Stability Risk Assessment of Underground Rock Pillars Using Logistic Model Trees

Area of Science:

• Geotechnical Engineering
• Rock Mechanics
• Computational Intelligence

Background:

• Underground pillars are critical for structural support in subterranean environments.
• Pillar instability can lead to catastrophic failures like roof collapse and rock bursts, necessitating accurate stability prediction.

Purpose of the Study:

• To develop and validate a novel approach for predicting underground pillar stability using Logistic Model Trees (LMT).
• To assess the performance of LMT against other machine learning techniques for this application.

Main Methods:

• Application of Logistic Model Trees (LMT) incorporating seven key parameters: pillar width, height, width-to-height ratio, uniaxial compressive strength, average pillar stress, underground depth, and bord width.
• Training LMT models on rock and coal pillar case histories using the LogitBoost algorithm.
• Validation through 10-fold cross-validation and testing on new case histories.

Main Results:

• The proposed LMT models demonstrated superior accuracy in predicting underground pillar stability compared to previously used machine learning methods.
• Sensitivity analysis identified average pillar stress and the pillar width-to-height ratio as the most significant parameters influencing stability predictions.

Conclusions:

• Logistic Model Trees offer a highly accurate and effective method for predicting underground pillar stability in both rock and coal formations.
• The findings provide valuable insights for geotechnical engineers in assessing and mitigating risks associated with underground excavations.