RESTRICTED – COMMERCIAL RS1067

INITIAL RESULTS FOR REFLUX CONDENSATION OF HYDROCARBON VAPOUR

MIXTURES

J M McNaught and A Bartleman1

HTFS/NEL
East Kilbride, Glasgow

ABSTRACT

This report follows on from RS1047 (1999). A further analysis is presented of the data for reflux condensation of
pure hydrocarbon vapours in the experimental facility at NEL. In addition, initial results for reflux condensation of
mixtures of pentane and iso-octane are presented. The single component data show the expected variation with the
condensate film Reynolds number and there is good agreement with the predicted condensate heat transfer
coefficient. An additional gas-phase resistance to heat transfer is apparent in the mixture data. The Colburn-Drew
model gave good agreement with the measured overall heat transfer coefficient, but underestimated the vapour
composition and temperature changes. Improved agreement with the data could be obtained by enhancing the
diffusive heat and mass transfer coefficients.

NOTATION E Boiler
f Condensate film
A Surface area
G Gas phase
Cp Specific heat
I Inlet
h Specific enthalpy i Inside
∆hV Latent heat L Liquid phase
M& Mass flow rate lm Logarithmic mean
Nu Nusselt number m Mean
Q& Heat transfer rate
O Outlet
Re Reynolds number
o Outside
T Temperature
P Pentane
t Thickness
U Overall heat transfer coefficient Sat Saturation
xg Vapour mass fraction T Top of tube
V Vapour
y Vapour phase mole fraction
W Tube wall
α Heat transfer coefficient
η Dynamic viscosity 1. INTRODUCTION
λ Thermal conductivity HTFS has initiated a research programme to acquire
ρ Density data to develop and validate thermal design methods
Subscripts for condensation in vertical tubes with reflux. A
programme of experimental work is being carried out
B Bottom of tube to examine the condensation of binary hydrocarbon
C Coolant mixtures in a single vertical reflux condenser tube.

1
Formerly of Department of Chemical and Process Engineering, Strathclyde University, Glasgow, UK.
(On attachment to NEL)

11
The previous report (RS1047, 1999) described an water from this bath is passed over the measuring cell
initial series of experiments to commission and to control the temperature. Thermostat controls on
validate the facility by condensing pure components. the bath allow adjustment of the cell temperature, and
In this report the pure component data are re-analysed it is measured by a thermocouple, connected to a
and new mixture data are presented. digital display, inserted in a port alongside the
measuring cell.
2. RESEARCH FACILITY
3. EXPERIMENTAL PROGRAMME
The reflux condensation research facility is fully
described by (Bartleman, 1999). A line diagram of Pure pentane tests were carried out operating with 3
the facility is shown in Figure 1. and 6 kW heat loads and pure iso-octane tests were
performed with a 2 kW heat load.
The main features of the facility are an electrically
heated boiler, the test condenser and a dump In the binary mixture tests the facility was filled with
condenser. The boiler contains six electrically a liquid pentane/iso-octane charge of known
powered cartridge heaters positioned well below the composition. Two sets of mixture data are reported
liquid level in the boiler. here, corresponding to initial charges of 25 and 35%
mole fraction of pentane. The mixture data were
The reflux condenser consists of three identical test obtained at nominal heat loads of 3 and 6 kW.
sections of copper tube of 45mm internal diameter
and 62 mm external diameter. Each test section is 4. DATA ANALYSIS – SINGLE
just over 0.5m in length, giving an overall condenser COMPONENT TESTS
length of approximately 1.5 m. The tube has a helical
groove 4 mm deep by 10 mm wide, with a 15 mm 4.1 Determination of Vapour and Condensate
pitch cut into its outer surface over a distance of 33.5 Flowrates
pitches. This creates a small rectangular flow The procedure described in RS1047 (1999) was
channel which provides a high coolant heat transfer modified to allow for the fact that the condensate at
coefficient whilst ensuring a measurable temperature the bottom of the tube is subcooled. The condensate
rise. outlet temperature TLO is obtained from

