Reduced Order Models For Unsteady Fluid Dynamic Optimization of Turbomachinery

The need to meet the increasingly demanding sustainability goals entails remarkable new challenges for technical innovation. Due to the crucial role of turbomachinery in present and future energy scenarios, advancements in turbomachinery performance by means of design methods represent a fundamental step towards global sustainable development.
Thanks to the progress in high-performance computing, automated turbomachinery design based on computational fluid dynamics is becoming more and more a viable option to tackle complex design problems. Because of the inherently unsteady nature of turbomachinery flows, optimization methods that are able to account for accurate time resolution of the flow features offer an increased level of simulation fidelity, if compared to methods that assume steady state flows. In this respect, unsteady-based optimization can lead not only to higher fluid dynamic performance, but it can also be seen as a key enabler to address complex multi-disciplinary design problems.
To date, however, most turbomachinery optimization methods are based on the assumption of steady state flows, as a consequence of the high computational cost associated with unsteady fluid dynamic simulations. Reduced order methods offer a computational efficient solution for shape optimization in unsteady flows.
This dissertation documents research on reduced order methods for unsteady adjointbased shape optimization of turbomachinery. In particular, the reduced order methods considered are: the harmonic balance and the look-up table method for the estimation of thermo-physical fluid properties.
The research work resulted in an optimization framework based on a novel harmonic balance discrete adjoint solver, implemented in the open source code SU2. Results show the computational efficiency and effectiveness of the proposed optimization method to deal with unsteady turbomachinery design problems. For the exemplary test cases considered, the unsteady-based optimization led to increased fluid dynamic performance if compared to the optimization results based on steady state computations. Furthermore, the method was successfully employed for design problems of turbomachinery operating with non-ideal compressible flows.