대한조선학회논문집 (Journal of the Society of Naval Architects of Korea)

  • 제57권4호
  • /
  • Pages.191-197
  • /
  • 2020
  • /
  • 1225-1143(pISSN)
  • /
  • 2287-7355(eISSN)
대한조선학회 (The Society of Naval Architects of Korea)
권호 표지
DOI QR코드

DOI QR Code

선박 초기 설계 단계에서의 강화학습을 이용한 배관경로 설계

Pipe Routing using Reinforcement Learning on Initial Design Stage

  • 신동선 (부산대학교 조선해양공학과) ;
  • 박병철 (부산대학교 조선해양공학과) ;
  • 임채옥 (부산대학교 조선해양공학과) ;
  • 오상진 (부산대학교 조선해양공학과) ;
  • 김기용 (부산대학교 조선해양공학과) ;
  • 신성철 (부산대학교 조선해양공학과)
  • Shin, Dong-seon (Department of Naval Architecture and Ocean Engineering, Pusan National University) ;
  • Park, Byeong-cheol (Department of Naval Architecture and Ocean Engineering, Pusan National University) ;
  • Lim, Chae-og (Department of Naval Architecture and Ocean Engineering, Pusan National University) ;
  • Oh, Sang-jin (Department of Naval Architecture and Ocean Engineering, Pusan National University) ;
  • Kim, Gi-yong (Department of Naval Architecture and Ocean Engineering, Pusan National University) ;
  • Shin, Sung-chul (Department of Naval Architecture and Ocean Engineering, Pusan National University)
  • 투고 : 2019.07.26
  • 심사 : 2020.05.04
  • 발행 : 2020.08.20
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

Pipe routing is a important part of the whole design process in the shipbuilding industry. It has a lot of constraints and many tasks that should be considered together. Also, the result of this stage affects follow-up works in a wide scope. Therefore, this part requires skilled designers and a lot of time. This study aims to reduce the workload and time during the design process by automating the pipe route design on initial stage. In this study, the reinforcement learning was used for pipe auto-routing. Reinforcement learning has the advantage of dynamically selecting routes, unlike existing algorithms. Therefore, it is suitable for the pipe routing design in ship design process which is frequently modified. At last, the effectiveness of this study was verified by comparing pipelines which were designed by piping designer and reinforcement learning results.

키워드

참고문헌

  1. Asmara, A., 2013. Pipe routing framework for detailed ship design. Ph.D. Delft: Delft University of Technology.
  2. Kim, S.H., Ruy, W.S. & Jang, B.S., 2013. The development of a practical pipe auto-routing system in a shipbuilding CAD environment using network optimization. International Journal of Naval Architecture and Ocean Engineering, 5(3), pp.468-477. https://doi.org/10.2478/IJNAOE-2013-0146
  3. Millan, J., D., R., & Torras, C., 1992. A reinforcement connectionist approach to robot path finding in non-maze-Like environments. Machine Learning, 8(3-4), pp.363-395. https://doi.org/10.1007/BF00992702
  4. Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D. & Riedmiller, M., 2013. Playing Atari with Deep Reinforcement Learning. arXiv:1312.5602.
  5. Nguyen, H., Kim, D.J. & Gao, J., 2016. 3D Piping route design including branch and elbow using improvements for dijkstra's algorithm, International conference on artificial intelligence: Technologies and Applications, pp.309-312.
  6. Park, J.H. & Storch, L., R., 2002. Pipe-routing algorithm development: case study of a ship engine room design. Journal of the Expert Systems with Applications, 23(3), pp.299-309. https://doi.org/10.1016/S0957-4174(02)00049-0
  7. Schulman, J., Levine, S., Moritz, P., Jordan, M. & Abbeel, P., 2015. Trust region policy optimization. arXiv:1502.05477v1.
  8. Schulman, J., Wolski, F., Dhariwal, P., Radford, A. & Klimov, O., 2017. Proximal policy optimization Algorithms. arXiv:1707.06347v2.