Simple Explicit Equations For Transient Heat Conduction in Finite Solids
Simple Explicit Equations For Transient Heat Conduction in Finite Solids
Simple Explicit Equations For Transient Heat Conduction in Finite Solids
冋 册
perature and concentration, respectively. The early engineering 1/2
problems included burning of bricks, steaming of wood, and vul- 1
1 = 1,⬁ 共1兲
canization of rubber. In most problems, the heat transfer coeffi- 1 + A/Bik
cient h 共W / m2 K兲 and the related Biot number
where 1,⬁ are the asymptotic values for Bi→ ⬁, 1,⬁ = / 2 for
hL the plate, 1,⬁ = 2.4048 for cylinder, and 1,⬁ = for the sphere.
Bi = = finite By fitting the straight line to the exact values, he obtained the
k following constants: 共i兲 A = 2.24 and k = 1.02 for plates 共ii兲 A
= 2.45 and k = 1.04 for cylinders, and 共iii兲 A = 2.7 and k = 1.07 for
are finite and can be assumed to be uniform. In the Biot number,
spheres. Beck et al. 关5兴 provided an accurate yet more involved
k 共W/m K兲 is the thermal conductivity and L is the characteristic
approximation for 1, valid for flat plates and Bi⬍ 2. Yovanovich
length.
关6兴 used the general expression of Churchill and Usagi 关3兴,
The solutions to the transient conduction in large plates, long
cylinders, and spheres with symmetric convection boundary con-
ditions are known since 1822 关1兴. The solution is comprised of Y = 共1 + Zn兲1/n 共2兲
Fourier infinite series, which are difficult to evaluate, although the with
series converge rapidly. For
y共z兲
␣t Y⬅
Fo ⬅ 2 ⬎ 0.2 y共z → 0兲
L
where Fo is the Fourier number, ␣ 共m2 / s兲 is thermal diffusivity, t y共z → ⬁兲
共s兲 is time, only one term is needed. The key obstacle to using the Z⬅
y共z → 0兲
Fourier’s solutions are the eigenvalues. For finite Bi, the boundary
condition of the third kind yields transcendental implicit eigen- for correlating functions y共z兲 that have asymptotically small and
value equations that have to be solved numerically or graphically. large values of the independent variable z. Yovanovich obtained
Therefore, in most heat transfer textbooks, eigenvalues are tabu-
lated as a function of Bi. 1,⬁
冋 冉 冊册
Because of the importance of the problem, a number of charts 1 = 共3兲
have been developed to provide a graphical solution of the Fourier 1,⬁ n 1/n
1+
equations. In 1947 Heisler 关2兴 published a set of semilog charts, 1,0
valid for 0.2艋 Fo⬍ ⬁, which has been reproduced in most heat
transfer textbooks and has been widely used by students and prac- where 1,0 is the asymptotic value of the first eigenvalues for
titioners. Bi→ 0, 1,0 = 冑Bi for plates, 1,0 = 冑2Bi for cylinders, and 1,0
Empirical correlations are usually more accurate than the semi- = 冑3Bi for spheres. The constant n was determined by fitting the
log charts and are more convenient for use in design and analysis above equation to the exact eigenvalues. Yovanovich determined
than tabulations and complex analytical results 关3兴. Several ex- n = 2.139 for plates, n = 2.238 for cylinders, and n = 2.314 for
plicit approximations of the transcendental eigenvalue equations spheres. Recently, Ostrogorsky and Mikic 关7兴 used the successive
substitution method to derive a set of explicit expressions for 1
for plates, cylinders, and spheres valid for 0 ⬍ Bi⬍ ⬁.
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received December 7, 2007; final manuscript
For BiⰆ 1, temperature gradients within solids are considered
received June 10, 2008; published online October 17, 2008. Review conducted by A. negligible, so that the lumped capacity solution provides an ex-
Haji-Sheikh. plicit expression for transient temperature,
Journal of Heat Transfer Copyright © 2009 by ASME JANUARY 2009, Vol. 131 / 011303-1
冉 冊
surrounding fluid. The transient temperature 共t兲 / i is function of
共t,x兲 x 2
Fo and Bi. A / V is the surface to volume ratio of the solid. For a = Bplate cos 1 e−1 Fo 共11兲
2L plate, the surface to volume ratio is A / V = 1 / L, resulting in i L
共t兲 −BiFo where Bplate is the Fourier coefficient,
=e 共4兲
i sin 1
Bplate = 2 共12兲
For a long cylinder having radius R, A / V = 2 / R, yielding 1 + sin 1 cos 1
共t兲 −2BiFo For Fo= 0.25, the error is approximately 1%, depending on Bi
=e 共5兲 and x / L. The key obstacle to getting 共t , x兲 / i is 1, which ac-
i
counts for the boundary conditions and thus contains the Biot
For a sphere having radius R, number. Equation 共9兲 is an implicit transcendental equation whose
solution requires iterations.
