Alexandria Engineering Journal (2014) 53, 705–715

HO ST E D BY

Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aejwww.sciencedirect.com

ORIGINAL ARTICLE

Removal of residual palm oil-based biodiesel

catalyst using membrane ultra-filtration technique:

An optimization study

* Corresponding author. Tel.: +234 81 36837147.

E-mail address: atadashimusa1@yahoo.com (I.M. Atadashi).

Peer review under responsibility of Faculty of Engineering, Alexandria

University.

http://dx.doi.org/10.1016/j.aej.2014.07.002

1110-0168 ª 2014 Production and hosting by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.

I.M. Atadashia,*, M.K. Aroua

b, A.R. Abdul Aziz

b, N.M.N. Sulaiman

b

a Adamawa State University, Mubi, Nigeriab Chemical Engineering Department, Faculty of Engineering, University Malaya, 50603 Kuala Lumpur, Malaysia

Received 2 September 2013; revised 28 June 2014; accepted 1 July 2014

Available online 2 August 2014

KEYWORDS

Transesterification;

Crude biodiesel;

Membrane refining;

Potassium content

Abstract In this research work, residual potassium hydroxide catalyst was removed from palm oil-

based alkyl esters (biodiesel) using membrane separative technique, with the aim of achieving high-

quality biodiesel thatmeets international standard specifications. Further, Central Composite Design

(CCD) coupled with Response SurfaceMethodology (RSM) was employed to study the effects of the

system variables such as flow rate, temperature and transmembrane pressure (TMP) on the retention

of potassium. At the optimum conditions, the coefficient of retention (%R) of the catalyst was 93.642,

and the content of the potassium was reduced from 8.328 mg/L to 0.312 mg/L; a value well below the

one specified by both EN 14214 and ASTM D6751 standards. In addition, the comparison between

predicted and experimental values for the catalyst retention offers a reasonable percentage error of

0.081%. Therefore, this study has proven that membrane technique can be used to post treat crude

biodiesel; in order to achieve high-quality biodiesel fuel that can be efficiently used on diesel engines.ª 2014 Production and hosting by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria

University.

1. Introduction

It is noteworthy to note that increasing energy demands,diminishing fossil fuel resources, concern over environmentalpollution, and the necessity to use clean fuels worldwide havemotivated investigations on the substitute renewable energy

sources such as wind energy, biofuels, and geothermal energy[1,2]. Biodiesel is ranked one of the most suitable substitute

fuels to petro-diesel fuel. It is usually derived from varietiesof feedstocks including among others; animal fats and vegeta-

ble oils. Because biodiesel is produced from biomass materials,it provides numerous advantages which include; renewability,biodegradability, reduction in global warming effects, higher

cetane numbers and higher flash points. In addition, biodieseloffers low aromatic and sulfur contents, high energy content,as well no net addition to the atmospheric CO2 levels [3]. Also,

the dependency on the importation of petroleum is reduced,since biodiesel is obtained from domestic sources [4].

Biodiesel is usually produced via four different techniquesnamely; direct/blend with oils [5,6], micro-emulsions [7], pyro-

lysis [6], and transesterification [1]. Among these techniques,

706 I.M. Atadashi et al.

biodiesel production is mostly achieved by means of transeste-rification. Transesterification is the reaction involving triglyc-erides with a short-chain alcohol such as methanol and

ethanol in the presence of a catalyst. The process of transeste-rification produces biodiesel as the main product and glycerolas by-product. Further, methanol is frequently used as alcohol

in biodiesel production, because of its availability, low price,good physicochemical properties, and it also leads to low vis-cous biodiesel fuel [8]. It is worth mentioning that, the process

of transesterification is employed to decrease the viscosity oftriglycerides; thus improving the physical properties of biodie-sel [9].

Transesterification is catalyzed through alkaline, acid, or

enzymes catalyst [6,8,10]. Because of the nonexistence of soapsformation and the ease to refine crude biodiesel; enzymes cat-alysts have received more attention from different researchers.

However, high cost and longer reaction rates have limited theirbiodiesel commercial application [10]. Although, a number ofresearchers preferred acid and alkaline catalysts compared

with enzymes catalysts in producing biodiesel fuel, becauseof their lower prices and availability. However, transesterifica-tion reaction involving acid catalysts require longer reaction

time and higher oil to alcohol ratios, as well the catalysts arevery corrosive and difficult to manage [11].

Biodiesel production mostly employed alkaline catalystssuch as sodium hydroxide and potassium hydroxide [9,12]. It

was reported that biodiesel is best produced using homoge-neous alkaline catalysts [11]. Alkali-catalyzed transesterifica-tion process presents several advantages which include;

operating at moderate temperature (40–60 �C), ambient pres-sure, and offering high biodiesel yields of 98–99% in shortreaction times. Compared with acid catalysts, alkali-catalyzed

transesterification is 4000 times faster [13]. Also, for alkalinecatalysts to effectively perform, the contents of free fatty acidsand water in any feedstock should not be more than 0.5 wt%

and 0.6%, respectively [4,14].Further, once the glycerin and biodiesel phases have been

separated, the excess alcohol in each phase is removed by dis-tillation. In other systems, the alcohol is removed and the mix-

ture neutralized before the glycerin and esters is separated. Ineither case, the alcohol is recovered and re-used. The by-prod-uct, glycerin contains unused catalyst and soaps formation

that are neutralized with an acid and sent to storage as crudeglycerin (water and alcohol are removed later, chiefly usingevaporation, to produce 80–88% pure glycerin). Once sepa-

rated from glycerin, the crude biodiesel undergo a number ofpurification steps such as excess methanol removal via rotaryevaporation, neutralization, wet washing and drying. Thesesteps are required to remove the contaminants in biodiesel

prior to its application as fuel on diesel engines [15]. Presenceof the remaining alkaline metals in biodiesel fuel can causeproblems in the fuel injection system components due to the

formation of carbon residue [9]. It is of great importance tomention that the tolerable limit specified by the internationalstandards (EN 14214 and ASTM D6751) for metal content

in biodiesel is 5 mg/kg [9]. For that reason, catalyst removalafter the completion of alkali-catalyzed transesterification isimportant.

Furthermore, crude biodiesel is frequently refined throughdifferent techniques namely wet washing, dry washing andmembrane extraction [16–18]. However water washing is themost conventionally used method in the biodiesel industry.

Usually, an acid such as H3PO4 is added to neutralize theremaining catalyst and split the soap before wet washing pro-cess commences [19]. The tribulations related to wet washing

process have resulted to the application of dry washingmethod. Dry washing method employs either ion exchange res-ins or silicates. This technique is introduced to purposely

replace wet washing technology in order to circumvent theproblems associated with wet washing method. On the con-trary, application of both wet and dry washing technologies

result to increase in the production cost due to the cost ofadded materials and in some situations overrun specified limitsas stipulated in EN 14214 Standard [6,9].

The recent use of membrane technology for the purification

of crude biodiesel has provided impetus for the achievement ofhigh-quality biodiesel fuel. Membrane systems exploit theinherent properties of high surface-area-per-unit-volume, high

selectivity, and their prospect to control the level of contactand/or mixing between two phases [20]. Membranes are classi-fied into two: inorganic and organic membranes. Inorganic

membranes are preferred because of the following reasons:high thermal and mechanical stability, resistance to acid andbase environments, and present energy savings [21,22]. Purifi-

cation of crude biodiesel via membrane technology wasreported to provide high-quality biodiesel [9]. But in all theprevious literatures published, no research work has been con-ducted on the optimization of ceramic membranes with pore

size of 0.05 lm for the removal of residual catalyst (KOH)from crude biodiesel. As a result, this paper is focused onthe removal of residual catalyst (KOH) from crude biodiesel

using ceramic membrane with pore size of 0.05 lm. As well acomprehensive optimization of the effects of the system vari-ables for the removal of residual catalyst is performed.

2. Methodology

2.1. Transesterification reaction for the production of biodiesel

The quantity of crude biodiesel required for the purification

process via membrane system was prepared using a batch reac-tor. Palm oil, methanol and potassium hydroxide (KOH) wereused as raw materials for the biodiesel production. Alkalinemetal hydroxide and potassium hydroxide (KOH) are

employed in producing crude biodiesel in comparison withmetal alkoxides such as CH3ONa and CH3OK because it ischeap and readily available. Besides high biodiesel yield more

than 98% is attained using moderate conditions of tempera-ture, pressure and molar ratios [6]. In this research work, abatch reactor with a size of 5 l was employed in producing

the crude biodiesel. A catalyst concentration of 1 wt%,KOH, and a molar ratio of methanol to oil of 6:1 was usedfor the biodiesel production. The reaction time for the biodie-

sel production was 1 h and the operating temperature was60 �C. The required quantity of KOH was thoroughly mixedin the required quantity of methanol. The carefully mixed cat-alyst and methanol were then fed into the reactor containing

2.5 l of palm oil. Several experimental runs were carried outusing 2.5 l of palm oil per batch. A water bath was used to heatthe reacting content to a temperature of 60 �C. The contents ofthe reactor were thoroughly mixed with a stirrer of stirringcapacity of 645 rpm to enhance its mixing. These reaction con-ditions were based on the reviewed literatures [6,8,23,24].

Removal of residual palm oil-based biodiesel catalyst using membrane ultra-filtration technique: An optimization study 707

After the completion of transesterification reaction, thecrude biodiesel consisting of residual catalyst, free glycerol,and other by-products was allowed to settle overnight. The set-

tled biodiesel mixture provided two distinct phases; the upperbiodiesel-rich phase and the bottom glycerol-rich phase. Thetwo distinct phases (biodiesel and glycerol) were then sepa-

rated using a decanter. After which, residual methanol wasremoved from the crude biodiesel using rotary evaporator.The biodiesel generated was then subjected to membrane

ultra-filtration process for the removal of the residual catalyst.In this work, in order to obtain sufficient biodiesel samples,quite a number of experimental runs were performed. Thesamples of biodiesel produced were properly stored in vessels

in a suitable place.

2.2. Membrane ultra-filtration technique

2.2.1. Separative membrane design

A tubular-type Al2O3/TiO2 ceramic membrane was used to

perform the refining experiments. The multi-channel mem-brane with pore size of 0.05 lm was acquired from JiangsuJiuwu Hitech CO., China. The ceramic membrane comprised

of filtration area of 0.031 m2. Fig. 1 shows schematic diagramof separative membrane system for the refining of the crudebiodiesel. The set up comprised of the feed tank, membranemodule, water bath, circulating pump, and digital balance,

etc. Pressure gauges and temperature indicator were providedto monitor the transmembrane pressures and temperature ofthe membrane system. The biodiesel samples were fed into a

5 l feed tank and circulated via the membrane system using acirculating pump. The membrane biodiesel refining processwas performed using operating conditions presented in Table 1.

