A function-behavior mapping approach for product conceptual design inspired by memory mechanism

Conceptual design is a pivotal stage for new product development that relies more on designers to solve open-ended and ill-defined problems. Situated function-behavior-structure ontology is an acknowledged method to facilitate conceptual design in a goal-oriented way. However, it depends on the subjective cognition abilities of designers, which are influenced by limited memory and reasoning capacities. Developing computer-aided methods grounded in this ontology holds significant promise in enhancing designers' cognitive abilities. This study delves into the function-behavior (F-B) mapping process. It explores the effect of working memory and long-term memory on design cognition and introduces a memory-inspired reinforcement learning framework for F-B mapping. The Markov decision process is then adopted to formalize F-B mapping while motivation-driven Q learning is employed by the design agent to learn knowledge from historical design cases. The learned state-action value matrix can be applied to guide the designer in selecting feasible behaviors for the specific function requirement. The proposed approach empowers design agents with self-learning and self-evolving capacities. A case study on the F-B mapping of a traction system is conducted to illustrate the feasibility and practicability of the proposed approach.