Computational Modeling of the Dynamics of Fluid-Filled Porous Media with Application to Road and Railway Structures

Licentiatavhandling, 2007

The dynamics of fluid-saturated porous media is a challenging topic with applications in many engineering fields like, e.g., road and railway mechanics. The complexity of the problem requires high computational effort when performing simulations of the coupled dynamic deformation and fluid flow in a saturated porous medium. Different possibilities for reducing the computational cost are discussed in the thesis. The porous medium is modeled as a binary mixture of a solid matrix phase and a fluid phase, whereby the thermodynamically consistent so-called Porous Media Theory is adopted. Reduction of the computational effort is achieved through simplifications of the continuum (physical) model as well as a novel reformulation of the variational format. As to the adopted modeling, different simplifying approximations for the relative fluid acceleration are compared. From the numerical examples, it is concluded that the contribution from the convective part of the relative fluid acceleration may be neglected without significant loss of accuracy; however, neglecting the complete relative fluid acceleration may lead to detrimental loss of accuracy. A problem of significant engineering importance, to which the developed model has been applied, is the coupled dynamic deformation and water flow in soils subjected to over-rolling of a high-speed train. As to the choice of computational method for the fully dynamic response, a novel two-field variational format, with the seepage velocity treated as a "local" field, was developed and compared to the more commonly used three-field format. This leads to a substantial reduction of the computational effort for given accuracy. A pressure gradient dependent permeability coefficient was introduced in the constitutive model for the seepage velocity. This was done essentially to illustrate the effectiveness of algorithmic structure for a nonlinear model.