Separating mixtures and solutions: Chromatography Mixtures and Solutions.
Quantification of Gas Mixtures Using Imaging ... - FlareNet
1
Rodrigo B. Miguel 1 , Johannes Emmert 1,2 , Dilan Avşar 1 , Jean-Philippe Gagnon 3 , and Kyle J. Daun 1 1 Department of Mechanical and Mechatronics Engineering, University of Waterloo; 2 Department of Reactive Flows and Diagnostics, Technical University Darmstadt; 3 Telops Inc. Quantification of Gas Mixtures Using Imaging Fourier Transform Spectrometry Motivation The oil and gas industry needs to quantify gaseous emissions in various scenarios, e.g. reporting and mitigating fugitive emissions and assessing flare combustion efficiency. Optical gas imaging provides a two-dimensional representation of concentration in the camera field-of-view. It is passive and nonintrusive, qualifying it for fenceline measurements. Most optical gas imaging techniques focus on quantifying a single species. However, many gas releases consist of mixtures (e,g, BTEX, natural gas, combustion gases, etc. ). These can be quantified using an imaging Fourier transform spectrometer (IFTS). Measurement model The spectral intensity I lη incident in each pixel is related to the gas concentration and temperature along the line- of-sight (LOS) by the radiative transfer equation (RTE). The IFTS data can be used to simultaneously infer the temperature and species column densities. Species column densities are combined with the velocity field computed from a sequence of images to obtain mass flux. Optical flow velocimetry Spectroscopic model Gas State Intensity Residual I Lη = f(X CH4 ,σ, ΔT) ||b meas - b mod (X CH4 ,σ, ΔT)|| 2 X CH4 s ΔT s Column Density L 0 b M X(s)p ds A k T(s) ρ= ∫ 6×σ Flare combustion efficiency The IFTS can also infer flare combustion efficiency (CE), the ratio of the mass of carbon affixed to CO 2 to the total mass of carbon in the fuel stream. Flare CE may be impacted by cross-winds (fuel stripping) and steam-assist 4 . Synthetic MW Hyper-Cam data was generated from a CFD large-eddy simulation of a flare in a crosswind at a resolution of 4 cm -1 . Peak species concentrations are inferred through nonlinear regression of modeled data to simulated measurements. CH 4 mass flow Conclusions and future work Imaging Fourier transform spectrometers can estimate mass fluxes of gas mixtures. Methane flux estimates from the Hyper-Cam LW were consistent with ground-truth data. More accurate results can be found using the Hyper-Cam LW Methane (more methane lines) or the Hyper-Cam MW (improved temperature sensitivity). The Hyper-Cam MW can also estimate CO 2 and CH 4 column densities in a flare, which can be used to assess combustion efficiency. The Hyper-Cam MW will be used measure flare CE in the Boundary Layer Wind Tunnel at Western University, and later in a flare at an industrial site. 1 2 3 4 5 6 7 **LES Parameters: Stack ID 10.3 cm, 25.6 g/s of CH4, crosswind at 1.9 m/s [5] CO2 pixel CO2 CH4 N CE N N ≈ + Imaging Fourier-transform spectrometer (IFTS) The IFTS generates two-dimensional interferograms, which are transformed into a hypercube of spectally-resolved images. The LW Hyper-Cam is used to measure a 10 slpm CH 4 release at 10 K above ambient temperature with a cloudy sky as background. The data was used to obtain CH 4 column densities in each pixel, which is combined with the estimated velocity field to infer the methane mass flow. OPD /2 Beam splitter Fixed mirror Moving mirror Sensor Source CH 4 column density in g/m 2 x in mm z in mm 50 100 150 200 25 20 15 10 5 0 50 100 150 200 250 normal velocity in m/s x in mm z in mm 50 100 150 200 0.35 0.3 0.25 0.2 0.15 0 50 100 150 200 250 0.1 0.05 Pixel interferogram FT OPD Intensity Intensity Wavenumber Model Resolution [pixels] FOV Spectral Resolution [cm -1 ] Spectral Range [μm ] NESR [W/m 2 sr cm -1 ] Hyper-Cam LW 320×256 6.4 o ×5.1 o Up to 0.25 7.7-11.8 (850-1300 cm -1 ) 20 ×10 -5 Hyper-Cam MW 320×256 6.4 o ×5.1 o Up to 0.25 3-5 (2000-3330 cm -1 ) 4 to 7 ×10 -5 Re-absorbed Blackbody emission Transmittance (1−α η ) Background () { } () () { } L L 0 0 L L b 0 S I I exp - s ds (s) I s exp - s ds ds η η η η η η = κ ′ + κ κ ∫ ∫ ∫ Absorption coefficient * X CH4 =0.2@298K, X CO2 =0.2, X H2O =0.4@400K 0 L I Lη I 0η 2000 2400 2800 3200 0 50 800 1000 1200 1400 0 3.5 κ η in cm -1 * η in cm -1 η in cm -1 CH 4 CO 2 H 2 O×100 H 2 O Outside (weather data) ΔT(fitted) Inside (IFTS sensor) s Temperature s Background H 2 O (fitted using BG pixels) CH 4 (fitted) 100 m L (fitted) Molar fraction The background intensity, I 0η , was obtained from pixels that did not contain CH 4 . The background temperature and ambient humidity were fitted assuming ambient temperature from meteorological data. Species molar fraction and temperature were parametrized as Gaussian profiles along the LOS 2 , and the species absorption coefficient is simulated using the HITRAN spectral line database. Peak molar fraction, peak gas temperature, and plume thickness at four heights were inferred from measured spectra of 4 cm -1 resolution. 