共t兲 −3BiFo
=e 共6兲
i 2.1.1 Transient Temperature 共t , x兲 / i in a 2L Plate. An ex-
plicit form of Eq. 共9兲 can be obtained by expanding the tangent,
since A / V = 3 / R. Incropera and De Witt 关8兴 advised that the error
associated with using the lumped capacitance model is small for 13 2 15
Bi⬍ 0.1. tan共1兲 = 1 + + +¯
3 15
Here, we convert the one-term Fourier series solutions into a set
of approximate explicit equations, which provide the transient and, for small 1,
temperature profiles in plates, cylinders, and spheres. The tran- tan共1兲 = 1 共13兲
sient temperature profiles can be evaluated without the use of
numerical procedures. We derive For small 1, Eqs. 共13兲 and 共9兲 yield 关4兴
共a兲 a low Biot number approximation, which is almost as 1 = 冑Bi 共14兲
simple as the lumped capacity solution yet accounts for Because of Eq. 共13兲, the error in Eq. 共14兲 grows with Bi to 1.65%
spatial variation and thus is more accurate
for Bi= 0.1 and to 4.98% for Bi= 0.3. Next we simplify the Bplate
共b兲 a more complex broad-range approximation valid for
by expanding sin and cos for small values of x 关9兴,
Fo⬎ 0.2 and 0 ⬍ Bi⬍ ⬁
x3 x5
We will consider only one-dimensional transients since the multi- sin共x兲 = x − + −¯ 共15兲
3! 5!
dimensional problems can be readily split into one-dimensional
ones using the separation of variables. x2 x4
cos共x兲 = 1 − + ¯ 共16兲
2! 4!
2 Low Biot Number Solution Equation 共12兲 is approximated as
2.1 Large Plate With 2L Thickness. The mathematical for- 1
mulation of the problem is Bplate ⬇
1 − Bi/4
1 2 Since for small ,
=
␣ t x2
1
where = T − T f . The boundary conditions are ⬇1+ 共17兲
1−
共L,t兲 one gets
BC共1兲:− k = h共L,t兲
x
Bi
Bplate ⬇ 1 + 共18兲
共0,t兲 4
BC共2兲: =0
x and
The initial condition is
IC:共x,0兲 = i
共t,x兲
i
冉 冊 冉 冊
= 1+
Bi
4
x 2
cos 1 e−1 Fo
L
共19兲
where = x / L and n = nL is a product of the eigenvalues n and Finally using Eq. 共16兲 again,
the plate thickness, defined by the transcendental equation,
n tan n = Bi 共8兲
共t,x兲
i
冉 冊冋 冉 冊 册
= 1+
Bi
4
1−
Bi x
2 L
2
e−BiFo 共21兲
However, for Fo艌 0.2 关2兴 Eq. 共7兲 may be approximated by one one gets an explicit equation, nearly as simple as the lumped
term. Hence, system equation.
Figure 1共a兲 shows the transient temperature calculated using
1 tan 1 = Bi 共9兲 Eqs. 共20兲 and 共21兲 together with the the exact Fourier series solu-
and tion and the lumped capacity model 共Eq. 共4兲兲. Equations 共20兲 and
共21兲 overlap, giving a less than 0.86% i error for Bi= 0.1. with
Figure 1共b兲, shows 共x , t兲 / i calculated for Bi= 0.3, which is at 共t,0兲
the border of the low Bi number approximation. The highest error =0
is less than 2.88% i 共for Fo= 3 at the centerline兲. The lumped r
冏 冏
capacity solution provides the highest error of 7.6% i. Since there
is no notable difference in accuracy between Eqs. 共20兲 and 共21兲, 共t,r兲
−k = hS
the simpler Eq. 共21兲 should be used. r r=R
2.1.2 Total Energy Released Q共t兲 / Q0 From a Plate. The total 共0,r兲 = i
energy content of the plate that can be removed by heat transfer
into the surroundings at T f is The temperature field in the infinite cylinder is
fines the energy that remains in the solid that can be removed by
where Fo= ␣t / R2, Bi= hR / k and, n = nR are roots of the tran-
heat transfer. Then, the exact solution for the dimensionless heat
scendental equation,
loss Q共t兲 / Q0 is obtained as:
nJ1共n兲 − BiJ0共n兲 = 0
where
2Bi
Bcylinder = 共33兲
共21 + Bi2兲J0共1兲
and
1J1共1兲 = J0共1兲Bi 共34兲
J 1共 1兲 =
1
2
− 冉 冊 冉 冊
1/2
1!2!
