Inclined to exchange: shocking gravity · PDF file · 2007-11-19X n odd 2 πn...
Transcript of Inclined to exchange: shocking gravity · PDF file · 2007-11-19X n odd 2 πn...
Inclined to exchange:shocking gravity currents
John Hinch (DAMTP)
and at FAST/Paris:
Jean-Pierre Hulin, Dominique Salin & Bernard Perrin
Thomas Seon & Jemil Znaien
Pipemix - Nottingham 2007 – p.1/25
Regimes
Vertical : turbulent
Inclined : nosed controlled gravity current
Horizontal : viscous counter-current
Pipemix - Nottingham 2007 – p.3/25
Speed of front Vf
Vt =√
Atgd (inertial nose) Vν =Atgd2
ν(viscous)
1.0
0.8
0.6
0.4
0.2
0.0
Vf/Vt
1501000
Vν sin α/Vt = Ret sin α
50
horizontal tube
Tilt angles 0 < α < 30o;At = 3.5 × 10−2 (•), 10−2
(�), 4× 10−3 (�), 10−3 (H),4 × 10−4 (N) for a viscosityµ = 10−3 Pa.s;and µ = 10−3 Pa.s (�), 4 ×10−3 Pa.s (⊞) for a densitycontrast At = 10−2
Vf ≈ 0.014Vν sin α
Pipemix - Nottingham 2007 – p.4/25
A little theory
Counter current in a circle at 50% fill
∇2u = f(θ) =
{
−12 in 0 ≤ θ < π,
+12 in π ≤ θ < 2π
and u = 0 on r = 1.
Fourier series
f(θ) = −∑
n odd
2
πnsin nθ,
so
u =∑
n odd
2
πn(n2 − 4)(r2 − rn) sin nθ.
Pipemix - Nottingham 2007 – p.5/25
a little theory continued
Hence flux
Q =∑
n odd
1
πn2(n + 2)2
=π
16− 1
2π
= 0.037195
Front velocity for area A = π2
Vf =Q
A= 0.012,
experiment 0.014 – eventually will do better.
Pipemix - Nottingham 2007 – p.6/25
Horizontal tube
Xf = Vf
∝ R2
µ
dp
dx
∝ R2
µ
∆ρgR
Xf
Hence
Xf =√
Dt
D = 0.0108Vνd in experiment
1.0
0.8
0.6
0.4
0.2
0.0
-20 -10 0 10 20x/t0.5
25 s
75 s
125 s
175 s
225 s
1-1 0.5-0.5 0
distance from gate valve (m)
1.0
0.8
0.6
0.4
0.2
0.0
Norm
alized co
ncen
tration
25 s
75 s
125 s
175 s
225 s
(a)
(b)
Norm
alized co
ncen
tration
Pipemix - Nottingham 2007 – p.7/25
2D spreading in a horizontal channel
µ∂2u
∂y2=
dp±dx
in 0 ≤ y ≤ h and h ≤ y ≤ a,
with u = 0 on y = 0 and a.
No net flux
∫
u dy = 0 and
[
dp
dx
]+
−= ∆ρg
∂h
∂x.
Hence∂h
∂t=
∂
∂x
(
∆ρgh3(a − h)3
3µa3
∂h
∂x
)
.
Pipemix - Nottingham 2007 – p.8/25
Spreading in a horizontal channel, continued
0
0.2
0.4
0.6
0.8
1
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
h
x/t0.5
HenceXf =
√Dt
with D = 0.017Vνa, cf experiment 0.0108.
Pipemix - Nottingham 2007 – p.9/25
Spreading in a horizontal circleFlat interface at y = h(x, t). Flow in cross-section
∇2u = G +
{
−12 in h ≤ y ≤ 1,
+12 in −1 ≤ y < h
and u = 0 on r = 1. G so no net flux.
Numerical (finite difference 80 × 80, sor) for flux Q(h) in h ≤ y ≤ 1.
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
h
Q
Pipemix - Nottingham 2007 – p.10/25
Thin layer asymptotics, 1 − |h| ≪ 1
No net flow ⇒ all pressure gradient in thin layer.
Thin layer ⇒ no stress by thick on thin.
Wide compared with thin ⇒ 2D, half parabolic profile.
Q(h) ∼ 32√
2
105(1 − |h|)7/2 .
Symmetric and smooth Q(h), so equally
Q(h) ∼ 4
105
(
1 − h2)7/2
.
