Fuel Processing Technology 133 (2015) 152–160

Contents lists available at ScienceDirect

Fuel Processing Technology

j ourna l homepage: www.e lsev ie r .com/ locate / fuproc

Biodiesel production from mixtures of waste fish oil, palm oil and wastefrying oil: Optimization of fuel properties

Vanessa F. de Almeida a, Pedro J. García-Moreno b,⁎, Antonio Guadix b, Emilia M. Guadix b

a College of Chemical Engineering, State University of Campinas — UNICAMP, PO Box 6066, 13081-970 Campinas, Brazilb Department of Chemical Engineering, University of Granada, 18071 Granada, Spain

⁎ Corresponding author. Tel.: +34 958241329; fax: +3E-mail address: pjgarcia@ugr.es (P.J. García-Moreno).

http://dx.doi.org/10.1016/j.fuproc.2015.01.0410378-3820/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 18 December 2014Received in revised form 29 January 2015Accepted 30 January 2015Available online 13 February 2015

Keywords:BiodieselFish oilPalm oilFrying oilOptimizationFuel properties

The presentwork studies the influence ofwastefish oil, palm oil andwaste frying oil as rawmaterial on biodieselproperties. The experimental planningwas executed through acid esterification (6:1methanol to oil ratio, 1wt.%sulfuric acid, at 60 °C, 1 h) followed by transesterification (9:1methanol to oil ratio, 0.5wt.% sodiumhydroxide, at60 °C for 1 h). Biodiesel samples showed yield higher than 82%, reaching 90% for palm oil (33.3 wt.%) and wastefrying oil (66.7 wt.%) biodiesel. FAME content was higher than 92.3% and had amaximum of 98.5% for waste fishoil (33.3 wt.%) and palm oil (66.7 wt.%) biodiesel. Special cubic models were used to fit experimental data, andwere optimized by response surface methodology and multi-objective optimization. Viscosity (4.3 mm2/s) andCOM (2.5 °C) were minimized when pure fish oil was used as raw material, whereas IP maximum (22.0 h)was found for palm oil biodiesel. Multi-objective optimization evidenced that although the use of the pure oilsas feedstock presented more advantages to biodiesel properties, the waste fish oil (42.1 wt.%) and waste fryingoil (57.9 wt.%) mix is beneficial, if the aim is IP (20%) and COM (80%) improvement.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Biodiesel is defined as mono-alkyl esters of vegetable oils or animalfats, obtained by transesterification of an oil or fat with an alcohol [1].It is a biodegradable andnontoxic biofuel, so is environmental beneficial[2]. That is one of the reasons why biodiesel has received increasingattention, besides the fact of petroleum reserves are diminishing andit is necessary to find other competitive energy sources [3].

Fish oil is recommended in a healthy diet because of its content inomega-3 polyunsaturated fatty acids such as eicosapentaenoic (EPA)and docosahexaenoic (DHA) acids. [4]. However, refining of fish oilextracted from fish wastes would result in low yields because of itshigh content of free fatty acids and oxidation products [5]. Moreover,waste fish oil might have a low amount of EPA and DHA reducing itsapplication on the pharmaceutical and functional food fields [6]. Thus,biodiesel using waste fish oil as feedstock has been recently researched[7–12]. Nevertheless, biodiesel from fish oil has low oxidative stability,mostly due to its high content of polyunsaturated fatty acids (PUFA)containing more allylic methylene positions [13]. Biodiesel oxidationis undesirable because it can increase viscosity and may lead to forma-tion of insoluble species, which can lead to clog fuel lines and pumps[14]. In order to improve this property, antioxidants can be added tobiodiesel [15] or fish oil can be mixed with more stable oils and usedas biodiesel feedstock [16].

4 958248992.

In this sense, palm oil, which is usually employed as rawmaterial forbiodiesel production, has a high resistance to oxidation due to its signif-icant content of saturated fatty acids (SFA) [17]. Hence, it can be anappropriated oil to be mixed with fish oil before transesterification.Furthermore, waste frying oil, which is also more oxidatively stablethan fish oil, is a substantial alternative of feedstock since it is cheapand diminishes the environmental impacts of inappropriate waste oildisposals [18]. Most of the scientific works on oil mixtures as feedstockare devoted to vegetable oils such as rapeseed, soybean and sunflower[19,20]. Some studies are also focused on mixtures of vegetable oilsand animal fats. For instance, Galvan et al. [21] evaluated the oxidativestability of biodiesel obtained from a ternary mixture of soybean oil(50%), beef tallow (20wt.%) and poultry fat (30wt.%). Nevertheless, ex-cept for thework of Costa et al. [16]who carried out a preliminary studyon the production of biodiesel from amixture of waste fish oil (20wt.%)and waste olive oil (80 wt.%), studies evaluating the fuel properties ofbiodiesel obtained from mixtures of waste fish oil and vegetable oilshave not been reported in the literature.

Therefore, the aims of this work were: a) to study, by means ofexperimental design and analysis of variance (ANOVA), the influenceof mixtures of waste fish oil, palm oil and waste frying oil as feedstockon the fuel characteristics of the biodiesel obtained, b) to obtain, bymeans of response surface methodology, the feedstock mixture thatoptimizes viscosity, oxidative stability and cold flow properties of thebiodiesel produced, and c) to determine, by means of multi-objectiveoptimization, the feedstock mixture that optimizes simultaneously thefuel properties of the biodiesel samples.

153V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

2. Materials and methods

2.1. Materials

Fish oil was purchased from Industrias Afines, S.L. (Pontevedra,Spain) and presented the following properties: acid value of 16.4 mg KOH/goil, peroxide value of 2.5 meq O2/kg oil and a composition of 32.7 wt.% inoleic acid, 11.7 wt.% in linoleic acid, 3.6 wt.% in EPA and 5.7 wt.% in DHA.Palm oil was donated by the company Lípidos Santiga S.A. (Barcelona,Spain) and presented an acid value of 16.3 mg KOH/g oil, peroxide value of2.1 meq O2/kg oil and a composition of 44.5 wt.% in palmitic acid and38.6wt.% inoleic acid.Waste fryingoilwas collected fromlocal domestic con-sumers (Granada, Spain) andhad an acid value of 2.4mgKOH/g oil, peroxidevalue of 4.6meq O2/kg oil and a composition of 55wt.% in oleic acid and26.7 wt.% in linoleic acid. Sodium hydroxide (≥99%), potassiumhydroxide in ethanol (0.1 M), n-heptane (≥99%) and methylheptadecanoate (≥99%) were purchased from Sigma-Aldrich, whilemethanol (≥99.8%) and diethyl ether (≥99%) were acquired fromScharlau. Ethanol (≥99%) and anhydrous sodium sulfate (≥99%) wereprovided by Panreac, whereas sulfuric acid (95%) was purchased fromVWR International. Phenolphthalein (1% in ethanol) was provided byJ. T. Barker.

