항공우주시스템공학회지 (Journal of Aerospace System Engineering)

  • 제18권5호
  • /
  • Pages.73-80
  • /
  • 2024
  • /
  • 1976-6300(pISSN)
  • /
  • 2508-7150(eISSN)
항공우주시스템공학회 (The Society for Aerospace System Engineering)
권호 표지
DOI QR코드

DOI QR Code

심층신경망을 이용한 CFRP 복합재 구조의 공정 유도 스프링백 예측

Prediction of Process-Induced Spring-Back of CFRP Composite Structure Using Deep Neural Network

  • 이유선 (한경국립대학교 ICT로봇기계공학부) ;
  • 김동협 (한국항공대학교 항공우주 및 기계공학과) ;
  • 김상우 (한국항공대학교 항공우주 및 기계공학과) ;
  • 이수용 (한국항공대학교 항공우주산업기술연구소)
  • Yuseon Lee (School of ICT, Robotics & Mechanical Engineering, Hankyong National University) ;
  • Dong-Hyeop Kim (Department of Aerospace and Mechanical Engineering, Korea Aerospace University) ;
  • Sang-Woo Kim (Department of Aerospace and Mechanical Engineering, Korea Aerospace University) ;
  • Soo-Yong Lee (Research Institute of Aerospace Engineering and Technology, Korea Aerospace University)
  • 투고 : 2024.08.02
  • 심사 : 2024.08.22
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 연구에서는 심층신경망을 활용하여 CFRP 복합재 스파의 경화 공정에 의한 스프링백을 예측하였다. 유한요소법 기반 경화 해석을 통해 적층 각, 적층 수, 플랜지 반경에 따른 총 816 개의 스프링백 데이터를 생성하여 심층신경망 모델을 학습시켰다. 학습된 모델의 R-squared 값은 0.99, 평균제곱오차는 0.00093으로 산출되어 모델 성능이 우수함을 확인하였다. 학습되지 않은 플랜지 반경에 대한 스프링백 예측 결과, 유한요소해석 결과와 비교하여 평균 상대오차는 2.18%로 나타났다. 또한, 유한요소해석은 약 20 min이 소요된 반면, 심층신경망을 통한 예측 시간은 약 14 ms에 불과하였다. 이를 통해 CFRP 복합재의 공정 유도 변형을 빠르게 예측하기 위한 심층신경망의 활용 가능성을 확인하였다.

A deep neural network (DNN) was employed to predict the spring-back of a CFRP composite spar induced by the curing process. A total of 816 spring-back data points, derived from varying stacking angles, layer counts, and flange radii, were generated through finite element method (FEM)-based curing analysis to train the DNN model. The trained model demonstrated an R-squared value of 0.99 and a mean squared error of 0.00093, indicating excellent performance. For untrained flange radii, the spring-back predicted by the DNN exhibited a mean relative error of 2.18% when compared to FEM results. Additionally, while FEM analysis required approximately 20 minutes, the DNN-based prediction required only about 14 milliseconds. These results highlight the potential of using DNNs for the rapid prediction of process-induced deformation in CFRP composites.

키워드

과제정보

본 연구는 2024년도 정부(교육부)의 재원으로 한국연구재단의 지원을 받아 수행된 기초연구사업임(No.2022R1A6A1A03056784). 또한 본 연구는 국토교통부/국토교통과학기술진흥원의 지원으로 수행되었음(과제번호 RS-2024-00444205).

