CN109800863B - Logging phase identification method based on fuzzy theory and neural network - Google Patents

Logging phase identification method based on fuzzy theory and neural network Download PDF

Info

Publication number
CN109800863B
CN109800863B CN201910042398.8A CN201910042398A CN109800863B CN 109800863 B CN109800863 B CN 109800863B CN 201910042398 A CN201910042398 A CN 201910042398A CN 109800863 B CN109800863 B CN 109800863B
Authority
CN
China
Prior art keywords
fuzzy
neural network
layer
logging
data
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.)
Active
Application number
CN201910042398.8A
Other languages
Chinese (zh)
Other versions
CN109800863A (en
Inventor
李忠伟
张卫山
宋弢
卢清华
崔学荣
刘昕
赵德海
何旭
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.)
China University of Petroleum East China
Original Assignee
China University of Petroleum East 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 China University of Petroleum East China filed Critical China University of Petroleum East China
Priority to CN201910042398.8A priority Critical patent/CN109800863B/en
Publication of CN109800863A publication Critical patent/CN109800863A/en
Application granted granted Critical
Publication of CN109800863B publication Critical patent/CN109800863B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/043Architecture, e.g. interconnection topology based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16ZINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
    • G16Z99/00Subject matter not provided for in other main groups of this subclass
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Fuzzy Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Automation & Control Theory (AREA)
  • Image Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a logging phase identification method based on a fuzzy theory and a neural network, firstly, a fuzzy area convolutional neural network is constructed, a given target hypothesis area and target identification are put into the same network, convolution calculation is shared, and the weight of the whole network is updated in a training process; and then, carrying out convolution and pooling operation on the logging data through a fuzzy area convolutional neural network, carrying out fuzzy operation on the convolution layer and the pooling layer in an interaction mode, starting from the first layer of the fuzzy area convolutional neural network, gradually increasing the number of fuzzy layers, adjusting the fuzzy layers according to different data sets, obtaining a feature vector at the last layer of the fuzzy area convolutional neural network, mapping the feature vector into a low-dimensional vector through a sliding window, and then inputting the feature into two fully-connected layers, wherein one fully-connected layer is used for positioning, and the other fully-connected layer is used for classifying.

