Response surface method based on uniform design and weighted least squares for non‐probabilistic reliability analysis
Pengya Fang,Shuhao Li,Xiao Guo,Zhenhua Wen
DOI: https://doi.org/10.1002/nme.6426
IF: 3.021
2020-06-24
International Journal for Numerical Methods in Engineering
Abstract:<p>The non‐probabilistic reliability theory is a promising methodology for implementing structural reliability analysis in case of scarce statistical data. One of the main obstacles to implementing non‐probabilistic reliability analysis is the implication of the limit state function (LSF) for complex structures. This paper aims to establish a surrogate model of the LSF with higher simulation precision, and whereby proposes a response surface method (RSM) based on the combination of uniform design (UD) and weighted least squares (WLS). At first, the UD method is selected as the sampling method of interval variables to realize the uniform space‐filling of the initial samples, and the sample set is updated by gradually adding the approximate optimal points to increase the sampling density of critical domain. Then, the WLS method is applied to improve the precision of the response surface by adjusting the importance of samples to the function fitting. Finally, a method of constructing sample weights is developed. Two examples are applied to validate the feasibility and efficiency of the proposed method. Results show that the proposed method is effective for non‐probabilistic reliability analysis of complex structures owning to high computational precision and low computational cost in both numerical and case study.</p>
engineering, multidisciplinary,mathematics, interdisciplinary applications