3
+
1/2
2!3!
5
¯ =
1 13 15
2
− +
16 386
−¯
共36兲
Using the first terms of Eqs. 共35兲 and 共36兲 and in Eq. 共34兲 gives
the expression 关4兴
1 = 冑2Bi 共37兲
Now using Eq. 共37兲 in Eq. 共33兲 yields
1 Bi
Bcylinder = ⬇1+ 共38兲
1 − Bi/2 2
A comparison to the exact solution shows that the approximation Fig. 2 Transient temperature „r , t… / i in cylinders calculated
used for the plates, using Eqs. „41… and „42…, along with the exact five-term Fourier
series solution „full lines… and the lumped capacity solution
Bi „Eq. „5……, dashed lines. „a… Bi= 0.1 „b… Bi= 0.3.
Bcylinder ⬇ 1 + 共39兲
4
is more precise. At Bi= 0.3, Eq. 共39兲 gives Bcylinder = 1.0750, while
the exact value is 1.0711. Thus,
冉 冊冉 冊
Q共t兲 Bi2 2
共t,r兲 Bi r =1−4 2 2 e−1 Fo 共44兲
1 共1 + Bi 兲
2 2
= 1+ J0 1 e−1 Fo 共40兲 Q0
i 4 R
or
and using Eq. 共37兲,
冉 冊冉 冊
Q共t兲 2
共t,r兲 = 1 − Ccylindere−1 Fo
= 1+
Bi
J0 冑2Bi r e −2BiFo
共41兲 Q0
i 4 R
where
Using the two-term approximation of J0共1兲 gives
冉 冊冋 冉 冊 册
Ccylinder ⬇ 1
共t,r兲 Bi Bi r 2
= 1+ 1− e −2BiFo
共42兲 For example, for Bi= 0.3, Ccylinder = 0.9952. For small Bi, using
i 4 2 R
1 = 冑2Bi
Figure 2 shows Eqs. 共41兲 and 共42兲 plotted along with the exact
solution 共full lines兲 and the lumped capacity solution 共dashed Q共t兲
lines兲. Since the two-term approximation of J0共1兲 is precise, Eqs. = 1 − e−2BiFo 共45兲
Q0
共41兲 and 共42兲 overlap. Hence, the simpler Eq. 共42兲 should be used.
For 共a兲 Bi= 0.1 共Fig. 2共a兲兲, the error in Eq. 共42兲 is below 0.88% i. The error grows with Bi. For 0 ⬍ Bi⬍ 0.3, the error is less than
For Bi= 0.3 共Fig. 2共b兲兲, the highest error for Eq. 共42兲 is 2.95% i 2% Q0.
