CN112991277A - Porous medium permeability prediction method based on three-dimensional convolutional neural network - Google Patents
Porous medium permeability prediction method based on three-dimensional convolutional neural network Download PDFInfo
- Publication number
- CN112991277A CN112991277A CN202110212709.8A CN202110212709A CN112991277A CN 112991277 A CN112991277 A CN 112991277A CN 202110212709 A CN202110212709 A CN 202110212709A CN 112991277 A CN112991277 A CN 112991277A
- Authority
- CN
- China
- Prior art keywords
- neural network
- dimensional
- convolutional neural
- porous medium
- permeability
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 103
- 238000000034 method Methods 0.000 title claims abstract description 79
- 230000035699 permeability Effects 0.000 title claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 50
- 238000013528 artificial neural network Methods 0.000 claims abstract description 47
- 238000012545 processing Methods 0.000 claims abstract description 17
- 230000006870 function Effects 0.000 claims description 38
- 238000012360 testing method Methods 0.000 claims description 26
- 238000012795 verification Methods 0.000 claims description 26
- 238000010586 diagram Methods 0.000 claims description 21
- 238000011176 pooling Methods 0.000 claims description 19
- 230000004913 activation Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 10
- 238000013138 pruning Methods 0.000 claims description 10
- 238000013135 deep learning Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 description 8
- 238000010998 test method Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 239000011148 porous material Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000020411 cell activation Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- QSHDDOUJBYECFT-UHFFFAOYSA-N mercury Chemical compound [Hg] QSHDDOUJBYECFT-UHFFFAOYSA-N 0.000 description 1
- 229910052753 mercury Inorganic materials 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T2207/10056—Microscopic image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a porous medium permeability prediction method based on a three-dimensional convolutional neural network. The method comprises the steps of carrying out classification training on a three-dimensional image sample based on a porous medium to obtain a three-dimensional convolution neural network, and processing a three-dimensional image of the porous medium to be detected by using the three-dimensional convolution neural network to obtain the permeability of the porous medium to be detected. The scheme provided by the application can simply and quickly predict the permeability of the porous medium, and the obtained prediction result has enough reliability.
Description
Technical Field
The application relates to the technical field of artificial neural networks, in particular to a porous medium permeability prediction method based on a three-dimensional convolutional neural network.
Background
Permeability is an important property of porous media that measures the ability of the porous media to impede fluid flow. Permeability is a key technical indicator in many engineering fields, such as coal and shale gas mining and CO2Deep geological storage and other fields. At present, macroscopic experiment testing methods for permeability of porous medium materials mainly comprise a mercury pressure method, a gas steady state method and a gas transient state method, and also comprise a finite volume method, a lattice boltzmann method and the like based on numerical analysis and calculation. Although the test results of the mature macroscopic test methods are very accurate, the test process is complicated, and non-professional operators are not easy to operate; and the test period of each method varies from several days to several months according to the porous medium material. Whereas permeability calculations based on numerical analysis are complex and time consuming. Therefore, the method for predicting the permeability of the porous medium, which is simple to operate, short in period and low in cost, is significant.
In the related art, patent with publication number CN109191423B provides a porous medium permeability prediction method based on machine image intelligent learning, which calculates five characteristic parameters including a gray mean, a gray variance, image energy, an image entropy and a fractal dimension of a two-dimensional image based on a two-dimensional image of a porous medium surface scanned by an SEM electron microscope, and obtains a network model capable of predicting porous medium permeability by learning a relationship between the characteristic parameters and corresponding real permeability. And the porous medium permeability can be predicted based on the network model.
The technical scheme has the following defects:
1. the two-dimensional image has contingency, and cannot comprehensively represent the porous medium, particularly the internal structure of the porous medium, so that the reliability of a neural network model for deep learning based on the two-dimensional image is not high;
2. the method needs to fit the nonlinear relation between the five image characteristic parameters and the real permeability, and the data processing process is complicated.
Disclosure of Invention
In order to overcome the problems in the related art, the application provides a porous medium permeability prediction method based on a three-dimensional convolutional neural network, and the method can simply and quickly predict the porous medium permeability and obtain a reliable prediction result.
The application provides a porous medium permeability prediction method based on a three-dimensional convolutional neural network, which comprises the following steps:
acquiring a three-dimensional image to be detected of the porous medium to be detected;
processing the three-dimensional image to be detected based on a three-dimensional convolutional neural network to obtain the permeability of the porous medium to be detected;
the three-dimensional convolutional neural network is used for carrying out classification training based on three-dimensional image samples to complete the deep learning of the nonlinear mapping from the three-dimensional images to the permeability; the three-dimensional image sample is a collection of three-dimensional images of a porous media sample, the three-dimensional image sample comprising: training and verifying sets;
and the three-dimensional convolution neural network expands the three-dimensional characteristic diagram after convolution and pooling into a vector form, classifies the three-dimensional characteristic diagram based on the vector form of the three-dimensional characteristic diagram, and outputs the permeability of the porous medium.
