Research on strength prediction of crack rock mass based on random forest algorithm

Rapid and accurate estimation of rock strength is of great practical significance for the stability of rock engineering. This article is based on the results of triaxial compression tests on crack rocks and comprehensively considers the interrelationships between crack geometric characteristics, confining pressure, and physical and mechanical parameters. A database of rock strength characteristic parameters from an experimental perspective is constructed, and a quantitative prediction model for the strength of crack rock masses under the influence of multiple characteristic parameters is established using the random forest algorithm. The results show that the accuracy of predicting rock strength in categories 2 of 21–30 MPa, 3 of 31–40 MPa, 4 of 41–50 MPa, and 5 of 51–60 MPa is 100%, and the accuracy of predicting rock strength in categories 1 of 11–20 MPa is 82% and in categories 0 of 0–10 MPa is 74%; through the study of parameter correlation and quantitative analysis with strength, the selected parameters are reliable and have a certain degree of correlation, and the influence mechanism of characteristic parameters on rock strength has been explored; the calculation results of parameter classification show that confining pressure and crack length are the key factors affecting rock strength. The importance of other parameters is ranked in the order of longitudinal wave velocity > Poisson’s ratio > saturated water content > porosity > saturated quality > dry quality > saturated density > crack dip angle > crack number > crack penetration. This provides an effective approach to consider the randomness and correlation of rock strength parameters, providing a new idea for rock mechanics research.