2.2. Production process

Waste fish oil, palm oil and waste frying oil were used to producebiodiesel. A simplex centroid design with 16 experiments was carriedout. It comprised 16 different oil contents as biodiesel feedstock,including pure oils andmixtures of two and three oils in different ratios.The content in free fatty acids (FFA) of the three oils was higher than2 mg KOH/g oil, what impairs the biodiesel production, since it reducesthe yield because of soap formation. Therefore, a two-step processwas performed [22]. The first step consists in an acid-catalyzed pre-treatment that converts the FFA content into methyl esters, reducingacid value below 2 mg KOH/g oil. In summary, 200 g of oil, methanol(molar ratio methanol to oil of 6:1) and 1 wt.% sulfuric acid as catalystwere placed into a flat-bottomed flask. The samples were stirred at300 rpm and kept at 60 °C for 60 min, using a hot plate controlled bya thermo regulator. After the reaction ending, the reactor content wastransferred to a separating funnel to settle for 1 h. The bottom phasecontaining mainly unconverted oil and methyl esters was collected tocarry out the next step [13].

In the second step the unreacted oil was submitted to an alkalinetransesterification. In brief, the collected phase from the first step wasintroduced into a flat-bottomed flask with methanol (molar ratiomethanol to oil of 9:1) and 0.5 wt.% sodium hydroxide. The sampleswere also stirred at 300 rpm and kept at 60 °C for 60 min, using a hotplate controlled by a thermo regulator. It should be noted that someamount of the alkali catalyst may have reacted in the neutralization ofthe first step samples, which had an increased acidity.

After the reaction completion, the content of the reactor was pouredinto a separating funnel, where it settled during 1 h. Then, the top phasecontaining mainly methyl esters was collected to be purified by wash-ing, three times, with distillated water at 50 °C [13]. The aim of thisstep is removing impurities, such as unreacted catalyst, soap, glyceroland unreactedmethanol. Finally, sampleswere vacuumfiltered throughanhydrous sodium sulfate and stored under nitrogen in amber bottles at−20 °C until analysis.

2.3. FAME content and yield

Gas chromatography was used to determine the fatty acid methylester (FAME) content in biodiesel samples, using the European standardEN 14103 [23]. Samples were analyzed according to the method fromCamacho-Paez et al. [24] in a gas chromatographer Agilent 7890A(Agilent Technologies, S.A., Santa Clara, California, USA) connected to a

capillary column of fused silica Omegawax (0:25 mm × 30 m, 0:25 μmstandard film; Supelco, Bellefonte, PA) and a flame-ionization detector.Nitrogen was used as the carrier gas, and the total column flow was44 ml/min. The oven temperature was initially set at 150 °C for 3 min,then increased at a rate of 10 °C/min until 240 °C, and then kept atthis temperature during 12 min. Matreya (Pleasant GAP, PA) n-3PUFAs standard (catalogue number 1177) was used for the qualitativefatty acid determination and methyl heptadecanoate was employed asanalytical standard for the quantitative determination. Firstly, theFAME of biodiesel samples was identified and then the peak areaswere utilized to quantify the FAME content, according to Eq. (1):

C ¼X

A� �

− AEI

AEI� CEI � VEI

m� 100: ð1Þ

Where C = fatty acid methyl content (%); ∑A = total peak area;AEI = peak area corresponding to the methyl heptadecanoate; CEI =concentration of methyl heptadecanoate solution in heptane (mg/ml);VEI = volume of methyl heptadecanoate solution (ml); m = mass ofbiodiesel sample (mg).

The biodiesel yield was calculated following the Eq. (2):

Yield ¼ MBiodiesel � CMOil

� 100: ð2Þ

Where MBiodiesel = mass of final biodiesel (g); MOil = mass of oilused as raw material (g); C = fatty acid methyl content descriptedabove (%).

2.4. Kinematic viscosity

Oils and biodiesel viscosities were evaluated using a rotationalviscometer Haake model VT500 and a NV sensor (Fisher Scientific,Aalst, Belgium). The samples were kept at 40 °C using a water bathand 10 ml of each sample was introduced between the two coaxialcylinders, maintaining the specified temperature. Shear stress (τ) wasmeasured using different shear rates (γ), from 0 to 3000 s−1. Then,the dynamic viscosity (μ, Pa.s) was determined by the slope of the lineobtained when plotting shear stress (τ, Pa) versus shear rate (γ, s−1)for all the samples. The density of all the samples was measured at40 °C. Using an electronic pipette, 5 ml of each sample were taken andthenweighted to determine itsmass. The dynamic viscosity and densitywere used to determine the kinematic viscosity at 40 °C, expressedin mm2/s.

2.5. Oxidative stability

The biodiesel oxidative stability was evaluated using the Rancimatmethod, a test of accelerated oxidation. The Metrohm Rancimat model743 (Methrom Instruments, Herisau, Switzerland) apparatus was usedto carry out the measurements as specified in the European standardEN 15751 [25]. A stream of filtered, cleaned and dried air in a rate of10 l/h was bubbled into 7.5 g of biodiesel samples in a glass tube, main-tained at 110 °C. The effluent air containing volatile organic compoundswas bubbled into a vessel containing 60 ml of distillated water wherethe conductivity was continually measured. The induction period (IP)was automatically calculated by the program. It represents the resis-tance time (in hours) of the biodiesel to oxidation.

2.6. Differential scanning calorimetry

Differential scanning calorimetry (DSC) was employed to evaluatethe cold flow properties of the biodiesel produced. DSC heating curveswere used to determine the completion of melt onset temperature(COM), minimal high melting peak temperature (P1) and melting

154 V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

point (MP) of biodiesel samples. The analysis was executed in a DSCapparatus Metler Todelo DSC 1 (Metler Toledo, Barcelona, Spain). Tenmg of biodiesel samples was hermetically sealed in an aluminum pan,and also an empty vessel was used as reference. To determine thebiodiesel transition phases, a ramp of 5 °C/minwas used under nitrogenatmosphere (50 ml/min). All the samples were cooled until −80 °C,kept at this temperature during 10 min, and then heated until 100 °C.From each DSC analysis, it obtained a graphic of heat flow (mW) versustemperature (°C), where it is possible to find the values of COM, P1and MP.

2.7. Acid value

The acid value shows the FFA content of the sample. It was deter-mined for the oils, the products from the first step reaction and finalbiodiesel, according to the European standard EN 14104 [26]. Themeth-od is based on titration of the sample, diluted in ethanol–diethyl ethermixed solvent, with potassium hydroxide solution in ethanolemploying phenolphthalein as indicator to detect the end point. Resultswere calculated according to Eq. (3) and they were expressed inmilligrams of potassium hydroxide per gram of sample (mg KOH/g).

Acid value ¼ 56:1 � V � Cm

ð3Þ

Where V= volume of potassium hydroxide solution (ml); C= con-centration of potassium hydroxide solution (M); m = mass of sample(g).

2.8. Statistical analysis and optimization

Statgraphics Centurion XVI was used to correlate biodiesel proper-ties (yield, viscosity, induction period, COM, P1, MP) to oil content andtype (waste fish oil, palm oil and waste frying oil) in the raw material.Special cubic models were developed by means of response surfacemethodology and were then used to predict biodiesel properties ac-cording to Eq. (4)

Y ¼X3

i¼1biXi þ

X3ib jbi jXiXj þ

X3ib jbk

bi jkXiXjXk: ð4Þ

Where Y = value of the specific property; X = content of oil in therawmaterial (%), bi is related to the linear effect, bij and bijk are relatedto the interaction between the respective oils. The analysis of varianceANOVAwas used to evaluate the biodiesel properties with a confidencelevel of 95% (p b 0.05). The estimated parameters were used to generatecontour plots and to define the feedstock composition that optimizesbiodiesel properties.

