点击次数：

论文类型：期刊论文

发表时间：2017-01-01

发表刊物：APPLIED MATHEMATICAL MODELLING

收录刊物：SCIE、EI、Scopus

卷号：41

页面范围：257-270

ISSN号：0307-904X

关键字：Reliability-based design optimization; Performance measure approach; Self-adaptive mean value; Hybrid self-adaptive mean value; Sufficient descent condition

摘要：Due to the efficiency and simplicity, advanced mean value (AMV) method is widely used to evaluate the probabilistic constraints in reliability-based design optimization (RBDO) problems. However, it may produce unstable results as periodic and chaos solutions for highly nonlinear performance functions. In this paper, the AMV is modified based on a self-adaptive step size, named as the self-adjusted mean value (SMV) method, where the step size for reliability analysis is adjusted based on a power function dynamically. Then, a hybrid self-adjusted mean value (HSMV) method is developed to enhance the robustness and efficiency of iterative scheme in the reliability loop, where the AMV is combined with the SMV on the basis of sufficient descent condition. Finally, the proposed methods (i.e. SMV and HSMV) are compared with other existing performance measure approaches through several nonlinear mathematical/structural examples. Results show that the SMV and HSMV are more efficient with enhanced robustness for both convex and concave performance functions. (C) 2016 Elsevier Inc. All rights reserved.