Application of Real Coded Genetic Algorithm for Ship Hull Surface Fitting With a Single Non-Uniform B-spline Surface
- Alternative Title
- 유전알고리즘을 이용한 단일 B-spline 선체 곡면 표현
- Abstract
- 디지털 방식으로 배를 설계하는 과정에서 선형의 곡면 모델링은 배의 생산과 성능수치해석에 있어서 매우 정밀함을 요구한다. 곡면 모델링의 전통적인 방법은 skinning 법이다. 이 방법에서 곡면 모델은 단면 곡선의 데이터 집합으로부터 만들어진다. 그러나 외판의 품질은 stern, bow profile, deck side lines, bottom tangential line 과 같은 특성곡선의 정확도와 단면곡선의 간격에 달려있으며 3 차원 diagonal lines, 불연속곡선과 같은 3 차원 교차곡선을 모두 포함하는 것이 skinning 법에서는 불가능하다. 이것은 배의 외판을 만드는 과정에서 유효한 형태의 정보가 무시된다는 것을 의미한다. 그러므로 고품질의 외판모델링은 쉽지 않은 설계과정이다.
이 연구의 목표는 선체곡면을 단일 B-spline 표현하는 것이다. 너클(Knuckle), 불연속조건, 곡률변화가 심한 구상선수(bulbous bow)와 같이 다양하고 독특한 형상을 고려하기 위해서 많은 최적화 기술이 사용되었다. 최근 몇 년사이에 GA 를 이용하여 multimodal 최적화 해결책을 얻을 수 있었으며 이 방법의 가장 큰 장점은 이전보다 간단하다는 것이다.
이 연구에서는, 최적화된 경계곡선을 찾아 곡면을 피팅(fitting)하기 위하여 GA 를 이용하였다. 선체의 곡면을 유전자 타입으로 가정하였으며 설계 변수의 입력 값은 조정점(vertex)과 노트(knot) 값이다. 이러한 변수들은 미리 정의한 정도(precision)값을 만족할 때까지 곡면의 붐질을 향상 시키기 위하여 수정되었다.
두 가지 알고리즘이 개발되었다. 첫 번째는 경계곡선을 고려한 알고리즘이다. Simultaneous multi-fitting GA 방법은 선수와 선미 경계곡선을 찾아내기 위해서 개발되었다. 이 방법은 선수, 선미곡선부의 공통적인 노트 값을 찾는데 사용되었으며 최적화된 노트 값과 조정점을 제공한다. 유사하게, bottom tangential line 이나 deck side line 같은 다른 경계곡선에도 같은 GA 기법이 적용된다. 두 번째 알고리즘은 경계곡선이 최적화된 후 내부 곡면 데이터를 피팅하기 위해서 개발되었으며 주어진 데이터 값을 만족시키기위해 조정점을 움직이는 GA 기술로 곡면이 생성된다.
네 가지 곡면에 대해 본 연구에서 개발된 기법을 적용하였으며 적용 결과 GA 기법은 좋은 결과를 보여주고 있다. 이를 이용하면 초기 설계 단계에서, 단일 NUB 곡면은 다른 CAD/CAM 프로그램으로 쉽게 변환할 수 있으며, 시각적 표현과 유한요소 방법 등에서 쉽게 이용될 수 있게 된다. 즉 다음 설계단계로 쉽게 전환이 용이하다. 본 연구의 결과는 초기설계 단계에서 곡면을 모델링 함에 있어서 강력한 도구를 제공한다.
In the digital ship design process, surface modeling is required to be as accurate as possible for the effective support of ship production as well as for numerical performance analysis. A traditional method for ship hull form reconstruction is skinning operation. In this method, a surface model is constructed from a set of cross-sectional data. However, the surface quality depends on the cross-sectional spacing and the accuracy of the characteristic curves, such as stern and bow profiles, deck side line, and bottom tangential line. In addition, it is impossible to include all of the intersection curves, such as three-dimensional diagonal lines and unconnected curves, into the skinning process. This means that valuable shape information is ignored during the ship hull form reconstruction process. Therefore, it is difficult to obtain a high quality of hull surface with this approach.
The aim of this research is to construct a single non-uniform B-spline (NUB) surface at the initial ship design stage. In order to consider many different shapes and features, such as knuckles, discontinuity condition, and bulbous bow with high curvature, various optimization techniques with multiple objectives have been widely used in the surface reconstruction process. In recent years, the genetic algorithm (GA) has gained increasing attention as a multimodal optimization solution for efficient surface reconstruction. One of the most powerful features of this method is its simplicity.
In this research, the GA has been used as a major tool to search for optimal boundary curves and to fit a surface. A preliminary hull surface is assumed to be a gene type. The encoded design variables for surface construction are the location of the vertices and knot values. Those variables are modified to improve the surface quality until the predefined precision is satisfied.
Two algorithms for the fitting method are developed. The first algorithm is to determine the boundary curves. The simultaneous multi-fitting GA method is developed as an approach to find boundary curves such as stem and stern profiles. This method considers both stem and stern curves simultaneously and finds the common knot values for both curves. Similarly, the same GA technique is applied for other boundary curves at the bottom and deck. The second algorithm is employed to fit the interior data points after fitting the boundary curves. The GA technique generates a final surface by manipulating the vertices to fit the interior given data points.
