A Knowledge Retrieval Method for Layout Design of Nuclear Power Reactor Coolant System Based on Hypernetwork and Deep Structured Semantic Model

Bin Wang,Bingtao Hu,Zhifeng Zhang,Yixiong Feng,Jianrong Tan

DOI: https://doi.org/10.1115/icone29-91827

2022-01-01

Abstract:Abstract The layout design of nuclear power reactor coolant system requires a large amount of knowledge that satisfies many disciplines, which will waste designers a lot of time retrieving relevant knowledge in the design process. In order to obtain the knowledge efficiently and accurately, a knowledge retrieval method based on hypernetwork and deep structured semantic model (DSSM) was proposed. The knowledge hypernetwork model consisted of a designer sub-network, a design task subnetwork, and a design knowledge resource sub-network. Nodes in each sub-network and between different sub-networks were connected through special edges, which represented correlation degree information. Then an improved DSSM model was used to evaluate relevance at the semantic level between user query information and knowledge elements in hypernetwork. Correlation scores will be obtained based on relevance at the semantic level, and knowledge elements with lower scores will be removed during the process. Finally, the Bayesian method was used to calculate the knowledge recommendation probability to obtain the most relevant knowledge retrieval results. The knowledge retrieval results were sorted from high to low according to the calculated probability. A case study conducted in this work showed that the proposed approach was effective in capturing relevance at the semantic level and supporting efficient and accurate knowledge retrieval services.