Abstract: In view of the nonlinear characteristics of joints due to the influence of roughness and compressive strength of rock mass joints， the traditional limit equilibrium method uses mean to characterize the physical and mechanical indexes of rock mass， so that the evaluation conclusion can not fully take into account the discrete nature of the indexes. A slope failure probability analysis method based on Barton-Bandis nonlinear failure criterion is proposed. Through field geological survey， the contour curve and rebound value of rock mass joints are obtained. The multi-group sample data of roughness coefficient JRC， compressive strength JCS and basic friction angle [?]b of joints are obtained after processing and calculating， and the probability distribution of parameters' sample data is verified. Based on the Barton-Bandis criterion， the stability analysis model of Baiyangmiao Mine slope is established， the multi- group stability coefficients of the slope are calculated by Morgenstern-Price method， and its statistical analysis is carried out by Monte Carlo method to determine the failure probability of the slope， and compared with the calculation results of the traditional limit equilibrium method. The study results show that the method of slope failure probability analysis based on Barton-Bandis criterion fully takes into account the nonlinear characteristics of shear strength of rock mass joints and the dispersion of shear strength parameters in slope stability analysis， and is an effective method for stability analysis of rock slope.

Key words: slope stability, rock mass joint, Monte Carlo method, failure probability, Barton-Bandis criterion