J. Phys. E: Sci. Instrum.. Vol. 18, 1985. Printed in Great Britain
`
`Imaging of hydrocarbon vapours
`and gases by infrared
`thermography
`
`D C Strachant, N A Heard?, W J Hossackt, J F Boycet and
`T M Cresswell+
`? Shell Research Ltd, Thornton Research Centre, P O Box 1 ,
`Chester C H 1 3SH, U K
`$ Physics Department, King's College, University of London,
`UK
`
`Received 8 May 1984, in final form 7 January 1985
`
`Abstract. The use of an infrared imaging technique for the
`visualisation of hydrocarbon gases and vapours has been
`demonstrated. The theoretical background of the method is
`discussed, as are some general instrumentation requirements.
`Laboratory investigations of the approach have been carried out
`with commercial infrared thermography equipment to examine
`sensitivity to different hydrocarbons. The studies have also
`clarified the requirements of and the potential for an instrument
`development programme. The potential of image processing
`techniques for the improvement of instrument sensitivity and
`image quality has also been examined.
`
`1. Introduction
`There is a major and continuing requirement within industry for
`instrumentation
`to monitor hydrocarbon gas and vapour
`concentrations. Such instrumentation is required particularly for
`safety monitoring purposes to avoid the build-up of flammable
`atmospheres and also to detect leaks and minimise product loss.
`Traditionally such instrumentation has been of the point
`measurement type and has used a wide variety of detection
`principles. In general. such instruments are capable of accurate
`quantitative measurements for a single hydrocarbon mixed with
`air.
`In this paper we describe a qualitative imaging approach
`to gas/vapour detection which, for a single hydrocarbon
`component, could produce some quantitative data also. The
`technique is based on real-time infrared
`imaging (thermo-
`graphy), which produces images of objects from their own
`radiation. By selecting
`spectral absorption
`infrared heat
`windows characteristic of hydrocarbon vapours and gases it is
`possible to visualise such gases against a background thermal
`scene. The approach and its limitations in terms of hydrocarbon
`detection and instrument development requirements for ambient
`temperature operations are discussed.
`
`2. Physical principles
`In this section we discuss the underlying principles upon which
`the proposed approach depends: namely infrared thermography
`and the absorption of infrared radiation by hydrocarbon gases
`and vapours. We then examine how these methods can be
`combined to produce an instrument for the visualisation of
`hydrocarbon vapours and gases.
`2.1. Infrared thermography
`All bodies at temperatures in excess of absolute zero emit
`thermal radiation as described by the Planck-Kirchoff
`law.
`0022-3735/85/060492 + 07 $02.25 0 1985 The Institute of Physics
`
`Perfect black-body sources have an emissivity e = 1. In general,
`however. ~(1.) < 1. producing
`M(A. T ) = E ( A ) ( C ~ / A ' ) [ ~ X ~ ( C ~ / A T ) - I]-' (W m-j)
`( I )
`where M(A, T ) is the spectral radiant excitance, C1(=2nhc2)
`is the first radiation constant (3.742 x
`J mz s-l) and
`CZ(=hc/k) is the second radiation constant (1.439 x lo-' m K).
`For most solid bodies it may be assumed that E is, to a
`first approximation, independent of wavelength. If this grey-
`body assumption is made, then provided E is known. surface
`temperatures may be calculated if some measure of radiant
`intensity is made. This is the principle of optical pyrometry,
`which underlies temperature measurements obtained thermo-
`graphically. One principal consideration underlying the choice
`of such a system must be the transmission characteristics
`of the atmosphere. The transmission of infrared radiation
`through the atmosphere over a 1 km pathlength is shown in
`figure 1. As may be seen, there are a number of strong
`absorption features in the infrared. notably between about 5 and
`8 pm. Most infrared imaging systems, as a result of this feature.
`employ detectors operating between 2-5 pm (short wave) and
`8-12pm (long wave). The spectral properties of gases and
`liquids in these wavelength bands may also be put to use in some
`industrial thermography applications, such as measurement
`through flame gases, measurement of molten glass temperatures
`and the application discussed in the present paper.
`
`1 . 0 r
`
`-
`
`I "
`
`Long-wave band
`
`L
`
`o L
`
`4
`
`"
`
` -
`
`"
`Short-wave band
`Wavelength lpm)
`Figure 1. Atmospheric transmission in the infrared (1 km
`pathlength at sea level). (Reproduced with permission
`of AGA Infrared Systems Ltd.)
`
`Figure 2 shows the relative response of the single-element
`detector thermography systems used in the present study. Two
`short-wave systems are shown based on a liquid-nitrogen-cooled
`indium antimonide detector, the broader band response as used
`here being obtained with coated optics. The long-wavelength
`system employs a cadmium mercury telluride detector. also
`cooled to liquid nitrogen temperatures. The use of cooling for
`both detectors provides a large increase in signal-to-noise ratio
`for the measurement of low object temperatures.
`2.2. Infrared absorption by hydrocarbon vapours and gases
`The value of infrared absorption instrumentation for the analysis
`of hydrocarbon vapours and gases is well known. Indeed, the
`is one of the most widely used
`infrared spectrophotometer
`instruments in the analytical chemist's armoury. There is a
`wealth of literature available on the infrared characteristics of
`hydrocarbons in gaseous as well as condensed phases. Much of
`this work is concerned with detailed spectral structure and
`comparisons with known catalogue spectra of pure compounds
`for the determination of composition. A review of this area is
`presented by Jones and Sandorfy (1956).
`
`FLIR Systems, Inc.
`Exhibit 1008-00001
`
`

