Back Paper No. T11-2.1, pp.

1-5
Contents
Home PSU-UNS International Conference on Engineering and
Environment - ICEE-2005, Novi Sad 19-21 May, 2005
University of Novi Sad, Faculty of Technical Sciences
Trg D. Obradovića 6, 21000 Novi Sad, Serbia & Montenegro

CLEANER PRODUCTION OF PALM OIL

MILLING BY PROCESS OPTIMIZATION
Suteera Prasertsan*, Charun Bunyakan, Juntima Chungsiriporn
Prince of Songkla University, Department of Mechanical Engineering,
Chemical Engineering, Faculty of Engineering,
Hat-Yai, Songkhla 90112 THAILAND. E-mail : Suteera@trf.or.th

Abstract: Palm oil milling is one of the main industries
in the south of Thailand. Milling process consumes large
amount of water, which consequently results in
wastewater treatment burden. The wastewater is high in
both BOD (>50,000 mg/l) and COD (>80,000 mg/l). Oil
generally presents in the wastewater. This does not only
hinder the treatment, but also causes the loss in
production. This study is an attempt to reduce the milling
wastewater by cleaner production concept. The milling
process was divided into 4 sub-systems and were studied
aiming to minimizing the water consumption.
Mathematical models of the processes were construction
based on the experiment results. The process
optimization was carried out by employing the mass
integration technique and Matlab package. It was found
that the milling process has a potential to reduce
freshwater consumption by 65%. Consequently, the
wastewater generation was reduced by 67%.
Key Words: Palm oil /Mass integration /Cleaner
production / process optimization
various stream are taken into consideration [6]. The
approach consists of water allocation alternatives
(recycle and reuse) derived from a graphical
representation of the flows.
This paper focuses on the reduction of freshwater
usage and wastewater effluent in palm oil milling
process within the conceptual framework of mass
integration and process optimization. The site of study is
the Thaksin Palm Oil Mill in the south of Thailand. The
normal milling capacity is 40 t FFB/h. The strategies of
recycle/reuse, mixing, and interception were used to
allocate source streams to sinks. Sources are the streams
that have potential to be recycled, while sinks are the
water-consuming units that can accept the source
streams. The concentration of non process element
(NPE) in the source is the allocation constraint.
Constrained mathematical model for water optimization
was formulated and solved by the Matlab optimization
toolbox.
2. MASS FLOW OF PALM OIL PROCESS

1. INTRODUCTION Palm oil milling is the process that extracts crude

Wet process of palm oil milling consumes a large
amount of water. Consequently, a larger amount of palm
oil mill effluent (POME) is generated with high organic
load. It was reported that for each ton of fresh fruit bunch
(FFB), 0.87 ton of POME is discharged with BOD as
high as 50,000 mg/l [1]. Many studies suggested various
water management approaches [2], [3], [4], and [5]), but
very rare was systematically. In recent years, the scarce
of freshwater supply and the increasing wastewater
treatment cost due to stringent environmental standard
became the major problem of the mills.
The appropriate strategy for water management to
reduce aqueous effluent and freshwater demand should
be implemented to make the process cleaner and
environment friendly. Mass integration technique is one
of the water utilization management tools, which is a
systematic approach for the management of global mass
flow within the system. It optimizes the allocation of
mass (water) streams. With this technique, the optimal
allocation of mixing and recycling scheme with
constraints of material balance and relative costs of
palm oil from fresh fruit bunches (FFB). The standard
wet process consists of bunch sterilization, fruit
stripping, digestion, screw pressing for liquid extraction
and centrifugation for oil separation. The FFB is
sterilized by steam. The fruits are stripped off by a rotary
thresher and subsequently mashed in a digester. A twin-
screw press is used to separate liquid, which is a mixture
of water, oil and solid. A vibrating screen removes large
size solid particles. Mixing water is added to the digester,
the screw press and the screening unit to improve
extraction efficiency and flow ability of the processing
stream. The extracted product, called palm oil mixture, is
the mixture of palm oil, water and fine solid particles.
The palm oil mixture is continuously fed to the
settling tank. Palm oil the lightest phase overflows from
the settling tank. Bottom sludge from the settling tank
consisting of water, solid and palm oil residue of 7-10%
is then passed through a series of a decanter and a
separator for oil recovery. Water is added to these units
to blend and balance the phase for efficient oil recovery.

