CN112307673A - Grid surface quality detection method based on deep learning - Google Patents
Grid surface quality detection method based on deep learning Download PDFInfo
- Publication number
- CN112307673A CN112307673A CN202011205869.1A CN202011205869A CN112307673A CN 112307673 A CN112307673 A CN 112307673A CN 202011205869 A CN202011205869 A CN 202011205869A CN 112307673 A CN112307673 A CN 112307673A
- Authority
- CN
- China
- Prior art keywords
- grid
- grid surface
- quality
- data set
- quality detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Fluid Mechanics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Mathematical Optimization (AREA)
- Algebra (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a grid surface quality detection method based on deep learning, which comprises the following steps: s1, constructing a grid surface data set, and dividing the established grid surface data set into a training data set and a testing data set; s2, marking training data and marking the quality of the grid surface; s3, constructing a grid surface quality detection network and detecting to obtain a trained grid surface quality detection network E'; s4, judging the quality of the grid surface, and outputting the quality judgment of the grid surface; in the process of detecting the quality of the grid surface, a deep learning algorithm is introduced, multi-level feature extraction can be automatically, efficiently and accurately carried out, the defect that the grid quality is judged by relying on complex manual operation in the prior art is effectively overcome, the quality of the grid can be predicted according to the given grid surface discrete coordinate value in the detection process, the judgment is carried out without relying on the complex manual operation, the prediction precision of 97% can be achieved through experiments, and the method has the characteristics of high automation degree and high detection precision.
Description
Technical Field
The invention relates to the technical field of analog simulation data processing, in particular to a grid surface quality detection method based on deep learning.
Background
In the process of processing simulation of analog data, the accuracy of the analog simulation data is determined by the CFD grid quality, so that the grid quality detection process is very important in the process of processing simulation of analog hardware, and the accuracy of the later-stage calculation work is indirectly determined;
the conventional grid quality distinguishing and detecting process can not get rid of manual participation all the time, and automatic distinguishing of the grid quality can not be realized at the present stage, so that the development of the automatic generation technology of the high-quality CFD grid at the next stage is restricted; the existing software basically depends on complex manual operation to solve various technical problems in the grid generation, so that the smooth implementation of the related CFD engineering is guaranteed, but the heavy operation process does not effectively reduce the working intensity of the grid generation and the time overhead of the grid generation; in addition, the method has the more negative influence that the method excessively depends on a manually operated grid generation method, so that the technical development of the automatic grid processing flow needed by the design optimization of the appearance of the aircraft based on computational fluid mechanics and the like is seriously hindered;
in a word, after years of development with engineering software as the leading factor, the grid generation technology reaches a technical watershed facing new problems, needing to introduce new technology and making new breakthroughs, and the grid quality detection technology is particularly important at this stage, is a process for judging the quality of the grid and indirectly determines the accuracy of later-stage work; on the other hand, in the automatic grid generation technology, the improvement of grid generation quality needs to be continuously iterated and converged, and finally, a grid meeting the CFD calculation requirement is obtained;
at present, automatic judgment of grid quality is not realized, and a CFD grid quality automatic judgment technology based on intelligent identification is still not researched by personnel or groups at home and abroad; the lack of the grid quality automatic discrimination technology will restrict the development of the CFD grid automatic generation technology with high quality and high efficiency at the next stage; after the automatic grid quality judging technology is combined, the automatic grid generation technology and the quality judging iteration are carried out, and positive promotion effect is brought to the development of automatic high-quality grid generation.
