Professional Documents
Culture Documents
27: 2005
__________________________________________________________________________
By
P.G. WRIGHT
Abstract
IN RESPONSE to requests from managers in overseas factories, a number of
technical benchmarks have been developed and applied to a number of factories,
including those involved in the manufacture of refined and sulphited white
sugars, as well as of raw sugar. These benchmarks are presented and discussed
in this paper. An expression for the overall pol recovery in sugar, modified from
the Australian CCS formula, has been used as a useful benchmark value
applicable to all factories. A related formula for expected losses can be used as a
benchmark for the pol losses % pol in cane. Benchmarks for the individual
losses of pol in bagasse and in filter cake are described. These are based on
available regression expressions for good practice performance, and, in
addition to the composition of the cane, these take into account the availability
of imbibition and wash water to the process as well as the number of crushing
mill stages and the rotary filter area. The benchmark loss in final molasses
requires that benchmark values for these upstream losses be known before the
flow to final molasses can be calculated. For this calculation, the apparent
disappearance of impurities during processing has to be allowed for,
interconversions made between true and apparent purity analyses, and
benchmark purity values for final molasses need to be obtained from selected
exhaustion formulae. The benchmark calculations could be packaged into a
spreadsheet program which can process laboratory information data and produce
appropriate plots of actual and benchmark data for each operating period. The
techniques are presented in the hope that they might enable factory performance
to be more critically analysed and better understood by management and
technical staff.
Introduction
Benchmarking is a tool to measure the manufacturing performance against a
standard/target which is achievable in the world or local sugar industry.
In the course of auditing of cane sugar factories overseas, repeated requests have been
to provide appropriate benchmark targets for the overall factory recovery performance as well
as the individual benchmark targets for the key process areas. Present benchmarks were
437
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
examined and considered to be deficient in taking into account variations in cane quality and
factory arrangements, and this situation has prompted work on the formulation of improved
benchmarks as put forward in this paper.
Past industry manuals (Anon., 1970) separated the chemical control of the sugar
process into two distinct phases.
The first phase assembles the values associated with the process:
the quantity of materials entering, leaving or in process,
the composition of the original, intermediate and final product streams,
the relevant details of the materials added to the process, and
the conditions under which the various stages are operated.
The second phase deals with the comparison of the performance of the process against
selected benchmarks, so that managers can evaluate how well the equipment is operating and
make adjustments to improve the performance of deficient areas and in establishing
manufacturing excellence in that area.
A major deficiency in present chemical control systems is perceived as being in the
lack of formulation of benchmarks for the key elements in the manufacturing process. The
benchmarks selected here are those that have important significance for the economic
viability of the sugar factory, notably the pol recovery as sugar, the overall losses of pol, and
the losses of pol from the individual output streams. This paper seeks to define and develop
more appropriate benchmarks so that the comparisons in this second phase can be more
effective.
For practical reasons there is an approximation of pol to sucrose and dry substance to
Brix. As well, the analysis of cane is often not done directly but is estimated from the pol and
brix measured in first expressed juice (FEJ)1. A recovery indicator widely reported was the
1
The cane analysis estimate is made by the 3 & 5 formula, with the following relationships: -
polcane = po1FEJ x Fpol
Bxcane = BxFEJ x FBx
Where Fpol = (100 - (Fibrecane + 5) / 100)
Where FBx = (100 - (Fibrecane + 3) / 100)
439
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Coefficient of Work (COW), which is the percentage ratio of 94 Net Titre (NT)2 sugar
produced to the measured CCS. However, in recent years the NT estimation of the refined
sugar yield from raw sugar has been supplanted by the IPS3 estimation (presumably a more
accurate assessment of refined sugar yield from a given raw sugar), and a Pool Sugar Index
(PSI) defined as the simple ratio of IPS sugar produced to the measured CCS. Historical
good practice values of the PSI and COW have been used as effective factory benchmarks
in Australian factories.
Yield benchmarks in overseas factories
Many factories overseas have been found not to report a benchmark against which their
reported yield of product sugar can be compared. They simply report the yield (or
rendimiento in Central and South American factories) of product sugar, sometimes
corrected to a nominated standard sugar pol, as a % on cane, along with the total losses in the
pol balance, and then compare these reported values against historical trend values and values
from neighbouring factories.
However, where overseas factories operate the recommended ISSCT Methods for cane
analysis (outlined above), it is found that a modification of the simple CCS formula can be
adapted as a benchmark against which the recovery of sugar of any reasonably high pol
specification can be judged. A useful adaptation of the formula is based on the pol % cane
and the apparent purity of the mixed raw juice (Pmj) as the RCS value in Equation (2):
100 50
Sugar , % cane ( RCS ) 0.945 x pol cane 1.50
( 2)
pol sugar P mj
The juice purity is taken from the mixed juice analysis as this is the reference point
for both the recommended ISSCT cane analysis systems. A value of the multiplier of 0.945
gives a reasonable benchmark for refined sugar production, though it is seen that this is
modified by the correction factor 100/(polsugar), to allow for the production of lower pol
sugars such as white sugar or raw sugars of higher pol.
For low purity sugars, it would be more correct to use a correction factor to reflect
the yield of pure sugar from the product sugar, using an SJM formula and assuming either a
molasses component of standard purity or a purity related to the ratio of the reducing
sugar/ash ratio of the sugar.4
An example of the variation of this benchmark during a season is given in Figure 1
for an overseas factory producing remelt refined sugar of 99.89 pol. In this particular case,
the average ratio of the measured sugar recovered to the RCS value was 0.9992, with a
standard deviation of 0.01.
2
The Net Titre (NT) of a sugar is calculated from its analysis (pol, reducing sugars (RS), and ash) as
follows:
NT = pol RS (5 x Ash)
3
The International Pool Sugar (IPS) of a sugar is calculated by an SJM-type formula as a function of
pol, moisture, impurities, and the expected true purity of the final molasses estimated at a RS/Ash ratio
of 0.9 that in the sugar.
4
The purity used could be the same as that used in the Australian IPS estimation, viz. the expected true
purity of the final molasses estimated at a RS/Ash ratio of 0.9 that in the sugar.
440
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Fig. 1An example comparison of actual total sugar recovery and the
benchmark recovery value (RCS).
A related formula for Expected pol loss is given as the difference between the pol in
cane and the RCS value, and this could be used as a benchmark for the total pol losses.
441
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
From recent experience, it would appear that a more appropriate benchmark for pol
extraction could be the value predicted from mill train regression models with fibrecane as
only one of several independent variables, others being pol % cane [polcane], the imbibition
water % fibre [WF], and the number of crushing mill stages [nmills].
It appears logical that these extra independent variables be included as having a
significant effect on the extraction in a crushing train.
Suitable regressions5 have been obtained by Dr Geoff Kent (pers. commun., Sugar
Research Institute, 1999) by exercising variables in the SRI MILSIM model over an
appropriate practical range (viz. number of crushing mills range 4 to 6, and an imbibition
range 180% to 300% fibre).
Kents regression for pol extraction (polextn, %polcane) for Australian type milling
units was a function of fibrecane, polcane, nmills, and WF had a multiple R squared value 0.9863
and a residual standard error of 0.11.
Kents regression for bagasse purity included Pcane along with the same independent
variables and had a multiple R squared value 0.9964 and a residual standard error of 0.44.
From the cane analysis, the imbibition water application and the number of crushing
stages, the pol extraction value can be calculated from an SRI regression and used as a
benchmark for the measured pol extraction for a factory extraction train.
The benchmark pol loss in bagasse per cent pol in cane is then given by Equation (3):
An example comparison of actual and benchmark pol loss in bagasse for an overseas
factory over a season is shown in Figure 2.
It is seen there that the actual reported pol loss was mostly about 0.45 units lower
than the benchmark value.
5
Being derived from the SRI MILSIM model, these regressions are the intellectual property of the Sugar
Research Institute SRI). It is probable that improved regressions could be obtained in the future if the
cane pol in open cells or preparation index was included as an extra independent variable.
442
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Fig. 2An example comparison of actual and benchmark pol loss in bagasse.
Fortunately, some reported regression analyses relate recorded good practice values
of pol % mud solids in cake (pol%MS) to the independent variables during a season of good
filter operation (Wright et al., 1997).
The independent variables examined included mud solids loading (MSL), wash water
loading (WWL), and wash water % mud solids in cake. Their regression on filter station data
from the 1994, 1995 and 1996 seasons at the Marian factory of the Mackay Sugar
Cooperative was as in Equation (4):
To use this regression to formulate a benchmark, the wash water loading has to be
calculated from knowledge of the applied wash water rate and the effective operating filter
area. The mud solids rate and loading is more difficult to estimate. At least two estimation
methods can be discerned, as listed.
443
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Method 1. The mud solids rate MSrate can be estimated directly from the
recorded filter cake production rate and the recorded analysis of the
approximate mud solids content in the filter cake;
Method 2. The mud solids rate can be estimated indirectly from knowledge
of the basic mud solids in dirt-free cane6, and the soil (non-bagacillo solids)
content in mixed juice taken before the addition of any recirculated materials.
An alternative to the latter would be knowledge of the average soil content of
the cane and the fraction of this soil passing to the mixed juice7. The filter
performance criteria (the filter area requirement, pol loss in mud cake, and
bagacillo filter aid requirement) are strongly related to level of total mud
solids in juice8.
Method 2 is considered a more fundamental approach for benchmarking (and
for modelling and prediction), but it requires a sampling and analysis step
(for soil in mixed juice or in cane) that is not normally carried out.
The calculated MSrate and known filter surface area (A) enable the mud solids loading
MSL to be determined. The wash water loading WWL can be also determined from the known
(or estimated) wash water application. The expected pol%MS can be calculated from
Equation (4), and the benchmark pol loss in mud per cent pol in cane given by Equation (5):
MS rate 100
Benchmark pol loss in mud (% pol in cane) = pol % MS ( ) (5)
canerate pol cane
Appendix A shows examples of the calculation of the benchmark pol loss in cake.
There the data values with the exception of the water application are held constant. It is seen
that the increasing water application has resulted in a reduction in the pol loss benchmark for
the filters.
Expected molasses purity, and benchmarks for the pol loss in final molasses
A suitable target true purity relationship (Miller et al., 1998), sometimes modified as
an Expected Apparent Purity, can be used as a target value for final molasses purity. This
relationship is usually a function of the RS/Ash ratio in the molasses, and sometimes allows a
correction for the presence of dextran in the extracted juice. As it is typical for molasses
purity in Australian factories to be up to four units above the target, the benchmark purity
6
This is developed as the precipitate in clarification and is normally in the range 0.2 to 0.3% on cane.
7
The data of Muller (Muller et al., 1982) suggests that, in normal milling, the fraction of the soil passing
to the mixed juice is around 0.36. In recent decades the increasing soil effect is becoming predominant
in determining the mud solids loading and in controlling the filter cake production and the pol losses in
filter cake. The effects of dirt on the process and on the bagasse heating values were discussed in a
previous ASSCT paper (Wright, 2003).
8
The situation with regards to cane quality has changed quite markedly since the 1970s, due to the
uptake of mechanical cane harvesting and, more recently, the return to green cane harvesting in many
areas. As stated (Steindl, 1998) the levels of soil (or dirt) are now relatively high, particularly in the early
weeks of the crushing season and in wet weather harvesting. The maintenance of a reasonable ratio of
bagacillo to mud solids in the filter feed becomes very important as the soil levels in cane fluctuate.
444
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
would be probably be better set at a more realistic value, for example at three units above the
value of the Miller et al. relationship.
In order to have a value which could be used as a benchmark for the sucrose (or pol)
loss in final molasses, the selected relationship for the molasses target purity has to be
combined with an estimate of the expected molasses quantity starting with the cane analysis.