In one embodiment, the three-dimensional convolutional neural network comprises: convolutional networks and artificial neural networks;
the convolutional network includes: the system comprises an input layer, five convolutional layers and three pooling layers, wherein the number of convolutional cores of the convolutional layers is 16, 32, 64, 128 and 256 in sequence, the sizes of the convolutional cores are all 3 multiplied by 3, the sizes of the pooling layers are all 2 multiplied by 2, and the step lengths are all 2; the convolution layer is used for extracting features and transmitting a feature map to the pooling layer; the pooling layer is used for compressing the characteristics of the obtained characteristic diagram, wherein the third pooling layer expands the obtained characteristic diagram into a vector form and transmits the vector form to the artificial neural network;
the artificial neural network includes: the device comprises an input layer, two hidden layers and an output layer, wherein the number of neurons of the two hidden layers is 1024; and the hidden layer is used for calculating the inner product of the input vector and the weight vector, obtaining a classification result through the processing of an activation function and transmitting the result to the output layer.
In one embodiment, the activation function of the artificial neural network is a modified linear cell activation function;
the modified linear unit activation function is a ReLU function, and its expression is f (x) max {0, x }.
In one embodiment, the weights of the hidden layers of the artificial neural network are processed using a regularization technique.
In one embodiment, the regularization technique processes, including:
adding an L2 regular penalty term on the basis of the loss function of the artificial neural network to obtain the loss function of the three-dimensional convolutional neural network;
the loss function of the three-dimensional convolutional neural network may be expressed by the following expression;
wherein,
as a function of the loss of the artificial neural network,
is a penalty term of L2 regular, theta is a hyper-parameter of the three-dimensional convolutional neural network, M is the number of training data,
is the square of the difference between the predicted value and the true value, λ is the regularization parameter, and takes a value of 0.01, l is the weight number of the hidden layer, ω isiIs the parameter to be learned of the artificial neural network layer.
In one embodiment, before the processing the three-dimensional image to be detected based on the three-dimensional convolutional neural network, the processing includes:
training based on a training set to obtain the three-dimensional convolutional neural network;
acquiring hyper-parameters of the three-dimensional convolutional neural network;
calling a verification set to test the three-dimensional convolutional neural network, judging whether the three-dimensional convolutional neural network meets a target condition, and if so, outputting the three-dimensional convolutional neural network; if not, resetting the hyper-parameters of the three-dimensional convolutional neural network until the three-dimensional convolutional neural network meets the target condition;
the target conditions include: the identification accuracy rate of the three-dimensional convolutional neural network is not lower than an identification threshold value;
the recognition threshold value ranges from 90% to 100%.
In one embodiment, said outputting said three-dimensional convolutional neural network comprises:
and optimizing the three-dimensional convolution neural network in a pruning optimization mode to reduce the size of the model.
In one embodiment, before the training based on the training set to obtain the three-dimensional convolutional neural network, the method includes:
scanning a porous medium sample by using an X-ray three-dimensional micro CT to obtain a three-dimensional image sample;
and performing data enhancement on the three-dimensional image sample to obtain a training set and a verification set.
In one embodiment, the data enhancement of the three-dimensional image sample is performed to obtain a training set and a validation set, and includes:
carrying out N data enhancement processing in rotation, turning, scaling, translation, scale transformation, contrast transformation, noise disturbance and random color on the three-dimensional image sample, wherein N is an integer with the value range of 1-8;
and splitting the three-dimensional image sample subjected to data enhancement into a training set and a verification set.
In one embodiment, the three-dimensional image acquisition to be performed on the porous medium to be measured includes:
scanning a porous medium to be detected by using an X-ray three-dimensional micro CT;
and removing irrelevant gaps of the scanned three-dimensional image by using three-dimensional visualization software AVIZO to obtain the three-dimensional image to be detected.
The technical scheme provided by the application can comprise the following beneficial effects:
according to the scheme, the three-dimensional image of the porous medium to be tested, which is obtained by scanning, is processed by only utilizing the three-dimensional convolutional neural network, the permeability of the porous medium to be tested can be predicted through the three-dimensional image to be tested on the basis of the deep learning of the nonlinear mapping from the three-dimensional image to the permeability, other characteristic parameters of the image do not need to be calculated, the operation is simple, and the test period is short; because the three-dimensional image can reliably represent the internal structure of the porous medium, and the permeability as the internal attribute of the porous medium is closely related to the internal structure thereof, the three-dimensional convolution neural network obtained by training by using the three-dimensional image of the porous medium as a training sample has enough reliability when predicting the permeability of the porous medium.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The foregoing and other objects, features and advantages of the application will be apparent from the following more particular descriptions of exemplary embodiments of the application, as illustrated in the accompanying drawings wherein like reference numbers generally represent like parts throughout the exemplary embodiments of the application.