Description

Logging phase identification method based on fuzzy theory and neural network
Statement of divisional application
The application is a divisional application of Chinese invention patent application with the application number of 201610762101.1, and the name of the invention is 'parallelization method of a fuzzy area convolutional neural network under a big data environment' submitted by the year 2016 and the month 08.
Technical Field
The invention relates to the technical field of petroleum logging, in particular to the field of big data logging.
Background
Logging information and depositions are reflecting and controlling factors of formation petrophysical properties, so logging information has been used as an important information source for the basis of the study of oil and gas reservoir depositions, and logging phases are bridges between logging information and reservoir depositional characteristics. For most oil and gas wells, logging data is the only comprehensive information source covering all well formations, so the logging phase identification analysis method has been used as the most important research means in oil and gas exploration and development geological research.
However, logging information is characterized by ambiguity, with geologic polynomials and ambiguity. Therefore, the identification and analysis of the logging phases are required to be established on the basis of comprehensive depth analysis of a large number of existing sediment characteristics and logging parameter relations (logging response), and meanwhile, modeling methods suitable for geological characteristics are selected by referring to results of field outcrop, core logging and seismic analysis, so that the accurate identification of the logging phases can be realized.
In addition, due to the lack of an effective automatic identification method and technology of the logging facies, the current logging facies identification is mainly realized through manual identification of geological staff, and due to factors such as personnel experience differences, subjective differences, systematic differences of logging data and the like, the geological staff is large in data volume and heavy in workload. Moreover, due to factors such as experience differences of geological personnel, subjective factors, systematic differences of logging data of different instruments in different periods and the like, the conventional logging phase identification accuracy is greatly reduced.
The advanced technologies such as big data analysis, deep learning and the like are applied to oil-gas geology research, so that the problem of idle exploration and attempt of big data analysis resources in the current petroleum industry is solved. In recent years, the petroleum industry has established a large number of cloud data centers, but the utilization rate is not high, and resources are seriously wasted. One of the important reasons is the lack of large data processing platforms and corresponding large data technologies to fully utilize these computing and storage resources.
Establishing an efficient and accurate logging phase identification method is an urgent need for current oil and gas geology research.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides a parallelization method of a fuzzy area convolutional neural network in a big data environment.
The technical scheme of the invention is realized as follows:
the parallelization method of the fuzzy area convolution neural network under the big data environment comprises the steps of firstly, constructing the fuzzy area convolution neural network, putting a given target hypothesis area and target identification into the same network, sharing convolution calculation, and updating the weight of the whole network in a training process;
next, dividing an input logging data set into a plurality of small data sets, parallelizing a plurality of workflows, performing convolution and pooling operations through a fuzzy area convolutional neural network, and training each small data set by gradient descent alone; after training is completed, outputting the result to a waiting queue, reading the output queue after one round of training is completed, performing synchronous updating operation of the shared weight, and performing the next round of training after updating is completed; in each training round, the calculation of each divided small data set is asynchronously carried out on a distributed basis, each gradient value calculated is added to a list, after all the small data sets are calculated, the weight and the offset value of the fuzzy area convolutional neural network are synchronously updated, and then the next training round is carried out; in the aspect of parallelization identification, collecting logging data by Spout, distributing the data to each Bolt node for parallelization logging phase identification, inputting an identification result to the next Bolt node by each Bolt node, and counting object information in the identification result;
