Development of Heat Exchangers for Use as Replacement Plates

In Distillation Columns: Proof of Concept and Collection of Design

Information

2
Graham Thomas Polley,
1
Edgar Enrique Vzquez Ramrez
1
Antonio Alberto Aguilar Moreno,
1
Jose Manuel
Riesco vila,
1
Carlos Omar Ros Orozco.

1
Departamento de Ingeniera Mecnica, Universidad de Guanajuato,
Carretera SalamancaValle de Santiago km. 3.5 + 1.8 km, Comunidad de Palo Blanco, Salamanca, Gto., Mxico
Telfono: 01 464 64 79940 ext. 2306

2
Departamento de Ingeniera Qumica, Universidad de Guanajuato

Introduction

Separation of chemical species by distillation is an energy intensive process. Whilst it has
been recognized for many years that the extraction of heat from intermediate stages
within a distillation column can result in significant energy savings the equipment
required to do this is generally expensive. One option is the use of side strippers [1].
These are often employed in columns used for the distillation of crude oil.

In more recent years the option of actually undertaking the distillation inside heat
exchangers has been examined [2]. However, this usually results in significant reduction
in the concentration driving forces used in the separation with significant effect upon
column size and cost.

Work being undertaken at FIMEE is directed at the examination of an inexpensive form
of heat exchanger (fabricated from parts manufactured for use in the motor industry) that
can be installed as a replacement for a plate within an existing distillation column.

Using such an exchanger positioned above the feed plate it becomes possible to extract
heat from the distillation column at a temperature that is higher than that present in the
column overheads. This heat can then be used to satisfy a suitable heat demand elsewhere
within the chemical process.

Using an exchanger positioned below the feed plate it becomes possible to provide part of
the heat needed to drive the distillation column at a lower temperature than that present in
the re-boiler at the base of the column. It may prove possible to provide such heat at
through a heat source available elsewhere in the chemical process.

By placing heat exchanger either side of the feed plate it becomes possible to extract heat
from above the feed plate and use a heat pump to provide boil up below the feed plate.





Exchanger Design

One possible design for a heat exchanger installed within a column was proposed by in
1993 [2]. A distillation column consisting of a number of sieve plates is illustrated in
Figure 1. A sieve plate is simply a flat plate through which a number of small diameter
holes have been drilled. The vapor traveling upward through the column has sufficient
velocity to prevent liquid flowing through these holes. As a consequence, a two phase
mixture is produced on top of the plate. It is maintained on the plate by a weir. Whilst the
vapor velocity is sufficient to prevent liquid flowing through the holes it is not high
enough to carry liquid to upward through the column. Consequently, liquid and vapor
separate above the reservoir formed by the weir with the liquid flowing down the plate
down-comer that feeds the next plate down the column.

The new exchanger fits in the space of two existing plates. The upper outlet then acts as a
sieve plate. The liquid over flows the weir and feeds the next plate down the column.

Sieve Plate
Down-comer
Weir
Standard Plate Spacings
0.15, 0.25, 0.3, 0.45, 0.6 & 0.9
Plate
Spacing
Replacement Plate
Down-comer from plate above
Down-comer to
plate below
Liquid Feeder
Heat Exchanger
vapour



Figure 1. Distillation Column Arrangement and New Exchanger










The proposed design has two significant features. The first is a liquid feeder plate that
distributes the liquid from the feed down-comer across the heat exchanger. This is
illustrated in Figure 2.

Liquid Feeder
Exchanger

Figure 2. Liquid Feeder

This base of this plate is sealed. The upper plate is perforated.

The second feature is the flow inducer and is illustrated in Figure 3.

9 mm
Flow Inducer
Vapour
Liquid
Exchanger Tube
Liquid Feeder

Figure 3, Flow Inducer

Vapor flowing from the sieve plate below the exchanger passes through tubes installed in
the base of the liquid feed plate. These tubes protrude into the individual heat exchanger
tubes. Liquid from the feed down-comer is then entrained by this vapor flow and drawn
through the heat exchanger.

Experimental Apparatus

The apparatus used in this study only considers hydraulic performance of the heat
exchanger. It consist consists of a single transparent tube (in order that flow patterns can
be observed using high speed photography) that represents a tube within the heat
exchanger that is fitted with flow inducer (as described above - Figure 3).

The rig (both schematic and photograph) is shown in Figure 4.

This tube sits inside a large diameter tube that provides a reservoir of liquid. The
reservoir has gauge marks that are used to measure the height of liquid above the flow
inducer.

Liquid is fed to the feed reservoir from tank A. The flow being controlled by a valve
positioned in the feed leg.

Air is injected into the tube at a controlled and measured rate (measured using
rotameters).

The two-phase mixture leaving the exchanger tube is separated in a tank (tank C), with
liquid flowing to a collection vessel (tank D). This vessel is used to measure the rate of
flow of liquid through the exchanger tube (through measuring the time to collect a given
quantity of liquid). On completion of the measurement this liquid is emptied into tank B.
Tanks A and B operate on a constant recycle in order to minimize pressure variation in
the feed leg and provide good flow control (confirmed by differential height calculations
during collection in tank D).






Figure 4. Experimental Rig

Analysis of Experimental Data

Data obtained in the experiments are given in Table 1.

The head required to provide the recirculation is assumed to consist of three terms:
gravitational head (a function of the voidage within the tube), frictional losses and a
kinetic energy term. This kinetic energy ter mis the difference between that of the two
phase flow leaving the tube and that of the liquid and vapor flow entering the tube. The
kinetic energy of the vapor entering the tube is calculated from the measured air flow.
The kinetic energy of the liquid entering the tube is assumed to be zero.

A homogeneous model is found to provide good correlation of the measurements for
conditions in which frictional lossses dominate. Accuracy of prediction falls as the % of
overall pressure drop due to kinetic effects increases. This suggest that assumption that
the incoming liquid has zero kinetic energy is incorrect.

Further experimentation is necessary to determine the significance of kinetic energy
effects within the wter injection system.



Flux msico
aire
kg/s m
2

Flux msico
agua
kg/s m
2

Variacin
P experimental
vs calculada %
% de P
Total
por
Friccin
% de P
Total
Energa
Cintica
% de P
Total
Energa
Potencial
3.22 701.9 8 48.5 29.9 21.6
3.22 609.7 11 48.6 29.1 22.3
3.22 516.2 13 48.8 28.3 22.9
3.22 429.1 16 49.0 27.5 23.5
3.22 333.5 21 49.4 26.5 24.1
1.93 728.0 18 39.4 24.0 36.6
1.93 631.0 20 38.9 23.0 38.1
1.93 520.8 25 38.4 21.8 39.8
1.93 430.1 27 38.0 20.7 41.3
1.93 316.0 34 37.6 19.3 43.1
0.64 651.5 23 19.9 11.6 68.5
0.64 497.5 29 17.2 9.4 73.3
0.64 366.0 36 14.9 7.6 77.5
0.64 251.8 44 13.0 6.1 81.0
0.64 158.9 54 8.9 5.0 86.1

Table 1. Experimental Data and Predictions of Homogeneous Flow Model


Evaluation of Thermal Performance

Heat transfer within the tubes of the exchanger will be controlled by the balance of two
mechanisms: two-phase convective heat transfer and nucleate boiling heat transfer.

The convective heat transfer coefficient is given by [4,5]:

45 . 0
|
|
.
|

\
|

=
L
TP
L fc
P
P


The predictions of this equation have been compared with a large data base of measured
heat transfer coefficients during upward flow boiling within circular tubes. The data are
well correlated.

This equation has been used to determine how much heat can be added through the
internal exchanger.

The principle problem that has been identified is that under many circumstances the
liquid Reynolds Number falls in the laminar flow region. This places uncertainty on the
validity of the calculation.

However, it has also led to the identification of two potential improvements in the
exchanger design. These are the provision of a liquid recycle and the use of tube inserts
to promote heat transfer within the exchanger.

Linking Column Design, Exchanger Design and Integration

The vapor flow thru the column is set by the column duty.

