Experimental and Numerical Study of Thermocouple-Induced Perturbation of the Flame
P.A. Skovorodko∗, 1, O.P. Korobeinichev2, A.G. Tereshchenko2, D.A. Knyazkov2,
Institute of Thermophysics, Siberian Branch of Russian Academy of Sciences, Novosibirsk 630090, Russia
Institute of Chemical Kinetics and Combustion Siberian Branch of Russian Academy of Sciences, Novosibirsk
In the measurements of temperature profiles in the flames using a thermocouple it was usually assumed that due to
relatively small sizes of the thermocouple the flame structure perturbations induced by it are negligible. Our study
indicates, however, that this statement may be wrong. The temperature in premixed atmospheric
methane/oxygen/argon flame obtained by several thermocouples was found to be systematically higher than
theoretical ones at low distances from the burner. To elucidate the observed discrepancy between experimental and
theoretical data the external flow of studied flame over the thermocouple was simulated in the frames of full set of
unsteady Navier-Stokes equations. The nature of the effect was found to be connected with deceleration of the flow
in the vicinity of the thermocouple that leads to additional heat release here due to the chemical reactions.
Introduction of experimental errors associated with heat gain or heat
Measurements of temperature profiles in the flames losses from junction to shoulders (upper part of Π) of
by thermocouples are widely used in experimental the thermocouple occur, because the shoulders and the
studies of the flame structure. A flame temperature junction are kept at different temperatures. These issues
profile allows one to evaluate the heat release in were studied in detail by A.A. Zenin (see Refs. 17, 19,
various flame zones and to distinguish the key reactions 22 – 25 in ).
in combustion process . The measured temperature It is known that to determine an actual value of
profile is usually used for combustion modeling. temperature, the errors related to heat exchange of the
Performing the measurements of the flame thermal thermocouple with environment as well as to
structure with a fair accuracy is one of the basic tasks in thermocouple emissivity should be taken into account.
combustion study. To measure the flame temperature As a rule, the researchers consider that the errors in
the thermocouples of two different cross-sectional thermocouple readings occur only due to the
shapes (circular and rectangular) are usually used. uncertainties associated with intrinsic properties of the
These thermocouples are called as circular and ribbon thermocouple as a temperature sensor, i.e. depend on its
thermocouples, respectively. Normally, the ribbon shape, material, emissivity and moving speed in the
thermocouples are used for studying the flame structure flame studied. Up to this point no particular attention
of condensed systems. The ribbon thermocouples are has been given to study of thermocouple-induced gas
generally manufactured by squashing the circular ones. dynamic perturbation of the gas flow, where the
At that, as a rule, the width of the ribbon thermocouple thermocouple is placed. A fixed thermocouple
exceeds 15-20 times its thickness. According to the represents always an obstacle for the gas flow which
researchers, who use such thermocouples, this allows temperature is to be measured by the thermocouple.
one to increase resolution capability of the When significant temperature gradients and high
thermocouple. In the gas flames usually stabilized on speeds of the gas flow take place, the thermocouple can
the flat burners, the time required for temperature affect considerably the gas flow in spite of its small
measurement using a thermocouple is unlimited. size. To our knowledge, there are no any works devoted
For studying the thermal structure of flames of to study of this issue. It was usually assumed that due
condensed systems the thermocouple is inserted into to relatively small sizes of the device the perturbations
the sample. As a rule, the thermocouples have Π- induced by the thermocouple in the flame structure are
shaped form and the thermocouple junction is in the negligible.
middle of upper part (working part of the A further analysis of data reported in our previous
thermocouple). The measurements are considered to be paper  showed that the flame temperature calculated
correct if the working part of the thermocouple is by CHEMKIN PREMIX code  in the vicinity of
parallel to flame isotherms. During the sample burning, burner is slightly underestimated in comparison with
the thermocouple moves relative to the burning surface the measured values that may be caused by
at a rate equal to burning velocity. Radiation heat losses thermocouple-induced gas dynamic perturbation of the
of the thermocouple were considered elsewhere . flow. The qualitative pattern is as described below.
