Session: 34-20 Solver methods 4

Paper Number: 128885

128885 - Flow Reconstruction in a Transonic Turbine Cascade Using Physics-Informed Neural Networks (PINNs) 

This paper investigates the application of Physics-Informed Neural Networks (PINNs) for the analysis of turbine blades in a transonic cascade. PINNs are a machine learning method trained on losses calculated from reconstructed governing equations, assigned boundary/initial conditions, and measured data. In this paper, we explore the 2-D flow field in a transonic turbine cascade in two ways: the traditional forward approach (without training/experimental data) and by training the PINN using experimental data. We then compare these results with measured data and a CFD model. The inclusion of measured data to guide the solution is a unique feature of PINNs and arguably their most advantageous quality. This process is repeated for three different turbine blades, each with distinct loading characteristics. All three blades have been assessed using a PINNs model trained with all available experimental data, showing very good agreement with the experimental data. The next step involves varying the amount of experimental data used for training to assess how much data is needed to obtain reasonable predictions with the PINN. Additionally, we will vary the locations of the experimental data used for training to evaluate which regions of the blade provide the most valuable insights for the efficient training of the PINN. The results from this paper will be of assistance to researchers using PINNs in the reconstruction of flow fields with sparse data or measurements.

Presenting Author: Ezra McNichols NGRC

Presenting Author Biography: Ezra McNichols is an aerospace research engineer who has worked at NASA Glenn Research Center since 2018. In his current role, he serves as the Turbomachinery Technical Lead within the Advanced Air Transport Technology (AATT) project. His professional expertise and research interests include aerodynamics and heat transfer in turbomachinery, thermal management solutions for electronics, fundamental heat and mass transfer phenomena, and the practical implementation of machine learning for design and analysis.

He holds a Master of Science degree in Mechanical Engineering from the University of Illinois at Urbana-Champaign, along with a Bachelor of Science degree in Mechanical Engineering from the University of Kentucky.

Authors:

Ezra McNichols NASA Glenn Research Center
Paht Juangphanich NASA Glenn Research Center
Mallory Hawke Johns Hopkins University Applied Physics Laboratory
Ethan Shoemaker Millennium Space Systems
Mackinnon Poulson Lockheed Martin
Meghan Brandt NASA Glenn Research Center
Jeffrey Bons The Ohio State University