CN110132626A - A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks - Google Patents
A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks Download PDFInfo
- Publication number
- CN110132626A CN110132626A CN201910419291.0A CN201910419291A CN110132626A CN 110132626 A CN110132626 A CN 110132626A CN 201910419291 A CN201910419291 A CN 201910419291A CN 110132626 A CN110132626 A CN 110132626A
- Authority
- CN
- China
- Prior art keywords
- multiple dimensioned
- neural networks
- convolutional neural
- feature
- network
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/007—Subject matter not provided for in other groups of this subclass by applying a load, e.g. for resistance or wear testing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of Fault Diagnoses of Oil Pump methods based on multiple dimensioned convolutional neural networks, conventional fault diagnosis method, which exists, depends on artificial selection feature, it calculates complicated, the not high problem of accuracy rate, the existing deep neural network applied to Fault Diagnoses of Oil Pump is completed in unipath, the size of filter is all single setting in each layer, limits the flexibility of parameter.Using multiple dimensioned convolution block as nuclear structure, a kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks is proposed.This method, avoid the Feature Engineering of conventional fault diagnosis method complexity and the uncertain influence to fault identification accuracy rate of feature selecting, meanwhile this method can be extracted that indicator card is more abundant, effective global and local feature, improve fault diagnosis accuracy rate.
Description
Technical field
The present invention relates to Fault Diagnoses of Oil Pump technologies, and in particular to a kind of oil pumping based on multiple dimensioned convolutional neural networks
Machine method for diagnosing faults.
Background technique
Sucker rod pumping machine is widely used in China Petroleum.Currently, the most widely used sucker rod pumping failure of pump
Diagnostic method is analyzed using indicator card data.Indicator card diagnostic method primarily now is to carry out feature to indicator card to mention
It takes, indicator card is analyzed using the methods of not bending moment, Freeman chain code, gray matrix, extraction can effectively reflect oil pumping
The feature vector of pump condition is diagnosed in conjunction with diagnostic models such as BP neural network, support vector machines (SVM).These failures
Diagnostic method all relies on artificial selection feature, calculates complexity, and fault identification accuracy rate is not high.
With the development of deep learning, powerful feature learning ability and classification capacity are attracted wide public concern.It is some
Deep learning method is gradually applied to the fault diagnosis of sucker rod pumping machine.Indicator card is carried out using depth confidence network automatic special
Sign is extracted, and is avoided complicated characteristic extraction procedure, is improved accuracy of identification and speed;Existed using improved Alexnet network
Normally, it is not fully filled, achieves good recognition effect on the disconnected four kinds of operating mode's switch of gases affect, sucker rod;Use convolutional Neural
Network (CNN) automatically extracts feature to indicator card realization, and SVM is recycled to carry out fault identification as classifier, these methods exist
In indicator card fault diagnosis, superiority has been embodied compared to traditional method for diagnosing faults.
But the existing deep neural network applied to indicator card fault diagnosis is completed in unipath, filtering
The size of device is all single setting in each layer, limits the flexibility of parameter.
Summary of the invention
The technical problem to be solved by the present invention is a kind of base in order to overcome the deficiencies in the prior art, the present invention provides
In the Fault Diagnoses of Oil Pump method of multiple dimensioned convolutional neural networks.
The technical scheme is that a kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks,
It is characterized in that including the following steps:
Step 1: collecting suspension point displacement and load data, initial data is pre-processed, according to network model to input
Indicator card is drawn in the requirement of data size;
Step 2: for indicator card data creating label, indicate the label of x-1 class indicator card with the number of 0~x, and by its
It is divided into training set and test set;
Step 3: construct multiple dimensioned convolution block, extract the global and local information of indicator card, multiple dimensioned convolution block structure by
(2m+3) * (2m+3), (2m+5) * (2m+5), three kinds of (2m+7) * (2m+7) different size convolution kernels are constituted, and m is just greater than 0
Integer, three kinds of various sizes of convolution kernels are extracted the feature of input feature vector figure by convolution operation, feature extraction result are carried out
Splice the output feature as multiple dimensioned convolutional layer;
Step 4: building multiple dimensioned convolutional neural networks model by core of multiple dimensioned convolution block, include p in network structure
A multiple dimensioned convolutional layer, q maximum pond layer, k convolutional layer, n be a to be averaged and pond layers and outputs and inputs layer, wherein p, q,
K, m is the positive integer more than or equal to 0, carries out nonlinear processing using Swish activation primitive after each convolution operation;
Wherein, Swish activation primitive are as follows:
β > 0 in formula;
Step 5: carrying out multiple dimensioned convolutional neural networks model training on the training set obtained in step 2, use intersection
Loss function of the entropy function as model minimizes loss function using optimizer, carries out network parameter training, saves best net
Network model parameter;
Wherein, cross entropy loss function are as follows:
In formula, y is the probability distribution of prediction label, and y ' is the probability distribution of true tag, is judged with intersection entropy function
Order of accuarcy of the model to true probability distribution estimation;
Step 6: testing characteristics of network on test set in step 2 saves network mould if test effect reaches expected
Type, otherwise return step 5, re-start network parameter adjustment.