FIG. 1 is a schematic flow chart of a porous medium permeability prediction method based on a three-dimensional convolutional neural network according to an embodiment of the present application;
fig. 2 is a schematic flowchart of a method for constructing a three-dimensional convolutional neural network according to an embodiment of the present application;
FIG. 3 is a flow chart of a testing method of a three-dimensional convolutional neural network according to an embodiment of the present disclosure;
fig. 4 is a flowchart illustrating a method for preparing a training set and a validation set according to an embodiment of the present application.
Detailed Description
Preferred embodiments of the present application will be described in more detail below with reference to the accompanying drawings. While the preferred embodiments of the present application are shown in the drawings, it should be understood that the present application may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It should be understood that although the terms "first," "second," "third," etc. may be used herein to describe various information, these information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present application. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
The permeability is a key technical index in many engineering fields, and the test result of the macroscopic test method for the permeability of the porous medium in the industry at present is very accurate, but has the defects of difficult operation, long test period and high cost. Although the operation difficulty is reduced compared with the macroscopic test method, the adopted two-dimensional image cannot comprehensively represent the porous medium and the nonlinear relation between five image characteristic parameters and the real permeability needs to be fitted, so that the defects of low reliability and complex data processing still exist.
Example 1
In view of the above problems, embodiments of the present application provide a porous medium permeability prediction method based on a three-dimensional convolutional neural network, which can simply and quickly predict the porous medium permeability, and an obtained prediction result has sufficient reliability.
The technical solutions of the embodiments of the present application are described in detail below with reference to the accompanying drawings.
Fig. 1 is a schematic flowchart of a porous medium permeability prediction method based on a three-dimensional convolutional neural network according to an embodiment of the present application.
Referring to fig. 1, the porous medium permeability prediction method based on the three-dimensional convolutional neural network is characterized by comprising the following steps:
101. acquiring a three-dimensional image to be detected of the porous medium to be detected;
in the embodiment of the application, the porous medium to be detected is scanned by using an X-ray three-dimensional micro CT to obtain the three-dimensional image to be detected.
Further, in order to make the porous medium permeability predicted based on the three-dimensional image to be measured more accurate, the three-dimensional image to be measured may be processed, for example, the three-dimensional image obtained by scanning may be processed to remove irrelevant gaps by using the three-dimensional visualization software AVIZO.
It is understood that the above description of the three-dimensional image to be measured is only one example of the embodiments of the present application, and is not necessarily taken as a limitation on the present invention.
102. And processing the three-dimensional image to be detected based on a three-dimensional convolution neural network to obtain the permeability of the porous medium to be detected.
In the embodiment of the application, the three-dimensional convolution neural network carries out classification training based on three-dimensional image samples to complete the deep learning of the nonlinear mapping from the three-dimensional image to the permeability; the three-dimensional image sample is a collection of three-dimensional images of a porous media sample, the three-dimensional image sample comprising: training set and validation set.
It should be noted that, the process of acquiring the three-dimensional image sample is as follows: acquiring a porous medium three-dimensional image with high resolution and high quality by X-ray three-dimensional micro CT; and removing irrelevant pores in the image by all three-dimensional image data in three-dimensional visualization software AVIZO, and cutting the obtained three-dimensional image of the porous medium into a proper size to form a sample data set.
It should be noted that the training set and the verification set are obtained by splitting the sample data set according to a certain ratio. The training set is used for training a three-dimensional convolution neural network, so that the three-dimensional convolution neural network can complete the deep learning of the nonlinear mapping from a three-dimensional image to permeability; and the verification set is used for testing the three-dimensional convolutional neural network after the three-dimensional convolutional neural network is built, and selecting the three-dimensional convolutional neural network meeting the requirement of accuracy rate.
It is to be understood that the above description of the three-dimensional image sample is only one example of the embodiments of the present application and is not necessarily intended as a limitation on the present invention.
In this embodiment of the application, a specific process of processing the three-dimensional image to be detected by the three-dimensional convolutional neural network may be represented as: and the three-dimensional convolution neural network expands the three-dimensional characteristic diagram after convolution and pooling into a vector form, classifies the three-dimensional characteristic diagram based on the vector form of the three-dimensional characteristic diagram, and outputs the permeability of the porous medium.
It is to be understood that the above description of a three-dimensional convolutional neural network is merely an example and is not necessarily intended as a limitation on the present invention.
