Data Imputation of Soil Pressure on Shield Tunnel Lining Based on Random Forest Model

Area of Science:

• Geotechnical Engineering
• Data Science
• Structural Health Monitoring

Background:

• Shield tunneling projects generate vast amounts of monitoring data for safety.
• Missing data in these datasets can compromise engineering project safety.
• Reliable data imputation is crucial for accurate analysis and decision-making.

Purpose of the Study:

• To develop and evaluate a missing data imputation technique for shield tunnel monitoring data.
• To assess the effectiveness of the Random Forest (RF) algorithm for this purpose.
• To compare RF imputation with traditional methods like median and mean.

Main Methods:

• Utilizing the Random Forest (RF) algorithm for missing data imputation.
• Optimizing RF parameters (decision trees, max depth, features) by minimizing Mean Squared Error (MSE).
• Creating artificial datasets with 20%, 40%, and 60% missing soil pressure data for testing.

Main Results:

• The proposed RF method demonstrated the smallest imputation error compared to median and mean imputation.
• MSE increased with higher missing data rates for all methods.
• The RF method showed a significant performance advantage, with error differences up to 70%.

Conclusions:

• The Random Forest method is a suitable and effective technique for imputing missing monitoring data in shield tunnels.
• Accurate data imputation enhances the reliability of structural health monitoring in underground engineering.
• The RF approach offers improved safety assurance for shield tunneling projects.