FTIR-ATR and Multivariate Calibration for the Prediction of Biodiesel Concentration in Petrodiesel Blends ()
1. Introduction
At present, most of the world’s energy consumption comes from oil, coal and natural gas, but these sources are limited because of their fossil origin, and it is a fact that they will be depleted in the near future. Therefore, it is essential to find alternative energy from renewable sources. Biodiesel has a promising potential as an alternative energy source. First, biodiesel represents a strategic renewable energy source that can replace diesel oil; second, biodiesel may be useful for reducing the emission of greenhouse gases [1]. It has been established that the combustion of diesel oil by engines produces mixtures of anhydride pollutants (COx, SOx and NOx), solid particles and unburned hydrocarbons that generate multiple environmental and health problems and contribute to global warming and the formation of acid rain [2-4].
Recent studies have shown that the use of petrodieselbiodiesel blends in combustion engines reduces the emission of pollutants, particularly of carbon dioxide. These blends are currently used in different proportions in biodiesel. The United States, Germany, France, Italy, Spain, Brazil, and others, commonly use blends of 98% - 2%, 85% - 15% and 80% - 20% petrodiesel-biodiesel [5].
Worldwide, the cultivated area of African palm (Elaeisguineensis Jacq) is increasing rapidly due to the growing demand for oil for the manufacture of biofuels. The African palm produces 5.5 tons of oil per·ha−1 [6], therefore, the African palm represents potential resources for the synthesis of biodiesel; currently 26% of the oil produced in the world comes from the African palm, but that proportion continues to increase [7].
The United States consumed 1.135 × 108 L of biodiesel in 2004, while the European Union consumed 9.274 × 108 L in 2006, with worldwide consumption forecasted to reach 2.65 × 109 L in 2015, this increase in biodiesel consumption is attributed to the demand for petrodiesel-biodiesel blends for diesel engines.
The need for economical, fast and efficient analytical techniques to assess the quality of petrodiesel-biodiesel blends has resulted in the development of instrumental techniques [8,9]; among the most widely used are gas chromatography, liquid chromatography, and spectroscopyinfrared spectroscopy (IR) is one of the most versatile techniques that have been used to monitor the transesterification processes of vegetable oils [10-16] and to quantify the concentration of biodiesel in blends with petrodiesel [17-22].
Due to the increased demand for petrodiesel-biodiesel blends worldwide, and the potential of African palm oils to serve as raw material for the synthesis of biodiesel, the aim of this study was to develop predictive models for estimating the concentration of African palm biodiesel in blends with petrodiesel by using chemometric techniques and infrared spectroscopy (FTIR-ATR).
2. Materials and Methods
2.1. Chemicals and Reagents
Reactive grade methanol and sodium hydroxide from Fermont were used for the synthesis of biodiesel. Commercial grade crude oil from African palms was donated by the company Agroindustrias de Palenque S.A., Chiapas, Mexico. The transesterification reaction was carried out in a round glass flask with a capacity of 1 liter; the temperature changes during the reaction were measured by a Scientific Thomas digital thermometer (±0.001˚C). The flask was fitted with a condenser to prevent any loss of methanol vapors. The mixing of the reagents was carried out by magnetic stirring. A Barnstead-Electrothermal electric blanket (BI), controlled manually was used to heat the reagents.
2.2. Instrument
The petrodiesel/biodiesel blends were analyzed using a FTIR-ATR spectrophotometer, model Agilent, Cary 660 equipped with a diamond-tip. The software Resolution 4.0 pro Variant served as an interface between the computer and the spectrophotometer. All spectra were obtained in the wavelength range between 549 and 4002 cm−1, with a resolution of 2 cm−1. The spectra were collected at 25˚C; sixteen scans were performed for each analysis. The cell was cleaned three times with acetone after each analysis. The data collected were exported to Microsoft Excel 200 and subsequently exported to the Infometrix Piruette V.4 software to build the model using partial least squares (PLS).
