A CN116108392 A CN 116108392A
- Authority
- CN
- China
- Prior art keywords
- random forest
- model
- identification
- seismic
- structure identification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a geological structure identification technology based on an improved random forest algorithm, which comprises the steps of firstly preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration; selecting a random forest parameter pair, optimizing the random forest parameter by utilizing improved step-by-step long grid search, and determining an optimal parameter pair; training an improved random forest classification model by utilizing the preprocessed seismic attribute data set, and analyzing the model evaluation index result to obtain an optimal random forest algorithm recognition model; the obtained recognition model is used for geological structure recognition, and a structure recognition result distribution diagram is obtained through visual processing; the method solves the problems of high requirements on the seismic attribute data set and complex data processing in geological structure identification, improves the accuracy of identification, and is suitable for identifying the collapse column and fault structure in three-dimensional seismic exploration.
Description
Technical Field
The invention belongs to the field of seismic exploration, and particularly relates to a geological structure identification technology based on an improved random forest algorithm.
Background
As coal resources are mined to deep, the difficulty of coal seam mining increases, and various complex geological structures seriously affect the safety of coal mining personnel. The seismic attribute is the geometric form, the kinematic characteristic, the dynamic characteristic and the statistical characteristic of the related seismic wave which are derived through mathematical transformation, and abnormal body (structure) information can be revealed by analyzing the seismic attribute and calibrating the seismic attribute. However, the complexity of the underground geological condition and the influence factors of the seismic information are too many, larger uncertainty or ambiguity exists, the structure identification can not be accurately carried out by applying any single seismic attribute, and the development of seismic multi-attribute fusion analysis is very necessary.
The research of seismic attribute fusion is various, and Balch displays seismic data in color in 1971, so that the identification capability of underground geological anomalies is improved. In 2002, our country Le Youxi teaches that the method of cluster analysis is preferentially applied to seismic attribute fusion, and that multiple linear regression can also be applied to attribute fusion (Ji Yuxin and euclidean, 2003). Along with the coming of big data age, the current development is faster based on fusion of seismic attribute data, namely, optimal seismic attributes are extracted through modes of mathematical statistics, artificial intelligence and the like, for example, in 2010, cao Lin applies a BP network technology based on particle swarm optimization to multi-attribute fusion for the first time. The neural network fusion attribute method has the advantages of high identification speed, strong self-adaptability and fault tolerance, and wide application range; however, this approach does not autonomously prefer properties while requiring sufficient sample data to train the network; in 2012, bruno uses PCA for fault identification, and the new fused attribute is obtained by fusing the seismic attribute with PCA, so that the accuracy of micro fault identification is greatly improved; however, PCA is a linear dimension reduction method, and when nonlinear relation exists in data, the effect of PCA is greatly reduced; in 2017, sun Zhenyu uses an SVM algorithm for identifying small faults of earthquakes, an SVM model fuses advantages of all attribute prediction faults, fault information is mined from different angles, influence of subjective factors of interpretation personnel on interpretation results is reduced, but when the SVM model is built, the structure of the model directly influences the model identification accuracy, and the influence of the selection of the earthquake attribute on the model accuracy is also great.
In recent years, the seismic attribute fusion technology has been rapidly developed, and has been widely applied to various fields such as reservoir prediction, sand prediction, and structure identification. In the seismic attribute fusion process, an algorithm which is high in accuracy and suitable for various sample data sets needs to be selected, so that the seismic attribute data are explained more effectively, and the accuracy of constructing an identification model is improved.
Aiming at the problems of high requirements on seismic attribute data sets and complex data processing in the existing structure identification research, the invention provides an improved random forest algorithm based on a classical machine learning random forest algorithm, combines a seismic multi-attribute fusion technology with the improved random forest algorithm, and establishes a geological structure identification model based on the improved random forest algorithm.
Disclosure of Invention
The random forest algorithm is popular in recent years as a highly flexible algorithm, and has wide application prospect. In all the current algorithms, the random forest algorithm as an integrated algorithm has better precision than most single algorithms, high accuracy and low requirements on data sets, and is suitable for various data sets (linear and nonlinear, high-dimensional data sets and the like). The random forest algorithm flow is shown in fig. 2, training sample sets which are mutually different are generated through a Bagging (integration) method, the algorithm is mainly used for classification and regression, and the random forest classification algorithm is used for geological structure identification.