each small data set is subjected to convolution and pooling operation through a fuzzy area convolutional neural network, and the method specifically comprises the following steps of: the convolution layer and the pooling layer are interacted, fuzzy operation is carried out on the convolution layer and the pooling layer, the number of fuzzy layers is gradually increased from the first layer of the fuzzy area convolution neural network, the number of fuzzy layers is adjusted for different data sets, the last layer of the fuzzy area convolution neural network obtains a feature vector, the feature vector maps features into a low-dimensional vector through a sliding window, then the features are input into two fully connected layers, one fully connected layer is used for positioning, and the other fully connected layer is used for classifying.
Optionally, the convolution layer formula is expressed as:
Figure GDA0004164076360000031
the pooling layer formula is expressed as:
Figure GDA0004164076360000032
wherein the bias is
Figure GDA0004164076360000033
And weight->
Figure GDA0004164076360000034
Are all fuzzy numbers, here using symmetric triangular fuzzy numbers,>
Figure GDA0004164076360000035
the j-th fuzzy number is a vector composed of fuzzy numbers>
Figure GDA0004164076360000036
The membership functions of (a) are:
Figure GDA0004164076360000037
optionally, during training of the fuzzy region convolutional neural network, a joint loss function is defined:
Figure GDA0004164076360000038
wherein ,pi Is the predicted probability that this sample is the log morphology,
Figure GDA0004164076360000039
a label of the sample, if it is the corresponding log morphology,/->
Figure GDA00041640763600000310
1, otherwise->
Figure GDA00041640763600000311
Is 0, L cls Is a two-class logical loss; t is t i Is a vector of four parameters predicting the boundary of an object,/->
Figure GDA00041640763600000312
For the vector composed of the labeling area parameters, they are respectively:
t x =(x-x a )/w a t h =(y-y a )/h a
t w =log(w/w a ) t h =log(h/h a )
Figure GDA0004164076360000041
Figure GDA0004164076360000042
wherein (x, y), w and h represent the center coordinates, width and length of the object, x, respectively a ,x * Respectively representPrediction region, anchoring region and labeling region, regression loss
Figure GDA0004164076360000043
R is the smooth loss function->
Figure GDA0004164076360000044
Figure GDA0004164076360000045
Figure GDA0004164076360000046
Indicating that only when the anchor region is positive sample +.>
Figure GDA0004164076360000047
Only then calculate regression loss, otherwise +.>
Figure GDA0004164076360000048
Not calculating, normalizing the parameter N cls and Nreg Representing the length of the low-dimensional vector mapped from the feature vector and the number of anchor regions, respectively. />
Optionally, normalization of logging data is performed first, original data are converted into dimensionless index evaluation values, the index evaluation values are all in the same number level, and comprehensive evaluation analysis is performed.
Optionally, the normalization of the logging data is performed using the following normalization method:
Sx=(x-M)/S,x∈{GR,AC,DEN,CNL,SDN,...}
wherein x represents the data of each log, sx represents normalized log data, M is the mean value of the corresponding log data, and S is the standard deviation of each log data.
The beneficial effects of the invention are as follows:
(1) According to the characteristic of data blurring in logging big data, a blurring theory is integrated, a blurring region convolution neural network FR-CNN is provided, a progressive blurring method is provided, and the number of blurring layers is gradually increased from the first layer of the blurring region convolution neural network, so that the network structure and parameters are optimized, and better analysis performance and precision are realized;
(2) The FR-CNN fuzzification layer number is adjusted for different logging data sets, so that the extracted characteristics better reflect the characteristics of an oil and gas reservoir, and the problem of logging data fuzziness can be solved;
(3) The invention uses multiple GPUs to perform the parallel training and execution of the FR-CNN so as to improve the efficiency of the FR-CNN.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a fuzzy region convolutional neural network of the present invention;
FIG. 2 is a schematic diagram of a symmetric triangle fuzzy number coordinate;
FIG. 3 is a schematic diagram of the parallel processing of real-time data by the fuzzy region convolutional neural network of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The ambiguity of the logging data is caused by various reasons including data space pollution of the logging data caused by noise, inconsistency, incompleteness and the like, and systematic data differences caused by logging of different instruments in different periods, and the ambiguity of the logging data caused by the problems restrict the accurate identification of logging phases.
