A Progressive Approach for Multi-objective Design through Inter-related Modeling of Value System and Meta-model

Abstract

Multi-objective optimization (MOP) has been increasingly interested in supporting agile and flexible manufacturing in complex and global decision environment under diversified customer demands. To avoid stiffness and shortcomings of the conventional methods and to facilitate wide application of MOP in designing tasks, we proposed a flexible approach in this paper. It relies on the progressive modeling of intent or value system and the artificial system in an inter-related manner. The value system is modeled based on easy subjective judgments regarding preference on the alternatives while the system model involving a meta-model is described as responsive surface by the virtue of design of experiments. Beginning with building rough model for each, the approach can lead to the preferentially optimal solution gradually through correcting both models especially concentrated on the niche around the tentative solution. We have shown such approach can improve the complex and complicated design process while reducing the designer's load to express his/her preference. After explaining the proposed method in detail, effectiveness is examined through illustrative case studies.