Yuseon Lee*, **, Dong-Hyeop Kim**, ***, Sang-Woo Kim**, ****†
*Department of Smart Air Mobility, Korea Aerospace University
**Department of Aerospace and Mechanical Engineering, Korea Aerospace University
***Research Institute for Aerospace Engineering and Technology, Korea Aerospace University
****Department of Aeronautical and Astronautical Engineering, Korea Aerospace University
이유선*, ** · 김동협**, *** · 김상우**, ****†
This article is an open access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
This paper presents a method for rapidly and accurately predicting the process-induced deformation (PID) of CFRP composite spars using a convolutional neural network (CNN). The deformed nodal coordinates of the composite structure were predicted to facilitate effective PID analysis. Finite element method (FEM)-based curing simulations were conducted to generate PID data for various manufacturing parameters, which were then used to train the CNN model. The validation results indicated that the CNN model attained a low mean squared error of 0.0092, with relative errors remaining within 1% in most regions of the CFRP spar compared to FEM-based analysis. Additionally, while FEM-based curing simulations required 15-20 minutes per case, the trained CNN model predicted PID in merely 10 milliseconds. These findings confirm that the CNN model effectively incorporates various manufacturing parameters and enables rapid PID prediction for CFRP composite structures.
본 논문에서는 합성곱 신경망(convolutional neural network, CNN)을 활용하여 CFRP 복합재 스파의 공정 유도 변형(process-induced deformation, PID)을 빠르고 정확하게 예측하는 기법을 제안한다. 특히, 변형된 복합재 구조의 절점 좌표를 예측하여 PID 형상을 효과적으로 분석할 수 있도록 하였다. 유한요소법(finite element method, FEM) 기반 경화 해석을 수행하여 다양한 공정변수에 따른 PID 데이터를 구축하고, 이를 학습 데이터로 활용하여 CNN 모델을 학습시켰다. 검증 결과, CNN 모델의 평균 제곱 오차는 0.0092로 낮게 나타났으며, FEM 기반 경화 해석과 비교했을 때 복합재 스파의 대부분의 영역에서 상대 오차가 1% 이내로 유지되었다. FEM 기반 경화 해석은 해석 조건당 15-20 min이 소요되는 반면, CNN 모델을 활용한 PID 예측에는 단 10 ms만 소요되었다. 이를 통해 CNN 모델이 다양한 공정변수를 반영하여 CFRP 복합재 구조의 PID를 신속하게 예측할 수 있음을 확인하였다.
Keywords: 탄소섬유강화플라스틱(Carbon fiber reinforced plastics), 경화 공정(Curing process), 공정유도 변형(Process-induced deformation), 딥러닝(Deep learning), 합성곱 신경망(Convolutional neural network)
This Article2025; 38(2): 106-113
Published on Apr 30, 2025
Correspondence to**Department of Aerospace and Mechanical Engineering, Korea Aerospace University
****Department of Aeronautical and Astronautical Engineering, Korea Aerospace University