`

`Infrared imaging of hydrocarbon vapours
`
`f l
`
`C M T
`
`
`10
`
`(standard pm peak 1
`
`-
`
`0
`
`a ? 50
`
`Indium antimonide
`
`
`A B r o a d b a n d loptionai I
`
`5 pm peak
`(standard]
`
`Wavelength (pml
`Figure 2. Relative response of thermography detection systems.
`(Reproduced with permission of AGA Infrared Systems Ltd.)
`
`In the current study we have made considerable use of the
`catalogues of infrared absorption spectra presented by Pierson
`et a1 (1956) for gases and by Welti (1970) for hydrocarbon
`vapours. These references are comprehensive, providing a total
`of 372 spectra, including the majority of common hydrocarbon
`products or their major components. For gases, absorption
`bands occur widely across the infrared spectrum between 2 and
`15 pm. Within these bands only one is common to the majority
`of hydrocarbon gases, that occurring generally between limits
`of 3.3 and 3.5pm in the short-wavelength region. This is the
`C-H stretching band of methyl (CH3) or methylene (CH2)
`groups. Jones and Sandorfy (1956) discuss the origins and
`characteristics of these bands in more detail than is presented
`here. The positions of the bands are fairly constant in all types of
`aliphatic hydrocarbons and in alicyclic hydrocarbons where no
`ring strain is involved. Band intensities increase with increasing
`chain length and molecular weight. This may be seen, for
`example, in the spectra of methane, ethane, propane and
`n-butane. In general terms, for vapours there is a trade-off
`in detectability between absorptivity and vapour pressure.
`
`2.3. Hydrocarbon imaging instrumentation
`to provide a
`Thermography employs
`infrared
`techniques
`television-type two-dimensional image of the source object. The
`way in which this is achieved in the AGA instrument, which we
`have employed, may be seen in figure 3. The detector is housed
`in its own Dewar flask, which contains a small quantity of liquid
`nitrogen coolant. Infrared radiation from the source object is
`imaged by a multi-element lens, generally silicon or germanium.
`
`Vertical and horizontal scanning of the test object is carried out
`by refractive scanning prisms before focusing onto the detector.
`The detector signal is then processed electronically to
`produce a real-time infrared television picture or thermogram.
`Figure 3 indicates schematically the operation of a hydrocarbon
`detection system, The camera views the thermal background
`scene around and through any intervening hydrocarbon cloud.
`Providing background and cloud are not in total thermal
`equilibrium with each other, then it is possible to visualise the
`gas cloud against the background. The current paper deals with
`the case where the background is warmer than the gas. This is
`frequently the case, for example, when volatile liquids evaporate
`and when pressurised gases leak from a container.
`Some critical factors in the successful operation of such a
`system are outlined. Firstly, the filter bandwidth and the
`absorption spectra of relevant hydrocarbons must be examined.
`Figure 4 shows the transmission characteristics of the two
`infrared filters that we have employed. The filters have nominal
`bandwidths of 5% and 10% centred approximately at 3.4 pm.
`
`80
`
`60
`
`C 0
`'"_ c
`2 40
`c
`a OI
`c
`a
`U
`L a a
`
`20
`
`0
`
`4
`
`Nomina(
`bandwidth
`
`5 Y o
`1 0 %
`
`4 0
`
`3.0
`
`j >
`Waveiength (pm)
`Figure 4. Infrared filter characteristics.
`
`\i Thermal background
`Y Scene
`
`\
`
`I
`
`
`
`Infrared viewer
`
`Figure 3. Schematic of hydrocarbon imaging system.
`(Reproduced with permission of AGA Infrared Systems Ltd.)
`
`d
`tor
`
`Figure 5 shows typical absorption spectra (Pierson et a1
`1956) for the lower alkanes - methane, ethane, propane and
`n-butane. As may be seen, the 3.46pm absorption feature is
`present in the three heavier gases but is shifted to 3.3pm for
`methane. When allowance is made for the lower concentrations
`employed in the heavier gas spectra, the absorption feature
`becomes stronger with increasing molecular weight. Com-
`parison of figure 4 with figure 5 indicates that the 10% filter may
`be used for methane in addition to the other gases. For the 5%
`filter, which is cutting-off rapidly at 3.3pm, much lower
`sensitivity may be expected.
`stretch absorption at
`In addition to the common C-H
`around 3.46 pm, hydrocarbon vapours and gases exhibit many
`other complex absorption features, as indicated in figure 5 .
`Significantly, a number of compounds show major absorptions
`within the 8-12 pm long-wavelength atmospheric window for
`which infrared cameras are available. Two important industrial
`gases displaying such strong absorptions are ethylene and
`
`493
`
`FLIR Systems, Inc.
`Exhibit 1008-00002
`
`

`

`D C Strachan et a1
`
`100
`
`al U : 80
`4- +
`5 60
`0
`* 40
`L
`4-
`C
`
`2
`
`20
`
`0
`
`IJm
`Figure 6. Infrared absorption spectra of ethylene and ammonia.
`(Reprinted with permission of Anal. Chem. from Pierson et a1
`(1 956))
`
`these studies was in outline as shown in figure 3; however, the
`hydrocarbon gas under investigation was contained within gas
`cells. The cells had short-wave infrared transparent sapphire
`windows and were of different lengths to enable quantitative
`measurements to be made. The absorption filters (see figure 4)
`for the measurements were centred on 3.4 pm. As expected, the
`5% filter resulted in lower sensitivity and therefore work
`reported here used the 10% filter. The gas absorption was
`monitored for a variety of background temperatures using a
`large black-body reference source. Absorption was measured in
`isotherm units at the output monitor. These normalised output
`units enable observed monitor images to be related to surface
`temperatures assuming black- or grey-body conditions.
`In order to determine the dependence of absorption upon gas
`concentration we consider a volume of gas containing N
`absorption centres per unit volume each of absorption cross
`section o. If we consider the absorption of monochromatic
`radiation of wavelength d, then, at a distance x within the
`volume, the intensity, Z(x, A), satisfies
`az(x, nyax = --Nu(;.)Z(x, 2)
`and hence the intensity transmitted by a length, 1, of gas is
`Z(1, i)=Z(O, A) exp[-No(d)I].
`(3)
`is the number density of the gas at atmospheric pressure
`If 1%’
`and K is its relative concentration in a gas/air mixture, then,
`since the absorption due to air is negligible at the wavelengths
`under observation, the net absorption, f(d, K ) is given by
`f(d, K ) = 1 - exp[-KNo(A)l].
`(4)
`The flux of infrared radiation falling upon the detector from
`a black-body source at temperature T via a cell containing a
`hydrocarbon gas of relative concentration K is
`
`(2)
`
`U, K ) = I
`
`”
`{M(A. TI[ 1 -f(A, K ) ]
`
`-,O
`
`P T
`Figure 5. Infrared absorption spectra of the lower alkanes in the
`gaseous phase (see note to figure 9). (Reprinted with permission
`ofdnal. Chem. from Pierson et a1 (1956).)
`
`ammonia, as may be seen from the spectra (Pierson et a1 1956)
`shown in figure 6. The most significant feature of these
`absorptions is their spectral width, which implies that filters are
`not necessary. Operation at ambient temperatures is therefore
`feasible with commercially available instrumentation. Long-
`wave studies have therefore been carried out in addition to the
`short-wave studies previously mentioned.
`
`3. Laboratory studies
`The initial feasibility studies have been carried out with an
`AGA 780 dual-wavelength thermography package. The system
`comprises a double-headed camera, one system operating in the
`8-12 pm long-wave band, the other system in the 2-5 pm short-
`wave band. The short-wave system has broadband coated optics
`to improve sensitivity at the lower wavelengths. The relative
`response curves of both 780 camera systems are as presented in
`figure 2. We have also used an AGA OSCAR digital framestore
`system for freeze frame and image digitisation (1 28 x 128 x 8 bit
`pixels).
`
`3.1. Short-wave incestigations
`3.1.1. Gas studies. The experimental arrangement used for
`
`494
`
`FLIR Systems, Inc.
`Exhibit 1008-00003
`
`

