ABSTRACT
The effect of round-off errors on the solution of numerical heat transfer is illustrated by a simple example both analytically and numerically. It is found that the upper bound of the round-off error under both conditions with or without an inner heat source is proportional to the square of grid number—n2. Increase in grid number might lead to larger round-off errors. The magnitude of relative round-off error is also determined by the specific problem. Proper treatment of the computation procedure can reduce the round-off error obviously. The precision can be improved with this method without occupation of additional computational resources.
Nomenclature
A, B, C, P, Q | = | coefficients in TDMA |
a, b | = | coefficients |
Bi | = | Biot number |
E | = | round-off error |
Er | = | relative round-off error |
e | = | exponent; relative error in basic operations |
F | = | set of floating-point number |
F | = | cross-sectional area, m2 |
f | = | arbitrary function |
fl | = | floating point |
h | = | convective heat transfer coefficient, W/m2 |
k | = | overall heat transfer coefficients, W/m2/K |
L | = | length of the slab, m |
M1 | = | grid number of the rightmost grid point |
m | = | significant |
op | = | operation |
p | = | precision |
R | = | set of real number |
ro | = | round function |
S | = | inner heat source, W/m3 |
s | = | sign |
T | = | temperature, °C |
x | = | arbitrary vector |
x | = | variable; coordinate along the slab, m |
δ | = | maximal absolute error for numbers close to zero |
δx | = | distance between grid points, m |
ε | = | relative error in basic operations |
ò | = | machine precision |
λ | = | thermal conductivity, W/m/K |
ϕ | = | solution of the PDE |
Subscripts | = | |
1 | = | left side of the slab |
2 | = | right side of the slab |
E | = | east |
f | = | fluid |
i | = | TDMA calculation step; grid number |
n | = | total grid number |
P | = | current grid point |
up | = | upper bound |
W | = | west |
Nomenclature
A, B, C, P, Q | = | coefficients in TDMA |
a, b | = | coefficients |
Bi | = | Biot number |
E | = | round-off error |
Er | = | relative round-off error |
e | = | exponent; relative error in basic operations |
F | = | set of floating-point number |
F | = | cross-sectional area, m2 |
f | = | arbitrary function |
fl | = | floating point |
h | = | convective heat transfer coefficient, W/m2 |
k | = | overall heat transfer coefficients, W/m2/K |
L | = | length of the slab, m |
M1 | = | grid number of the rightmost grid point |
m | = | significant |
op | = | operation |
p | = | precision |
R | = | set of real number |
ro | = | round function |
S | = | inner heat source, W/m3 |
s | = | sign |
T | = | temperature, °C |
x | = | arbitrary vector |
x | = | variable; coordinate along the slab, m |
δ | = | maximal absolute error for numbers close to zero |
δx | = | distance between grid points, m |
ε | = | relative error in basic operations |
ò | = | machine precision |
λ | = | thermal conductivity, W/m/K |
ϕ | = | solution of the PDE |
Subscripts | = | |
1 | = | left side of the slab |
2 | = | right side of the slab |
E | = | east |
f | = | fluid |
i | = | TDMA calculation step; grid number |
n | = | total grid number |
P | = | current grid point |
up | = | upper bound |
W | = | west |
Acknowledgment
This work has been financially supported by the National Key Research and Development Program — China (2016YFB0601201).