Impacts of Type-2 Fuzzy Modeling Approach on Set-Based Design, Based On Results from an Existing Hybrid Agent Design Experiment

摘要

Set-based design is an advance design method based on the Toyota design process. Unlike point-based design, set-based design focuses on keeping a large design space open for a long period of time, thus increasing the possibility of achieving a globally optimal design versus a locally optimal one. Maintaining a large design space is critical for large, complex, system integration intensive, concurrent team engineering projects since the success of the final product is dependent on the integration of dispiriting systems, not the optimization of a single component. Previous work by the authors focused on a set-based design approach to the cross-functional team design activity in early ship design. A hybrid agent analogy, with human agents aided by intermediate computer agents, was introduced. A new implementation of a fuzzy logic system was developed to enable these designers to communicate and negotiate within the team about their preferences over ranges of design independent values. The output of the system is a crisp valued function of the active design variable, indicating the joint preference of the team of human design agents over the current set of each design independent variable. The conclusions from the previous work were that the fuzzy logic agent software successfully facilitated set-based design, thus, increasing the probability of reaching a more globally optimal ship design. The set-based design paradigm can replace classical point-based design approach, allowing more of the design to proceed concurrently. Additionally, set-based design defers detailed specifications until the implied tradeoffs are more fully understood. The next phase of the fuzzy logic assisted set-based design research is to first determine if the type-1 fuzzy logic system can be improved, by changing the definition of the agent membership functions. The second phase requires introducing uncertainty into the design space through die introduction of a type-2 modeling approach for the fuzzy logic system. Early stages of design are plagued with uncertainty due to a lack of knowledge, under-defined specifications, and incomplete tools. If more information can be conveyed, and uncertainty can be expressed within the fuzzy system, human designers will be able to make better decisions, thereby increasing the speed of the design process and increasing the probability of reaching globally optimal decisions. An investigation of the impacts of type-1 membership functions, and a type-2 fuzzy modeling approach, on the original research is presented.