.png){kind=icon}
.png){kind=icon}
Return
Innovation design oriented functional knowledge integration framework based on reinforcement learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.aei.2023.102122.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
According to the basic law of Design Science, new product design is based on existing design knowledge. Knowledge integration can be applied to product function design to shorten design time and improve the design quality through effective use of the existing knowledge. With the increase of the product design complexity and the number of design knowledge, it is harder and harder for traditional traversal-based algorithms to complete knowledge integration under acceptable time cost. A Reinforcement Learning (RL) based functional knowledge integration framework is proposed. The functional knowledge is represented by its input and output, and organized using a knowledge graph. The Q-network is constructed and trained for the deep Monte Carlo methodbased functional unit chain generation algorithm. The performance experiments show that comparing with the traditional searching algorithms, the RL based algorithm can provide same quality design scheme with much shorter time. The proposed algorithm is promising to realize real-time functional knowledge integration in largescale knowledge bases.
Keywords:
Functional knowledge integration
Computational design synthesis
Reinforcement learning
Knowledge graph
Journal
IF:
9.9
Papers:
4.0K
Citations:
1.7W