Knowledge graph-based multi-granularity tacit design knowledge reuse for product design

Abstract

Decomposing product design into multiple granularities is an effective strategy to implement complex product design. Tacit design knowledge (TDK), adhering to designers’ experience and intuition, plays an essential role in the product design at various granularities, but it is difficult to acquire and reuse due to its implicity. Aiming at this issue, a knowledge graph-based multi-granularity TDK reuse framework for product design is presented. First, the design issue-solving oriented knowledge graph (DSKG) model is employed to indirectly capture TDK, where the design result data are represented as design solution nodes and the design issues that refine design solution are represented as relation edges. An improved ontological concept model is designed as the schema of the DSKG to represent the multi-granularity design knowledge on functions, behaviours, and structures of products. Afterwards, the framework supports two design modes, i.e., top–down design and bottom–up design, where the captured TDK is reused to create the design context and predict design solutions. DSKG embedding is employed for design solution prediction in both design modes, while the design context creation is achieved distinctively. In top–down design, the design issue-solving process starts from the design requirements and then gradually downward refines the design solutions at different granularities with breadth-first strategy, which naturally creates the continuous design context, until moving to the detailed design. In bottom–up design, the target design solution is commonly isolated at specific fine-grained granularity. This paper presents a novel approach to constructing design context based on path-ranking algorithm. Path-ranking algorithm can identify the design solution influence paths to build connections between the design requirements and the isolated design solution and thus construct the design context. Finally, a case study for stamping die design is implemented, which illustrates the feasibility and effectiveness of the framework. The comparisons with the state-of-the-art design methods are also made to demonstrate the merits of the presented framework.