Professional Documents
Culture Documents
www.elsevier.com/locate/ijhmt
a
Laboratory of Structural Engineering and Building Physics, Department of Civil and Environmental Engineering,
Helsinki University of Technology, PL 2100, FIN-02015 HUT, Espoo, Finland
b
Andritz Group, Tammasaarenkatu 1, FIN-00180 Helsinki, Finland
Abstract
An analytical method leading to the solution of transient temperature filed in multi-dimensional composite circular
cylinder is presented. The boundary condition is described as time-dependent temperature change. For such heat con-
duction problem, nearly all the published works need numerical schemes in computing eigenvalues or residues. In this
paper, the proposed method involves no such numerical work. Application of Ôseparation of variablesÕ is novel. The
developed method represents an extension of the analytical approach derived for solving heat conduction in composite
slab in Cartesian coordinates. Close-formed solution is provided and its agreement with numerical result is good which
demonstrates a good accuracy of the developed solution form.
2005 Elsevier Ltd. All rights reserved.
Keywords: Multi-dimension; Composite circular cylinder; Heat conduction; Analytical method; Close-formed solution
0017-9310/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2005.06.019
342 X. Lu et al. / International Journal of Heat and Mass Transfer 49 (2006) 341–348
Nomenclature
2
where Xj(x) is a variable-separated function which satis- o ujm 1 oujm oujm
fies the homogeneous form of Eq. (3.3a). Then by substi- kj þ l2jm ujm ¼ þ bm T 01 ðtÞ
or2 r or ot
tuting Xj(x) into Eq. (3.3a) results in r 2 ½rj1 ; rj ; j ¼ 1; . . . ; n ð3:10aÞ
k j X 00j
function of t and r ¼ ð3:5Þ with boundaries
Xj
ou1m
Setting each side of the above equation equal l2j k1 ðt; r0 Þ ¼ aþ ðu1m ðt; r0 Þ þ bm T 1 ðtÞ bm T þ ðtÞÞ
or
gives
ð3:10bÞ
l2j ujm ðt; rj Þ ¼ uðjþ1Þm ðt; rj Þ; j ¼ 1; . . . ; n 1 ð3:10cÞ
X 00j þ Xj ¼ 0 ð3:6aÞ
kj ouðjþ1Þm
oujm
kj ðt; rj Þ ¼ kjþ1 ðt; rj Þ j ¼ 1; . . . ; n 1
The general solution of (3.6a) is then obtained as or or
! ð3:10dÞ
ljm
X jm ¼ Ajm sin pffiffiffiffi x ð3:6bÞ ounm
kj kn ðt; rn Þ ¼ a1 unm ðt; rn Þ ð3:10eÞ
or
Combining the boundaries (3.3f–g), Xjm(0) = 0 and ujm ð0; rÞ ¼ bm T 1 ð0Þ r 2 ½rj1 ; rj ;
Xjm(1) = 0, leads to j ¼ 1; . . . ; n ð3:10fÞ
l pffiffiffiffi
pjmffiffiffiffi ¼ mp or ljm ¼ mp k j and Note that Eq. (3.10f) is obtained by expressing 1 as a
kj
sum of Xm in Eq. (3.3h). It can be observed that the de-
X jm ðxÞ ¼ X m ðxÞ ¼ sinðmpxÞ; m ¼ 1; . . . ; 1 ð3:7Þ rived one-dimensional transient heat equation for r-var-
Solution Uj in Eq. (3.4) can then be expressed as iable exhibits quite different mathematical form than the
original multi-dimensional equation does.