compared to the 6.9% i error for the lumped capacity. 2.3 Sphere With Radius R. The mathematical formulation is
2.2.2 Total Energy Released Q共t兲 / Q0 From Cylinders. The di-
mensionless heat loss from cylinders is obtained by substituting
the temperature profile 共30兲 into Eq. 共23兲, setting u = r / R and n
1 1 2
=
␣ t r2 r
r
r
冉 冊 共46兲
The transient temperature field in the sphere is 共t , x兲 / i 共Eq. No.兲 Q共t兲 / Q0 共Eq. No.兲
冉 冊 r
共 兲关 共 兲 兴
Bi Bi x 2
1 − e−BiFo 共28兲
冉 冊
⬁ sin n Plate 1+ 1− e−BiFo 共21兲
共t,r兲 sin n − n cos n
兺
R 2 4 2 L
=2 e −n Fo
共47兲
i n − sin n cos n
共 兲关 共 兲 兴 1 − e−2BiFo 共45兲
Cylinder 2
r Bi Bi r
n=1
n 1+
4
1−
2 R
e−2BiFo 共42兲
R
共 兲关 共 兲 兴 1 − e−3BiFo 共60兲
Sphere 2
Bi Bi r
where the eigenvalues n = nR are defined by 1+ 1− e−3BiFo 共55兲
4 2 R
n cot n = 1 − Bi
The one-term approximation of Eq. 共47兲 is
冉 冊
4
Q0 = 3 R3c共Ti − T f 兲 共57兲
共t,r兲 r R − 2Fo
= Bsphere sin 1 e 1 共49兲 The one-term approximation of Eq. 共56兲 is
i R 1r
where Q共t兲 1 共sin 1 − 1 cos 1兲2 − 2Fo
=1−6 3 e 1 共58兲
sin 1 − 1 cos 1 Q0 1 1 − sin 1 cos 1
Bsphere = 2 共50兲
1 − sin 1 cos 1 or
and Q共t兲 2
= 1 − Cspheree−1 Fo 共59兲
1 cot 1 = 1 − Bi 共51兲 Q0
where
2.3.1 Transient Temperature in Spheres. For small values of
1 and Bi, Csphere ⬇ 1
1 1 13 1 1 For example, for Bi= 0.3, Csphere = 0.9985. Replacing Csphere = 1
cot 1 = − − − ¯ ⬇ − 共52兲
1 3 45 1 3 and 1 = 冑3Bi in Eq. 共59兲 gives
Combining Eqs. 共51兲 and 共52兲 provides a simple approximation
Q共t兲
for 1 关4兴, = 1 − e−3BiFo 共60兲
Q0
1 = 冑3Bi 共53兲
2.4 Summary. The equations for transient temperature 共 , t兲
Using small number expansions for sin and cos, one gets Bsphere and heat released Q共t兲 / Q0, derived in Sec. 2, are listed in Table 1.
⬇ 1. However, the approximation used for plates provides an ac-
curate approximation, 2.4.1 Low Biot Number Solution (LBi). The lumped capacity
solution is normally used up to Bi= 0.1 关8,10兴, where it gives
Bi ⬃3% i error or up to Bi= 1 / 6 关12兴, where it gives ⬃5% i error
Bsphere ⬇ Bplate ⬇ 1 +
4 共see Figs. 1–3 and Table 3兲. This is justified since in many cases
one does not know the convection coefficient better than ⫾10%.
giving for Bi= 0.1 Bsphere = 1.0250, while the exact value is 1.0298.
The low Biot number solution presented herein accounts for
Therefore,
nonzero gradients within the solid. Based on Figs. 1–3 and Tables
共t,r兲
i
冉 冊冑
= 1+
Bi
4
R
3Bi · r
冉冑 冊
sin 3Bi
r − 2Fo
R
e 1 共54兲
1–3,
• it is more precise than the lumped capacity solution by a
factor of ⬃3,
Equation 共54兲 is further simplified by using a two-term approxi- • it can be used up to Bi⬃ 1 / 3, where it gives ⬃3% error, and
mation for sin, • it has a general form valid for plates, cylinders, and spheres,
共t,r兲
i
= 1+
4
冉 冊冋 冉 冊 册
Bi
1−
Bi r
2 R
2
e−3BiFo 共55兲 共t, 兲
i
冉 冊冋
= 1+
Bi
4
1−
2
e 册
Bi 2 −nBiFo
= B · F · e−nBiFo 共61兲
Figure 3 shows the transient temperature calculated using Eq.
共55兲 along with the exact solution 共full lines兲 and the lumped where n = 1 for plates, n = 2 for cylinders, and n = 3 for spheres.
capacity solution 共dashed lines兲. For Bi= 0.1, the error is less than Equation 共61兲 is almost as simple as the lumped capacity solution.
0.97% i. For Bi= 0.3, the error is less than 3.1% i. For Bi→ 0, B and F asymptotically approach 1, reducing Eq. 共61兲
to the lumped capacity model. Therefore, one could consider the
2.3.2 Total Energy Released Q共t兲 / Q0 From Sphere. The di- lumped capacity solution to be a limiting case of the low Biot
mensionless heat loss from spheres is obtained by substituting the number solution presented herein.