`

`Infrared imaging of hydrocarbon uapours
`
`where M(A, T ) is the spectral radiant excitance, defined by
`equation (l), f(d, K ) is the gas absorption, given by equation (4),
`F(L) is the spectral response of the camera and filter and Tg is
`the gas temperature.
`Since observations are limited to source temperatures in the
`range 303-373 K and wavelengths from 2-5 pm, C2/AT= 10,
`and consequently Wien’s approximation may be made to the
`spectral radiant excitance
`
`The integral of the excitance times the filter response was
`approximated by the total flux transmitted by the filter times
`the excitance at io=3.45,um. For the relevant range of
`temperatures, 303 K < T < 373 K, the error of the approxi-
`mation was less than 1% (Hossack 1984). If the absorption
`cross section is also approximated by its mean over the
`wavelength aperture of the filter then the flux of radiation at
`the detector becomes
`
`+ exp(-KNul)[M(Ao, T)-M(iv0. T,)]) (7)
`where F is the total flux transmitted by the filter.
`An initial series of calibration observations was made by
`varying the source temperature in the absence of hydrocarbon
`vapour. For this case, where T, is the ambient temperature,
`independent of T. it may be shown that equation (7) reduces
`to the form of equation (8). Figure 7 shows the experimental
`results together with a least squares fit using equation (8) with
`a=4.235 x 10’ and b=-0.495:
`I(T, O)=a exp(-Cz/AoT) + b.
`(8)
`(Note that constants a and b have been fitted for I expressed
`in the AGA instrument’s own units of intensity change -
`isothermal units.) Besides testing the initial narrow-band approxi-
`mation of equation (8), the above relation may be employed
`to convert from isothermal units to equivalent black-body
`temperature changes although due to the larger scatter at lower
`temperatures, figure 7, accuracy is limited. This scatter was
`principally due to detector noise and drift.
`Results for the four lowest alkane hydrocarbons are
`presented in table 1. Two gas cell lengths. 50 and 200 mm, were
`used, the gases being undiluted and at atmospheric pressure.
`From the table, we may see that each gas can be detected by the
`system whilst the signal-to-noise ratio was observed to increase
`temperature. As previously
`with
`increasing background
`described, the absorption observed increases with carbon chain
`length, i.e. butane is the most readily detected and methane the
`least.
`
`300
`
`320
`
`360
`
`0
`
`3 40
`Temperoiure I K )
`Figure 7. Measured radiation flux change against black-body
`source temperature.
`
`flux absorption with background
`The variation of
`temperature for the series of ambient temperature hydrocarbon
`gases, with the 200 mm cell, is shown in figure 8. The data have
`been fitted by curves of the form
`
`I(T, 1)=uf(&, 1) exp --
`
`I i ?TI
`
`where &(A) is the emissivity of the source at wavelength A.
`The constants of proportionality, uf(Ao, l), relate to the
`mean absorption cross sections, .(A,),
`of the vapours, while the
`difference between the zero levels of equations (8) and (9) is due
`to the infrared emission from the vapour. Using T, = 293 K, the
`curves of figure 8 were obtained by fitting the net absorption,
`f(&. l), of each vapour to the data.
`The dependence of absorption upon vapour concentration
`was observed by varying the concentration of methane in the
`200 mm cell while maintaining a fixed source temperature of
`373 K. Figure 9 shows that the results are in agreement with the
`
`Table 1. Gas absorption in isotherm units A I and equivalent temperature changes AT(K).
`
`Background temp. (K)
`
`373
`
`353
`
`323
`
`313
`
`303
`
`Cell length (mm)
`
`200
`
`50
`
`200
`
`50
`
`200
`
`50
`
`200
`
`50
`
`200
`
`50
`
`CHp methane
`
`C2H6 ethane
`
`C3Hs propane
`
`C4HI0 n-butane
`
`AI
`AT
`A I
`AT
`A I
`AT
`AI
`AT
`
`3.50
`34
`4.55
`51
`4.90
`59
`5.05
`> 6 0
`
`2.05
`19
`3.75
`38
`4.00
`41
`4.10
`43
`
`1.76
`21
`2.31
`31
`2.45
`33
`2.55
`36
`
`~~
`
`1.32
`15
`1.89
`22
`2.00
`24
`2.10
`26
`
`0.48
`> 15
`0.60
`> 15
`0.60
`> 15
`0.68
`> 15
`
`~~~~~
`
`0.42
`7
`0.53
`> 15
`0.54
`> 15
`0.54
`> 15
`
`0.20
`0.12
`0.18
`- - -
`0.3
`0.25
`0.14
`- - -
`0.28
`0.28
`0.16
`- - -
`0.34
`0.24
`0.16
`- - -
`
`0.08
`-
`0.10
`-
`0.14
`-
`0.16
`-
`
`495
`
`FLIR Systems, Inc.
`Exhibit 1008-00004
`
`

