CN115393279A - Sandstone CT image permeability prediction method based on deep learning model - Google Patents

Sandstone CT image permeability prediction method based on deep learning model Download PDF

Info

Publication number
CN115393279A
CN115393279A CN202210878240.6A CN202210878240A CN115393279A CN 115393279 A CN115393279 A CN 115393279A CN 202210878240 A CN202210878240 A CN 202210878240A CN 115393279 A CN115393279 A CN 115393279A
Authority
CN
China
Prior art keywords
sandstone
permeability
image
deep learning
learning model
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
Application number
CN202210878240.6A
Other languages
Chinese (zh)
Inventor
扶海鹰
王帅
丁德馨
贺桂成
张辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of South China
Original Assignee
University of South China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of South China filed Critical University of South China
Priority to CN202210878240.6A priority Critical patent/CN115393279A/en
Publication of CN115393279A publication Critical patent/CN115393279A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation
    • G06T2207/30184—Infrastructure

Abstract

The invention discloses a sandstone CT image permeability prediction method based on a deep learning model, which mainly comprises the following steps: (1) Carrying out X-ray micro-tomography scanning on different types of sandstone to obtain a series of three-dimensional gray images; (2) Carrying out binarization processing on the sandstone three-dimensional gray image, and segmenting the sandstone three-dimensional gray image into gaps and solid spaces to obtain a three-dimensional binarization image; (3) And (3) carrying out blocking processing on the three-dimensional binary image by adopting sliding window sampling to obtain a large number of small-size three-dimensional binary images suitable for the deep learning model. The sandstone CT image permeability prediction method based on the deep learning model disclosed by the invention can be used for rapidly predicting the permeability of a large-size sandstone CT image, and the prediction result has reliability. Meanwhile, the resolution of the image is converted into pixel units in the calculation process of the permeability, so that the permeability of the sandstone can be flexibly predicted under wide spatial resolution without changing or retraining the effect of the model.

