scholarly journals Machine Learning Models of Errors in Large Eddy Simulation Predictions of Surface Pressure Fluctuations

2017 ◽  
Author(s):  
Matthew F. Barone ◽  
Julia Ling ◽  
Kenny Chowdhary ◽  
Warren Davis ◽  
Jeffrey Fike
Keyword(s):  
Machine Learning ◽  
Large Eddy Simulation ◽  
Surface Pressure ◽  
Pressure Fluctuations ◽  
Learning Models ◽  
Eddy Simulation ◽  
Large Eddy ◽  
Machine Learning Models

Large-Eddy Simulation of Unsteady Surface Pressure Over a LP Turbine Blade Due to Interactions of Passing Wakes and Inflexional Boundary Layer

2005 ◽  
Author(s):  
S. Sarkar ◽  
Peter R. Voke
Keyword(s):  
Large Eddy Simulation ◽  
Turbine Blade ◽  
Coherent Structures ◽  
Pressure Fluctuations ◽  
Time Dependent ◽  
Suction Surface ◽  
Eddy Simulation ◽  
Large Eddy ◽  
Grid Points ◽  
Lp Turbine

The unsteady pressure over the suction surface of a modern low-pressure (LP) turbine blade subjected to periodically passing wakes from a moving bar wake generator is described. The results presented are a part of detailed Large-Eddy Simulation (LES) following earlier experiments over the T106 profile for a Reynolds number of 1.6×105 (based on the chord and exit velocity) and the cascade pitch to chord ratio of 0.8. The present LES uses coupled simulations of cylinder for wake, providing four-dimensional inflow conditions for successor simulations of wake interactions with the blade. The three-dimensional, time-dependent, incompressible Navier-Stokes equations in fully covariant form are solved with 2.4×106 grid points for the cascade and 3.05×106 grid points for the cylinder using a symmetry-preserving finite difference scheme of second-order spatial and temporal accuracy. A separation bubble on the suction surface of the blade was found to form under the steady state condition. Pressure fluctuations of large amplitude appear on the suction surface as the wake passes over the separation region. Enhanced receptivity of perturbations associated with the inflexional velocity profile is the cause of instability and coherent vortices appear over the rear half of the suction surface by the rollup of shear layer via Kelvin-Helmholtz (K-H) mechanism. Once these vortices are formed, the steady-flow separation changes remarkably. These coherent structures embedded in the boundary amplify before breakdown while traveling downstream with a convective speed of about 37 percent of the local free-stream speed. The vortices play an important role in the generation of turbulence and thus to decide the transitional length, which becomes time-dependent. The source of the pressure fluctuations on the rear part of the suction surface is also identified as the formation of these coherent structures. When compared with experiments, it reveals that LES is worth pursuing as an understanding of the eddy motions and interactions is of vital importance for the problem.

Prediction of Pressure Fluctuation on a Vehicle by Large Eddy Simulation

2015 ◽  
Author(s):  
Yoshinobu Yamade ◽  
Chisachi Kato ◽  
Akiyoshi Iida ◽  
Shinobu Yoshimura ◽  
Keiichiro Iida
Keyword(s):  
Structural Analysis ◽  
Large Eddy Simulation ◽  
Unsteady Flow ◽  
Pressure Fluctuation ◽  
Pressure Fluctuations ◽  
Sound Fields ◽  
Eddy Simulation ◽  
Acoustical Analysis ◽  
Wide Range ◽  
Large Eddy

The objective of this study is to predict accurately interior aeroacoustics noise of a car for a wide range of frequency between 100 Hz and 4 kHz. One-way coupled simulations of computational fluid dynamics (CFD), structural analysis and acoustical analysis were performed to predict interior aeroacoustics noise. We predicted pressure fluctuations on the outer surfaces of a test car by computing unsteady flow around the car as the first step. Secondary, the predicted pressure fluctuations were fed to the subsequent structural analysis to predict vibration accelerations on the inner surfaces of the test car. Finally, acoustical analysis was performed to predict sound fields in the test car by giving vibration accelerations computed by the structural analysis as the boundary conditions. In this paper, we focus on the unsteady flow computations, which is the first step of the coupled simulations. Large Eddy Simulation (LES) was performed to predict the pressure fluctuations on the outer surfaces of the test car. We used the computational mesh composed of approximately 5 billion hexahedral grids with a spatial resolution of 1.5 mm in the streamwise and spanwise directions to resolve the dynamics of the small vortices in the turbulence boundary layer. Predicted and measured pressure fluctuation at several sampling points on the surface of the test car were compared and they matched well in a wide range of frequency up to 2 kHz.

