Professional Documents
Culture Documents
Tw,c = 70F
th = 0.065 inch kt = 0.22 W/m-K
hw = 250 W/m -K
2
Tw,h = 135F
hw = 250 W/m -K
2
w = 1.25 inch
Do = 0.75 inch
Figure P2.1-1: Tubes buried in soil.
a.) Estimate the heat transfer from the hot water to the cold water due to their proximity to one
another.
The heat transfer is resisted by convection within the tube, conduction through the tube, and
conduction in the soil. The convection resistance is calculated according to:
1
Rconv = (1)
hw L ( Do 2 th )
The conduction resistance associated with the tube is calculated according to:
Do
ln
Do 2 th
Rcond = (2)
2 kt L
The shape factor for parallel, buried tubes (SF) is obtained using EES internal library and used
to compute the resistance due to conduction through the soil:
1
Rtubetotube = (3)
SF ks
Tw,h Tw,c
q = (4)
2 Rconv + 2 Rcond + Rtubetotube
b.) To do part (a) you should have needed to determine a shape factor; calculate an approximate
value of the shape factor and compare it to the accepted value.
The shape factor can be thought of as the ratio of the effective area for conduction, Aeff, to the
effective length required for conduction, Leff. The effective area for conduction can be estimated
according to:
Aeff w L (5)
Leff w Do (6)
Aeff
SFapp (7)
Leff
The exact value of the shape factor returned by the function is SF = 17.4 m while the
approximate value is SFapp = 15.2 m; these are sufficiently close for a sanity check.
c.) Plot the rate of heat transfer from the hot water to the cold water as a function of the center to
center distance between the tubes.
The heat transfer rate is shown in Figure 2 as a function of the center to center distance between
the tubes.
Figure 2: Heat transfer rate as a function of the center to center distance between the tubes
Problem 2.1-2
Currently, the low-pressure steam exhausted from a steam turbine at the power plant is condensed
by heat transfer to cooling water. An alternative that has been proposed is to transport the steam via
an underground pipe to a large building complex and use the steam for space heating. You have
been asked to evaluate the feasibility of this proposal. The building complex is located 0.2 miles
from the power plant. The pipe is made of uninsulated PVC (thermal conductivity of 0.19 W/m-K)
with an inner diameter of 8.33 in and wall thickness 0.148 in. The pipe will be buried underground
at a depth of 4 ft in soil that has an estimated thermal conductivity of 0.5 W/m-K. The steam leaves
the power plant at 6.5 lbm/min, 8 psia with a 95% quality. The outdoor temperature is 5F.
Condensate is returned to the power plant in a separate pipe as, approximately, saturated liquid at 8
psia. You may neglect the resistance due to convection to the air and the steam.
a.) Neglecting the inevitable pressure loss, estimate the state of the steam that is provided to the
building complex.
$UnitSytem SI K J Pa
$TabStops 0.2 0.4 3.5 in
"known information"
L=0.2 [mile]*convert(mile,m) "pipe length"
D_i=8.33 [in]*convert(in,m) "pipe inner diameter"
thk=0.148 [in]*convert(in,m) "pipe wall thickness"
k_soil=0.5 [W/m-K] "soil thermal conductivity"
k_PVC=0.19 [W/m-K] "PVC thermal conductivity"
T_air=convertTemp(F,K,5 [F]) "air temperature in K"
P_steam=8 [psia]*convert(psia,Pa) "steam pressure"
T_steam=T_sat(Steam,P=P_steam) "steam temperature in K"
W=4 [ft]*convert(ft,m) "distance of pipe below ground level"
m_dot=6.5 [lb_m/min]*convert(lb_m/min,kg/s) "mass flow rate of the steam"
Do = Di + 2 th (1)
where Di is the inner diameter and th is the pipe thickness. The resistance to conduction through the
pipe is:
D
ln o
R pipe = Di (2)
2 k PVC L
where kPVC is the conductivity of the PVC pipe and L is the length of the pipe. The resistance
between the pipe surface and the surface of the soil is obtained using the shape factor (S) obtained
using the function SF_2 in EES.
1
Rsoil = (3)
S k soil
The heat transfer from the pipe to the soil surface is:
q =
(Tsteam Tair ) (4)
R pipe + Rsoil
where Tsteam is the temperature of the steam and Tair is the temperature of the air.
which leads to q = 107.3 kW. An energy balance on the steam leads to:
where is,in and is,out are the inlet and outlet enthalpies of the steam, respectively. The outlet enthalpy
is used to determine the outlet quality, xs,out:
b.) Are the thermal losses experienced in the underground pipe transport process significant in your
opinion? Do you recommend insulating this pipe?
Yes, the losses are significant. The quality of the steam is reduced from 95% to 67.3% and
therefore you have lost approximately 30% of the heating content of the steam. The pipe should be
insulated.
c.) Provide a sanity check on the shape factor that you used to solve this problem.
The shape factor represents, approximately, the ratio of the cross-sectional area for conduction to
the length for conduction. A crude estimate of the shape factor for this problem is:
W
Do + L
S app = 2
(6)
W
which leads to Sapp = 687.3 m; this is approximately in line with the value provided by the EES
function, S = 651.8 m.
Problem 2.1-3 (2-2 in text)
A solar electric generation system (SEGS) employs molten salt as both the energy transport and
storage fluid. The molten salt is heated to Tsalt = 500C and stored in a buried semi-spherical
tank. The top (flat) surface of the tank is at ground level. The diameter of the tank before
insulation is applied Dt = 14 m. The outside surfaces of the tank are insulated with tins = 0.30 m
thick fiberglass having a thermal conductivity of kins = 0.035 W/m-K. Sand having a thermal
conductivity of ksand = 0.27 W/m-K surrounds the tank, except on its top surface. Estimate the
rate of heat loss from this storage unit to the Tair = 25C surroundings. You may neglect the
resistance due to convection.
$UnitSytem SI K J Pa
$TabStops 0.2 0.4 3.5 in
"known information"
t_ins=0.3 [m] "insulation thickness"
D_t=14 [m] "tank diameter"
k_sand=0.27 [W/m-K] "sand thermal conductivity"
k_ins=0.035 [W/m-K] "insulation thermal conductivity"
T_air=convertTemp(C,K,25 [C]) "air temperature in K"
T_salt=converttemp(C,K,500 [C]) "salt temperature in K"
The resistance between the bottom (buried, hemispherical) surface of the tank and the air is due
to the conduction through the spherical shell of insulation
1 1
+
Dt / 2 ( Dt / 2 + tins )
Rins ,bottom = (1)
2 kins
1
Rsand = (2)
k sand S
where S is the shape factor, obtained using the EES function SF_22. The resistance from the top,
unburied surface of the tank is only due to conduction through the insulation:
tins
Rins ,top = (3)
D2
kins t
4
The heat transfer from the bottom and top surfaces are calculated according to:
qbottom =
(Tsalt Tair ) (4)
Rsoil + Rins ,bottom
and
qtop =
(Tsalt Tair ) (5)
Rins ,top
R_ins_bottom=(1/(D_t/2)-1/(D_t/2+t_ins))/(2*pi*k_ins)
"resistance to conduction through insulation on bottom"
S=SF_22(D_t+2*t_ins) "shape factor"
R_sand=1/(k_sand*S) "resistance of sand"
R_ins_top=t_ins/(k_ins*pi*D_t^2/4)
"resistance to conduction through insulation on top"
q_dot_bottom=(T_salt-T_air)/(R_sand+R_ins_bottom) "heat loss from bottom surface of tank"
q_dot_top=(T_salt-T_air)/(R_ins_top) "heat loss from top surface of tank"
q_dot_total=q_dot_bottom+q_dot_top "total heat loss"
L=1 [m]
D=1 [cm]*convert(cm,m) "diameter of hole"
W=3 [cm]*convert(cm,m) "width of extrusion"
k=0.5 [W/m-K] "conductivity"
h_bar_a=50 [W/m^2-K] "heat transfer coefficient on air-side"
T_a=converttemp(C,K,20 [C]) "air temperature"
h_bar_w=150 [W/m^2-K] "heat transfer coefficient on water-side"
T_w=converttemp(C,K,80 [C]) "water temperature"
1
Rconv ,a = (1)
ha 4W L
1
Rconv , w = (2)
hw D L
The shape factor between the inner and outer surfaces of the extrusion (SF) is obtained from the
SF_12 function in EES. The resistance to conduction is:
1
Rcond = (3)
SF k
q =
(Tw Ta ) (4)
Rconv ,a + Rcond + Rconv , w
SF=SF_12(D,W,L) "shape factor"
R_cond=1/(SF*k) "conduction resistance"
q_dot=(T_w-T_a)/(R_conv_a+R_cond+R_conv_w) "heat transfer rate"
b.) Carry out a sanity check on the value of the shape factor that you used in (a).
The shape factor can be thought of as the ratio of the effective area to the effective length for
conduction. The area for conduction is smallest at the inner surface and largest at the outer
surface. The approximate area can be taken to be the average of these two:
D L + 4W L
Aapp = (5)
2
The effective length is largest from the surface of the hole to the corner and smallest from the
surface of the hole to the middle of the edge. The average of these values is used:
Lapp =
1 (W D )
+
( )
2W D
(6)
2 2 2
Aapp
SFapp = (7)
Lapp
which leads to SFapp = 5.78 m which compares well to the shape factor obtained using the EES
library, SF = 5.35 m.
$UnitSystem SI MASS RAD PA K J
$Tabstops 0.2 0.4 0.6 3.5 in
"Inputs"
th=1.5 [mm]*convert(mm,m) "thickness"
k=75 [W/m-K] "conductivity of collector plate"
e=1 [-] "emissivity of collector plate"
q``_s=900 [W/m^2] "solar flux"
T_infinity=converttemp(C,K,10[C]) "ambient temperature"
h_bar=5 [W/m^2-K] "heat transfer coefficient"
L=8 [cm]*convert(cm,m) "half-width between tubes"
T_w=converttemp(C,K,50 [C]) "water temperature"
W=1 [m] "per unit length of collector"
Nodes are placed along the length of the collector. Only the region from 0 < x < L is considered
due to the symmetry of the system. Therefore, the position of each node is:
L ( i 1)
xi = for i = 1..N (8)
( N 1)
The distance between adjacent nodes is:
L
x = (9)
( N 1)
N=21 [-] "number of nodes"
Dx=L/(N-1)
duplicate i=1,N
x[i]=L*(i-1)/(N-1) "location of each node"
end
where
k W th
q LHS ,i = (Ti 1 Ti ) (11)
x
k W th
q RHS ,i = (Ti +1 Ti ) (12)
x
qs ,i = W x q s (13)
qs ,i = W x h (T Ti ) (14)
"internal nodes"
duplicate i=2,(N-1)
q_dot_LHS[i]=k*W*th*(T[i-1]-T[i])/Dx "conduction from left hand side node"
q_dot_RHS[i]=k*W*th*(T[i+1]-T[i])/Dx "conduction from right hand side node"
q_dot_s[i]=W*Dx*q``_s "absorbed solar radiation"
q_dot_conv[i]=W*Dx*h_bar*(T_infinity-T[i]) "convection"
q_dot_rad[i]=W*Dx*e*sigma#*(T_infinity^4-T[i]^4) "radiation"
q_dot_LHS[i]+q_dot_RHS[i]+q_dot_s[i]+q_dot_conv[i]+q_dot_rad[i]=0 "energy balance"
end
The temperature of node 1 is assumed to be equal to the water temperature (neglecting any
resistance to convection on the water-side):
T1 = Tw (16)
where
k W th
q LHS , N = (TN 1 TN ) (18)
x
W x
qs , N = q s (19)
2
W x
qs , N = h (T TN ) (20)
2
W x
qrad , N = (T4 TN4 ) (21)
2
"node 1"
T[1]=T_w
"node N"
q_dot_LHS[N]=k*W*th*(T[N-1]-T[N])/Dx "conduction from left hand side node"
q_dot_s[N]=W*Dx*q``_s/2 "absorbed solar radiation"
q_dot_conv[N]=W*Dx*h_bar*(T_infinity-T[N])/2 "convection"
q_dot_rad[N]=W*Dx*e*sigma#*(T_infinity^4-T[N]^4)/2 "radiation"
q_dot_LHS[N]+q_dot_s[N]+q_dot_conv[N]+q_dot_rad[N]=0 "energy balance"
331
Temperature (K)
329
327
325
323
0 0.02 0.04 0.06 0.08 0.1
Position (m)
Figure 2: Temperature as a function of position in the collector.
An energy balance on node 1 provides the rate of energy transfer to the water:
where
k W th
q RHS ,1 = (T2 T1 ) (23)
x
W x
qs ,1 = q s (24)
2
W x
qs ,1 = h (T T1 ) (25)
2
W x
qrad ,1 = (T4 T14 ) (26)
2
"node 1"
q_dot_RHS[1]=k*W*th*(T[2]-T[1])/Dx "conduction from right hand side node"
q_dot_s[1]=W*Dx*q``_s/2 "absorbed solar radiation"
q_dot_conv[1]=W*Dx*h_bar*(T_infinity-T[1])/2 "convection"
q_dot_rad[1]=W*Dx*e*sigma#*(T_infinity^4-T[1]^4)/2 "radiation"
q_dot_water=q_dot_RHS[1]+q_dot_s[1]+q_dot_conv[1]+q_dot_rad[1] "energy balance"
b.) Determine the efficiency of the collector; efficiency is defined as the ratio of the energy
delivered to the water to the solar energy incident on the collector.
qwater
= (27)
LW
eta=q_dot_water/(L*W*q``_s) "efficiency"
c.) Plot the efficiency as a function of the number of nodes used in the solution.
Figure 3 illustrates the efficiency as a function of the number of nodes and shows that at least
20-30 nodes are required for numerical convergence.
0.409
0.408
0.407
0.406
Efficiency (-)
0.405
0.404
0.403
0.402
0.401
2 10 100 225
Number of nodes
Figure 3: Efficiency as a function of the number of nodes.
0.9
0.8
0.7
0.6
Efficiency
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80
Water to ambient temperature difference (K)
Figure 4: Solar collector efficiency as a function of the water-to-ambient temperature difference.
Problem 2.1-5
A pipe carrying water for a ground source heat pump is buried horizontally in soil with
conductivity k = 0.4 W/m-K. The center of the pipe is W = 6 ft below the surface of the ground.
The pipe has inner diameter Di = 1.5 inch and outer diameter Do = 2 inch. The pipe is made of
material with conductivity kp = 1.5 W/m-K. The water flowing through the pipe has temperature
Tw = 35F with heat transfer coefficient hw = 200 W/m-K. The temperature of the surface of the
soil is Ts = 0F.
a.) Determine the rate of heat transfer between the water and the air per unit length of pipe.
L=1 [m]
D_i=1.5 [inch]*convert(inch,m) "inner diameter of tube"
D_o=2.0 [inch]*convert(inch,m) "outer diameter of tube"
k_p=1.5 [W/m-K] "conductivity of plastic"
k=0.4 [W/m-K] "conductivity of soil"
W=6 [ft]*convert(ft,m) "depth of tube beneath soil"
T_s=converttemp(F,K,0 [F]) "temperature of soil surface"
T_w=converttemp(F,K,35 [F]) "water temperature"
h_bar_w=200 [W/m^2-K] "convection coefficient with water"
1
Rconv , w = (1)
hw Di L
D
ln o
= i
D
Rcond , p (2)
2 k p L
The shape factor between the outer surface of the pipe and the surface of the soil (SF) is
obtained from the SF_2 function in EES. The resistance to conduction through the soil is:
1
Rcond , s = (3)
SF k
b.) Plot the heat transfer as a function of the depth of the pipe.
Figure 1 illustrates the rate of heat transfer as a function of the depth of the pipe.
17
16
15
Rate of heat transfer (W)
14
13
12
11
10
7
0 1 2 3 4 5 6 7 8 9 10 11
Depth of pipe (m)
Figure 1: Rate of heat transfer as a function of the depth of the pipe (per unit length).
c.) Carry out a sanity check on the value of the shape factor that you used in (a).
The shape factor can be thought of as the ratio of the effective area to the effective length for
conduction. Assuming that the heat spreads at an angle of approximately 45 as it moves from
the pipe to the surface, the approximate area for conduction is:
Aapp = 2 W L (5)
The effective length is the average of the length from the pipe to the surface perpendicularly and
at 45.
2W +W
Lapp = (6)
2
which leads to SFapp = 1.17 m which compares well to the shape factor obtained using the EES
library, SF = 1.26 m.
Problem 2.2-1: Model of Welding Process
You are evaluating a technique for controlling the properties of welded joints by using
aggressive liquid cooling. Figure P2.2-1 illustrates a cut-away view of two plates that are being
welded together. Both edges of the plate are clamped and effectively held at temperatures Ts =
25C. The top of the plate is exposed to a heat flux that varies with position x, measured from
joint, according to: qm ( x ) = q j exp ( x / L j ) where q j =1x106 W/m2 is the maximum heat flux
(at the joint, x = 0) and Lj = 2.0 cm is a measure of the extent of the heat flux. The back side of
the plates are exposed to aggressive liquid cooling by a jet of fluid at Tf = -35C with h = 5000
W/m2-K. A half-symmetry model of the problem is shown in Figure P2.2-1. The thickness of
the plate is b = 3.5 cm and the width of a single plate is W = 8.5 cm. You may assume that the
welding process is steady-state and 2-D. You may neglect convection from the top of the plate.
The conductivity of the plate material is k = 38 W/m-K.
both edges are
clamped and held heat flux
at fixed temperature
joint
qm
k = 38 W/m-K
W = 8.5 cm
y b = 3.5 cm Ts = 25C
T f = 35C
h = 500 W/m -K
2
Figure P2-2-1: Welding process and half-symmetry model of the welding process.
a.) Develop a separation of variables solution to the problem. Implement the solution in EES
and prepare a plot of the temperature as a function of x at y = 0, 1.0, 2.0, 3.0, and 3.5 cm.
The problem shown in Figure P2.2-1 is governed by the partial differential equation:
2T 2T
+ =0 (1)
x 2 y 2
T
=0 (2)
x x =0
Tx =W = Ts (3)
T
k = h (Ty =0 T f ) (4)
y y =0
T
k = qm (5)
y y =b
Notice that the problem has 3 non-homogeneous boundary conditions and cannot, as it is written,
be solved using separation of variables. However, the simple transformation:
= T Ts (6)
2 2
+ =0 (7)
x 2 y 2
d
=0 (8)
dx x =0
x =W = 0 (9)
k = h ( y =0 f ) (10)
y y =0
k = qm (11)
y y =b
where
f = T f Ts (12)
T = TX TY (13)
and substituted into Eq. (7); the process leads to two ordinary differential equations:
d 2TX
2
+ 2 TX = 0 (14)
dx
d 2TY
2
2 TY = 0 (15)
dy
We will solve the eigenproblem first (i.e., the problem in the homogeneous direction). The
boundary condition at x = 0, Eq. (8), leads to:
dTX
= C1 cos ( x ) C2 sin ( x ) x =0 = C1 = 0 (18)
dx x =0
TX = C2 cos ( x ) (19)
TX x =W = C2 cos ( W ) = 0 (20)
i =
(1 + 2 i ) for i = 1.. (21)
2W
TX = C2,i cos ( i x ) (23)
i =0
and the solution for the temperature is:
TX = C2,i cos ( i x ) C3,i sinh ( i y ) + C4,i cosh ( i y ) (24)
i =0
= cos ( i x ) C3,i sinh ( i y ) + C4,i cosh ( i y ) (25)
i =0
or
k cos ( i x ) C3,i i = h cos ( i x ) C4,i f (27)
i =0 i =0
W
W W
k C3,i i cos 2 ( i x ) dx = h C4,i cos 2 ( i x ) dx f cos ( x ) dx
i (28)
0 0 0
> restart;
> assume(i,integer);
> lambda:=(1+2*i)*Pi/(2*W);
( 1 + 2 i~ )
:=
2W
> int((cos(lambda*x))^2,x=0..W);
W
2
> int(cos(lambda*x),x=0..W);
2 ( -1 ) i~ W
( 1 + 2 i~ )
which leads to:
k C3,i i W C4,i W 2 ( 1) W
i
= h f (29)
2 2 (1 + 2 i )
or
4 ( 1)
i
k
f = C4,i i C3,i (30)
(1 + 2 i ) h
which is one equation for each of the two pairs of unknown coefficients.
Equations (25) and Error! Reference source not found. are substituted into the final boundary
condition at y = b, Eq. (11):
x
k cos ( i x ) C3,i i cosh ( i b ) + C4,i i sinh ( i b ) = q j exp
(31)
Lj
i =0
W W
x
k C3,i i cosh ( i b ) + C4,i i sinh ( i b ) cos 2 ( i x ) dx = q j exp cos ( i x ) dx (32)
Lj
0 0
or
2 q
W
x
C3,i i cosh ( i b ) + C4,i i sinh ( i b ) = exp L cos ( i x ) dx (33)
Wkj 0 j
which is the 2nd relationship between the coefficient pairs. The integral on the right hand side of
Eq. (32) is accomplished using Maple:
> int(exp(-x/L_j)*cos(lambda*x),x=0..W);
L_j
W W
+ L_j ( -1 ) + 2 L_j ( -1 ) i~ e
L_j i~ i~
2 L_j W 2 W e
4 W2 + 2 L_j2 + 4 2 L_j2 i~ + 4 2 L_j2 i~ 2
"Inputs"
b=3.5 [cm]*convert(cm,m) "plate thickness"
W=8.5 [cm]*convert(cm,m) "plate width"
k=38 [W/m-K] "plate conductivity"
T_s=converttemp(C,K,25) "edge temperature"
T_f=converttemp(C,K,-35) "fluid temperature"
h=5000 [W/m^2-K] "heat transfer coefficient"
q``_j=1e6 [W/m^2] "heat flux at x=0"
L_j=2.0 [cm]*convert(cm,m) "extent of the heat flux"
"Position"
x_bar=0.5
x=x_bar*W
y_bar=0.5
y=y_bar*b
Note that the bold portion of the code was pasted from Maple.
Figure 2 illustrates the temperature as a function of dimensionless axial position (x/W) for
various values of y.
Figure 2: Temperature as a function of x/W for various values of y.
T1
a
T2
k
y
b
x
q
th
h ,Tf
h ,T f
Figure P.2.2-2: Thin plate exposed to air.
a.) The temperature distribution within the plate can be considered 2-D (i.e., temperature
variations in the z-direction can be neglected) if the plate is thin and conductive. How would
you determine if this approximation is valid?
The Biot number characterizes the ratio of conduction in the z direction (which I want to neglect)
to convection from the plate surface (which I will consider):
Rcond , z th h a b th h
Bi = = = (1)
Rconv 2k ab 1 2k
b.) Derive the partial differential equation and boundary conditions that would need to be solved
in order to obtain an analytical solution to this problem.
A differential control volume has extend dx x dy and extends through the thickness of the plate,
as shown in Figure 2.
Figure 2: Differential control volume used to derive the governing partial differential equation.
q x + q y = q x + dx + q y + dy + qconv (2)
or
q x q y
0= dx + dy + qconv (3)
x y
Substituting the rate equations into the energy balance leads to:
T T
0= k th dy dx + k th dx dy + 2 dx dy h (T T f ) (4)
x x y y
or
2T 2T 2 h
+
x 2 y 2 k th
(T T f ) = 0 (5)
Tx = a = T2 (6)
Ty =b = T1 (7)
The boundary condition at the left edge is obtained from an interface balance which requires that
the conduction heat flux in the x-direction at this insulated edge be zero:
q x , x =0 = 0 (8)
or
T
k =0 (9)
x x =0
T
=0 (10)
x x =0
The boundary condition at the bottom edge is also obtained from an interface energy balance:
T
q = k (11)
y y =0
or
T q
= (12)
y y =0
k
Problem 2.2-3 (2-3 in text): Heat Transfer Coefficient Measurement Device
You are the engineer responsible for a simple device that is used to measure heat transfer
coefficient as a function of position within a tank of liquid (Figure P2.2-3). The heat transfer
coefficient can be correlated against vapor quality, fluid composition, and other useful quantities.
The measurement device is composed of many thin plates of low conductivity material that are
interspersed with large, copper interconnects. Heater bars run along both edges of the thin
plates. The heater bars are insulated and can only transfer energy to the plate; the heater bars are
conductive and can therefore be assumed to come to a uniform temperature as a current is
applied. This uniform temperature is assumed to be applied to the top and bottom edges of the
plates. The copper interconnects are thermally well-connected to the fluid; therefore, the
temperature of the left and right edges of each plate are equal to the fluid temperature. This is
convenient because it isolates the effect of adjacent plates from one another which allows each
plate to measure the local heat transfer coefficient. Both surfaces of the plate are exposed to the
fluid temperature via a heat transfer coefficient. It is possible to infer the heat transfer
coefficient by measuring heat transfer required to elevate the heater bar temperature a specified
temperature above the fluid temperature.
a = 20 mm
b = 15 mm
The nominal design of an individual heater plate utilizes metal with k = 20 W/m-K, th = 0.5 mm,
a = 20 mm, and b = 15 mm (note that a and b are defined as the half-width and half-height of the
heater plate, respectively, and th is the thickness as shown in Figure P2-3). The heater bar
temperature is maintained at Th = 40C and the fluid temperature is T = 20C. The nominal
value of the average heat transfer coefficient is h = 50 W/m2-K.
a.) Develop an analytical model that can predict the temperature distribution in the plate under
these nominal conditions.
The problem can be simplified and tackled using the quarter-symmetry model of a single plate,
shown in Figure P2.2-3, is specified mathematically in Figure 2(a).
(a)
(b)
Figure 2: Quarter symmetry model of a plate.
The lines of symmetry (x =0 and y = 0) are adiabatic, the top edge (y = b) is held at the heater
temperature (Th) and the right edge (x = a) is held at the fluid temperature (T). These boundary
conditions are expressed below:
T
=0
x x =0
Tx =a = T
T
=0
y y =0
Ty =b = Th
The differential control volume and associated first law balance are shown in Figure 2(a):
q x + q y = q x + dx + q y + dy + qconv
q x q y
0= dx + dy + qconv
x y
Substituting the rate equations into the energy balance leads to:
T T
0= k th dy dx + k th dx dy + 2 dx dy h ( T T )
x
x y y
or
2T 2T 2 h
+
x 2 y 2 k th
T T = 0 ( ) (1)
= T T (2)
Substituting Eq. (2) into Eq. (1) transforms the governing partial differential equation into a
homogeneous equation; i.e., any multiple of will satisfy transformed partial differential
equation:
2 2
+ 2 m2 = 0 (3)
x 2
y
where
2h
m=
k th
Substituting Eq. (2) into the boundary conditions leads to:
=0 (4)
x x =0
x =a = 0 (5)
=0 (6)
y y =0
y =b = h (7)
where
h = Th T
The transformed problem is shown in Figure 2(b); note that the transformed problem is linear
and consists of a homogeneous partial differential equation and homogeneous boundary
conditions in the x-direction. Therefore, we can apply a separation of variables solution to the
problem and x is our homogeneous direction.
( x, y ) = X ( x ) Y ( y )
Substituting the product solution into the differential equation leads to:
d 2 X d 2 Y
Y + X m2 X Y = 0
dx 2
dy 2
d 2 X d Y
2
dx 2 + dy m 2 = 0
2
X Y
2 2
Recall that x is our homogeneous direction; therefore, we need the X group in the equation to
equal a negative constant (-2); the remaining part of the equation must equal the positive value
of the same constant (2). Therefore, the two ordinary differential equations that result from the
separation process are:
d 2 X
+ 2 X = 0 (8)
dx 2
and
d 2 Y
( 2 + m2 )Y = 0 (9)
dy 2
The next step is to solve the eigenproblem; the solution to the ordinary differential equation for
X is:
The 1st boundary conditions in the homogeneous direction, Eq. (4), leads to:
d X
=0 (11)
dx x =0
d X
= C1 sin ( 0 ) + C2 cos ( 0 ) = 0
dx x =0
=0 =1
which can only be true if C2 = 0. the 2nd boundary condition in the x-direction, Eq. (5), leads to:
x =a = C1 cos ( a ) = 0
which can only be true if the argument of the cosine is 0. Therefore, the eigenfunctions and
eigenvalues for the problem are:
The solution to the ordinary differential equation in the non-homogeneous direction, Eq. (9) can
be obtained using Maple:
> restart;
> assume((lambda^2+m^2),positive);
> ODEy:=diff(diff(thetaY(y),y),y)-(lambda^2+m^2)*thetaY(y)=0;
d
2
ODEy := 2 thetaY( y ) ( 2 + m~ 2 ) thetaY( y ) = 0
dy
> qYs:=dsolve(ODEy);
( 2 + m~2 y ) ( 2 + m~2 y )
qYs := thetaY( y ) = _C1 e + _C2 e
Note that Maple can convert this exponential form to an equivalent form involving hyperbolic
sines and cosines with the convert command.
> convert(qYs,'trigh');
thetaY( y ) = ( _C1 + _C2 ) cosh( 2 + m~ 2 y ) + ( _C1 _C2 ) sinh( 2 + m~ 2 y )
Yi = C3,i cosh ( )
i2 + m 2 y + C4,i sinh ( i2 + m 2 y )
The solution associated with the ith eigenvalue is therefore:
where the constant C1,i is absorbed into the constants C3,i and C4,i. It is good practice to verify
that Eq. (12) satisfies both boundary conditions in the x-direction and the partial differential
equation for any value of i:
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*a);
( 2 i~ 1 )
:=
2a
> theta:=(x,y)->cos(lambda*x)*(C_3*cosh(sqrt(lambda^2+m^2)*y)+C_4*sinh(sqrt(lambda^2+m^2)*y));
:= ( x, y ) cos( x ) ( C_3 cosh( 2 + m 2 y ) + C_4 sinh( 2 + m 2 y ) )
> eval(diff(theta(x,y),x),x=0);
0
> theta(a,y);
0
> diff(diff(theta(x,y),x),x)+diff(diff(theta(x,y),y),y)-m^2*theta(x,y);
( 2 i~ 1 ) x
cos
1
( 2 i~ 1 )
2 2
4 2a
( 2 i~ 1 ) 2 2 ( 2 i~ 1 ) 2 2
+ 4 m 2 y + 4 m 2 y
a 2
a 2
C_3 cosh + C_4 sinh
2 2
a 2 + cos
( 2 i~ 1 ) x
2a
( 2 i~ 1 ) 2 2
+ 4 m 2 y
1 a 2
( 2 i~ 1 ) 2 2
C_3 cosh + 4 m 2
4 2 a 2
( 2 i~ 1 )
2 2
+ m 2
y
4
1 a2 ( 2 i~ 1 ) 2 2
+ C_4 sinh
+ 4 m m 2
2
a
4 2 2
i~ x
cos
( 2 1 )
2 a
( 2 i~ 1 )2 2 ( 2 i~ 1 ) 2 2
+ m 2
y + 4 m 2 y
4
a 2
a 2
C_3 cosh + C_4 sinh
2 2
> simplify(%);
0
The general solution is the sum of the solution for each eigenvalue:
( ) ( )
= i = cos ( i x ) C3,i cosh i2 + m 2 y + C4,i sinh i2 + m 2 y (13)
i =1 i =1
The general solution must satisfy the boundary conditions in the non-homogeneous direction.
Substituting Eq. (13) into Eq. (6) leads to:
( ) ( )
= cos ( i x ) C3,i i + m sinh i + m 0 + C4,i i + m cosh i + m 0 = 0
2 2 2 2 2 2 2 2
y i =1
y =0
=0 =1
or
C
i =1
4,i i2 + m 2 cos ( i x ) = 0
( )
= Ci cosh i2 + m 2 y cos ( i x ) (14)
i =1
where the subscript 3 has been removed from C3,i because it is the only remaining undetermined
constant. Substituting Eq. (14) into Eq. (7) leads to:
( )
y =b = Ci cosh i2 + m 2 b cos ( i x ) = h
i =1
( )
a a
Orthogonality causes all of the terms in the summation to integrate to zero except for the one in
which i = j:
( ) cos ( x ) dx = cos ( x ) dx
a a
Ci cosh +m b
i
2 2 2
i h i
0 0
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*a);
( 2 i~ 1 )
:=
2a
> int((cos(lambda*x))^2,x=0..a);
a
2
> int(cos(lambda*x),x=0..a);
( 1 + i~ )
2 ( -1 ) a
( 2 i~ 1 )
h 4 ( 1)
1+ i
Ci =
cosh ( )
i2 + m 2 b ( 2 i 1)
"Inputs"
th_mm=0.5 [mm] "thickness of plate in mm"
th=th_mm * convert(mm,m) "thickness of plate"
k=20 [W/m-K] "conductivity of plate"
a_mm=20 [mm] "half-width of plate in mm"
a= a_mm * convert(mm,m) "half-width of plate"
b= 15 [mm] *convert(mm,m) "half-height of plate"
T_h = converttemp(C,K,40 [C]) "heater temperature"
T_infinity=converttemp(C,K,20 [C]) "fluid temperature"
h_bar=50 [W/m^2-K] "heat transfer coefficient"
"position"
x_bar=0.5
y_bar=0.5
x=x_bar*a
y=y_bar*b
A parametric table is created and used to generate the contour plot shown in Figure 3.
Figure 3: Contour plot of temperature on the plate.
b.) The measured quantity is the rate of heat transfer to the plate from the heater ( qh ) and
therefore the relationship between qh and h (the quantity that is inferred from the heater
power) determines how useful the instrument is. Determine the heater power.
The heater power can be computed by integrating the conduction heat transfer along the top
surface according to:
a
qh = 4 k th dx (15)
0
y y =b
where the factor of 4 comes from the quarter symmetry of the model. Equation (14) is
substituted into Eq. (15):
( ) cos ( x ) dx
a
qh = 4 k th Ci i2 + m 2 sinh i2 + m 2 b i
i =1 0
( )
qh = 2 k th a Ci i2 + m 2 sinh i2 + m 2 b (16)
i =1
"heater power"
duplicate i=1,N
q_dot_h[i]=2*k*th*a*C[i]*sqrt(lambda[i]^2+m^2)*sinh(sqrt(lambda[i]^2+m^2)*b)
end
q_dot_h=sum(q_dot_h[1..N])
c.) If the uncertainty in the measurement of the heater power is qh = 0.01 W, estimate the
uncertainty in the measured heat transfer coefficient ( h ).
A parametric table is used to explore the relationship between h and qhtr ; the two columns in the
table are the variables h_bar and q_dot_h. Figure 4 illustrates the heat transfer coefficient as a
function of the heater power, all else held constant.
Note that a good measuring device would show a strong relationship between the physical
quantity being measured ( h ) and the measurement ( qh ); this would be indicated by a large
partial derivative of heat transfer coefficient with respect to heater power. Given the uncertainty
in the heater power measurement, qh (related to, for example, the resolution of the data
acquisition system, noise, etc.), it is possible to estimate the uncertainty in the measurement of
the heat transfer coefficient, h , according to:
h
h = qh
qh
For example, if the uncertainty in the heater power is 0.01 W then Fig. 3 suggests that the partial
derivative of heat transfer coefficient with respect to heater power is approximately 125 W/m2-
K-W and the uncertainty in the heat transfer coefficient measurement would be h = 1.8 W/m2-
K.
