Egyptian Journal of Petroleum (2015) 24, 383–396

H O S T E D BY
Egyptian Petroleum Research Institute

Egyptian Journal of Petroleum

www.elsevier.com/locate/egyjp
www.sciencedirect.com

FULL LENGTH ARTICLE

Crude biodiesel refining using membrane

ultra-filtration process: An environmentally
benign process
I.M. Atadashi a,*, M.K. Aroua b, A.R. Abdul Aziz b, N.M.N. Sulaiman b

a
Adamawa State University, Mubi, P.M.B 25, Mubi, Nigeria
b
Chemical Engineering Department, Faculty of Engineering, University Malaya, 50603 Kuala Lumpur, Malaysia

Received 17 June 2014; revised 12 July 2014; accepted 20 July 2014

Available online 6 November 2015

KEYWORDS Abstract Ceramic membrane separation system was developed to simultaneously remove free
Biodiesel; glycerol and soap from crude biodiesel. Crude biodiesel produced was ultra-filtered by multi-
Ceramic membrane; channel tubular membrane of the pore size of 0.05 lm. The effects of process parameters: trans-
Separation; membrane pressure (TMP, bar), temperature (°C) and flow rate (L/min) on the membrane system
Purification; were evaluated. The process parameters were then optimized using Central Composite Design
Optimization (CCD) coupled with Response Surface Methodology (RSM). The best retention coefficients
(%R) for free glycerol and soap were 97.5% and 96.6% respectively. Further, the physical proper-
ties measured were comparable to those obtained in ASTMD6751-03 and EN14214 standards.
Ó 2015 The Authors. Production and hosting by Elsevier B.V. on behalf of Egyptian Petroleum Research
Institute. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction greenhouse gases such as CO2 in the atmosphere can be con-

siderably reduced by substituting petro-diesel with biodiesel
Global increasing demands for energy, declining fossil fuel [5–7].
reserves, environmental concerns, and price hike have resulted The most commonly adopted technique to produce biodie-
in a growing interest in the development of alternative renew- sel fuel is transesterification [8,9]. Other techniques used in
able energy source [1–4]. Presently the energy sources being producing biodiesel include: direct/oil blends, microemulsion,
explored include water, wind, geothermal, and biofuels. Biofu- and pyrolysis. Transesterification reaction is catalyzed by
els are generally known to present numerous advantages over either acid, base or enzyme catalysts. Transesterification reac-
fossil fuels such as sustainability, lower gaseous emissions, tion catalyzed by alkaline catalysts such as NaOH, KOH,
social structure and agriculture development, regional develop- CH3ONa and CH3OK, provides higher conversion and faster
ment, and fuel security supply. Besides accumulation of reaction rates [10,11]. However the process requires raw mate-
rials with low water content (0.6 wt%) and less free fatty acids
* Corresponding author. Tel.: +234 08123214691. content (0.5–3.0 wt%). The presence of free fatty acids and
E-mail address: [email protected] (I.M. Atadashi).
water could lead to soap formation. Soap formation could
Peer review under responsibility of Egyptian Petroleum Research
deactivate the catalyst, lower its catalytic performance and ren-
Institute. der biodiesel separation and purification difficult [12–14].
http://dx.doi.org/10.1016/j.ejpe.2015.10.001
1110-0621 Ó 2015 The Authors. Production and hosting by Elsevier B.V. on behalf of Egyptian Petroleum Research Institute.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
384 I.M. Atadashi et al.

After transesterification reaction is completed, the biodiesel 2. Materials and methods

