Kierkegaard A.,MWL Marcus Wallenberg Laboratory for Sound and Vibration Research |
Kierkegaard A.,Linne Center |
Kierkegaard A.,KTH Royal Institute of Technology |
Allam S.,MWL Marcus Wallenberg Laboratory for Sound and Vibration Research |
And 7 more authors.
Journal of Sound and Vibration | Year: 2012
This paper demonstrates a linear aeroacoustic simulation methodology to predict the whistling of an orifice plate in a flow duct. The methodology is based on a linearized NavierStokes solver in the frequency domain with the mean flow field taken from a Reynolds-Averaged NavierStokes (RANS) solution. The whistling potentiality is investigated via an acoustic energy balance for the in-duct element and good agreement with experimental data is shown. A Nyquist stability criterion based on the simulation data was applied to predict whistling of the orifice when placed in a finite sized duct and experiments were carried out to validate the predictions. The results indicate that although whistling is a nonlinear phenomena caused by an acoustic-flow instability feed-back loop, the linearized NavierStokes equations can be used to predict both whistling potentiality and a duct systems ability to whistle or not. © 2011 Elsevier Ltd. All rights reserved.
Muld T.W.,Linne Center |
Muld T.W.,KTH Royal Institute of Technology |
Efraimsson G.,Linne Center |
Efraimsson G.,KTH Royal Institute of Technology |
And 2 more authors.
Computers and Fluids | Year: 2012
In this paper, Proper Orthogonal Decomposition (POD) and Dynamic Mode Decomposition (DMD) are used to extract the most dominant flow structures of a simulated flow in the wake of a high-speed train model, the Aerodynamic Train Model (ATM). The use of decomposition methods to successfully identify dominant flow structures for an engineering geometry is achieved by using a flow field simulated with the Detached Eddy Simulation model (DES), which is a turbulence model enabling time accurate solutions of the flows around engineering geometries. This paper also examines the convergence of the POD and DMD modes for this case. It is found that the most dominant DMD mode needs a longer sample time to converge than the most dominant POD mode. A comparison between the modes from the two different decomposition methods shows that the second and third POD modes correspond to the same flow structure as the second DMD mode. This is confirmed both by investigating the spectral content of the POD mode coefficients, and by comparing the spatial modes. The flow structure associated with these modes is identified as being vortex shedding. The identification is performed by reconstructing the flow field using the mean flow and the second DMD mode. A second flow structure, a bending of the counter-rotating vortices, is also identified. Identifying this flow structure is achieved by reconstructing the flow field with the mean flow and the fourth and fifth POD modes. © 2012 Elsevier Ltd.
Monokrousos A.,KTH Royal Institute of Technology |
Monokrousos A.,Linne Center |
Lundell F.,KTH Royal Institute of Technology |
Lundell F.,Linne Center |
And 2 more authors.
AIAA Journal | Year: 2010
An experimental demonstration of feedback control of bypass transition is presented. The data from an earlier experiment will be used here as reference in a numerical study aimed at reproducing the disturbance conditions in the experiment, as well as the control performance. A linear feedback control scheme was employed in order to reduce the disturbance growth and, consequently, delay transition. The case of bypass transition represents an extension of the linear control approach to flows characterized by strong nonlinearities. The results showed that the control was able to delay the growth of the streaks in the region where it was active. The flow-field can be estimated from wall measurements alone: the structures occurring in the real flow are reproduced correctly in the region where the measurements are taken. The LQR with time-and-space varying blowing/suction gives much larger initial disturbance attenuation than the experiments and a considerable transition delay.
