Surrogate-based optimization

Description/Abstract

In surrogate-based optimization, the designer's full precision computer (or even physical) models are replaced by curve fits - the surrogates. Optimization is then carried out using these curve fits, which by their nature are much quicker and simpler to use than the full models. Care needs to be taken to ensure that the surrogates used to replace the full analyses are suitably accurate, at least in the areas of interest, but provided they are, a significant speedup in the design optimization process can be achieved. The surrogates can also be used in a variety of supporting roles such as visulaization and system integration tasks. This topic has an extensive research literature - for a more detailed coverage, the reader is referred to the dedicated text by Forrester, Sobester and Keane (2008). Here we refer to a more limited number of publications on the subject - the cited text contains a much more extensive bibliography