Professional Documents
Culture Documents
201406980
ME 505: Assignment 1
Problem 1:
Write the generalized transport equation in conservative form and non-conservative form (in
Vector notation) and show that both forms are equivalent.
The generalized transport equation for a conserved quantity in conservative form is given by:
+ . () = . ( ) +
(1)
To write Eq. (1) in non-conservative from, we need to expand the derivative in the LHS of Eq. (1):
+
+ . () + (. ) = . ( ) +
(2)
+ . ()) + ( + (. )) = . ( ) +
(3)
+ . ()) = 0
(4)
+ (. )) = . ( ) +
(5)
(
Using Eq. (4) in Eq. (3) gives:
(
identifying that
+ (. ) =
gives:
1
= . ( ) +
(6)
Problem 2:
Fill the table for the generalized conservation equation as discussed in lecture 3.
Quantity
Mass
x-Momentum
y-Momentum
z-Momentum
Enthalpy
Temperature
Species mass
fraction
Symbol
0
.
)
(
+ ( + )
.
)
(
+
+ ( + )
.
)
(
+
+ ( + )
+ +
( +
) /
Meaning
Dynamic viscosity
Volumetric body force
Bulk viscosity
Thermal conductivity
Specific heat
Density
Viscous dissipation
Problem 3:
We like to solve the 1D heat conduction using 3 methods for the following 2 problems:
Case 1: A flat plate of thickness L=2cm, k=0.5W/(mK) and heat generation of 1000 kW/m3.
The plate is maintained between Tleft=100C and Tright of 200C.
Case 2: A circular fin of constant cross-sectional area A, perimeter P, thermal conductivity k,
heat transfer coefficient h and length L=1m. The fin base temperature is TB=100C and the other
tip is insulated. The ambient temperature is 20C, and hP/(kA)=25m-2.
Solution Methods:
1. Exact
2. Finite Volume with cell length i) L/5, ii) L/10 and iii) L/50
3. Finite difference method with node spacing of i) L/5, ii) L/10 and iii) L/50
For both FV and FD use direct solution method (Matrix inversion) and iterative method (e.g. Gauss
Sidel)
Deliverable
a) Please show your derivation of the discretized equation for Finite volume and finite
difference for a typical finite volume (or typical node) and for boundary nodes if needed.
b) Compare the temperature results for exact, FV (3 grids, direct, iterative) and FD (3 grids,
direct, iterative). Use tables and plots and indicate % difference
c) Do the same for the heat fluxes.
d) Compare the time of execution between 2iii) and 3iii) and the matrix inversion part. (If small
difference in time, use L/100.)
e) Discuss your results briefly.
Case 1:
Part a)
Derivation of the discretized equation for finite volume:
The governing equation is:
( ) =
(7)
Integrating Eq.(7) along the volume P (see Fig. 1) and using the divergence theorem to convert
the volume integral of the LHS of Eq.(7) to surface integral gives:
e
P
) | ( ) | =
(8)
Where is the average of over the volume, and is the area of the volume cell faces.
Approximating the derivatives at faces e and w by central difference approximation gives:
(
Using =
) (
) =
(9)
(10)
(11)
= + = , =
Finite volume left boundary discretized equation:
Integrating Eq.(7) along the volume P (see Fig. 2) and discretizing gives:
e
B
/2
) (
) =
/2
(12)
(13)
= + , = , = + , = 2
Finite volume right boundary discretized equation:
The discretized equation are the same as Eq. (13) with the coefficients being:
= + , = , = + , = 2
Discretizing the second derivative in the LHS of Eq.(7) with central difference approximation
yields (see Fig. 3)
i-1
i+1
(1 + 2 +1 )
()2
= = 1 + 2 +1 =
()2
(14)
()2
22 3 =
+ 1
(15)
()2
=
+
(16)
Part b)
The exact solution of Eq.(7) is:
( ) =
2 2
2
+ (( ) +
) +
2
2
(17)
Where = /.
