Chemtcd Brgineering Science. Vol. 40. No. 1 I. pp. 2073-2078. 1985. 0009-2X19/85 53.00+0.

00
Printed in Great Britain. CQ 1985. Pe~mon Press Ltd.

MODELLING AND OPTIMIZATION OF THE STEADY

STATE OPERATION OF A VAPOUR RECOMPRESSION
DISTILLATION COLUMN

E. BROUSSE, B. CLAUDEL+ and C. JALLUT

Laboratoire de CinCtique et Genie Chimiques, 404-INSA, 20, Avenue A. Einstein, F-69621 Villeurbanne,
France

(Received 7 November 1984)

Abstract-A data processing system adapted to a vapour recompression distillation unit is described. It is
shown that it allows the identification of mass transfer parameters of the distillation column, of heat exchange
of the boiler-condenser, of manometric and flow coefficients of compressor and of its isentropic yield. This
last parameter and the overall output of the unit pass through a maximum for the same value of the
compressor rotation speed.

INTRODUCTION
A distillation column is a mounting which works with
two heat sources at different temperatures: the boiler at
the higher one and the condenser at the lower one.
Therefore, its coupling with a heat pump allows one to
raise the thermal level of the heat available at the
condenser and to reinject it at the boiler. In this way, a
substantial energy saving can be achieved. There are
different coupling possibilities:
-use of an auxiliary transfer fluid, the circuit of
which is separated from that of the distilled mixture by
two heat exchangers. A mounting using Forane R 114
has been set up and operated in our laboratory
(Brousse, 1977, 1978);
*xpansion and compression of the boiler fluid.
The bottom heat exchanger is thus suppressed;
--compression and expansion of condenser vapour.
The top heat exchanger is thus removed.
The last rig has actually been set up in our labora-
tory, giving, in the first experiments, improved coef-
ficients of performance (COP), of the order of 6
(Brousse et al., 1982,1983). It seemed to us that the use
of a microcomputer for the command of such a system
could lead to a still better performance, and this is why Fig. 1. General scheme of the assembly: A, compressor; B,
boiler-condenser; C. distillation column; D, H, valves; E,
we present here a computational tool for analysis and
auxiliary boiler; F. G, heat exchangers; R,, mass flow rate
identification of the system, based upon a model of it measurement; N, rotation speed measurement, Pi, pressure
with a view to its optimization in a stationary regime measure_mentat point i; T,, temperature measurement at
(Jallut,, 1984). point j; V,. volumic flow rate measurement at point 4, VAN,
electro-pneumatic transmitter: p4, densimeter at point 4.

DESCRIPTION OF THE UNIT AND OF ITS

INSTRUMENTATION
multiplier of ratio 13.45. The compressor’s nominal
Characteristics of the installation (Fig. 1)
speed is ‘36,000 r.p.m. The rotor diameter is 132 mm.
Basically, the system distils a cyclohexane-n-
The compressor is fitted with a by-pass which plays a
heptane mixture at total reflux. Column C is 5 m high,
two-fold role:
its internal diameter is 0.444 m. It is packed over a
-avoiding compressor pumping (i.e. instability
height of 3.2 m with stainless steel Pall rings of 1 in,
occurring at low flow rates), the flow in derivation
intended to provide eight theoretical stages at total
being adjusted by valve H;
reflux. Compressor A is of centrifugal type. It is driven
setting the fluid overheating at compressor inlet in
by a d.c. motor of variable speed through a speed
order to avoid the incoming or inner formation of
droplets. This setting is made with the help of the water
‘Author to whom correspondence should be addressed. heat exchanger F.

2073
2074 E. BROWSE ef al.