Due to a restricted size of samples, when using Π- Since the gas flow decelerates in front of the
shaped thermocouples, some difficulties in estimation thermocouple, the flow perturbations in the vicinity of
Corresponding author: email@example.com
Proceedings of the European Combustion Meeting 2009
the thermocouple occur. The temperature gradients near shapes placed into the flame in the area with high
the burner surface or near burning surface of solid temperature gradients is studied experimentally and
propellants are quite high, therefore, in this region the numerically. Particular attention is given to gas flows
thermocouple-induced distortion can sufficiently with high mass flow rate (combustion of condensed
change the original temperature field of the flame. systems at high pressure). The results obtained will
Fristrom and Westenberg  noted that the help to analyze the correctness of using the ribbon
thermocouple perturbs locally the flame velocity profile thermocouples for studying the structure of flames of
and, actually, measures the temperature downstream gaseous and condensed systems.
from thermocouple position. A trace associated with
thermocouple perturbation was estimated to be about Experimental Technique
4 – 5 thermocouple junction diameters. Temperature profile in one-dimensional premixed
Microthermocouple technique is used to study methane/oxygen/argon (CH4/O2/Ar – 6/15/79 vol. %)
thermal structure of flames of gaseous and condensed flame stabilized on a Botha-Spalding flat burner at
systems. Application of this technique for studying the atmospheric pressure was selected as a subject of
flame structure of condensed systems is the only way to research. This flame is well reproducible object as
obtain any information about flame structure in a wide compared with, for example, condensed system flame.
range of pressures (from subatmospheric up to 500 During the experiment, the temperature of the burner
atm). As distinguished from gaseous systems where the surface was kept at 368 K using a thermostat. The
flame is burner-stabilized and gas mass flow rate near burner top represented a porous brass disk 16 mm in
the burner surface is about 0.01 – 0.03 g/cm2·s, in the diameter and 5 mm in height. It was made of sintered
case of burning of condensed systems the mass flow balls ~0.1 mm in diameter. The relative porosity of the
rates of gaseous products are greater by an order of burner disk was 46%. The flow rate of unburnt mixture
magnitude. They are about 0.1 – 0.5 g/cm2s at was 25 cm3/s at standard conditions. Flow rates of gas
atmospheric pressure and rise to higher values at higher components of the unburnt mixture were set by mass
pressures: about 4.9 g/cm2·s for RDX burning at 300 flow controllers (MKS Instruments Inc.) with the
atm . The ribbon thermocouples have the thickness accuracy of ±1%. The burner was mounted on a moving
of about 2 – 7 µm and the width of about 40 – 140 µm; mechanism, which allowed it to move in vertical
the thickness of the gas phase of combustion wave direction relative to a fixed thermocouple with
(measured using the thermocouple technique) at positioning accuracy of 0.01 mm. The shoulders of all
pressure 300 atm comprises 130 µm, and the thermocouples were parallel to burner surface. In case
temperature gradients is about 4×105 K/cm. It is clear of ribbon thermocouples, the broad part of the
that study of such an extreme processes falls into field thermocouple was parallel to burner surface.
of serious criticism by experimentalists and The temperature profiles were measured using
theoreticians who develop models of combustion of Pt/Pt+10%Rh thermocouple coated with a thin layer of
condensed systems . One of the criticized parameters SiO2 (~ 2 – 3 µm) to prevent catalytic processes. The
of the technique is spatial resolution of the technique in thermocouple had rectangular cross-section; its
the area close to burning surface in both cases of dimensions were 20x125 µm (including coating layer).
condensed and gaseous phase. In practice, one can not The length of the shoulders of the thermocouple was
provide an ideality of one-dimensional character of about 10 mm. The thermocouple junction was placed
combustion of condensed systems, therefore, Beckstead into the area close to flame center. Stretching
 assumed that in condensed phase an actual construction for lead wires of the thermocouple 
resolution capability of the thermocouples used is more provided a parallel alignment of the thermocouple
than 2 – 3 times lower in comparison with the relative to burner surface during the experiments. The
thermocouple size (in this case, thickness). It should be stretching construction allowed one also to prevent
pointed out that another thermocouple characteristic deformation of the ribbon thermocouples in the flame.