1 2 3 20 40 60 80 100 20 40 60 80 100 120 vertical pixels horizontal pixels 2 4 6 8 10 12 14 16 18 75 85 25 35 45 55 65 75 85 95 105 115 CO 2 g/m 2 pixels pixels 75 85 25 35 45 55 65 75 85 95 105 115 0.2 0.4 0.6 0.8 1 1.2 CH 4 g/m 2 pixels 0.3 0.4 0.5 0.6 0.7 0.8 0.9 75 85 25 35 45 55 65 75 85 95 105 115 CE PIXEL pixels Single point sample overestimates global CE The velocity field is estimated by the evolution of brightness patterns in the interferogram data-cube. Changes in pixel brightness between successive timeframes are caused by the CH 4 flow. Image data-cubes are used to derive the temporal and spatial derivatives by the brightness constancy assumption 3 . The underdetermined system of equations is solved by combining the gradient constraint with Laplacian priors on the flow velocities. u v 0 x y t ∂Ι ∂Ι ∂Ι + + = ∂ ∂ ∂ References [1] “Thermal Infrared Cameras.” Telops, https://www.telops.com/. [2] Grauer, S. J. et all., “Gaussian model for emission rate measurement of heated plumes using hyperspectral data,” JQRST, 2018. [3] Horn, B. K., & Schunck, B. G. “Determining optical flow,” Artificial Intelligence, 1981. [4] Johnson, M. R. et all., “A Fuel Stripping Mechanism for Wake-Stabilized Jet Diffusion Flames in Crossflow,” Combust. Sci. Technol., 2001. [5] Flare LES provided by Jeremy Thornock from the Department of Chemical Engineering at The University of Utah. [6] Quantifying & Reducing Impacts of Global Gas Flaring. (2019, October 9). Retrieved from https://www.flarenet.ca/. [1] [6] nozzle Video of the optical flow results Hyper-Cam LW Hyper-Cam MW
Transcript of Quantification of Gas Mixtures Using Imaging ... - FlareNet
Rodrigo B. Miguel1, Johannes Emmert1,2, Dilan Avşar1, Jean-Philippe Gagnon3, and Kyle J. Daun11Department of Mechanical and Mechatronics Engineering, University of Waterloo; 2Department of Reactive Flows and Diagnostics, Technical University Darmstadt; 3Telops Inc.
Quantification of Gas Mixtures Using Imaging Fourier Transform Spectrometry
Motivation The oil and gas industry needs to quantify gaseous emissions in various scenarios, e.g. reporting and
mitigating fugitive emissions and assessing flare combustion efficiency. Optical gas imaging provides a two-dimensional representation of concentration in the camera field-of-view.
It is passive and nonintrusive, qualifying it for fenceline measurements. Most optical gas imaging techniques focus on quantifying a single species. However, many gas releases
consist of mixtures (e,g, BTEX, natural gas, combustion gases, etc. ). These can be quantified using an imaging Fourier transform spectrometer (IFTS).
Measurement model The spectral intensity Ilη incident in each pixel is related
to the gas concentration and temperature along the line-of-sight (LOS) by the radiative transfer equation (RTE).
The IFTS data can be used to simultaneously infer the temperature and species column densities.
Species column densities are combined with the velocity field computed from a sequence of images to obtain mass flux.
Optical flow velocimetrySpectroscopic model
Gas State
Intensity ResidualILη = f(XCH4,σ, ΔT) ||bmeas - bmod(XCH4,σ, ΔT)||2XCH4 s
ΔT s Column DensityL
0b
M X(s)p dsA k T(s)
ρ = ∫6×σ
Flare combustion efficiency The IFTS can also infer flare combustion efficiency (CE), the ratio of the mass of
carbon affixed to CO2 to the total mass of carbon in the fuel stream. Flare CE may be impacted by cross-winds (fuel stripping) and steam-assist4.
Synthetic MW Hyper-Cam data was generated from a CFD large-eddy simulation of a flare in a crosswind at a resolution of 4 cm-1.
Peak species concentrations are inferred through nonlinear regression of modeled data to simulated measurements.
CH4 mass flow Conclusions and future work Imaging Fourier transform spectrometers can estimate
mass fluxes of gas mixtures. Methane flux estimates from the Hyper-Cam LW were
consistent with ground-truth data. More accurate results can be found using the Hyper-Cam LW Methane (more methane lines) or the Hyper-Cam MW (improved temperature sensitivity).