For finite volume method, the temperature was solved at the locations:
1
= [0 ,
3
2
(21)
2
For finite differences method, the temperature was solved at the locations:
1
| |
100
(18)
Figures 5 and Error! Reference source not found. and Table 1 and Table 2 show comparison
between the finite volume and finite differences solutions with the exact solution. Note that in Fig.
5 and Error! Reference source not found. only direct method solution are shown. This is because
the difference between the direct method and iterative method
a)
is
b)
Fig. 5 Comparison of temperature results for finite volume method using multiple cell sizes. a)
Temperature distribution b) error percentage compared with exact solution
7
Table 1. Comparison between the exact and finite volume method for different cell sizes.
0.1
0.5
0.9
146
250
226
0.05
0.45
0.85
124
244
236
0.01
0.35
0.69
104.96
226.00
254.56
= /5
150
254
230
149.9999969
253.9999937
229.9999986
= /10
125
245
237
124.9999946
244.9999727
236.9999897
= /50
105.00
226.04
254.60
104.9999765
226.0393554
254.5994214
2.7397260
1.6000000
1.7699115
0.806451613
0.409836066
0.423728814
0.038109756
0.017699115
0.015713388
2.7397239
1.5999975
1.7699109
0.806447265
0.40982489
0.423724441
0.038087346
0.017413908
0.015486107
a)
b)
Fig. 6 Comparison of temperature results for finite differences method using multiple node
spacing. a) Temperature distribution b) error percentage compared with exact solution
8
Table 2. Comparison between the exact and finite differences method for different node
spacing.
0.2
0.6
0.8
184
256
244
0.1
0.4
0.9
146
236
226
0.02
0.36
0.98
109.84
228.16
205.84
= /5
184
256
244
183.9999969
255.9999967
243.9999983
= /10
146
236
226
145.9999909
235.9999759
225.9999939
= /50
109.84
228.16
205.84
109.8399535
228.1593524
205.8399577
0
1.11E-14
0
0
2.41E-14
0
1.04E-13
7.47E-13
6.90E-14
1.71E-06
1.30E-06
6.84E-07
6.23E-06
1.02E-05
2.69E-06
4.23E-05
0.000283853
2.05E-05
Part c)
The exact heat flux is given by:
( )
(19)
= = (
+ )
2
For finite volume method, the flux was obtained using central difference approximation to
approximate the temperature derivative (except at the boundaries in which forward or backward
differences is used as appropriate). The locations of the approximate flux is at:
Figures 7 and Table 3 and 4 show comparison between the finite volume and finite differences
flux solutions with the exact flux solution.
a)
b)
Fig. 7 Comparison of flux results for finite volume method using multiple cell sizes. a) flux
distribution b) error percentage compared with exact solution
Table 3. Comparison between the exact and finite volume method for different cell sizes.
0
0.4
0.8
-12500
-4500
3500
0
0.3
0.6
-12500
-6500
-500
0
0.34
0.68
-12500
-5700
1100
= /5
-12500
-4500
3500
-12499.9992
-4500.00000
3499.99965
= /10
-12500
-6500
-500
-12499.9973
-6499.99893
-500.000916
= /50
-12500
-5700
1100
-12499.9412
-5699.97417
1099.969079
0
0
2.08E-13
0
-1.12E-13
0
-5.68E-13
-5.58E-13
6.82E-13
-6.28E-06
0
9.97E-06
-2.16E-05
-1.64E-05
-0.00018
-0.00047
-0.00045
0.002811
10
Fig. 8 Comparison of flux results for finite differences method using multiple node spacing. a)
flux distribution b) error percentage compared with exact solution
Table 4. Comparison between the exact and finite differences method for different node
spacing.
