N. R.

Keltner
Thermal Test and Analysis Division,
Surface Temperature
Sandia National Laboratories,
Albuquerque, N. M. 87185 Measurement Errors
J.V. Beck Mathematical models are developed for the response of surface mounted ther-
mocouples on a thick wall. These models account for the significant causes of errors
Department of Mechanical Engineering, in both the transient and steady-state response to changes in the wall temperature.
Michigan State University,
In many cases, closed form analytical expressions are given for the response. The
East Lansing, Mich. 48824
Mem. ASME cases for which analytical expressions are not obtainable can be easily evaluated on
a programmable calculator or a small computer.

Introduction
The analysis of and correction for the errors involved when from this solution. The solutions will be in either closed form
making surface temperature measurements with ther- or in a form for easy numerical evaluation.
mocouples attached externally is an important problem in The geometry of the problem being considered is shown in
experimental heat transfer. The errors result from the ther- Fig. 1. Two homogeneous, isotropic bodies with temperature
mocouple installation altering the local surface temperature independent thermal properties are in thermal contact over a
distribution and the effects of heat transfer in the ther- circular region of radius, a. The interface is considered to
mocouple attachment and the thermcouple itself. The errors have an imperfect contact that can be thermally modeled by a
may be transient, steady-state, or both. contact coefficient, h; if h goes to infinity there is perfect
There are many separate causes of the errors but the most contact. Except for the contact area, the surface of the semi-
significant are: infinite body is considered to be adiabatic.
1 Thermal constriction effects in the body to which the The thermocouple is modeled as a single semi-infinite
thermocouple is attached cylinder with lateral surface heat loss characterized by a heat
transfer coefficient, hc, with the ambient at the initial ther-
2 Thermal inertia of the thermocouple [1-14]
mocouple temperature. If the cylinder is uninsulated, the
3 Imperfect contact between the thermocouple and the
coefficient is that due to convection and/or linearized
surface [7, 10,11,15-18]
radiation. If it is insulated, axial conduction is assumed to
4 Heat loss from the thermocouple to the ambient [7, 11,
occur only in the wire and radial conduction only in the in-
15-18]
sulation. Following the simplications utilized by Moffat [15],
5 The effective junction location being displaced from the
Sparrow [16], and others for insulated thermocouples, the
surface [10, 11]
lateral resistance due to both surface losses and the insulation
A number of methods and simplifying assumptions have can be given by
been utilized in modeling the transient response of ther- 1 \n(r0/a)
mocouples. An early paper by Henning and Parker [1] used R- (1)
separation of variables for an idealized problem involving a hr2irr 2wk,
connecting hemisphere of infinite conductivity and zero heat where r0 is the outer radius of the insulation layer and £, is its
capacity. If a modification of their model suggested in an thermal conductivity. An overall heat transfer coefficient can
attached review is not used, the model gives large errors in the be defined from equation (1) and can be used in the same
early to middle times. Others have solved less idealized models manner as hc.
by utilizing (a) finite differences [4, 5, 14], {b) Duhamel's In general, the temperature and heat flux at the contact
integral with numerical evaluation [5, 6, 11, 13], and (c) vary with radial location as well as with time. In this paper the
Duhamel's integral with Laplace transforms [7, 8, 10,13]. emphasis is upon providing an average heat flux or tem-
A recent paper by Cassagne et al. [7] gives a thorough perature across the interfaces. Two complementary assump-
analysis of the transient response problem, except for the tions are used: radially constant (/) heat flux, and («) tem-
effect of junction displacement. The analytical approach is perature. For the thermocouple application the analysis given
similar to the approach used herein. However, because the for radially constant interface heat flux is particularly ap-
exact solution for uniform heat flux is used as the kernel, the propriate for small dimensionless times. The analysis for a
solution must be evaluated numerically. The present method
uses easily evaluated kernels in Duhamel's equation and in
many cases analytical expressions for the response are ob-
tainable.
The analysis is developed for the case of a thermocouple
mounted on a thick wall by using the Unsteady Surface
Element Method (USEM) [8, 19]. This method utilizes
fundamental solutions in the two bodies (i.e., the surface or coefficient adiabatic
surface
substrate and the thermocouple or wire) as kernels in
Duhamel's convolution equation. The solutions may be
developed using either temperature or heat flux kernels. An
overall solution which considers all of the above effects will be
developed first and then certain special cases will be obtained

