ABSTRACT
In the present study, a feed-forward artificial neural network (ANN) was developed to estimate the Nusselt number (Nu), friction factor (f), and thermal performance (η) in a tube equipped with perforated twisted tape. The MSE and R2 values of the best network (4-10-3-3) are 0.04 and 0.9965, respectively. In addition, empirical equations were developed using the genetic algorithm. The MSE values of power-law equations for Nu, f and η are 2.7369, 4.13E-06, and 5.09E-05, respectively. Estimation of the developed ANN was found to be superior in comparison with the corresponding power-law equations.
Nomenclature
A | = | heat transfer surface area |
ANN | = | artificial neural network |
BP | = | back-propagation |
CP | = | specific heat of fluid |
D | = | inside diameter of the test tube |
d | = | diameter of perforated tape |
f | = | friction factor = ΔP/((L/D)(ρU2/2)) |
h | = | heat transfer coefficient |
I | = | current |
k | = | thermal conductivity of fluid |
L | = | length of test section |
M | = | mass flow rate |
Nu | = | Nusselt number |
P | = | flow pressure in stationary tube |
ΔP | = | pressure drop |
Pr | = | Prandtl number |
Q | = | heat transfer rate |
Re | = | Reynolds number = ρUD/µ |
s | = | spaced-pitch length of perforated tape, mm |
t | = | width of test tube |
T | = | temperature |
= | mean temperature | |
U | = | mean axial flow velocity |
V | = | voltage |
W | = | twisted tape width, weight in ANN |
X | = | input value of ANN |
Y | = | output value of ANN |
y | = | twisted tape pitch |
ρ | = | fluid density |
δ | = | twisted tape thickness |
δi | = | vector of errors for each hidden layer neuron |
δk | = | vector of errors for each output neuron |
µ | = | fluid dynamic viscosity |
η | = | thermal performance factor, momentum factor |
θj | = | threshold between input and hidden layers |
θk | = | threshold connecting hidden and output layers |
fh() | = | logistic sigmoid activation function from input layer to hidden layer |
fk() | = | logistic sigmoid activation function from hidden layer to output layer |
∝ | = | learning rate |
Subscripts | = | |
b | = | bulk |
c | = | convection |
i | = | inlet |
o | = | outlet |
p | = | plain |
s | = | surface |
t | = | twisted tape |
w | = | Water |
Nomenclature
A | = | heat transfer surface area |
ANN | = | artificial neural network |
BP | = | back-propagation |
CP | = | specific heat of fluid |
D | = | inside diameter of the test tube |
d | = | diameter of perforated tape |
f | = | friction factor = ΔP/((L/D)(ρU2/2)) |
h | = | heat transfer coefficient |
I | = | current |
k | = | thermal conductivity of fluid |
L | = | length of test section |
M | = | mass flow rate |
Nu | = | Nusselt number |
P | = | flow pressure in stationary tube |
ΔP | = | pressure drop |
Pr | = | Prandtl number |
Q | = | heat transfer rate |
Re | = | Reynolds number = ρUD/µ |
s | = | spaced-pitch length of perforated tape, mm |
t | = | width of test tube |
T | = | temperature |
= | mean temperature | |
U | = | mean axial flow velocity |
V | = | voltage |
W | = | twisted tape width, weight in ANN |
X | = | input value of ANN |
Y | = | output value of ANN |
y | = | twisted tape pitch |
ρ | = | fluid density |
δ | = | twisted tape thickness |
δi | = | vector of errors for each hidden layer neuron |
δk | = | vector of errors for each output neuron |
µ | = | fluid dynamic viscosity |
η | = | thermal performance factor, momentum factor |
θj | = | threshold between input and hidden layers |
θk | = | threshold connecting hidden and output layers |
fh() | = | logistic sigmoid activation function from input layer to hidden layer |
fk() | = | logistic sigmoid activation function from hidden layer to output layer |
∝ | = | learning rate |
Subscripts | = | |
b | = | bulk |
c | = | convection |
i | = | inlet |
o | = | outlet |
p | = | plain |
s | = | surface |
t | = | twisted tape |
w | = | Water |