material 版 (精华区)
发信人: boat (——★船儿★——), 信区: material
标 题: 复合材料8
发信站: 听涛站 (Tue Feb 6 17:20:51 2001), 转信
4.2 Sequential Quadratic Programming Algorithm
For the present problem, the method of sequential quadratic programming
is aa suitable candidate to search for the optimum design. This method can s
olve the general nonlinear programming problem [8] as follows:
minimize f(x)
s.t. g (x)=0 for j=1,....,m
g (x)>=0 for j=m ,.......,m
(224)
x <=x <=x i=1,.....,n
where all problem functions are assumed to be continuously differentiable. T
he mmethod, based on the iterative formulation and solution of quadratic pro
gramming subproblems, obtains subproblems by using a quadratic approximation
of the Lagrangian and by
linearizing the constraints. This is expressed as
minimize _____________
s.t. __________________
_________________________ (25)
_________________
where __ is a positive definite approximation of the Hessian, and __ is the
curr
nt iterate. Let __ be the solution of the subproblem. A line search is used
to f
nd a new point __
x_=x__+__ ____
(26)
such that a "merit function" will have a lower function value at the new ite
rati
n point. Here the augmented Lagrange function is used as the merit function.
Whe
optimality is not achieved, __ is updated according to the modified BFGS fo
rmu
. The
computer program POCSQP (Probabilistic Optimization of Composites using Sequ
enti
l Quadratic Programming) developed by the authors to implement the method de
scr
ed in this paper uses the subroutine NCONG in the IMSL MATH/library [15] for
the
application of sequential
quadratic programming.
--
※ 来源:.听涛站 cces.net.[FROM: 匿名天使的家]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间:1.011毫秒