0
0.4
1
-12500
-4500
7500
0
0.3
1
-12500
-6500
7500
0
0.34
1
-12500
-5700
7500
= /5
-10500
-4500
5500
-10500
-4500
5500
= /10
-11500
-6500
6500
-11500
-6500
6499.998
= /50
-12300
-5700
7300
-12299.9
-5699.97
7299.947
-16
-4.04E-14
26.66667
-8
-1.12E-13
13.33333
-1.6
-1.80E-12
2.666667
-16
-2.58E-07
26.66667
-8.00002
-1.47E-05
13.33335
-1.60046
-0.00045
2.667371
11
Part d)
Figure 9 Comparison of time of execution between finite volume and finite differences shows
comparison between time of execution of finite volume and finite differences using direct
inversion method.
Fig. 9 Comparison of time of execution between finite volume and finite differences
Part e)
Finite volume method gives accurate results for the temperature distribution even with only 5
volume cells. As can be seen in Fig. 5, the error is less than 3%. Table 1 show that the direct and
iterative methods of solution given essentially same results. Finite differences is more accurate
than finite volume for this problem as can be seen by comparing Fig. 5 and Fig. 6. For heat flux
values, both finite volume and finite differences are very accurate. However, the boundary flux for
the finite differences has substation error probably because of first order derivative approximation
that is used for calculation. In terms of computational cost, Fig. 9 shows that finite differences is
little bit more costly compared with finite volume. In general, an order of magnitude increase in
the size of the system gives order of magnitude increase in terms of computation time.
12
Case 2:
Part a)
Derivation of the discretized equation for finite volume:
The governing equation is:
( ) =
= ,
(20)
=
Integrating Eq.(20) along the volume P (see Fig. 1) and using the divergence theorem to convert
the volume integral of the LHS of Eq.(20) to surface integral gives:
) | ( ) | =
(21)
Where is the average of over the volume, and is the area of the volume cell faces.
Approximating the derivatives at faces e and w by central difference approximation gives:
(
Using =
) (
) =
1 +
Evaluating the integral in Eq. (24) using trapezoidal approximation gives:
=
=
+
2
(22)
+
2
(23)
(24)
(25)
13
(26)
=
(27)
= + = , =
Finite volume left boundary discretized equation:
Integrating Eq.(20) along the volume P (see Fig. 2) and discretizing gives:
) (
) =
/2
(28)
(29)
= , = , = = , = 2
Finite volume right boundary discretized equation:
The discretized equation are the same as Eq.(29) with the coefficients being:
= , = , = = 0
Note that the right boundary value = where N is the last cell volume. This due to the zero
flux boundary condition at the right boundary.
Derivation of the discretized equation for finite differences:
Typical node discretized equation:
Discretizing the second derivative in the LHS of Eq.(20) with central difference approximation
yields (see Fig. 3)
14
(1 2 + +1 )
= 0 = 1 + (2 + 2 ) +1 = 0
()2
(30)
(31)
|
= =/2
/2
|
= 0, |=/2 =
=
(32)
( 1 )
|=
|
=/2 = 2( 1 )
/2
2
(33)
(34)
Part b)
The exact solution of Eq.(20) is:
( )
cosh ((1 ))
cosh()
(35)
15
Where = /.
For finite volume method, the temperature was solved at the locations:
1
= [0 ,
3
2
(21)
2
For finite differences method, the temperature was solved at the locations:
1
| |
100
(36)
Figures 10 and 11 and Tables 5 and 6 show comparison between the finite volume and finite
differences solutions with the exact solution.
Fig. 10 Comparison of temperature results for finite volume method using multiple cell sizes. a)
Temperature distribution b) error percentage compared with exact solution
16
Table 5. Comparison between the exact and finite volume method for different cell sizes.
