.png){kind=icon}
Return
High-dimensional reliability-based structural design optimization: A surrogate model-assisted decoupled method based on importance sampling quantiles
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engstruct.2024.119549.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Reliability-based design optimization aims at maximizing the performance of a structural design while ensuring the reliability with respect to desired criteria in uncertain environments. Decoupling methods have been developed to alleviate the computational complexity by gradually approximating probabilistic constraints with deterministic ones. Surrogate model techniques, entailing building a cheaper-to-evaluate surrogate model as a substitute for the original performance function, have also received much attention due to their high computational efficiency. However, the state-of-the-art methods still face challenges when handling certain problems with high-dimensionality, discrete training data and very low failure probabilities. To address the above gaps, the work proposes a novel Surrogate model-assisted Quantile-based Sequential Optimization and Reliability Assessment method, called SQ-SORA. Different from the existing methods, this work proposes transforming the constraint for the performance function to the one for the decision function of a Support Vector Machine surrogate model. Therefore, the constraint can be continuously shifted even with a discrete training database. Moreover, this work proposes a new strategy for shifting the constraint based on the quantile function with importance sampling. A bisection algorithm is developed to gradually approach the threshold with respect to the target reliability index. Substantially higher efficiency can be achieved compared with traditional MCS-based shifting, especially for rare events. Several numerical examples and a practical application of reliability-based cable design optimization for a long-span cable-stayed bridge illustrate the performance and advantages of the proposed method.
Keywords:
Reliability-based design optimization
Support vector machines
Quantile
Decoupled method
Importance sampling
Journal
IF:
6.4
Papers:
2.1W
Citations:
8.7W
Organization
Cited Papers
ShareSaveShareSaveShareSaveFalse reactive HIV-1 diagnostic test results in an individual from Kenya on multiple testing platforms-A case report.png)
IDCases
OAAI
IDCasesIF0
OAAIShareSave
ShareSave
View more