scholarly journals Stochastic backscatter of turbulence energy and scalar variance by random subgrid-scale fluxes

1995 ◽  
Vol 451 (1941) ◽  
pp. 293-318 ◽  
Keyword(s):  
Large Eddy Simulation ◽  
Isotropic Turbulence ◽  
Pressure Fluctuations ◽  
Decay Rates ◽  
Temperature Fields ◽  
Turbulence Energy ◽  
Eddy Simulation ◽  
Large Eddy ◽  
Scalar Variance ◽  
Energy Decay Rates

For large-eddy simulation with a finite-difference scheme, a simple stochastic subgridscale (SGS) model is introduced which describes the effects of random SGS motions on the resolved (filtered) scales of incompressible turbulent motions. The model extends the Smagorinsky-Lilly model by adding realizable random stresses and fluxes which are constructed as quadratic expressions of Gaussian random velocity and temperature fields. The random components reduce the correlations between stresses and strain rates to in between 0.16 and 0.5, in agreement with observations. The random stresses (fluxes) also induce random accelerations (temperature changes) with a k 4 power spectrum. Such random sources backscatter energy (variance) from SGS motions to resolved scale motions when temporally correlated with finite timescales. The timescales are different for momentum and heat flux. The analysis of the model provides an upper estimate of the magnitude of backscatter which is close to previous predictions. The analysis identifies the influence of the quasi-normal assumption and of numerical filters and determines the variance of the pressure fluctuations induced by the random accelerations at grid scales. Backscatter increases the SGS turbulent Prandtl number to a degree depending strongly on the numerical filter. Tests of the model in large-eddy simulation of isotropic turbulence show energy decay rates in close agreement with expected rates when the stochastic SGS model is included. Backscatter cannot be simulated with reduced diffusivities or filter widths.

scholarly journals Towards the Large-Eddy Simulation of a full engine: Integration of a 360 azimuthal degrees fan, compressor and combustion chamber. Part II: Comparison against stand-alone simulations

2021 ◽  
pp. 1-16
Author(s):  
Carlos Pérez Arroyo ◽  
Jérôme Dombard ◽  
Florent Duchaine ◽  
Laurent Gicquel ◽  
Benjamin Martin ◽  
...  
Keyword(s):  
Large Eddy Simulation ◽  
Combustion Chamber ◽  
Hot Spot ◽  
Pressure Fluctuations ◽  
Forward Region ◽  
Impeller Blade ◽  
Eddy Simulation ◽  
Blade Passing Frequency ◽  
Large Eddy ◽  
The Impact

Unsteady simulations of various components of a gas-turbine engine are often carried out independently and only share averaged quantities at the component interfaces. In order to study the impact and interactions between components, this work compares results from sectoral stand-alone simulations of a fan, compressor and annular combustion chamber of the DGEN-380 demonstrator engine at take-off conditions against an integrated 360 azimuthal degrees large-eddy simulation with over 2.1 billion cells of all previously listed components. Note that, at take-off conditions the compressor works at transonic conditions and generates an upstream-propagating shock that interacts with the fan modifying the shape of its wake with respect to the stand-alone simulation. Furthermore, the shock is seen as a tone in the pressure spectra at half the impeller blade passing frequency in the forward region of the engine. In the aft region, time-averaged fields are overall similar between stand-alone and integrated simulations but show a deviation in the azimuthal position of the hot-spot at the exit of the combustion chamber due to the addition of the diffuser. Pressure fluctuations generated in the compressor are captured in the combustion chamber as tones in the temperature and pressure spectra at the impeller blade-passing frequency and harmonics as well as an increase in the root-mean-square pressure.

Sign in / Sign up

Export Citation Format

Share Document