As an engineer designing this measurement device, you would like to calculate not the heater
power but rather the partial derivative in the heat transfer coefficient with respect to heater power
in order to automate the process of computing h and therefore evaluate how design changes
affect the instrument. This would be difficult to accomplish analytically; note that both m and
the constants Ci in Eq. (16) are functions of h . It is more convenient to determine the partial
derivative numerically. That is, set the heat transfer coefficient at its nominal value plus a small
amount ( h + h ) and evaluate the heater power ( qh , h +h ); note that h should be small relative
to the nominal value of the heat transfer coefficient. Then set the heat transfer coefficient at its
nominal value less a small amount ( h h ) and evaluate the heater power ( qh , h h ). The partial
derivative is approximately:
h 2 h
qh qh ,h +h qh ,h h
Since qh needs to be evaluated at several values of the heat transfer coefficient, it is convenient
to have the computation of the heater power occur within a function that is called twice within
the equation window. Because the solution is in the form of an EES code, this is an ideal
problem to use a MODULE. A MODULE is a stand-alone EES program that can be called from
the main EES equation window. The MODULE is provided with inputs and it calculates
outputs. The protocol of a call to a MODULE involves the name of the MODULE followed by a
series of inputs separated by a colon from a series of outputs. For example, we will create a
MODULE Heaterpower that calculates the value of the variable q_dot_h. It is important to note
that all of the variables from the main equation window that are required by the MODULE must
be passed to the MODULE; the MODULE can only access variables that are passed to it as
parameters or using the $COMMON directive.
The MODULE Heaterpower must be placed at the top of the EES code and is a small self-
contained EES program; we create the MODULE by copying those lines of the main EES code
that are required to calculate the variable q_dot_h.
MODULE Heaterpower(th,k,a,b,T_h,T_infinity,h_bar:q_dot_h)
theta_h=T_h-T_infinity "heater temperature difference"
m=sqrt(2*h_bar/(k*th)) "solution constant"
N=100 "number of terms"
duplicate i=1,N
lambda[i]=(2*i-1)*pi/(2*a) "eigenvalues"
C[i]=theta_h*(4*(-1)^(1+i)/(2*i-1)/Pi)/(cosh(sqrt(lambda[i]^2+m^2)*b))
q_dot_h[i]=2*k*th*a*C[i]*sqrt(lambda[i]^2+m^2)*sinh(sqrt(lambda[i]^2+m^2)*b)
end
q_dot_h=sum(q_dot_h[1..N])
end
The calling protocol for the MODULE consists of a series of inputs (the variables d, k, a, b,
T_htr, T_f, and h) that are separated by a series outputs (in this case only the variable q_dot_htr)
by a colon. MODULES are most useful where a certain sequence of code must be executed
multiple times; in this case, the MODULE Heaterpower enables the partial derivative to be easily
computed.
CALL Q_dot_heater(d,k,a,b,T_htr,T_f,h-dh:q_dot_htr_minus)
CALL Q_dot_heater(d,k,a,b,T_htr,T_f,h+dh:q_dot_htr_plus)
dhdqdot=2*dh/(q_dot_htr_plus-q_dot_htr_minus)
delta_h=dhdqdot*delta_q_dot_htr
The modifications to the EES code verifies that the uncertainty in heat transfer coefficient,
delta_h, is 1.26 W/m2-K and provides a convenient tool for assessing the impact of the various
design parameters on the performance of the measurement system. For example, Figure 5
illustrates the uncertainty in the heat transfer coefficient as a function of the plate thickness (th)
for various values of its half-width (a).
Figure 5: Uncertainty in the measured heat transfer coefficient as a function of the plate thickness
for various values of the plate half-width.
Problem 2.2-4: Local Heat Transfer Coefficient Measurement
Figure P2.2-4(a) illustrates a proposed device to measure the local heat transfer coefficient from
a surface undergoing a boiling heat transfer process. Micro-scale heaters and temperature
sensors are embedded in a substrate in a regularly spaced array, as shown.
evaporating fluid
heaters computational
temperature domain
sensors
Figure P2.2-4(a): An array of micro-scale heaters and temperature sensors embedded in a substrate.
The heaters are activated, producing a heat flux that is removed primarily from the surface of the
substrate exposed to evaporating fluid. The temperature sensors are embedded in sets of two
located very near the surface. Each set of thermocouples are used to infer the local heat flux to
the surface ( qs ) and the surface temperature (Ts); these quantities are sufficient to measure the
heat transfer coefficient. A half-symmetry model of the region of the substrate between two
adjacent heaters (see Figure 2.2-4(a)) is shown in Figure P2.2-4(b).
T = 20C
y1 = 1.9 mm h = 2500 W/m -K
2
b = 2 mm
W = 5 mm
y
x
y2 = 1.8 mm
k = 3.5 W/m-K
Figure P2.2-4(b): A half-symmetry model of the region of the substrate between two adjacent
heaters.
The thickness of the substrate is b = 2 mm and the half-width between adjacent heaters is W = 5
mm. The substrate has conductivity k = 3.5 W/m-K and you may assume that the presence of the
temperature sensors does not affect the temperature distribution. The heat flux exposed to the
computational domain at x = W is qh = 1x105 W/m2. The heat transfer coefficient between the
evaporating fluid at T = 20C and the surface is h = 2500 W/m2-K.
a.) Develop a separation of variables solution based on the computational domain shown in
Figure P2.2-4(b). Implement your solution in EES.
"Inputs"
k=3.5 [W/m-K] "conductivity"
T_infinity=converttemp(C,K,20 [C]) "fluid temperature"
h=2500 [W/m^2-K] "heat transfer coefficient"
W=5 [mm]*convert(mm,m) "half-width"
b=2 [mm]*convert(mm,m) "substrate thickness"
y_1=1.9 [mm]*convert(mm,m) "temperature sensor #1 position"
y_2=1.8 [mm]*convert(mm,m) "temperature sensor #2 position"
q_h=1e5 [W/m^2] "edge heat flux"
The problem shown in Figure P2.2-4(b) is governed by the partial differential equation:
2T 2T
+ =0 (1)
x 2 y 2
T
=0 (2)
x x =0
T
k = qh (3)
x x =W
T
=0 (4)
y y =0
T
k = h (Ty =b T ) (5)
y y =b
= T T (6)
2 2
+ =0 (7)
x 2 y 2
k = qh (9)
x x =W
=0 (10)
y y =0
k = h y =b (11)
y y =b
= X Y (12)
and substituted into Eq. (7); the process leads to two ordinary differential equations:
d 2 X
2 X = 0 (13)
dx 2
d 2 Y
+ 2 Y = 0 (14)
dy 2
Notice that the sign of the constant is selected so that the solution in the homogeneous direction
(y) is sines and cosines. The solution to Eq. (14) is:
d Y
= C1 cos ( 0 ) C2 sin ( 0 ) = 0 (16)
dy y =0
Y = C2 cos ( y ) (17)
hb
tan ( i b ) = (19)
k ( i b )
The roots of Eq. (19) occur in regular intervals of the argument of the tangent function. A
residual is defined by rearranging Eq. (19):
hb
Res = tan ( i b ) (20)
k ( i b )
Figure P2.2-4-2 illustrates the residual defined in Eq. (20); the eigenvalues correspond to the
points where the residual is zero.
1
0.75
0.5
0.25
Residual
-0.25
-0.5
-0.75
eigenvalues
-1
0 1.57 3.14 4.71 6.28 7.85 9.42 10.99 12.56 14.13 15.7
b
Figure P2.2-4-2: Residual as a function of b.
The eigenvalues lie in regular intervals of b, as shown in Figure 2.2-4-2 and are identified
automatically in EES by defining arrays and using them as the lower limit, upper limit, and guess
values for the entries in the eigenvalue array in the Variable Information window (Figure P2.2-4-
3).
The sum of these solutions is, itself, a solution (note that the constant C2,i is absorbed into the
other constants):
= i = cos ( i y ) C3,i sinh ( i x ) + C4,i cosh ( i x ) (24)
i =1 i =1
= cos ( i y ) C3,i i cosh ( i 0 ) + C4,i i sinh ( i 0 ) (25)
x x =0 i =1
or
cos ( y ) C
i =1
i 3,i i = 0 (26)
= Ci cos ( i y ) cosh ( i x ) (27)
i =1
where the constant C4,i is referred to simply as Ci since it is the only remaining undetermined
constant associated with each eigenvalue. Equation (27) is substituted into the boundary
condition at x = W, Eq. (9):
k Ci cos ( i y ) i cosh ( i W ) = qh (28)
i =1
The eigenfunctions are orthogonal; therefore, Eq. (28) is multiplied by an arbitary eigenfunction
and integrated from y = 0 to y = b. The result is:
b b
Ci k i cosh ( i W ) cos ( y ) dy = q cos ( y ) dy
2
i i (29)
y =0 y =0
> restart;
> int((cos(lambda*y))^2,y=0..b);
1 cos( b ) sin( b ) + b
2
> int(cos(lambda*y),y=0..b);
sin( b )
sin ( i b )
Ci = q (30)
k cosh ( i W ) Integrali
i
2
where
cos ( i b ) sin ( i b ) + i b
Integrali = (31)
2 i
duplicate i=1,N
Integral[i]=(cos(lambdab[i])*sin(lambdab[i])+lambdab[i])/(2*lambda[i])
C[i]=q_h*sin(lambdab[i])/(k*lambda[i]^2*cosh(lambdab[i])*Integral[i])
end
x_bar=0.5 [-]
y_bar=0.5 [-]
x=x_bar*W
y=y_bar*b
duplicate i=1,N
theta[i]=C[i]*cos(lambda[i]*y)*cosh(lambda[i]*x)
end
T=T_infinity+sum(theta[i],i=1,N)
T_C=converttemp(K,C,T)
0.5 61.54
44.08
0.4 32.45
0.3 38.27 67.36
0.2 73.17
0.1
78.99
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/W
Figure P2.2-4-4: Contour plot of temperature.
The position of temperature sensors #1 and #2 at a particular value of x are y1 = 1.9 mm and y2 =
1.8 mm, respectively (see Figure P2.2-4(b)). The surface temperature measurement extracted
from these measured temperatures is associated with a linear extrapolation to the surface at y = 0:
Ts ,m = T2 + (T1 T2 )
( b y2 ) (32)
( y1 y2 )
The heat flux measurement extracted from these measured temperatures is obtained from
Fourier's law according to:
qs,m = k
(T2 T1 ) (33)
( y1 y2 )
c.) What is the heat transfer coefficient measured by the device at x/W = 0.5? That is, based on
the temperatures T1 and T2 predicted by your model at x/W = 0.5, calculate the measured heat
transfer coefficient according to:
qs,m
hm = (34)
(Ts ,m T )
and determine the discrepancy of your measurement relative to the actual heat transfer
coefficient.
The temperatures measured by the sensor set at x/W = 0.5 are evaluated:
duplicate i=1,N
theta_1[i]=C[i]*cos(lambda[i]*y_1)*cosh(lambda[i]*x)
theta_2[i]=C[i]*cos(lambda[i]*y_2)*cosh(lambda[i]*x)
end
T_1=T_infinity+sum(theta_1[i],i=1,N)
T_2=T_infinity+sum(theta_2[i],i=1,N)
The surface temperature, heat flux, and heat transfer coefficient are evaluated according to Eqs.
(32) through (34):
T_s_m=T_2+(T_1-T_2)*(b-y_2)/(y_1-y_2)
q_s_m=k*(T_2-T_1)/(y_1-y_2)
h_m=q_s_m/(T_s_m-T_infinity)
which leads to hm = 2358 W/m2-K. This is 5.6% in error relative to the actual heat transfer
coefficient.
d.) Plot the % error associated with the device configuration (i.e., the discrepancy between the
measured and actual heat transfer coefficient from part (c)) as a function of axial position, x.
deltah=100*abs(h-h_m)/h
18
16
14
12
10
8
6
4
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/W
Figure P2.2-4-5: Error as a function of axial location.
Problem 2.2-5
Three heater blocks provide heat to the back-side of a fin array that must be tested, as shown in
Figure P2.2-5.
k = 30 W/m-K
q2 = 3x10 W/m
4 2
c = 1 cm d = 4 cm
L = 7 cm
y
x h = 20 W/m -K
2
th = 1 cm T = 20C
h f = 35 W/m -K
2
T f = 35C
N fins = 150
As, fin = 8 cm
2
fin = 0.85
Figure P2.2-5: An array of fins installed on a base plate energized by three heater blocks.
The fins are installed on a base plate that has half-width of L= 7 cm, thickness of th = 1 cm, and
width W = 20 cm (into the page). The base plate material has conductivity k = 30 W/m-K. The
edge of the base plate (at x = L) is exposed to air at T = 20C with heat transfer coefficient h =
20 W/m2-K. The middle of the plate (at x = 0) is a line of symmetry and can be modeled as
being adiabatic. The bottom of the plate (at y = 0) has an array of Nfin = 150 fins installed. Each
fin has surface area As,fin= 8 cm2, base area Ab,fin = 0.8 cm2, and efficiency fin = 0.85. The fin
and the base material are exposed to fluid at Tf = 35C with heat transfer coefficient h f = 20
W/m2-K. The top of the plate (at y = th) is exposed to the heat flux from the heater blocks. The
heat flux is distributed according to:
q1 if x < c
q y =th = 0 if c < x < ( d + c )
q if x > ( d + c )
2
"PROBLEM 2.2-5"
$UnitSystem SI MASS RAD PA K J
$TABSTOPS 0.2 0.4 0.6 0.8 3.5 in
"Inputs"
k=30 [W/m-K] "thermal conductivity"
h_bar=20 [W/m^2-K] "heat transfer coefficient to ambient air"
W=20 [cm]*convert(cm,m) "width of base"
L=7 [cm]*convert(cm,m) "half-length of base"
th=1 [cm]*convert(cm,m) "thickness of base"
h_bar_f=35 [W/m^2-K] "heat transfer coefficient to fluid"
A_s_fin=8 [cm^2]*convert(cm^2,m^2) "surface area of each fin"
A_b_fin=0.8 [cm^2]*convert(cm^2,m^2) "base area of each fin"
N_fin=150 [-] "number of fins"
eta_fin=0.85 [-] "fin efficiency"
T_infinity=converttemp(C,K,20[C]) "ambient air temperature"
T_f=converttemp(C,K,35 [C]) "fluid temperature"
c=1 [cm]*convert(cm,m) "width of first heating zone"
d=4 [cm]*convert(cm,m) "space between heating zones"
q``_1=3e4 [W/m^2] "heat flux in 1st heating zone"
q``_2=3e4 [W/m^2] "heat flux in 2nd heating zone"
The effective heat transfer coefficient is defined so that it provides the same rate of heat transfer
as the finned surface:
W*L*h_bar_eff=(W*L-N_fin*A_b_fin)*h_bar_f+eta_fin*N_fin*A_s_fin*h_bar_f
"effective heat transfer coefficient"
b.) Develop a separation of variables solution for the temperature distribution within the fin base
material.
The problem shown in Figure P2.2-5 is governed by the transformed partial differential equation:
2 2
+ =0 (2)
x 2 y 2
=0 (3)
x x =0
k = h x=L (4)
y x=L
q1 if x < c
k = 0 if c < x < ( d + c ) (5)
y y =th
q2 if x > ( d + c )
heff f y =0 = k (6)
y y =0
where
= T T (7)
and
f = T f T (8)
= X Y (9)
and substituted into Eq. (2); the process leads to two ordinary differential equations:
d 2 X
+ 2 X = 0 (10)
dx 2
d 2 Y
2 Y = 0 (11)
dy 2
Notice that the sign of the constant is selected so that the solution in the homogeneous direction
(x) is sines and cosines. The solution to Eq. (10) is:
X = C2 cos ( x ) (13)
The boundary condition at x = L, Eq. (4), leads to the eigencondition for the problem:
hL
tan ( i L ) = (14)
k ( i L )
The eigenvalues lie in regular intervals of L and are identified automatically in EES by defining
arrays and using them as the lower limit, upper limit, and guess values for the entries in the
eigenvalue array in the Variable Information window.
The sum of these solutions is, itself, a solution (note that the constant C2,i is absorbed into the
other constants):
= i = cos ( i x ) C3,i sinh ( i y ) + C4,i cosh ( i y ) (18)
i =1 i =1
heff f C4,i cos ( i x ) = k C3,i i cos ( i x ) (19)
i =1 i =1
or
L L
heff f cos ( i x ) dx = heff C4,i k C3,i i cos 2 ( i x ) dx (21)
0
0
Integral1
The interals in Eq. (21) are evaluated in Maple:
> restart;
> int(cos(lambda[i]*x),x=0..L);
sin( i L )
i
> int((cos(lambda[i]*x))^2,x=0..L);
1 cos( i L ) sin( i L ) + i L
2 i
The integrals are put into EES, leading to a relationship between C3,i and C4,i :
duplicate i=1,Nterm
Int1[i]=1/2*(cos(lambda[i]*L)*sin(lambda[i]*L)+lambda[i]*L)/lambda[i]
h_bar_eff*(T_f-T_infinity)*sin(lambda[i]*L)/lambda[i]=(h_bar_eff*C4[i]-k*C3[i]*lambda[i])*Int1[i]
end
Equation (18) is substituted into the boundary condition at y = th, Eq. (5):
q1 if x < c
i k cos ( i x ) C3,i cosh ( i th ) + C4,i sinh ( i th ) = 0 if c < x < ( d + c ) (22)
i =1
q2 if x > ( d + c )
> int(cos(lambda[i]*x),x=0..c);
sin( i c )
i
> int(cos(lambda[i]*x),x=(d+e)..L);
sin( i ( d + e ) ) + sin( i L )
i
duplicate i=1,Nterm
Int2[i]=(-sin(lambda[i]*(d+c))+sin(lambda[i]*L))/lambda[i]
k*(C3[i]*lambda[i]*cosh(lambda[i]*L)+C4[i]*lambda[i]*sinh(lambda[i]*L))*Int1[i]=&
q``_1*sin(lambda[i]*c)/lambda[i]+q``_2*Int2[i]
end
x_bar=0.5 [-]
y_bar=0 [-]
x=x_bar*L
y=y_bar*th
duplicate i=1,Nterm
theta[i]=cos(lambda[i]*x)*(C3[i]*sinh(lambda[i]*y)+C4[i]*cosh(lambda[i]*y))
end
T=sum(theta[1..Nterm])+T_infinity
T_C=converttemp(K,C,T)
83
y/th = 1
82
y/th = 0.75
Temperature (C)
81
y/th = 0.5
80
y/th = 0.25
79
y/th = 0
78
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless axial position, x/L
Figure 2: Temperature as a function of x/L for various values of y/th.
d.) The goal of the base plate is to provide an uniform heat flow to each fin. Assess the
performance of the base plate by plotting the rate of heat flux transferred to the fluid as a
function of x at y = 0.
Figure 3 shows the heat flux to the fluid as a function of axial location.
19000
3 W/m-K
18000
Heat flux to fluid (W/m )
2
17000
16000
15000
14000 10 W/m-K
13000
30 W/m-K
12000 100 W/m-K
11000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless axial location, x/L
Figure 3: Heat flux to fluid as a function of x/L.
e.) Overlay on your plot for (d) the rate of heat flux transferred to the fluid for various values of
the base plate conductivity.
Figure 3 includes various values of conductivity and shows that as conductivity increases, more
energy is lost to the edge and the heat flux along the bottom is more uniform.
Problem 2.2-6 (2-4 in text)
q = 10000 W/m
2
H = 3 cm
Tset = 20C
W = 6 cm
Tset = 20C
kx = 50 W/m-K
ky = 4 W/m-K
Figure P2.2-6: Composite structure exposed to a heat flux.
The structure is anisotropic. The effective conductivity of the composite in the x-direction is kx =
50 W/m-K and in the y-direction it is ky = 4 W/m-K. The top of the structure is exposed to a heat
flux of q = 10,000 W/m2. The other edges are maintained at Tset = 20C. The height of the
structure is H = 3 cm and the half-width is W = 6 cm.
a.) Develop a separation of variables solution for the 2-D steady-state temperature distribution in
the composite.
"Inputs"
W=6 [cm]*convert(cm,m) "width of laminate"
H=3 [cm]*convert(cm,m) "height of laminate"
q``=10000 [W/m^2] "heat flux"
k_x=50 [W/m-K] "conductivity in the x-direction"
k_y=4 [W/m-K] "conductivity in the y-direction"
T_set=converttemp(C,K,20[C]) "specified edge temperatures"
A half-symmetry model of the problem shown in Figure P2.2-6 is governed by the transformed
partial differential equation:
2 2
kx + k =0 (1)
x 2 y 2
y
=0 (2)
x x =0
x =W = 0 (3)
ky = q (4)
y y=H
y =0 = 0 (5)
where
= T Tset (6)
= X Y (7)
d 2 X d 2 Y
kx Y + y
k X =0 (8)
dx 2 dx 2
d 2 X d 2 Y
dx 2 + k y dx 2 = 0 (9)
X
k Y
x
2 2
d 2 X
2
+ 2 X = 0 (10)
dx
d 2 Y
2 2 Y = 0 (11)
dy 2
where
kx
2 = (12)
ky
beta=sqrt(k_x/k_y)
Notice that the sign of the constant is selected so that the solution in the homogeneous direction
(x) is sines and cosines. The solution to Eq. (10) is:
X = C1 sin ( x ) + C2 cos ( x ) (13)
X = C2 cos ( x ) (14)
The boundary condition at x = L, Eq. (3), leads to the eigencondition for the problem:
cos ( i W ) = 0 (15)
i =
(1 + 2 i ) for i = 0,1,... (16)
2W
The sum of these solutions is, itself, a solution (note that the constant C2,i is absorbed into the
other constants):
= i = cos ( i x ) C3,i sinh ( i y ) + C4,i cosh ( i y ) (20)
i =1 i =1
= i = Ci cos ( i x ) sinh ( i y ) (21)
i =1 i =1
Equation (21) is substituted into the boundary condition at y = H, Eq. (4):
i k y Ci cos ( i x ) cosh ( i H ) = q (22)
i =1
> restart;
> assume(i,integer);
> lambda:=(1+2*i)*Pi/(2*W);
( 1 + 2 i~ )
:=
2W
> int((cos(lambda*x))^2,x=0..W);
W
2
> int(cos(lambda*x),x=0..W);
2 ( -1 )i~ W
( 1 + 2 i~ )
duplicate i=0,N_term
k_y*C[i]*beta*lambda[i]*cosh(beta*lambda[i]*H)*W/2=q``*2*(-1)^i*W/((1+2*i)*Pi)
end
1
20
23.12
Dimensionless y position, y/H 0.8
26.24
29.36
0.6 32.48
35.6
38.72
0.4
41.84
44.96
0.2 48.08
51.2
0
0 0.2 0.4 0.6 0.8 1
rout = 10 cm L = 10 cm
rin = 6 cm TLOx = 125 K
Figure P2.3-1: Cryogen transfer pipe connecting two liquid oxygen tanks.
a.) Develop an analytical model that can predict the temperature distribution within the pipe.
Prepare a contour plot of the temperature distribution within the pipe.
qx + qr = qx + dx + qr + dr
or
qx q
0= dx + r dr
x r
T
qx = k 2 r dr
x
and
T
qr = k 2 r dx
r
or
2T T
r + r =0
x 2 r r
A half-symmetry model of the pipe will be generated; the boundary conditions are therefore:
T
=0
x x =0
Tx = L = TLOx
T
=0
r r = rin
T
k = q
r r = rout
As stated, there are two non-homogeneous boundary conditions; however, the boundary
condition at x = L can be made homogeneous by defining the temperature difference:
= T TLOx
The partial differential equation and boundary conditions are written in terms of :
2
r + r =0 (1)
x 2 r r
=0 (2)
x x =0
x= L = 0 (3)
=0 (4)
r r = rin
k = q (5)
r r = rout
Note that the two homogeneous boundary conditions are in the x-direction and so the
eigenfunctions of the problem will be in this direction. We assume that the solution is separable;
that is, the solution is the product of a function only of x (X) and r (R):
( x, y ) = X ( x ) R ( r )
Substituting the product solution into the governing partial differential equation, Eq. (1), leads to:
d 2 X d d R
r R + X r =0
dx 2
dr dr
d 2 X d d R
r
dx + dr dr = 0
2
X r R
2 2
Note that the 1st term is a function only of x while the 2nd term is a function only of r; these two
quantities must be equal and opposite constants (2). The choice of the sign is again important;
the eigenfunctions must be in x and therefore the two ordinary differential equations must be:
d 2 X
+ 2 X = 0 (6)
dx 2
d d R
r 2 r R = 0 (7)
dr dr
The eigenproblem will be solved first; the solution to Eq. (6) is:
X = C1 cos ( x ) + C2 sin ( x )
The boundary condition at x = 0, Eq. (2), eliminates the sine term. The boundary condition at x =
L, Eq. (3), leads to:
C1 cos ( L ) = 0
The ordinary differential equation for R, Eq. (7), is solved by Bessel functions, as discussed in
Section 1.8. Equation (7) is a form of Bessel's equation:
d p d 2 s
x c x =0
dx dx
where s = 1 and p = 1; the quantity s p +2 is therefore not equal to zero and we are directed
toward the left-side of the Bessel function flow chart presented in Section 1.8.4 where the
parameters n = 0, a = 1, and n/a = 0 are computed. The last term is negative and therefore the
general solution to Eq. (7) is:
> restart;
> ODEr:=diff(r*diff(thetar(r),r),r)-lambda^2*r*thetar(r)=0;
d
2
d
ODEr := thetar( r ) + r 2 thetar( r ) 2 r thetar( r ) = 0
dr dr
> thetars:=dsolve(ODEr);
thetars := thetar( r ) = _C1 BesselI( 0, r ) + _C2 BesselK( 0, r )
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*L);
( 2 i~ 1 )
:=
2L
> theta:=(x,r)->cos(lambda*x)*(C3*BesselI(0,lambda*r)+C4*BesselK(0,lambda*r));
:= ( x, r ) cos( x ) ( C3 BesselI( 0, r ) + C4 BesselK( 0, r ) )
Verify that it solves both boundary conditions in the x-directions, Eqs. (2) and (3):
> eval(diff(theta(x,r),x),x=0);
0
> theta(L,r);
0
> r*diff(diff(theta(x,r),x),x)+diff(r*diff(theta(x,r),r),r);
( 2 i~ 1 ) x
r cos
1
( 2 i~ 1 )
2 2
4 2 L
( 2 i~ 1 ) r + C4 BesselK 0, ( 2 i~ 1 ) r L 2 +
C3 BesselI 0,
2L 2L
( 2 i~ 1 ) r ( 2 i~ 1 )
C3 BesselI 1,
( 2 i~ 1 ) x 1 2 L
cos
2 L 2 L
( 2 i~ 1 ) r
C4 BesselK 1, ( 2 i~ 1 )
1 2L + r cos ( 2 i~ 1 ) x
2 L 2 L
( 2 i~ 1 ) r
2 L BesselI 1,
( 2 i~ 1 ) r 2L ( 2 i~ 1 ) 2 2
C3 BesselI 0,
( 2 i~ 1 ) r
1 2L
4 2
L
( 2 i~ 1 ) r
2 L BesselK 1,
( 2 i~ 1 ) r ( 2 i~ 1 ) 2 2
C4 BesselK 0, 2L
1 2L ( 2 i~ 1 ) r
2
4 L
> simplify(%);
0
= i = cos ( i x ) C3,i BesselI ( 0,i r ) + C4,i BesselK ( 0,i r ) (8)
i =1 i =1
The solution must satisfy the boundary conditions in the non-homogeneous direction; the
boundary condition at r = rin, Eq. (4), leads to:
d
= cos ( i x ) C3,i BesselI ( 0,i r ) + C4,i BesselK ( 0,i r ) r = r = 0 (9)
r r = rin i =1 dr in
The derivatives of the Bessel functions may either be evaluated using the equations provided in
Section 1.8.4 or, more conveniently, using Maple:
> restart;
> diff(BesselI(0,lambda*r),r);
BesselI( 1, r )
> diff(BesselK(0,lambda*r),r);
BesselK( 1, r )
cos ( x ) C
i =1
i 3,i i BesselI (1,i rin ) C4,i i BesselK (1,i rin ) = 0
Unlike most of the problems we have previously encountered, neither of the constants is
eliminated; instead we see that there is a relationship between the two constants:
Equation (11) is substituted into the remaining non-homogeneous boundary condition at r = rout ,
Eq. (5), in order to obtain:
This equation is multiplied by an arbitrary eigenfunction, cos(j x), and integrated between the
homogeneous boundary conditions (from x = 0 to x= L); using the orthogonality property of the
eigenfunctions we obtain:
2 q sin ( i L )
Ci =
BesselI (1,i rin )
i2 k L BesselI (1,i rout ) BesselK (1,i rout )
BesselK (1,i rin )
"Inputs"
q``_dot=8000 [W/m^2] "Heat flux on pipe surface"
r_in=6.0 [cm]*convert(cm,m) "Pipe inner radius"
r_out=10.0 [cm]*convert(cm,m) "Pipe outer radius"
L = 10.0 [cm]*convert(cm,m) "Pipe half-length"
T_LOx=125 [K] "Pipe end temperature"
k=10 [W/m-K] "Pipe conductivity"
"dimensionless position"
r_bar=0.5
x_bar=0.5
r=r_in+(r_out-r_in)*r_bar
x=x_bar*L
The temperature distribution is computed over a range of positions and the results are used to
generate the contour plot shown in Figure 2.
Figure 2: Contour plot of temperature.
Problem 2.3-2 (2-6 in text)
Figure P2.3-2 illustrates a cylinder that is exposed to a concentrated heat flux at one end.
extends to infinity
k = 168 W/m-K
rout = 200 m
Ts = 20C
rexp = 21 m
q = 1500 W/cm
2 adiabatic
Figure P2.3-2: Cylinder exposed to a concentrated heat flux at one end.
The cylinder extends infinitely in the x-direction. The surface at x = 0 experiences a uniform
heat flux of q = 1500 W/cm2 for r < rexp = 21 m and is adiabatic for rexp < r < rout where rout =
200 m is the outer radius of the cylinder. The outer surface of the cylinder is maintained at a
uniform temperature of Ts = 20C. The conductivity of the cylinder material is k = 168 W/m-K.
a.) Develop a separation of variables solution for the temperature distribution within the
cylinder. Plot the temperature as a function of radius for various values of x.
"Inputs"
r_out=200 [micron]*convert(micron,m) "outer radius of domain"
q``_dot=1500 [W/cm^2]*convert(W/cm^2,W/m^2) "exposure flux"
k=168 [W/m-K] "conductivity of work piece"
r_exp=21 [micron]*convert(micron,m) "radius of exposure zone"
qx + qr = qx + dx + qr + dr
or
qx q
0= dx + r dr
x r
and
T
qr = k 2 r dx
r
T T
0= k 2 r dr dx + k 2 r dx dr
x x r r
or
2T T
r + r =0
x 2 r r
Tx = Ts
Tr =0 must be finite
Tr = rout = Ts
As stated, there are two non-homogeneous boundary conditions; however, the boundary
condition at r = rout can be made homogeneous by defining the temperature difference:
= T Ts
The partial differential equation and boundary conditions are written in terms of :
2
r + r =0 (1)
x 2 r r
r =r = 0
out
(5)
Note that the two homogeneous boundary conditions are in the x-direction and so the
eigenfunctions of the problem will be in this direction. We assume that the solution is separable;
that is, the solution is the product of a function only of x (X) and r (R):
( x, y ) = X ( x ) R ( r )
Substituting the product solution into the governing partial differential equation, Eq. (1), leads to:
d 2 X d d R
r R + X r =0
dx 2
dr dr
d 2 X d d R
r
dx + dr dr = 0
2
X r R
2 2
Note that the 1st term is a function only of x while the 2nd term is a function only of r; these two
quantities must be equal and opposite constants (2). The choice of the sign is again important;
the eigenfunctions must be in r and therefore the two ordinary differential equations must be:
d 2 X
2 X = 0 (6)
dx 2
d d R
r + 2 r R = 0 (7)
dr dr
The eigenproblem will be solved first; the solution to Eq. (7) is:
R = C1 BesselJ ( 0, r ) + C2 BesselY ( 0, r )
R = C1 BesselJ ( 0, r )
The boundary condition at r = rout, Eq. (5), leads to:
The 0th order Bessel function of the 1st kind (i.e., Bessel_J(0,x)) oscillates about zero every time
the argument changes by 2 in the same way that sine and cosine do; therefore, there are an
infinite number of eigenvalues i that will satisfy Eq. (8) associated with an infinite number of
eigenfunctions. The eigencondition for this problem cannot be used to explicitly solve for the
eigenvalues; rather, an implicit equation for the eigenvalues results from Eq. (8):
duplicate i=1,N
BesselJ(0,lambda[i]*r_out)=0 "solve for eigenvalues"
end
The arrays lowerlimit[], upperlimit[], and guess[] are used to constrain the solution for the array
lambda[] in the Variable Information window. The solution to the ordinary differential equation
for X, Eq. (6), is:
X i = C4,i exp ( i x )
Substituting Eq. (10) into the boundary condition at x = 0, Eq. (2), leads to:
rout rexp
k Ci i r BesselJ ( 0,i r ) dr = q
2
r BesselJ ( 0, r ) dr
i (12)
0 0
Integral 1 Integral 2
> restart;
> Integral1:=int(r*(BesselJ(0,lambda[i]*r))^2,r=0..r_out);
Integral1 :=
2 2
1 r_out ( i r_out BesselJ( 0, i r_out ) + i r_out BesselJ( 1, i r_out ) )
2 i
> Integral2:=int(r*BesselJ(0,lambda[i]*r),r=0..r_exp);
r_exp BesselJ( 1, r_exp i )
Integral2 :=
i
duplicate i=1,N
Integral1[i]=1/2*r_out/Pi^(1/2)/lambda[i]*(Pi^(1/2)*lambda[i]*r_out*BesselJ(0,lambda[i]*r_out)^2+&
Pi^(1/2)*lambda[i]*r_out*BesselJ(1,lambda[i]*r_out)^2)
Integral2[i]= 1/lambda[i]*r_exp*BesselJ(1,r_exp*lambda[i])
k*lambda[i]*C[i]*Integral1[i]=q``_dot*Integral2[i]
end
1.2
0.4
0.2
0
0 20 40 60 80 100 120 140 160 180 200
Radius (m)
Figure 2: Temperature difference as a function of r for various values of x.
b.) Determine the average temperature of the cylinder at the surface exposed to the heat flux.
The average temperature of the cylinder over the region x = 0 and 0 < r < rexp is:
rexp
1
T =
rexp
2 2 r T
0
x =0 dr (13)
or:
rexp
1
T = Ts +
rexp
2 2 r
0
x =0 dr (14)
c.) Define a dimensionless thermal resistance between the surface exposed to the heat flux and
Ts. Plot the dimensionless thermal resistance as a function of rout/rin.
A dimensionless thermal resistance is defined by normalizing the actual resistance against the
reistance to axial conduction through a cylinder that is rout in radius and rout long:
k (T Ts ) rout k (T Ts ) rout
2
R rout
2 2
R= = = k (17)
rout q rexp rout
2
q rout rexp
actual resistance
80
Dimensionless thermal resistance
10
0.1
1 10 100
Ratio of cylinder to exposure radii, rout/rexp
Figure 3: Dimensionless thermal resistance as a function of rout/rexp.
d.) Show that your plot from (c) does not change if the problem parameters (e.g., Ts, k, etc.) are
changed.
The values of various parameters were changed and did not affect Figure 3 (the plots are overlaid
onto Figure 3).
Problem 2.3-3
A disk-shaped window in an experiment is shown in Figure P2.3-3.
h = 50 W/m -K
2
T = 20C
th = 1 cm k = 1.2 W/m-K
Rw = 3 cm
x
r Tedge = 25C
= 1000 W/m2
qrad
Figure P2.3-3: Window.
The inside of the window (the surface at x = 0) is exposed to vacuum and therefore does not
experience any convection. However, this surface is exposed to a radiation heat flux qrad = 1000
2
W/m . The window is assumed to be completely opaque to this radiation and therefore it is
absorbed at x = 0. The edge of the window at Rw = 3 cm is maintained at a constant temperature
Tedge = 25C. The outside of the window (the surface at x = th) is cooled by air at T = 20C with
heat transfer coefficient h = 50 W/m2-K. The conductivity of the window material is k = 1.2
W/m-K.
a.) Is the extended surface approximation appropriate for this problem? That is, can the
temperature in the window be approximated as being 1-D in the radial direction? Justify
your answer.
