Professional Documents
Culture Documents
by
D.P.Correia, A. Caldeira-Pires, P. FerrHo and M.V.Heitor
/----x
HorizontalPlatform
particles, soot, (usually between lOOA and 2000A in
diameter). The solid carbon particles, which are formed in
a flame as a result of incomplete combustion of
hydrocarbons, increase the emissivity and enhance flame
radiation, which is to be used for temperature diagnostics
on the present paper.
However, as carbon particles are frequently smaller than
the wavelength of light, their emissivity increases for
shorter wavelengths and, as a consequence, the spectral
intensity distribution in the continuum is different from
YWdeo Slgnalr
that of a black body at the flame temperature.
Figure 1. Experimental Rig.
The tomographic reconstruction software, developed for RADIATION ABSORPTION
the reconstruction of the local emitted radiation field of
the flame, included correction terms for the radiation Radiation emitted inside a flame is attenuated in an extent
attenuation inside the flame, in order to accurately that depends on the path in the attenuating environment
estimate the local temperature of soot particles. and on the physical properties of the environment itself.
Therefore, radiant energy, at a given wavelength h,
propagating through the reacting gases, is absorbed and
3) Experimental study of the flames studied, which scattered in the gas. For monochromatic radiation, the
included the characterisation of the scalar and vector absorption by each chemical specie in the gas, a,(T),
fields of non-premixed flames, making use of which is quantified as the ratio between the
conventional diagnostic techniques. monochromatic radiance at wavelength h, N, (T) , and
the spectral radiance of a blackbody at the same
4) Calibration of the tomographic pyrometer, by temperature, N i (T) , depends on the characteristics of
comparing the data obtained from the new sensor with the absorbing particles, according to the Beer-Lambert
accurate results obtained with conventional methods, in law:
order to evaluate the adequacy of CT to flame
diagnostics.
This paper discusses the approach adopted on each of the where K,is the extinction coefficient and L the optical
stages and, in particular, presents a theoretical description path on the absorbing gas.
of the relevant aspects of radiative heat transfer, radiation Several attempts have been made in order to model the
absorption and pyrometry in flames, and the description
value of K,, based on different principles and
of the tomography procedures used. Results of
tomographic reconstruction of the axissymmetric flame simplifications. The most common are:
studied are presented and compared with the temperature i) Neglecting the absorption effect, which corresponds to
measurements performed. assume that Ka = O (as performed by Fischer [2]).
Tomography is performed assuming that all emitted
radiation reaches the radiation collector equipment
SPECTRAL EMISSION OF FLAMES without attenuation.
According to Gaydon [lo] there are three basic types of ii) Assuming a given relation of K, with A ,as initially
flame spectra: line spectra, emitted or absorbed by free
suggested by Hottel and Broughton [Ill and later
atoms, band spectra, corresponding to electronic
described by Siddal and McGrath [12], considering
transitions in molecules and changes in rotational or
vibrational energy, and continuous spectra, which is that the absorption coefficient, a,(T), is constant
associated with emission or absorption by solid particles inside the flame.
or liquid droplets, but may also be due to processes such
as dissociation, ionisation or recombination in molecules iii) Determining the local K, through means of an
or atoms. Flame radiation may consist of a few spectral emission-absorption method, as performed by Best et
lines, many lines arranged in band systems, continua, or al. [5] and Barrag and Lawton [6], and then using the
combinations of lines, band and continua depending on determined values of the extinction coefficient for the
the fuels and oxidisers present, and their interaction, pyrometry calculations. This method requires the use
which is determined by complex aerothermodinamic of an external radiation source, or a mirror of known
phenomena. Luminous flames show strong continuous reflectivity (as proposed by Porter [131).