The structure of the building of multiple dimensioned convolution block in step 2, multiple dimensioned convolution block can plant different size rolls by n (n >=1)
Product core is constituted.
Beneficial outcomes of the invention are:
(1) multiple dimensioned convolutional neural networks do not need manually to carry out compared with traditional sucker rod pumping machine method for diagnosing faults
Feature extraction, but deep neural network independently extracts feature, avoids the uncertain of complicated Feature Engineering and feature selecting
Influence of the property to fault identification accuracy rate.
(2) compared with existing deep neural network method, multiple dimensioned convolutional neural networks using multiple dimensioned convolution block with
Various sizes of convolution kernel carries out convolution operation to input feature vector figure, and it is more abundant, effective can to extract indicator card data
Global and local feature, so that being improved using the network model fault diagnosis accuracy rate that multiple dimensioned convolution block is built as core.
Detailed description of the invention
Fig. 1 is flow chart of the invention, is normalized to initial data, is smoothly equal to processing, according to the depth built
The input size requirements of neural network are spent, the indicator card data of corresponding size size is drawn, data is divided into data set and test
Collection saves network model, then test on test set the network model using training set training network, if test
Effect differs too big with the test effect on training set, and re -training model parameter, otherwise preservation model is as final mask;
Fig. 2 is multiple dimensioned convolution block structural diagram, defeated to upper one layer using the convolution kernel of 32 3*3,5*5,7*7 sizes respectively
Enter characteristic pattern and carry out convolution operation, convolution results are carried out to splice the output feature as multiple dimensioned convolution block;
Fig. 3 is multiple dimensioned convolutional neural networks model structure, when inputting as " gases affect " fault type indicator card, warp
Multiple dimensioned convolutional neural networks model is crossed, then output is to indicate the one-hot coding form of the digital label of the fault type;
Fig. 4 be accuracy result figure of the multiple dimensioned convolutional neural networks model on training set and test set, training set and
When train epochs are 400 step, accuracy tends towards stability test set, and the fault identification accuracy rate on test set reaches 98%
Left and right.
Specific embodiment
The present invention is described in detail below in conjunction with attached drawing.Fig. 1 is overall flow figure of the invention.Tool of the invention
Body implementation steps are as follows:
Step 1: collecting suspension point displacement and load data, initial data is pre-processed, specific pretreatment mode are as follows:
(1) smooth and normalized is carried out to data;
(2) two-dimentional indicator card is drawn with displacement and load data, the input of network model is 32*32, so by indicator card
Size is also handled as 32*32 pixel.
Step 2: for indicator card data creating label, and it being divided into training set and test set by a certain percentage.
Indicator card label is converted into one-hot coding form, such as: indicator card data totally 8 class, gases affect operating condition are the
4 classes, original tag 3, one-hot coding form label are [0,0,0,1,0,0,0,0].
Step 3: construct multiple dimensioned convolution block:
Multiple dimensioned convolution block structure is by (2m+3) * (2m+3), (2m+5) * (2m+5), three kinds of (2m+7) * (2m+7) different rulers
Very little convolution kernel is constituted, and m is the positive integer greater than 0, this example m takes 1.Multiple dimensioned convolution block structure is as shown in Fig. 2, be by 3* at this time
3, tri- kinds of size convolution kernels of 5*5,7*7 are constituted, and three kinds of various sizes of convolution kernels extract the feature of input feature vector figure, by feature
Result is extracted to carry out splicing the output feature as multiple dimensioned convolutional layer.Each size convolution kernel number is 32, multiple dimensioned convolution
The output of block is 96.
Step 4: build multiple dimensioned convolutional neural networks model:
Include p (p >=0, positive integer) a multiple dimensioned convolutional layer, a maximum pond q (q >=0, positive integer) in network structure
A average pond layer of layer, a convolutional layer of k (k >=0, positive integer), n (n >=0, positive integer) and output and input layer, this example
In, p=2, q=2, k=1, n=1.As shown in figure 3, it includes 8 layers that multiple dimensioned convolutional neural networks model, which has altogether,.In network structure
Comprising 2 multiple dimensioned convolutional layers, 2 maximum pond layers, 1 convolutional layer, 1 average pond layer and output and input.