The invention provides a parallelization method of a Fuzzy area convolutional neural network in a big data environment, a multidimensional data space is constructed for logging data, a Fuzzy theory and a deep learning network R-CNN are fused, a method for identifying logging phases under the condition of solving the Fuzzy data is provided, the Fuzzy theory is fused according to the characteristic of data blurring in the logging big data, the Fuzzy area convolutional neural network FR-CNN (Fuzzy R-CNN) is provided, a progressive Fuzzy method is further provided, the number of Fuzzy layers is gradually increased from a first layer of the convolutional neural network, network structures and parameters are optimized, and finally, the theory and the method of the FR-CNN are established, so that better analysis performance and accuracy are achieved.
The design of a suitable fuzzy area convolutional neural network is an important point of the invention, and the construction of the fuzzy area convolutional neural network is described in detail below.
The fuzzy region convolution neural network FR-CNN is established on the basis of the deep learning network R-CNN, as shown in fig. 1, the FR-CNN puts the given target hypothesis region and target identification into the same network, so as to share convolution calculation, avoid complex calculation steps, update the weight of the whole network only by one training process, and accelerate the detection speed at the same time, thereby achieving the purpose of rapid processing.
In fig. 1, logging data is rolled and pooled through a fuzzy region convolutional neural network. The core of the fuzzy area convolutional neural network training is the interaction of a convolutional layer and a pooling layer, so that fuzzy operation is carried out on the convolutional layer and the pooling layer. In order to avoid excessive information loss caused by excessive blurring, and considering that the refinement degree of the extracted features of the fuzzy region convolution neural network is reduced layer by layer, the blurring of each layer of the traditional fuzzy neural network is changed.
The last layer of the fuzzy region convolutional neural network gets feature vectors that map features into a low-dimensional vector through a small sliding window, then input features into two fully connected layers, one for localization and the other for classification. At each sliding window several target hypothesis regions, which may be referred to as anchor regions, are simultaneously presented, centered on the sliding window, with different aspect ratios and scales.
The convolutional layer formula for convolutional neural network R-CNN can be expressed as:
Figure GDA0004164076360000071
wherein ,
Figure GDA0004164076360000072
representing the value at the (x, y) position of the j-th eigenvector of the i-th layer neuron,/->
Figure GDA0004164076360000073
Representation->
Figure GDA0004164076360000074
The convolution kernel connected to the mth eigenvector weights at locations (p, q). P (P) i and Qi Representing the height and width of the convolution kernel, b ij For the bias term, f (x) represents the activation function of the neuron.
The R-CNN pooling layer formula is expressed as:
x ij =f(β ij down(x i-1j )+b ij ) (2)
down (-) represents a downsampling function, typically by summing all information of different n x n blocks of input data, such that the output data is scaled down n times in both dimensions, each output map corresponding to an own multiplicative offset β and an additive offset b.
The input and calculation processes of the convolutional neural network are real numbers, the obtained results are deterministic, and for the condition of data ambiguity such as data loss, the fuzzy theory is introduced into the fuzzy region convolutional neural network, and the improved formula is as follows:
the convolution layer formula is expressed as:
Figure GDA0004164076360000081
the pooling layer formula is expressed as:
Figure GDA0004164076360000082
wherein the bias is
Figure GDA0004164076360000083
And weight->
Figure GDA0004164076360000084
Are all fuzzy numbers, here using symmetric triangular fuzzy numbers,>
Figure GDA0004164076360000085
the j-th fuzzy number is a vector composed of fuzzy numbers>
Figure GDA0004164076360000086
Membership function of +.>
Figure GDA0004164076360000087
As shown in FIG. 2, w j Is the center of symmetry of the fuzzy number,
Figure GDA0004164076360000088
is half length of the fuzzy number, +.>
Figure GDA0004164076360000089
Representing the membership at w.
In the training process of the fuzzy area convolutional neural network, a joint loss function is defined:
Figure GDA00041640763600000810
wherein pi Is the predicted probability that this sample is the log morphology,
Figure GDA00041640763600000811
a label of the sample, if it is the corresponding log morphology,/->
Figure GDA00041640763600000812
1, otherwise->
Figure GDA00041640763600000813
Is 0, N cls Is a binary class (0 or 1) logic loss.
t i Is a vector of four parameters predicting object boundaries,
Figure GDA00041640763600000814
for the vector composed of the labeling area parameters, they are respectively:
t x =(x-x a )/w a t h =(y-y a )/h a (7)
t w =log(w/w a ) t h =log(h/h a )
Figure GDA00041640763600000815
Figure GDA00041640763600000816
wherein (x, y), w and h each representCenter coordinates, width and length of object, x a ,x * Representing the predicted region, the anchor region and the labeled region, respectively (y, w, h are the same). Regression loss
Figure GDA0004164076360000091
R is a smooth loss function
Figure GDA0004164076360000092
Figure GDA0004164076360000093
Figure GDA0004164076360000094
Representing +.>
Figure GDA0004164076360000095
Only calculate regression loss, otherwise +.>
Figure GDA0004164076360000096
And not calculated. Normalized parameter N cls and Nreg Representing the length of the low-dimensional vector mapped from the feature vector and the number of anchor regions, respectively.
Different logging tools may be used to generate different data. If natural Gamma (GR), compensating sound wave (AC), compensating Density (DEN), compensating Neutrons (CNL), neutron apparent porosity and density apparent porosity difference (SDN) are adopted, and the like, the data are of different dimensions, and no comparability exists between the data, therefore, the invention needs to normalize logging data, convert original data into dimensionless index evaluation values, namely, all index evaluation values are in the same number level, and then perform comprehensive evaluation analysis.
The following normalization method is adopted:
Sx=(x-M)/S,x∈{GR,AC,DEN,CNL,SDN,...}
wherein x represents the data of each logging curve, and Sx represents the normalized logging curve data; m is the mean value of the corresponding log data, and S is the standard deviation of each log data.
The invention performs calibration from the existing logging data, establishes the training data set of the FR-CNN, and on the basis, as the information disclosed by different logging methods is different, selects the combination of different logging data as the input of the FR-CNN, thereby determining the optimal logging data combination of the FR-CNN and optimizing the network parameters and the structure of the FR-CNN when logging phase identification is performed.
The FR-CNN has more two types of full-connection layers than the traditional convolutional neural network, and has more operations such as calculation of region coordinates, and the calculation amount of the operations is large. In the fuzzy neural network, fuzzy operation exists in each layer of the network, that is, the deeper the network is, the more calculation amount is increased, which makes the network which is required to be heavy and calculation is heavy. The increase of the calculated amount leads to the great increase of the training time of the network, prolongs the updating period of the network model, weakens the flexibility of the system, and prolongs the detection time.
According to the invention, FR-CNN training and operation efficiency are improved through parallelization, an input logging data set is firstly divided into a plurality of small data sets, a plurality of workflows are operated simultaneously, and each part is trained by gradient descent alone. After training is completed, the result is output to a waiting queue, and after one round of training is completed, the output queue is read, and synchronous updating operation of the sharing weight is carried out. After the updating is completed, the next training round is performed.
In each training round, the calculation of each divided small data set is asynchronously performed on a distributed basis, each gradient value calculated is added to the list, and after all the small data sets are calculated, the weight and the bias value of the network are synchronously updated, and then the next training round is performed.
As shown in fig. 3, in the aspect of parallelization identification, the solution adopted is: collecting logging data by Spout, distributing the data to each Bolt node for logging phase identification in parallel, inputting an identification result to the next Bolt node by each Bolt node, and counting object information in the next Bolt node.
According to the characteristic of data blurring in logging big data, a blurring theory is integrated, a blurring region convolution neural network FR-CNN is provided, a progressive blurring method is provided, the number of blurring layers is gradually increased from the first layer of the blurring region convolution neural network, and therefore network structures and parameters are optimized, and better analysis performance and accuracy are achieved; in addition, the invention adjusts the FR-CNN fuzzification layer number aiming at different logging data sets, so that the extracted characteristics better reflect the characteristics of an oil and gas reservoir, and the problem of logging data fuzziness can be solved; aiming at the problem of large operation calculation amount, the invention utilizes multiple GPUs to carry out the parallel training and execution of the FR-CNN so as to improve the efficiency of the FR-CNN.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.

