content classical types of taylor vortices important parameters mesh independence study ...
TRANSCRIPT
Summary for Validation Period
Qian Wentao
24/05/2011
Content
Classical Types of Taylor Vortices
Important Parameters
Mesh Independence Study
Taylor-Couette Validation
Wavy Taylor Validation
Turbulent Validation
Thermal Validation
Simple Model Test
Plans for Next Period
Classical Types of Taylor Vortices
Laminar Couette Flow
According to the stability analysis without considering the viscous, the flow inside the cylinders should be always instable when outer cylinder is kept fixed. However, viscosity has an important stabilizing influence at low Reynolds numbers. Stability will be broken only if the angular velocity of inner cylinder exceeds a critical value.
Classical Types of Taylor Vortices
Taylor vortex forms when Re exceeds Rec
When the angular velocity of the inner cylinder is increased above a certain threshold, Couette laminar flow becomes unstable and a secondary steady state characterized by axis-symmetric toroidal vortices, known as Taylor vortex flow, emerges.
Classical Types of Taylor Vortices
Wavy taylor vortex forms when Re exceeds Rec2
Subsequently increasing the angular speed of the cylinder the system undergoes a progression of instabilities which lead to states with greater spatio-temporal complexity, with the next state being called as Wavy Vortex Flow. The rotational speed approximately 20% higher than the critical speed for transition to Taylor vortices.
Classical Types of Taylor Vortices
Chaotic Flow Happens Before Fully Turbulence
The flow undergoes a series of transitions before it becomes fully turbulent. By using spectral method, there is only one peak at the initial period of the wavy vortex flow, while more peaks will be formed when Re is increasing, this period is called Chaotic Flow until no peaks appear.
Terminology
k- Thermal Conductivity or circumferential wave numberν - Kinematic Viscosityuθ- Azimuthal Velocityu0- Tangential Velocity of Inner Cylinderr- Distance from Centre Axis N-number of cellsm-number of azimuthal wavesω-fundamental angular frequency of the wave
Radius Ratio η=R1/R2 Gap Width d=b-aAspect Ratio Г=H/d Re=R1Ωd/νAxial Wavelength λ=2H/N=(2Г/N)dWave Speed s=ω/(m Ω)Mean Equivalent Conductivity Keq= -h*r*ln(R1/R2)/k
h- Convective Heat Transfer CoefficientR1- Radius of Inner CylinderR2- Radius of Outer CylinderH- heightΩ- angular velocity of inner cylinder
Important Parameters η
Critical Reynolds number for forming taylor vortices is strongly influenced by the radius ratio of two cylinders.
Important Parameters
Aspect ratio
It is noted that aspect ratio is expected to have little effect on the quantitative behavior of the flow for aspect ratios above 40. [J.A. Cole 1976]
Important Parameters
Axial wavelength & Wave number
[D. Coles (1965)] entailed the observation of different flow states with the same Taylor number and concluded that the flow depends not only on Re, but also on previous flow history. The states are defined by number of axial and azimuth waves.
Important Parameters
Wave speed
[KIKG,G. P. (1984)] concluded that there was much weaker dependence of wave speed on axial wavelength, azimuthally wave number, and the aspect ratio.
Important Parameters Flow structure
[M. Fenot (2011)] concluded that the structure of the combined flow in annular space hinges not only on the operation point (axial Reynolds and Taylor numbers), but also – and strongly- on geometry and, to a lesser degree, on parietal thermal condition.
Mesh Independence Study
Objectives & Methodology After comparing with
experiment data, we can find the minimum mesh density that can appropriately simulate the physical phenomenon. With such confidence, a more reasonable mesh structure would be selected as a final reference for the later mesh building.
All the boundary conditions are kept the same with changing the mesh grid numbers only.
The length was cut into half to save computing time.
Radial Grids
Axial Grids
Circle Grids
Mesh Independence Study
Boundary Conditions
R1=0.1906m R2=0.2622m H=1.6714m (full length)
Re=4Rec Ω=0.0244 rad/s
End walls are rotating with inner cylinder while outer was fixed
Laminar mode
Fluid: Water (µ=0.001003kg/m-s ρ=998kg/m3 ν=1.005*e-6 m2/s)
Mesh Independence Study
Circle Grids
Radial Grids
Axial Grids
Cells Numb
erResidual for Continuity
Accordingly Wavelength
200 60 250 28 9.98E-04 1.569100 60 250 28 1.26E-03 1.575
*Axial grid number here is for full length. In computing convenience, the number was cut into half of that in the form.*The experimental value of wavelength is 1.59
No influence on the cells number for decreasing the circle grid number. So keep it as 100.