produced contains various impurities, such as soap, catalyst,
free glycerol, and alcohol etc that must be removed for the 2.1. Materials
resultant biodiesel product to meet the strict international
standards (ASTMD6751 and EN14214) specifications. Fur-
Palm oil used for the production of crude biodiesel was
ther, free glycerol removal from biodiesel is important due to
obtained from a commercial local store. Anhydrous methanol
its negative effects on diesel engines and on the quality of bio-
(99.85% purity) and potassium hydroxide (KOH, reagent
diesel fuel. These negative effects include: higher aldehydes and
grade) used were purchased from MERCK. All other chemical
acrolein emissions, fuel settling problems, tank bottom depos-
reagents employed to wash the membrane and analyze the free
its, decantation, injector fouling, storage problem, and engine
glycerol and soap contents were purchased from Global
durability problems [15,16]. Furthermore the amount of soap
Science Resources Sdn, Bhd, Malaysia.
in biodiesel is another critical issue in biodiesel production.
Higher amount of soap in biodiesel could damage injectors,
2.2. Methods
pose corrosion problem in diesel engines, plugging of filters
and weakening of engines [17].
2.2.1. Production of biodiesel
Conventionally, biodiesel is purified using wet and dry
washing processes. Although wet washing process provides The required crude biodiesel samples were prepared using a 5
high-quality biodiesel with physicochemical properties meeting liter batch reactor. The reactor was operated using methanol
the values prescribed by ASTMD6751 and EN14214 standards to oil molar ratio of 6:1, catalyst concentration of 1 wt%
specifications, the process involves large amount of water and (KOH) based on vegetable oil, reaction time of 1 h and the
high energy usage. Besides it can result in an increased cost and operating temperature was maintained at 60 °C. The required
production time, loss of biodiesel yield, and disposal of huge quantity of KOH was thoroughly mixed in the required quan-
amount of wastewater [18,19]. Wastewater disposal is the main tity of alcohol (methanol). The mixture of methanol and KOH
disadvantage of wet washing process. About 20–120 liters of was then charged into the reactor together with palm oil and
wastewater is generated per 100 liters of biodiesel [20]. The dif- heated to 60 °C using a water bath. A stirrer with a capacity
ficulties generally encountered with wet washing process have of 645 rpm was used to improve the mixing of the reactor con-
resulted in the development of dry washing process such as tent. The selection of the experimental conditions for the pro-
ion exchange resins (amberlite or purolite), silicates (magnesol duction of the biodiesel samples was based on the reviewed
or trisyl), cellulosics, activated carbon, activated clay, and acti- literature [22,25].
vated fiber etc for the purification of crude biodiesel. Like wet After the reaction was completed, the transesterified pro-
washing process, dry washing technique provides high-quality duct consisting of biodiesel, glycerol, and other by-products
biodiesel with very good physicochemical properties; however was allowed to settle overnight and then decanted. After
the inability to regenerate the spent adsorbents has discour- removing the bottom glycerol-rich phase, the upper
aged its use. Besides the understanding of the chemistry of biodiesel-rich phase was then transferred to the feed tank for
the adsorbents is still skeletal [21]. Thus the problems associ- the purification process. In this work, several runs were per-
ated with the conventional wet and dry washing techniques formed to produce adequate biodiesel samples for the sched-
have resulted in the current study on the application of mem- uled experiments. The produced biodiesel samples were put
brane technology for the purification of crude biodiesel. So far in appropriate vessels and then properly stored in a cold room.
very few studies have been conducted using membrane tech-
nology for the purification of crude biodiesel [22,23]. Applica- 2.2.2. Biodiesel membrane separation and purification process
tion of membranes to purify crude biodiesel has provided 2.2.2.1. Ceramic membranes. A multi-channel tubular-type
promising results with high-quality biodiesel achieved. In addi- Al2O3/TiO2 ceramic membrane was used for the experiments.
tion membrane biodiesel purification technique does not The total filtration area is 0.031 m2. The membrane with the
require water, hence no wastewater disposal is required [24]. pore size of 0.05 lm was purchased from Jiangsu Jiuwu Hitech
Absence of wastewater generation indicates environmental CO., China and the module was fabricated in-house.
friendliness of the membrane biodiesel separation process.
It is worth mentioning that in the previous published liter- 2.2.2.2. Determination of initial permeate fluxes. The prelimi-
ature no research has been carried out to simultaneously nary permeate fluxes were obtained with distilled water (Clean
remove soap and free glycerol from crude biodiesel using membrane water flux). The preliminary values of the permeate
multi-channel tubular membrane with pore size of 0.05 lm, fluxes were obtained using distilled water at 50 °C, transmem-
and optimize the effects of the main process parameters such brane pressure of 1, 2 and 3 bars and flow rate of 150 L/min. In
as transmembrane pressure, flow rate and temperature. There- order to monitor the effectiveness and the performance of the
fore, the goals of this investigation are: to employ membrane membrane cleaning process, the preliminary conditions
ultra-filtration process to simultaneously remove free glycerol obtained were used as a reference points.
and soap from crude biodiesel in the presence of acidified
water; to conduct rigorous optimization on the main process 2.3. Ultra-filtration process for the refining of biodiesel
operating parameters such as transmembrane pressure, tem-
perature and flow rate in order to determine the optimum 2.3.1. Operating parameters
operating conditions of the membrane system; and to deter- The key operating parameters evaluated for the membrane
mine the physical properties of the biodiesel produced at the separation method are transmembrane pressure, temperature
best operating conditions. and flow rate. The values of the operating parameters were
Crude biodiesel refining using membrane ultra-filtration process 385

varied as follows: transmembrane pressure (1–3 bar), tempera-

Table 1 Operating conditions for biodiesel membrane sepa-
ture (30–50 °C), and flow rate (60–150 L/min). Additionally,
ration process.
the content of free glycerol was expressed as percentage, con-
tent of soap as part per million (ppm), and the unit of the per- Run TMP (bar) Temperature (°C) Flow rate (L/min)
meate flux as kg/m2 hr. 1 1(
1) 30(
1) 60(
1)
2 3(+1) 30(
1) 60(
1)
2.3.2. Separation and purification method 3 1(
1) 50(+1) 60(
1)
4 3(+1) 50(+1) 60(
1)
Fig. 1 shows schematic diagram of ceramic membrane separa-
5 1(
1) 30(
1) 150(+1)
tion system for the purification of crude biodiesel. The set up 6 3(+1) 30(
1) 150(+1)
consists of a membrane module, feed and product tanks, water 7 1(
1) 50(+1) 150(+1)
bath, circulating pump, digital weighing balance, and a stirrer 8 3(+1) 50(+1) 150(+1)
etc. In addition pump tubing (Chem-Durance chemical resis- 9 1(
1) 40(0) 105(0)
tant) with a size of 16 (ID = 44 mm, OD = 2.36 mm) was 10 3(+1) 40(0) 105(0)
provided for the experiment. The pressure and temperature 11 2(0) 30(
1) 105(0)
of the membrane system were monitored via pressure gauges 12 2(0) 50(+1) 105(0)
and temperature indicator. The crude biodiesel was charged 13 2(0) 40(0) 60(
1)
into a 5 liter feed tank and a pump was used to circulate the 14 2(0) 40(0) 150(+1)
15 2(0) 40(0) 105(0)
crude biodiesel via the membrane tube at the conditions shown
16 2(0) 40(0) 105(0)
in Table 1. The pressures at the inlet and outlet were achieved 17 2(0) 40(0) 105(0)
by adjusting the valves at the ends of the membrane tube. The 18 2(0) 40(0) 105(0)
temperature of the system was monitored by means of a water 19 2(0) 40(0) 105(0)
bath. In this work, a membrane with pore size of 0.05 lm was 20 2(0) 40(0) 105(0)
used to carry out 20 experimental runs. In addition, retention
of the impurities in the membrane system was enhanced by
adding acidified water to the biodiesel samples. The membrane
module operation was based on the recycling of the biodiesel
concentrate. Further, a digital balance was used to automati- using design of experiment software Version 8.0.0 (Stat-
cally record the mass permeate fluxes every 10 min throughout Ease Inc., USA) and also the effects of the process parame-
the filtration process. The separation time for each experimen- ters; transmembrane pressure, temperature, and flow rate
tal run was 1 h. The initial samples were denoted as original for the ultrafiltration process were optimized using the soft-
biodiesel samples, and the permeates were taken after the com- ware. Hence, Response Surface Method (RSM) coupled with
pletion of the experimental runs and examined. The use of Central Composite Design (CCD) was chosen in this work.
membranes in the separation of the contaminants from biodie- The main responses are the soap and free glycerol contents.
sel minimizes quite a number of steps that are essential in bio- Further, six replicated center points were selected and per-
diesel treatment via conventional techniques, as well it formed in a randomized order, so as to provide a true mea-
consumes a lesser amount of water [26]. sure of error due to natural variations. The selection of the
number of replicates was meant to proffer a broad region
2.4. Experimental design and optimization where the standard error of prediction remains considerably
steady. Table 2 presents coded and actual levels of the pro-
There are numerous techniques available for the optimization cess parameters. The process parameters are flow rate (60–
of a process. In this study, the experiments were designed 150 L/min), transmembrane pressure (1–3 bar) and tempera-

