Publication detail

Gradient vs. approximation design optimization techniques in low-dimensional convex problems

FEDORIK, F.

Original Title

Gradient vs. approximation design optimization techniques in low-dimensional convex problems

Type

conference paper

Language

English

Original Abstract

Design Optimization methods' application in structural designing represents a suitable manner for efficient designs of practical problems. The optimization techniques' implementation into multi-physical softwares permits designers to utilize them in a wide range of engineering problems. These methods are usually based on modified mathematical programming techniques and/or their combinations to improve universality and robustness for various human and technical problems. The presented paper deals with the analysis of optimization methods and tools within the frame of one to three-dimensional strictly convex optimization problems, which represent a component of the Design Optimization module in the Ansys program. The First Order method, based on combination of steepest descent and conjugate gradient method, and Supbproblem Approximation method, which uses approximation of dependent variables' functions, accompanying with facilitation of Random, Sweep, Factorial and Gradient Tools, are analyzed, where in different characteristics of the methods are observed.

Keywords

constraints; Convex optimization; efficiency; FEM/FEA; First Order Method; robustness; Subproblem Approximation Method

Authors

FEDORIK, F.

RIV year

2013

Released

21. 9. 2013

ISBN

978-0-7354-1184-5

Book

ICNAAM 2013

ISBN

0094-243X

Periodical

AIP conference proceedings

Year of study

1558

Number

2

State

United States of America

Pages from

2175

Pages to

2178

Pages count

4

BibTex

@inproceedings{BUT107571,
  author="Filip {Fedorik}",
  title="Gradient vs. approximation design optimization techniques in low-dimensional convex problems",
  booktitle="ICNAAM 2013",
  year="2013",
  journal="AIP conference proceedings",
  volume="1558",
  number="2",
  pages="2175--2178",
  isbn="978-0-7354-1184-5",
  issn="0094-243X"
}