1
The schematic diagram of palm oil milling process, with (w13). It is noted that two sterilizer influx (w17) and
designated mass flows, is shown in Fig.1. turbine cooling water (w18) are not regarded as sinks
Cooling water Condensate
because they require freshwater only. The source streams
M ixing water
include sterilizer condensate (w3), separator outlet (w14),
w16 w3 w4
cooling water (w16), and freshwater (if needed). The NPE
Superheated Steam
Steam w17 Turbine w Sterilizer Stripping, Digestion, contents (z values) in the source streams are obtained
2
w18 FFB
Screw press, and screening from field measurement.
w1
Cooling Sources Sinks
water inlet Empty bunch, z = 0.093 Mixing water
Palm oil mixture w5 Nut, and Fibres Separator
Crude w6 w = 24 t/h outlet 0 < z < 0.020 Processing
Palm Oilw15 Settling tank 14
w 4 = 14 t/h
Palm oil recovery
w10 w11 z = 0.01 Sterilizer
Blending water
Sludge w7 w = 6 t/h Condensate
3 0 < z < 0.12
Decanter Separator Wastewater Decanter
w12 w14 z=0 Cooling w 8 = 2.5 t/h
w8
w9 w = 4 t/h water
w13 16 Phase balancing water
Blending water z=0 0 < z < 0.093
Cake Balancing water Separator
Freshwater w = 7.5 t/h
13
Fig.1. Mass flow in palm oil mill process.
Fig. 3. Quality and quantity of sources and sinks
3. WATER MANAGEMENT (symbols of stream flows “wn“ are refered to Fig.1).
Mass integration technique is used to design the Source-sink diagram that shows the mass flow
mass flow network to minimize wastewater discharge rate and solid content (NPE) is presented in Fig.4(a). In
and freshwater usage. The strategy includes mixing, the complex process, where many potential sources are
recycling, and interception. Recycle is employed when available, the method called “smallest possible lever arm
the source properties and compositions meet the for a sink” is used as a tool to determine the most
receiving conditions of the sink. The mixing of two or suitable source [7]. For a simple process, the relative
more sources is carried out to obtain sufficient flow rate position of source points and sink lines easily gives the
and acceptable compositions for the sink. Interception is recycle potential. For instance, the NPE contents of the
the separation of NPE from the targeted source. NPE in condensate (6t/h) and cooling water (4 t/h) fall in the
this study are fine particles, fiber and other suspended acceptable range of the mixing water (14 t/h), thus
solids. Internal and external sources are the aqueous allows direct recycle of these two sources. However, the
streams within the process and freshwater respectively. total quantity is insufficient and, as a result, 4 ton/h of
The graphical technique gives a guideline for stream mixing water is still needed as shown in Fig.4(b).
mixing to meet the requirements of different sinks.

.
(a) 30
The procedure starts with the decision for the
worthiness of conducting water recycle scheme. Overall 25
Separator
Mass flow (ton/h)

water balance of the palm oil mill under this study is 20

effluent
shown in Fig.2. Since the quantity of wastewater (34 t/h)

..
15
is less than that of freshwater usage (36 t/h), the ideal Mixing water
targeted of net wastewater generated would be none (all 10
Phase balancing
wastewater substitutes for sink streams), and 94% of 5
Condensate water
Cooling water Blending
freshwater could be eventually reduced. It is, therefore, water
0
concluded that the existing process is attractive enough 0 0.03 0.06 0.09 0.12 0.15
to warrant conducting the mass integration analysis. Solid content
sink source
w4 w16
14 ton/h

.
4 ton/h (b) 30
w8
2.5 ton/h
w13 Palm oil w3 6 ton/h
25
Separator
7.5 ton/h
flow (ton/

effluent
w17 milling process 20
8 ton/h w14
h)

15
4 ton/h w18 24 ton/h
Mass

10
Phase balancing
water
Total FW usage = 36 ton/h POM E = 34 ton/h 5 Mixing water Blending
water
0
Fig.2. Overall water balance for palm oil mill process 0 0.03 0.06 0.09 0.12 0.15
Solid content
(symbols of stream flows “wn“ are refered to Fig.1 ). sink source

