28/06/2018 14:00 - Salle D
The intent of this talk is to summarize development of a methodology used for carrying out geometric shape optimization studies using the TELEMAC-MASCARET numerical modelling system. This methodology was developed during the course of my post-doc. The overall intent of the methodology is to identify an optimum geometric shape for design problems that are typically studied using free surface flow, wave and sediment transport numerical modeling (i.e., determining length or spacing of groins in a river, or finding an orientation or layout of breakwaters in a harbour, for example). Such design problems are solved by specifying problem’s variables, objective function to be optimized, and relevant constraints. The shape optimization problem is solved numerically by coupling an optimization algorithm to results produced by a numerical simulation model. The developed tool chain then defines, tests and evaluates a multitude of new shapes, which are internally generated and simulated using a numerical model. The above steps thus require a tool that automatically generates new model geometries, as well as extracting desired outputs from the simulations. In other words, the tool chain must be able to create new meshes, interpolate topography/bathymetry, assign initial and boundary conditions, as well as manage model output on the fly. A form of scripting is required to recover from numeric simulation results, which are then used in evaluation of the problem’s objective function. A demonstration will be presented during the talk, as well as two example test cases.