Automatic geometrical optimization by way of numerical flow models

The potentialities in numerical flow modeling available today allow performing more
and more representative and accurate computations of an increasing part of hydraulic engineering
practical problems. In some cases, numerical simulation can be used as an alternative, or at least an
efficient pre-design tool, to more classic physical modeling. In this framework, the paper presents
the application of integrated numerical tools to the automatic optimization of the geometry of a
guide-wall at the entrance of a channel. The simulations have been performed by using a Genetic
Algorithm based optimization tool coupled to a 2D free surface flow model. This finite volume
multibock flow solver, WOLF2D, solves the conservative form of the classical shallow water
equations together with a depth-integrated k-ε type turbulence model. The calculations have been
performed on a cluster of 12 processors to reach a satisfactory solution in less than 2.5 days (96
solutions tested). On the basis of the numerical results and of the value of the suitability indicators,
the optimal solution is objectively better than the other tested geometries. This confirms the
efficiency of the automatic optimization procedure.