Contributed by the Heat Transfer Division for publication in the JOURNAL OF

HEAT TRANSFER. Manuscript received by the Heat Transfer Division May 7, Fig. 1 Idealized geometry for a thermocouple attached externally to a
1982. solid surface

3 1 2 / V o l . 105, MAY 1983 Copyright © 1983 by ASME Transactions of the A S M E

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 radially constant interface temperature gives accurate results 77 = </>,/[•**+1/B], (9a)
for middle to late times.
Ti = 42/K[s€+UB] (.9b)
Heat Flux Based U S E Solution where
The heat flux based USE method starts with Duhamel's 77 = (r 1 -r 1/ )/(r„-r 2/ ), (10a)
integral for the average interface temperature for body, j , in n = (T2-T2i)/(Tu-T2i) (106)
the form
In further use of these equations, the " * " superscripts are
omitted.
7}(0,0 = Tji + ~^'Q qj(\)<t>j(0,t- \)d\ (2) The relations given by equation (7) and (9a), (9b) are ter-
med transfer functions. Though there is specific reference to
where qj(t) is the flux into body, j . The kernel <t>j(Q,t) is the the geometry shown in Fig. 1, these relations can be employed
contact area-average temperature rise in body j at Zj = 0 for a for many other cases of two connecting geometries, both
unit, qj. Notice that qx and q2 can be related by semi-infinite and finite. The task now is to invert (9b) for the
Fig. 1 geometry which is distinguished from others by its
<7i(0 = -qi(t) = h[T2{Q,t)-Ti(0,t)] (3) particular <j> functions.
where again area averages are implied.
Taking the Laplace transform of (2) for j = 1 and 2 and Temperature Based U S E Solution
solving for qx (s) using the Laplace transform of (3) gives
The temperature based USE method starts with Duhamel's
Qi'(Tu - T2i) = - \/s[s(ix + 0 2 ) + l/h] (4) integral for the area-average heat flux through the disk
contact
where the Laplace transform notation is used

QM- J>(0,X)- •TJi]6a,Q,t-\)d\ (11)

o e-*flr,(0<ft (5) dt
where Oqj(Q,t) is the area average heat flux for a unit increase
This equation can be made dimensionless by defining in surface temperature
t* = uxt/a2 (6a)
ddj(Zj,t)
qi M0,0 = A J/ dA (12)
a* " (6b) dZj ZJ=0
* kx[Txi-T2i]
K=k2/kx,A = a2/ax (6c, d) The kernel 6j(Zj,t) is the temperature at (zj,t) for a unit rise in
surface (r = 0 to a) temperature. Taking the Laplace trans-
4>*=<i>jkj/a (fie) form of (3) and using (11) gives
**= <t>\ + tf/K (6/) <?, =s(Tx-Tu)Oql = -s(T2-T2i)6q2 (13)
= ha/kx,s* =a2s/ax (6g,h) The Laplace transform of (3) can be rearranged to
The B symbol in (6g) is used to denote a Biot type modulus
7 W 7 , = 7^7^,- + (T2i - Tu)/s - qx/h (14)
based on the contact heat transfer coefficient. Using the above
notation with the " * " superscripts dropped yields for Employing this expression in (13) produces
equation (4) (T2 - T2i)/(TU - T2i) = S s8qlql ~6q2
0,iql /s(6ql + 6q2 + sd Q2/h)
/h) (15)
qx = -l/s(s4>+l/B) (7) Similar expressions can be derived for Tx and qx. Rather than
For the perfect contact case, 5—oo, and thus giving these expressions, dimensionless forms are given
below. Let
qx = -l/(s2$) (8)
9*qJ = dqja/kj (16a)
The same procedure yields for the interface area-average
temperatures TqX +K6q2 (166)

Nomenclature

q* dimensionless heat flux -

a = thermocouple wire radius equation (6b)
thermal diffusivity ratio rerfc(x) exp (x2)erfc(x) 13 = K/<A
A = equation (6d) r„ outside radius of an in- e = temperature kernel
contact Biot modulus sulated thermocouple function
B = equation (6) s = Laplace transform par- % = heat flux function-
lateral surface Biot ameter equation (12)
Bi = modulus t* = dimensionless time - equa- p = density
specific heat tion (6a) <t> = heat flux kernel func-
c = equation (lib) temperature tion - equation (2)
D = contact heat transfer T* dimensionless temperature * = equation (6f)
h = coefficient x = position along the ther-
Subscripts
lateral heat transfer mocouple
coefficient x* = dimensionless position, 1 = related to the substrate
k thermal conductivity x/a 2 = related to the ther-
K thermal conductivity mocouple
ratio - equation (6c) Greek Symbols
Ji = initial value of a pa-
heat flux a = thermal diffusivity rameter for body/'