In the application examples, four different surfaces are presented. The GA technique developed in this research has proven to provide good single NUB surfaces with high efficiency. Therefore, the single NUB surface can be translated into other CAD/CAM (Computer Aided Design/Computer Aided Manufacturing) programs easily. In the early design stage, the single NUB surface is more convenient for visualization performance and finite element methods.
The contributions of this dissertation are a simultaneous fitting of the two different NUB boundary curves and interior NUB surface using the GA and an innovative construction of a single NUB surface for geometrically complicated ship hull forms. The simultaneous fitting technique provides optimal knot values and vertices arrangement for the surface reconstruction. The single NUB surface can be readily translated into many CAD/CAM packages, which facilitate the smooth data transition across the different design stages. These factors provide a powerful tool for the hull form construction at the initial design stage.
- Author(s)
- Le, Tat-Hien
- Issued Date
- 2009
- Awarded Date
- 2009. 8
- Type
- Dissertation
- Keyword
- Surface Fitting Genetic Algorithm Multimodal Optimization Hull Form Reconstruction Simultaneous Multi-Fitting
- Publisher
- 부경대학교 대학원
- URI
- https://repository.pknu.ac.kr:8443/handle/2021.oak/11230
http://pknu.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000001955066
- Affiliation
- 부경대학교 대학원
- Department
- 대학원 조선해양시스템공학과
- Advisor
- 김동준
- Table Of Contents
- Chapter 1 INTRODUCTION = 1
1.1 State of the Art = 1
1.2 About This Work = 7
Chapter 2 CLASSIFICATION OF SURFACE MODELING = 9
2.1 Boundary Interpolating Patch Models = 9
2.1.1 Ruled Surfaces = 9
2.1.2 Lofted Surfaces = 10
2.1.3 Bilinear Blended Coons Patch = 10
2.1.4 Bicubic Coons Patches = 12
2.2 Irregular Patch = 14
2.3 Parametric Polynomial Patch Model = 16
2.3.1 Standard Polynomial Surface Patch = 17
2.3.2 Ferguson Surface Patch = 18
2.3.3 B?zier Surface Patch = 19
2.3.4 Uniform B-Spline Surface Patch = 20
2.3.5 Non-Uniform B-Spline Surface Patch = 21
2.3.6 Definition and Properties of Knot Vector = 22
2.3.7 Definition and Properties of Non-Uniform B-Spline Basis Function = 23
2.3.8 Non-Uniform B-spline Surface from 3D Data Array = 24
Chapter 3 OVERVIEW OF THE NON-UNIFORM B-SPLINE FITTING ALGORITHM = 27
3.1 Non-Uniform B-Spline Surface Fitting Application in Ship Hull Design = 27
3.2 Hull Form Modeling Requirements = 29
3.2.1 Shape Requirements = 30
3.2.2 Continuity between Patches = 31
3.2.3 End Condition = 33
3.2.4 Irregular Patch Constraints = 34
3.2.5 Effect of Multiple Knot Vector and Multiple Vertex Point = 35
3.3 Matrices Inversion Problems in Non-Uniform B-spline Surface Fitting = 37
Chapter 4 APPLICATION OF REAL CODED GENETIC ALGORITHM FOR SURFACE FITTING = 40
4.1 The Goals of Optimization = 40
4.1.1 What Is Optimization? = 40
4.1.2 Local and Global Optimization = 41
4.2 Overview of Real Coded Genetic Algorithm = 42
4.3 Fitness Function For Non-Uniform B-spline Surface Fitting = 43
4.4 Encoding for Initial Population = 43
4.5 Reproduction Process = 44
4.6 Crossover Process = 46
4.7 Mutation Process = 47
4.8 Crossover and Mutation Probability = 48
Chapter 5 SINGLE NON-UNIFORM B-SPLINE SURFACE FITTING = 50
5.1 Non-Uniform B-spline Curve Fitting for Boundary Curves = 50
5.1.1 Yoshimoto?s Method for Boundary Curves = 50
5.1.2 Different Sets of Knot Value at Stern and Bow Boundaries = 51
5.1.3 The Other Problems of Boundary Curves Fitting = 53
5.2 New Approach to Boundary Curves Fitting = 54
5.2.1 Simultaneous GA Fitting for Multiple Curves = 54
5.2.2 Handling Weakly Knuckle Point and Twist Problem = 56
5.3 New Approach to Surface Fitting for the Given Interior Data Point by Using GA = 58
5.3.1 Vertices Encoding for Initial Population = 58
5.3.2 Reproduction Process = 60
5.3.3 Crossover Process = 61
5.3.4 Mutation Process = 62
5.3.5 Nearest Point Finding for Fitness Function = 63
5.4 Summary = 65
Chapter 6 APPLICATION EXAMPLES = 68
6.1 Simple Surface = 68
6.2 Yacht Hull Surface = 72
6.3 Complicated Surface = 75
6.4 Container Ship Hull Form = 78
Chapter 7 CONCLUSIONS = 83
REFERENCES = 86
APPENDICES = 91
- Degree
- Doctor
-
Appears in Collections:
- 대학원 > 조선해양시스템공학과
- Authorize & License
-
- Files in This Item:
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.