Crude biodiesel

Retentate

Valve

P1 P2
Heating equipment Pump Membrane
Valve Valve

Permeate (Biodiesel)
Digital balance Computer

P1= Inlet Pressure gauge, P2= Outlet Pressure gauge

Figure 1 Schematic diagram of biodiesel ceramic membrane separation unit.

386 I.M. Atadashi et al.

2.6. Biodiesel characterization

Table 2 Process parameters levels in actual and coded forms.
Parameters Unit Low level High level 2.6.1. Determination of free glycerol content in biodiesel
TMP bar 1(
1) 3(+1) The amounts of free glycerol in both the original and the per-
Temp °C 30(
1) 50(+) meate samples were determined using modified version of the
Flow rate L/min 60(
1) 150(+)
AOCS technique for the analysis of free glycerol in fats and
oils (Ca 14-56). The titration technique used is based on the
glycerol reaction in aqueous medium with excess sodium peri-
odate to form iodic acid, formic acid, and formaldehyde, and
ture (30–50 °C). Whereas
1 (low), 0 (medium), +1 (high), then addition of potassium iodate to react with the iodic acid

a and +a, are designated as coded values. Alpha (a) is and sodium periodate formed. The periodate titration tech-
the distance from the center point which might either be nique for the determination of free glycerol affords low cost
inside or outside the range, with the high value of 2k/4 (where compared to gas chromatography. Besides, the technique is
k is the number of factors) [8]. In comparison to the tradi- easy and uncomplicated, rapid and satisfactorily dependable.
tional method in which one parameter is considered at a The technique which involved periodate as an oxidant reagent
given time, this technique can determine the interaction for free glycerol determination was thoroughly assessed, and it
between the parametric effects, as well as giving good estima- was concluded that analysis of glycerol by this technique is
tions of the errors. In addition, the cost and time of experi- very promising in terms of precision and accuracy [15]. Fur-
ments are decreased since the overall number of trials is thermore, Naviglio et al. [29] used periodate technique to
reduced [27]. The levels of each parameter in this study were determine the contents of glycerides in oils and esterified glyc-
selected based on the data available in the published litera- erol after transesterification, and obtained very good results.
ture [15,24,26]. The authors revealed that the technique is easily reproducible
In this study two dependent parameters are to be optimized with precision.
simultaneously. Thus second-order polynomial model
presented in Eq. (1) is used to explain the relationship between 2.6.2. Determination of soap content in biodiesel
the dependent parameters. The amount of soap in biodiesel was determined by stirring the
liquid in the beaker containing isopropyl alcohol and 12 drops
X ¼ b0 þ b1 Y1 þ b 2 Y2 þ b3 Y3 þ b12 Y1 Y2 þ b13 Y1 Y3 of bromophenol blue indicator solution. After which the solu-
þ b23 Y2 Y3þ b11 Y21 þ b22 Y22þ b33 Y23 ð1Þ tion turned dark blue. At that point, the weight of beaker and
the solution was recorded. Certain quantity of biodiesel was
where: X = dependent variable; Y1, Y2 and Y3 = independent then added. The solution was then stirred continuously while
variables; b0 = intercept; b1, b2 and b3 = linear coefficients; slowing adding HCl solutions until the stirred solution turned
b12, b13 and b23 = interaction coefficients; b11, b22 and from bluish color to yellowish color. The final weight of the
b33 = quadratic coefficients. beaker and solution was recorded. And the difference between
the weights of the beakers and the solutions was then deter-
2.5. Cleaning of the membrane module mined. The difference indicates the amount of HCl added.

The process of membrane cleaning entails disruption of the

Table 3 Concentrations of free glycerol and soap in permeate
membrane separation process. Due to the extensive nature of
(final biodiesel).
cleaning required, quite substantial time losses may occur.
Therefore, a typical cleaning method would necessitate flush- Run Free glycerol (wt) Soap (ppm)
ing with filtered water at 35–50 °C to displace residual reten- 1 0.0378 71.34
tate. Further back-flushing or recirculation with cleaning 2 0.0417 87.95
agent, probably at higher temperatures and rinsing with water 3 0.0401 76.09
could aid complete module cleaning. Nevertheless, this is 4 0.0423 82.57
rarely totally effective, thus chemical cleaning is eventually 5 0.0347 96.83
6 0.0452 97.63
required [28]. In this work, the membrane module after each
7 0.0205 96.57
run was carefully cleaned to restore, protect, or preserve the
8 0.0256 56.60
membrane performance in relation to its permeability. The 9 0.0302 61.52
washing process of the membrane was done using water and 10 0.0291 89.76
detergent until the biodiesel was completely removed. After- 11 0.0281 62.79
ward 1% NaOH solution at 70 °C was circulated through 12 0.0137 57.67
the module for 45 min. The module was then thoroughly 13 0.0357 78.89
cleaned with water and rinsed with warm distilled water [15]. 14 0.0253 71.45
Also after each cleaning process, the permeate fluxes were 15 0.0183 59.32
determined by means of distilled water so as to determine 16 0.0117 50.89
17 0.0122 62.38
the reproducibility of the experiments. The distilled water per-
18 0.0108 49.67
meate fluxes obtained after the membrane cleaning process
19 0.0143 56.87
were observed to be almost similar to the initial permeate 20 0.0137 61.07
fluxes.
Crude biodiesel refining using membrane ultra-filtration process 387

120

100

Permeate flux (kg/m hr)

2
80

60

40

20

0
0 10 20 30 40 50 60
Time (min )

Figure 2 Permeate flux vs time for biodiesel ultrafiltration.

