Hexagonal honeycomb cell optimisation by way of meta-model techniques

Sadeghi Esfahlani, Shabnam and Shirvani, Hassan and Shirvani, Ayoub and Nwaubani, Sunny O. and Mebrahtu, Habtom and Chirwa, Clive (2013) Hexagonal honeycomb cell optimisation by way of meta-model techniques. International Journal of Crashworthiness, 18 (3). pp. 264-275. ISSN 1754-2111

Full text not available from this repository.
Official URL: https://doi.org/10.1080/13588265.2013.776337


This paper presents the result of an optimisation study by linear, quadratic, Kriging and radial basis meta-models in order to augment the crashworthiness characteristic of cellular structures. Thin-walled cellular structures (honeycomb) have the ability to absorb impact energy during crashing, thus it is important to enhance the crushing efficiency and optimise the structural reliability. The optimisation carried out in this study is aimed at maximising the energy absorption characteristics using meta-models while considering some limitation on the maximum force as a constraint. Achieving these characteristics is an important factor in crashworthiness analysis, which minimises the damage in dynamic performance. The objective of using various meta-models is to qualify the meta-model in crashworthiness analysis using different point selection schemes and different number of points. The optimisation is performed in two stages; through experimental design methods in which a set of sampling points is selected from design space and polynomial fitting in order to optimise the objective. It is concluded that D-optimal best suits response surface method for model approximation. Kriging performed best by space filling and the best point selection scheme for radial basis surrogate is latin hypercube design. The results show that for optimising the crashworthiness characteristics of honeycomb, Kriging and quadratic response surface (RS) are best, in terms of accuracy and robustness point of view and the radial basis neural network would be the second best. During this optimisation, the RS was combined with detailed geometrically simplified finite element model of honeycomb cell using ANSYS/LS-DYNA, LS-DYNA and LS-opt packages. Approximated functions combined with the finite element analysis were an effective tool to optimise highly non-linear impact problems. This has led to the development of validated algorithm that enabled the development of the optimised solutions.

Item Type: Journal Article
Keywords: linear response surface method (L-RSM), quadratic response surface method (Q-RSM), radial basis neural network (RBNN), radial basis function (RBF), latin hypercube design (LH-DOE), crashworthiness
Faculty: ARCHIVED Faculty of Science & Technology (until September 2018)
Depositing User: Repository Admin
Date Deposited: 03 Feb 2014 15:34
Last Modified: 03 Feb 2021 13:04
URI: https://arro.anglia.ac.uk/id/eprint/312147

Actions (login required)

Edit Item Edit Item