Shape optimization of a metallic flywheel using an evolutive system method: Design of an asymmetrical shape for mechanical interface

Abstract

In this work, a novel optimization approach is used to define the shape of a flywheel for energy storage. The procedure acts on a two-dimensional axisymmetric finite element method model, in which many parameters are used to describe the geometry. The optimization is performed by an evolutive system method, whereby the population’s genome is described with statistical quantities. An accurate definition of the fitness function allows for a broad spectrum of objectives. The evolution of the fitness value during population generation is discussed. The procedure does not benefit from parallelization: an alternative way of parallelizing the optimization process is presented, where the population is equally divided among the cores and calculated independently, allowing for an approximately linear performance scaling with the number of cores. The method is applied to the minimization of the mass in a flywheel for energy storage application, displaying great flexibility to the variation of the parameters describing the rotational speed, geometry constraints, and material properties. The true potential of the evolutive method is then demonstrated by optimizing an asymmetrical flywheel, where a mechanical interface lies on one side only. Usually additional parts are added manually on the optimized shape, increasing the thickness where necessary; this method permits to directly optimize to the final shape. The script used in this work is available upon request. Matlab and Ansys APDL software are needed.