2.9. Multi-objective optimization

Multi-objective optimization is required if an issue with severalobjectives is proposed. In this work, the objectives are minimizingviscosity and COM and maximizing IP, which is a controversial issue.The Pareto Front is defined as a set of non-inferior solutions, with theaim of satisfying all the three objectives [27]. The weighted-summethod was used to generate the Pareto Front [28]. This methodconsists of expressing a comprehensive objective function (OBJ) as alinear combination of the individual objectives (viscosity, IP andCOM), by means of weight factors (wi), which quantifies the relativeimportance given to the accomplishment of each individual objective,Eq. (5):

OBJ ¼ w1: viscosityð Þ þw2: −IPð Þ þw3: COMð Þ: ð5Þ

Where w1 + w2 + w3 = 1 and 0 ≤ wi ≤ 1.

It should be noted that the IP contribution is negative in order to setthe problem as aminimization of the objective function. The Solver Toolin MS Excel software was used to execute all the calculations for themulti-objective optimization.

3. Results and discussion

3.1. Characterization of biodiesel samples

Table 1 shows the experimental values of FAME, SFA, MUFA andPUFA content, yield, kinematic viscosity, IP, COM, P1 and MP of thebiodiesel samples.

3.1.1. FAME content and yieldHigh FAME content was found in all biodiesel samples, since the

minimum FAME content was 92.3% for the experiments 5 (50 wt.%palm oil and 50 wt.% frying oil) and 16 (33.3 wt.% fish oil and66.7 wt.% frying oil). The highest FAME content was 98.5%, found forthe experiment 12 (33.3 wt.% fish oil and 66.7 wt.% palm oil). However,just the experiments 12 (33.3 wt.% fish oil and 66.7 wt.% palm oil), 14(33.3 wt.% palm oil and 66.7 wt.% frying oil) and 15 (66.7 wt.% fish oil,33.3 wt.% frying oil) were within the required value by the Europeanstandard EN 14214 (≥96.5%) [29].

The fatty acids methyl esters of the waste fish oil biodiesel werecomposed of 15.0 wt.% saturated fatty acids (SFA), 42.2 wt.% monoun-saturated fatty acids (MUFA) and 26.1wt.% PUFA. The palm oil biodieselhad a considerably higher content of methyl esters of SFA (44.3 wt.%),similar content of methyl esters of MUFA (35.8 wt.%) and lower contentof methyl esters of PUFA (10.0 wt.%). Last, the composition of wastefrying oil biodiesel was of 13.7 wt.% methyl esters of SFA, methyl estersof 51.4 wt.% MUFA and 24.3 wt.% methyl esters of PUFA (Table 1). Thesum of methyl esters of SFA, MUFA and PUFA content in each biodieselsample were not 100 wt.%, because some unrecognized peaks, found atthe biodiesel chromatograms, which prevented all the fatty acids to beidentified. As expected, biodiesel from oil mixtures showed methylesters content of SFA, MUFA and PUFA proportional to the content ofthese fatty acid types in the oils and to the amount of oil employed inthe mixture.

Yield was also high in all experiments, and varied between 82% forthe experiment 16 (33.3 wt.% fish oil and 66.7 wt.% frying oil) and 90%for the experiment 14 (33.3 wt.% palm oil and 66.7 wt.% frying oil)(Table 1). For biodiesel obtained from fish oil a yield of 85% was found(Table 1). García-Moreno et al. [13] obtained a lower yield (79.9%) forbiodiesel from waste fish oil having similar fatty acid composition(17.7 wt.% SFA, 42.8 wt.% MUFA and 33.2 wt.% PUFA) and using thesame operating conditions, except the catalyst content (1 wt.% ofNaOH). Another work also demonstrated that using 0.5 wt.% of NaOHincreases the biodiesel yield, comparing to 1 wt.% of NaOH [30]. Thus,the difference in yield may be related to the catalyst content, which inexcess, contributes to soap formation. The yield found for biodieselobtained from palm oil (87.1%) was in the line of previous works.Hayyan et al. [31] studied the production of biodiesel from sludgepalm oil using an acid esterification process (0.75 wt.% of sulfuric acid,molar ratio of methanol to oil of 8:1, 60 min reaction time at 60 °C)previous to the transesterification. Despite these authors employeddifferent transesterification operating conditions (1 wt.% KOH, molarratio ofmethanol to oil of 10:1, 60min of reaction at 60 °C), they obtain-ed a similar biodiesel yield (83.7%). It should be noted that the fatty acidcomposition of the sludge palm oil employed (47.2 wt.% SFA and52.8 wt.% unsaturated fatty acids) is similar to the palm oil used inthis study (50.3 wt.% SFA and 49.7 wt.% unsaturated fatty acids). Uzunet al. [32] obtained the highest yield (96.0%) for biodiesel obtainedfrom waste frying oil by one-step transesterification process (0.5 wt.%of NaOH, methanol to oil molar ratio of 7.5:1, 30 min of reaction at50 °C). This value is considerably higher than the yield (87.6%) obtainedin this work for experiment 2 (100% waste frying oil). Besides the

Table 1Experimental design and measured values for the response variables.

Exp# Fish oil(%)

Palm oil(%)

Frying oil(%)

FAME(%)

SFA(%)

MUFA(%)

PUFA(%)

Yield(%)

Viscosity(mm2/s)

IP(h)

COM(°C)

P1(°C)

MP(°C)

1 0.0 100.0 0.0 94.8 44.3 35.8 10.0 87.1 5.6 22.9 21.0 19.0 −54.22 0.0 0.0 100.0 96.4 13.7 51.4 24.3 87.6 5.3 6.1 3.7 −35.1 −56.43 100.0 0.0 0.0 95.1 15.0 42.2 26.1 85.0 4.2 1.6 2.8 −37.4 −67.54 50.0 0.0 50.0 94.5 14.5 47.4 25.3 84.3 5.5 3.9 3.2 −37.4 −64.45 0.0 50.0 50.0 92.3 29.5 43.5 17.0 85.4 5.5 14.7 14.1 −34.2 −56.66 50.0 50.0 0.0 93.4 30.1 39.5 18.1 82.6 5.4 7.0 14.6 −37.1 −65.87 16.7 66.7 16.7 94.4 35.0 39.8 15.1 84.6 5.4 15.3 17.1 −35.0 −60.78 16.7 16.7 66.7 93.0 19.4 47.7 22.3 84.0 5.5 9.3 7.5 −35.3 −60.89 66.7 16.7 16.7 94.9 19.8 43.1 23.1 84.8 5.3 5.3 7.3 −38.7 −66.910 33.3 33.3 33.3 94.7 24.7 43.4 20.1 82.2 5.4 9.1 11.0 −37.0 −62.611 66.7 33.3 0.0 96.0 25.1 40.7 20.7 88.2 5.2 5.6 8.5 −38.6 −67.612 33.3 66.7 0.0 98.5 35.2 38.4 15.3 87.3 5.5 10.8 17.0 −37.5 −63.613 0.0 66.7 33.3 95.2 34.7 40.8 14.6 88.0 5.6 12.9 16.8 −34.1 −56.214 0.0 33.3 66.7 97.0 24.4 46.2 19.4 90.0 5.7 12.1 11.2 −34.7 −56.815 66.7 0.0 33.3 98.1 14.5 45.9 25.6 84.7 5.5 3.0 2.6 −38.7 −67.016 33.3 0.0 66.7 92.3 14.2 48.8 25.0 82.0 5.5 5.1 2.9 −36.4 −62.7

155V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

operating conditions, the yield differences are probably associated tothe differences in the MUFA and PUFA content of the frying oil usedby these authors (36.7wt.%MUFA and 52.2wt.% PUFA)when comparedto the waste frying oil employed in this study (56.6 wt.% MUFA and26.6wt.% PUFA). Furthermore, while these authors allowed the reactionmixture to separate overnight, in thiswork the period of phase's separa-tion lasted 1 h.

