Research on a Self-enhancing Maintenance Knowledge Base Method based on Fault Causal Association Probabilities

Linjiao Hou; Bin Liu

doi:10.54691/wdwdnq69

Authors

Linjiao Hou National-recognized Enterprise Technology Center, Sunward Equipment Co., Ltd., Changsha, Hunan 410100, China
Bin Liu National-recognized Enterprise Technology Center, Sunward Equipment Co., Ltd., Changsha, Hunan 410100, China

DOI:

https://doi.org/10.54691/wdwdnq69

Keywords:

Intelligent Maintenance; Fault Diagnosis; Knowledge Base; High-incidence Fault Probability Model; Large Language Model; Closed-loop Learning.

Abstract

To address the problems of over-reliance on expert experience, knowledge fragmentation, and low diagnostic efficiency in the intelligent maintenance of widely distributed complex equipment, this paper proposes an intelligent maintenance method based on a self-enhancing dynamic knowledge base. First, a structured integration framework for multi-source heterogeneous maintenance knowledge is established to achieve the systematic organization of fault cases and service records. Second, a "high-incidence fault probability model" based on statistical analysis of historical fault data is proposed, forming a priority-based diagnostic strategy of “high-incidence faults first, low-incidence faults later . Furthermore, a large language model is integrated to enable natural language interaction and accurate knowledge retrieval. In addition, a knowledge credibility evolution and closed-loop optimization mechanism based on maintenance log feedback is introduced, driving the knowledge base to evolve from static storage into a dynamic self-enhancing system. On this basis, a progressive knowledge architecture consisting of “Level 1: authoritative knowledge, Level 2: knowledge to be verified, and Level 3: AI-derived knowledge”, together with its self-enhancing interaction workflow is designed. The proposed method is further validated through practical after-sales maintenance scenarios covering three stages: work-order submission, fault analysis, and case accumulation. The proposed method is further validated through practical after-sales maintenance scenarios covering three stages: : work-order submission, fault analysis, and case accumulation . The results show that the proposed method can significantly improves the overall efficiency of fault description, localization, and troubleshooting, thereby providing a systematic solution with continuous self-enhancing capability for the intelligent maintenance of complex electro-mechanical-hydraulic integrated equipment.

Downloads

Download data is not yet available.

References

[1] Dai, A. S., Yang, D., Luo, Z., & He, S. G. (2025). The digital and intelligent operation and maintenance mechanism of high-end equipment driven by information technology – A multi-case study. Industrial Engineering and Management, 30(5), 206–218.

[2] Liu, B. M. (2025). Intelligent scheduling of remote communication and control for electromechanical power distribution. Smart China, 28(12), 44–45.

[3] Min, Y., Wang, Y. Q., Ma, X., & Dong, Y. N. (n.d.). Intelligent fault diagnosis in hybrid chemical processes under limited samples based on multi-feature fusion learning. Computers & Chemical Engineering, 207(1), 960–968.

[4] Guan, H. Y. (2025). Research on construction and optimization of intelligent diagnosis model for remote oil pipeline leakage. Computer Knowledge and Technology, 21(33), 106–108.

[5] Zhang, X. J. (2025). Remote operation and maintenance and fault diagnosis methods for smart manufacturing equipment driven by digital twins. Electronic Technology, 54(11), 13–15.

[6] Sun, S. J. (2025). Analysis of fault diagnosis and intelligent maintenance technology for electromechanical equipment. Electronic Technology, 54(11), 166–167.

[7] Kumar, L., & Khedlekar, U. K. (2025). A hybrid AI and blockchain framework for intelligent supply chain optimization with reinforcement learning-based inventory management, stochastic demand forecasting, and sustainable supplier selection. Unidentified Journal, 30(12), 1–29.

[8] Li, Z. Q., & Zuo, T. T. (2025). Integrated model of intelligent predictive inventory management considering data fluctuations. Computer Simulation, 42(6), 220–224, 378.