The maximum cross section of the heat exchanger is dependent on the area taken up by
the down-comers. This cross section can be related to the number of tubes installed in the
heat exchanger allowing calculation of the minimum mass flux of vapor flowing through
an exchanger handling the specified column duty. Higher mass fluxes are obtained by
reducing the number of tubes in the exchanger.

The mass flux of liquid flowing through the exchanger can be set independently by
adjusting the positioning of recycle vanes installed in the liquid down-comer.

So, it is possible (using a quite simple spreadsheet) to determine the heat transfer
coefficient achieved in the heat exchanger for different combinations of vapor and liquid
mass fluxes. The homogeneous pressure drop model will then allow calculation of
minimum tube length for the heat exchanger.

Knowing the overall heat transfer coefficient for the heat exchanger the heat load
obtained for a given temperature difference between the column and the fluid with which
it is integrated.

Case Study: Methanol-Water Separation

An overview of column integration is provided in Appendix 1.

The first step in the analysis is an evaluation of exchanger performance as a function of
vapor and liquid mass flex through the unit. This is presented in Table 2.

It is found that the vapor mass flux should be at least 50 kg/m
2
she higher the liquid mass
flux the higher is the quantity of heat that can be transferred through the tubes.

The exchanger has a square face and the tube length equates with the plate spacing. The
use of a square face provides space for the provision of liquid recycle lanes.

The column diameter is assumed to be 1 metre. The vapor mass flux within the column is
assumed to be 1.8 kg/m
2
. Therefore, to obtain the vapor mass flux necessary for efficient
heat transfer in the exchanger the exchanger dimensions need to be 0.25 m x 0.25 m x 0.9
m (yielding a vapor mass flux of 57.3 kg/m
2
s).

The next step is examination of Table 2 to determine the heat duty that can be obtained as
a function of liquid mass flux.


mass flux mass flow mass flux x rho-tp vis-tp Re-L f-L Re-tp f-tp dP-tp dP-L multiplier Nu h-l htc Q/dT
vapour liquid per tube
23 0.001463 50 0.315068 3.858492905 3.25541E-05 933.9975 0.017131 20181.82 0.006628 1830.8174 11.45098 9.81095 3.7 77.7 762.3108 19.40096
50 0.003181 100 0.333333 3.648099781 3.09375E-05 1818.182 0.0088 43636.36 0.005466 6742.3417 23.52941 12.75674 3.7 77.7 991.1991 25.22621
50 0.003181 200 0.2 6.060461162 4.85294E-05 4363.636 0.00972 46363.64 0.005384 11104.194 103.9569 8.182484 42.57413 894.0568 7315.605 186.1836
50 0.003181 400 0.111111 10.83857686 7.81579E-05 9696.97 0.007961 51818.18 0.005236 19565.426 340.5779 6.189745 80.64469 1693.539 10482.57 266.7836

100 0.006362 100 0.5 2.436026801 2.12903E-05 1363.636 0.011733 84545.45 0.004633 15214.667 31.37255 16.16481 3.7 77.7 1256.006 31.9656
100 0.006362 200 0.333333 3.648099781 3.09375E-05 3636.364 0.010173 87272.73 0.004596 22678.444 108.805 11.05437 36.79602 772.7164 8541.894 217.3929
100 0.006362 400 0.2 6.060461162 4.85294E-05 8727.273 0.008173 92727.27 0.004527 37349.909 349.668 8.182484 74.12587 1556.643 12737.21 324.1645
100 0.006362 800 0.111111 10.83857686 7.81579E-05 19393.94 0.006694 103636.4 0.004403 65809.987 1145.563 6.189745 140.4106 2948.622 18251.22 464.4972

50 0.003181 200 0.2 6.060461162 4.85294E-05 4363.636 0.00972 46363.64 0.005384 11104.194 103.9569 8.182484 42.57413 894.0568 7315.605 186.1836
50 0.003181 400 0.111111 10.83857686 7.81579E-05 9696.97 0.007961 51818.18 0.005236 19565.426 340.5779 6.189745 80.64469 1693.539 10482.57 266.7836
50 0.003181 800 0.058824 20.21252866 0.000121957 20534.76 0.006599 62727.27 0.004992 35687.012 1129.31 4.730028 146.9801 3086.583 14599.62 371.5633
50 0.003181 1200 0.04 29.3510704 0.000152778 31418.18 0.005934 73636.36 0.004796 51059.915 2284.666 4.047304 206.5436 4337.416 17554.84 446.7742


Table 2. Exchanger Performance

The results of these calculations are given in Table 3.