The Selection of the levels of each design variables was basedon the data available in the open literatures [25–27]. The inletand outlet pressures at the ends of the membrane tube were

carefully adjusted using the valves in order to achieve thedesigned experimental conditions. In addition, the system tem-perature was monitored through water bath. Using the mem-brane with the pore size of 0.05 lm, quite a number of

experimental runs were carried out. These experiments werecarried out based on the design of experiment. Further, toimprove the separation and retaining of impurities such as

Figure 1 Membrane system for t

residual catalyst, acidified water was added. During the refin-ing process, the biodiesel concentrate was continuously circu-lated through the membrane module. Afterward, a digital

weighing balance was used to automatically measure the masspermeate fluxes at an interval of 10 min. The refining processeswere halted after 1 h of operation. At the end of every refining

process, both the crude and the refined biodiesel samples (per-meates) were taken and analyzed.

2.2.2. Operating system variables

The main operating variables assessed for the refining of crudebiodiesel via the ceramic membrane system are TMP, temper-ature, and flow rate. These operating variables are varied: flow

rate of 60–150 L/min, temperature of 30–50 �C and transmem-brane pressure of 1–3 bar. The potassium contents weredenoted as percentage (%) and the permeate fluxes generated

were expressed as kg/m2 h. The selection of the ranges; temper-atures of 30–70 �C and flow rates of 50–150 L/min for therefining of the crude biodiesel was based on the published lit-erature in which considerable potassium content was retained,

with biodiesel having low potassium content being achieved[27]. Further, transmembrane pressure of 1–3 bar was reportedto be a good operating condition for the purification of crude

biodiesel by means of a membrane system [28].

2.2.3. Experimental design

Design Expert software, version 8.0.0 (Stat-Ease Inc., USA)

was employed in this work, to conduct the experimental designand optimize the system operating variables. Central Compos-ite Design (CCD) coupled with RSM was selected for the pur-

pose of this design. The three independent design variableswere transmembrane pressure, flow rate and temperature,and the response was the catalyst removal (%) based on mem-

brane ultra-filtration process. In this design, six replications ofcenter points in a randomized order were applied; this isbecause of the natural variations, and to have a true measure-ment of error and stable predictions. Comparing with classical

technique which uses one variable at a time, this technique canresolve the interaction of effects of variables, as well as provid-ing good errors estimation. Additionally, the time and cost of

experiments are reduced since the overall number of trials isreduced [29]. The second-order polynomial presented in Eq.

he refining of crude biodiesel.

Table 1 Variable levels in the experimental design.

Parameters Symbol Values

Low (�1) Center (0) High (+1)

Transmembrane pressure (bar) X1 1 2 3

Temperature (�C) X2 30 40 50

Flow rate (L/min) X3 60 105 150

708 I.M. Atadashi et al.

(1) elucidates the relationship between dependent and indepen-

dent variables:

Y ¼ a0 þ a1Z1 þ a2Z2 þ a3Z3 þ a12Z1Z2 þ a13Z1Z3

þ a23Z2Z3 þ a11Z21 þ a22Z

22 þ a33Z

23 ð1Þ

where Y is the dependent variable; Z1, Z2 and Z3 are the inde-

pendent variables; a0 is the intercept, a1, a2, a3, are linear coef-ficients and a12, a13 and a23 are the interaction coefficients, a11,a22, a33 are the quadratic coefficients, respectively.

2.2.4. Membrane cleaning process

After each run, themembranemodule was scrupulously cleanedwith detergent and water until complete biodiesel removal was

achieved. Subsequently a solution consisting of 1 wt% NaOHat a temperature of 70 �C was circulated for 45 min throughthe membrane module. Water was then used to clean the mem-brane module which was then rinsed with warm distilled water.

It is important to mention that, proper cleaning of the mem-brane module is necessary to protect the equipment, restore itspermeability and maintain its performance.

2.2.5. Biodiesel characterization

The contents of potassium were measured before and aftermembrane ultra-filtration process via the PerkinElmer Optima

7000 DV ICP-OES (PerkinElmer, Inc., Shelton, CT, USA).The instrument was outfitted with WinLab32 and a carriergas (Argon (99.99%)). For the potassium analysis, the wave-

length was position at 766.486 nm. The oil standard for potas-sium, 500 mg/kg, and based oil were supplied by ConostanCompany. To dilute the standard at six different concentra-

tions (0.5 ppm, 1 ppm, 2 ppm, 5 ppm, 7 ppm and 15 ppm),Xylene solution (Merck, >99%) was used. To carry out thecalibration, the resultant solutions were injected directly intothe ICP-OES. The correlation coefficient (R2) was obtained

to be 0.99 which showed good linearity. The Instrument Detec-tion limit (IDL) which was found to be 0.03 mg/kg was deter-mined by three times the standard deviation of the blank (ten

replicates). Furthermore, seven portions of calibration stan-dard of 1 ppm were papered and analyzed for the calculationof method detection limit (MDL). Then the standard deviation

of seven results was calculated and multiplied by 3.14. Theattained MDL value was found to be 0.1 mg/kg. All samplesof biodiesel were injected to ICP-OES and after the determina-

tion of potassium peak area, the content of potassium in eachsample of biodiesel was calculated with reference to calibration[9].

The coefficient of retention (%R) of potassium content was

determined using Eq. (2):

%R ¼ ðYf � YperÞYf

� 100 ð2Þ

where Yf and Yper are the mass fractions of potassium in thecrude biodiesel and the permeate, respectively.

