The required aerodynamic simulation fidelity to usefully support a gas turbine Digital Twin for manufacturing

With the imminent digitalisation of the manufacturing processes of gas turbine components, a large volume of geometric data of as-manufactured parts is being generated. This geometric data can be used in aerodynamic simulations to predict component performance. Both the cost and accuracy of these simulations increase with their fidelity. To efficiently exploit Digital Twin technology, one must therefore understand how realistic the aerodynamic simulations need to be to give useful performance predictions. This paper considers this issue for a sample of scrapped high-pressure turbine rotor blades from a civil aero engine. The measured geometric data was used to build aerodynamic models of varying degrees of realism, ranging from quasi-three-dimensional blade sections for an Euler solver to three-dimensional, multi-passage and multi-stage Reynolds-Averaged-Navier-Stokes models. The flow near the tip of these shrouded blades is sensitive to manufacturing variability and can switch between two quasi-stable horseshoe vortex modes. In general, capacity and exit flow angle can be adequately predicted by three-dimensional, single-passage calculations: averaging single-passage calculations gives a good prediction of the multi-passage behaviour. For efficiency and stage loading, the approach of averaging single-passage calculations is less accurate as the multi-passage behaviour requires an accurate prediction of the horseshoe vortex modes.