In this paper, an advanced back analysis methodology is proposed so as to identify inhomogeneous behavior of rock masses, which is suitable for a large-scale cavern excavation into complex rock masses. This allows us to estimate the extent and characteristics of inhomogeneous zones by iterative calculations. It employs the reduction factor of elastic modulus Ai by strain outputs from back analysis and also the back analysis formulation to take initial stress parameters as well as non-elastic strains as unknowns with minimum norm solution. This method is applied both to numerical experiments and to an actual cavern construction problem and the results suggest that more reasonable strain distributions with proper reduction factors Ai can be appropriately estimated by this proposed method than those by others methods.


Du e to inherent complex conditions of jointed rock masses, an observational construction principle based on field measurements has been employed for the purpose to properly construct underground caverns. Back analysis is clearly one key tool, which allows us to estimate rock mass deformability or large-scale initial stresses from measured displacements. Sakurai and Takeuchi (1983) initially developed the back analysis method assuming a linear elastic medium and it has been successfully applied to various tunnel constructions (called DBAP here after). Then, Sakurai et al. (1994a) proposed the upgraded procedure, which can take into non- elastic strains as unknowns of back analysis (N-DBAP hereafter). Although this is more advanced than the original, the assumption of overall rock mass homogeneity around a cavern remains and there is still a need on back analysis to appropriately identify in homogeneity, especially for a large-scale underground cavern. In this paper, the formulation of this advanced back analysis is briefly presented (I-N-DBAP hereafter) and then its applications follow as case studies.

This content is only available via PDF.
You can access this article if you purchase or spend a download.