Session: 32-01 High Fidelity CFD

Paper Number: 125856

125856 - High-Fidelity Investigation of Vortex Shedding From a Highly-Loaded Turbine Blade 

High-fidelity numerical simulations of the flow in the linear turbine cascade VKI LS59 (Sieverding, 1972) are presented. The LS59 geometry is representative of a highly loaded rotor blade and has served as a prototype for numerous past studies, including measurements in 4 different wind tunnels (Kiock, 1986), and Laser Doppler Velocimetry (LDV) measurements at the University of Graz (1998). This configuration was also used as a test case for turbomachinery simulations with (U)RANS, but unlike other configurations such as LS89 or LS94 cascades, no high-fidelity simulations were performed. One question that remains unanswered in the literature is the laminar or turbulent nature of the boundary layers and how the latter affects the overall flow dynamics, including cascade losses. In order to shed light on these aspects, a numerical campaign was carried out. The calculations relied on a high-order solver for the compressible Navier-Stokes equations, using tenth-order finite differences and a fourth-order Runge-Kutta time stepping with high-order implicit residual smoothing to relax stability constraints on the time step. The computational grid contained approximately 152 million points (with 240 points in the span, corresponding to 10% of the chord), satisfying the criteria for wall-resolved large-eddy simulations. Synthetic turbulence was injected at the inlet with a given intensity Tu. The calculation campaign covered several outlet Mach numbers (subsonic and transonic) and several experimental configurations, including the setup used in the Göttingen (GO) wind tunnel (chord c=6cm) and the setup of Graz University (c=5.8 cm). A first surprising result concerned experiments in the subsonic regime. At M₂ =0.8 (GO case), a lattice of shock waves is induced by the alternate vortex shedding. The flow thus locally reaches supersonic speeds when the vortices are rolled up: this regime is called 'transonic vortex shedding' (Melzer & Pullan, 2019). In the Graz experiment at M₂ =0.6, however, there is no shock wave: this corresponds to the 'detached vortex shedding' regime, characterized by a less coherent wake and a higher shedding frequency. In the Graz case at M₂ =0.7,we found that, depending on the level of inlet turbulence (Tu=2.5 and 5%), two different regimes are established: the 'detached vortex shedding' regime at 2.5% (long bubble, St ∼0.24) and the 'transonic vortex shedding' regime for Tu=5% (short bubble, St ∼0.19). The vertical profiles in the wake for the Tu=2.5% case are in good agreement with LDV measurements (even if Tu~5% in experiments). Such a change in flow regime, solely due to an increase in freestream turbulence, has not been reported to our knowledge. Furthermore, while Melzer & Pullan found that both boundary layers on suction and pressure sides must be turbulent to switch to the transonic regime, both boundary layers were laminar in the present case. In the 'transonic' regime, the vortices are more two-dimensional and very intense, yielding much higher levels of fluctuations, so that losses are multiplied by a factor of 2. Finally, we set out to reproduce the transonic GO case (M_2,is =0.99). Kiock et al. reported that the incoming turbulence level is almost negligible for this case (~1%) and the results were not affected by placing a wire trip at 60% of the suction side. However, simulations without inlet turbulence and without trip wire could not match the experimental Schlieren field. In turn, the simulated shock system returned to good agreement with experiments with trip wire or turbulence. In the untripped case, the recirculation bubble was shorter leading to a vortex shedding Strouhal number of 0.17, whereas in simulations with trip wire or inlet turbulence the bubble was longer, with St~0.24. Such change of regime was reminiscent of one observed at subsonic conditions. In this case, the boundary layer on the suction side remained laminar without trip wire, whereas it transitioned short after the trip wire, and was transitional in the case with freestream turbulence. The switch between the long or short bubble regimes strongly affects the shock system and thus the losses.

Presenting Author: Xavier Gloerfelt DynFluid Laboratory - Arts et Métiers Insitute of Technology

Presenting Author Biography: Xavier Gloerfelt is Professor at DynFluid Laboratory in Arts and Metiers Instiute of Technology in Paris (France). His research interest are the study of compressible flows including turbulence, aeroacoustics and complex physics such as dense gases.

Authors:

Xavier Gloerfelt DynFluid Laboratory - Arts et Métiers Insitute of Technology
Paola Cinnella Institut Jean Le-Rond D'Alembert - Sorbonne Universite