Claims (3)

1. A logging phase identification method based on fuzzy theory and neural network is characterized in that,
firstly, constructing a fuzzy region convolutional neural network, putting a given target hypothesis region and target identification into the same network, sharing convolutional calculation, and updating the weight of the whole network in a training process;
then, the logging data is subjected to convolution and pooling operation through a fuzzy area convolutional neural network, the convolution layer and the pooling layer are interacted, fuzzy operation is performed on the convolution layer and the pooling layer, the number of fuzzy layers is gradually increased from the first layer of the fuzzy area convolutional neural network, the number of fuzzy layers is adjusted for different data sets, the last layer of the fuzzy area convolutional neural network obtains a feature vector, the feature vector maps features into a low-dimensional vector through a sliding window, then the features are input into two fully connected layers, one fully connected layer is used for positioning, and the other fully connected layer is used for classifying;
the convolution layer formula is expressed as:
Figure FDA0004164076350000011
wherein ,
Figure FDA0004164076350000012
representing the value at the (x, y) position of the j-th eigenvector of the i-th layer neuron, P i and Qi Respectively representing the height and width of the convolution kernel, f (x) representing the activation function of the neuron;
the pooling layer formula is expressed as:
Figure FDA0004164076350000013
where down (-) represents a downsampling function, offset
Figure FDA0004164076350000014
And weight->
Figure FDA0004164076350000015
All are fuzzy numbers +.>
Figure FDA0004164076350000016
Representation->
Figure FDA0004164076350000017
Weights at positions (p, q) connected to the convolution kernel of the mth eigenvector, where symmetric triangle ambiguity is used, < >>
Figure FDA0004164076350000018
The j-th fuzzy number is a vector composed of fuzzy numbers>
Figure FDA0004164076350000019
The membership functions of (a) are:
Figure FDA00041640763500000110
wherein wj Is the center of symmetry of the fuzzy number,
Figure FDA00041640763500000111
is half length of the fuzzy number, +.>
Figure FDA00041640763500000112
Represents the membership at w; in the training process of the fuzzy area convolutional neural network, a joint loss function is defined:
Figure FDA0004164076350000021
wherein ,pi Is the predicted probability that this sample is the log morphology,
Figure FDA0004164076350000022
a label of the sample, if it is the corresponding log morphology,/->
Figure FDA0004164076350000023
1, otherwise->
Figure FDA0004164076350000024
Is 0, L cls Is a two-class logical loss; t is t i Is a vector of four parameters predicting the boundary of an object,/->
Figure FDA0004164076350000025
For the vector composed of the labeling area parameters, they are respectively:
t x =(x-x c )/w a t h =(y-y a )/h a
t w =log(w/w a )t h =log(h/h a )
Figure FDA0004164076350000026
Figure FDA0004164076350000027
wherein (x, y), w and h represent the center coordinates, width and length of the object, x, respectively a ,x * Representing the predicted region, the anchor region and the labeled region, respectively, regression loss
Figure FDA0004164076350000028
R is the smooth loss function->
Figure FDA0004164076350000029
Figure FDA00041640763500000210
Figure FDA00041640763500000211
Indicating that only when the anchor region is positive sample +.>
Figure FDA00041640763500000212
Only then calculate regression loss, otherwise +.>
Figure FDA00041640763500000213
Not calculating, normalizing the parameter N cls and Nreg Representing the length of the low-dimensional vector mapped from the feature vector and the number of anchor regions, respectively. />
2. The method for identifying the logging phases based on the fuzzy theory and the neural network according to claim 1, wherein the normalization of logging data is performed first, raw data are converted into dimensionless index evaluation values, the index evaluation values are all in the same number level, and then comprehensive evaluation analysis is performed.
3. The method for identifying the logging phases based on the fuzzy theory and the neural network as claimed in claim 2, wherein the normalization of the logging data is carried out by adopting the following normalization method:
Sx=(x-M)/S,x∈{GR,AC,DEN,CNL,SDN,...}
wherein x represents data of each log, sx represents normalized log data, M is an average value of the corresponding log data, S is a standard deviation of each log data, GR is natural gamma, AC is a compensating sound wave, DEN is a compensating density, CNL is a compensating neutron, and SDN is a density viewing porosity difference.
CN201910042398.8A 2016-08-30 2016-08-30 Logging phase identification method based on fuzzy theory and neural network Active CN109800863B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910042398.8A CN109800863B (en) 2016-08-30 2016-08-30 Logging phase identification method based on fuzzy theory and neural network