(thermocouple width) is not discussed in this context, During the measurements of temperature profiles, the
although the width of the thermocouple is a main factor distance between the thermocouple junction and the
leading to perturbation of the gas flow. The majority of burner surface was controlled using a cathetometer with
the works devoted to the modeling of flames of an accuracy of 0.01 mm. The error of the thermocouple
condensed systems at high pressures are based measurements was ±30 К.
generally on the data obtained using
microthermocouple technique, including data on Numerical Simulation of the Flow
temperature gradients at burning surface of a The simulation of the plane external flow field over
propellant. Thus, the reliability of the data obtained is the thermocouple inserted in the flow at some distance
of great importance. Up to now, in literature there are from the burner was performed in the frames of the full
no any estimations of the thermocouple perturbing set of unsteady Navier-Stokes equations. The finite-
effect on the gas flow structure in the flame, where the difference representation of the governing equations
thermocouple is placed. was made on a staggered grid that allows one to develop
In this paper, perturbation of a temperature profile an effective algorithm for simulation of viscous flows
caused by flow-around the thermocouples of different . For approximate accounting for the heat release
due to the chemical reactions the source term is At the burner surface (left boundary) the conditions
introduced in the energy equation providing the given for all the values were specified on the basis of the
temperature distribution in the plane undisturbed solution for undisturbed isobaric flame obtained by
isobaric flame i. e. in the absence of the body in the CHEMKIN PREMIX code  for considered mixture
flame. The corresponding algorithm is described in (T=368 K, w = 15.68 cm/s, u = 0 ).
detail in Ref.  where it was used to simulate the At the right boundary both components of velocity
perturbation of the flow caused by axisymmetric as well as the temperature were obtained by
sampling probe inserted in the flame. interpolation from internal points of the domain while
The simulations were performed to describe the flow the pressure on this surface was prescribed to be
of burner-stabilized premixed methane-oxygen-argon constant (isobaric flame).
flame with initial mixture composition 6%СН4 + At the plane of symmetry (bottom boundary passing
15%О2 + 79%Аr used in our experiments and the flow through the middle of the thermocouple) the transverse
of RDX burning at pressure 20 atm experimentally velocity as well as the transverse derivatives of the
studied in . longitudinal velocity and the temperature are prescribed
to be zero.
Methane flame To avoid the problems with boundary conditions at
The considered combustion-gas mixture is assumed the top boundary the latter was also treated as the plane
to be a monocomponent perfect gas with molecular of symmetry that is equivalent to the assumption that we
mass 37.32 kg/kmole and the value of the specific heats have an infinite row of thermocouples placed on the
ratio κ = 1.5747 calculated from the real parameters of same distance from the burner surface. The step of this
the flame at the burner surface. row i. e. the distance between the middle points of the
The temperature dependence of dynamic viscosity µ adjacent thermocouples was chosen large enough to
of the gas was assumed to be described by Lennard- minimize the distortion of the flow field caused by this
Jones (6 – 12) potential with parameters σ = 3.418 Å, ε assumption.
/ k = 124 K typical for argon . Since the main At the thermocouple surface the normal component
component of considered mixture is argon, the value of of velocity was prescribed to be zero while the
Prandtl number (Pr = µ Cp / λ) was assumed to be the tangential component of velocity was defined taking
same as for monatomic gas, i. e. Pr = 2/3. into account the velocity slip  though the slip effect at
For numerical solution the variables ρ (density), p considered conditions is very small. The temperature of
(pressure), T (temperature), u (transverse component of the surface of the thermocouple was defined from the
total velocity) and w (longitudinal component of total condition that the heat flux to the surface from the gas is
velocity) have been normalized by the corresponding equal to the heat flux from the surface due to radiation
parameters at the burner surface including the sound with assumed value of emissivity factor ε (for ε = 0 the
speed c 0 . The spatial variables x and y appear to be surface is in adiabatic conditions). It should be noted
conveniently expressed directly in mm. that the described approach allows one obtaining the
The mentioned above source term Q source for thermocouple temperature that may be directly
compared with the results of measurements delivering
energy equation was obtained from the relation written from any semi-empirical corrections of data for
in non-dimensional form radiation.
