CFD 해석용 형상 스케치 딥러닝을 이용한 OpenFOAM 연동
Interoperation of OpenFOAM by Using Shape Sketch Deep Learning for a Use of CFD Analysis
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This focuses on the possibility of automatically performing CFD (computational fluiddynamics) analysis with hand-written sketches on paper(i.e. the easier way). The sketch isrecognized and predicted as the target of the CFD analysis model by applying Deep ConvolutionalNeural Network suitable for classification of abstract sketch images. The sketch is only possiblewithin the range of primitive geometric shapes in the field of CFD. The CFD analysis model needsto be prepared in advance in parametric format. This study shows that one-stop solution of CFDanalysis is possible with the composition of the following procedures: 1) Sketch image recognition,2) Sketch intention prediction for CFD analysis by using deep learning A.I., 3) Adaptation ofreusable OpenFOAM case, 4) Interoperation with CFD application. Reusable OpenFOAM cases arecoded in a parametric template format. Through this study, it was confirmed that the CFDanalysis can be performed well by recognizing sketch written on paper. In this study, an elbowshape sample case at the prototype level was created and tested. As a result of the performanceevaluation of the sketch recognition artificial neural network model, the accuracy of the sketchintention decision was relatively good at 87.5% (0.8750), and the loss was 37.3% (0.3729).
Abstract
This focuses on the possibility of automatically performing CFD (computational fluiddynamics) analysis with hand-written sketches on paper(i.e. the easier way). The sketch isrecognized and predicted as the target of the CFD analysis model by applying Deep ConvolutionalNeural Network suitable for classification of abstract sketch images. The sketch is only possiblewithin the range of primitive geometric shapes in the field of CFD. The CFD analysis model needsto be prepared in advance in parametric format. This study shows that one-stop solution of CFDanalysis is possible with the composition of the following procedures: 1) Sketch image recognition,2) Sketch intention prediction for CFD analysis by using deep learning A.I., 3) Adaptation ofreusable OpenFOAM case, 4) Interoperation with CFD application. Reusable OpenFOAM cases arecoded in a parametric template format. Through this study, it was confirmed that the CFDanalysis can be performed well by recognizing sketch written on paper. In this study, an elbowshape sample case at the prototype level was created and tested. As a result of the performanceevaluation of the sketch recognition artificial neural network model, the accuracy of the sketchintention decision was relatively good at 87.5% (0.8750), and the loss was 37.3% (0.3729).
참고문헌 (9)
- 지명근, "A Sketch-based 3D Object Retrieval Approach for Augmented Reality Models Using Deep Learning", 인터넷정보학회논문지, 21(1), pp.33~43 (2020), 한국인터넷정보학회[DOI]
- Liu, L., , Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.2862~2871 (2017)
- Park, D. G., , Engineering Applications of Computational Fluid Mechanics, 14(1), pp.180~201 (2019)
- Seddati, O., , 13th International Workshop on ContentBased Multimedia Indexing (2015)
- 김연중, "뉴럴 네트워크의 최적화에 따른 유사태풍 예측에 관한 연구", 한국해양공학회지, 33(5), pp.427~434 (2019), 한국해양공학회[DOI]