"Inputs"
th=1 [cm]*convert(cm,m) "thickness of window"
R_w=3 [cm]*convert(cm,m) "outer radius of window"
q``_rad=1000 [W/m^2] "radiation heat flux"
k=1.2 [W/m-K] "conductivity"
T_edge=converttemp(C,K,25 [C]) "edge temperature"
T_infinity=converttemp(C,K,20 [C]) "ambient temperature"
h_bar=50 [W/m^2-K] "heat transfer coefficient"
A Biot number that characterizes the resistance to conduction in the axial direction to convection
is:
h th
Bi = (1)
k
b.) Assume that your answer to (a) is no. Develop a 2-D separation of variables solution to this
problem.
qx + qr = qx + dx + qr + dr
or
qx q
0= dx + r dr
x r
T
qx = k 2 r dr
x
and
T
qr = k 2 r dx
r
T T
0= k 2 r dr dx + k 2 r dx dr
x x r r
or
2T T
r + r =0
x 2 r r
T
k
= qrad
x x =0
T
k = h (Tx =th T )
x x = th
Tr =0 must be finite
Tr = Rw = Tedge
As stated, there are two non-homogeneous boundary conditions; however, the boundary
condition at r = Rw can be made homogeneous by defining the temperature difference:
= T Tedge
The partial differential equation and boundary conditions are written in terms of :
2
r 2 + r =0 (2)
x r r
k
= qrad (3)
x x =0
k = h ( x =th ) (4)
x x = th
r=R = 0 w
(6)
where
= T Tedge (7)
and
= T Tedge (8)
theta_infinity=T_infinity-T_edge
Note that the two homogeneous boundary conditions are in the x-direction and so the
eigenfunctions of the problem will be in this direction. We assume that the solution is separable;
that is, the solution is the product of a function only of x (X) and r (R):
( x, y ) = X ( x ) R ( r )
Substituting the product solution into the governing partial differential equation, Eq. (2), leads to:
d 2 X d d R
r R + X r =0
dx 2
dr dr
d 2 X d d R
r
dx + dr dr = 0
2
X r R
2 2
Note that the 1st term is a function only of x while the 2nd term is a function only of r; these two
quantities must be equal and opposite constants (2). The choice of the sign is again important;
the eigenfunctions must be in r and therefore the two ordinary differential equations must be:
d 2 X
2 X = 0 (9)
dx 2
d d R
r + 2 r R = 0 (10)
dr dr
The eigenproblem will be solved first; the solution to Eq. (10) is:
R = C1 BesselJ ( 0, r ) + C2 BesselY ( 0, r )
R = C1 BesselJ ( 0, r )
The 0th order Bessel function of the 1st kind (i.e., Bessel_J(0,x)) oscillates about zero every time
the argument changes by 2 in the same way that sine and cosine do; therefore, there are an
infinite number of eigenvalues i that will satisfy Eq. (11) associated with an infinite number of
eigenfunctions. The eigencondition for this problem cannot be used to explicitly solve for the
eigenvalues; rather, an implicit equation for the eigenvalues results from Eq. (11):
duplicate i=1,N
BesselJ(0,lambda[i]*R_w)=0 "solve for eigenvalues"
end
The arrays lowerlimit[], upperlimit[], and guess[] are used to constrain the solution for the array
lambda[] in the Variable Information window. The solution to the ordinary differential equation
for X, Eq. (9), is:
= Ri X i = BesselJ ( 0,i r ) C3,i sinh ( i x ) + C4,i cosh ( i x )
i =1 i =1
k BesselJ ( 0,i r ) C3,i i = qrad
(13)
i =1
Rw Rw
> restart;
> Int1[i]:=int(r*(BesselJ(0,lambda[i]*r))^2,r=0..R_w);
2 2
1 R_w ( i R_w BesselJ( 0, i R_w ) + i R_w BesselJ( 1, i R_w ) )
Int1i :=
2 i
> Int2[i]=int(r*BesselJ(0,lambda[i]*r),r=0..R_w);
R_w BesselJ( 1, i R_w )
Int2i =
i
and pasted into EES in order to obtain the constant C3,i for each eigenvalue:
duplicate i=1,N
Int1[i]=(R_w/Pi^(1/2)/lambda[i]*(Pi^(1/2)*lambda[i]*R_w*BesselJ(0,lambda[i]*R_w)^2+&
Pi^(1/2)*lambda[i]*R_w*BesselJ(1,lambda[i]*R_w)^2))/2
Int2[i] = 1/lambda[i]*R_w*BesselJ(1,lambda[i]*R_w)&
-k*C3[i]*lambda[i]*Int1[i]=q``_rad*Int2[i]
end
Rw
h r BesselJ ( 0, r ) dr =
0
i
Integral 2
Rw
duplicate i=1,N
h_bar*theta_infinity*Int2[i]=Int1[i]*(C4[i]*(k*lambda[i]*sinh(lambda[i]*th)+&
h_bar*cosh(lambda[i]*th))+C3[i]*(k*lambda[i]*cosh(lambda[i]*th)+&
h_bar*sinh(lambda[i]*th)))
end
x_bar=0 [-]
r_bar=0.5 [-]
x=x_bar*th
r=r_bar*R_w
duplicate i=1,N
theta[i]=BesselJ(0,lambda[i]*r)*(C3[i]*sinh(lambda[i]*x)+C4[i]*cosh(lambda[i]*x))
end
theta=sum(theta[1..N])
T=theta+T_edge
T_C=converttemp(K,C,T)
38 x/th = 0
Temperature (C)
36 x/th = 0.25
34 x/th = 0.5
x/th = 0.75
32
x/th = 1
30
28
26
24
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless radius, r/Rw
Figure 2: Temperature as a function of r for various values of x.
0.9 26.33
Dimensionless x position, x/th
27.66
0.8
28.98
0.7
30.31
0.6
31.64
0.5 32.97
0.4 34.3
0.3 35.62
0.2 36.95
38.28
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless radial position, r/Rw
Figure 3: Contour plot of the temperature in the window.
Problem 2.3-4
Reconsider Problem 2.3-3. The window is not opaque to the radiation but does absorb some of
it. The radiation that is absorbed is transformed to thermal energy. The volumetric rate of
exp ( x ) where = 100 m-1 is the
thermal energy generation is given by: g = qrad
absorption coefficient. The radiation that is not absorbed is transmitted. Otherwise the problem
remains the same.
a.) Develop a separation of variables solution to this problem using the techniques discussed in
Section 2.3.
"Inputs"
th=1 [cm]*convert(cm,m) "thickness of window"
R_w=3 [cm]*convert(cm,m) "outer radius of window"
q``_rad=1000 [W/m^2] "radiation heat flux"
k=1.2 [W/m-K] "conductivity"
T_edge=converttemp(C,K,25 [C]) "edge temperature"
T_infinity=converttemp(C,K,20 [C]) "ambient temperature"
h_bar=50 [W/m^2-K] "heat transfer coefficient"
alpha=100 [1/m] "absorption coefficient"
where g is the rate of generation of thermal energy. Equation (1) can be expanded:
q x q
g = dx + r dr (2)
x r
T
q x = k 2 r dr
x
T
qr = k 2 r dx
r
and
exp ( x )
g = 2 r dr dx qrad
into the differential energy balance leads to:
T T
exp ( x ) =
2 r dr dx qrad k 2 r dr dx + k 2 r dx dr
x x r r
or
2T T qrad
r + r + exp ( x ) r = 0 (3)
x 2
r r k
T
k =0 (4)
x x =0
T
k = h (Tx =th T ) (5)
x x = th
Tr = Rw = Tedge (7)
Note that the radiant heat flux is no longer absorbed at x = 0 but rather is absorbed volumetrically
throughout the domain; it shows up as the volumetric generation in the partial differential
equation rather than as a heat flux at a boundary. With the simple transformation:
= T T (8)
2 qrad
r + r + exp ( x ) r = 0 (9)
x 2
r r k
k =0 (10)
x x =0
k = h x =th (11)
x x = th
r =0 must be finite (12)
r = R = edge
w
(13)
where
theta_edge=T_edge-T_infinity
The partial differential equation is non-homogeneous and there is no simple transformation that
can rectify this. Therefore, it is necessary to use the solution technique provided in Section 2.3.
The first step is to split the solution into a homogeneous, 2-D solution that can be solved using
separation of variables and one or more 1-D particular solution:
= h + X + R (15)
2 h d 2 X h d dR qrad
r + r + r + r + exp ( x ) r = 0 (16)
x 2
dx 2
r r dr r k
In order to solve the problem for h using separation of variables it is necessary that:
2 h h
r 2 + r =0 (17)
x r r
d 2 X qrad
2
+ exp ( x ) = 0 (18)
dx k
and
d dR
r =0 (19)
dr r
dX qrad
= exp ( x ) + C1 (20)
dx k
qrad
X = exp ( x ) + C1 x + C2 (21)
k
R = C3 ln ( r ) + C4 (22)
In order to be able to solve h using separation of variables it is necessary that one direction have
homogeneous boundary conditions. This problem only works if x is the homogeneous direction.
Substituting Eq. (15) into Eq. (10) leads to:
h dX
k k =0 (23)
x x =0 dx x =0
Therefore:
h
k =0 (24)
x x =0
and
dX
k =0 (25)
dx x =0
qrad
+ C1 = 0 (26)
k
or:
qrad
C1 = (27)
k
C_1=-q``_rad/k
Therefore:
dX qrad q
= exp ( x ) rad (28)
dx k k
qrad q
X = exp ( x ) rad x + C2 (29)
k k
h
= h ( h , x =th + X x =th + R )
dX
k k (30)
x x =th dx x = th
h
k = h h , x =th (31)
x x = th
as well as:
dX
k = h X x =th (32)
dx x =th
and
R=0 (33)
Therefore C3 and C4 must both be zero. Substituting Eqs. (28) and (29) into Eq. (32) leads to:
q q q q
k rad exp ( th ) rad = h rad exp ( th ) rad th + C2 (34)
k k k k
-k*(q``_rad*exp(-alpha*th)/k-q``_rad/k)=h_bar*(-q``_rad*th/k+C_2-q``_rad*exp(-alpha*th)/(k*alpha))
Finally, Eq. (15) is substituted into Eqs. (12) and (13) in order to determine the non-
homogeneous boundary conditions for h:
h,r = R + X = edge
w
(36)
Therefore:
and
qrad q
h ,r = R = edge exp ( x ) + rad x C2 (38)
w
k k
The problem for h is solved using the typical steps for separation of variables. The
eigencondition is:
Bi
tan ( i th ) = (39)
i
where
h th
Bi = (40)
k
The eigenvalues occur in regular intervals of th and are automatically identified using EES:
= cos ( i x ) C3,i BesselI ( 0,i r ) + C4,i BesselK ( 0,i r ) (41)
i =1
= Ci cos ( i x ) BesselI ( 0,i r ) (42)
i =1
Substituting Eq. (42) into the boundary condition at r = Rw, Eq. (38), leads to:
qrad q
C cos ( x ) BesselI ( 0, R ) =
i =1
i i i w edge
k
exp ( x ) + rad x C2
k
(43)
0
0
Integral 1 Integral 2
(44)
q q
th th
+ rad
k 0 x cos ( i x ) dx krad 0 exp ( x ) cos ( i x ) dx
Integral 3 Integral 4
> restart;
> Int1[i]:=int((cos(lambda[i]*x))^2,x=0..th);
1 cos ( i th ) sin( i th ) + i th
Int1i :=
2 i
> Int2[i]:=int(cos(lambda[i]*x),x=0..th);
sin( i th )
Int2i :=
i
> Int3[i]=int(x*cos(lambda[i]*x),x=0..th);
1 + cos ( i th ) + th sin( i th ) i
Int3i = 2
i
> Int4[i]=int(exp(-alpha*x)*cos(lambda[i]*x),x=0..th);
( th ) ( th )
e cos ( i th ) + i e sin( i th )
Int4i = 2
2 + i
duplicate i=1,N
Int1[i] = 1/2*(cos(lambda[i]*th)*sin(lambda[i]*th)+lambda[i]*th)/lambda[i]
Int2[i] = 1/lambda[i]*sin(lambda[i]*th)
Int3[i] = (-1+cos(lambda[i]*th)+th*sin(lambda[i]*th)*lambda[i])/lambda[i]^2
Int4[i] = (alpha-alpha*exp(-alpha*th)*cos(lambda[i]*th)+&
lambda[i]*exp(-alpha*th)*sin(lambda[i]*th))/(alpha^2+lambda[i]^2)
C[i]*BesselI(0,lambda[i]*R_w)*Int1[i]=(theta_edge-C_2)*Int2[i]-C_1*Int3[i]+q``_rad*Int4[i]/(k*alpha)
end
x_bar=0.5 [-]
r_bar=0.5 [-]
x=x_bar*th
r=r_bar*R_w
Xp=C_1*x+C_2-q``_rad*exp(-alpha*x)/(k*alpha)
duplicate i=1,N
theta_h[i]=C[i]*cos(lambda[i]*x)*BesselI(0,lambda[i]*r)
end
theta_h=sum(theta_h[1..N])
theta=theta_h+Xp
T=theta+T_infinity
T_C=converttemp(K,C,T)
31
x/th = 0
30 x/th = 0.25
x/th = 0.5
x/th = 0.75
Temperature (C)
29
x/th = 1.0
28
27
26
25
24
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless radius r/Rw
Figure 2: Temperature as a function of dimensionless radius for various values of x/th.
c.) Show that your solution limits to the solution from Problem 2.3-3 in the limit that .
Figure 3 illustrates the temperature at x = 0 predicted by Problem 2.3-3 and by the numerical
model developed here with .
35
34 P2.3-4 with
P2.3-3
33
Temperature (C)
32
31
30
29
28
27
26
25
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless radius r/Rw
Figure 3: Temperature at x = 0 as a function of dimensionless radius predicted by the solution developed in
Problem 2.3-3 overlaid onto the solution developed here with approaching infinity.
Problem 2.4-1 (2-7 in text)
The plate shown in Figure P2.4-1 is exposed to a uniform heat flux q = 1x105 W/m2 along its
top surface and is adiabatic at its bottom surface. The left side of the plate is kept at TL = 300 K
and the right side is at TR = 500 K. The height and width of the plate are H = 1 cm and W = 5
cm, respectively. The conductivity of the plate is k = 10 W/m-K.
q = 1x10 W/m
5 2
TL = 300 K W = 5 cm
TR = 500 K
y H = 1 cm
x
k = 10 W/m-K
Figure P2.4-1: Plate.
a.) Derive an analytical solution for the temperature distribution in the plate.
"Inputs"
H=1 [cm]*convert(cm,m) "height of plate"
W=5 [cm]*convert(cm,m) "thickness of plate"
k=10 [W/m-K] "conductivity"
q``=100000 [W/m^2] "heat flux"
T_R=500 [K] "temperature of right hand surface"
T_L=300 [K] "temperature of left hand surface"
2 2
+ =0 (1)
x 2 y 2
x =0 = 0 (2)
x =W = R (3)
=0 (4)
y y =0
k = q (5)
y y=H
where
= T TL (6)
and
R = TR TL (7)
The problem has two, non-homogeneous boundary condition and therefore must be solved using
superposition.
= A + B (8)
2 A 2 A
+ 2 =0 (9)
x 2 y
A, x = 0 = 0 (10)
A, x =W = R (11)
A
=0 (12)
y y =0
A
=0 (13)
y y=H
x
A = R (14)
L
2 B 2 B
+ 2 =0 (15)
x 2 y
B , x =0 = 0 (16)
B , x =W = 0 (17)
B
=0 (18)
y y =0
B
k = q (19)
y y=H
The solution for B is 2-D and can be obtained using separation of variables. The eigenfunctions
are:
X B ,i = sin ( B ,i x ) (20)
iW
B , i = for i = 1, 2,.. (21)
"sub-problem B solution"
N_term=11 [-] "number of terms"
duplicate i=1,N_term
lambda_B[i]=i*pi/W "i'th eigenvalue"
end
k B ,i Ci sin ( B ,i x ) sinh ( B ,i H ) = q (24)
i =1
> restart;
> assume(i,integer);
> lambda_B:=i*Pi/W;
i~
lambda_B :=
W
> int((sin(lambda_B*x))^2,x=0..W);
W
2
> int(sin(lambda_B*x),x=0..W);
W ( 1 + ( -1 )i~ )
i~
duplicate i=1,N_term
C[i]*k*lambda_B[i]*sinh(lambda_B[i]*H)*W/2=q``*(-W*(-1+(-1)^i)/i/Pi) "i'th constant"
end
duplicate i=1,N_term
theta_B[i]=C[i]*sin(lambda_B[i]*x)*cosh(lambda_B[i]*y) "i'th term"
end
theta_B=sum(theta_B[1..N_term]) "solution to sub-problem B"
345.5
0.8
390.9
436.4
0.6 481.8
527.3
572.7
0.4
618.2
663.6
0.2 709.1
754.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless x-position, x/W
Figure 2: Contour plot of the temperature distribution.
Problem 2.5-1 (2-8 in text): A Heating Element
Figure P2.5-1 illustrates an electrical heating element that is affixed to the wall of a chemical
reactor. The element is rectangular in cross-section and very long (into the page). The
temperature distribution within the element is therefore two-dimensional, T(x, y). The width of
the element is a = 5.0 cm and the height is b = 10.0 cm. The three edges of the element that are
exposed to the chemical (at x = 0, y = 0, and x = a) are maintained at a temperature Tc = 200C
while the upper edge (at y = b) is affixed to the well-insulated wall of the reactor and can
therefore be considered adiabatic. The element experiences a uniform volumetric rate of thermal
energy generation, g = 1x106 W/m3. The conductivity of the material is k = 0.8 W/m-K.
reactor wall
k = 0.8 W/m-K
Tc = 200C
g = 1x10 W/m
6 3
a = 5 cm
y
Tc = 200C
x
Tc = 200C b = 10 cm
Figure P2.5-1: Electrical heating element.
"Inputs"
a=5.0 [cm]*convert(cm,m) "width of element"
b=10.0 [cm]*convert(cm,m) "height of element"
k=0.8 [W/m-K] "conductivity"
T_c=converttemp(C,K,200 [C]) "chemical temperature"
g```_dot=1e6 [W/m^3] "rate of volumetric generation"
L=1 [m] "unit length of element into the page"
The computational domain of the element with the regularly spaced grid of nodes is shown in
Figure 2.
Figure 2: The regularly spaced grid used to obtain a numerical solution.
The first step in obtaining a numerical solution is to position the nodes throughout the
computational domain. We will use grid with nodes placed on the edges and distributed
uniformly throughout. The x and y distance between adjacent nodes (x and y) are:
L
x = (1)
( m 1)
b
y = (2)
( n 1)
and the x and y positions of any node i,j are given by:
xi =
( i 1) a (3)
( m 1)
yj =
( j 1) b (4)
( n 1)
where m and n are the number of nodes used in the x and y directions.
"Setup nodes"
m=11 [-]
"number of nodes in the x-direction"
n=11 [-]
"number of nodes in the y-direction"
Dx=a/(m-1)
"distance between nodes in the x-direction"
Dy=b/(n-1)
"distance between nodes in the y-direction"
duplicate i=1,m
x[i]=(i-1)*Dx "x-position of each node"
end
duplicate j=1,n
y[j]=(j-1)*Dy "y-position of each node"
end
The next step in the solution is to write an energy balance for each node. Figure 3 illustrates a
control volume and the associated energy transfers for an internal node (see Figure 2) which
include conduction from each side ( q RHS and q LHS ), the top ( qtop ), and the bottom ( qbottom ). Note
that the direction associated with these energy transfers is arbitrary (i.e., they could have been
taken as positive if energy leaves the control volume), but it is important to write the equation in
a manner consistent with the chosen directions.
q RHS [i, j ] + q LHS [i, j ] + qtop [i, j ] + qbottom [i, j ] + g [i, j ] = 0 (5)
The next step is to approximate each of the terms in the energy balance; the material separating
the nodes is assumed to behave as a plane wall resistance and therefore:
k y L
q RHS [i, j ] =
x
(T [i + 1, j ] T [i, j ]) (6)
where L is the length of the element (assumed to be 1 m in order to do the problem on a unit
length basis); therefore, y L is the area for conduction and x is the distance over which the
conduction heat transfer occurs. Note that the temperature difference is consistent with the
direction of the arrow in Figure 3; if Ti+1,j is greater than Ti,j then energy is leaving the node and
q RHS is positive. The other heat transfers are approximated using a similar model:
k y L
q LHS [i, j ] =
x
(T [i 1, j ] T [i, j ]) (7)
k xW
qtop [i, j ] =
y
(T [i, j + 1] T [i, j ]) (8)
k xW
qbottom [i, j ] =
y
(T [i, j 1] T [i, j ]) (9)
The generation is the product of the volume of the control volume and the volumetric rate of
generation:
g [i, j ] = g x y L (10)
Note that each time the outer duplicate statement iterates once (i.e., i is increased by 1), the inner
duplicate statement iterates (n-1) times (i.e., j runs from 2 to n-1). Therefore, all of the internal
nodes are considered with these two nested duplicate loops. Also note that the unknowns are
placed in an array rather than a vector. The entries in the array T is accessed using two indices
contained in square brackets.
The boundary nodes have to be treated separately. The left, right, and bottom boundaries are
easy as the temperature is specified:
The upper boundary nodes must be considered using energy balances. Figure 4 illustrates an
energy balance associated with a node that is located on the top, insulated boundary (see Figure
2).
k y L
q RHS [i, n ] =
2x
(T [i + 1, n] T [i, n]) (15)
k y L
q LHS [i, n ] =
2 x
(T [i 1, n] T [i, n]) (16)
The factor of 2 in the denominator appears because there is half the available area for conduction
through the sides of the control volume on the top boundary. The other conduction term is
approximated as before:
k xW
qbottom [i, n ] =
y
(T [i, n 1] T [i, n]) (17)
The generation term is:
x y L g
g [i, n ] = (18)
2
"Upper edge"
duplicate i=2,(m-1)
q_dot_LHS[i,n]+q_dot_RHS[i,n]+q_dot_bottom[i,n]+gen[i,n]=0
q_dot_LHS[i,n]=k*(Dy/2)*L*(T[i-1,n]-T[i,n])/Dx
q_dot_RHS[i,n]=k*(Dy/2)*L*(T[i+1,n]-T[i,n])/Dx
q_dot_bottom[i,n]=k*Dx*L*(T[i,n-1]-T[i,n])/Dy
gen[i,n]=Dx*Dy*L*g```_dot/2
end
Notice that the control volumes at the top left corner (i.e., i =1, j =n) and the top right corner
(i.e., i =m, j =n) have already been taken addressed by the equations for the left and right
boundaries, Eqs. (11) and (12). Therefore, we have to make sure not to write additional
equations related to this node or the problem will be over-specified and so the equations for the
top boundary can only be written for i = 2...(m-1).
b.) Plot the temperature as a function of x at various values of y. What is the maximum
temperature within the element and where is it located?
The solution is obtained for n=11 and the columns Ti,1 (corresponding to y =0) through Ti,11
(corresponding to y = 10 cm) are plotted in Figure 5 as a function of x.
Figure 5: Temperature as a function of x for various values of y.
The hottest spot in the element is at the adiabatic wall (y = 10 cm) and the center (x = 2.5 cm);
the hottest temperature is about 860 K.
c.) Prepare a reality check to show that your solution behaves according to your physical
intuition. That is, change some aspect of your program and show that the results behave as
you would expect (clearly describe the change that you made and show the result).
There are several possible answers to this; I increased the conductivity by a factor of 10 and
examined the temperature distribution. Figure 6 illustrates the temperature as a function of x for
the original conductivity (k = 0.8 W/m-K) and the increased conductivity (k = 8.0 W/m-K);
notice that the increased conductivity has had the expected effect of reducing the temperature
rise.
Figure 6: Temperature as a function of x for y = 10 cm and two values of conductivity.
Problem 2.5-2
Figure P2.5-2 illustrates a composite material that is being machined on a lathe.
composite
thins = 100 m
kins = 1.5 W/m-K
thm = 200 m
km = 35 W/m-K h depends on RS
T = 20C
q
l
Tchuck = 20C
y
x
W = 12 cm
H = 3 cm
The composite is composed of alternating layers of insulating material and metal. The insulating
layers have thickness thins= 100 m and conductivity kins = 1.5 W/m-K. The metal layers have
thickness thm = 200 m and conductivity km = 35 W/m-K. The workpiece is actually cylindrical
and rotating. However, because the radius is large relative to it thickness and there are no
circumferential variations we can model the workpiece as a 2-D problem in Cartesian
coordinates, x and y, as shown in Figure 2.5-2. The width of the workpiece is W = 12 cm and the
thickness is H = 3 cm. The left surface of the workpiece at x = 0 is attached to the chuck and
therefore maintained at Tchuck = 20C. The inner surface at y = 0 is insulated. The outer surface
(at y = H) and right surface (at x = W) are exposed to air at T = 20C with heat transfer
coefficient h that depends on the rotational speed of the chuck, RS in rev/min, according to:
W min 2 2
h = 2 2 2
RS
m K rev
In order to extend the life of the tool used for the machining process, the workpiece is preheated
by applying laser power to the outer surface. The heat flux applied by the laser depends on the
rotational speed and position according to:
x x 2
ql = a RS exp - c
pw
The inputs are entered in EES and a function is defined in order to provide the heat flux:
$UnitSystem SI MASS RAD PA K J
$TABSTOPS 0.2 0.4 0.6 0.8 3.5 in
function q_f(RS,x)
"Inputs
RS - rotational speed (rpm)
x - position (m)
Output
q_f - laser heat flux (W/m^2)"
a=5000 [W-min/m^2-rev]
x_c=8 [cm]*convert(cm,m)
pw=1 [cm]*convert(cm,m)
q_f=a*RS*exp(-((x-x_c)/pw)^2)
end
"Inputs"
th_ins=100 [micron]*convert(micron,m) "insulation layer thickness"
k_ins=1.5 [W/m-K] "insulation conductivity"
th_m=200 [micron]*convert(micron,m) "metal thickness"
k_m=35 [W/m-K] "metal conductivity"
W=12 [cm]*convert(cm,m) "width of work piece"
H=3 [cm]*convert(cm,m) "thickness of work piece"
T_chuck=20 [C] "chuck temperature"
T_infinity=20 [C] "ambient air temperature"
RS=60 [rev/min] "rotational speed"
The heat transfer coefficient ( h ) is evaluated using the equation provided in the problem
statement.
The effective conductivities of a lamination stack was derived in Section 2.9.1. The effective
conductivity in the x-direction is:
keff , x =
( km thm + kins thins ) (1)
( thm + thins )
and the conductivity in the y-direction is:
keff , y =
( thm + thins ) (2)
thm thins
+
km kins
A regularly spaced grid of nodes is uniformly distributed with the first and last nodes in each
dimension placed on the boundaries of the domain; the x- and y-positions of any node (i,j) are
given by:
where M and N are the number of nodes used in the x- and y-directions, respectively. The x- and
y-distance between adjacent nodes (x and y, respectively) are:
W
x = (5)
( M 1)
H
y = (6)
( N 1)
This information is entered in EES:
"Setup nodes"
M=21 [-] "number of nodes in the x-direction"
N=11 [-] "number of nodes in the y-direction"
Dx=W/(m-1) "distance between nodes in the x-direction"
Dy=H/(n-1) "distance between nodes in the y-direction"
duplicate i=1,m
x[i]=(i-1)*Dx "x-position of each node"
end
duplicate j=1,n
y[j]=(j-1)*Dy "y-position of each node"
end
The next step in the solution is to write an energy balance for each node. The energy transfers
for an internal node include conduction from each side ( q RHS and q LHS ), the top ( qtop ), and the
bottom ( qbottom ). The direction associated with these energy transfers is arbitrary but they are all
taken as being into the control volume, as was done in Section 2.5.2.
The next step is to approximate each of the terms in the energy balance:
y y x x
keff , x
x
( Ti +1, j Ti , j ) + keff , x
x
( Ti 1, j Ti , j ) + keff , y
y
( Ti , j +1 Ti , j ) + keff , y
y
(Ti, j 1 Ti, j ) = 0 (8)
for i = 2... ( M 1) and j = 2... ( N 1)
where Tchuck is the chuck temperature. These equations are entered in EES:
"left boundary"
duplicate j=1,N
T[1,j]=T_chuck
end
The remaining boundary nodes do not have specified temperatures and therefore must be
determined using energy balances. An energy balance on a node that is located on the top
boundary (at y= H) is:
x h
+
k
(T Ti, N ) + qlx=0 (2-11)
i = 2... ( M 1)
"top boundary"
duplicate i=2,(M-1)
Dy*k_x*(T[i+1,N]-T[i,N])/(2*Dx)+Dy*k_x*(T[i-1,N]-T[i,N])/(2*Dx)+&
Dx*k_y*(T[i,N-1]-T[i,N])/Dy+Dx*h_bar*(T_infinity-T[i,N])+q_f(RS,x[i])*Dx=0
end
A similar procedure for the nodes on the lower boundary leads to:
"bottom boundary"
duplicate i=2,(M-1)
Dy*k_x*(T[i+1,1]-T[i,1])/(2*Dx)+Dy*k_x*(T[i-1,1]-T[i,1])/(2*Dx)+Dx*k_y*(T[i,2]-T[i,1])/Dy=0
end
Energy balances for the nodes on the right-hand boundary (x = L) lead to:
"right boundary"
duplicate j=2,(N-1)
Dx*k_y*(T[M,j+1]-T[M,j])/(2*Dy)+Dx*k_y*(T[M,j-1]-T[M,j])/(2*Dy)+&
Dy*k_x*(T[M-1,j]-T[M,j])/Dx+Dy*h_bar*(T_infinity-T[M,j])=0
end
The two corners (right upper and right lower) have to be considered separately. A control
volume and energy balance for node (M,N) leads to:
keff , y x keff , x y
2 y
(T M , N 1 TM , N ) +
2 x
(T
M 1, N TM , N )
(2-14)
x y
+h
2
( T TM , N ) +h
2
(T TM , N ) =0
"upper right corner"
Dx*k_y*(T[M,N-1]-T[M,N])/(2*Dy)+Dy*k_x*(T[M-1,N]-T[M,N])/(2*Dx)+&
h_bar*Dx*(T_infinity-T[M,N])/2+h_bar*Dy*(T_infinity-T[M,N])/2=0
The energy balance for the right lower boundary, node (M,1), leads to:
x y y
2 y
( TM ,2 TM ,1 ) +
2 x
( TM 1,1 TM ,1 ) + h
2
(T TM ,1 ) = 0 (2-15)
y = 30 mm
Temperature (C) 200
150
y = 24 mm
y = 18 mm
100
50
y = 0 mm
y = 6 mm
y = 12 mm
0
0 0.02 0.04 0.06 0.08 0.1 0.12
Axial position (m)
Figure 2: Temperature as a function of x for various values of y.
c.) Plot the maximum temperature in the workpiece as a function of the rotational speed, RS. If
the objective is to preheat the material to its maximum possible temperature, then what is the
optimal rotational speed?
The maximum temperature (Tmax) occurs on the upper boundary (j = N) and is identified
according to:
T_max=max(T[1..M,N])
Figure 3 illustrates the maximum temperature as a function of rotational speed and shows that
the maximum preheating occurs when the rotational speed is approximately 12 rpm.
90
Maximum temperature (C)
80
70
60
50
40
30
0 10 20 30 40 50 60
Rotational speed (rev/min)
Figure 3: Maximum temperature as a function of the rotational speed.
Problem 2.5-3
In Section 1.6 the constant cross-section, straight fin shown in Figure P2.5-3 was analyzed under
the assumption that it could be treated as an extended surface (i.e., temperature gradients in the y
direction can be neglected). In this example, the 2-D temperature distribution within the fin will
be determined using separation of variables.
T , h
W
insulated tip
y
x
g
Tb L
th
Figure P2.5-3: Straight, constant cross-sectional area fin.
Assume that the tip of the fin is insulated and that the width (W) is much larger than the
thickness (th) so that convection from the edges can be neglected. The length of the fin is L =
5.0 cm. The fin base temperature is Tb = 20C and the fin experiences convection with fluid at
T= 100C with average heat transfer coefficient, h = 100 W/m2-K. The fin is th = 2.0 cm thick
and has conductivity k = 1.5 W/m-K.
a.) Develop a numerical solution for the temperature distribution in the fin using a finite
difference technique.
A regularly spaced grid of nodes is uniformly distributed with the first and last nodes in each
dimension placed on the boundaries of the domain; the x- and y-positions of any node (i,j) are
given by:
xi =
( i 1) L (1)
( M 1)
yj =
( j 1) th (2)
2 ( N 1)
where M and N are the number of nodes used in the x- and y-directions, respectively. The x- and
y-distance between adjacent nodes (x and y, respectively) are:
L
x = (3)
( M 1)
th
y = (4)
2 ( N 1)
where
k yW
q RHS ,i , j =
x
(Ti+1, j Ti, j ) (6)
k yW
q LHS ,i , j =
x
(Ti 1, j Ti, j ) (7)
k xW
qtop ,i , j =
y
(Ti, j +1 Ti, j ) (8)
k xW
qbottom ,i , j =
y
(Ti, j 1 Ti, j ) (9)
g i , j = x yW g (10)
"Internal nodes"
duplicate i=2,(M-1)
duplicate j=2,(N-1)
q_dot_LHS[i,j]+q_dot_RHS[i,j]+q_dot_top[i,j]+q_dot_bottom[i,j]+g_dot[i,j]=0
q_dot_LHS[i,j]=k*W*Dy*(T[i-1,j]-T[i,j])/Dx
q_dot_RHS[i,j]=k*W*Dy*(T[i+1,j]-T[i,j])/Dx
q_dot_top[i,j]=k*W*Dx*(T[i,j+1]-T[i,j])/Dy
q_dot_bottom[i,j]=k*W*Dx*(T[i,j-1]-T[i,j])/Dy
g_dot[i,j]=W*Dx*Dy*g```_dot
end
end
where
k yW
q RHS ,i , N =
2 x
(Ti+1, N Ti, N ) (13)
k yW
q LHS ,i , N =
2 x
(Ti1, N Ti, N ) (14)
qconv ,i , N = h W x (T Ti , N ) (15)
k xW
qbottom ,i , N =
y
(Ti, N 1 Ti, N ) (16)
y
g i , N = x W g (17)
2
This process is continued for all of the edge and corner nodes:
b.) Use the numerical solution to predict and plot the temperature distribution in the heater.
180
y = 0 (center)
160
Temperature (C)
140
y = th/2 (edge)
120
100
80
60
40
20
0 0.01 0.02 0.03 0.04 0.05
Axial position (m)
Figure 2: Temperature as a function of x for various values of y.
c.) Use the numerical solution to predict the heater efficiency; the heater efficiency is defined as
the ratio of the rate of heat transfer to the fluid to the total rate of thermal energy generation
in the fin.
x
qconv ,1, N = h W
2
(T T1, N ) (18)
and the total convective heat transfer from the heater is:
M
qhtr = 2 qconv ,i , N (19)
i =1
qhtr
htr = (20)
g W L th
d.) Plot the heater efficiency as a function of the fin length for various values of the fin
thickness. Explain your plot.
Figure 3 illustrates the heater efficiency as a function of length for various values of thickness.
Note that at small lengths the fin efficiency can become negative because the heater thermally
communicates with the wall and is unable to elevate the temperature above the fluid
temperature. Therefore, the fluid is actually cooled by the heater.
1
0.8
th = 3 cm
0.6 th = 5 cm
Heater efficiency 0.4
0.2 th = 7 cm
0 th = 9 cm
-0.2
-0.4
-0.6
-0.8
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Heater length (m)
Figure 3: Heater efficiency as a function of length for various values of thickness.
