Development of a data-driven design tool for turbomachinery blades: Exploration of secondary flows in 3D turbine cascades

Gas turbines, particularly those in aircraft propulsion, represent a cutting-edge engineering field. Although advances in cooling technologies and materials have led to mechanical efficiency above 90 percent, economic and physical constraints limit future increases. As a result, improving machine performance still heavily relies on minimising aerodynamic losses. Among these, secondary-flow-induced losses — arising from the complex interaction between the main passage flow and endwall boundary layers — still represent a challenge in the design phase. Despite empirical correlations and simplified models, no comprehensive analytical framework currently exists to quantify accurately these losses across different design configurations. This thesis develops a Python-based computational tool that automates the three-dimensional aerodynamic analysis of turbine cascades through NUMECA Cadence FINETurbo software. The tool integrates key stages of the CFD (Computational Fluid Dynamics) workflow, including blades CAD (Computer Aided Desgin) generation through profile extrusion, mesh generation, boundary conditions computation, based on operational Reynolds and Mach numbers, and the setting of desired inlet boundary layer profiles. Following this, the tool automates simulation execution, its monitoring, and post-processing data, extracting key aero-thermodynamic quantities to evaluate aerodynamic losses. Through data collected from arbitrary blade configurations, this framework allows the preliminary investigation of secondary losses and their sensitivity to design parameters, drawing from methodologies established in prior research. The study is structured as followed: a literature review on secondary flows-induced and endwall losses; a methodology section outlining the tool development and key computational processes; a further validation and comparative analysis of processed blade configurations; and a final discussion of results, literature comparisons, and future research directions. The long-term objective is to deliver quality data useful to establish loss correlations with design parameters and develop a breakdown of losses, aiming to evaluate the secondary-flow induced ones. Thus, advancing current analytical approaches for turbine cascade performance optimisation through machine-learning based models.