等方乱流に埋め込まれた周期運動

Statistics of Passive Vectors in
an Unstable Periodic Flow of
Couette System
S. Kida (Kyoto Univ.)
with T. Watanabe, T. Taya
Couette System
U
2h
-U
Couette System
U
2h
-U
Simulation range ( 5.513h, 2h , 3.770h )
Number of modes (16×31×16)
Minimal box
Jimenez & Moin 1991
Number of independent variables 15,422
Re = Uh /ν= 400
Unstable periodic flow
Kawahara & Kida 2001
Couette Turbulence
U
- 0.3U
by G. Kawahara
Unstable Periodic Flow
U
- 0.3U
by G. Kawahara
Turbulence vs UPF
Same
Low-Order
Statistics
Unstable Periodic Flow
0.3
0.4
0.7
0.2
0.1
t/T=0
0.5
0.8
0.6
0.9
Turbulence vs UPF
Chaotic
Unpredictable
Periodic
Repeatable
Same Low-Order Statistics
Not easy
Easier to increase the
accuracy of statistics
Subject
We consider turbulence mixing
by examining the statistics of passive vectors
advected in an unstable periodic flow (UPF)
for a Couette system.
We focus our attention on the relation between
the coherent structures (streamwise vortices, lowspeed streaks, ejection regions, sweep regions)
and the mixing properties of turbulence.
Orientation of Passive Vectors
Laminar flow
Lyapunov exponent = 0
Aligned slowly
Turbulent flow
Aligned quickly
Lyapunov exponent > 0
Alignment of Directions
A unit sphere
represents the
direction of
passive vectors
z
y
x
Initially, we distribute many passive vectors at
a same point with random orientations.
Alignment of passive vectors
planar
linear
Alignment of passive vectors
20
38
56
We distribute 5320 sets of passive vectors uniformly
in space at an initial instant.
Each set is composed of 1000 passive vectors with
random orientations.
Alignment of Passive Vectors
Alignment of Passive Vectors
Direction Field of Passive Vectors
Passive vectors which start at a same position
with random orientations will be aligned in a
few periods of UPF.
Then, can we expect that the direction field of
passive vectors in UPF is unique irrespective
of the initial condition ?
Construction of Direction Field
Direction distribution of passive vectors
2h
20
38
56
Divide the computational box into many cubes of side 0.1h.
Track the motion of many(5320)passive vectors for a
long time, and
Store the position and direction of each vector fallen
in the individual cubes every time phase of the UPF.
Calculate the direction distribution of vectors in each cube.
Classification of Directional Distribution
linear
planar
scattered
Classification of direction
distribution of passive vectors
Number of PV in a cube: N
Directional Vector
Direction matrix
Direction Matrix
Number of PV in the
∝
direction of
Eigenvalues of
Eigenvectors:
Classification of the Type of Directional Distribution
A reference distribution
to be compared
Classification of Directional Distribution
linear
planar
scattered
Directional Distribution
planar
scattered
linear
Classification of Directional Distribution
linear
planar
scattered
Relative Population
Cube size dependence
Relative Population
Cube size dependence
Where are the linear and planar
types located in the flow field ?
Direction field and coherent structure
linear
planar
scattered
Vortex center
Periphery of vortices
Direction field and coherent structure
linear
planar
scattered
Vortex center
Ejection region
Periphery of vortices
Sweep region
Summary(1/2)
Passive vectors which start at a same position with
different orientations in UPF will be aligned in a few
time periods.
The direction field of passive vectors can be defined
uniquely in UPF.
It is found that in most of small cubes the passive
vectors align either in a single direction (linear
type) or on a plane (planar type).
Summary(2/2)
The linear type is observed in streamwise vortices and
low-speed streaks (or the ejection region near
the wall), the planar type in the periphery of vortices
and in the sweep region near the wall.
The dispersion of orientation of passive vectors may
be used to quantify turbulent mixing.
