Prediction of the uniaxial compressive strength of rocks from simple index tests using a random forest predictive model

Uniaxial compressive strength (UCS) is an important mechanical parameter for stability assessments in rock mass engineering. In practice, obtaining the UCS simply, accurately and economically has attracted substantial attention. In this paper, studies related to UCS estimation using indirect tests were reviewed, it was found that regression techniques and soft computing techniques were mainly used to evaluate the UCS value, and theses soft computing techniques can accurately and effectively predict the UCS. To select the proper indirect parameters to predict the UCS, statistical analysis was performed on the relationships between UCS and indirect parameters, and based on the analysis, two indirect parameters (the Schmidt hammer rebound value (L-type) and ultrasonic P-wave velocity) were deemed adequate to predict UCS. To establish the UCS predictive model, the random forest algorithm was employed, the predictive model was verified by data collected from references. To further verify the validity of the predictive model, laboratory tests were performed, and the predictive results were consistent with the measured results, thus the UCS value predictive model can be applied to the fields of rock mechanics and engineering geology.