Disclosure of Invention
Aiming at the existing problems, the invention aims to provide a grid surface quality detection method based on deep learning, which can realize automatic, efficient and accurate multi-level feature extraction by introducing a deep learning algorithm in the grid surface quality detection process, effectively overcomes the defect that the grid quality is judged by relying on complex manual operation in the prior art, and has the characteristics of high automation degree and high detection precision.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a grid surface quality detection method based on deep learning comprises the steps of
S1, making a grid surface by using grid generation software NNW-GridStar, constructing a grid surface data set, and dividing the established grid surface data set into a training data set and a test data set;
s2, marking training data: performing quality detection on the grid surfaces in the grid surface training data set obtained in the step S1 by using a grid surface checking function of grid generation software NNW-GridStar, and marking the quality of the grid surfaces;
s3, constructing a grid surface quality detection network: constructing a neural network E for grid surface quality detection based on the VGG neural network, and training the neural network by using the training data set obtained in the step S2 to obtain a trained grid surface quality detection network E';
s4, judging the quality of the grid surface: and (4) testing the grid plane quality detection network E' trained in the step (S3) by using the grid plane data in the test training set obtained in the step (S1), and outputting quality judgment of the grid plane.
Preferably, the process of constructing the grid plane data set in step S1 includes:
s101, importing a digital model and corresponding three-dimensional grid data into grid generation software NNW-GridStar;
s102, extracting two-dimensional grid data from the three-dimensional grid data by utilizing NNW-GridStar, and adding the extracted two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s103, optimizing the two-dimensional grid by using a surface optimization function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s104, projecting the digital model imported in the step S101 by the two-dimensional grid by using a surface projection function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s105, repeating the steps S101 to S104, importing different digital models and corresponding three-dimensional grid data, generating a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in the form of grid surface discrete coordinate values until the number of the data sets is larger than or equal to 5000;
and S106, dividing the obtained grid surface data set into a training data set and a testing data set, wherein 80% of the training data set and 20% of the testing data set are used as the training data set.
Preferably, the labeling process of the training data in step S2 includes:
s201, importing one grid surface in a grid surface training data set into grid generation software NNW-GridStar;
s202, detecting the quality of the grid surface by using a grid surface quality detection function of grid generation software NNW-GridStar;
s203, judging whether the quality of the grid surface is good or not through grid surface quality index information fed back by grid generation software NNW-GridStar;
s204, if the quality of the grid surface is judged to be good through the judging process of the step S203, marking the quality index of the grid surface to be 1; if it is determined in step S203 that the quality of the mesh surface is not good, the quality index of the mesh surface is marked as 0, and label data of the quality of the mesh surface is obtained.
Preferably, the process of constructing the mesh plane quality detection network in step S3 includes:
s301, constructing a neural network E for grid surface quality detection by using a VGG neural network;
s302, randomly selecting grid surface data from the training data set obtained in the step S2, and training the neural network E constructed in the step S301 100000 times by using the selected grid surface data and the corresponding label data of the grid surface quality to obtain a trained grid surface quality detection network E'.
Preferably, the design process of constructing the neural network E for grid plane quality detection by using the VGG neural network in step S301 includes:
(1) constructing a convolutional layer model based on a VGG neural network, wherein the convolutional layer is a core layer for constructing the convolutional neural network and generates most of calculated amount in the network, and the form of the convolutional layer model is as follows:
wherein:
represents the nth characteristic diagram of the l layer,
represents the mth characteristic diagram of the l-1 st layer,
representing the convolution kernel acting between two feature maps, f is the activation function,
represents a bias term;
(2) designing a pooling layer, wherein the model of the pooling layer is as follows:
wherein:
represents the nth characteristic diagram of the l layer,
representing the nth feature map of the l-1 layer, s is a selected downsampling template,
the weight of the template is used as the weight of the template,
represents a bias term;
(3) periodically inserting the pooled layer obtained in step (2) into the convolutional layer model obtained in step (1).
Preferably, the process of determining the quality of the grid plane in step S4 includes:
s401, randomly selecting grid surface data from the test data set obtained in the step S1;
and S402, inputting the grid surface data obtained in the step S401 into the grid surface quality detection network E' trained in the step S3, and outputting the judgment of the quality of the grid surface, wherein the output 1 shows that the quality of the grid surface is good, and the output 0 shows that the quality of the grid surface is not good.
