Optimization of methyl ester production from Prunus

Renewable Energy 130 (2019) 61e72

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

Optimization of methyl ester production from Prunus Amygdalus seed oil using response surface methodology and Artificial Neural Networks Chizoo Esonye a, *, Okechukwu Dominic Onukwuli b, Akuzuo Uwaoma Ofoefule c a b c

Department of Chemical and Petroleum Engineering, Federal University, Ndufu-Alike Ikwo, P.M.B 1010, Abakaliki, Ebonyi State, Nigeria Department of Chemical Engineering, Nnamdi Azikiwe University, P.M.B 5025 Awka, Anambra state, Nigeria Department of Pure and Industrial Chemistry, University of Nigeria, 410001 Nsukka, Enugu State, Nigeria

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 June 2017 Received in revised form 21 April 2018 Accepted 12 June 2018 Available online 14 June 2018

This research work investigated the optimization of biodiesel production from Sweet Almond (Prunus amygdalus) Seed oil (SASO) using Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) models through base (NaOH) transesterification. The Central Composite Design (CCD) optimization conditions were temperature (30 C to 70 C), catalyst concentration (0.5%w/w to 2.5% w/w), reaction time (45 mine65 min) and oil/methanol molar ratio (1:3 mol/mol to 1:7 mol/mol). The physicochemical properties of the seed oil and the methyl ester were carried out using standard methods. The fatty acids were determined using GC-MS and characterized using FT-IR techniques. An optimized biodiesel yield of 94.36% from the RSM and 95.45% from the ANN models respectively were obtained at catalyst concentration of 1.5w/w%, reaction time of 65 min, oil/methanol molar ratio of 1:5 mol/mol and temperature of 50 C. The quality of the RSM model was analyzed using Analysis of Variance (ANOVA). Model statistics of the ANN showed comfortable values of Mean Squared Error (MSE) of 6.005, Mean Absolute Error (MAE) of 2.786 and Mean Absolute Deviation (MAD) of 1.89306. The RSM and ANN models gave coefficient of determination (R2) of 0.9446 and correlation coefficient (R) of 0.96637 respectively. © 2018 Elsevier Ltd. All rights reserved.

Keywords: Transesterification Biodiesel Optimization Models Prunus Amygdalus

1. Introduction Modern industrialized society has for many decades relied basically on single natural resource: petroleum. The total dependence on petroleum for liquid transportation, heating fuels, plastics and other petrochemicals, asphalt for road construction, packaging materials and modern medical devices have made it very difficult for the modern man to live without petroleum. Energy has become a crucial determinant for continued economic growth and improved standard of living especially after the inauguration of the industrial revolution in the late 18th and 19th centuries [1]. According to the International Energy Agency (IEA) report, the world will need 50% more energy in 2030 than today of which 45% will be accounted by China and India [2,3]. But for every positive benefit

* Corresponding author. E-mail addresses: [email protected] (C. Esonye), [email protected] (O.D. Onukwuli), [email protected] (A.U. Ofoefule). https://doi.org/10.1016/j.renene.2018.06.036 0960-1481/© 2018 Elsevier Ltd. All rights reserved.

that petroleum has provided there seems to be a negative environmental ramification [4]. It has been reported that the deep water horizon oil spill released 4 million barrels of oil into Gulf of Mexico which had series of negative ecological and economic impacts on the states [5]. Extension use of asphalt have created a phenomenon known as islands resulting in increase of energy consumption and high mortality rate in urban centres [6]. Combustion engine emissions have resulted in growing concern over air quality and green house effects. It is believed that climate change is currently the most pressing global environmental problem. If the average global temperature increases by more than 20 C, over one million species could become extinct and hundreds of millions of people could lose their lives [7]. It is expected that 4.1billion metric tonnes of carbon dioxide will be released to the atmosphere from 2020 to 2035 and this is estimated to be about 43% increase for the aforementioned projected period [8]. Globally, the awareness of energy issues and environmental problems associated with burning fossil fuel has encouraged researches on the possibility of using alternative renewable sources of energy. Among them

62

C. Esonye et al. / Renewable Energy 130 (2019) 61e72

Abbreviations ANN ANOVA CCD FFA FT-IR GC-MS HHV MAD MSE RSM SASO SASOME SSE

Artificial neural Network Analysis of Variance Central Composite Design Free Fatty Acid Fourier Transform Infra red Gas Chromatography-Mass Spectroscopy Higher Heating Value Mean Absolute Error Mean Squared Error Resonance Surface Methodology Sweet Almond Seed Oil Sweet Almond Seed Oil Methyl Ester Sum of Squares Error

biodiesel appears very interesting for several obvious reasons: It is highly biodegradable and has minimal toxicity, it can replace diesel fuel in many different applications such as in boilers and internal combustion engines without major modification, almost zero emissions of sulphates, aromatic compounds and other chemical substances that are destructive to the environment, a small net contribution to CO2 when the whole life-cycle is considered and it appears to cause significant improvement of rural economic potentials [7]. It has been observed that the fuel properties determined by fuel components of biodiesel play a major role in the combustion processes. This highlights the importance of developing suitable models in order to cover an exact view of the phenomenon and sequences occurring when biodiesel is used. Equally, optimization of relevant parameters in transesterification reaction and the biodiesel fuel properties by laboratory tests are much time consuming and costly. The response surface methodology (RSM) model has been known as a powerful tool for optimization in several fields such as chemical engineering process control and chemical analysis among many other applications. The response surface methodology (RSM) is used to determine the optimum conditions for the transesterification reaction. In order to achieve maximum production of biodiesel, process variables including reaction temperature, oil/ methanol molar ratio, amount of catalyst and reaction time have been used simultaneously in RSM method by performing central composite design (CCD) [9]. The main advantage of RSM is its capability to minimize the number of experimental runs needed to give adequate evidence for statistically acceptable result [10]. It has been applied in methanolysis optimization of some vegetable oil to biodiesel: Sesame oil [11], Moringa oleifera oil [12], waste rape seed oil [13], Zanthoxylum bungeanum seed oil using CaO as catalyst [14] and cotton seed oil [15]. Additionally, artificial neural network (ANN) approach has been one of the well known types of evolutionary computation method in the last decades. Artificial neural networks have shown great ability in solving complex nonlinear systems identification and control problems and can be described either as mathematical and computational models for non-linear function approximation, data classification, clustering and nonparametric regression or as simulations of the behavior of collections of model biological neurons. Various studies have proved that ANN and RSM are powerful techniques for process modeling of biofuels to save energy, cost and time [16]. Currently, RSM and ANN are still being applied and seen as the best statistical approach in modeling of transesterification processes and the most efficient method for empirical modeling [17]. In this study, these models are applied to sweet almond seed oil transesterification process as they