The boiler-condenser B is of compact type. It

Graphic plotter
consists of parallel rectangular brazed aluminum
plates. Between these plates, waves serve both as braces
and as fins. Fluids flow countercurrently in the spaces
System diskette
between the 18 plates. The heat exchange area is, on the Lint printer
condensation side, 118 m’. PDP 11/03
In order to start the installation, an auxiliary boiler
E is needed. It is made of a vessel heated by three
18 kW resistances, one of which is regulated as a
function of temperature at the column top.
The water cooled heat exchanger G allows one
to undercool the liquid issued from condenser.
Isenthalpic expansion of liquid is performed by
electro-pneumatic valve D.
d’=*j Thtrmocouole nraamoIifier (8 channels
I Hulti~lcxcr I

Unit instrumentation
We use our microcomputer in order to realize in real
time the acquisition of 19 sets of data. We therefore Fig. 2. The computing system.
chose every measurement line in order for the signal
delivered to be compatible with its interface. The data
monitored are temperatures, pressures, and flow rates Steps of the computerized process control. They iue
at locations marked on Fig. 1, and top fluid com- shown in Fig. 3. The results of every step can retroact
position, compressor rotation speed and expansion on one or many of the preceding ones. Our immediate
valve setting. Temperatures are measured at 11 lo- objective is the static optimization of the system. The
cations TX to T,, with screened copper-constantan input variables are the compressor speed and the
thermocouples. Pressures are measured at three points expansion valve setting. The output variables are the
Pi, P2 and Ps with sensors fitted with a membrane, the compositions at column top and bottom. In the case of
deformation of which is monitored by constraint an industrial installation, other output variables could
gauges. Two flow rates are measured: a volumic flow be set such as distillate and bottoms flow rates. A third
rate v., for the liquid reflux, with a turbine flowmeter, input variable is then available, namely the reflux ratio.
and a mass flow rate mz for the gas flowing through As we aim at static optimization (with energy
the by-pass, with a diaphragm flowmeter. The measure- consumption as a criterion), we need a model d&b-
ment of the liquid reflux composition is based upon ing the steady states: we chose a “white box” model
the relation between this composition and the density with parameter experimental identification.
P,,, of a binary mixture. If the latter is supposed to be
ideal, this relation is Identification computerization. Figure 4 depicts all
the tasks insured by the software, which is written in
x&f,, + “&fb
Pm= &j (1) FORTRAN IV. Its thermodynamic section is based
M.
x&+ + Xb--
PII Pb

where x, + xb = 1. Measurement of p,,, is carried out

with a vibrating tube densimeter, which rests upon the
I AND
CHOlCE

OUTPUT
OF lNPUt

VARIABLES
I

following principle: every tube containing a liquid and

vibrating transversally resonates at a frequency which
depends on the liquid density. The compressor ro- CHOlCE OF A YODEL

tation speed is deduced from the measurement of

electric motor rotation speed, carried out with a
IDENTIFICATION
tachymetric dynamo fitted with two collectors: the first
OF THEPARAMETERS
serves for regulating the motor speed and the second
independently measures this speed. The setting of the
expansion valve is known by the feed current of the
electro-pneumatic transmitter which rules this setting: OPTlYlLATlON CRITERION

it linearly varies between zero (closed valve) and 20 mA

(completely open valve).

The computing system

Hardware configuration and use. The MINC system
of Digital Equipment, shown in Fig. 2, has as a central
unit, a PDP 1 l/O3 microprocessor, under RT 11
exploitation system. Fig. 3. Steps of the computerized system control.
 Steady state operation of a distillationcolumn 2075
TASK
SEQUENCE

Acquisition

Preliminary
Treatments

Identification

Plotting

Compressor Column Data from process Heat flux Heat tmnsfer coeff

Fig. 4. Expanded software.