Table 4 Retention coefficients (%R1 and %R2) of glycerol and soap.

Run order TMP (bar) Temp. °C Flow rate (L/min) %R1 (free glycerol) %R2 (Soap)
Experimental Predicted Experimental Predicted
1 1.00 30.00 60.00 93.87 94.02 95.82 96.45
2 3.00 30.00 60.00 92.54 92.25 2.24 93.35
3 1.00 50.00 60.00 89.23 90.57 93.15 93.33
4 3.00 50.00 60.00 89.92 90.16 94.52 94.89
5 1.00 30.00 150.00 93.85 93.51 93.25 92.77
6 3.00 30.00 150.00 89.45 89.38 84.15 85.37
7 1.00 50.00 150.00 96.57 96.78 95.94 96.72
8 3.00 50.00 150.00 95.60 95.36 96.20 95.47
9 1.00 40.00 105.00 96.77 96.78 95.72 96.14
10 3.00 40.00 105.00 94.83 94.19 93.93 93.97
11 2.00 30.00 105.00 96.32 96.88 95.06 94.60
12 2.00 50.00 105.00 97.65 97.46 96.43 97.35
13 2.00 40.00 60.00 93.67 93.60 94.68 95.92
14 2.00 40.00 150.00 95.50 95.94 95.17 94.38
15 2.00 40.00 105.00 97.93 97.67 96.45 96.42
16 2.00 40.00 105.00 97.92 97.67 96.57 96.42
17 2.00 40.00 105.00 97.82 97.67 95.79 96.42
18 2.00 40.00 105.00 98.08 97.67 96.66 96.42
19 2.00 40.00 105.00 97.35 97.67 96.15 96.42
20 2.00 40.00 105.00 97.65 97.67 97.79 96.42

The amount of soap in biodiesel was then determined using ½ðCf
Þ  100
%R ¼ ð3Þ
Eq. (2). Finally the result of Eq. (2) was multiplied by one mil- Cf
lion to obtain grams of soap per million grams of the sample
where Cf and Cper are free glycerol and soap mass fractions in
(ppm).
the feed and the permeate, respectively.
Px0:01xQ soapðgramsÞ
¼ ð2Þ
1000xR sampleðgramsÞ 2.6.3. Determination of physical properties of biodiesel
where P = HCl (grams), Q = Catalyst Factor (304.4 for American Society for Testing and Materials (ASTM) was
NaOH, 320.56 for KOH), R = biodiesel in solution (grams) used to determine the physical properties of the produced
and Sample = difference in the weights. biodiesel. Some of the physical properties determined
Both free glycerol and soap contents in both the feed and include among others: viscosity at 40 °C (ASTM D445-06),
permeate were determined after each run. The coefficients of density at 15 °C (ASTM D4052-96), cloud point (ASTM
retention (%R) of free glycerol and soap were calculated using D2500), pour point (ASTM D97-93), and flash point
Eq. (3): (ASTM D93-07).
388 I.M. Atadashi et al.

Predicted vs. Actual 3. Results and discussions

100.00 3.1. Separation and purification of biodiesel using membrane

ultrafiltration process
98.00
The separation of free glycerol and soap from biodiesel was
achieved using ceramic membrane with pore size of 0.05 lm.
96.00
Predicted Due to the immiscibility of free glycerol and biodiesel as well
as the surface activity of soap, the soap exists in the form of
94.00 reversed micelle which is very similar to the form of phospho-
lipids in the hexane miscella whose size is larger than a single
92.00 solute molecule [24]. The hydrophilic end of the soap is bound
to the droplets of free glycerol while the hydrophobic end is
submerged into the biodiesel. The reversed micelle of free glyc-
90.00
erol and soap is too large to pass through the pores of the
membrane, and therefore easily retained during biodiesel mem-
88.00 brane separation process [21]. The presence of methanol and
soap during biodiesel purification process leads to an increase
88.00 90.00 92.00 94.00 96.00 98.00 100.00 in the contact surface area and decrease in the interface tension
Actual between glycerol and biodiesel, thus enhancing the dissolution
between the two phases and reducing glycerol molecular size.
Figure 3 Predicted and experimental values for %R1 (free
For that reason, before starting the membrane purification
glycerol).
process, the crude biodiesel was subjected to rotary evapora-
tion for the complete removal of the residual methanol. In
Predicted vs. Actual addition the membrane surface immersed by the biodiesel
has various surface forces to prevent the polar molecules from
98.00 passing through the membrane pores. It has been noted that
free glycerol and soap are the major substances in biodiesel
96.00
causing membrane fouling. Ghasem [28] noted that the degree
of membrane fouling is dependent on the properties of the pro-
cess feed and on the nature of the membrane employed. There-
94.00 fore membrane fouling is firstly controlled by careful choice of
Predicted membrane type. Secondly, a good choice of module design will
offer appropriate hydrodynamic conditions for the particular
92.00
application. The author revealed that when membrane fouling
occurred, the permeation rate can be substantially restored
90.00 through back-flushing of the membrane. However, in this
work, less membrane fouling effect was observed during the fil-
tration process due to low contents of soap [24].
88.00
In order to improve the separation of biodiesel from the
88.00 90.00 92.00 94.00 96.00 98.00
contaminants, acidified water was added to the crude biodiesel
before commencing the membrane ultrafiltration process. The
Actual
acid neutralizes the residual catalyst and also converts the soap
Figure 4 Predicted and experimental values for %R2 (soap). into water-soluble salts. The addition of water to the crude

Table 5 ANOVA for the response surface models.

