Many-objective optimisation based on decomposition strategies; enhancing, contrasting and visualising pareto front approximations of evolutionary algorithms

Many-objective optimisation analysis is frequently carried out today on problems of ever-increasing complexity. Metaheuristic evolutionary algorithms represent effective and practical tools in solving such issues. The current PhD thesis describes four contributions to the field of study.

A new method for the creation of reference points and an indexing system of reference vectors are proposed, generating more evenly distributed reference points than the popular uniform design method.

A new quality metric/indicator for diversity evaluation on MaOP approximations is proposed. The numerical studies show that the new indicator varies more systematically with the diversity change of PF approximations compared to how frequently applied existing diversity indicators behave.

A new visualisation method for revealing high dimensional MaOP approximations is proposed. The new method satisfies all the desired requirements of a visualisation method in a balanced manner.

A real-life application case study is performed on an additive manufacturing problem: optimisation analysis on process parameters of a selective laser melting manufacturing process. The three proposed methods and related algorithms are utilised to enhance, contrast, and visualise the Pareto Front approximations. The outcome shows that the proposed methods are readily used to enhance the quality of the Pareto front and the Pareto optimal process parameters of the manufacturing operation.