Intelligent calibration method for microscopic parameters in the discrete element method based on ensemble learning.

Area of Science:

• Geotechnical Engineering
• Computational Mechanics
• Machine Learning Applications

Background:

• The Block Discrete Element Method (BDEM) is crucial for modeling fractured rock masses.
• Accurate BDEM simulations require precise microscopic parameters, which are difficult to obtain directly from macroscopic tests.
• Traditional calibration methods are inefficient and computationally intensive.

Purpose of the Study:

• To develop an efficient and accurate method for calibrating microscopic parameters in BDEM simulations.
• To establish a correlation between microscopic parameters and macroscopic rock behavior.
• To validate the proposed method against experimental data.

Main Methods:

• Random generation of microscopic parameters for discrete block elements.
• Development and validation of computational models for various rock mechanics tests (uniaxial compression, Brazilian splitting, triaxial compression).
• Construction of a macroscopic-microscopic parameter dataset and correlation analysis.
• Application and optimization of a Stacking ensemble learning model.

Main Results:

• The Stacking ensemble model demonstrated high accuracy in predicting macroscopic rock properties.
• Achieved low prediction errors for uniaxial compressive strength (0.6%), elastic modulus (6.6%), indirect tensile strength (10.6%), friction angle (8.6%), and cohesion (5.1%).
• The model's simulation results closely matched experimental values, confirming its reliability.

Conclusions:

• The proposed Stacking ensemble learning method provides a highly accurate and reliable approach for predicting discrete element microscopic parameters.
• This method significantly improves the efficiency of BDEM calibration compared to traditional techniques.
• The findings offer valuable support for practical engineering applications involving fractured rock masses.