3.1.2. ViscosityViscosity is a key property for biodiesel, since high viscosity can

cause operational problems such as engine deposits [33]. Waste fishoil, palm oil and waste frying oil had viscosity at 40 °C of 33.7 mm2/s,56.1 mm2/s and 45.2 mm2/s, respectively. After transesterification,waste fish oil, palm oil and waste frying oil biodiesels had viscosity of4.2 mm2/s, 5.6 mm2/s and 5.3 mm2/s, respectively (Table 1). Viscosityreduction demonstrates that the transesterification process waseffective.

Table 1 shows that kinematic viscosity of the biodiesels obtainedranged from 4.2 mm2/s for experiment 3 (100% fish oil) to 5.7 mm2/sfor experiment 14 (33.3wt.% palmoil, 66.7wt.% frying oil). The viscosityvalue found for experiment 14 was not expected, since it was higherthan palm oil and waste frying oil biodiesel viscosity. It should benoticed that, although the biodiesels obtained were close to meet theEuropean standard EN 14214 (3.5–5 mm2/s at 40 °C) [24], only theviscosity of the biodiesel produced using 100% fish oil as feedstock(4.2 mm2/s) was in the range required.

Viscosity increases with chain length and decreases with thenumber of double bonds in the chain [33]. Therefore, the lower viscosityfound for fish oil biodiesel (4.2 mm2/s) can be explained by its highcontent in PUFA (26.1 wt.%), comparing to the biodiesel from palm oil(10.0 wt.% PUFA) and waste frying oil (24.3 wt.% PUFA) (Table 1).Moreover, cis double bonds present lower viscosity than trans doublebonds, which shows similar viscosity to a correspondent saturatedcompound [33]. Waste frying oil is usually partially hydrogenated andcontains higher amounts of trans fatty acid chains [33]. This explainsthe fact that frying oil biodiesel, with contents in SFA, MUFA and PUFAsimilar to fish oil biodiesel, had a higher viscosity (5.3 mm2/s)(Table 1). The higher viscosity for palm oil biodiesel (5.6mm2/s) is asso-ciated to its high content in SFA (44.3%) (Table 1).

The viscosity values obtained in this work are in the line of previousreported studies. For instance, García-Moreno et al. [13] found for fishoil biodiesel (17.7 wt.% SFA and 33.2 wt.% PUFA), produced under thesame process conditions (except 1 wt.% NaOH), viscosities that variedfrom 6.05 mm2/s to 6.66 mm2/s at 30 °C. These values are higher thanfound in this work (4.2 mm2/s) for waste fish oil biodiesel (15.0 wt.%SFA and 26.1 wt.% PUFA) at 40 °C. Taking into account that viscositydecreases with temperature increase and that fish oil composition is

similar, the difference in viscositymay be reduced if both fish oil biodie-sel viscositieswere compared at 40 °C. Sarin et al. [34] found for palm oilbiodiesel a viscosity at 40 °C of 4.5 mm2/s, very similar to the valuefound in this work (4.4 mm2/s) for the same biodiesel. Although theydid not show the fatty acid composition of biodiesel, the palm oil usedas feedstock has high content of saturated fatty acids (43.4 wt.% SFA),as palm oil used in this study (50.3 wt.% SFA), which may explain theclose viscosities achieved. For waste cooking oil biodiesel (10.1 wt.%SFA and 55.7 wt.% PUFA), Uzun et al. [32] found a viscosity value of4.4 mm2/s at 40 °C, lower than the viscosity value found in this work(5.3 mm2/s) for the biodiesel obtained from waste frying oil(13.7 wt.% SFA and 24.3 wt.% PUFA). This is explained by the differencein the PUFA content between both oils [33].

3.1.3. Oxidative stabilityOxidative stability of biodiesel determines its fuel quality, especially

when storing biodiesel for a long period of time. The fatty acid compo-sition of biodiesel is the main factor affecting its oxidative stability[14]. Rancimat induction period (IP) was assayed as a measure of bio-diesel resistance to oxidation. The lowest value of IP (1.6 h) was foundfor fish oil biodiesel, while the highest value of IP (22.9 h) was obtainedfor palm oil biodiesel (Table 1).

Similar values of IP for fish oil biodiesel were obtained by García-Moreno et al. [13]. These authors obtained IP values ranging from1.05 h to 2.22 h. In the case of palm oil biodiesel, Sarin et al. [34]found an IP value of 13.4 h, lower when comparing to the IP achievedin this work for the same biodiesel (22.9 h), but much greater thanthe required for the standard EN 14214 [29]. This is attributed to thedifferent fatty acid compositions of the palm oil employed as feedstock.These researchers used palm oil with 40.3 wt.% palmitic acid and43.4 wt.% oleic acid, whereas palm oil used in this study had 44.5 wt.%in palmitic acid and 38.6 wt.% in oleic acid. In general terms, it wasobserved that while IP decreases with the content of fish oil in the feed-stock, IP increases with the content of palm oil. This is explained due toadding palm oil (50.3 wt.% SFA, 38.6 wt.% MUFA and 11.1 wt.% PUFA)increased the SFA content in the mixture, improving is resistance tooxidation [33].

For biodiesel produced from waste frying oil, an IP of 6.1 h wasobtained. This value is considerably higher than the IP (0.9 h) foundby Uzun et al. [32], who produced waste frying oil biodiesel with ahigher unsaturated fatty acid content (89.8 wt.%) in comparison to theproduced in this study (75.7 wt.%). The mixtures of waste frying oil(16.8 wt.% SFA, 56.6 wt.% MUFA and 26.7 wt.% PUFA) and waste fishoil (16.2 wt.% SFA, 46.6 wt.% MUFA and 37.2 wt.% PUFA) produced afuel that showed IP values higher than pure waste fish oil biodiesel.Both oils have similar content of unsaturated fatty acids (about83.6 wt.%). However, waste fish oil had more polyunsaturated fatty

156 V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

acids, which are more unstable, because of the bis-allylic positions. Thisfact explains why fish oil biodiesel is more susceptible to oxidation andthe mix of waste fish oil and waste frying oil as biodiesel feedstockincreases the biodiesel oxidation resistance.

It should be noticed that most IP values are up the Europeanstandard EN 14214 specifications (minimum IP of 8 h) (Table 1) [29].Thus, antioxidant addition is not necessary.

Fig. 1. DSC heating curve of biodiesel: (a) Exp. 1, (b) Exp. 9 and (c) Exp. 10.

3.1.4. Cold flow propertiesBiodiesel often presents poor cold flow properties, due to the high

melting point ofmethyl esters of SFA found in biodiesel [35]. DSC curveswere analyzed to determine COM, P1 and MP for the 16 biodieselsamples (Table 1).