2.3. Procedure
2.3.1. Synthesis of Biodiesel
Biodiesel from African palms was synthesized at atmospheric pressure under the following reaction conditions: oil-methanol ratio of 1:6; catalyst concentration of 2% by weight based on the weight of the oil; reaction temperature of 60˚C and reaction time of 150 min. The catalyst was prepared by reacting anhydrous methanol with sodium hydroxide in a 500 mL Erlenmeyer flask; the mixture was stirred until sodium hydroxide was completely dissolved. The catalyst was added to the oil in the reactor and the mixture was heated to 60˚C; to facilitate the transfer of mass between the reagents, the reagents were mixed by magnetic stirring. After the reaction, the mixture was poured into a separating funnel.
The biodiesel was washed with distilled water to eliminate the presence of soap in the aqueous phase; two immiscible phases were formed in the washing process: the lower phase, containing water, glycerol, methanol and soap was separated by gravity; the upper phase, containing the methyl esters, was transferred to a distillation device in which it was distilled at 100˚C to remove the methanol and water remaining in the biodiesel. In the next purification phase, the biodiesel was refined by passing it through a glass column packed with 100 g of Amberlite Dry ion resin 10, which allowed for the removal of trace contaminants. The biodiesel was packed in amber bottles and kept refrigerated at 5˚C until further use.
2.3.2. Biodiesel/Petrodiesel Blends Preparation
Nineteen petrodiesel-biodiesel blends were prepared in triplicate for the construction of the prediction model (57 samples in total). The blends were prepared using biodiesel synthesized in the laboratory and commercial petrodiesel, purchased from a certified PEMEX franchise. An analytical balance (Ohaus ± 0.001 g) and a 100 - 1000 µL automatic pipette were used to weigh the amounts corresponding to the blends: B0, B5, B10, B15, B20, B25, B30, B35, B40, B50, B53, B60, B70, B75, B80, B82, B90, B95 and B100, which were placed into 250 mL Erlenmeyer flasks.
3. Results and Discussion
3.1. Interpretation of the Spectra
The biodiesel produced from palm oil were analyzed using FTIR-ATR spectrophotometry. The spectra as shown in Figure 1. The band located at approximately 3012 cm−1 is attributed to the C-H stretching of the double bonds of the C-H group which make up the olefins [12]; the bands located at 2924 and 2852 cm−1 are related to C-H vibrations of the methylene groups and to stretching and contraction vibrations of the methyl group. The intense peak located at 1745 cm−1 corresponds to the carbonyl radical and is characteristic of esters. The band located at 1450 cm−1 corresponds to the asymmetric stretching of the C-H bond and the asymmetric bending
Figure 1. Absorbance bands in the infrared spectrum of samples from: (a) African palm biodiesel and (b) Petrodiesel.
of the functional group. There is a set of bands between 1000 and 1300 cm−1 that are associated with asymmetric vibrations of the C-C(=O)-O and O-C-C bonds; the high intensity bands located at 1200 cm−1 are attributed to the stretching of the methyl group O-CH3 and to the axially asymmetric deformation of the C-C=O bonds. The bands between 1100 - 1170 cm−1 correspond to the vibrations of the C-CH2-O group, the asymmetric stretching of C-O-C and C-C bond stretching [13]. The high intensity peak located at 950 cm−1 corresponds to deflections out of the plane of the C-O group and the one located at 1000 cm−1 is attributed to the symmetric angular deformation of the C-H bond of olefins [19].
The IR spectrum of petrodiesel (Figure 1) presents fewer bands compared with biodiesel. There are three overlapping bands that correspond to the methylene and methyl groups in the region between 2985 and 2870 cm−1; the first peak is attributed to the stretching vibrations of the terminal CH2 group in the olefins chain. The second and the third peaks correspond to stretching vibrations and contractions of the C-H and CH2 bonds of the methyl and methylene groups. The most relevant folding vibrations of the methyl groups are those corresponding to the phase folding deformation (1378 cm−1 band) and the outof-phase degenerate phase folding deformation (1463 cm−1 band). The foldings arise from twisting and balancing that appear at low frequencies. The methylene group presents scissors vibrations at 1463 and torsion vibrations at 721 cm−1, respectively.