Liq. Mass Flux dQ/dT Exchanger dx/dT
Each tube Duty MW



It is concluded that given a heat source that is 10 degrees greater than local column
temperature 0.9 MW of heat can be exchanged. 32% of recycled liquid would be
vaporised.

The final stage of the design would be the identification and validation of the integration
options (see Appendix 1).

Conclusions

The concept of an integrated internal heat exchanger that takes position previously
occupied by distillation plates has been demonstrated using a simple apparatus.

It has been found that the hydraulic behavior of such an exchanger is well modeled by a
homogeneous flow model.

There is significant scope to improve the heat transfer performance of such exchangers
by installing tube inserts.

The flexibility of such exchangers can be improved by installing vanes in the liquid
down-comers in order to provide liquid recycle and increased liquid flow through the
exchanger.

A procedure that allows the engineer to link column duty, heat exchanger design and
process integration opportunity has been developed.

REFERENCES
[1] S. Bandyopadhyay, Thermal Integration of a Distillation Column Through Side-Exchangers, Chemical
Engineering Research and Design, Vol. 85, No.1, 2007, pp. 155-166.
[2] M.Nakaiwa, K.Huang, A.Endo,T.Ohmori,T.Takamatsu, Internally heat integrated columns: A
review, ChERD, 81, 162-177, 2003
[3] G.T.Polley, Heat Exchanger Design and Process Integration, The Chemical Engineer, 8 April 1993
[4] J. C. Chen Correlations for boiling heat transfer to saturated fluids in convective flow, Ind. Eng. Chem.
Process Design Development, Vol. 5, no. 3 (1966) pp. 322-329
[5] E.E. Vzquez Ramrez, G.T Polley , J.M. Riesco Avila Engineering equations for the estimation of
heat transfer in pool boiling and flow boiling during upward flow in tubes, AMIDIQ XXX, (2009),
Mxico.




Appendix 1

Integration of Simple Columns

The scope for integrating a single simple distillation column into a process is easily
identified by superimposing the temperature/heat load lines representing the column
reboiler and condenser on the Grand Composite Curve for the background process
(example given in Figure 1).


















Figure 1. Integration of Simple Column

This lines can be considered to make up a block. If the block fits well, integration is
straight forward. A column positioned above the Pinch would usually be driven by hot
utility and the heat released in the condenser used for other process heating duties. The
column could be controlled using the boil-up rate.

A column positioned below the Pinch would usually be driven by heat from the process
with the heat given up in the condenser being absorbed by cold utility. The column could
be controlled using the reflux.

On occasions, the block may fit partially (on rare occasions totally) into a pocket
(Figure 2) in the Grand Composite Curve. In the example shown, the column would be
partially driven by heat recovery and partially driven by hot utility. The heat released
through the condenser would be recovered within the process. Such an arrangement could
Ti
H
provide control and operability challenges. These would have to be examined and an
acceptable control strategy identified before the integration scheme could be accepted.



















Figure 2. Integration into a Pocket

Improving Integration

It is sometimes found that a simple column does not easily fit into the Grand Composite
Curve. For instance, consider the situation illustrated in Figure 3.
Here a heat process demand positioned at a temperature level intermediate between the
column re-boiler and the condenser is restricting the ability of the process to absorb the
full condenser heat load.

In this situation the designer has a number of options. The choice may be to only re-
absorb part of the column heat duty (i.e. accept the energy penalty). Alternatively, the
designer may seek to develop a process modification (e.g. raise the column pressure and
hence the temperature at which the condenser operates, or modify the background
process in order to remove the restriction placed upon heat recovery from the column).
Another option would be to introduce an intermediate condenser such that some of the
column heat is rejected at a higher temperature level than the main overhead condenser
(Figure 4).