From 3 to 5 distinct peaks were found at the DSC curves (Fig. 1). Thefirst peak varied from−60.54 °C for the experiment 11 (66. 7 wt.% fishoil and 33.3 wt.% palm oil) to −50.98 °C for the experiment 1 (100%palm oil) and it may be associated to the methyl esters of PUFA. Thelast peak varied from −4.88 °C for the experiment 8 (16.7 wt.% wastefish oil, 16.7 wt.% palm oil and 66.6 wt.% frying oil) to 18.99 °C for theexperiment 1 (100% palm oil) and it is associated to the methyl estersof SFA.

The number of peaks between the first and the last peaks variedfrom 1–3 (intermediated peaks) (Fig. 1). They are associated to thecontent of methyl esters of unsaturated fatty acids. There may bemore than one intermediated peak when the methyl esters of unsatu-rated fatty acids have considerably different melting point tem-peratures. Otherwise, there would be just one larger peak. Theintermediated peaks ranged between −50.05 °C for experiment 15(66.7 wt.% fish oil and 33.3 wt.% frying oil) to−23.58 °C for experiment1 (100% palm oil). For the experiments 1, 2, 5, 7, 13 and 14, threeintermediated peaks were found, being one of them an exothermicpeak (Fig. 1a). According to Foon et al. [36], the exothermic peak occursdue to a simultaneousmelting and crystallization of the palm oil methylesters in the samples. The fatty acid methyl esters that are alreadymelted start to re-crystallize in the crystals that have higher MP, andthen at a superior temperature they all melt together. Therefore, there-crystallization may cause the exothermic peaks found in some DSCheating curves.

COM is defined as the temperature at which the melting phenome-non ends, varied from 2.6 °C for the experiment 15 (66.7 wt.% fish oiland 33.3 wt.% frying oil) to 21 °C for the experiment 1 (100% palmoil). For biodiesels obtained from pure oils (experiments 1–3), COMvalue decreases (21.0, 3.7 and 2.8 °C) with the polyunsaturatedcompounds content (10.0, 24.3 and 26.1 wt.%, respectively) (Table 1).García-Moreno et al. [13] found for waste fish oil biodiesel COM valuesthat varied from 3.31 °C to 3.83 °C, which are in the range of the valuefound in this work for waste fish oil biodiesel (2.8 °C). The similarcontent in SFA explains the result accordance. Foon et al. [36] foundfor palm oil biodiesel (49.41 wt.% SFA) a DSC heating curve with thesame profile to the achieved here, with four endothermic peaks andone exothermic peak. The COM value was 24.7 °C, very close to theCOM value for palm oil biodiesel (44.3 wt.% SFA) determined in thiswork (21.0 °C). The superior COM value found for these authors isprobably related to the slight higher SFA content in palm oil biodiesel.

P1 values were around −36.5 °C, in exception of P1 value for thebiodiesel obtained from only palm oil (19 °C) (Table 1). Foon et al.[36] found for palm oil biodiesel P1 value of −24.32 °C. The P1 valuedifference may be due to the content of SFA methyl esters that melt inhigher temperature, which seems to be greater in the palm oil biodieselused in this work, and so is expressed for a higher peak at the DSCheating curve. MP values ranged from −67.6 °C for experiment 11(66.7 wt.% fish oil and 33.3 wt.% palm oil) to −54.1 °C for the experi-ment 1 (100% palm oil) (Table 1). Foon et al. [36] found MP value forpalm oil biodiesel (49.41 wt.% SFA) of −61.55 °C, similar to that foundin this study (−54.1 °C) for the palm oil biodiesel (44.3 wt.% SFA).

157V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

3.1.5. Acid valueThe acid value is a measure of the content of free fatty acids in the

biodiesel and it determines its corrosiveness and long-term stability.All the biodiesel samples are in accordance to the European standardEN 14214 (b0.5 mg KOH/g) [29], having acid values lower than0.39 mg KOH/g of biodiesel.

3.2. Statistical modeling and optimization of biodiesel properties

Special cubic models were selected to relate the content of fish oil,palm oil and frying oil in the raw material to yield, viscosity, IP, COM,P1 andMP values. Table 2 shows the polynomial coefficients, calculatedbymultiple regressions for each response variable, as well as the associ-ated p-values and R2 for each model obtained. All the models werestatistically significant (p b 0.05), except the model found for yield(p = 0.2325). Furthermore, it should be noted that, apart for yield, allthe models explain the experimental data to a large extent with coeffi-cients of determination (R2) higher than 0.8616, reaching 0.9935 forCOM.

Viscosity, IP and COM were chosen as most important output vari-ables indicating the fuel properties of the biodiesels. Thus, they wereoptimized using response surface methodology. Firstly, the goodnessof the fit of the models obtained for these variables was also provedby plotting measured values against the predicted ones, for viscosity(Fig. 2a), IP (Fig. 2b) and COM (Fig. 2c). A regression line correlatedthe data, whose equation and coefficient of determination wereinserted in each figure. A high coincidence was observed between thisregression line and the diagonal. Furthermore, almost all points wereinside the region delimited by the dotted lines, which show a deviationof ±10% between predicted and measured values (Fig. 2). Both factsindicate a high correlation.

Afterwards, through the polynomial equations obtained and usingresponse surface methodology, contour maps were generated whereviscosity, IP and COM values were determined by the raw materialcomposition in waste fish oil, palm oil and waste frying oil (Fig. 3a–c).For biodiesel properties of improvement, viscosity and COM were setto minimize, whereas IP was set to be maximum. Viscosity influencesthe atomization of a fuel and upon injection into the combustionchamber and thus the formation of engine deposits. The higher theviscosity, the greater is the appearance of such problems [37]. COM isthe temperature inwhich the fatty acidmethyl esters start to crystallize.The crystals rapidly grow and agglomerate, clogging fuel lines andfilters[37]. Therefore, COMwas set to beminimum, so biodiesel can be used atlow temperatures without causing operational problems. The biodieseloxidative stability is also an important issue, mainly when biodiesel isstored during an extended period of time [37]. As IP is a measure of

Table 2Polynomial coefficients, p-values and R2 for the models of the response variables.

Yield(%)

Viscosity(mm2/s)

IP(h)

Coefficient Standarderror

Coefficient Standarderror

Coefficient Standarderror

A: Fish oil,%

8.62E-01 1.9138 4.32E-02 0.1232 1.83E-02 1.2824

B: Palm oil,%

8.68E-01 1.9138 5.49E-02 0.1232 2.20E-01 1.2824

C: Fryingoil, %

8.73E-01 1.9138 5.30E-02 0.1232 6.92E-02 1.2824

AB −2.35E-04 8.0178 1.97E-04 0.5163 −1.76E-03 5.3728AC −1.29E-03 8.0178 3.07E-04 0.5163 −1.35E-04 5.3728BC 2.25E-04 8.0178 7.10E-05 0.5163 −4.55E-04 5.3728ABC −7.79E-05 54.1942 −6.35E-06 3.4895 6.18E-05 36.3162p-Value 0.2325 0.0004 0.0000R2 0.5282 0.9057 0.9625

biodiesel oxidative stability and biodiesel requires high resistance tooxidation in order to avoid the formation of high-molecular-weightproducts from polymerization reactions, IP was set to maximize.

