ISSN 0862-5468 (Print), ISSN 1804-5847 (online) 

Ceramics-Silikáty 69, (3) 389 - 403 (2025)


PERFORMANCE EVALUATION OF DEEPSEEK-R1 IN UNDERSTANDING AND DESIGNING Al₂O₃ POROUS CERAMICS WITH THE SACRIFICIAL BIOMASS TEMPLATE METHOD
 
Sun Chaowei 1, Du Danfeng 1, Liu Zexin 1, Xu Guoxu 2, Hao Zewei 1, Guo Xiurong 1
 
1 College of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin, P. R., China.
2 College of Civil Engineering and Transportation, Northeast Forestry University, Harbin, P. R., China.

Keywords: Porous ceramic, Artificial intelligence, DeepSeek-R1, Natural language processing
 

DeepSeek has demonstrated preliminary success in multidisciplinary applications over recent months. This paper investigates the level of understanding of DeepSeek-R1 on Al₂O₃ porous ceramics from the sacrificial biomass template method (APCBT) and provides a preliminary study of its application to porous ceramics, including wood ceramics, porous ceramics and APCBT ceramics. The study was initiated with 48 questions of increasing complexity, the purpose of which was to ascertain the readability and accuracy of DeepSeek’s responses to APCBT. Based on this, the study explored the feasibility of using DeepSeek to identify and analyse the APCBT material mix ratio design, and filtration performance analysis. The results showed that DeepSeek-R1 is highly proficient in acquiring fundamental knowledge of APCBT materials, a capability validated through an interactive human-computer model, for analysing APCBT precursors, formulating preliminary material compositions, and providing feedback on filtration performance. However, it was also found that DeepSeek-R1 has limitations in understanding the complex preparation design and mechanics of APCBT and lacks the ability to synthesise and integrate interdisciplinary knowledge, sometimes resulting in arbitrary and reluctant responses. These findings emphasise the need for further improvements in the algorithmic complexity and database comprehensiveness. The paper concludes by predicting the future applicability of DeepSeek-R1 in the field of porous ceramic materials, pointing the way for subsequent research and development.


PDF (4.2 MB)

doi: 10.13168/cs.2025.0023
 
 
Licence Creative Commons © 2015 - 2025
Institute of Rock Structure and Mechanics of the CAS & University of Chemistry and Technology, Prague
Webmaster | Journal Contact