Description

Sandstone CT image permeability prediction method based on deep learning model
Technical Field
The invention relates to the technical field of deep learning, in particular to a sandstone CT image permeability prediction method based on a deep learning model.
Background
The permeability of sandstone is an important index concerned in many engineering technical fields, such as sandstone uranium ore in-situ leaching exploitation, sandstone reservoir oil and gas exploitation, carbon dioxide geological disposal and other technical fields. Sandstone permeability acquisition can generally be measured by laboratory tests, but the testing process is time-consuming and labor-consuming, and irreversible damage may be caused to the core, affecting other subsequent tests. Semi-empirical equations, such as the Kozeny-Carman equation, have been summarized on the basis of a large number of experiments to allow rapid prediction of permeability based on a limited number of characteristic parameters, however, the application range of empirical equations is very limited and empirical expressions are difficult to handle with strong heterogeneity.
Permeability reflects the ability of the rock mass to allow the passage of fluids and is obviously a parameter related to the pore structure characteristics of the rock mass. In recent years, thanks to the development of a high-precision three-dimensional imaging technology, a digital core technology can completely and accurately present three-dimensional pore structure characteristics of a rock mass, so that a numerical simulation technology based on a three-dimensional digital core, such as a Lattice Boltzmann Method (LBM), a pore scale finite volume method (PFVS), and the like, can accurately simulate permeability values of the rock mass. However, the calculated volume is larger than the typical unit body volume, the calculation result has application value, and the volume of the typical unit body is often larger, and the calculation amount of the simulation process is often huge. For example, a cell with a voxel of 1000 × 1000 × 1000 may require a supercomputer to calculate permeability by LBM or PFVS. Another computational method, the Pore Network Model (PNM), was introduced to reduce the computational cost of numerical simulation, but there is a simplification of the pore structure by PNM, which affects accuracy.
In recent years, the rapid development of artificial intelligence algorithms makes rapid prediction of permeability based on images possible. In the related art, patent publication No. 202110212709.8 provides a porous medium permeability prediction method based on a three-dimensional convolution neural network. But the technical scheme has the following defects: (1) The set resolution ratios of different CT images in the scanning process may be different, and if the resolution ratio of the CT image of the rock mass to be measured is not consistent with the resolution ratio of the CT image used for training the three-dimensional convolutional neural network model, the prediction result has no reliability; (2) The typical unit body with the representative volume is large in size and limited by video memory, and generally cannot be directly used for training a three-dimensional convolution neural network model, and if the model trained by the small-size unit body is adopted, the model can only be used for predicting the permeability of the unit body with the same size, and the predicted value is not representative.
Disclosure of Invention
The invention discloses a sandstone CT image permeability prediction method based on a deep learning model, which can be used for rapidly predicting permeability of large-size sandstone CT images with different resolutions in order to solve the problems and has reliable results.
In order to achieve the purpose, the invention adopts the following technical scheme:
a sandstone CT image permeability prediction method based on a deep learning model specifically comprises the following steps:
s1: establishing a sandstone three-dimensional binary image permeability data set:
s2: training a deep learning model;
s3: predicting the permeability of the large-size sandstone CT image;
when the permeability data set is established, actual CT images of different types of sandstone are adopted, so that the data set samples have enough diversity, the number of the samples is sufficient, and the model has enough generalization capability.
In a preferred embodiment, the step S1 specifically includes the following steps:
s11: carrying out X-ray micro-tomography scanning on different types of sandstone to obtain a series of high-resolution three-dimensional gray images, and cutting the images into regular large-size cubes;
s12: carrying out binarization processing on the sandstone three-dimensional gray image, and segmenting the sandstone three-dimensional gray image into gaps and solid spaces, wherein 0 pixel point represents the gap, and 1 pixel point represents a solid framework, so as to obtain a three-dimensional binarization image;
s13: the method comprises the steps of (1) carrying out blocking processing on a large-size sandstone three-dimensional binary image by adopting a sliding window sampling method, and segmenting the large-size sandstone three-dimensional binary image into small-size (such as a voxel size of 128 x 128);
s14: calculating the permeability of the small-size sandstone image by using an LBM (local binary matrix) method to obtain a permeability data set of the small-size sandstone image;
the step S2 specifically comprises the following steps:
s21: dividing the small-size three-dimensional binary sandstone image obtained in the step S1 into a training set, a verification set and a test set according to a certain proportion (for example, 80% of the small-size sandstone image is used as the training set, 10% is used as the verification set and 10% is used as the test set);
s22: training the three-dimensional deep learning model by using a permeability data set to obtain a weight parameter;
the step S3 specifically comprises the following steps:
s31: the sandstone to be detected is subjected to the same small-step processing of S11, S12 and S13 in the step S1, and the scanning resolution can be different;
