Role of automatic differentiation in concurrent subspace optimization method

Jongsoo Lee, Changkue Park

Research output: Contribution to journalConference article

Abstract

The paper explores the implementation of automatic differentiation in sensitivity based concurrent subspace optimization (CSSO) for coupled multidisciplinary systems. A FORTRAN program, ADIFOR has been used as a tool to evaluate first order derivatives during the iterative processes in global sensitivity equation, design variable allocation, optimal sensitivity analysis and coordination of problem. A coupled multidisciplinary mechanical element design problem is adopted as benchmarking in the present study. Automatic differentiation based CSSO has shown the effectiveness over both traditional non-decomposed approach and finite difference based CSSO in terms of number of iterations and objective function values. The paper also discusses how much of move limits on design variable affect each of final optimization results.

Fingerprint

Benchmarking
Sensitivity analysis
Derivatives

All Science Journal Classification (ASJC) codes

  • Architecture
  • Materials Science(all)
  • Aerospace Engineering
  • Mechanics of Materials
  • Mechanical Engineering

Cite this

@article{70fa5a802f594f70864d0d5ae53ad563,
title = "Role of automatic differentiation in concurrent subspace optimization method",
abstract = "The paper explores the implementation of automatic differentiation in sensitivity based concurrent subspace optimization (CSSO) for coupled multidisciplinary systems. A FORTRAN program, ADIFOR has been used as a tool to evaluate first order derivatives during the iterative processes in global sensitivity equation, design variable allocation, optimal sensitivity analysis and coordination of problem. A coupled multidisciplinary mechanical element design problem is adopted as benchmarking in the present study. Automatic differentiation based CSSO has shown the effectiveness over both traditional non-decomposed approach and finite difference based CSSO in terms of number of iterations and objective function values. The paper also discusses how much of move limits on design variable affect each of final optimization results.",
author = "Jongsoo Lee and Changkue Park",
year = "2000",
month = "1",
day = "1",
language = "English",
volume = "1",
pages = "2175--2184",
journal = "Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference",
issn = "0273-4508",
publisher = "American Institute of Aeronautics and Astronautics Inc. (AIAA)",
number = "III",

}

TY - JOUR

T1 - Role of automatic differentiation in concurrent subspace optimization method

AU - Lee, Jongsoo

AU - Park, Changkue

PY - 2000/1/1

Y1 - 2000/1/1

N2 - The paper explores the implementation of automatic differentiation in sensitivity based concurrent subspace optimization (CSSO) for coupled multidisciplinary systems. A FORTRAN program, ADIFOR has been used as a tool to evaluate first order derivatives during the iterative processes in global sensitivity equation, design variable allocation, optimal sensitivity analysis and coordination of problem. A coupled multidisciplinary mechanical element design problem is adopted as benchmarking in the present study. Automatic differentiation based CSSO has shown the effectiveness over both traditional non-decomposed approach and finite difference based CSSO in terms of number of iterations and objective function values. The paper also discusses how much of move limits on design variable affect each of final optimization results.

AB - The paper explores the implementation of automatic differentiation in sensitivity based concurrent subspace optimization (CSSO) for coupled multidisciplinary systems. A FORTRAN program, ADIFOR has been used as a tool to evaluate first order derivatives during the iterative processes in global sensitivity equation, design variable allocation, optimal sensitivity analysis and coordination of problem. A coupled multidisciplinary mechanical element design problem is adopted as benchmarking in the present study. Automatic differentiation based CSSO has shown the effectiveness over both traditional non-decomposed approach and finite difference based CSSO in terms of number of iterations and objective function values. The paper also discusses how much of move limits on design variable affect each of final optimization results.

UR - http://www.scopus.com/inward/record.url?scp=0033672149&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033672149&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:0033672149

VL - 1

SP - 2175

EP - 2184

JO - Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference

JF - Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference

SN - 0273-4508

IS - III

ER -