= Qconv + Qvis + Qsource = 0 , (1)
∂t RDX flame
where the convective terms entering the energy equation The RDX flame was simulated in the same approach
are schematically denoted as Qconv while the viscous as the methane flame.
The considered combustion-gas mixture is again
terms are denoted as Qvis . The obtained dependence of
assumed to be a monocomponent perfect gas with
Q source ( x) reported in  is characterized by sharp molecular mass 24.642 kg/kmole and the value of the
maximum at x = 0.5 mm reflecting the peculiarities of specific heats ratio κ = 1.382 calculated from the
undisturbed methane flame at considered conditions. In parameters behind the flame front thus neglecting the
contrast to approach with spatially fixed sources changing of mixture composition inside the front. The
Q source ( x) used in  in this study we apply another viscosity of mixture was estimated based on assumed
approach Qsource (T ) that seemed to provide more mixture composition by relations of , the value of
Prandtl number was found to be 0.6145. The source
adequate description of nonlinear interaction between term Qsource (T ) for energy equation was found based
chemical and gas dynamics processes. The undisturbed
flame is well reproduced by both these approaches. on the approximation of the measured temperature
The domain of simulation was of rectangular shape profile for pressure 20 atm reported in . The sizes of
with 2.4 mm size in longitudinal direction x and 1.5 mm domain of simulation were 0.1 mm in longitudinal
in transverse direction y with mesh sizes of the grid direction x and 0.4 mm in transverse direction y with
dx = dy = 2.5 µm. The boundary conditions for the mesh sizes of the grid dx= 0.2 µm, dy= 1 µm.
considered flow were set as follows.
Results and Discussion
Т, K 0
1400 1 -100 4
1200 4 2 -150
400 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
200 Distance from burner, mm
0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 Fig. 2. The difference between perturbed and
Distance from burner, mm unperturbed values of temperature.
0) that may be treated as the exact values of radiation
Fig. 1. Experimental and numerical temperature corrections for performed numerical experiment.
profiles in methane flame. First of all let’s note that for considered flame
maximum values of overestimation of temperature by
In this paper a methane flame was chosen as one of
thermocouple with ε = 0 achieves the value about 100 K
the studied flames, because a detailed mechanism of
at x ~ 0.4 mm. At small distances from the burner (≤ 0.3
chemical reactions for this flame is believed to be well-
mm) the experimental data are also higher than
known; using the mechanism and PREMIX code from
unperturbed ones (curve 3). The radiation losses lead to
the CHEMKIN suite  the flame temperature profile
significant decrease of both measured (curve 3) and
can be predicted with appropriate accuracy. For our
calculated (curve 2) temperatures that are in evident
calculations we chose GRI 3.0 mechanism  for
agreement between each other from the point of view of
methane oxidation. This mechanism reproduces a large
visible peculiarities of these profiles (the experimental
number of various experimental data including the data
curve 3 reveal even non-monotonous dependence of x).
on burning velocity and structure of methane flames.
The radiation corrections for performed numerical
Fig. 1 illustrates the obtained experimental and
experiment (curve 4) reveals also non-monotonous
numerical temperature profiles. The curve 1 represents
dependence on x that reflects complex character of the
the profile obtained by the calculations based on the
flow over thermocouple in the region of the flame front.
detailed chemical kinetics that was considered as a
Figures 3, 4 and 5 show the numerical results for
temperature profile of unperturbed flame and will be
streamlines picture, the fields of longitudinal velocity
denoted later as Tflame, curve 2 corresponds to the
and isotherms, respectively, obtained for the case of
meaning temperature profile measured using ribbon
flowing around the ribbon thermocouple (20x125 µm),
thermocouple 20x125 µm without any corrections for
the center of which is placed at the distance x = 400 µm
radiation losses. Numerical data denoted by symbols
from the burner surface and for ε = 0. This position
were obtained for two values of emissivity factor ε –
points 3 correspond to the value ε = 0, i. e. to the case of
adiabatic conditions on the thermocouple surface, while y, µm
points 4 – to the value ε = 0.4 providing satisfactory 1500
description of post-flame temperature.