temperature profile 共47兲 into Eq. 共17兲, setting u = r / R and n = 3 and The coefficient F accounts for the spatial variation of tempera-
integrating. The result of this integration is ture. Surprisingly, the two-term approximation,
Bcylinder = 1 + Bi/4.5
2Bi
1,cylinder2 =
1 + Bi/4
3Bi
1,sphere2 = 共65兲
1 + Bi/5
Using Eqs. 共64兲 and 共65兲 in Eqs. 共11兲, 共32兲, and 共49兲 gives a set of
extended low Biot number approximations listed in Table 2 for
Fig. 3 Transient temperature „r , t… / i in spheres calculated
using Eq. „55… along with the exact five term Fourier series so- plates, cylinders, and spheres. This set is more precise than the
lution „full lines… and the lumped capacity solution „Eq. „6……, lumped capacity solution by one order of magnitude 共Table 3兲,
dashed lines. „a… Bi= 0.1. „b… Bi= 0.3. and can be used up to Bi⬃ 1.
冑
3.1.1 Coefficients Bplate and Cplate. The coefficient Bplate de-
Plate Bi
1 + Bi/ 3 共 兲 共 兲
1+
Bi
7
x
L
2
cos 1 e−1 Fo fined by Eq. 共12兲 asymptotically approaches constant values,
Bplate共Bi→ 0兲 = 1 and Bplate共Bi→ ⬁兲 = 1.273. Hence, expression 共2兲
Cylinder
冑 2Bi
1 + Bi/ 4 共 兲共 兲
1+
Bi
J
4 0 1R
r − 2Fo
e 1
is constant,
y共z → ⬁兲
冑
Z⬅
共 兲 共 兲
= 1.273
Sphere 3Bi
1+
Bi R r 2
sin 1 e−1 Fo y共z → 0兲
1 + Bi/ 5 3.5 1r R
This violates Churchill and Usagi’s requirement that “y共z → 0兲 and
y共z → ⬁兲 cannot be both constant” 关3兴. However, the difference
Table 3 Error % i in LBi and ELBi solutions „see Tables 1 and 2… and in the lumped capacity solution
Bi= 0.03 −0.95 −0.27 0.09 0.74 0.12 −0.1 0.89 0.25 0.04
Bi= 0.1 −2.98 −0.86 0.03 2.4 0.88 −0.38 2.9 0.97 0.11
Bi= 0.3 −7.67 2.9 0.45 6.9 2.9 −0.40 8.0 3.1 0.20
Bi= 1 / 2 −11.0 4.8 −0.40 9.2 5.3 −0.87 12.3 6.2 0.22
Bi= 2 / 3 −12.9 5.3 −0.87 12.4 8.6 −3.4 15.3 8.7 0.23
Bi= 1 −2.1 −2.3 0.27
Bi= 2 −6.9 −5.5 0.61
which holds both limits and gives less than 1.5% error. The form
of approximation 共66兲 with one exponent n did not reduce the
error sufficiently. However,
1
1 − Cplate = 0.189 共69兲
1 + K/Bim
with K = 9.5 and m = 1.25 gives less than 0.5% error. Then, intro-
ducing the second independent exponent
1
1 − Cplate = 0.189 共70兲
共1 + 3.8/Bi1.116兲1.613
gives less than 0.14% error in the 0 ⬍ Bi⬍ ⬁ range 共Fig. 4共b兲兲.
Equation 共70兲 can be written in the form of Churchill and Usagi’s
Eq. 共2兲,
1
Cplate = 1 − 0.189 共71兲
关1 + 共8.61/Bi1.8兲0.62兴1/0.62
冋 册
Fig. 4 Error: „a… B coefficients, error
„%… = „Bexact − Bapprox… / Bexact à 100%; „b… Coefficients, error Q共t兲 1 −12Fo
„%… = „Cexact − Capprox… / Cexact à 100% = 1 − 1 − 0.189 1.116 1.62 e 共72兲
Q0 共1 + 3.8/Bi 兲
3.1.2 First Eigenvalue for Plates 1. For Bi⬍ 0.1, Eq. 共14兲
provides the simplest approximation for 1. This can be seen in
Bplate − Bplate共0兲 is subject to a simple approximation, Fig. 5, which shows the error exact − approx = ⌬1. For 0 ⬍ Bi
1 ⬍ ⬁, the first eigenvalue can be approximated as 关7兴
冑
Bplate − 1 = 0.273
1 + 1/Bi
Bi /2
冉冊
1共Bi兲 = = 共73兲
which holds both limits and gives less than 2% error. To further
reduce the error, we employ a more complex approximation,
2 2 冑1 + 2.4674/Bi
1 + Bi
1
Bplate − 1 = 0.273 共66兲 Approximation 共73兲 is precise at both limits 共error ⬍1% for Bi
共1 + K/Bin兲1/n
⬍ 0.5 and Bi⬎ 25兲 but gives ⌬1 = 0.029 共or ⌬1 / 1 ⫻ 100%
which also holds both limits. To minimize error in the 0.1⬍ Bi = 2.29%兲 error at Bi⬇ 4, see Fig. 5共a兲兲. For most applications this
⬍ 10 range, the constants are adjusted to n = 1.5 and K = 2.42, is acceptable because
1 • interpolating between the 1/Bi lines in the Heisler charts
Bplate = 1 + 0.273 共67兲
共1 + 2.45/Bi1.5兲2/3 should result in a higher error;
• most correlations for convective heat transfer coefficient h,
Figure 4共a兲 shows the error in the Fourier constants for transient have uncertainty of about ⫾10%, which causes uncertainty
temperature in plates, cylinders, and spheres. Equation 共67兲 gives in Bi and 1.