The small circles at the diagrams (Fig. 3a–c) are indicating thecomposition that optimizes biodiesel properties. The circle allocated atthe diagram tip (Fig. 3a), which represents 100% fish oil, indicates thatbiodiesel viscosity is minimized (4.3 mm2/s) when pure fish oil isused as feedstock. The bold contour line in Fig. 3a shows the maximumviscosity value at 40 °C specified in EN 14124 (5.0 mm2/s). Fig. 3b indi-cates that biodiesel IP is maximum (22.0 h) when just palm oil isemployed as raw material. A bold contour line was also added toFig. 3b showing in this case the minimum IP value according to EN14124 (8.0 h). At Fig. 3c, the indication is evidencing that pure fish oilas raw material minimizes COM (2.5 °C). The bold contour line in thisfigure shows the minimum COM value required for grade D biodieselin the European standard EN 14214. According to this standard, gradeD biodiesel should have a maximum cloud point of 5 °C, which denotesa COM value lower than 10.2 °C when calculated by Eq. (6) [38]:

COM ¼ 7:06 þ 0:6377� CP: ð6Þ

Where COM: completion of melt onset temperature, °C; and CP:cloud point, °C.

The optimum values found fit well with theminimum experimentalvalues obtained for viscosity (4.2 mm2/s) and COM (2.6 °C) and themaximum for IP (22.9 h). According to Knothe et al. [37], fuel propertiesare strongly influenced by the properties of individual fatty esters inbiodiesel. SFAmethyl ester content increases IP [39], while unsaturatedmethyl ester content decreases viscosity and COM [33,35]. The highestcontent of methyl esters of SFA was found in palm oil, whereas highestcontent of polyunsaturatedmethyl esters in fish oil biodiesel, explainingthe fact that contourmaps indicateminimumviscosity and COM for fishoil biodiesel and maximum IP for palm oil biodiesel.

3.3. Multi-objective optimization

The aim of determining a feedstock composition that optimizesbiodiesel properties (viscosity, IP and COM) simultaneously promptsthe use of multi-objective optimization. Table 3 displays the ParetoFront and the decision space. The Pareto Front was generated by theweight-sum method and then transferred to the decision space, by thedetermination of input variables (waste fish oil, palm oil and wastefrying oil contents) for each selection of weight factors (w1, w2, w3).

Viscosity, IP and COM weight factors were w1, w2 and w3,respectively. In rows 1–3 two weight factors had a value of zero,which is equivalent to a single objective optimization. In rows 4–30,just oneweight factor had a value of zero, so a bi-objective optimization

COM(°C)

P1(°C)

MP(°C)

Coefficient Standarderror

Coefficient Standarderror

Coefficient Standarderror

2.45E-02 0.6935 −3.45E-01 6.1881 −6.75E-01 0.4405

2.13E-01 0.6935 1.13E-01 6.1881 −5.44E-01 0.4405

3.67E-02 0.6935 −3.14E-01 6.1881 −5.65E-01 0.4405

6.90E-04 2.9056 −1.15E-02 25.9255 −2.03E-03 1.8456−6.66E-05 2.9056 −1.59E-03 25.9255 −1.18E-03 1.84566.81E-04 2.9056 −1.07E-02 25.9255 −5.24E-04 1.84561.66E-05 19.6397 1.58E-04 175.2360 1.07E-05 12.47470.0000 0.0019 0.00000.9910 0.8616 0.9935

a)

b)

c)

y = 0.9057x + 0.5074

R² = 0.9057

4.0

4.5

5.0

5.5

6.0

4.0 4.5 5.0 5.5 6.0

Pred

icte

d Vi

scos

ity (m

m2/s)

Measured Viscosity (mm2/s)

y = 0.9625x + 0.3391

R² = 0.9625

0

5

10

15

20

25

0 5 10 15 20 25

Pred

icte

d IP

(h)

Measured IP (h)

y = 0.9910x + 0.0903

R² = 0.9910

0

5

10

15

20

25

0 5 10 15 20 25

Pred

icte

d C

OM

(C

)

Measured COM ( C)

Fig. 2. Correlation between predicted and measured values of (a) Viscosity, (b) IP and(c) COM.

4.5

5.5

5.5

5.0

4.0

8.0

12.0

20.0

16.0

5.5

10.2

14.9

19.6

a)

b)

c)

Fig. 3. Contour plots for (a) Viscosity, mm2/s; (b) IP, h; and (c) COM, °C.

158 V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

was carried out. Finally, in rows 31–34, all theweight factors had valueshigher than zero. The weight factors sum is always one for each row.Thus, the following cases are possible:

i) Single-objective optimization: in rows 1–3 viscosity, IP and COMwere the objective, respectively. In row 1, 100% fish oil was thefeedstock obtained to optimize biodiesel viscosity with a valueof 4.3mm2/s, while IP and COMwere 1.8 h and 2.5 °C, respective-ly. In row 2, 100% palm oil was determined as the feedstock thatmaximizes biodiesel IP, which valued 22.0 h, while viscosity and

COM valued 5.5 mm2/s and 21.3 °C, respectively. Last, in row 3,in order to optimize biodiesel COM, 100% fish oil was the feed-stock defined, and variables took the same values as obtainedfor row 1. The single objective results are in accordance to theoptimization by response surface methodology showed beforefor viscosity (100% fish oil), IP (100% palm oil) and COM (100%fish oil).

ii) Bi-objective optimization for viscosity and IP: in rows 4–12,w3 waszero. For all the weight factors combination, 100% palm oilwas defined as the feedstock that optimizes both viscosity andIP. It is known by the single-objective optimization that whenviscosity weight factor is one, fish oil is the raw material thatoptimizes this property. However, even thoughwhen theweightfactors for viscosity and IP are 90% and 10%, respectively, palm oil

Table 3Set of optimal solutions (Pareto Front) and decision space for the multi-objectiveoptimization problem.

Row w1 w2 w3 Viscosity(mm2/s)

IP(h)

COM(°C)

Fish oil(%)

Palm oil(%)

Frying oil(%)