The variation of the concentration of biodiesel in petrodiesel blends is reflected in the absorbance intensity of the functional groups or chemical bonds, which generates multiple overlapping bands at different wavelengths of the infrared spectrum in Figure 2. This information contained in the spectral fingerprints allowed us to develop a prediction model through chemometrics analysis that was employed to determine the concentration of biodiesel in blends with petroleum diesel with high accuracy.
Figure 2. The variations in the intensity of absorbance in different infrared spectral bands reflect the biodiesel concentrations in petrodiesel-biodiesel blends.
3.2. Analysis of the PLS Model
Fifty seven infrared spectral fingerprints were used to build the calibration model and were within the wavelength range of 651 - 4001 cm−1. Each spectrum consisted of 1738 variables, which gave a total of 99,066 variables forming the data matrix.
To build the calibration model, the data were pretreated using a central mean and then normalized. To determine the robustness of the calibration model, the model was evaluated using cross validation with “leave five out” data and by external validation with twelve blends prepared in the laboratory; the blends were not used to build the model.
To evaluate the predictive ability of the calibrationmodel, some statistical parameters were obtained (Table 1): optimal number of principal components (PCs); explained variance (Var %) in the X matrix; standard error of calibration (SEC); standard error of validation (SEV); correlation coefficient between the actual concentration of biodiesel and the concentration of biodiesel predicted in the validation(Val r); correlation coefficient between the actual concentration of biodiesel and the concentration of biodiesel predicted in the calibration (r Cal); standard error of calibration (SEC) and standard error validation (SEV); standard error of prediction (SEP) and (SEC/ SEV) similarity criterion [23].
The model predictions with six PCs explain 99.306% of the total variance. The calibration (r Cal) and validation (r Val) correlation coefficients had values to the unity, which indicates that a high correlation exists between the actual values of the biodiesel concentration and the concentration predicted by the model; with six PCs, the values of the standard error of calibration (SEC) and the standard error of validation (SEV) were lower than unity, indicating that the model has a high predictive ability (Table 1). The value of SEC/SEV similarity criterion was 0.935, which is near unity, indicating that the model has a good fit, as the identified values are within 0.5 - 1 [23]. The low value presented by the standard error of predictction (SEP = 0.736) shows that the model has a high predictctive ability and high precision; the cross validation using “leave five out” data shows that the model is robust, as the values of the statistical parameters were good in both cases. This is shown in Figure 3.
4. Conclusion
Fourier transform infrared spectroscopy (FTIR-ATR) and chemometric analysis are useful tools for building a PLS prediction model. The statistical parameters obtained for the prediction model show that this approach allows for the determination of the concentration of biodiesel from
Table 1. Cross validation using leave five out data to evaluate the PLS prediction model.
Figure 3. Calibration graph of the PLS regression model built with cross-validation using “leave five out” data; used to predict the concentration of biodiesel of African palm in blends with petrodiesel.
African palm biodiesel in blends with petrodiesel with high precision. The spectroscopic technique can be used together with the traditionally developed model to control the quality of petrodiesel-biodiesel fuel blends. It has the advantage of being quick, inexpensive, efficient and environmentally friendly.
5. Acknowledgements
We are grateful to PROMEP for providing financial resources throughout the project PROMEP/103.5/09/4363 and the Research Line 3 of Alternate Energy and Biomaterials of Colegio de Postgraduados. The Nanomaterials Laboratory Faculty of Chemistry, Universidad Autonó- madel Carmen, Mexico, for performing the IR analysis of petrodiesel-biodiesel blends.
NOTES
#Corresponding author.