Journal of Heat Transfer MAY 1983, Vol. 105/313

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 and use the dimensionless terms defined by equations (6) and , . ,25 ^ ^ ^ ^ ^ ^ ^ = -
(10). One then can write with the " *" being omitted
Q\ --K6ql0q2/D (17a)
D=dq+sKeqXeql/B (176) t . ^ ^ / / / /
f, = Kdq2/(sD) (18a) i.o____-'/'/ /
T2 = dql/(sD) (186) 1.5_____—^ /
Notice that for perfect contact the expressions for qx, Tx> a n d
t2 are unchanged but D becomes simply 6q. Furthermore,
1.0 __ ^
TI=T2.
Evaluation
To evaluate the sensor temperatures, the unit step response
(or kernel) functions are required. For the case of a semi-
infinite body with a step change in temperature over a disk r
Fig. 2 Ideal intrinsic thermocouple response for a step input
region, this function has been developed in the oblate
spheroidal coordinate system which is a natural coordinate
system for the problem.
A "late" time asymptotic solution developed by Norminton In the <?-based USE solution for the full thermocouple
and Blackwell [20] is used. Analyses [5, 21] have shown that problem, <j>] is the average temperature over the disk resulting
this solution gives good results for dimensionless times greater from a unit step change in heat flux [24]
than 0.1 (at/a1 > 0.1, where units are used.) The heat flux f>, = 2 V ? / 7 r - 2 / / 7 r (27a)
based solutions are simpler to use and more accurate over the 3/2
time range 0 < t < 0.3. j[ — li/Aj -—2/J// mno2 (276)
Using the late time asymptotic solution, it can be shown and 0 2 is the temperature at the end of a semi-infinite rod for
that the average heat flux over the disk is a unit step change in heat flux [22].
0gl=[8a7r3r1/2+4]/7r (19)
which is in dimensionless form [8]. Taking the Laplace 4>2 = 2-jAt/ir (28a)
transform of (19) gives 0 2 =vC4A 3 / 2 (286)
4 For a cylinder with heat loss, the step temperature change
.+ — (20)
ins solution gives the exact Laplace transform for 6q
For a cylinder with lateral heat loss, that heat flux into the 6q2 = ^Js + 4ABi/s\fA (29)
end for a unit step temperature change at the end is [22] It can be shown for one-dimensional problems that 6qi and
eXp( Bi0 4>t are related by
"92- -^ +V44Blerf(V44BT0 (21)
V7T? dqi = l/s2^ (30)
which has a Laplace transform of thus
0»,2=-VJTt/lDI/SV/l
~ =^s + AABi/s^A (22) VA
< f2t , =
• = (31)
From equation (186), the temperature at the thermocouple W s + 4/4Bi
junction is given by The expressions given by (28) are exact, but those given by
?2 (0,5) = 9ql M6ql + K6q2 + sK6ql §q2 /B) (23) (27) are valid only for small values of / or equivalently large
where 6ql and dq2 are given by equations (20) and (22) and values of s.
The effect of junction displacement can be accounted for by
Bi = hca/2k2 equations (24) and (25) in the temperature based development.
hc = lateral heat transfer coefficient Combining results gives
To account for the junction being at a location, x, inside the ~ 02
cylinder (see Fig. 1), the transform of the temperature at x is • „exp(
. r v -. x^J(s
. . , „ .+ 4ABi)/A
,... (32)
related to the transform of the surface temperature by K[s<b+\/B]
It does not appear that the inverse Laplace transform of the
T2(x,s)^s[T2(0,sye2(x,s)] (24) full equations, equations (26) and (32), can be obtained
with directly. The inversion has been performed numerically using
a ten-term, Gaver-Stehfest method programmed for a desktop
§2 = exp( - x-J{s + 4Am)/A)/s (25) computer and for a programmable calculator [25, 26]. Before
where x* = x/a and the " * " was dropped in equations (24) considering the full problem, certain limiting cases (some of
and (25). The quantity, 02> is the temperature in a semi- which have convenient closed-form solutions) will be con-
infinite rod with lateral heat loss when exposed to a step sidered. These use limiting values for three parameters; x, Bi,
temperature change on its end [22, 23]. Using equations (20), and B.
(22), (23), and (25) in (24) yields
F,(s) Case I - Ideal Intrinsic Thermocouple
T2(x,s)= *
F, (s) + KF2 (s)/^As + KFI (s)F2 (s)/\[AB The ideal intrinsic thermocouple problem has been con-
sidered by numerous authors [1, 4-8]. For this case the
parameter values are: x = 0, Bi = 0, and B=oo, When these
exp(-xF2(s)/VA) values are used in equation (26), the result can be inverted
(26)
exactly to give the late time solution of [8]
where F{ (s) = 8/TC2JS + 4/TTSF2(S) = Vs + 4.4Bi. T2(0,t) = 1 - C, rerfc(C2 vT) (33)

