| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 62.27 MB | Adobe PDF |
Advisor(s)
Abstract(s)
The solution of inverse problems based on experimental data is itself an important research issue. In this context and assuming that an experimental sample is available, rather than trying to find a specific deterministic solution for the inverse problem, one aims to determine the probabilistic distribution of the modelling parameters, based on the minimization of the dissimilarity between the empirical cumulative distribution function of an experimental solution and its simulation counterpart. The present paper presents na innovative framework, where Differential Evolution is extended in order to estimate not only an optimal set of modelling parameters, but to estimate their optimal probabilistic distributions. Additionally, the Adaptive Empirical Distributions optimization scheme is here introduced. Both schemes rely on the two samples Kolmogorov-Smirnov goodness-offit test in order to evaluate the resemblance between two empirical cumulative distribution functions. A numerical example is considered in order to assess the performance of the proposed strategies and validity of their solutions.
Description
Keywords
Inverse problem Extended differential evolution Adaptive empirical distributions
Citation
SILVA, Tiago A. N.; [et al]. – A global optimization approach based on adaptive populations. In SYMCOMP 2015. Faro, Portugal: ECCOMAS, 2015. ISBN 978-989-96264-6-1. Pp. 389-401.