402
It is known that Ka depends on both the wavelength and
the soot concentration across the flame, which would
imply spatial differences in the local extinction
coefficient. A reasonable compromise between the
determination of the local absorption, which imposes the
use of an external radiation source, precluding its use on L
These authors report values for the constant p ranging On the present work, the spectral radiances from the
from 0.95 (infrared) to 1.39 on the visible wavelengths, flames were experimentally obtained for two
from direct measurement on steady-state luminous flames. wavelengths, kept close together in order to validate the
Siddall and McGrath [12] quantify p as: gray body approach. Equation 6 can be used to quantify
the temperature, resulting in Equation 7, where c2 is the
p = 0.906 + 0.283.In il (3)
second radiation constant (= 1,4388cmK).
for h = [l - l O ] p . The same authors have studied the
emissivity and absorption in luminous flames for different
fuels, and concluded that the extinction coefficient
depended not only on the wavelength but also on the
carbonhydrogen ratio of the fuel and on soot
concentration. They suggested a large set of mean values
for the constant p (for various fuels), which, for propane
flames, as used in this paper, is 1.00.
TOMOGRAPHY ALGORITHMS
COLOR PYROMETRY The tomographic process consists on the reconstruction of
the two or three dimensional internal structure of a given
The monochromatic spectral radiance of a blackbody object from one or several projections.
N i ( T ) is modelled by the Plank radiation law (as stated,
The use of analytical integral equations, for the
for example, in Tourin [16]). For a radiating body, the reconstruction of the tridimensional structure of
monochromatic spectral radiance is obtained by axissymmetric objects, is well known and has been
multiplying the spectral emissivity { A ( T ) by the applied to a large set of different subjects, as in medical
blackbody spectral radiance, as stated in Eq. 4: diagnostic, geological survey, astronomy, fluid mechanics
and combustion.
The mathematical reconstruction equation was first
developed by Radon in [l], and latter optimised and
applied to radio astronomy by Bracewell [17]. From a
Two-color pyrometry is based on the ratio of the spectral
finite set of projections the Radon Transform, R[F],
radiance equations obtained for the same conditions but
for different wavelengths, assuming a given relation provides the mathematical equation (Eq.8) that allows the
between the two monochromatic emissivities. The general reconstruction of the original property distribution, F,
approach is to assume that a flame behaves like a gray based on a set of one or more projections, represented by
body, which is to say that the emissivity is independent of the function P, as illustrated on Figure 2.
the wavelength, as represented in Eq. 5 : 403
~
INCLUDING ABSORPTION ON AN
AXISSYMMETRIC TOMOGRAPHIC
RECONSTRUCTION ALGORITHM
The tomographic reconstruction of any given object is
based on the assumption that any projection integrates
radiation (or some other property, depending on the
application) that is either emitted or transmitted through
the object. The reconstruction process basically
deconvolves the retrieved data in order to obtain the
initial property field.
For the specific flame in study, illustrated in Figure 3, the
projection acquired by the detector equipment integrates
the radiation emitted inside the body, along a given
Figure 2. Sample object and its projection for a given optical path (in this case assumed to be a parallel beam),
angle @ . but the radiation emitted by the soot particles is largely
For the specific case of an axissymmetric object, all the absorbed inside the flame, by the soot layers standing
radial projections of the object become equal, between the emitter particles and the detector.
F(x, y) = F(r) and P(p, $) = P(p) , and therefore the
Radon integral equation can be developed into the Abel
integral equation that states as follows:
Ill1
ALGORITHM EVALUATION
The numerical discretisation of the tomographic
reconstruction was implemented on dedicated computer
programs in order to evaluate the various algorithms.
The procedure, similar to the one performed by Fischer
[3], was to use a set of different mathematical functions of
known projection and to check the performance of the
algorithms, both in terms of computing time and quality
of the reconstruction.
Several different functions were used, which were meant
to test the algorithm for different aspects, such as, the
response to discontinuities on the function and on its first C> d)
derivative.
Figure 5 . Algorithm response for the linear function, for
The testing procedure was to generate a projection different numbers of integration points: a) 10, b) 25, c)
function from a given mathematical function, and then 100 and d) 1000 points.
reconstruct the original function using the tomographic
The improvement on the spatial resolution results in
reconstruction algorithms. Sample results are presented in
higher computational costs, although, as displayed on the
Figures 4 a) and b), for two of the tested functions using chart in Figure 6 , for small numbers of discretisation
the optimised algorithm. The domain of the test functions points the processing time is reduced, therefore limiting
is [-1,1]. Results are presented on a [0,1] scale due to the problem to larger arrays.
object axissymmetry. The results demonstrate the ability
of the optimised algorithm to reproduce the original
functions, i),based on theirs projections, ii).