314/Vol. 105, MAY 1983 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 an electrical output that went negative in the early to middle
times.
For the case /3 = 4/3, the agreement between the present
model and the results of two large finite difference codes was
± 3 percent [8]. In [5], an earlier version of the model was
compared with data from nine experiments. The experiments
involved Type K (Chromel/Alumel) thermocouples on a 304
stainless steel plate with temperature steps of 55K to 275K.
The larger temperature steps produced a significant change in
thermal properties. However, when the model was evaluated
using the property values at a single intermediate temperature
the agreement was better than ± 6 percent.

Case II - Ideal Beaded Thermocouple

The ideal beaded thermocouple has no lateral surface heat
Fig. 3 Response of an ideal beaded thermocouple to a step change in loss and no contact resistance at the interface. It differs from
substrate temperature (K = A = 1) the intrinsic thermocouple by having the effective junction
displaced from the surface by the thickness of the ther-
1
mocouple bead. The parameter values for this case are x^O,
1.0
=
0.001
' Bi = 0,fl=oo.
lil = 0 ^ ^
0,01
The response is obtained by using equations (26) and (32)
i).a with x^0 to account for the junction displacement. The late
time response is
0.1
0.6
r 2 (*,0 = e r f c ( ^ ) - C , e x p ( c 2 ^ + Ci,)*

0.1 -
erfc
2v'At
~+C yr) (35)
0.2
where C, and C 2 are defined below equation (33).
The results are shown in Fig. 3 for the case where the bodies
have identical properties (i.e., A =K=P=1), and the junction
is displaced from the surface by an amount up to the wire
diameter. The result of the junction being displaced from the
Fig. 4 Response of an intrinsic thermocouple to a step change in surface is a zero response at zero time and an increase in the
substrate temperature when there is lateral heat loss (K = A = 1) error for all but very late times. The delay between the step
input and the first response can be obtained from the results
for a semi-infinite body subjected to a step temperature
where C, = /3/(8/TT 2 +/S), C 2 = 4/(8/TT+/3TT), rerfc(x) = change. Note that the very late time response approaches that
exp(x2)erfc(x), and /3 = K/VA. The parameter, /3, figures of the ideal intrinsic thermocouple.
prominently in the analytical expressions for the ther-
mocouple temperature. The interface temperature given by
equation (33) is applicable for t* > 0.1. Case III - Intrinsic Thermocouple with Lateral Heat
The early time response is obtained from the heat flux Loss
method equation (32). The solution is
The case of lateral heat loss is more realistic because
T2 =Di [1 +expCC>^)erf(JD2v7)] (34)
radiative losses and/or convective losses are always present.
where £), = 1/(1 +/3), D2 = 2/3/TT(1 +(3). For this case the parameter values are: x = 0, Bi^O, and
The two solutions have been evaluated for a range of B=<x>. It can be shown that the Laplace transform of the late
property values for the two bodies which cover most possible time solution can be simplified to the form
combinations of metals. It can be seen that the level of
response at zero time and the transient response up to the final T- / •> VJ+ C7
value are governed by a single parameter, /3. These results are r , ( i ) = —j= , (36)
shown in Fig. 2 where equation (34) was used for / < 0 . 1 and where C 7 , C 8 , and C 9 are constants depending on the par-
equation (33) for t > 0.1. Notice that the matching between the ticular problem.
two solutions is excellent. Amos [27] has shown that the inverse Laplace transform of
From Fig. 2 one can observe that the error (the difference equation (36) cannot be obtained in closed form. The inverse
between the undisturbed temperature, Tu, and the ther- transform in the form of complex expression of constants,
mocouple temperature, T2(0,t), is always largest at zero time polynomials of vT, exponentials, complementary error
and goes to zero at large times. The zero time error and the functions, modified Bessel functions and their various
dimensionless time for which a significant error exists are products, along with two easily evaluated integrals, has been
both directly proportional to |8. obtained. This expression can be used to look at parametric
Both single-wire and two-wire intrinsic thermocouples are effects directly. The numerical inversion procedure is easier to
used in practice. For most common thermocouple pairs, there evaluate and produces excellent agreement.
are significant thermal property differences. During a The numerical inversion procedure was to evaluate the
transient for the two-wire thermocouples, this results in a response for varying values of the Biot modulus. These results
temperature difference between the two junctions, and the are shown in Fig. 4 for Bi = 0, .001, .01, .1 an&K=A = 1. This
total thermoelectric circuit has three elements. Depending on range of the Biot modulus covers much of that estimated to
the thermoelectric power of the substrate, very strange results bound thermocouple applications [181. Note that while there
could be produced as in [12], where a thermal pulse produced is little difference in the early time response, the lateral heat