The randomness of the random forest algorithm is in two aspects: the training sample of each tree is random, and the splitting attribute set of each node in the tree is also randomly selected and determined. Because of the two randomness, the random forest is insensitive to noise data, and the problem of over fitting is overcome. However, the number k of decision trees in the random forest, the maximum feature number m of a single decision tree and other parameters are optimized and selected relatively less, and in general, the parameters are selected through experience, and may not be optimal parameters, so that how to select the optimal feature number is critical.
Aiming at the problems, the invention provides an improved grid search algorithm, which utilizes improved step-by-step long grid search to optimize parameters in a random forest algorithm model, selects an optimal parameter value, improves the accuracy of model identification, overcomes the defect of selecting parameters according to experience in the past, and obtains an optimal random forest algorithm identification model through model evaluation index result analysis; carrying out geological structure identification by using the obtained algorithm model; the method is suitable for identifying the collapse column and fault structures in three-dimensional seismic exploration.
Preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration, determining attributes suitable for a subsequent geological structure identification model, and inputting the preprocessed seismic attribute data set as a random forest classification model; secondly, selecting a random forest parameter pair, carrying out parameter optimization on an algorithm model by utilizing improved step-by-step long grid search, and selecting two parameter composition parameter pairs of the number (n_identifiers) of random forest classifiers and the maximum feature number (max_features) of a single decision tree to carry out step-by-step long grid search optimization; and firstly, carrying out large step search, and then, carrying out small step search to determine the optimal parameter pair.
Training an improved random forest classification model by utilizing the preprocessed seismic attribute data set, and comparing the model evaluation index result analysis with a classical random forest algorithm model to obtain an optimal random forest algorithm construction identification model; performing a comparison experiment by using a logistic regression, decision tree and GBDT machine learning algorithm model, and verifying the identification effect of the model through Accuracy (Accuracy), precision and f1 score evaluation indexes; and finally, respectively using the obtained structure recognition model for geological structure recognition of the experimental mining area and the verification mining area, and obtaining a structure recognition result distribution diagram through visual processing. The method comprises the following specific steps:
1. a geologic structure identification technique based on an improved random forest algorithm, the technique comprising:
s1: preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration, wherein the specific process is as follows:
s11: sample marking is carried out on a data set containing various seismic attributes obtained by three-dimensional seismic exploration according to the disclosed different geological structure types, a collapse column is marked as 1, a fault is marked as 2, a non-structural mark is marked as 0, and a training set and a testing set are proportionally divided;
s12: performing feature correlation analysis and feature importance analysis on the marked seismic attribute data to determine attributes suitable for subsequent geological structure identification;
s2: using the seismic attribute data set obtained in the step S1 as input to construct a geological structure identification model for improving a random forest algorithm, wherein the geological structure identification model comprises the following steps:
s21: selecting a random forest parameter pair, optimizing the random forest parameter by utilizing improved step-by-step long grid search, and determining an optimal parameter pair;
s22: performing a comparison experiment on the modified random forest algorithm model based on the S21 and the classical random forest algorithm model on different test sets, and verifying the superiority of the modified algorithm;
s3: the improved random forest algorithm model is used for identifying geological structures, and comprises the following specific steps:
s31: comparing the improved random forest algorithm model with the model identification results of logistic regression, decision trees, GBDT and the like, and verifying the model identification effect;
s32: and respectively carrying out geological structure identification on the experimental mining area and the verification mining area by utilizing the improved random forest algorithm model, and obtaining a structural distribution diagram through visual processing, wherein the faults are displayed as lines, and the collapse columns are displayed as faces.
2. The geologic structure identification technology according to claim 1, wherein in S12, the correlation coefficient is between-1 and 1, the larger the value is, the stronger the correlation between two features is, and the redundancy exists in the features, otherwise, the smaller the correlation is proved, the weaker the correlation between two features is;
the larger the value of the characteristic importance is between 0 and 1, the more important the attribute is for classifying the sample marks, and the attribute with smaller attribute relevance and larger characteristic importance for the sample is selected by sequencing the coefficients of the two analysis results of S12, namely, the attributes with the best classifying effect for the sample are taken as attribute data for constructing the recognition model.