upon an equation of state for the mixture, namely the assuming a constant volatility a, the relation between
virial equation: the packed height L, the cyclohexane mass fractions
y,(O) at the bottom and y,,(L) at the top, the molal
Pu = RT+ BP (2) vapour flow rate p and the product kvpa of the mass
in connection with Pitzer’s correlation for the compu- transfer coefficient and interfacial area is the following
tation of B (Reid et al., 1977). Its process section is (Pigford and Colburn, 1950):
based upon models in which some of the parameters
are identified. On the axis “task sequence” of this L=
graphical representation, we find the succession in
which the different programs are carried out: We identified the coefficient k,,a, and Fig. 5 gives its
--acquisition of data during a test: it is a sequential variation as a function of i? This variation can be
acquisition, adapted to stationary regimes of the expressed by the following relation, obtained by linear
aforementioned 19 sets of data; regression:
-preliminary calculations of average values of data
issued from the same working point and of thermo- k,,a = 1.1644tO.886 mole s-’ m-i (4)
dynamic functions; in the range: 1.5 < 7 C 4.6 mole s-i.
-identification, i.e. computation of the model par-
This result can be compared with those of bib-
ameters from the experimental data;
liographic correlations (Bolles and Fair, 1982) and
-visual display: it is possible to record on a graphic
plotter the evolution of all the calculated parameters as those given by the packing supplier. The following
functions of any variable. table gives, in m, the averaged HTU values:
On the axis “themes”, we find a cutting which Our result 0.61
corresponds to the different components, each one According to supplier 0.45
with its own model: According to correlations 0.16
-distillation column;
-heat pump (compressor, boiler-ondenser). The supplier’s estimate differs from our result by about
There are coupling equations between these dif- 26*/& which can be explained by the use of a different
ferent sub-models, particularly between the thermal mixture or by a different packing of the rings. On the
energy receiver (the column) and the thermal generator contrary, the correlations give unacceptable values,
and should be applied with circumspection.
(the heat pump). These equations express the energy
and mass conservations. From the user’s viewpoint, the
set of identification and display programs appear as Heat exchange at the boiler-condenser
menus which give access to computations and piot- Examination of the different thermal transfers oc-
tings of their choice. curring in this exchanger: condensation, evaporation,
heating or cooling of a vapour or a liquid allows one to
RESULTS AND DISCUSS1ON propose a simple model in which evaporation of
Distillation column bottoms results from condensation and undercooling
The packed column is amenable to a kinetic (mass of the top mixture. We therefore identify two
transfer) model. Within the Lewis hypothesis, and parameters:
2076 E. BROUSSE er al.

k v& MOLE/M/S

I 1

140 2.20 31)O 3.80 4.60 v MOLE/S

Fig. 5. Variation of k,a vs 7.

B221 the thermal power exchanged between two -thermodynamic characteristics, -expressed by the
diphasic mixtures, per degree of temperature function:
difference;
&(S) = %
flS2, the thermal power exchanged between liquid
and diphasic mixture, per degree of temperature
difference. called the isentropic yield.
Figure 7 depicts the experimental points obtained
Figure 6 depicts their variations as functions of 7, for for r(s). Except for four of them, which lie above the
which the following expressions hold: theoretical ~(6) line (obtained from Euler’s theorem),
fi22 = 2.474 P- 2.475 kW/“C they are close to a straight line, the equation of which
is:
(5)
B 32 = 0.486 7 - 0.160 kW/“C.
p = - 12.806 + 0.92. (8)
Figure 8 sketches the experimental points obtained for
Compressor t,~ (6). They fit best on a cubic curve, the equation of
Compressor characterization is two-fold: which is:

-flow rate, pressure characteristics through non- ?jti = 60.96-861_65P- 113346’. (9)
dimensional coefficients: Although there is some scatter of the points, due to the
indire& way they were obtained, the maximum of Fig.
(manometric coefficient) (6) 8 appears to be real, and can be interpreted by a better
adaptation between flow rate and speed, which mini-
6=& (flow coefficient).
mixes the irreversibilities due to friction and knocks on
blades at inlet.

7.

I
05 l.5 2s 3S 45 t MOLE/S

Fig. 6. Exchanged thermal powers, per degree of temperature difference, vs E

 Steady state operation of a distillationcolumn 2077

.I .P .3 .4 .s 6x10

Fig. 7. Compressor manometric coefficient vs flow coefficient: p,* is the theoretical variation. derived from
Euler’s theorem.

.a -
.I-

.6 -

.s -

A -

.3-

I
.l .2 .3 .4 .5 61 IO

Fig. 8. Compressor isentropic yield vs flow coefficient.