As can be seen from Fig. 1 the numerical results 1200
obtained taking into account radiation losses (points 4)
are in good agreement with experimental data (curve 2). 900
For ε = 0 the calculated temperature values in
temperature gradient zone are significantly higher than 600
temperature values in unperturbed flame.
For more detailed analysis of thermocouple-induced 300
temperature perturbations the same data as in Fig. 1 are
replotted in Fig. 2 in the form of difference between 0
perturbed and unperturbed values (∆T = T - Tflame, for 0 300 600 900 1200 1500
curves 1– 3), where curve 1are the data for ε = 0, curve Distance from burner, µm
2 – the data for ε = 0.4, curve 3 – experimental data.
The curve 4 represents the difference T(ε = 0.4) – T(ε = Fig. 3. Field of streamlines.
0 200 400 600 0 200 400 600
Distance from burner, µm
Distance from burner, µm
Fig. 5. Field of isotherms.
Fig. 4. Field of longitudinal velocities.
readings overestimate the actual temperature values in
corresponds to the flame zone with the most significant
the certain respect. Therefore, the numerical results
perturbation of temperature field by the thermocouple
show that the ribbon thermocouples in the flame zone,
(see curve 1 in Fig. 2; the flow direction in these figures
where the temperature gradient takes place, disturb
is from left to right). In unperturbed flame (without
significantly the gas flow, in which the measurements
thermocouple), all the streamlines are parallel among
themselves. As can be seen from Fig. 3, the flow
represents typical laminar flow enveloping the obstacle
(thermocouple). The distortions of the streamlines RDX flame
The simulation of the combustion-gases flow over
become more significant in small vicinity of the
thermocouple for RDX flame was performed to estimate
thermocouple surface. The parallel alignment of
the level of the effect of overestimation the temperature
streamlines is distorted by the thermocouple inserted
measured by the thermocouple for the flame with high
into the flow at a large distance from the plane of
mass flow rate. Additional reasons for such simulation
symmetry (streamline at y ~ 1200 µm). Therefore, when
are connected with discussions concerning the
a thermocouple with the width of 125 µm is inserted
correctness of thermocouple measurements of the
into the flow, the streamlines suffer distortion at a
temperature profiles in the flames of condensed systems
distance ~20 times greater than the thermocouple width.
. The simulation was performed for regime with
An analysis of the field of longitudinal velocities (in
pressure 20 atm (g = 0.83 g/cm2·s) from those studied in
cm/s) at y = 0 (Fig. 4) showed that the velocity of the
. To clarify the effect of thermocouple shape on the
flow directed to the central part of the thermocouple,
where its junction is placed, decelerates from ~20 cm/s results two thermocouples, ribbon 3x60 µm and near
practically to zero as the thermocouple surface is quadratic 13x14 µm with similar transverse area were
approached and increases slowly when moving away
from the thermocouple. One can see that the flow T, K ∆T, K
velocity behind the thermocouple at the distance of 3000 1200
~200 µm from it is several times less than in 2500 3 1000
unperturbed flame (x = 600 µm, y = 700 µm). Therefore, 2
as it can be seen from Fig. 4, the region of distortion of 2000 800
longitudinal velocity field is about 250 µm in front and 1
behind the thermocouple. The dimensions of the region 1500 600
of distortion (~500 µm) are more than 20 times greater 4
than the thermocouple thickness and comprise in this 1000 400
particular case almost half of the width of the flame 500
Thus, the thermocouple forms a deceleration zone in 0 0
the gas flow, and reflects the temperature inside this 0 10 20 30 40 50 60
zone. Since the thermocouple is in the gas flow with
intensive sources of heat release due to the chemical Distance from burning surface, µm
reactions, the flow deceleration by the thermocouple
causes the distortion of the field of isotherms (see Fig. Fig. 6. Temperature profiles for two
5) at the region of thermocouple position. The thermocouples with similar transverse
thermocouple appears to be in the region with higher area (1 – Tflame, 2 - T3x60; 3 - T13x14; 4 –
temperature (~100 degrees) in comparison with ∆T3x60; 5 – ∆T13x14).
unperturbed flame and, consequently, the thermocouple
reveals small effect of radiation on thermocouple
temperature for this flame – maximum difference T(ε =
0) – T(ε = 0.4) was about 40 K that is an order of
150 magnitude less than the discussed effect of overheating
of the thermocouple. Small effect of radiation on
120 thermocouple temperature in RDX flame is explained
by large values of Reynolds number determining the
heat exchange between the flow and the body.