are required to ascertain the viability of the industrial application through the knowledge of the optimum conditions for the operating variables. These models have not been applied to sweet almond seed oil transesterification process and this study is undertaken to fill this gap. Almond plants are included in the family Rosaceae in addition to promoideae (apples, pears), prunoideae (apricot, cherry, peach and plum) and rosoideae (blackberry, strawberry) fruits [18]. There are two major varieties of almonds, the bitter almond (Prunus amygdalus“amara”) and the sweet almond (Prunus amygdalus“dulcis”) used mainly for culinary purposes and making of oils and flavourings respectively [19]. Almond fruit consists of four portions: comprising the kernel or meat, middle shell, outer green shell cover or almond hull and a thin leathery layer known as brown meat or seed coat [20]. World production of almond was 2.9 million tonnes in 2013 with United States as the largest producer of 1.8 million tones [21]. Sweet almond tree is found in the south eastern and south southern parts of Nigeria where they are basically grown to provide shades to homes, offices and the environment. Their fruits litter the environment and are picked either by children or disposed off as wastes and as such their use as feedstock for biodiesel production would also serve as a waste disposal option in these areas. Although, Almond seed oil is from an edible feedstock, its application as a viable feedstock for biodiesel production may not likely compete with its use as food since it is not a staple food in so many parts of the world and not widely consumed in Africa. Sweet almond though edible feedstock has been previously recommended for biodiesel in Nigeria (where its underutilization was reported) and Saudi Arabia [18,22] as well as for industrial production of cosmetics [23]. This application is supported by the fact that some researchers have advised against the consideration of only the first, second and third generation classification of biodiesel feedstock rather to pay more attention to the fuel related qualities of the derived biodiesel [24]. Jatropha oil which is seen as one of the leading nonedible feedstocks for biodiesel has been reported to possess poorer cold flow properties than biodiesel derived from edible feedstocks such as rape seed, palm and soybean seed oils [3]. Also, sweet almond seed oil though applied as food supplement has substitutes like Canola and Jajoba oils and these feedstocks are equally considered as viable feedstocks together with other edible oils [3,25]. Giwa and Ogunbona [18], studied the extraction and characterization of the seed oil biodiesel from sweet almond obtained from Nigeria. Their study revealed that the seed oil has an oil yield of 51.45%, acid value of 1.07 mg KOH/g and fatty acid composition of oleic acid (69.7%), linoleic acid (18.2%) and palmitic acid (9.3%). Their result equally showed that the cold flow properties were 3 and 9 for the cloud point and pour point respectively with the specific fuel properties found to satisfy both EN 14214 and ASTM D6751 biodiesel standards. Mehdic and Kariminia [26], also studied the optimization of biodiesel production from Iranian bitter almond oil using statistical approach. Their investigation revealed that at the following optimal conditions: temperature of 35 C, catalyst concentration of 1.4 wt% and methanol to oil molar ratio of 9.7 mol/mol, the actual values of the product yield, biodiesel yield and biodiesel purity were 96.7, 94.7 and 97.9 wt% respectively while the predicted values were 98.1, 96.3 and 98.2 wt% respectively. It is therefore evident that there has not been any detailed published work on the application of RSM with CCD and ANN models to optimize the biodiesel production from Sweet almond seed oil obtained from Nigeria. Consequently, this work seeks to optimize biodiesel production from prunus amygdalus oil through the application of Central Composite Design of Response Surface Methodology (RSM) and Artificial Neural Network (ANN) models to ascertain the commercial viability of the process.


2. Materials and methods 2.1. Materials Sodium hydroxide (99% Sigma-aldrich), potassium hydroxide (lobachemie, gmbH 85%), methanol (Merck, Germany 99.5% purity), carbon tetrachloride (chloroform), Wij’s solution (iodine monochloride), potassium iodide solution and phenolphthalein (Merck Germany) were all of analytical grade. 2.2. Test materials/sample collection 2.2.1. Source of biomass The fresh fruits were sourced from Onitsha City in Anambra State of Nigeria. 2.2.2. Biomass preparation The fruits were washed properly and separated into seeds and pulp. The husks containing the seeds were sun-dried for 5 days to ensure free movement of the seed (an indication of readiness for seed separation). The seed were manually separated from hulls by cracking and the seeds collected were sun-dried in the open for 7 days. The seeds were ground into meal using electric grinder. 2.2.3. Oil extraction A known weight (3.5 kg) of the dried seed meal was packed in a big fractionating column up to three quarter level and n-hexane was poured well above the level of the meal in the column. It was closed with aluminum foil and sealed with masking tape and then left for a period of 24 h. The mixture of oil and solvent was collected and filtered into a beaker. This was repeated to extract more oil from the meal. After which the oil was recovered using rotary evaporator to distill off the solvent. After distillation, the oil was left in the open to dry up the remaining solvent completely. The oils were degummed to remove phosphosphides and lysophasidic acids. 2.2.4. Physico-chemical characterization of the seed oil The properties of the sweet almond seed oil were determined in accordance with Association of Official Analytical Chemists [27] method: the acid value by AOAC Ca5a-40, saponification value by AOAC 920:160; iodine value by AOAC 920:158 and peroxide value by AOAC 965.33) while the viscosity was determined by using Oswald viscometer apparatus, the density by using density bottle and moisture content by the Rotary evaporator oven (BTOV 1423). The ash content by heating to dryness in Veisfar muffle furnace and the refractive index by using Abbe refractometer (Model: WAY-25, Search tech. Instruments). 2.3. Biodiesel production 2.3.1. Preheating the oil The oil was heated fairly at 80 C for 30 min using Gallenkamp magnetic stirrer thermostat hot plate (WeissTechnik England) to reduce the viscosity of the oil. 2.3.2. Preparation of catalyst This was prepared by adding NaOH (2% weight of the oil) of to 175 ml of methanol and stirred at 200 rpm until it dissolved completely for about two minutes in the reaction vessel. 2.3.3. Base transesterification The SASO was subjected to base transesterification. The calculated amount of NaOH (catalyst) and methanol were added to the oil for each reaction for the temperature and reaction time

63

specified. The Base transesterification was carried out in a Sohxlet extractor fitted with thermo-regulator heater and stirrer. The calculated amount of oil was measured into the flask and was heated to the specified temperature. The Sodium methoxide was then poured into the flask containing oil and was immediately covered. The temperature was maintained for the specified time at constant agitation. 2.3.4. Separation of biodiesel After the base transesterification process the biodiesel was allowed to settle for at least 24 h inside a separating funnel to allow clear separation of biodiesel from glycerin by gravity. The layer on the top is the biodiesel while the bottom layer is the glycerol. The Biodiesel separation was carried out by decanting as the glycerol was drained off while the biodiesel remained. 2.3.5. Biodiesel washing Warm distilled water at 50 C was added to the separated biodiesel and the mixture was shaken vigorously. The water was allowed to drain through the bottom of the separating funnel. This was carried out five times until a clear biodiesel was obtained. 2.3.6. Biodiesel drying Anhydrous CaCl2 was added to the biodiesel and heated gently at 50 C. The anhydrous CaCl2 was later separated from the biodiesel to obtain a clean dry biodiesel. The volume of the biodiesel obtained from each sample was determined while the percentage yield of biodiesel was calculated. 2.3.7. Physico-chemical characterization of the biodiesel The fuel properties of the synthesized biodiesel were determined by ASTM standards: the kinematic viscosity was determined by ASTM D-445 method, the density was determined by ASTM D1298 method, and the pour point determination was made using ASTM D-97 methods. The flash point of the fuel was determined as ASTM D-93, the value of cloud point was estimated according to ASTMD-2500, and Acid value was measured following the ASTM D664 method. The Calorific value and cetane number were calculated according to the correlation developed by Patel [28]. 2.3.8. Chemical characterization of seed oil and biodiesel 2.3.8.1. FTIR spectroscopic analysis of the oil. The mid Infrared spectra of oil and biodiesel samples were obtained in Fourier Transform Spectrometer by IR Affinity-1 Shimadzu, Japan FJS: JPN patent No: 2115670, No 3613171 and JPN reg. of utility model No: 3116465. The FTIR has SN ratio of its class of 30,000:1, 1 min accumulator in the neighborhood of 2,100 cm1 peak to peak with a maximum resolution of 0.5 cm1 in the region of 400 cm1 4000 cm1. 2.3.8.2. GC-MS analysis of the fatty acid profile of the biodiesel. The fatty acid composition of the biodiesel samples was analyzed by Gas eChromatography coupled with Mass Spectrometer. The gas chromatographic analysis was made using GCMS-QP2010 plus, Shimadzu, Japan instrument. The GC column used was calibrated by injecting methyl ester standards, good separations were achieved by diluting the sample in a small amount of ethyl acetate. The carrier gas used was hydrogen and its flowrate was regulated at 41.27 ml/min while the column flows at 1.82 ml/min. The oven temperature was set at 80 C before ramping up at 6 C/min until 340 C. The identification of peaks was done by comparison of their retention time and mass spectra with mass Spectra Library (NIST05s LIB.).

64


2.4. Experimental design and statistical analysis by RSM

Table 2 The CCD for five-level-four-factor response surface analysis.

In order to optimize the central composite experimental design (CCD), a five-level-four-factor central design was employed for this study, which generated twenty five (25) runs. The factors investigated in this study were reaction time (minutes), catalyst amount (wt %), Temperature (ºC), and oil/methanol ratio (mol/mol). The coded and uncoded factors (X1, X2, X3 and X4) and levels used are shown in Table 1 and Table 2. The variables ranges were selected based on results obtained from preliminary studies and literature [29]. The data obtained in the experiments were analyzed by response surface methodology by following a polynomial regression first order model. The equation is shown below in equation (1) as generated by MATLAB 8.5 software 2015 version.

Y ¼ a0 þ

n X i¼1

ai X i þ

k XX i¼1

aij X ij

(1)

Where a0 is the constant, ai is the linear coefficient, aij-interactive coefficients, Xi and Xij are the uncoded independent variables and Y is the predicted response (% biodiesel yield). The linear coefficients and the cross-products are significant model terms while the effects of the factors like reaction time, catalyst concentration, temperature and oil/methanol ratio on the percentage seed oil biodiesel yield are shown on the 3D representations. The quality of the fit of the model was evaluated using analysis of variance (ANOVA). By applying sum of squares error (SSE), probability value (t-values), general probability (p-values), degree of freedom (df) and correlation coefficient (R).

Run

X1 (C)

X2 (wt %)

X3 (min.)

X4(mol/mol)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 0 0 0 0 0 0 0

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 2 2 0 0 0 0 0

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 2 2 0 0 0

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 2 2 0

MAPE ¼

. n X MAD ¼ yi yî n

(2)

i¼1

Table 1 Factors and their levels of CCD for the biodiesel productions.

X1 X2 X3 X4

(4)

Where MSE e Mean Squared Error (Standard Deviation)

A consolidated data set comprising of twenty five (25) data set are compiled and parameters like temperature at which the reaction is carried out, the time of reaction in minutes, the catalyst concentration as weight percent and the oil to methanol ratio were used as the independent input parameters. In this study, a threelayered feed-forward neural network with tangent sigmoid transfer function (tansig) at hidden layer and linear transfer function (purelin) at output layer was used. The models developed are used for the production of the one dependent parameter: the biodiesel yield in each of the 25 independent runs. ANN model is developed using MATLAB 8.5 software 2015 version. The propagation algorithm was used for network training, 52% of the data was taken for training set, 13% for validation and the rest of the data for the test set. The accuracy of the models was determined by using equations (2)e(4) as applied by Ahmadian -Morghadam et al., [30].