According to the scheme, the three-dimensional image of the porous medium to be tested, which is obtained by scanning, is processed by only utilizing the three-dimensional convolutional neural network, the permeability of the porous medium to be tested can be predicted through the three-dimensional image to be tested on the basis of the deep learning of the nonlinear mapping from the three-dimensional image to the permeability, other characteristic parameters of the image do not need to be calculated, the operation is simple, and the test period is short; because the three-dimensional image can reliably represent the internal structure of the porous medium, and the permeability as the internal attribute of the porous medium is closely related to the internal structure thereof, the three-dimensional convolution neural network obtained by training by using the three-dimensional image of the porous medium as a training sample has enough reliability when predicting the permeability of the porous medium.
Example 2
In practical applications, the embodiment of the present application designs the method for constructing the three-dimensional convolutional neural network described in embodiment 1 above.
Fig. 2 is a flow chart diagram of a method for constructing a three-dimensional convolutional neural network.
With specific reference to fig. 2, the method for constructing the three-dimensional convolutional neural network includes:
201. building a network structure of a three-dimensional convolutional neural network model;
in an embodiment of the present application, the three-dimensional convolutional neural network includes: convolutional networks and artificial neural networks;
the convolutional network includes: the system comprises an input layer, five convolutional layers and three pooling layers, wherein the number of convolutional cores of the convolutional layers is 16, 32, 64, 128 and 256 in sequence, the sizes of the convolutional cores are all 3 multiplied by 3, the sizes of the pooling layers are all 2 multiplied by 2, and the step lengths are all 2; the convolution layer is used for extracting features and transmitting a feature map to the pooling layer; the pooling layer is used for compressing the characteristics of the obtained characteristic diagram, wherein the third pooling layer expands the obtained characteristic diagram into a vector form and transmits the vector form to the artificial neural network;
the artificial neural network includes: the device comprises an input layer, two hidden layers and an output layer, wherein the number of neurons of the two hidden layers is 1024; the number of the neurons of the two hidden layers is 1024; and the hidden layer is used for calculating the inner product of the input vector and the weight vector, obtaining a classification result through the processing of an activation function and transmitting the result to the output layer.
It is to be understood that the above description of convolutional networks and artificial neural networks is only one example of the embodiments of the present application and is not necessarily intended as a limitation on the present invention.
202. Constructing a loss function and an activation function of the three-dimensional convolutional neural network;
in the embodiment of the application, the activation function of the artificial neural network is a modified linear unit activation function; specifically, the modified linear unit activation function is a ReLU function, and its expression is f (x) max {0, x }.
It should be noted that, in practical applications, other activation functions may also be adopted as the activation function of the artificial neural network according to practical situations, for example, a sigmoid function, whose expression is f (x) 1/(1+ e)x)。
It is to be understood that the above description of the activation function is only an example of the embodiments of the present application and is not necessarily intended as a limitation on the present invention.
In the embodiment of the application, the weights of the hidden layer of the artificial neural network are processed by a regularization technology.
The following are exemplary:
adding an L2 regular penalty term on the basis of the loss function of the artificial neural network to obtain the loss function of the three-dimensional convolutional neural network;
the loss function of the three-dimensional convolutional neural network may be expressed by the following expression;
a regular penalty term, theta is a hyper-parameter of the three-dimensional convolution neural network, M is the number of training data,
is the square of the difference between the predicted value and the true value, λ is the regularization parameter, and takes a value of 0.01, l is the weight number of the hidden layer, ω isiIs the parameter to be learned of the artificial neural network layer.
It should be noted that the above-mentioned L2 regularization process is only an example in the embodiment of the present application, and in an actual application process, the regularization technical process may also adopt an L1 regularization process.
It is to be understood that the above description of regularization technique is only one example of an embodiment of the present application and is not intended as a limitation on the present invention.
203. Training a three-dimensional convolutional neural network based on a training set;
in the embodiment of the application, the training set is a data set of a three-dimensional image with porous medium permeability as a label; the permeability of the porous medium is obtained by adopting a numerical model of a pore size finite volume method.
It should be noted that, in the embodiment of the present application, there is no strict limitation on the method for obtaining the permeability of the porous medium in the training set, and any method that can calculate the permeability of the porous medium sample is applicable to this step, for example, the lattice boltzmann method or the inversion method.
It should be noted that, in practical applications, the three-dimensional convolutional neural network may also be trained by using a method of unsupervised learning without acquiring a tag.
It is to be understood that the above description of the training process of the three-dimensional convolutional neural network is only an example in the embodiments of the present application, and is not necessarily taken as a limitation of the present invention.
