Design of a Relational Database for Turbine Center Frames With Application for Geometry Optimization

The flow path between high- and low-pressure turbines (TCF) is a key component of modern aero engines; the radial offset between the two stages and the small axial length result in a strong curvature and complex flow behavior. A solid database is an important foundation for design and further investigations, especially when applying modern artificial intelligence (AI) methods. Due to a long research history, the Institute for Thermal Turbomachinery and Machine Dynamics (ITTM) at Graz University of Technology owns an extensive TCF data collection. In order to reuse these data samples, a database application has been developed, providing the functionalities to normalize and store the data uniformly. The TCF data samples are mapped into a common parametric space during import, making them comparable and accessible to AI applications. Furthermore, statistical evaluation tools and an automated CFD export are available. In the second part of the paper, the database is used for an exemplary geometry optimization task. The approach is to train a surrogate model with data from the database and then use the model for optimization. The results agree with the findings of other authors, and the surrogate model's predictions coincide well with CFD results.