Yield design for porous media subjected to unconfined flow: construction of approximate pressure fields

We consider the stability of a porous medium submitted to a steady-state flow with free-boundary. Assuming some hypotheses, it is possible to implement the kinematic method by using an approximate pressure field bounding the true pressure field from below. We are interested in finding such approximate pressure fields and in proving that they bound the true pressure field from below without knowing the true pressure field. We use fields which are solutions of a problem with relaxed conditions with regard to the real problem. Under a uniqueness condition of the solution of a weak formulation of the problem, such fields are lower bounds for the true pressure field. Finally, we give the example of a vertical dam.

💡 Research Summary

The paper addresses the stability assessment of porous media subjected to steady‑state flow with a free surface, a situation commonly encountered in geotechnical problems such as dams, levees, and slopes. Traditional yield design (or kinematic) methods evaluate stability by comparing internal virtual work with external loads, but they require knowledge of the pore‑pressure distribution, which is difficult to obtain when a free boundary is present. The authors propose a novel framework that circumvents the need for the exact pressure field by constructing an approximate pressure field that is provably a lower bound of the true pressure.

The core idea is to replace the exact boundary and continuity conditions that define the true pressure problem with relaxed conditions, thereby formulating a “relaxed problem”. In this relaxed problem the free‑surface condition is expressed as an inequality (pressure non‑negative and zero on the free surface) rather than a strict Dirichlet condition, and continuity across internal interfaces or drainage boundaries is allowed to have jumps. The governing flow equation (typically Laplace’s equation for incompressible flow in a homogeneous porous medium) is also weakened, for example by permitting source/sink terms or by using Neumann conditions on parts of the boundary.

A weak formulation of the relaxed problem is introduced, and under a uniqueness assumption for its solution (which holds for the standard Laplacian with the chosen boundary conditions) the authors invoke the maximum and comparison principles. These mathematical tools guarantee that the solution of the relaxed problem, denoted (p^{}), satisfies (p^{} \leq p) everywhere, where (p) is the true pressure field. Consequently, any stability analysis that uses (p^{}) will be conservative: the effective stresses computed from (p^{}) are larger (more tensile) than those from the real pressure, leading to a lower bound on the safety factor.

To illustrate the methodology, the paper presents a detailed example of a vertical dam. The dam geometry creates a groundwater flow from the upstream side to the downstream side, with a free surface that rises on the upstream side and falls on the downstream side. The authors construct a simple linear interpolation of pressure between the upstream head and the downstream outlet, enforcing the relaxed boundary conditions. They then solve the relaxed problem analytically (or with a low‑order finite‑element scheme) to obtain (p^{}). A comparison with a full numerical solution of the exact free‑surface problem shows that (p^{}) is always below the true pressure, yet the discrepancy is modest.

With the approximate pressure field in hand, the kinematic yield design proceeds as follows: (1) a plausible failure mechanism (e.g., a sliding block or a shear band) is selected; (2) a virtual velocity field compatible with that mechanism is defined; (3) the internal virtual work is computed using the effective stress tensor (\sigma’ = \sigma - p^{} I); (4) external virtual work (gravity, water pressure on the downstream face, etc.) is evaluated; and (5) the yield condition (e.g., Mohr‑Coulomb) is applied to obtain a safety factor. Because (p^{}) underestimates the true pore pressure, the resulting safety factor is conservative, providing a reliable lower bound for design.

The paper’s contributions are threefold. First, it offers a systematic way to generate lower‑bounding pressure fields without solving the full free‑surface flow problem. Second, it supplies a rigorous mathematical proof—based on uniqueness and comparison principles—that these fields are indeed lower bounds, thereby strengthening the theoretical foundation of yield design in porous media. Third, it demonstrates the practical applicability of the approach on a classic engineering problem, showing that the method yields realistic and safe design values while significantly reducing computational effort.

The authors conclude by suggesting extensions to more complex scenarios, such as transient flows, heterogeneous permeability, non‑linear Darcy behavior, and multiphase conditions. They also recommend experimental validation and integration with advanced numerical tools (e.g., mixed finite‑element or lattice‑Boltzmann methods) to broaden the scope of the technique. Overall, the work bridges a critical gap between rigorous stability theory and the practical limitations of pressure‑field determination in porous media with free surfaces.