3. Results and discussions

3.1. Residual catalyst separation process

Ultra-filtration is a well-known membrane separation processwhich consists of different module configurations such as hol-

low fiber, flat plate, tubular, spiral wound and rotating devices.Dead-end and cross-flow configurations are the two standardmodes of ultra-filtration operation. In the cross flow approach

due to a pressure difference, the targeted fluid permeatesthrough the membrane. This process reduces the formationof filter cake and keeps it at a low level [30], and this makes

it suitable for different refining applications. Thus, cross flowroute was adopted in the refining of the crude biodiesel pro-duced to remove the residual catalyst, in order to ensure thatthe final biodiesel fuel meet the standards specifications of

ASTM D6751 and EN 14214 for use on diesel engines. Therefining process was started by separating biodiesel from glyc-erol via decantation and subjecting the biodiesel to rotary

evaporation for the complete removal of the excess methanol.Acidified water was then added to biodiesel which lead to theformation of aqueous phase containing catalyst and other

related water-soluble substances. This process aided the reduc-tion of biodiesel solubility in glycerol, and led to the retentionof the impurities by the membrane [25]. The process of adding

acidified water helped in breaking down the soap formationinto water soluble salt and free fatty acids, in addition to theacid neutralization process of the alkaline catalyst [20].

After membrane ultra-filtration process was done, the anal-

ysis carried out showed that application of membrane withpore size of 0.05 lm could provide biodiesel with low potas-sium contaminant, with some of the biodiesel samples meeting

the international standard specification as shown in Table 2.As well, addition of acidified water generated higher permeatesflux but sharp drop in the fluxes were observed. This could be

attributed to the initial permeate flux stabilization.It is worth mentioning that removal of catalyst (potassium)

in biodiesel is essential, because the presence of catalyst sub-stance is strictly limited by the biodiesel international stan-

dards specifications. The main difficulty associated withpotassium impurity is that when burning fuel in the dieselengine, the remaining metal contained in the catalyst will be

transformed to ash, and this may cause sulfated ash limit tobe deviated. Both ASTM D 6751 and EN 14241 have specifiedcatalyst and ash content of biodiesel to be 0.5 mg/kg and

0.02 wt% respectively. The presence of this metal is also asso-ciated with biodiesel instability. In addition large effects onbiodiesel fuel oxidative stability could be noticed even when

Table 2 Final levels of potassium in permeate (final biodiesel).

Run TMP (bar) Temperature (�C) Flow rate (L/min) Potassium (mg/L)

1 (�1) (�1) (�1) 0.547

2 (+1) (�1) (�1) 0.987

3 (�1) (+1) (�1) 1.344

4 (+1) (+1) (�1) 1.539

5 (�1) (�1) (+1) 1.150

6 (+1) (�1) (+1) 1.191

7 (�1) (+1) (+1) 0.855

8 (+1) (+1) (+1) 0.965

9 (�1) (0) (0) 0.580

10 (+1) (0) (0) 0.691

11 (0) (�1) (0) 0.541

12 (0) (+1) (0) 0.477

13 (0) (0) (�1) 0.873

14 (0) (0) (+1) 0.821

15 (0) (0) (0) 0.387

16 (0) (0) (0) 0.291

17 (0) (0) (0) 0.298

18 (0) (0) (0) 0.387

19 (0) (0) (0) 0.356

20 (0) (0) (0) 0.387

Removal of residual palm oil-based biodiesel catalyst using membrane ultra-filtration technique: An optimization study 709

trace amounts of this metal are present. Furthermore presence

of K can also increase corrosion rate of fuel systems andincrease in sludge and deposits build up. Many of these spentpotassium salts have low melting temperatures and may cause

engine damage in combustion chambers [31]. It was noticedthat presence of potassium in the final biodiesel product causedformation of insoluble soap and as well lead to deposit forma-tion in the vehicles filters. Consequently the residual catalyst

ought to be eradicated after which the quantification of thismetal becomes essential to guarantee the quality of the finalbiodiesel product [32].

3.2. Determination of permeate fluxes

To ascertain the suitability and applicability of the membrane

applied, the initial permeate fluxes were determined using dis-tilled water at a flow rate of 150 L/min, temperature of 50 �C,and transmembrane pressures of 1, 2 and 3 bars, and werefound to be 67 kg/m2 h, 72 kg/m2 h and 81 kg/m2 h respec-

tively. These conditions were employed as point of referenceto examine the cleaning efficiency of the membrane. The per-meate fluxes of the membrane system are presented in

Fig. 2(a–d) respectively. It can be seen that Fig. 2(a–d) presentsreasonable steady state permeate fluxes over the course of theruns, indicating that a steady state gel layer thickness was

attained. The gel layer accumulated on the membrane surfacedominated the flux decline and the effect of viscosity becamenegligible [24].

Further the permeate fluxes generated were presented as afunction of time for the ultra-filtration process. The permeatefluxes found vary and were based on the combination of thesystem variables. The permeate fluxes decreased in the first

(30 min) after which the fluxes attained a steady state valuesas shown in Fig. 2. The optimum operating conditions forthe membrane were: Temp = 40 �C, TMP = 2 bar, and flow

rate = 105 L/min, which were determined at the center points.

3.3. Application of Response Surface Methodology

This study investigated the relationship between three indepen-dent variables (Temperature, TMP and flow rates) and the

response (catalyst removal). Hence, RSM was employed toinvestigate both single and interactive effects of the variablesand also facilitate optimization of the post-treatment process.