Dependence on the cube size
Vortex Center
Vortex Periphery
線素の方向の変動と流線
αβ
t=0.5T
方位角とその変化
渦中心域
渦外縁域
向壁面域
Past history
過去
距離
現在
方向
向壁面
大
大
渦外縁
小
大
渦中心
大
小
壁面沿い
小
小
領域
方向の偏差値
距離の偏差値
A cross-section of the direction field
of passive vectors
線素の方向場の断面
線素の方向の標準偏差
線素の方向の標準偏差
Unstable Periodic Flow
Kawahara & Kida 2001
UPF vs Turbulence
UPF
Turbulence
UPF vs Turbulence
UPF
Turbulence
UPF vs Turbulence
UPF
Turbulence
Alignment of passive vectors in direction
Many (5320)passive vectors which start with random
orientations will align in direction after a few periods
of UPF
planar
random
linear
対称性
(1)スパン方向に垂直な平面 z = 0 に関する
鏡映と主流方向への半周期(Lx/2)移動
(すべり鏡映)。
(2) チャネルの中央を通るスパン方向の直線
(x = y = 0)に関する 180o回転とスパン方向
への半周期(Lz/2)移動。
y
活発なよどみ点
静穏なよどみ点
x
z
向壁面域
向壁面域
Alignment of Directions
z
y
x
After a while, all the test particles align in a single direction.
Passive Vector Field
The alignment of orientation of passive vectors in an unstable
periodic flow (perhaps in a turbulent flow too) implies that their
orientation (therefore their stretching rate as well) is a function
of space and time determined uniquely irrespective of their
history earlier than some finite time.
Thus we can define the fields of various physical quantities
associated with passive vectors, which are called the passive
vector fields.
Passive Vector Field
In the case of unstable periodic flows, in particular, the passive
vector field is also periodic in time.
This property enables us to calculate the passive vector
fields by particle-pair simulation.
For example, the orientation field can be obtained by statistical
average, over a sufficiently long time, of the relative position
vector of many particle pairs near each simulation grid point.
Orientation Field
20
38
56
Divide the computational box into many cubes.
Integrate the motion of many particle pairs for a long
time, and store the relative position vector of the pairs
in each cube at each temporal phase of the unstable
periodic flow.
Estimate the orientation vector at the center of each
cube by a linear interpolation.
Linear Interpolation
Let us assume there are N (≧ 4) particle pairs
in a cube centered at (x, y, z). Let (xi, yi, zi) be
the center of i-th pair (i = 1, 2, …, N),and (pi,
qi, ri) be the orientation vector (pi2+qi2+ri2=1) .
Then, the orientation field at point (x, y, z) may be estimated
by a, which minimizes the following function:
where fi is pi, qi or ri .
Stretching Field
Threshold is 0.35
0.8 T
Temporal Variation of Mean Stretching Rate
Time average
0.8
Strain Characteristics
Rate-of-strain tensor
Eigenvalues
Eigenvectors
Strain Characteristics
Rate-of-strain tensor
Eigenvalues
Eigenvectors
Correlation bet. Stretching & Strain Rates
λ1
λ2
γ
γ
λ3
0.8 T
γ
Spatial Symmetry
(1) Reflection w.r.t. plane z=0 and translation
along x axis by Lx/2.
(2) 180o rotation around line (x = y = 0) and
translation along z by Lz/2.
y
x
Active S.P.
Gentle S.P.
z
Alignment of Directions
ある基準粒子の近傍に,等しい距離 (d) だけ離れ,
方向がランダムに分布した多数のテスト粒子を配置
する。数値シミュレーションの各ステップで基準粒子
との距離を d に戻す。しばらく(3周期程度)すると,
テスト粒子の方向がそろってくる。
線素の方向の整列
時間位相 0.7 T において,ある立方体(一辺 0.1 h)に落ち
こんだ粒子対の方向(左図)と中心の位置座標(右図)。
Main Results
Unstable periodic flow varies periodically in time.
That is, the flow state repeats in a constant period at
all the spatial points.
In this flow we submerge passive line elements at
arbitrary time, at arbitrary position and in arbitrary
direction.
After a while (about three periods), the direction of
line elements (therefore their stretching rate) is
determined uniquely depending upon the time and
place independent of their initial directions.
Strain and Stretching
Rates
Dark blue : Strain rate (0.26)
Light blue : 1-st Eigenvalue (0.55)
0.8 T