참고문헌

  1. B. Dhiman, V. Guleria, and P. Sharma, "Applications and future trends of carbon fiber reinforced polymer composites: a review," International Research Journal of Engineering and Technology, vol. 7, no. 10, pp. 1883-1889, Oct. 2020.
  2. M. Al-Dhaheri, K. A. Khan, R. Umer, F. Van Liempt, and W. J. Cantwell, "Process-induced deformation in U-shaped honeycomb aerospace composite structures," Composite Structures, vol. 248, pp. 112-503, Apr. 2020.
  3. D. H. Kim, S. W. Kim, H. K. Kim, S. C. Kim, and B. G. Shin, "A study on the verification method of bird strike certification requirements for civilian helicopters through numerical analysis," Journal of the Korean Society for Aeronautical & Space Sciences, vol. 13, no. 6, pp. 70-79, Dec. 2019.
  4. M. Fiorina, A. Seman, B. Castanie, K. M. Ali, C. Schwob, and L. Mezeix, "Spring-in prediction for carbon/epoxy aerospace composite structure," Composite Structures, vol. 168, pp. 739-745, May 2017.
  5. D. H. Kim, S. W. Kim, and I. Lee, "Evaluation of curing process-induced deformation in plain woven composite structures based on cure kinetics considering various fabric parameters," Composite Structures, vol. 287, pp. 115-379, May 2022.
  6. D. H. Kim and S. W. Kim, "Process-induced distortion of triaxially braided composites considering different geometric parameters using simplified constitutive model with effective property," Finite Elements in Analysis and Design, vol. 222, pp. 103-974, Jul. 2023.
  7. A. Bhaduri, A. Gupt, and L. Graham-Brady, "Stress field prediction in fiber-reinforced composite materials using a deep learning approach," Composites Part B: Engineering, vol. 238, Jun. 2022.
  8. Y. Shokrollahi, M. M. Nikahd, K. Gholami, and G. Azamara, "Deep learning techniques for predicting stress fields in composite materials: A superior alternative to finite element analysis," Journal of Composites Science, vol. 7, no. 8, pp. 311-323, Jul. 2023.
  9. Y. Wang, C. Soutis, D. Ando, Y. Sutou, and F. Narita, "Application of deep neural network learning in composites design," European Journal of Materials, vol. 2, no. 1, pp. 117-170, Apr. 2022.
  10. J. Hou, B. You, J. Xu, and T. Wang, "Prediction of curing process for thermosetting prepreg compression molding process based on machine learning," Polymer Composites, vol. 43, no. 3, pp. 749-1762, Jan. 2022.
  11. S. Fan, J. Zhang, B. Wang, J. Chen, W. Yang, W. Liu, and Y. Li, "A deep learning method for fast predicting curing process-induced deformation of aeronautical composite structures," Composites Science and Technology, vol. 232, pp. 109-844, May 2023.
  12. W. I. Lee, A. C. Loos, and G. S. Springer, "Heat of reaction, degree of cure, and viscosity of Hercules 3501-6 resin," Journal of Composite Materials, vol. 16, no. 6, pp. 510-520, Jun. 1982.
  13. X. Liu, Z. Guan, X. Wang, et al., "Study on cure-induced residual stresses and spring-in deformation of L-shaped composite laminates using a simplified constitutive model considering stress relaxation," Composite Structures, vol. 272, p. 114-203, Oct. 2021.
  14. W. Chen and D. Zhang, "A micromechanics-based processing model for predicting residual stress in fiber-reinforced polymer matrix composites," Composite Structures, vol. 204, pp. 153-166, Aug. 2018.
  15. B. J. Kang, S. W. Cheon, H. S. Cho, Y. J. Ki, and T. S. Kim, "A deep neural network-based model for predicting sectional structural stiffness of rotorcraft blades," Journal of the Korean Society for Aeronautical & Space Sciences, vol. 18, no. 1, pp. 21-28, Jan. 2024.
  16. V. Sze, Y. H. Chen, T. J. Yang, and J. S. Emer, "Efficient processing of deep neural networks: A tutorial and survey," Proc. of the IEEE, vol. 105, no. 12, pp. 2295-2329, Dec. 2017.
  17. D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," arXiv preprint arXiv:1412.6980, Dec. 2014.
  18. S. Ruder, "An overview of gradient descent optimization algorithms," arXiv preprint arXiv:1609.04747, Sep. 2016.