3. The geologic structure identification technology according to claim 1, wherein when the step-by-step long grid search is performed in S21, two parameter composition parameter pairs, namely the number of random forest classifiers (n_estimators) and the maximum feature number (max_features) of a single decision tree, are selected respectively for step-by-step long grid search optimization; and firstly, carrying out large step search, and then, carrying out small step search to determine the optimal parameter pair.
4. The geological structure identification technology according to claim 1, wherein the structure in the experimental mining area is mainly faults, the structure in the experimental mining area is mainly collapse columns in the experimental mining area is verified in the step S32, and the accuracy of the model on the collapse columns and fault structure identification is verified by combining two mining area identification results.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow of geologic structure identification technique based on improved random forest algorithm
FIG. 2 is a schematic flow chart of a random forest algorithm;
FIG. 3 is a schematic diagram of an improved grid search algorithm of the present invention;
FIG. 4 is a schematic illustration of the feature importance of an attribute to a sample tag;
FIG. 5 is a schematic diagram of a large step search model score;
FIG. 6 is a diagram of a small step search model score;
FIG. 7 is a CAD graph of an actual exposure construction of an experimental mining area;
FIG. 8 is a graph of experimental mining area structure identification results (x and y coordinate values on the abscissa and ordinate, respectively);
FIG. 9 is a CAD graph of an actual exposure construction of a validated mine;
fig. 10 is a diagram showing the result of verifying the mining area structure identification (x, y coordinate values corresponding to points on the abscissa and ordinate axes).
Detailed Description
S1: firstly, preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration; the method comprises the following steps:
s11: the method utilizes the processing results of the three-dimensional seismic exploration in the earlier stage of the ore side, extracts x and y coordinates and related attribute data corresponding to a 3# coal bed of a detection area according to a 5X 5 grid, selects twelve seismic attributes sensitive to geological structures based on the existing research results and expert recommendations, constructs a CAD graph according to the actual disclosure provided by the ore side, classifies and marks attribute data in the area according to the structure type, marks a fault area as 2, marks a collapse column area as 1 and marks a non-structure area as 0. Through attribute sensitivity test, a 1397 data set containing x and y coordinates and a marker is finally obtained, wherein the data set contains twelve seismic attributes: variance volume slice, coherence volume slice, frequency division, root mean square amplitude, average energy, dip, curvature, instantaneous phase, instantaneous amplitude, instantaneous frequency, minimum amplitude, and maximum amplitude.
S12: and comprehensively analyzing the characteristics by combining the characteristic correlation analysis and the characteristic importance analysis by utilizing the influence of the characteristic correlation analysis and the characteristic importance analysis on the classification effect of the random forest algorithm. Firstly, carrying out feature correlation analysis on twelve attributes, wherein the larger the correlation coefficient is, the stronger the correlation between two features is, the redundancy exists in the features, and the smaller the correlation is, the weaker the correlation between the two features is; the characteristic correlation analysis in the data set is shown in table 1, and it can be seen that the correlation between the maximum amplitude and the average energy, and the correlation between the instantaneous amplitude and the root mean square amplitude are large;
then, in a random forest algorithm, carrying out feature importance analysis on the attributes, and determining the importance of the twelve attributes to classifier construction and algorithm identification, as shown in table 2 and fig. 5; selecting attribute data according to correlation analysis and feature importance analysis among features; as can be seen from table 2 and fig. 4, the influence of different features on the classification effect is different, and the feature importance difference between the four attributes is not large; through further algorithm test comparison experiments, the influence of four features on the classification effect of the data set is found to be large, and after one feature is deleted, the accuracy of algorithm identification is reduced (about 3 percent of the reduction), and the data set has fewer features, so that the original 12 features are selected to be reserved for subsequent algorithm optimization.