Journal of Heat Transfer MAY 1983, Vol. 105/315

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 l.U l.U
l "^^—
/ 6
0 S
0.8
/ / "" 8
W// A 7
/ /
0,6 / 0,6

B- no - — ^ ^ / / CURVE 1/B B; x
1 0 0 0
0.4 /m /\ / 0.1 2 0 .001 0
3 1 0 0
1 1 .001 0
0,2 0.2

^S
/ 1/IJ
6
7
3
0
0
1
1
0
,001
0
.001
2
2
2
2
no nn __——^^^ ~J/
f
Fig. 5(a) Response of an intrinsic thermocouple to a step change in
Fig. 6 Effects of the various parameters on the response o f a ther-
substrate temperature when there is contact resistance (K = 1.5486,
mocouple to a step change in the substrate temperature (K = A 1)
A = 1.3485)

For this case the parameter values are Bi = 0, 5-finite and

x = 0, and then x-finite. For these values, the late time solution
(equation (26)) can be inverted to give
B
7-2(0,0= j [-^(l-rerfctfC.V/^)

•(l-rerfc(C2Vf72)j (38)
C,
where C = (B/fi + ir2B/8 + TT/2)

C, ,C2=(C±^IC2 - 2irB/(3)/2

Dl = (C, - 7r/2)/(C, - C2),D2 = (C2 - TT/2)/(C 2 - C,)

Note the C,, C 2 , Du and D2 are different than those

previously defined.
Fig. 5(b) Response of a beaded thermocouple to a step change in
substrate temperature for finite contact resistance (K = A = 1) Now if xis finite, then starting from equations (24) and (26)
with Bi = 0 the resulting transform equation can be inverted to
give
loss produces a steady-state offset (i.e., there is a temperature W ) =
measurement error at all times). Note that as Bi increases, the j [^r[ erfc (2^)- exp ( ClX
V2/1
time at which the response begins to deviate from the ideal
ch
><^Tt+ c
^)]+JtV<2k)
case decreases.
One surprising result observed in Fig. 4 is that the tem-
peratures reach a maximum and then decline slightly. This
drop occurs for all values of Bi, although it is very small at
Bi = .001. The behavior may not be all bad in that the steady-
state result is reached earlier than for the zero loss case. For
example, the response is within 1 percent of its steady-state The corresponding early time solutions can be obtained,
although the late time solutions are the most useful, par-
value at / = 20, while the Bi = 0 response is within 1 percent of
its steady-state value at the much larger time of t = 2000. ticularly for the beaded thermocouple. Figures 5(a) and 5(b)
The steady-state offset can be obtained by evaluating the show the effects of varying the contact resistance for x = 0 and
full model at very large times (r~ 10000). Alternatively, the x = 2. The effects of finite values of B are similar to the effects
steady-state portions of equations (19) and (21) can be used of the ideal beaded thermocouple (Figure 3) at late times. The
directly to develop the expression early time response rises faster. Note that if B is not infinite in
the x = 0 case that the initial measured temperature is zero.
T2 (0, oo) = 1 / ( l + Tr/rVBi/2) (37) For reference, approximate values of B for a 24 gage (.02 in.)
which presumes that T2i = 0 and the ambient temperature is thermocouple on a mild steel plate are 0.05-0.1 for a one mil
also zero. (.001 in.) epoxy bond and 0.02-0.4 for a pressed contact
Evaluating T^O.oo) from either the full model or equation depending on the pressure.
(37) produces essentially the same results and these agree (to
the accuracy of reading charts) with the results presented in Case V - Combined Case
[16] or [18].
This case includes any or all of the possible effects.
Reasonable ranges for the parameter values are 0 < x < 2 ,
Case IV - Contact Resistance at the Interface
0 < B i < 0 . 5 , and . l < f i < o o . The possible combinations for
The interface contact resistance results when the ther- this case are endless, particularly if the ranges of substrate
mocouple is simply pressed against the surface or adhesively and thermocouple properties are included. To demonstrate
bonded with a thin bond line whose capacity can be neglected. how the various parameters and their combinations affect the