(1) the 1st layer: input layer inputs as pretreated 32*32 size indicator card image data.
(2) the 2nd layers: multiple dimensioned convolutional layer carries out feature extraction to input feature vector figure with multiple dimensioned convolution block, exports feature
Figure remains as 32*32 size.Swish activation primitive is used after each convolution operation.
Wherein: Swish activation primitive are as follows:
This example β value is 1.
(3) the 3rd layers: pond layer reduces calculation amount, portrays translation invariance.This example uses the maximum pond of 2*2, defeated
Characteristic pattern is 16*16 size out.
(4) the 4th layers: multiple dimensioned convolutional layer, parameter setting is the same as (2).
(5) the 5th layers: pond layer, using the maximum pond of 2*2, output characteristic pattern is 8*8 size.
(6) the 6th layers: convolutional layer, convolution kernel is having a size of 5*5, and step-length is set as 1, and output characteristic pattern is 8*8 size, convolution
Non-linearization variation is carried out using Swish activation primitive after operation.
(7) the 7th layers: pond layer, using the average pond of 8*8, output characteristic pattern is 1*1 size.
(8) the 8th layers: upper one layer output is converted into 1D vector by output layer, as the input of softmax classifier, output
Classification results.
Step 5: the multiple dimensioned convolutional neural networks that step 4 is built are trained:
Network training is carried out using the training set that step 2 obtains.It is lost using the output that intersection entropy function carrys out computation model,
Using Adam optimization method, loss function is minimized, network parameter training is carried out, saves optimum model parameter.
Wherein, intersect entropy function are as follows:
In formula, y is the probability distribution of indicator card prediction label, and y ' is the probability distribution of indicator card true tag, with intersection
Entropy function carrys out judgment models to the order of accuarcy of true probability distribution estimation.
Step 6: test set in step 2 being tested using the network model that step 5 training obtains.Fig. 4 is model instruction
Practice accuracy and test accuracy result figure.In training to 400 steps or so, training set and test set tend towards stability model,
Training set accuracy is 100%, and test set accuracy is 98% or so.
Claims (2)
1. a kind of working conditions of oil extractor diagnostic method based on multiple dimensioned convolutional neural networks, it is characterised in that the following steps are included:
Step 1: collecting suspension point displacement and load data, initial data is pre-processed, according to network model to input data
Indicator card is drawn in the requirement of size;
Step 2: for indicator card data creating label, indicating the label of x-1 class indicator card with the positive integer of 0~x, and by its point
At training set and test set;
Step 3: constructing multiple dimensioned convolution block, extract the global and local information of indicator card, multiple dimensioned convolution block structure is by (2m+
3) * (2m+3), (2m+5) * (2m+5), three kinds of (2m+7) * (2m+7) different size convolution kernels are constituted, and m is the positive integer greater than 0,
Three kinds of various sizes of convolution kernels extract the feature of input feature vector figure by convolution operation, and feature extraction result is carried out splicing work
For the output feature of multiple dimensioned convolutional layer;
Step 4: building multiple dimensioned convolutional neural networks model by core of multiple dimensioned convolution block, include p more in network structure
Scale convolutional layer, q maximum pond layer, k convolutional layer, n be a to be averaged and pond layers and outputs and inputs layer, wherein p, q, k, m
It is the positive integer more than or equal to 0, carries out nonlinear processing using Swish activation primitive after each convolution operation;
Wherein, Swish activation primitive are as follows:
β > 0 in formula;
Step 5: carrying out multiple dimensioned convolutional neural networks model training on the training set obtained in step 2, use cross entropy letter
Loss function of the number as model minimizes loss function using optimizer, carries out network parameter training, saves optimum network mould
Shape parameter;
Wherein, cross entropy loss function are as follows:
In formula, y is the probability distribution of prediction label, and y ' is the probability distribution of true tag, with intersection entropy function come judgment models
To the order of accuarcy of true probability distribution estimation;
Step 6: testing characteristics of network on test set in step 2 saves network model if test effect reaches expected,
Otherwise return step 5 re-start network parameter adjustment.