The scope for identifying the possibility of using an intermediate condenser (or, re-boiler)
cannot be determined through arbitrarily drawing lines to form new blocks representing
the distillation column. An examination of the thermodynamics of the distillation process
is required.























Figure 3. Process Restricts Column Integration















Ti





Figure 4. Use of Intermediate Condenser?




Distillation Column Thermodynamics

The thermodynamics of the distillation process have been appreciated for many years
(see, for example, Benedict [1947]). A good description has been given by Flower &
Jackson [1964].

An appreciation can be gained by first considering the concept of minimum reflux ratio.
Using a McCabe-Thiele plot, this can be determined from the slope of line connecting the
required overhead composition to a point at which the feed q-line cuts the equilibrium
line (or, if this point cannot be reached without cutting the equilibrium line, a line from
this point tangential to the equilibrium line). This is the limiting operating line for the
rectification section.

















Figure 5. Minimum Reflux for Separation

The operating line of minimum reflux is illustrated in Figure 5. As the rectification takes
place the concentration of more volatile component in both liquid and vapour phases
H
Rectification
x
y
A
B
increase. The distillation process moves up the equilibrium line. The amount of reflux
required to complete the separation therefore changes. For the equilibrium line illustrated,
it is observed that the minimum reflux required to complete the separation falls as the
rectification process continues (moving up the curve from line A to line B. to line C etc.).
Consequently, not all of the heat that needs to be removed from the column has to
extracted through the top condenser (which otherwise provides the full reflux down the
column). Some heat can be extracted via an intermediate condenser. The quantity of heat
that can be extracted in this way can be determined (as a function of temperature) from
the differences in minimum reflux ratio.

A similar analysis can be conducted for the stripping section of the column. The
minimum boil-up required for a separation can be determined from the slope of a line
joining the required bottom concentration with the point at which the feed q-line cuts the
equilibrium line (Figure 6). It is found that some heat can be provided through the re-
boiler positioned part way up the column (using an intermediate re-boiler). Possible
intermediate load against temperature level can be determined in a manner similar to that
use for intermediate condensers.





Lecture 1 Slide 16
x
y
Boil-up required to
complete the
separation easily
determined
Heat can be added at
Intermediate temperatures
McCabe-Thiele Diagram
Required Base Condition



Figure 6. Column Boil-up Requirements


The Q-Curve

Flowers & Jackson proposed that the heat rejection profile for the rectification part of the
column and the heat absorption profile for the stripping section be brought together to
form a single curve, that they named the Q curve.

They showed that for the rectification the quantity by which the heat rejection from the
column between the feed plate and the top condenser exceeds that needed to condense
the top product is given by:


where:

For the stripping section the quantity of heat that must be added to the column between
the feed plate and a local point is given by:




The Q Curve is a plot of excess against local temperature. It can be used to identify how
heat can be added and rejected at different temperature levels. A typical example is
shown in Figure 7.










Q D
x x
x x
h
D
v
=

|
\

|
.
|
*
*
D molal fowrate of top product
x concentration of top product
x local liquid concentration
x local equilibrium concentration
h molal heat of vaporisation
D
v
=
=
=
=
=
. . . .
. . .
. .
. .
. . .
*

Q B
x x
x x
h
B
v
=

|
\

|
.
|
| *






















Figure 7. Q Curve



Development of T-Q Plot from Column Simulation Results

The McCabe-Thiele construction is based on the assumption that the molal latent heat
does not vary with composition. For many systems that is not a good approximation.
For these systems the Q-curve is better developed using the Enthalpy-Concentration
Diagram for the separation. The construction (developed by Flower & Jackson) for the
rectification section is shown in Figure 8. That for the stripping section is shown in
Figure 9.