Temperature (ºC) Catalyst conc. (w/w %) Reaction time (minutes) Oil/methanol molar ratio (mol/mol)

o. n n X ^ ^ n yi yi 2 i¼1

2.5. Development of artificial neural network (ANN) model

Symbols

(3)

i¼1

MSE ¼

Variable

. o. n n X n yi yî yi

Coded levels 2

1

0

1

2

30 0.5 45 1:3

40 1.0 50 1:4

50 1.5 55 1:5

60 2.0 60 1:6

70 2.5 65 1:7

MAD- Mean Absolute Deviation MAPE- Mean Absolute percentage Error yi e Actual biodiesel yield (%) yî- Predicted biodiesel yield 3. Results and discussion The fuel related properties of the biodiesel obtained from this work at the optimum conditions are contained in Table 4 while Table 3 contains the physico-chemical properties of the seed oil. The density, flash point, kinematic viscosity, iodine value, cetane Table 3 Physico-chemical properties of SASO. Parameters

Results

Oil yield (%) Colour Density (kg/m3) Moisture content (%) Refractive index Saponification value (mg KOH/g) Iodine value (g/100 g) Peroxide value (milli eq. oxy/kg) Acid value (mg KOH/g) Free fatty acid as oleic (%) Ash Content (%) Viscosity (cP) Smoke point (oC) Titre point (oC) Flash point (oC) Cloud point (oC)

60.15 Golden 855.2 0.57 1.4472 165.50 35.77 1.48 2.805 1.402 1.02 6.05 40 52 157 2


65

Table 4 Physico-chemical properties of the SASO FAME at optimum conditions. Parameter

Biodiesel Yield (%) Density (kg/m3) Moisture content (%) Refractive Index Acid value (mgKOH/g) Free fatty acid (%) Iodine value (mgKOH/g) Saponification value (mgKOH/g) Ash Content (%) Kinematic viscosity (mm2/s) Smoke point (ºC) Fire point (ºC) Flash point (ºC) Cloud point (ºC) Pour point (ºC) Calorific value (KJ/Kg) Conductivity (Us/CM) Cetane Index Cetane Number Higher Heating Value (HHV)a (MJ/kg) Higher Heating Value (HHV)b (MJ/kg) Higher Heating Value (HHV)c (MJ/kg) a b c

SASO FAME

94.90 849.1 0.02 1.4402 0.46 0.23 28.02 161.05 0.01 2.52 34 40 136 2 6 31,178.39 0.40 73.0 70.40 34.72 40.76 63.75

Standards ASTM D 9751

ASTM D 6751

DIN 14214

e 850 e e 0.062 0.31 42e46 e 0.01 2.6 e e 60e80 20 35 42e46 e e 40e55 e e e

e 880 e e 0.50 0.25 e e 0.02 1-9-6.0 e e 100e170 3 to 12 15 to 16 e e e 47min. e e e

e 860e900 e e 0.50 0.25 120max. e 0.02 3.5e5.0 e e 120 e e 35 e e 51min. e e e

Based on flash point. Based on viscosity. Based on density, min-minimum, max- maximum.

number, pour point, cloud point, moisture content and calorific value, of the biodiesel compared well with the American standard (ASTMD 6751) and European specification (EN 14214). The oil characteristics contained in Table 3 showed improved fuel qualities upon transesterification. The fatty acid chromatogram showing the different components present in SASOME is shown in Fig. 1. Twenty four (24) peaks were recorded which showed different fatty acid methyl esters and few non fatty acid components present. The major fatty acid component present in SASOME is oleic acid followed by a-linolenic acid, palmitic acid and stearic acid. Other organic compounds detected by the GCeMS in SASOME include hexadecane, hexadimethylacetal, octanal, pentanal and octane. These results are in line with the results obtained by Botinestean et al. [31], who identified decane, tetralin and hexadimethylacetal in tomato seed oil by GC-MS and those obtained by Sharmila and Jeyanthi [32], who identified over six non e fatty acid methyl esters through GC-MS of Cladophora vagabund. From Table 5 and Fig. 2, SASOME contains a total of 37.74% saturated fatty acid, 41.42% monounsaturated fatty acid and 13.90% polyunsaturated fatty acids. Therefore, the SASOME is expected to have low thermal efficiency and viscosity, emit lower HC, CO and less smoke than other higher saturated biodiesel fuels [29]. The specific peak 891.42 cm1 indicates the presence of ¼CeH

functional groups and possesses bending type of vibrations appearing at low energy and frequency region in the spectra (Figs. 3 and 4). They are all double bonded and attributed to as unsaturated. They are part of fatty acid methyl ester with unsaturated bond in the triglyceride and ester (Oleate and linoleate) [33]. The following bonds are typical of an ester C¼ O, CeO, CeC, CeH and OeH. The characteristics peaks found in the region of 1076.70 cm1 and 1196.40 cm1 show split stretching of CeO and rocking vibration of CeO as carbonyl groups for SASO and SASOME [33,34]. It could be observed that 1188.64 cm1 in the oil sample got split into two concrete signals at 1134.60 and 1196.36 cm1. The band region between 1319.88 and 1501.30 cm1 and 1319.88e1566.92 cm1 for SASO and SASOME spectral respectively can be ascribed to the bending and rocking vibrations of methyl group in the glyceride and ester [35]. The band region between 1721.32 and 1840.98 cm1 and 1721.32 and 1813.96 cm1 for SASO and SASOME spectral respectively can be ascribed to the stretching vibrations of C]O group indicating the conversion of the triglyceride, to methyl esters. The characteristic bands of 2419.98 cm1 and 2400.68 cm1 appear with C]C (alkynes group) for SASO and SASOME while the band region between 3384.98 and 3597.28 cm1 and 3384.98e3608.86 cm1 for SASO and SASOME can be ascribed to OeH stretching vibrations which are single bonded and appear at high energy levels. 3.1. Response surface methodology (RSM) optimization

Fig. 1. GC-MS chromatogram of fatty acid methyl ester of SASO.