Problem 2.6-1 (2-9 in text): Model of Welding Process (revisited)
Figure P2.6-1 illustrates a cut-away view of two plates that are being welded together. Both
edges of the plate are clamped and effectively held at temperatures Ts = 25C. The top of the
plate is exposed to a heat flux that varies with position x, measured from joint, according to:
qm ( x ) = q j exp ( x / L j ) where q j =1x106 W/m2 is the maximum heat flux (at the joint, x = 0)
and Lj = 2.0 cm is a measure of the extent of the heat flux. The back side of the plates are
exposed to liquid cooling by a jet of fluid at Tf = -35C with h = 5000 W/m2-K. A half-
symmetry model of the problem is shown in Figure P2.6-1. The thickness of the plate is b = 3.5
cm and the width of a single plate is W = 8.5 cm. You may assume that the welding process is
steady-state and 2-D. You may neglect convection from the top of the plate. The conductivity
of the plate material is k = 38 W/m-K.
both edges are held at fixed temperature
heat flux
joint
qm impingement cooling
k = 38 W/m-K
W = 8.5 cm Ts = 25C
b = 3.5 cm
y
Figure P2.6-1: Welding process and half-symmetry model of the welding process.
a.) Develop a separation of variables solution to the problem (note, this was done previously in
Problem 2.2-1). Implement the solution in EES and prepare a plot of the temperature as a
function of x at y = 0, 1.0, 2.0, 3.0, and 3.5 cm.
b.) Prepare a contour plot of the temperature distribution.
See the solution for Problem 2.2-1 for parts (a) and (b).
c.) Develop a numerical model of the problem. Implement the solution in MATLAB and
prepare a contour or surface plot of the temperature in the plate.
The input parameters are entered in the MATLAB function P2p2_1; the input arguments are m
and n, the number of nodes in the x and y coordinates while the output arguments are the x and y
positions of each node and the predicted temperature at each node.
function[xm,ym,T]=P2p6_1(m,n)
end
function[qflux]=qf(x)
qflux=qf_j*exp(-x/L_j);
end
A 2-D numerical model will be generated using a grid in which the x and y coordinates of each
node are:
xi =
( i 1)W for i = 1..m (1)
( m 1)
yi =
( j 1)W for j = 1..n (2)
( m 1)
The distance between adjacent nodes is:
W
x = (3)
( m 1)
W
y = (4)
( n 1)
%Setup grid
for i=1:m
x(i,1)=(i-1)*W/(m-1);
end
Dx=W/(m-1);
for j=1:n
y(j,1)=(j-1)*W/(n-1);
end
Dy=b/(n-1);
The problem will be solved by setting up and inverting a matrix containing the algebraic
equations that enforce the conservation of energy for each control volume. A control volume for
an internal node includes conduction from the left and right sides ( q LHS and q RHS ) and top and
bottom ( qtop and qbottom ). The energy balance is:
k y L
q RHS =
x
(Ti1, j Ti, j )
k y L
q LHS =
x
(Ti+1, j Ti, j )
k x L
qbottom =
y
(Ti, j 1 Ti, j )
k x L
qtop =
y
(Ti, j +1 Ti, j )
where L is the depth of the plate (into the page); L is set to 1.0 m which is consistent with doing
the problem on a per unit length basis. Combining these equations leads to:
k y L k y L k x L k x L
x
( Ti 1, j Ti , j ) +
x
( Ti +1, j Ti , j ) +
y
( Ti , j 1 Ti , j ) +
y
(Ti, j +1 Ti, j ) = 0 (5)
for i = 2.. ( m 1) and j = 2.. ( n 1)
The equation is rearranged to make it clear what the coefficient is for each unknown
temperature:
k y L k x L k y L k y L k x L k x L
Ti , j 2 2 + Ti 1, j + Ti +1, j + Ti , j 1 + Ti , j +1 =0
x y x x y y
for i = 2.. ( m 1) and j = 2.. ( n 1)
(6)
The control volume equations must be placed into the matrix equation:
AX =b
where the equation for control volume i,j is placed into row m(j-1)+i of A and Ti,j corresponds to
element Xm (j-1)+i in the vector X . Therefore, each coefficient in Eq. (6) (i.e., each term
multiplying an unknown temperature on the left side of the equation) must be placed in the row
of A corresponding to the control volume being examined and the column of A corresponding
to the unknown in X . The matrix assignments consistent with Eq. (6) are:
k y L k x L
Am( j 1)+i ,m( j 1)+i = 2 2 for i = 2.. ( m 1) and j = 2.. ( n 1) (7)
x y
k y L
Am( j 1)+i ,m( j 1)+i 1 = for i = 2.. ( m 1) and j = 2.. ( n 1) (8)
x
k y L
Am( j 1)+i ,m( j 1)+i +1 = for i = 2.. ( m 1) and j = 2.. ( n 1) (9)
x
k x L
Am( j 1)+i ,m( j 11)+i = for i = 2.. ( m 1) and j = 2.. ( n 1) (10)
y
k x L
Am( j 1)+i ,m( j +11)+i = for i = 2.. ( m 1) and j = 2.. ( n 1) (11)
y
A sparse matrix is allocated in MATLAB for A and the equations derived above are
implemented using a nested for loop. The spalloc command requires three arguments, which are
the number of rows and columns and the maximum number of elements in the matrix. Note that
there are at most 5 non-zero entries in each row of A , corresponding to Eqs. (7) through (11);
thus the last argument in the spalloc command which corresponds to the maximum number of
non-zero entries in the sparse matrix.
The nodes along the upper edge must be considered separately, leading to:
k y L k x L k y L k y L k x L
Ti ,n
x
y + Ti 1, n + Ti +1, n + Ti ,n 1 = qm x L
2 x 2 x y (12)
for i = 2.. ( m 1)
k y L k x L
Am( n 1)+i ,m( n 1)+i = for i = 2.. ( m 1) (13)
x y
k y L
Am( n 1)+i ,m( n 1)+i 1 = for i = 2.. ( m 1) (14)
2 x
k y L
Am( n 1)+i ,m( n 1)+i +1 = for i = 2.. ( m 1) (15)
2 x
k x L
Am( n 1)+i , m( n 11)+i = for i = 2.. ( m 1) (16)
y
%upper edge
for i=2:(m-1)
A(m*(n-1)+i,m*(n-1)+i)=-k*Dy*L/Dx-k*Dx*L/Dy;
A(m*(n-1)+i,m*(n-1)+i-1)=k*Dy*L/(2*Dx);
A(m*(n-1)+i,m*(n-1)+i+1)=k*Dy*L/(2*Dx);
A(m*(n-1)+i,m*(n-1-1)+i)=k*Dx*L/Dy;
bm(m*(n-1)+i,1)=-qf(x(i))*Dx*L;
end
The node at the upper left corner must be considered separately, leading to:
k y L k x L k y L k x L x L
T1,n + T2,n + T1,n 1 = qm (18)
2 x 2 y 2 x 2 y 2
k y L k x L
Am( n 1)+1,m( n 1)+1 = (19)
2 x 2 y
k y L
Am( n 1) +1,m( n 1)+1+1 = (20)
2 x
k x L
Am( n 1)+1, m( n 11)+1 = (21)
2 y
x L
bm( n 1)+1,1 = qm (22)
2
The node along the left edge must be considered separately, leading to:
k y L k x L k y L k x L k x L
T1, j + T2 +1, j + T1, j 1 + T1, j +1 =0
x y 2 x 2 y 2 y (23)
for j = 2.. ( n 1)
k y L k x L
Am( j 1)+1,m( j 1)+1 = for j = 2.. ( n 1) (24)
x y
k y L
Am( j 1)+1,m( j 1)+1+1 = for j = 2.. ( n 1) (25)
x
k x L
Am( j 1) +1,m( j 11)+1 = for j = 2.. ( n 1) (26)
2 y
k x L
Am( j 1)+1,m( j +11)+1 = for j = 2.. ( n 1) (27)
2 y
%left edge
for j=2:(n-1)
A(m*(j-1)+1,m*(j-1)+1)=-k*Dy*L/Dx-k*Dx*L/Dy;
A(m*(j-1)+1,m*(j-1)+1+1)=k*Dy*L/Dx;
A(m*(j-1)+1,m*(j-1-1)+1)=k*Dx*L/(2*Dy);
A(m*(j-1)+1,m*(j+1-1)+1)=k*Dx*L/(2*Dy);
end
The node at the lower left corner must be considered separately, leading to:
k y L k x L x L k y L k x L x L
T1,1 h + T2,1 + T1,2 = h Tf (28)
2 x 2 y 2 2 x 2 y 2
k y L k x L x L
Am(11) +1,m(11)+1 = h (29)
2 x 2 y 2
k y L
Am(11)+1, m(11)+1+1 = (30)
2 x
k x L
Am(11)+1, m(1+11)+1 = (31)
2 y
x L
bm(11)+1,1 = h Tf (32)
2
The nodes along the bottom edge must be considered separately, leading to:
k y L k x L k y L k y L k x L
Ti ,1 h x L + Ti 1,1 + Ti +1,1 + Ti ,1+1 = h x LT f
x y 2 x 2 x y (33)
for i = 2.. ( m 1)
k y L
Am(11) +i , m(11)+i 1 = for i = 2.. ( m 1) (35)
2 x
k y L
Am(11) +i , m(11)+i +1 = for i = 2.. ( m 1) (36)
2 x
k x L
Am(11) +i , m(1+11)+i = for i = 2.. ( m 1) (37)
y
%lower edge
for i=2:(m-1)
A(m*(1-1)+i,m*(1-1)+i)=-k*Dy*L/Dx-k*Dx*L/Dy-h*Dx*L;
A(m*(1-1)+i,m*(1-1)+i-1)=k*Dy*L/(2*Dx);
A(m*(1-1)+i,m*(1-1)+i+1)=k*Dy*L/(2*Dx);
A(m*(1-1)+i,m*(1+1-1)+i)=k*Dx*L/Dy;
bm(m*(1-1)+i,1)=-h*Dx*L*T_f;
end
The vector of unknowns X is obtained through matrix manipulation and then placed into matrix
format. Matrices for the x and y positions of each node are also created.
X=A\bm;
for i=1:m
for j=1:n
xm(i,j)=x(i);
ym(i,j)=y(j);
T(i,j)=X(m*(j-1)+i);
end
end
>> [x,y,T]=P2p6_1(20,40);
>> surf(x,y,T);
b.) Plot the temperature as a function of x at y = 0, b/2, and b and overlay on this plot the
separation of variables solution obtained in part (a) evaluated at the same locations.
Figure 4: Temperature as a function of axial position for y = 0, b/2, and b predicted using the
separation of variables solution and the numerical solution.
Problem 2.6-2
Prepare a solution to Problem 2.3-3 using a finite difference technique.
a.) Plot the temperature as a function of r for various values of x.
clear all;
th=0.01; % thickness of window (m)
h_bar=50; % heat transfer coefficient (W/m^2-K)
T_infinity=20; % ambient temperature (C)
T_edge=25; % edge temperature (C)
k=1.2; % conductivity (W/m-K)
R_w=0.03; % window radius (m)
qf_rad=1000; % radiation heat flux (W/m^2)
A 2-D numerical model will be generated using a grid in which the r and x coordinates of each
node are:
ri =
( i 1) RW for i = 1..M (1)
( M 1)
xj =
( j 1) th for j = 1..N (2)
( N 1)
The distance between adjacent nodes is:
RW
r = (3)
( M 1)
th
x = (4)
( N 1)
% Setup grid
M=51; % number of r-nodes
for i=1:M
r(i,1)=(i-1)*R_w/(M-1);
end
Dr=R_w/(M-1); % distance between r-nodes
N=101; % number of x-nodes
for j=1:N
x(j,1)=(j-1)*th/(N-1);
end
Dx=th/(N-1); % distance between x-nodes
The problem will be solved by setting up and inverting a matrix containing the algebraic
equations that enforce the conservation of energy for each control volume. A control volume for
an internal node includes conduction from the left and right sides ( q LHS and q RHS ) and top and
bottom ( qtop and qbottom ). The energy balance is:
r
k 2 ri + x
2
q RHS =
r
(Ti +1, j Ti, j )
r
k 2 ri x
2
q LHS =
r
(Ti1, j Ti, j )
k 2 ri r
qbottom =
x
(Ti, j 1 Ti, j )
k 2 ri r
qtop =
x
(Ti, j +1 Ti, j )
Combining these equations leads to:
r r
k 2 ri + x k 2 ri x
r
2
( Ti +1, j Ti , j ) +
r
2
(Ti1, j Ti, j )
k 2 ri r k 2 ri r
+
x
( Ti , j 1 Ti , j ) +
x
(Ti, j +1 Ti, j ) = 0 (5)
The equation is rearranged to make it clear what the coefficient is for each unknown
temperature:
r r
k 2 ri 2 x k 2 ri + 2 x k 4 r r
Ti , j i
+
r r x
r r
k 2 ri 2 x k 2 ri + 2 x
Ti 1, j +T + (6)
i +1, j
r r
k 2 ri r k 2 ri r
Ti , j 1 + Ti , j +1 =0
x x
for i = 2.. ( M 1) and j = 2.. ( N 1)
The control volume equations must be placed into the matrix equation:
AX =b
where the equation for control volume i,j is placed into row M (j-1) + i of A and Ti,j corresponds
to element XM(j-1)+i in the vector X . Therefore, each coefficient in Eq. (6) (i.e., each term
multiplying an unknown temperature on the left side of the equation) must be placed in the row
of A corresponding to the control volume being examined and the column of A corresponding
to the unknown in X . The matrix assignments consistent with Eq. (6) are:
r r
k 2 ri x k 2 ri + x k 4 r r
AM ( j 1)+i , M ( j 1)+i = 2
2
i
(7)
r r x
for i = 2.. ( M 1) and j = 2.. ( N 1)
r
k 2 ri x
AM ( j 1)+i , M ( j 1)+i 1 = 2
for i = 2.. ( M 1) and j = 2.. ( N 1) (8)
r
r
k 2 ri + x
AM ( j 1)+i , M ( j 1)+i +1 = 2
for i = 2.. ( M 1) and j = 2.. ( N 1) (9)
r
k 2 ri r
AM ( j 1)+i , M ( j 11)+i = for i = 2.. ( M 1) and j = 2.. ( N 1) (10)
x
k 2 ri r
AM ( j 1)+i , M ( j +11)+i = for i = 2.. ( M 1) and j = 2.. ( N 1) (11)
x
A sparse matrix is allocated in MATLAB for A and the equations derived above are
implemented using a nested for loop. The spalloc command requires three arguments, which are
the number of rows and columns and the maximum number of elements in the matrix. Note that
there are at most 5 non-zero entries in each row of A , corresponding to Eqs. (7) through (11);
thus the last argument in the spalloc command which corresponds to the maximum number of
non-zero entries in the sparse matrix.
A=spalloc(M*N,M*N,5*M*N);
b=zeros(M*N,1);
for i=2:(M-1)
for j=2:(N-1)
A(M*(j-1)+i,M*(j-1)+i)=-k*2*pi*Dx*(r(i)-Dr/2)/Dr-...
k*2*pi*Dx*(r(i)+Dr/2)/Dr-k*4*pi*r(i)*Dr/Dx;
A(M*(j-1)+i,M*(j-1)+i-1)=k*2*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(j-1)+i,M*(j-1)+i+1)=k*2*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(j-1)+i,M*(j+1-1)+i)=k*2*pi*r(i)*Dr/Dx;
A(M*(j-1)+i,M*(j-1-1)+i)=k*2*pi*r(i)*Dr/Dx;
end
end
for j=1:N
A(M*(j-1)+M,M*(j-1)+M)=1;
b(M*(j-1)+M,1)=T_edge;
end
The nodes along the bottom edge must be considered separately, leading to:
k x r k x r k 2 ri r
Ti ,1 ri ri +
r 2 r 2 x
k x r k x r k 2 ri r
+Ti 1,1 ri + Ti +1,1 ri + + Ti ,2 2 ri r
= qrad (12)
r 2 r 2 x
for i = 2.. ( M 1)
for i=2:(M-1)
A(M*(1-1)+i,M*(1-1)+i)=-k*pi*Dx*(r(i)-Dr/2)/Dr-...
k*pi*Dx*(r(i)+Dr/2)/Dr-k*2*pi*r(i)*Dr/Dx;
A(M*(1-1)+i,M*(1-1)+i-1)=k*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(1-1)+i,M*(1-1)+i+1)=k*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(1-1)+i,M*(2-1)+i)=k*2*pi*r(i)*Dr/Dx;
b(M*(1-1)+i,1)=-qf_rad*2*pi*r(i)*Dr;
end
The nodes along the top edge must be considered separately, leading to:
k x r k x r k 2 ri r
Ti , N ri ri +
r 2 r 2 x
k x r k x r k 2 ri r
+Ti 1, N r ri 2 + Ti +1, N r ri + 2 + Ti , N 1 x = h 2 ri r T (13)
for i = 2.. ( M 1)
for i=2:(M-1)
A(M*(N-1)+i,M*(N-1)+i)=-k*pi*Dx*(r(i)-Dr/2)/Dr-k*pi*Dx*(r(i)+Dr/2)/Dr-
k*2*pi*r(i)*Dr/Dx-h_bar*2*pi*r(i)*Dr;
A(M*(N-1)+i,M*(N-1)+i-1)=k*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(N-1)+i,M*(N-1)+i+1)=k*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(N-1)+i,M*(N-1-1)+i)=k*2*pi*r(i)*Dr/Dx;
b(M*(N-1)+i,1)=-h_bar*2*pi*r(i)*Dr*T_infinity;
end
k 2 x r k 2 x r
T1, j r1 + + T2, j r1 + =0
r 2 r 2 (14)
for j = 2.. ( N 1)
for j=2:(N-1)
A(M*(j-1)+1,M*(j-1)+1)=-k*2*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(j-1)+1,M*(j-1)+2)=k*2*pi*(r(1)+Dr/2)*Dx/Dr;
end
The node at the lower and upper left corners lead to:
k x r k x r
T1,1 r1 + + T2,1 r1 + =0 (15)
r 2 r 2
k x r k x r
T1, N r1 + + T2, N r1 + =0 (16)
r 2 r 2
A(M*(1-1)+1,M*(1-1)+1)=-k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(1-1)+1,M*(1-1)+2)=k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(N-1)+1,M*(N-1)+1)=-k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(N-1)+1,M*(N-1)+2)=k*pi*(r(1)+Dr/2)*Dx/Dr;
The vector of unknowns X is obtained through matrix manipulation and then placed into matrix
format.
X=A\b;
for i=1:M
for j=1:N
T(i,j)=X(M*(j-1)+i);
end
end
40
x/th = 0
38
x/th = 0.25
36 x/th = 0.5
Temperature (C)
34
32
30
x/th = 0.75
28 x/th = 1
26 Problem 2.6-2
Problem 2.3-3
24
0 0.005 0.01 0.015 0.02 0.025 0.03
Radius (m)
Figure 1: Temperature as a function of radius at various values of axial position.
36
0.025
34
0.02
Axial position (m)
32
0.015
30
0.01
28
0.005
26
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Radius (m)
c.) Verify that your solution agrees with the analytical solution from Problem 2.3-3.
Figure 1 illustrates the solution obtained from Problem 2.3-3 and shows that the agreement is
nearly exact.
Problem 2.6-3
Prepare a solution to Problem 2.3-4 using a finite difference technique.
a.) Plot the temperature as a function of r for various values of x.
clear all;
th=0.01; % thickness of window (m)
h_bar=50; % heat transfer coefficient (W/m^2-K)
T_infinity=20; % ambient temperature (C)
T_edge=25; % edge temperature (C)
k=1.2; % conductivity (W/m-K)
R_w=0.03; % window radius (m)
qf_rad=1000; % radiation heat flux (W/m^2)
alpha=100; % absorption coefficient (1/m)
A 2-D numerical model will be generated using a grid in which the r and x coordinates of each
node are:
ri =
( i 1) RW for i = 1..M (1)
( M 1)
xj =
( j 1) th for j = 1..N (2)
( N 1)
The distance between adjacent nodes is:
RW
r = (3)
( M 1)
th
x = (4)
( N 1)
% Setup grid
M=51; % number of r-nodes
for i=1:M
r(i,1)=(i-1)*R_w/(M-1);
end
Dr=R_w/(M-1); % distance between r-nodes
N=101; % number of x-nodes
for j=1:N
x(j,1)=(j-1)*th/(N-1);
end
Dx=th/(N-1); % distance between x-nodes
The problem will be solved by setting up and inverting a matrix containing the algebraic
equations that enforce the conservation of energy for each control volume. A control volume for
an internal node includes conduction from the left and right sides ( q LHS and q RHS ) and top and
bottom ( qtop and qbottom ). The energy balance is:
r
k 2 ri + x
2
q RHS =
r
(Ti +1, j Ti, j )
r
k 2 ri x
2
q LHS =
r
(Ti1, j Ti, j )
k 2 ri r
qbottom =
x
(Ti, j 1 Ti, j )
k 2 ri r
qtop =
x
(Ti, j +1 Ti, j )
exp ( x j )
g = 2 ri r x qrad
r r
k 2 ri + x k 2 ri x
2
r
( Ti +1, j Ti , j ) +
r
2
(Ti1, j Ti, j )
k 2 ri r k 2 ri r
+
x
( Ti , j 1 Ti , j ) +
x
(Ti, j +1 Ti, j ) = 2 ri r x qrad
exp ( x j ) (5)
The equation is rearranged to make it clear what the coefficient is for each unknown
temperature:
r r
k 2 ri 2 x k 2 ri + 2 x k 4 r r
Ti , j i
+
r r x
r r
k 2 ri 2 x k 2 ri + 2 x
Ti 1, j +T + (6)
i +1, j
r r
k 2 ri r k 2 ri r
Ti , j 1 + Ti , j +1 = 2 ri r x qrad exp ( x j )
x x
for i = 2.. ( M 1) and j = 2.. ( N 1)
The control volume equations must be placed into the matrix equation:
AX =b
where the equation for control volume i,j is placed into row M (j-1) + i of A and Ti,j corresponds
to element XM(j-1)+i in the vector X . Therefore, each coefficient in Eq. (6) (i.e., each term
multiplying an unknown temperature on the left side of the equation) must be placed in the row
of A corresponding to the control volume being examined and the column of A corresponding
to the unknown in X . The matrix assignments consistent with Eq. (6) are:
r r
k 2 ri x k 2 ri + x k 4 r r
AM ( j 1)+i , M ( j 1)+i = 2
2
i
(7)
r r x
for i = 2.. ( M 1) and j = 2.. ( N 1)
r
k 2 ri x
AM ( j 1)+i , M ( j 1)+i 1 = 2
for i = 2.. ( M 1) and j = 2.. ( N 1) (8)
r
r
k 2 ri + x
AM ( j 1)+i , M ( j 1)+i +1 = 2
for i = 2.. ( M 1) and j = 2.. ( N 1) (9)
r
k 2 ri r
AM ( j 1)+i , M ( j 11)+i = for i = 2.. ( M 1) and j = 2.. ( N 1) (10)
x
k 2 ri r
AM ( j 1)+i , M ( j +11)+i = for i = 2.. ( M 1) and j = 2.. ( N 1) (11)
x
exp ( x j ) for i = 2.. ( M 1) and j = 2.. ( N 1)
bM ( j 1)+i = 2 ri r x qrad (12)
A sparse matrix is allocated in MATLAB for A and the equations derived above are
implemented using a nested for loop. The spalloc command requires three arguments, which are
the number of rows and columns and the maximum number of elements in the matrix. Note that
there are at most 5 non-zero entries in each row of A , corresponding to Eqs. (7) through (11);
thus the last argument in the spalloc command which corresponds to the maximum number of
non-zero entries in the sparse matrix.
A=spalloc(M*N,M*N,5*M*N);
b=zeros(M*N,1);
for i=2:(M-1)
for j=2:(N-1)
A(M*(j-1)+i,M*(j-1)+i)=-k*2*pi*Dx*(r(i)-Dr/2)/Dr-
k*2*pi*Dx*(r(i)+Dr/2)/Dr-k*4*pi*r(i)*Dr/Dx;
A(M*(j-1)+i,M*(j-1)+i-1)=k*2*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(j-1)+i,M*(j-1)+i+1)=k*2*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(j-1)+i,M*(j+1-1)+i)=k*2*pi*r(i)*Dr/Dx;
A(M*(j-1)+i,M*(j-1-1)+i)=k*2*pi*r(i)*Dr/Dx;
b(M*(j-1)+i,1)=-2*pi*r(i)*Dr*Dx*qf_rad*alpha*exp(-alpha*x(j));
end
end
for j=1:N
A(M*(j-1)+M,M*(j-1)+M)=1;
b(M*(j-1)+M,1)=T_edge;
end
The nodes along the bottom edge must be considered separately, leading to:
k x r k x r k 2 ri r
Ti ,1 ri ri +
r 2 r 2 x
k x r k x r k 2 ri r
+Ti 1,1 ri + Ti +1,1 ri + + Ti ,2 = (13)
r 2 r 2 x
exp ( x1 )
ri r x qrad
for i = 2.. ( M 1)
for i=2:(M-1)
A(M*(1-1)+i,M*(1-1)+i)=-k*pi*Dx*(r(i)-Dr/2)/Dr-k*pi*Dx*(r(i)+Dr/2)/Dr-...
k*2*pi*r(i)*Dr/Dx;
A(M*(1-1)+i,M*(1-1)+i-1)=k*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(1-1)+i,M*(1-1)+i+1)=k*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(1-1)+i,M*(2-1)+i)=k*2*pi*r(i)*Dr/Dx;
b(M*(1-1)+i,1)=-pi*r(i)*Dr*Dx*qf_rad*alpha*exp(-alpha*x(1));
end
The nodes along the top edge must be considered separately, leading to:
k x r k x r k 2 ri r
Ti , N ri ri +
r 2 r 2 x
k x r k x r k 2 ri r
+Ti 1, N r ri 2 + Ti +1, N r ri + 2 + Ti , N 1 x = (14)
exp ( xN )
h 2 ri r T ri r x qrad
for i = 2.. ( M 1)
for i=2:(M-1)
A(M*(N-1)+i,M*(N-1)+i)=-k*pi*Dx*(r(i)-Dr/2)/Dr-k*pi*Dx*(r(i)+Dr/2)/Dr-...
k*2*pi*r(i)*Dr/Dx-h_bar*2*pi*r(i)*Dr;
A(M*(N-1)+i,M*(N-1)+i-1)=k*pi*Dx*(r(i)-Dr/2)/Dr;
A(M*(N-1)+i,M*(N-1)+i+1)=k*pi*Dx*(r(i)+Dr/2)/Dr;
A(M*(N-1)+i,M*(N-1-1)+i)=k*2*pi*r(i)*Dr/Dx;
b(M*(N-1)+i,1)=-h_bar*2*pi*r(i)*Dr*T_infinity-...
pi*r(i)*Dr*Dx*qf_rad*alpha*exp(-alpha*x(N));
end
k 2 x r k 2 x r
T1, j r1 + + T2, j r1 + =0
r 2 r 2 (15)
for j = 2.. ( N 1)
for j=2:(N-1)
A(M*(j-1)+1,M*(j-1)+1)=-k*2*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(j-1)+1,M*(j-1)+2)=k*2*pi*(r(1)+Dr/2)*Dx/Dr;
end
The node at the lower and upper left corners lead to:
k x r k x r
T1,1 r1 + + T2,1 r1 + =0 (16)
r 2 r 2
k x r k x r
T1, N r1 + + T2, N r1 + =0 (17)
r 2 r 2
A(M*(1-1)+1,M*(1-1)+1)=-k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(1-1)+1,M*(1-1)+2)=k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(N-1)+1,M*(N-1)+1)=-k*pi*(r(1)+Dr/2)*Dx/Dr;
A(M*(N-1)+1,M*(N-1)+2)=k*pi*(r(1)+Dr/2)*Dx/Dr;
The vector of unknowns X is obtained through matrix manipulation and then placed into matrix
format.
X=A\b;
for i=1:M
for j=1:N
T(i,j)=X(M*(j-1)+i);
end
end
31
x/th = 0
30 x/th = 0.25
x/th = 0.5
x/th = 0.75
Temperature (C)
29
x/th = 1.0
28
27
26
Problem 2.3-4
25 Problem 2.6-3
24
0 0.005 0.01 0.015 0.02 0.025 0.03
Radius (m)
Figure 1: Temperature as a function of radius at various values of axial position.
30
0.025 29.5
29
0.02 28.5
Axial position (m)
28
0.015
27.5
27
0.01
26.5
26
0.005
25.5
0 25
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Radius (m)
Figure 2: Contour plot of the temperature.
c.) Verify that your solution agrees with the analytical solution from Problem 2.3-4.
Figure 1 illustrates the solution obtained from Problem 2.3-4 and shows that the agreement is
nearly exact.
Problem 2.7-1 (2-10 in text): A Double Paned Window
Figure P2.7-1(a) illustrates a double paned window. The window consists of two panes of glass
each of which is tg = 0.95 cm thick and W = 4 ft wide by H = 5 ft high. The glass panes are
separated by an air gap of g = 1.9 cm. You may assume that the air is stagnant with ka = 0.025
W/m-K. The glass has conductivity kg = 1.4 W/m-K. The heat transfer coefficient between the
inner surface of the inner pane and the indoor air is hin = 10 W/m2-K and the heat transfer
coefficient between the outer surface of the outer pane and the outdoor air is hout = 25 W/m2-K.
You keep your house heated to Tin = 70F.
width of window, W = 4 ft
tg = 0.95cm
Tin = 70F
tg = 0.95 cm
hin = 10 W/m -K
2
g = 1.9 cm
H = 5 ft Tout = 23F
hout = 25 W/m -K
2
ka = 0.025 W/m-K
kg = 1.4 W/m-K
The average heating season in Madison lasts about time = 130 days and the average outdoor
temperature during this time is Tout = 23F. You heat with natural gas and pay, on average, ec =
1.415 $/therm (a therm is an energy unit =1.055x108 J).
a.) Calculate the average rate of heat transfer through the double paned window during the
heating season.
"Inputs"
tg=0.375*convert(inch,m) "glass thickness"
g=0.75*convert(inch,m) "air gap"
k_g=1.4 [W/m-K] "glass conductivity"
k_a=0.025 [W/m-K] "air conductivity"
H=5 [ft]*convert(ft,m) "height of window"
W=4[ft]*convert(ft,m) "width of window"
T_in=converttemp(F,K,70 [F]) "indoor air temperature"
h_in=10 [W/m^2-K] "heat transfer coefficient on inside of window"
T_out=converttemp(F,K,23 [F]) "outdoor air temperature"
h_out=25 [W/m^2-K] "heat transfer coefficient on outside of window"
time=130 [day]*convert(day,s) "heating season duration"
ec=1.415 [$/therm]*convert($/therm,$/J) "cost of energy"
The heat transfer is resisted by convection on the inner and outer surfaces:
1
Rconv ,in = (1)
hin W H
1
Rconv ,out = (2)
hout W H
tg
Rcond , g = (3)
kg W H
g
Rcond ,a = (4)
ka W H
Tin Tout
q = (5)
Rconv ,in + 2 Rcond , g + Rcond ,a + Rconv ,out
q_dot=(T_in-T_out)/(R_conv_in+2*R_cond_g+R_cond_a+R_conv_out)
"rate of heat transfer through the window"
b.) How much does the energy lost through the window cost during a single heating season?
The total amount of energy lost over the course of a heating season is:
Q = q time (6)
cost = ec Q (7)
1.9 cm
Tin = 70F 0.95 cm Tout = 23F
hout = 25 W/m -K
2
hin = 10 W/m -K
2
air 2 cm
4 cm
0.5 cm 3 cm
metal casing
km = 25 W/m-K 0.4 cm
wood
All surfaces of the casing that are adjacent to glass, wood, or the air between the glass panes can
be assumed to be adiabatic. The other surfaces are exposed to either the indoor or outdoor air.
c.) Prepare a 2-D thermal analysis of the casing using FEHT. Turn in a print out of your
geometry as well as a contour plot of the temperature distribution. What is the rate of energy
lost via conduction through the casing per unit length (W/m)?
Each of the points are selected individually and their exact coordinates are entered, which leads
to the more precise geometry shown in Figure 4.
Figure 5: Geometry.
The material properties are specified by selecting the outline and selecting Material Properties
from the Specify menu. The boundary conditions are specified by selecting each type of
boundary and then selecting Boundary Conditions from the Specify menu. A crude mesh is
generated, as shown in Figure 6.
d.) Show that your numerical model has converged by recording the rate of heat transfer per
length for several values of the number of nodes.
The mesh is refined several times and each time the heat transfer rate per unit length of casing is
recorded; the results are shown in Figure 8.
Figure 8: Heat transfer through the casing per unit length as a function of the number of nodes.
e.) How much does the casing add to the cost of heating your house?
) is entered in EES:
The result from FEHT, the heat transfer per unit length ( qcasing
qcasing = qcasing 2 (W + H ) (8)
stack structure
The system is fabricated from a high temperature material in the form of a ring; 16 fluid channels
are integrated with the stack in a circular array. The water to be heated flows through these
channels. The inner surface of the liner is finned to increase its surface area and is exposed to
the hot gas while the outer surface is cooled externally by ambient air. A cut-away view of the
stack is shown in Figure P2.7-2(b).
unit cell
Tair = 20C
hair = 20 W/m -K
2
Tw = 30C
hw = 500 W/m -K
2
Tgas = 800C
hgas = 50 W/m -K
2
k = 25 W/m-K
Figure P2.7-2(b): Problem specification for the energy recovery problem.
At a particular section, the water can be modeled as being at a uniform temperature of Tw = 30C
with a heat transfer coefficient, hw = 500 W/m2-K. The hot gas is at Tgas = 800C with a heat
transfer coefficient, hgas = 50 W/m2-K. The ambient air external to the stack is at Tair = 20C
and hair = 20 W/m2-K. The stack material has conductivity k = 25 W/m-K.
The liner geometry is relatively complex and includes curved segments as well as straight
sections. Only a single unit cell of the structure (see Figure P2.7-2(b)) needs to be simulated.
FEHT does not allow curved sections to be simulated; rather, a curved section must be
approximated as a polygon. Rather than attempting to draw the geometry manually, it is
preferable to import a drawing (e.g., from a computer aided drawing package) and trace the
drawing in FEHT. More advanced finite element tools will have automated processes for
importing geometry from various sources. A drawing of the unit cell with a scale can be copied
onto the clipboard (from the website for this text, using Microsoft Powerpoint) and pasted into
FEHT.
a.) Use FEHT to develop a finite element model of the stack and determine the heat transfer to
the water per unit length of stack.
Open FEHT and specify that you are doing a steady state heat transfer problem in cartesian
coordinates using the C temperature scale using the Setup menu (Figure 3).
Open the program that contains the unit cell drawing and copy it to the clip board. In FEHT,
select Paste from Clipboard from the File menu; the drawing should appear in the drawing space
(Figure 4).
Move the template to the center of the screen and select Size/Move Template from the draw
menu; this will allow the grid to show through the template (Figure 5).
Figure 5: Template and drawing grid.
Finally, resize your drawing grid so that 1 cm on your template corresponds to 1 cm on your
model. First, identify the size of the square that is contained on the template for this purpose.
Select Scale and Size from the Setup menu and then adjust the Grid Spacing option until the size
of grid matches the size of the square (Figures 6 and 7).
Figure 6: Adjust Grid Spacing until a grid matches the scale on the template.
Figure 7: Grid spacing adjusted to match the template scale.
In Figure 7, the grid spacing is 1.45 cm x 1.45 cm; therefore, the 1.45 cm of screen is equivalent
to 1 cm of drawing distance. Select Scale and Size from the Setup menu and set the scale such
that 1 cm of screen is equal to 1/1.45 cm (or 0.69 cm) of drawing distance (Figure 8).