X
1
U j ðt; r; xÞ ¼ ujm ðt; rÞX m ðxÞ
3.4. Closed form solution
m¼1
X
1
¼ ujm ðt; rÞ sinðmpxÞ ð3:8Þ For any jth layer, we introduce the following new
m¼1 variable as
Note that the coefficient Ajm in Eq. (3.6b) is embedded in ix1 bm l2jm bm 2
I 1 ðg1 Þ h0 I 0 ðg1 Þ K 1 ðg1 Þ h0 K 0 ðg1 Þ 0 0 ... 0 0 0 0
I 0 ðn1 Þ K 0 ðn1 Þ I 0 ðg2 Þ K 0 ðg2 Þ ... 0 0 0 0
I 1 ðn1 Þ K 1 ðn1 Þ h1 I 1 ðg2 Þ h1 K 1 ðg2 Þ ... 0 0 0 0
DðsÞ ¼ .. . .. . . .. ... ... . .. .. . ... .. .
0 0 0 0 . . . I 0 ðnn1 Þ K 0 ðnn1 Þ I 0 ðgn Þ K 0 ðgn Þ
0 0 0 0 . . . I 1 ðnn1 Þ K 1 ðnn1 Þ hn1 I 1 ðgn Þ hn1 K 1 ðgn Þ
0 0 0 0 ... 0 0 I 1 ðnn Þ hn I 0 ðnn Þ K 1 ðnn Þ hn K 0 ðnn Þ
ð3:14aÞ
DðsÞ with 3.5. Solution to the second x-boundary condition
row 1 column 2j 1
deleted The second x-boundary condition requires that
D1 ðsÞ ¼ h0 ;
DðsÞ a0 = 1, a1 = 0. Then the boundary condition (2.1f–g)
DðsÞ ð3:14bÞ
with becomes (see equations)
row 2n column 2j 1
T j ðt; r; 0Þ ¼ T 1 ðtÞ; r 2 ½rj1 ; rj ; j ¼ 1; . . . ; n ð3:17aÞ
deleted
D2 ðsÞ ¼ hn oT j
DðsÞ ðt; r; 1Þ ¼ 0; r 2 ½rj1 ; rj ; j ¼ 1; . . . ; n ð3:17bÞ
ox
DðsÞ with
Eq. (3.3) keeps the same except (3.3g) which becomes
row 1 column 2j
deleted
D3 ðsÞ ¼ h0 ; oU j
DðsÞ ðt; r; 1Þ ¼ 0 ð3:18Þ
ð3:14cÞ ox
DðsÞ with
row 2n column 2j
Separating variables results in Eqs. (3.5) and (3.6).
deleted
D4 ðsÞ ¼ hn Boundary condition (3.17a–b) requires that Xj(0) = 0
DðsÞ oX
and oxj ð1Þ ¼ 0. Therefore, Eq. (3.7) is changed as
346 X. Lu et al. / International Journal of Heat and Mass Transfer 49 (2006) 341–348
lj 1 1 pffiffiffiffi
pffiffiffiffi ¼ mþ p or ljm ¼ mþ p kj
kj 2 2
and
1
X jm ðxÞ ¼ X m ðxÞ ¼ sin mþ px ; m ¼ 1; . . . ; 1
2
ð3:19Þ
Using the orthogonal property of Xm to express 1 as a
sum of Xm, we get exactly the same heat conduction
equation as Eq. (3.9) except that ljm and bm are given
in Eq. (3.19) and the following, respectively:
2ð1 cosðmp þ p2ÞÞ
bm ¼ ; m ¼ 1; . . . ; 1 ð3:20Þ
ð2m þ 1Þp
The solution procedure follows exactly the same meth-
odology developed for the first x-boundary condition.