TLhemtdynamic aspect: heat pump subsystem cycle control necessitates the elaboration of models. We
Using the graphical processing tool and thermo- chose a physical model for vapour recompression
dynamic computation, we have been able to represent distiLlation process and realized a data processing tool
the cycle described by the top mixture, in an able to perform the identification of the process
enthalpy-pressure chart. An example is given in Fig. 9. parameters. After completion with smoothing
The coetficient of performance corresponding to this methods, this software could be applied to computer
example, valuated from the cycle, is about 15.2. aided industrial installations.

Overall output of the unit and optimization Ackmwkd@rtnent--One of us (C. J.) is indebtedto the French
Figure 10 represents, for one run, the variation of Ministry of Industry and Reaearc h for the grant that he
COP vs the compressor rotation speed. Although receivedfor the accomplishmentof this research.
lower values of the latter are la&in& it appears that
this curve passes through a maxi&urn. If the value of NOTATION
parameter 6 is evaluated for this maximum, it is found a interfacial area, per unit of packed volume,
that it corresponds to the maximum of isentropic yield. m2 mm3
This finding, which requires further confirmation, B second virial coefficient, m” mole- ’
would mean that the optimization of the whole unit COP coeEcient of performance, dimensionless
largely depends on that of the compressor. HTU height of transfer unit, m
k mass transfer coefiicient, mole s - 1
CONCLUSION L packed height, m
Perfo rmance improvement of energy converters molal mass, g mole- ’
passes through computerized process control. Such a mass flow rate, kg s- ’
2078 E. BROUSSE et AI.

P em

30 40 50 60 70 II KJ/UOLE

Fig. 9. Heat pump subsystem cycle in the X-P chart, Points 1,2, 3.4 correspondto the locations of Fig. 2.

C0.P
I

14-

12-

11-

lo-

9-

L
26 28 30 32 34 36 N/ 1000
r.p.m

Fig. 10. Coefficient of performance vs compressor rotation speed.

0 volumic flow rate, m3 s-l is isentropic

ideal gas constant, J mole- ’ V relative to vapour
R
radius of compressor wheel 9 5-l
T temperature, K (or “C) REFERENCES

u peripheral speed of compressor blades, m s- I Bolles W. L. and Fair J. R., 1982, Improved mass transfer
vapour molal volume, m3 mole-’ model enhances packed whtmn design. Chem. Engng 12
b vapour molal flow rate, mole s-l 109-l 16.
Browse E., 1977, Etude d’un syst&nethermopompe-wlonne
W mechanical work, J & distillerde 72 kW. Doctoral Dissertation, Lyon.
x molal fraction in liquid phase Brousw E., 1978, Study of a system using a heat-pump and a
Y molal fraction in the gas phase 72 kW distillationwlumn. Rev. G&n.Z%ewn. u13 847-866.
Brousse E., Capuwi A., Claude1 B., Stamatopoulos A. and
Wang S. Y., 1982, Systt?me colonne & distiller-pompe H
Greek letters chaleur & recompression dirwte des vapeurs. Reu. G&n.
volatility, dimensionless 7?mm. 246 509-519.
; exchanged thermal Power, Per degree K of Browse E., Claude1 B. and Jallut C., 1983, Le disgramme
temperature, kW K- ’ enthalpiepression composition. Application en distil-
lation aux pompes B cbaleur. Entropic 112 29-34.
6 compressor flow coefficient, dimensionless
Jailut C., 1984, Mise au point d’un outil infonnatique
tl yield, dimensionless d’analyse d’un syst&ne wlonne B distiller-pompe &
cc compressor manometric coefficient, dimen- chaleur par recompressionm&auique des vapeurs en vue
sionless de son optimisationstatique.Doctoral Dissertation,Lyon.
Pigford R. L. and Colburn A. P., 1950, Packed Distillation
P density, kg m - ’
Colunms, Chemical Engineers’ Wondbook, Perry J. H. (Ed.),
3rd edn, pp. 618621. McGraw-Hill, New York.
Subscripts
Reid R. C., Prausnitz J. M. and Sherwood T. K., 1977, The
a relative to oyclohexane Properties of Gases and Liquids, 3rd edn, p. 83. McGraw-
b relative to n-heptane Hill, New York.