Ramamoorthy, Vivek T.
(2023)
Heuristics and metaheuristics in the design of sound-absorbing porous materials.
PhD thesis, University of Nottingham.
Abstract
Inexact optimisation techniques such as heuristics and metaheuristics that quickly find near-optimal solutions are widely used to solve hard problems. While metaheuristics are well studied on specific problem domains such as travelling salesman, timetabling, vehicle routing etc., their extension to engineering domains is largely unexplored due to the requirement of domain expertise. In this thesis, we address a specific engineering domain: the design of sound-absorbing porous materials. Porous materials are foams, fibrous materials, woven and non-woven textiles, etc., that are widely used in automotive, aerospace and household applications to isolate and absorb noise to prevent equipment damage, protect hearing or ensure comfort. These materials constitute a significant amount of dead weight in aircraft and space applications, and choosing sub-optimal designs would lead to inefficiency and increased costs. By carefully choosing the material properties and shapes of these materials, favourable resonances can be created making it possible to improve absorption while also reducing weight. The optimisation problem structure is yet to be well-explored and not many comparison studies are available in this domain. This thesis aims to address the knowledge gap by analysing the performance of existing and novel heuristic and metaheuristic methods. Initially, the problem structure is explored by considering a one-dimensional layered sound package problem. Then, the challenging two-dimensional foam shape and topology optimisation is addressed. Topology optimisation involves optimally distributing a given volume of material in a design region such that a performance measure is maximised. Although extensive studies exist for the compliance minimisation problem domain, studies and comparisons on porous material problems are relatively rare. Firstly, a single objective absorption maximisation problem with a constraint on the weight is considered. Then a multi-objective problem of simultaneously maximising absorption and minimising weight is considered. The unique nature of the topology optimisation problem allows it to be solved using combinatorial or continuous, gradient or non-gradient methods. In this work, several optimisation methods are studied, including solid isotropic material with penalisation (SIMP), hill climbing, constructive heuristics, genetic algorithms, tabu search, co-variance matrix adaptation evolution strategy (CMA-ES), differential evolution, non-dominated sorting genetic algorithm (NSGA-II) and hybrid strategies. These approaches are tested on a benchmark of seven acoustics problem instances. The results are used to extract domain-specific insights. The findings highlight that the problem domain is rich with unique varieties of solutions, and by using domain-specific insights, one can design hybrid gradient and non-gradient methods that consistently outperform state-of-the-art ones.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
