kumamoto univ. x-earth center toshifumimukunoki chiaki nagaisfb837.sd.rub.de/downloads/presentation...
Post on 18-Aug-2019
218 Views
Preview:
TRANSCRIPT
Evaluation of oil contamination in porous media by X-ray CT image analysis and LBM
simulation
Kumamoto Univ. X-Earth CenterToshifumi Mukunoki
Chiaki Nagai
1
BASICS OF X-RAY CT METHOD
GENERAL PRINCIPLE, VOXEL AND CT VALUE
n voxels
n voxels
thickness of the slice
CT image is a digital image.
Figure 17
Bulk density (t/m3)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
CT
valu
e
-1000
-500
0
500
1000CT value = 943.2 - 987.18AirWaterRatio of fine grains: 0 %Ratio of fine grains: 30 %Ratio of fine grains: 50 %Ratio of fine grains: 70 %Ratio of fine grains: 100 %Equation [4]
10001000 dvalueCT
…………
Reconstruction of 3-D CT image
IMPORTANT FEATURES IN IMAGE ANALYSIS
•THRESHOLD VALUE•BEAM HARDENING EFFECT•PARTIAL VOLUME EFFECT
THRESHOLDING VALUE
(a) 0<CT-value<500
(b) –1000<CT-value<500(II) CT image
Figure 3
(I) Histogram of CT value
0
3000
6000
9000
-1500 -1000 -500 0 500 1000 1500CT-values
Freq
uenc
y
(b) 256levels
WW
WL
(a) 256levels
PARTIAL VOLUME EFFECT
The size of all particles is greater than a voxel dimension
Baseline
X-ray beamSoil particle
CASE1
BaselineThe size of some particles is greater or smaller than a voxel dimension
CASE2
BaselineThe size of all particles is smaller than a voxel dimension
CASE3
(a)
Figure 14
Location along the diameter of spesimen (mm)
-30 -20 -10 0 10 20 30C
T-va
lue
-1500
-1000
-500
0
500
1000
1500
2000
(b)
CASE1
CT-value for air
(a)
CASE3
Location along a diameter of image (mm)
CT-
valu
e
Figure 15
(b)
-2000
-1000
0
1000
2000
-40 -20 0 20 40
IMAGE PROCESSING ANALYSIS TO X-RAY CT
IMAGE
12
X-ray source
Detectors
Specimen table
①
②
③ ④
⑤⑥
⑦
⑧ ⑨
scan area
100
300
840
Fig. 4
⑩
⑪⑫
⑬
Initial step B step C step D step Estep A
70mm
55mm
40mm
25mm
10mm
Hei
ght f
rom
the
botto
m o
f spe
cim
en
Strain level
Specimen
SMOOTHING TECHNIQUE
Basic analysis of X-ray CT image(2) 11pixel
11pi
xel
(a) (b) (c)
Fig. 15
0
200
400
600
800
1000
-40 -20 0 20 40
step Astep F
CT-
valu
e
Location of a voxel from the center of the specimen
25mm
500
550
600
650
700
750
800
-30 -25 -20 -15 -10 -5 0location along radius(mm)
CT-
valu
e
initialstep Astep Bstep Cstep Dstep E
MAIN TOPIC
Introduction
It is important to know the mechanism of LNAPL migration
LNAPL:Light- non aqueous phase liquids (gasoline, diesel, etc)
Residual LNAPL causes long-term pollution
Ground contamination
Source
Vadoze zone
Saturated zone
Soil particlesResidualLNAPL
pore waterPore scale
17 Motivation
(~1μm)
Grain
Water
LNAPL
Air
Pore-Scale
Core-Scale
Small-size blockLarge-size block
(1cm~)
(1m~) (10m~)Representative
parameter
Ex.Pore-Scale
(C. Pan et al,2004)
Numerical Analysis for Micro scale
Avergae
0)( gpKk
tS r
Numerical Analysis for Macro scale
Introduction
Soil Particles ResidualLNAPL
Pore waterPore scale
19
Transfer mechanism of oil flow
Pore structure
Pore Pore size
Pore Pore shape Connectivity
Fluid Fluid distribution
Velocity and
distribution
Velocity and Pressure
distribution
Image analysis
CFD analysis
Objective
To clarify the transfer mechanism of oil and water in pore structure of sandy soil using Image analysis and CFD analysis
Image analysisGeneration mechanism of blobs in sandy soils
LBM analysisLBM simulation can be performed using MXCT image
MXCT image
20
APPLICATION OF M-FOCUSED X-RAY CT SCANNER FOR THIS
STUDY
M-FOCUSED X-RAY CT SCANNER GIVES THE SPATIAL DISTRIBUTION OF MATERIAL DENSITY. A CT IMAGE IS NOT A
MICROSCOPE PHOTOGRAPH BUT A DIGITAL IMAGE.