204. And testing the three-dimensional convolution neural network based on the verification set.
In the embodiment of the application, after the three-dimensional convolutional neural network is trained, the trained three-dimensional convolutional neural network is repeatedly tested by using a verification set, and the three-dimensional convolutional neural network is adjusted according to a result obtained by the test until the test result meets the requirement.
It should be noted that the objects of the above adjustment may be M of the following objects: the number of convolutional layer layers, convolutional kernel size, pooling layer size, and learning rate, where M is a positive integer.
It is to be understood that the above description of the test procedure is only one example of an embodiment of the present application and does not necessarily constitute a limitation of the present invention.
The embodiment of the application provides a method for constructing a three-dimensional convolutional neural network, wherein the three-dimensional convolutional neural network constructed based on the method comprises a convolutional network and an artificial neural network, and the convolutional network can be used for extracting and compressing the characteristics of an input three-dimensional image; the artificial neural network adopts a modified linear unit activation function as an activation function, and punishment is carried out on a loss function by adopting regularization processing, so that overfitting of the model is effectively prevented, and the generalization performance of the model is improved; meanwhile, the method utilizes the verification set to repeatedly test and adjust the three-dimensional convolutional neural network, so that the finally obtained test result of the three-dimensional convolutional neural network is close to the actual result, and the high-precision porous medium permeability prediction can be realized through the three-dimensional convolutional neural network.
Example 3
In practical applications, the present embodiment is designed for step 204 in embodiment 2.
Fig. 3 is a flow chart of a testing method of a three-dimensional convolutional neural network.
Referring to fig. 3, the method for testing a three-dimensional convolutional neural network includes:
301. acquiring hyper-parameters of the three-dimensional convolutional neural network;
in the embodiment of the application, the acquisition of the hyper-parameters is not strictly limited, in practical application, a three-dimensional image sample model can be expanded by adopting data augmentation, an expanded three-dimensional image set is used as the input of a three-dimensional convolution neural network to obtain a prediction result, and the hyper-parameters of the countermeasure network model are determined based on the prediction result; or the user sets the hyper-parameters by himself.
It is understood that the above description of the hyper-parameters of the three-dimensional convolutional neural network is only one example of the embodiments of the present application, and does not necessarily constitute a limitation of the present invention.
302. Calling a verification set to test the three-dimensional convolutional neural network to obtain the three-dimensional convolutional neural network meeting target conditions;
the target conditions include: the identification accuracy rate of the three-dimensional convolutional neural network is not lower than an identification threshold value;
the recognition threshold value ranges from 90% to 100%.
In the embodiment of the present application, since the verification set used for the test is subjected to the preprocessing of removing the irrelevant pore, the data noise of the verification set used for the test is low, and therefore, when the test of the three-dimensional convolutional neural network is performed based on the verification set, the obtained accuracy value is high, and therefore, in the embodiment of the present application, the identification threshold value is 90%.
It should be noted that, in practical application, the data sample for testing may be directly acquired by the scanning device without being processed by the three-dimensional visualization software, and thus, the value of the identification threshold may be adjusted according to the actual situation and the picture quality of the verification set.
It should be understood that the above description of the identification threshold is only an example of the embodiment of the present application, and is not necessarily taken as a limitation on the present invention.
The embodiment of the application provides a test method of a three-dimensional convolutional neural network, which is characterized in that the three-dimensional convolutional neural network is tested by using a verification set, and the optimal value of the hyper-parameter of the three-dimensional convolutional neural network is determined based on the test result, so that the three-dimensional convolutional neural network capable of meeting the requirement of accuracy is selected, and the finally obtained three-dimensional convolutional neural network has excellent accuracy and reliability in predicting the permeability of a porous medium.
Example 4
In practical applications, after step 302 of the above embodiment 3, the embodiment of the present application performs design of an optimization step on the three-dimensional convolutional neural network.
The optimization method comprises the following steps: and optimizing the three-dimensional convolution neural network in a pruning optimization mode to reduce the size of the model.
It should be noted that, in the embodiments of the present application, there is no strict limitation on the pruning method, and different pruning methods, for example, feasible pruning or optimal pruning, may be adopted according to actual requirements.
It should be noted that, in practical applications, besides the pruning optimization mode, the quantization or binarization optimization mode may be used to process the three-dimensional convolutional neural network, so as to reduce the calculation amount and memory occupation of the three-dimensional convolutional neural network and accelerate the inference rate of the model.
It should be understood that the above description of model optimization is only an example of the embodiments of the present application, and should not be construed as limiting the present invention.
The embodiment of the application provides a model optimization method, which processes a three-dimensional convolution neural network in a pruning optimization mode to reduce the memory occupation of the three-dimensional convolution neural network. For the three-dimensional convolutional neural network, pruning removes neurons with a small weight ratio in a model, and because the weight of the removed neurons is small, the influence on the accuracy of the model is small, but the calculated amount of the model can be greatly reduced, so that the prediction efficiency of the three-dimensional convolutional neural network is improved, and the prediction period is shortened.
