The blade cascade LS89 turbine vane configuration tested in the experimental facility at the Von Karman Institute, see Fig. 5.6. This configuration consists of a set of highly loaded tran- sonic nozzle guide vanes in a linear configuration [7]. The vane profile was especially designed for this experiment and consists of a 2D extruded profile. More details related to the original geometry provided are given in Fig. 5.7 and Table 5.1. Although it is an academic configura- tion, the values encountered could easily be found in current aeroengines today and it represents the most valued database available for numerical comparisons in turbomachinery due to the reduced data provided by industrial partners. This has led to the thorough study already per- formed numerically by a large number of authors as Bhaskaran and Lele [18], Gourdain et al. [72], Wheeler et al. [192] or Collado Morata et al. [38].
The first simulations that used this configuration were of RANS type like in Smirnov and Smirnovsky [168], requiring complex turbulence models to capture transition and other phe- nomena. Results when compared to experimental data had hence an improvement margin and LES showed its capabilities to capture unsteady effects that RANS can not. The increase in numerical power as well as the large number of operating points, has led in the last years to an increase in the number of publications concerning this test case. RANS as in Emory et al. [62] for Uncertainty Quantification analyses but also DNS studies in Wheeler et al. [192] have been performed nearly simultaneously, showing that there are many ways to approach the problem. Focus is set on different parameters such as the certainty in low-cost simulations or the effect of high resolution are reported of crucial importance for simulation assessment. This exempli- fies the fact that a simplified 2D extruded profile vane still represents a challenge in terms of numerical reproduction due to the complex physics it entails and that different methodologies are still today possible towards the resolution of this problem.
Although experiments available for comparison have been available for a few decades al- ready, numerical simulations and finer aspects of them such as sensitivity to boundary condi- tions, shock capturing techniques have not been evaluated thoroughly in a LES context. This can be explained by the lack of computational power available until recent years. Latest pub- lications show that one of the most critical issues points towards the evaluation of free-stream
5.2 LS89: Experiments and literature review
Figure 5.6: Isentropic Light Piston Compression Tube facility at VKI facilities. c 67.674 mm cax 38.81 mm g/c 0.85 55.0 deg o/g 0.2597 rLE/c 0.061 rT E/c 0.0105
Table 5.1: Geometrical parameters characterizing LS89 geometry.
turbulence and the parameters that characterize it. Turbulence characterization requires both intensity measurements but also correlations to estimate for example, the integral length scale which in turn, will influence the dissipation rate and the turbulence evolution. This type of information is not provided in Arts et al. [7] and has been a problem towards characterizing the real conditions that the blade encounters. Studies in Consigny and Richards [40] give esti- mations but these have proven to be insufficient. Recent experimental tests on the same test bench at VKI have provided complementary and valuable information concerning these missing parameters [66].
Many of the first simulations performed show large differences. The most notable of them all is the heat transfer field which represents the most challenging task as it requires to correctly solve all phenomena (thermal and aerodynamic) involved. This has been improved during the last few years due partly to an increased quality in the geometry description. Indeed, based on the original data available to describe the blade geometry, one of the concerns when first performing the simulations was the existence of high-frequency oscillations around the profiles when investigating heat transfer or shear stress fields from the predictions. This problem has also been reported by Wheeler et al. [192], stating that the geometry provided by the original authors [7], although accurate, is not adapted to high fidelity CFD simulations. By taking a set
Figure 5.7: Geometrical detail of the LS89 blade.
of coordinates of the contour of the vane under manufacturing tolerance, it is possible to obtain a curvature that avoids the observed "bumps" and renders a geometrical profile adapted to such numerical simulations. For this study, the geometry was provided by the CENAERO group under the same constraints leading to an appropriate curvature profile. Note that with the original geometry, numerical oscillations were observed whenever diagnosing the LES predic- tions obtained with AVBP, and these were always found at the same position on the curvilinear abscissa. This led to a study of the curvature around the vane, which even under eye inspection can be found to be uneven confirming previous conclusions.
To summarize; difficulties are known to exist to predict the heat transfer coefficient profile around the blade or more generally the overall aerodynamics. This specific field can hence serve to evaluate the quality of the simulation since it is easily accessible experimentally. Different experimental observations on this specific field will then be associated to different boundary layer states or flows around the blade profile. Indeed, the dynamics present in the near-wall re- gion is the most probable reason for these differences as they affect both the momentum and the thermal boundary layers. One of the sources of fluctuations in these boundary layers are due to the free-stream turbulence in the mainstream. This value is fixed experimentally but different turbulent intensities and integral length scales are tested to check its impact and importance. From a numerical point of view, turbulence has to be created and different methods to generate this turbulence may lead to different results, this point is further investigated. Finally, different LES predictions when modifying the turbulent parameters require a mesh convergence analysis. Once the grid effect is assessed and the flow prediction is considered of good quality, a more detailed analysis of the boundary layer statistics is performed to discriminate the findings and
5.3 LS89: Operating points