Deep learning modelling of manufacturing and build variations on multi-stage axial compressors aerodynamics

TL;DR

This work tackles the difficulty of predicting turbomachinery aerodynamics under manufacturing variations by introducing C(NN)FD, a physics-based dimensionality reduction that reformulates a high-dimensional CFD regression into a structured, reduced problem. A 3D U-Net–based surrogate predicts a 6-channel flow field on a structured grid across 24 axial locations and 64×64 planes, enabling real-time predictions with accuracy comparable to CFD while providing explainability by tracing outputs to aerodynamic drivers. The approach explicitly handles tip clearance and surface roughness variations in a 10-stage compressor, delivering accurate flow-field, radial profile, and stage-wise performance predictions, including PR and ηp, with strong generalization to out-of-tolerance cases. The framework is designed for integration into the manufacturing/build process to assess engine builds analytically, reduce expensive physical testing, and inform robust design and corrective actions.

Abstract

Applications of deep learning to physical simulations such as Computational Fluid Dynamics have recently experienced a surge in interest, and their viability has been demonstrated in different domains. However, due to the highly complex, turbulent, and three-dimensional flows, they have not yet been proven usable for turbomachinery applications. Multistage axial compressors for gas turbine applications represent a remarkably challenging case, due to the high-dimensionality of the regression of the flow field from geometrical and operational variables. This paper demonstrates the development and application of a deep learning framework for predictions of the flow field and aerodynamic performance of multistage axial compressors. A physics-based dimensionality reduction approach unlocks the potential for flow-field predictions, as it re-formulates the regression problem from an unstructured to a structured one, as well as reducing the number of degrees of freedom. Compared to traditional "black-box" surrogate models, it provides explainability to the predictions of the overall performance by identifying the corresponding aerodynamic drivers. The model is applied to manufacturing and build variations, as the associated performance scatter is known to have a significant impact on $CO_2$ emissions, which poses a challenge of great industrial and environmental relevance. The proposed architecture is proven to achieve an accuracy comparable to that of the CFD benchmark, in real-time, for an industrially relevant application. The deployed model is readily integrated within the manufacturing and build process of gas turbines, thus providing the opportunity to analytically assess the impact on performance with actionable and explainable data.

Figures (17)

• Figure 1: Overview of the C(NN)FD framework
• Figure 2: Overview of the CFD domain and axial velocity contours at the mixing-plane locations
• Figure 3: Box plots of the input variables (tip clearance and surface roughness) at each blade row
• Figure 4: C(NN)FD architecture overview
• Figure 5: Training history: Training loss and Test loss