The invention has the beneficial effects that: the invention discloses a grid surface quality detection method based on deep learning, compared with the prior art, the improvement of the invention is as follows:
the invention designs a grid surface quality detection method based on deep learning, and the method can realize automatic, efficient and accurate multi-level feature extraction by introducing a deep learning algorithm in the grid surface quality detection process, effectively overcomes the defect that the grid quality is judged by relying on complex manual operation in the past, can predict the quality of a grid according to given grid surface discrete coordinate values in the detection process, does not depend on complex manual operation for judgment, and is proved by experiments that the method can reach 97% of prediction accuracy and has the advantages of high automation degree and detection accuracy;
meanwhile, the method fills the blank of the automatic grid surface quality judging technology, and can intelligently detect the quality of the grid surface based on the deep neural network.
Drawings
Fig. 1 is a flowchart of a grid surface quality detection method based on deep learning according to the present invention.
FIG. 2 is an algorithm process diagram of the grid surface quality detection method based on deep learning.
FIG. 3 is a diagram of a different quality mesh plane dataset according to the present invention.
FIG. 4 is a diagram of the VGG neural network of the present invention.
Wherein: in fig. 3: fig. (a) and (d) are good quality grid maps, and fig. (b), fig. (c), fig. (e), and fig. (f) are poor quality grid maps.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the following further describes the technical solution of the present invention with reference to the drawings and the embodiments.
In the CFD numerical simulation, the quality of grid quality directly influences the calculation precision and the calculation efficiency, and how to check the grid quality and evaluate the grid quality is crucial; the grid quality of a set of computational grids needs to be judged, and the three dimensions of geometric characteristics, flow characteristics, solver adaptability and the like are mainly involved. In the specific research, research is respectively carried out from the three dimensions, and a comprehensive evaluation method for the quality of the computational grid is established;
referring to fig. 1-4, a grid surface quality detection method based on deep learning can realize automatic, efficient and accurate grid surface quality detection, and includes the steps of:
s1, making a grid surface by using grid generation software NNW-GridStar, constructing a grid surface data set, and dividing the established grid surface data set into a training data set and a testing data set, wherein the method specifically comprises the following steps:
s101, importing a digital-to-analog (CAD digital-to-analog) and corresponding three-dimensional grid data into grid generation software NNW-GridStar;
s102, extracting two-dimensional grid data from the three-dimensional grid data by utilizing a grid derivation function of NNW-GridStar, and adding the extracted two-dimensional grid into a grid surface data set (initially, an empty set) in a grid surface discrete coordinate value form;
s103, optimizing the two-dimensional grid by using a surface optimization function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s104, projecting the digital model imported in the step S101 by the two-dimensional grid by using a surface projection function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s105, repeating the steps S101 to S104, importing different digital models and corresponding three-dimensional grid data, generating a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in the form of grid surface discrete coordinate values until the number of the data sets is larger than or equal to 5000;
s106, dividing the obtained grid surface data set into a training data set and a testing data set, wherein 80% of the training data set is the training data set, and 20% of the testing data set is the testing data set;
s2, marking training data: and (3) performing quality detection on the grid surfaces in the grid surface training data set obtained in the step S1 by using a grid surface checking function of grid generation software NNW-GridStar, and marking the quality of the grid surfaces, wherein the method specifically comprises the following steps:
s201, importing one grid surface in a grid surface training data set into grid generation software NNW-GridStar;
s202, detecting the quality of the grid surface by using a grid surface quality detection function of grid generation software NNW-GridStar;
s203, manually judging whether the quality of the grid surface is good or bad according to judgment indexes of different examples through grid surface quality index information fed back by grid generation software NNW-GridStar, such as orthogonality and smoothness of the grid;
s204, if the quality of the grid surface is judged to be good through the judging process of the step S203, marking the quality index of the grid surface to be 1; if the quality of the grid surface is judged to be not good through the step S203, marking the quality index of the grid surface as 0 to obtain label data of the quality of the grid surface;