2. in claim 1, the structure of the building of the multiple dimensioned convolution block of step 3, multiple dimensioned convolution block can be by n kind difference size
Convolution kernel is constituted, and n is the positive integer greater than 0.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910419291.0A CN110132626A (en) | 2019-05-20 | 2019-05-20 | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910419291.0A CN110132626A (en) | 2019-05-20 | 2019-05-20 | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110132626A true CN110132626A (en) | 2019-08-16 |
Family
ID=67571723
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910419291.0A Pending CN110132626A (en) | 2019-05-20 | 2019-05-20 | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110132626A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110595775A (en) * | 2019-09-19 | 2019-12-20 | 电子科技大学 | Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network |
CN111144548A (en) * | 2019-12-23 | 2020-05-12 | 北京寄云鼎城科技有限公司 | Method and device for identifying working condition of pumping well |
CN111444871A (en) * | 2020-04-01 | 2020-07-24 | 北京信息科技大学 | Fault diagnosis method for multi-scale deep convolution neural network planetary gearbox |
CN111627253A (en) * | 2020-06-12 | 2020-09-04 | 浙江驿公里智能科技有限公司 | Anti-collision system and method for actively guiding parking by using camera |
CN111810124A (en) * | 2020-06-24 | 2020-10-23 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic re-calibration residual convolution neural network model |
CN113095414A (en) * | 2021-04-15 | 2021-07-09 | 中国石油大学(华东) | Indicator diagram identification method based on convolutional neural network and support vector machine |
CN113567159A (en) * | 2021-06-17 | 2021-10-29 | 太原理工大学 | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation |
CN113865859A (en) * | 2021-08-25 | 2021-12-31 | 西北工业大学 | Multi-scale multi-source heterogeneous information fusion gearbox state fault diagnosis method |
CN114898126A (en) * | 2022-01-24 | 2022-08-12 | 南京富岛油气智控科技有限公司 | Method for improving fault diagnosis precision of composite working condition of oil production machine |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107631867A (en) * | 2017-09-07 | 2018-01-26 | 天津工业大学 | A kind of rotating machinery fault intelligent method for classifying based on deep learning |
CN108122075A (en) * | 2017-12-22 | 2018-06-05 | 胜利油田鲁明油气勘探开发有限公司 | A kind of sucker rod pump operating mode intelligent monitoring method based on convolutional neural networks intensified learning |
CN109029989A (en) * | 2018-09-28 | 2018-12-18 | 西北工业大学 | A kind of Fault Diagnosis of Gear Case method based on infrared thermal imaging |
CN109057776A (en) * | 2018-07-03 | 2018-12-21 | 东北大学 | A kind of oil well fault diagnostic method based on improvement fish-swarm algorithm |
US20190122113A1 (en) * | 2017-10-19 | 2019-04-25 | International Business Machines Corporation | Pruning Redundant Neurons and Kernels of Deep Convolutional Neural Networks |
CN109765053A (en) * | 2019-01-22 | 2019-05-17 | 中国人民解放军海军工程大学 | Utilize the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index |
-
2019
- 2019-05-20 CN CN201910419291.0A patent/CN110132626A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107631867A (en) * | 2017-09-07 | 2018-01-26 | 天津工业大学 | A kind of rotating machinery fault intelligent method for classifying based on deep learning |
US20190122113A1 (en) * | 2017-10-19 | 2019-04-25 | International Business Machines Corporation | Pruning Redundant Neurons and Kernels of Deep Convolutional Neural Networks |
CN108122075A (en) * | 2017-12-22 | 2018-06-05 | 胜利油田鲁明油气勘探开发有限公司 | A kind of sucker rod pump operating mode intelligent monitoring method based on convolutional neural networks intensified learning |
CN109057776A (en) * | 2018-07-03 | 2018-12-21 | 东北大学 | A kind of oil well fault diagnostic method based on improvement fish-swarm algorithm |
CN109029989A (en) * | 2018-09-28 | 2018-12-18 | 西北工业大学 | A kind of Fault Diagnosis of Gear Case method based on infrared thermal imaging |
CN109765053A (en) * | 2019-01-22 | 2019-05-17 | 中国人民解放军海军工程大学 | Utilize the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index |
Non-Patent Citations (4)
Title |
---|
仲志丹等: ""石油生产中有杆抽油机故障诊断研究"", 《计算机仿真》 * |
宋青松等: ""基于多尺度卷积神经网络的交通标志识别"", 《湖南大学学报(自然科学版 )》 * |
张文达等: ""基于多尺度分块卷积神经网络的图像目标识别算法"", 《计算机应用》 * |
罗仁泽等: ""基于功图分析的有杆抽油泵故障诊断技术研究"", 《计算机测量与控制》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110595775B (en) * | 2019-09-19 | 2021-03-30 | 电子科技大学 | Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network |
CN110595775A (en) * | 2019-09-19 | 2019-12-20 | 电子科技大学 | Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network |
CN111144548B (en) * | 2019-12-23 | 2023-09-01 | 北京寄云鼎城科技有限公司 | Method and device for identifying working condition of oil pumping well |