Mesh Independence Study
Circle Grids
Radial Grids
Axial Grids
Cells Numb
erResidual for Continuity
Accordingly Wavelength
100 60 120 28 1.38E-03 1.588100 60 180 28 2.42E-03 1.610100 60 250 28 1.26E-03 1.575100 60 400 28 2.05E-03 1.541
*Axial grid number here is for full length. In computing convenience, the number was cut into half of that in the form.*The experimental value of wavelength is 1.59
No influence on the cells number for an interval of circle grid numbers.
Mesh Independence Study
100 60 120 100 60 180 100 60 250 100 60 400
Cross Section View with Mesh
Mesh Independence Study
According to the above figures, simulation shall be kept reasonable with each cell shares at least 3 grid axial. In order to achieve relatively accurate result, the one with axial grid number of 400 is selected as the reference.
Mesh Independence Study
Circle Grids
Radial Grids
Axial Grids
Cells Numb
erResidual for Continuity
Accordingly Wavelength
100 80 180 28 5.21E-02 N/A100 60 180 28 2.42E-03 1.610100 50 180 24 3.03E-03 1.852100 40 180 24 5.21E-02 N/A
*Axial grid number here is for full length. In computing convenience, the number was cut into half of that in the form.*The experimental value of wavelength is 1.59.
The minimum for the radial grid number is 60.
Mesh Independence Study
According to the real computing condition, the mesh with circle 100, radial 60 and axial 400 (full length) is good enough which can be regarded as the reference for later mesh building.
Conclusion
Taylor-Couette Validation
Objectives & Methodology After comparing with the
experiment data from J.E. Burkhalter & E.L. koschmieder (1974) and check wavelength versus taylor number. We shall confirm that fluent is able to simulate the Taylor- Couette phenomena. λ
Taylor-Couette Validation
Boundary Conditions
Since author have mentioned that neither the end wall nor the column length have significant influence on the experiment result, so we kept the same boundary condition with former mesh independence study, only changing the Taylor number which is exact a non-dimension value.
1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.21
2
4
8
16
32
64
J.E. Burkhalter & E.L. koschmieder (1974)Simulation Data
Taylor-Couette Validation
Wavelength
T/T
c
Full Length
Typical wavelengths after sudden starts
Taylor-Couette Validation Conclusion
From above figure, we tried both full length and half length geometries to run cases with different taylor numbers. The results under 8Tc are well fit with the experiment data. Those above 8Tc are absolutely among the wavy taylor vortices period which will possibly cause the unstable measurement for the wavelength. However, the discrepancy is acceptable
Wavy Taylor Validation Objectives & Methodology
Taking KIKG,G. P., Lr, Y ., LEE, W., SWINNEY, H. L. & MARCUS, P.S . 1984 Wave speeds in wavy Taylor-vortex flow. J . Fluid Mech. 141, 365-390. as the reference paper to compare with. Check wave speed versus different radius ratio which is the most significant factor.
Set measure point located in the gap centre from 2D cross section view to record the z-velocity during the time interval which can derive the fundamental frequency of the azimuthal waves. m can be observed by applying Tecplot to stretch out the cylinder.
Wavy Taylor Validation Boundary Conditions
*All the meshes are built with reference of “100 60 400” which has been mentioned in the former part.
η(R1/R2)R1(cm)
R2(cm) H(cm)
Re(11Rc)
Ω(rad/s)
Upper Boun
dcircu
m radial axial
0.868 2.205 2.54010.05
0 1266.1 17.222 Free 100 30 2500.900 2.286 2.540 7.620 1447.6 25.050 Free 100 25 1900.950 5.649 5.946 8.910 2036.1 12.194 Free 200 25 200
Laminar mode
Fluid: Water (µ=0.001003kg/m-s ρ=998kg/m3 ν=1.005*e-6 m2/s)
Keep Г=30, Re=11Rc, λ/d=2.4
Wavy Taylor Validation
η(R1/R2)
R1(cm)
R2(cm) H(cm)
Re(11Rc)
Ω(rad/s)
Upper Bound circum
radial
axial
0.868 2.205 2.54010.05
0 1266.1 17.222 Free 100 30 250
Fundamental angular frequencyω=17.279s=ω/(m Ω)=0.334
Wavy Taylor Validation
Two fundamental frequenciesω=27.227s=ω/(m Ω)=0.362
η(R1/R2)
R1(cm)
R2(cm) H(cm)
Re(11Rc)
Ω(rad/s)
Upper Bound circum
radial
axial
0.900 2.286 2.540 7.620 1447.6 25.050 Free 100 25 190
Wavy Taylor Validation
Fundamental angular frequencyω=50.265s=ω/(m Ω)=0.458
η(R1/R2)
R1(cm)
R2(cm) H(cm)
Re(11Rc)
Ω(rad/s)
Upper Bound circum
radial
axial
0.950 5.649 5.946 8.910 2036.1 12.194 Free 200 25 200
Wavy Taylor ValidationConclusion
η(a/b) Computed S1
Measured S1
0.868 0.334 0.320±0.005
0.900 0.362 0.360±0.010
0.950 0.458 0.450±0.001
The difference is located in the reasonable region of uncertainty
Need to be calculated longer.
