Development of evolutionary based techniques with applications to engineering.

Every possible problem can be considered to have a set of possible states by which amongst them, some are considered better than others by some chosen measure. It is the intention of optimisation to discover such states that perform better than all others for any given problem. It is an important tool within an array of subject areas, arguably all, in particular engineering, which tackles such applications as shape optimisation and industrial scheduling to name but a few. The aims of this work, are to increase the performance of the in-house general-purpose particle swarm optimiser designed at the department of engineering at Swansea University. This is to be achieved through its hybridisation with a local search, considering both solution refinement and early triggering mechanisms. In the discrete domain, an ant colony algorithm is to be chosen and evaluated by way of a parameter study and comparison against other leading ant colony algorithms made for the purpose of development for the future application to scheduling problems. Objectives are achieved through the increased refinement properties of the particle swarm optimiser with its hybridisation with local search. Additionally, an early switching mechanism is derived for the local search, resulting on average in a 20% reduction in the number of function evaluations required for constrained problems. With the highly unpredictable responses to unconstrained problems, only stagnation measures are derived. This study bridges the gap between the in-house optimiser and other hybrid particle swarm techniques available in the literature, resulting in competitive performance. An extensive literature review of ant colony identified the population-based ant colony algorithm (PACO) for further investigation. A detailed parameter study is conducted, resulting in the realisation of the strongly coupled parameters present. Following this, a hybrid off-line tuning method is devised, hybridising a simple particle swarm optimiser with the ant colony algorithm, resulting in an overall better performing algorithm. This indicated clear strengths in some cases over the more popular of ant colony algorithms