Now it is possible to draw the geometry by tracing the template. You need to specify the outer
region and then punch holes in it (rather than drawing the flow passage and then overwriting it
with the outer region). Select Outline from the Draw menu and trace the outside edge of the
drawing. It is important to recognize that the curved surfaces must be represented by a discrete
number of straight lines that are defined by the nodes. The more nodes you use, the better you
can simulate the curve but the more difficult it will be to generate a mesh. Start with just a few
nodes, as shown in Figure 9.
Figure 9: Adjust scale to match template.
Select Hide Patterns from the Display menu so that the material is transparent (Figure 10).
Select Outline from the Draw menu and trace the circle (use an octagon to represent the curve),
as shown in Figure 11.
Figure 11: Circular flow passage defined in FEHT.
Select Hide Template from the Display menu and the geometry will remain. Save your file.
Having defined the geometry it is necessary to define the material properties, boundary
conditions, and mesh in order to solve the problem. The liquid within the flow passage can be
represented conveniently as a single temperature. In this case it is possible to mesh the liquid
and specify a very high conductivity; however, this is a waste of computational resources given
that we know the answer. Instead, FEHT allows you to specify that a region is a Lumped/Fluid
element which has a single temperature. This is a nice feature for many problems that reduces
the complexity of the model without affecting the accuracy of the predictions. Select the outline
of the fluid passage and then select Material Properties from the Specify menu (Figure 12).
Select not specified from the list and then give the element the name Lumped Water, click on
Type until Fluid/Lump appears.
Select Show Patterns from the Display menu to see the water. Next select the outline of the
stack material and specify the conductivity.
The edges that define the unit cell are adiabatic; select these edges and then select Boundary
Conditions from the Specify menu. Set the heat flux to 0 W/m2 (Figure 13).
Set the boundary conditions on the internal surface (exposed to combustion gas) and external
surface (exposed to ambient air) in the normal way. You also need to specify a heat transfer
coefficient between the water and the fluid passage; select the octagon that defines this boundary
and then select Boundary Conditions from the Specify menu. Set the heat transfer coefficient to
hwater (Figure 14).
Figure 14: Specify Boundary Conditions for the boundary of the lumped element.
You will need to specify the temperature of the lumped element; click on the lumped element
and select Lumped Information from the Specify menu (Figure 15). Set the temperature to be
Twater
Figure 15: Specify Lumped Information dialog.
It is necessary to specify a crude mesh; you do not need to mesh the lumped element because the
governing equations will not be solved within this element (it is just assigned a temperature).
The mesh is specified by selecting Element Lines from the Draw menu; all of the mesh elements
must be triangular. A reasonable initial mesh is shown in Figure 16.
Select Calculate from the Run menu and then select Temperature Contours from the View menu
(Figure 17).
Figure 17: Solution with crude mesh.
It is possible to determine the total heat transferred to the water by selecting Heat Flows from the
View menu and then selecting the 8 boundaries that define the water passage (Figure 18). Notice
that the sum of the total heat flow is displayed on the toolbar.
The total heat flow to a single tube is 3076 W/m; therefore, the total heat transfer to the 16
passages filled with water will be 49.2 kW.
Figure 19: Predicted heat flows per passage per unit length as a function of the number of elements.
The geometry is too irregular to allow an analytical solution. Therefore, it is possible to compare
the answer to a very rough resistance calculation. The known information is entered in EES:
"Inputs"
T_gas=converttemp(C,K,800) "hot gas"
h_gas=50 [W/m^2-K] "hot gas heat transfer coefficient"
T_water=converttemp(C,K,30) "water temperature"
h_water=500 [W/m^2-K] "water heat transfer coefficient"
k=25 [W/m-K] "material conductivity"
L=1 [m] "per unit length of stack"
There are, approximately, three resistances between the hot gas and the water. The resistance to
convection with the flue gas, Rconv,gas, is:
1
Rconv , gas = (1)
hgas L pgas
where pgas is the perimeter of the stack material that is exposed to gas - approximately 7.0 cm.
p_gas=7.0 [cm]*convert(cm,m) "perimeter of stack exposed to gas"
R_conv_gas=1/(h_gas*L*p_gas) "resistance due to convection to the gas"
1
Rconv , water = (2)
hwater L Dwater
where Dwater is the diameter of the water passage, approximately 1.5 cm.
Finally, the resistance due to conduction within the stack material, Rcond, is approximately:
Lc
Rcond = (3)
k L wc
where Lc is the distance heat must be conducted and L wc is the area for conduction; Lc is
approximately 3.0 cm and wc is approximately 2.5 cm.
q
(T gas Twater )
(4)
Rconv , gas + Rconv , water + Rcond
The approximate model predicts that Rconv,gas= 0.29 K/W, Rconv,water = 0.04 K/W, Rcond = 0.05
K/W which leads to q =2050 W; this is 30% less than the FEHT model prediction. The
agreement is not expected to be substantially better than this given the extremely crude
approximations involved in the resistance model. However, the calculations suggest that no
substantial errors were made in the finite element model. Furthermore, the magnitude of the
three resistances provides some insight into what thermal resistance controls the problem (the
convection with the air) and therefore how the design could be most effectively improved.
P2.7-3 (2-11 in text): A Spacecraft Radiator
A radiator panel extends from a spacecraft; both surfaces of the radiator are exposed to space (for
the purposes of this problem it is acceptable to assume that space is at 0 K); the emittance of the
surface is = 1.0. The plate is made of aluminum (k = 200 W/m-K and = 2700 kg/m3) and has
a fluid line attached to it, as shown in Figure 2.7-3(a). The half-width of the plate is a=0.5 m
wide while the height of the plate is b=0.75m. The thickness of the plate is a design variable and
will be varied in this analysis; begin by assuming that the thickness is th = 1.0 cm. The fluid
lines carry coolant at Tc = 320 K. Assume that the fluid temperature is constant although the fluid
temperature will actually decrease as it transfers heat to the radiator. The combination of
convection and conduction through the panel-to-fluid line mounting leads to an effective heat
transfer coefficient of h = 1,000 W/m2-K over the 3.0 cm strip occupied by the fluid line.
k = 200 W/m-K
space at 0 K
= 2700 kg/m3
= 1.0
a = 0.5 m 3 cm
th = 1 cm
b = 0.75 m
fluid at Tc = 320 K
The radiator panel is symmetric about its half-width and the critical dimensions that are required
to develop a half-symmetry model of the radiator are shown in Figure 2.7-3(b). There are three
regions associated with the problem that must be defined separately so that the surface conditions
can be set differently. Regions 1 and 3 are exposed to space on both sides while Region 2 is
exposed to the coolant fluid one side and space on the other; for the purposes of this problem, the
effect of radiation to space on the back side of Region 2 is neglected.
Region 1 (both sides exposed to space)
Region 2 (exposed to fluid - neglect radiation to space)
Region 3 (both sides exposed to space)
(0.50,0.75)
(0.50,0.55)
(0.50,0.52)
y
x
(0,0) (0.50,0)
(0.25,0) line of symmetry
(0.22,0)
Figure 2.7-3(b): Half-symmetry model.
a.) Prepare a FEHT model that can predict the temperature distribution over the radiator panel.
The radiator panel can be modeled as 2-D problem because it is thin and has high conductivity.
The Biot number compares the ratio of the internal conduction resistance to the external
resistance; in this case due to radiation.
Rcond
Bi = (1)
Rrad
where
th
Rcond = (2)
2 k As
1
Rrad = (3)
As s (T + Tsur
s
2 2
)(Ts + Tsur )
The Biot number is therefore:
th s (Ts2 + Tsur
2
)(Ts + Tsur )
Bi = (4)
2k
The surrounding temperature in Eq. (4), Tsur corresponds to space which is essentially 0 K and
therefore Eq. (4) reduces to:
th s Ts3
Bi = (5)
2k
We dont know the value of the panel surface temperature, Ts in Eq. (5), but we can assume that
it will be near the fluid temperature for any well-designed radiator. Using Ts = Tf in Eq. (5)
results in a Biot number of 4.6x10-5 which is very small; clearly for any reasonable value of Ts
the problem will be 2-D.
FEHT can simulate 2-D problems of this type using the Extended Surface mode. Start FEHT
and select Extended Surface from the Subject menu. Each of the three regions must be drawn
separately using the Outline option from the Draw menu. It is easiest to set an appropriate grid
using the Scale and Size selection from the Setup menu (Figure 2).
Figure 2: Scale and Size dialog window
Generate outlines for the 3 regions that are close to the correct scale and then double-click on
each node and position it exactly (Figure 3).
The material properties for all three regions can be set by selecting each while holding down the
shift key and then selecting Specify Properties from the Specify menu. Select Aluminum from
the list of materials and then modify the conductivity to be 200 W/m-K and the thickness to be
0.01 m to match the problem statement (Figure 4).
Figure 4: Specify material properties.
The surface conditions for Regions 1 and 3 can be set by selecting these regions with the shift
key held down and then selecting Surface Conditions from the Specify menu. These regions are
radiating to space but there is no option listed in the Extended Surface Conditions for a radiative
surface condition. The radiation heat flux is:
= T 4
qrad (6)
= T 3 (T 0 )
qrad (7)
= hrad (T T f )
qrad (8)
where fluid temperature (Tf) is 0 K and the convection coefficient (hrad) is a function of
temperature:
hrad = T 3 (9)
FEHT allows the specification of the surface conditions in terms of position (X and Y) as well as
temperature (T); therefore, the radiative surface condition can be modeled as shown in Figure 5.
Figure 5: Extended Surface Conditions Dialog window for Regions 1 and 3.
Region 2 experiences a convection boundary condition with the coolant, but only from one side
(as opposed to the double sided condition assumed by FEHT). The heat transfer from a
differential element in Region 2 is given by:
q = h dA (T Tc ) (10)
whereas the convection surface condition in FEHT assumes convection from both sides of the
plate and therefore:
q = hF 2 dA (T Tc ) (11)
where hF is the heat transfer coefficient that should be set in FEHT in order to simulate the
single-sided convection coefficient represented by Eq. (10). Comparing Eqs. (10) and (11) leads
to:
h
hF = (12)
2
Click on Region 2 and select Surface Condition from the Specify menu; specify the surface
conditions as shown in Figure 6.
Finally, the boundary conditions along each edge of the computational domain must be specified;
the line of symmetry is adiabatic and the remaining edges are also assumed to be adiabatic.
Generate a reasonable but crude mesh, as shown in Figure 7(a) and then refine it. Note that the
mesh within Region 2 will start relatively refined due to its small width and need not be refined
as much as the mesh in Regions 1 and 3. However, it is possible to refine the mesh in a single
region by selecting that region and then selecting Reduce Mesh from the Draw menu. The result
should be similar to Figure 7(b).
(a) (b)
Figure 7: (a) coarse and (b) refined mesh.
Solve the problem by selecting Calculate from the Run menu; the problem is non-linear due to
the temperature dependent heat transfer coefficient and therefore the solution process will be
iterative. Plot the temperature distribution by selecting Temperature Contours from the View
menu (Figure 8).
Figure 8: Temperature distribution for a 1 cm thick plate.
b.) Export the solution to EES and calculate the total heat transferred from the radiator and the
radiator efficiency (defined as the ratio of the radiator heat transfer to the heat transfer from
the radiator if it were isothermal and at the coolant temperature).
Select Tabular Output from the View menu and then Select All and Save As. Save the data as a
file called 1 cm. Open EES and select Open Lookup Table from the Tables menu; navigate the
file 1 cm and open it; the solution (the temperature at each node together with the locations of
the nodes) is contained in the lookup table. Using the technique discussed in EXAMPLE 16-1 it
is possible to use the Interpolate2D function to obtain the temperature at an arbitrary x and y
location on the radiator plate.
"INPUTS"
e=1.0 [-] "emittance of panel surface"
a=0.5 [m] "half-width of panel"
b=0.75 [m] "height of panel"
T_f=320 [K] "fluid temperature"
th=0.01 [m] "thickness"
rho=2700 [kg/m^3] "density"
k=200 [W/m-K] "conductivity"
"location on plate"
x=0.1 [m]
y=0.5 [m]
"2-D interpolation from table of nodal data"
T=Interpolate2D('1 cm',X,Y,T,X=x,Y=y) "temperature interpolated from data"
In order to obtain the total heat transferred from the panel it is necessary to integrate the heat flux
over the entire area of the plate. This process can be accomplished manually by dividing the
plate into many small integration areas, calculating the heat flux within each area, and summing
the result. The plate is divided into Nx segments in the x direction, each with width:
a
x = (13)
Nx
xi = ( i 0.5 ) x (14)
The plate is divided into Ny segments in the y direction, each with height:
b
y = (15)
Ny
y j = ( j 0.5 ) y (16)
The total heat transferred from the plate ( q ) is calculated using the double integral:
b a
q = q x , y dx dy (17)
0 0
The numerical summation that approximates the integration in Eq. (17) is:
N y Nx
q = q xi , y j x y (19)
j =1 i =1
where
q xi , y j = Tx4i , y j (20)
duplicate i=1,Nx
duplicate j=1,Ny
"2-D interpolation from table of nodal data"
T[i,j]=Interpolate2D('1 cm',X,Y,T,X=x[i],Y=y[j]) "temperature interpolated from data"
q``_dot[i,j]=2*e*sigma#*T[i,j]^4 "heat flux"
end
end
q_dot=sum(q``_dot[1..Nx,1..Ny])*Dx*Dy
Note that the same process can be accomplished using EES native Integral command. The
Integral command carries out numerical integration using a more sophisticated algorithm than
simply assuming a constant value over each step.
EES Integral command requires 4 arguments and allows a 5th, optional argument; the protocol
for calling the function is:
F=Integral(Integrand,VarName,LowerLimit,UpperLimit,StepSize)
where Integrand is the EES variable or expression that must be integrated, VarName is the
integration variable, and LowerLimit and UpperLimit define the limits of integration. StepSize
is optional and defines the numerical step that is used to accomplish the integration; a small
value of StepSize will lead to more accurate results but take longer to calculate.
The double integral in Eq. (17) can be accomplished by calling the Integral function twice; Eq.
(17) is rewritten as:
b
q = qy dy (21)
0
where
a
q y = q x , y dx (22)
0
Equation (22) is evaluated using the Integral command as shown in the EES code below:
Note that the value of y was set as q y is a function of y but the value of x is not set as x is the
integration variable. To evaluate Eq. (21), integrate q y from y=0 to y=b:
Note that the value of y is no longer set as y is the integration variable in the 2nd integral. The
results of the manual integration should agree with the use of the Integral function; either should
yield 339 W. The radiator efficiency () and mass (M) are computed according to:
q
= (23)
2 a b T f4
and
M = a b th (24)
c.) Explore the effect of thickness on the radiator efficiency and mass.
The solution procedure described above is repeated for several values of the radiator thickness.
Figure 10 shows the predicted temperature distribution for several values of the thickness. Note
that the same contour levels were selected for each plot by selecting User in the Temperature
Contour Information dialog window and specifying the range from 200 K to 320 K (Figure 9);
this allows the different cases to be compared directly and shows clearly that the reduced
thickness increases the resistance to conduction along the plate and therefore leads to
progressively larger temperature gradients through the plate.
Figure 9: Temperature Contour Information dialog window.
Figure 10: Temperature distribution in the panel for thicknesses of (a) 1.0 cm, (b) 0.75 cm, (c) 0.5 cm, (d) 0.3
cm, (e) 0.2 cm, and (f) 0.1 cm.
Figure 11 illustrates the radiator efficiency and mass as a function of the thickness. As the
thickness is reduced, the efficiency drops but so does the mass; clearly there must be a trade-off
between these effects.
Figure 11: Efficiency and mass as a function of thickness.
Figure 12 shows the efficiency as a function of mass and makes the trade-off clearer; above a
panel with a mass of nominally 2 kg there is a region of diminishing return where additional
mass provides only a small gain in efficiency.
A complete analysis would require more information then is given; specifically, how should the
radiator performance be compared with its mass to determine a true optimal thickness. Barring
additional constraints related to, for example, structural stability, Figure 12 suggests that any
optimization process will result in a panel that is approximately 0.2 cm thick with a mass of 2 kg
and an efficiency of 60%.
Problem 2.7-4
Gas turbine power cycles are used for the generation of power; the size of these systems can
range from 10s of kWs for the microturbines that are being installed on-site at some commercial
and industrial locations to 100s of MWs for natural gas fired power plants. The efficiency of a
gas turbine power plant increases with the temperature of the gas entering from the combustion
chamber; this temperature is constrained by the material limitations of the turbine blades which
tend to creep (i.e., slowly grow over time) in the high temperature environment in their high
centrifugal stress state. One technique for achieving high gas temperatures is to cool the blades
internally; often the air is bled through the blade surface using a technique called transpiration.
A simplified version of a turbine blade that will be analyzed in this problem is shown in Figure
P2.7-4.
combustion gas
k = 15 W/m-K
Tg = 1800 K
hg= 850 W/m2-K
The high temperature combustion gas is at Tg = 1800 K and the heat transfer coefficient between
the gas and blade external surface is hg = 850 W/m2-K. The blades are cooled by three internal
air passages. The cooling air in the passages is at Tc = 500 K and the air-to-blade heat transfer
coefficient is ha= 250 W/m2-K. The blade material has conductivity k = 15 W/m-K. The
coordinates of the points required to define the geometry are indicated in Figure P2.7-4.
a.) Generate a symmetry model of the blade in FEHT. Generate a figure showing the
temperature distribution in the blade predicted using a very crude mesh.
The half-symmetry model is generated by selecting Scale and Size from the Setup menu and
setting up a convenient grid to approximately place the outer nodes; the scale was set to 1 cm =
0.5 cm with grid spacing 0.5 cm x 0.5 cm (Figure 1). The temperatures are set to K and the
problem set to a steady-state problem in cartesian coordinates by making the appropriate
selections in the Setup menu.
Figure 2: Scale and Size window.
The nodes required to define the shape are approximately positioned using the grid by selecting
Outline from the Draw menu (Figure 3(a)); each node is subsequently double-clicked on and the
exact coordinates are entered (Figure 3(b)).
(a)
(b)
Figure 3: FEHT model with (a) nodes approximately placed on the grid and (b) nodes exactly positioned.
The material properties are specified by clicking on the outline and selecting Material Properties
from the Specify menu (Figure 4). It is best to set a lighter color for the material so that the grid
can be seen.
Figure 4: Specifying material properties.
The boundary conditions along the external surfaces are specified by clicking on these two lines
and selecting Boundary Conditions from the Specify menu (Figure 5).
The remaining boundary conditions are specified using the same process (notice that the
boundaries along the line of symmetry are adiabatic); the result is shown in Figure 6.
A very crude grid is generated (Figure 7); recall that each element has to be a triangle.
Figure 7: Crude mesh.
The mesh can be checked by selecting Check from the Run menu. If there are no errors then
select Calculate from the Run menu and view the temperature distribution by selecting
Temperature Contours from the View menu; the result is shown in Figure 8.
b.) Refine your mesh and keep track of the temperature experienced at the trailing edge of the
blade (i.e., at position 9 in Figure P2.7-4) as a function of the number of nodes in your mesh.
Prepare a plot of this data that can be used to establish that your model has converged to the
correct solution.
The nodal data can be viewed by selecting either Tabular Output or Temperatures from the View
menu. The number of entries in the Tabular Output table leads to the number of nodes and the
temperature at the trailing edge can be read from the model. The number of unknown
temperatures is also displayed each time the model is calculated. The mesh is refined over an
over again (by selecting Reduce Mesh from the Draw menu) in order to obtain the results listed
in Table 1 and plotted in Figure 9.
Table 1 and Figure 9 suggest that the solution has converged at 657 nodes. The temperature
distribution for 657 nodes is shown in Figure 10.
Figure 9: Trailing edge temperature as a function of the number of nodes.
Figure 10: Temperature distribution in the turbine blade with 657 nodes.
c.) Do your results make sense? Use a very simple, order-of-magnitude analysis based on
thermal resistances to decide whether your predicted blade surface temperature is reasonable
(hint there are three thermal resistances that govern the behavior of the blade, estimate each
one and show that your results are approximately correct given these thermal resistances).
From a very approximate standpoint, the problem is governed by the thermal resistance to
convection on the internal and external surfaces (Rconv,in and Rconv,ext) and conduction through the
blade (Rcond). These resistances are estimated here:
1
Rconv ,c = (1)
hc pc L
where pc is the perimeter of the cooling passages (7.75 cm, from Figure 2.7-4) and L is a unit
length of the turbine blade.
1
Rconv , g = (2)
hg pg L
Lcond
Rcond = (3)
k Acond
where Lcond is the approximate length that must be conducted through (about 0.5 cm from Figure
2.7-4) and Acond is the approximate area for conduction (about 5.0 cm x L from Figure 2.7-4).
The calculations are carried out in EES:
"Inputs"
T_c=500 [K] "cooling air temperature"
h_c=250 [W/m^2-K] "cooling air heat transfer coefficient"
T_g=1800 [K] "combustion gas temperature"
h_g=850 [W/m^2-K] "combustion gas heat transfer coefficient"
k=15 [W/m-K] "conductivity"
L=1 [m] "calculations are per meter"
Tg Tc
q = (4)
Rconv , g + Rcond + Rconv ,c
and can be used to estimate the surface temperature of the blade (Ts,g):
Ts , g = Tg q Rconv , g (5)
The result is 1570 K which is approximately consistent with the results shown in Figure 10.
Problem 2.7-5
a.) Show how the construction of the finite element problem changes with the addition of
volumetric generation.
T T
k + k g = 0 (1)
x x y y
T T
f x k x dx dy + f y k y dx dy f g dx dy = weighted average residual (2)
A
A A
Integral 1 Integral 2
f T f T
A x x y y dA A f g dx dy = B f qn ds
k + k (3)
T ( x, y ) = wT T (4)
f f
k x x ( w T ) + k y y ( w T ) dA f g dx dy = f qn ds
T T
(5)
A A B
Using the Galerkin technique for the weighting functions leads to:
w1 w1 w w w w w w
k + k 1 1 dA T1 + k 1 2 + k 1 2 dA T2 + ...
A x x y y A x x y y
w w w w
+ k 1 N + k 1 N dA TN w1 g dx dy = w1 qn ds
A x x y y A B
w2 w1 w w w w w w
k + k 2 1 dA T1 + k 2 2 + k 2 2 dA T2 + ...
A x x y y A x x y y
(6)
w wN w wN
+ k 2 +k 2 dA TN w2 g dx dy = w2 qn ds
A x x y y A B
...
wN w1 w w w w2 w w2
k + k N 1 dA T1 + k N +k N dA T2 + ...
A x x y y A x x y y
w wN w wN
+ k N +k N dA TN wN g dx dy = wN qn ds
A x x y y A B
w wT w wT
k + dA T w g dx dy = w qn ds (7)
A x x y y
A
B
K g q
The construction of the matrix K and the vector q are discussed in Section 2.7.2. This problem
focuses on the construction of the generation matrix g . The generation vector is assembled by
the summation of element generation vectors according to:
Ne
g = ge (8)
i =1
g e = w g edx dy (9)
Ae
where g e is the volumetric generation within element e. Within element e, only the weighting
functions wi, wj, and wk are nonzero (where i, j, and k are the indices of the nodes that define the
element):
The integral of a weighting function over the element is obtained from Eq. (S2.7.2-55):
a !b !c !
w wbj wkc dA =
a
2A (S2.7.2-55)
Ae
i
( a + b + c + 2 )! e
or
1!0!0! 1
w
Ae
i dA = wi1 w0j wk0 dA =
Ae (1 + 0 + 0 + 2 )!
2 Ae = Ae
3
(12)
.
row i
1
.
e Ae
g
ge = row j
3
1
.
1 row k
.
(13)
b.) Re-solve the problem discussed in Section 2.7.2 assuming that the material experiences a
volumetric generation rate of g = 1x103 W/m3, as shown in Figure P2.7-5.
ht = 100 W/m -K
2
T,t = 300 K
(0,0.75) (0.5,0.75)
k = 0.25 W/m-K
q = 100 W/m
2
g = 1500 W/m
3
(0.5,0.25)
hb = 50 W/m -K
2
(0,0) (0.25,0)
T,b = 350 K
Figure P2.7-5: Two-dimensional conduction problem used to illustrate the finite element solution with
generation. The coordinates of points are shown in m.
The solution proceeds as discussed in Section 2.7.2. The information about the sub-domains is
contained in a matrix sd . The number of columns in the matrix sd is equal to the number of
sub-domains in the computational domain. The information about the sub-domains is contained
in the column. The first row of each column contains the conductivity of the sub-domain and the
second row the generation. The matrix sd is setup in the MATLAB script S2p7d5 according to:
clear all;
bcm=[0,0,0; 50,350,0; 0,0,0; 100,300,0; 0,0,100]'; % boundary condition for each boundary
The global conduction matrix and generation vector are setup together, element-by-element.
Both are initialized:
for e=1:N_e
The indices of the three nodes that define the element and the subdomain that the element
belongs to are obtained from the triangle matrix:
The x- and y-distances separating the nodes (xij, xik, etc.) are computed; the coordinates of the
nodes are obtained from the point matrix:
The parameter bijk and the area of the element are computed:
bijk=xij*yjk-xjk*yij;
Ae=abs(bijk)/2; % area of element e
The conductivity of the element is obtained from the matrix sd and the elements of the element
conduction matrix are added to the global conduction matrix.
The generation within the element is obtained from the matrix sd and the elements of the
element conduction matrix are added to the global conduction matrix according to Eq. (13).
Figure 3: Solution to the problem shown in Figure 2.7-5 with the mesh shown in Figure 2.
Note that the PDE toolbox can be used to generate a more refined mesh, as shown in Figure 4.
Figure 4: Refined mesh generated by the PDE toolbox.
The mesh is exported to the workspace and the script is altered as shown below (changes are
highlighted in bold):
%clear all;
Figure 5: Solution to the problem shown in Figure 2.7-5 with the mesh shown in Figure 4.
Problem 2.7-6
Figure P2.7-6 illustrates a tube in a water-to-air heat exchanger with a layer of polymer coating
that can be easily etched away in order to form an array of fin-like structures that increase the
surface area exposed to air.
ha = 10 W/m -K
2
hw = 250 W/m -K
2
Tw = 320 K
tube, kt = 15 W/m-K unit cell of fin structure
Figure P2.7-6: Tube coated with polymer and a unit cell showing fin-like structures etched into polymer.
The water flowing through the tube has temperature Tw = 320 K and heat transfer coefficient hw =
250 W/m2-K. The air has temperature Ta = 290 K and heat transfer coefficient ha = 10 W/m2-K.
The thermal conductivity of the polymer and tube material is kp = 1.5 W/m-K and kt = 15 W/m-
K, respectively.
a.) Generate a finite element solution for the temperature distribution within the unit cell shown
in Figure 2.7-6 using the mesh shown in Figure 2.7-6(b). The coordinates of the nodes are
listed in Table P2.7-6.
13 14
12
11
10
7 8 9
6 5 4
1 2 3
Figure P2.7-6(b): Mesh for finite element solution.
function[]=P2p7d6(h_a)
% Input:
% h_a - heat transfer coefficient on the air-side (W/m^2-K)
%
subdomain 2 boundary 3
boundary 4
boundary 2
subdomain 1
boundary 1
Figure 3:Sub-domains and boundaries.
The information about the boundary conditions is contained in a matrix bc . The number of
columns in the matrix bc is equal to the number of boundaries in the computational domain.
The information about the boundaries is contained in the column: the first row holds the heat
transfer coefficient, the second row holds the adjacent fluid temperature, and the third row holds
the specified heat flux into the computational domain. There are four boundaries for the
problem, as shown in Figure 3, and therefore the matrix bc is setup according to:
The mesh shown in Figure P2.7-6(b) is defined by Nn = 14 nodes. Information about the nodes
is stored in the point matrix, p . There are Nn columns in the point matrix. The first row of each
column contains the x-coordinate of the corresponding node and the second row contains the y-
coordinate. The point matrix is setup:
The mesh is defined by Nb = 12 boundary segments. Information about the boundary segments
is stored in the edge matrix, e . There are Nb columns in the edge matrix. The first two rows of
each column contain the indices of the nodes that define the boundary segment. The third row
contains the index of the boundary that the boundary segment lies on.
The mesh is defined by Ne = 14 elements. Information about the edge is stored in the triangle
matrix, t . There are Ne columns in the triangle matrix. The first three rows contain the indices
of the three nodes that define the elements, given in clockwise order. The fourth row contains
the number of the subdomain that the element lies in.
The element conduction matrix was derived in Eq. (S2.7.2-66), which is repeated below:
. . . . . . .
(x jk + y jk ) . (x jk xik + y jk yik ) . (x xij + y jk yij )
2 2
. jk . row i
. . . . . . .
K e = 2 . (xik x jk + yik y jk ) (x + yik2 ) . (xik xij + yik yijk ) . row j
Ae ke 2
.
bijk
ik
. . . . . . .
. (xij x jk + yij y jk ) (xij xik + yij yik ) . (x + yij2 )
2
. ij . row k
. . . . . . .
The elements associated with the element conduction matrix are added, element by element, to
the global conduction matrix according to:
Ne
K = Ke (S2.7.2-1)
i =1
The global conduction matrix is initialized:
for e=1:N_e
The indices of the three nodes that define the element and the subdomain that the element
belongs to are obtained from the triangle matrix:
The x- and y-distances separating the nodes (xij, xik, etc.) are computed; the coordinates of the
nodes are obtained from the point matrix:
The parameter bijk and the area of the element are computed:
bijk=xij*yjk-xjk*yij;
Ae=abs(bijk)/2; % area of element e
The conductivity of the element is obtained from the matrix sd and the elements of the element
conduction matrix are added to the global conduction matrix according to Eq. (S2.2-66)..
The vector Qb and matrix H b were derived in Eq. (S2.7.2-84), which is repeated below:
row i . . . . .
. .
. 1 1
1 1 . . row i
qs,b sij hb T,b sij 3 6
qb = . + . hb sij . . . . . T
2 2
1 1 . 1
.
1
. row j
. . 6 3
row j . . . . .
Qb Hb column i
column j (S2.7.2-84)
for b=1:N_b
The indices of the two nodes that define the boundary segment and the boundary that the
segment belongs to are obtained from the edge matrix:
The x- and y-distance separating the nodes and the linear distance between the nodes, sij, are
computed:
The specified heat flux, heat transfer coefficient, and ambient temperature associated with the
boundary segment is obtained from the matrix bc and the elements of Qb and H b are added to
the global vector Q and matrix H according to Eq. (S2.7.2-84).
pdeplot(pm,em,tm,'xydata',T,'contour','on');
b.) Plot the average heat flux at the internal surface of the tube as a function of the air-side heat
transfer coefficient with and without the polymer coating. You should see a cross-over point
where it becomes disadvantageous to use the polymer coating; explain this.
The heat transfer into the different boundaries can be evaluated by summing the heat transfer to
the boundary segments that lie on that boundary. The heat transfer from the water and the air is
set to 0 and then each boundary segment is considered one by one:
q_dot_w=0;
q_dot_a=0;
for b=1:N_b
The indices of the nodes that define the boundary segment and the boundary that the segment lies
on are obtained from the edge matrix:
If the boundary segment lies on boundary 1 then it is exposed to water and the heat transfer into
that boundary is added to the total heat transfer to the water. The heat transfer is the product of
the length of the segment, the heat transfer coefficient and the difference between the adjacent
water temperature and the average temperature on the boundary.
if(bndry==1)
hb=bcm(1,bndry); % heat transfer coefficient (W/m^2-K)
Tinfb=bcm(2,bndry); % ambient temperature (K)
T_b=(T(i)+T(j))/2; % average temperature on boundary (K)
q_dot_w=q_dot_w+sij*hb*(Tinfb-T_b); % rate of heat transfer per unit length (W/m)
end
If the boundary segment lies on boundary 3 then it is exposed to air and the heat transfer into that
boundary is added to the total heat transfer to the air.
if(bndry==3)
hb=bcm(1,bndry); % heat transfer coefficient (W/m^2-K)
Tinfb=bcm(2,bndry); % ambient temperature (K)
T_b=(T(i)+T(j))/2; % average temperature on boundary (K)
q_dot_a=q_dot_a+sij*hb*(Tinfb-T_b); % rate of heat transfer per unit length (W/m)
end
end
The heat flux from the water is the heat transfer over the area:
which leads to 570.6 W/m2. The function P2p7d6_makefig runs the finite element solution over a
range of air-side heat transfer coefficients:
h_a=1;
for i=1:101
h_av(i)=h_a;
[q_dot_fluxv(i)]=P2p7d6(h_a);
h_a=h_a*1.08;
end
Figure 5 illustrates the heat flux as a function of air-side heat transfer coefficient predicted by the
finite element solution.
7000
100
10
1 10 100 1000 2250
2
Air-side heat transfer coefficient (W/m -K)
Figure 5: Heat flux as a function of air-side heat transfer coefficient with the polymer fins and without the
polymer (bare tube).
If the polymer is removed, then the problem is a 1-D conduction problem that can be solved in
EES. The inputs are entered:
"Inputs"
{h_bar_a=10 [W/m^2-K]} "heat transfer coefficient on air-side"
s=0.015 [m] "width of tube"
h_bar_w=250 [W/m^2-K] "heat transfer coefficient on water-side"
k_t=15 [W/m-K] "conductivity of tube"
th_t=0.008 [m] "tube thickness"
T_w=320 [K] "water temperature"
T_a=290 [K] "air temperature"
The resistance to conduction through the tube, per unit area, is:
tht
Rtube =
kt
The resistance to convection on the air- and water-sides, per unit area, are computed according
to:
1
Rconv ,a =
ha
1
Rconv , w =
hw
and the heat flux associated with the bare tube is:
q =
(Tw Ta )
Rtotal
The heat flux associated with the bare tube (without the polymer fins) as a function of the air-
side heat transfer coefficient is shown in Figure 5. The addition of the polymer fins adds a
resistance to conduction through the polymer but reduces the resistance to air-side convection. If
the resistance to air-side convection is sufficiently small, then the reduction of the air-side
convetion resistance is not sufficient to make up for the additional resistance due to conduction
through the polymer. Therefore, there is a cross-over point in Figure 5.
Problem 2.7-7
Figure P2.7-7 illustrates a power electronics chip that is used to control the current to a winding
of a motor.
ha = 10 W/m -K
silicon chip 2
ks = 80 W/m-K Ta = 20C
spreader
k = 45 W/m-K
0.5 cm sp
0.2 cm
0.5 cm
0.1 cm 1.4 cm
dielectric layer
hw = 1000 W/m -K
2
kd = 1 W/m-K
Tw = 10C
Figure P2.7-7: Power electronics chip.
The silicon chip has dimensions 0.2 cm by 0.5 cm and conductivity ks = 80 W/m-K. A
generation of thermal energy occurs due to losses in the chip; the thermal energy generation can
be modeled as being uniformly distributed with a value of g = 1x108 W/m3 in the upper 50% of
the silicon. The chip is thermally isolated from the spreader by a dielectric layer with thickness
0.1 cm and conductivity kd = 1 W/m-K. The spreader has dimension 0.5 cm by 1.4 cm and
conductivity ksp = 45 W/m-K. The external surfaces are all air cooled with ha = 10 W/m 2 -K and
Ta = 20C except for the bottom surface of the spreader which is water cooled with
hw = 1000 W/m 2 -K and Tw = 10C.
a.) Develop a numerical model of the system using FEHT.
FEHT is opened and Heat Transfer is selected from the Subject menu. Steady-state and
Temperatures in C are selected from the setup menu and a grid of dimension 0.2 cm x 0.2 cm is
specified. A rough geometry is obtained using the Outline selection from the Draw menu, as
shown in Figure 2.