Source Retention of free glycerol Retention of Soap
F-value p-value Remarks F-value p-value Remarks
Model 120.50 <0.0001 Significant 12.21 0.0002 Significant
A-TMP 42.68 <0.0001 9.99 0.0101
B-Temp 5.84 0.0363 16.01 0.0025
C-Flow rate 93.06 <0.0001 5.04 0.0485
AB 25.07 0.0005 16.11 0.0025
AC 18.88 0.0015 7.92 0.0184
BC 209.60 <0.0001 31.20 0.0002
A2 52.75 <0.0001 4.33 0.0640
B2 4.66 0.0564 0.46 0.5150
C2 156.21 <0.0001 3.72 0.0046
Lack of Fit: 3.43 0.1014 Not significant 4.12 0.0731 Not significant
Crude biodiesel refining using membrane ultra-filtration process 389

biodiesel leads to the formation of aqueous phase containing addition of acidified water generated higher permeates fluxes
glycerol, salt, catalyst and other related water-soluble sub- but a sharp drop in the fluxes was observed. This was due to
stances, which differs from the phase rich in unreacted oil initial permeate flux stabilization. Saleh et al. [21] conducted
and biodiesel. The presence of water reduces the solubility of experimental study for the removal of free glycerol using poly-
biodiesel in the glycerol by forming agglomeration of glycerol meric membrane with a total filtration surface area of
in larger droplets [26]. As can be seen in Table 3, biodiesel sam- 0.0276 m2. The authors reported that addition of water
ples with low concentrations of glycerol and soap were (0.06–0.2 wt%) led to the formation of an aqueous phase dis-
achieved with some of the samples presenting glycerol concen- persed in the biodiesel which provided effective free glycerol
trations that are well below the limit of 0.02 wt% as prescribed retention.
by ASTMD6751. Although both ASTMD6751-03 and
EN14214 have not stated soap limit in biodiesel product, 3.2. Permeate fluxes
achievement of low soap levels in biodiesel could result in
ASTM specification for sulfated ash (0.02% weight) not to The initial permeate fluxes obtained using distilled water at a
be exceeded [30]. The achievement of low values of impurities temperature of 50 °C, transmembrane pressure of 1, 2 and 3
in the purified biodiesel samples demonstrated the efficiency bars, and flow rate of 150 L/min were 67 kg/m2 hr, 72 kg/
and suitability of the membrane system developed. As well, m2 hr and 81 kg/m2 hr respectively. To ascertain the efficiency

Figure 5a Plots for Response surface and contour presenting the effects of flow rate (L/min) and TMP (bar) on the retention of free
glycerol by biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).
390 I.M. Atadashi et al.

and performance of the membrane system, after each mem- The permeate fluxes were presented as a function of time.
brane cleaning process, the membrane system was ran with dis- The fluxes obtained varied and were based on the operating
tilled water and the permeate fluxes were compared with initial parameters. During cross flow filtration, continuous gel layer
permeate fluxes. The fluxes obtained in all the cases showed lit- building to a certain thickness is observed, then accumulation
tle variations. of hydrophilic compounds (polar-rich) on the membrane sur-
Furthermore during the course of the biodiesel membrane face which forms larger droplets [23]. As can be seen in
ultra-filtration process, different permeate fluxes were gener- Fig. 2 reasonable steady state thickness of the gel layer was
ated with some of the permeate fluxes presented in Fig. 2. achieved over the course of the experimental runs, demonstrat-
Steady state permeate fluxes were achieved over the course ing that the thickness of the gel layer above the membrane had
of the experimental runs, indicating that the thickness of the become stable. Fig. 2 showed decrease in the permeate flux in
gel layer above the membrane had stabilized [22]. The perme- the first 30 min which later stabilized. It was reported that the
ate fluxes obtained were derived using Eq. (4): continuous flux reduction with time indicates that other ‘‘in-
crustation” phenomena, such as pore blocking or molecule
Q
J¼ ð4Þ adsorption on the membrane surface, must have occurred
Ax [31]. The best experimental conditions for the membrane
where J = flux, Q = mass (kg), A = area (m2) and t = time experimented (TMP = 2 bar, Temp = 40 °C and flow rate =
(hr). 105 L/min) were obtained at the center point.

Figure 5b Plots for Response surface and contour presenting the effects of temperature (°C) flow rate (L/min) on the retention of free
glycerol by biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).
Crude biodiesel refining using membrane ultra-filtration process 391

3.3. Response surface analysis separation and refining of biodiesel depend on all the operat-
ing parameters. The coefficients of retention (%R) of free glyc-
This study is centered on the optimization of the operating erol and soap were found to vary from 89.45–98.08% and
parameters (transmembrane pressure, flow rate and tempera- 89.52–96.66% respectively.
ture) using RSM so as to enhance biodiesel membrane separa- Regression analysis is the general approach to fit the empir-
tion process. The Response Surface Methodology is chosen to ical model with the collected response variable data [32]. By
optimize the process because it is adequate enough to evaluate means of multiple regression analysis, the responses obtained
the parametric effects on the performance of the membrane for in Table 4 were correlated with the three independent operat-
the separation and refining of biodiesel. In addition the oper- ing parameters using the polynomial equations (Eqs. (5) and
ating parameters chosen were employed to discover the best (6)). The coefficients of the full regression model equations
operating conditions that have effects on the separation and and their statistical significance were determined and evaluated
refining of biodiesel using RSM. The coefficients of retention using Design-Expert 8.0.0 software from State-Ease Inc, USA.
(%R) of free glycerol and soap contents were determined for The quadratic models of the retention coefficients for free glyc-
each experimental run. Table 4 presents coefficients of reten- erol and soap as a function of transmembrane pressure (A),
tion of the contaminants (free glycerol and soap) by the mem- temperature (B) and flow rate (C) are presented in Eqs. (5)
brane system. Based on the data obtained, it was found that and (6), respectively.