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910042398.8A CN109800863B (en) 2016-08-30 2016-08-30 Logging phase identification method based on fuzzy theory and neural network
CN201610762101.1A CN106372402B (en) 2016-08-30 2016-08-30 The parallel method of fuzzy region convolutional neural networks under a kind of big data environment

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
CN201610762101.1A Division CN106372402B (en) 2016-08-30 2016-08-30 The parallel method of fuzzy region convolutional neural networks under a kind of big data environment

Publications (2)

Publication Number Publication Date
CN109800863A CN109800863A (en) 2019-05-24
CN109800863B true CN109800863B (en) 2023-05-23

Family

ID=57902416

Family Applications (2)

Application Number Title Priority Date Filing Date
CN201910042398.8A Active CN109800863B (en) 2016-08-30 2016-08-30 Logging phase identification method based on fuzzy theory and neural network
CN201610762101.1A Active CN106372402B (en) 2016-08-30 2016-08-30 The parallel method of fuzzy region convolutional neural networks under a kind of big data environment

Family Applications After (1)

Application Number Title Priority Date Filing Date
CN201610762101.1A Active CN106372402B (en) 2016-08-30 2016-08-30 The parallel method of fuzzy region convolutional neural networks under a kind of big data environment

Country Status (1)

Country Link
CN (2) CN109800863B (en)

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106446514A (en) * 2016-08-31 2017-02-22 中国石油大学(华东) Fuzzy theory and neural network-based well-log facies recognition method
CN106980895A (en) * 2017-02-22 2017-07-25 中国科学院自动化研究所 Convolutional neural networks Forecasting Methodology based on rotary area
CN106980874B (en) * 2017-03-13 2019-09-10 南京邮电大学 A kind of multi-time Scales dimension data fusion method towards distribution big data
CN107168527B (en) * 2017-04-25 2019-10-18 华南理工大学 The first visual angle gesture identification and exchange method based on region convolutional neural networks
US11244226B2 (en) 2017-06-12 2022-02-08 Nvidia Corporation Systems and methods for training neural networks with sparse data
CN109034385A (en) * 2017-06-12 2018-12-18 辉达公司 With the system and method for sparse data training neural network
CN107678059B (en) * 2017-09-05 2019-06-28 中国石油大学(北京) A kind of method, apparatus and system of reservoir gas-bearing identification
US10892953B2 (en) * 2018-02-23 2021-01-12 Nec Corporation Network-based application acceleration
CN108629072B (en) * 2018-03-12 2019-07-09 山东科技大学 Convolutional neural networks study and prediction technique towards the distribution of earthquake oil and gas reservoir
CN108648495A (en) * 2018-06-08 2018-10-12 华南理工大学 A kind of method and system of the intelligence real-time display bus degree of crowding
CN109171774B (en) * 2018-09-30 2021-07-06 合肥工业大学 Personality analysis method and system based on multi-channel data
CN109920013B (en) * 2019-01-30 2020-11-10 北京交通大学 Image reconstruction method and device based on progressive convolution measurement network
CN109919313B (en) * 2019-01-31 2021-06-08 华为技术有限公司 Gradient transmission method and distributed training system
CN109816042B (en) * 2019-02-01 2020-11-24 北京达佳互联信息技术有限公司 Data classification model training method and device, electronic equipment and storage medium
CN111236932B (en) * 2020-01-13 2022-12-13 中国石油天然气股份有限公司大港油田分公司 Shale oil lithology evaluation method based on indication curve reconstruction
CN111462060A (en) * 2020-03-24 2020-07-28 湖南大学 Method and device for detecting standard section image in fetal ultrasonic image
CN112784980B (en) * 2021-01-05 2024-05-28 中国石油天然气集团有限公司 Intelligent logging horizon dividing method
CN113537458B (en) * 2021-06-21 2022-12-30 中国科学技术大学 Rational function neural network construction method, system and readable storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104217225A (en) * 2014-09-02 2014-12-17 中国科学院自动化研究所 A visual target detection and labeling method
CN104573669A (en) * 2015-01-27 2015-04-29 中国科学院自动化研究所 Image object detection method
CN104657748A (en) * 2015-02-06 2015-05-27 中国石油大学(华东) Vehicle type recognition method based on convolutional neural network
CN105654729A (en) * 2016-03-28 2016-06-08 南京邮电大学 Short-term traffic flow prediction method based on convolutional neural network
CN105825511A (en) * 2016-03-18 2016-08-03 南京邮电大学 Image background definition detection method based on deep learning