`

`D C Strachan et a1
`
`Figure 8. Variation of absorbed flux with temperature for the
`low alkanes. A, methane; B, ethane; C, propane; D, n-butane.
`
`Figure 9. Dependence of absorption on methane concentration.
`
`least squares fitted variation predicted from equation (7), namely
`
`Figure 10. Single frame images of butane leaking from
`small-bore pipe in front of black-body radiator. Short-wave
`camera system. Black-body temperature: (a), 303 K; (b), 353 K.
`
`353 K. As may be seen, image quality improves with higher
`background temperature. At 303 K, however, the image is
`marginally usable. As with the gas cell tests, visibility improved
`with increasing molecular weight.
`3.1.2. Vapour studies. The same procedure was carried out for
`selected volatile hydrocarbon liquid products. In these tests
`several drops of the liquid were introduced into a gas cell and
`allowed to vaporise. The results are shown in table 2 for two
`selected background temperatures, and indicate that significant
`absorptions do indeed occur.
`
`3.2. Long-wave investigations
`Comparing the long-wave spectral response of the thermo-
`graphy system (figure 2) with the infrared absorption spectra
`of, for example, ethylene (figure 6), it is apparent that some
`gases have strong absorption in that band, so that no additional
`
`( M(no’ T 8 ) ) { 1 - exp[ - KNu(,~o)\l 1
`
`AI(T, K)=I(T, 0) 1 -
`W O ,
`
` T )
`(10)
`=a[ 1 - exp(-KNul)].
`The mean absorption cross section of methane arising from
`figure 9 was 4.8 x lo-’’ m2.
`In addition to the quantitative studies outlined above, a
`number of qualitative tests were undertaken. In a typical
`example, gas leaking from a fine hole in a small-bore supply
`pipe has been imaged. A typical result for butane is shown in
`figure 10, with thermal background temperatures of 303 and
`
`496
`
`FLIR Systems, Inc.
`Exhibit 1008-00005
`
`