Figure 5c Plots for Response surface and contour presenting the effects of TMP (bar) and temperature (°C) on the retention of free
glycerol by biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).
392 I.M. Atadashi et al.

%R1 ðFree glycerolÞ ¼ þ97:67
0:80A þ 0:29B þ 1:17C soap. In addition, using Design Expert software analysis
þ 0:68AB
0:59AC þ 2:02BC of variance (ANOVA), the significance and fitness of the
models were studied. The ANOVA for the Response Sur-

1:69A2
0:50B2
2:90C2 ð5Þ face Methodology is presented in Tables 5. Significance
means that the approximated value of the variable coeffi-
%R2 ðSoapÞ ¼ þ96:42
1:08A þ 1:37B
0:77C cient is bigger than a value that would be achieved from
þ 1:54AB
1:08AC þ 2:14BC
1:36A2 experimental noise. Also the p value is the probability
values employed to determine the significance of each of

0:44B2
1:26C2 ð6Þ the coefficient which may invariably indicate the pattern
The positive sign in front of the terms shows synergistic of association between the parameters [34]. The significance
effect whereas the negative sign indicates antagonistic effect of the result obtained is evaluated by the closeness of its p-
[33]. Figs. 3 and 4 present the predicted and experimental value to zero (0.00). For the effect to be statistically signif-
values for free glycerol and soap retention by means of icant, the confidence level should be 95%, this indicates that
the developed models. These figures demonstrate that the the p-value should be less than or equal to 0.05 [8]. Thus
models represent a relatively good description of the exper- the p-values from the ANOVA tables showed that all the
imental data regarding the retention of free glycerol and linear terms are statistically significant. To minimize error,

Figure 6a Plots for Response surface and contour presenting the effects of TMP (bar) and temperature (°C) on the retention of soap by
biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).
Crude biodiesel refining using membrane ultra-filtration process 393

all of the coefficients were considered in the design. It can Therefore it is desirable to have a ratio larger than 4. In this
also be observed that the statistical analysis of variance study, model ratios of 31.090, and 15.624 were achieved for
revealed that the overall models p-values were lower than free glycerol and soap retention. The model ratios are much
0.0003, which shows high significance. larger than 4; this demonstrate the model adequacy. As well,
Moreover, to validate the models, the goodness of fit was coefficient of variation (CV) with lower values ranging from
determined by evaluating coefficients of determination (R2). 0.43% to 1.59%, usually suggests reliability and good preci-
It was reported that coefficients of determination (R2) with val- sion of the experiments [34].
ues more than 80% should be achieved to ensure good agree- Additionally, lack of fit of the models was used to test the
ment between experimental data and predicted values [34]. In adequacy of all the models generated, thus, lack of fit of a
this research work, higher coefficients of determination for quadratic model is the weighted sum of squared deviations
the retention of free glycerol (R2 = 99%) and for the retention between the mean response at each factor level and the
of soap (R2 = 92%) were obtained. Furthermore in the exper- corresponding fitted value [27,35]. Consequently the lack of
iments conducted, the predicted R-squared which is a measure fit of the models was not statistically significant since the
of goodness of the model, for all the responses was above 80%. probability values were all more than 0.05 (lack of fit
This confirms that there is no problem with either the experi- p-value > 0.05 is not significant). This shows that the model
mental data or the models [27]. As well, the signal to noise equations are well fitted to all the data (Non-significant lack
ratio is being measured by adequate precision of the model. of fit is good).

Figure 6b Plots for Response surface and contour presenting the effects of flow rate (L/min) and TMP (bar) on the retention of soap by
biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).
394 I.M. Atadashi et al.

Figure 6c Plots for Response surface and contour presenting the effects of temperature (°C) flow rate (L/min) on the retention of soap
by biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).

3.4. Effect of operating parameters on the retention of biodiesel can as well be visualized as a two and three-dimensional plots
contaminants that show the response as a function of two factors, with one
factor kept constant. Fig. 5a–c present surface plots for 3D
3.4.1. Effects of parameters on free glycerol retention and 2D. It was found that the retention of glycerol increased
The ANOVA table (Table 5) indicates that the three operating when TMP and flow rate were increased. Optimum glycerol
parameters, transmembrane pressure (A), temperature (B) and retention of 98% was achieved at TMP (2 bar) and flow rate
flow rate (C) and the interactions of transmembrane pressure– (105 L/min). Further increase in TMP or flow rate resulted
temperature (AB), temperature-flow rate (BC) and transmem- in the decrease in the glycerol retention. Wang et al. [24] fil-
brane pressure-flow rate (AC) and the quadratic terms of reac- tered crude biodiesel using feed flow rate (50–150 (L/min))
tion temperature (A2 and C2) are statistically significant for and transmembrane pressure of 0.05 and 0.20 MPa. The
free glycerol retention during biodiesel separation and refining, authors reported a significant result for the retention of free
with reference to the p-values less than 0.05. Also, based on the glycerol and obtained 0.0108 + 0.0034 wt% free glycerol in
model developed, it was observed that visually all the param- the permeate. It was also found that at a higher TMP, the glyc-
eters played a vital role during the retention of free glycerol. erol retention coefficients were low.
The retention of the free glycerol varied considerably upon The shapes of surface response plot either elliptical or circu-
changing the flow rate and TMP. Further the response surfaces lar indicate significant or insignificant interaction between the
Crude biodiesel refining using membrane ultra-filtration process 395