s3, constructing a grid surface quality detection network: constructing a neural network E for grid plane quality detection based on the VGG neural network, and training the neural network by using the training data set obtained in the step S2 to obtain a trained grid plane quality detection network E', wherein the specific steps comprise:
s301, constructing a neural network E for grid surface quality detection by using the VGG neural network, wherein the specific process comprises the following steps:
(1) constructing a convolutional layer model based on a VGG neural network, wherein the convolutional layer is a core layer for constructing the convolutional neural network and generates most of calculated amount in the network, and the form of the convolutional layer model is as follows:
wherein:
represents the nth characteristic diagram of the l layer,
represents the mth characteristic diagram of the l-1 st layer,
representing the convolution kernel acting between two feature maps, f is the activation function,
represents a bias term;
(2) designing a pooling layer, wherein the model of the pooling layer is as follows:
wherein:
represents the nth characteristic diagram of the l layer,
representing the nth feature map of the l-1 layer, s is a selected downsampling template,
the weight of the template is used as the weight of the template,
represents a bias term;
(3) periodically inserting the pooling layer obtained in the step (2) into the convolutional layer model obtained in the step (1) to obtain a neural network E for grid surface quality detection;
s302, randomly selecting grid surface data from the training data set obtained in the step S2, and training the neural network E constructed in the step S301 100000 times by using the selected grid surface data and the corresponding label data of the grid surface quality to obtain a trained grid surface quality detection network E'.
S4, judging the quality of the grid surface: and (3) testing the grid surface quality detection network E' trained in the step (S3) by using the grid surface data in the test training set obtained in the step (S1), and outputting quality judgment of the grid surface, wherein the method specifically comprises the following steps:
s401, randomly selecting grid surface data in the test data set obtained in the step S1 by using a random function of python;
and S402, inputting the grid surface data obtained in the step S401 into the grid surface quality detection network E' trained in the step S3, and outputting the judgment of the quality of the grid surface, wherein the output 1 shows that the quality of the grid surface is good, and the output 0 shows that the quality of the grid surface is not good.
Example 1: s5, experiment and comparison
S501.CNN model training
In order to verify the grid surface quality detection method based on deep learning, the total number of grid surface sample sets generated in the method is 5024, 21X21, 41X41, 81X81 and 21X121 are used as training samples, and data of each scale are disordered sequentially; 71x71 as a test sample, and testing the prediction capability of the CNN;
the training method selects a random gradient descent algorithm, the batch size of image input is set to be 8 or 16, namely 8 or 16 grid surface samples with uniform scale are input in each batch for training; the number of iterations is set to 100000 times; the learning rate was set to 0.0001; the initialization weight of the convolution kernel is set to be a Gaussian distribution random number with the mean value of 0 and the standard deviation of 0.01, and the bias is initialized to be 0;
the computer used herein for modeling and simulation is configured to: intel (R) Xeon (R) E5-2698 v42.20GHz CPU, NVIDIATeslaV100 video card;
as shown in table 2, different training strategies have a great influence on accuracy, only one 121 × 121 scale grid is trained, only data of the scale has generalization capability, and data of other scales has the same effect as random prediction; training is carried out by adopting data of multiple scales of 21x21, 41x41, 81x81 and 121x121, and great difference exists among different strategies; all data of each scale are used for training an epoch, all data of the next scale are trained for training an epoch, and the test effect is very poor; the data of all scales have a generalization mode, one scale selects one batch training, and then the data of other scales selects one batch training;
table 2: comparison of results of different training modes
As can be seen from table 2, the testing accuracy of the grid surface quality detection method based on deep learning of the present invention is 97.13%, which is much greater than that of other grid surface quality detection methods;
table 3 shows that the batch size is 8 and 16 different sizes of the batch training results, and the results show that the impact of the size of the batch size on the test results is not great;
table 3: comparison of results for different BatchSize
S502.CNN prediction
Table 4 shows that different depth neural networks use the same training parameters and training strategies, and the result shows that the test effect of VGG16 is the best;
