Abstract: Taking the actual rock mass joint information from a gold mine in Shandong Province as samples， a method is studied to determine the lower limit of the statistical joints trace length in underground engineering rock mass by cluster analysis and its evaluating indicators with the full-scale test. Starting from the definition and characteristics of the lower limit， the DBI (Davies-Bould in index) is used to divide dominant joint sets and determine the best setting scheme of dominant joints. The joint set can become the dominant one if the discrete coefficient k is greater than 20， and the discrete coefficient k with its probability distribution index n together is used to analyze and determine the lower limit value. Furthermore， the method of determining the lower limit value of the statistical joints trace length under different rock mass conditions is developed in order to eliminate random joints and optimize the sample data of rock mass joints， thus getting spatial distribution characteristics of rock mass joints.

Key words: underground engineering rock mass, joint trace length, statistical lower limit, random joint, cluster evaluation index, joint data optimization