Applicability of a 3D Numerical Model for Flow Simulation of Spillways

Very often, the design of spillways requires physical models, or the use of results of previous ones reported as design aids in the general technical literature. The application of three-dimensional flow models may, in many cases, replace the need of physical models leading to significant savings of time and money, as well as becoming a more economical option for the design of hydraulic works.

The application of fluid calculation programs with the 3D equations for the study of the behavior of water in spillways and in energy dissipating structures has been studied with greater emphasis in the last 20 years. In the literature there are examples of application of Eulerian models such as SSIIM, OpenFOAM, Flow 3D, Phoenics, ANSYS, TELEMAC, just to mention some of them. More recently, the Lagrangian approach, in the so called SPH models, has also allowed 3D calculation in complex situations.

The present work constitutes a first phase of a very comprehensive evaluation of the open source software OpenFOAM, developed under a finite volume discretization, in the calculation of spillway structures. For that aim, the geometry of standard Creager type crests was evaluated, for which a wealth of data from physical models is available, so it is possible to compare the results obtained with the program. More than 800 cases were run with the aim to evaluate the effects of mesh sizes, turbulence models, mesh geometries, crest geometry, slope of the upstream face, heads to design head ratios, and height of the approach channel, for spillways with and without intermediate piers. To our knowledge, there has not been such comprehensive comparison between 3D model and empirical results, for a family of cases of standard Creager type crests.

The program showed good results for the flow profile, crest pressures and discharge capacity for cases with P/H0 (height of approach channel to design head) higher than 1.0. Smaller P/HO values give unstable results. With the correct mesh size and mesh geometry, most results are within few percentage (less than 5%) from the empirical results.