(i) (ii) (iii)
RESULTS
The laboratory experiments resulted in a large set of time-
averaged images, for the two wavelengths used. These
images represent the monochromatic flame radiation,
406
integrated along an optical path (assumed parallel), for by the CCIR cameras. The right hand sides characterise
burner distances of 25 to 55 burner diameters. The time- their tomographic reconstructions, which represent a
averaged images have shown to be axissymmetric, as radial distribution of the radiation emitted at the given
Figure 10 illustrates. wavelength.
a) b) a) b)
Figure 10. a) Instantaneous, and b) Time-averaged images Figure 12. Tomography reconstructed images: a) 800nm
(16 seconds = 400 images) for 800nm. and b) 900nm.
In Figure lob), the grayscale level corresponding to each The 3D temperature field evaluation process consists on
picture element (pixel) is a function of the radiation the combined use of the tomography reconstruction
emitted along the optical path, taking into consideration techniques and the color pyrometry equation (Eq. 7). A
the radiation absorbed along the same optical path. Two conversion term is multiplied to the ratio of the spectral
average images for the wavelengths considered behind are radiances, N,, (T)/NA2(T) , so that the grayscale images
represented on Figure 11. The image units are in can be translated into IS units. This conversion term is
grayscale values and can be correlated to radiance units
obtained through calibration with a blackbody furnace.
by calibration over a blackbody furnace. Acquisition
Applying the pyrometry equation to the tomography
times, camera settings, focal distances, shutter speeds and
images results in the local temperature field, as illustrated
combustion parameters were maintained constant on Figure 13a). These results are compared against the
throughout the experiments, allowing for the pyrometry data obtained from the experiments obtained by Caldeira-
calculations.
Pires et al. [21], represented on Figure 13b).
a) b)
Figure 11. Monochromatic images: a) 800nm and
900nm.
These images were used to perform the calculations a> b)
necessary for the tomographic reconstruction of the
monochromatic images. This was done according to the Figure 13. a) Temperature field obtained with the
algorithm including absorption correction. As stated by temperature tomographic sensor, and b) Measurements
Eq. 2, the extinction coefficient was considered to be by Caldeira-Pires et al..
proportional to the inverse of the wavelength (with The qualitative agreement suggests that the experimental
p = 1) and therefore the extinction coefficient should be procedure adopted in obtaining these preliminary results
higher for the 900nm image. The results for the is adequate, although more detailed experiments are
tomographic reconstruction of the monochromatic images required in order to validate the technique and to optimise
are presented in Figure 12. The left hand side of Figures the modelling of the extinction coefficient.
12a) and 12b) represent the original image, as retrieved
407r
RCSCalLII L U l l I l L L C I C I ) rxpu *-rlr”) yy. /rr ,, .
sensor for combustion systems, making use of radiation
pyrometry, radiation absorption compensation and [8] Philipp, H., Plimon, A., Fernitz, G., Hirsch, A., Fraidl,
tomographic reconstruction algorithms, was discussed G., and Winklhofer, E., (1995). “A Tomographic
and assessed. The results obtained for the 3D temperature Camera System for Combustion in SI Engines”, SAE
field of a non-premixed propane flame and compared International Congress and Exposition, February 27 -
against the data obtained from measurements performed March 2 1995.
with conventional techniques (time-resolved
thermometry), suggest that the new sensor is adequate for [93 Waterfall, R., He, R., White, N. and Beck, C. (1996)
the characterisation of the temperature field and the “Combustion Imaging From Electrical Impedance
radiative heat transfer process on combustion Measurements” Measurement, Science and
environments, although detailed experiments are required Technology, n 7 , pp.1-6.
for optimising the flame radiation absorption modelling.
[lo] Gaydon, A. G. (1957) The Spectroscopy of FZames.
Chapman & Hall.
ACKNOLEDGEMENTS
[ 111 Hottel, H.C. and Broughton, F.P. (1932)
Financial support provided by the European Union under “Determination of True Temperature and Total
the project CLEAN-GLASS, Contract noBRPR-CT95- Radiation from Luminous Gas Flames”. Industrial
0033, is gratefully acknowledged. Engineering and Chemistry, Analytical Edition. Vo1.4,
pp. 166-175.
409