It has not been common practice to record estimates of molasses quantity, though
there have been attempts in Hawaii and Java to compare molasses produced with molasses
calculated (Anon., 1955).
However, recent experience with factory modelling has encouraged improvements in
estimation procedures. It is considered that mass balance calculations can be used to estimate
a benchmark for the final molasses per cent cane, given the analysis of the cane, the
benchmark values for sugar production and for the other losses, and some other operational
factors.
Table 1 lists and gives an example the required factory data for the period, and the
benchmark sucrose (and pol) losses in molasses as a percent of pol in cane. The many
calculations involved in determination of the estimated molasses quantity are not included in
this table, but the details of the calculations are given in Appendix B.
Table 1An example of the calculation of the benchmark pol loss in final
molasses.
Required factory data for the period
Pol in cane. [polcane, %cane] 14.76
Brix in cane. [Bxcane, %cane] [Given directly, or by = 3* polcane 2*CCS] 16.82
RS/ash ratio in molasses, [RS/Ash] 0.88
Pol of the product sugar, [polsugar] 97.90
Purity of product sugar, [Psugar] 98.44
Fibre in cane, [fibrecane, %cane] 13.12
Number of crushing mills, [nmills] 5
Added water % fibre, [WF] 200
Final molasses true purity, [TPmol] 46.68
Final molasses apparent purity, [APmol] 43.7
*** ***
Fr apparent (non-pol) impurity loss, from cane to clarified juice 0.20
Fr impurity loss (due to thermal destruction) in the pan station [FILpans] 0.02
Filter area per unit cane rate, [m2/(t/h)] 0.70
Filter cake, [%cane] 3.82
Mud solids, [%filter cake] 17.58
Wash water application, [% filter cake] 200.0
Results from impurity balance
Estimated flow in molasses @78DS, %cane. [based on true impurity 2.56
flow]
Est. flow in molasses @85 Bx, %cane. [based on apparent impurity 2.35
flow]
Benchmark sucrose loss in molasses 6.05
Benchmark pol loss in molasses 5.63
445
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
The eleventh data item in Table 1 (shown with a triple asterisk), the fraction loss of
apparent non-pol impurity from cane to clarified juice, deserved special mention. This is a
key input value correction to allow for the fact that some of apparent (non-pol) impurities
extracted to the juice disappear in processing. Most factory balance determinations carried
out by the writer suggest a value of 0.20 for this data item. This input value would have to be
standardised when formulating a suitable benchmark.
5670
log (k) = 16.91 pH + log (d c) (8)
T
where T is the absolute temperature, o K ,
kg kg
d is the solution density, , and c is the sucrose concentratiion,
L L
The Vukov expression is considered the best currently available for technical sugar
solutions. The pH of the juice or syrup has to be corrected for temperature elevation above
20oC. A simple correction, subtracting 0.009 for each oC above 20oC, is commonly used.
Spreadsheet programs have been developed for the application of the Vukov formula to
estimate the sucrose loss in clarification, and also in the various stages of the evaporator
bodies, in the syrup storage tanks and in the boiling pans. The sucrose destruction can be
calculated for each individual stage of the process as a function of the prevailing temperature,
pH, juice brix, juice purity and of the residence time in the stage. The total sucrose loss by
inversion and destruction is then the sum of the individual loss estimates.
In applications where sulphitation of syrup is operated, the pH values in the syrup
may be quite low, sometimes around a pH of 5.0, and very significant sucrose losses are then
estimated for the syrup tank and high grade pan stages.
the filter loadings and the value of the total water additions to the process are effectively
imposed by the factory equipment (steam boiler capacity, fuel availability, evaporator/heating
vapour scheme) and the benchmark should respond to these variables as well. The
benchmarks selected should then be more realistic than those currently-used good practice
benchmarks as they take into account the equipment and the imposed operational factors.