Example 5
The establishment of the three-dimensional convolutional neural network used in the above-described embodiments 1, 2, and 3 relies on deep learning using a training set and screening a network model with high accuracy using a validation set, and thus, the embodiments of the present application design the preparation of the training set and the validation set.
Fig. 4 is a flow chart diagram of a method for preparing a training set and a validation set.
Referring to fig. 4, the preparation of the training set and validation set includes:
501. scanning a porous medium sample by using an X-ray three-dimensional micro CT to obtain a three-dimensional image sample;
in this embodiment of the application, the specific operation process of step 501 is consistent with that of step 101 in embodiment 1, and is not described herein again.
502. And performing data enhancement on the three-dimensional image sample to obtain a training set and a verification set.
The following are exemplary:
carrying out N data enhancement processing in rotation, turning, scaling, translation, scale transformation, contrast transformation, noise disturbance and random color on the three-dimensional image sample, wherein N is an integer with the value range of 1-8;
and splitting the three-dimensional image sample subjected to data enhancement into a training set and a verification set.
In the embodiment of the application, the three-dimensional image sample after data enhancement is split according to a certain proportion to obtain a training set and a verification set.
It should be noted that, in the embodiment of the present application, the splitting ratio is not strictly limited, and in practical applications, the splitting ratio may be set according to circumstances, for example, in the embodiment of the present application, the three-dimensional image sample may be split into a training set and a verification set according to a ratio of 1:1 or 2: 3.
It is to be understood that the above description of the training set and the validation set is only an example of the embodiments of the present application and is not necessarily intended as a limitation on the present invention.
The embodiment of the application provides a method for preparing a training set and a verification set. The data enhancement processing increases the data volume of the training set and the verification set, so that the training set and the verification set are diversified as much as possible, and the trained model has stronger generalization capability; deep learning is carried out based on diversified training sets, so that the prediction performance and accuracy of the three-dimensional convolutional neural network can be effectively improved; and a three-dimensional convolutional neural network with high reliability can be screened out through a network model obtained by a large number of diversified verification set tests.
The aspects of the present application have been described in detail hereinabove with reference to the accompanying drawings. In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments. Those skilled in the art should also appreciate that the acts referred to in the description are not necessarily required in this application. In addition, it can be understood that the steps in the method of the embodiment of the present application may be sequentially adjusted, combined, and deleted according to actual needs.
Furthermore, the method according to the present application may also be implemented as a computer program or computer program product comprising computer program code instructions for performing some or all of the steps of the above-described method of the present application.
Alternatively, the present application may also be embodied as a non-transitory machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) having stored thereon executable code (or a computer program, or computer instruction code) which, when executed by a processor of an electronic device (or electronic device, server, etc.), causes the processor to perform part or all of the various steps of the above-described method according to the present application.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems and methods according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Having described embodiments of the present application, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (10)
1. A porous medium permeability prediction method based on a three-dimensional convolutional neural network is characterized by comprising the following steps:
acquiring a three-dimensional image to be detected of the porous medium to be detected;
processing the three-dimensional image to be detected based on a three-dimensional convolutional neural network to obtain the permeability of the porous medium to be detected;
the three-dimensional convolutional neural network is used for carrying out classification training based on three-dimensional image samples to complete the deep learning of the nonlinear mapping from the three-dimensional images to the permeability; the three-dimensional image sample is a collection of three-dimensional images of a porous media sample, the three-dimensional image sample comprising: training and verifying sets;
and the three-dimensional convolution neural network expands the three-dimensional characteristic diagram after convolution and pooling into a vector form, classifies the three-dimensional characteristic diagram based on the vector form of the three-dimensional characteristic diagram, and outputs the permeability of the porous medium.
2. The method for predicting the permeability of porous media based on three-dimensional convolutional neural network of claim 1,
the three-dimensional convolutional neural network, comprising: convolutional networks and artificial neural networks;
the convolutional network includes: the system comprises an input layer, five convolutional layers and three pooling layers, wherein the number of convolutional cores of the convolutional layers is 16, 32, 64, 128 and 256 in sequence, the sizes of the convolutional cores are all 3 multiplied by 3, the sizes of the pooling layers are all 2 multiplied by 2, and the step lengths are all 2; the convolution layer is used for extracting features and transmitting a feature map to the pooling layer; the pooling layer is used for compressing the characteristics of the obtained characteristic diagram, wherein the third pooling layer expands the obtained characteristic diagram into a vector form and transmits the vector form to the artificial neural network;
the artificial neural network includes: the device comprises an input layer, two hidden layers and an output layer, wherein the number of neurons of the two hidden layers is 1024; and the hidden layer is used for calculating the inner product of the input vector and the weight vector, obtaining a classification result through the processing of an activation function and transmitting the result to the output layer.