table 4: comparison of different network test results
| Network model | Training data | Test accuracy |
| Resnet50 | 21x21～121x121 | 0.7808 |
| 3Conv+2Dense | 21x21～121x121 | 0.9743 |
| VGG16 | 41x41～121x121 | 0.9863 |
And (4) conclusion: according to the verification process, the grid surface quality detection method based on deep learning does not depend on complex manual operation for judgment, can predict the quality and the defect of a grid in a given grid surface discrete coordinate value, and has high prediction precision;
meanwhile, through the research in the text, the deep learning has a good application prospect in the aspect of grid quality detection, and the fitting capability of the network can be increased by increasing the number of network layers, the number of convolution kernels, the number of iteration times, the number of full-connection layers, the Dropout technology and other measures, so that the method is suitable for grid quality detection of different scales.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (6)
1. A grid surface quality detection method based on deep learning is characterized in that: comprises the steps of
S1, making a grid surface by using grid generation software NNW-GridStar, constructing a grid surface data set, and dividing the established grid surface data set into a training data set and a test data set;
s2, marking training data: performing quality detection on the grid surfaces in the grid surface training data set obtained in the step S1 by using a grid surface checking function of grid generation software NNW-GridStar, and marking the quality of the grid surfaces;
s3, constructing a grid surface quality detection network: constructing a neural network E for grid surface quality detection based on the VGG neural network, and training the neural network by using the training data set obtained in the step S2 to obtain a trained grid surface quality detection network E';
s4, judging the quality of the grid surface: and (4) testing the grid plane quality detection network E' trained in the step (S3) by using the grid plane data in the test training set obtained in the step (S1), and outputting quality judgment of the grid plane.
2. The grid surface quality detection method based on deep learning of claim 1, wherein: the process of constructing the mesh plane data set in step S1 includes:
s101, importing a digital model and corresponding three-dimensional grid data into grid generation software NNW-GridStar;
s102, extracting two-dimensional grid data from the three-dimensional grid data by utilizing NNW-GridStar, and adding the extracted two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s103, optimizing the two-dimensional grid by using a surface optimization function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s104, projecting the digital model imported in the step S101 by the two-dimensional grid by using a surface projection function of grid generation software NNW-GridStar to obtain a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in a grid surface discrete coordinate value form;
s105, repeating the steps S101 to S104, importing different digital models and corresponding three-dimensional grid data, generating a new two-dimensional grid, and adding the new two-dimensional grid into a grid surface data set in the form of grid surface discrete coordinate values until the number of the data sets is larger than or equal to 5000;
and S106, dividing the obtained grid surface data set into a training data set and a testing data set, wherein 80% of the training data set and 20% of the testing data set are used as the training data set.
3. The grid surface quality detection method based on deep learning of claim 1, wherein: the labeling process of the training data in step S2 includes:
s201, importing one grid surface in a grid surface training data set into grid generation software NNW-GridStar;
s202, detecting the quality of the grid surface by using a grid surface quality detection function of grid generation software NNW-GridStar;
s203, judging whether the quality of the grid surface is good or not through grid surface quality index information fed back by grid generation software NNW-GridStar;
s204, if the quality of the grid surface is judged to be good through the judging process of the step S203, marking the quality index of the grid surface to be 1; if it is determined in step S203 that the quality of the mesh surface is not good, the quality index of the mesh surface is marked as 0, and label data of the quality of the mesh surface is obtained.
4. The grid surface quality detection method based on deep learning of claim 1, wherein: the process of constructing the mesh plane quality detection network described in step S3 includes:
s301, constructing a neural network E for grid surface quality detection by using a VGG neural network;
s302, randomly selecting grid surface data from the training data set obtained in the step S2, and training the neural network E constructed in the step S301 100000 times by using the selected grid surface data and the corresponding label data of the grid surface quality to obtain a trained grid surface quality detection network E'.
