Session: 36-03 Deep Learning based applications

Paper Number: 151295

C(NN)FD - Deep Learning Modelling of Multi-Stage Axial Compressors Aerodynamics 

The field of scientific machine learning and its applications to numerical analyses such as Computational Fluid Dynamics have recently experienced a surge in interest. While its viability has been demonstrated in different domains across a wide range of engineering applications, it has not yet reached a level of robustness and scalability to make it practical for industrial applications in turbomachinery field.  The highly complex, turbulent and three-dimensional flows of multi-stage axial compressors for gas turbine applications represent a remarkably challenging case. This is due to the high-dimensionality of the regression of the flow-field from geometrical and operational variables, and the high computational cost associated with the large scale of the CFD domains. This paper demonstrates the development and application of a generalised deep learning framework for predictions of the flow field and aerodynamic performance of multi-stage axial compressors, which is also potentially applicable to any type of turbomachinery. A physics-based dimensionality reduction unlocks the potential for flow-field predictions for large-scale domains, re-formulating the regression problem from an un-structured to a structured one and reducing the number of degrees of freedom. The relevant physical equations are then used to calculate the corresponding radial distributions, stage-wise and overall performances. Compared to “black-box” approaches, the proposed framework has the advantage of explainable predictions of overall performance, as the corresponding aerodynamic drivers can be identified on a 0D/1D/2D/3D level. Moreover, the associated uncertainty is also estimated at each level of the predictions, providing a quantifiable measure of the confidence in resolving the relevant flow features, without a significant increase in computational cost. This is applied to model the effect of manufacturing and build variations on the compressor aerodynamic performance, considering both tip clearance and surface roughness. The model is trained on a dataset including different engine designs and operating conditions, demonstrating the capability to predict the flow-field and the overall performance in a generalisable manner, for different engines across their compressor maps, with accuracy comparable to the CFD benchmark.

Presenting Author: Giuseppe Bruni Siemens Energy Industrial Turbomachinery Ltd

Presenting Author Biography: Giuseppe Bruni is currently employed as a Principal Aerodynamicist by Siemens Energy Industrial Turbomachinery Ltd, Lincoln, UK, where he has been working since 2016. He received the BSc and MSc degrees in Mechanical Engineering from University of Padova, Italy, respectively in 2014 and 2016. He received the MSc in Gas Turbine Technology from Cranfield University, UK in 2016, and is currently working towards a PhD at the University of Lincoln, UK, on the topic of Machine learning modelling of manufacturing variations on compressor aerodynamic performance. His research interest include the aerodynamic, aero-mechanical analysis and design of axial compressors, as well as development and applications of optimisation and machine learning methods. He is a Chartered Engineer and member of the IMechE.

Authors:

Giuseppe Bruni Siemens Energy Industrial Turbomachinery Ltd
Sepehr Maleki University of Lincoln
Senthil K. Krishnababu Siemens Energy Industrial Turbomachinery Ltd