3. The method for predicting the permeability of porous media based on three-dimensional convolutional neural network of claim 2,
the activation function of the artificial neural network is a modified linear unit activation function;
the modified linear unit activation function is a ReLU function, and its expression is f (x) max {0, x }.
4. The method for predicting the permeability of porous media based on three-dimensional convolutional neural network of claim 2,
the weight of the hidden layer of the artificial neural network is processed by adopting a regularization technology.
5. The porous medium permeability prediction method based on the three-dimensional convolutional neural network as claimed in claim 4, wherein the regularization technology process comprises:
adding an L2 regular penalty term on the basis of the loss function of the artificial neural network to obtain the loss function of the three-dimensional convolutional neural network;
the loss function of the three-dimensional convolutional neural network may be expressed by the following expression;
wherein,
as a function of the loss of the artificial neural network,
is a penalty term of L2 regular, theta is a hyper-parameter of the three-dimensional convolutional neural network, M is the number of training data,
is the square of the difference between the predicted value and the true value, λ is the regularization parameter, and takes a value of 0.01, l is the weight number of the hidden layer, ω isiIs the parameter to be learned of the artificial neural network layer.
6. The method for predicting the permeability of the porous medium based on the three-dimensional convolutional neural network as claimed in claim 1, wherein before the three-dimensional convolutional neural network processes the three-dimensional image to be measured, the method comprises the following steps:
training based on a training set to obtain the three-dimensional convolutional neural network;
acquiring hyper-parameters of the three-dimensional convolutional neural network;
calling a verification set to test the three-dimensional convolutional neural network, judging whether the three-dimensional convolutional neural network meets a target condition, and if so, outputting the three-dimensional convolutional neural network; if not, resetting the hyper-parameters of the three-dimensional convolutional neural network until the three-dimensional convolutional neural network meets the target condition;
the target conditions include: the identification accuracy rate of the three-dimensional convolutional neural network is not lower than an identification threshold value;
the recognition threshold value ranges from 90% to 100%.
7. The method for predicting the permeability of the porous medium based on the three-dimensional convolutional neural network as claimed in claim 6, wherein after outputting the three-dimensional convolutional neural network, the method comprises:
and optimizing the three-dimensional convolution neural network in a pruning optimization mode to reduce the size of the model.
8. The method for predicting the permeability of the porous medium based on the three-dimensional convolutional neural network as claimed in claim 6, wherein before the training based on the training set to obtain the three-dimensional convolutional neural network, the method comprises:
scanning a porous medium sample by using an X-ray three-dimensional micro CT to obtain a three-dimensional image sample;
and performing data enhancement on the three-dimensional image sample to obtain a training set and a verification set.
9. The method for predicting the permeability of the porous medium based on the three-dimensional convolutional neural network as claimed in claim 8, wherein the performing data enhancement on the three-dimensional image sample to obtain a training set and a verification set comprises:
carrying out N data enhancement processing in rotation, turning, scaling, translation, scale transformation, contrast transformation, noise disturbance and random color on the three-dimensional image sample, wherein N is an integer with the value range of 1-8;
and splitting the three-dimensional image sample subjected to data enhancement into a training set and a verification set.