The precise coordinates of each point are specified by double-clicking on each one in turn. The
result is shown in Figure 3.
Figure 3: Geometry completely specified.
The material properties are specified by selecting an outline and then selecting Material
Properties from the Specify menu. The boundary conditions and generation are specified in the
same way. A very crude mesh is obtained by selecting Element Lines from the Draw menu.
The result is shown in Figure 4.
The problem is solved by selecting Calculate from the Run menu. The temperatures are shown
at each node by selecting Temperatures from the View menu, as shown in Figure 5. The
maximum temperature is 153C.
Figure 5: Temperature predicted at each node.
b.) Plot the maximum temperature in the system as a function of the number of nodes.
The mesh is successively refined (by selecting Reduce Mesh from the Draw menu. Figure 6
illustrates a highly reduced mesh and Figure 7 illustrates the maximum temperature in the
silicon as a function of the number of nodes in the mesh.
154.2
153.8
153.6
153.4
153.2
153
0 100 200 300 400 500 600
Number of nodes
Figure 7: Maximum temperature as a function of the number of nodes.
c.) Develop a simple sanity check of your results using a resistance network.
Figure 8 shows that the temperature is reduced primarily by conduction across the dielectric and
convection from the surface of the spreader to the water. The temperature gradients caused by
conduction within the spreader and silicon are small. These observations are consistent with a
simple resistance network in which the heat transfer related to generation:
0.002 m mK K
Rcond , s = = 0.005 (2)
0.005 m 1 m 80 W W
0.001 m mK K
Rcond ,d = = 0.2 (3)
0.005 m 1 m 1 W W
0.005 m mK K
Rcond , sp = = 0.008 (4)
0.014 m 1 m 45 W W
m2 K K
Rconv , w = = 0.071 (5)
1000 W 0.014 m 1 m W
The resistance to conduction through the dielectric and convection to the water are important,
the others are not. The temperature difference between the silicon and the spreader is
approximately:
0.2 K 500 W
Td = Rcond , d q = = 100 K (6)
W
and the temperature difference between the water and the spreader is approximately:
0.071 K 500 W
Tconv ,v = Rconv , w q = = 35.7 K (7)
W
These values are consistent with the FEHT predictions, providing some confidence in the
results.
Problem 2.7-8
Figure P2.7-8(a) illustrates a heat exchanger in which hot fluid and cold fluid flows through
alternating rows of square channels that are installed in a piece of material.
C C C C C C
H H H H H H
unit cell of
C C C C C C
heat exchanger
H H H H H H
You are analyzing this heat exchanger and will develop a model of the unit cell shown in Figure
P2.7-8(a) and illustrated in more detail in Figure 2.7-8(b).
L = 6 mm h = 150 W/m -K
2
Th = 80C
th/2 =0.4 mm
p/2 = 4 mm
thb = 0.8 mm
h = 150 W/m -K
2
Tc = 25C k = 12 W/m-K
Figure 2.7-8(b): Details of unit cell shown in Figure 2.7-8(a).
The metal struts separating the square channels form fins. The length of the fin (the half-width
of the channel) is L = 6 mm and the fin thickness is th = 0.8 mm. The thickness of the material
separating the channels is thb = 0.8 mm. The distance between adjacent fins is p = 8 mm. The
channel structure for both sides (hot and cold) are identical. The conductivity of the metal is k =
12 W/m-K. The hot fluid has temperature Th = 80C and heat transfer coefficient h = 150
W/m2-K. The cold fluid has temperature Tc = 25C and the same heat transfer coefficient.
a.) Prepare a numerical model of the unit cell shown in Figure 2.7-8(b) using FEHT.
The geometry is entered in FEHT using the Outline selection from the draw menu. Each point is
precisely positioned by double-clicking on it and entering its coordinates. The result is shown in
Figure 3.
Figure 3: Geometry entered in FEHT.
The material properties and boundary conditions are entered. Note that the lines of symmetry
(i.e., the lines at the ends and centers of the fins) are adiabatic. A crude mesh is drawn. The
result is shown in Figure 4.
b.) Plot the rate of heat transfer from the hot fluid to the cold fluid within the unit cell as a
function of the number of nodes.
The boundary adjacent to the cold fluid is selected and Heat Flows is selected from the View
menu. The mesh is successively refined by selecting Reduce Mesh from the Draw menu and the
heat flow as a function of the number of nodes is shown in Figure 6.
32.2
32
31.6
31.4
31.2
31
0 100 200 300 400 500 600 700
Number of nodes
Figure 6: Heat transfer rate (per length of heat exchanger) as a function of the number of nodes.
c.) Develop a simple model of the unit cell using a resistance network and show that your result
from (b) makes sense.
The fin efficiency (f) for an adiabatic tipped constant cross-section fin is obtained using the
function eta_fin_straight_rect in EES. The resistance of each of the fins is:
1
Rf = (1)
h LW f
where W is a unit length of heat exchanger and only the area of the fin within the unit cell is
used. The resistance of the unfinned region of the base is:
2
Ruf = (2)
h ( p th ) W
q =
(Th Tc ) (4)
Rtotal
which leads to q = 32.6. Note that this is somewhat larger than the prediction from (b) because
the conduction resistance of the base (both across the base and along the base) is not considered.
Problem 2.8-1 (2-12 in text): Cryogenic Thermal Switch
There are several cryogenic systems that require a thermal switch, a device that can be
used to control the thermal resistance between two objects. One class of thermal switch
is activated mechanically and an attractive method of providing mechanical actuation at
cryogenic temperatures is with a piezoelectric stack; unfortunately, the displacement
provided by a piezoelectric stack is very small, typically on the order of 10 microns. A
company has proposed an innovative design for a thermal switch, shown in Figure P2.8-
1(a). Two blocks are composed of th = 10 m laminations that are alternately copper (kCu
= 400 W/m-K) and plastic (kp = 0.5 W/m-K). The thickness of each block is L = 2.0 cm
in the direction of the heat flow. One edge of each block is carefully polished and these
edges are pressed together; the contact resistance associated with this joint is Rc = 5x10-4
K-m2/W.
th = 10 m plastic laminations
kp = 0.5 W/m-K
L = 2 cm L = 2 cm direction of actuation
TH TC
Figure P2.8-1(b)
on position off position
contact resistance, Rc = 5x10 m -K/W
-4 2
th = 10 m copper laminations
kCu = 400 W/m-K
Figure P2.8-1(a): Thermal switch in the on and off positions.
Figure P2.8-1(a) shows the orientation of the two blocks when the switch is in the on
position; notice that the copper laminations are aligned with one another in this
configuration which provides a continuous path for heat through high conductivity
copper (with the exception of the contact resistance at the interface). The vertical
location of the right-hand block is shifted by 10 m to turn the switch "off". In the off
position, the copper laminations are aligned with the plastic laminations; therefore, the
heat transfer is inhibited by low conductivity plastic. Figure P2.8-1(b) illustrates a closer
view of half (in the vertical direction) of two adjacent laminations in the on and off
configurations. Note that the repeating nature of the geometry means that it is sufficient
to analyze a single lamination set and assume that the upper and lower boundaries are
adiabatic.
L = 2 cm L = 2 cm
TH TC
on position
TH TC
off position
Figure P2.8-1(b): A single set consisting of half of two adjacent laminations in the on and "off
positions.
The key parameter that characterizes a thermal switch is the resistance ratio (RR) which is
defined as the ratio of the resistance of the switch in the off position to its resistance in
the on position. The company claims that they can achieve a resistance ratio of more
than 100 for this switch.
a) Estimate upper and lower bounds for the resistance ratio for the proposed thermal
switch using 1-D conduction network approximations. Be sure to draw and clearly
label the resistance networks that are used to provide the estimates. Use your results
to assess the companys claim of a resistance ratio of 100.
"Inputs"
th = 10 [micron]*convert(micron,m) "thickness of laminations"
k_Cu=400 [W/m-K] "conductivity of copper laminations"
k_p=0.5 [W/m-K] "conductivity of plastic laminations"
L = 2.0 [cm]*convert(cm,m) "lengt of laminations"
R``_c=5e-4 [K-m^2/W] "area specific contact resistance of interface"
W=1 [m] "unit depth into page"
The isothermal and adiabatic models provide two limiting cases. Both result in a 1-D
problem that can be represented using a resistance network. The adiabatic
approximation does not allow heat transfer in the y-direction (i.e., perpendicular to the
laminations). The resistance network for the adiabatic approximation is shown in Figure
3(a) for the on position and in Figure 3(b) for the off position.
(a) (b)
Figure 3: Resistance network for the adiabatic approximation in the (a) on and (b) off
positions.
Note that the parameter Alam in Figure 3 is the cross-sectional area associated a single
lamination:
Alam = W th
where W is the depth into the page (assumed to be 1 m). The resistance associated with
the adiabatic approximation in the on state, Rad,on from Figure 3(a) is:
1
2 1 1
Rad ,on = +
Alam 2 L + R 2 L + R
k c
kCu
c
p
which leads to Rad,on = 119.1 K/W. The resistance associated with the adiabatic
approximation in the off state, Rad,off from Figure 3(b) is:
1
L L
+ + Rc
2 1 1 k p kCu
Rad ,off = + =
Alam L + L + R L + L + R Alam
k c c
p kCu kCu k p
Rad ,off
RRad =
Rad ,on
RR_ad=R_ad_off/R_ad_on "resistance ratio estimated using adiabatic limit"
(a) (b)
Figure 4: Resistance network for the isothermal approximation in the (a) on and (b) off
positions.
By inspection, the two resistance networks shown in Figures 4(a) and 4(b) will yield the
same resistance and therefore the isothermal assumption will predict a resistance ratio,
RRiso, of 1.0. Clearly the companys claim of a resistance ratio of 100 is not possible as it
does not lie between the two bounding quantities associated with the isothermal and
adiabatic approximations.
b) Provide one or more suggestions for design changes that would improve the
performance of the switch (i.e., increase the resistance ratio). Justify your
suggestions.
The resistance ratio would be increased by any change that causes the two resistance
networks in Figure 3 to be more different. Possible improvements include using
materials with a larger ratio of conductivities (i.e., lower conductivity plastic and, to a
lesser degree, higher conductivity copper) or eliminating the contact resistance. From a
more practical standpoint, any design change that causes the actual the device to behave
more like the adiabatic approximation in Figure 3 than the isothermal approximation in
Figure 4 will improve the performance; for example, increasing the contact resistance
between adjacent laminations would be important.
c.) Sketch the temperature distribution through the two parallel paths associated with the
adiabatic limit of the switchs operation in the off position. Do not worry about the
quantitative details of the sketch, just make sure that the qualitative features are
correct.
Figure 5: Qualitative temperature distribution through paths 1 and 2 consistent with the adiabatic
approximation when the switch is in the off position.
d.) Sketch the temperature distribution through the two parallel paths associated with the
adiabatic limit in the on position. Again, do not worry about the quantitative details
of your sketch, just make sure that the qualitative features are correct.
Figure 6: Qualitative temperature distribution through paths 1 and 2 consistent with the adiabatic
approximation when the switch is in the on position.
Problem 2.8-2 (2-13 in text): Resistance of a Bus Bar
Figure P2.8-2 illustrates a thermal bus bar that has width W = 2 cm (into the page).
H1 = 5 cm L2 = 7 cm
TH = 80C h = 10 W/m -K
2
T = 20C
k = 1 W/m-K
L1 = 3 cm H2 = 1 cm
Figure P2.8-2: Thermal bus bar.
The bus bar is made of a material with conductivity k = 1 W/m-K. The middle section is L2 = 7
cm long with thickness H2 = 1 cm. The two ends are each L1 = 3 cm long with thickness H1 = 3
cm. One end of the bar is held at TH = 80C and the other is exposed to air at T = 20C with h
= 10 W/m2-K.
a.) Use FEHT to predict the rate of heat transfer through the bus bar.
The geometry is entered in FEHT, a mesh is generated, and the boundary conditions and material
properties are specified (Figure 2).
The mesh is refined several times and then the heat transfer at the convective boundary is
obtained; this leads to q = 5.40 W/m or q = 0.108 W.
b.) Obtain upper and lower bounds for the rate of heat transfer through the bus bar using
appropriately defined resistance approximations.
"Inputs"
L_1=3 [cm]*convert(cm,m) "length of edge pieces"
H_1=5 [cm]*convert(cm,m) "height of edge pieces"
L_2=7 [cm]*convert(cm,m) "length of center piece"
H_2=1 [cm]*convert(cm,m) "height of center piece"
W=2 [cm]*convert(cm,m) "width"
k=1 [W/m-K] "conductivity"
h_bar=10 [W/m^2-K] "heat transfer coefficient at right side"
T_infinity=20 [C] "ambient temperature at right side"
T_H=80 [C] "left side temperature"
An upper bound on the heat transfer rate is obtained using the isothermal limit. The isothermal
limit is calculated according to:
qiso =
(TH T )
1 L1 L2
+2 +
h H1 W k H1 W k H 2 W
"Isothermal limit"
R_conv_iso=1/(h_bar*W*H_1)
R_cond_1_iso=L_1/(k*W*H_1)
R_cond_2_iso=L_2/(k*W*H_2)
R_total_iso=(R_conv_iso+2*R_cond_1_iso+R_cond_2_iso)
q_dot_iso=(T_H-T_infinity)/R_total_iso
A lower bound on the heat transfer rate is obtained using the adiabatic limit. The adiabatic limit
is calculated according to:
qad =
(TH T )
1 L1 L2
+2 +
h H2 W k H2 W k H2 W
"Adiabatic limit"
R_conv_ad=1/(h_bar*W*H_2)
R_cond_1_ad=L_1/(k*W*H_2)
R_cond_2_ad=L_2/(k*W*H_2)
R_total_ad=R_conv_ad+2*R_cond_1_ad+R_cond_2_ad
q_dot_ad=(T_H-T_infinity)/R_total_ad
superconducting strands,
km,normal = 50 W/m-K
km,sc = 0.2 W/m-K
magnet
a = 5 mm TC b = 1 mm
TH
b = 1 mm polymer matrix
kp = 2.5 W/m-K
Figure P2.8-3: Superconducting heat switch.
The heat switch is made by embedding eight square superconducting strands in a polymer
matrix. The width of switch is W = 10 mm (into the page). The size of the strands are a = 5 mm
and the width of polymer that surrounds each strand is b = 1 mm. The conductivity of the
polymer is kp = 2.5 W/m-K. The heat switch is surrounded by a magnet. When the heat switch
is on (i.e., the thermal resistance through the switch in the x-direction is low, allowing heat flow
from TH to TC), the magnet is on. Therefore, the magnetic field tends to drive the
superconductors to their normal state where they have a high thermal conductivity, km,normal = 50
W/m-K. To turn the heat switch off (i.e., to make the thermal resistance through the switch high,
preventing heat transfer from TH to TC), the magnet is deactivated. The superconductors return to
their superconducting state, where they have a low thermal conductivity, km,sc = 0.2 W/m-K. The
edges of the switch are insulated.
a.) Develop a model using a 1-D resistance network that provides a lower bound on the
resistance of the switch when it is in its off state (i.e., km = km,sc).
"Inputs"
W=10 [mm]*convert(mm,m) "width of switch"
a=5 [mm]*convert(mm,m) "size of strands"
b=1 [mm]*convert(mm,m) "width of polymer layer"
k_p=2.5 [W/m-K] "polymer conductivity"
k_m_normal=50 [W/m-K] "metal conductivity in its normal state"
k_m_sc=0.2 [W/m-K] "metal conductivity in its superconducting state"
The isothermal limit provides a lower bound on the resistance. The isothermal limit is shown in
Figure P2.8-3-2.
Riso,2
Riso,1
TH TC
Riso,3
Figure P2.8-3-2: Isothermal limit resistance network.
The resistance Riso,1 represents the conduction through the polymer layers that stretch across the
width of the switch:
5b
Riso ,1 = (1)
k p W ( 2 a + 3b)
The resistance Riso,2 represents the conduction through the polymer layers that stretch between
adjacent superconducting fibers:
4a
Riso ,2 = (2)
k p W 3b
The resistance Riso,3 represents the conduction through the superconducting fibers:
4a
Riso ,3 = (3)
km W 2 a
"isothermal model"
R_iso_1=5*b/(k_p*W*(2*a+3*b)) "conduction through polymer layers"
R_iso_2=4*a/(k_p*W*3*b) "conduction through polymer between superconducting strands"
R_iso_3=4*a/(k_m*W*2*a) "conduction through superconducting strands"
The adiabatic limit provides an upper bound on the resistance. The adiabatic limit is shown in
Figure P2.8-3-3.
Rad,2 Rad,3
TH TC
Rad,1
Figure P2.8-3-3: Adiabatic limit resistance network.
The resistance Rad,1 represents the conduction through the polymer layers that stretch across the
entire length of the switch in the x-direction:
Rad ,1 =
(4 a + 5b) (5)
k p W 3b
The resistance Rad,2 represents the conduction through the polymer layers that lie between
adjacent superconducting fibers in the x-direction:
5b
Rad ,2 = (6)
kp W 2 a
The resistance Rad,3 represents the conduction through the superconducting fibers:
4a
Rad ,3 = (7)
km W 2 a
"adiabatic model"
R_ad_1=(4*a+5*b)/(k_p*W*3*b) "conduction through polymer layers in x-direction"
R_ad_2=5*b/(k_p*W*2*a) "conduction through polymer between strands"
R_ad_3=4*a/(k_m*W*2*a) "conduction through superconducting strands"
1
1 1
Rad = + (8)
Rad ,1 ( Rad ,2 + Rad ,3 )
Figure P2.8-3-4 illustrates Riso and Rad as a function of km. Notice that Rad is always larger than
Riso and that regardless of which model you use, the resistance of the switch decreases
substantially when the superconducting strands go from their superconducting to their normal
state.
500
adiabatic limit
Thermal resistance (K/W)
100
isothermal limit
10
0.2 1 10 50
Conductivity of strands (W/m-K)
Figure P2.8-3-4: Thermal resistance as a function of km.
d.) The performance of a heat switch is provided by the resistance ratio; the ratio of the
resistance of the switch in its off state to its resistance in the on state. Use your model to
provide an upper and lower bound on the resistance ratio of the switch.
Setting km = km,sc leads to Riso = 225.9 K/W and Rad = 251.2 K/W. Setting km = km,normal leads Riso
= 19.33 K/W and Rad = 22.39 K/W. Therefore, the maximum possible value of the resistance
ratio is 251.2/19.33 = 12.995 and the minimum possible value of the resistance ratio is
225.9/22.39 = 10.089.
e.) Plot the ratio of your answer from part (b) to your answer from part (a) as a function of km for
km,sc < km < km,normal. Explain the shape of your plot.
1.14
1.12
1.1
1.08
1.06
1.04
1.02
1
0.2 1 10 70
Strand conductivity (W/m-K)
Figure P2.8-3-5: Thermal resistance as a function of km.
Notice that the adiabatic and isothermal models predict the same value when km = kp (i.e., Rad/Riso
= 1 at km = 2.5 W/m-K in Figure P2.8-3-5). This is because the problem is 1-D in this limit and
therefore both resistance network representations are exactly correct.
Problem 2.9-1: Composite Equivalent Conductivity
A composite material is formed from laminations of high conductivity material (khigh = 100
W/m-K) and low conductivity material (klow = 1 W/m-K) as shown in Figure P2.9-1. Both
laminations have the same thickness, th.
low conductivity laminations
klow = 1 W/m-K
th
th
y
a.) Do you expect the equivalent conductivity of the composite to be higher in the x or y
directions? Note from Figure P2.9-1 that the x-direction is parallel to the laminations while
the y direction is perpendicular to the laminations.
The heat can flow entirely through high conductivity laminations in the x direction whereas it
must pass through the low conductivity laminations in order to travel in the y direction.
Therefore, the equivalent conductivity of the composite will be much higher in the x direction.
b.) Estimate the equivalent conductivity of the composite in the x direction. You should not
need to any calculations to come up with a good estimate for this quantity.
You are essentially forcing the heat to flow through a block of high conductivity material with
half the area of the composite in order to move in the x direction. Therefore, the equivalent
conductivity of the composite in the x direction will be approximately half the conductivity of
the high conductivity laminations, keff,x ~ 50 W/m-K.
Problem 2.9-2 (2-14 in text)
A laminated stator is shown in Figure P2.9-2. The stator is composed of laminations with
conductivity klam = 10 W/m-K that are coated with a very thin layer of epoxy with conductivity
kepoxy = 2.0 W/m-K in order to prevent eddy current losses. The laminations are thlam = 0.5 mm
thick and the epoxy coating is 0.1 mm thick (the total amount of epoxy separating each
lamination is thepoxy = 0.2 mm). The inner radius of the laminations is rin= 8.0 mm and the outer
radius of the laminations is ro,lam = 20 mm. The laminations are surrounded by a cylinder of
plastic with conductivity kp = 1.5 W/m-K that has an outer radius of ro,p = 25 mm. The motor
casing surrounds the plastic. The motor casing has an outer radius of ro,c = 35 mm and is
composed of aluminum with conductivity kc = 200 W/m-K.
h = 40 W/m -K
2
Rc = 1x10 K-m /W
-4 2
rin = 8 mm
ro,lam = 20 mm
ro,p = 25 mm
ro,c = 35 mm
Figure P2.9-2: Laminated stator.
A heat flux associated with the windage loss associated with the drag on the shaft is q = 5x104
W/m2 is imposed on the internal surface of the laminations. The outer surface of the motor is
exposed to air at T = 20C with a heat transfer coefficient h = 40 W/m2-K. There is a contact
resistance Rc = 1x10-4 K-m2/W between the outer surface of the laminations and the inner
surface of the plastic and the outer surface of the plastic and the inner surface of the motor
housing.
a.) Determine an upper and lower bound for the temperature at the inner surface of the
laminations (Tin).
An upper bound on the temperature allows no heat spreading; therefore, the heat must take two
parallel paths that pass through the iron and the epoxy. The total resistance associated with one
lamination/epoxy pair is:
1 1
=
Rtotal ,ad r r r
ln o ,lam ln o , p ln o ,c
rin + Rc
+
ro ,lam
+
Rc
+ ro , p + 1
2 klam thlam 2 ro ,lam thlam 2 k p thlam 2 ro , p thlam 2 kc thlam 2 ro, p thlam h
1
+
r r r
ln o ,lam ln o , p ln o ,c
r
rin + Rc ro ,lam Rc o, p + 1
+ + +
2 kepoxy thepoxy 2 ro ,lam thepoxy 2 k p thepoxy 2 ro , p thepoxy 2 kc thepoxy 2 ro, p thepoxy h
(1)
b.) You need to reduce the internal surface temperature of the laminations and there are a few
design options available, including: (1) increase the lamination thickness (up to 0.7 mm), (2)
reduce the epoxy thickness (down to 0.05 mm), (3) increase the epoxy conductivity (up to 2.5
W/m-K), or (4) increase the heat transfer coefficient (up to 100 W/m-K). Which of these
options do you suggest and why?
Examination of the Solution Window (Figure 2) shows that the resistance of the lamination is
much less than the resistance of the epoxy; therefore, the resistance of the lamination dominates
the resistance of the epoxy. The resistance due to convection is much larger than the resistance
of the lamination, contact resistance, or conduction through the plastic and casing. Therefore,
the resistance to convection dominates the problem and the most effective mechanism for
reducing the temperature is to increase the heat transfer coefficient.
w = 0.2 mm
w = 0.2 mm
s = 1.0 mm
L = 10 cm
s = 1.0 mm
km = 100 W/m-K
kp = 2 W/m-K
z
b = 2 cm
x
y
Figure P2.9-3: Composite material.
The metal bars are square with edge width w = 0.2 mm and are aligned with the z-direction. The
bars are arrayed in a regularly spaced matrix with a center-to-center distance of s = 1.0 mm. The
conductivity of the metal is km = 100 W/m-K. The length of the material in the z direction is L =
10 cm. The polymer fills the space between the bars and has a thermal conductivity kp = 2.0
W/m-K. The cross-section of the material in the x-y plane is square with edge width b = 2.0 cm.
a.) Determine the effective conductivity in the x- , y- , and z-directions.
"Inputs"
w = 0.2 [mm]*convert(mm,m) "size of metal bars"
s = 1.0 [mm]*convert(mm,m) "center-to-center distance between bars"
k_m=100 [W/m-K] "metal conductivity"
L = 10 [cm]*convert(cm,m) "length of composite"
k_p=2.0 [W/m-K] "polymer conductivity"
b=2.0 [cm]*convert(cm,m) "size of composite"
The effective conductivity in the x and y directions will be the same as the energy must be
transferred through the same structure. The effective conductivity is calculated by considering a
unit cell of the structure, shown in Figure 2.
Figure 2: Unit cell of the structure in the x or y direction
The conduction through the unit cell can be approximated using three resistances, shown in
Figure 2. These resistances are:
w
Rm, x = (1)
km L w
R p ,1, x =
( s w) (2)
kp L w
s
R p ,2, x = (3)
k p L ( s w)
1
1 1
Reff , x = + (4)
Rm, x + R p ,1, x R p ,2, x
The resistance must be equal to the resistance of a homogeneous material with the same size that
has the effective thermal conductivity, keff,x:
s
= Reff , x (5)
keff , x L s
and indicate that keff,x= keff,y = 2.1 W/m-K (the matrix behaves essentially like a polymer with
respect to conduction in the x or y directions because the energy must transfer, primarily, through
the polymer).
The same type of analysis is accomplished for conduction in the z direction using the same unit
cell shown in Figure 2. Conduction in the z direction occurs through the metal bar and
surrounding polymer in parallel. The resistances of these two paths are:
L
Rm , z = (6)
k m w2
L
Rp, z = (7)
k p ( s w2 )
2
1
1 1
Reff , z = + (8)
Rm, z R p , z
The resistance must be equal to the resistance of a homogeneous material with the same size that
has the effective thermal conductivity, keff,z:
L
= Reff , z (9)
keff , z s 2
b.) The outer edges of the material are insulated and the faces of the material at z = 0 and z = L
are exposed to a convective boundary condition with h = 10 W/m2-K. Is it appropriate to
treat this problem as a lumped capacitance problem?
The Biot number is the ratio of the conductive resistance associated with getting energy from
within the structure to the surface (Rcond) to the convective resistance associated with transferring
the energy from the surface to the surrounding fluid (Rconv). The conduction resistance is
associated with transferring energy from z = L/2 to z = L:
L/2
Rcond , z = (10)
keff , z b 2
1
Rconv , z = (11)
h b2
Rcond , z
Bi = (12)
Rconv , z
which leads to a Biot number of 0.08; the lumped capacitance assumption is justified for this
problem.
Problem 2.1-1 (2-1 in text): Buried Tubes
Figure P2.1-1 illustrates two tubes that are buried in the ground behind your house that transfer
water to and from a wood burner. The left hand tube carries hot water from the burner back to
your house at Tw,h = 135F while the right hand tube carries cold water from your house back to
the burner at Tw,c = 70F. Both tubes have outer diameter Do = 0.75 inch and thickness th =
0.065 inch. The conductivity of the tubing material is kt = 0.22 W/m-K. The heat transfer
coefficient between the water and the tube internal surface (in both tubes) is hw = 250 W/m2-K.
The center to center distance between the tubes is w = 1.25 inch and the length of the tubes is L =
20 ft (into the page). The tubes are buried in soil that has conductivity ks = 0.30 W/m-K.
ks = 0.30 W/m-K
Tw,c = 70F
th = 0.065 inch kt = 0.22 W/m-K
hw = 250 W/m -K
2
Tw,h = 135F
hw = 250 W/m -K
2
w = 1.25 inch
Do = 0.75 inch
Figure P2.1-1: Tubes buried in soil.
a.) Estimate the heat transfer from the hot water to the cold water due to their proximity to one
another.
The heat transfer is resisted by convection within the tube, conduction through the tube, and
conduction in the soil. The convection resistance is calculated according to:
1
Rconv = (1)
hw L ( Do 2 th )
The conduction resistance associated with the tube is calculated according to:
Do
ln
Do 2 th
Rcond = (2)
2 kt L
The shape factor for parallel, buried tubes (SF) is obtained using EES internal library and used
to compute the resistance due to conduction through the soil:
1
Rtubetotube = (3)
SF ks
Tw,h Tw,c
q = (4)
2 Rconv + 2 Rcond + Rtubetotube
b.) To do part (a) you should have needed to determine a shape factor; calculate an approximate
value of the shape factor and compare it to the accepted value.
The shape factor can be thought of as the ratio of the effective area for conduction, Aeff, to the
effective length required for conduction, Leff. The effective area for conduction can be estimated
according to:
Aeff w L (5)
Leff w Do (6)
Aeff
SFapp (7)
Leff
The exact value of the shape factor returned by the function is SF = 17.4 m while the
approximate value is SFapp = 15.2 m; these are sufficiently close for a sanity check.
c.) Plot the rate of heat transfer from the hot water to the cold water as a function of the center to
center distance between the tubes.
The heat transfer rate is shown in Figure 2 as a function of the center to center distance between
the tubes.
Figure 2: Heat transfer rate as a function of the center to center distance between the tubes
Problem 2.1-3 (2-2 in text)
A solar electric generation system (SEGS) employs molten salt as both the energy transport and
storage fluid. The molten salt is heated to Tsalt = 500C and stored in a buried semi-spherical
tank. The top (flat) surface of the tank is at ground level. The diameter of the tank before
insulation is applied Dt = 14 m. The outside surfaces of the tank are insulated with tins = 0.30 m
thick fiberglass having a thermal conductivity of kins = 0.035 W/m-K. Sand having a thermal
conductivity of ksand = 0.27 W/m-K surrounds the tank, except on its top surface. Estimate the
rate of heat loss from this storage unit to the Tair = 25C surroundings. You may neglect the
resistance due to convection.
$UnitSytem SI K J Pa
$TabStops 0.2 0.4 3.5 in
"known information"
t_ins=0.3 [m] "insulation thickness"
D_t=14 [m] "tank diameter"
k_sand=0.27 [W/m-K] "sand thermal conductivity"
k_ins=0.035 [W/m-K] "insulation thermal conductivity"
T_air=convertTemp(C,K,25 [C]) "air temperature in K"
T_salt=converttemp(C,K,500 [C]) "salt temperature in K"
The resistance between the bottom (buried, hemispherical) surface of the tank and the air is due
to the conduction through the spherical shell of insulation
1 1
+
Dt / 2 ( Dt / 2 + tins )
Rins ,bottom = (1)
2 kins
1
Rsand = (2)
k sand S
where S is the shape factor, obtained using the EES function SF_22. The resistance from the top,
unburied surface of the tank is only due to conduction through the insulation:
tins
Rins ,top = (3)
D2
kins t
4
The heat transfer from the bottom and top surfaces are calculated according to:
qbottom =
(Tsalt Tair ) (4)
Rsoil + Rins ,bottom
and
qtop =
(Tsalt Tair ) (5)
Rins ,top
R_ins_bottom=(1/(D_t/2)-1/(D_t/2+t_ins))/(2*pi*k_ins)
"resistance to conduction through insulation on bottom"
S=SF_22(D_t+2*t_ins) "shape factor"
R_sand=1/(k_sand*S) "resistance of sand"
R_ins_top=t_ins/(k_ins*pi*D_t^2/4)
"resistance to conduction through insulation on top"
q_dot_bottom=(T_salt-T_air)/(R_sand+R_ins_bottom) "heat loss from bottom surface of tank"
q_dot_top=(T_salt-T_air)/(R_ins_top) "heat loss from top surface of tank"
q_dot_total=q_dot_bottom+q_dot_top "total heat loss"
a = 20 mm
b = 15 mm
The nominal design of an individual heater plate utilizes metal with k = 20 W/m-K, th = 0.5 mm,
a = 20 mm, and b = 15 mm (note that a and b are defined as the half-width and half-height of the
heater plate, respectively, and th is the thickness as shown in Figure P2-3). The heater bar
temperature is maintained at Th = 40C and the fluid temperature is T = 20C. The nominal
value of the average heat transfer coefficient is h = 50 W/m2-K.
a.) Develop an analytical model that can predict the temperature distribution in the plate under
these nominal conditions.
The problem can be simplified and tackled using the quarter-symmetry model of a single plate,
shown in Figure P2.2-3, is specified mathematically in Figure 2(a).
(a)
(b)
Figure 2: Quarter symmetry model of a plate.
The lines of symmetry (x =0 and y = 0) are adiabatic, the top edge (y = b) is held at the heater
temperature (Th) and the right edge (x = a) is held at the fluid temperature (T). These boundary
conditions are expressed below:
T
=0
x x =0
Tx =a = T
T
=0
y y =0
Ty =b = Th
The differential control volume and associated first law balance are shown in Figure 2(a):
q x + q y = q x + dx + q y + dy + qconv
q x q y
0= dx + dy + qconv
x y
Substituting the rate equations into the energy balance leads to:
T T
0= k th dy dx + k th dx dy + 2 dx dy h ( T T )
x
x y y
or
2T 2T 2 h
+
x 2 y 2 k th
T T = 0 ( ) (1)
= T T (2)
Substituting Eq. (2) into Eq. (1) transforms the governing partial differential equation into a
homogeneous equation; i.e., any multiple of will satisfy transformed partial differential
equation:
2 2
+ 2 m2 = 0 (3)
x 2
y
where
2h
m=
k th
Substituting Eq. (2) into the boundary conditions leads to:
=0 (4)
x x =0
x =a = 0 (5)
=0 (6)
y y =0
y =b = h (7)
where
h = Th T
The transformed problem is shown in Figure 2(b); note that the transformed problem is linear
and consists of a homogeneous partial differential equation and homogeneous boundary
conditions in the x-direction. Therefore, we can apply a separation of variables solution to the
problem and x is our homogeneous direction.
( x, y ) = X ( x ) Y ( y )
Substituting the product solution into the differential equation leads to:
d 2 X d 2 Y
Y + X m2 X Y = 0
dx 2
dy 2
d 2 X d Y
2
dx 2 + dy m 2 = 0
2
X Y
2 2
Recall that x is our homogeneous direction; therefore, we need the X group in the equation to
equal a negative constant (-2); the remaining part of the equation must equal the positive value
of the same constant (2). Therefore, the two ordinary differential equations that result from the
separation process are:
d 2 X
+ 2 X = 0 (8)
dx 2
and
d 2 Y
( 2 + m2 )Y = 0 (9)
dy 2
The next step is to solve the eigenproblem; the solution to the ordinary differential equation for
X is:
The 1st boundary conditions in the homogeneous direction, Eq. (4), leads to:
d X
=0 (11)
dx x =0
d X
= C1 sin ( 0 ) + C2 cos ( 0 ) = 0
dx x =0
=0 =1
which can only be true if C2 = 0. the 2nd boundary condition in the x-direction, Eq. (5), leads to:
x =a = C1 cos ( a ) = 0
which can only be true if the argument of the cosine is 0. Therefore, the eigenfunctions and
eigenvalues for the problem are:
The solution to the ordinary differential equation in the non-homogeneous direction, Eq. (9) can
be obtained using Maple:
> restart;
> assume((lambda^2+m^2),positive);
> ODEy:=diff(diff(thetaY(y),y),y)-(lambda^2+m^2)*thetaY(y)=0;
d
2
ODEy := 2 thetaY( y ) ( 2 + m~ 2 ) thetaY( y ) = 0
dy
> qYs:=dsolve(ODEy);
( 2 + m~2 y ) ( 2 + m~2 y )
qYs := thetaY( y ) = _C1 e + _C2 e
Note that Maple can convert this exponential form to an equivalent form involving hyperbolic
sines and cosines with the convert command.
