Abstract
The knowledge of experienced users in solving real-world optimization problems can be formulated as inter-variable relationships to guide an optimization algorithm towards good solutions faster. Alternatively, such interactions can be learned algorithmically during the optimization by analysing good solutions—a process called innovization. Any common pattern extracted from good solutions can be used as a repair operator to modify candidate solutions. The key aspect is to strike a balance between the relevance of the pattern and the extent of its use in the repair operator. This article proposes a multi-objective evolutionary algorithm framework that combines problem-specific knowledge and online innovization approaches to solve two real-world large-scale multi-objective problems: a 879- and 1479-variable truss design and a 544-variable solid fuel rocket design. Four repair operators suitable for uncovering monotonic relations involving multiple decision variables are proposed. Performance variations resulting from different combinations of initial user knowledge and repair operators are also presented.
Disclosure statement
No potential conflict of interest was reported by the authors.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon request.
Notes
1 Comparable variables mean that they have identical units and vary in a similar range. For example, two size-related variables varying within are defined as comparable variables here.