less than 0.16% for 0 ⬍ Bi⬍ ⬁. Note that Eq. 共67兲 has the form of
Eq. 共2兲, yet the constant, 1.8011.5 = 2.45, does not seem to be re- At Bi= 4, Eq. 共73兲 gives a maximum error 1,approx = 1.2354 in-
lated to the limits 共zero or 0.273兲. Combining Eqs. 共67兲 and 共11兲 stead of 1,exact = 1.2646, a ⌬1 = 0.029 error. Using 1 = 1.235 in
gives the transient temperature, Eq. 共9兲, one gets Bi= 3.543, so that the error introduced by Eq.
共t,x兲
i
冋
= 1+
0.273
共1 + 2.42/Bi1.5兲2/3
册 冉 冊
x 2
cos 1 e−1 Fo
L
共68兲
共73兲 is equivalent to making a 共4 − 3.543兲 / 4 ⫻ 100= 11.4% error in
h or Bi. Yet, Eq. 共73兲 is considerably improved by adjusting the
constant and the exponent,
The Fourier coefficient Cplate for transient heat loss defined by
Eq. 共27兲 also approaches asymptotically constant values for small /2
and large values of Bi. The coefficient decreases from Cplate共Bi 1共Bi兲 = 共74兲
冑1 + 2.3/Bi1.035
→ 0兲 = 1 to Cplate共Bi→ ⬁兲 = 0.8106. The expression Z is again con-
stant, Equation 共74兲 gives error ⌬1 = 0.0044 at Bi⬇ 1 where exact
y共z → 0兲 ⬇ 0.85 共or ⬇0.5% error兲. It has the same form as the Luikov’s
Z⬅ = 1.234 equation but contains different constants. Luikov’s Eq. 共1兲 is sig-
y共z → ⬁兲 nificantly less accurate in the 0.1⬍ Bi⬍ 5 range. It is more accu-
so that Eq. 共2兲 cannot be readily implemented. Instead, the differ- rate for Bi⬍ 0.05 where 1 = 冑Bi is yet more precise. The approxi-
ence Cplate共0兲 − Cplate is subject to a simple approximation, mation proposed by Yovanovich 关6兴,
0.602
Bcylinder − 1 =
共1 + K/Bin兲1/n
Adjusting the constants K and n brings the error down to less than
0.23% 共see Fig. 4共a兲兲,
0.602
Bcylinder = 1 + 共78兲
共1 + 4.8/Bi1.64兲1/1.64
Combining Eqs. 共78兲 and 共32兲 gives
共t,r兲
i
冋
= 1+
0.602
共1 + 4.8/Bi1.64兲0.61
册冉 冊
r 2
J0 1 e−1 Fo
R
共79兲
0.3083
1 − Ccylinder =
1 + 7/Bi
which holds both limits and gives less than 2.5% error. As for
Cplate, we assume
Fig. 5 Error= exact − approx = ⌬1 as a function of the Biot num-
ber for „a… plate, „b… cylinder, and „c… sphere 0.3083
1 − Ccylinder = 共80兲
共1 + K/Bim兲n
Adjusting the exponents to K = 2.58, m = 1.08, and n = 2.35 gives
/2 less than 0.12% error in the 0 ⬍ Bi⬍ ⬁ range 共see Fig. 4共b兲兲. Us-
冋 冉冑 冊 册
1共Bi兲 = 共75兲 ing Eq. 共80兲 in Eq. 共44兲 provides an explicit expression for
/2 2.139 1/2.139
Q共t兲 / Q0 valid for 0 ⬍ Bi⬍ ⬁,
1+
冋 册
Bi
Q共t兲 0.308 −12Fo
is the most accurate approximation for 0.01⬍ Bi⬍ 2 共Fig. 5共a兲兲. =1− 1− 1.08 2.35 e 共81兲
Q0 共1 + 2.58/Bi 兲