0.1
0.5
0.9
0.05
0.45
0.85
62.3059
8.466013
1.395698
0.01
0.35
0.69
48.52624
6.610747
1.215606
76.09872
13.92218
2.65394
= /5
44.22764
6.504065
1.300813
44.22764
6.504066
1.300813
= /10
60.5991
8.404587
1.417561
60.5991
8.404595
1.417566
= /50
76.00535
13.915
2.654584
76.00536
13.91515
2.654807
8.858291
1.613758
7.009404
2.739384
0.725558
1.566481
0.122686
0.051608
0.024294
8.858289
1.613741
7.009434
2.739381
0.725465
1.566856
0.122679
0.05053
0.032697
Fig. 11 Comparison of temperature results for finite differences method using multiple node
spacing. a) Temperature distribution b) error percentage compared with exact solution
17
Table 6. Comparison between the exact and finite differences method for different node
spacing
0.2
0.6
0.8
29.43889
4.055732
1.663476
0.1
0.4
0.9
48.52624
10.85317
1.215606
0.02
0.36
0.98
72.38772
13.24528
1.083417
= /5
30.56911
4.552846
1.95122
30.56911
4.552847
1.951221
= /10
48.77309
11.07705
1.275805
48.77309
11.07706
1.275812
= /50
72.39074
13.25526
1.085669
72.39075
13.25541
1.085905
3.839187
12.25706
17.29774
0.508691
2.062835
4.952162
0.004166
0.075368
0.207881
3.839192
12.2571
17.29781
0.508698
2.06291
4.952745
0.004179
0.076498
0.229631
Part c)
The exact heat flux is given by:
sinh ((1 ))
1
= =
cosh()
(37)
For finite volume method, the flux was obtained using central difference approximation to
approximate the temperature derivative (except at the left boundary in which forward differences
is used). The locations of the approximate flux is at:
1
18
Fig. 12 Comparison of flux results for finite volume method using multiple cell sizes. a) flux
distribution b) error percentage compared with exact solution
Table 7. Comparison between the exact and finite volume method for different cell sizes.
0
0.4
0.8
399.9637
53.99748
6.334467
0
0.3
0.6
399.9637
89.16663
19.54919
0
0.34
0.68
399.9637
72.97069
12.80458
= /5
357.7236
52.03252
6.504065
357.7236
52.03252
6.504067
= /10
388.018
87.82304
19.53576
388.018
87.82303
19.53577
= /50
399.4645
72.93094
12.80599
399.464
72.9306
12.80589
10.56099
3.638978
2.677383
2.986689
1.506834
0.06866
0.124804
0.054469
0.010982
10.56099
3.638976
2.677412
2.986699
1.506846
0.06863
0.12493
0.054935
0.010219
19
Fig. 13 Comparison of flux results for finite differences method using multiple node spacing. a)
flux distribution b) error percentage compared with exact solution
Table 8. Comparison between the exact and finite differences method for different node
spacing.
0
0.4
0.8
399.9637
53.99748
6.334467
0
0.3
0.9
399.9637
89.16663
2.808763
0
0.34
0.98
399.9637
72.97069
0.53991
= /5
247.1545
65.04065
8.130081
247.1545
65.04065
8.130083
= /10
312.2691
93.31197
3.012317
312.2691
93.31196
3.012325
= /50
380.4632
73.11327
0.541484
380.4627
73.11293
0.541537
38.20577
20.45128
28.34673
21.92563
4.648989
7.247139
4.87557
0.195395
0.291582
38.20577
20.45128
28.34676
21.92564
4.648974
7.247402
4.87569
0.194928
0.301383
20
Part d)
Figure 14 shows comparison between time of execution of finite volume and finite differences
using direct inversion method.
Fig. 14 Comparison of time of execution between finite volume and finite differences
Part e)
Both finite volume and finite differences method give reasonable accurate results for the
temperature distribution. The accuracy increases quickly with the number of cells and with the
decrease in number of nodes. Generally Case 2 accuracy is less than Case1 for the same cell
numbers. The same trends goes for heat flux values with the finite volume being more accurate
than finite differences. Errors in heat flux are maximum at the left boundary for both finite volume
and finite differences. In terms of computational cost, Fig. 14 shows that finite differences is an
order of magnitude slower compared with finite volume. This is in contrast with Case 1 in which
both finite volume and finite differences have essentially the same time of execution.
21