Pipemix - Nottingham 2007 – p.11/25
Lucky break
Thin layer asymptotic’s4
105= 0.038095
50% fill Q(0) = 0.037195
Hence good approximation (within 1%):
Q(h) ≈ Q(0)(1 − h2)7/2
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
h
Q
Pipemix - Nottingham 2007 – p.12/25
Nonlinear diffusion equation
∂A(h)
∂t+
∂Q(h)
∂x= 0,
with cross-sectional area A(h) = 2 cos−1 h − h√
1 − h2, sodA/dh = −2
√1 − h2. Hence
∂h
∂t=
1
2√
1 − h2
∂
∂x
(
(1 − h2)7/2 ∂h
∂x
)
.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4
h
x/√
Dt
Xf =√
Dt
with D = 0.0108Vνd,experiment 0.0108 ± 0.0008.
Pipemix - Nottingham 2007 – p.13/25
Spreading in nearly horizontal tubesFlow now driven by difference in pressure gradients in two fluids
∆ρg
(
cos α∂h
∂x+ sin α
)
.
So with suitable rescaling
∂h
∂t=
1
2√
1 − h2
∂
∂x
(
(1 − h2)7/2
(
∂h
∂x+ 1
))
.
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6
h
x
t = 0.5, (0.5), 5.0.
Pipemix - Nottingham 2007 – p.15/25
Motion of front Xf(t)
0
1
2
3
4
5
6
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Xf
t
Diffusive spreading only at very short times.
Slow to attain long-time velocity.
Pipemix - Nottingham 2007 – p.16/25
Experimental motion of front Xf(t)
6
5
4
3
2
1
0
2 X
f/(d c
ot
α)
6543210
t (8 F(0)Vν sin
2α) /(d cos α)
α = 1o ◦,2o △,3o ▽.
Experiments straighter due to different early behaviour – gate release and a short time
limited by inertia
Pipemix - Nottingham 2007 – p.17/25
Long time approximation?
Xf (t) ∼ 0.856t + 1.01 4 < t < 5
0.830t + 1.19 7 < t < 11
0.774t + 1.83 20 < t < 25
0.750t + 2.82 90 < t < 100
0.746t + 3.36 145 < t < 150
Linear fit fails.
Pipemix - Nottingham 2007 – p.18/25
Logarithms at long timeCrude model: horizontal + inclined spreading
Xf =VfL
Xf+ Vf ,
with solutionVf t = Xf − L log(1 +
Xf
L).
Vf L
0.745 0.798 0 < t < 5
0.735 0.851 0 < t < 20
0.740 0.812 0 < t < 50
0.742 0.791 0 < t < 100
0.742 0.789 0 < t < 200
Slow logarithmic approach also in experimental data.
Pipemix - Nottingham 2007 – p.19/25
Test Vf = 0.742
20
15
10
5
0
0.10.080.060.040.020
Vf
(mm/s)
sinα
0 α (°)2 4
∞
Continuous line is new 0.742 (0.014Vν sin α),dashed is old 0.637 (0.012Vν sin α).
Pipemix - Nottingham 2007 – p.20/25
But why Vf = 0.742?
∂h
∂t=
1
2√
1 − h2
∂
∂x
(
(1 − h2)7/2
(
∂h
∂x+ 1
))
.
For hx ≪ 1
∂h
∂t+ c(h)
∂h
∂x= 0 with kinematic wave speed c(h) = 7
2h(1 − h2)2.
Test rarefactionwave solution byplotting h vs x/t.
But must shock 0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
t=2550
100200
h
x/tPipemix - Nottingham 2007 – p.21/25
Shock waves
Speed of shock V (h) =Q(h)
A(h)must equal speed of arriving characteristics c(h).
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
t=2550
100200
h
x/t
Equal at h = 0.23817 with c = V = 0.74172.
Pipemix - Nottingham 2007 – p.22/25
Form of shock waves
Switch to frame moving with V , then steady form governed by
∂h
∂x= V
A(h)
Q(h)− 1
0
0.2
0.4
0.6
0.8
1
-80 -70 -60 -50 -40 -30 -20 -10 0 10
t=2550
100200
h
x − Xf (t)
Pipemix - Nottingham 2007 – p.23/25
Conclusions
Front is a shockwave:
speed selected by matching onto a rarefaction wave.
Logarithmic approach to long-time:
value depends on initial condition.
Pipemix - Nottingham 2007 – p.24/25