The maximum error is ⌬1 = 0.0026 at Bi⬇ 3 where exact ⬇ 1.2
共or ⬇0.22% error兲. However, it is more complex than Eq. 共14兲 or 3.2.2 First Eigenvalue for Cylinders 1. For Bi⬍ 0.1, Eq. 共37兲
共74兲. Equation 共75兲 can be further simplified as provides the simplest and yet very accurate approximation. For
0.1⬍ Bi⬍ ⬁, the approximation based on Refs. 关4,7兴
/2
1共Bi兲 = 共76兲 2.4048
共1 + 2.62/Bi1.07兲0.468 1共Bi兲 = 共82兲
How much accuracy in 1 is needed? In calculating 共 , t兲 / i,
冑1 + 2.62/Bi1.05
the main source of error is the uncertainty in the heat transfer yields sufficient accuracy, provided that the constants are adjusted
coefficient h. Therefore, one can expect typically ⫾10% error in to minimize the error. It can be seen in Fig. 5共b兲 that Eq. 共82兲 is
the Biot number. However, the relation between Bi and 1 is more precise than Eq. 共1兲 with A = 2.45 and k = 1.04, except for
exceedingly nonlinear. For example, using Bi= 1.1 instead of Bi⬍ 0.07. The approximation proposed by Yovanovich 关6兴,
Biexact = 1 due to 10% error in h will result in error ⌬1
= 0.89035− 0.86033= 0.03002. Using Eq. 共74兲 will give approxi- 2.4048 2.4048
冋 冉冑 冊 册 冉 冊
mately ten times less error 共see Fig. 5共a兲兲. Yet, using Bi= 110 1共Bi兲 = 2.238 1/2.238
= 0.4468
2.4048 3.005
instead of Biexact = 100 will give ⬃30 times lower error in the 1+ 1 + 1.119
eigenvalue, ⌬1 = 1.55665− 1.55525= 0.0014. Based on the 2Bi Bi
above, one could conclude that for BiⰇ 1, the correlations need to 共83兲
provide 1 with four digits of accuracy, or the error in h will be
augmented. However, for BiⰇ 1, the resistance within the solid yields the maximum error of ⌬1 = 0.0082 at Bi⬇ 3 where exact
controls the transient, and a relatively large error in h 共or Bi兲 will = 1.8 共or 0.46% error兲. By further simplifying Eq. 共83兲 and adjust-
cause a small error in 1 and 共 , t兲 / i. ing the constants,
1共Bi兲 =
2.4048
共1 + 3.28/Bi1.125兲0.446
共84兲
共t,r兲
i
冋
= 1+
1 R
共1 + 8.87/Bi1.76兲0.568 1r
册
sin
R
冉 冊
1r − 2Fo
e 1
the error is reduced to 0.00455 also at Bi⬇ 3 共or ⬇0.25% error兲 共86兲
共see Fig. 5共b兲兲. However for Bi⬎ 10, the error is slightly in- The Fourier coefficient Csphere for the heat loss asymptotically
creased. approaches Csphere共Bi→ 0兲 = 1 and Csphere共Bi→ ⬁兲 = 0.392. The ap-
proximation
3.3 Spheres
0.392
3.3.1 Coefficients Bsphere and Csphere. The coefficient Bsphere 1 − Csphere =
1 + 7.7/Bi
has asymptotic values: Bsphere共Bi→ 0兲 = 1 and Bsphere共Bi→ ⬁兲 = 2.