316/Vol. 105, MAY 1983 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1.0 perfect, and there is no lateral heat loss, and the junction is at
the surface.
The model uses equations (23) and (20) but replaces (22)
with
1 y V §q2 =y[A/b2sin + 7(1,0,b)/s (42)
1
where b = cylinder radius/contact radius and l(\,0,b) is a
constant [29]. The expression was obtained from the response
CURVE t/l of a semi-infinite cylinder heated by a disk source by using the
1 1.0 relation given in equation (30); the original expression is valid
2 0.9
3 0.8 for t/b2 > 1. The inversion has not been performed
1 0.7 analytically. Numerical evaluations for several contact area
ratios are shown in Fig. 7. These results show that even
though initial and final values are identical to the Case I
results, that the partial contact significantly slows the
response in the middle times.
Fig. 7 Response of an intrinsic thermocouple with partial contact
(K = A = 1) Applications
While using the models can help to develop a feeling for the
early time response, the late time response, the time to reach errors in temperature measurements, few experiments involve
steady state, and the steady state error, the full solutions were a step temperature change. A useful adjunct to the step
evaluated for A =K= 1, x = 0 and 2, Bi = 0 and .001, and B= 1 response is the response to a temperature ramp.
and oo. The results shown in Fig. 6 indicate that having a The development of the step response models assumed
beaded thermocouple (x = 2) most strongly affects the early constant thermal properties. Thus the equations are linear,
time response, that contact resistance slows the response, and and the principle of superposition can be utilized for inputs of
that the effect of lateral heat loss is at the late times and the arbitrary form. The response, R(t), to a forcing function or
steady-state error. substrate temperature change, F(t), is given by
The case of x = 0 but Bi finite and 5 < o o , has been
developed by Cassagne et al. [7] using a method similar to the W) = ~X F(\)C{t-\)d\ (43)
dt Jo
g-based USE. There appears to be excellent agreement bet-
ween their work and the present model. For the case where where C(t) is the unit step response (equations (26) or (32)).
K = A = l,Bi = .00l,B=l, andx = 0, a comparison of the two For the case where F(t) = Ht, H = constant, t > 0, the
models showed exact agreement at early times, a slight response of an ideal intrinsic thermocouple (using equation
divergence in the middle time range (maximum difference of (33)) is
1.7 percent at / ~ 1 ) and agreement within 0.5 percent at late
times. R(t) = H\t- -—^[rerfc(C 2 v7)- 1 +2C2V777r]l (44)
For large times with both imperfect contact and lateral heat L C2 J
loss and for an ideal beaded thermocouple
r 0 exp(-2*VBi) R(t) = //[^/ 2 erfc(C 3 /v7) - - ^ - [(exp(2C 2 C 3
T20c,0»- l+2A"VBi(l/B+7r/4)