Power of voltage (kV) 150Current (mA) 180
Numver of views 1500Number of integration
treatment10
Voxel dimension (mm) 4.43 x 4.43 x5.0Number of voxel 1024 x 1024 x 1000
Imgae processing
23
NOWADAYS, SO MANY USEFUL APPLICATIONS ARE DISTRIBUTED IN THE
IMAGE ENGINEERING• IMAGE J (ONE OF POPULAR APPLICATION FOR CT USERS)
• ITK (IMAGE TOOL KIT: LANGUAGE IS C++
• MOST OF CASE, WE DON’T NEED TO DEVELOP VERY NEW ALGORISM WHICH WE WANT TO DO IMAGE ANALYSIS.
24
Decision of analyzing area
100voxels 200voxels 300voxels
400voxels 500voxels 600voxels 700voxels
25
Decision of analyzing area Porosity
38.63
37.65
38.7138.82
39.1439.30 39.26
37.5
38
38.5
39
39.5
0 100 200 300 400 500 600 700
poro
sity
(%)
range of analysis (voxel)
試料作製時
の間隙率
39.40
The are for evaluation of porosity is changed and then, 500 voxels should be enough dimensions because errors are within 1 %.
26
27
100 190200 195~200300 200~205400 200~205500 200~205600 200~205700 200~205
試料作製時の平均粒径
200
一辺のサイズ(voxel)
平均粒径(μm)
Grain size distribution
Image J has the option called “3D object counter” and then, its data can be output as text data, such as number of particles, long and short diameter. We analyze those value using Excel, whatever.
As for grain size distribution, dimension of 300 cubic voxel is enough for evaluation of grain size distribution.
0102030405060708090
100
0.025 0.050 0.100 0.200 0.400
Pass
mas
s per
cent
age (
%)
Diameter(mm)
100200300400500600700Sieving test
When 1voxel =5μmWe concluded that we should use 500 cubic voxels for image analysis and numerical analysis.
28
3D distribution of pore size
XBBX XXX
B |Mathematical Morphology: Opening and Closing
3D binary image
Pore
Sphere element
1 5
10 15
Pore sizeParticles Large pore It is possible to extract pore structure and
pore size
3D distribution of pore size
Toyoura sand
Glass beads
29
Small
Large
Pore size
02468
1012141618
5 15 25 35 45 55 65 75 85 95 105
115
125
135
145
155
165
175
185
195
205
215
225
Freq
uenc
y, %
Pore diameter, μm
02468
1012141618
5 15 25 35 45 55 65 75 85 95 105
115
125
135
145
155
165
175
185
195
205
215
225
Freq
uenc
y, %
Pore diameter, μm
Toyoura sand
Glass beads
NAPL INJECTION TEST
30
1D INJECTION TEST
Water or NAPL is injected1㎝
ParticlesResidualLNAPL
Pore water
Particles
Pore water
Particles
Pore water
LNAPL
20cm
31
Sample Mean diameter (mm)
Particle density (t/m3)
Porosity (%)
Case 1 Glass beads 300 2.45 36.0
Case 2 Toyoura sands 200 2.65 39.4
Test Method
土粒子密度 乾燥密度
(t/m3) (t/m
3)
CASE1 ガラスビーズ 250-350 2.45 1.57 0.89
CASE2 ガラスビーズ2種 250-350と1000の混合 2.45 1.75 0.89
CASE3 ガラスパウダー 250-450 2.45 1.49 0.89
CASE4 豊浦砂 200 2.64 1.6 0.87
試料 粒径(μm) 相対密度
Injection test apparatus Illustration
Flow
Sample should be fully
saturatedLNAPL was
injectedKI solution was
injected CT scan
Precision balance
200mm
Syringe pump
Pressure sensor10mm
KI solution was used.For saturation, we injected CO2 gas using 120 kPa pressure.
