Metamodels of optimal quality for stochastic structural optimization

摘要

Metamodels or response surface models are frequently chosen to reduce computational cost for structural optimization problems. These methods are also very popular for structural reliability analysis. It is therefore not surprising that response surface models are very attractive for reliability-based structural optimization. The paper discusses strategies to obtain a suitable response surface model, to assess its quality concerning prediction, and to use the response surface mode to identify important and unimportant variables. Several mathematical and structural examples illustrate the applicability of the presented approach. (C) 2017 Elsevier Ltd. All rights reserved.