1 1.00 0.00 0.00 4.3 −1.8 2.5 100.0 0.0 0.02 0.00 1.00 0.00 5.5 −22.0 21.3 0.0 100.0 0.03 0.00 0.00 1.00 4.3 −1.8 2.5 100.0 0.0 0.04 0.10 0.90 0.00 5.5 −22.0 21.3 0.0 100.0 0.05 0.20 0.80 0.00 5.5 −22.0 21.3 0.0 100.0 0.06 0.30 0.70 0.00 5.5 −22.0 21.3 0.0 100.0 0.07 0.40 0.60 0.00 5.5 −22.0 21.3 0.0 100.0 0.08 0.50 0.50 0.00 5.5 −22.0 21.3 0.0 100.0 0.09 0.60 0.40 0.00 5.5 −22.0 21.3 0.0 100.0 0.010 0.70 0.30 0.00 5.5 −22.0 21.3 0.0 100.0 0.011 0.80 0.20 0.00 5.5 −22.0 21.3 0.0 100.0 0.012 0.90 0.10 0.00 5.5 −22.0 21.3 0.0 100.0 0.013 0.10 0.00 0.90 4.3 −1.8 2.5 100.0 0.0 0.014 0.20 0.00 0.80 4.3 −1.8 2.5 100.0 0.0 0.015 0.30 0.00 0.70 4.3 −1.8 2.5 100.0 0.0 0.016 0.40 0.00 0.60 4.3 −1.8 2.5 100.0 0.0 0.017 0.50 0.00 0.50 4.3 −1.8 2.5 100.0 0.0 0.018 0.60 0.00 0.40 4.3 −1.8 2.5 100.0 0.0 0.019 0.70 0.00 0.30 4.3 −1.8 2.5 100.0 0.0 0.020 0.80 0.00 0.20 4.3 −1.8 2.5 100.0 0.0 0.021 0.90 0.00 0.10 4.3 −1.8 2.5 100.0 0.0 0.022 0.00 0.10 0.90 4.3 −1.8 2.5 100.0 0.0 0.023 0.00 0.20 0.80 5.6 −4.5 3.0 42.1 0.0 57.924 0.00 0.30 0.70 5.3 −6.9 3.7 0.0 0.0 100.025 0.00 0.40 0.60 5.3 −6.9 3.7 0.0 0.0 100.026 0.00 0.50 0.50 5.3 −6.9 3.7 0.0 0.0 100.027 0.00 0.60 0.40 5.5 −22.0 21.3 0.0 100.0 0.028 0.00 0.70 0.30 5.5 −22.0 21.3 0.0 100.0 0.029 0.00 0.80 0.20 5.5 −22.0 21.3 0.0 100.0 0.030 0.00 0.90 0.10 5.5 −22.0 21.3 0.0 100.0 0.031 0.50 0.25 0.25 5.3 −6.9 3.7 0.0 0.0 100.032 0.25 0.25 0.50 5.3 −6.9 3.7 0.0 0.0 100.033 0.25 0.50 0.25 5.5 −22.0 21.3 0.0 100.0 0.034 0.33 0.33 0.33 5.3 −6.9 3.7 0.0 0.0 100.0

159V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

was the optimum feedstock. For all the combinations, viscosity,IP and COM values were the same as those of row 2: 5.5 mm2/s,22.0 h and 21.3 °C, respectively.

iii) Bi-objective optimization for viscosity and COM: in rows 13–21,w2

was zero. As result, it was obtained that 100% fish oil as feedstockoptimizes both viscosity and COM for all weight factors values.It was previously expected, since fish oil used as feedstock hadoptimized both viscosity and COM, separately. Viscosity, IP andCOM were respectively maintained as 4.3 mm2/s, 1.8 h and2.5 °C, equal to the values obtained in rows 1 and 3.

iv) Bi-objective optimization for IP and COM: in rows 22–30, w1 wasset as zero. In row 22, the defined feedstock composition was100% fish oil, in order to optimize both IP (w2 = 0.10) andCOM (w3 = 0.90), and it was obtained the same results fromrows 1, 3 and 13–21. For the row 23 the feedstock that optimizesIP (w2=0.20) and COM(w3=0.80) is composed by 42.1wt.% ofwaste fish oil and 57.9 wt.% of waste frying oil. Viscosity, IP andCOM showed the respective values: 5.6 mm2/s, 4.5 h and 3 °C.This oilmixture as feedstock for optimization of biodiesel proper-ties is a singular result, and it shows that waste frying may be anappropriated oil to bemixedwith fish oil in the proper ratios. Forrows 24–26, waste frying oil was defined as the feedstock thatoptimizes IP (w2 from 0.30 to 0.50) and COM (w3 from 0.50 to0.70), showing values for viscosity, IP and COM of 5.3 mm2/s,6.9 h and 3.7 °C, respectively. Last, in rows 27–30, the feedstockestablished to optimize both IP (w2 from 0.60 to 0.90) and COM(w3 from 0.10 to 0.40) was 100% palm oil, and variables showedthe same optimum values found in rows 2, 4–12.From these results, it is deduced that when IP has much lessweight (10%) than COM, fish oil is the best raw material. Then,when IPweight increases (20%), amix of fish oil andwaste frying

oil is the best option as rawmaterial. If IP weight grows from 30to 50%, waste frying oil is the recommended feedstock. However,if IP weight is in the range of 60 to 90%, palm oil is the best choiceto be used as raw material.

v) Tri-objective optimization: for the rows 31–34, the weight factorsvalue varied from 33% to 50% (Table 3). In row 33 (w1 = 0.25,w2 = 0.50 andw3 = 0.25) the determined feedstock to optimizebiodiesel propertieswas 100% palm oil, and the variables showedthe same optimum values as those of rows 2, 4–12, 29 and 30,viscosity value of 5.5 mm2/s, IP value of 22.0 h and COM valueof 21.3 °C. Finally, in rows 31, 32 and 34, 100% waste frying oilwas obtained, and the biodiesel propertiesmaintained the valuesfrom rows 24–26, with the respective values for viscosity, IP andCOM: 5.3 mm2/s, 6.9 h and 3.7 °C.

Therefore, the importance given to viscosity, IP and COM willdetermine the selection of a single optimum solution from the ParetoFront. It will define the composition of the oil mixture (waste fish oil,palm oil andwaste frying oil) employed as rawmaterial for the produc-tion of biodiesel.

4. Conclusion

The present study showed that biodiesel properties are highlyinfluenced by the fatty acid composition of the oil mixture used asfeedstock. Viscosity, IP and COM are properties that depend on the oilcomposition employed as raw material. From the response surfacemethodology it was concluded that viscosity and COM are minimized(4.3 mm2/s and 2.5 °C, respectively) when waste fish oil is the biodieselfeedstock, whereas IP ismaximized (22.0 h)when palm oil is employed.This is attributed to the SFAmethyl ester content of the mixture, whichincreases biodiesel viscosity, IP and COM. Consequently, the conflictingbehavior of these variables (viscosity, IP and COM) suggests the use ofmulti-objective optimization. This analysis reveals that although theuse of the pure oils studied (waste fish oil, palm oil and waste fryingoil) as feedstock showed more advantages to biodiesel properties, themixture of waste fish oil (42.1 wt.%) and waste frying oil (57.9 wt.%) isbeneficial when the importance given to IP and COM is 20 and 80%,respectively.

Acknowledgments

This work was supported by the Spanish National Plan I + D + I(project CTQ2011-23009) and by the Andalusian Government (projectP12-AGR-1993). V. F. de Almeida acknowledges a grant from the“Science without Frontiers” program, supported by CNPq, Brazil(project 207461/2013-9).

References

[1] G. Knothe, Biodiesel and renewable diesel: a comparison, Progress in Energy andCombustion Science 36 (2010) 364–373.

[2] F. Ma, M.A. Hanna, Biodiesel production: a review, Bioresource Technology 70(1999) 1–15.

[3] H. Fukuda, A. Kondo, H. Noda, Biodiesel fuel production by transesterification of oils,Journal of Bioscience and Bioengineering 92 (2001) 405–416.

[4] P.M. Kris-Etherton, W.S. Harris, L.J. Appel, Fish consumption, fish oil, omega-3 fattyacids, and cardiovascular disease, Circulation 106 (2002) 2747–2757.

[5] T.H. Wu, P.J. Bechtel, Salmon by-product storage and oil extraction, Food Chemistry111 (2008) 868–871.

[6] A.P. Bimbo, Current and future sources of raw materials for the long‐chain omega‐3fatty acid market, Lipid Technology 19 (2007) 176–179.