error occurrence. Fig. 6(a) presents the effects of TMP and

Table 6 Constrains for the parameters and responses in
temperature on the retention of soap. At any designed TMP
numerical optimization.
from 1–3 bar and temperature of 30–50 °C, an increase in
Parameters Ultimate goal Experimental region TMP and temperature led to an increase in soap retention to
TMP (bar) In range 1–3 the optimum point after which the retention coefficients were
Temp (°C) In range 30–50 observed to decrease. Further, Fig. 6(b) and (c) shows the
Flow rate (L/min) In range 60–150 interaction between flow rate and TMP, and flow rate and tem-
Membrane pore size (0.05 lm) perature. In Fig. 6(b), retention of soap increased with an
Free glycerol retention (%) In range 89.23–98.08 increase in both flow rate and TMP up to the optimum point.
Soap retention (%) In range 89.58–96.66 However gradual decrease in the soap retention was noticed
when the TMP exceeded 2 bar. From Fig. 6(b) it can be seen
that higher temperatures could significantly affect the retention
of soap. At higher temperatures and low TMP, the retention of
soap is favoured. Further the retention of soap is much
Table 7 Optimization results and model evaluation.
favoured at higher temperatures and higher flow rates.
TMP Temp Flow rate %R (free %R (Soap)
(bar) (°C) (L/min) glycerol) 3.5. Optimization
Predicted Predicted
experimental experimental Based on the predicted quadratic models which were validated
2 40 105 97.669 97.523 96.416 96.568 statistically, numerical hill-climbing algorithms were employed
to search for the most desirable outcome [8]. The operating
parameters and responses (free glycerol and soap retention)
with respect to low and high limits satisfy the criterion defined
operating parameters. The shapes of Fig. 5(b) indicate that the for the optimum conditions shown in Table 6. Thus the mem-
mutual interactions between the temperature and flow rate are brane system was optimized based on the limits of operating
significant on the retention of free glycerol. In this case the effect parameters and the responses generated. The optimum operat-
of temperature above 40 °C is significant with more glycerol ing conditions were then used to evaluate the precision of the
retention coefficient being observed. Also the interaction quadratic model; hence the experimental and the predicted val-
between TMP and temperature is significant for the retention ues are shown in Table 7. Comparing the predicted and exper-
of free glycerol as shown in Fig. 5(c). At temperatures above imental results, it can be observed that the errors between the
40 °C, the effect of temperature on glycerol retention is not predicted and experimental results are less than 0.2%. For that
much significant. This results in low coefficients of free glycerol reason it can be stated that the generated quadratic models
retention. Thus the maximum retention of free glycerol was have sufficient accuracy to predict the retention of free glycerol
positioned at the center of the experimental region. It was also and soap. In addition the accuracy of the models has been val-
observed that elliptical contour plots indicate perfect relation- idated by p-values presented in the ANOVA tables (Table 5).
ships between independent parameters [36]. The transmem- The results obtained demonstrated that Response Surface
brane pressure, temperature and flow rate are identified as key Methodology can be effectively employed to optimize the
parameters in the dynamics of biodiesel separation and purifica- operating parameters for the membrane biodiesel separation
tion using membrane technology [24,26]. process. Also under the optimum operating conditions, the
value of free glycerol (0.012 (wt%)) is well below ASTM inter-
3.4.2. Effects of parameters on the retention of soap national standard specifications for biodiesel. In addition, a
significant reduction in the concentration of soap (52.48
Figs. 6a–c present surface response plots and contour plots for (ppm)) was achieved. As well the physical properties of biodie-
the retention of soap by the membrane. The retention of soap sel produced at the optimum conditions met both EN 14214
showed a trend with the optimum retention of the soap being and ASTM D6751 as depicted in Table 8.
also achieved at the center points. The soap retention based on
the results obtained in the ANOVA table (Table 5), indicated 3.6. Cleaning of the membrane module
that the p-values (<0.05) of the three linear terms (A, B, C),
the interaction terms (AB, AC, and BC) and the quadratic The procedure of cleaning the membrane module was fast and
term (C2) were statistically significant. During the design, all effective. The cleaning of membrane system is quite essential; it
the coefficients were considered to minimize the possibility of is as important as the ultrafiltration process itself, since it is

Table 8 Physical properties of the biodiesel produced.

Properties Test method Unit Determined value Limits (ASTM)
Viscosity ASTM D445-06 mm2/s 4.91 1.9–6.0
Density ASTM D4052-96 kg/m3 878 –
Pour point ASTM D97-93 °C 5 –
Cloud point ASTM D2500 °C 14 Report
Flash point ASTM D93-07 °C 179 130 min
396 I.M. Atadashi et al.