With the help of RSM, a model was fitted to determine theremoval of the catalyst. The variation between the predictedand actual removal of catalyst signified that the model does

not require any transformation to augment its accuracy.Additionally, to get high-purity biodiesel, this work was

centered on optimizing system variables such as temperature,flow rate, and TMP via RSM. This is to enhance the mem-

brane biodiesel separation and purification processes whichprovide less costly biodiesel refining process in an industrial-ized level. The Response Surface Methodology is selected to

optimize the system variables as it is adequate to evaluatethe effects of system variables on the performance of the mem-brane refining of crude biodiesel. Thus the system variables

chosen were applied to obtain the most favorable conditionsthat have effects on the refining process using Response Sur-face Methodology. For each experiment carried out, the coef-ficient of retention (%R) was determined. Table 3 presents the

coefficients of retention of the membrane system. Based on thedata obtained; it was found that refining of crude biodiesel isdependent on all the system variables. The coefficient of reten-

tion (%R) of potassium which enabled high-quality biodieselduring membrane refining process was found to vary from79.9% to 95.1%. And the best retention of potassium was

obtained at the center points. The main advantage of ResponseSurface Methodology is that the interactions of system vari-ables are measured in the experimental design and is observed

easily based on the resultant regression equations. Diagnosticsof the residuals, the variation between the actual and the pre-dicted potassium content, demonstrated that transformingpotassium model is not required to enhance its retention. Thus,

Time (min)0 10 20 30 40 50 60

Perm

eate

flux

(kg/

m2 hr

)

0

20

40

60

80

100

Run 1 Run 2 Run 3 Run 4 Run 5

Perm

eate

flux

(kg/

m2 hr

)

0

20

40

60

80

100

Run 6 Run 7 Run 8 Run 9 Run 10

Time (min)0 10 20 30 40 50 60

Time (min)0 10 20 30 40 50 60

Perm

eate

flux

(kg/

m2

hr)

0

20

40

60

80

100

120

Run 11Run 12Run 13Run 14Run 15

Time (min)0 10 20 30 40 50 60 70

Perm

eate

flux

(kg/

m2 hr

)

0

20

40

60

80

100

Run 16Run 17Run 18Run 19Run 20

(a) (c)

(b) (d)

Figure 2 (a–d) Permeate flux vs time.

Table 3 Retention coefficients (%R) of potassium.

Run TMP (bar) Temperature (�C) Flow rate (L/min) %R (potassium) experimental

1 (�1) (�1) (�1) 91.27

2 (+1) (�1) (�1) 86.36

3 (�1) (+1) (�1) 84.96

4 (+1) (+1) (�1) 79.86

5 (�1) (�1) (+1) 88.76

6 (+1) (�1) (+1) 86.98

7 (�1) (+1) (+1) 88.86

8 (+1) (+1) (+1) 88.13

9 (�1) (0) (0) 91.46

10 (+1) (0) (0) 90.03

11 (0) (�1) (0) 92.26

12 (0) (+1) (0) 93.31

13 (0) (0) (�1) 90.06

14 (0) (0) (+1) 90.55

15 (0) (0) (0) 94.96

16 (0) (0) (0) 95.12

17 (0) (0) (0) 94.32

18 (0) (0) (0) 92.23

19 (0) (0) (0) 94.19

20 (0) (0) (0) 92.23

710 I.M. Atadashi et al.

using Design Expert software to determine the curvatureeffects, higher degree polynomial equations for instance qua-dratic and two factor interactions (2FI) models, etc. were fitted

with the data obtained. Based on ANOVA analysis, the exper-imental data were best fitted using quadratic model [2]. TheANOVA was performed to ascertain its significance. Conse-

Table 4 ANOVA for the response surface model.

Source Retention of potassium (K)

F-value p-Value Remarks

Model 19.86 <0.0001 Significant

X1-TMP 9.45 0.008 Significant

X2-Temp 5.36 0.032 Significant

X3-Flow rate 5.63 0.0288 Significant

X1X2 0.052 0.8244 InSignificant

X1X3 3.95 0.0751 InSignificant

X2X3 11.99 0.0042 Significant

X12 15.12 0.0019 Significant

X22 0.94 0.3557 InSignificant

X32 19.69 0.0007 Significant

Lack of fit: 1.14 0.4437 Not significant

Removal of residual palm oil-based biodiesel catalyst using membrane ultra-filtration technique: An optimization study 711

quently, the designed experiments were investigated viaANOVA at 95% level of confidence.

3.3.1. The quadratic model for the membrane system

The quadratic models to predict the retention of potassium (K)in terms of coded factors is given in Eq. (3):

%R3ðK3Þ ¼ þ93:72� 1:39X1 � 1:05X2 þ 1:08X3

þ 0:11X1X2 þ 0:94X1X3 þ 1:76X2X3

� 2:82X21 � 0:78X2

2 3:56X23 ð3Þ

where %R is a function of transmembrane pressure (X1), tem-

perature (X2) and flow rate (X3). The negative sign indicatesantagonistic effect whereas the positive sign in front of theterms shows synergistic effect. The analysis of variance

(ANOVA) was performed to investigate the fitness and signif-icance of the predictive model. The analysis of variance showsthe influence of each variable and their interactions on the

removal of catalyst. As shown in Table 4, the analysis of var-iance on the basis of small p-value which is less than 0.0001,showed that the predictive model is very significant. The pvalue is the probability value employed to determine the signif-

Actual

Pred

icte

d

Predicted vs. Actual

75.00

80.00

85.00

90.00

95.00

100.00

75.00 80.00 85.00 90.00 95.00 100.00

Figure 3 Predicted and experimental values.

icance of each of the coefficient which may invariably show the

pattern of the relationship between the variables [2]. The p-val-ues from the ANOVA tables showed that all the linear termswere statistically significant. Furthermore, to minimized error

all coefficients were considered in the design. The significanceof the result being obtained is judged by the closeness of itsp-value to zero (0.00). For the effects to be statistically signif-icant, the confidence level should be 95%, this shows that the

p-value is less than or equal to 0.05 [33]. The significancemeans that a value that would be achieved from experimentalnoise is smaller than the approximated value of the variable

coefficient [34]. It can also be observed that the statistical anal-ysis of variance shows the overall model p-values to be lowerthan 0.0003, which presents high significance.