Fig.4. Source-sink diagram of palm oil milling

Sources and sinks of the palm oil mill process
have been identified and given in Fig.3. The palm oil
milling process has 3 sink streams with total amount of
24 ton/h namely, processing mixing water (w4), decanter
blending water (w8) and separator phase balancing water
(a) Initial condition (b) After first allocation.
Similarly, the source from the separator (24 t/h) is
directly recycled as phase balancing water (in separator,
7.5 t/h) and blending water (in decanter, 2.5 t/h). Finally,
it leaves only 14 t/h of the effluent and only 4 t/h of
freshwater is required for mixing water. The reduction of
wastewater and freshwater is substantially significant in
comparison to the corresponding current figure of 34 and
36 t/h, respectively.
From the NPE view point, it seems that a certain
amount of the final effluent (from separator) can be
mixed with the condensate and cooling water and supply
as mixing water because the mixing water can tolerate
NPE up to 0.03. In such ideal practice, no freshwater is
required for sinks and only 10 t/h of POME is generated.
If this is true, 10 t/h of wastewater actually comes from
the moisture content in the fruits and steam absorbed by
the sterilized fruits. However, it was found that, unless
the interception process is applied, this practice will
consequently increase the NPE in the downstream
processes. As a result, a certain quantity of freshwater is
needed to adjust the phase of the material in the decanter
and separator. Consequently, the final quantity of
wastewater was not significantly reduced. Therefore,
mass integration gives only a simple and quick guideline
for water management. The real solution must take into
account of the consequence of the NPE and its
acceptable range in the sinks. This will again be
discussed in more details in the next section. It is also
obvious that there are possibilities to combine the
sources to meet the requirement of the sinks. The
optimum solution will be described later.
3.1 Source-sink allocation alternatives
To simplify the optimization problem, the palm
oil milling process is split into 4 sub-systems namely,
processing unit, settling tank, decanter, and centrifugal
separator. The processing unit comprises of the FFB
sterilizer, rotary thresher, digester, screw press and
vibrating screen.
The water usage in conventional palm oil milling
process was shown previously in Figs 1 and 2. In such
conventional process, 24 t/h of freshwater is used as
(external) sinks for mixing, blending and balancing
water. Source-sink allocation strategies could possibly be
implemented in two schemes as shown in Fig.5.
Process 1 FW
FFB Settling
Processing Decanter Separator
feed tank
a
FW
Sterilizer FW
condensate Cooling
water
Process 2
FW
FFB Settling
Processing Decanter Separator
feed tank
a options)
Sterilizer FW
condensate Cooling water
Fig.5. Alternative schemes of source-sink allocation for
palm oil milling process.
In these schemes the freshwater at the processing,
decanter and the separator was partly replaced by the
effluent (source) streams. The optimum operating
condition is solved from mathematical representatives of
these two schemes. It was aimed to minimize the
summation of the freshwater added in whole milling
process.
3.2 Optimization models
The optimization models consist of objective
functions and constraints. The constraints are formed
from the source-sink material balance, unit operation
material balance, technical specification, process
modeling and environmental restriction.
The processes of the four milling sub-systems are
optimized with respect to their constraints as given in
Tables 1. The problem statements aim for minimizing
freshwater consumption of the sinks. The settling model
and separator model, which has been reported previously
[7] and operating conditions/constraints are used for the
process optimization. It was found that, under the force
of gravity, the residue oil in the settling sludge was 7-
10%. Therefore, the decanter feed is controlled at 7% oil
(by adding blending water) and this becomes the inlet
constraint of the decanter. In addition, the solid
concentration in decanter discharge (x27 ≤ 0.115) is the
outlet constraint.
The design of the centrifugal separator has the
optimum feed flow rate of 24 ton/h and the upper limit of
solid concentration of 9.3% (by weight) or x39 ≤ 0.093.
Since the solid concentration in the decanter outlet
(separator inlet) is 9-11%, balancing water is needed for
the separator. Together with the recycled streams,
freshwater is used to adjust the oil and solid
concentration of the sources to satisfy the limitation of
the sinks.
The optimization problem is solved by the Matlab
program. The optimization toolbox consists of
commands for option parameters, input and output
arguments, and functions that perform minimization. The
algorithm used to find the minimum of a constrained
nonlinear multivariable function is fmincon. The
objective function and constraints are written in function
form called M-file as shown below.
Objective function min f ( x )
x
subject to
c( x) ≤ o, ceq( x) = 0
Separator
A.x ≤ 0, Aeq.x = beq, lb ≤ x ≤ ub
Effluent
where x, b, beq, lb, and ub are vectors, A and Aeq are
Wastewater
recycle matrices, c(x) and ceq(x) are functions that return
vectors, and f(x) is a function that returns a scalar. f(x),
c(x), and ceq(x) can be nonlinear functions. Syntax
command line is written in the form of
Separator
[x]=fmincon(@myfun,x0,A,b,Aeq,beq,lb,ub,@nonlcon,
Effluent
Wastewater
recycle Constrained optimization routine that consists of
starting guess, syntax command, options, lower bound
and upper bound is invoked to solve the problem. The
results of optimization of the whole system are given as
total freshwater demand for the 2 schemes (in Fig.5).
 water is added at the processing. It is, therefore,
Table 1. Optimization problems and constraints of palm suggested at this moment that the allocation of
oil milling process. wastewater recycle should adopt scheme 2. However, it
Objective function Min f(x) = x4 + x23 + x29 seems unjustifiable to unlimitedly recycle the wastewater
to the processing because this will add more solid in the
Items Subject to Constraints
upstream process and adverse effects could be
Material balance anticipated. The inclusion of decanter and separator
x1(0.86- x5(1- x2- x3)+ x11) = x14
constraint conditions could help deciding for the quantity
Processing unit

