Professional Documents
Culture Documents
IMECE2014
November 14-20, 2014, Montreal, Quebec, Canada
IMECE2014-36866
ABSTRACT pressure turbine (LPT) profiles, and to a smaller degree for com-
A correlation based approach for estimation of the turbu- pressor profiles. The significance is even accentuated due to the
lence length scale lT at the inflow boundary is proposed and drop of Reynolds numbers at high flight altitudes, that means
presented. This estimation yields reasonable turbulence de- at cruise conditions. Some turbine profiles operate at Reynolds
cay, supporting the transition model in accurately predicting the numbers as low as Re2th ≤ 1.0 · 105 , showing a large laminar
laminar-turbulent transition location and development. As an boundary layer patch with subsequent separation and turbulent
additional element of the approach, the sensitivity of the tur- reattachment, cf. Mayle [8] and Hourmouziadis [9].
bulence model to free stream values is suppressed by limiting There exist many approaches for the prediction of laminar-
the eddy viscosity in non-viscous regions. Therefore a criterion turbulent transition. For turbomachinery flows, a combination of
to detect those regions, based only on local variables, is de- a RANS (Reynolds-averaged Navier-Stokes) turbulence model
rived. The method is implemented in DLR’s turbomachinery flow with a correlation-based transition model is widely spread. In
solver TRACE in the framework of the k − ω turbulence model by this work, the combination of the k-ω turbulence model by
Wilcox 1988 [1] and the γ −Reθ transition model by Langtry and Wilcox [1] and the γ -Reθ transition model by Langtry and
Menter [2] in combination with a cross flow (CF) induced tran- Menter [2] in combination with a cross flow (CF) induced transi-
sition criterion after Müller [3]. The improved model is tested to tion criterion after Müller [3] is applied. This transition model is
the T161 turbine test case [4], [5] and validated at the Durham developed and validated for several generic and turbomachinery
turbine Cascade [6] and an outlet guide vane for low pressure testcases, cf. Langtry and Menter [2]. Herein common turbo-
turbine configurations [7]. machinery testcases are turbine and compressor cascades which
provide a wide range of overall characteristics as well as detailed
boundary layer values to validate the distinct transition modes.
INTRODUCTION
Laminar-turbulent transition plays a significant role in the The transition process is influenced by many parameters,
boundary layer development of modern highly-loaded low- e.g. turbulence intensity, Reynolds number, pressure gradient,
etc.. Most transition models capture this influence by taking the
corresponding parameters into account, in one way or another.
∗ Address all correspondence to this author.
COMPUTATIONAL METHOD
TRACE Code
An up-to-date numerical method, the parallel CFD-solver
TRACE of DLR Cologne has been applied, cf. Nürnberger [15],
Kügeler [16], Marciniak et al. [17] and Becker et al. [18]. In this
solver, the three-dimensional Reynolds-averaged Navier-Stokes
equations are solved on multi-block meshes by a finite volume
technique. The convective fluxes are discretized by the 2nd or-
der TVD upwind scheme of Roe [19] and the diffusive fluxes by
a central differencing scheme. An implicit predictor-corrector
time integration algorithm has been used for the steady simu-
lations. The turbulence is modelled by the two-equation k − ω
model of Wilcox [1], together with the Kato-Launder [20] fix for
the stagnation point anomaly. Integral boundary layer parameters
are determined by integration of the velocity field perpendicular
to the blade surface up to a point where the total pressure has
increased by 99% of the whole velocity defect. A more detailed
description of the method is given in Kožulović [21]. To cap-
FIGURE 1. Computational Domain for T161
ture the streamline curvature effects the k − ω turbulence model
has been modified using local variables only, cf. Kožulović [22].
Furthermore, non-reflecting boundary conditions by Saxer and
Giles [23] have been applied to the inlet and outlet boundaries. wise positions are marked, figure 2 gives an idea of the measuring
The boundary layer transition has been modelled by the two- planes which are used for experimental and numerical investiga-
equation γ − Reθ model of Langtry and Menter [2]. The model tions for this test case. Today linear eddy viscosity models are
evaluates the local flow features to facilitate natural, bypass and the workhorse for aerospace engineers. Those linear models use
separation induced transition. For cross flow (CF) induced tran- the Boussinesq assumption to get the relationship between the
sition an extinction after Müller [3] is used if needed. For a more mean strain rate and the Reynolds stress tensor, cf. equation 1.
