Deep learning modelling of tip clearance variations on multi-stage axial compressors aerodynamics

These guidelines, often consider a trade-off between product cost and engine performance, with tolerance ranges specified accordingly. However, the impact on performance used to define these ranges is typically based on previous experience and simplified correlations. Significant advancements in accuracy and computational cost, have also made CFD analyses an integral part of industrial design processes of turbomachinery components. Despite these advances, analytical models are generally limited to simplified scenarios (not considering real-world effects). Build-specific CFD models are not traditionally used as part of the manufacturing and build process as a day-to-day occurrence, due to the associated computational cost and requirement for specialized engineers to carry out the analyses. The proposed deep learning framework aims to model the effect of manufacturing and build variations on engine performance, achieving similar accuracies to standard CFD solvers, in a significantly shorter timescale. The focus of the current work is on tip clearance variations, one of the main sources of performance variability, but the framework is readily generalizable to other manufacturing variations. This framework is envisioned to be incorporated into the manufacturing and build process, providing instantaneous feedback, which can potentially be used to reduce the requirements for expensive physical testing when clearances go outside the acceptance limits. The challenge: The impact of manufacturing and build variations on the overall performance of gas turbines is known to be significant [1] [2], with the axial compressor, as shown in Figure 1, being a large contributor.