The estimation of the expected flow of final molasses and then of a benchmark
sucrose (or pol) loss in final molasses is not trivial, as all the benchmark values for the other
individual losses have to be calculated before the flow to final molasses can be calculated. As
well, the apparent disappearance of impurities during processing has to be allowed for, and
interconversions made between true and apparent purity analyses.
The application of the regressions and benchmark calculations to the factory
chemical control data for the period can be organised in a simple spreadsheet.
The regression expressions suggested here as the basis for the benchmarking of
individual losses in the factory are tentative and can no doubt be revised and improved in the
future. Even in their present form, the process benchmarks outlined in this paper can give
factory managers better bases for discussing factory performance with their technical staff,
and can give factory technologists useful insights into the equipment and operational factors
that control the magnitude of process losses. It is recommended they be routinely calculated
and used in the future technical control of sugar factories.
REFERENCES
Anon. (1955). System of Cane Sugar Factory Control. Int. Soc. Sugar Cane Technol., Special
Committee on Uniformity in Reporting Factory Data, 2nd Edition.
Anon. (1970). Laboratory Manual for Queensland Sugar Mills. Bureau of Sugar Experiment
Stations, Brisbane, 5th Edition, 151.
Miller, K.F., Ingram, G.D. and Murry, J.D. (1998). Exhaustion characteristics of
Australian molasses. Proc. Aust. Soc. Sugar Cane Technol., 20: 506513.
Muller, R.L., Player, M.R. and Wise, M.B. (1982). An examination of input deposition and
effect of dirt in Queensland sugar mills. Proc. Aust. Soc. Sugar Cane Technol., 4: 19.
Steindl, R.J. (1998). DirtIts implications for the clarifier and filter stations. Proc. Aust.
Soc. Sugar Cane Technol., 20: 484490.
Vukov, K. (1965). Kinetic aspects of sucrose hydrolysis. International Sugar J., 67: 172175.
Wright, P.G., Steggles, C.C. and Steindl, R.J. (1997). The balance between capacity and
performance of rotary mud filters, Proc. Aust. Soc. Sugar Cane Technol., 19: 438.
Wright, P.G. (2003). The effect of dirt on bagasse quantity and heating value. Proc. Aust.
Soc. Sugar Cane Technol., 25 (CD-ROM).
Appendix A
Example calculations to determine the benchmark pol loss in filter cake
Example calculations of the benchmark pol loss in filter cake are shown in Table A1.
There the data values with the exception of the water application are held constant. It is seen
that the increasing water application has resulted in a reduction in the pol loss benchmark for
the filters. The results are plotted in Figure A1.
447
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Table A1Examples of the calculation of the benchmark pol loss in filter cake.
0.45 9.0
0.30 7.5
Pol % MS in cake
0.25 7.0
0.20 6.5
0.15 6.0
0.10 5.5
0.05 5.0
0.00 4.5
100 125 150 175 200 225 250
Wash water application, % cake
Fig. A1Plot of filter benchmark examples showing the effect of varying wash
water application on the expected filter performance/
448
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
Appendix B
Calculations to determine the benchmark sucrose and pol in final molasses
This section gives an example of a mass balance calculation procedure that can be
used to estimate the benchmark value of final molasses per cent cane, given the analysis of
the cane, the benchmark values for sugar production and the other losses, and other selected
operational factors.
In this example the input soil content of the can or juice is not known, so that the
mud solids rate MSrate was estimated directly from the recorded filter cake production rate and
the recorded analysis of the approximate mud solids content in the filter cake. The inputs
were from factory data for the period and the calculations were conveniently programmed in
a simple spreadsheet.
The details of a possible set of data input items as used here are shown in Table B1.
Data items are listed from 1 to 16 in the table. A further explanation of the items is as
follows:
1. Data item (1) is the pol % cane [polcane];
2. Data item (2) is the brix % cane. One method of calculation for Australian
factories is to estimate it from the pol and CCS [ = 3* polcane -2*CCS];
3. Data item (3) is RS/ash ratio in final molasses, [RS/Ash, (required for the
expected purity regression)];
4. Data item (4) is the pol of product sugar, [polsugar];
449
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
9
The recorded values for final molasses flow (@85 Bx) and molasses loss for the period were
2.17%cane and 5.08 respectively.