detailed description of the original transition model see Langtry
and Menter [2]. Throughout this paper the boundary layers of all 00 00 2
τT,i j = −ρ ui u j = 2µT si j − ρ kδi j (1)
no-slip boundaries are highly resolved with a dimensionless wall 3
distance of the wall adjacent cells down to y+ = 1. Depending on
the test case, the convergence of steady simulations was achieved This equation 1 is reduced to the closure problem of the scalar
after 2000-10, 000 iterations, and was characterized by a density eddy viscosity µT . The main task of turbulence modelling is to
( x )− 5
7
Tu = 1.12 (4)
b
Figure 3 shows the correlation after Baines and Peterson [25] for
several experimental data. The free stream turbulence intensity
(TuFS ) dependent on the normalized distance from the turbulence
generating grid (x/b) is shown. In figure 3 the good agreement of
all experimental results, including the inhouse experiments, with
the theory of Baines and Peterson is seen. Figure 3 provides also
confidence in the grid design process and the measurement tech-
nique, cf. Aufderheide et al. [26]. To get the information about
FIGURE 2. Measuring Planes Definitions for T161 (contour lines are
free stream velocity)
ρk
µT = Cµ , (2)
ω
√ [( ) ]−1.2
k k (t/t0 ) t
ωFS = (3) = +1 (6)
lT k0 t0
(a) Surface Pressure Coefficient c p FIGURE 7. Wake Traverse Measurements and Numerical Simula-
tions at Measuring Plane ME = 40%lax for T161 Turbine Cascade
[ [( ) ] ]
ω
bv = min max , 0.1 , 1.0 . (10)
ωFS
- ζ2th /ζExp.
EXP 1.0 FIGURE 8. Turbulence Intensity Decay
VB 0.946
lT = 0.02m 2.225
lT = 0.002m 1.027 is taken into account by specifying a streamtube height distribu-
lT = 0.0002m 0.709 tion in accordance with the endwall boundary layer development.
lT = 0.00002m 0.682 A corresponding method has been proposed by Stark and Ho-
TABLE 1. Normalized Total Pressure Loss ζ2th /ζExp. heisel [34]. Essentially, this method yields a potential solution
of the endwall displacement effect on the midspan. It has been
used in combination with different blade-to-blade codes to cal-
culate Q2D flows in compressor cascades, cf. for example Bode
behaviour and the turbulence intensity decay correctly compared et al. [35]. The applied grid consists of 18.000 cells and shows a
to the experiments. A closer look at figure 7 shows that the to- high resolution of the boundary layers. Computations conducted
tal pressure loss peaks are better predicted and the mean levels with finer grids offered essentially the same solutions. For more
of the turbulence intensity is best predicted by the new approach informations the reader is referred to Koch et al. [7].
(VB).
Inlet Flow Figure 8 shows again a comparison of experimental
data and numerical simulations for the turbulence intensity at the
inlet block. Again only the simulations with the high value of the
Validation of the New Viscous Blending Approach (VB)
turbulence length scale and the simulation with the new approach
The remaining part of this paper will show the accuracy of
(VB) fits the experimental data in a well appropriate way where
the new approach (VB) by presenting test cases that have been
the simulations with the small turbulence length scales fail.
computed with the TRACE solver and the new approach (VB)
described in the previous section. As one test cases an outlet
Transition Behaviour In figure 9 and figure 10 detailed exper-
guide vane airfoil is presented. This test case is characterized
imental and numerical results for different boundary layer val-
by detailed midspan measurements regarding transition start and
end location as well as boundary layer development and overall
characteristics. As a second test case the Durham turbine cascade
is shown, were the focus lies on the three dimensional flow and
the effects to transition. Therefore the previous mentioned cross
flow (CF) extinction for the transition modell is used.
ACKNOWLEDGMENT
The authors gratefully acknowledge the substantial contri-
butions of the DLR Institute of Propulsion Technology, MTU
Aero Engines and Prof. Dr.-Ing. Reinhard Niehuis from the Uni-
versity of the Armed Forces in München are thanked for pro-
viding the T161 turbine cascade hardware. Dipl.-Ing. Christoph
Müller from the Leibniz Univeristy of Hannover is thanked for
using and discussing the cross flow transition criterion extension
in TRACE. Furthermore, the authors thank MTU Aero Engines
for the permission to publish this work.
(a) Exp. (b) Sim. lT = 9.4mm
NOMENCLATURE
a1 constant in equation 9
b turbulence grid parameter
bv blend viscous in equation 10
cp pressure coefficient, (p − p1 )/(pt1 − p1 )
c p,2th pressure coefficient, (p − p2th )/(pt1 − p2th )
i incidence
k turbulent kinetic energy
l chord
lT turbulent length scale
p pressure
∆p static pressure rise
q dynamic pressure
(c) Sim. lT = 0.94mm (d) Sim. lT = 0.094mm
si j strain rate, 0.5 (δ jUi + δiU j )
tθ pitch
t time in equation 5
x,y,z coordinates
y+ dimensionless wall distance, ρ uτ y/µ
C chord Durham turbine cascade
Cµ coefficient in viscosity equation
Ctke turbulent kinetic energy coefficient
F2 Blending Function in equation 9
M mesh size
Ma Mach number, U/a
Re Reynolds number, ρ ·U · l/µ
Re2th Reynolds number, ρ2 ·U2th · l/µ2
S strain rate, si j s ji
√
(e) VB (f) VB + CF 2k
3
Tu turbulence intensity, U
FIGURE 16. Endwall Intermittency Behaviour for Durham Turbine U velocity
Cascade
Greek
α constant in equation 8
(b) U-Duct
(c) ReΘ
FIGURE 19. Boundary Layer Values for Curved Bend And U-Duct
FIGURE 17. Boundary Layer Values for T3X Flat Plate
Additionally Koch [33], Rüd [47] and Aufderheide [24] did measurements without a turbulence grid. The data for Michalek et
al. [48] is obtained without any turbulence grid.