Robust design optimization of imperfect stiffened shells using an active learning method and a hybrid surrogate model

Abstract

There are many uncertain factors in aerospace structures, such as variations in manufacturing tolerance, material properties and environmental aspects. Although conventional robust design optimization (RDO) can effectively take into account these uncertainties under the specified robust requirement, it is less satisfactory in addressing these difficulties owing to the prohibitive numerical cost of finite element analysis of stiffened shells. To improve the efficiency of RDO of imperfect stiffened shells, a new hybrid surrogate model (HSM), taking full advantage of the efficiency of the smeared stiffener method and the accuracy of the finite element method, is developed in this article. Then, a new active learning method is constructed based on the HSM. Furthermore, a hybrid bi-stage RDO framework is proposed to alleviate the computational burden incurred by repeated structural analysis. An example of a typical 3 m diameter stiffened shell demonstrates the high efficiency and accuracy of the proposed method.

KEYWORDS:

• Robust design optimization
• imperfect stiffened shells
• hybrid surrogate model
• efficiency
• accuracy

Acknowledgements

This work was supported by the National Natural Science Foundation of China [grant numbers 11972143 and 11602076], the Natural Science Foundation of Anhui Province [grant number 1708085QA06] and the Fundamental Research Funds for the Central Universities of China [grant numbers JZ2018HGTB0231 and PA2018GDQT0008].

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Huanlin Zhou http://orcid.org/0000-0003-3678-7688

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [grant numbers 11972143 and 11602076], the Natural Science Foundation of Anhui Province [grant number 1708085QA06] and the Fundamental Research Funds for the Central Universities of China [grant numbers JZ2018HGTB0231 and PA2018GDQT0008].