Abstract: It is an important work in digital mine research to construct the three-dimensional （3D） geological models of strata based on multi-source data of mines，and these models will facilitate the mine resource assessment and production planning process.The processing of discrete sampling of non-uniform point data by interpolation method is the key to accurately construct the strata model.However，the interpolation methods commonly used in geological modeling do not consider the non-uniformity of the data.Based on the multi-scale CSRBFs interpolation method，a 3D stratum modeling process that can employ non-uniform data is proposed. The process is divided into three steps.Firstly，non-uniform geological discrete points are extracted from multi-source data.Then，the multi-scale CSRBFs interpolation is used to construct the stratum surface model，including four steps:①sample points and unit normal vectors acquisition；②construction of hierarchical point sets by octree；③ multi-scale CSRBFs method is used to interpolate layer by layer from coarse to fine on the hierarchical point sets；④visualization of the implicit surface of the stratum.Finally，the proposed fault modeling method is used to construct normal，reverse and cross-fault models on the continuous stratum surface.The stduy results show that the average error obtained by multi-scale CSRBFs interpolation method （Multi-scale CSRBF，MsCSRBF） through ten-fold cross-validation is 0.943 m，which is lower than the error obtained by inverse distance weighted （IDW） and ordinary kriging （OK）.And the MsCSRBF method is also more time-efficient than the IDW and OK interpolation methods.The above stduy results further indicated that multi-scale CSRBF interpolation method proposed in this paper can carry out high-precision and rapid reconstruction of the formation surface model for non-uniform data.

Key words: Digital mine, Stratum modeling, Multi-source data, Uneven distributed data, Radial basis function, Fault modeling