The Hyper-Cam MW can also estimate CO2 and CH4column densities in a flare, which can be used to assess combustion efficiency.
The Hyper-Cam MW will be used measure flare CE in the Boundary Layer Wind Tunnel at Western University, and later in a flare at an industrial site.
1 2
3 4
56 7
**LES Parameters: Stack ID 10.3 cm, 25.6 g/s of CH4, crosswind at 1.9 m/s [5]
CO2pixel
CO2 CH4
NCE N N
≈+
Imaging Fourier-transform spectrometer (IFTS)
The IFTS generates two-dimensional interferograms, which are transformed into a hypercube of spectally-resolved images.
The LW Hyper-Cam is used to measure a 10 slpm CH4 release at 10 K above ambient temperature with a cloudy sky as background.
The data was used to obtain CH4 column densities in each pixel, which is combined with the estimated velocity field to infer the methane mass flow.
OPD /2
Beam splitter
Fixed mirror
Moving mirror
Sensor
Source
CH
4co
lum
n de
nsity
in g
/m2
x in mm
z in
mm
50
100
150
200
25
20
15
10
5
050 100 150 200 250
norm
al v
eloc
ity in
m/s
x in mm
z in
mm
50
100
150
200
0.35
0.3
0.25
0.2
0.15
050 100 150 200 250
0.1
0.05
Pixel interferogram FT
OPD
Inte
nsity
Inte
nsity
Wavenumber
Model Resolution[pixels] FOV Spectral Resolution
[cm-1]Spectral Range
[μm ]NESR
[W/m2 sr cm-1]Hyper-Cam LW 320×256 6.4o×5.1o Up to 0.25 7.7-11.8 (850-1300 cm-1) 20 ×10-5
Hyper-Cam MW 320×256 6.4o×5.1o Up to 0.25 3-5 (2000-3330 cm-1) 4 to 7 ×10-5
Re-absorbedBlackbodyemission
Transmittance (1−αη)Background
( ){ }( ) ( ){ }
L
L 0 0L L
b0 S
I I exp - s ds
(s) I s exp - s ds ds
η η η
η η η
= κ
′+ κ κ
∫∫ ∫
Absorption coefficient
* XCH4=0.2@298K, XCO2=0.2, XH2O=0.4@400K
0L
ILη
I0η
2000 2400 2800 32000
50
800 1000 1200 14000
3.5
κ ηin
cm
-1 *
η in cm-1 η in cm-1
CH4 CO2
H2O×100H2O
Outside (weather data)
ΔT(fitted)Inside (IFTS sensor)
sTem
pera
ture
s
Background H2O (fitted using BG
pixels)
CH4 (fitted)
100 m L (fitted)
Mol
ar fr
actio
n
The background intensity, I0η, was obtained from pixels that did not contain CH4.
The background temperature and ambient humidity were fitted assuming ambient temperature from meteorological data.
Species molar fraction and temperature were parametrized as Gaussian profiles along the LOS2, and the species absorption coefficient is simulated using the HITRAN spectral line database.
Peak molar fraction, peak gas temperature, and plume thickness at four heights were inferred from measured spectra of 4 cm-1 resolution.
1 2 3
20 40 60 80 100
20
40
60
80
100
120
verti
cal p
ixel
s
horizontal pixels
2
4
6
8
10
12
14
16
18
75 85
25
35
45
55
65
75
85
95
105
115
CO2 g/m2
pixe
ls
pixels75 85
25
35
45
55
65
75
85
95
105
115
0.2
0.4
0.6
0.8
1
1.2
CH4 g/m2
pixels
0.3
0.4
0.5
0.6
0.7
0.8
0.9
75 85
25
35
45
55
65
75
85
95
105
115
CEPIXEL
pixels
Single point sample overestimates global CE
The velocity field is estimated by the evolution of brightness patterns in the interferogram data-cube.
Changes in pixel brightness between successive timeframes are caused by the CH4 flow.
Image data-cubes are used to derive the temporal and spatial derivatives by the brightness constancy assumption3.
The underdetermined system of equations is solved by combining the gradient constraint with Laplacian priors on the flow velocities.
u v 0x y t∂Ι ∂Ι ∂Ι
+ + =∂ ∂ ∂
References[1] “Thermal Infrared Cameras.” Telops, https://www.telops.com/. [2] Grauer, S. J. et all., “Gaussian model for emission rate measurement of heated plumes using hyperspectral data,” JQRST, 2018. [3] Horn, B. K., & Schunck, B. G. “Determining optical flow,” Artificial Intelligence, 1981. [4] Johnson, M. R. et all., “A Fuel Stripping Mechanism for Wake-Stabilized Jet Diffusion Flames in Crossflow,” Combust. Sci. Technol., 2001.[5] Flare LES provided by Jeremy Thornock from the Department of Chemical Engineering at The University of Utah.[6] Quantifying & Reducing Impacts of Global Gas Flaring. (2019, October 9). Retrieved from https://www.flarenet.ca/.
[1]
[6]
nozzle
Video of the optical flow results
Hyper-Cam LW Hyper-Cam MW
