Tunnel construction worker safety state prediction and management system based on AHP and anomaly detection algorithm...

Area of Science:

• Civil Engineering
• Occupational Safety and Health
• Machine Learning Applications

Background:

• Tunnel construction presents significant safety risks due to complex environments and demanding conditions.
• Traditional safety analysis relying on static factors is insufficient for modern tunnel engineering.
• Worker safety is influenced by physiological indicators and internal environmental conditions.

Purpose of the Study:

• To comprehensively analyze factors impacting tunnel worker safety.
• To develop a novel safety assessment model, Tunnel-APH-AD, incorporating key identified factors.
• To enhance the accuracy and reliability of safety alerts in tunnel construction.

Main Methods:

• Utilized the Analytic Hierarchy Process (AHP) to identify seven critical safety factors.
• Employed data augmentation and four distinct machine learning models for anomaly detection.
• Applied ensemble learning techniques to aggregate model predictions for improved safety state detection.

Main Results:

• The ensemble learning model demonstrated superior performance over individual models in detecting safety states.
• Models were evaluated on out-of-distribution samples, showing robustness and generalizability.
• Experimental validation confirmed the model's interpretability, scalability, and scientific generalization.

Conclusions:

• Ensemble learning effectively enhances the accuracy and reliability of safety alerts for tunnel workers.
• The Tunnel-APH-AD model offers a scientifically generalized application of machine learning for proactive safety management.
• Findings contribute to advancing safety practices in tunnel engineering, ensuring worker well-being.