Design spaces that consist of millions or billions of design combinations pose a challenge to current methods for identifying optimal solutions. Complex analyses can also lead to lengthy computation times that further challenge the effectiveness of an algorithm in terms of solution quality and run-time. This work explores combining the design space exploration approach of a Multi-Objective Genetic Algorithm with different instance-based, statistical, rule-based and ensemble classifiers to reduce the number of unnecessary function evaluations associated with poorly performing designs. Results indicate that introducing a classifier to identify child designs that are likely to push the Pareto frontier toward an optima reduce the number of function calculations by 75–85%, depending on the classifier implemented.

This content is only available via PDF.
You do not currently have access to this content.