표제지
요약문
ABSTRACT
목차
NOMENCLATURE 11
GREEK LETTERS 12
SUBSCRIPTS 13
1. 서론 14
2. 모델링 21
2.1. 초기 형상 및 설계 범위 21
2.2. 라틴 하이버큐브 샘플링(Latin hypercube sampling) 23
2.3. 격자 구성 및 수치해석 조건 30
2.4. 격자 민감도 분석 및 수치해석 방법 검증 33
2.5. 인공신경망 35
2.6. 유전 알고리즘 37
3. 결과 41
3.1. 형상 변수 영향 분석 41
3.2. 인공신경망 학습 성능 평가 및 metamodel 생성 47
3.3. 형상최적화 51
4. 결론 59
참고문헌 61
Table 1. Design parameters range of the honeycomb (HC) labyrinth seal 22
Table 2. Design points by Latin hypercube sampling 24
Table 3. Numerical methods and boundary condition 32
Table 4. Reference & optimized parameters of the honeycomb labyrinth seal geometry 54
Fig. 1. Schematic of effective clearance 16
Fig. 2. Flowchart summarizing the research process in 3 steps 20
Fig. 3. Reference schematic geometry and parameters of the honeycomb labyrinth seal 22
Fig. 4. Example of 2D LHS sample 24
Fig. 5. Computational domain and grids 32
Fig. 6. Example of grid dependence test results of the reference seal 34
Fig. 7. CFD validation of honeycomb labyrinth seal (PR=1.3) 34
Fig. 8. Split original data (training, validation, test) 36
Fig. 9. Structure of the genetic algorithm 39
Fig. 10. Example of cross over process 40
Fig. 11. Example of mutation process 40
Fig. 12. Effect of first tooth angle on discharge coefficient 43
Fig. 13. Velocity vector for first tooth angle effect on discharge coefficient (β₁=60°) 43
Fig. 14. Effect of second tooth angle on discharge coefficient 44
Fig. 15. Velocity vector for second tooth angle effect on discharge coefficient (β₂=70°) 44
Fig. 16. Effect of step position on discharge coefficient 45
Fig. 17. Effect of pitch on discharge coefficient 45
Fig. 18. Effect of tooth height on discharge coefficient 46
Fig. 19. Effect of step height on discharge coefficient 46
Fig. 20. Structure of feed forward neural network 48
Fig. 21. Number of hidden layers on the performance of the neural network 48
Fig. 22. Neural network performance evaluation 49
Fig. 23. Regression and coefficient(R) of discharge coefficient by ANN & CFD 49
Fig. 24. Comparison predicted results ANN & new unseen data 50
Fig. 25. Genetic algorithm convergence process 53
Fig. 26. Honeycomb stepped labyrinth seal - a) reference seal, b) optimized seal 53
Fig. 27. Velocity vector for honeycomb stepped labyrinth seal (PR=2.5) 55
Fig. 28. Velocity plots for honeycomb stepped labyrinth seal (PR=2.5) 56
Fig. 29. Static pressure plots for honeycomb stepped labyrinth seal (PR=2.5) 57
Fig. 30. Velocity vector for tooth angle effect 58