Prediction of Rock Mechanical Properties Based on Numerical Simulation and Machine Learning
A novel approach for predicting rock mechanical properties,which combines accuracy and efficiency,is proposed based on finite element numerical simulation and multi-objective machine learning to address the shortcomings of traditional methods in the study of rock mechanics performance.A microscale finite element simulation of rock mechanics experiments is conducted using the continuum damage mechanics method to obtain the sample data for machine learning.Uniaxial tensile/compressive strength,elastic modulus,Poisson's ratio,and confining pressure are used as feature variables,while stress-strain matrix and damage variable matrix are used as target variables.A rock triaxial mechanical performance prediction model is established using the particle swarm-random forest algorithm.Based on this,a rock triaxial mechanical performance prediction soft is developed,which enables the intuitive display of stress-strain curves and failure patterns.This approach achieves good results in rock mechanics experimental teaching and provides effective references for engineering design and calculations.
万方数据
维普
