Hybrid neural network and support vector machine method for optimization
원문보기
IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0274744
(2005-11-14)
|
등록번호 |
US-7293001
(2007-11-06)
|
발명자
/ 주소 |
|
출원인 / 주소 |
- The United States of America as represented by the Administrator of the National Aeronautics and Space Administrator (NASA)
|
인용정보 |
피인용 횟수 :
0 인용 특허 :
4 |
초록
▼
System and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (NN/SVM) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. As a first example, the NN/SVM analysis is applied
System and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (NN/SVM) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. As a first example, the NN/SVM analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. As a second example, the NN/SVM analysis is applied to data classification of a sequence of data points in a multidimensional space. The NN/SVM analysis is also applied to data regression.
대표청구항
▼
What is claimed is: 1. A method for aerodynamic design optimization of an aerodynamic component, the method comprising: providing a computer that is programmed: (1) to provide a group of M parameters that define a design shape of a selected aerodynamic component, and to provide a vector p=p0 of ini
What is claimed is: 1. A method for aerodynamic design optimization of an aerodynamic component, the method comprising: providing a computer that is programmed: (1) to provide a group of M parameters that define a design shape of a selected aerodynamic component, and to provide a vector p=p0 of initial values for each parameter in the group, where M is a selected integer ≧1; (2) to provide data point values and at least one numerical criterion for an optimal design at one or more selected location values of the aerodynamic component; (3) to provide an M-simplex in parameter space, centered at a vector location p=p0 and having a selected diameter d0; (4) to provide a design function fk at each location value k on the aerodynamic component, depending on the choice of the parameter vector p, for the parameter vector p0 and for each parameter vector corresponding to a vertex of the M-simplex, and for an expanded M-simplex centered at p0; (5) to provide a selected first objective function OBJ(p;p0;1), dependent upon the parameter vector p and upon a difference between the optimal design data point value and a first design function data point value at one or more of the location values; (6) to determine a parameter vector p=p(min) within the expanded M-simplex for which the first objective function attains a minimum value; (7) to compute a selected second objective function OBJ(p;p0;2) for p=p(min) and for p=p0; (8) when OBJ(p(min);p0;2) is not less than OBJ(p0;p0;2), to provide a modified expanded M-simplex, with a modified diameter d' satisfying d0k is one of said selected location values, fk(rk;p) is one of said design function data point values, fk(rk;opt) is one of said optimal design data point values, wk is a selected non-negative weight coefficient, q is a selected positive number, and K is a selected positive integer. 3. The method of claim 1, further comprising choosing said first objective function to be the same as said second objective function. 4. The method of claim 1, wherein said second objective function depends upon said second design function, upon said parameter vector p and upon a difference between said optimal design data point value and a data point value computed using computer simulation of a response of said design, at one or more of said location values. 5. The method of claim 1, further comprising choosing said design function to correspond to pressure on an airfoil at said selected locations on a perimeter of the airfoil. 6. The method of claim 1, wherein said computer is further programmed: (13) to provide data point values and at least a second numerical criterion for a second optimal design at one or more selected location values; and (14) to cause said computer to apply steps (1) and (3)-(11) of claim 1 to obtain an optimal set of design parameters for the second optimal design, where the second optimal design has selected third and fourth objective functions that are independent of said first and second objective functions. 7. The method of claim 1, further comprising selecting said aerodynamic component to be a turbine blade or compressor blade and said design to be a shape, viewed in plan view, of the blade. 8. The method of claim 1, further comprising selecting said aerodynamic component to be an aircraft wing and said design to be a shape, viewed in plan view, of the wing.
이 특허에 인용된 특허 (4)
-
Jin,Yaochu; Sendhoff,Bernhard, Approximate fitness functions.
-
Freedman Barry A. (Lincroft NJ) Meketon Marc S. (Middletown NJ) Vanderbei Robert J. (Morganville NJ), Methods and apparatus for efficient resource allocation.
-
Vanderbei Robert J. (Red Bank NJ), Methods and apparatus for efficient resource allocation.
-
Karmarkar Narendra K. (North Plainfield NJ) Ramakrishnan Kajamalai G. (Hillsborough NJ), Preconditioned conjugate gradient system.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.