Fig. 2. Schematic picture of the five-layer circular cylinder
3.6. Solution to more generally time-dependent calculated in the example.
boundary condition
Table 1
For more generally time-dependent boundary condi-
Material properties and dimensions of the five-layer circular
tions, presentPthe boundaries as Fourier series as cylinder
T þ ðtÞ ¼ aþ0 þ 1
P k¼1 aþk cosðxþk t þ uþk Þ, T 1 ðtÞ ¼ a10 þ
1 Layer Thermal Thermal Thickness
k¼1 a1k cosðx1k t þ u1k Þ. By linear property, the corre-
conductivity diffusivity (mm)
sponding solution can be expressed as the sum of solu-
(W/m/K) (m2/s)
tions with constant boundary temperatures and with
infinitely sums of cosines. Solution to the constant 1 0.23 4.11 · 107 50
2 0.0337 1.47 · 106 100
boundary temperature is approximated by linearisation
3 0.9 3.75 · 107 100
of hyperbolic functions sinh and cosh in F, G in Eq.
const
4 0.147 1.61 · 107 200
(3.15a) to obtain Rjm ðsÞ const1sþconst2 . Hence 5 0.12 1.5 · 107 20
const const2
Rjm ðtÞ ¼ const1 expð const1 tÞ. If studies do not focus very
much on the initial temperature change, the steady state
solution may also be a good approximation to the con-
stant boundary temperature. The solution to infinitely Table 2
sums of cosines is easily obtained from the previous the- Parameters of Eq. (4.1)
ory due to the linear property. x1 x2 x3 x4
30.0 5.0 2.0 1.0
u+1 u+2 u+3 u+4
4. Calculation example 5.149231 16.77994 0.67884 4.381328
a+0 a+1 a+2 a+3 a+4
A five-layer composite circular cylinder was selected 17.0 1.919486 0.732953 0.25824 0.132831
as the calculation example demonstrated in Fig. 2. Ther- u11 u12 u13 u14
mal properties and dimensions of the slab are given in 5.607506 13.59596 1.451539 5.418717
Table 1. Surface heat transfer coefficients were set as a10 a11 a12 a13 a14
5.0 2.72217 5.019664 1.084058 0.4648
a1 = 25 W/m2/K and a+ = 6 W/m2/K.
The boundary temperatures were taken from mea-
surements and fitted with periodic functions with peri-
Calculated points were made in the central points of
ods 30, 5, 2 and 1 days as
each layer (e.g. represented as layer 3 to layer 4 in the
X4
2pt
T þ ðtÞ ¼ aþ0 þ aþi cos uþi ð4:1aÞ figure). Fig. 4 displays the comparison of the transient
i¼1
xi temperature variation using the analytical and numerical
X 4 methods. The temperatures were stored in files as hourly
2pt
T 1 ðtÞ ¼ a10 þ a1i cos u1i ð4:1bÞ values and shown in figures as hourly and daily values.
xi
i¼1 In the figure, boundary conditions are presented as
where fitting parameters are listed in Table 2 and Fig. 3 T+(t) T1(t) (represented as T+ Tinf in the figure)
shows the values. for convenience. The corresponding temperatures in lay-
X. Lu et al. / International Journal of Heat and Mass Transfer 49 (2006) 341–348 347
Fig. 4. Comparison of analytical and numerical results, layers three and four.
ers are plotted as Tj(t) T1(t). It can be seen that the • Calculation includes only simple computation of
numerical agrees with analytical results. The results for matrix determinant which can be easily accomplished
the first two days are provided in Fig. 5. More calcula- by commercial mathematical packages like Maple,
tion results in other layers did not show any substantial Matlab and Mathematica and even by hands. No
change. Therefore, we only illustrate the results in layer numerical work is necessitated. For any jth layer,
three and four. only five sparse matrices are involved. The calcula-
tion load is small and the computing time is short.
• Compared with numerical methods, the developed
5. Discussions method is easier to complement and a possible insta-
bility in numerical method is avoided. This is espe-
The solution of transient multi-dimensional heat con- cially important for multi-dimensional heat
duction of n-layer circular cylinder is explicitly expressed conduction problem as imaginary eigenvalues may
through Eq. (3.16). We make some observations. exist which cause instability of the numerical program.
348 X. Lu et al. / International Journal of Heat and Mass Transfer 49 (2006) 341–348
Acknowledgement