CN111144548A (en) * | 2019-12-23 | 2020-05-12 | 北京寄云鼎城科技有限公司 | Method and device for identifying working condition of pumping well |
CN111444871A (en) * | 2020-04-01 | 2020-07-24 | 北京信息科技大学 | Fault diagnosis method for multi-scale deep convolution neural network planetary gearbox |
CN111627253A (en) * | 2020-06-12 | 2020-09-04 | 浙江驿公里智能科技有限公司 | Anti-collision system and method for actively guiding parking by using camera |
CN111627253B (en) * | 2020-06-12 | 2021-05-04 | 浙江驿公里智能科技有限公司 | Anti-collision system and method for actively guiding parking by using camera |
WO2021248833A1 (en) * | 2020-06-12 | 2021-12-16 | 浙江驿公里智能科技有限公司 | Anti-collision system and method for actively guiding vehicle stopping using camera |
CN111810124A (en) * | 2020-06-24 | 2020-10-23 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic re-calibration residual convolution neural network model |
CN111810124B (en) * | 2020-06-24 | 2023-09-22 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic recalibration residual convolutional neural network model |
CN113095414A (en) * | 2021-04-15 | 2021-07-09 | 中国石油大学(华东) | Indicator diagram identification method based on convolutional neural network and support vector machine |
CN113567159A (en) * | 2021-06-17 | 2021-10-29 | 太原理工大学 | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation |
CN113567159B (en) * | 2021-06-17 | 2023-06-23 | 太原理工大学 | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation |
CN113865859A (en) * | 2021-08-25 | 2021-12-31 | 西北工业大学 | Multi-scale multi-source heterogeneous information fusion gearbox state fault diagnosis method |
CN113865859B (en) * | 2021-08-25 | 2024-05-14 | 西北工业大学 | Gear box state fault diagnosis method for multi-scale multi-source heterogeneous information fusion |
CN114898126B (en) * | 2022-01-24 | 2023-09-01 | 南京富岛油气智控科技有限公司 | Method for improving fault diagnosis precision of oil extraction machine under composite working condition |
CN114898126A (en) * | 2022-01-24 | 2022-08-12 | 南京富岛油气智控科技有限公司 | Method for improving fault diagnosis precision of composite working condition of oil production machine |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110132626A (en) | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks | |
CN106555788B (en) | Application based on the deep learning of Fuzzy Processing in hydraulic equipment fault diagnosis | |
WO2022077901A1 (en) | Bearing failure mode diagnosis method using small sample data sets, and system | |
CN110070091B (en) | Semantic segmentation method and system based on dynamic interpolation reconstruction and used for street view understanding | |
CN108830285A (en) | A kind of object detection method of the reinforcement study based on Faster-RCNN | |
CN110533086A (en) | The semi-automatic mask method of image data | |
CN104751472A (en) | Fabric defect detection method based on B-spline wavelets and deep neural network | |
CN107633272A (en) | A kind of DCNN textural defect recognition methods based on compressed sensing under small sample | |
CN110163069A (en) | Method for detecting lane lines for assisting driving | |
CN107423747A (en) | A kind of conspicuousness object detection method based on depth convolutional network | |
CN112766283B (en) | Two-phase flow pattern identification method based on multi-scale convolution network | |
CN110288257A (en) | A kind of depth transfinites indicator card learning method | |
CN109389170B (en) | Gradual change type working condition early warning method based on 3D convolutional neural network | |
CN112308850A (en) | Multi-scale feature fusion power transmission line detection method and system | |
CN112560948A (en) | Eye fundus map classification method and imaging method under data deviation | |
CN116416237A (en) | Power transmission line defect detection method based on improved YOLOv5 and fuzzy image enhancement | |
CN117422695A (en) | CR-deep-based anomaly detection method | |
CN115908793A (en) | Coding and decoding structure semantic segmentation model based on position attention mechanism | |
CN113361496A (en) | City built-up area statistical method based on U-Net | |
CN117576038A (en) | Fabric flaw detection method and system based on YOLOv8 network | |
CN117435992A (en) | Fault prediction method and system for hydraulic propulsion system of shield tunneling machine | |
CN116721291A (en) | Metal surface defect detection method based on improved YOLOv7 model | |
CN115240057A (en) | Overhead transmission line monitoring image detection method based on deep learning | |
CN115273009A (en) | Road crack detection method and system based on deep learning | |
CN114494236A (en) | Fabric defect detection method and system based on over-complete convolutional neural network |
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 |
Application publication date: 20190816 |
|
RJ01 | Rejection of invention patent application after publication |