ZfVBi 1 C
1+ (40) + Ci/)erfc(-^+C2Vr)
W T T ? 1 + 2#VBi(l/.B+7r/4)-
Notice that this expression shows a decrease of T2{x,t) with t
for nonzero Bi. Also, the steady-state result is - erfc(C3 /vT) + 2C 2 v7/erfc(C3 /v7)]l (45)
T2(x,<x) = r„exp( - 2 W I i ) / [ l + 2irVBi(l/fi + TT/4)] (41) where C 3 = x/lVA.
This expression shows that there is no steady-state error unless Note that the error in the slope or the temperature rise rate
Bi is greater than zero; if Bi>0, T2(x,oo) is reduced both by is given by
dF
the presence of Bi in the exponential and in the term in the (0 dR(t)
brackets. Slope error = —— — (46a)
at at
Case VI - Partial Contact = H[\-C(t)) (466)
Thus the slope error at time, /, is equal to the undisturbed rise
The previous cases have assumed that there is uniform
rate multiplied by (1-unit step response at t).
contact over the entire end of the thermocouple. If the
A second problem is the estimation of the forcing function
thermocouple is pressed against the surface (instead of being
(i.e., the undisturbed temperature). Rearranging the Laplace
welded or bonded) the contact will be at the high points of the
transform of equation (43) gives
two surfaces or if the weld is poor it will have voids. While
this will not affect the response of beaded thermocouples F=R/sC (47)
(except as it affects the value of the contact coefficient), it If the inverse Laplace transform exists, then a good estimate
could have a serious effect on intrinsic thermocouples [4]. An of F(t) can be obtained. As an example consider a response
alternate problem — enlargement of the junction during function modeled by i ^ v T + H2t for the case of the ideal
welding [9] - will not be considered. intrinsic thermocouple. The Laplace transform of C(t), from
For the partial contact case only the late time model will be equation (26) can be put in the form
utilized because the very early time response will be identical
C=l/5-//4/V]r+//4(Vir+//6) (48)
to that in Case I. The problem is approximated by having the
contact radius be smaller than the cylinder radius. Also, it is The resulting forcing function can be obtained analytically
assumed that the contact over the smaller radius disk is and is