Any uncondensed vapour passes to the dump T LO = 0.62T Sat + 0.38TW (1)
condenser where it is completely condensed and
returned to the bottom of the tube bundle in the where TSat is the saturation temperature at the boiler
boiler. pressure and TW is the measured wall temperature in
the bottom test section.
The liquid can be sampled at two locations to
measure the composition of the condensate from the The mass flow of condensate produced in the reflux
test and dump condensers. The locations of these two condenser is found from
sampling points are (i) a channel in the boiler to
which condensate from the test condenser is directed, Q& C
M& L = (2)
and (ii) the return leg below the dump condenser. ∆hV + C pL ( T Sat − T LO )

The measured composition is derived by measuring

The mass flow of vapour entering the reflux
the density of the sample in a calibrated densitometer.
condenser is obtained from
The sample is pumped through the vibrating U-tube
densitometer by a variable speed gear pump.
Samples are cooled in a cooling water bath, and the

12 RS1067
 Q& • Obtain the condensate mass flowrates from each
M& V = E (3)
∆hV condenser from
The coolant heat loads are determined from
measurement of the coolant flow rate, and the inlet M& L = Q& C (hVO − h LO ) (7)
and outlet coolant temperatures across the condenser.
• Obtain the vapour inlet flowrate to the test
Q& C = M& C Cp C (TCO − TCI ) (4) condenser by summing the condensate flowrates
from the two condensers
4.2 Determination of the Mean Experimental
• Calculate the vapour inlet mass fraction yI from a
Condensate-film Heat Transfer Coefficient
mass balance involving the two measured
The mean overall heat transfer coefficient for the test compositions and the condensate and vapour
condenser is obtained from flowrates

Q& C 6. RESULTS
U = (5)
Ai ∆Tlm
6.1 Single Component

The mean condensate-film heat transfer coefficient is The data are presented in Figure 2 in the form of a
then calculated using the equation plot of a dimensionless mean heat transfer coefficient
(see HTFS Handbook Sheet CM8) plotted against the
1 1 A  1  Ai  xW condensate film Reynolds number at the bottom of
= − i  
 α − A
 (6)
α f U  Ao  C  W
λ
 W
the tube. The Nusselt analysis for gravity controlled
laminar filmwise condensation, and the method of
Sheet CM8, which incorporates a correction to the
where the coolant heat transfer coefficient αC is
Nusselt analysis for the effects of waves and
obtained as described in RS1047 (1999).
turbulence, are also shown.
5. DATA ANALYSIS – MIXTURE TESTS
As expected the mean condensate heat transfer
5.1 Determination of Vapour and Condensate coefficient is significantly greater than the Nusselt
Flowrates and Compositions value. The coefficient is on average about 12%
greater than the CM8 value. These results support
The main steps in the process to determine the vapour
the validity of the measurements, and they can also be
and condensate flowrates from the heat load and
utilised in the analysis of the mixture data. The
composition measurements are as follows:
measured mean condensate heat transfer coefficient
can be well represented by the CM8 method
• Estimate the condensate temperature at the
multiplied by the following correction factor F given
bottom of the tubes using equation (1)
by
• Use the measured condensate composition from
F = 1.01 ×10 -4 Re f + 1.03 (8)
the test and dump condensers to obtain the
condensate outlet properties
The fact that the CM8 method gives good overall
• Calculate the condensate specific enthalpy hLO at
accuracy in predicting the data indicates that vapour
the outlet of the test and dump condensers using
shear effects are small. This is in line with the
the measured condensate compositions
commonly held view that shear effects on the
• For each condenser, calculate the vapour specific
condensate heat transfer coefficient can be ignored
enthalpy hVO corresponding to the condensate
below the flooding point.
outlet composition and the vapour inlet
temperature

