Simulation of cross-correlated non-Gaussian random fields for layered rock mass mechanical parameters

Abstract

In order to represent the spatial variability of layered rock mass mechanical parameters in Finite Difference Method (FDM) grids, an approach based on the numerical simulation of cross-correlated non-Gaussian random fields was proposed in this paper. First, the discretization of a Gaussian random field was performed using the Karhunen-Loève (K-L) expansion. Then, the Gaussian random field for layered rock mass parameter was built by geostatistics and the random field size calculation. Next, the cross-correlated non-Gaussian random fields were simulated using Nataf transformation and isoprobabilistic transformation methods. Furthermore, a grid model coupled to the random fields was finally built using regroupment and element traversal methods. The engineering application showed that the spatial variability, autocorrelation, cross-correlation and non-Gaussianity described by the random fields agreed closely with the statistical characteristics. The correspondence between the random fields and grid model verified that the random parameters in the grid model can represent the anisotropy of layered rock mass parameters. This study provides a computational basis for quantitative risk assessment during tunnel excavation in layered rock mass.