As for plates and cylinders, the approximation holds both limits but gives less than 3.5% error. Adjusting the
1 constant and the exponents as in Eq. 共70兲 for the plate gives
Bsphere − 1 = 共85兲
共1 + K/Bin兲1/n
0.392
1 − Csphere = 共87兲
holds both limits. Adjusting the constant to K = 8.87 and the ex- 共1 + 2.36/Bi1.09兲2.85
ponents to n = 1.76 gives less than 0.21% error in the 0 ⬍ Bi⬍ ⬁
range 共Fig. 4共a兲兲. The transient temperature in spheres is obtained Approximation 共87兲 yields less than 0.21% error in the 0 ⬍ Bi
by substituting Eq. 共85兲 into Eq. 共49兲, ⬍ ⬁ range 共Fig. 4共b兲兲. Then using Eq. 共87兲 with Eq. 共59兲 gives
关 兴 共 兲
Plate 共t , x兲 0.273 x 2
= 1+ cos 1 e−1 Fo 共68兲
i 共1 + 2.42/ Bi1.5兲2/3 L
关 兴
Q共t兲 1 2
= 1 − 1 − 0.189 e−1 Fo 共72兲
Q0 共1 + 3.8/ Bi1.116兲1.62
关 兴共 兲
Cylinder 共t , r兲 0.602 r − 2Fo
= 1+ J e 1 共79兲
i 共1 + 4.8/ Bi1.64兲0.61 0 1 R
关 兴
Q共t兲 0.308 2
=1− 1− e−1 Fo 共81兲
Q0 共1 + 2.58/ Bi1.08兲2.35
共t , r兲
关 共 兲 兴 共 兲
1r − 2Fo
Sphere 8.87 −0.568
R
= 1+ 1+ sin e 1 共86兲
i Bi1.76 1r R
关 兴
Qsphere共t兲 0.392 2
=1− 1− e−1 Fo 共88兲
Q0 共1 + 2.36/ Bi1.09兲2.85
Q共t兲
Q0
=1− 1− 冋
0.392
1.09 2.85 e
共1 + 2.36/Bi 兲
−12Fo
册 共88兲
Fig. 7 Error in midplane temperature, error
„%… = „exact − approx… / i à 100%, for 2L plates as a function of Bi
3.3.2 First Eigenvalue for Spheres 1. For Bi⬍ 0.1, Eq. 共53兲 and Fo, calculated using equations listed in Table 4, with cor-
provides the simplest and yet very accurate approximation. For relations for 1 listed in „a… Table 5 and „b… Table 6.
0.1⬍ Bi⬍ ⬁, the approximation based on Refs. 关4,7兴 with adjusted
constants,
3.4 Results for the Broad-Range Correlations
1共Bi兲 = 共89兲
冑1 + 2.93/Bi1.08 Figure 6 shows the transient temperature 共t兲 / i at = 0, plotted
in the format of the Heisler charts 关2兴. The exact Fourier series
yields sufficient accuracy 共see Fig. 5共c兲兲. Luikov’s Eq. 共1兲 with solution is given by the full lines. The approximate solutions were
A = 2.7 and k = 1.07 for spheres is less accurate than Eq. 共89兲, calculated using 共i兲 the correlations listed in Table 4 共ii兲 the eigen-
except for Bi⬍ 0.2. The approximation proposed by Yovanovich values given in Table 5. The error in 共t兲 / i is well below 1%.
关6兴, This level of precision cannot be obtained from the Heisler charts
or by interpolating the tabulated values for 1. Since the log-
冋 冉冑 冊 册 冉 冊
1共Bi兲 = = 0.43215 linear charts do not provide a sufficient resolution, the error in
2.314 1/2.314 3.9662 midplane temperature for a 2L plate is shown in Fig. 7. For Bi
1+ 1 + 1.157
3Bi Bi ⬎ 10, the error is particularly low 共⬎0.1% 兲 because conduction
within the solid 共not convection at the boundary兲 controls the
共90兲
transient.
yields maximum error of ⌬1 = 0.0180 at Bi⬇ 4, where exact
⬇ 2.45 共or 0.71% error兲. The constants and the exponents can be 4 Conclusions and Recommendations
further adjusted,
The low Biot number solution 共LBi, Table 1兲 is approximately
three times more accurate than the lumped capacity solution. It is
1共Bi兲 = 共91兲 recommended for the 0.03⬍ Bi⬍ 0.3 range, where the error of the
共1 + 4.1/Bi1.18兲0.4238
lumped capacity solution is considerable 共see Table 3兲. The ex-
to give less than ⌬1 = 0.051 or 0.21% error. tended low Biot number solution 共ELBi, Table 2兲 is approximately
ten times more accurate than the lumped capacity solution but is
Table 5 Explicit correlations for the first eigenvalue: simpler Table 6 Explicit correlations for the first eigenvalues: the gen-
option eral option