Investigation of the Flow in a linear high pressure compressor Cascade using scale resolving Simulations

Hybrid RANS/LES (HRLES) is one SRS category, which bridges the gap between RANS and LES in regard to prediction accuracy of the results and required computing resources. The HRLES methods (i.e. various Detached Eddy Simulation (DES) formulations), with RANS modelling of the flow near the wall, and eddy-resolving simulation away from the wall, are believed to represent the mixing in turbulent flows and other secondary flow features formed in the end wall regions better than the state-of-the-art RANS. 
The major goals targeted within this project can be concluded as follows:
1.    Quantify the suitability of HRLES methods (one method in specific) to accurately predict the 3D flow field of high-pressure compressor cascade operating at high Reynolds number.
2.    Analyzing the 3D flow of the high-pressure compressor cascade in terms of secondary flow structures, boundary layer development and loss mechanism, especially in the wake field.
The first goal is important because such HRLES methods are a promising methodology for simulating flows with higher Reynolds numbers in the near future. The reason is that wall resolved LES for such Reynolds numbers is still challenging in terms of computing resources. The second goal is important because a thorough understanding of secondary (near-wall) flow, which has been mostly neglected in the past, will help designers develop more sophisticated designs. The focus of the current study is the high-pressure compressor cascade (HPC cascade) operated at the Institute of Jet Propulsion and Turbomachinery (IST) of the RWTH Aachen University.
There are three simulations (3 operating points) planned to be performed within this computing period.
The mesh study for the HRLES was based on three different meshes, which were used to quantify the required mesh elements to resolve the turbulent structures. Additionally, the Pope criteria, in which 80% of the turbulent kinetic energy should be resolved, was sought for in the free stream regions. The second mesh, which consist of 47 mio. mesh elements, was found to be a good compromise between accuracy and computing resources.
At the lowest Reynolds number 490,000, the HRLES results show marginal improvement regarding the total pressure loss in the wake at midspan. Both RANS and HRLES show an offset regarding the peak of the total pressure wake loss but the shape (width) of the wake loss fits very well with the measurements. The total pressure wake loss of HRLES show almost no improvement in the near wall area, compared to RANS results. 
The blade pressure distribution of both simulations matches very well with measurements.
The reason for the only marginal improvement of the HRLES in the wake area has been shown to be the mesh density and/or quality, especially in the near wall and trailing edge areas. A further mesh refinement in the near wall area will result in higher computing resources and is not justified for HRLES. The HRLES methodology used in this project, switches between RANS and LES based on the turbulent length scale ratio of the corresponding method (l_RANS/l_IDDES) and is therefore highly coupled with the near wall mesh density.
At the highest Reynolds number 790,000, the results show nearly the same trend as for the lower Reynolds number. Also, at this operating point, there is a good match between both RANS and HRLES and the measurements regarding the blade pressure distribution. Here, the experience from the previous simulations regarding the mesh quality was adapted to come up with a more sophisticated mesh quality. The HRLES results show a better agreement with the measurement results, when comparing the total pressure wake loss (both peak values and shape) at midspan. Unfortunately, also at this operating point, the losses in the near wall area show almost no improvement, compared to RANS results. Also, here the near wall mesh density is believed to be responsible for the failure of the method.