Dependence of s1 on radius ratio
Turbulent Validation Objectives & Methodology
Comparison of normalized mean angular momentum profiles between presentsimulation (Re=8000) and the experiment of Smith & Townsend (1982).
Using two models (k-epsilon and k-omega) to compare with the experiment data (left), after which an appropriate model would be selected for later calculation.
Couple of measure points are located in the midline across the gap from 2D view of the cross section. They are applied to record the tangential velocity during time interval and calculate the average value.
Turbulent Validation Boundary Conditions
R1 = 0.1525 mR2 = 0.2285 mΩ = 22.295 rad/s (Re=17295)H = 1.80 m
End walls are free surfaces
k- epsilon and k- omega were chosen to compare
Fluid: Water (µ=0.001003kg/m-s ρ=998kg/m3 ν=1.005*e-6 m2/s)
Mesh DensityAxial = 400Circle = 100Radial = 60
Turbulent Validation Comparing with Experiment Data
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.2
0.3
0.4
0.5
0.6
0.7
0.8
Experiment DataSimulation Data k-epsilonSimulation Data k-omega
z/d
U*r
/(U
1*R
1)
Turbulent Validation Conclusion
Flow time interval is not enough
ΔTepsilon=27.68s ΔTomega=20.48s
Sampling frequency
fexperiment=10kHz fsimulation=200Hz
Mesh density
Tip: W. M. J. Batten used k-epsilon as the turbulent model
Thermal Validation Objectives & Methodology
Variation of mean equivalent conductivity with Reynolds number for different Grashof numbers
By comparing with experiment data from K.S. Ball (1989) , we would confirm the capability of fluent on simulating the heat transfer for taylor-couette flow.
We simplified the condition by ignore the effect of both conduction and radiation which will cause at most 5% error but saving much calculating time. Average heat transfer coefficient in the inner cylinder surface can be directly achieved from fluent.
Thermal Validation Boundary Conditions
Keq= -h*r*ln(R1/R2)/kRe = Ω* (R1-R2)*R1/ν
R1 = 1.252 cmR2 = 2.216 cmH = 50.64 cmGr= 1000 ΔT= 7.582 KTi = 293K To= 300.582KEnd walls are fixed and insulated
Fluid: Air (k=0.0257w/m-k β=3.43*e-3 1/k ν=15.11*e-6 m2/s)
Re=[40 120 280]Ω=[5.008 15.023 35.054] rad/s
Since for η=0.565 Rec= 70,All the three cases are in laminar mode.
Mesh DensityAxial = 1000Circle = 100Radial = 60
Comparing with Experiment Data
Re2 h(w/m2k) keq R. Kedia Experiment Data
Residue
1600 3.600 1.001 N/A 1.080 9.8e-04
6400 4.636 1.289 N/A 1.450 1.4e-03
14400
6.206 1.725 1.65 1.500 2.0e-03
40000
8.545 2.376 2.30 1.750 1.7e-03
78400
10.228 2.845 2.70 2.120 1.8e-03
Comparing with Experiment Data
Possible Reasons for Difference
Boundary condition set-up
ideal gas, pressure based, real apparatus error (axial temperature gradient, end walls effect)
“This discrepancy arises because of the variation of axial wave length with axial distance, which results from the thermal conditions at the ends of the experimental apparatus.” [R. Kedia (1997)]
Simple Model Test
R1 = 96.85 mmR2 = 97.5 mmHeight = 140 mm Q=4 L/min Vin= 0.000168 m/sTin = 308K Tout= 551KΩ=29.311 rad/sEnd walls are fixed and insulated
Measure points are located in the vertical lines close to the inner cylinder.
Since for η=0.975 Rec= 260.978,In this case Re=1837.075So, it is in laminar mode.
Plans for Next Period
Keep running both of the turbulent cases
Finish the thermal validation
Repeat Taylor-couette validation with full length
Wavy validation should be finished with running 0.95 case long enough
More validation of the thermal part (optional)
Keep turbulent case running
Finish simple model test
Check geometry related paper