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1758398A1 (en) * 2005-08-23 2007-02-28 Syneola SA Multilevel semiotic and fuzzy logic user and metadata interface means for interactive multimedia system having cognitive adaptive capability
US20080033991A1 (en) * 2006-08-03 2008-02-07 Jayanta Basak Prediction of future performance of a dbms
US11256982B2 (en) * 2014-07-18 2022-02-22 University Of Southern California Noise-enhanced convolutional neural networks
CN105354568A (en) * 2015-08-24 2016-02-24 西安电子科技大学 Convolutional neural network based vehicle logo identification method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104217225A (en) * 2014-09-02 2014-12-17 中国科学院自动化研究所 A visual target detection and labeling method
CN104573669A (en) * 2015-01-27 2015-04-29 中国科学院自动化研究所 Image object detection method
CN104657748A (en) * 2015-02-06 2015-05-27 中国石油大学(华东) Vehicle type recognition method based on convolutional neural network
CN105825511A (en) * 2016-03-18 2016-08-03 南京邮电大学 Image background definition detection method based on deep learning
CN105654729A (en) * 2016-03-28 2016-06-08 南京邮电大学 Short-term traffic flow prediction method based on convolutional neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
基于 BP 神经网络的测井相分析及沉积相识别;吴灿灿等;《煤田地质与勘探》;20120229;第 40 卷(第 1 期);全文 *
应用神经网络模式识别技术进行测井沉积学研究;张福明等;《石油勘探与开发》;20030630;第30卷(第3期);全文 *
苏北盆地测井层序与沉积相研究;程道解;《中国优秀硕士学位论文全文数据库基础科学辑》;20090615;全文 *

Also Published As

Publication number Publication date
CN109800863A (en) 2019-05-24
CN106372402A (en) 2017-02-01
CN106372402B (en) 2019-04-30

Similar Documents

Publication Publication Date Title
CN109800863B (en) Logging phase identification method based on fuzzy theory and neural network
CN106529667B (en) Well-log facies recognition and analysis method based on fuzzy deep learning under a kind of big data environment
CN112083498B (en) Multi-wave earthquake oil and gas reservoir prediction method based on deep neural network
CN110674841B (en) Logging curve identification method based on clustering algorithm
CN101271469B (en) Two-dimension image recognition based on three-dimensional model warehouse and object reconstruction method
CN107545577B (en) Sedimentary facies image segmentation method based on neural network
CN109711636A (en) River water level prediction method based on chaotic firefly and gradient lifting tree model
CN111596978A (en) Web page display method, module and system for lithofacies classification by artificial intelligence
US8090538B2 (en) System and method for interpretation of well data
CN110490100A (en) Ground automatic identification based on deep learning names method and system
CN116665067B (en) Ore finding target area optimization system and method based on graph neural network
CN106446514A (en) Fuzzy theory and neural network-based well-log facies recognition method
CN109345007B (en) Advantageous reservoir development area prediction method based on XGboost feature selection
CN116704137B (en) Reverse modeling method for point cloud deep learning of offshore oil drilling platform
US11899148B1 (en) Seismic imaging free gas structure identification method and system
CN113902861A (en) Three-dimensional geological modeling method based on machine learning
CN111914478A (en) Comprehensive geological drilling well logging lithology identification method
CN110363299A (en) Space reasoning by cases method towards delamination-terrane of appearing
Yi et al. An Effective Approach for determining Rock Discontinuity sets using a modified Whale optimization Algorithm
CN112634447B (en) Outcrop stratum layering method, device, equipment and storage medium
CN117408167A (en) Debris flow disaster vulnerability prediction method based on deep neural network
US20230141334A1 (en) Systems and methods of modeling geological facies for well development
CN114036829B (en) Geological profile generation method, system, equipment and storage medium
CN114239937A (en) Reservoir oil-gas-containing property prediction method and device, computer equipment and storage medium
CN106227959B (en) Method and device for predicting lithologic reservoir favorable area based on four-graph superposition 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
GR01 Patent grant
GR01 Patent grant