> convert(qYs,'trigh');
thetaY( y ) = ( _C1 + _C2 ) cosh( 2 + m~ 2 y ) + ( _C1 _C2 ) sinh( 2 + m~ 2 y )
Yi = C3,i cosh ( )
i2 + m 2 y + C4,i sinh ( i2 + m 2 y )
The solution associated with the ith eigenvalue is therefore:
where the constant C1,i is absorbed into the constants C3,i and C4,i. It is good practice to verify
that Eq. (12) satisfies both boundary conditions in the x-direction and the partial differential
equation for any value of i:
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*a);
( 2 i~ 1 )
:=
2a
> theta:=(x,y)->cos(lambda*x)*(C_3*cosh(sqrt(lambda^2+m^2)*y)+C_4*sinh(sqrt(lambda^2+m^2)*y));
:= ( x, y ) cos( x ) ( C_3 cosh( 2 + m 2 y ) + C_4 sinh( 2 + m 2 y ) )
> eval(diff(theta(x,y),x),x=0);
0
> theta(a,y);
0
> diff(diff(theta(x,y),x),x)+diff(diff(theta(x,y),y),y)-m^2*theta(x,y);
( 2 i~ 1 ) x
cos
1
( 2 i~ 1 )
2 2
4 2a
( 2 i~ 1 ) 2 2 ( 2 i~ 1 ) 2 2
+ 4 m 2 y + 4 m 2 y
a 2
a 2
C_3 cosh + C_4 sinh
2 2
a 2 + cos
( 2 i~ 1 ) x
2a
( 2 i~ 1 ) 2 2
+ 4 m 2 y
1 a 2
( 2 i~ 1 ) 2 2
C_3 cosh + 4 m 2
4 2 a 2
( 2 i~ 1 )
2 2
+ m 2
y
4
1 a2 ( 2 i~ 1 ) 2 2
+ C_4 sinh
+ 4 m m 2
2
a
4 2 2
i~ x
cos
( 2 1 )
2 a
( 2 i~ 1 )2 2 ( 2 i~ 1 ) 2 2
+ m 2
y + 4 m 2 y
4
a 2
a 2
C_3 cosh + C_4 sinh
2 2
> simplify(%);
0
The general solution is the sum of the solution for each eigenvalue:
( ) ( )
= i = cos ( i x ) C3,i cosh i2 + m 2 y + C4,i sinh i2 + m 2 y (13)
i =1 i =1
The general solution must satisfy the boundary conditions in the non-homogeneous direction.
Substituting Eq. (13) into Eq. (6) leads to:
( ) ( )
= cos ( i x ) C3,i i + m sinh i + m 0 + C4,i i + m cosh i + m 0 = 0
2 2 2 2 2 2 2 2
y i =1
y =0
=0 =1
or
C
i =1
4,i i2 + m 2 cos ( i x ) = 0
( )
= Ci cosh i2 + m 2 y cos ( i x ) (14)
i =1
where the subscript 3 has been removed from C3,i because it is the only remaining undetermined
constant. Substituting Eq. (14) into Eq. (7) leads to:
( )
y =b = Ci cosh i2 + m 2 b cos ( i x ) = h
i =1
( )
a a
Orthogonality causes all of the terms in the summation to integrate to zero except for the one in
which i = j:
( ) cos ( x ) dx = cos ( x ) dx
a a
Ci cosh +m b
i
2 2 2
i h i
0 0
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*a);
( 2 i~ 1 )
:=
2a
> int((cos(lambda*x))^2,x=0..a);
a
2
> int(cos(lambda*x),x=0..a);
( 1 + i~ )
2 ( -1 ) a
( 2 i~ 1 )
h 4 ( 1)
1+ i
Ci =
cosh ( )
i2 + m 2 b ( 2 i 1)
"Inputs"
th_mm=0.5 [mm] "thickness of plate in mm"
th=th_mm * convert(mm,m) "thickness of plate"
k=20 [W/m-K] "conductivity of plate"
a_mm=20 [mm] "half-width of plate in mm"
a= a_mm * convert(mm,m) "half-width of plate"
b= 15 [mm] *convert(mm,m) "half-height of plate"
T_h = converttemp(C,K,40 [C]) "heater temperature"
T_infinity=converttemp(C,K,20 [C]) "fluid temperature"
h_bar=50 [W/m^2-K] "heat transfer coefficient"
"position"
x_bar=0.5
y_bar=0.5
x=x_bar*a
y=y_bar*b
A parametric table is created and used to generate the contour plot shown in Figure 3.
Figure 3: Contour plot of temperature on the plate.
b.) The measured quantity is the rate of heat transfer to the plate from the heater ( qh ) and
therefore the relationship between qh and h (the quantity that is inferred from the heater
power) determines how useful the instrument is. Determine the heater power.
The heater power can be computed by integrating the conduction heat transfer along the top
surface according to:
a
qh = 4 k th dx (15)
0
y y =b
where the factor of 4 comes from the quarter symmetry of the model. Equation (14) is
substituted into Eq. (15):
( ) cos ( x ) dx
a
qh = 4 k th Ci i2 + m 2 sinh i2 + m 2 b i
i =1 0
( )
qh = 2 k th a Ci i2 + m 2 sinh i2 + m 2 b (16)
i =1
"heater power"
duplicate i=1,N
q_dot_h[i]=2*k*th*a*C[i]*sqrt(lambda[i]^2+m^2)*sinh(sqrt(lambda[i]^2+m^2)*b)
end
q_dot_h=sum(q_dot_h[1..N])
c.) If the uncertainty in the measurement of the heater power is qh = 0.01 W, estimate the
uncertainty in the measured heat transfer coefficient ( h ).
A parametric table is used to explore the relationship between h and qhtr ; the two columns in the
table are the variables h_bar and q_dot_h. Figure 4 illustrates the heat transfer coefficient as a
function of the heater power, all else held constant.
Note that a good measuring device would show a strong relationship between the physical
quantity being measured ( h ) and the measurement ( qh ); this would be indicated by a large
partial derivative of heat transfer coefficient with respect to heater power. Given the uncertainty
in the heater power measurement, qh (related to, for example, the resolution of the data
acquisition system, noise, etc.), it is possible to estimate the uncertainty in the measurement of
the heat transfer coefficient, h , according to:
h
h = qh
qh
For example, if the uncertainty in the heater power is 0.01 W then Fig. 3 suggests that the partial
derivative of heat transfer coefficient with respect to heater power is approximately 125 W/m2-
K-W and the uncertainty in the heat transfer coefficient measurement would be h = 1.8 W/m2-
K.
As an engineer designing this measurement device, you would like to calculate not the heater
power but rather the partial derivative in the heat transfer coefficient with respect to heater power
in order to automate the process of computing h and therefore evaluate how design changes
affect the instrument. This would be difficult to accomplish analytically; note that both m and
the constants Ci in Eq. (16) are functions of h . It is more convenient to determine the partial
derivative numerically. That is, set the heat transfer coefficient at its nominal value plus a small
amount ( h + h ) and evaluate the heater power ( qh , h +h ); note that h should be small relative
to the nominal value of the heat transfer coefficient. Then set the heat transfer coefficient at its
nominal value less a small amount ( h h ) and evaluate the heater power ( qh , h h ). The partial
derivative is approximately:
h 2 h
qh qh ,h +h qh ,h h
Since qh needs to be evaluated at several values of the heat transfer coefficient, it is convenient
to have the computation of the heater power occur within a function that is called twice within
the equation window. Because the solution is in the form of an EES code, this is an ideal
problem to use a MODULE. A MODULE is a stand-alone EES program that can be called from
the main EES equation window. The MODULE is provided with inputs and it calculates
outputs. The protocol of a call to a MODULE involves the name of the MODULE followed by a
series of inputs separated by a colon from a series of outputs. For example, we will create a
MODULE Heaterpower that calculates the value of the variable q_dot_h. It is important to note
that all of the variables from the main equation window that are required by the MODULE must
be passed to the MODULE; the MODULE can only access variables that are passed to it as
parameters or using the $COMMON directive.
The MODULE Heaterpower must be placed at the top of the EES code and is a small self-
contained EES program; we create the MODULE by copying those lines of the main EES code
that are required to calculate the variable q_dot_h.
MODULE Heaterpower(th,k,a,b,T_h,T_infinity,h_bar:q_dot_h)
theta_h=T_h-T_infinity "heater temperature difference"
m=sqrt(2*h_bar/(k*th)) "solution constant"
N=100 "number of terms"
duplicate i=1,N
lambda[i]=(2*i-1)*pi/(2*a) "eigenvalues"
C[i]=theta_h*(4*(-1)^(1+i)/(2*i-1)/Pi)/(cosh(sqrt(lambda[i]^2+m^2)*b))
q_dot_h[i]=2*k*th*a*C[i]*sqrt(lambda[i]^2+m^2)*sinh(sqrt(lambda[i]^2+m^2)*b)
end
q_dot_h=sum(q_dot_h[1..N])
end
The calling protocol for the MODULE consists of a series of inputs (the variables d, k, a, b,
T_htr, T_f, and h) that are separated by a series outputs (in this case only the variable q_dot_htr)
by a colon. MODULES are most useful where a certain sequence of code must be executed
multiple times; in this case, the MODULE Heaterpower enables the partial derivative to be easily
computed.
CALL Q_dot_heater(d,k,a,b,T_htr,T_f,h-dh:q_dot_htr_minus)
CALL Q_dot_heater(d,k,a,b,T_htr,T_f,h+dh:q_dot_htr_plus)
dhdqdot=2*dh/(q_dot_htr_plus-q_dot_htr_minus)
delta_h=dhdqdot*delta_q_dot_htr
The modifications to the EES code verifies that the uncertainty in heat transfer coefficient,
delta_h, is 1.26 W/m2-K and provides a convenient tool for assessing the impact of the various
design parameters on the performance of the measurement system. For example, Figure 5
illustrates the uncertainty in the heat transfer coefficient as a function of the plate thickness (th)
for various values of its half-width (a).
Figure 5: Uncertainty in the measured heat transfer coefficient as a function of the plate thickness
for various values of the plate half-width.
Problem 2.2-6 (2-4 in text)
q = 10000 W/m
2
H = 3 cm
Tset = 20C
W = 6 cm
Tset = 20C
kx = 50 W/m-K
ky = 4 W/m-K
Figure P2.2-6: Composite structure exposed to a heat flux.
The structure is anisotropic. The effective conductivity of the composite in the x-direction is kx =
50 W/m-K and in the y-direction it is ky = 4 W/m-K. The top of the structure is exposed to a heat
flux of q = 10,000 W/m2. The other edges are maintained at Tset = 20C. The height of the
structure is H = 3 cm and the half-width is W = 6 cm.
a.) Develop a separation of variables solution for the 2-D steady-state temperature distribution in
the composite.
"Inputs"
W=6 [cm]*convert(cm,m) "width of laminate"
H=3 [cm]*convert(cm,m) "height of laminate"
q``=10000 [W/m^2] "heat flux"
k_x=50 [W/m-K] "conductivity in the x-direction"
k_y=4 [W/m-K] "conductivity in the y-direction"
T_set=converttemp(C,K,20[C]) "specified edge temperatures"
A half-symmetry model of the problem shown in Figure P2.2-6 is governed by the transformed
partial differential equation:
2 2
kx + k =0 (1)
x 2 y 2
y
=0 (2)
x x =0
x =W = 0 (3)
ky = q (4)
y y=H
y =0 = 0 (5)
where
= T Tset (6)
= X Y (7)
d 2 X d 2 Y
kx Y + y
k X =0 (8)
dx 2 dx 2
d 2 X d 2 Y
dx 2 + k y dx 2 = 0 (9)
X
k Y
x
2 2
d 2 X
2
+ 2 X = 0 (10)
dx
d 2 Y
2 2 Y = 0 (11)
dy 2
where
kx
2 = (12)
ky
beta=sqrt(k_x/k_y)
Notice that the sign of the constant is selected so that the solution in the homogeneous direction
(x) is sines and cosines. The solution to Eq. (10) is:
X = C1 sin ( x ) + C2 cos ( x ) (13)
X = C2 cos ( x ) (14)
The boundary condition at x = L, Eq. (3), leads to the eigencondition for the problem:
cos ( i W ) = 0 (15)
i =
(1 + 2 i ) for i = 0,1,... (16)
2W
The sum of these solutions is, itself, a solution (note that the constant C2,i is absorbed into the
other constants):
= i = cos ( i x ) C3,i sinh ( i y ) + C4,i cosh ( i y ) (20)
i =1 i =1
= i = Ci cos ( i x ) sinh ( i y ) (21)
i =1 i =1
Equation (21) is substituted into the boundary condition at y = H, Eq. (4):
i k y Ci cos ( i x ) cosh ( i H ) = q (22)
i =1
> restart;
> assume(i,integer);
> lambda:=(1+2*i)*Pi/(2*W);
( 1 + 2 i~ )
:=
2W
> int((cos(lambda*x))^2,x=0..W);
W
2
> int(cos(lambda*x),x=0..W);
2 ( -1 )i~ W
( 1 + 2 i~ )
duplicate i=0,N_term
k_y*C[i]*beta*lambda[i]*cosh(beta*lambda[i]*H)*W/2=q``*2*(-1)^i*W/((1+2*i)*Pi)
end
1
20
23.12
Dimensionless y position, y/H 0.8
26.24
29.36
0.6 32.48
35.6
38.72
0.4
41.84
44.96
0.2 48.08
51.2
0
0 0.2 0.4 0.6 0.8 1
rout = 10 cm L = 10 cm
rin = 6 cm TLOx = 125 K
Figure P2.3-1: Cryogen transfer pipe connecting two liquid oxygen tanks.
a.) Develop an analytical model that can predict the temperature distribution within the pipe.
Prepare a contour plot of the temperature distribution within the pipe.
qx + qr = qx + dx + qr + dr
or
qx q
0= dx + r dr
x r
T
qx = k 2 r dr
x
and
T
qr = k 2 r dx
r
or
2T T
r + r =0
x 2 r r
A half-symmetry model of the pipe will be generated; the boundary conditions are therefore:
T
=0
x x =0
Tx = L = TLOx
T
=0
r r = rin
T
k = q
r r = rout
As stated, there are two non-homogeneous boundary conditions; however, the boundary
condition at x = L can be made homogeneous by defining the temperature difference:
= T TLOx
The partial differential equation and boundary conditions are written in terms of :
2
r + r =0 (1)
x 2 r r
=0 (2)
x x =0
x= L = 0 (3)
=0 (4)
r r = rin
k = q (5)
r r = rout
Note that the two homogeneous boundary conditions are in the x-direction and so the
eigenfunctions of the problem will be in this direction. We assume that the solution is separable;
that is, the solution is the product of a function only of x (X) and r (R):
( x, y ) = X ( x ) R ( r )
Substituting the product solution into the governing partial differential equation, Eq. (1), leads to:
d 2 X d d R
r R + X r =0
dx 2
dr dr
d 2 X d d R
r
dx + dr dr = 0
2
X r R
2 2
Note that the 1st term is a function only of x while the 2nd term is a function only of r; these two
quantities must be equal and opposite constants (2). The choice of the sign is again important;
the eigenfunctions must be in x and therefore the two ordinary differential equations must be:
d 2 X
+ 2 X = 0 (6)
dx 2
d d R
r 2 r R = 0 (7)
dr dr
The eigenproblem will be solved first; the solution to Eq. (6) is:
X = C1 cos ( x ) + C2 sin ( x )
The boundary condition at x = 0, Eq. (2), eliminates the sine term. The boundary condition at x =
L, Eq. (3), leads to:
C1 cos ( L ) = 0
The ordinary differential equation for R, Eq. (7), is solved by Bessel functions, as discussed in
Section 1.8. Equation (7) is a form of Bessel's equation:
d p d 2 s
x c x =0
dx dx
where s = 1 and p = 1; the quantity s p +2 is therefore not equal to zero and we are directed
toward the left-side of the Bessel function flow chart presented in Section 1.8.4 where the
parameters n = 0, a = 1, and n/a = 0 are computed. The last term is negative and therefore the
general solution to Eq. (7) is:
> restart;
> ODEr:=diff(r*diff(thetar(r),r),r)-lambda^2*r*thetar(r)=0;
d
2
d
ODEr := thetar( r ) + r 2 thetar( r ) 2 r thetar( r ) = 0
dr dr
> thetars:=dsolve(ODEr);
thetars := thetar( r ) = _C1 BesselI( 0, r ) + _C2 BesselK( 0, r )
> restart;
> assume(i,integer);
> lambda:=(2*i-1)*Pi/(2*L);
( 2 i~ 1 )
:=
2L
> theta:=(x,r)->cos(lambda*x)*(C3*BesselI(0,lambda*r)+C4*BesselK(0,lambda*r));
:= ( x, r ) cos( x ) ( C3 BesselI( 0, r ) + C4 BesselK( 0, r ) )
Verify that it solves both boundary conditions in the x-directions, Eqs. (2) and (3):
> eval(diff(theta(x,r),x),x=0);
0
> theta(L,r);
0
> r*diff(diff(theta(x,r),x),x)+diff(r*diff(theta(x,r),r),r);
( 2 i~ 1 ) x
r cos
1
( 2 i~ 1 )
2 2
4 2 L
( 2 i~ 1 ) r + C4 BesselK 0, ( 2 i~ 1 ) r L 2 +
C3 BesselI 0,
2L 2L
( 2 i~ 1 ) r ( 2 i~ 1 )
C3 BesselI 1,
( 2 i~ 1 ) x 1 2 L
cos
2 L 2 L
( 2 i~ 1 ) r
C4 BesselK 1, ( 2 i~ 1 )
1 2L + r cos ( 2 i~ 1 ) x
2 L 2 L
( 2 i~ 1 ) r
2 L BesselI 1,
( 2 i~ 1 ) r 2L ( 2 i~ 1 ) 2 2
C3 BesselI 0,
( 2 i~ 1 ) r
1 2L
4 2
L
( 2 i~ 1 ) r
2 L BesselK 1,
( 2 i~ 1 ) r ( 2 i~ 1 ) 2 2
C4 BesselK 0, 2L
1 2L ( 2 i~ 1 ) r
2
4 L
> simplify(%);
0
= i = cos ( i x ) C3,i BesselI ( 0,i r ) + C4,i BesselK ( 0,i r ) (8)
i =1 i =1
The solution must satisfy the boundary conditions in the non-homogeneous direction; the
boundary condition at r = rin, Eq. (4), leads to:
d
= cos ( i x ) C3,i BesselI ( 0,i r ) + C4,i BesselK ( 0,i r ) r = r = 0 (9)
r r = rin i =1 dr in
The derivatives of the Bessel functions may either be evaluated using the equations provided in
Section 1.8.4 or, more conveniently, using Maple:
> restart;
> diff(BesselI(0,lambda*r),r);
BesselI( 1, r )
> diff(BesselK(0,lambda*r),r);
BesselK( 1, r )
cos ( x ) C
i =1
i 3,i i BesselI (1,i rin ) C4,i i BesselK (1,i rin ) = 0
Unlike most of the problems we have previously encountered, neither of the constants is
eliminated; instead we see that there is a relationship between the two constants:
Equation (11) is substituted into the remaining non-homogeneous boundary condition at r = rout ,
Eq. (5), in order to obtain:
This equation is multiplied by an arbitrary eigenfunction, cos(j x), and integrated between the
homogeneous boundary conditions (from x = 0 to x= L); using the orthogonality property of the
eigenfunctions we obtain:
2 q sin ( i L )
Ci =
BesselI (1,i rin )
i2 k L BesselI (1,i rout ) BesselK (1,i rout )
BesselK (1,i rin )
"Inputs"
q``_dot=8000 [W/m^2] "Heat flux on pipe surface"
r_in=6.0 [cm]*convert(cm,m) "Pipe inner radius"
r_out=10.0 [cm]*convert(cm,m) "Pipe outer radius"
L = 10.0 [cm]*convert(cm,m) "Pipe half-length"
T_LOx=125 [K] "Pipe end temperature"
k=10 [W/m-K] "Pipe conductivity"
"dimensionless position"
r_bar=0.5
x_bar=0.5
r=r_in+(r_out-r_in)*r_bar
x=x_bar*L
The temperature distribution is computed over a range of positions and the results are used to
generate the contour plot shown in Figure 2.
Figure 2: Contour plot of temperature.
Problem 2.3-2 (2-6 in text)
Figure P2.3-2 illustrates a cylinder that is exposed to a concentrated heat flux at one end.
extends to infinity
k = 168 W/m-K
rout = 200 m
Ts = 20C
rexp = 21 m
q = 1500 W/cm
2 adiabatic
Figure P2.3-2: Cylinder exposed to a concentrated heat flux at one end.
The cylinder extends infinitely in the x-direction. The surface at x = 0 experiences a uniform
heat flux of q = 1500 W/cm2 for r < rexp = 21 m and is adiabatic for rexp < r < rout where rout =
200 m is the outer radius of the cylinder. The outer surface of the cylinder is maintained at a
uniform temperature of Ts = 20C. The conductivity of the cylinder material is k = 168 W/m-K.
a.) Develop a separation of variables solution for the temperature distribution within the
cylinder. Plot the temperature as a function of radius for various values of x.
"Inputs"
r_out=200 [micron]*convert(micron,m) "outer radius of domain"
q``_dot=1500 [W/cm^2]*convert(W/cm^2,W/m^2) "exposure flux"
k=168 [W/m-K] "conductivity of work piece"
r_exp=21 [micron]*convert(micron,m) "radius of exposure zone"
qx + qr = qx + dx + qr + dr
or
qx q
0= dx + r dr
x r
and
T
qr = k 2 r dx
r
T T
0= k 2 r dr dx + k 2 r dx dr
x x r r
or
2T T
r + r =0
x 2 r r
Tx = Ts
Tr =0 must be finite
Tr = rout = Ts
As stated, there are two non-homogeneous boundary conditions; however, the boundary
condition at r = rout can be made homogeneous by defining the temperature difference:
= T Ts
The partial differential equation and boundary conditions are written in terms of :
2
r + r =0 (1)
x 2 r r
r =r = 0
out
(5)
Note that the two homogeneous boundary conditions are in the x-direction and so the
eigenfunctions of the problem will be in this direction. We assume that the solution is separable;
that is, the solution is the product of a function only of x (X) and r (R):
( x, y ) = X ( x ) R ( r )
Substituting the product solution into the governing partial differential equation, Eq. (1), leads to:
d 2 X d d R
r R + X r =0
dx 2
dr dr
d 2 X d d R
r
dx + dr dr = 0
2
X r R
2 2
Note that the 1st term is a function only of x while the 2nd term is a function only of r; these two
quantities must be equal and opposite constants (2). The choice of the sign is again important;
the eigenfunctions must be in r and therefore the two ordinary differential equations must be:
d 2 X
2 X = 0 (6)
dx 2
d d R
r + 2 r R = 0 (7)
dr dr
The eigenproblem will be solved first; the solution to Eq. (7) is:
R = C1 BesselJ ( 0, r ) + C2 BesselY ( 0, r )
R = C1 BesselJ ( 0, r )
The boundary condition at r = rout, Eq. (5), leads to:
The 0th order Bessel function of the 1st kind (i.e., Bessel_J(0,x)) oscillates about zero every time
the argument changes by 2 in the same way that sine and cosine do; therefore, there are an
infinite number of eigenvalues i that will satisfy Eq. (8) associated with an infinite number of
eigenfunctions. The eigencondition for this problem cannot be used to explicitly solve for the
eigenvalues; rather, an implicit equation for the eigenvalues results from Eq. (8):
duplicate i=1,N
BesselJ(0,lambda[i]*r_out)=0 "solve for eigenvalues"
end
The arrays lowerlimit[], upperlimit[], and guess[] are used to constrain the solution for the array
lambda[] in the Variable Information window. The solution to the ordinary differential equation
for X, Eq. (6), is:
X i = C4,i exp ( i x )
Substituting Eq. (10) into the boundary condition at x = 0, Eq. (2), leads to:
rout rexp
k Ci i r BesselJ ( 0,i r ) dr = q
2
r BesselJ ( 0, r ) dr
i (12)
0 0
Integral 1 Integral 2
> restart;
> Integral1:=int(r*(BesselJ(0,lambda[i]*r))^2,r=0..r_out);
Integral1 :=
2 2
1 r_out ( i r_out BesselJ( 0, i r_out ) + i r_out BesselJ( 1, i r_out ) )
2 i
> Integral2:=int(r*BesselJ(0,lambda[i]*r),r=0..r_exp);
r_exp BesselJ( 1, r_exp i )
Integral2 :=
i
duplicate i=1,N
Integral1[i]=1/2*r_out/Pi^(1/2)/lambda[i]*(Pi^(1/2)*lambda[i]*r_out*BesselJ(0,lambda[i]*r_out)^2+&
Pi^(1/2)*lambda[i]*r_out*BesselJ(1,lambda[i]*r_out)^2)
Integral2[i]= 1/lambda[i]*r_exp*BesselJ(1,r_exp*lambda[i])
k*lambda[i]*C[i]*Integral1[i]=q``_dot*Integral2[i]
end
1.2
0.4
0.2
0
0 20 40 60 80 100 120 140 160 180 200
Radius (m)
Figure 2: Temperature difference as a function of r for various values of x.
b.) Determine the average temperature of the cylinder at the surface exposed to the heat flux.
The average temperature of the cylinder over the region x = 0 and 0 < r < rexp is:
rexp
1
T =
rexp
2 2 r T
0
x =0 dr (13)
or:
rexp
1
T = Ts +
rexp
2 2 r
0
x =0 dr (14)
c.) Define a dimensionless thermal resistance between the surface exposed to the heat flux and
Ts. Plot the dimensionless thermal resistance as a function of rout/rin.
A dimensionless thermal resistance is defined by normalizing the actual resistance against the
reistance to axial conduction through a cylinder that is rout in radius and rout long:
k (T Ts ) rout k (T Ts ) rout
2
R rout
2 2
R= = = k (17)
rout q rexp rout
2
q rout rexp
actual resistance
80
Dimensionless thermal resistance
10
0.1
1 10 100
Ratio of cylinder to exposure radii, rout/rexp
Figure 3: Dimensionless thermal resistance as a function of rout/rexp.
d.) Show that your plot from (c) does not change if the problem parameters (e.g., Ts, k, etc.) are
changed.
The values of various parameters were changed and did not affect Figure 3 (the plots are overlaid
onto Figure 3).
Problem 2.4-1 (2-7 in text)
The plate shown in Figure P2.4-1 is exposed to a uniform heat flux q = 1x105 W/m2 along its
top surface and is adiabatic at its bottom surface. The left side of the plate is kept at TL = 300 K
and the right side is at TR = 500 K. The height and width of the plate are H = 1 cm and W = 5
cm, respectively. The conductivity of the plate is k = 10 W/m-K.
q = 1x10 W/m
5 2
TL = 300 K W = 5 cm
TR = 500 K
y H = 1 cm
x
k = 10 W/m-K
Figure P2.4-1: Plate.
a.) Derive an analytical solution for the temperature distribution in the plate.
"Inputs"
H=1 [cm]*convert(cm,m) "height of plate"
W=5 [cm]*convert(cm,m) "thickness of plate"
k=10 [W/m-K] "conductivity"
q``=100000 [W/m^2] "heat flux"
T_R=500 [K] "temperature of right hand surface"
T_L=300 [K] "temperature of left hand surface"
2 2
+ =0 (1)
x 2 y 2
x =0 = 0 (2)
x =W = R (3)
=0 (4)
y y =0
k = q (5)
y y=H
where
= T TL (6)
and
R = TR TL (7)
The problem has two, non-homogeneous boundary condition and therefore must be solved using
superposition.
= A + B (8)
2 A 2 A
+ 2 =0 (9)
x 2 y
A, x = 0 = 0 (10)
A, x =W = R (11)
A
=0 (12)
y y =0
A
=0 (13)
y y=H
x
A = R (14)
L
2 B 2 B
+ 2 =0 (15)
x 2 y
B , x =0 = 0 (16)
B , x =W = 0 (17)
B
=0 (18)
y y =0
B
k = q (19)
y y=H
The solution for B is 2-D and can be obtained using separation of variables. The eigenfunctions
are:
X B ,i = sin ( B ,i x ) (20)
iW
B , i = for i = 1, 2,.. (21)
"sub-problem B solution"
N_term=11 [-] "number of terms"
duplicate i=1,N_term
lambda_B[i]=i*pi/W "i'th eigenvalue"
end
k B ,i Ci sin ( B ,i x ) sinh ( B ,i H ) = q (24)
i =1
> restart;
> assume(i,integer);
> lambda_B:=i*Pi/W;
i~
lambda_B :=
W
> int((sin(lambda_B*x))^2,x=0..W);
W
2
> int(sin(lambda_B*x),x=0..W);
W ( 1 + ( -1 )i~ )
i~
duplicate i=1,N_term
C[i]*k*lambda_B[i]*sinh(lambda_B[i]*H)*W/2=q``*(-W*(-1+(-1)^i)/i/Pi) "i'th constant"
end
duplicate i=1,N_term
theta_B[i]=C[i]*sin(lambda_B[i]*x)*cosh(lambda_B[i]*y) "i'th term"
end
theta_B=sum(theta_B[1..N_term]) "solution to sub-problem B"
345.5
0.8
390.9
436.4
0.6 481.8
527.3
572.7
0.4
618.2
663.6
0.2 709.1
754.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Dimensionless x-position, x/W
Figure 2: Contour plot of the temperature distribution.
Problem 2.5-1 (2-8 in text): A Heating Element
Figure P2.5-1 illustrates an electrical heating element that is affixed to the wall of a chemical
reactor. The element is rectangular in cross-section and very long (into the page). The
temperature distribution within the element is therefore two-dimensional, T(x, y). The width of
the element is a = 5.0 cm and the height is b = 10.0 cm. The three edges of the element that are
exposed to the chemical (at x = 0, y = 0, and x = a) are maintained at a temperature Tc = 200C
while the upper edge (at y = b) is affixed to the well-insulated wall of the reactor and can
therefore be considered adiabatic. The element experiences a uniform volumetric rate of thermal
energy generation, g = 1x106 W/m3. The conductivity of the material is k = 0.8 W/m-K.
reactor wall
k = 0.8 W/m-K
Tc = 200C
g = 1x10 W/m
6 3
a = 5 cm
y
Tc = 200C
x
Tc = 200C b = 10 cm
Figure P2.5-1: Electrical heating element.
"Inputs"
a=5.0 [cm]*convert(cm,m) "width of element"
b=10.0 [cm]*convert(cm,m) "height of element"
k=0.8 [W/m-K] "conductivity"
T_c=converttemp(C,K,200 [C]) "chemical temperature"
g```_dot=1e6 [W/m^3] "rate of volumetric generation"
L=1 [m] "unit length of element into the page"
The computational domain of the element with the regularly spaced grid of nodes is shown in
Figure 2.
Figure 2: The regularly spaced grid used to obtain a numerical solution.
The first step in obtaining a numerical solution is to position the nodes throughout the
computational domain. We will use grid with nodes placed on the edges and distributed
uniformly throughout. The x and y distance between adjacent nodes (x and y) are:
L
x = (1)
( m 1)
b
y = (2)
( n 1)
and the x and y positions of any node i,j are given by:
xi =
( i 1) a (3)
( m 1)
yj =
( j 1) b (4)
( n 1)
where m and n are the number of nodes used in the x and y directions.
"Setup nodes"
m=11 [-]
"number of nodes in the x-direction"
n=11 [-]
"number of nodes in the y-direction"
Dx=a/(m-1)
"distance between nodes in the x-direction"
Dy=b/(n-1)
"distance between nodes in the y-direction"
duplicate i=1,m
x[i]=(i-1)*Dx "x-position of each node"
end
duplicate j=1,n
y[j]=(j-1)*Dy "y-position of each node"
end
The next step in the solution is to write an energy balance for each node. Figure 3 illustrates a
control volume and the associated energy transfers for an internal node (see Figure 2) which
include conduction from each side ( q RHS and q LHS ), the top ( qtop ), and the bottom ( qbottom ). Note
that the direction associated with these energy transfers is arbitrary (i.e., they could have been
taken as positive if energy leaves the control volume), but it is important to write the equation in
a manner consistent with the chosen directions.
q RHS [i, j ] + q LHS [i, j ] + qtop [i, j ] + qbottom [i, j ] + g [i, j ] = 0 (5)
The next step is to approximate each of the terms in the energy balance; the material separating
the nodes is assumed to behave as a plane wall resistance and therefore:
k y L
q RHS [i, j ] =
x
(T [i + 1, j ] T [i, j ]) (6)
where L is the length of the element (assumed to be 1 m in order to do the problem on a unit
length basis); therefore, y L is the area for conduction and x is the distance over which the
conduction heat transfer occurs. Note that the temperature difference is consistent with the
direction of the arrow in Figure 3; if Ti+1,j is greater than Ti,j then energy is leaving the node and
q RHS is positive. The other heat transfers are approximated using a similar model:
k y L
q LHS [i, j ] =
x
(T [i 1, j ] T [i, j ]) (7)
k xW
qtop [i, j ] =
y
(T [i, j + 1] T [i, j ]) (8)
k xW
qbottom [i, j ] =
y
(T [i, j 1] T [i, j ]) (9)
The generation is the product of the volume of the control volume and the volumetric rate of
generation:
g [i, j ] = g x y L (10)
Note that each time the outer duplicate statement iterates once (i.e., i is increased by 1), the inner
duplicate statement iterates (n-1) times (i.e., j runs from 2 to n-1). Therefore, all of the internal
nodes are considered with these two nested duplicate loops. Also note that the unknowns are
placed in an array rather than a vector. The entries in the array T is accessed using two indices
contained in square brackets.
The boundary nodes have to be treated separately. The left, right, and bottom boundaries are
easy as the temperature is specified:
The upper boundary nodes must be considered using energy balances. Figure 4 illustrates an
energy balance associated with a node that is located on the top, insulated boundary (see Figure
2).
k y L
q RHS [i, n ] =
2x
(T [i + 1, n] T [i, n]) (15)
k y L
q LHS [i, n ] =
2 x
(T [i 1, n] T [i, n]) (16)
The factor of 2 in the denominator appears because there is half the available area for conduction
through the sides of the control volume on the top boundary. The other conduction term is
approximated as before:
k xW
qbottom [i, n ] =
y
(T [i, n 1] T [i, n]) (17)
The generation term is:
x y L g
g [i, n ] = (18)
2
"Upper edge"
duplicate i=2,(m-1)
q_dot_LHS[i,n]+q_dot_RHS[i,n]+q_dot_bottom[i,n]+gen[i,n]=0
q_dot_LHS[i,n]=k*(Dy/2)*L*(T[i-1,n]-T[i,n])/Dx
q_dot_RHS[i,n]=k*(Dy/2)*L*(T[i+1,n]-T[i,n])/Dx
q_dot_bottom[i,n]=k*Dx*L*(T[i,n-1]-T[i,n])/Dy
gen[i,n]=Dx*Dy*L*g```_dot/2
end
Notice that the control volumes at the top left corner (i.e., i =1, j =n) and the top right corner
(i.e., i =m, j =n) have already been taken addressed by the equations for the left and right
boundaries, Eqs. (11) and (12). Therefore, we have to make sure not to write additional
equations related to this node or the problem will be over-specified and so the equations for the
top boundary can only be written for i = 2...(m-1).
b.) Plot the temperature as a function of x at various values of y. What is the maximum
temperature within the element and where is it located?
The solution is obtained for n=11 and the columns Ti,1 (corresponding to y =0) through Ti,11
(corresponding to y = 10 cm) are plotted in Figure 5 as a function of x.
Figure 5: Temperature as a function of x for various values of y.
The hottest spot in the element is at the adiabatic wall (y = 10 cm) and the center (x = 2.5 cm);
the hottest temperature is about 860 K.
c.) Prepare a reality check to show that your solution behaves according to your physical
intuition. That is, change some aspect of your program and show that the results behave as
you would expect (clearly describe the change that you made and show the result).