10. The method for predicting the permeability of the porous medium based on the three-dimensional convolutional neural network as claimed in claim 1, wherein the acquiring of the three-dimensional image to be detected of the porous medium to be detected comprises:
scanning a porous medium to be detected by using an X-ray three-dimensional micro CT;
and removing irrelevant gaps of the scanned three-dimensional image by using three-dimensional visualization software AVIZO to obtain the three-dimensional image to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110212709.8A CN112991277A (en) | 2021-02-25 | 2021-02-25 | Porous medium permeability prediction method based on three-dimensional convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110212709.8A CN112991277A (en) | 2021-02-25 | 2021-02-25 | Porous medium permeability prediction method based on three-dimensional convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112991277A true CN112991277A (en) | 2021-06-18 |
Family
ID=76350738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110212709.8A Pending CN112991277A (en) | 2021-02-25 | 2021-02-25 | Porous medium permeability prediction method based on three-dimensional convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112991277A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118136181A (en) * | 2024-03-25 | 2024-06-04 | 西北有色金属研究院 | Porous material characteristic parameter prediction method based on deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140236486A1 (en) * | 2013-02-21 | 2014-08-21 | Saudi Arabian Oil Company | Methods, Program Code, Computer Readable Media, and Apparatus For Predicting Matrix Permeability By Optimization and Variance Correction of K-Nearest Neighbors |
| CN109191423A (en) * | 2018-07-18 | 2019-01-11 | 中国矿业大学 | A kind of porous media Permeability Prediction method based on machine image intelligence learning |
| CN110335294A (en) * | 2019-07-11 | 2019-10-15 | 中国矿业大学 | Mine water pump house leakage detection method based on frame difference method Yu 3D convolutional neural networks |
| CN111561313A (en) * | 2020-05-15 | 2020-08-21 | 中国地质大学（北京） | Compact sandstone reservoir parameter prediction method based on physical model and machine learning |
-
2021
- 2021-02-25 CN CN202110212709.8A patent/CN112991277A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140236486A1 (en) * | 2013-02-21 | 2014-08-21 | Saudi Arabian Oil Company | Methods, Program Code, Computer Readable Media, and Apparatus For Predicting Matrix Permeability By Optimization and Variance Correction of K-Nearest Neighbors |
| CN109191423A (en) * | 2018-07-18 | 2019-01-11 | 中国矿业大学 | A kind of porous media Permeability Prediction method based on machine image intelligence learning |
| CN110335294A (en) * | 2019-07-11 | 2019-10-15 | 中国矿业大学 | Mine water pump house leakage detection method based on frame difference method Yu 3D convolutional neural networks |
| CN111561313A (en) * | 2020-05-15 | 2020-08-21 | 中国地质大学（北京） | Compact sandstone reservoir parameter prediction method based on physical model and machine learning |
Non-Patent Citations (3)
| Title |
|---|
| M.B. BIRD,ET AL: ""Numerical modeling of fluid and electrical currents through geometries based on synchrotron X-ray tomographic images of reservoir rocks using Avizo and COMSOL"，M.B. Bird,et al,《Computers&Geosciences》，正文第6-16页", 《COMPUTERS&GEOSCIENCES》 * |
| M.B. BIRD,ET AL: "Numerical modeling of fluid and electrical currents through geometries based on synchrotron X-ray tomographic images of reservoir rocks using Avizo and COMSOL", 《COMPUTERS&GEOSCIENCES》 * |
| 王依诚 等: "基于机器学习的数字岩心孔渗预测方法研究", 《石油科学通报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118136181A (en) * | 2024-03-25 | 2024-06-04 | 西北有色金属研究院 | Porous material characteristic parameter prediction method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110533631B (en) | SAR image change detection method based on pyramid pooling twin network | |
| Kanarachos et al. | Detecting anomalies in time series data via a deep learning algorithm combining wavelets, neural networks and Hilbert transform | |
| Hou et al. | Deep features based on a DCNN model for classifying imbalanced weld flaw types | |
| CN111723732B (en) | Optical remote sensing image change detection method, storage medium and computing equipment | |
| Singh et al. | Textural identification of basaltic rock mass using image processing and neural network | |
| CN110189255A (en) | Method for detecting human face based on hierarchical detection | |
| CN105608698A (en) | Remote image change detection method based on SAE | |
| CN111323228A (en) | Fault diagnosis method based on time series mapping and convolutional neural network | |
| CN116046810B (en) | Nondestructive testing method based on RPC cover plate damage load | |
| CN112598657A (en) | Defect detection method and device, model construction method and computer equipment | |
| CN117011274A (en) | Automatic glass bottle detection system and method thereof | |
| CN117911879B (en) | SAM-fused fine-granularity high-resolution remote sensing image change detection method | |
| CN115295098A (en) | Method and system for predicting permeability characteristics of spherical particle system based on machine learning | |
| CN112991277A (en) | Porous medium permeability prediction method based on three-dimensional convolutional neural network | |
| CN111242028A (en) | Remote sensing image ground object segmentation method based on U-Net | |
| Argilaga | Fractal Informed Generative Adversarial Networks (FI-GAN): Application to the generation of X-ray CT images of a self-similar partially saturated sand | |
| CN117849193A (en) | Online crack damage monitoring method for neodymium iron boron sintering | |
| CN118015839A (en) | Expressway road domain risk prediction method and device | |
| CN115292538A (en) | Map line element extraction method based on deep learning | |
| CN113591608A (en) | High-resolution remote sensing image impervious surface extraction method based on deep learning | |
| Machairas et al. | Application of dynamic image analysis to sand particle classification using deep learning | |
| CN116680988A (en) | Porous medium permeability prediction method based on Transformer network | |
| CN116563251A (en) | Intelligent processing method and system for tie rod ball head for automobile | |
| CN113592829B (en) | Deep learning sediment particle identification method and device based on segmentation and recombination | |
| CN113449626A (en) | Hidden Markov model vibration signal analysis method and device, storage medium and terminal |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210618 |