Moreover, testing of lack of fit generated p-value greaterthan 0.01 which signifies that the model was able to explainall the data well. Lack of fit F-test was used to test the ade-quacy of the model generated [34]. Since the probability value

was more than 0.05, the lack of fit of the quadratic model wasnot statistically significant. This implies that the model equa-tion has almost accounted all the variations that might have

existed. It is worthy to state that lack of fit is the differenceof the data around the fitted model. A higher order model ora transformation is required when a significant lack of fit is

obtained; this perhaps indicates that the model is not well fit-ted with the design points. Furthermore, for the reliability ofthe model to be ascertained, the data generated should be

Figure 4a Plots for response surface and contour presenting the

effects of temperature (�C) TMP (bar) on the retention of K by

biodiesel membrane separation: (a) response surface 3D and (b)

contour plot (2D).

712 I.M. Atadashi et al.

predicted with reasonable accuracy by the model compared tothe experimental data [2]. Fig. 3 presents predicted and exper-imental values for the retention of K (potassium) using the

developed model equation. The Fig. 3 showed that the modelwas able to successfully and adequately capture the correlationbetween the system variables and the response, %R (K).

Also, to validate the model, coefficient of determination(R2) was evaluated to determine the goodness of fit of themodel. It was reported that for a good agreement between pre-

dicted and experimental data to be achieved, values greaterthan 80% should be achieved [33]. In this work higher coeffi-cient of determination (R2 = 0.9343) for potassium wasobtained. This indicates that the agreement between the exper-

imental values and the predicted values by the models is good[33]. The coefficient of variation (CV) obtained was 1.48%,lower values of the coefficient of variation (CV) ranging from

Figure 4b Plots for response surface and contour presenting the effe

biodiesel membrane separation: (a) response surface 3D and (b) conto

0.43% to 1.59% suggested good precision and reliability of theexperiments [35,36].

3.3.2. Effects of system variable on the retention of potassium

The ANOVA Table 4 indicates that the linear terms, trans-membrane pressure (X1), temperature (X2), flow rate X3, inter-action term, X2X3, and quadratic terms, X1

2, and X32 were

statistically significant for the retention of catalyst (K) sincethe p-values for these terms are less than 0.05. Even thoughit is good to attempt the removal of the insignificant terms,

however the terms were allowed to remain in order to improvethe lack of fit p-value of the model. Figs. 4a–c illustrate the twoand three dimensional surface plots of predicted K retention.

The figures showed that the retention of K is affected byincreasing the TMP, temperature and flow rates. Nevertheless,at very high temperature and TMP, the retention of K is low.

cts of TMP (bar) and flow rate (L/min) on the retention of K by

ur plot (2D).

Figure 4c Plots for response surface and contour presenting the effects of temperature (�C) flow rate (L/min) on the retention of K by

biodiesel membrane separation: (a) response surface 3D and (b) contour plot (2D).

Table 5 Constrains for the parameters and responses in

numerical optimization.

Parameters Ultimate goal Experimental region

TMP (bar) In range 1–3

Temp. (�C) In range 30–50

Flow rate (L/min) In range 60–150

Potassium retention (%) In range 79.9–96.1

Removal of residual palm oil-based biodiesel catalyst using membrane ultra-filtration technique: An optimization study 713

It was observed by Wang et al. [27] that fatty acid is trans-

formed into soaps formation due to combination of cationssuch as calcium, magnesium, potassium and sodium with thecarboxyl anion. The soap molecule formed between fatty acid

group with potassium and sodium is less hydrophobic than thesoap formed with divalent ion such as calcium. Hence, due tothe increasing miscibility with temperature and the possibilityof size distribution of the membrane, not all polar molecules

were cut-off by the membrane.

3.4. Optimization

There are numerous techniques available for optimizing achemical processes. Based on the models predicted, numericalhill-climbing algorithms were employed to search for the most

desirable outcome [37]. A comprehensive and up-to-datedescription of the most effective methods in continuous opti-mization was provided by numerical optimization. The knowl-

edge of process optimization has played a vital role inaddressing problems especially in the field of business, science,and engineering which focuses on the techniques that are mostsuited to practical problems [33]. Furthermore, system

variables and response (potassium) with respect to high and

low levels that satisfy the conditions defined for the optimumconditions are depicted in Table 5. Hence the optimizationwas done based on the levels of system variables and responses

generated; thus the optimum conditions are presented inTable 6. At the optimum point, the comparison between thepredicted and experimental values for the catalyst retentionoffers a reasonable accuracy with percentage error of

0.081%. Consequently, this study has proven that membranetechnology can be employed to post treat crude biodiesel, inorder to achieve high-quality biodiesel fuel that can be success-

fully used on diesel engines.

