LightGBM integration with modified data balancing and whale optimization algorithm for rock mass classification.

Area of Science:

• Geotechnical Engineering
• Artificial Intelligence in Civil Engineering
• Rock Mechanics

Background:

• Accurate rock mass classification is vital for intelligent tunnel-boring machine (TBM) operations.
• Geological variability and complexity pose significant challenges to traditional rock mass classification methods.
• Existing methods often struggle with minority rock mass classes, impacting overall prediction accuracy.

Purpose of the Study:

• To develop an innovative predictive model for accurate rock mass classification in TBM tunneling.
• To improve the prediction accuracy, particularly for minority rock mass classes (II, IV, V).
• To enhance the efficiency and intelligence of TBM construction through improved geological understanding.

Main Methods:

• Integration of an improved Emperor Penguin Optimization Algorithm (IEWOA) with a novel parameter 'l' and sine functions for optimized search.
• Application of a minority class technique enhanced with a random walk strategy (MCT-RW) to expand minority class boundaries.
• Utilizing the Light Gradient Boosting Machine (LightGBM) for the classification task.

Main Results:

• The proposed IEWOA-LightGBM model achieved a superior accuracy of 94.74% in rock mass classification.
• Significant improvements in recall and F1-score were observed for minority rock mass classes.
• The MCT-RW strategy effectively extended the boundaries of challenging rock classes.

Conclusions:

• The IEWOA-LightGBM model offers a robust and accurate solution for rock mass classification in TBM tunneling.
• This approach significantly enhances the intelligent operation and efficiency of TBM construction.
• The study provides a valuable contribution to addressing key challenges in underground engineering.