s32: predicting the permeability of the small-size three-dimensional binary image by using the three-dimensional deep learning model trained in the step S2;
s33: and calculating the average value of the permeability of the small-size image to be used as the permeability of the large-size sandstone to be detected.
In a preferred embodiment, when using LBM to calculate the permeability of small-size images, the following method should be used to convert the permeability unit into Pixel square Pixel 2 The generalization capability of the subsequent model under the wide spatial resolution is improved:
Figure BDA0003763046720000041
where k is the permeability in Pixel 2 ；
Figure BDA0003763046720000042
Is the average velocity vector in line with the direction of pressure drop;
Figure BDA0003763046720000043
the pressure gradient with similar velocity vector and mu is the viscosity of the fluid and the calculation method is
Figure BDA0003763046720000044
Wherein, omega is the relaxation frequency of LBM, and is set to be 1.0 in order to ensure the density convergence of the LBM and reduce the error;
the permeability units can be converted to millidarcy mD by providing useful permeability values for subsequent permeability predictions as follows:
Figure BDA0003763046720000045
k s in terms of permeability, millidarcy (mD);
Figure BDA0003763046720000046
the resolution (μm/pixel) of the CT image.
In a preferred scheme, the deep learning model adopts a residual error network model ResNet, and mainly comprises a single convolution layer, four groups of residual error blocks and a full connection layer; wherein, four groups of residual blocks can be ResNet _18, resNet _34 and ResNet _50 models according to different numbers and types of each group of residual blocks, when the target value distribution is unbalanced, log10 (k) transformation is adopted to enable the target value distribution to approximate normal distribution, and the fitting degree of the models is improved.
In a preferred scheme, a deep learning model is trained based on a training set and a verification set, model hyper-parameters, a training period, a learning rate, batch size and an L2 regularization coefficient are continuously adjusted until the model has good prediction capability, a test set is adopted to test the generalization capability of the model, and when sliding window sampling is carried out, the sliding step length is smaller than the side length of a sampling window, namely, any two adjacent sub-small samples are ensured to have repeated parts, so that structural information loss in large-size image samples is avoided, and when sliding window sampling is carried out, the sampling quantity is ensured to be enough, so that a final result is representative.
Therefore, the sandstone CT image permeability prediction method based on the deep learning model specifically comprises the following steps: s1: establishing a sandstone three-dimensional binary image permeability data set: s2: training a deep learning model; s3: predicting the permeability of the large-size sandstone CT image; when the permeability data set is established, actual CT images of different types of sandstone are adopted, so that the data set samples have enough diversity, the number of the samples is sufficient, and the model has enough generalization capability. The sandstone CT image permeability prediction method based on the deep learning model provided by the invention can be used for rapidly predicting the permeability of a large-size sandstone CT image, and the prediction result has reliability. Meanwhile, the resolution of the image is converted into pixel units in the calculation process of the permeability, so that the permeability of the sandstone can be flexibly predicted under wide spatial resolution without changing or retraining the technical effect of the model.
Drawings
Fig. 1 is a schematic overall structure diagram of a sandstone CT image permeability prediction method based on a deep learning model provided by the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Referring to fig. 1, a sandstone CT image permeability prediction method based on a deep learning model specifically includes the following steps:
s1: establishing a sandstone three-dimensional binary image permeability data set:
s2: training a deep learning model;
s3: predicting the permeability of the large-size sandstone CT image;
when the permeability data set is established, actual CT images of different types of sandstone are adopted, so that the data set samples have enough diversity, the number of the samples is sufficient, and the model has enough generalization capability.
In a preferred embodiment, the step S1 specifically includes the following steps:
s11: carrying out X-ray micro-tomography scanning on different types of sandstone to obtain a series of high-resolution three-dimensional gray images, and cutting the images into regular large-size cubes;
s12: carrying out binarization processing on the sandstone three-dimensional gray image, and segmenting the sandstone three-dimensional gray image into gaps and solid spaces, wherein 0 pixel point represents the gap, and 1 pixel point represents a solid framework, so as to obtain a three-dimensional binarization image;
s13: the method comprises the steps of (1) carrying out blocking processing on a large-size sandstone three-dimensional binary image by adopting a sliding window sampling method, and segmenting the large-size sandstone three-dimensional binary image into small-size (such as a voxel size of 128 x 128);
s14: and calculating the permeability of the small-size sandstone image by using an LBM (local binary matrix) method to obtain a permeability data set of the small-size sandstone image.
In a preferred embodiment, the step S2 specifically includes the following steps:
s21: dividing the small-size three-dimensional binary sandstone image obtained in the step S1 into a training set, a verification set and a test set according to a certain proportion (for example, 80% of the small-size sandstone image is used as the training set, 10% is used as the verification set and 10% is used as the test set);
