Sensitivity analyses on the influence of constitutive parameters on the numerical simulation of the behavior of a cavern in rock salt

Nowadays, due to facing the limitation of fossil fuel resources more attention is drawn to the use of renewable energy sources. The amount of electrical-based energy generated by these new resources fluctuates regularly. Due to this, the availability of adequate energy storage systems is an important issue to balance out the renewable energy's demand-generation process. One option for energy storage is the storage of pressurized air or hydrogen in caverns located in the subsoil. Rock salt has adequate properties like low permeability and self healing behavior to host such caverns. For designing caverns in rock salt, several constitutive models have been developed describing rock salt behavior. Some of these material models developed based on rheological concepts and some more advanced ones are defined on the basis of micro structural behavior of rock salt. For the application of these models to real projects, a number of constitutive parameters have to be determined. Basis for this determination is a limited number small scale experiments. A more accurate knowledge about these parameters and their influence on the results of calculations seems essential for transferring them to real scale and practical design. This paper presents results of a numerical study modeling the behavior of a typical salt cavern using a Finite Element (FE) code. A visco-elastic constitutive model has been used which is able to describe the long term behavior of rock salt (creep). For performing global sensitivity analysis on different parameters of the constitutive model, a metamodel has been generated. The metamodel allows to reduce the number of calculations with the FE model which are needed for sensitivity analysis significantly. In sake of evaluating the behavior of the cavern, related displacements around the cavern's wall have been assessed along time. As a result, the sensitivity index of each related parameter has been evaluated. This study shows the ability of global sensitivity analysis methods in determining the efficiency and importance level of initial parameters of constitutive models.