S2: and (3) constructing a geological structure identification model for improving a random forest algorithm by using the seismic attribute data set obtained in the step (S1) as input, wherein an improved grid search flow is shown in a figure 3. The method comprises the following steps:
s21: firstly, searching the number of classifiers in a large step length, setting the initial searching range of the number n_identifiers of the random forest classification trees as [50, 1000], setting the step length as 50, setting the range of max_features as [1,12], and setting the step length as 1; the algorithm model is evaluated by using the model score parameters in Python, and the curve of the model score affected by the two parameters in the large-step search process is shown in FIG. 5.
From the graph peak value 0.9605 in FIG. 5, the parameter values ('max_features': 3, 'n_detectors': 50) are output, and it can be seen from FIG. 5 that when the number of base classifiers exceeds a certain value, the score of the algorithm model is basically converged, the number of base classifiers is increased again, the effect is not basically improved, and the code running speed is slowed down.
Next, performing small-range search, preliminarily setting the range of n_optimizers to be [1,100], and the range of max_features to be [1,12], and obtaining an optimal parameter pair ('max_features': 3, 'n_optimizers': 58); since there are many parameter data, 20 sets of data near the optimal parameter point are selected as references in the present invention, as shown in table 3; when the number of classifiers is 58 and the maximum feature number is 3, the model score is 0.9633, and the small-step search model score is schematically shown in fig. 6.
S22: in order to further verify the reliability of the algorithm model, the method utilizes seismic attribute data sets of other mining areas acquired in the field and processed for verification; in verification, in order to save work, directly performing grid search with a large step length (the step length is set to be 50), primarily obtaining the number of better classifiers, performing search with a small range (the step length is set to be 10) with a second step to obtain more accurate classifier number values, and finally performing refinement division with the step length of 1 to obtain final model parameters, wherein the scores of each dataset in a random forest model for improving the grid search are shown in a table 4; it can be seen that the accuracy of the improved random forest algorithm model is improved to different degrees.
S3: geological structure identification is carried out based on the random forest algorithm model with improved S1 and S2, and the specific steps are as follows:
s31: and comparing the improved random forest algorithm with the recognition results of three algorithm models of GBDT (Gradient Boosting Decision Tree, gradient lifting tree), logistic regression and decision tree on the data set, and evaluating the classification effect of the algorithm model by using Accuracy (Accuracy), precision and f1 score evaluation indexes in order to evaluate the classifier more comprehensively when comparing the recognition classification effect.
In training samples, true Positive (TP): the positive example data which is correctly classified by the classifier is indicated; true Negative (True Negative, TN): negative example data correctly classified by the classifier; false Positive (FP): negative example data that is erroneously marked as positive example data; false Negative (FN): positive example data that is erroneously marked as negative example data.
For the whole training sample, the Accuracy (Accuracy) is calculated by:
for the positive example: the accuracy rate calculation formula is:
the calculation formula of the recall rate is as follows:
f1 score is defined as the harmonic mean of accuracy and recall, which is used to comprehensively evaluate model performance harmonic mean:
the identification results of the different algorithm models are shown in table 5, and the improved random forest algorithm model has higher score, higher identification Accuracy and correspondingly improved classification effect from the aspect of algorithm Accuracy, accuracy Precision and f1 score.
S32: carrying out construction identification on the experimental mining area-new coal mine two-strip two-mining area by utilizing the improved random forest algorithm model; according to the exploration result of the new element coal mine provided by the mining party after the actual disclosure of the two strip and two mining areas, the method comprises the following steps: drawing a 5000 scale to obtain an actual exposure construction CAD graph of the region, as shown in FIG. 7; in the structural diagram, the faults are lines (linear structure in fig. 7), the trapping columns are planes (elliptical area in the right lower corner in fig. 7), and as can be seen from fig. 7, the areas are mainly constructed as faults, and the trapping columns are relatively fewer.
Seismic attribute data were taken in the experiment as 7:3, dividing the model into a training set and a testing set, training and classifying and identifying the model to obtain seismic attribute data coordinate points and mark type files, and obtaining a geological structure distribution map through visual processing to obtain a structure identification result distribution as shown in fig. 8, wherein faults are displayed as lines, and collapse columns are displayed as faces.
Comparing with the three-dimensional seismic structure CAD graph (figure 7) obtained by the regional exploration, the method can see that the structure quantity is accurately identified in the identification result, and the coordinate points corresponding to the structures in the generated text invention piece are accurate, so that the geographic position of the regional structure can be effectively identified.