`

`Infrared imaging of hydrocarbon vapours
`
`filtering is necessary. In experiments with ethylene, gas was
`again allowed to leak from an open tube in front of the back-
`ground radiator, in this case at ambient temperature. The gas
`could be readily detected at flow rates below 1 litre per minute.
`Tests were performed using the leaking pipe assembly also used
`for the short-wavelength butane results of figure 10. Figure 1 I
`shows long-wavelength ethylene results with a background
`scene temperature of 303 K. As may be seen, cloud resolution
`is significantly
`improved
`in comparison with
`the lower-
`temperature butane image.
`
`Figure 11. Image of ethylene leaking from a pipe.
`Background temperature 303 K; long-wave system.
`
`4. Instrument development and image processing
`The experimental studies have demonstrated the feasibility of
`visualising normally invisible hydrocarbon vapour and gas
`clouds using an infrared technique. In particular, studies using
`the common C-H stretch absorption of 3.46 pm have confirmed
`the wide applicability of a short-wave approach. The prin-
`cipal drawbacks of this short-wave approach, using typical
`thermography equipment, are insufficient sensitivity allied to
`high capital cost. For practical hydrocarbon imaging, a scene
`temperature in excess of 303 K is required with the laboratory
`system employed by the authors. Methods of redressing the
`limitations apparent in general purpose thermal imagers are the
`subject of continuing work. Possible improvements include
`cheaper infrared optics optimised for the relatively narrow
`wavelength response required for the instrument with newer and
`better detection and signal processing.
`
`4.1. Image processing improvemenrs in performance
`The digitised output from the imaging system consists of a 128
`square, 256 grey-level image, scanned as two interlaced frames.
`Typical images of butane plumes against a background of
`constant temperature are shown in figure IO. Images taken
`against a temperature above 303 K are of usable S/N, but at
`303 K some image processing is required. The effect of temporal
`averaging of
`ten successive frames followed by contrast
`enhancement by grey-level stretching is shown in figure 12(a).
`The additional effect of 3 x 3 spatial averaging is shown in
`figure 12(b) whilst figure 12(c) shows the results of applying a
`binary threshold. This final image is of practical quality. This
`
`Figure 12. Simple image processing of butane image for a
`303 K background. (a), Ten-frame average followed by contrast
`stretch. (b), 3 x 3 spatial averaging of (a). (c), Binary threshold
`of (b).
`
`image processing is all of a very simple nature, and can be
`implemented in real time by a micro-based system. Thus initial
`results suggest that with the addition of simple image process-
`ing, imaging of hydrocarbon vapours against background
`temperatures of 303 K and below is practicable.
`
`5. Conclusions
`invisible
`imaging normally
`to
`approach
`( I ) A novel
`hydrocarbon clouds has been demonstrated successfully in the
`laboratory. Instrument development is required to produce a
`general purpose short-wave instrument for use under lower
`ambient temperature conditions.
`(2) For certain hydrocarbons, most notably the lower olefins,
`long-wave instrumentation currently available is suitable for
`imaging at ambient temperatures. Using such equipment and a
`suitable hydrocarbon, such as ethylene, field use of imaging
`equipment may usefully be examined. Such work will enable a
`more critical assessment of the potential of this new technique to
`be made.
`
`491
`
`FLIR Systems, Inc.
`Exhibit 1008-00006
`
`

`

`D C Strachan et a1
`
`References
`Hossack W J 1984 Applications of image processing and
`pattern recognition techniques
`PhD Thesis Queen Elizabeth College, University of London
`Jones R N and Sandorfy C 1956 The application of infra-red
`and Raman spectra to the elucidation of molecular structure
`Chemical Applications of Spectroscopy, Techniques of Organic
`Chemistry vol. 9, ed. W West (London: Interscience)
`Pierson R H, Fletcher A N and Gantz E S C 1956 Catalogue of
`infrared spectra for qualitative analysis of gases
`Anal. Chem. 28 1218-39
`Welti D 1970 Infra-red Vapour Spectra (London:
`HaydenjSadler)
`
`498
`
`FLIR Systems, Inc.
`Exhibit 1008-00007
`
`