Vinuesa R.,Illinois Institute of Technology |
Rozier P.H.,Illinois Institute of Technology |
Schlatter P.,KTH Royal Institute of Technology |
Schlatter P.,Linne Center |
Nagib H.M.,Illinois Institute of Technology
AIAA Journal | Year: 2014
The study of high-Reynolds-number wall-bounded turbulent flows has become a very active area of research in the past decade, where several recent results have challenged current understanding. In this study, four different localized pressure gradient configurations are characterized by computing them using four Reynolds-averaged Navier-Stokes turbulence models (Spalart-Allmaras, κ-ε, shear stress transport, and the Reynolds stress model) and comparing their predictions with experimental measurements of mean flow quantities and wall shear stress. The pressure gradients were imposed on high-Reynolds-number, two-dimensional turbulent boundary layers developing on a flat plate by changing the ceiling geometry of the test section. The computations showed that the shear stress transport model produced the best agreement with the experiments. It was found that what is called "numerical transition" (a procedure by which the laminar boundary conditions are transformed into inflow conditions to characterize the initial turbulent profile) causes the major differences between the various models, thereby highlighting the need for models representative of true transition in computational codes. Also, both experiments and computations confirm the nonuniversality of the von Kármán coefficient κ. Finally, a procedure is demonstrated for simpler two-dimensional computations that can be representative of flows with some mild three-dimensional geometries. Copyright © 2013 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
Na W.,KTH Royal Institute of Technology |
Na W.,Linne Center |
Na W.,2 Vehicle Design Center |
Efraimsson G.,KTH Royal Institute of Technology |
And 6 more authors.
22nd International Congress on Sound and Vibration, ICSV 2015 | Year: 2015
The paper presents a numerical methodology for the prediction of the thermoacoustic instabilities with the effects of the mean-flow as well as the viscosity. As an academic standard test case, the configuration within the flame sheet located in the middle of the duct is investigated. First, the ducted flame numerical reference case is solved by the inhomogeneous Helmholtz equations in combination of the n - τ flame model assuming that the flow is at rest. Then, we derive the linearized Navier-Stokes equations (LNSE) in frequency domain in combination of the flame model. The unsteady effect of the flame is modeled by the n - τ flame model in harmonic form, which is essentially a 1D formulation relating the rate of heat release and the acoustic velocity at the reference point.
Tempelmann D.,KTH Royal Institute of Technology |
Tempelmann D.,Linne Center |
Schrader L.-U.,KTH Royal Institute of Technology |
Schrader L.-U.,Linne Center |
And 7 more authors.
6th AIAA Theoretical Fluid Mechanics Conference | Year: 2011
Direct numerical simulations (DNS) of the flow over a wing with 45° sweep and -4° angle-of-attack are presented. This flow configuration was investigated in a series of wind-tunnel experiments at the Arizona State University (ASU). Here, we examine the boundary-layer receptivity to surface roughness and to single vortical free-stream modes. The roughness is modeled by a shallow circular disk and is identical with one single element of the spanwise roughness array considered in the ASU experiments. The boundary layer develops a steady crossflow mode downstream of the roughness. The spatial evolution of the modal amplitude obtained by the DNS is in excellent agreement with a solution to the nonlinear parabolized stability equations (NPSE) while being lower than that measured in the experiments. The reasons for this discrepancy are yet to be determined. Possible explanations are the presence of traveling crossflow waves due to background free-stream turbulence in the experiments or the slight difference between the numerical and experimental pressure gradients at the roughness site. Stationary crossflow vortices can also be triggered by zero-frequency free-stream vortical modes. We consider two types of mode, carrying streamwise and vertical vorticity. Both modes give rise to nonmodal disturbances near the leading edge, which soon evolve into a steady crossflow mode. The boundary layer is found to be somewhat more receptive to the streamwise-vorticity mode than to the chordwise vorticity. Copyright © 2011 by David Tempelmann.
Von Stillfried F.,Linne Center |
Kekesi T.,Linne Center |
Wallin S.,Linne Center |
Wallin S.,Swedish Defence Research Agency |
Johansson A.V.,Linne Center
Journal of Physics: Conference Series | Year: 2011
Separating turbulent boundary-layers can be energized by streamwise vortices from vortex generators (VG) that increase the near wall momentum as well as the overall mixing of the flow so that flow separation can be delayed or even prevented. In general, two different types of VGs exist: passive vane VGs (VVG) and active VG jets (VGJ). Even though VG s are already successfully used in engineering applications, it is still time-consuming and computationally expensive to include them in a numerical analysis. Fully resolved VGs in a computational mesh lead to a very high number of grid points and thus, computational costs. In addition, computational parameter studies for such flow control devices take much time to set-up. Therefore, much of the research work is still carried out experimentally. KTH Stockholm develops a novel VGJ model that makes it possible to only include the physical influence in terms of the additional stresses that originate from the VGJs without the need to locally refine the computational mesh. Such a modelling strategy enables fast VGJ parameter variations and optimization studies are easliy made possible. For that, VGJ experiments are evaluated in this contribution and results are used for developing a statistical VGJ model.