TOPOLOGY OPTIMIZATION UNDER UNCERTAINTY FOR GAS TURBINE INTERNAL FLOWS

Aerospace industries rely heavily on computational fluid dynamic simulations before to prototype and manufacture components. Consequently, it is important that these simulations represent with as much fidelity as possible real conditions. New regulations in the aerospace industry on carbon dioxide emissions have also increased the need for more efficient gas turbines. Increased efficiency can be obtained by increasing the turbine entry temperature, well above the melting point of the alloys used in aerospace. This requires more advanced coolant systems able to reduce the metal temperature and improve the components life.
With the introduction of additive manufacturing (AM), it is possible to have any possible complex geometries designed and manufactured, without the constraints brought by traditional manufacturing methods such as milling or drilling. For example, the idea is that in the structural domain, to develop bones similar structure designs would enable the industries to get lighter components. However, no design methods are currently able to exploit the flexibility offered by AM.
The overall goal of my research is to develop codes able to automatically design such geometries. We managed to obtain bio-inspired structures, similar to the body vein system, that can be used to cool down the components of real gas turbines. Moreover, we managed to obtain new valves without moving parts that can be used in different parts of the engine, such as the internal coolant system. Valves without moving parts have for property to facilitate the flow in one direction while impeding it in the reverse direction, without the use of mechanical components.
An important aspect of my research is to also achieve reliable geometries able to cope with manufacturing errors and uncertainties brought by flow conditions or material properties: a variation of 20 degree C of metal temperature can change the life of hot components by 50%. For these reasons, I am also developing an uncertainty quantification framework which will be able to deal with these problems. The algorithms are developed on an in-house solver TOffee.