Table 6 Overall optimization results and model evaluation.

TMP (bar) Temp. (�C) Flow rate (L/min) %R (potassium (K)) Predicted experimental

2 40 105 93.718 93.642

714 I.M. Atadashi et al.

In addition good agreement between experimental and pre-dicted results validated the model and the existence of the opti-mum conditions. Also the p-values presented in the ANOVA

table have justified the accuracy of the model. Hence, theresults achieved demonstrated that Response Surface Method-ology with suitable design of experiment can be effectivelyapplied for the optimization of the system variables in a refin-

ing process. As a result this study was centered on the optimi-zation via RSM for the refining of crude biodiesel usingmembrane technology. The optimization route can proffer

valuable information pertaining to the improvement of eco-nomic and efficient process for the refining of crude biodieselvia membrane processes.

3.5. Process of membrane clean-up

The process of membrane cleaning was quite efficient and fastusing 1% NaOH solution at 70 �C which was circulated for

45 min and thereafter washed with warm deionized water.The cleaning of the membrane is equally important as theultra-filtration process itself, since it is essential in the analysis

of the technical and economic viability of the process on anindustrialized scale, where repeatability and effectiveness arereasonably significant [27]. Thus, to ascertain the performance

of the membrane, after each membrane cleaning process iscompleted; the membrane is run with distilled water to deter-mine the permeate fluxes. The permeate fluxes generated in

Fig. 5 signified that the performance of the membrane waswholly maintained.

3.6. Membrane resistibility to degradation

A fundamental concern when dealing with high base or acidcatalyst concentration is the life of the ceramic membrane usedduring the membrane purification process. The ceramic

Time (min)0 10 20 30 40 50 60 70

Perm

eate

flux

(kg/

m2 hr

)

0

20

40

60

80

100

120

TMP 1 barTMP 2 barTMP 3 bar

Figure 5 Permeate flux vs time using distilled water at flow rate

of 150 (L/min) and temperature of 60 �C.

membrane used in our experiments was able to resist the highbase and acid environments. Biodiesel also presents verystrong solvent qualities [24]. After several months of operation

and contact with different basic and acidic solutions, no clearevidence of degradation of the membrane was detected. Theabsence of defects on the membrane used indicated its suitabil-ity for the refining of crude biodiesel.

4. Conclusion

This study was mainly carried out to develop a membrane sys-

tem that can yield biodiesel that meets the international stan-dards (ASTM D6751 and EN 14214) specification in termsof residual catalyst content. Thus, the ceramic membrane sys-

tem developed was able to provide high-quality biodiesel withresidual catalyst content in biodiesel reduced from 8.328 mg/Lto 0.312 mg/L; this value is well below the value specified by

both ASTM D651 and EN 14214 standards. The ResponseSurface Methodology (RSM) was used for the optimizationof the system variables of membrane post treatment of crude

biodiesel. Three variables, flow rate, temperature and TMPwere found to be significant on the catalyst retention via mem-brane system. The best coefficient of retention (%R) of thecontaminant was 93.642. At the optimum conditions, compar-

ison between predicted and experimental values for the reten-tion of catalyst showed a reasonable accuracy with apercentage error of 0.081% obtained. The results of the pres-

ent study showed that the membrane post treatment of crudebiodiesel could be environmentally benign.

Acknowledgments

The authors wishes to immensely thank Adamawa State Uni-

versity, Mubi-Nigeria and the University of Malaya, Malaysia,for their support to this research work.

References

[1] K. Jacobson, R. Gopinath, L.C. Meher, A.K. Dalai, Solid acid

catalyzed biodiesel production from waste cooking oil, Appl.

Catal. B 85 (2008) 86–91.

[2] Saeid Baroutian, Kaveh Shahbaz, Farouq Sabri Mjalli, Inas

Muen Alnashef, M.A. Hashim, Adsorptive removal of residual

catalyst from palm biodiesel: application of response surface

methodology, Hemijska industrija 66 (3) (2012) 373–380.

[3] M. El_Kassaby, M.A. Nemit_allah, Studying the effect of

compression ratio on an engine fueled with waste oil produced

biodiesel/diesel fuel, Alexandria Eng. J. 52 (2013) 1–11.

[4] F. Ma, M.A. Hanna, Biodiesel production: a review, Bioresour.

Technol. 70 (1999) 1–15.

[5] M. Canakci, J.V. Gerpen, Biodiesel production by acid catalysis,

Am. Soc. Agric. Eng. 42 (1999) 1203–1210.

[6] M. Balat, H. Balat, A critical review of bio-diesel as a vehicular

fuel, Energy Convers. Manage. 49 (2008) 2727–2741.

[7] A. Demirbas, Biodiesel fuels from vegetable oils via catalytic

and non-catalytic supercritical alcohol transesterifications and

Removal of residual palm oil-based biodiesel catalyst using membrane ultra-filtration technique: An optimization study 715

other methods: a survey, Energy Convers. Manage. 44 (2003)

2093–2109.

[8] A. Demirbas, Progress and recent trends in biodiesel fuels,

Energy Convers. Manage. 50 (2009) 14–34.

[9] K. Shahbaz, F.S. Mjalli, M.A. Hashim, I.M. AlNashef, Eutectic

solvents for the removal of residual palm oil-based biodiesel

catalyst, Sep. Purif. Technol. 81 (2011) 216–222.

