High-cycle thermal fatigue in mixing Tees.
Large-Eddy simulations compared to a new
validation experiment
Johan Westin Pascal Veber, Lars Andersson
Vattenfall Research and Development AB Onsala Ingensjörsbyrå AB
SE-81426 Älvkarleby, Sweden SE-43437 Kungsbacka, Sweden
Carsten ‘t Mannetje Urban Andersson, Jan Eriksson,
Forsmarks Kraftgrupp AB Mats Henriksson
SE-74203 Östhammar, Sweden Vattenfall Research and Development AB
SE-81426 Älvkarleby, Sweden
Farid Alavyoon Claes Andersson
Forsmarks Kraftgrupp AB Ringhals AB
SE-74203 Östhammar, Sweden SE-43022 Väröbacka , Sweden
© Vattenfall AB
Outline
• Background
– Previous work by the authors
• New experimental validation test case (2006)
• Computational results (Fluent)
– Mesh dependence study
– Large Eddy Simulations (LES) compared with Detached Eddy
Simulations (DES)
• Concluding remarks
© Vattenfall AB
2
Background (1)
Introduction
• Temperature fluctuations can cause
thermal fatigue
• Interesting case for CFD-validation
(unsteady flow, large fluctuation levels)
• Static mixers or thermal sleeves can be
installed to reduce the risk for thermal
fatigue (but expensive) Static mixer (MIX-331)
• Desirable with accurate predictions of the
risk for thermal fatigue
• Structural analysis require boundary
conditions
1) Amplitudes of temperature fluctuations
2) Frequencies of temperature fluctuations
3) Heat transfer to the wall
© Vattenfall AB
3
Background (2)
Previous work by the present authors
• Model test of a plant specific T-junction performed in 2002 3000 mm
– Geometry including upstream bends D=600 mm
R=1100 mm
– Temperature fluctuations near the wall 2840 mm
measured with thermocouples D=200 mm
4450 mm
– Several test cases (flow ratios) D2=190 mm
1168 mm
• Computational studies R=225 mm
1000 mm
D1=123 mm
– Unsteady RANS failed to predict the temp. fluctuations
– LES showed promising results
– Still discrepancies (amplitude and frequencies overpredicted)
– Complicated and uncertain inflow boundary conditions
• Need for more validation data and well-documented inflow
boundary conditions
© Vattenfall AB
4
New model tests for validation of CFD (2006)
Test rig overview
• Fully developed pipe flow in the cold • Focus on one test case
water inlet pipe – 'T|15qC
(>80 diameters straight section – Constant flow ratio Qcold/Qhot=2
upstream the T-junction) – Equal inlet velocities in the cold
• Pipe diameters 140 mm (cold) and and hot water pipe
100 mm (hot)
High-level
reservoir
Stagnation
>2000 (>20D) Hot chamber
water DN300
Stagnation chamber
DN400
Cold water
z
x
>12200 (>80D)
© Vattenfall AB
5
T-junction model and measurements
T11-T12
LDV measurements
T-junction
T31-T32
2D 4D 6D 8D 10D 15D 20D
0°
90° 270°
180°
• Thermocouples • Lased Doppler Velocimetry (LDV)
(0.13 and 0.07 mm) – Inlet-BCs at x/D=-3 and z/D=-3.1
– Located 1 mm from the – Profiles at x/D=2.6 and 6.6
wall – Measurements at 'T|15qC and isothermal
– Frequency response
• Single-point Lased Induced Fluorescence (LIF)
30-45 Hz
– Conc. measurements at isothermal conditions
© Vattenfall AB
6
Flow visualization: 50%, 100% and 200% flow
(Reynolds number: 0.5u105, 1.0u105 and 2.0u105)
© Vattenfall AB
7
Spectra of temperature fluctuations at x=4D
Various flow rates
x=4D
P vs f (Hz)
No normalization
f=4 Hz f=4 Hz
P fD1
vs
'T 2 U3
© Vattenfall AB
8
Performed simulations and numerical settings
Influence of computational mesh Influence of unsteady inflow-BC
Case #cells tsamp Note Organization 1) Vortex method
(s)
2) No perturbation
T1vm- 0.52M 29.0 4 boundary Forsmark
3) Scaled isotropic turbulence
FKA layer cells
from separate input files
T1Bvm 0.45M 21.8 no BLcells Onsala
Vattenfall R&D
T2vm 0.93M 19.6 More Onsala Comparison LES-DES
uniform Vattenfall R&D
LES: WALE (Wall-Adaptive
T3vm 9.5M 8.3 Similar, but Onsala Local Eddy viscosity model)
refined DES: SST k-Z model
Numerical settings
• Non-iterative time advancement (NITA): 2nd order, implicit
• Pressure-velocity coupling: Fractional step
• Momentum eq: Bounded central differences
• Pressure: PRESTO
• “Law-of-the-wall” applied near the wall (y+ typically 20-50)
© Vattenfall AB
9
Instantaneous temperature fields
Case T2vm (0.93 Mcell) Case T3vm (9.5 Mcell)
© Vattenfall AB
10
Simulation, 9.5 Mcell
© Vattenfall AB
11
Temperatures near the pipe wall
Different computational mesh
Top
Left Right
Bottom
T Tcold
T*
Thot Tcold
Trms Trms
'T Thot Tcold
© Vattenfall AB
12
Velocity fluctuations at the pipe centerline
Different computational mesh
urms
U bulk
wrms
U bulk
© Vattenfall AB
13
Velcoity spectra, pipe centreline at x=2.6D
Different computational mesh
u-component (=streamwise) v-component (=spanwise)
(not measured)
f=4 Hz
© Vattenfall AB
14
Mean velocity profiles, x/D=2.6
Different computational mesh
z
x
y
© Vattenfall AB
15
Velocity fluctuations, x/D=2.6
Different computational mesh
z
x
y
© Vattenfall AB
16
Mean and fluctuating temperatures
Comparison LES vs DES (mesh 0.93 Mcell)
Left Right
T Tcold
T*
Thot Tcold
Trms Trms
'T Thot Tcold
© Vattenfall AB
17
Time signals
Comparison LES vs DES (x/D=4)
LES
DES
Experiment
© Vattenfall AB
18
Modelled turbulent viscosity
Comparison LES vs DES
LES DES
Color scale: 0-0.2 kg/(ms)
© Vattenfall AB
19
Concluding remarks
• Good agreement between simulation and model test results, also
with fairly coarse computational mesh
– Both fluctuation amplitude and spectral distribution show good
agreement
– Considerably better than in previous experiment/simulations
– (Indicate that the current flow case is quite ”forgiving” for LES)
• Insensitive to variations in the (unsteady) inlet boundary conditions
• However: Clear improvement of the results with a refined mesh
– Improved results in the entire computational domain with 9.5 Mcell
• Still insufficient resolution near the walls
– Erroneous prediction of the near-wall mean velocity profile and the
wall-shear stress as compared to fully developed turbulent pipe flow
– Detached Eddy Simulations (DES) results in better near-wall profiles,
but the tested model is too dissipative in order to give good predicition
of the temperature fluctuations
© Vattenfall AB
20
Interested in the Vattenfall T-junction test case?
• The experimental data can be made available for those who are interested
to perform simulations
• In return we expect to get access to the computational results
• No restrictions to publish your results (reference to the source of the data)
• Presently computations are carried out by NRG, The Netherlands (Ed
Komen et al.) and ANSYS, Germany (Frank et al.)
If you are interested to use this test case for CFD-validation, contact
Johan Westin, Vattenfall Research and Development AB
E-mail: johan.westin@vattenfall.com
© Vattenfall AB
21