5. The grid surface quality detection method based on deep learning of claim 4, wherein: the design process of constructing the neural network E for grid plane quality detection by using the VGG neural network in step S301 includes:
(1) constructing a convolutional layer model based on a VGG neural network, wherein in the network model, a convolutional layer is a core layer for constructing the convolutional neural network and has the following form:
wherein:
represents the nth characteristic diagram of the l layer,
represents the mth characteristic diagram of the l-1 st layer,
representing the convolution kernel acting between two feature maps, f is the activation function,
represents a bias term;
(2) designing a pooling layer, wherein the model of the pooling layer is as follows:
wherein:
represents the nth characteristic diagram of the l layer,
representing the nth feature map of the l-1 layer, s is a selected downsampling template,
the weight of the template is used as the weight of the template,
represents a bias term;
(3) periodically inserting the pooled layer obtained in step (2) into the convolutional layer model obtained in step (1).
6. The grid surface quality detection method based on deep learning of claim 1, wherein: the process of determining the quality of the mesh surface in step S4 includes:
s401, randomly selecting grid surface data from the test data set obtained in the step S1;
and S402, inputting the grid surface data obtained in the step S401 into the grid surface quality detection network E' trained in the step S3, and outputting the judgment of the quality of the grid surface, wherein the output 1 shows that the quality of the grid surface is good, and the output 0 shows that the quality of the grid surface is not good.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205869.1A CN112307673A (en) | 2020-11-02 | 2020-11-02 | Grid surface quality detection method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205869.1A CN112307673A (en) | 2020-11-02 | 2020-11-02 | Grid surface quality detection method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112307673A true CN112307673A (en) | 2021-02-02 |
Family
ID=74333892
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011205869.1A Pending CN112307673A (en) | 2020-11-02 | 2020-11-02 | Grid surface quality detection method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112307673A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115618694A (en) * | 2022-12-15 | 2023-01-17 | 博志生物科技(深圳)有限公司 | Image-based cervical vertebra analysis method, device, equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108009222A (en) * | 2017-11-23 | 2018-05-08 | 浙江工业大学 | Method for searching three-dimension model based on more excellent view and depth convolutional neural networks |
| CN108345856A (en) * | 2018-02-09 | 2018-07-31 | 电子科技大学 | The SAR automatic target recognition methods integrated based on isomery convolutional neural networks |
| CN109064459A (en) * | 2018-07-27 | 2018-12-21 | 江苏理工学院 | A kind of Fabric Defect detection method based on deep learning |
| CN109934216A (en) * | 2017-12-19 | 2019-06-25 | 华为技术有限公司 | The method, apparatus of image procossing, computer readable storage medium |
| CN110334450A (en) * | 2019-07-09 | 2019-10-15 | 中国空气动力研究与发展中心计算空气动力研究所 | A kind of multi-blocked structure grid generate in object plane erroneous projection restorative procedure |
| CN110889111A (en) * | 2019-10-23 | 2020-03-17 | 广东工业大学 | Power grid virtual data injection attack detection method based on deep belief network |
| CN111783964A (en) * | 2020-08-11 | 2020-10-16 | 中国人民解放军国防科技大学 | Grid quality detection method facing GPU and neural network |
| WO2020211007A1 (en) * | 2019-04-17 | 2020-10-22 | 大连理工大学 | Form and position deviation feature library establishment method for aerospace thin-housing structure |
-
2020
- 2020-11-02 CN CN202011205869.1A patent/CN112307673A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108009222A (en) * | 2017-11-23 | 2018-05-08 | 浙江工业大学 | Method for searching three-dimension model based on more excellent view and depth convolutional neural networks |
| CN109934216A (en) * | 2017-12-19 | 2019-06-25 | 华为技术有限公司 | The method, apparatus of image procossing, computer readable storage medium |
