High-Lift High-Work LPT Blades and Separation

Document Type

Article

Publication Date

2023

Find this in a Library

Catalog Record

Abstract

Identifying boundary layer separation is important in the design and analysis of high-lift high-work (HLHW) low-pressure turbine (LPT) blades, as separation can severely degrade blade performance. This paper presents a novel method for identifying separation/reattachment lines that is not algorithmic in nature. Rather, it belongs to a new class of flow feature identification techniques ushered in by the era of “big data”: machine learning. Of the various types of machine learning, only supervised machine learning is considered. The present work proposes to use a deep feed-forward neural network to identify separation/reattachment lines on the surface of HLHW LPT blades. Network inputs and architecture are discussed at length. The network is shown to generalize well and produce physically realistic results when asked to extrapolate beyond its training set. Network results are critically assessed and compared with existing techniques. The proposed approach is advantageous (relative to traditional algorithmic techniques) as it does not demand precise mathematic definitions for separation/reattachment lines.

DOI

10.2514/6.2023-0113

Link to Full Text

Catalog Record

Share

COinS