|
Abstract:
|
When solving real engineering design problems, exploring the decision space for alternative and near-optimal solutions, and identifying non-unique solutions are useful. An efficient way to conduct this search is to identify alternative solutions that are as far apart as possible in the decision space. This will help not only provide meaningful design choices, if any are available, but also assess the degree of non-uniqueness present in the problem. This paper describes the development and testing of a new evolutionary algorithm that extends evolution strategies to generate simultaneously a set of maximally different alternatives that are efficient solutions for engineering design problems. The new method, niched co-evolution strategies (NCES), is demonstrated for a groundwater pollutant source characterization problem that is known to have a high degree of non-uniqueness. Through an illustrative case study, NCES is shown to successfully assess and address the non- uniqueness for an instance of this relatively complex engineering design problem.
|