In order to further verify the algorithm model, verifying by using the western seismic attribute data of a verification mining area, namely a new element north mining area, wherein the three-dimensional seismic structure of the mining area is shown in fig. 9, and the verification mining area has more collapse column structures and fewer faults; the obtained structure recognition result is shown in fig. 10, and it can be seen that the improved algorithm model recognition result is basically consistent with the mining area structure type; substantial agreement of coordinates is observed by comparing the data in the generated text inventories. The two strip mining areas have more fault structures and fewer collapse column structures; while the western mining area of the north mining area has more collapse column structures and relatively fewer fault structures. However, through experimental results, the algorithm model can be seen to have more accurate identification effect on faults and collapse columns.
TABLE 1 characterization correlation analysis
TABLE 2 feature importance analysis
Table 3 parameter pairs and scores
Table 4 random forest parameter optimization algorithm verification
Table 5 comparison of different algorithmic model identification results
Aiming at the problems of high requirements on seismic attribute data sets and complex data processing in the existing structure identification, the invention provides a structure identification technology based on an improved random forest algorithm; firstly, preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration; then selecting two parameter composition parameter pairs of the number (n_identifiers) of random forest classifiers and the maximum feature number (max_features) of a single decision tree to perform step-by-step long grid search optimization, and determining an optimal parameter pair; training an improved random forest classification model by utilizing the preprocessed seismic attribute data set, and analyzing the model evaluation index result to obtain an optimal random forest algorithm recognition model; finally, the obtained recognition model is used for geological structure recognition, and a structure recognition result distribution diagram is obtained through visual processing; comparing the algorithm model with several algorithm models of logistic regression, decision tree and GBDT, wherein the identification effect of the improved random forest algorithm model is superior to that of other algorithm models; meanwhile, the improved random forest algorithm model is respectively tested on seismic attribute data sets of different mining areas, and the accuracy and the applicability of classification results of the model in fault and collapse column structure identification are verified.
The method solves the problems of high requirements on the seismic attribute dataset and complex data processing in the geological structure identification, and improves the accuracy of the identification.
Claims (4)
1. A geologic structure identification technique for three-dimensional seismic exploration, the technique comprising:
s1: preprocessing a data set containing various seismic attributes obtained by three-dimensional seismic exploration, wherein the specific process is as follows:
s11: sample marking is carried out on a data set containing various seismic attributes obtained by three-dimensional seismic exploration according to the disclosed different geological structure types, a collapse column is marked as 1, a fault is marked as 2, a non-structural mark is marked as 0, and a training set and a testing set are proportionally divided;
s12: performing feature correlation analysis and feature importance analysis on the marked seismic attribute data to determine attributes suitable for subsequent geological structure identification;
s2: using the seismic attribute data set obtained in the step S1 as input to construct a geological structure identification model for improving a random forest algorithm, wherein the geological structure identification model comprises the following steps:
s21: selecting a random forest parameter pair, optimizing the random forest parameter by utilizing improved step-by-step long grid search, and determining an optimal parameter pair;
s22: performing a comparison experiment on the modified random forest algorithm model based on the S21 and the classical random forest algorithm model on different test sets, and verifying the superiority of the modified algorithm;
s3: the improved random forest algorithm model is used for identifying geological structures, and comprises the following specific steps:
s31: comparing the improved random forest algorithm model with the model identification results of logistic regression, decision trees, GBDT and the like, and verifying the model identification effect;
s32: and respectively carrying out geological structure identification on the experimental mining area and the verification mining area by utilizing the improved random forest algorithm model, and obtaining a structural distribution diagram through visual processing, wherein the faults are displayed as lines, and the collapse columns are displayed as faces.
2. The geologic structure identification technology according to claim 1, wherein in S12, the correlation coefficient is between-1 and 1, the larger the value is, the stronger the correlation between two features is, and the redundancy exists in the features, otherwise, the smaller the correlation is proved, the weaker the correlation between two features is;
the larger the value of the characteristic importance is between 0 and 1, the more important the attribute is for classifying the sample marks, and the attribute with smaller attribute relevance and larger characteristic importance for the sample is selected by sequencing the coefficients of the two analysis results of S12, namely, the attributes with the best classifying effect for the sample are taken as attribute data for constructing the recognition model.
