Abstract: In geological engineering, dominant partitioning of discontinuities of rock mass is a fundamental work for mechanical and hydraulic behaviors analysis of rock mass. In common methods, only two characteristic parameters (dip and dip angle) are selected to identify discontinuity sets. Trace length, joint opening and surface morphology of discontinuities also influence the mechanical behaviors of rock mass. Therefore, a novel scheme for discontinuities classification is proposed based on multivariate parameters and artificial bee colony algorithm. The sum of deviations squares of the entire sample data is taken as an objective function. The artificial bee colony algorithm is used to search the optimal solution which makes the objective function achieve the minimum value. The optimal solution is the cluster centers. On this basis, a mathematical model is established. At the same time, the boundaries between different sets are determined automatically. The validation of the novel scheme is proved by results based on artificial data. The calculation precision of the method is satisfactory. Finally, the proposed method is applied to multivariate parameter dominant partitioning of discontinuities of rock mass at Songta dam site on the NuJiang River. The classification result verifies that the method is efficient and practical.

Key words: rock mechanics, discontinuities in rock mass, data partitioning, clustering method, artificial bee colony algorithm