A linear regression model (equation (1)) was used to predict the response and determine the input value at which an optimized output is obtained. The regression coefficients a0 to a10 after simulation gives; a0 ¼ 322.5668, a1 ¼ 1.9178, a2 ¼ 57.4717, a3 ¼ 2.4426, a4 ¼ 170.5154, a5 ¼ 0.1375, a6 ¼ 0.0138, a7 ¼ 0.6875, a8 ¼ 0.3750, a9 ¼ 18.75, a10 ¼ 1.8750, R2 ¼ 0.9446. With these parameters, the response was predicted to get the best model for maximizing biodiesel production by response surface methodology. The final equation in terms of coded factors for

66


Table 5 Fatty acid profile of SASO FAME. Fatty acid

Systematic Name

Structure

Formula

Molecular Weight (g/mol)

Amount (wt%)

Capric Caprylic Stearic Erucic Palmitic Lignoceric Oleic a- linolenic Palmitoleic Margaroleic Arachidic Behenic Myristic Lauric Linoleic Gadoliec Y- linolenic Othersa

Decanoic Octanoic Octadecanoic Cis-13-docosenoic Hexadecanoic Tetracosanoic Cis-9-octadecenoic Cis,cis,cis-9,12,15-octadactrienoic Cis-9-hexadecenoic Cis-9-heptadecenoic Eicosanoic Docosanoic Tetradecanoic Dodecanoic Cis-9-cis-12-octadecedianoic Cis-9-eicosenoic Cis,cis,cis-9,12,15- octadactrienoic

10:0 8:0 18:0 22:1 16:0 24:0 18:1 18:3 16:1 17:1 20:0 22:0 14:0 12:0 18:2 20:1 18:3 e

C10H20O2 C8H16O2 C18H36O2 C22H42O2 C16H32O2 C24H48O2 C18H34O2 C18H30O2 C16H30O2 C17H32O2 C20H40O2 C22H44O2 C14H28O2 C12H24O2 C18H32O2 C20H38O2 C18H30O2 e

172.27 144.21 284.48 356.21 256.43 368.63 282.48 278.44 354.41 268.48 312.54 340.59 228.38 200.32 280.45 310.51 280.45 e

1.06 1.36 7.14 0.73 7.88 4.75 40.34 13.07 0.58 0.11 4.30 4.71 3.69 1.53 0.83 0.24 5.01 6.94

a

Non-fatty acids identified by the GC-MS.

Table 6 Central composite design of five-level four-factor response surface study SASOME. Run

Factor 1 X1 (oC)

Factor 2 X2 (wt %)

Factor 3 X3 (mins)

Factor 4 X4 (mol/mol)

Actual Value (%)

Predicted Value (%)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

40 60 40 60 40 60 40 60 40 60 40 60 40 60 40 60 30 70 50 50 50 50 50 50 50

1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.5000 1.5000 0.5000 2.5000 1.5000 1.5000 1.5000 1.5000 1.5000

50 50 50 50 60 60 60 60 50 50 50 50 60 60 60 60 55 55 55 55 45 65 55 55 55

1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:5000 1:5000 1:5000 1:5000 1:5000 1:5000 1:3000 1:7000 1:5000

80.7700 70.5400 66.8400 69.6100 84.8860 87.6560 83.9560 86.7260 67.7700 71.5400 65.8400 69.7100 85.8860 86.6560 85.9560 86.7260 74.4780 80.0180 78.1780 86.3180 60.1320 94.3640 77.2480 76.9420 76.8431

75.2854 72.6804 68.3137 68.4587 86.3597 86.5047 83.1380 86.0330 69.1823 69.3273 65.9606 68.8556 84.0066 86.9016 84.5349 90.1799 76.3376 79.3776 79.7043 76.0110 61.6583 94.0570 78.8357 77.3425 76.5521

Fig. 3. FTIR spectrum for SASO. Fig. 2. Composition of saturated, monounsaturated and polyunsaturated fatty acids of SASOME.


Fig. 4. FTIR spectrum for SASO biodiesel.

the central composite response surface linear model is shown in Equation (5).

maximum value of 95.66% at 65 min and 2.45% catalyst concentration on biodiesel was obtained with the effects of reaction time being more pronounced than catalyst concentration. Fig. 5D showed the effect of temperature and oil/methanol ratio on the biodiesel yield. A maximum value of 80.03% at 69.17 C and oil/ methanol ratio of 1.58 was obtained the two factors have equal effects on the yield. Fig. 5E showed the effect of reaction time and temperature on biodiesel yield with a maximum value of 92.93%, The effects time is higher than that of temperature. Fig. 5F showed the effect of catalyst concentration and temperature on the yield of biodiesel. A maximum value of 80.43% at 0.54% catalyst concentration and 32.5 C temperature was obtained. The effects of catalyst concentration and temperature are almost the same. Taking all together the result, an optimized highest yield of 94.37% was obtained at temperature of 50 C, Catalyst concentration of 1.5 g, reaction time of 65 min and oil/methanol ratio of 1.5 which was very

Y ¼ 322:5667 1:9178X 1 57:471X2 2:4426X3 170:5154X4 þ 0:1375X 1 X 2 þ 0:0138X 1 X 3 þ0:6875X 1 X 4 þ 0:3750X 2 X 3 þ 18:75X 2 X 4 þ 1:8750X 3 X 4

