Full convolutional neural network and corresponding microstructure identification method

A convolutional neural network and network technology, which is applied in the field of mesoscopic structure recognition of ceramic matrix composite material preforms, can solve the problems of large memory usage and long training time of full convolutional neural network, reduce time and improve accuracy. , the effect of enhancing the ability of expression

Active Publication Date: 2019-11-08
NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF11 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0012] Aiming at the deficiencies in the prior art, the present invention provides a fully convolutional neural network and a corresponding mesostructure recognition method for the mesostructure recognition of XCT slices of ceramic matrix composite material preforms, because it is applied in the field of CMC, so Called CMCs_Net, it solves the problem of large memory usage and long training time of the full convolutional neural network, and further improves the accuracy of the network

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Full convolutional neural network and corresponding microstructure identification method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] The present invention is described in further detail now in conjunction with accompanying drawing.

[0043] Such as figure 1 The CMCs_Net full convolutional neural network shown is composed of an encoder network, a decoder network and a classification layer. The convolution kernels in the full convolutional neural network are all 3x3, and the padding operation is performed. The padding size is 1 , to ensure that the size of the feature remains unchanged during the image convolution process.

[0044] The encoder network part consists of five encoders, which are sequentially recorded as encoder one, encoder two, encoder three, encoder four, and encoder five. Each encoder is composed of multiple convolutional layers and a layer of pooling layer in turn. The convolutional layer is used to extract the features of the picture; the pooling layer is used to reduce the resolution of the picture, and the pooling kernel is 2*2, that is The length and width are respectively reduc...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention relates to a full convolutional neural network and a corresponding microstructure identification method, by comprehensively adopting a maximum pooling index connection method and a channel connection method in the network, serial processing of the two methods is carried out, shallow low-level edge information of an encoder and deep semantic information of a decoder are fully utilized, and a better effect is achieved. In the process of amplifying the image resolution, only an up-sampling mode is adopted, so that the consumption of a video memory is reduced, and the time of networktraining is also reduced. Batch regularization is adopted through cross symmetry, the original distribution state of the feature map after convolution is ensured as much as possible, the expression ability of the convolutional neural network is further enhanced, semantic segmentation of CMCs preform XCT slices is achieved, and the accuracy of microstructure recognition is improved. The same fullconvolutional neural network is adopted to carry out microstructure identification on braided structures of different ceramic matrix composite material complex preforms, different full convolutional neural network weight files adopted by different braided structures are abandoned, and identification operation is simplified.

Description

technical field [0001] The invention belongs to the field of mesoscopic structure recognition of prefabricated ceramic matrix composite materials, and in particular relates to a fully convolutional neural network for semantic segmentation of XCT slices of woven ceramic matrix composite materials. Background technique [0002] Ceramic matrix composites (CMCs, Ceramic Material Composites) usually use continuous ceramic fibers as the reinforcement phase, which is a material that can be applied under extreme conditions, and has the advantages of high temperature resistance, corrosion resistance, low density, and wear resistance. [0003] The skeleton composed of continuous ceramic fibers is called a complex preform, and its weaving structure includes plain weave, 2.5D, three-dimensional four-way, etc. The preparation process of CMCs includes chemical vapor deposition (CVI), precursor impregnation cracking (PIP), reaction melting (MI) and other methods. Usually, CMCs are prepare...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06N3/04
CPCG06N3/045G06F18/2415
Inventor 宋迎东贾蕴发高希光张盛于国强韩笑谢楚阳董洪年
Owner NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products