Journal of Heat Transfer MAY 1983, Vol. 105/317

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 r ^H^H6^R- 2H2H, + 2H2H4H6 -, Contact," International Journal of Heat and Mass Transfer, Vol. 23, 1980, pp.
F(t)=\ L [l-rerfc(//7v7)] 1207-1217.
J 8 Keltner, N. R., and Beck, J. V., "Unsteady Surface Element Method,"
2H6
JOURNAL OF HEAT TRANSFER, Vol. 103, 1981, pp. 759-764.
9 Maglic, K. D., and Marsicanin, B. S., "Factors Affecting the Accuracy of
+ [//,+ 2H2H4 /v^r] vT+ i/ 2 / (49) Transient Response of Intrinsic Thermocouples in Thermal Diffusivity
Measurement," High Temperatures-High Pressures, Vol. 5, 1973, pp. 105-110.
10 Quandt, E. R.., and Fink, E. W., "Experimental and Theoretical Analysis
where H7 = H6/(l -H4H6). Similar results can be obtained of the Transient Response of Surface Bonded Thermocouples," Bettis
for polynomials in vT of increasing order; however, the Technical Review, WAPD-BT-19, Reactor Technology, June, I960, p. 31.
algebra becomes increasingly complex. The simplest approach 11 Wally, K., "The Transient Response of Beaded Thermocouples Mounted
on the Surface of a Solid," Proceedings of the 23rd International In-
is to numerically invert R(s). strumentation Symposium, Instrument Society of America, 1977, pp. 127-132.
12 Heckman, R. C , "Intrinsic Thermocouples in Thermal Diffusivity
Experiments," Proceedings of the Seventh Symposium on Thermophysical
Summary Properties, ASME, New York, 1977.
13 Litkouhi, B., "Surface Element Method for Transient Heat Conduction
Using the unsteady surface element method, easily Problems," Ph.D. dissertation, Department of Mechanical Engineering,
evaluated models have been developed for estimating the Michigan State University, E. Lansing, Michigan, 1982.
errors in surface temperature measurements with ther- 14 Shewen, E. C , "A Transient Numerical Analysis of Conduction Between
mocouples for a variety of conditions. The models can be Contacting Circular Cylinders and Halfspaces Applied to a Biosensor," MS
thesis, University of Waterloo, Waterloo, Ontario, Canada, 1976.
adapted to a particular situation by specifying relatively few 15 Moffat, R. J., "Temperature Measurement in Solids," Paper 68-514,
parameters. For specific limiting cases, it is shown that the Advances in Test Measurement, Vol. 5, ISA, Pittsburgh, Pa., 1968.
results are in excellent agreement with the literature. 16 Sparrow, E. M., "Error Estimates in Temperature Measurements," ch.
1, Measurements in Heat Transfer, 2nd ed., edited byE.R.G. Eckert and R. J.
Goldstein, Hemisphere Publishing Corp. 1976.
Acknowledgments 17 Green, S. J., and Hart, T. W., "Accuracy and Response of Ther-
mocouples for Surface and Fluid Temperature Measurements," Temperature,
The research was sponsored by Sandia National Its Measurement and Control in Science and Industry, Vol. 3, pt. 2, Reinhold,
Laboratories, which is operated by the Western Electric New York, p. 695.
Corporation for the Department of Energy under Contract 18 Hennecke, D. K., and Sparrow, E. M., "Local Heat Sink on a Con-
vectively Cooled Surface—Application to Temperature Measurement Error,"
No. DE-AC04-76DP00789 and by the National Science International Journal of Heat and Mass Transfer, Vol. 13, 1970, pp. 287-304.
Foundation under Grant No. CME-79-20103. The computer 19 Beck, J. V., and Keltner, N. R., "Transient Thermal Contact of Two
code used for estimating thermocouple response for Cases Semi-Infinite Bodies over a Circular Area," Paper AIAA-81-1162, AIAA 16th
I-V is available from the first author. The code, ap- Thermophysics Conference, Palo Alto, Calif., June, 1981.
20 Norminton, E. J., and Blackwell, J. H., "Transient Heat Flow from
proximately 100, lines, is written in HP BASIC. Constant Temperature Spheroids and the Thin Circular Disk," Quarterly
Journal of Mechanics and Applied Mathematics, Vol. 17, 1964, pp. 65-72.
21 Marder, B. M., and Keltner, N. R., "Heat Flow From a Disk by
References Separation of Variables," Numerical Heat Transfer, Vol. 4, 1981, pp. 485-497.
22 Luikov, A. V., Analytical Heat Diffusion Theory, Academic Press, New
1 Henning, C. D., and Parker, R., "Transient Response of an Intrinsic York, 1968.
Thermocouple," JOURNAL OF HEAT TRANSFER, Vol. 39, 1967, p. 146. 23 Carslaw, H. S., and Jaeger, J. C , Conduction of Heat in Solids, 2nd ed.,
2 Burnett, D. R., "Transient Measurement Errors in Heated Slabs for Oxford University Press, London, 1967.
Thermocouples Located at an Insulated Surface," JOURNAL OF HEAT TRANS- 24 Beck, J. V., "Average Transient Temperature Within a Body Heated by a
FER, Vol. 83, 1961, p. 505. Disk Heat Source," Heat Transfer, Thermal Control, and Heat Pipes, Vol. 70
3 Larson, J. B., and Nelson, E., "Variables Affecting the Dynamic of Progress in Astronautics and Aeronautics, AIAA, New York, (1980).
Response of Thermocouples Attached to Thin Skinned Models," JOURNAL OF 25 Stehfest, H., "Numerical Inversion of Laplace Transforms," Com-
HEAT TRANSFER, Vol. 41, 1969, p. 166. munications of the ACM, Vol. 13, 1970, pp. 47-49 and p. 624.
4 Gat, U., et al., "The Effect of Temperature Dependent Properties on 26 Woo, K., "TI-59 Inverts Laplace Transformation for Time-Domain
Transients Measurement with Intrinsic Thermocouple," International Journal Analysis," Electronics, Oct. 9, 1980, pp. 178-179.
of Heat and Mass Transfer, Vol. 18, 1975, p. 1337. 27 Amos, D. E., On the Inverse of a Laplace Transform and its Numerical
5 Keltner, N. R., "Intrinsic Thermocouple Response," Proceedings of the Evaluation, SAND81-2470, Sandia National Laboratories, Albuquerque,
20th International Instrumentation Symposium, Instrument Society of N.M.,Nov. 1981.
America, 1974, p. 91. 28 Cetinkale, T. N., et al., "Thermal Conductance of Metal Surfaces' in
6 Keltner, N. R., and Bickle, L. W., "Intrinsic Thermocouple Contact." Proceedings of the International Conference of Heat Transfer,
Measurement Errors," Paper 76-HT-65, ASME-AIChE Heat Transfer Con- London, 1951, pp. 271.
ference, St. Louis, MO., Aug. 1976. 29 Beck, J. V., "Transient Temperatures in a Semi-Infinite Cylinder Heated
7 Cassagne, B., et al., "Theoretical Analysis of the Errors Due to Stray by a Disk Heat Source," International Journal of Heat and Mass Transfer,
Heat Transfer During the Measurement of a Surface Temperature by Direct Vol.24, 1981, pp. 1631-1640.

318/Vol. 105, MAY 1983 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use