s22: and training the three-dimensional deep learning model by using the permeability data set to obtain a weight parameter.
In a preferred embodiment, the step S3 specifically includes the following steps:
s31: performing small same processing steps S11, S12 and S13 in the step S1 on the sandstone to be detected, wherein the scanning resolutions can be different;
s32: predicting the permeability of the small-size three-dimensional binary image by using the three-dimensional deep learning model trained in the step S2;
s33: and calculating the average value of the permeability of the small-size image to be used as the permeability of the large-size sandstone to be detected.
In a preferred embodiment, when using LBM to calculate the permeability of small-size images, the following method should be used to convert the permeability unit into Pixel square Pixel 2 Is improvedGeneralization ability of subsequent models at wide spatial resolution:
Figure BDA0003763046720000071
where k is the permeability in Pixel 2 ；
Figure BDA0003763046720000072
Is the average velocity vector that is coincident with the pressure drop direction;
Figure BDA0003763046720000073
the pressure gradient with similar velocity vector and mu is the viscosity of the fluid and the calculation method is
Figure BDA0003763046720000081
Where ω is the relaxation frequency of the LBM, and is set to 1.0 to ensure LBM density convergence and reduce errors.
In a preferred embodiment, the permeability units can be converted to mmd by providing a useful permeability value for subsequent permeability prediction as follows:
Figure BDA0003763046720000082
k s as permeability, in millidarcy (mD);
Figure BDA0003763046720000083
the resolution (μm/pixel) of the CT image.
In a preferred embodiment, the deep learning model adopts a residual error network model ResNet, and mainly comprises a single convolution layer, four groups of residual error blocks and a full connection layer; wherein, four groups of residual blocks can be ResNet _18, resNet _34 and ResNet _50 models according to the different number and types of each group of residual blocks, when the target value distribution is not balanced, log10 (k) transformation is adopted to enable the target value distribution to approximate normal distribution, and the fitting degree of the models is improved.
In a preferred embodiment, the deep learning model is trained based on a training set and a validation set, model hyper-parameters, training period, learning rate, batch size and L2 regularization coefficient are continuously adjusted until the model has good prediction capability, and a test set is adopted to test the generalization capability of the model.
In a preferred embodiment, when sliding window sampling is performed, the sliding step length should be smaller than the side length of the sampling window, i.e. it is ensured that any two adjacent sub-small samples have repeated portions, thereby avoiding the loss of structural information in the large-size image sample, and when sliding window sampling is performed, it is ensured that the sampling number is sufficient, so that the final result is representative.
Application examples. In order to prove the creativity and the technical value of the technical scheme of the invention, the part is the application example of the technical scheme of the claims on specific products or related technologies.
(1) The permeability data set is established based on the sandstone in the 10 digital core portal website opening source database by adopting the method. The 10 kinds of sandstone are respectively: bandera Gray, parker, kirby, berea Sister Gray, berea Upper Gray, berea, castegate, buffberea, leopard, bentheim, with an image resolution of 2.25 μm/pixel, are three-dimensional binary images that have been subjected to binarization processing. The large size image volume pixels are 1000 × 1000, sampling is performed using a sliding window of 128 × 128, the sliding step is 96 pixels, and each large size image can be divided into 1000 small size images of 128 × 128 voxels, totaling 10000 subsamples.
(2) The permeability calculation is carried out on 10000 subsamples by adopting the method, the calculation result is used as a real permeability value, a permeability data set is established, and a training set, a verification set and a test set are divided, wherein 8800 are training sets, 800 are verification sets, and the rest 400 are test sets.
(3) Because the target value distribution is not balanced, the label is subjected to log10 (k + 0.1) transformation, so that the target value distribution is approximate to normal distribution, and the fitting degree of the model is improved. Training a ResNet _34 model, and setting the hyper-parameters as follows: the activation function is ReLU, adam is selected by the optimizer, MSE is selected by the loss function, a decision coefficient R2 is used as a measurement index, and an L2 regular penalty coefficient is 1 x 10-4. Setting a dynamic adjustment learning rate, wherein the initial learning rate is 0.001, when the verification loss is not reduced in three iterations (epoch), the learning rate is attenuated by 0.4, the total iteration (epoch) times are 60, and the weight parameter when the verification loss is the lowest is saved as the weight parameter of the final model. The model generalization ability is preferably tested using a test set.
(4) The prediction method was examined by selecting a sandstone (Bandera Brown) having a permeability test value of 63.0mD. First, a large-size three-dimensional binary image of 1000 × 1000 voxels is sampled with a sliding window scale of 128 × 128 and a sliding step size of 96, and the large-size image is segmented into 1000 images of 128 × 128. And predicting the permeability of 1000 samples by adopting the trained model weight parameters, and finally taking an average value as the permeability of the large-size Bandra Brown sandstone.
(5) The predicted value is 69.9mD, which has reliable accuracy compared to the experimental value (63.0 mD), and the prediction process takes only about 20 seconds. And the model training data set does not contain Bandera Brown sandstone, which shows that the model has good generalization capability
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.

