ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ, 2015, том 70, № 8, с. 821-827
ОРИГИНАЛЬНЫЕ СТАТЬИ =
УДК 543
QUANTIFICATION OF BIODIESEL IN BIODIESEL- DIESEL BLENDS USING SPECTROFLUORIMETRY AND MULTIVARIATE CALIBRATION
© 2015 C. N. C. Corgozinho1, P. J. S. Barbeira
Laboratorio de Ensaios de Combustiveis, Departamento de Quimica, ICEx — UFMG Av. Antonio Carlos, 6627, 31270-901, Belo Horizonte, MG, Brazil 1E-mail: camilancc@ufmg.br Received 13.10.2011; in final form 05.02.2015
With the aim to increase the percentage of biofuel in Brazil's energy assets, biodiesel has been compulsorily added to regular diesel oil since January 2008, currently at a concentration of 5% (v/v). The method used to quantify the biodiesel in biodiesel—diesel blends has limitations with respect to their sensitivity to contaminants containing carbonyl groups. In this context, the objective was to develop a simple analytical and low cost method for the quantification of biodiesel in biodiesel—diesel blends combining multivariate calibration to spectrofluorimetry. The proposed method provided the determination of biodiesel in the range of 0.4 to 19.0% (v/v) with prediction errors less than 0.35% (v/v) using synchronized spectrofluorimetry and partial least squares. In the range of 4.5 to 20.0% (v/v) the average relative error was less than 4%. The method is simple, fast, requires no pre-treatment of samples and can be performed on site, featuring low cost and good accuracy.
Keywords: diesel oil, biodiesel, spectrofluorimetry, partial least squares.
DOI: 10.7868/S004445021508006X
Environmental issues regarding the use of fossil fuels, as well as political and economic issues have raised increasing concerns towards renewable energy sources with lower environmental impact [1]. Biodiesel has been compulsorily added to regular diesel since 2008 to increase the percentage of biofuels into Brazil's energy assets [2]. The current concentration is 5% (v/v) [3] and the blend is designated as B5. Despite monitoring and regulation practices carried out continuously, B5 diesel with quality parameters out of specification has been commercialized in Brazil. According to ANP's Quality Bulletin [4], one of the highest rates of non-conformities in these samples was registered for the parameter content of biodiesel, representing approximately 17% of non-compliance of B5 diesel in 2013. The commercialization of out of specification blends has also been found in the United States, where the biodiesel content in diesel is 20% (v/v) [5].
The biodiesel content in the samples analyzed by the Program for Monitoring Fuel Quality is determined according to European Norm EN 14 078 [6]. The procedure is to obtain IR spectra of samples in cyclohexane. Quantification is performed using a univariate calibration curve that relates the concentration of biodiesel to the intensity of the band (1745 ± 5) cm-1, which is characteristic of the stretching of C=O bond of esters. The method has the advantage of not requiring any prior knowledge about the type of biodiesel in the mixture.
However, no distinction between the esters derived from transesterification reactions (biodiesel) from those present in the mono, di and triglycerides (vegetable oils) is noticed. That is, the test is not adequate to inform that the mixture actually contains biodiesel.
A more specific method for this measurement was proposed by the Brazilian National Standards Organization [7]. In this method, the multivariate calibration is applied to absorbance measurements in the mid infrared. Samples can be read directly without dilution, however, the presence of contaminants containing carbonyl groups can cause distorted results [7].
Besides the standard methods cited, several techniques were used, but most of the work was carried out using infrared spectroscopy [1, 8]. Spectrofluorimetric techniques are widely used for characterization and detection of tampered petroleum derived products [9], identification and classification of crude oil [10] and for the classification and characterization of edible oils [11].
Considering the applicability of spectrofluorimetry to the samples of both diesel and vegetable oil, our effort was made to develop a simple analytical method with low cost for the quantification of biodiesel in biodieseldiesel blends, combining this analytical technique with multivariate calibration.
EXPERIMENTAL
Chemicals and reagents. Biodiesel—diesel blends were prepared with pure diesel type S500, provided by Gabriel Passos Refinery (REGAP), and soybean, cotton, soybean/tallow mixture biodiesels were supplied by other refineries. It is noteworthy that in the case of biodiesel from tallow, 30% of soybean oil is added in order to avoid solidification of the final product. These three types of biodiesel were chosen because they represent about 98% of the varieties produced in Brazil.
