Buckling optimization of variable stiffness composite cylinders by using multi-fidelity surrogate models

Qi Guo,Jiutao Hang,Suian Wang,Wenzhi Hui,Zonghong Xie

DOI: https://doi.org/10.1016/j.tws.2020.107014

2020-11-01

Abstract:<p>An efficient design optimization method assisted by multi-fidelity surrogate models is presented for the buckling design of thin-walled variable stiffness composite cylinders made by fiber steering. To reduce the computational burden, a multi-fidelity surrogate model called hierarchical kriging model is constructed through a few expensive high-fidelity samples and many cheap low-fidelity samples. Fine and coarse finite element (FE) analysis is performed respectively to calculate the structural responses for corresponding datasets. The Efficient Global Optimization based on a modified Expected Improvement criterion is introduced to adaptively select new sampling points of variable fidelity. Two case studies, a circular composite cylinder under pure bending and an elliptical composite cylinder under torsion, are considered. The effects of different number of design variables on the optimum configuration are studied to demonstrate the validity and robustness of the method. The effects of the structural parameters of the elliptical cylinder on the buckling load improvement are also investigated. The results indicate that the optimization efficiency is remarkably improved by the present method when compared with standard Efficient Global Optimization as well as available methods in the literature. Additionally, the investigation in the mechanism of load carrying capacity improvement shows that the increase is mainly due to the load redistribution.</p>

engineering, mechanical,mechanics, civil