13 RS1067
6.2 Binary Mixtures – Heat Transfer equilibrium-based process. In the limit of very
high heat transfer rate, the temperature range in
A mean temperature difference ∆Tm was evaluated
the gas phase would tend to zero.
using the measured vapour and coolant inlet and
(c) it is probably conservative to use the dry-gas
outlet temperatures. A mean overall heat transfer
coefficient for heat transfer from the gas phase to
coefficient was then obtained from
the condensate surface, since waves at the liquid
QC surface may enhance the heat transfer rate
U = (9)
Ai ∆Tm
The next step is to investigate the application of the
Colburn-Drew method, which utilises film-theory to
A mean heat transfer coefficient for the condensing
determine heat and mass transfer rates explicitly.
stream was then obtained using the coolant and wall
heat transfer coefficients. The CM8 method together
A Colburn-Drew model was programmed as
with equation (8) was used to evaluate the condensate
described in HTFS Handbook Sheets CM33 and
heat transfer coefficient, which was then utilised to
CM35. The model was set up to calculate the tube
derive a mean heat transfer coefficient for the gas
length required for the measured condensation rate
phase from
from the data. The simulation used 20 increments
along the tube length, starting at the vapour inlet.
1 1 1  Ai  1  Ai  tW
= − −  
 α − A
 (10) The initial conditions at the bottom of the tube were
α G U α f  Ao  C  W
λ
 W the measured vapour temperature and condensate
outlet composition, together with the vapour
The mean heat transfer coefficient αG is plotted
composition that was obtained from the heat and
against the mean vapour Reynolds number in Figure mass balance calculations on the data. It is possible
3. This shows, as expected, that αG increases as the to perform the calculations with either the no-mixing
mean vapour Reynolds number increases. or perfect-mixing models for the condensate. With
the perfect-mixing model the simulation is iterative
Figure 3 also shows the overall condensing side heat because the local condensate composition depends on
transfer coefficient and the estimated condensate heat the condensation rates higher up the tube. The
transfer coefficient. iterations were initiated using an estimated value of
the condensate outlet composition, and the simulation
An approximate overall analysis of the data showed was repeated until the initial and calculated values of
that application of Silver’s method with the integral the outlet composition converged.
condensation curve generated a very conservative
prediction of αG . The performance of the Colburn-Drew method
against the data was quantified using the following
7. MASS TRANSFER MODELS parameters:
Section 6.2 indicates that the basic Silver method
gives over-conservative predictions of the data. This (1) Actual/required condenser length.
can be attributed to several reasons: (2) Measured and predicted temperature drop in
the gas phase
(a) the counter-current vapour-liquid flow, and the (3) Measured and predicted values of the
presence of gas-phase heat and mass transfer parameter E defined by
resistances, mean that the mixture does not
follow the equilibrium-based heat release curve. y P, B − y P,T
E= (11)
(b) finite mass transfer rates cause the temperature y P, B − y *P, B
range to be less than predicted from an

