JZUS - Journal of Zhejiang University SCIENCE

Journal of Zhejiang University SCIENCE A

Accepted manuscript available online (unedited version)

A virtual sample diffusion generation method guided by LLM-generated knowledge for enhancing information completeness and zero-shot fault diagnosis in building thermal systems

Author(s): Zhe SUN¹, Qiwei YAO¹, Ling SHI¹, Huaqiang JIN³, Yingjie XU¹, Peng YANG¹, Han XIAO¹, Dongyu CHEN⁴, PanPan ZHAO⁵, Xi SHEN^1, ²
Affiliation(s): ¹College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310023, China; more
Corresponding email(s): sx@zjut.edu.cn, Jhq@zjut.edu.cn
Key Words: Information completeness; Large language models; Virtual sample generation; Knowledge-guided; Building air conditioning systems

Share this article to： More <<< Previous Paper \|Next Paper >>>

Zhe SUN¹, Qiwei YAO¹, Ling SHI¹, Huaqiang JIN³, Yingjie XU¹, Peng YANG¹, Han XIAO¹, Dongyu CHEN⁴, PanPan ZHAO⁵, Xi SHEN^1,². A virtual sample diffusion generation method guided by LLM-generated knowledge for enhancing information completeness and zero-shot fault diagnosis in building thermal systems[J]. Journal of Zhejiang University Science A,in press.Frontiers of Information Technology & Electronic Engineering,in press.https://doi.org/10.1631/jzus.A2400560

@article{title="A virtual sample diffusion generation method guided by LLM-generated knowledge for enhancing information completeness and zero-shot fault diagnosis in building thermal systems",
author="Zhe SUN¹, Qiwei YAO¹, Ling SHI¹, Huaqiang JIN³, Yingjie XU¹, Peng YANG¹, Han XIAO¹, Dongyu CHEN⁴, PanPan ZHAO⁵, Xi SHEN^1,²",
journal="Journal of Zhejiang University Science A",
year="in press",
publisher="Zhejiang University Press & Springer",
doi="https://doi.org/10.1631/jzus.A2400560"
}

%0 Journal Article
%T A virtual sample diffusion generation method guided by LLM-generated knowledge for enhancing information completeness and zero-shot fault diagnosis in building thermal systems
%A Zhe SUN¹
%A Qiwei YAO¹
%A Ling SHI¹
%A Huaqiang JIN³
%A Yingjie XU¹
%A Peng YANG¹
%A Han XIAO¹
%A Dongyu CHEN⁴
%A PanPan ZHAO⁵
%A Xi SHEN^1
%A²
%J Journal of Zhejiang University SCIENCE A
%P
%@ 1673-565X
%D in press
%I Zhejiang University Press & Springer
doi="https://doi.org/10.1631/jzus.A2400560"

TY - JOUR
T1 - A virtual sample diffusion generation method guided by LLM-generated knowledge for enhancing information completeness and zero-shot fault diagnosis in building thermal systems
A1 - Zhe SUN¹
A1 - Qiwei YAO¹
A1 - Ling SHI¹
A1 - Huaqiang JIN³
A1 - Yingjie XU¹
A1 - Peng YANG¹
A1 - Han XIAO¹
A1 - Dongyu CHEN⁴
A1 - PanPan ZHAO⁵
A1 - Xi SHEN^{1
A1 -}²
J0 - Journal of Zhejiang University Science A
SP -
EP -
%@ 1673-565X
Y1 - in press
PB - Zhejiang University Press & Springer
ER -
doi="https://doi.org/10.1631/jzus.A2400560"

Abstract
Chinese Summary
Academic Network
Reviewer Comment

Abstract: In the era of big data, data-driven technologies are increasingly leveraged by industry to facilitate autonomous learning and intelligent decision-making. However, the challenge of �small samples in big data� emerges when datasets lack comprehensive information necessary for addressing complex scenarios, which hampers adaptability. Thus, enhancing data completeness is essential. Knowledge-guided virtual sample generation transforms domain knowledge into extensive virtual datasets, thereby reducing de-pendence on limited real samples and enabling zero-sample fault diagnosis. This study used building air conditioning systems as a case study. We innovatively used the LLM model to acquire domain knowledge for sample generation, significantly lowering knowledge acquisition costs and establishing a generalized framework for knowledge acquisition in engineering applications. This acquired knowledge guided the design of diffusion boundaries in mega-trend diffusion, while the Monte Carlo method was used to sample within the diffusion function to create information-rich virtual samples. Additionally, a noise-adding technique was introduced to enhance the information entropy of these samples, thereby improving the robustness of neural networks trained with them. Exper-imental results showed that training the diagnostic model exclusively with virtual samples achieved an accuracy of 72.80%, signifi-cantly surpassing traditional small-sample supervised learning in terms of generalization. This underscores the quality and com-pleteness of the generated virtual samples.

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference

Open peer comments: Debate/Discuss/Question/Opinion

<1>

- Go to

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference