48 Proceedings of The South African Sugar Technologists' Association - June 1985

PROGRAM FOR· SIMULATING AND EVALUATING A

CONTINUOUS A-SUGAR PAN
By R. G. HOEKSTRA
Tongaat-Hulett Sugar Limited, La Lucia
Abstract (4) Each of the compartments is to be considered as a stirred
The Continuously Operating Raw Pan Simulation and Eval- tank reactor. For the continuous pans on which Rein et
uation (CORPSE) is a steady-state computerised simulation of al'? have performed lithium tracer tests, it was found that
such a pan, based on the solution by the Newton-Raphson one compartment could effectively represent anything from
method of simultaneous algebraic equations which describe the I to 2 equivalent stirred tanks. Such evidence was not
mass and heat transfer within and the material balances over sufficientlyconclusive to justify the complicated modelling
each compartment. Each compartment is considered as a stirred necessary for assuming anything other than one stirred
tank reactor, with unknown components being: Total solids, tank per compartment.
total sucrose, crystal sucrose and water. Other unknowns could (5) The number of crystals is assumed to remain constant
include compartment volumes and process values such as pan from entry to exit of the pan. When operating an A-pan,
vapour pressure, supersaturation and mass and heat transfer the crystals in the seed stream will have already grown to
coefficients. The user of the model has considerable freedom a fair size in a batch pan. If excessive under- and super-
of choice of input data. . saturation respectively are avoided, there should be
This model has been used for: neither disappearance of crystals through dissolution nor
creation of new crystals through spontaneous nucleation.
• Evaluating process parameters from observations of an op-
erating pan. If more than the minimum amount of data is (6) Within each compartment, the processes of heat transfer
submitted, the program performs a least-squares fit. from the calandria into the massecuite and of mass trans-
fer of dissolved sucrose from the mother liquor onto the
• Sizing of compartments and calandria heat transfer areas for growing crystals are simulated. Mass balances across that
design purposes. compartment are also taken.
• Showing how different values of controllable process vari- (7) The simulation could be based on one of the following
ables affect the operation of the pan. methods:
(a) Simulation of each compartment in turn:
The firstcompartment is simulated by using the values
1. Object
of the input streams, thus calculating the values ofthe
The computer program for the Continuously Operating Raw output streams, and from there on to the second com-
Pan Simulation and Evaluation (CORPSE) was developed with partment, etc. However, some of the necessary data
the following objectives in mind: for the simulation will usually be unknown, such as
(a) Evaluation Mode: parameters (when in Evaluation mode) or compart-
To determine the main process parameters of an existing ment volumes and areas (when in Design mode). This
pan from measurements taken during factory operation. would require using initial estimates of the unknowns
and repeating the calculating cycle over the successive
(b) Design Mode: compartments several times.
To size the volumes and heat transfer areas offuture pans,
using the above obtained parameters together with the (b) Simultaneous simulation of all compartments:
operating specifications. This is performed by setting up the appropriate equa-
tions for each ofthe compartments, andl solving them
(c) Operation Mode: all as a set of simultaneous algebraic equations. Be-
To show the quality and quantity of product which will cause not all the equations are linear, they have to be
be produced from the interaction of a given set of con- solved by an interative mathematical method, with
trollable operating variables. initial estimates of the unknowns.
It was decided to use the method (b) of simultaneous
equations, using the Newton-Raphson method of solution.
2. Approach
Although it requires more computer capacity, it is a
(1) CORPSE is to be a steady-state simulation. This require- more orderly method and allowsgreater flexibility of choice
ment is obvious from objectives (a) and (b) above, and of which shall be the given input values and which the
the object of (c) was also to show steady-state operation unknown values to be determined by CORPSE. For each
rather than to demonstrate dynamic response to changes given value in the input data, CORPSE sets up an appro-
in controllable process variables. priate equation. Furthermore, when in the Evaluation
(2) The number of compartments can be varied, to make mode, it is possible (and indeed desirable) to over-specify
CORPSE applicable to various installations. the system with whatever plant observations are available,
meaning that there will be more relationships (equations)
(3) Many simplifying assumptions have to be made: some- than unknowns.
times for lack of accurately known relationships; some-
times in order to keep the mathematics manageable. In such a case, the Newton-Raphson method can be
Because the same simplifying assumptions are used in all adapted to perform a least-squares fit to determine the
three modes, it is reasonable to expect that, by applying most likely values of unknowns from all the data.
the same values of parameters obtained in the Evaluation
mode to the Design and Operation modes, the results ob- 3. Schematic representation of a Continuous Pan
tained in the simulations will be sufficientlyin accord with This is shown in Figure 1 for a pan of N compartments,
practice. numbered 1 to N.
Proceedings of The South African Sugar Technologists' Association - June 1985 49
o

Pan Vapour

0) 0)
Seed'
)
_._--~---~
G
~ f;\
---[~r -~p-r\V-o-du-c-t~
Heating
~team

Condensate
Dilution
water

Syru
- - -- -----'
o Mainstream ref. no. ~ Compartment ref. no.

FIGURE 1 Diagrammatic representation of a continuous pan.

The compartments are connected in series, and share a com- TABLE 1

mon pan vapour space and a common calandria steam supply. Primary unknowns
The mainstream enters compartment I as the seed stream.
Symbolic
The contents of compartment I overflow via a weir into com- Representation
Description Dimensions Indexing
partment 2, and so on. The mainstream leaving compartment
N is the product stream. The following will always be unknowns:
Because ofthe assumption that each compartment, say No. SOLIDS(J) Mass rate of total solids in tons/h J = 0 to N
J, is a perfectly stirred tank, it follows that its contents will mainstream
SUC(J) Mass rate of total sucrose in tons/h J= 0 to N
have the same composition as the mainstream leaving it, which mainstream
should logically be called mainstream No. J. For convenience WATER(J) Mass rate of water in mainstream tons/h J = 0 to N
the seed stream is referred to as mainstream No. O. XTAL(J) Mass rate of crystal sucrose in tons/h J = 0 to N
mainstream
The composition and mass rate of the mainstream are com- N_XTALS No. of crystals in mainstream XI09/h None
pletely specified by the following components, with dimension The following can be additional unknowns for the Evaluation Mode:
tons/h:
K_GROWTH Crystal growth rate parameter mm/h None
• water K_UB Intercept} Para~eters o.flinear kW/(m"C) None
function of heat
• crystal sucrose K_UM transfer coeff.w.r.t. kW/(m"C) None
• total sucrose, which consists of dissolved as well as crystal Slope compartment no.
sucrose K_SUPERSATN Common value of supersaturation None
The following can be additional unknowns for the Design Mode:
• total solids, which consists of total sucrose plus non-sucrose.
The non-sucrose is all in dissolved form. CELL_YOL(J) Effective volume per compartment m' J = 1 to N
STEAM_TEMP Temp. of heating vapour to 'C None
Each of the N compartments has 2 auxilliary feeds to it, syrup calandria
and dilution water. These feeds share a common control valve
per compartment, and either the one or the other is fed to the
compartment. The syrup feed has the components water, total CORPSE sets up the equations in terms of these primary
sucrose and total solids, but no crystal sucrose. unknowns and solves for them. From this solution, all the other
unknown values can be calculated.
4. The Unknowns An illustration of the variables w.r.t. one compartment, say
compartment J, is shown in Fig. 2.
There are many variables in a continuous pan. Initially the
values of most are unknown, and have to be calculated by
running the mathematical model. All these unknowns can how- 5. Data which can be available as input to CORPSE
ever be reduced to a set of primary unknowns, which are tab- The user has considerable choice of what data to input to
ulated as follows: the model. This will depend on the purpose he wants to use
50 Proceedings of The South African Sugar Technologists' Association - June 1985

Pan construction specifications:

YAP TEMP @ Volume } per compartment, for compart-
Mainstream(J-1 ) Mainstream J: Rates/h: Heat transfer area ments I to N.
1i\1l\!f\ TOT_MASS(J)
I I I Process parameters:
SOLIDS(J)*
I I I Crystal growth rate.
SUC(J)*
Heat transfer coefficient per compartment, for compartments
Compartment J: WATER(J)* I to N.
KGROWTH@ XTAL(J)*
CELL_VOL(J)@ NXTALS* 6. Concept of Mean Crystal Sizes
RET_TIME(J) As the massecuite moves from compartment to compart-
SUPERSATN(J)@ Properties:
ment, the linear dimension (length) of each crystal will increase
MSC_BX(J) during its stay in that compartment, dependent mainly on the
AREA(J) MSC_POL(J) degree of supersaturation and the retention time in that com-
HTC(J)=K_UB@ + J.K_UM@ MSC_PUR(J) partment, and on the growth rate parameter for sucrose.
MOL_BX(J) One obviously cannot consider each crystal on its own, so
MOL_PUR(J) it is therefore necessary to find some general measure of crystal
I I II size which is appropriate to the technical context in which it
STEAM TEMP XTAL~SIZE(J) is used in the simulation. This requires a knowledge of the
\I - II XTAL_PC_SOLIDS(J) grain size distribution in each of the compartments.
For the purposes of this simulation, 3 different bases of fre-
quency distributions are used:
J, (a) Size-frequency-by-number
WASH(J) ....V ..
.........
SYRUP(J) ,[, This is the fraction dn/n w.r.t. the total number of crystals
~
CONDENSATE n with sizes (lengths) within the range of Land L + dL,
where n = total number of crystals in sample
TOT SYRUP t I, t__ . L = size (length) of crystal
dL = small increment in size L
* Will always be a primary unknown One would obtain this basis of frequency distribution
@ Can be a primary unknown when examining a representative sample of the massecuite
under a microscope with a grid and counting the number
FIGURE 2 Variables w.r.t. compartment J of a continuous pan. of crystals in each size category.
Its application is the following:
CORPSE for as well as what data are available. The amount Because each individual crystal is subjected to conditions
of data must ofcourse be sufficient so that the number of equa- of growth, it is logical when considering growth to define
tions which CORPSE will set up will be at least equal to the the size of the average crystal as the sum of the sizes of all
number of unknowns. That in itself is no absolute guarantee the crystals divided by the number of crystals, which will
that the equations can be solved, because the user must still be called the mean-size-by-number, and is calculated from
beware of the possibility that the system is overspecified in one the size-frequency-by-number distribution.
part and underspecified in another. The following list gives the
possible data which can be specified as input: (b) Volume-frequency-by-number
This is the fraction by dn/n w.r.t. the total number of
Performance Data: crystals n with volumes within the range of V and V +
Feed rate of seed stream. dV, where V = volume of a crystal.
Output rate of massecuite stream. dV = small increment in volume V.
Total feed rate of syrup.
Total production rate of calandria condensate. If the crystals are all of uniform shape, one can have the
relationship between volume and length of crystal, namely
Solids % Total mass V = cLJ,
Sucrose % Solids of massecuite, for mainstreams 0 where c = shape factor
Crystal % Solids to N = I for cubes,
Purity
= n/6 for spheres, etc.
Crystal sizes
For CORPSE, the crystals were taken to be cubic, ie c = 1.
Solids % Total Mass} of mother liquor (molasses), for One could equally well define the volume-frequency-by-
Sucrose % Solids mainstreams 0 to N number as the fraction by number dn/n w.r.t. the total
Purity number of crystals n with volumes within the range of CL3
and c(L + dL)3. Note that this is not the same as the range
Supersaturation of mother liqUO~} per compartment, for CL3 and cLJ + cdfl.'). Its application is when it is required
Feed rate of dilution water compartments I to N. to calculate the total mass of crystals in any compartment,
Solids % Total Mass} f given their total number.
Purity 0 syrup. It should be clear that:
Mass of crystals
Pressure or temperature of calandria vapour. = Mean-volume-by-number X Density of crystal X No
Pressure or temperature of pan vapour. of crystals, where
Proceedings of The South African Sugar Technologists' Association - June 1985 51

Mean-volume-by-number can be defined as the sum of the in terms of the same parameters. This last mentioned p.d.f
volumes of all the crystals divided by the no. of crystals, wasfitted to the sieving resultsby non-linearregression analysis
and can be calculated if the volume-frequency-by-number and the values of its parameters determined.
distribution is known. From theseparametersthe mean-size-by-number, mean-vol-
Re-arranging: ume-by-number and mean-size-by-mass can be calculated.
Mass of crystals Appendix 1 describes the GNE p.d.f., its transformationand
Mean-volume-by-number , the expressions of the above means sizes in terms of its
= (M . b . b )3 X (Mean size-by-number)' parameters.
ean-size- y-num er
X Density of crystals X No. of crystals.
Ed.f. per compartment
Mean-volume-by-number , ,
The factor (Mean-size- ize-b y-numb) w111 be abbreviated Theoretically, one could start offwith a p.d.f of the crystals
er 3 in the seedstream and, withconsideration of the residence time
by VL3 and can be calculated for each compartment. distributions in each of the theoretical stirred tank reactors,
Symbolically, the above expression for crystal mass model the size dispersion as well as the increase in mean size
becomes: of the crystals leaving that compartment. This would however
XTAL(J) = VL3 * XTAL_SIZE(J) ** 3 * XTAL_DEN- make the mathematics extremely complicated, plus the fact
SITY * N_XTALS that the changes in size dispersions obtained in practice might
in any case not agree with the theoretical.
(c) Size-frequency-by-mass
For those reasons, the following. procedure was decided on:
This is the fraction by mass dm/m w.r.t. the total mass
of crystals m with sizes within the range of Land L + dL, • Take a sample from each of several compartments of an op-
where m = mass of sample. erating pan (in practice, every 2nd compartment).
This basis of frequency distribution is obtained from • On each sample do a sieving analysis and use non-linear
sieving analyses. regression to calculate the shape parameter of the GNE p.d.f
(See Appendix 1).
Although it is not directly used in the process relation-
ships of CORPSE, its application is to calculate the mean- • Fit a linear regression relationship between the shape para-
size-by-mass, or more commonly known as the mean ap- meter and the compartment no., so that this parameter can
erture (MA). This measure is often used in the sugar in- be estimated for each compartment. The value of the shape
dustryand is defined as the sum of (Size X mass per crystal) parameterfor eachcompartmentgives a measure of the crys-
divided by the total mass of all crystals. tal size dispersion, and is used in whatever of the 3 modes
(Evaluation, Design, or Operation) it is intended to run the
program.
7. Probability Density Functions:
Types
8. Setting up equations on mainstream properties
Various probability density functions (p.d.f.'s) have been
considered to describe distribution based on size-frequency- These equations can all be transformed into a pure linear
by-number: form, and do not pose any problem.
• Gammap.d.f. The procedure is to let the streamreference index J gothrough
• Log-normal p.d.f a DO-loopfrom 0 to N, examinefor which properties are there
values given in the input and, if so, set. up the appropriate
• Rosin-Rammler p.d.f. equations. The mainstream properties considered are:
• Generalised Negative Exponential p.d.f • Crystal size (mm):
In deciding which of these p.d.f.'s to use, the following must XTAL(J) = 109 * N_XTALS * XTAL_DENSITY * VL3 *
be considered: «Given) SIZE(J» ** 3
(1) It must be a good fit to the results obtained in practice. This is a linear equation in the primary unknowns XTAL(J)
(2) It should be possible to transform mathematically the p.d.f. and N_XTALS.
on size-frequency-by-number to corresponding p.d.f''s • Solids % Total mass in massecuite:
which describe the volume-frequency-by-number and the SOLIDS(J)/(SOLIDS(J) + WATER(J»
size-frequency-by-mass distributionsper (b)and (c) above. = (Given) MSC_BX(J)/ 100
(3) In practice, the size-frequency-by-number distribution is This can be linearised in terms of the primary unknowns:
skewed and thereobviously cannot be negative sizevalues. SOLIDS(J) - WATER(J) * MSC_BX(J)/(IOO - MSC_BX(J»
This rules out the normal (Gaussian) p.d.f which is sym- =0
metrical and would show a significant proportion of neg-
ative sizes at the comparatively high coefficients of Other properties can similarly be reduced to linear forms:
variation which crystal size distributions exhibit. • Sucrose % Total mass in massecuite
From work which has thus far been done, the generalised • Purity of massecuite
negative exponential (GNE) p.d.f answers fairly well to the • Solids % Total mass in molasses
above requirements.
• Sucrose % Total mass in molasses
Fitting the p.dj in practice • Purity of molasses
When performing factory tests it is, for practical reasons, far • Crystal % Total solids
easierto obtain size-frequencies-by-mass through sieving than
size-frequencies-by-number through counting under a micro- • Meancrystal size. If this is given as mean-size-by-mass (Mean
scope. To deal with this, the GNE p.d.f which describes the Aperture), it will be converted to mean-size-by-number.
size-frequency-by-number was transformed to a p.d.f which • Total mass flow rate (usually only available for seed or pro-
describes the corresponding size-frequency-by-mass, but still duct stream). '
52 Proceedings of The South African Sugar Technologists' Association - June 1985

9. Setting up equations on compartment material balances K_GROWTH = Growth rate parameter (mm/h). In
the Evaluation mode this is one of
These equations can also be linearised, and are set up in a the process parameters.
DO-loop over compartments 1 to N.
TEMP_EFFECT = Temperature effect on the reaction
• Sucrose mass balance: rate. This is an Arrhenius type
SUC(J)- SUC(J-l) = (SOLIDS(J) - SOLIDS(J-l» * (Given) function of the massecuite temper-
SYRUP_PURl 100 ature MSC_TEMP, per Wright. 14
• Solids mass balance: MSC_TEMP = Massecuite tempf'C),
SOLIDS(J) = SOLIDS(J-l) + (Given)SYRUP(J) * (Given) This is calculated from the boiling
SYRUP_BX/100 point elevation (BPE) relativeto the
This equation can only be set up for compartments where given pan vapour-temperature
the amount of syrup feed is known, even if it is zero. YAP_TEMP. The BPE(J) is ex-
pressedin terms ofthe primary un-
10. Setting up equations on processes within compartments knowns in an empirical relationship
derived by the author and given in
These equations are very much non-linear, and express the Appendix 2.
fundamental mass and heat transfer processes taking place.
CW_RATIO = Impurity-to-water ratio of molas-
Because they have to be solved simultaneously with all the ses,whichcan directlybe expressed
previouslydiscussed pure linear equations, the Newton-Raph- in terms of the primary unknowns.
son method is used, whereby the partial derivatives with regard
= (SOLIDS(J)-SUC(J»/WATER(J)
to the primary unknowns form part of the coefficient matrix
of the linear equations to be solved. Fortunately these equa- The way in which the impurity-to-
tions, in spite of their complexity, are differentiable, so that the water ratio affects the rate of mass
computer program can calculatethe values for the expressions transfer, as determined by Wright!-
of the partial derivatives. In another publication, Hoekstra- for Queensland syrups, is used.
describes the general principles of the method. SUPERSATN(J) = Supersaturation (dimensionless).
The procedure again is to deal with the compartments in In the Designand Operation modes
turn through a DO-loop with index J going from 1 to N. this willbe given per compartment,
and in the Evaluation mode it will
(a) Growth rate equation be unknown.
The basic assumption is that the increase in the linear For equation (1) the unknown SIZE must be expressed
dimension of a crystal is equal to the growth rate (meas- in terms of primary unknowns:
ured as increase in length per unit time) multiplied by its XTAL(J) = 109 X N_XTALS *
retention time. According to Wright!' and Austmeyer', the XTAL_DENSITY * VL3 *
rate of crystallisation is dependent on crystal size, in that SIZE(J) ** 3
it is proportional to (crystal size)", where Wrightestimated
n = 0,8. The relationship proposed by Pot" is far more where
complicated. In view of the seed and product sizesin most XTAL_DENSITY = crystal density (tons/m-)
applications being of roughly the same respective sizes, VL3 = factor of mean volume to cubed
and the increasein sizeusuallybeingby a factorof between mean size mentioned in Section
1,5 and 2, it was felt that, for the way in which CORPSE 6(b).
was to be used, the error in letting the growth rate in the
model be independent of crystal size would be sufficiently In this way it is possibleto expressequation (1) entirely
small to justify the gain in simplicity. Gross also ignored in terms of the primary unknowns.
the size-dependence effect in his model.
(b) Supersaturation equation
Expressing it in an equation:
Whether the supersaturation is itself known or un-
SIZE(J) SIZE(J-l) + GROWTH_RATE *
= known, it has to be expressed in terms of the primary
RET_TIME (1) unknowns. The ICUMSN definition of supersaturation
where SIZE(J) = Size (length) of crystal, mm can be transformed to:
RET_TIME = Retention time (h) SUPERSATN(J) = SueW_RATIO * (100 - SOLUB)/
(SOLUB * (1 - 0.088 *
= CELL_VOL(J)/MSC_VOL.
C W_RATIO». . . . . . . (2)
= Compartment volume (m-). In the
where
Design mode this is one of the pri-
mary unknowns, and assumed to SUC_W_RATIO = Sucrose-to-water ratio of mother
be equal for all compartments. A liquor (molasses)
dead-volume fraction of zero is = (SUC(J) - XTAL(J»/WATER(J)
assumed. SOLUB = Solubility of sucrose in pure solu-
MSC_VOL = Massecuite volume (mvh) which tion (%). This can be expressed in
can be empirically expressed in an empirical polynomial equation
terms of the primary composition in terms of MSC_TEMP, per
variables. Charles'.
GROWTH_RATE = Growth rate (mm/h) The effect of the impurity-to-water ratio C W_RATIO
= K..GROWTH * (SUPERSATN(J)- on the solubility of pure sucrose is per Wright>.
1.0046) * TEMP_EFFECT * EXP To obtain meaningful results from plant data when in
(-1.75 * CW_'RATIO) the Evaluation mode, it is necessary to make the assump-
Proceedings of The South African Sugar Technologists' Association - June 1985 53
N
tion that the unknown supersaturation is the same in all (3.6/LAT_HT(STEAM_TEMP» I (STEAM_TEMP -
compartments, ie at a common value K_SUPERSATN = 1=1
SUPERSATN(J) for all J = 1 to N. MSC TEMP) * (AREA(J) * HTC(J» (Given)
CONDENSATE
(c) Heat Transfer Equation This relationship is likely to arise in the Evaluation mode,
This is another important non-linear relationship in the when HTC as well as BPE will be functions of the primary
vacuum pan, which is expressed in terms of some of the unknowns. This will make the equation non-linear in the
process parameters. primary unknowns, so that it will also be part of the Newton-
Taking a water balance: Raphson method, similar to the equations described in Sec-
tion 10.
Total water entering compartment J - Total water leav-
ing compartment J
12. Solution of the equations
= Water evaporated

= Heat transferred through the calandria walls -;- La-

For a pan with 12 compartments,and the mainstreams there-
tent heat of evaporation. fore indexed 0 - 12, there will be about 56 primary unknowns.
When in the Design or Operation mode, the number of equa-
More specifically, for compartment J: tions should exactly equal the number of primary unknowns,
WATER(J-l) + (Given) WASH(J) + SYRUP(J) * (1 - so that the equations of each iteration can be solved directly.
SYRUP_BX/lOO) - WATER(J) When in the Evaluation mode, there will usually be more equa-
= 3.6 * HTC(J) * AREA(J) * tions than unknowns because of deliberate over-specification,
(STEAM_TEMP - and the program will first calculate the cross-product matrices
MSC_TEMP(J)/ as part of the linear regression fitting procedure and then per-
LAT_HT(VAP_TEMP) (3) form the solution of simultaneous linear equations. In another
publication Hoekstra- explains the principle behind this ap-
where WASH(J) = amount of dilution water added
plication of multiple linear regression analysis.
(tons/h)
The simultaneous linear equations are solved by the Gaus-
SYRUP(J) = amount of syrup feed added
sian elimination procedure. The size of the matrix of coeffi-
(tons/h) cients and being in double precision, make it too large for the
SYRUP_BX = solids % mass of syrup computer memory, and instead it is stored on disc in a record
HTC(J) = heat transfer coefficient direct access file. Each time the program sets up an equation,
(kW(m 2°C» its coefficients are written to file. During the course of the so-
AREA(J) lution procedure, the relevant row is read from file, the neces-
= heat transfer area of calandria
sary manipulations done on it, and written back to the disc file.
for compartment J (m-)
Because the iterative Newton-Raphson method of solution
LAT_HT(VAP_TEMP) = latent heat (kJ/kg) of pan va-
is used, the whole procedure ofsetting up and solving the equa-
pour, taken to be a linear func-
tions has to be repeated several times. To indicate how far the
tion of VAP_TEMP ro process has converged, the intermediate results can be printed
= temperature of boiling masse- out after each iteration. To assist in tracing errors, the user can
cuite CC), as mentioned in Sec- also request the entire matrix of coefficients of the equations
tion 10(a). to be printed out.
STEAM_TEMP = temperature of the heating va-
pour CC), which could be a pri-
mary unknown when run in the 13. Input and Output
Design or the Operation mode. The input data to the program is in the form of a stream
Equation (3) is only applied to compartments for which the sequential file of a standard format, which the user fills in via
amount of dilution water WASH(J) is known. the keyboard. The file format provides prompts and other in-
The equation can be used in the following ways: dicators. The data is in free-format, meaning that an item does
(a) HTC's and AREA's given: This is for the Operation mode. not have to be positioned exactly within certain columns on
the screen. Where a data item is not known, the user inputs a
(b) HTC's unknown, AREA's given: This is for the Evaluation
mode, where the HTC is assumed to be a linear function
o which the program interprets as "no data given" and will
therefore not set up an equation for it. The file must also be
of the compartment index J, with the primary unknowns given initial estimates of the priinary unknowns, which are
slope K_UM and intercept K_UB as parameters. required to start off the Newton-Raphson method.
(c) HTC's given, AREA's unknown: This is for the Design For the convenience of the user, CORPSE reproduces the
mode. AREA(J) is specified as proportional to primary abovementioned input to the computer in the same format as
unknown CELL_VOL(J), and can thus be calculated. the final output, but shows values only for the given data items
- the rest are not shown as zeros, but as blank spaces.
Figure 3 and Figure 4 respectively show a typical input and
ll. Setting up equations on the overall pan output of a run.
Of these equations there can be at most only one of each.
• Overall Syrup Feed balance: 14. Comments on the results
SOLIDS(N)- SOLIDS(O) = (Given) TOT_SYRUP * (Given)
SYRUP_BX/IOO When in the Evaluation mode and applying the least squares
fit, many of the final results might not be quite the same as
• Pan Exhaustion: their corresponding values in the input data. This will be be-
XTAL(N)/SUC(N) = (Given) PAN_EXH/IOO cause they are back-calculated from the values of the primary
• Overall Calandria Condensate balance: unknowns which were solved in the final iteration, and will be
 S4 Proceedings of The South African Sugar Technologists' Association - June 1985

ClIIlIMtmilY OPERAlING RAW PAM SIKUlAlIOli AMD ElJtllUATlOM ( C.O.R.P.S.L )

tUN Mlh 26)0) lULL: KAlDSTOIIE OBJECT: A-PAN TRAILS: EVAl-Of 26/10/83, SHIFT 08 - 16.
S\lMY OF INPUT DATA.
=========================:
oVERA LL PAN.
=====================
110. OF COKPARTKENTS (N h 12 PAN EXHAUSTN:
KOLA$ES OR SYRUP FEEDS: SOLIDS %SOLUTION: 67.3 PURITY (%) : 83.3 KASS <TONS/HR): 44.2
IIEATIMG STEAK TO PAN : PRESSURE( "'Po ABS. h 93.0 TEKPERATURE( BEG.C h 97.6 KASS <TONS/HR h 26.0
VAflO~ ABOVE PAM : PRESSURE( IlPo ABS. h 15.0 TEKPERATURE( BEG.C >: 54.0 KASS (TOMS/HR h

PAflAllETERS OF PAN: NO.OF XTALS (*1E+9): 75.00 EKPIRICAL EQUATIONS: HTC(J) = -0.010 *J
GROliTH (K_GROWTH) : DISTR}ARAK( J) = -0.080*( J-6) + 4.07~
SOLUBILITY (K_SOLUB): 1.00 ASSUKED GRAIN SIZE DISTRIBUTION: GENERALISED NEGATIVE EXPONENTIAL
KASS ECUI TEST REA KS.
ee: ::::===============================

STREAK REF.NO.:! 0 1 2 3 4. 5 6 7 8 9 10 11 12
-------------------!----------------------------------------------------------------------------------------------------------------
KASS RATES( TOIIS/IIl h
TOTAL KASS
SOLI IS
SUCROSE(IMCL.XTAL:
WATER
XTH.
IJOL.FLOWRATE,CU.K/H:
XTH. SIZES (KK)
BY KASS FRED: KA : .437 .456 .478 .554 .530 .582 .602 .604
BY NR. FREQ: L1: .237 .246 .256 .294 .279 .302 .309 .305
nASSEQU.PROPERTIES :
SOLIDSXKASS<BRIX}: 86.80 89.00 88.70 88.80 89.00 89.00 89.10 89.60
SUCRllSEXKASS<PQL}: 73.60 75.90 75.40 75.30 75.50 75.30 75.10 76.10
PURITY (X)
XTAL %SOLIDS
DENSITY (GK/CC) :
MOLA$ES PROPERTIES:
SOLIDS%KASS<BRIX}: 78.20 80.20 80.00 80.50 80.30 80.20 80.60 80.90
SUCROSE%KASS<POL>: 57.90 57.80 57.90 58.40 57.60 58.00 57.60 54.60
PURITY %<NUTSCH}:
SUPERSATN (FRACN):
IKPUR/WATER RATIO:
C0 KPAR TKEN TS.
::::,::====================
toHPARTKENT REF.NO.! 1 2 3 4 5 6 7 8 9 10 11 12
-------------------!---------------------------------~------------------------------------------------------------------------------
SPACE DETAilS
IJOLUKE (CU.K.) ·
DEAD VOL.(FRACTN):
9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70
RETENTION (HRS.) :
CONTENTS PROPERTIES:
GR(IITH RATE( KK/H h
KSC lEKP.( DEG.C. h
IIA9S RATES<TOtlSlHR h
SYRUP FEEDS 0.00 0.00 0.00 0.00 0.00
IIOIJEI£NT WATER
EIJAPORATIOtl
· 0.00 0.00 0.00 0.00 0.00 0.00 0.00
HEAT TRANSFER
HEAT TR.AREA,SO.K: 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0
H.T.C.,KW/SQ.K-C :
HEAT TR., KW
FIGURE 3 Input data to a CORPSE run in Evaluation mode.
Proceedings of The South African Sugar Technologists' Association - June 1985 55
COHTINUOUSLY OPERATING RAIl PAN SIIIUlATION AND EVAlUATION ( C.O.R.P.S.E. ) DATE:27KAR85 TIKE: 17:49:04

RUN NO: 26.10.2 KILL: KAIDSTONE OBJECT: A-PAN TRAILS: EVAL.OF 26/10/83, SHIFT 08 - 16.

tE\illTS OF RUN. ITERATION 6 OF 6 56 UNKNOWNS, 90 RELATIONSHIPS.

=================
o VERA LL PAN.
=====================
1Kl. OF COKPARTKENTS (N): 12 STD.DEV.OF RES. OF PREV.ITERATION = 0.128 TONS/H PAN EXHAUSTN: 59.5
KOLASSES OR SYRUP FEEDS: SOLIDS XSOLUTION ~ 67.3 PURITY (X) : 83.3 KASS (TONS/HR): 43.6
HEATING STEAK TO PAN : PRESSURE(kPo ABS.): 93.0 TEKPERATURE(DEG.C): 97.6 KASS (TONS/HR): 25.7
lJAPOlR ABOVE PAN : PRESSURE(kPo ABS.}: 15.0 TEKPERATURE(DEG.C}: 54.0 KASS (TONS/HR): 24.4

PARAKETERS OF PAN: NO.OF XTALS (*lE+9): 211.84 EKPIRICAL EQUATIONS: iiTC( J} = -0.010 *J +.0.480
GROlIlll (KJROIITH): .82841 DISTR}ARAK(J} = -0.080*(J-6} + 4.074
SOLUBILITY (K_SOLUB): 1.00 ASSUKED GRAIN SIZE DISTRIBUTION: GENERALISED NEGATIVE EXPONENTIAL

KASS ECUI TEST REA KS.

=::: ~~================================
STREAK REF.NO.:! 0 1 2 3 4 5 6 7 8 9 10 11 \2
....._--------------- !-- --------------------------- ------------------------------------------------------------------------'------ ---
HASS RATES( TONS/IIl >:
TOTAL KASS 24.29 24.01 30.63 37.00 42.01 41.91 42.66 46.55 51.19 56.52 56.44 56.07 56.29
SOLIrs 21.09 21.40 27.17 32.72 37.31 38.09 37.96 41.52 45.54 50.30 50.27 50.34 50.42
SUCROSE(INCL.XTAL: 18.16 18.58 23.45 28.44 32.05 32.33 32.38 35.55 38.92 42.72 42.65 42.36 42.27
WATER 3.20 2.61 3.46 4.28 4.70 3.82 4.70 5.03 5.65 6.22 6.17 5.73 5.87
Wi.. 9.80 11.41 13.48 15.92 18.17 18.34 18.68 20.39 22.14 23.17 24.16 24.67 25.17
lJOL.FLOIIRATE,CU.K/H: 16.78 16.39 20.96 25.35
. 28.72 28.35 29.14 31.77 34.98 38.61 38.54 38.15 38.34
Xl tL SIZES (11K)
BY KASS FREQ: HA : .431 .455 .482 .511 .535 .539 .544 .562 .580 .591 .601 .608 .615
BY MR. FREQ: L1 : .233 .245 .259 .273 .284 .285 .286 .294 .301 .305 ~309 .310 .311
MSSEQU.PROPERTIES :
SOlIDSXKASS<BRIX): 86.84 89.13 88.70 88.44 88.81 90.87 88.99 89.19 88.96 88.99 89.06 89.77 89.57
SUCROSE%KASS<POL): 74.78 77.36 76.57 76.86 76.31 77.14 75.91 76.36 76.02 75.59 75.56 75.55 75.09
PURITY (%) 86.12 86.79 86.33 86.91 85.92 84.89 85.31 85.61 85.46 84.94 84.84 84.16 83.83
XTAL %SOLIDS 46.45 53.30 49.61 48.65 48.69 48.15 49.21 49.11 48.62 46.07 48.07 49.01 49.92
DFt6ITY (GK/CC) : 1.45 1.46 1.46 1.46 1.46 1.48 1.46 1.47 1.46 1.46 1.46 1.47 1.47
KOlASSES PROPERTIES:
SOLIOS%KASS<BRIX): 77.94 79.29 79.81 79.71 80.28 83.78 80.40 80.77 80.54 81.35 80.87 81.74 81.14
SUCROSEXKASS<POl): 57.73 56.87 58.17 59.40 58.25 59.37 57.14 57.94 57.75 58.63 57.26 56.34 54.95
PURITY %<NUTSCH): 74.07 71.72 72.88 74.52 72.56 70.86 71.07 71.73 71.70 72.07 70.80 68.93 67.72
SUPERSATN (FRACN): .98 1.04 1.09 1.10 1.12 1.41 1.11 1.14 1.13 1.19 1.14 1.19 1.12
IKPUR/IIATER RATIO: .92 1.08 1.07 1.00 1.12 1.50 1.19 1.19 1.17 1.22 1.23 1.39 1.39
C 0 KPAR TKEN TS.
:::'.::====================
tolf'ARTKENT REF .NO.! 1 2 3 4 5 6 7 8 9 10 11 12
-------------------!-~_ .._-----------------------------------------------------------------------------------------------------------
Sf'ACE DETAILS
VOlUKE (CU.K.) : 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70 9.70
DEAD VOL.( FRACTN ):
RETENTION (HRS.) : .592 .463 .383 .338 .342 .333 .305 .277 .251 .252 .254 .253
CONTENTS PROPERTIES:
GROIITH RATE(IIK/H): .022 .022 .025 .021 .012 .019 .019 .019 .018 .017 .014 .014
KSC l£KP.( BEG.C.): 61.4 61.5 61.4 61.7 63.4 61.8 61.9 61.9 62.2 62.0 62.5 62.3
IlASS RATES( TONSlHR >:
SYRUP FEEDS .46 8.57 8.25 6.81 1.16 -.19 5.30 5.97 7.07 -.04 .11 .13
KOlJEl£NT WATER : 1.60 .33 .36 .37 .76 3.00 .61 .64 .15 1.83 1.31 1.84
EVAPORATION 2.34 2.28 2.24 2.17 2.02 2.06 2.01 1.96 1.90 1.86 1.78 1.75
HEAT TRANSFER
HEAT TR.AREA,SG.K: 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0 91.0
H.T.C.,KII/SQ.K-C : .470 .460 .450 .440 .430 .420 .410 .400 .390 .380 .370 .360
HEAT TR., KII 1548.8 1509.8 1481.5 1436.4 1336.1 1366.6 1329.6 1300.6 1256.6 1229.3 1180.9 1155.7
FIGURE 4 Results from a CORPSE run.
S6 Proceedings of The South African Sugar Technologists' Association - June 1985

