A cognitive analysis-based key concepts derivation approach for product design

In the development of a new product, concept generation is a crucial stage, as it determines much of the cost of the product lifetime. The process of generating concepts that meet customer requirements, however, can be challenging due to incomplete information, chaotic design constraints, and complicated design iterations. Deep learning-based pretrained language models can be a potential solution to overcome these limitations. But current methods are limited by the gap between the language used in design-related documents such as patent and the casual spoken language as the latter reflects the thinking process of concept generation. In this paper, we propose a key concepts derivation approach based on cognitive analysis to address challenges in product design and to effectively incorporate prior knowledge. This approach leverages a pretrained language model to extract information from multiple sources of textual data and represents design thinking in a vector space. We conduct a Design-by-Analogy based verbal protocol analysis experiment for high-speed elevators to collect verbal data and establish a design thinking corpus. Based on this, our approach is employed for key concepts derivation in product design and explore how solutions are generated to realize specific customer-required functionality. By analyzing the types of design stimuli, the effectiveness of our approach in deriving key design concepts are demonstrated with respect to novelty and feasibility.