| CN108345856A (en) * | 2018-02-09 | 2018-07-31 | 电子科技大学 | The SAR automatic target recognition methods integrated based on isomery convolutional neural networks |
| CN109064459A (en) * | 2018-07-27 | 2018-12-21 | 江苏理工学院 | A kind of Fabric Defect detection method based on deep learning |
| WO2020211007A1 (en) * | 2019-04-17 | 2020-10-22 | 大连理工大学 | Form and position deviation feature library establishment method for aerospace thin-housing structure |
| CN110334450A (en) * | 2019-07-09 | 2019-10-15 | 中国空气动力研究与发展中心计算空气动力研究所 | A kind of multi-blocked structure grid generate in object plane erroneous projection restorative procedure |
| CN110889111A (en) * | 2019-10-23 | 2020-03-17 | 广东工业大学 | Power grid virtual data injection attack detection method based on deep belief network |
| CN111783964A (en) * | 2020-08-11 | 2020-10-16 | 中国人民解放军国防科技大学 | Grid quality detection method facing GPU and neural network |
Non-Patent Citations (2)
| Title |
|---|
| 周娇媚: "分布式并行网格生成软件框架研制及其应用", 《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》 * |
| 黄尚坤: "基于卷积神经网络的网格质量判别技术研究", 《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115618694A (en) * | 2022-12-15 | 2023-01-17 | 博志生物科技(深圳)有限公司 | Image-based cervical vertebra analysis method, device, equipment and storage medium |
| CN115618694B (en) * | 2022-12-15 | 2023-03-21 | 博志生物科技(深圳)有限公司 | Image-based cervical vertebra analysis method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109948165B (en) | Fine granularity emotion polarity prediction method based on mixed attention network | |
| CN109558487A (en) | Document Classification Method based on the more attention networks of hierarchy | |
| CN107590565A (en) | A kind of method and device for building building energy consumption forecast model | |
| CN107784380A (en) | The optimization method and optimization system of a kind of inspection shortest path | |
| CN111619755B (en) | Hull profile design method based on convolutional neural network | |
| CN103336869A (en) | Multi-objective optimization method based on Gaussian process simultaneous MIMO model | |
| CN111080356A (en) | Method for calculating residence price influence factors by using machine learning regression model | |
| CN101055631A (en) | Space data fuzzy evidence weight analysis method | |
| CN104573333A (en) | Method for optimizing of model selection based on clustering analysis | |
| CN112948932A (en) | Surrounding rock grade prediction method based on TSP forecast data and XGboost algorithm | |
| CN111785326B (en) | Gene expression profile prediction method after drug action based on generation of antagonism network | |
| CN111833310A (en) | Surface defect classification method based on neural network architecture search | |
| CN110942182A (en) | Method for establishing typhoon prediction model based on support vector regression | |
| CN110765788A (en) | Knowledge graph embedding method based on implicit translation model | |
| CN115392477A (en) | Skyline query cardinality estimation method and device based on deep learning | |
| CN112307673A (en) | Grid surface quality detection method based on deep learning | |
| CN110633868A (en) | Method for predicting properties of oil and gas in exploration well oil testing layer by optimizing neural network through genetic algorithm | |
| CN101241520A (en) | Model state creation method based on characteristic suppression in finite element modeling | |
| CN112862063A (en) | Complex pipe network leakage positioning method based on deep belief network | |
| Valente et al. | Parameter identification and shape optimization: An integrated methodology in metal forming and structural applications | |
| CN109214447A (en) | Model training method and device, disk life-span prediction method and device | |
| CN114862007A (en) | Short-period gas production rate prediction method and system for carbonate gas well | |
| CN108960347A (en) | A kind of recruitment evaluation system and method for convolutional neural networks handwriting recongnition Ranking Stability | |
| Yun et al. | [Retracted] Quality Evaluation and Satisfaction Analysis of Online Learning of College Students Based on Artificial Intelligence | |
| CN114266953B (en) | Target detection flow self-optimization method based on deep learning |
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: 20210202 |
|
| RJ01 | Rejection of invention patent application after publication |