Claims (9)

1. A sandstone CT image permeability prediction method based on a deep learning model is characterized by comprising the following steps:
s1: establishing a sandstone three-dimensional binary image permeability data set:
s2: training a deep learning model;
s3: predicting the permeability of the large-size sandstone CT image;
when the permeability data set is established, actual CT images of different types of sandstone are adopted, so that the data set samples have enough diversity, the number of the samples is sufficient, and the model has enough generalization capability.
2. The sandstone CT image permeability prediction method based on the deep learning model according to claim 1, wherein the S1 step specifically comprises the following steps:
s11: carrying out X-ray micro-tomography scanning on different types of sandstone to obtain a series of high-resolution three-dimensional gray images, and cutting the images into regular large-size cubes;
s12: carrying out binarization processing on the sandstone three-dimensional gray image, and segmenting the sandstone three-dimensional gray image into gaps and solid spaces, wherein 0 pixel point represents the gap, and 1 pixel point represents a solid framework, so as to obtain a three-dimensional binarization image;
s13: the method comprises the steps of adopting a sliding window sampling method to carry out blocking processing on a large-size sandstone three-dimensional binary image, and dividing the large-size sandstone three-dimensional binary image into small-size three-dimensional binary sandstone images;
s14: and calculating the permeability of the small-size sandstone image by using an LBM (local binary matrix) method to obtain a permeability data set of the small-size sandstone image.
3. The sandstone CT image permeability prediction method based on the deep learning model according to claim 2, wherein the S2 step specifically comprises the following steps:
s21: dividing the small-size three-dimensional binary sandstone image obtained in the step S1 into a training set, a verification set and a test set according to a certain proportion;
s22: and training the three-dimensional deep learning model by using the permeability data set to obtain a weight parameter.
4. The sandstone CT image permeability prediction method based on the deep learning model according to claim 3, wherein the S3 step specifically comprises the following steps:
s31: the sandstone to be detected is subjected to the same small-step processing of S11, S12 and S13 in the step S1, and the scanning resolution can be different;
s32: predicting the permeability of the small-size three-dimensional binary image by using the three-dimensional deep learning model trained in the step S2;
s33: and calculating the average value of the permeability of the small-size image to be used as the permeability of the large-size sandstone to be detected.
5. A permeability data set according to claim 1, further characterized in that when using LBM to calculate the permeability of small-sized images, the permeability units are converted to Pixel squared pixels by the following method 2 The generalization capability of the subsequent model under the wide spatial resolution is improved:
Figure FDA0003763046710000021
where k is the permeability in Pixel 2 ；
Figure FDA0003763046710000022
Is the average velocity vector in line with the direction of pressure drop;
Figure FDA0003763046710000023
the pressure gradient with similar velocity vector and mu is the viscosity of the fluid and the calculation method is
Figure FDA0003763046710000024
Where ω is the relaxation frequency of the LBM, and is set to 1.0 to ensure LBM density convergence and reduce errors.
6. The sandstone CT image permeability prediction method based on the deep learning model as claimed in claim 5, wherein the permeability unit can be converted into mD in millidarcy in the following way, so as to provide a usable permeability value for the subsequent permeability prediction:
Figure FDA0003763046710000031
k s as permeability, in millidarcy (mD);
Figure FDA0003763046710000032
the resolution (μm/pixel) of the CT image.
7. The sandstone CT image permeability prediction method based on the deep learning model is characterized in that the deep learning model adopts a residual error network model ResNet and mainly comprises a single convolution layer, four groups of residual error blocks and a full connection layer; wherein, four groups of residual blocks can be ResNet _18, resNet _34 and ResNet _50 models according to the different number and types of each group of residual blocks, and when the target value distribution is not balanced, log10 (k) transformation is adopted to enable the target value distribution to approximate normal distribution.
8. The sandstone CT image permeability prediction method based on the deep learning model as claimed in claim 1, wherein the deep learning model is trained based on a training set and a validation set, model hyper-parameters, training period, learning rate, batch size and L2 regularization coefficient are continuously adjusted until the model has good prediction capability, and a test set is adopted to test the generalization capability of the model.
9. The sandstone CT image permeability prediction method based on the deep learning model as claimed in claim 1, wherein, when the sliding window sampling is performed, the sliding step length is smaller than the side length of the sampling window, that is, any two adjacent sub-small samples are ensured to have repeated parts, and when the sliding window sampling is performed, the sampling quantity is ensured to be sufficient, so that the final result is representative.
CN202210878240.6A 2022-07-25 2022-07-25 Sandstone CT image permeability prediction method based on deep learning model Pending CN115393279A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210878240.6A CN115393279A (en) 2022-07-25 2022-07-25 Sandstone CT image permeability prediction method based on deep learning model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210878240.6A CN115393279A (en) 2022-07-25 2022-07-25 Sandstone CT image permeability prediction method based on deep learning model

