A redesign strategy for high-pressure steam turbine stages

The main topic of this PhD thesis is the description and the analysis of a redesign strategy for high-pressure steam turbine stages. The overall procedure will be presented, from the choices made for the stage optimization to the numerical verifications need for the new product industrialization. The optimization technique relies on a response surface method, coupling the high-fidelity CFD RANS computations with an optimization algorithm based on meta-models. The emphasis will be mainly focused on two meta-models in particular, the support vector machines and the artificial neural networks: an entire chapter will be devoted to their description, and a comparison between these two methodologies will be reported in the section concerning the results of the stage optimization. A wide section will be devoted to the impact of surface roughness, since it is a relevant and active research area and it plays a role of main importance into blades design.