somewhere "in-between" the over-specified input data, in the tions often suffice, and the run time is down to about 15 minutes
same way that a least-squares fit of a straight line to a number because of this and because no-least-squares fit is required.
of points will not, in general, pass directly through anyone of
the points. 16. Concluding Remarks
As an example, part of the results ofa series of6 tests which
were taken on each of 3 shifts on 26 and 27 October 1983 on CORPSE makes no pretence at being a rigorous model of
a continuous A-pan at Maidstone and subsequently run on the basic principles of sucrose crystallisation. It must be con-
CORPSE in the Evaluation mode are shown: sidered as a tool to be used not only for comparing the relative
performances' of different continuous pans when in the Eval-
Date: 26.10.83 27.10.83 uation mode, but also for applying the empirically obtained
Shift:
--------------
2 3 2 3 process parameters of growth rate constant and heat transfer
coefficients to simulation purposes when in the Design or Op-
Growth rate constant eration mode. The previously mentioned self-compensating ef-
K_GROWTH (mm/h): 0.835 0.828 0.646 0.865 0.885 0.521 fect of errors resulting from certain assumptions or
Heat transfer coefficient (kW/(m 2'C)): simplifications would apply, among others, to grain size dis-
HTCofCompartment 1:0.472 0.470 0.478 0.392 0.400 0.456 tributions- and their changes from compartment to compart-
HTC of Compartment ment, the cubic shape of the crystals, the growth rate
12: 0.362 0.362 0.368 0.282 0.290 0.346 independence of crystal size, constancy in the number of crys-
tals, the zero dead volume per compartment and the impurity
Common supersaturation: effects on mass transfer.
K_SUPERSATN 1.20 1.17 1.16 1.19 1.21 1.26
Considering the practical problems in measuring some of the 17. Acknowledgement
flow rates accurately and in obtaining representative samples
from the compartments of a plant which has a long hold-up I want to pay tribute to Dr Ross Broadfoot- for his Ph.D.
time, the above variations in results seem reasonable. thesis on continuous crystallisers. The orderly way in which he
discussed the relationships was a great help in the writing of
The following remarks apply: the program.
Growth rate constant 18. REFERENCES
The growth rate constant in the formula proposed by Wright!' 1. Austmeyer, K. E. (1983). Betrachtungen zur kontinuierlichen
is 7.418 mm/h, which is higher than that obtained here by a Nachproduktverdampfungskristallisation in der Zuckerfabrik Lage. Zuck-
factor of about 10. This could be due to the many assumptions erind 108 (10) p. 927-934.
and simplifications which have had to be made in the com- 2. Batterharn.R. J. and Norgate, T. E. (1986). Boiling Point Elevation and
superheat in impure cane sugar solutions. Int. Sugar J. 78: 359-364.
puterised model, compared with the rigorous analysisby Wright. 3. Broadfoot, R. J. (1980). Modelling and optimum design of continuous sugar
pans. Ph.D. thesis, University of Queensland.
Heat transfer coefficients 4. Charles, D. F. (1960). The solubility of pure sucrose in water. Int. Sugar
Because the factory installations on which the evaluations 1. 62: 126.
were performed thus far had no facilities for measuring the 5. Gros, H. (1979). A Modelfor Vacuum Pan Crystallisers, 12th Symposium
Computer Appl. in Chem. Eng. (2) 1010-1020.
amounts ofsyrup and dilution water going into individual com- 6. Hoekstra, R. G. (1983). A flexible computer program for four-component
partments, except of course where the amount was zero, it was material balances in sugar industry boiling houses. Int. Sugar J. 85 (1016):
not feasible for CORPSE to calculate a heat transfer coefficient 227-232, (1017): 262-265.
for each compartment separately. To use the same value of heat 7. Honig, P. (1953). Principles of Sugar Technology. Elsevier Amsterdam,
Vol. I. p. 60-67.
transfer coefficient for all compartments would be too severe 8. Hugot, E. (1972). Handbook of Cane Sugar Engineering. Elsevier Am-
an approximation, and it was decided to let the HTC decrease sterdam, p. 491-493.
in a linear manner from the first to the last compartment. The 9. ICUMSA (1962). Rept, Proc. 13th Session, Hamburg, 40.
rate of decrease was estimated for a 12-compartment pan to 10. Lyle, O. (1970). Technology for sugar refinery workers. Chapman & Hall,
London p. 627. .
be 0.0IkW/(m2'C) per compartment, and was specified as an II. Moore, W. J. (1963). Physical Chemistry. Longmans London, p. 132.
input for the slope K-UM of the line, so that CORPSE then 12. Pot, A. (1983). Industrial Sucrose Crystallisation, Ph.D. dissertation Delft
merely had to calculate the intercept K-UB of the line. University of Technology. '
13. Rein, P. W.,Cox M. S. G., Love D. 1. (1985). Analysisof crystal residence
Supersaturation time distribution and size distribution in continuous boiling vacuum pans.
Proc S Afr Sug Technol Ass 59.
Each" supersaturation value given in the table is the solution 14. Wright, P. G. (1971). A model of industrial sugar crystallisation Ph.D.
to the primary unknown K_SUPERSATN which is taken to Thesis, University of Queensland. '
be common for all compartments, per Section 10(b). They are
generally at the level one would expect. When re-calculating it APPENDIX 1
back for individual compartments J from the applicable mass GENERALISED NEGATIVE EXPONENTIAL PROBABILITY DENSITY
values SOLIDS(J), SUC(J), WATER(J) and XTAL(J), there is FUNCTION
considerable variation amongst compartments, as can be seen .The probability density function (p.d.f.) for the size-frequency-by-number
in Figure 4. distribution IS:

I\x) == £ . exp(-(x/k)p)
15. The Computer k J'(l/p)
The computer used is a Data General Eclipse S/140 with where x = size (length), > 0
3/4 Megabyte of memory and 75 Megabyte of disc storage. It P = shape parameter, > 0, dimensionless
operates under the Data General Advanced Operating System,
and the program was written in the language PLlI. Because of k = scale parameter, > 0, with same dimension as x
its size, the program had to be broken up into smaller proce- J'(r) = Gamma function with parameter r
dures, and overlays used.
When in the Evaluation mode, 6 iterations are usually re- == 1" t,-I eo' dt, for r > 0
quired, which require about 40 minutes of run time for a 12-
compartment pan. In the Design or Operation mode 3 itera- = (r-) ! for r a positive integer.
 Proceedings of The South African Sugar Technologists' Association - June 1985 57
The corresponding cumulative distribution function (c.d.f.), which is used in From thermodynamic principles, the following relationship can be approx-
determining the parameters p and k by fitting to experimental results of grain imated for pure, dilute solutions from the; Gibbs-Helmholtz equation (see
counting, is obtainedby puttingt = (x/k)« in the above equationand integrating: Moore"):
Probability (Randomly selected grain size « x)
BPE = K mB T'
= F(x)
rnA
e-t
f
' Xl kJP t(l/p - I)
= dt where K = constant
" T(I/p)
m, = mass of solute
The mean-size-by-number is obtained from:

L, = f' x /tx)dx
m, = mass of solvent
T = boiling point of pure solvent, 'K
= kT(2/p)/T(l/p) This suggested using the form

r
BPE = AO*(S_W_RATIO)A.' * (T_ABS)A'
The p.d.f. for the size-frequency-by-mass is given by:
where S_W_RATIO = Solids-to-Water ratio of mother liquor
t3f(t) dt
SOLIDS(J) - XTAL(J)
g(x) = f'" WATER(J)
t3/tt) dt
" T_ABS = Absolute temp. of pan vapour
= £ . (~)3 exp(-(x/k)p) AO, A I and A2 = Regression constants
k k T(4/p)

G(x) == r
and the corresponding c.d.f. by:
l kiP

t(4~~;~r dt
The mean-size-by-mass is obtained from:
In a pure, dilute solution one should expect that Al = I and A2 = 2, but
lettingthem findtheir ownbest-fit levelwouldhelpcompensate for non-ideality.
Impurity of the mother liquor was brought into account by the additional
product term (NUTSCH_PUR)A3
where NUTSCH PUR Purity of mother.liquor (as a fraction)
- SUC(J) - XTAL(J)
MA = f'" xg(x)dx
SOLIDS(J) - XTAL(J)
= kT(S/p)/T(4/p) A3 = Regression constant.
Note that, for a pure solution, this term = I, whatever the value of A3.
It can similarly also be shown that mean-volume-by-number
Bytaking logs, the entire equation could be Iinearised for regression analysis
V= f'" x3 f(x)dx purposes:
2n BPE = 2n AO + Al * 2n (S W RATIO) + A2 * 2n(T ABS) + A3 *
2n(NUTSCH_PUR) - - -
= k 3T(4/p)/T(l/p)
Data from Honig', whichdealt with comparatively pure solutions, was used
This gives the relationships which are required in the running of CORPSE: to determine the constants AO, Al and A2. Data from Batterham & Norgate',
VL3 = V/Lf Hugot' and Lyle!" was used to determine A3.
The final equation obtained was:
T(4/p)(T(l/p»'
BPE = 0.1379 * (S_W_RATIO)o.,o, * (T_ABs),·m * (NUTSCH_PUR)-o",
(T(2/p»'
and MAIL = T(S/p) T(I/p)
I T(2/p) T(4/p)
Note that the scale parameter k cancels out, so that these relationships are
only a function of the shape parameter p. .

APPENDIX 2
DERIVATION OF EMPIRICAL EXPRESSION FOR BOILING POINT
ELEVATION OF CONCENTRATED SUGAR SOLUTIONS
The task was to derive an empirical relationship for boiling point elevation
(BPE) in terms of the concentration of the mother liquor (molasses), the boiling
point of pure water at the particular vapour space operating pres~ure and the
degree of impurity, by applying multiple linear regression analysis to tabular
data.