x1(x3-0.135+ x11) = x10

x1(x2-0.005 + x19( x11-0.25)/0.09) = x9 of water allocation. It gives the maximum allowable
x1( x11-0.25)(0.093- x20)/0.093 = x4 quantity of recycling water adding in the processing unit
x1(1- x2- x3)(1- x5) + x1x20( x11-0.25)/0.9 = x6 that will not violate the constraint conditions. Higher
Boundary limit milling capacity can receive less (specific) wastewater
34 ≤ x1 ≤ 50 recycle. Table 2 gives comparison between the
Material balance conventional practice and scheme 2. While the specific
x9 /( x9+ x10+ x6) = x12 processing mixing water decreases with the milling
x10 /( x9+ x10+ x6)- x13 (x14( x12/0.932+ x13 capacity for scheme 2, the specific freshwater
+( x35)/1.929)+( x25+ x31)/0.932) x16 = x15 consumption and POME are both increased.
x8 x12+ x13 = x8
1- x12- x13 = x35 Table 2. Optimized freshwater consumption and POME
x14-( x12 x14- x14x7(1- x12) /(1- x7)) = x21 usage in palm oil mill process.
x14( x12/0.932+ x13+ x35/1.929) + x16(x25+
S Settling tank

x31)/0.932 = x15 Milling Processing Freshwater POME

Settling model Process capacity unit water consumption Generation
(91.28-1.486 x18+2.851 x18 x12 (t FFB/h) (t/t FFB) (t/t FFB) (t/h) (t/t FFB) (t/h)
-367.43 x12 x8+95.98 x82 ) = x17
Conventional

(1/100)(0.00407+0.04949 x16-0.000165 x172 )-1 = x7

practice

Boundary limit 40 0.35 0.90 36 0.870 34.0

x 13 > 0.55
x 11 > 0.25
90 ≤ x 18 ≤ 97 36 0.45 0.304 10.9 0.267 9.6
0.07 ≤ x7 ≤ 0.15
(Scheme 2)
Optimum

40 0.40 0.317 12.7 0.285 11.4

process

Material balance
x23+ x21+ x22- x25- x24 x6 = x26 45 0.34 0.333 15.0 0.306 13.8
x7 x21+s x36 x22-0.011 x26 = x25
x24 x6 + x27 x26 - x37 x22 = x6 50 0.29 0.348 17.4 0.324 16.2
Decanter

