Research on Railway Survey and Design Review Technology Based on Knowledge Graph.

Saved in:

Bibliographic Details
Title:	Research on Railway Survey and Design Review Technology Based on Knowledge Graph.
Authors:	Liu, Beisheng¹ 295168157@qq.com, Xu, Jing² 1540606195@qq.com, Lyu, Xiangru³ 347098260@qq.com, Li, Hui⁴ 11373692@qq.com, Li, Zhaoxiang⁵ 2330559390@qq.com
Source:	IAENG International Journal of Applied Mathematics. Jun2026, Vol. 56 Issue 6, p2222-2228. 7p.
Subjects:	Knowledge graphs, Inspection & review, Deep learning, Digital technology, Surveying (Engineering), Standardization, Model validation
Abstract:	Traditional railway survey and design review processes are plagued by incomplete data, non-standardized formats, and fragmented review criteria, which severely reduce review efficiency and accuracy. This study focuses on the digital delivery and intelligent review of railway survey and design outcomes, and proposes a systematic technical framework integrating digital delivery standardization and knowledge graph-based intelligent review. First, a four-dimensional data catalog is established, and unified digital delivery standards are formulated. Second, a BERT+Bi-LSTM+CRF hybrid deep learning model is proposed for the construction of railway design review knowledge graph; the model is optimized for railway through domain-specific pre-training, network parameter adjustment and custom label transition rules, and realizes the extraction of "entity-constraint-parameter" triples from specification provisions. Multi-source knowledge fusion is further achieved via a weighted entity alignment algorithm, credibility-based conflict resolution and confidence-filtered relationship mapping, transforming scattered review rules into a structured and reusable knowledge network. Finally, comprehensive validation experiments were conducted using preliminary design data of 3 railway projects, with multi-dimensional evaluation indicators and ablation experiments to verify the model performance. Results show that the model achieves an F1-score of over 90% in specialties with clear review rules and standardized data; for specialties with partial unstructured data and ambiguous rules, the F1-score is 0.87-0.88; for the Alignment specialty with a large number of unstructured data and subjective judgments, the F1-score is 0.799. The proposed method is 21 times more efficient than manual review and outperforms BIM+Rule Engine and Bi-LSTM single model in both accuracy and efficiency, providing a complete technical solution for the automation and digitalization of railway survey and design review. [ABSTRACT FROM AUTHOR]
	Copyright of IAENG International Journal of Applied Mathematics is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Traditional railway survey and design review processes are plagued by incomplete data, non-standardized formats, and fragmented review criteria, which severely reduce review efficiency and accuracy. This study focuses on the digital delivery and intelligent review of railway survey and design outcomes, and proposes a systematic technical framework integrating digital delivery standardization and knowledge graph-based intelligent review. First, a four-dimensional data catalog is established, and unified digital delivery standards are formulated. Second, a BERT+Bi-LSTM+CRF hybrid deep learning model is proposed for the construction of railway design review knowledge graph; the model is optimized for railway through domain-specific pre-training, network parameter adjustment and custom label transition rules, and realizes the extraction of "entity-constraint-parameter" triples from specification provisions. Multi-source knowledge fusion is further achieved via a weighted entity alignment algorithm, credibility-based conflict resolution and confidence-filtered relationship mapping, transforming scattered review rules into a structured and reusable knowledge network. Finally, comprehensive validation experiments were conducted using preliminary design data of 3 railway projects, with multi-dimensional evaluation indicators and ablation experiments to verify the model performance. Results show that the model achieves an F1-score of over 90% in specialties with clear review rules and standardized data; for specialties with partial unstructured data and ambiguous rules, the F1-score is 0.87-0.88; for the Alignment specialty with a large number of unstructured data and subjective judgments, the F1-score is 0.799. The proposed method is 21 times more efficient than manual review and outperforms BIM+Rule Engine and Bi-LSTM single model in both accuracy and efficiency, providing a complete technical solution for the automation and digitalization of railway survey and design review. [ABSTRACT FROM AUTHOR]
ISSN:	19929978