Block failure is one of the most common types of failure in underground excavations. The volumes of possible unstable blocks are evaluated using different methods such as kinematic analysis and Discrete Fracture Network (DFN) methods. The forces acting on blocks also are estimated using different analytical and numerical methods such as Limit Equilibrium and Distinct Element Method (DEM) approaches. Probabilistic Kinematics analysis is based on Monte Carlo simulations from the distribution of geometry parameters of fractures. In order to evaluate the rock block volumes around an excavation using the stochastic DFN approach, a large number of DFN realizations should be generated. The problem in DFN is the calibration for making a representative realization of DFN. Using some statistical tests such as William-Watson (W-W) test, the most compatible DFN models with mapped fracture in the field are selected for block stability analysis. The purpose of this study is to evaluate the unstable block volumes and required support system such as bolt patterns to estimate the model uncertainty of Probabilistic Kinematics Limit Equilibrium (PKLE) compared with Discrete Fracture Network-Distinct Element Method (DFN-DEM) analysis which provides a closer representation of reality. The basic geometry parameters of fractures and also mechanical properties of rocks are taken from a large excavated cavern in southeast of Sweden . A FORTRAN code was developed based on the William-Watson’s (W-W) statistical test. The confirmed DFN's by the W-W test was assigned to the DEM model as the geometric basis for the block stability analysis using UDEC code. For the PKLE analysis the commercial software Unwedge, which is based on key block theory has been used. Model uncertainty is calculated using mean difference and variance ratio of two normal distribution functions. The results show that the PKLE underestimates the unstable block volumes. The calculated mean value of unstable block volumes using PKLE is much smaller than DFN-DEM simulation results. Although the variance ratio of two normal distribution of required total bolt length from PKLE and DFN-DEM methods are equal, the mean values of two distributions are significantly different which means again that using PKLE method cause a large degree of uncertainty for block stability analysis around a cavern.
