유리섬유 복합재 내부 미세결함의 두께 및 위치 분석을 위한 합성곱 신경망 기반의 테라헤르츠 신호 학습 기법Training Method of Terahertz Signal based on Convolutional Neural Network to Predict thickness and depth of Micro-Delamination inside Glass Fiber Reinforced Polymer
- Other Titles
- Training Method of Terahertz Signal based on Convolutional Neural Network to Predict thickness and depth of Micro-Delamination inside Glass Fiber Reinforced Polymer
- Authors
- 김헌수; 박동운; 김상일; 김학성
- Issue Date
- Jun-2021
- Publisher
- 대한기계학회
- Citation
- 대한기계학회 재료 및 파괴부문 2021년도 춘계학술대회 논문집, pp.120 - 120
- Indexed
- OTHER
- Journal Title
- 대한기계학회 재료 및 파괴부문 2021년도 춘계학술대회 논문집
- Start Page
- 120
- End Page
- 120
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/191347
- Abstract
- By training the terahertz (THz) signal through the convolutional neural network (CNN), a micro-delamination inside the glass fiber reinforced polymer (GFRP) was detected. First, GFRP specimens with micro-delamination of 25, 50, 75, and 110 um were prepared, and THz signals for each size of micro-delamination were measured through the reflection mode of the THz-TDS system. Each pulse reflected from surface, bottom, and delamination of GFRP was labeled by investigating the behavior of the THz wave in the GFRP specimens. In order to remove signal noise, the high frequency domain was filtered by performing Fast Fourier Transform (FFT) of the labeled signals, and then it was reverted to the original form using Inverse Fast Fourier Transform (IFFT) [1]. After that, 1D THz signals were transformed into 2D spectrograms through Short-Term Fourier Transform (STFT), which were trained through a convolutional neural network (CNN) [2]. Based on the trained CNN model, a probability map representing the probability distribution of delamination thickness and location was derived for classifying each pulse of actual THz signal. As a
result, it was found that signals according to the thickness and depth of micro-delamination could be successfully identified through the CNN training Model.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 기계공학부 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/191347)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.