fundamental in determining both economic and technical via- [4] T.S. Gendy, S.A. El-Temtamy, Egypt. J. Pet. 22 (1) (2013) 43–
bility of the process on an industrial scale, where repeatability 51.
and efficiency are quite important. [5] M. Berrios, R.L. Skelton, Chem. Eng. J. 144 (3) (2008) 459–465.
[6] A.V. Bridgwater, Chem. Eng. J. 91 (2–3) (2003) 87–102.
[7] A. Santana, J. Maçaira, M.A. Larrayoz, Fuel Process. Technol.
4. Conclusion 96 (2012) 214–219.
[8] S. Baroutian, M.K. Aroua, A.A.A. Raman, N.M.N. Sulaiman,
Although biodiesel is believed to be the best alternative to Bioresour. Technol. 102 (2) (2011) 1095–1102.
[9] R. Luque, l. Herrero-Davila, J.M. Campelo, J.H. Clark, J.M.
petro-diesel, it must be purified to attain ASTMD6751-03
Hidalgo, D. Luna, J.M. Marinas, A.A. Romero, Energy
and EN14214 standards specifications. Therefore to achieve Environ. Sci. 1 (2008) 542–564.
high-quality biodiesel fuel, selection of a suitable technique [10] D.M. Alonso, R. Mariscal, M.L. Granados, P. Maireles-Torres,
to purify crude biodiesel is necessary. Based on the investiga- Catal. Today 143 (1–2) (2009) 167–171.
tions conducted, the following conclusions were made: [11] M. Benzies, Biodiesel purification techniques. www.filtertech-
nikcouk/biodiesel, 2011.
(1) Application of ceramic membrane with pore size of [12] F. Ma, M.A. Hanna, Bioresour. Technol. 70 (1) (1999) 1–15.
0.05 lm has allowed effective separation of biodiesel [13] A. Kelloway, W. Alex Marvin, Lanny D. Schmidt, P. Daoutidis,
from free glycerol and soap. Chem. Eng. Res. Des. (2013).
(2) Application of acidified water was instrumental to the [14] Y.M. Sani, W.M.A.W. Daud, A.R. Abdul Aziz, J. Environ
Chem. Eng. (2013).
successful application of the membrane system for the
[15] M.C.S. Gomes, N.C. Pereira, J. Membr. Sci. 352 (1–2) (2010)
purification of biodiesel. 271–276.
(3) It was found that at the optimum conditions, the mem- [16] J.V. Gerpen, Fuel Process. Technol. 86 (10) (2005) 1097–1107.
brane process developed offered biodiesel that met both [17] I.M. Atadashi, M.K. Aroua, A.R. Abdul Aziz, Appl. Energy 88
ASTMD6751 and EN14214 standards specifications. (12) (2011) 4239–4251.
(4) The application of Central Composite Design (CCD) [18] T.L. Chew, S. Bhatia, Bioresour. Technol. 99 (17) (2008) 7911–
couple with Response Surface Methodology (RSM) pre- 7922.
sented clear understanding to the interaction of various [19] J. Kwiecien, M. Hajek, F. Skopal, Bioresour. Technol. 100
operating process parameters for the separation of (2009) 5555–5559.
biodiesel. [20] O. Chavalparit, M. Ongwandee, J. Environ. Sci. 21 (11) (2009)
1491–1496.
(5) The best conditions obtained for the membrane with
[21] I.M. Atadashi, M.K. Aroua, A.A. AbdulAziz, Renewable
pore size of 0.05 lm were TMP of 2 bar, temperature Energy 36 (2) (2011) 437–443.
of 40 °C and flow rate of 105 L/min and permeate flux [22] J. Saleh, M.A. Dubé, A.Y. Tremblay, Fuel Process. Technol. 92
of 22.17 kg/m2 hr. (7) (2011) 1305–1310.
(6) Overall assessment showed that membrane technique [23] J. Saleh, A.Y. Tremblay, M.A. Dubé, Fuel 89 (9) (2010) 2260–
has the potential to completely replace both biodiesel 2266.
water and dry wishing processes, for the process being [24] Y. Wang, X. Wang, Y. Liu, S. Ou, Y. Tan, S. Tang, Fuel
operated effectively without water being consumed. Process. Technol. 90 (3) (2009) 422–427.
The absence of water indicated that no wastewater was [25] M. Balat, H. Balat, Energy Convers. Manage. 49 (10) (2008)
discharged. 2727–2741.
[26] M. Gomes, A.A. Pedro, C. Nehemias, J. Membr. Sci. 378 (2011)
(7) In addition the membrane system developed provided
453–461.
clean biodiesel without using any water washing. This [27] M. Zabeti, W.M.A.W. Daud, M.K. Aroua, Appl. Catal. A 366
certainly indicates environmental friendliness of mem- (1) (2009) 154–159.
brane biodiesel separation. Thus membrane develop- [28] D.N. Ghasem, Biochemical Engineering and Biotechnology,
ment could in the future serve as a technology that Elsevier, Amsterdam, 2007, pp. 351–389.
can wholly replace the conventional biodiesel purifica- [29] D. Naviglio, R. Romano, F. Pizzolongo, A. Santini, A. de Vito,
tion processes. L. Schiavo, G. Nota, S.S. Musso, Food Chem. 102 (2007) 399–
405.
[30] J. Wall, Comparison of Methods for the Purification of
Biodiesel-Thesis, Biological and Agricultural Engineering,
Acknowledgments
University of Idaho, 2009.
[31] H. Choi, Z. Kai, D.D. Dionysios, B.O. Daniel, A.S. George, J.
The authors wish to immensely thank the Management of Membr. Sci. 248 (2005) 189–199.
Adamawa State University, Mubi-Nigeria and the University [32] N. Dizge, C. Aydiner, D.Y. Imer, M. Bayramoglu, A.
of Malaya, Malaysia and, for their support to this research Tanriseven, B. Keskinler, Bioresour. Technol. 100 (2009)
work. 1983–1991.
[33] V. Caballero, F.M. Bautista, J.M. Campelo, D. Luna, J.M.
Marinas, A.A. Romero, J.M. Hidalgo, R. Luque, A. Macario,
References
G. Giordano, Process Biochem. 44 (2009) 334–342.
[34] P. Tamunaidu, S. Bhatia, Bioresour. Technol. 98 (2007) 3593–
[1] J. Dupont, Paulo A.Z. Suarez, M.R. Meneghetti, S.M.P.
3601.
Meneghetti, Energy Environ. Sci. 2 (2009) 1258–1265.
[35] D.C. Montgomery, Design and Analysis of Experiment, fifth
[2] A.K. Tiwari, A. Kumar, H. Raheman, Biomass Bioenergy 31
ed., Wiley Inc, New York, USA, 1997.
(2007) 569–575.
[36] R.V. Muralidhar, R.R. Chirumamila, R. Marchant, P. Nigam,
[3] Anton Radu, Chem. Eng. Res. Des. (2013).
Biochem. Eng. J. 9 (1) (2001) 17–23.