LAUSR

This paper intends to obtain a sub-level classification of fully weathered red sandstone rock mass during the Cretaceous period based on physical property indices. An automatic prediction system for mechanical properties of rock mass is also established in which the values of rock mass grade and mechanical properties can be returned when physical indices are given. In this study, water content, relative compaction, and fine content, which are relatively easy to measure on tunnel sites, are selected as physical property indices, while cohesion, internal friction angle, and compression modulus, which affect mechanical strength, are taken as mechanical indices. For the convenience of analysis, rock mass of different physical states is divided into 18 different combinations based on the results of lab tests and rock mass stability of each state is evaluated using a stability evaluation criterion we put forward. The numerical simulation of tunnel excavation is further carried out for determining the sub-level classification. A decision tree model is applied for further studying the relationship between physical and mechanical indices to ensure accurate prediction. The result shows that the prediction system works with high accuracy and can be used for engineering reference for fully weathered red sandstone rock mass.