Jean-Michel Hervouet

A high resolution 2-D dam-break model using parallelization

The evolution of laws on dam safety in France is briefly described. Most simulation studies for such problems are currently attempted using one-dimensional models, and we investigate here the possibility of moving to two-dimensional simulations. To date, these have been constrained by lack of available computational power. Parallel processing is an obvious solution, however, to date, an efficient parallel processing method that does not require major software recoding has proved elusive. Domain-decomposition is identified as being capable of overcoming these problems, as it allows pre-existing software to be adapted to run on clusters of supercomputer processors, workstations or PCs. The development of this new parallel method is outlined and its accuracy and efficiency tested in terms of the TELEMAC-2D model applied to the Malpasset dam break accident, which occurred in 1959 in the south of France. After a discussion of the data available, a sensitivity study is performed to evaluate some physical parameters in the equations, mainly the diffusion coefficient and the bottom friction. The friction appears to be the most important. A discrepancy of 3·4% on the wave celerity is obtained between the model and field observations for a number of locations where the arrival time of the flood wave is perfectly known. The efficiency of different scalar and parallel machines is assessed. The study concludes that 2-D simulations of flood waves are already possible on domains with a length of some 10 s of kilometres. Larger domains (100 to 400 km) are also within reach with supercomputers, or with parallel architectures. Copyright

Morphodynamic modeling using the Telemac finite-element system

The open-source finite-element system Telemac has been applied to simulate various complex hydrodynamic and morphodynamic situations including waterways, curved channels, recirculating flows and wave-induced littoral transport. In the applications presented here, the sediment transport model is mainly restricted to the transport of non-cohesive sediments, which relies on classical semi-empirical concepts including sand grading effects, parameterization of secondary currents and wave effects. In comparison with other comprehensive modeling systems (Delft-3D, Mike-21, etc.), the main originality lies in the efficiency and flexibility of the finite elements. Thanks to the optimization of numerical schemes, parallelism, as well as tremendous progress in the performance of computers, bed evolution can be calculated on basin scale (10-100km) and for the medium term (years to decades), without the use of hydrodynamic filtering methods. As a novelty in release 6.0, we present a method of feedback for the bed roughness, which reduces uncertainty in the prediction of both transport rates and flow velocities.

The computation of free surface flows with TELEMAC: An example of evolution towards hydroinformatics

Presentation of the different computer programmes in the TELEMAC system, dedicated to free-surface flows. This system includes user interfaces and all pre and post-processors, along with fluid mechanics programmes using the same Finite Element library. After a period of internal use at Electricite de France, this system is now available for other users.

Finite volume schemes and residual distribution schemes for pollutant transport on unstructured grids

In this work, a recent residual distribution scheme and a second-order finite volume method are compared to model the transport of a pollutant in free surface flows. The phenomenon is described in two dimensions using the shallow water (SW) system augmented by a scalar conservation law for the pollutant transport. The two numerical methods are developed to minimize the numerical diffusion which is a critical problem for transport phenomena. The conservation of the mass and the monotonicity of the solution are two other important numerical requirements necessary to reproduce the physics of the problem. These three features (low numerical diffusion—mass conservation—monotonicity) will be theoretically analyzed and then numerically verified through a series of test cases. Both methods are used on completely unstructured grids, but the regularity of the grid with respect to the streamlines direction produces for the two methods different behaviours which will be studied. This work has been realized within the Telemac-2D system, constituted by a finite element kernel and a finite volume kernel. Telemac2D uses several numerical schemes, including the new schemes presented here. The aim of this paper is to present state-of-the-art research in the field of finite volumes (FV) and of residual distribution schemes for advection problems.