Each of the three sets of calibration, i.e. a set for each type of biodiesel, was formed by a series of 30 biodiesel—diesel blends, with concentrations ranging from 0.20 a 20.00% (v/v). Each series was divided into three tracks: 0.20 to 5.00% (v/v) with increments of 0.20% (v/v); 5.00 to 10.00% (v/v) with increments of 1.00% (v/v) and 10.00 to 20.00% (v/v), with increments of 2.00% (v/v). Besides these samples, 10 other mixtures were prepared from each type of biodiesel with varying concentrations, which were used as the validation set.
Molecular emission spectra of B0 and mixtures were obtained in conventional and in synchronous mode and used for linear regression and partial least squares (PLS).
Instrumentation. The experimental measurements were performed on a Shimadzu spectrofluorometer model RF-5301 PC equipped with a xenon lamp of 150 W, using quartz cuvettes with path length of 1 cm. The spectra used for the treatment of data were obtained from the average of three spectra for each sample.
The emission spectra were obtained with the fluorescent molecular synchronous mode while maintaining a difference with fixed wavelength (AX) of 40 nm between the monochromators for excitation and emission. The initial excitation wavelength (XEX) was equal to 300 nm and ranged up to 760 nm, with increments of 1 nm and scanning the wavelength of emission (XEM) was done from 340 to 800 nm. The openings of the windows of excitation and emission were equal to 3 nm and the scan rate was approximately 20 nm/s.
For readings of the conventional mode, the samples were excited at one XEX of 430 nm and scanning the emission monochromator was of 400—800 nm. The windows of excitation and emission were both of 3 nm for the opening and the scan rate was of approximately 20 nm/s.
The software Solo 5.0.3b (Eigenvector Research Inc.) was used for the chemometric processing data.
In multivariate calibration the ranges from 440 to 800 nm were used and 340 to 800 nm of the spectra were obtained in conventional and synchronous modes, respectively. Three calibration models were built for each set ofblends, using PLS regression. In the first model the spectra were obtained in the conventional mode, mean centered. In the other two models, the spectra used were obtained in the synchronous mode — one of the models
built with mean centering and the other with data normalized by peak I. Normalization was performed dividing the spectra by the intensity of synchronized emission corresponding to the wavelength of peak I.
The number of latent variables used in the PLS models was determined by cross validation leave-one-out as the one that gave the lowest prediction error for the set of cross-validation (RMSECV). Calibration errors (RMSEC) and prediction (RMSEP) were calculated from the actual concentrations and those provided by the models.
RESULTS AND DISCUSSION
Wavelength definitions. To define the XEM for the excitation of the samples in the conventional mode and the AX for the synchronized mode, 3D graphics were built in which the axes x, y and z correspond, in conventional mode, to the wavelengths of excitation (XEX) and emission (XEM) and to the emission intensity, respectively, and for the synchronized mode, the XEM, the AX and the intensity of the synchronized spectrum. Analyzing these graphs, it was possible to observe that the excitation wavelength around 430 nm provided the maximum emission intensity, which was observed at about 474 nm. For the synchronized mode, it was observed that most of the samples showed a great AX around 40 nm. This result is in agreement with the optimized AX by Patra and Mishra [12, 13] for samples of diesel from India and therefore was used for the readings of the samples.
Molecular emission spectra. The molecular emission spectra of pure biodiesel, diesel and their blends were obtained in conventional and synchronized mode (Figs. 1, 2). In these figures the arrows indicate the increasing content of biodiesel within the blends. Comparing the spectra obtained in the conventional way, it seems that biodiesel—diesel blends containing cotton (Fig. 1a1) show the emission band at about 475 nm. For blends containing soybean biodiesel (Fig. 1a2) there is a split of the emission band when content of biodiesel increases, an effect that is even more pronounced when the biodiesel is blended with soybean/tallow biodiesel (Fig. 1a3). In these cases, the bands appear around 470 and 500 nm.
For the emission spectra obtained in the synchronous mode (Fig. 2) the mixtures have a more intense emission band between 400 and 450 nm and a shoulder in the region of 485 nm. For samples containing biodiesel from cotton (Fig. 2a1) this shoulder became more defined with the increasing content of biodiesel, which can be attributed to the emission at 500 nm by biodiesel made from pure cotton. For other samples of biodiesel, the emission bands occur in the region which coincides with the emission of pure diesel.
These results show that each biodiesel provides different analytical signals, which can hinder the development of a single model applicab
Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.