14 RS1067
where the subscripts T and B refer to the top and be improved by enhancing the diffusive heat and
bottom of the tube respectively and y* is the vapour mass transfer coefficients.
composition that is in equilibrium with the
condensate composition at the bottom of the tube. Additional mixture data over a wider range of
This parameter gives a measure of the separation compositions, reflux ratios and heat loads are
achieved in the condenser. required to further investigate the performance of the
model.
The results of the Colburn-Drew simulations are
compared with the data in Figures 4-6 for the no- ACKNOWLEDGEMENTS
mixing and perfect mixing models. The comparisons A. Bartleman was an Associate of the Postgraduate
show that the predicted condenser length is in good Training Partnership (PTP) between NEL and
agreement with the actual length. However the Strathclyde University. The PTP scheme is a joint
vapour temperature change and the predicted initiative of the Department of Trade and Industry
separation of the components are much smaller than (DTI) and the Engineering and Physical Sciences
was measured. This suggests that the Colburn-Drew Research Council (EPSRC) in the UK, and is
model gives a good measure of the additional financially assisted by the DTI. A. Bartleman
resistance to heat transfer in the gas phase, but the gratefully acknowledges grant support from both
relative contributions of diffusive and convective heat NEL and EPSRC.
and mass transfer are not in agreement with the data.
Professor C D Grant of the Department of Chemical
In order to test the effect of increasing the effective and Process Engineering at Strathclyde University
gas-phase heat and mass transfer coefficients, the was the Academic Supervisor of the project.
Colburn-Drew method was implemented with the
film-theory mass-transfer corrections omitted. This REFERENCES
has the effect of enhancing the diffusive contributions
BARTLEMAN, A. (1999), "Condensation of Pure
to the heat and mass fluxes from the gas-phase. The
Hydrocarbons in a Single Vertical-tube Reflux
results for the perfect-mixing model, also shown in
Condenser", HTFS RS1047.
Figures 4-6 and referred to as “Perfect Mixing, No
MTC”, show considerably improved predictions of LIST OF FIGURES
the composition change, though the vapour
temperature change is now over-predicted. 1. Diagram of the HTFS Reflux Condensation
Facility, NEL
8. CONCLUSIONS 2. Measured and predicted mean condensate-film
heat transfer coefficients from single component
Heat transfer data for reflux condensation of single
data
hydrocarbons in the new Reflux Condensation Test
3. Mean experimental heat transfer coefficients from
Facility demonstrate are well predicted by the HTFS
mixture data
method for the condensate heat transfer coefficient in
4. Actual/required condenser lengths predicted by
gravity-controlled conditions.
Colburn-Drew models
5. Comparison of vapour composition parameters
New reflux condensation data for mixtures of pentane
predicted by Colburn-Drew models with
and iso-octane show, as expected, an additional
measurements
resistance to heat transfer in the gas-phase. This
6. Comparison of vapour temperature change
overall magnitude of the resistance is well
predicted by Colburn-Drew models with
represented by the Colburn-Drew model, but the
measurements
vapour composition and temperature changes are
significantly under-predicted. The predictions could

15 RS1067
 To Atmosphere
Primary Loop
Head Tank

Reflux Condenser

Dump Condenser
Secondary Loop
Head Tank

Air-cooler
Boiler

Intercooler
Densitometer
To Drain To Drain

Cooling Bath

Figure 1: Diagram of the HTFS Reflux Condensation Facility, NEL

0.35

0.3

0.25

0.2
Nuf

0.15
Pentane (Exp)
0.1 Pentane(CM8)
Iso_oct (Exp)
0.05 iso_Oct (CM8)
Nusselt Theory

0
0 500 1000 1500 2000

Ref
Figure 2: Measured and predicted mean condensate-film heat transfer coefficients from
single component data

16 RS1067
 1200

Mean experimental heat transfer coefficients 1000

800
(W/m 2K)

600

Condensate
400
Gas-Phase
200
Overall
Condensing
0
0 10000 20000 30000 40000
Mean ReG

Figure 3: Mean experimental heat transfer coefficients from mixture data

1.4

1.3

1.2
Actual/Required Tube Length

1.1

0.9

0.8

0.7
No Mixing
0.6 Perfect Mixing
Perfect Mixing, No MTC
0.5

0.4
0 5 10 15 20
2
Heat Flux (kW/m )

Figure 4: Actual/Required condenser lengths predicted by Colburn-Drew models

17 RS1067
 1.50

Measurements
1.25

Vapour composition parameter

No Mixing
Perfect Mixing
Perfect Mixing, No MTC
1.00

0.75

0.50

0.25

0.00
0 5 10 15 20

Heat Flux (kW/m2)

Figure 5: Comparison of vapour composition parameters predicted by Colburn-Drew models with measurements

20
Measurements
No Mixing
Perfect Mixing
Vapour temperature change (K)

Perfect Mixing, No MTC

15

10

0
0 5 10 15 20

Heat Flux (kW/m 2)

Figure 6: Comparison of vapour temperature change predicted by Colburn-Drew models with measurements

18 RS1067