The values of SASOME content obtained ranged from 66.84 wt% to 94.36 wt % (see Table 6). The goodness of the model was checked by different criteria: the coefficient of determination (R2) is 0.9446 which implies that 94.46% of the sample variation in the biodiesel production is attributed to the independent variables. The corresponding Analysis of Variance (ANOVA) is presented in Table 5. The significant of each coefficient was determined by student’s t-test and p-values which are also listed in Table 7. The larger the magnitude of the t-value and the smaller the p-value, the more significant is the corresponding coefficient [36]. The 3 D response surface plots described by the regression model were drawn to illustrate the effects of each variable upon the response variable (Fig. 5). The values of Prob.> F was less than 0.05 which indicates that the model terms are significant. The model terms X2 and X4 significantly affected the measured response of the system. Fig. 5A showed the effect of reaction time and oil/methanol ratio on the biodiesel yield. Maximum of 94.51% was obtained at 64.58 min and 1.58 oil/methanol ratio with reaction time having more effect than oil/methanol ratio. Fig. 5B showed the effect of catalyst concentration and oil/methanol ratio on the biodiesel yield. A maximum yield of 83.63% at 1.3 oil/methanol ratio and 0.59% catalyst concentration was obtained. Their effects are almost balanced or equal on the biodiesel yield. Fig. 5C showed the effect of reaction time and catalyst concentration on biodiesel yield. A

Table 7 ANOVA for response surface linear model for SASOME. Xi

Coefficients

SSE

t-values

P-values

Dfe

ao a1X1 a2X2 a3X3 a4X4 a5X1X2 a6X1X3 a7X1X4 a8X2X3 a9X2X4 a10X3X4

322.5668 1.9178 57.4717 2.4426 170.5154 0.1375 0.0138 0.6875 0.3750 18.7500 1.8750

123.5003 1.1719 23.9602 2.0027 77.5669 0.1243 0.0124 0.6214 0.2486 12.4283 1.2428

2.6119 1.6364 2.3986 1.2196 2.1983 1.1063 1.1063 1.1063 1.5086 1.5086 1.5086

0.0227 0.1277 0.0336 0.2460 0.0483 0.2903 0.2903 0.2903 0.1573 0.1573 0.1573 Pval ¼ 1.0386 108

12

67

(5)

much close to the predicted value (94.05%). It is observed that reaction time has the greatest effect followed by catalyst concentration and temperature as this is a homogeneous catalyzed reaction. The result of this study compared well with the response surface methodology application results reported by Mehdic and Kariminia [26] on Iranian bitter almond, Awolu and Layokun [29] on Neem oil and De Lima et al. [37] on corn oil. It could be inferred that the process of alkaline transesterification via the methylic route is controllable in a catalyst concentration range of 0.5e2.5% and molar ratio (oil: methanol) between 1:3 to 1:7. It means that the tolerance for the process is 1.5 ± 0.25 w/w% for catalyst concentration and 1:5 ± 1.0 for molar ratio. This information is very important for the process control because that will determine the type of controller and sensors to be installed [37]. The result of this study showed that the maximum yield was not obtained at high factor values. This will definitely have advantages on the economic aspects of the biodiesel production. 3.2. Artificial Neural Network (ANN) optimization Fig. 6 and Table 8 show the relationship between output and target with high coefficient of determination (R ¼ 0.96637) and low standard error of estimation using ANN as contained in Table 9. The values on the X-axis are the target values, or the experimental values input to develop the model whereas the values on the Y-axis are the values predicted by the ANN developed. As can be seen from the high regression value, the values predicted are very close to the actual yield values for all data set and an indication of successful development of the ANN model. The regression coefficients of training, test, validation and overall model developed using ANN is shown in Fig. 7. Also Table 9 summarizes the statistical results for training and validation sets of artificial neural network models. These results indicate forecasting error measurements based on difference between the model and actual values. By this consideration, these training data, the lowest standard deviation, mean absolute deviation, mean absolute percentage error and the highest R2 were calculated for SASOME yield. For validation data, however, the lowest standard deviation, mean absolute deviation mean absolute percentage error and the higher R2 were observed for

68


Table 8 Sample of data of Artificial Neural Network of SASOME. Run

Factor 1 X1 (oC)

Factor 2 X2 (wt%)

Factor 3 X3 (mins)

Factor 4 X4 (mol/mol)

Actual Value (%)

Predicted Value (%)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

40 60 40 60 40 60 40 60 40 60 40 60 40 60 40 60 30 70 50 50 50 50 50 50 50

1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.0000 1.0000 2.0000 2.0000 1.5000 1.5000 0.5000 2.5000 1.5000 1.5000 1.5000 1.5000 1.5000

50 50 50 50 60 60 60 60 50 50 50 50 60 60 60 60 55 55 55 55 45 65 55 55 55

1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:4000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:6000 1:5000 1:5000 1:5000 1:5000 1:5000 1:5000 1:3000 1:7000 1:5000

80.7700 70.5400 66.8400 69.6100 84.8860 87.6560 73.9560 86.7260 67.7700 71.5400 65.8400 69.7100 74.8860 86.6560 85.9560 86.7260 74.4780 80.0180 78.1780 86.3180 60.1320 94.3640 77.2480 76.9420 76.8431

83.0520 70.8070 61.9403 73.3347 87.9489 85.8663 74.4695 87.1048 63.6320 75.6868 68.1964 74.0382 74.2280 87.9548 90.1428 82.6761 73.1443 80.1113 78.2802 81.3596 57.3792 95.4555 76.4045 75.3852 77.8612

95 X: 64.58 Y: 1.588 Z: 94.51

SASO FAME (%W/W)

100

90 85

90 80

80

70 75 60 70

50 1.6 65

1.5 55

1.4 Oil/Methanol ratio

65

60 50 1.3

45

60

Reaction Time (mins)

Fig. 5A. The 3D response surface plots of oil/methanol ratio and reaction time effects on FAME yield.

biodiesel yield. The optimal value of the response is 95.45% calculated at 60 C, catalyst concentration of 1.5 g, reaction time of 65 min and methanol/oil ratio of 1.5. Calculated statistics as shown in Table 9 indicates that ANN provides a desirable means of

efficiently recognizing the patterns in data and predicting biodiesel yield in agreement with Fig. 8, (MSE Vs Epochs). It showed high coefficient of determination (R ¼ 0.9664) and low standard of error of estimation as shown in Table 7. The results obtained are in close


69

84 83

SASO FAME (%W/W)

100

82 81

90

80 80

79 X: 0.5833 Y: 1.306 Z: 83.63

70

78 77 76

60 1.6 2.5

1.5

2 1 1.3

Oil/Methanol ratio

74

1.5

1.4

75

0.5

Catalyst Concentration (wt%)

Fig. 5B. The response surface plots of oil/methanol ratio and catalyst concentration FAME yield.

