Large-Eddy Simulation of a Linear Compressor Cascade with Tip Gap: Aerodynamic and Acoustic Analysis

2021 ◽  
Author(s):  
Regis Koch ◽  
Marlène Sanjosé ◽  
Stephane Moreau
2015 ◽  
Vol 27 (8) ◽  
pp. 085105 ◽  
Author(s):  
Feng Gao ◽  
Wei Ma ◽  
Gherardo Zambonini ◽  
Jérôme Boudet ◽  
Xavier Ottavy ◽  
...  

Author(s):  
Wei Ma ◽  
Feng Gao ◽  
Xavier Ottavy ◽  
Lipeng Lu ◽  
A. J. Wang

Recently bimodal phenomenon in corner separation has been found by Ma et al. (Experiments in Fluids, 2013, doi:10.1007/s00348-013-1546-y). Through detailed and accurate experimental results of the velocity flow field in a linear compressor cascade, they discovered two aperiodic modes exist in the corner separation of the compressor cascade. This phenomenon reflects the flow in corner separation is high intermittent, and large-scale coherent structures corresponding to two modes exist in the flow field of corner separation. However the generation mechanism of the bimodal phenomenon in corner separation is still unclear and thus needs to be studied further. In order to obtain instantaneous flow field with different unsteadiness and thus to analyse the mechanisms of bimodal phenomenon in corner separation, in this paper detached-eddy simulation (DES) is used to simulate the flow field in the linear compressor cascade where bimodal phenomenon has been found in previous experiment. DES in this paper successfully captures the bimodal phenomenon in the linear compressor cascade found in experiment, including the locations of bimodal points and the development of bimodal points along a line that normal to the blade suction side. We infer that the bimodal phenomenon in the corner separation is induced by the strong interaction between the following two facts. The first is the unsteady upstream flow nearby the leading edge whose angle and magnitude fluctuate simultaneously and significantly. The second is the high unsteady separation in the corner region.


Author(s):  
Susumu Teramoto ◽  
Takuya Ouchi ◽  
Hiroki Sanada ◽  
Koji Okamoto

Fully resolved large eddy simulation (LES) is applied to two simple geometry flowfields with well-defined boundary conditions. The LES results are compared with simulations based on a Reynolds-averaged Navier-Stokes (RANS) model with turbulence, and pros and cons of using high-resolution LES for turbomachinery flows are discussed. One flow is a linear compressor cascade flow composed of the tip section of GE rotor B at Rec = 4 × 105 with a clearance, and the other is a Mach 1.76 supersonic turbulent boundary layer at Reδ = 5000 that laminerizes through a 12-degree expansion corner. The grids are prepared fine enough to resolve the turbulent boundary layer through a grid sensitivity study. The liner cascade result shows that all the turbulent shear layers and boundary layers including those in the small tip clearance are well resolved with 800 million grid points. The Reynolds stress derived from the LES results are compared directly with those predicted from the Spalart-Allmaras one-equation RANS turbulence model. The two results agreed qualitatively well except for the shear layer surrounding the tip leakage vortex, demonstrating that the RANS model performs well at least for flowfields near the design condition. From the simulation of the turbulent boundary layer experiencing sudden expansion, noticeable decreases of both Reynolds stress and local friction coefficient were observed, showing that the turbulent boundary layer has relaminarized through the sudden expansion. The boundary layer downstream of the expansion exhibits a nonequilibrium condition and was different from the laminar boundary layer.


Author(s):  
Zifei Yin

Abstract Delayed detached eddy simulations and wall-modeled eddy simulations using the adaptive DES model were performed to simulate corner separation in the Ecole Centrale de Lyon linear compressor cascade. The adaptive DES model directly uses length scale to define eddy viscosity, which makes it nature to compute the model constant CDES dynamically. The dynamic procedure adapts viscosity to local flow and grid. Delayed detached eddy simulations, with and without the dynamic procedure, were performed to demonstrate the benefit of adapting viscosity to local flow. Recycling method was adopted to generate inflow unsteady turbulent boundary layer for wall-modeled eddy simulations. The wall-modeled eddy simulation showed improvement over delayed-DES, in terms of static pressure coefficient around the blade and total pressure loss at downstream locations.


AIAA Journal ◽  
2006 ◽  
Vol 44 (4) ◽  
pp. 741-750 ◽  
Author(s):  
Charles E. Martin ◽  
Laurent Benoit ◽  
Yannick Sommerer ◽  
Franck Nicoud ◽  
Thierry Poinsot

Author(s):  
Man Zhang

A diffusion swirling flame under external forcing and self-excitation within a single swirler combustor have been studied in this paper with the large-eddy simulation and linear acoustic method. The combustor features pre-vaporized kerosene as the fuel, a single radial air swirler for flame stabilization and a square cross section chamber with adjustable length. Firstly, self-sustained pressure oscillation has been achieved by using of a chocked nozzle on the chamber outlet with large-eddy simulation. Dynamic pressure oscillations are analyzed in frequency domain through Fast Fourier Transform. The major pressure oscillation is identified as the 1st order longitudinal mode of the chamber. Further, the same frequency in the form of harmonic velocity oscillation is imposed on the inlet of the combustor while the chamber length has been changed. Based on this approach, a comparative study of the flame response with different excitation method but same frequency is carried out. In both self-excited and forced cases, global and local flame responses as well as Rayleigh index have been analyzed and compared. With the flame response function, the excited acoustic modes under the influence of dynamic heat release have been predicted with linear acoustic method and compared with the results obtained from large-eddy simulation. Results show that the flame response presents a great difference in the spacial distribution with different excitation approaches. Thermo-acoustic interaction distributes along the flame front with the expansion of the flame under self-excitation while it damps with the acoustic propagating downstream under forcing condition. As the ratio of flame length to acoustic wave length could not be neglected for the diffusion swirling flame, the global flame response under forcing cannot represent the local response feature of the flame accurately, thus influencing the instability prediction.


Sign in / Sign up

Export Citation Format

Share Document