A Real Time AI-Based Strategy for the Design of a Low-Pressure Turbine Profile

This paper focuses on the implementation of a data-driven, AI-powered design strategy for Low Pressure turbine (LPT) profiles. The objective is to develop an effective LPT profile optimization environment, capable of rapidly generating a CFD calculation database to feed AI-based tools for subsequent phases of design space exploration. To achieve this purpose, an automatic workflow has been developed to create profile geometries, computational meshes, and perform URANS calculations. CFD results are then collected both in terms of overall performance (losses) and detailed flow field. AI-models are then trained over this large database, resulting in an accurate metamodel that can be effectively explored looking for specific design solutions. The paper will present and discuss the AI-based tool, showing how in addition to suggesting optimal geometries, it offers additional outputs to inform the designer. Indeed, the tool can provide optimal geometries that match mechanical and geometrical constraints providing performance at varying Reynolds number. Moreover, the blade-loading and the entire 2D flow field are predicted in real-time for the optimal geometry at a given operating condition.