95 X: 2.458 Y: 65 Z: 95.66

SASO FAME (%W/W)

100

90 85

90 80

80

70

75

60

70

50 65

65 60

2.5 2

55 50 Reaction Time (mins)

60

1.5 1 45

0.5

Catalyst Concentration (wt%)

Fig. 5C. The 3D response surface plots of reaction time and catalyst concentration effects on FAME yield.

70


80 X: 69.17 Y: 1.588 Z: 80.03

SASO FAME (%W/W)

100

79

90 78 80 77

70

76

60 1.6 70

1.5

60 50

1.4

40 1.3

Oil/Methanol ratio

75

30

Temperature (oC)

Fig. 5D. The 3D response surface plots of oil/methanol ratio and temperature effects on FAME yield.

agreement with the previous research results [16,38].

100 SASO FAME (%W/W)

95

X: 69.17 Y: 65 Z: 97.93

90

90

85

80

80

70

75

60 65

70 60

70 60

55 50 Reaction Time (mins)

65

50 40 45

30

Temperature (oC)

Fig. 5E. The 3D response surface plots of reaction time and temperature effects on FAME yield.

4. Conclusion The maximum SASOME conversion yield were validated as 94.36% using RSM and 95.45% using ANN under the optimal conditions of catalyst concentration of 1.5 w/w%, reaction time of 65 min, oil/methanol molar ratio of 1:5 mol/mol and temperature of 50 C. The fuel properties of SASOME satisfied both ASTMD 6751 and DIN EN 1424 standards. FT-IR characterization showed clear conversion of SASO to SASOME. Therefore, the prediction by ANN and RSM have been proven as usually much faster than the conventional simulation program without lengthy iteration methods of calculation in order to solve differential equations using numerical methods. The results of ANN and RSM models are observed to be comparable. Also the models applied in this research work developed an accurate, time saving, simple and non-destructive methods for estimation of biodiesel yield in batch transesterification from sweet almond seed oil and this has helped in established the conformity of SASOME to standards and the sustainability of the seed oil as an effective feedstock for biodiesel. Also, the high seed oil content and percentage conversion of the oil into biodiesel together with the low factor values obtained after optimization


71

80

SASO FAME (%W/W)

100

79 78

90

77 80 76 X: 32.5 Y: 0.54 Z: 80.43

70

75 74

60 2.5 2

70

73

60

1.5

50

1 Catalyst Concentration (wt%)

72

40 0.5

30

Temperature (oC)

Fig. 5F. The 3D response surface plot of the effect of catalyst concentration and temperature on FAME yield.

Fig. 6. ANN plot showing the predicted output and the actual output.

Table 9 Model statistics and information for the artificial neural network model of SASOME. Performance

Biodiesel

MSE MAE MAD Minimum Absolute Error Maximum Absolute Error R

6.0035 2.786 1.89306 0.0933 4.3282 0.9663

Fig. 7. ANN Regression values for training data, test data, validation data and overall model.

72


Fig. 8. Plot indicating the best performance validation in terms of MSE with respect to the number of iterations.

indicate that the biodiesel production process will be viable on commercial scale. Funding This research did not receive any specific grant from funding agencies in public, commercial, or not-for-profit sectors. Competing interest The authors have declared no competing interest. Acknowledgement We would like to thank the anonymous reviewers for their valuable comments that helped in improving the manuscript. The authors are grateful to the National Centre for Energy Research and Development (NCERD), University of Nigeria Nsukka and National Research Institute for Chemical Technology (NARICT), Ahmmadu Bello University (ABU), Zaria Nigeria for the availability of the laboratory facilities, apparatus and analytical equipment. References [1] International Energy Agency, Oil Marketer Reports, Total OECD12 Month Moving Average Demand Vs Year. 1year Growth, Organization for economic Co-operation and development, Pans, France, 2011. [2] E.M. Shahid, J. Jama, Production of biodiesel: a technical review, Renew. Sustain. Energy Rev. 25 (9) (2011) 4732e4745. [3] A.E. Atabani, A.S. Silitonga, A.B. Irfan, T.M.I. Mahlia, H.H. Masjuki, S. Mekhilef, A comprehensive review on biodiesel as an alternative energy resource and its characteristics, Renew. Sustain. Energy Rev. 16 (2012) 2070e2093. [4] C.N. Stewart, B.L. Joyce, Designing the perfect plant feedstock for biofuels production; using the whole buffalo to diversify fuels and products, Biotechnol. Adv. 30 (2012) 1011e1022. [5] R. Camilli, C.M. Reddy, D.R. Coercer, B.A.S. Van Mooy, M.V. Kinsey, J.C. Jakuba, Tracking hydrocarbon plume transport and biodegradation at deepwater horizon, Science 330 (2010) 201e204. [6] A.M. Rizwan, L.Y.C. Dennis, C. Liu, A review on the generation determination and mitigation of urban Heat Island, J. Environ. Sci. 20 (2008) 120e128. [7] M. Ahmad, H. Sadia, M. Zafar, S. Sultana, M.A. Khan, Z. Khan, The production and quality assessment of mustard oil biodiesel. A cultivated potential oil seed crop, Energy Sources Part A 34 (2012) 1480e1490. [8] U.S Energy Information, International Energy Outlook, 2011. Available from: http//www. Eia.doe.gov/oiaf/ieo/pdf (accessed 20.02.17). [9] A.Z. Abdullah, N. Razali, K.T. Lee, Optimization of mesoporous K/SBA e 15 catalyzed transesterification of palm oil using response surface methodology,

