Machine learning approaches for predicting the construction time of drill-and-blast tunnels

Area of Science:

• Civil Engineering
• Geotechnical Engineering
• Computational Science

Background:

• Accurate prediction of construction duration is vital for efficient project management in tunneling.
• Drill-and-blast tunneling involves complex geological and operational factors influencing project timelines.
• Existing methods for duration estimation often lack the precision required for dynamic project adjustments.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting drill-and-blast tunnel construction duration.
• To identify key parameters significantly impacting tunnel construction timelines.
• To create a practical tool for real-time duration estimation and project monitoring.

Main Methods:

• Compilation of a comprehensive dataset of 500 data points from eight tunnels, including 20 diverse parameters.
• Feature selection to identify 17 crucial parameters for ML model training, focusing on overbreak and cross-section.
• Application and hyperparameter tuning of various ML models, including Gaussian process regression.
• Development of a machine learning-based graphical user interface (GUI) for practical application.

Main Results:

• Identification of overbreak and tunnel cross-section as highly influential parameters on construction duration.
• Gaussian process regression demonstrated superior performance, achieving an average R-squared of 0.89.
• The developed ML-based GUI enables accurate, real-time duration predictions and dynamic updates.
• Hyperparameter tuning significantly enhanced the predictive accuracy of the ML models.

Conclusions:

• Machine learning techniques, particularly Gaussian process regression, are highly effective for predicting drill-and-blast tunnel construction duration.
• Key parameters like overbreak and cross-section are critical for accurate time estimations.
• The developed GUI offers a valuable, practical tool for enhancing project management and real-time decision-making in tunneling projects.