A machine learning framework for predicting shear strength properties of rock materials.

Area of Science:

• Geotechnical Engineering
• Computational Intelligence

Background:

• Rock shear strength parameters (cohesion and internal friction angle) are vital for designing stable rock structures.
• Accurate prediction of these parameters reduces engineering time, costs, and material waste from physical testing.

Purpose of the Study:

• To develop a novel machine learning model for predicting rock shear strength characteristics.
• To enhance the accuracy and efficiency of rock strength prediction in geotechnical engineering.

Main Methods:

• Utilized a dataset of 199 Himalayan rock samples with six input variables.
• Developed and implemented the Improved Sparrow Search Algorithm (ISSA) combined with Extreme Gradient Boosting (XGBoost) - the ISSA-XGBoost model.
• Benchmarked ISSA-XGBoost against other models and performed feature importance analysis.

Main Results:

• The ISSA-XGBoost model demonstrated superior predictive accuracy, achieving R² = 0.982 for cohesion and R² = 0.932 for internal friction angle.
• Uniaxial compressive strength and P-wave velocity were identified as key predictors for cohesion.
• Density and P-wave velocity were found to be the most influential factors for internal friction angle.

Conclusions:

• The ISSA-XGBoost model offers a highly accurate and efficient method for predicting rock shear strength.
• The findings provide valuable insights into the key factors influencing rock mechanical properties, aiding in more reliable geotechnical designs.