Publications (1)

Publication Number Publication Date
CN115393279A true CN115393279A (en) 2022-11-25

Family

ID=84117688

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210878240.6A Pending CN115393279A (en) 2022-07-25 2022-07-25 Sandstone CT image permeability prediction method based on deep learning model

Country Status (1)

Country Link
CN (1) CN115393279A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117152373A (en) * 2023-11-01 2023-12-01 中国石油大学(华东) Core-level pore network model construction method considering cracks

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117152373A (en) * 2023-11-01 2023-12-01 中国石油大学(华东) Core-level pore network model construction method considering cracks
CN117152373B (en) * 2023-11-01 2024-02-02 中国石油大学(华东) Core-level pore network model construction method considering cracks

Similar Documents

Publication Publication Date Title
CN110533631B (en) SAR image change detection method based on pyramid pooling twin network
Bai et al. Hybrid geological modeling: Combining machine learning and multiple-point statistics
CN112819207B (en) Geological disaster space prediction method, system and storage medium based on similarity measurement
CN109214470B (en) Image visibility detection method based on coding network fine adjustment
CN109191423B (en) A kind of porous media Permeability Prediction method based on machine image intelligence learning
CN111325403B (en) Method for predicting residual life of electromechanical equipment of highway tunnel
CN114154427B (en) Volume fracturing fracture expansion prediction method and system based on deep learning
CN112396130A (en) Intelligent identification method and system for rock stratum in static sounding test, computer equipment and medium
CN111950140B (en) Dam seepage behavior analysis method considering various uncertainties
CN115393279A (en) Sandstone CT image permeability prediction method based on deep learning model
CN113591700A (en) Prediction method for potential landslide and river blockage
CN113792372A (en) Ground continuous wall deformation dynamic prediction method based on CV-LSTM combined model
CN114998246A (en) Method for calculating absolute permeability of porous medium based on image layering processing technology
Liu et al. Image feature recognition and gas permeability prediction of Gaomiaozi bentonite based on digital images and machine learning.
Sun et al. EPCI: A new tool for predicting absolute permeability from computed tomography images
Liu et al. Deep learning in frequency domain for inverse identification of nonhomogeneous material properties
Li et al. Reconstructing the 3D digital core with a fully convolutional neural network
CN113112447A (en) Tunnel surrounding rock grade intelligent determination method based on VGG convolutional neural network
Tang et al. History matching complex 3D systems using deep-learning-based surrogate flow modeling and CNN-PCA geological parameterization
Juda et al. Comparison of three recent discrete stochastic inversion methods and influence of the prior choice
CN112307536A (en) Dam seepage parameter inversion method
CN112884739B (en) Deep learning network-based method for rapidly detecting filling compactness of rock-fill body
Sattarin et al. GeoPixAI: from pixels to intelligent, unbiased and automated fast track subsurface characterization
Li et al. Deep learning electromagnetic CO 2 sequestration monitoring using the NANO-IP effect with convolutional neural network
WO2021160980A1 (en) An improved micro-ct image acquisition method

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