[7] C.Y. Lin, R.J. Li, Fuel properties of biodiesel produced from the crude fish oil from thesoapstock of marine fish, Fuel Processing Technology 90 (2009) 130–136.

[8] C.Y. Lin, R.J. Li, Engine performance and emission characteristics of marine fish-oilbiodiesel produced from the discarded parts of marine fish, Fuel ProcessingTechnology 90 (2009) 883–888.

[9] R. Behçet, Performance and emission study of waste anchovy fish biodiesel in adiesel engine, Fuel Processing Technology 92 (2011) 1187–1194.

160 V.F. de Almeida et al. / Fuel Processing Technology 133 (2015) 152–160

[10] D. Madhu, B. Singh, Y.C. Sharma, Studies on application of fish waste for synthesis ofhigh quality biodiesel, RSC Advances 4 (2014) 31462–31468.

[11] A.B. Fadhil, M.M. Dheyab, L.A. Saleh, Conversion of fish oil into biodiesel fuels viaacid-base catalyzed transesterification, Energy Sources, Part A: Recovery, Utilization,and Environmental Effects 36 (2014) 1571–1577.

[12] P. Jayasinghe, K. Hawboldt, A review of bio-oils from waste biomass: focus onfish processing waste, Renewable and Sustainable Energy Reviews 16 (2012)798–821.

[13] P.J. García-Moreno, M. Khanum, A. Guadix, E.M. Guadix, Optimization of biodieselproduction from waste fish oil, Renewable Energy 68 (2014) 618–624.

[14] G. Knothe, Some aspects of biodiesel oxidative stability, Fuel Processing Technology88 (2007) 669–677.

[15] L. CherngYuan, L. Jung-Chi, Oxidative stability of biodiesel produced from the crudefish oil from the waste parts of marine fish, Journal of Food, Agriculture & Environ-ment 8 (2010) 992–995.

[16] J.F. Costa, M.F. Almeida, M.C.M. Alvim-Ferraz, J.M. Dias, Biodiesel production usingoil from fish canning industry wastes, Energy Conversion and Management 74(2013) 17–23.

[17] S. Mekhilef, S. Siga, R. Saidur, A review on palm oil biodiesel as a source of renewablefuel, Renewable and Sustainable Energy Reviews 15 (2011) 1937–1949.

[18] A.N. Phan, T.M. Phan, Biodiesel production from frying oils, Fuel 87 (2008)3490–3496.

[19] G. Martínez, N. Sánchez, J.M. Encinar, J.F. González, Fuel properties of biodiesel fromvegetable oils and oil mixtures. Influence of methyl esters distribution, Biomass andBioenergy 63 (2014) 22–32.

[20] Z. Jurac, V. Zlatar, Optimization of raw material mixtures in the production ofbiodiesel from vegetable and used frying oils regarding quality requirements interms of cold flow properties, Fuel Processing Technology 106 (2013) 108–113.

[21] D. Galvan, J.R. Orives, R.L. Coppo, E.T. Silva, K.G. Angilelli, D. Borsato, Determinationof the kinetics and thermodynamics parameters of biodiesel oxidation reactionobtained from an optimized mixture of vegetable oil and animal fat, Energy andFuels 27 (2013) 6866–6871.

[22] G. Knothe, J.V. Gerpen, J. Krahl, The Biodiesel Handbook, first ed. AOCS Press, Illinois,2005.

[23] UNE-EN 14103, Fat and oil derivatives-Fatty Acid Methyl Esters (FAME)-determina-tion of ester and linolenic acid methyl ester contents, European Committee for Stan-dardization, 2003.

[24] B. Camacho-Páez, A. Robles-Medina, F. Camacho-Rubio, P. González-Moreno, E.Molina-Grima, Production of structured triglycerides rich in n-3 polyunsaturat-ed fatty acids by the acidolysis of cod liver oil and caprylic acid in a packed-bedreactor: equilibrium and kinetics, Chemical Engineering Science 57 (2002)1237–1249.

[25] UNE-EN 15751, Automotive fuels-Fatty acid methyl ester (FAME)-fuel and blendswith diesel fuel-Determination of oxidation stability by accelerated oxidation meth-od, European Committee for Standardization, 2008.

[26] UNE-EN 14104, Oil and fat derivatives-Fatty Acid Methyl Esters (FAME)-determina-tion of acid value, European Committee for Standardization, 2003.

[27] H. Halsall-Whitney, J. Thibault, Multi-objective optimization for chemical processesand controller design: approximating and classifying the Pareto domain, Computers& Chemical Engineering 30 (2006) 1155–1168.

[28] I.Y. Kim, O.L. DeWeck, Adaptive weighted summethod for multiobjective optimiza-tion: a new method for Pareto front generation, Structural and MultidisciplinaryOptimization 31 (2006) 105–116.

[29] UNE-EN 14214, Automotive fuels-Fatty acid methyl esters (FAME)-for diesel engines.Requirements and test methods, European Committee for Standardization, 2014.

[30] A.B. Fadhil, L.H. Ali, Alkaline-catalyzed transesterification of Silurus triostegus Heckelfish oil: optimization of transesterification parameters, Renewable Energy 60(2013) 481–488.

[31] A. Hayyan, M.Z. Alam, M.E. Mirghani, N.A. Kabbashi, N.I.N.M. Hakimi, Y.M. Siran, S.Tahiruddin, Reduction of high content of free fatty acid in sludge palm oil via acidcatalyst for biodiesel production, Fuel Processing Technology 92 (2011) 920–924.

[32] B.B. Uzun, M. Kılıç, N. Özbay, A.E. Pütün, E. Pütün, Biodiesel production from wastefrying oils: optimization of reaction parameters and determination of fuel proper-ties, Energy 44 (2012) 347–351.

[33] G. Knothe, K.R. Steidley, Kinematic viscosity of biodiesel fuel components andrelated compounds. Influence of compound structure and comparison topetrodiesel fuel components, Fuel 84 (2005) 1059–1065.

[34] R. Sarin, M. Sharma, S. Sinharay, R.K. Malhotra, Jatropha–palm biodiesel blends: anoptimum mix for Asia, Fuel 86 (2007) 1365–1371.

[35] G. Knothe, R.O. Dunn, A comprehensive evaluation of the melting points of fattyacids and esters determined by differential scanning calorimetry, Journal of theAmerican Oil Chemists' Society 86 (2009) 843–856.

[36] C.S. Foon, Y.C. Liang, N.L. Dian, C.Y. May, Crystallization and melting behavior ofmethyl esters of palm oil, American Journal of Applied Sciences 3 (2006) 1859.

[37] G. Knothe, Dependence of biodiesel fuel properties on the structure of fatty acidalkyl esters, Fuel Processing Technology 86 (2005) 1059–1070.

[38] R.O. Dunn, M.O. Bagby, Low-temperature filterability properties of alternative dieselfuels from vegetable oils, in: J.S. Cundiff, E.E. Garett, C. Hansen, C. Peterson, M.A.Sanderson, H. Shapouri, D.L. Van Dyne (Eds.), Proceedings of the Third LiquidFuels Conference: Liquid Fuels and Industrial Products From Renewable Resources,American Society of Agricultural Engineers, St. Joseph, MI, 1996, pp. 95–103.

[39] R.O. Dunn, Effect of antioxidants on the oxidative stability of methyl soyate (biodiesel),Fuel Processing Technology 86 (2005) 1071–1085.