59277 - Cascade With Sinusoidal Leading Edges: Identification and Quantification of Losses With Unsupervised Machine Learning 

One of the key issues in fan design is the identification of loss mechanisms and their quantification, both during preliminary design and in all subsequent optimization loops.

Over the years, many correlations have been proposed, taking into account different dissipative mechanisms that occur in blade-to-blade passages. Literature reports on correlations to account for friction in the boundary layer that develops over aerodynamic surfaces, turbulent wake mixing, shockwaves, secondary flows, off-design incidence and other.

However, in recent years, the fan industry started the production of more complex rotor geometries, characterized by sinusoidal leading and trailing edges, mostly to extend stall margin and to reduce noise emissions. In particular, sinusoidal leading edges have proven able to change the evolution of trailing edge separation and stall dynamic.

In this case, literature still lacks a quantification of losses introduced by the secondary motions released by the leading-edge sinusoid. In this paper we investigate a design of experiments that entails 76 cases of a 3D flow cascade with NACA 4digit profiles with sinusoidal leading edges to measure losses according to Lieblein approach. Losses are here quantified according to changes in the geometry of the blade characterized by different amplitude and wavelengths, Reynolds number, cascade solidity and cascade pitch (input parameters).

To do so, we simulated the DoE with RANS approach using OpenFOAM v18.12 and analysed the flow field using a machine learning approach to classify and isolate the turbulent wake downstream of the cascade with a combination of Principal Component Analysis and K-Means clustering. Then we used a gradient boosting regressor to derive the correlation between input parameters and losses [GT2020-15337].