Advanced

On Latin hypercube sampling for structural reliability analysis

Olsson, Anders LU ; Sandberg, Göran LU and Dahlblom, Ola LU (2003) In Structural Safety 25(1). p.47-68
Abstract
Latin hypercube sampling is suggested as a tool to improve the efficiency of different importance sampling methods for structural reliability analysis. In simple importance sampling, where the sampling centre is moved from the origin to the design point, standard Monte Carlo sampling can be replaced by Latin hypercube sampling. The efficiency improvement is then highly dependent on the choice of sampling directions. Different versions of Latin hypercube sampling are also successfully employed to improve the more efficient axis orthogonal importance sampling method. By means of different numerical examples, it is shown that more than 50% of the computer effort can be saved by using Latin hypercubes instead of simple Monte Carlo in... (More)
Latin hypercube sampling is suggested as a tool to improve the efficiency of different importance sampling methods for structural reliability analysis. In simple importance sampling, where the sampling centre is moved from the origin to the design point, standard Monte Carlo sampling can be replaced by Latin hypercube sampling. The efficiency improvement is then highly dependent on the choice of sampling directions. Different versions of Latin hypercube sampling are also successfully employed to improve the more efficient axis orthogonal importance sampling method. By means of different numerical examples, it is shown that more than 50% of the computer effort can be saved by using Latin hypercubes instead of simple Monte Carlo in importance sampling. The exact savings, however, are dependent on details in the use of Latin hypercubes and on the shape of the failure surfaces of the problems. (C) 2002 Elsevier Science Ltd. All rights reserved. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
orthogonal, reliability, Latin hypercube sampling, FORM, axis, importance sampling, directional sampling
in
Structural Safety
volume
25
issue
1
pages
47 - 68
publisher
Elsevier
external identifiers
  • wos:000179343400003
  • scopus:0037212955
ISSN
0167-4730
DOI
10.1016/S0167-4730(02)00039-5
language
English
LU publication?
yes
id
f57d2041-ae68-4d01-954c-3edbff895034 (old id 322970)
date added to LUP
2007-09-19 15:33:17
date last changed
2018-02-18 04:22:19
@article{f57d2041-ae68-4d01-954c-3edbff895034,
  abstract     = {Latin hypercube sampling is suggested as a tool to improve the efficiency of different importance sampling methods for structural reliability analysis. In simple importance sampling, where the sampling centre is moved from the origin to the design point, standard Monte Carlo sampling can be replaced by Latin hypercube sampling. The efficiency improvement is then highly dependent on the choice of sampling directions. Different versions of Latin hypercube sampling are also successfully employed to improve the more efficient axis orthogonal importance sampling method. By means of different numerical examples, it is shown that more than 50% of the computer effort can be saved by using Latin hypercubes instead of simple Monte Carlo in importance sampling. The exact savings, however, are dependent on details in the use of Latin hypercubes and on the shape of the failure surfaces of the problems. (C) 2002 Elsevier Science Ltd. All rights reserved.},
  author       = {Olsson, Anders and Sandberg, Göran and Dahlblom, Ola},
  issn         = {0167-4730},
  keyword      = {orthogonal,reliability,Latin hypercube sampling,FORM,axis,importance sampling,directional sampling},
  language     = {eng},
  number       = {1},
  pages        = {47--68},
  publisher    = {Elsevier},
  series       = {Structural Safety},
  title        = {On Latin hypercube sampling for structural reliability analysis},
  url          = {http://dx.doi.org/10.1016/S0167-4730(02)00039-5},
  volume       = {25},
  year         = {2003},
}