There are several possible answers to this; I increased the conductivity by a factor of 10 and
examined the temperature distribution. Figure 6 illustrates the temperature as a function of x for
the original conductivity (k = 0.8 W/m-K) and the increased conductivity (k = 8.0 W/m-K);
notice that the increased conductivity has had the expected effect of reducing the temperature
rise.
Figure 6: Temperature as a function of x for y = 10 cm and two values of conductivity.
Problem 2.6-1 (2-9 in text): Model of Welding Process (revisited)
Figure P2.6-1 illustrates a cut-away view of two plates that are being welded together. Both
edges of the plate are clamped and effectively held at temperatures Ts = 25C. The top of the
plate is exposed to a heat flux that varies with position x, measured from joint, according to:
qm ( x ) = q j exp ( x / L j ) where q j =1x106 W/m2 is the maximum heat flux (at the joint, x = 0)
and Lj = 2.0 cm is a measure of the extent of the heat flux. The back side of the plates are
exposed to liquid cooling by a jet of fluid at Tf = -35C with h = 5000 W/m2-K. A half-
symmetry model of the problem is shown in Figure P2.6-1. The thickness of the plate is b = 3.5
cm and the width of a single plate is W = 8.5 cm. You may assume that the welding process is
steady-state and 2-D. You may neglect convection from the top of the plate. The conductivity
of the plate material is k = 38 W/m-K.
both edges are held at fixed temperature
heat flux
joint
qm impingement cooling
k = 38 W/m-K
W = 8.5 cm Ts = 25C
b = 3.5 cm
y
Figure P2.6-1: Welding process and half-symmetry model of the welding process.
a.) Develop a separation of variables solution to the problem (note, this was done previously in
Problem 2.2-1). Implement the solution in EES and prepare a plot of the temperature as a
function of x at y = 0, 1.0, 2.0, 3.0, and 3.5 cm.
b.) Prepare a contour plot of the temperature distribution.
See the solution for Problem 2.2-1 for parts (a) and (b).
c.) Develop a numerical model of the problem. Implement the solution in MATLAB and
prepare a contour or surface plot of the temperature in the plate.
The input parameters are entered in the MATLAB function P2p2_1; the input arguments are m
and n, the number of nodes in the x and y coordinates while the output arguments are the x and y
positions of each node and the predicted temperature at each node.
function[xm,ym,T]=P2p6_1(m,n)
end
function[qflux]=qf(x)
qflux=qf_j*exp(-x/L_j);
end
A 2-D numerical model will be generated using a grid in which the x and y coordinates of each
node are:
xi =
( i 1)W for i = 1..m (1)
( m 1)
yi =
( j 1)W for j = 1..n (2)
( m 1)
The distance between adjacent nodes is:
W
x = (3)
( m 1)
W
y = (4)
( n 1)
%Setup grid
for i=1:m
x(i,1)=(i-1)*W/(m-1);
end
Dx=W/(m-1);
for j=1:n
y(j,1)=(j-1)*W/(n-1);
end
Dy=b/(n-1);
The problem will be solved by setting up and inverting a matrix containing the algebraic
equations that enforce the conservation of energy for each control volume. A control volume for
an internal node includes conduction from the left and right sides ( q LHS and q RHS ) and top and
bottom ( qtop and qbottom ). The energy balance is:
k y L
q RHS =
x
(Ti1, j Ti, j )
k y L
q LHS =
x
(Ti+1, j Ti, j )
k x L
qbottom =
y
(Ti, j 1 Ti, j )
k x L
qtop =
y
(Ti, j +1 Ti, j )
where L is the depth of the plate (into the page); L is set to 1.0 m which is consistent with doing
the problem on a per unit length basis. Combining these equations leads to:
k y L k y L k x L k x L
x
( Ti 1, j Ti , j ) +
x
( Ti +1, j Ti , j ) +
y
( Ti , j 1 Ti , j ) +
y
(Ti, j +1 Ti, j ) = 0 (5)
for i = 2.. ( m 1) and j = 2.. ( n 1)
The equation is rearranged to make it clear what the coefficient is for each unknown
temperature:
k y L k x L k y L k y L k x L k x L
Ti , j 2 2 + Ti 1, j + Ti +1, j + Ti , j 1 + Ti , j +1 =0
x y x x y y
for i = 2.. ( m 1) and j = 2.. ( n 1)
(6)
The control volume equations must be placed into the matrix equation:
AX =b
where the equation for control volume i,j is placed into row m(j-1)+i of A and Ti,j corresponds to
element Xm (j-1)+i in the vector X . Therefore, each coefficient in Eq. (6) (i.e., each term
multiplying an unknown temperature on the left side of the equation) must be placed in the row
of A corresponding to the control volume being examined and the column of A corresponding
to the unknown in X . The matrix assignments consistent with Eq. (6) are:
k y L k x L
Am( j 1)+i ,m( j 1)+i = 2 2 for i = 2.. ( m 1) and j = 2.. ( n 1) (7)
x y
k y L
Am( j 1)+i ,m( j 1)+i 1 = for i = 2.. ( m 1) and j = 2.. ( n 1) (8)
x
k y L
Am( j 1)+i ,m( j 1)+i +1 = for i = 2.. ( m 1) and j = 2.. ( n 1) (9)
x
k x L
Am( j 1)+i ,m( j 11)+i = for i = 2.. ( m 1) and j = 2.. ( n 1) (10)
y
k x L
Am( j 1)+i ,m( j +11)+i = for i = 2.. ( m 1) and j = 2.. ( n 1) (11)
y
A sparse matrix is allocated in MATLAB for A and the equations derived above are
implemented using a nested for loop. The spalloc command requires three arguments, which are
the number of rows and columns and the maximum number of elements in the matrix. Note that
there are at most 5 non-zero entries in each row of A , corresponding to Eqs. (7) through (11);
thus the last argument in the spalloc command which corresponds to the maximum number of
non-zero entries in the sparse matrix.
The nodes along the upper edge must be considered separately, leading to:
k y L k x L k y L k y L k x L
Ti ,n
x
y + Ti 1, n + Ti +1, n + Ti ,n 1 = qm x L
2 x 2 x y (12)
for i = 2.. ( m 1)
k y L k x L
Am( n 1)+i ,m( n 1)+i = for i = 2.. ( m 1) (13)
x y
k y L
Am( n 1)+i ,m( n 1)+i 1 = for i = 2.. ( m 1) (14)
2 x
k y L
Am( n 1)+i ,m( n 1)+i +1 = for i = 2.. ( m 1) (15)
2 x
k x L
Am( n 1)+i , m( n 11)+i = for i = 2.. ( m 1) (16)
y
%upper edge
for i=2:(m-1)
A(m*(n-1)+i,m*(n-1)+i)=-k*Dy*L/Dx-k*Dx*L/Dy;
A(m*(n-1)+i,m*(n-1)+i-1)=k*Dy*L/(2*Dx);
A(m*(n-1)+i,m*(n-1)+i+1)=k*Dy*L/(2*Dx);
A(m*(n-1)+i,m*(n-1-1)+i)=k*Dx*L/Dy;
bm(m*(n-1)+i,1)=-qf(x(i))*Dx*L;
end
The node at the upper left corner must be considered separately, leading to:
k y L k x L k y L k x L x L
T1,n + T2,n + T1,n 1 = qm (18)
2 x 2 y 2 x 2 y 2
k y L k x L
Am( n 1)+1,m( n 1)+1 = (19)
2 x 2 y
k y L
Am( n 1) +1,m( n 1)+1+1 = (20)
2 x
k x L
Am( n 1)+1, m( n 11)+1 = (21)
2 y
x L
bm( n 1)+1,1 = qm (22)
2
The node along the left edge must be considered separately, leading to:
k y L k x L k y L k x L k x L
T1, j + T2 +1, j + T1, j 1 + T1, j +1 =0
x y 2 x 2 y 2 y (23)
for j = 2.. ( n 1)
k y L k x L
Am( j 1)+1,m( j 1)+1 = for j = 2.. ( n 1) (24)
x y
k y L
Am( j 1)+1,m( j 1)+1+1 = for j = 2.. ( n 1) (25)
x
k x L
Am( j 1) +1,m( j 11)+1 = for j = 2.. ( n 1) (26)
2 y
k x L
Am( j 1)+1,m( j +11)+1 = for j = 2.. ( n 1) (27)
2 y
%left edge
for j=2:(n-1)
A(m*(j-1)+1,m*(j-1)+1)=-k*Dy*L/Dx-k*Dx*L/Dy;
A(m*(j-1)+1,m*(j-1)+1+1)=k*Dy*L/Dx;
A(m*(j-1)+1,m*(j-1-1)+1)=k*Dx*L/(2*Dy);
A(m*(j-1)+1,m*(j+1-1)+1)=k*Dx*L/(2*Dy);
end
The node at the lower left corner must be considered separately, leading to:
k y L k x L x L k y L k x L x L
T1,1 h + T2,1 + T1,2 = h Tf (28)
2 x 2 y 2 2 x 2 y 2
k y L k x L x L
Am(11) +1,m(11)+1 = h (29)
2 x 2 y 2
k y L
Am(11)+1, m(11)+1+1 = (30)
2 x
k x L
Am(11)+1, m(1+11)+1 = (31)
2 y
x L
bm(11)+1,1 = h Tf (32)
2
The nodes along the bottom edge must be considered separately, leading to:
k y L k x L k y L k y L k x L
Ti ,1 h x L + Ti 1,1 + Ti +1,1 + Ti ,1+1 = h x LT f
x y 2 x 2 x y (33)
for i = 2.. ( m 1)
k y L
Am(11) +i , m(11)+i 1 = for i = 2.. ( m 1) (35)
2 x
k y L
Am(11) +i , m(11)+i +1 = for i = 2.. ( m 1) (36)
2 x
k x L
Am(11) +i , m(1+11)+i = for i = 2.. ( m 1) (37)
y
%lower edge
for i=2:(m-1)
A(m*(1-1)+i,m*(1-1)+i)=-k*Dy*L/Dx-k*Dx*L/Dy-h*Dx*L;
A(m*(1-1)+i,m*(1-1)+i-1)=k*Dy*L/(2*Dx);
A(m*(1-1)+i,m*(1-1)+i+1)=k*Dy*L/(2*Dx);
A(m*(1-1)+i,m*(1+1-1)+i)=k*Dx*L/Dy;
bm(m*(1-1)+i,1)=-h*Dx*L*T_f;
end
The vector of unknowns X is obtained through matrix manipulation and then placed into matrix
format. Matrices for the x and y positions of each node are also created.
X=A\bm;
for i=1:m
for j=1:n
xm(i,j)=x(i);
ym(i,j)=y(j);
T(i,j)=X(m*(j-1)+i);
end
end
>> [x,y,T]=P2p6_1(20,40);
>> surf(x,y,T);
b.) Plot the temperature as a function of x at y = 0, b/2, and b and overlay on this plot the
separation of variables solution obtained in part (a) evaluated at the same locations.
Figure 4: Temperature as a function of axial position for y = 0, b/2, and b predicted using the
separation of variables solution and the numerical solution.
Problem 2.7-1 (2-10 in text): A Double Paned Window
Figure P2.7-1(a) illustrates a double paned window. The window consists of two panes of glass
each of which is tg = 0.95 cm thick and W = 4 ft wide by H = 5 ft high. The glass panes are
separated by an air gap of g = 1.9 cm. You may assume that the air is stagnant with ka = 0.025
W/m-K. The glass has conductivity kg = 1.4 W/m-K. The heat transfer coefficient between the
inner surface of the inner pane and the indoor air is hin = 10 W/m2-K and the heat transfer
coefficient between the outer surface of the outer pane and the outdoor air is hout = 25 W/m2-K.
You keep your house heated to Tin = 70F.
width of window, W = 4 ft
tg = 0.95cm
Tin = 70F
tg = 0.95 cm
hin = 10 W/m -K
2
g = 1.9 cm
H = 5 ft Tout = 23F
hout = 25 W/m -K
2
ka = 0.025 W/m-K
kg = 1.4 W/m-K
The average heating season in Madison lasts about time = 130 days and the average outdoor
temperature during this time is Tout = 23F. You heat with natural gas and pay, on average, ec =
1.415 $/therm (a therm is an energy unit =1.055x108 J).
a.) Calculate the average rate of heat transfer through the double paned window during the
heating season.
"Inputs"
tg=0.375*convert(inch,m) "glass thickness"
g=0.75*convert(inch,m) "air gap"
k_g=1.4 [W/m-K] "glass conductivity"
k_a=0.025 [W/m-K] "air conductivity"
H=5 [ft]*convert(ft,m) "height of window"
W=4[ft]*convert(ft,m) "width of window"
T_in=converttemp(F,K,70 [F]) "indoor air temperature"
h_in=10 [W/m^2-K] "heat transfer coefficient on inside of window"
T_out=converttemp(F,K,23 [F]) "outdoor air temperature"
h_out=25 [W/m^2-K] "heat transfer coefficient on outside of window"
time=130 [day]*convert(day,s) "heating season duration"
ec=1.415 [$/therm]*convert($/therm,$/J) "cost of energy"
The heat transfer is resisted by convection on the inner and outer surfaces:
1
Rconv ,in = (1)
hin W H
1
Rconv ,out = (2)
hout W H
tg
Rcond , g = (3)
kg W H
g
Rcond ,a = (4)
ka W H
Tin Tout
q = (5)
Rconv ,in + 2 Rcond , g + Rcond ,a + Rconv ,out
q_dot=(T_in-T_out)/(R_conv_in+2*R_cond_g+R_cond_a+R_conv_out)
"rate of heat transfer through the window"
b.) How much does the energy lost through the window cost during a single heating season?
The total amount of energy lost over the course of a heating season is:
Q = q time (6)
cost = ec Q (7)
1.9 cm
Tin = 70F 0.95 cm Tout = 23F
hout = 25 W/m -K
2
hin = 10 W/m -K
2
air 2 cm
4 cm
0.5 cm 3 cm
metal casing
km = 25 W/m-K 0.4 cm
wood
All surfaces of the casing that are adjacent to glass, wood, or the air between the glass panes can
be assumed to be adiabatic. The other surfaces are exposed to either the indoor or outdoor air.
c.) Prepare a 2-D thermal analysis of the casing using FEHT. Turn in a print out of your
geometry as well as a contour plot of the temperature distribution. What is the rate of energy
lost via conduction through the casing per unit length (W/m)?
Each of the points are selected individually and their exact coordinates are entered, which leads
to the more precise geometry shown in Figure 4.
Figure 5: Geometry.
The material properties are specified by selecting the outline and selecting Material Properties
from the Specify menu. The boundary conditions are specified by selecting each type of
boundary and then selecting Boundary Conditions from the Specify menu. A crude mesh is
generated, as shown in Figure 6.
d.) Show that your numerical model has converged by recording the rate of heat transfer per
length for several values of the number of nodes.
The mesh is refined several times and each time the heat transfer rate per unit length of casing is
recorded; the results are shown in Figure 8.
Figure 8: Heat transfer through the casing per unit length as a function of the number of nodes.
e.) How much does the casing add to the cost of heating your house?
) is entered in EES:
The result from FEHT, the heat transfer per unit length ( qcasing
qcasing = qcasing 2 (W + H ) (8)
b = 0.75 m
fluid at Tc = 320 K
The radiator panel is symmetric about its half-width and the critical dimensions that are required
to develop a half-symmetry model of the radiator are shown in Figure 2.7-3(b). There are three
regions associated with the problem that must be defined separately so that the surface conditions
can be set differently. Regions 1 and 3 are exposed to space on both sides while Region 2 is
exposed to the coolant fluid one side and space on the other; for the purposes of this problem, the
effect of radiation to space on the back side of Region 2 is neglected.
Region 1 (both sides exposed to space)
Region 2 (exposed to fluid - neglect radiation to space)
Region 3 (both sides exposed to space)
(0.50,0.75)
(0.50,0.55)
(0.50,0.52)
y
x
(0,0) (0.50,0)
(0.25,0) line of symmetry
(0.22,0)
Figure 2.7-3(b): Half-symmetry model.
a.) Prepare a FEHT model that can predict the temperature distribution over the radiator panel.
The radiator panel can be modeled as 2-D problem because it is thin and has high conductivity.
The Biot number compares the ratio of the internal conduction resistance to the external
resistance; in this case due to radiation.
Rcond
Bi = (1)
Rrad
where
th
Rcond = (2)
2 k As
1
Rrad = (3)
As s (T + Tsur
s
2 2
)(Ts + Tsur )
The Biot number is therefore:
th s (Ts2 + Tsur
2
)(Ts + Tsur )
Bi = (4)
2k
The surrounding temperature in Eq. (4), Tsur corresponds to space which is essentially 0 K and
therefore Eq. (4) reduces to:
th s Ts3
Bi = (5)
2k
We dont know the value of the panel surface temperature, Ts in Eq. (5), but we can assume that
it will be near the fluid temperature for any well-designed radiator. Using Ts = Tf in Eq. (5)
results in a Biot number of 4.6x10-5 which is very small; clearly for any reasonable value of Ts
the problem will be 2-D.
FEHT can simulate 2-D problems of this type using the Extended Surface mode. Start FEHT
and select Extended Surface from the Subject menu. Each of the three regions must be drawn
separately using the Outline option from the Draw menu. It is easiest to set an appropriate grid
using the Scale and Size selection from the Setup menu (Figure 2).
Figure 2: Scale and Size dialog window
Generate outlines for the 3 regions that are close to the correct scale and then double-click on
each node and position it exactly (Figure 3).
The material properties for all three regions can be set by selecting each while holding down the
shift key and then selecting Specify Properties from the Specify menu. Select Aluminum from
the list of materials and then modify the conductivity to be 200 W/m-K and the thickness to be
0.01 m to match the problem statement (Figure 4).
Figure 4: Specify material properties.
The surface conditions for Regions 1 and 3 can be set by selecting these regions with the shift
key held down and then selecting Surface Conditions from the Specify menu. These regions are
radiating to space but there is no option listed in the Extended Surface Conditions for a radiative
surface condition. The radiation heat flux is:
= T 4
qrad (6)
= T 3 (T 0 )
qrad (7)
= hrad (T T f )
qrad (8)
where fluid temperature (Tf) is 0 K and the convection coefficient (hrad) is a function of
temperature:
hrad = T 3 (9)
FEHT allows the specification of the surface conditions in terms of position (X and Y) as well as
temperature (T); therefore, the radiative surface condition can be modeled as shown in Figure 5.
Figure 5: Extended Surface Conditions Dialog window for Regions 1 and 3.
Region 2 experiences a convection boundary condition with the coolant, but only from one side
(as opposed to the double sided condition assumed by FEHT). The heat transfer from a
differential element in Region 2 is given by:
q = h dA (T Tc ) (10)
whereas the convection surface condition in FEHT assumes convection from both sides of the
plate and therefore:
q = hF 2 dA (T Tc ) (11)
where hF is the heat transfer coefficient that should be set in FEHT in order to simulate the
single-sided convection coefficient represented by Eq. (10). Comparing Eqs. (10) and (11) leads
to:
h
hF = (12)
2
Click on Region 2 and select Surface Condition from the Specify menu; specify the surface
conditions as shown in Figure 6.
Finally, the boundary conditions along each edge of the computational domain must be specified;
the line of symmetry is adiabatic and the remaining edges are also assumed to be adiabatic.
Generate a reasonable but crude mesh, as shown in Figure 7(a) and then refine it. Note that the
mesh within Region 2 will start relatively refined due to its small width and need not be refined
as much as the mesh in Regions 1 and 3. However, it is possible to refine the mesh in a single
region by selecting that region and then selecting Reduce Mesh from the Draw menu. The result
should be similar to Figure 7(b).
(a) (b)
Figure 7: (a) coarse and (b) refined mesh.
Solve the problem by selecting Calculate from the Run menu; the problem is non-linear due to
the temperature dependent heat transfer coefficient and therefore the solution process will be
iterative. Plot the temperature distribution by selecting Temperature Contours from the View
menu (Figure 8).
Figure 8: Temperature distribution for a 1 cm thick plate.
b.) Export the solution to EES and calculate the total heat transferred from the radiator and the
radiator efficiency (defined as the ratio of the radiator heat transfer to the heat transfer from
the radiator if it were isothermal and at the coolant temperature).
Select Tabular Output from the View menu and then Select All and Save As. Save the data as a
file called 1 cm. Open EES and select Open Lookup Table from the Tables menu; navigate the
file 1 cm and open it; the solution (the temperature at each node together with the locations of
the nodes) is contained in the lookup table. Using the technique discussed in EXAMPLE 16-1 it
is possible to use the Interpolate2D function to obtain the temperature at an arbitrary x and y
location on the radiator plate.
"INPUTS"
e=1.0 [-] "emittance of panel surface"
a=0.5 [m] "half-width of panel"
b=0.75 [m] "height of panel"
T_f=320 [K] "fluid temperature"
th=0.01 [m] "thickness"
rho=2700 [kg/m^3] "density"
k=200 [W/m-K] "conductivity"
"location on plate"
x=0.1 [m]
y=0.5 [m]
"2-D interpolation from table of nodal data"
T=Interpolate2D('1 cm',X,Y,T,X=x,Y=y) "temperature interpolated from data"
In order to obtain the total heat transferred from the panel it is necessary to integrate the heat flux
over the entire area of the plate. This process can be accomplished manually by dividing the
plate into many small integration areas, calculating the heat flux within each area, and summing
the result. The plate is divided into Nx segments in the x direction, each with width:
a
x = (13)
Nx
xi = ( i 0.5 ) x (14)
The plate is divided into Ny segments in the y direction, each with height:
b
y = (15)
Ny
y j = ( j 0.5 ) y (16)
The total heat transferred from the plate ( q ) is calculated using the double integral:
b a
q = q x , y dx dy (17)
0 0
The numerical summation that approximates the integration in Eq. (17) is:
N y Nx
q = q xi , y j x y (19)
j =1 i =1
where
q xi , y j = Tx4i , y j (20)
duplicate i=1,Nx
duplicate j=1,Ny
"2-D interpolation from table of nodal data"
T[i,j]=Interpolate2D('1 cm',X,Y,T,X=x[i],Y=y[j]) "temperature interpolated from data"
q``_dot[i,j]=2*e*sigma#*T[i,j]^4 "heat flux"
end
end
q_dot=sum(q``_dot[1..Nx,1..Ny])*Dx*Dy
Note that the same process can be accomplished using EES native Integral command. The
Integral command carries out numerical integration using a more sophisticated algorithm than
simply assuming a constant value over each step.
EES Integral command requires 4 arguments and allows a 5th, optional argument; the protocol
for calling the function is:
F=Integral(Integrand,VarName,LowerLimit,UpperLimit,StepSize)
where Integrand is the EES variable or expression that must be integrated, VarName is the
integration variable, and LowerLimit and UpperLimit define the limits of integration. StepSize
is optional and defines the numerical step that is used to accomplish the integration; a small
value of StepSize will lead to more accurate results but take longer to calculate.
The double integral in Eq. (17) can be accomplished by calling the Integral function twice; Eq.
(17) is rewritten as:
b
q = qy dy (21)
0
where
a
q y = q x , y dx (22)
0
Equation (22) is evaluated using the Integral command as shown in the EES code below:
Note that the value of y was set as q y is a function of y but the value of x is not set as x is the
integration variable. To evaluate Eq. (21), integrate q y from y=0 to y=b:
Note that the value of y is no longer set as y is the integration variable in the 2nd integral. The
results of the manual integration should agree with the use of the Integral function; either should
yield 339 W. The radiator efficiency () and mass (M) are computed according to:
q
= (23)
2 a b T f4
and
M = a b th (24)
c.) Explore the effect of thickness on the radiator efficiency and mass.
The solution procedure described above is repeated for several values of the radiator thickness.
Figure 10 shows the predicted temperature distribution for several values of the thickness. Note
that the same contour levels were selected for each plot by selecting User in the Temperature
Contour Information dialog window and specifying the range from 200 K to 320 K (Figure 9);
this allows the different cases to be compared directly and shows clearly that the reduced
thickness increases the resistance to conduction along the plate and therefore leads to
progressively larger temperature gradients through the plate.
Figure 9: Temperature Contour Information dialog window.
Figure 10: Temperature distribution in the panel for thicknesses of (a) 1.0 cm, (b) 0.75 cm, (c) 0.5 cm, (d) 0.3
cm, (e) 0.2 cm, and (f) 0.1 cm.
Figure 11 illustrates the radiator efficiency and mass as a function of the thickness. As the
thickness is reduced, the efficiency drops but so does the mass; clearly there must be a trade-off
between these effects.
Figure 11: Efficiency and mass as a function of thickness.
Figure 12 shows the efficiency as a function of mass and makes the trade-off clearer; above a
panel with a mass of nominally 2 kg there is a region of diminishing return where additional
mass provides only a small gain in efficiency.
A complete analysis would require more information then is given; specifically, how should the
radiator performance be compared with its mass to determine a true optimal thickness. Barring
additional constraints related to, for example, structural stability, Figure 12 suggests that any
optimization process will result in a panel that is approximately 0.2 cm thick with a mass of 2 kg
and an efficiency of 60%.
Problem 2.8-1 (2-12 in text): Cryogenic Thermal Switch
There are several cryogenic systems that require a thermal switch, a device that can be
used to control the thermal resistance between two objects. One class of thermal switch
is activated mechanically and an attractive method of providing mechanical actuation at
cryogenic temperatures is with a piezoelectric stack; unfortunately, the displacement
provided by a piezoelectric stack is very small, typically on the order of 10 microns. A
company has proposed an innovative design for a thermal switch, shown in Figure P2.8-
1(a). Two blocks are composed of th = 10 m laminations that are alternately copper (kCu
= 400 W/m-K) and plastic (kp = 0.5 W/m-K). The thickness of each block is L = 2.0 cm
in the direction of the heat flow. One edge of each block is carefully polished and these
edges are pressed together; the contact resistance associated with this joint is Rc = 5x10-4
K-m2/W.
th = 10 m plastic laminations
kp = 0.5 W/m-K
L = 2 cm L = 2 cm direction of actuation
TH TC
Figure P2.8-1(b)
on position off position
contact resistance, Rc = 5x10 m -K/W
-4 2
th = 10 m copper laminations
kCu = 400 W/m-K
Figure P2.8-1(a): Thermal switch in the on and off positions.
Figure P2.8-1(a) shows the orientation of the two blocks when the switch is in the on
position; notice that the copper laminations are aligned with one another in this
configuration which provides a continuous path for heat through high conductivity
copper (with the exception of the contact resistance at the interface). The vertical
location of the right-hand block is shifted by 10 m to turn the switch "off". In the off
position, the copper laminations are aligned with the plastic laminations; therefore, the
heat transfer is inhibited by low conductivity plastic. Figure P2.8-1(b) illustrates a closer
view of half (in the vertical direction) of two adjacent laminations in the on and off
configurations. Note that the repeating nature of the geometry means that it is sufficient
to analyze a single lamination set and assume that the upper and lower boundaries are
adiabatic.
L = 2 cm L = 2 cm
TH TC
on position
TH TC
off position
Figure P2.8-1(b): A single set consisting of half of two adjacent laminations in the on and "off
positions.
The key parameter that characterizes a thermal switch is the resistance ratio (RR) which is
defined as the ratio of the resistance of the switch in the off position to its resistance in
the on position. The company claims that they can achieve a resistance ratio of more
than 100 for this switch.
a) Estimate upper and lower bounds for the resistance ratio for the proposed thermal
switch using 1-D conduction network approximations. Be sure to draw and clearly
label the resistance networks that are used to provide the estimates. Use your results
to assess the companys claim of a resistance ratio of 100.
"Inputs"
th = 10 [micron]*convert(micron,m) "thickness of laminations"
k_Cu=400 [W/m-K] "conductivity of copper laminations"
k_p=0.5 [W/m-K] "conductivity of plastic laminations"
L = 2.0 [cm]*convert(cm,m) "lengt of laminations"
R``_c=5e-4 [K-m^2/W] "area specific contact resistance of interface"
W=1 [m] "unit depth into page"
The isothermal and adiabatic models provide two limiting cases. Both result in a 1-D
problem that can be represented using a resistance network. The adiabatic
approximation does not allow heat transfer in the y-direction (i.e., perpendicular to the
laminations). The resistance network for the adiabatic approximation is shown in Figure
3(a) for the on position and in Figure 3(b) for the off position.
(a) (b)
Figure 3: Resistance network for the adiabatic approximation in the (a) on and (b) off
positions.
Note that the parameter Alam in Figure 3 is the cross-sectional area associated a single
lamination:
Alam = W th
where W is the depth into the page (assumed to be 1 m). The resistance associated with
the adiabatic approximation in the on state, Rad,on from Figure 3(a) is:
1
2 1 1
Rad ,on = +
Alam 2 L + R 2 L + R
k c
kCu
c
p
which leads to Rad,on = 119.1 K/W. The resistance associated with the adiabatic
approximation in the off state, Rad,off from Figure 3(b) is:
1
L L
+ + Rc
2 1 1 k p kCu
Rad ,off = + =
Alam L + L + R L + L + R Alam
k c c
p kCu kCu k p
Rad ,off
RRad =
Rad ,on
RR_ad=R_ad_off/R_ad_on "resistance ratio estimated using adiabatic limit"
(a) (b)
Figure 4: Resistance network for the isothermal approximation in the (a) on and (b) off
positions.
By inspection, the two resistance networks shown in Figures 4(a) and 4(b) will yield the
same resistance and therefore the isothermal assumption will predict a resistance ratio,
RRiso, of 1.0. Clearly the companys claim of a resistance ratio of 100 is not possible as it
does not lie between the two bounding quantities associated with the isothermal and
adiabatic approximations.
b) Provide one or more suggestions for design changes that would improve the
performance of the switch (i.e., increase the resistance ratio). Justify your
suggestions.
The resistance ratio would be increased by any change that causes the two resistance
networks in Figure 3 to be more different. Possible improvements include using
materials with a larger ratio of conductivities (i.e., lower conductivity plastic and, to a
lesser degree, higher conductivity copper) or eliminating the contact resistance. From a
more practical standpoint, any design change that causes the actual the device to behave
more like the adiabatic approximation in Figure 3 than the isothermal approximation in
Figure 4 will improve the performance; for example, increasing the contact resistance
between adjacent laminations would be important.
c.) Sketch the temperature distribution through the two parallel paths associated with the
adiabatic limit of the switchs operation in the off position. Do not worry about the
quantitative details of the sketch, just make sure that the qualitative features are
correct.
Figure 5: Qualitative temperature distribution through paths 1 and 2 consistent with the adiabatic
approximation when the switch is in the off position.
d.) Sketch the temperature distribution through the two parallel paths associated with the
adiabatic limit in the on position. Again, do not worry about the quantitative details
of your sketch, just make sure that the qualitative features are correct.
Figure 6: Qualitative temperature distribution through paths 1 and 2 consistent with the adiabatic
approximation when the switch is in the on position.
Problem 2.8-2 (2-13 in text): Resistance of a Bus Bar
Figure P2.8-2 illustrates a thermal bus bar that has width W = 2 cm (into the page).
H1 = 5 cm L2 = 7 cm
TH = 80C h = 10 W/m -K
2
T = 20C
k = 1 W/m-K
L1 = 3 cm H2 = 1 cm
Figure P2.8-2: Thermal bus bar.
The bus bar is made of a material with conductivity k = 1 W/m-K. The middle section is L2 = 7
cm long with thickness H2 = 1 cm. The two ends are each L1 = 3 cm long with thickness H1 = 3
cm. One end of the bar is held at TH = 80C and the other is exposed to air at T = 20C with h
= 10 W/m2-K.
a.) Use FEHT to predict the rate of heat transfer through the bus bar.
The geometry is entered in FEHT, a mesh is generated, and the boundary conditions and material
properties are specified (Figure 2).
The mesh is refined several times and then the heat transfer at the convective boundary is
obtained; this leads to q = 5.40 W/m or q = 0.108 W.
b.) Obtain upper and lower bounds for the rate of heat transfer through the bus bar using
appropriately defined resistance approximations.
"Inputs"
L_1=3 [cm]*convert(cm,m) "length of edge pieces"
H_1=5 [cm]*convert(cm,m) "height of edge pieces"
L_2=7 [cm]*convert(cm,m) "length of center piece"
H_2=1 [cm]*convert(cm,m) "height of center piece"
W=2 [cm]*convert(cm,m) "width"
k=1 [W/m-K] "conductivity"
h_bar=10 [W/m^2-K] "heat transfer coefficient at right side"
T_infinity=20 [C] "ambient temperature at right side"
T_H=80 [C] "left side temperature"
An upper bound on the heat transfer rate is obtained using the isothermal limit. The isothermal
limit is calculated according to:
qiso =
(TH T )
1 L1 L2
+2 +
h H1 W k H1 W k H 2 W
"Isothermal limit"
R_conv_iso=1/(h_bar*W*H_1)
R_cond_1_iso=L_1/(k*W*H_1)
R_cond_2_iso=L_2/(k*W*H_2)
R_total_iso=(R_conv_iso+2*R_cond_1_iso+R_cond_2_iso)
q_dot_iso=(T_H-T_infinity)/R_total_iso
A lower bound on the heat transfer rate is obtained using the adiabatic limit. The adiabatic limit
is calculated according to:
qad =
(TH T )
1 L1 L2
+2 +
h H2 W k H2 W k H2 W
"Adiabatic limit"
R_conv_ad=1/(h_bar*W*H_2)
R_cond_1_ad=L_1/(k*W*H_2)
R_cond_2_ad=L_2/(k*W*H_2)
R_total_ad=R_conv_ad+2*R_cond_1_ad+R_cond_2_ad
q_dot_ad=(T_H-T_infinity)/R_total_ad
h = 40 W/m -K
2
Rc = 1x10 K-m /W
-4 2
rin = 8 mm
ro,lam = 20 mm
ro,p = 25 mm
ro,c = 35 mm
Figure P2.9-2: Laminated stator.
A heat flux associated with the windage loss associated with the drag on the shaft is q = 5x104
W/m2 is imposed on the internal surface of the laminations. The outer surface of the motor is
exposed to air at T = 20C with a heat transfer coefficient h = 40 W/m2-K. There is a contact
resistance Rc = 1x10-4 K-m2/W between the outer surface of the laminations and the inner
surface of the plastic and the outer surface of the plastic and the inner surface of the motor
housing.
a.) Determine an upper and lower bound for the temperature at the inner surface of the
laminations (Tin).
An upper bound on the temperature allows no heat spreading; therefore, the heat must take two
parallel paths that pass through the iron and the epoxy. The total resistance associated with one
lamination/epoxy pair is:
1 1
=
Rtotal ,ad r r r
ln o ,lam ln o , p ln o ,c
rin + Rc
+
ro ,lam
+
Rc
+ ro , p + 1
2 klam thlam 2 ro ,lam thlam 2 k p thlam 2 ro , p thlam 2 kc thlam 2 ro, p thlam h
1
+
r r r
ln o ,lam ln o , p ln o ,c
r
rin + Rc ro ,lam Rc o, p + 1
+ + +
2 kepoxy thepoxy 2 ro ,lam thepoxy 2 k p thepoxy 2 ro , p thepoxy 2 kc thepoxy 2 ro, p thepoxy h
(1)
b.) You need to reduce the internal surface temperature of the laminations and there are a few
design options available, including: (1) increase the lamination thickness (up to 0.7 mm), (2)
reduce the epoxy thickness (down to 0.05 mm), (3) increase the epoxy conductivity (up to 2.5
W/m-K), or (4) increase the heat transfer coefficient (up to 100 W/m-K). Which of these
options do you suggest and why?
Examination of the Solution Window (Figure 2) shows that the resistance of the lamination is
much less than the resistance of the epoxy; therefore, the resistance of the lamination dominates
the resistance of the epoxy. The resistance due to convection is much larger than the resistance
of the lamination, contact resistance, or conduction through the plastic and casing. Therefore,
the resistance to convection dominates the problem and the most effective mechanism for
reducing the temperature is to increase the heat transfer coefficient.