Learning and adapting fuzzy set-based trade-off strategy in engineering design synthesis

Multi-criteria decision methods are common in engineering design solution synthesis to accomplish trade-offs among competing objectives. Since design is an evolving interactive process with less precise information available in earlier stages than in later stages, the trade-off strategy could also change as design stages progress and more information is added. This paper provides the rationale and advantages of choosing design trade-off strategies based on fuzzy set-based preference aggregation, which not only relies on specifying parameters about importance weights of design attributes, but also the degree of compensation among them. A neural network function approximation method and procedure, devised to learn and adapt the trade-off strategies according to the current preference information available from the environmental evaluation feedback, is then provided. As the design process evolves, this adaptation should lead to more suitable and stabilized trade-off strategies. A numerical example of experimentation is included to demonstrate the approach.