(x7 x21+ x36 x22)/( x21+ x22+ x23) = 0.07

Boundary limit It is obvious that when compared with the
x27 ≤ 0.115 conventional practice, scheme 2 substantially reduces the
x26 ≤ 23 freshwater consumption at the order of approximately
Material balance 65% for the same milling capacity. Consequently, the
x26(0.093 x30-(1- x30)( x27-0.093))/(x37(1- x30)) = x32 effluent discharged from the mill has been drastically
(x26+ x32+ x29-0.011x26- x36 x32)/(1- x36) = x33 reduced from 0.87 t/t FFB to 0.285 t/t FFB (67%
(x27 x26)/( x33- x32) = x37 reduction for 40 t/h capacity). This would save
0.011x26+ x36( x32- x33) = x31
freshwater resource and reduces the wastewater
( x26 x30-(1- x30) x32)/(1- x30) = x29
treatment cost. In addition, less wastewater means less
Separator

x32+ x29+ x26 = 24

x33 - x32 - x22 - x1(x20/0.093)( x11-0.25)/0.09 = x28
(x4+ x23+ x29)/x1 = x38
(x27x26)+ x20 x32)/( x26+ x29+ x32) = x39
Separator model
(2.1875/100)ln(4.318-0.0655 x34 + 0.000392 x342
-0.0156 x34 x30+1.785 x302 ) = x36
Boundary limit
x39 ≤ 0.093
3.3 Optimization results
The higher productivity naturally requires more
freshwater and consequently generates larger amount
POME. For the case of scheme 1, it increases in both
freshwater consumption and POME generation as more
water is added in the processing. Scheme 2 contrarily
requires less freshwater and generate less POME as more
oil loss to the wastewater treatment system. It is
estimated that, with this optimum scheme, 4 kg of oil/t
FFB, that would, otherwise, have been lost, is recovered.
4. CONCLUSION
This work has presented an attempt to reduce
freshwater demand and wastewater generation for palm
oil mills by mass integration of source and sink analysis.
Optimization was carried out mathematically. In
conclusion, it was found that,
1. The recycle of sterilizer condensate and cooling
water to processing mixing water and the channeling of
separator effluent to the processing, the decanter
(blending water) and the separator (balancing water) are
recommended. Consequently, freshwater consumption
and wastewater generation were reduced by 65% and
67%, respectively.
 2. Increasing the production rate required higher
specific freshwater in the process and generates higher
specific POME, even though the process is optimized.
3. There is a limit for water recycle into the
processing unit, which is the main controlling parameter
for the optimum operation in the aspect of freshwater
consumption and POME generation.
5. ACKNOWLEDGEMENT
The authors greatly appreciate the financial
support of the Graduate School of Energy and
Environment, Thailand. We also would like to thank
Taksin Palm Oil Company Ltd. for their support of
laboratory set up at their site.
6. REEERENCE
[1] Prasertsan, S., and Prasertsan, P. (1996), Biomass
residues from palm oil mills in Thailand: an
overview on quantity and potential usage, Biomass
and Bioenergy, vol.11, No.5, pp.387-395.
[2] Kittikun, A. H., Prasertsan, P., Srisuwan, G., and
Krause, A. (2000), Environmental Management for
Palm oil mill, Internet conference on material flow
analysis of Integrated Bio-Systems.
[3] Department of Industrial works and Deutsche
Gesellschaft fur Technische Zusammenarbeit
(GTZ) GmbH, (1997), Environmental Management
Guideline for the palm oil industry, Thailand, PN
2000.2266.5-001.00.
[4] Leong, S. T., Muttamara1, S. and Laortanakul, P.
(2002), Sustainable Water Conservation and
Wastewater Reuse in a Palm Oil Mill: A Case
Study in Southern Thailand, Water Quality
Research Journal of Canada, vol.37, No.4, pp.711-
728
[5] Barrantes, J. O. (2001), Systems of innovation and
cleaner technologies in the palm oil sector, Costa
Rica, DRUID’s Nelson and Winter Conference.
[6] Parthasarathy, G. (2001), Minimization of cost of
recycling in chemical processes, Chemical
Engineering Journal, vol.81, pp.137-151.
[7] Chungsiriporn J., Prasertsan S. and Bunyakan C.
Modeling of Palm Oil Separation for Horizontal
Settling Tank: A Key Answer toward Cleaner
Technology, The PSU-UNS International
Conference 2003: Energy and the Environment, 11-
12 Dec. 03.