[10] Z. Helwani, M.R. Othman, N. Aziz, W.J.N. Fernando, J. Kim,

Technologies for production of biodiesel focusing on green

catalytic techniques: a review, Fuel Process. Technol. 90 (2009)

1502–1514.

[11] H. Fukuda, A. Kondo, H. Noda, Biodiesel fuel production by

transesterification of oils, J. Biosci. Bioeng. 92 (2001) 405–416.

[12] S. Semwal, A.K. Arora, R.P. Badoni, D.K. Tuli, Biodiesel

production using heterogeneous catalysts, Bioresour. Technol.

102 (2011) 2151–2161.

[13] D.Y.C. Leung, X. Wu, M.K.H. Leung, A review on biodiesel

production using catalyzed transesterification, Appl. Energy 87

(2010) 1083–1095.

[14] M. Berrios, R.L. Skelton, Comparison of purification methods

for biodiesel, Chem. Eng. J. 144 (2008) 459–465.

[15] I.M. Atadashi, Design and construction of a pilot plant for the

production of biodiesel from cottonseed oil, M.Sc Thesis, 2007.

[16] B. Cooke, S. Abrams, B. Bertram, Purification of biodiesel with

adsorbent materials, US Patent 0509959P, 2003.

[17] P. Cao, M.A. Dube, A.Y. Tremblay, High-purity fatty acid

methyl ester production from canola, soybean, palm, and yellow

grease lipids by means of a membrane reactor, Biomass

Bioenergy 32 (2008) 1028–1036.

[18] M.A. Dube, A.Y. Tremblay, J. Liu, Biodiesel production using a

membrane reactor, Bioresour. Technol. 98 (2007) 639–647.

[19] J. Van Gerpen, B. Shanks, R. Pruszko, D. Clements, G. Knothe,

Biodiesel Production Technology, NREL/SR-510-36244, 2004.

[20] L.Y. Jiang, Jia Ming Zhu, Separation technologies for current

and future biorefineries-status and potential of membrane-based

separation: Membrane-based separation for biorefinery-status

and potential, Wiley Interdisciplinary Reviews Energy and

Environment, 2 (2013) 673–690.

[21] G.Q. Lu, J.C. Diniz da Costa, M. Dukec, S. Giessler, R.

Socolowe, R.H. Williams, T. Kreutze, Inorganic membranes for

hydrogen production and purification: a critical review and

perspective, J. Colloids Interface Sci. 314 (2007) 589–603.

[22] C. Catherine, Membrane processes in biotechnology: an

overview, Biotechnol. Adv. 24 (2006) 482–492.

[23] K.A. Avinash, Biofuels (alcohols and biodiesel) applications as

fuels for internal combustion engines, Prog. Energy Combust.

Sci. 33 (2007) 233–271.

[24] J. Saleh, M.A. Dube, A.Y. Tremblay, Separation of glycerol

from FAME using ceramic membranes, Fuel Process. Technol.

92 (2011) 1305–1310.

[25] M. Gomes, A.A. Pedro, C. Nehemias, Biodiesel production

from degummed soybean oil and glycerol removal using ceramic

membrane, J. Membr. Sci. 378 (2011) 453–461.

[26] J. Saleh, A.Y. Tremblay, M.A. Dube, Glycerol removal from

biodiesel using membrane separation technology, Fuel 89 (2010)

2260–2266.

[27] Y. Wang, X. Wang, Y. Liu, S. Ou, Y. Tan, S. Tang, Refining of

biodiesel by ceramic membrane separation, Fuel Process.

Technol. 90 (2009) 422–427.

[28] M.C.S. Gomes, N.C. Pereira, S.T.D.d. Barros, Separation of

biodiesel and glycerol using ceramic membranes, J. Membr. Sci.

352 (2010) 271–276.

[29] M. Zabeti, W.M.A.W. Daud, M.K. Aroua, Optimization of the

activity of CaO/Al2O3 catalyst for biodiesel production using

response surface methodology, Appl. Catal. A 366 (2009) 154–

159.

[30] C. Catherine, Membrane processes in biotechnology: an

overview, Biotechnol. Adv. 24 (2006) 482–492.

[31] D. Ayhan, Progress and recent trends in biodiesel fuels, Energy

Convers. Manage. 50 (2009) 14–34.

[32] A. de Jesus, M.M. Silva, M.G.R. Vale, The use of

microemulsion for determination of sodium and potassium in

biodiesel by flame atomic absorption spectrometry, Talanta 74

(2008) 1378–1384.

[33] P. Tamunaidu, S. Bhatia, Catalytic cracking of palm oil for the

production of biofuels: optimization studies, Bioresour.

Technol. 98 (2007) 3593–3601.

[34] Y.M. Sani, W.M.A.W. Daud, A.R. Abdul Aziz, Solid acid-

catalyzed biodiesel production from microalgal oil––the dual

advantage, J. Environ. Chem. Eng. 1 (2013) 113–121.

[35] D.C. Montgomery, Design and Analysis of Experiment, fifth

ed., Wiley Inc., New York, USA, 1997.

[36] S. Baroutian, M.K. Aroua, A.A.A. Raman, N.M.N.

Sulaiman, Methanol recovery during transesterification of

palm oil in a TiO2/Al2O3 membrane reactor: experimental

study and neural network modeling, Sep. Purif. Technol. 76

(2010) 58–63.

[37] S. Baroutian, M.K. Aroua, A.A.A. Raman, N.M.N. Sulaiman,

A packed bed membrane reactor for production of biodiesel

using activated carbon supported catalyst, Bioresour. Technol.

102 (2011) 1095–1102.