Fuel Process. Technol. 90 (2009) 958e964. [10] T.F. Adepoju, O. Olawale, Acid-catalyzed Esterification of waste cooking oil with high FFA for Biodiesel Production, Chem. Process Eng. Res. 21 (2014) 80e84. [11] E. Betiku, T.F. Adepoju, Methanolysis optimization of Sesame (Sesamum Indicum) oil to biodiesel and fuel quality characterization, Int. J. Energy Environ. Eng. 4 (9) (2013) 1e8. [12] U. Rashid, F. Anwar, M. Ashraf, M. Saleem, S. Yusuf, Application of Response surface methodology for optimization transesterification of Moringa Oleifera oil, Biodiesel production, Energy Convers. Manag. 52 (2011) 3034e3042. [13] X. Yuan, Liu, G. Zeng, J. Shi, J. Tong, G. Huang, Optimization of conversion of waste rape seed oil with high FFA to biodiesel using response surface methodology, Renew. Energy 33 (2008) .1678e1684. [14] X. Fan, X. Wang, F. Chem, Biodiesel production from crude cotton seed oil: an optimization process using Response surface Methodology, Open Fuels Energy. Sci. 4 (2011) 1e8. [15] J. Zhang, S. Che, R. Yang, Y. Yan, Biodiesel production from vegetable oil using heterogeneous acid and alkali catalyst, Fuel 89 (10) (2010) 2939e2944. [16] M. Rajendra, C.P. Jena, H. Rahaman, Prediction of optimized pretreatment process parameters for biodiesel production using ANN and GA, Fuel 88 (5) (2009) 868e875. [17] A. Sarve, M.N. Varma, S.S. Sonawane, Response surface optimization and artificial neural network modeling of biodiesel production from crude Mahua (Mahua indica) oil under supercritical ethanol condition using CO2 as co-solvent, R. Soc.Chem. Adv. (85) (2015) article inpress. [18] S. Giwa, C. Ogunbona, Sweet almond (Prunus amygdalus) “dulcis”) seeds as a potential feedstock for Nigerian Biodiesel Automotive Project, Rev. Ambient Aqua 9 (1) (2014), https://doi.org/10.4136/1980-9934. [19] S.O. Agunbiade, J.O. Olanlokun, Evaluation of some nutritional characteristics of Indian almond (Prunus amygdalus) nut, Pakistan J. Nutr. 5 (2006) 316e318. [20] J.E. Ali, J. Rashid, J.E. Rana, The importance of almond (Prunus amygdalus L.), and its bye-products, Food Chem. 120 (2) (2010) 349e360. [21] FAOSTAT, Production/Crops for almond with shell, Food and agriculture Organization of United Nations, Statistics Division, 2014, pp. 65e120. [22] N.H. Abu-Hamdeh, K.A. Alne face, Biodiesel from almond oil as an alternative fuel for diesel Engines, Appl. Mech. Mater. 575 (2014) 624e627. https://doi. org/10.4028/www.scientific.net/AMM.575.624. [23] S.A. Hernandez, C.M.Z. Flavia, Sweet almond oil: extraction, characterization and application, Quim. Nova 32 (5) (2009) 1342e1345, https://doi.org/ 10.1590/50100-40422009000500044. [24] G. Knothe, Biodiesel and renewable diesel: a comparison, Prog. Energy Combust. Sci. 36 (3) (2010) 364e373. [25] C.G. Jun, K.Y. Sam, J.C. Nag, Using Canola oil biodiesel as alternative fuel in diesel engines: a review, Appl. Sci. 7 (9) (2017) 881, https://doi.org/10.3390/ app7090881. [26] A. Mehdic, H. Kariminia, Optimization of biodiesel production from Iranian bitter almond oil using statistical approach, Appl. Energy 88 (7) (2011) 2377e2381. [27] AOAC, Official Methods of Analysis, thirteenth ed., Association of Official Analytical Chemists, Washington, DC, 1990. [28] V. Patel, Cetane Number of New Zealand Diesel, Report, Office of Chief Gas Engineer, Energy Inspection Group, Ministry of Commerce Press, Wellington, New Zealand, 1999. [29] O.O. Awolu, S.K. Layokun, Optimization of two-step transesterification production of biodiesel from Neem (Azadirachta indica) oil, Int. J Energy Environ. Eng. 4 (2013) 39. [30] H. Ahmadian-Morghadam, F.B. Elegado, R. Nayve, Prediction of ethanol concentration in biofuel production using artificial neural networks, Am. J. Model. Optim. 1 (3) (2013) 31e35. [31] C. Botinestean, N. Ghadaruga, D.J. Hadaruga, I. Jiami, Fatty acids composition by Gas Chromatography-Mass spectrometry(GC-MS) and most important physicoechemical parameters of tomato seed oil (TSO), J. Agroalimentary Process Technol. 18 (1) (2012) 89e94. [32] S. Sharmila, L.R. Jeyanthi, GC-MS Analysis of esters of fatty acid present in biodiesel produced from Cladophora vagabunda, J. Chem. Pharmaceut. Res. 4 (11) (2012) 4883e4887. [33] N. Saifuddin, H. Refai, Spectroscopy analysis of structural transesterification in biodiesel degradation, Res. J. Appl. Sci. Eng. Technol. 8 (9) (2014) 1149e1159. [34] M.M. Conceicao, V.J. Fermardes, A.S. Araiyo, M.F. Farais, I.M. Santos, A.G. Souza, Thermal and Oxidative degradation of castor oil biodiesel, J. Energy Fuel 21 (1) (2007) 1522e1527. [35] E.D. Gunstone, Rapeseed and Canola oil: production process, properties and Uses, JAOCS 33 (2004) 132e139. [36] S.B. Imadi, V.R. Bandaru, S.R. Somalanka, H.R. Garapati, Optimization of medium constituents for the production of citric acid from byproducts using Doehlert experimental design, Enzym. Microb. Technol. 40 (2007) 1367e1372. [37] A.P. De Lima, A.L. De lima, D.Q. Santos, W.B. Neto, Application of factorial design and response surface mechanism to optimize ethyl biodiesel production from corn oil, Rev. Virtual Quim. 5 (5) (2013) 817e882. [38] Y.C. Bhattacharyulu, V.N. Ganvir, A. Aditaya, A. Ramming, Modeling of Neem oil methyl esters production using ANN, Int. J. Comput. Appl. 70 (27) (2013) 0925e8887.