Population-based techniques for the multiple objective optimisation of sandwich materials and structures

Abstract

Sandwich materials, consisting of two thin, stiff facings separated by a low density core, can be used to produce structures that are both light and flexurally rigid. Such assemblies are attractive for applications in transport and construction. However, their optimisation is rarely straightforward. Not only is this due to the complex equations that govern their mechanics, but also because multiple design variables and objectives are often present. The work in this thesis identifies population-based optimisation techniques as a novel solution to this challenge. Three of these techniques have been developed in MATLAB specifically for this purpose and are based on particle swarm optimisation (sandwichPSO), ant colony optimisation (sandwichACO), and simulated annealing (sandwichSA). To assess their suitability, a benchmark problem considered the application of these techniques to a multiple objective sandwich beam optimisation. Optimised for stiffness mass and cost, a selection of 16 materials for both facing and core were available. Several constraints were also present. The sandwichACO technique demonstrated superior ability as it was able to obtain all optimal solutions in most cases. However, the sandwichPSO and sandwichSA techniques struggled to identify local optimum solutions for the multi-ply, fibre-reinforced polymer sandwich facing laminates. A further case study then applied sandwichACO to the optimisation of a sandwich plate for a rail vehicle floor panel. In addition to the benchmark, the problem was extended to include 40 materials. Also, the material and thickness of the top face was allowed to be different to the bottom. Furthermore, orthotropic fibre-reinforced facing constructions were included, as well as a localised load constraint. A broad range of optimal solutions were identified for the applied minimum mass and cost objectives. Sandwich constructions provided a significant (approximately 40%) saving in both mass and cost compared to the existing plywood design. More significant mass saving designs were also identified (of over 40%), but with a cost premium. Overall, population-based techniques have demonstrated successful application to the design of sandwich materials and structures.