The discovered effect of overestimation of the flame
temperature by thermocouple due to deceleration of the
flow in the vicinity of the thermocouple seemed to be
very important and should be taken into account while
0 experimental measuring of the temperature profiles in
0 15 30 45 60 the flames.
Distance from burning
surface, µm References
1. A.A.Zenin, Progress In Astronautics and
Fig. 7. Field of isotherms for RDX flame Aeronautics 143 (1992) 197-231.
flow over ribbon (3x60 µm) 2. W.E. Kaskan, Proc. Combust. Inst. 6 (1957) 134.
thermocouple. 3. A.A. Zenin, J. Prop. Power 11 (4) (1995) 752-759.
4. A.G.Tereshchenko, P.A.Skovorodko, O.P.
tested. While plotting the temperature profiles the Korobeinichev, D.A. Knyazkov, A.G. Shmakov, in:
calculated values of thermocouple temperature were Proc. 5th Int. Seminar on Flame Structure. CD
ascribed to the middle point of the thermocouple. version, O.P. Korobeinichev (Ed.), Parallel’ Ltd.,
The obtained temperature profiles are illustrated in Novosibirsk, Russia, 2005, OP-06.
Fig. 6, where the line 1 represents the profile Tflame that 5. R. J. Kee, F. M. Rupley, and J. A. Miller.
is considered as undisturbed one and was used to CHEMKIN-II: A Fortran Chemical Kinetics
Package for the Analysis of Gas Phase Chemical
deduce the source term Qsource (T ) . The results for
Kinetics. Sandia National Laboratories Report No.
considered thermocouples are shown by symbols (2 – SAND89 - 8009B (1989).
T3x60; 3 – T13x14). The differences between the 6. R.M. Fristrom, A.A. Vestenberg, Flame Structure.
thermocouple and undisturbed temperatures (4 – ∆T3x60 Metallurgia, Moscow 1969.
= T3x60 – Tflame, 5 – ∆T13x14 = T13x14 – Tflame) are also 7. А.A. Zenin, S.V. Finjakov, Comb. Explos. Shock
shown for comparison. As it is seen from these profiles Waves 42 (5) (2006) 32-45.
the ribbon thermocouple leads to significant 8. M.W. Beckstead, Comb. Explos. Shock Waves 43
overestimation of temperature compared to quadratic (2) (2007) 134-136.
thermocouple. Maximum values of temperature 9. Korobeinichev O.P. et. al., Comb. Sc. & Tech. 116-
difference ∆T3x60 is about 400 K, while for quadratic 117 (1996) 51.
thermocouple the maximum overestimation ∆T13x14 is 10. Broc., S. De Benedictis, G. Dilecce, M. Vigliotti,
about two times lower. R.G Sharafutdinov, and P.A. Skovorodko, J. Fluid
Therefore, to obtain true (undisturbed) temperature Mech. 500 (2004) 211 – 237.
gradient in the combustion wave with high mass flow 11. Hirschfelder J. O., Curtiss Ch. F., and Bird R. B.
rates (0.83 g/cm2·s and more) it is necessary to take into Molecular Theory of Gases and Liquids. Wiley,
account the overestimation of thermocouple temperature New York, 1954.
due to the disturbances induced in the flow field by 12. http://www.me.berkeley.edu/gri_mech/.
thermocouple itself. For rectangular thermocouples with
the same transverse area the disturbances are increased
with increasing the thermocouple width. As a result of
gas dynamic disturbances induced by the thermocouple
the temperature gradients measured by the
thermocouple exceed true (undisturbed) values.
Fig. 7 illustrates the field of isotherms for RDX
flame flow over ribbon (3x60 µm) thermocouple. The
distortions of the field induced by the thermocouple in
the flow field are quite evident and are similar to those
for methane flame (see Fig. 5).
All the reported results for RDX flame were
obtained for ε = 0. Simulation of the flow with ε = 0.4