T
Stripping
Section
Rectification
Feed
Q











Figure 8. Rectification: Excess Heat Profile


For the rectification section, the following equation gives the excess heat profile:


For the stripping section the excess heat can be calculated from:



Prior to the development of distillation column simulators and physical property
packages, this more accurate approach was limited by lack of Enthalpy-Concentration
Diagrams. Now, as discussed by Dhole [1991], the information required for the solution
of these equations is to be found in the standard output from commercial process
simulators.






( )
Q D H H H
x x
x x
H h
L v L
D
c vD
= +

|
\

|
.
|

|
\

|
.
|
*
*
[ ] [ ] [ ]
Q B H H H H
x x
x x
H H
x x
x x
v vF v L vF Lf
f B
f f
B
=

(
(
+

(
(
{ }
* * *
*
*
*
*
*
Tie
Line
x
*
x x
D
H
L
H
v
*
Q












Figure 9. Heat Input Through Stripping Section


Most significantly, Dhole has further suggested that, as a first approximation, the analysis
can be extended to cover multi-component as well as binary separations. All that is
needed is a converged column simulation.

Dhole also pointed out that in addition to the Q curve, a curve relating heat requirements
to plate position can be directly derived for the converged simulation.

Interpretation of Q Curve

The Q curve (also known as the Column Grand Composite Curve, CGCC) indicates how
heat can be added and extracted from the column at intermediate points. Columns have to
operate at reflux ratios above the minimum value in order to complete the separation with
a finite number of trays. Typically economic reflux ratios fall in the range 1.05 to 1.25
times the minimum value. Consequently a heat content block for a distillation column
would consist of the heat load associated with the excess reflux plus the Q

curve profile.

In the case shown in Figure 10 a single re-boiler at the base of the column (Figure 10a)
could be used. Alternatively, re-boiler at the base and a number of intermediate re-boilers
(Figure 10b) could be used. Since, the satisfaction of overall heat requirements using the
intermediate addition of heat effects the internal reflux ratio, the use of intermediate re-
Feed Tie
x
*
x
H
L
H
v f
*
Local
Tie Line
H
L f
H
v
*
Q
x
f
x
vf
*
boilers increases the number of plates required within the column. The effect upon
column size and cost needs to be determined before a decision to use intermediate units is
taken.

























The Q Curve also provides us with guidance on how the column feed can be conditioned.
If the curve shows a sharp increase in enthalpy immediately below the feed plate
(decreasing temperature), cooling of the feed could be beneficial (Figure 11). So, in an
integrated plant context, the feed to the column could be a valuable heat source.

If the curve shows a marked increase in enthalpy immediately above the feed plate
(Figure 12), partial or complete vaporization of a saturated feed would result in reduced
re-boiler load. In an integrated context the feed stream would be treated as a heat sink in
the overall process.

Vaporization of the feed changes the temperature at which it enters the column. A
different thermodynamic profile exists and a new Q Curve should be developed (Figure
13).


T
H
Reboiler
Figure 10a. The Use of a Single Reboiler























T
H
Figure 10b. The use of multiple reboilers




















T
H
Feed
Location
Scope for
feed cooling
Figure 11. Profile Indicating feed cooling

The Q Curve is particularly useful in identifying where heat pumping across the feed
plate is an option (see Figure 14).


























T
H
Figure 13. CGCC with feed vaporisat ion.
Saturated Feed
Vaporised Feed





















T
H
Figure 12. Profile indicating that feed pre-
heating would be advantageous






Benedict M. Multi-stage Separation Processes, Trans.A.I.Chem.E. 1947, 43(2), 41-60

Flower J.R. & Jackson R. Energy requirements in the separation of mixtures by
distillation, Trans.I.Chem.E. 1964, 42,249-258

Dhole V.R. Ph.D.Thesis, UMIST, 1991

Dhole V.R. & Linnhoff B. Distillation Column Targets, Paper presented at ESCAPE-1
Conf., Denmark, May 1992

Rajah W. & Polley G.T. Synthesis of Practical Distillation Schemes, ChERD, 73A,953-
966, 1995
































T
H
Intermediate
Condenser
Intermediate
Reboiler
Heat Pump
Figure 14. Application of Heat Pump