451
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
5. The mud solids loading [MSL] can be calculated directly from the recorded
filter cake % cane, the mud solids analysis in cake [MScake], and the rotary
vacuum filter surface area per unit cane. It may be preferable, however, that
the expected cake%cane can be estimated indirectly if a value for the non-
bagacillo mud solids in extracted juice is available (along with the MScake
value).
6. The wash water loading [WWL] can be calculated from the filter cake %
cane, the wash water application to the cake, and the rotary vacuum filter
surface area per unit cane.
7. The expected pol % MS in cake can then be calculated using Equation (4),
using the calculated values of MSL and WWL.
8. The benchmark pol loss filter cake can then be calculated from Equation (5).
9. The purity of juice in filter cake has to be estimated. Here it is arbitrarily
taken as 5 units less that the purity of cane.
10. The target true purity of final molasses can be estimated from the RS/Ash
ratio using an SRI expression. This is corrected to a reasonable
benchmarking value by adding a constant 3.0, and a benchmark apparent
purity calculated from this value using a true-apparent conversion
expression.
The steps used in the impurity losses calculations of Table B2 are taken on a per
100 cane basis and are as follows:
1. The impurity entering in cane is calculated as the difference between the brix
and pol in cane.
2. The impurity loss to bagasse is as calculated before in a general calculation step.
3. The impurity entering clarification (in the primary and secondary extracted
juices} is calculated as the difference of the two preceding items.
4. The impurity loss to filter cake is calculated as a function of the pol in cane,
the benchmark pol loss in filter cake, and the purity of the cake.
5. The impurity eliminated in clarification is calculated as the apparent impurity
flow multiplied by the elimination factor, the fraction of the apparent
impurity which disappears during juice processing.
6. The impurity flow entering the pan/crystallisation station to be calculated as
the impurity entering clarification less the losses in clarification.
7. An estimate of the impurity eliminated in pans (due to thermal effects in
sugar boiling) is then calculated, along with the calculated impurity flow
passing to the sugar.
8. The apparent impurity flow to final molasses can then be calculated as the
impurity flow entering the pan/crystallisation station less the impurities
eliminated in pan boiling and in the sugar stream.
9. The flow of true impurities to final molasses can then be estimated using the
true-apparent conversion factor [K].
10. The flow in molasses solids can be calculated as a function of the flow of
true impurities and the benchmark true purity of the molasses (the target true
purity plus 3 units).
452
Wright, P.G. Proc. Aust. Soc. Sugar Cane Technol., Vol. 27: 2005
__________________________________________________________________________
a. This value can be converted to a standard dry substance [DS] value (e.g. 78
DS) to enable easy comparison with the reported molasses % cane value for
the factory.
11. An alternative value for the flow in molasses solids can be calculated as a
function of the flow of apparent impurities and the benchmark apparent
purity of the molasses.
a. This value can be converted to a standard brix value (e.g. 85 Bx) to enable
easy comparison with the reported molasses % cane value for the factory.
The steps used in the Benchmark sucrose and pol losses in molasses calculations
of Table B2 are taken on a per 100 cane basis and are as follows:
1. A benchmark sucrose loss in final molasses can then be calculated as a
function of the molasses solids flow (based on true impurities) and the
benchmark true purity.
2. A benchmark pol loss in final molasses can then be calculated as a function
of the molasses solids flow (based on apparent impurities) and the
benchmark apparent purity.
It is seen that the process of estimating the expected flow of final molasses and then
calculating a benchmark sucrose (or pol) loss in final molasses is not trivial. The benchmark
values for the other individual losses have to be known before the remaining flow to final
molasses can be calculated. The apparent disappearance of impurities during processing has
to be allowed for, and interconversions made between true and apparent purity analyses.
453