Related Experiment Video
Updated: May 15, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.6K
Automatic construction of risk transmission network about subway construction based on deep learning models
Yanxiang Liang1, Na Xu2, Hong Chang3
1School of Mechanics and Civil Engineering, China University of Mining and Technology, Xuzhou, 221116, China.
Scientific Reports
|May 11, 2025
Summary
This study introduces advanced AI models to automatically identify subway construction safety risks and their causal links from accident reports. The findings enhance safety management by revealing risk patterns and interrelationships.
Area of Science:
- Artificial Intelligence
- Natural Language Processing
- Civil Engineering Safety
Background:
- Conventional subway construction safety risk management relies heavily on expert experience.
- Existing methods struggle to identify complex relationships between risk factors and events within accident texts.
- This limitation hinders effective guidance for subway safety risk management.
Purpose of the Study:
- To develop domain-specific AI models for automated extraction of safety risks and causal relationships from subway construction accident texts.
- To improve the identification and understanding of safety hazards and their interconnections in subway projects.
Main Methods:
- A Bidirectional Long Short-Term Memory Network with Conditional Random Fields (BiLSTM-CRF) was used for named entity recognition (NER) of safety hazards.
- Convolutional Neural Networks (CNN) were employed for domain-specific entity causal relation extraction.
- A domain dictionary was developed for entity normalization, leading to the creation of a causal relationship database.
Main Results:
- The Metro Construction Safety Risk Named Entity Recognition Model (MCSR-NER-Model) achieved >77% in precision, recall, and F1-scores.
- The Metro Construction Safety Risk Domain Entity Causal Relationship Extraction Model (MCSR-CE-Model) demonstrated excellent performance with 98.96% accuracy, recall, and F1-scores.
- 533 domain entity causal relation triplets were extracted, forming a complex network and database of subway construction risks.
Conclusions:
- The developed AI models effectively convert accident texts into a structured causal chain of 'safety risk factors to risk events'.
- The research provides detailed categorization of safety risks and events, revealing interrelationships and historical patterns.
- The created database supports project managers in making informed safety risk management decisions.