4. Test materials
2.5mm
ρ=1.02t/m3 ρ=1.05t/m3 ρ=1.25t/m3
KI solution Water
Density (t/m3) 1.25 1.00
Surface tension (mN/m) 72.45 72.94
Interfacial tension withLNAPL (mN/m)
Contact angle (°) 61.5 53.7
Viscosity (mPa・sec) 0.966 1.002
KI ρ=1.25t/m3 Main properties (18℃)
54.5 52.9
SoilKI solution
SoilKI solution
2.5mm
CT images of sample before and after KI solution injection10
CASE1( Ca = 1.57×10-6)
CASE2( Ca = 3.14×10-5)
5mm
5mm
Resolution:1024×1024×800
1voxel : 5×5×5μm
Resolution:1024×1024×800
1voxel : 5×5×5μm
3D pore distribution
Binary Image
Pore
Pore with residual LNAPL NAPL
3D pore structure
185μm
0μm
間隙径
35
Evaluation of residual LNAPL and KI solution
36
Residual LNAPL KI solution Residual LNAPL KI Solution
Glass beads Toyoura sand
Sample should be
fully saturatedLNAPL was
injectedKI solution
injected CT scan
ResidualLNAPL
KI solution
Conclusions
• Micro X-ray CT scanner is a powerfultool to evaluate pore structure ofsandy soil and its techniques can beapplied to the evaluation of porestructure with LNAPL.
37
Ongoing work• To perform LBM simulation using X-
ray CT image obtained from these test.
Lattice Boltzmann Method (LBM)
38
LBM - INTRODUCTIONLBM models the fluid consisting of fictive particles, and such particlesperform consecutive propagation and collision processes over adiscrete lattice mesh.
Advantages of LBM over traditional methods:Allows modelling of multi-phase behaviour at local scaleAllows dealing with complex boundariesAllows incorporation of microscopic interactionsAllows parallelization of algorithm (for example using GPU)
Disadvantages of LBM:Limited memory and mesh size (depends on efficiency ofparallelization modelling and hardware)Lack of use of classic physical parameters
LNAPL
Water
Glassbead
Wall
parameters LNAPL WaterDensity ratio Relaxation time (τ) 1.167 1fluid-fluid Interaction (gf)fluid-solid Interaction (gs) -0.020 0.020
1
0.0015
ρN=100
ρW=90
kk
kP7
3
λk:流体の圧縮性に関係するパラメータ
40
LBM simulation for two-phase flow in porous materials
33
0
20
Poresize(voxels)
LBM simulation 3D distribution of pore size
02468
1012141618
0.005 0.025 0.045 0.065 0.085 0.105 0.125 0.145 0.165
Nor
mal
izin
g va
lue(
%)
poresize(mm)
ALLLNAPLEntire pore size
Pore size with LNAPL
41
LBM simulation for two-phase flow in porous materials
Water
LNAPL
parameters LNAPL WaterDensity ratio Relaxation time (τ) 1.167 1fluid-fluid Interaction (gf)fluid-solid Interaction (gs) -0.020 0.020
1
0.0015k
kkP
73
λk:流体の圧縮性に関係するパラメータ
42
Pipe model for LBM simulation
Comparison porous material with pipe modelGlass beads Pipe model
43
Discussion of results for pipe model
1 2 3 4 5
1 2 3 4 5
1 2 3 4 5
AB C
D E FG H I
Definition of each areaPipe model
5
5
1nnX
Mean diameter along the flow direction =
X:location(A~I)
Aspect ratio along the flow direction =
4
4
11
nnn XX
X:場所(A~I)
44
db
ca
0
3
6
0.06 0.08 0.1
インクビン効果係数
平均間隙径(㎜)
a b
c d
45
Discussion of results for pipe model
5
5
1nnX
Mean diameter along the flow direction =
X:location(A~I)
Aspect ratio along the flow direction =
4
4
11
nnn XX
X:場所(A~I)0
3
6
0.06 0.08 0.1
インクビン
効果係数
平均間隙径(㎜)
ABCDEFGHI
0
50
100
150
200
250
300
A B C D E F G H I
Pene
tratio
n di
stan
ce (v
oxel
)
Area各領域(A~I)におけるLNAPLが浸透した距離
46
a
c d
b
a b
c dConnectivity of pore structure for LNAPL flow means the large mean diameter and less aspect ratio along flow direction.
Discussion of results for pipe model
47
Conclusions
poresmall large
X-ray CT image analysis and LBM simulationcan Evaluate LNAPL migration in porous media.
Conclusions
(~1μm)
Grain
Water
LNAPL
Air
Pore-Scale
Core-Scale
Small-size blockLarge-size block
(1cm~)
(1m~) (10m~)Representative
parameter
Average
0)( gpKk
tS r
We don’t say X-ray CT scanner can solve many geotechnical and geoenvironmental issues. X-ray CT scanner is a powerful tool to observe the inner condition of materials. Of course, the sample size is too small to evaluate real condition. Hence, we need to develop the model to connect micro observation and macro observation.
top related