3. The geologic structure identification technology according to claim 1, wherein when the step-by-step long grid search is performed in S21, two parameter composition parameter pairs, namely the number of random forest classifiers (n_estimators) and the maximum feature number (max_features) of a single decision tree, are selected respectively for step-by-step long grid search optimization; and firstly, carrying out large step search, and then, carrying out small step search to determine the optimal parameter pair.
4. The geologic structure identification technology of claim 1, wherein S32 is characterized by a fault-based structure in the experimental mining area and S32 is characterized by a trapping column in the validation mining area, and the accuracy of the model is validated by a combination of two mining area identification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111320576.2A CN116108392A (en) | 2021-11-09 | 2021-11-09 | Geological structure identification technology based on improved random forest algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111320576.2A CN116108392A (en) | 2021-11-09 | 2021-11-09 | Geological structure identification technology based on improved random forest algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116108392A true CN116108392A (en) | 2023-05-12 |
Family
ID=86262484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111320576.2A Pending CN116108392A (en) | 2021-11-09 | 2021-11-09 | Geological structure identification technology based on improved random forest algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116108392A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117113259A (en) * | 2023-10-19 | 2023-11-24 | 华夏天信智能物联(大连)有限公司 | Coal mine state data processing method and system for predicting potential safety hazards |
-
2021
- 2021-11-09 CN CN202111320576.2A patent/CN116108392A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117113259A (en) * | 2023-10-19 | 2023-11-24 | 华夏天信智能物联(大连)有限公司 | Coal mine state data processing method and system for predicting potential safety hazards |
| CN117113259B (en) * | 2023-10-19 | 2023-12-22 | 华夏天信智能物联(大连)有限公司 | Coal mine state data processing method and system for predicting potential safety hazards |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635461B (en) | Method and system for automatically identifying surrounding rock grade by using while-drilling parameters | |
| CN110674841B (en) | Logging curve identification method based on clustering algorithm | |
| US8090538B2 (en) | System and method for interpretation of well data | |
| US7519476B1 (en) | Method of seismic interpretation | |
| CN103487832A (en) | Method for classifying supervised waveforms in three-dimensional seismic signal | |
| CN101266299B (en) | Method for forecasting oil gas utilizing earthquake data object constructional features | |
| CN104280771A (en) | Three-dimensional seismic data waveform semi-supervised clustering method based on EM algorithm | |
| CN111931505A (en) | Cross-language entity alignment method based on subgraph embedding | |
| CN109165672A (en) | A kind of Ensemble classifier method based on incremental learning | |
| CN105893876A (en) | Chip hardware Trojan horse detection method and system | |
| CN109345007A (en) | A kind of Favorable Reservoir development area prediction technique based on XGBoost feature selecting | |
| CN113902861A (en) | Three-dimensional geological modeling method based on machine learning | |
| CN110837115A (en) | Seismic identification method and device for lithology of land-facies mixed rock compact reservoir | |
| CN112084761A (en) | Hydraulic engineering information management method and device | |
| CN104570109A (en) | Method for reservoir petroleum gas prediction | |
| CN116108392A (en) | Geological structure identification technology based on improved random forest algorithm | |
| CN117272841B (en) | Shale gas dessert prediction method based on hybrid neural network | |
| CN115330007A (en) | Reservoir multi-classification prediction method based on deep fusion model | |
| CN111105041B (en) | Machine learning method and device for intelligent data collision | |
| CN116796231A (en) | Method and device for automatically dividing lithology and comparing lithology based on logging curve | |
| CN103048688A (en) | Seismic attribute optimal selection method based on three-step method | |
| CN112835620B (en) | Semantic similar code online detection method based on deep learning | |
| CN113946790A (en) | Method, system, equipment and terminal for predicting height of water flowing fractured zone | |
| Yu et al. | Training image optimization method based on convolutional neural network and its application in discrete fracture network model selection | |
| CN112990320A (en) | Lithology classification method and device, electronic equipment and storage medium |
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 |