Advanced Postprocessing for High Fidelity Simulation

The thesis is part of a research project carried out by AvioAero Ge in collaboration with academic research groups and Morfo Design Srl, with the purpose of improving the investigation of Low-Pressure Turbine (LPT) cascades performances through high fidelity CFD simulation. In particular, the aim of this work is to highlight the power of Large Eddy Simulation (LES) of a complex environment such as a LPT cascade with unsteady incoming wakes by introducing and developing new techniques of postprocessing that can be placed within the framework of a preliminary design. The main topic regards the developing of a Python tool to compute the Proper Orthogonal Decomposition (POD) on unsteady LES data able to evaluate unsteady losses due to larger and smaller flow structures carried by the wakes in different parts of the cascade. Moreover, it is also illustrated a deep investigation of the boundary layer and the acceleration parameter calculation (K) on time averaged LES data compared with experimental measurements. After an introduction about the technology readiness levels related with a low-pressure turbine design and an overview of computational fluid dynamics methodologies, the work is divided into two parts. The former is related to the mathematical framework behind the Proper Orthogonal Decomposition to which it follows a literature review regarding the loss generation mechanisms in a low-pressure turbine cascade by the application of the POD to experimental measurements. The latter presents both the steps and the results analysis of the POD tool, developed for this work, applied to the LES data. The advanced postprocessing techniques were developed through Python scripts as they aim to be as much automatic as possible in order to produce results in an accessible time scales. The first step is the production of data by interacting with the High-Fidelity CFD Software. Secondly it follows the postprocessing of this data by applying the POD. This allows to pass from a physical plane to a basis representation thanks to which it is feasible to evaluate different dynamic features and the unsteady losses in terms of production of turbulent kinetic energy associated to them. The main results were the ranking of the initial bigdata set based on the turbulent kinetic energy of each mode, the evaluation of the dynamic behavior of each flow field realization, the amount of production of turbulent kinetic energy (i.e., losses) due to unsteadiness and where this source of losses occur. All these information provides further inputs to the designers for future optimization strategies.