Study of the Aerodynamic Performances of LPT Profiles and Use of Machine-Learning Techniques for Data Analysis and Optimization

The quest to enhance the power-to-weight ratio and performance of aircraft engines is pivotal for maximizing payload efficiency. The LPT, constituting a substantial fraction of turbogas mass, faces challenges related to Reynolds number effects, leading to notable losses, especially during cruise conditions. Viscous effects play a crucial role, contribut- ing to profile losses. Advanced Machine Learning (ML) techniques, applied to extensive experimental and numerical databases, offer a promising avenue to deepen understanding and optimize LPT performance. The present thesis explores different methods to evaluate LPT blade performance, emphasizing profile loss evaluation. Experimental campaigns utilizing wind tunnels and innovative measurement techniques provide comprehensive in- sights into blade behaviour across varied operating conditions. A dataset of N = 629 tests allows tuning the ML-based correlation of profile losses, considering a multitude of varia- tions of the parameters. Parallel experimental investigations employ hot wire anemometer (HWA) and laser Doppler velocimetry (LDV) to enhance understanding further. The re- sults showcase the potential of ML algorithms in aerodynamic optimization coupled with a deep experimental insight shedding light on complex phenomena within LPT cascades.